2.1. Materials

The cathode powders (LiCoO₂: LCO or LiCo_0.33Ni_0.33Mn_0.33O₂: NMC 111) were either obtained from MSE Supplies or synthesized through the ceramic route. X-ray diffraction (XRD) characterizations and chemical analyses of these powders are presented in Supporting Information Figures S1 and S2. The black mass (BM) was obtained from mobile phone batteries (Xiaomi model BN44), by discharging, opening in a glovebox, shredding in a cutting mill (Retsch SM 300), and sieving in a vibratory sieve shaker (Fritsch Analysette 3 Spartan) using a 100 μm grid size. XRD characterization of the sieved BM and inductively coupled plasma (ICP) analysis are shown in Supporting Information Figure S3. As graphite peaks are observed (ICDD file No. 65-6212), one can conclude that sieving was not efficient to remove graphite present in the BM. In fact, other separation techniques exist or are currently being developed to extract graphite from black mass before any processing, such as froth flotation [25, 26]. Furthermore, ICP and Energy-dispersive X-ray spectroscopy (EDX) analysis (Supporting Information Figure S4) indicate that the cathode active material is a NMC-type.

The process is performed in a Teflon-lined bomb provided by Parr Instruments (4744 model, 45 mL) or by Top Industrie (2148 5000, 500 mL or 2148 4000, 1 L), depending on the amount of solvent used. The solvent used is ethylene glycol (Merck 99%) in the presence of carbonate ions (e.g. potassium carbonate Alfa Aesar 99.0% min), with an amount corresponding to 1.2× the stoichiometric amount of carbonate ions necessary to form Li₂CO₃ from Li⁺ cations contained in the cathode powder. A pressure calculation based on the saturation vapor pressure of ethylene glycol and the air pressure at 225 °C, indicates that in the 45 mL reactor using 30 mL of ethylene glycol, the pressure is expected to be 3.9 bars (i.e. an overpressure of 2.9 bars). Table 1 summarises the operating conditions for the treatments performed in this study.

2.2. Experimental procedure

After introducing the cathode powder or the black mass, the reactor was filled with ethylene glycol in which potassium carbonate was previously dissolved. The reactor was then closed and heated to 225 °C for 10 h in an oven. This choice of experimental conditions comes from the conditions usually applied during polyol reduction treatments of metal oxides under solvothermal conditions (temperature higher than the boiling point of the solvent). After cooling, the reaction mixture comprises of both powder and liquid, and several steps of centrifugation, rinsing and magnetic separation (if needed) were carried out. These steps are illustrated in Figures 1a and b for the LCO and NMC cathode powder treatments, respectively. For LCO (Figure 1a), the solvent was first separated from the powder by centrifugation and stored for possible reuse. Then, the extracted powder was rinsed with distilled cold water while undergoing ultrasound dispersion, and separated again from water by a second step of centrifugation (P2). The rinse water was evaporated in an oven to recover a white powder (P1).

For NMC (Figure 1b), a subsequent step was added, in which the rinsed black powder was magnetically separated to isolate magnetic and non-magnetic fractions, each of which was dried. Magnetic separation was performed by mixing, under ultrasonic dispersion, the powder to be separated with water or ethanol in a beaker, and by approaching a Nd₂Fe₁₄B magnet close to the side of the beaker, which attracts magnetic part of the powder (this operation was repeated twice). The fraction that was not attracted to the magnet was a beige powder (P2′), while the fraction that was attracted to the magnet was a black powder (P3′).

2.3. Characterization methods

The crystallinity and purity of the obtained materials were characterized using an Empyrean diffractometer (PAnalytical) or a D8 diffractometer (Bruker AXS), using Co Kα radiations. The intensity was measured with a 2𝜃 step of 0.01°. Phase identification was done using DIFFRAC.EVA (Bruker AXS). Powders were observed by Scanning Electron Microscopy (SEM) equipped with an Energy-dispersive X-ray spectroscopy (EDX) detector (Zeiss 1530). Infrared spectroscopy analyses were performed with a Shimadzu IR affinity-1 spectrometer. ATR spectra of liquid samples were obtained using a GladiATR instrument equipped with a monolithic diamond by accumulating 32 scans in the spectral range from [4000–400] cm⁻¹ with a resolution of 4 cm⁻¹. Elemental quantification was performed by Crealins (Lyon, France) using inductively coupled plasma-atomic emission spectrometry (ICP-AES) iCAP 6500 Duo (Thermo Scientific) after acidic dissolution of the samples in the presence of sulfuric acid/nitric acid mixtures in an open environment at 220–240 °C. The relative error on the mass % of the elements is 5%.

