Recovery of cobalt from lithium-ion batteries using ﬂuidised cathode molten salt electrolysis

The future need to recycle enormous quantities of Li-ion batteries is a consequence of the rapid rise in electric vehicles required to decarbonise the transport sector. Cobalt is a critical element in many Li-ion battery cathode chemistries. Herein, an electrochemical reduction and recovery process of Co from LiCoO 2 is demonstrated that uses a molten salt ﬂuidised cathode technique. For the Li-Co-O-Cl system, speciﬁc to the experimental process, a predominance diagram was developed to aid in understanding the reduction pathway. The voltammograms indicate two 2-electron transfer reactions and the reduction of CoO to Co at − 2.4 V vs. Ag/Ag + . Chronoamperometry revealed a Faradaic current eﬃciency estimated between 70- 80% for the commercially-obtained LiCoO 2 and upwards of 80% for the spent Li-ion battery. The molten salt electrochemical process route for the recycling of spent Li-ion batteries could prove to be a simple, green and high-throughput route for the eﬃcient recovery of critical materials.


Introduction
Lithium-ion batteries (LiBs) have found varied use in portable energy storage devices [1 , 2] , power tools and electric vehicles, and have the potential for larger-scale stationary electric storage [3] . Compared with alternative battery chemistries, they possess high energy and power densities, long cycle lifespans and flexible operating conditions [4 , 5] . The use of LiBs is set to increase substantially with the electric vehicle revolution, driven by the need to decarbonise the transportation sector to meet global targets of reduced greenhouse gas emissions [6] . Although the growth of the energy storage market is a positive development for society, it brings with it a number of challenges; such as, the scarcity of raw materials like cobalt [7][8][9] , and the environmental footprint of production, towards which the extraction of these materials is a large contributor [10 , 11] . The recycling of LiBs, and the recovery of valuable materials, would provide a solution to this, and is an essential part of the future life cycle of energy storage devices.
There are currently five main recycling methods: pyrometallurgical recovery, physical materials separation, hydrometallurgical metals reclamation, direct recycling and biological metals recla- Fig. 1. Flowsheet of the recycling and regeneration process of spent LiCoO 2 -based batteries, proposed by Zhang et al. [34] , dashed red lines highlight the steps that could be forgone following the process described in this work.

Fig. 2. (a)
Schematic of the electrolytic cell used for the recovery process of Co from LiCoO 2 particles using a fluidised cathode [22] ; (b) oxide particles suspended in the electrolyte melt collide with the current collector and undergo reduction to the metal [23] ; (c) the particle/electrode/molten salt interaction sets up a three-phase interline (3PI) with minimal O 2 − ion build-up adapted from Deng et al. [45] . as a range of other studies which explore electrochemical methods and process optimisation and kinetics for leaching of metals and cobalt [39][40][41][42].
This manuscript describes how cobalt-containing particles [ 43 , 44] can be suspended in a molten salt fluidised cathode process [24] to produce cobalt metal from commercially-obtained and spent LiCoO 2 from LiB cells. LiCl-KCl eutectic has been selected as a suitable electrolyte, with high Faradaic efficiency, whilst eliminating the leaching and regeneration steps ( Fig. 1 ), with associated cost reductions. The decision to forgo the regeneration step was made based on the fact that the market share of LiCoO 2 cathode material is declining [12] ; thus, producing high-quality Co metal for other uses and to make other types of cathode materials (e.g. NMC) is considered to be advantageous. The fluidised cathode, which was previously developed by the authors for the production of tungsten, and for nuclear reprocessing applications, which include the formation of uranium [ 22 -24] , is a robust and dynamic system, in which kinetics are enhanced by the agitation of particles in the electrolyte melt. A schematic of the process is presented in Fig. 2 (a), (b) illustrates the pathway for metal oxide particle reduction; (c) describes the three-phase interline (3PI) of metal, metal oxide, and salt necessary for reduction reactions to take place.

