Interactions of molten salts with cathode products in the FFC Cambridge Process

Molten salts play multiple important roles in the electrolysis of solid metal compounds, particularly oxides and sulfides, for the extraction of metals or alloys. Some of these roles are positive in assisting the extraction of metals, such as dissolving the oxide or sulfide anions, and transporting them to the anode for discharging, and offering the high temperature to lower the kinetic barrier to break the metal-oxygen or metal-sulfur bond. However, molten salts also have unfavorable effects, including electronic conductivity and significant capability of dissolving oxygen and carbon dioxide gases. In addition, although molten salts are relatively simple in terms of composition, physical properties, and decomposition reactions at inert electrodes, in comparison with aqueous electrolytes, the high temperatures of molten salts may promote unwanted electrode-electrolyte interactions. This article reviews briefly and selectively the research and development of the Fray-Farthing-Chen (FFC) Cambridge Process in the past two decades, focusing on observations, understanding, and solutions of various interactions between molten salts and cathodes at different reduction states, including perovskitization, non-wetting of molten salts on pure metals, carbon contamination of products, formation of oxychlorides and calcium intermetallic compounds, and oxygen transfer from the air to the cathode product mediated by oxide anions in the molten salt.


Background
In December 1999, the University of Cambridge published an international patent on what is now known widely as the Fray-Farthing-Chen (FFC) Cambridge Process. It is about electrolytic extraction of metals and alloys directly from their solid compounds in molten salts [1]. Preliminary findings from testing the FFC Cambridge Process were soon reported in a Letter to Nature for the extraction of titanium from titanium dioxide (TiO 2 ) in molten calcium chloride (CaCl 2 ) [2]. The report aroused enthusiastic responses, both positive and critical, from global communities of titanium technologists and researchers [3][4][5]. In the past two decades, world-wide research and development have confirmed the scientific principle and technical feasibility and flexibility of the process for the extraction of almost all metals listed in the periodic table and their alloys from the respective oxide or sulfide precursors [6][7][8][9][10][11][12][13][14]. In addition, the FFC Cambridge Process has been shown to have versatile applications in other fundamental and industrial areas such as near-net shape manufacturing of metallic artefacts of complex structures, medical implants, oxygen generation on the Moon, capture and electrolytic conversion of carbon dioxide CO 2 to various forms of solid carbon, e.g. carbon nanotubes, carbon monoxide (CO) and hydrocarbon fuels (C n H 2n+2 , n < 10), and rechargeable molten salt metal-air batteries [15][16][17][18][19][20][21][22].

Basic electrochemistry
The main claim of the FFC Cambridge Process is very general and states a method "for removing a substance (X) from a solid metal, a metal compound or semi-metal compound (M 1 X) by electrolysis in a fused salt of M 2 Y or mixture of salts, which comprises conducting the electrolysis under conditions such that reaction of X rather than M 2 deposition occurs at an electrode surface, and that X dissolves in the electrolyte M 2 Y." This claim makes it very clear that the cathode can be a metal containing another substance (e.g. impurity) or a metal A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d or semi-metal compound, including but not limited to metal oxides which have been mostly studied in the past two decades. It is interesting to point out that later research has demonstrated that the electrolysis of metal sulfides is actually quicker and more efficient than that of metal oxides, but the invention of FFC Cambridge Process was mostly based on the initial study of electrochemical reduction of TiO 2 to Ti metal in molten CaCl 2 , which is now known as one of the few most difficult metal oxides to electrolyse.
