Recovering Lithium Cobalt Oxide, Aluminium, and Copper from Spent Lithium- ion Battery via Attrition Scrubbing

In this manuscript, the results show that the single-stage liberation by using a cutting mill is sub-optimum. From the analysis, that the size fraction of < 850 μm only recovers 43.7 wt% LiCoO2. With the recovery of 9.0 wt% aluminium and 10.6 wt% copper the remainder of the copper being in the > 850 μm size fraction. The low recovery of LiCoO2 is caused by the particles that are still adhering on to the surface of the aluminium current collector. This lack of liberation prompted the use of attrition scrubbing as a secondary stage of mechanical treatment. 2.5 min Attrition scrubbing improves the selective liberation of cobalt towards aluminium and copper by 36.6 % and 42.6 % respectively. Attrition induces abrasion and it is shown to liberate the LiCoO2 particles. Results show a minimum of 80 wt% LiCoO2 particles can be recovered in the size fraction of < 38 μm with 7.0 wt% aluminium


Introduction
Lithium-ion battery (LIB) technology has become the dominant energy storage for many consumer electronics and electric grids (Blomgren, 2017;Dunn et al., 2011). Despite the advancement of battery technology, present LIBs meet most of the requirements dictated by the large volume of the application linked to renewable energy and electric transportation field (Winslow et al., 2018). The spent LIBs from electric vehicles will emerge as the future waste problem with at least 25 billion units and 500 thousand tonnes of spent LIBs would be generated by 2020 (Richa et al., 2014;Zeng et al., 2014). When considering the natural scarcity and the demand projected for materials used in LIB production, cobalt is the most critical material as the demand for the future types of LIB is likely to contain embedded cobalt (Zubi et al., 2018).
In recent years, much research has been focused on developing efficient recovery methods for the materials found in spent LIBs. With the positive electrode active materials as the main targeted component as it is where the incentive of LIBs recycling come from (Gaines, 2018). Current research to recover positive electrode active materials are focused on leaching processes (Li et al., 2018).
The components that makeup LIBs can be generalised into two major components, leachable and non-leachable. This is based on whether it can be dissolved or deconstructed to its elemental form during leaching. The positive electrode active materials, iron, and current collectors are of the leachable components. Whereas, other components such as graphite and polymeric materials are non-leachable components. The current practice is to obtain the positive electrode active materials via manual dismantling Roshanfar et al., 2019;Yu et al., 2019), which is not practical on an industrial scale. In the mechanical treatment of spent LIBs leachable contamination by iron, copper, and aluminium are expected. The challenge in LIB recycling is to produce a positive electrode active material concentrate (LiCoO2) that is suitable for the hydrometallurgical process without involving manual dismantling of LIB cell. Impurities such as iron aluminium and copper in the leach liquor can be effectively precipitated by adjusting the pH value between 4.5 (Joo et al., 2016) to 5.5 (Chen et al., 2011) using NaOH. Sa et al. (2015) argue that the removal of Cu 2+ is more difficult than the removal of Al 3+ and Fe 3+ due to the higher solubility constant of Cu 2+ and led to the study of how the presence of copper may affect the performance of LiNi1/3Mn1/3Co1/3O2 (NMC) positive electrode active materials. The results suggest that 5 wt% copper impurity is acceptable for NMC battery (Sa et al., 2015).
The mechanical treatment of LIBs has been reported to be a selective phenomenon (Widijatmoko et al., 2020). The positive and negative electrode active materials can be concentrated in the finer size region without over crushing of other battery components in both wet and dry grinding (Zhang et al., 2013). The occurrence of selective liberation can then allow sizebased separation to be carried out. The sieve size of the acts as the cut point to concentrate the positive electrode active materials. The positive and negative electrode active materials are concentrated below the cut point.
Whereas, the copper, aluminium, and iron are predominantly found above the cut point. To concentrate positive electrode active materials, the cut point reported varies from 250 µ m (He et al., 2017) to 2000 µ m (Li et al., 2009). The smaller cut point of 250 µm has been reported to give high purity of positive electrode active material, but it only recovers 56.38% LiCoO2 (He et al., 2017). Moreover, the cut point size of 250 µ m is substantially greater when compared to the positive electrode active materials powder size found in LIBs (1.50 µ m -7.80 µ m) (Pavoni et al., 2018). This is due to the active materials are still aggregated or attached to the current collectors (Widijatmoko et al., 2020).
The occurrence of selective liberation also depends on the comminution technique being used (Hesse et al., 2017). Different techniques may result in different size distributions due to the predominant force acting during comminution as well as the milling conditions being employed (Gao and Forssberg, 1995). Selective liberation occurs when the breakage of a component is dependent on physical and mechanical properties (Mariano et al., 2016). Hesse et al. (2017) demonstrate that different predominant load being applied to a mineral would result in different liberation selectivity.
Considering the LIBs assembly, the active materials cast on the surface of current collectors can be scoured and liberated from the current collectors.
Therefore, the use of liberation technique based on impact and abrasion is hypothesized to be a suitable method to promote selective libation of positive electrode active material. The use of attrition scrubbing is proposed to liberate the positive electrode active materials while minimising the breakage of copper and aluminium components.
Attrition scrubber has been designed to induce impact and shearing action between particles that promote surface abrasion and produces fine particles (Bayley and Biggs, 2005). Attrition scrubbing is conventionally used to upgrade minerals by removing surface impurities such as sand for glass making (METSO, 2018) and shown to be applicable for environmental remediation purposes such as the decontamination of storm water sediment (Petavy et al., 2009).
The original contribution of this work is related to the application of attrition scrubbing in the liberation of LiCoO2 particles. The novel use of inert silica sand media as the abrasive allows the liberation LiCoO2 which then allows for the concentration of LiCoO2 in the finer size region. The LiCoO2 product contains low copper contamination that is below the reported maximum tolerable contamination when undergoing hydrometallurgical processes. This paper will firstly describe the cause of poor liberation of LiCoO2 when using a single-stage size-reduction. The proof of concept of using attrition scrubbing as a secondary liberation technique is then discussed and the mechanism is conveyed. Additionally, the breakage kinetics of LiCoO2, copper, and aluminium are studied to understand the relative breakage of different components. Finally, a demonstration for the use of electrostatic separator to recover the copper and aluminium current collector is presented.