The electrochemical cell
A tall-form beaker made from borosilicate glass (250 ml, Duran, SLS) was used as a primary container for both the molten salt system and metal oxide, in this case LiCoO 2 , throughout the experiment. This beaker was secured within a borosilicate flanged glass beaker (internal diameter 70 mm, outer diameter 120 mm, height 300 mm, GPE Scientific Limited). It was found that borosilicate glass proved a suitable material not only in withstanding these intermediate temperatures (~450 °C) but also in avoiding unwanted chemical interactions with the metal and metal oxide. The cell was heated using a Vecstar programmable furnace and was raised to 450 °C at a linear ramp rate of 100 °C h −1 . An inert high purity argon gas (99.998%, BOC) flowed through a ceramic tube (internal diameter 5 mm, Multi-lab, Quartz and Ceramic Technology) at a constant rate. The assembly of the cell was carried out inside an inert argon gas glovebox. The molten salts in this study comprised of 150 g anhydrous lithium chloride (Extra Pure, SLR, Fisher Scientific) and potassium chloride (Certified AR for Analysis, Fisher Scientific) in the ratio of 59-41 mol.%, respectively. Both salts were dried under vacuum at 150 °C for 24 h prior to use to ensure all residual moisture from the salts was adequately removed.

Electrodes and cathode materials
For the fluidised cathode setup, a tungsten rod (99.95% metals basis, 1.5 mm diameter, Alfa Aesar) was used with a glass sheath that had been fused around the shaft of the rod prior to experimental studies. The rod was modified because of the fluctuating height of the melt due to agitation. The working electrode was compared to an Ag/AgCl reference electrode which is known for its stable and well-defined potential [46] . Specifically, a silver wire (0.203 mm diameter, 99.95% metals basis, Alfa Aesar) was placed inside a glass tube with 1 wt% AgCl (99.999% trace metals basis, Sigma Aldrich) in 1 g LiCl-KCl (59 −41 mol.%). The auxiliary electrode, a high-density graphite rod (99.9995% metal basis, 3.05 mm diameter, Alfa Aesar), was inserted into a glass fritted (porosity 5) anode compartment with 10 g LiCl-KCl which contained a small aperture to allow any anodic gases (CO 2 from the oxide ion reaction with the graphite anode) produced to escape and to prevent the re-oxidation of any reduced cobalt particles. It should be noted that containment of the counter electrode behind a glass sheath and frit will add to the IR drop in the cell. However, this was not seen to affect the efficacy of the approach at this scale, for this high conductivity salt and the compliance of the potentiostat used. For scale-up to a practical system, further cell engineering will be required that includes optimisation of the working and counter electrode shape and placement, compartment separation and electrolyte agitation mechanism. An identical setup was used to conduct the 'static' experiments (quiescent electrolyte). In this setup, the LiCoO 2 particles were allowed to settle at the bottom of the crucible, the working and reference electrodes were immersed through them.
Spent commercial LiBs (Polymer Li-ion cell, 3.7 V, 400 mA h, AA Portable Power Corp) were also used in the trials to determine the efficiency of recovering Co from spent LiBs compared with commercially-obtained pure-phase LiCoO 2 (97%, Alfa Aesar). Initially, the Li-ion pouch cell was discharged using a battery cycler (BioLogic, BCS-805) to the manufacturer's safe discharge-limit (2.75 V). The pouch cell was then dismantled inside a glovebox ( < 0.5 ppm O 2 /H 2 O), where the cathode layer was separated from the anode and washed in dimethyl carbonate solution (DMC) for the initial removal of electrolyte. After this, the obtained cathode scrap was heated using a modular horizontal tube furnace (Carbolite) at 450 °C for 1 h under an inert argon atmosphere to remove the polyvinylidene fluoride (PVDF) binder. The cathode powder was then calcined at 800 °C in air for a further hour to burn off acetylene black and any residual organic additives. An agate mortar and pestle was used to manually grind the thermally-treated LiCoO 2 to a fine powder for further processing.

Procedure
Electroanalytical studies, including cyclic voltammetry and chronoamperometry, were undertaken using a computer-controlled potentiostat (Reference 30 0 0, Gamry instruments). The electrochemical procedure was carried out with a constant flow of argon gas, resulting in a homogeneous distribution of electrolyte and metal oxide particles. The product materials were characterised using a scanning electron microscope (Zeiss EVO MA10, accelerating voltage: 15.0 kV) and energy dispersive X-ray spectroscopy (Oxford Instruments, INCAx-act, PentaFET Precision). A three-electrode electrolysis setup was employed to perform electrochemical experiments and only absolute currents are reported to account for varying electrode surface area. The current collector was immersed ∼4 cm (1.90 cm 2 ) within the melt during both the fluidised and static measurements. An identical process was used for the recovery of commercially-available LiCoO 2 powder and spent Li-ion pouch cell, any differences between the studies are made clear in the manuscript.