In terms of thermodynamics [23], electrolysis of TiO 2 should be fairly feasible, following Reaction (1) or (2) and (3) on an inert or carbon anode, respectively. Note that although Reaction (3) seems more feasible than Reaction (2) on a carbon anode, the highly spontaneous Reaction (4) makes CO 2 effectively the main product on the carbon anode. This view can be further explained. With log K = 2log p CO2  (2log p CO + log p O2 ) = 16.1, it can be established that p CO2 >> (p CO + p O2 ) and (p CO2 + p CO + p O2 )  1 atm for bubbles of the mixed anode gases to escape from the anode surface. Therefore, log (p CO  Carbon anode: C + x O 2-= CO x + 2x e (x = 1, 2) (7) Reaction (5) represents electrochemical reduction, or deoxidation of TiO 2 . The same can be written for other metal oxides. Thus, electro-deoxidation and electro-reduction are both used in the literature as the scientific terms in place of the FFC Cambridge Process. Figure 1a illustrates schematically a typical laboratory molten salt electrolysis cell for studying the FFC Cambridge Process. This two electrode cell is suitable for electro-reduction of metal oxides and other compounds at the gram scale. Note that the lower part of the steel vessel (or retort) extrudes below and outside the furnace so that it remains at temperatures much lower than that of the molten salt. In this way, any molten salt dripping or condensing at the welding joint between the wall and bottom of the retort will solidify and become noncorrosive [18]. It can also be readily modified into a three electrode cell for fundamental analyses by, for example, cyclic voltammetry and chronoamperometry.  Process. For the cathode, Figure 1b and 1c show the pellets of TiO 2 (white) and mixed TiO 2 , Al 2 O 3 and V 2 O 5 (brown) that were to attach to, and be electro-reduced on the cathode current collector. The products in Figure 1d and 1e were the Ti-6Al-4V alloy before and after light polishing. For the anode, because of their easy availability, ease of shaping and low cost, commercial graphite products, particular rods (see Figure 1f) and plates, have been commonly used to make the anode or counter electrode in various studies on the FFC Cambridge Process, although quality of commercial graphite varies significantly. Poor quality graphite may suffer from the attack of oxidation and gas bubbling, leading to graphite erosion as shown clearly by comparison between Figure 1g and 1h, and unwanted carbon debris off the anode. The carbon debris can float on, or suspend in the molten salts, causing electronic conduction to lower electrolysis efficiency and contamination of the product on the cathode.
More discussions are given later on the issues from using a graphite anode. Glassy carbon can also be used to make the anode or working electrode in studies of molten salts, particularly for fundamental analysis by cyclic voltammetry [26,27]. Like graphite, glassy carbon also suffers from electrochemical oxidation in presence of oxide ions, and hence is too expensive to use in bulk electrolysis.
In addition, it has been recognised that the discharge (electro-oxidation) of oxide ions (O 2-) on a carbon electrode suffers from serious kinetic difficulties [26,27]. These complex kinetic steps lead to a fast increase of polarisation with increasing the current density on the anode. The kinetics are partly responsible for the practically applied cell voltage for electrolysis of TiO 2 to be at or greater than 3.00 V, in contrast to the thermodynamic predictions of around 1.00 V according to Reactions (2) and (3). To minimise such kinetic barriers, the surface area of the graphite anode in contact with the molten salt should be as large as realistically possible. It was found that the anodic polarisation could be reduced by about 1.0 V when the graphite anode surface area was increased by 10 times [27].
Alternatively, inert anodes should be considered as an option.
A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d By definition, inert anodes should be inactive and non-consumable, and would work without all the problems resulting from using carbon. Although the potential for anodic formation of O 2 is about 1.00 V higher than that for CO 2 formation, see Reactions (1) and (2), this benefit of using carbon anodes could be largely lost to the above mentioned kinetic polarisation. There are two types of ceramic based inert anode: ion blocking and ion conducting as explained in Figure 2a and 2b, respectively. Several materials have been tested for making ion blocking inert anodes, including tin oxide (SnO 2 ) with or without copper doping, and calcium ruthenate (CaRuO 3 ) with or without titanium substitution [28,29]. Ion conducting inert anodes are constructed with a membrane of oxide ion conductor, typically yttria stabilized zirconia (YSZ) in the form a tube with a closed end. One side of the membrane, the wet side, faces the molten salt to connect O 2conduction into the membrane.