Spent LIBs sample
The spent prismatic LIBs used in this study were collected from local electronic repair shops in Ningbo, China. Only LIBs containing LiCoO2 as positive electrode active materials were used. The type of positive electrode active materials was later confirmed as LiCoO2 using an X-ray Diffraction (XRD, Bruker-AXS D8 Advance) and the result is presented in Figure 1. All spent LIBs were firstly discharged by connecting to a 56-ohm resistor until the voltage is below 0.3 V to render them safe.
The spent LIBs were crushed using a cutting mill (Retsch SM2000) with 8 mm grid. The samples then dried in an oven at 80 o C until a constant weight was achieved to remove the volatile organic electrolytes. The bulk dried sample was then split into aliquots by using a static rifle with chute size of 31 mm x 160 mm with 16 alternating chutes.
The representative samples were then screened for ferromagnetic materials by using a cylindrical rare earth magnet enclosed in polyvinyl chloride (PVC) pipe. The ferromagnetic materials were found to be less than 2 wt%.

Attrition scrubbing experiment
Attrition scrubbing experiments were carried out using a WEMCO 1L labscale attrition scrubber with a constant impeller speed of 1000 rpm. Clean low iron silica sand in size range of 2360 µ m -850 µ m was used. In this study, the pulp density of 70 wt% with 10 wt% ratio of spent LIBs to silica sand media is used. The attrition time then varied from 2.5 min to 20 min.
Following attrition scrubbing, the product was wet sieved using 5 L water.
Since the LiCoO2 particles found in LIBs are in size range of 1.5 µ m -7.8 µ m (Pavoni et al., 2018), a 38 µ m sieve was used as the cut point. Moreover, to prevent damage to the 38 µ m sieve, the attrition product was firstly sieved by using a 212 µ m sieve.
Following wet sieving, the products were dried in an oven at 80 o C until a constant weight was achieved.
The size fraction of > 212 µ m was further dry sieved into different size fractions. In this study, the attrition products were sieved into the size