Thermodynamic analysis
To help understand the reduction pathway of LiCoO 2 in a LiCl-KCl eutectic system a predominance diagram [47][48][49][50] was generated for the Li-Co-O-Cl system. Predominance diagrams are analogous to Pourbaix diagrams; however, the former compare the standard electrode potential to the negative logarithm of O 2 − ion activity, pO 2 − rather than pH. All thermodynamic data used in producing the diagram has been obtained from HSC Chemistry 6.0 database. The diagram in Fig. 3 indicates the stability for different cobalt oxides (II/III) and cobalt metal at 500 °C. From the thermodynamic analysis, one can assume that the lithium in LiCoO 2 ionises with the rest of the ions in the LiCl-KCl melt, as per Eq. (1 ), and that it would require a relatively high (negative) potential to electroplate, in the same manner as the Li + from the fused salt. From the diagram, we may also deduce that a greater concentration of O 2 − ions in the eutectic melt will retard the reduction process of Co 3 O 4 to Co metal. As the O 2 − ion activity fluctuates throughout the experiment, the required reduction potential may also vary. It is therefore important that the O 2 − ion concentration is kept as low as possible. With respect to the reduction of Co 3 O 4 , there are two steps in the formation of Co metal. These steps are outlined in Eqs. (2 ) and (3) ; Eqs. (4 ) and (5) represent how the standard equilibrium potential, E , was calculated, where G is the Gibbs free energy of the reaction, n is the number of moles of participating electrons, F is the Faraday constant, R is the universal gas constant, and T is the temperature.
The thermodynamic analysis stipulates that the electrochemical reduction process for cobalt oxide involves two reactions, Eqs. (2 ) and (3) , and that the O 2 − ion activity level ultimately has little to no effect on the reduction procedure, as cobalt metal can be produced, with adjustment to the potential applied, at all oxide ion activity levels.

Cyclic voltammetry
Cyclic voltammetry measurements were recorded using a tungsten working electrode on a static setup first; the voltammogram at 450 °C is presented in Fig. 4 (a). This set-up was initially used to enable the retrieval of high signal-to-noise data showing clear reaction peaks and to understand the electrochemical system before agitation using a fluidised cathode setup. The working electrode was swept between 0 and −2.7 V at a scan rate of 50 mV  4 g) and (c) single cycle for commercially-obtained LiCoO 2 powder (5 g). The difference in loading results in no Li deposition and stripping peak being observed this is due to a 'shielding' effect caused by the higher particle loading [23] . Temperature: 450 °C; scan rate: 50 mV s −1 ; argon gas flowrate: 200 cm 3 min −1 . s −1 thereby avoiding the salt's decomposition potential ( ∼−2.8 V); three potential cycles were performed. The coupled redox potentials (described in Eq. (2 )) at c1 and a1 represent the reduction of Co 3 O 4 to CoO and the reoxidation of CoO back to Co 3 O 4 . The second coupled redox potentials (described in Eq. (3 )) at c2 and a2 represent the second reduction step where CoO is converted to Co metal and Co is reoxidised back to CoO, respectively. The general form of the electrochemical response is consistent with previous results [34] ; the difference in redox potentials is due to kinetic limitations as the temperatures and molten salt eutectics used in this system (750 °C vs. 450 °C used here) are different to others.
Next, cyclic voltammetry measurements were performed on the fluidised cathode arrangement, this is presented in Fig. 4 (b) and (c). The tungsten working electrode was scanned between 0 and −2.55 V in Fig. 4 (b) and from 0 to −2.7 V in Fig. 4 (c), and then back to the starting potential (vs. Ag/Ag + ). Again, the coupled redox potentials at c1 and a1 represent the reduction of Co 3 O 4 to CoO, and its reoxidation, as described in Eq. (2 ). The second coupled redox potentials at c2 and a2 represent the reduction of CoO to Co, and its reoxidation, as described in Eq. (3 ). The fluidised cathode system reveals an interesting electrochemical feature, the apparent 'noise' in the cyclic voltammetry has been associated with particle/current collector reactions, as has been observed in previous studies in the fluidised cathode system [23] . These collision-reaction features (i.e. illustrated by a 'noisy' system) are present at potentials where Co 3 O 4 reduction takes place and may extend into the decomposition potential of the Li salt, as shown by Fig. 4 (c). Upon observing the anodic sweep this phenomenon does not occur; therefore, the re-oxidation of Co metal to Co x O x is primarily a surface-based process. In this system the electrochemical process proceeds from approximately −0.33 to −0.52 V (vs. Ag/Ag + ) in Fig. 4 (b) and (c), respectively.
When comparing the voltammograms of the static arrangement against the fluidised cathode setup, there is significantly more current that is passed when the reaction mixture is fluidised, where the current collector size grows as the reaction proceeds, although some spalling occurs. This is to be expected as the current is influenced by the rate of collision and reactions become more favourable and frequent, when compared to a solid precursor (pellet) or a still (quiescent) set-up.