The other side, or dry side of the membrane can be made in contact with a liquid or solid metal, e.g. tin (Sn) or silver (Ag), with relatively high diffusivity and solubility of atomic oxygen [30,31] with the produced O 2 gas escaping through the pores in the paste into air [32]. The main issues of ceramic based anodes include their highly temperature dependent resistivity, thermal cyclability and related costs.
It is worth mentioning that graphite anodes are almost fully inert for discharging sulfide ions (S 2-) to the sulfur vapour (S 2 ) [33,34], and chloride ions (Cl -) to the chlorine gas (Cl 2 ) [35,36] in molten chlorides. In electrolytes with high oxide ion activity, such as molten carbonates and mixed oxide melts, refractive metals with high tendency to form a stable surface oxide layer may also be used as ion blocking inert anodes [37][38][39]. However, a few metals, e.g. iridium (Ir) and silver (Ag) do not form stable oxides at high temperatures, and may behave sufficiently stable as an inert anode in some molten salts [12,40,41].

Perovskitisation of metal oxides on cathode
In a laboratory cell as shown in Figure 1a, the cell voltage applied for electrolysis of TiO 2 is usually about 3.00 V with a graphite anode, although the thermodynamic prediction from Reactions (2) and (3) is less than 1.00 V. Apart from the kinetic difficulties on a carbon anode as mentioned above, cathodic processes also contribute.   was almost twice faster as electrolysis of TiO 2 to achieve Ti products of comparable purities.
The higher speed of electrolysis was also a reflection of faster transportation of O 2ions in the CaTiO 3 cathode than in the TiO 2 cathode. This is because electro-reduction of CaTiO 3 removes not only O 2but also Ca 2+ ions, leaving behind increased porosity. This is in contrast to electro-reduction of TiO 2 in which perovskitisation brings about increased volume of the solid phase and hence blockage of the pores in the oxide cathode, impeding removal of O 2ions and the whole electrolysis. Note that continuous CaTiO 3 electrolysis via Reaction (16) will lead to accumulation of CaO in the molten salt. However, one can in principle combine Reactions (16) and (13) to form a closed loop in which CaO is cycled and functions like a "phase change catalyst" to accelerate the electrolysis of TiO 2 .
The second approach is to simply increase the porosity of the TiO 2 cathode, using the low cost and recyclable NH 4 HCO 3 as the fugitive porogenic agent. In most previous studies of the FFC Cambridge Process, the TiO 2 cathode had usually a porosity of 40 to 50 %.
Because the molar volumes of TiO 2 and CaTiO 3 are 18.9 and 34.2 mL/mol, respectively, perovskitisation can lead to a volume increase up to 81 %. Obviously, when it happens inside the pores of the TiO 2 cathode of 40 to 50 % in porosity, partial blockage of the ion channels is inevitable. Because perovskitisation of the TiO 2 cathode proceeds with electro-reduction A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d which follows the 3PI propagation mechanism [46][47][48], it is possible to bypass the effect of perovskitisation by increasing the TiO 2 cathode porosity to 60 to 80 % so that the formation rate of TiO and the pseudo-oxide phases surpasses the rate for perovskitisation. In other words, the perovskite phase, if formed, is unable to grow in size before further reduction.
This was indeed confirmed by experiments in which electro-reduction was found to be most effective when the porosity of the TiO 2 precursor was made about 68 % [49]. It was thought that at higher porosities, the cathode volume also increased, making the path and time longer as well for O 2ion to diffuse out of porous cathode. Another benefit was that the porous TiO 2 cathode prepared from using NH 4 HCO 3 as the fugitive agent presented a micromacro-bimodal porosity as shown in Figure 3a. It is understood that the macropores of over 100 m in length were left by the evaporation of the NH 4 HCO 3 granules that were mixed, pressed and sintered together with the TiO 2 powder which alone would only form micropores A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d  proposed increase of 10 to 20% in cathode porosity may translate to reduced volumetric productivity, which should be well balanced by the benefits in both speed increase and energy saving from using a more porous cathode.