Elemental analysis
The elemental analysis is adapted from BS EN 62321-5:2014 (Standard, 2014). It is important that the sample tested is a representative of the entire aliquots and particle size of <250 µ m is suggested in the standard. To fulfil this requirement, calcination was firstly carried for particles with a size greater than 212 µ m to remove polymers that are difficult to mill. The calcination was carried out in multiple stages to prevent a sudden release of gas with the final stage at 500 o C for 3 hours. The samples are allowed to cool down to room temperature and milled using a centrifugal mill (Retsch ZM200) with 0.25 mm grid. All product was sieved with a nominal aperture size of 212 µ m. The size fraction of >212 µ m was re-milled until the mass recovery rate of <212 µ m is greater than 95 wt%. Approximately 0.2000 g ± 0.010 g of sample were weighed to four decimal places using an analytical balance.
The digestion is carried out in multiple stages of acid addition. The first stage The sample then analysed using an Inductively Coupled Plasma -Mass Spectrometry (ICP-MS, Nexion 300x).

Morphology observation
Scanning Electron Microscopy -Energy Dispersive X-Ray (SEM-EDX, Zeiss-Sigma-VP). The backscattered detector allows a different compound to be identified based on the average molecular weight. The heavier average molecular weight being brighter than the lighter average molecular weight.
The backscattered detector was used to observe the morphological characteristics with the EDX identifies the different element present. Prior to morphological analysis, the surface is mace conductive by applying 4 nm gold layer by using gold sputtering machine (LEICA EM SCD 500).
There are currently two types of binder that are widely used; which are PVDF and SBR-CMC. The morphology analysis with EDX shows the presence of fluorine atoms and the absence of sodium atoms. This observation suggest that the binder used in positive and negative electrode is PVDF.

Electrostatic separation experiment
The electrostatic separator allows the separation of materials based on the difference in surface conductivity or by the preferential charging and attraction materials to an electric field of opposing charge potential (Kelly and Spottiswood, 1989). The electrodynamic mode involves the use of ionizing electrode and static electrode, whereby all particles receive a positive or negative charge. The separation then occurs by leakage of this assumed charge by the conductive materials compared to the retention of charge by the non-conductors. Thus, the electro-dynamic mode is able to produce fractions that concentrate non-conductive, middling, and conductive components (Kelly and Spottiswood, 1989). In this manuscript, a roll-type electrostatic separator (Carpco,HT (15,25,36)) was used to separate the attrition products > 38 µ m. The size of the static electrode is 71.5 cm X 12.5 cm X 5 cm (length x width x height), ionization electrode is a 0.010 mm wire of 10 cm in length, and roll radius of 12.7 cm. The experiment was performed at air relative humidity of 40% -50% and temperature of 20 o C. The feeder vibration was maintained constant at 30% of its maximum power.

Characteristics of spent LIBs sample
Three representative samples that were crushed using a cutting mill only were classified. The average particle diameter (d50) was found to be 1.5 mm.
Each different size fraction then subjects to elemental analysis for the desired element content. The size-based recovery rate of the key elements is presented in Figure 1.     suggests that the initial liberation of spent LIBs by using a cutting mill only leads to a sub-optimum result.

