Inert anodes for aluminum electrolysis and method of production thereof

FIELD OF THE INVENTION

[0001] The present invention relates to inert anodes for aluminum electrolysis. More specifically, the present invention is concerned with a composition of anodes for aluminum electrolysis and a method of production thereof.

BACKGROUND OF THE INVENTION

[0002] The primary aluminum industry is a high producer of greenhouse gases with mean emissions of 5.7 to 19.2 tons of C02-equivalent per ton of produced Al, depending on the electric power source. A significant contribution, i.e. about 3.7 tons C02-eq /ton Al, originates from the use of consumable carbon anodes in the Hall-Heroult electrolysis process. In this context, the development of alternative cells consisting in a combination of inert anodes, also referred to as 02-evolving anodes, and wetted cathodes is a first R&D priority of primary aluminum producers. Successful research in this field promises significant environmental benefits, energy savings and cost reductions.

[0003] Among possible inert anode materials, i. e. metals, ceramics and cermets, metal- based anodes are currently promising candidates because they offer high electrical conductivity, excellent thermal shock resistance, mechanical robustness, ease of manufacture and simplicity of electrical connection to current leads. However, obtaining a metal-based inert anode with a long-term viability, i.e. typically of at least several months, in the highly corrosive conditions of the Al electrolysis, is very challenging. The use of a low-temperature, i.e. between about 700 and about 800°C instead of a temperature of about 950° C as standardly used for cryolite electrolyte for Al electrolysis should significantly increase the utility of metal-based anodes and would offer a larger selection of alloys that could be used as inert anodes. However, the decrease of the alumina solubility in low-temperature NaF- AIF ₃ electrolytes causes operational difficulties. KF-AIF ₃ -based electrolytes have been proposed as a way to operate at lower temperatures due to its relatively high alumina solubility at low-temperatures, in amounts of about 5 wt. % at 700 °C.

[0004] Metals are chemically unstable in cryolitic bath and, as a result, metallic anodes must be permanently covered by a protective, self-repairing and relatively thin oxide layer during Al electrolysis. For that purpose, the metallic anode composition must be optimized in order to achieve an adequate balance between the oxidation rate of the metal substrate and the dissolution rate of the oxide layer in the electrolyte. Cu-Ni-Fe based alloys have shown promising properties as inert anodes US patent 5,284,562 to Beck et al., 1994) due to their ability to form an adherent, electronically conducting nickel ferrite plus copper scale during the operation of the electrolysis cell. However, Cu-Ni-Fe alloys present a two-phased microstructure, comprising a Cu-rich phase and a Fe-Ni-rich phase, over a large composition range. This chemical inhomogeneity decreases their corrosion resistance because the iron- rich phase is preferentially corroded upon Al electrolysis inducing the formation of iron fluoride corrosion tunnels in the anode scale as recently shown by Beck et al. (T.R. Beck, CM. MacRae and N.C. Wilson, Metall. Mat Trans. B, 42, 807 (201 1 )).

[0005] Homogenization of the alloys through an appropriate thermal treatment is said to improve their corrosion resistance for Al production (T.R. Beck, CM. MacRae and N.C. Wilson, Metall. Mat. Trans. B, 42, 807 (201 1 ), US Patent 7,7077,945 to Bergsma et al., 2006).

[0006] There is still a need in the art for a composition for anodes for aluminum electrolysis and a method of production thereof.

SUMMARY OF THE INVENTION

[0007] More specifically, in accordance with the present invention, there is provided an inert anode for Al electrolysis, made of Cu-Ni-Fe-0 based materials, comprising Fe in a range between about 10 and 20 % by weight, Cu in a range between about 60 and about 80% by weight, Ni in a range between about 20 and about 30% by weight, and oxygen in a range between about 1 and about 3% by weight.

[0008] There is further provided a method for producing metallic inert anodes, comprising mechanically alloying metallic elements; oxygen doping; and consolidation.