Non-wetting of molten salts on fully electro-reduced metals
It was found in the world's first 99.8 % pure Ti product from electro-reduction of TiO 2 in molten CaCl 2 that the porous oxide precursor turned into a porous metal, as shown in Figure   4a. A technical question asked then on the FFC Cambridge Process was how to remove the salt that had solidified inside the porous metallic products. Since CaCl 2 has a high solubility in water (100 g/mL) and most transition metals and their alloys are relatively stable in water, most, if not all, porous and powdery FFC metal samples reported in the literature were washed in water. Could it be fully effective to remove solidified salts hidden deep inside the pores? Also, because the nodular Ti particles were much larger than the spherical particles in the TiO 2 precursor, growth of the Ti particles must have happened. Therefore, would it be possible that in the course of Ti particle growth, molten CaCl 2 may be enclosed in the metal?
Analyses of the FFC Ti samples by SEM and XRD revealed interesting behaviour. As shown in Figure 4b, with careful selection and inspection, nanometre pores were found in the cross section of some relatively large but broken Ti nodules, but these were empty [50]. It could be argued that CaCl 2 were removed by washing the sample in water and drying before the SEM examination. However, such nanopores should also exist in many more unbroken Ti nodules. If these closed nanopores were filled with CaCl 2 , the salt should have been detected by XRD, but it was not [43,49]. A hypothesis had attributed the formation of nanopores to the space left after removal of the fairly large amount of oxygen in the initially formed metallic phases [50]. For example, the pseudo-oxide phase of Ti 3 O 2 has an oxygen content of 18.2 wt%. This rationale is acceptable but it is not an account for the undetected CaCl 2 in the pores of the porous FFC Ti product.
A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d Another explanation came to light when the cause for an incidental observation was considered. In an experiment for electro-deoxygenation of a Ti foil sample in a Ti crucible, both the foil and crucible were polarised negatively at -3.0 V against a common graphite anode. When electrolysis was near the designated time of completion, the Ti foil was moved forward and backward to drive away a thin layer of floating carbon debris. The manual operation incidentally led to the contact between the foil and crucible. What happened next was surprisingly unexpected. The Ti foil was stuck to the Ti crucible, and the two were only separated after cooling and washing away the solidified salt, and forcing a screwdriver between the two with a hammer. Obviously, the Ti foil was welded to the Ti crucible. It is known that in the absence of impurities, such as oxide, pure metal to metal mixing occurs, resulting in an integral bond. This principle is the foundation of an industrial technique called friction welding. Although the friction generates heat and hence high temperatures to promote atom mobility during friction welding, the metals are still solid without melting [51]. In the case of Ti foil and Ti crucible, because both were negatively polarised for a sufficiently long time, their surfaces must have become oxygen-free and hence were able to weld or bond.
However, there was the molten salt between the two Ti surfaces and why the welding still proceeded could only be explained by the molten salt being non-wetting to pure Ti surfaces. revealing nanometre pores as shown in the inset [50].
This understanding of non-wetting of molten salts on a pure metal surface accounts well for another observation of the absence or very little of salt being present in a well electro-reduced and very lightly washed pellet of mixed nickel, manganese and gallium oxide powders [52]. Later, purposely designed experiments studied the wetting of molten CaCl 2 on the surface of terbium (Tb) metal [53]. In the first test, a fully electro-reduced (metallised) porous pellet of Tb 2 O 3 (converted from Tb 4 O 7 ) powder was cooled to room temperature and broken, without washing, into two halves to reveal the cross section which was directly analysed by SEM and EDX. As expected, CaCl 2 was only detected in the surface region (labelled A), whilst the contents of both Cl and Ca decreased quickly into the metallised pellet (from B to E) as shown in Figure 5a. The SEM image of nodules in Figure 5b was taken from the sand-paper ground surface of the fully metallised sample after rinsing in dimethyl sulfoxide (DMSO) to remove debris from grinding. It can be seen that these nodules were very clean, whilst CaCl 2 is insoluble in dry DMSO.