3.1. Application of the process to LCO material

3.1.1. Use of virgin ethylene glycol

For this first experiment (LCO1), XRD analysis and Scanning Electron Microscopy results of P1 and P2 are presented in Figure 2. P1 powder (Figure 2a) consists of grains of few tens of micrometers and contains only one phase of lithium carbonate (monoclinic, ICDD file No. 22-1141). P2 powder (Figure 2b) contains particles of a few hundreds of nanometers which are formed through the aggregation of smaller particles. Phase analysis shows that two phases are present: a face-centered cubic (fcc) phase and a hexagonal close-packed (hcp) corresponding to metallic cobalt (ICDD files No. 15-0806 and 89-4308 respectively). This morphology and the coexistence of both phases is classically observed in the case of synthesis of cobalt nanoparticles by polyol way from cobalt salts [17, 27].

Chemical analysis of the residual EG, after separation from the solid by centrifugation, shows that it contains dissolved lithium, with an amount corresponding to approximately 30 wt% of the initial lithium present in the LCO compound.

At this stage, the mechanism involved in the process can be explained as follows: LiCoO₂ phase is destroyed by the joint action of EG which reduces Co²⁺ cations to Co⁰, and carbonate ions which extract lithium cations to form lithium carbonate, according to (1)

LiCoO 2 +CH 2 OHCH 2 OH+0.5CO 3 2- →0.5LiCO 3 +Co+[EG] ox · product

(1)

where [EG]_ox⋅product represents the possible oxidation products of EG as it will be discussed in Section 3.1.2.

The reduction is performed without any dissolution of cobalt cations, as cobalt was not detected in the residual EG. Lithium carbonate is partially soluble in EG, its solubility has been evaluated, under our solvothermal experimental conditions, to be approximately 3 g/L. After separation of the residual EG by centrifugation, the solid contains Co_(s) and Li₂CO_3(s). Li₂CO₃ is soluble in water, especially cold water (13.1 g/L at 20 °C and 15.2 g/L at 0 °C [24]), so it dissolves in water during the rinsing, while Co remains as a powder (P2). Lithium carbonate can be recovered by evaporating water, this is P1.

Recovery rates for lithium and cobalt are calculated according to Equations (2) and (3) where m_LCO treated is the mass of LCO treated (1.5, 15 or 30 g), m_P1 and m_P2 are the recovered mass of P1 and P2 powders, %mass(Li) is the Li content in P1 and %mass(Co) is the Co content in P2, these two values being determined by ICP analysis. The values 7.4% and 62.1% are the content of respectively Li and Co in the initial cathode powder obtained by ICP analysis (Supporting Information Figure S1).

Li recovery rate = %mass (Li)× m P1 7 . 4 % × m LCO treated × 1 0 0

(2)

Co recovery rate = %mass (Co)× m P2 6 2 . 1 % × m LCO treated × 1 0 0

(3)

Recovery rates of Co and Li (respectively 84.1% and 40%), and purities of Co and Li₂CO₃ (respectively 96.7% and 90.5%) are indicated in Table 2 for this first experiment. Concerning the recovery rate of lithium in the form of carbonate powder, it should be remembered that part of this carbonate is dissolved in EG (around 30%). This part will be recovered during subsequent solvent reuse (see Section 3.1.2). These values, especially for the recovery rate of lithium, are quite low, if one examines the new directive concerning battery recycling [3]. This value of around 70% is below the threshold to be reached for lithium (80%).

Table 2.

Recovery rates of Li, Co, Ni and Mn and purity of P1, P2, P2′ and P3′ powders, calculated from ICP-AES analyses of the different powders extracted

Sample	LCO cathode
	Recovery rate (%)					Purity (%)
	Lia		Co			Li2CO3	Co metal
LCO1	40.0		84.1			90.5	96.7
LCO2	69.7		99.1			95.2	99.9
LCO3	55.9		98.1			97.1	99.9
LCO4	101.3b		99.3			95.4	99.9
Sample	NMC 111 cathode
	Recovery rate (%)					Purity (%)
	Lia	Co	Ni	Mn		Li2CO3	NixCo1−x alloy	MnCO3
NMC	24	79	80.5	72		94.4	91	95.2

^aConsidering only P1 powder, without lithium dissolved in EG, ^bup to 100% due to the presence of Li ions in EG coming from the previous treatment.

As this first experiment was performed on a small amount of powder (1.5 g), unavoidable materials losses during the several steps (centrifugation, drying) have a strong influence on the recovery rates. The impact of these losses on the recovery rate is minimized by increasing the quantity of powder treated. The recycling process was performed on higher amounts of powder (15 g and 30 g, samples LCO2 and LCO3), with the same steps as described previously. As for the first experiment, a powder of Co was isolated, and a powder of Li₂CO₃ was recovered after evaporating the rinsing water. As it can be seen in Table 2, values of recovery rates and purity are greatly improved. The recovery rate of lithium in Li₂CO₃ form is between 55.9 and 69.7% with a purity between 95.2 and 97.1%. The recovery rate of cobalt is between 98.1 and 99.1%, i.e. identical for LCO2 and LCO3 taking into account measurement uncertainties, and with a purity of 99.9%.