[0009] Other objects, advantages and features of the present invention will become more apparent upon reading of the following non-restrictive description of specific embodiments thereof, given by way of example only with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS [0010] In the appended drawings:

[0011] Figure 1 a is a diagrammatic representation of a method for producing

(Cu65Ni2oFei5)ioo-xO powders according to an embodiment of an aspect of the present invention;

[0012] Figure 1b is a diagrammatic representation of an electrochemical reactor according to an embodiment of an aspect of the present invention;

[0013] Figure 2 shows XRD patterns of (Cu65Ni ₂ oFei5)ioo-xO _x materials for different values of x in a) as-milled state; and b) after powder consolidation treatment; ;

[0014] Figure 3 shows evolution of the lattice parameter of the γ-phase as function of x in as-milled and consolidated (Cu ₆ 5Ni ₂ oFei5)ioo-xO _x samples of Figure 2; [0015] Figures 4 show BSE surface images of the consolidated (Cu65Ni ₂ oFei ₅ )ioo-xOx materials with a) x = 0.3, b) x= 1.4, c) x= 3.3 and d) x= 7.2;

[0016] Figure 5 shows variation of the mass gain Am/m (%) with respect to the oxidation time of the consolidated (Cu65 i ₂₀ Fei5)ioo-xOx materials with x= 0.3, 1.4, 3.3 and 7.2;

[0017] Figures 6 show cell voltage versus electrolysis time for (Cu ₆ 5 i ₂ oFei5)ioo-xOx anodes with a) x = 0.3, b) x= 1.4, c) x= 3.3 and d) x= 7.2;

[0018] Figure 7 shows the evolution of the ohmic drop with the electrolysis time at

(Cu65 i ₂ oFei ₅ )ioo-xOx anodes with x= 0.3, 1.4, 3.3 and 7.2;

[0019] Figures 8 show BSE cross-sectional images of the (Cue5Ni ₂ oFei5)ioo-xO _x anodes with a) x = 0.3, b) x = 1.4, c) x = 3.3 and d) x = 7.2 after 20h of electrolysis;

[0020] Figures 9 are a diagrammatic representations of the composition of the scale formed on the Cu65Ni ₂₀ Feis)ioo-xOx anodes with a) x = 0.3, b) x = 1.4, c) x = 3.3 and d) x = 7.2 after 20h of electrolysis;

[0021] Figure 10 shows Cu, Ni and Fe concentrations (wt.%) in the produced Al versus x in (Cu ₆ 5Ni ₂ oFei5)ioo-xOx anodes after 20 hours of electrolysis;

[0022] Figure 11 shows the evolution of the Cu, Ni and Fe concentrations (ppm) in the electrolyte as a function of the electrolysis time with the (Cu65Ni ₂ oFei ₅ )g8.60i.4 anode; and

[0023] Figure 12 shows BSE surface image of a consolidated composite anode made from a mixture milled for 4 hours of 95.3 wt.% Cu6 ₇ .iNi ₂ o.6Fei2.3 (pre-milled for 10 h) + 4.7 wt.% Fe ₇ o03o (50 nm in size).

DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0024] In a nutshell, there is provided an anode composition and a method for producing metallic anodes having an improved resistance to corrosion, by preparing the anodes starting from alloys synthesized by mechanical alloying, optimizing the stoichiometry, and oxygen doping.

[0025] In particular, it was found that mechanically preparing Cu6sNi2oFei5 alloys under an oxygen atmosphere yielded nanostructured alloys with a resistance to corrosion in cryolitic environment at 700°C increased compared to corrosion resistance of Cu65Ni ₂ oFei5 alloys synthesized under inert atmosphere.

[0026] (Cu65Ni2oFei ₅ )ioo-xOx materials were prepared by mechanical alloying under oxygen atmosphere. Their structural and chemical characteristics were studied at different stages of their preparation and after 20 hours of electrolysis in a low-temperature (700°C) KF-AIF ₃ electrolyte. It was shown that oxygen, when added in appropriate amount during the mechanical alloying process, has a significant beneficial effect on the electrode corrosion resistance.

[0027] (Cu65Ni2oFei5)ioo-xOx materials with different 0 contents were synthesized by grinding powders in two steps as shown in Figure 1 a.

[0028] In a first step, shown in grey in Figure 1 a, the metallic Cu65Ni2oFei5 alloy was synthesized by high energy ball milling (HEBM) of Cu, Ni and Fe powders (Cu purity >99.5%, Ni and Fe purity≥ 99.9%, -325 mesh). The powder mixture (11.35 g) was weighted in a glove box under Ar atmosphere and placed in a vial (55 ml) with three hardened steel balls (two balls with a diameter of 14 mm and one ball with a diameter of 11 mm for a total mass of 22.7 g). The ball-to-powder mass ratio (BPR) was 2:1. Stearic acid (0.5 wt. %) was also added in order to prevent excessive cold welding. The vial was sealed under Ar atmosphere and placed in a vibratory miller (SPEX™ 8000M). The milling time was set at 10 h, leading to the completion of the alloying process between the Cu, Ni and Fe elements.