Constant cell voltage electrolysis
Plotting current versus time curves provides an overview of the reduction pathway assumed in a given electrochemical process. For this study, the LiCoO 2 cathode material was subjected to a fixed potential of −2.4 V (vs. Ag/Ag + ). This potential was chosen using the data acquired from Fig. 4 , and a value of −2.4 V, represented the most likely reduction potential to LiCoO 2 . The reduction potential is more positive than the expected decomposition potential of the salt and should therefore avoid Li electroplating. As observed from Fig. 5 (a) and (d), the current increased, quickly approaching  nearly 80 mA in Fig. 5 (a) and 90 mA in Fig. 5 (d) due to the deposit growth of Co metal on the substrate electrode surface. One explanation for the current increase, which has been evidenced in previous work [22] , is that it could arise due to the difference in particle size distribution which means some particles undergo partial reduction. This enables a higher probability of contact between the metal oxide and salt and improves the reduction rate. Fig. 5 (c) shows a photograph of the product after it was washed with distilled water compared with the starting material photographed in Fig. 6 . The final product appeared to be a homogeneous grey-black powder, indicative of cobalt metal, and both SEM/EDS analyses, in Fig. 7 (a)-(c), and the EDS spectrum in the Supplementary Material, indicated that oxygen was absent in the final Co product. At higher magnifications, Fig. 7 (a) and (b), the Co metal particles produced appear homogeneous in size and porosity, depicting a 'corallike' structure, which indicate a feature size of 2-3 μm. The same is true of Fig. 7 (c) although there is also some presence of cobalt silicate particles which are assumed to be due to the reaction of borosilicate glass with the cobalt ions in the molten salt system. The current versus time curves for both studies can be separated into two segments. The first segment is associated with a rapid rate of reduction and is observed in Fig. 5 (b) and (d) where the current rises. This rise in current is a result of the increasing electrode surface area. The second segment is where the current begins to fall. Fig. 5 (d) is where the periodic spalling is most noticeable and is denoted by a rapid loss of current for short periods towards the end of the process. The depletion of available oxide from the reaction is realised once the current reaches the background level or zero. From calculations, 1640 C has been deter-mined as the total theoretical charge that would be required to reduce the LiCoO 2 (5 g), assuming 100% current efficiency. Acknowledging incomplete reduction and extrapolating towards the background current, resulted in a current efficiency between 70% and 80% in Fig. 5 (a) and was upwards of 80% in Fig. 5 (d) [22] .

Conclusions
This work demonstrates that the recovery of Co metal is feasible using the fluidised cathode technique from commerciallyobtained and spent LiCoO 2 cathode material. It serves as an important step for the electroreduction of metal oxides to their elemental state. The results from the chronoamperometry reveal that the fluidised cathode process possesses a high Faradaic efficiency of > 70%. By using this process, several steps in the flowsheet presented in Fig. 1 can therefore be eliminated; namely the leaching and regeneration steps. It also utilises a molten salt eutectic (LiCl-KCl) at a much lower temperature (450 °C) leading to considerable cost reductions overall. The decision to forgo the regeneration step was made on the fact that the market share of LiCoO 2 cathode material is falling; therefore, producing high-quality pure-phase Co metal is practical for other uses which include the formation of lithiated cobalt-oxides in LiBs (e.g. NMC and NCA). Given the success of this work, the authors envision this technology to be applied to the aforementioned battery electrode materials, with the possibility of selectively reducing and separating different materials, other components of a battery and other types of batteries.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement
This work was supported by the STFC Network in Batteries and Electrochemical Devices (ST/R006873/1). Shearing acknowledges financial support from the Royal Academy of Engineering ( CiET1718/59 ). The authors also acknowledge support from the EPSRC ( EP/R023581/1 ).

Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.electacta.2021.138846 .