In the other experiment, a small sheet of pure Tb metal was drilled with 6 small holes of 1 mm in diameter, and immersed into molten CaCl 2 in air. Upon removing the Tb sample from the molten salt, cooled in air, and scraping away the skin of solidified salt, it was seen that all the 6 holes were filled with solidified salt. This change, as presented in Figure 5c and 5d, is understandable because the Tb metal surface was covered with a thin oxide layer on which the molten salt could wet. The Tb sheet was then placed back into the molten salt, electro-reduced against a graphite anode at 3.2 V for 30 min under argon. After electrolysis, the electro-reduced Tb sheet was removed from the molten salt and cooled in air. It was then seen that 5 of the 6 holes were empty. These empty holes must have resulted from the electroreduced Tb sheet surface being free of oxygen. As a result, the walls of the small holes were unwettable by the molten salt which was driven out of the holes by gravity when the Tb sheet was lifted above the molten salt. This finding also allows the use of some organic solvents that have a low but sufficient solubility of the salts, but much lower reactivity to the as-produced metals, particularly rare earths [54]. A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d

Carbon contamination
The use of a carbon anode is commonplace in molten salt electrolysis, and has been the case in most reported studies of the FFC Cambridge Process. Graphite is mostly used, although glassy carbon also helped a few fundamental studies. The advantages of using a graphite anode are basically commercial availability, shaping and processing convenience, thermal and chemical stability in molten salts, and good electronic conductivity. Obviously, according to Reactions (2) and similar ones, emission of CO 2 from the FFC Cambridge Process using a carbon anode is inevitable. However, if cost of materials is correlated with CO 2 impact in the manufacturing stage, carbon anodes may still be a more acceptable choice in comparison with those inert anodes which are still under development. In fact, the more negative impact from using a carbon anode is carbon contamination on the cathodic products, which follows two mechanisms as explained below.

The carbonate cycling mechanism
It has been long recognised that in operation of the FFC Cambridge Process, the CO 2 gas produced on a carbon anode could be re-absorbed back into molten CaCl 2 and contaminate possibly the cathode product via Reactions (17) and (18) below [42,44,55].
CO 2 + CaO = CaCO 3 CO 3 2-+ 4 e = C + 3 O 2- It is worth noting that Reaction (18)   . This affinity pushes the metal deposition potential negatively away from that for carbon deposition. The same does not apply to Na + and K + ions, leading to preference for metal deposition [60].
Unlike in molten carbonate salts or others with high CO 3 2activities, Reaction (18) in molten chloride salts is highly likely diffusion controlled. Also, the fate of CO 3 2in molten CaCl 2 will depend on the experimental conditions. Thermal decomposition of CaCO 3 occurs at temperatures beyond 887 o C, which would reduce (but not likely eliminate) the CO 3 2activity. At a sufficiently high activity, CO 3 2may also compete with O 2to discharge on the carbon anode according to Reaction (19) at a potential about 400 mV more positive than that of the reverse of Reaction (18) [60].
A systematic study of the products from electro-reduction of TiO 2 precursors (cylindrical pellets) with different porosities revealed an interesting trend of carbon contamination in products from low porosity precursors, but not in those high porosity cases [61,62]. XRD patterns of these products are presented in Figure 6a. It should be noted that the XRD patterns of the electrolysed dense precursors were taken from the samples' surface materials because the cores were partially or not reduced. In another experiment, TiO 2 precursors of the same porosity (70%) were electrolysed in molten CaCl 2 with or without added CaO. The products were then analysed by XRD as shown in Figure 6b. In both cases, carbon contamination was represented by the detection of TiC in the electrolysis products.