[0029] In a second step, the Cu6sNi2oFei5 powder was oxidized by a subsequent HEB performed under O2 atmosphere. The oxygen amount in the samples was varied with the number of times (0, 4, 9 and 18 times) that the vial was filled with O2 (Figure 1a). For each O2 filling, the vial was opened and then sealed in a glove bag under oxygen atmosphere at a pressure of 1 atm. The milling duration after each O2 filling was 30 minutes. The oxygen contents x (measured with a LECO™ oxygen analyzer) in the resulting (Cu65Ni2oFei5)ioo-xO _x samples were 0.3, 1.4, 3.3 and 7.2 wt. %, respectively, as indicated in Figure 1a. The O contamination in the sample only milled under Ar atmosphere (x = 0.3 wt.%) may originate from the native oxide layer present on the starting powders and/or to the oxidation of the powder surface with ambient atmosphere once the sample is taken out of the vial. The Cu, Ni and Fe concentrations measured by energy dispersive X-ray (EDX) analysis in the four samples were in accordance (within 1-2 wt. %) with their nominal composition. The structure of the (Cu65Ni2oFeis)ioo-xOx powders was determined by X-ray diffraction (XRD) using a Brucker™ D8 diffractometer with Cu Ka radiation.

[0030] Powder consolidation was then performed to obtain pellet samples for oxidation and electrolysis tests. The as-milled powder was first sieved to select only a powder fraction with a particle size between 20 and 75 |jm. Then, it was introduced into a quartz cylinder pre-form and heated from room temperature to 1000 ° C under Ar atmosphere for a thermal softening treatment. The resulting sample was cold pressed at 26 tons cm- ² for 10 minutes and then sintered at 1000 °C under Ar atmosphere for one hour. The pellet was removed from the heating zone of the furnace and left to cool down to room temperature. The obtained pellets had a diameter of about 11.3 mm and a thickness of about 5 mm for the electrolysis tests, and of about 1 mm for the oxidation tests. Their porosity was assessed according to the following equation: (d _: - i _ej™ r )

Oorosity [ = — x 100

where d _t is the theoretical density determined from XRD measurements, and dexp is the experimentally measured density obtained by weighing and measuring the thickness of the pellet. The porosity of the samples was thus determined to be of 5 ± 2%. The structure of the consolidated samples was determined by XRD. Backscattered electron (BSE) images of the sample surface were carried out using a JEOL™ JSM-6300F scanning electron microscope (SEM).

[0031] Thermogravimetric analyses (TGA) were performed using a Thermax™ 500 equipment. The samples were first heated up from room temperature to 700°C at 10°C min- ¹ under Ar atmosphere. Oxidation experiments were then conducted at 700°C under Ar-20%O2 with a flow rate of 240 cc min ^-1 . The mass variation of the samples was recorded for 20 hours. The nature of the oxides formed during these oxidation tests was determined by XRD analyses.

[0032] For electrolysis tests, a hole was drilled and tapped into the edge of the pellet in order to insert an electrical connection rod protected by an alumina-based cement coating. Electrolyses were performed at about 700 °C under argon atmosphere using a two-electrode configuration cell controlled by a V P3 Multichannel Potentiostat/Galvanostat (by BioLogic Instruments). The electrochemical reactor contained three electrochemical cells and thus, three electrolysis tests could be conducted in parallel.

[0033] Figure 1 b is a diagrammatic representation of such electrochemical reactor, which allows running three experiments at the same time with three electrochemical cells controlled by a multichannel potentiostat/galvanostat. Only one cell set-up is represented here for clarity.

[0034] The reactor comprises a stainless steel container 4 receiving crucibles 11 , provided with a stainless steel cover 3, which temperature was controlled by a furnace controller and a water cooling system.