Attrition liberation proof of concept
The suitability of an attrition scrubber as a second liberation stage to selectively liberate LiCoO2 particles is initially measured by assessing the attrition from a short duration of 2.5 min attrition time with 70 wt% pulp density and 10 wt% of LIBs to silica sand ratio. The product was then sieved to produce different size fractions and digested for elemental analysis as previously described in the experimental method. The aluminium, cobalt, and copper elements were detected, and the size-based recovery rate is presented in Figure 4.    Otherwise, the recovery curve below the diagonal line indicates the enrichment of valuable component in the bigger size fraction.
From Figure 5, the first stage of liberation using a cutting mill does induce selective liberation of cobalt in the finer size region. The result suggests that selective liberation of LiCoO2 in the finer size region is improved using an attrition scrubber. This is indicated by the recovery line of Co-Al and Co-Cu that are further from the diagonal following attrition. Furthermore, the Fuerstenau plot which takes account of the interaction between aluminium and copper (Al-Cu) is also shown in Figure 5 indicates that the aluminium is concentrated in the larger size fraction relative to copper.
From the Al-Cu plot shown in Figure 5, the result suggests that there is a minimum improvement in the enrichment of the two components based on size. Therefore, the use of size-based separation to separate copper and aluminium is a challenging task. There is only a slight increase in the selective liberation efficiency between aluminium and copper by 2.2 %. Thus, the separation of aluminium and copper based on size for attrition scrubbing product is still unlikely.
From the proof of concept laid out, the initially sub-optimum liberation of LiCoO2 particles using cutting mill only can be further improved by the use of attrition scrubbing as a second stage for liberation. The attrition products concentrate the majority of LiCoO2 particles in the fine size region of less than 38 µ m in 2.5 min attrition time. Furthermore, it is also expected that the graphite is also concentrated together with the LiCoO2 particles due to the weaker attachment of graphite laminate onto the copper current collector as compared to the LiCoO2 laminate counterparts (Dai et al., 2019).
This weaker attachment of the negative electrode active materials may also help to explain the Fuerstenau Al-Cu curve that is below the diagonal line.
The LiCoO2 laminate may help in maintaining the aluminium current collector shape and preventing further breakage. Whereas, the copper current collector may not have this benefit. Morphology observation of attrition product is presented in Figure 6. The weaker attachment of graphite towards copper current collector as compared to the aluminium counterparts can be clearly seen by comparing Figure 6a to To confirm that the copper and aluminium contaminating the surface of the larger particles, elemental mapping by using EDX was carried out for the attrition product of < 38 µ m. The EDX-elemental mapping was set to detect the cobalt, silica, copper and aluminium elements and the results of the elemental mapping is presented in Figure 7. From the observation, it was found that some of the graphite, LiCoO2 and SiO2 particles are contaminated by copper and aluminium fine particles. From the elemental mapping results, it is understood that further mechanical separation of copper and aluminium from this powder may be challenging.
Comparing the particle morphology of before and after attrition, the impact and shearing load appears to liberate the active materials that laminate the and aluminium current collector is deduced to be slower than that of the active materials. To confirm this, the study related to the breakage kinetics of LiCoO2 laminate as compared to the copper and aluminium current collector was carried out.

The breakage kinetics and its implication
The pulp density and the LIBs to silica sand ratio were kept constant at 70 wt% and 10 wt% respectively. Samples from the size fractions 4750 µ m - been reported to be a first-order (Sadler III et al., 1975). The rate of disappearance, by breakage, from a given narrow size, is given by Equation Where w is the weight of material in the given size fraction, t is time, and k is milling rate constant for the given size fraction. k is, in general, different for each size fraction present and is dependent on operating parameters, mill design, the material being milled, and the environment inside the mill.