Considering findings from both experiments, it appears that the TiC phase formed disregarding the porosity of the TiO 2 precursors, but appeared in products from electrolysis for longer times. In addition, as expected, the XRD patterns in Figure 6b suggest convincingly a correlation of carbon contamination with the O 2activity in the molten salt.
It was thought that the anodically generated CO 2 could travel to the cathode through both the gas and liquid phases. In the gas phase route, CO 2 entered the molten salt via

A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d
Reaction (17) at the gas/liquid interface, particularly near the cathode. In the liquid path, the as-produced CO 2 immediately reacts with O 2via Reaction (17) in the molten salt near the anode, and then transports to the cathode via convection and diffusion. Figure 7a illustrates schematically this understanding. In this mechanism, both time for mass transport of O 2and CO 3 2and O 2activity for conversion of CO 2 to CO 3 2play important roles.

The carbon debris mechanism
The other mechanism for carbon contamination is unique to the quality of commercial     and eventually drift to contaminate the cathode [12].
Although there has not yet been a systematic study on the formation mechanisms of carbon debris, two reasonable assumptions are worth more discussion here. Firstly, the reduction of CO 3 2-, and likely directly CO 2 , to carbon can proceed at the "gas/electrolyte/cathode current collector" three phase lines. Such produced carbon may detach from the cathode and float as debris on the molten salt surface. However, this cathodic carbon formation mechanism may have only played a secondary role in the formation of carbon debris, if any. This is because dedicated cathodic deposition of carbon in molten carbonates has very high current efficiency [18][19][20]60], which means that most deposited carbon remained on the cathode upon collection.
Secondly, it was commonly noticed that the graphite anode was not attacked or eroded uniformly after electrolysis, as shown in Figure 1h. In other words, the oxidative attack was selective. The theoretical density of graphite is 2.26 g/cm 3 whilst commercial graphite rods and plates are manufactured by densification of graphitic particles with densities commonly below 2.0 g/cm 3 . Thus, commercial graphite has always a certain level of porosity with the pores, voids or cracks existing more likely in the boundaries between the packed particles. Additional surface pores or cracks on the anode surface can result from oxidative attack by oxygen atoms and molecules formed from O 2discharge, particularly to the boundaries. Such pores or cracks on the surface of a graphite anode would be filled with molten salt in which CO 2 forms. This process leads to two consequences. Firstly, the pore becomes deeper and wider due to carbon loss to CO 2 formation. Secondly, when the speed of CO 2 formation is faster than that for CO 2 to escape from the pore, pressure builds up in the pores. Eventually, the increased CO 2 gas pressure in the pores and the weakened connection between graphite particles jointly force the detachment of carbon debris into the molten salt. Figure 7b illustrates schematically this debris formation process on the carbon anode.
Obviously, like the carbonate cycling mechanism, the carbon debris mechanism also takes A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d time to proceed (to erode the anode gradually) and depends on the O 2activity in the molten salt (to enable CO 2 formation in pores and cracks).

Prevention of carbon contamination
The above discussion indicates clearly that carbon contamination to the products in the FFC Cambridge Process originates from the use of a carbon anode. Therefore, as already discussed above, the use of either an ionic blocking or conducting inert anode will eliminate this problem. Unfortunately, inert anodes are still under development, making graphite the more favoured choice in both laboratory and industry. Further, there are a few approaches that can help reduce the negative impact on the cathode from using a carbon anode. The followings discuss these approaches.
As mentioned above, the key step in the carbonate cycling mechanism is the reaction between CO 2 and O 2in the molten salt near the anode, or at the gas/molten salt interface.
However, decomposition of CaCO 3 to CaO and CO 2 occurs at temperatures above 887 o C.