[0035] In the cell geometry and electrode arrangement illustrated in Figure 1 b, the geometric surface area of the inert anode 7 immersed in the KF-AIF3-AI2O3 electrolyte 9 was about 4 cm ² . The counter electrode 8 was a graphite rod of about 13 cm ² immersed in the electrolyte 9. The anode-cathode distance was 2.3 cm. The crucibles 11 containing the electrolyte were made of sintered alumina. The electrolyte composition was 50 wt. % AIF ₃ -45 wt. % KF-5 wt. % AI2O3. No alumina was added during the electrolysis since its consumption is assumed to be compensated by the dissolution of the alumina crucibles. Electrolyses were performed at an anode current density of 0.5 A cm- ² for 20 hours. Before measurement, the anode 7 was maintained above the electrolyte 9 for 30 minutes and then immersed in the electrolyte 9 at open circuit conditions for 10 minutes.

[0036] Current interruption measurements were performed after 0.25, 5 and 20 hours of electrolysis to determine the different voltage components, i.e., Nernst potential, polarization potential and ohmic drop. The current was interrupted for 30 seconds with a voltage sampling rate of 1 ms. The Nernst potential was defined as the voltage measured at 30 seconds after the current interruption. The ohmic drop was defined as the difference between the operating voltage measured before the current interruption and the voltage taken at 1 ms after the current interruption. The polarization potential was defined as the difference between the Nernst potential and the voltage taken at 1 ms after the current interruption.

[0037] The figure shows a graphite connecting rod 5 connected to the sintered alumina tubes 2 supported by stainless steel rods 1 connected to a thermocouple 12. As the connection rod 6 was partially immersed in the electrolyte, an undesired part of the contamination of the produced aluminum 10 came from the corrosion of the connection rod 6. Thus, in order to determine the Cu, Ni and Fe contamination levels only coming from the anode corrosion, two series of electrolysis tests were performed: a first series using an Inconel 718 rod for Cu quantification and a second series using an aluminum-bronze (C63000) rod for Ni and Fe quantifications. The Cu, Ni and Fe contents in the produced Al and electrolyte were measured by neutron activation. The experiment was repeated at least twice for each anode composition. The annual wear rate of the anode was calculated according to the equation:

( m _b w _b + _Al w _Al )x 36 S x 24

Wear rate (cm year- 1}— [2 }

100 x p _a x S _a x t ^{K J} where /7¾ is the mass of electrolyte (g); w¾ is the mass fraction of contaminants (Cu+Ni+Fe) in the electrolyte (wt.%); rriAi is the mass of produced Al (g); WAI is the mass fraction of contaminants (Cu+Ni+Fe) in the produced Al (wt.%); p _a is the anode density (g cm- ³ ); S _a is the geometric surface area of the anode immersed in the electrolyte (cm ² ); and f is the electrolysis time (h).

[0038] The composition and structure of the oxide layers formed on the anode during Al electrolysis were determined by EDX and XRD analyses recorded after polishing the electrode for different times in order to reveal the successive oxide layers. The surface and cross section of the electrodes were observed by SEM.

[0039] Figure 2a shows the XRD patterns of as-milled (Cu6sNi2oFei5)ioo-xOx powders for x

= 0.3, 1.4, 3.3 and 7.2 wt. %. All XRD patterns exhibit one series of peaks which corresponds to a face- centered-cubic (fee) phase (γ-phase) attributed to a solid solution of Cu(Ni,Fe,0). The lattice parameter of the γ-phase (calculated from the (11 1 ) peak position) increases slightly with x, as shown in Figure 3, which may reflect the insertion of O atoms in the γ-phase.

[0040] Figure 2b displays the XRD patterns of the (Cu65Ni ₂ oFei5)ioo-xO _x materials after the powder consolidation treatment. As expected, this treatment generates grain growth and strain release as illustrated by a decrease of the full width at half maximum (FWHM) of the diffraction peaks. On the basis of Williamson-Hall plots (not shown), the lattice strain is about 0.3% and the crystallite size is about 30 nm for the consolidated samples compared to about 0.5% and about 15 nm before consolidation. Furthermore, a new series of peaks for x = 7.2 was observed, which correspond to a Fe20 ₃ phase. This phase is also observable for x= 3.3 but the intensity of the peaks is much smaller. In addition, it is noted that the diffraction peaks of the γ-phase slightly shift towards higher 2Θ angles with increasing the oxygen content in the material, indicating a decrease of the γ-phase lattice parameter as x increases in the consolidated samples (see Figure 3). This can be attributed to the decrease of the Fe content in the γ- phase due to the formation of iron oxides during the consolidation treatment.