Equation 2
Where wo is the initial amount of material present in the specific size range.
Equation 2 Figure 8c. Similarly, to the phenomenon described for Figure 8a, the aluminium breaks slower than that of copper.
From Figure 8d, it can be seen that the LiCoO2 laminate breakage rate towards the cut point of 38 µ m is faster than that of copper and aluminium.
The gradient of the cobalt breakage kinetics is much higher than that of copper and cobalt. The substantially faster breakage of LiCoO2 laminates towards the cut point of less than 38 µ m is expected. The LiCoO2 particle size range found in spent LIBs is between 1.50 µ m -7.80 µ m (Pavoni et al., 2018). This indicates that the liberation of LiCoO2 particles that are still held together by the binder is much faster than the fines produced by the breakage of aluminium and copper.   µ m is calculated based on the ratio of cobalt to graphite in LiCoO2 batteries from the published data by . By this way, the concentration of the silica sand in the < 38 µ m attrition product can also be estimated and summarised in Table 2. This is attributed to the contamination caused by the silica sand media as attrition time increases. This is caused following the breakage of silica particles. The contamination from the attrition media is expected, and low iron silica sand has been chosen for it is chemically resistance to the lixiviant which has been proposed by researchers to leach the LiCoO2 particles.
Therefore, the main leachable contamination of attrition products is copper and aluminium. From previously reported literature, a proportion of 5 wt% copper relative to LiCoO2 is a tolerable contamination for leaching and resynthesizing (Sa et al., 2015). Aluminium can initially be removed via dissolution by using NaOH. From Table 2, the 20 min attrition time results in only 3 wt% copper relative to LiCoO2. Therefore, the attrition product can be concluded to be suitable for subsequent hydrometallurgical processes.
Remaining graphite can be separated from silica sand by using froth flotation due to the hydrophilicity difference (i.e. graphite is hydrophobic and silica sand is hydrophilic) (Lu and Forssberg, 2001). Diekmann et al. (2017) have proposed a second stage liberation by using a cutting mill, the active material fine particles concentrate were recovered by using air classifier, this method resulted in 75% recovery. Comparing the method proposed, higher recovery rate of active materials from second stage wet liberation by using an attrition scrubbing can be found 80.0 wt% -89.8 wt%. Moreover, the use of air classifier to recover spent LIBs active materials may impose serious hazard related to respiratory health.

The separation of attrition scrubber > 38 µm product
The majority of the LiCoO2 particles are concentrated in the size fraction of less than 38 µ m. The size fraction > 38 then concentrates the copper and aluminium current collectors. The > 38 µ m sample obtained then subject to the electrostatic separation to separate polymer, silica sand and the current collector. A roll-type electrostatic separator (Carpco,HT (15,25,36)) was used in this study.
The electrostatic separator comprises of two main components which are the beam and static electrode. The ionizing electrode pinned nonconductive materials on to the roll and collected at the end of the roll by a static brush.
For the particles that are heavier than the pinning action are collected as middlings. The static electrode attracts conductive materials while leaving the non-conductive to fall through the roll and collected. Considering the composition of the sample, both electrodes were used to separate the sample into three different fractions, such that the polymeric separator and fine silica sand would be pinned on to the roll and collected on the left hand side of the receptacle, the voltage is adjusted so that the silica sand > 850 µ m would not be pinned as strongly as the polymeric separator and collected as middling, and the conductive materials are thrown into the right side of the receptacle.
The optimum conditions reported by Silveira et al. (2017) were initially used and need to be adjusted for a workable parameter. The particles were flowing unevenly and arching between the electrodes towards copper and aluminium foils was observed. Moreover, that the silica sand with particle size of > 850 µ m was strongly pinned onto the roll and can be found in the non-conductor fraction. To overcome these issues, adjustment trial and error based on visual inspection was carried out to adjust the parameters and minimise these issue. The key parameters of the electrostatic separator are shown in Table 3.  In a one-pass electrostatic separation, only copper and aluminium foils are registered as a conductive fraction. However, it was observed that the middling still contains a substantial amount of conductor materials and therefore the middling was re-introduced into the feeder. The middling was re-introduced to the feeder for five times. To assess the grade of the resulting separation, manual picking was done. The schematic diagram of the electrostatic separator, as well as the resulting separated products, are shown in Figure 10. The results from this exploratory study suggest that electrostatic separator is a useful technique in separating the current collectors and polymeric materials from the attrition media. It is also observed that the attrition media that has undergone size reduction < 850 µ m are registered as nonconductive and collected together with the polymeric materials. The recovered attrition media are visually clean from the separator and with minimum contamination from copper and aluminium, and therefore can be re-used for the subsequent attrition media.

Conclusion
Mineral processing technique is an important part in LIBs recycling to liberate and concentrate valuable metals due to its high throughput (Al-Thyabat et al., 2013). The recovery technique proposed in this research allows the majority of LiCoO2 particles to be recovered in the size fraction of < 38 µ m with minimum contamination from copper and aluminium components and therefore reducing the need for leachate purification during hydrometallurgical process. This is achieved by using attrition scrubbing as the second stage of liberation.
From the morphological analysis of particles before attrition, above and