Thus, the carbonate cycling mechanism should be ineffective in molten CaCl 2 at higher electrolysis temperatures [23,64]. Nevertheless, cautions should be applied when using this simple approach if the molten salts used contains other alkali and/or alkaline earth metal cations which can stabilise CO 3 2against decomposition. As shown by Reactions (20) to (25), CaCO 3 decomposes, but Na 2 CO 3 remains very stable, at 950 o C [23]. Because CaCl 2 has been often mixed with other chloride salts to lower the liquidus temperature, the carbonate cycling mechanism would not change in such melts by raising the electrolysis temperatures. The Physically reducing the direct contact between carbon debris and the oxide cathode was attempted to mitigate carbon contamination from the cathode. An alumina tube (sheath) was used to enclose the oxide pellet cathode from the gas phase and block most mass movement between the anode and cathode in the molten salt, leaving only a small hole in the tube side wall to continue the ionic conduction path, as shown in Figure 8a [35,65,66]. This approach worked highly effectively in elimination of the effect of carbon debris. Figure 8b and 8c compare the electrolysis products from the cell with and without using the alumina tube. Without using the tube, the as-electrolysed cathode was covered by solidified mixture of salt and carbon debris, whilst that electrolysed inside the tube showed clear solidified salt only. In addition, it was noticed that using the tube in the electrolysis led to a significantly lower current flow than that without using the tube, as shown in Figure 8d, but the oxide pellet that was electro-reduced inside the tube could reach an oxygen content A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d current and energy efficiencies. The efficiency derived from the electrolysis data in Figure 8d was almost twice improved by using the alumina tube [66].  Without using the sheathing tube, the graphite anode suffered serious erosion and thinning, see Figure 8e which is similar to that in Figure 1h, but much less erosion occurred to the graphite anode, shown in Figures 8f and 8g, that was coupled with the alumina tube sheathed cathode. It is interesting to note that erosion occurred around the graphite anode in absence of the sheathing tube, but it was more notable in the frontside facing the hole of the tube than on the backside, as shown in Figure 8f and 8g, respectively. These are strong evidence that the O 2flux was smaller from the tube sheathed cathode, than that from the naked cathode, to the graphite anode, which can be explained by reduced CO 3 2cycling. Also, because the discharge of the O 2is driven by the electric field in the interfacial layer between the electrode and electrolyte, i.e. the electric double layer, the difference between Figure 8f and 8g is indicative of difference in the electric field strength. This was less likely caused by the potential distribution around the graphite anode which is a good electronic conductor. It can be attributed to the greater O 2activity and flux in front of the anode facing the hole of the sheathing tube. It is expected that if the anode was placed sufficiently distant away from the hole of the tube, the frontside and backside of the anode would have behaved more similarly as diffusion and convection should even out the O 2distribution around the anode.

Other interactions
A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d The above discussed three main interactions are common to all metal oxides to be reduced in the FFC Cambridge Process. However, some interactions are specific to the metal oxides, or the metals produced. For example, SiO 2 can be electro-reduced in molten CaCl 2 to Si which can react with Ca or Ca 2+ at appropriate cathodic potentials to form various calcium silicides, CaSi x with x varying between 0.5 and 2.0 [67][68][69]. Reactions (26) and (27)  Because formation of CaSi x occurs at potentials more positive than that for Ca deposition, the potential window for electro-reduction of SiO 2 to pure Si is very much limited [67]. Whilst a robust reference electrode [70][71][72] could help perform constant potential electrolysis in laboratory to avoid CaSi x formation, its application can be a great challenge in an industrial cell. The use of a computer-aided control (CAC) of the cell voltage programme may offer a promising direction in industrial production [45]. This approach is simple, energy saving and low cost. The principle of CAC is basically to programme the cell voltage -time profile following that recorded in a successful test of constant potential electrolysis.
Aluminium (Al) is difficult to produce from its oxide, Al 2 O 3 , in the FFC Cambridge Process, although thermodynamics could predict clear feasibility as shown by Reactions (28) and (29).