[0041 ] Moreover, it appears in Figure 3 that the γ-phase lattice parameter after consolidation decreases linearly with x. Thus, it is assumed that all the consolidated (Cu65Ni ₂ oFei5)ioo-xO _x materials milled under O2 contain some amount of iron oxides even if the XRD pattern for x = 1.4 does not show any discernable FeOx diffraction peaks (see Figure 2b).

[0042] This is supported by BSE images of the surface of the consolidated

(Cu65Ni2oFei ₅ )ioo-xO _x materials (see Figures 4). Indeed, for x = 1.4, 3.3, 7.2, the micrographs supported by EDX analyses reveal the presence of micrometric Fe203 precipitates, shown as dark grey areas, well distributed in the Cu(Ni,Fe) phase matrix, which appears as clear grey areas. It can be seen that the number and size of the iron oxide precipitates increase with x. Their sizes are typically at most 0.5, 0.5-3 and 1 -5 μ m for x = 1.4, 3.3 and 7.2, respectively.

[0043] Figure 5 shows TGA curves expressed as the mass gain (Am/m ₀ ) with respect to the oxidation time performed at 700°C under 1 atm Ar:02 (80:20) for the different consolidated (Cu ₆ 5Ni2oFei ₅ )ioo-xOx materials. For x = 0.3, a very fast increase of the sample mass is observed during the first hour of oxidation, which is attributed to the formation of CuO as confirmed by XRD analysis (not shown). The oxidation rate drastically slows down for further oxidation time due to the formation of NiO and NiFe204 phases which act as barrier to the copper flux at the oxide-alloy interface. After 20 hours of oxidation, the mass gain reaches 5.2%. The addition of a small amount of 0 in the Cu-Ni-Fe alloy (x = 1.4) decreases drastically the oxidation kinetics by preventing the rapid mass gain during the first stage of oxidation related to the formation of CuO, which leads to a mass gain of only 1.1% after 20 hours of oxidation.

[0044] A possible explanation is that the presence of finely dispersed Fe203 inclusions in the Cu-Ni-Fe matrix (Figures 4) favors the rapid formation of NiFe ₂ C>4 from NiO + Fe203 thanks to a minimization of the diffusion distance between each compounds. The formation of NiFe204 is assumed to limit the outward diffusion of Cu in Cu oxides, resulting in a lower oxidation rate as shown in Figure 5. More added oxygen in the Cu-Ni-Fe alloy (x = 3.3 and 7.4) induces minor additional improvement of the alloy oxidation resistance, with a mass gain of 0.7% for both samples after 20 hours of oxidation. The XRD patterns of the (Cu6sNi2oFei5)ioo-xO _x samples after the oxidation test (not shown) confirm the formation of CuO, NiO and NiFe20 ₄ in all cases.

[0045] Figures 6 show the evolution of the cell voltage for 20 hours of electrolysis in a low-temperature (700°C) KF-AIF ₃ electrolyte at 0.5 A cm- ² with the different (Cu65Ni ₂ oFei5)ioo-xO _x electrodes. For x = 0.3, the cell voltage gradually increases from 3.8 to 4.4 V during the 20 hours of electrolysis. For x =1.4 and 3.3, the cell voltage is more stable with a slight increase from ca. 3.9 to 4.1 V. For x = 7.2, the cell potential is less stable and higher with a rapid decrease from 4.5 to 4.1 V during the first hour of electrolysis followed by a slow increase to reach 4.4 V at the end of electrolysis.

[0046] The current interruption method was performed after 0.25, 5 and 20 hours of electrolysis in order to determine the different voltage components, i.e., Nernst potential, polarization potential and ohmic drop. For all electrodes, the measured Nernst potential is initially around 2.1 V and reaches a stable value of ca. 2.4 V after a few hours of electrolysis which is in accordance with the theoretical voltage (E = 2.37 V) for the decomposition reaction of alumina (AI2O3 = 2AI + 3/20 ₂ ) at 700°C under 1 atm. O2. The polarization potential at the end of the electrolysis is in the range 0.15-0.3V. [0047] The evolution of the ohmic drop with the electrolysis time for the four electrodes is showed in Figure 7. Assuming that the external circuit, electrolyte and electrode connection resistance are similar and stable for all the electrodes, the evolution of the ohmic drop is an indication of the variation of the electrical resistance of the anode. After 0.25 hours of electrolysis, the ohmic drop is assumed to be mainly due to the bulk electrical resistance of the anode and thus, its increase with x observed in Figure 7 (from 1.33V for x = 0.3 to 1.69 V for x = 7.2) may be explained by the larger amount of insulating iron oxide particles with increasing x (see Figure 4). On the other hand, the variation of the ohmic drop with the electrolysis time is considered to reflect the growth of the oxide scale on the electrode surface. It clearly appears in Figure 7 that the increase of the ohmic drop with the electrolysis time is less marked as x increases with a mean variation of 18, 7, 5 and 3 mV IT ¹ for x= 0.3, 1.4, 3.3 and 7.2, respectively. This could reflect the fact that, as x is increased, the thickness of the oxide scale decreases or that its conductivity is increased. It was already shown that the gain of mass, as measured during TGA experiments, decreases as x is increased, indicating that the thickness of the oxide scale varies with x (see Figure 5). However, it will be hereinbelow that a thick oxide scale is formed at the surface of the electrode with x = 7.2, suggesting that the conductivity of this oxide scale is large or that it is highly porous.

[0048] Figures 8 show BSE cross-section images of the four (Cu65Ni2oFei5)ioo-xO _x anodes after 20 hours of electrolysis. The presence of an oxide scale is easily discernable, which is delaminated from the bulk alloy for x= 3.3 and 7.2 probably due to the thermal shock when the electrode is taken out of the electrolyte. In all cases, the surface scale is composed of three main layers but their thickness and nature depend on the electrode composition.

[0049] The schematic representation of these layers determined from EDX and XRD analyses after polishing the electrodes for different times but not shown here is presented for each electrode in Figures 8.For x = 0 (Figures 8a and 9a), the surface scale appears dense, which is supported by the fact that no electrolyte salts was detected into it. The outermost layer is a Cu20-rich scale about 200 μ ιτι thick containing NiO and FeO inclusions. The intermediate layer (about 100 m in thickness) consists of a mixture of CU2O and NiFe204. This oxide scale structure results from the outward diffusion of Cu in Cu oxides and internal oxidation of Fe and Ni with the subsequent formation of NiFe20 ₄ . Near the bulk alloy, a non-continuous layer of FeF ₂ (about 50 μ m in thickness) is observed, which is assumed to be mainly formed when the anode was maintained above the electrolyte for 30 minutes and during the first minutes of electrolysis.

[0050] For x = 1.4 (Figures 8b and 9b) and x = 3.3 (Figures 8c and 9c), the surface scale appear less dense than for x = 0.3 with the presence of few pores but no electrolyte infiltration was observed into it. Both electrodes present the same layer arrangement with approximately the same thickness. The outermost layer (about 120 and about 135 m thick for x = 1.4 and 3.3, respectively) is composed of a mixture of NiFe20 ₄ , FeO _x and CU2O with NiFe20 ₄ as major constituent. Underneath this one, a layer (about 80 and about 95 μ m thick for x = 1.4 and 3.3, respectively) containing CU2O and NiFe20 ₄ is present. Finally, the inner layer (about 50 μ m thick) is constituted of FeF ₂ inclusions inside the alloy matrix as observed for x= 0.3. The fact that the outermost layer is thinner and much poorer in CU2O for x = 1.4 and 3.3 than for x = 0.3 indicate that the outward diffusion of Cu in Cu oxides is significantly slowed down. As discussed before, this is attributed to the more favorable formation of NiFe20 ₄ , as confirmed by the observation of a NiFe20 ₄ -rich outermost layer, for x = 1.4 and 3.3 due to the presence of finely dispersed Fe ₂ C>3 inclusions in the Cu-Ni-Fe matrix acting as nucleation sites for the formation of NiFe ₂ 0 ₄ .

[0051] For x = 7.2 (Figures 8d and 9d), the electrode presents the thickest surface scale with a total thickness of about 440 μ m, which is due to the formation of a thick Cu20-rich outermost layer (about 340 μ m). Moreover, a significant amount of electrolyte was detected into it. In addition, NiFe20 ₄ is not observed in the outermost layer in contrast to the two other samples milled under O2 (x = 1.4 and 3.3). Lastly, the intermediate (CU2O + NiFe20 ₄ ) layer is thinner with a thickness of about 50 μ ιη compared to about 80 μ m to about 100 μ m for the three other electrodes. A possible explanation is that for x = 7.2, the Fe203 inclusions formed in the Cu-Ni-Fe matrix during the consolidation treatment are present in a too larger amount (Figure 4d), inducing an important chemical inhomogeneity in the sample in addition to produce an inadequate balance between the amounts of NiO and Fe203 to favor the formation of a protective NiFe204-rich surface layer as observed for x = 1.4 and 3.3. As a result, this electrode displays a lower corrosion resistance compared to x= 1.4 and 3.3, as discussed hereinbelow.