The literature is lack of sufficient research on electro-reduction of solid Al 2 O 3 , although one report claimed successful production of pure Al [73], but the other described both thermodynamic and experimental findings of formation of CaAl x [74]. A more recent study found that when the Al 2 O 3 powder was used on the cathode, electroreduction produced A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d As it has been discussed above, research on the FFC Cambridge Process in the past 20 years has been successful in making targeted metallic products, thanks to the relatively high stability of the relevant metal oxides and sulfides in CaCl 2 (or LiCl) based molten salts.
However, it was also found that formation of oxychlorides could occur during electroreduction of metal oxides to different extends. The awareness of oxychloride formation came from an early observation of yellowish condensate [76] on the surface of the upper portions of a long graphite anode for electrolysis of TiO 2 in molten CaCl 2 , see Figure 1f. A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d Last, but not the least, in the FFC Cambridge Process, air-isolated high temperature molten salts (e.g. CaCl 2 , LiCl and their mixtures with other salts) are used as the electrolyte.
The applied high temperature helps lower the kinetic barriers for electrochemical reduction of semiconducting and insulating solid metal oxides or sulfides to the respective metals. Air isolation, together with inert gas purging, is the most essential key to the success of the FFC Cambridge Process. To achieve air-isolation, it is crucial that the molten salt electrolyser is strictly sealed in a reaction vessel. There has been a general thought that because argon (Ar) is heavier than air, it can sink in the vessel and protect the electrolyser and cathode product from oxidation by the oxygen in air. This is an unfortunate misunderstanding and needs clarification. It can be derived from the Gibbs energy changes of Reactions (1) and (8) that the equilibrium partial pressures of oxygen are 2.50 x 10 -33 atm and 4.60 x 10 -39 atm, respectively, whilst the oxygen partial pressure in air is 0.21 atm. This ultra large difference in oxygen partial pressure means a huge rate for oxygen to diffuse from air into the reaction vessel, even just through small a leaking gap.
Another misunderstanding is that the molten salt may provide a physical barrier for direct contact between the cathode product and air, and hence protecting the former from oxidation. The truth is that when O 2ions are present in a CaCl 2 or LiCl based melt, which is inevitable in the FFC Cambridge Process, they could function as a phase transfer catalyst and transfer oxygen from the gas phase to the metal on cathode in the molten salt by formation of peroxide (O 3 2-) and superoxide (O 2 -) anions via Reactions (34) and (35) below [78].
3 Ti + 4 O 2 -= 3 TiO 2 + 2 O 2-It can be seen from these two reactions that the O 2ion transfers oxygen, as a vehicle offering the return service, from the gas phase to the Ti metal on cathode in the molten salt. After reaction with the metal, the O 2ion is released back to the melt to bring more oxygen into the A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d melt. Therefore, these two factors will together make it impossible to electro-reduce a metal oxide to the metal with a sufficiently low oxygen content.

Summary
The FFC Cambridge Process has been in research and development for over two decades, embracing various successes but also challenges. Three generic interactions between the cathode product and molten salt, specifically CaCl 2 , have been identified as (1) perovskitisation and similar reactions that include Ca into the cathode without oxygen removal, (2) non-wetting of molten salts on pure metal surfaces, which helps separation of In addition to what has been discussed on using more porous oxide precursor to speed up electro-reduction, it is worth mentioning that a few studies were carried out on oxide pellets of very low porosity ( 30%), capturing more details of intermediate phases [79][80][81].
Another uncovered area of interest and technological importance is nuclear fuel processing [82,83], which deserves a separate specialist analysis. Further, the literature of the past two decades has included numerous comprehensive or specific topic review articles on the FFC Cambridge Process, all showing positive and encouraging views on the prospects [6][7][8][9][10][11][12]40,50,69,[84][85][86]. What has also emerged in the literature is the scaling-up studies of the FFC Cambridge Process with promising results for eventual commercialisation [13,14,[85][86][87].  A c c e p t e d M a n u s c r i p t N o t C o p y e d i t e d