[0052] Figure 10 shows the evolution of the Cu, Ni and Fe contamination (wt. %) in the produced aluminum as a function of x in (Cu65Ni2oFei5)ioo-xO _x electrodes after 20 hours of electrolysis. Two additional compositions (x = 1.0 and 4.5) were evaluated in order to confirm the tendency of the Al contamination curves. All the contaminants display the same evolution with a minimum for x = 1.4 with the presence in the produced Al of 0.13, 0.08 and 0.03 wt. % Cu, Fe and Ni, respectively. This corresponds to an aluminum purity of 99.76 wt. %, which meets the chemical specification of P1020A grade Al (purity≥ 99.7%). It can be noted that (Cu ₆ 5Ni2oFei ₅ )ioo-xO _x electrodes with x = 1.0 and 3.3 also give good results with a total impurity level of about 0.3 wt.% compared to about 0.7 wt.% for x = 0. Further addition of oxygen (x = 4.5 and 7.2) induces an increase of the Al contamination with a total impurity content of about 0.6 wt. %. This highlights the fact that oxygen must be added in appropriate amount during the mechanical alloying process in order to induce the formation of finely dispersed Fe2C"3 inclusions in the Cu-Ni-Fe matrix during the subsequent consolidation treatment (see Figure 4), which is essential to favor the formation of a protective NiFe2C>4-rich layer at the surface of the electrode during the electrolysis process as shown before (Figure 9). This is also in accordance with the absence of significant improvement of the corrosion resistance for composite anodes prepared from a ball-milled mixture of (Cu6sNi2oFei5 + x Fe ₂ 0 ₃ ) or (Cu6sNi ₂ oFei5 + x NiFe ₂ 04) powders of micrometric iron oxide size, i.e. generally of at least 1 m due to the formation of too large Fe203 or NiFe ₂ 04 inclusions in size in such composite materials (results not shown). In contrast, a significant improvement of the corrosion resistance is observed for composite anodes prepared from a ball-milled mixture of Cu-Ni-Fe alloy + nanometric iron oxide particles, i.e. generally of at most 100 nm. For instance, aluminum with a purity of 99.5 wt.% was produced with a composite anode made from a mixture milled for 4 hours of 95.3 wt.% Cu67.iNi ₂ o.6Fei2.3 (pre-milled for 10 h) + 4.7 wt.% Fe7o03o (50 nm in size).

[0053] Electrolyte sampling was performed after 0, 1 , 2, 4, 15 and 20 hours of electrolysis for the (Cu6s i2oFeis)98.60i.4 electrode. The evolution of the Cu, Ni and Fe concentrations in the bath is plotted as a function of the electrolysis time in Figure 11. The electrolyte contamination strongly increases during the first 4 hours of electrolysis but tends to stabilize over this period. Thus, after 20 hours of electrolysis, steady-state conditions are assumed to be established with Cu, Fe and Ni concentrations in the bath of ca. 80 (0.008), 40 (0.004) and 5 (0.0005) ppm (wt.%), respectively. This stabilization indicates that equilibrium between the oxide dissolution rate at the anode and their reduction rate at the cathode or by Al is reached or that the Cu, Fe and Ni oxides have reached their saturation level in the bath. From the amounts of Cu, Ni and Fe impurities in the electrolyte and in the produced Al after 20 hours of electrolysis, the wear rate of the electrode was calculated according to equation (2) and is estimated at about 0.8 cm year ¹ , which is below the target of 1 cm year ¹ specified in the Aluminum industry technology roadmap, by the Aluminum Association, Aluminum industry technology roadmap, Washington DC (2003).

[0054] As people in the art will now be in a position to appreciate, there is provided an anode composition and a method of production thereof, for inert anodes of a high resistance to corrosion, with an erosion rate of at most 1 cm year ¹ , during electrolysis of aluminum at low temperature, i.e. at about 700°C. There are provided mechanically alloyed Cu-Ni-Fe-0 based materials for inert anodes.

[0055] A Cu-Ni-Fe alloy with a composition comprising Cu between about 65 and 70%.

Ni-Fe alloy with a composition comprising Cu in the range between 60 and 80 wt. %, Ni in the range between 20 and 30 wt. % and Fe in the range between 10 and 20 wt. % can be considered as appropriate for obtaining an anode with a good corrosion resistance after subsequent O doping in the range between 1 to 3% by weight. For example, an optimized composition was about 15% by weight Fe, about 64% by weight Cu, about 20 % by weight, and about 1.5% by weight oxygen.

[0056] Oxidization by grinding under oxygen atmosphere after an initial grinding under inert atmosphere to allow a proportion of oxygen between about 1 and 3% by weight is found to be efficient in increasing the anode corrosion resistance (see Figure 10). In this O concentration range, the size, dispersion and concentration of Fe203 precipitates in the consolidated powder are optimized to promote the formation of a protective NiFe204-rich layer on the anode surface.

[0057] As an alternative, Cu-Ni-Fe-0 based anodes with dispersed Fe ₂ 03, precipitates can also be produced by ball milling Cu-Ni-Fe alloy with nanometric iron oxide particles, i.e. of a size of at most 100 nm (see Figure 12).

[0058] With anodes of (Cu6sNi2oFei5)ioo-xO with x comprised in the range between about

1 and about 3, aluminum could be produced with a purity of 99.7 %. The rate of corrosion of the anodes is very low, at about 8 mm/year, which is well below current industry target of typically at most 10 mm/year. Moreover, the present anodes have good thermal, and mechanical stability, and low electric resistivity. They also have a stable potential and a low overvoltage for the reaction of oxygen, for example less than 0.4 V at 0.5 A/cm ² .

[0059] The addition of a small concentration, i.e. at most 5 wt. % and preferably at most 1 wt. %, of rare earth elements (such as Y or Ce for example) to the composition is expected to further increase resistance to corrosion (see for instance, works of R. Cueff et al in Corrosion Science 45 (2003) 1815-10831 ).

[0060] The consolidation procedure for producing nanostructured anodes from ball-milled

Cu-Ni-Fe-0 powders can be done through a cold pressing-sintering procedure as described hereinabove before. Other techniques characterized by their ability to produce nanostructured bulk materials or coatings from ball-milled powders, such cold spray or spark plasma sintering for example, can also be used.

[0061] In order to induce the formation of a protective NiFe20 ₄ -rich layer at the surface of the Cu-Ni-Fe-0 electrode before Al electrolysis and then, to prevent the formation of metal fluorides (e.g., FeF2) at the electrode surface during the first minutes of electrolysis, a pre-treatment of the electrode can be performed through an air oxidation step (e.g., oxidation under air atmosphere at 700°C for 3 h).

[0062] Although the present invention has been described hereinabove by way of embodiments thereof, it may be modified, without departing from the nature and teachings of the subject invention as recited herein.

[0063] References

B. Assouli, M. Pedron, S. Helle, A. Carrere, D. Guay and L. Roue, Light Metals, 1141 (2009)

S. Helle, M. Pedron, B. Assouli, B. Davis, D. Guay and L. Roue, Corros. Sci, 52, 3348 (2010)

S. Helle, B. Davis, D. Guay and L. Roue, J. Electrochem. Soc, 157, E173 (2010)

T.R. Beck, Light Metals, 355 (1995)

T.R. Beck, CM. MacRae and N.C. Wilson, Metali. Mat. Trans. B, 42, 807 (2011 )

US 5,284, 562 (A) Beck et al.

V. de Nora et al., Light Metals, 501 (2007)

V.A. Kovrov, A.P. Khramov, A.A. Redkin and Y.P. Zaikov, ECS Transactions, 16, 7 (2009)

S. Helle, B. Brodu, B. Davis, D. Guay and L Roue, Corros. Sci., 53, 3248 (2011 )

US 6,692,631 Bergsma et al.

US 7,077,945 Bergsma et al.

Aluminum industry technology roadmap, by the Aluminum Association, Aluminum industry technology roadmap, Washington DC (2003)

R. Cueff et al, Corrosion Science 45 (2003) 1815-10831