3.1.2. Ethylene glycol reuse

In order to save EG and to recover more lithium, the solvent was reused (up to 4 times, but the result of only 1 time is shown here, sample LCO4). The phases obtained are the same that in Section 3.1.1, and the values of recovery rates and purities are indicated in Table 2. In this case, the recovery rate of lithium (calculated by comparing with the lithium amount present in 15 g of LCO treated) is higher than 100% because Li⁺ cations present in the solvent (and coming from a previous treatment of LCO) have precipitated as carbonate during the second treatment as saturation was reached more easily for this salt. This experiment also shows that EG is not degraded during the process, and that it remains efficient as a reducer for possible reuse.

In order to confirm the integrity of the ethylene glycol, infrared spectrometry was carried out on the solvent after extraction of the solid phase by centrifugation. Figure 3 shows the infrared spectra of virgin ethylene glycol, of the solvent after extraction of samples LCO1 and LCO4, and of a solvent after three successive reuses. The spectrum of the ethylene glycol presents absorption bands which correspond perfectly to what is expected [28]. After treatment, the solvent spectra still show all the absorption bands characteristic of ethylene glycol. Additional absorption bands appear at 1595, 1355 and 1125 cm⁻¹, whose intensities increase with the number of times the solvent is reused. The two bands located at 1595, 1355 cm⁻¹ could be attributed to asymmetric and symmetric elongation vibrations of the COO⁻ group in a glycolate (CH₂OHCOO⁻), or glyoxylate (CHOCOO⁻), both of which are oxidation products of ethylene glycol [29, 30, 31]. Doubt remains as to the attribution of the weak absorption band at 1125 cm⁻¹, whose wavenumber is slightly higher than already observed for these two compounds [29]. These results confirm that the reduction of Co(III) to cobalt metal during treatment does indeed involve a redox mechanism with ethylene glycol. The exact nature of the ethylene glycol oxidation product(s) remains to be confirmed. In addition, the solvent being in very large excess compared with the cathode compound (EG/LCO molar ratio = 36/1 in the case of LCO1, LCO2 and LCO3), it can be reused several times before being significantly oxidized.

3.2. Application of the process to NMC material

The process applied to NMC 111 powder, and following the scheme of Figure 1b, leads to the obtaining of 3 fractions of powder named P1, P2′ and P3′. XRD analysis and Scanning Electron Microscopy results of these 3 powders are presented in Figure 4. As in the treatment of LCO, P1 (Figure 4a) only contains Li₂CO₃, with the same morphology that was previously observed (grains of few tens of micrometers). P2′ powder (Figure 4c) consists in octahedral-shaped grains of around 10 micrometers and contains MnCO₃ as a single phase (rhombohedral, ICDD file No. 44-1472). P3′ powder (Figure 4b) mainly contains spherical particles with heterogeneous sizes between few hundreds of nanometers and several micrometers, which are formed through the aggregation of smaller grains. Some octahedral grains are also observed. Concerning P3′ powder, the XRD structural characterization shows the presence of diffraction peaks whose positions are between those of cobalt (ICDD file No. 15-0806) and nickel (ICDD files No. 87-0712), indicating the presence of a single-phase Ni–Co alloy. No hcp structure was identified, as usually observed while forming Ni–Co alloys by polyol way [17]. Peaks of MnCO₃ with a low intensity were also observed in P3′ due to an incomplete magnetic separation, which explains the observation of octahedral-shaped grains in the SEM picture (Figure 4b).

The mechanism explained in Section 3.1.1 can be completed as follows for the treatment of NMC. After extraction of Li₂CO₃ by its dissolution in water, the remaining powder contains MnCO₃ (which is insoluble in water [32]) and the Ni–Co alloy. As the alloy is ferromagnetic, it is attracted by permanent magnet, whereas MnCO₃ is not.

Recovery rates of Co, Ni, Mn and Li are reported in Table 2 (respectively 79%, 80.5%, 72% and 24%). The value for cobalt is of a similar magnitude as when treating an equivalent quantity of LCO (84.1%). The purity of lithium carbonate (94.4%) is also comparable to what was obtained for LCO (90.5%). The low recovery rate for lithium means that Li⁺ cations remain in EG, due to the precipitation of MnCO₃. These results show that, in EG, manganese carbonate precipitates preferentially and that it is therefore necessary to increase the amount of carbonate ions in order to improve the lithium recovery rate. The purity in the Ni–Co alloy (91%) is lower than for metallic cobalt in LCO, which can be explained by the presence of MnCO₃ in the alloy. The magnetic separation, which was performed by a very simple method, is the reason of this low value. The purity could be considerably increased by using more suitable techniques adapted for magnetically separating powders, such as high-gradient magnetic separator (HGMS) which creates strong magnetic forces developed in association with the intense fields delivered by a superconducting magnet [33, 34].

3.3. Application of the process to a black mass

Despite the presence of graphite in addition to the active cathode material, the recycling treatment was applied to this BM to ensure that the graphite did not affect the process. It should be noted that in this case, magnetic separation was not carried out, which leads to the isolation of only two fractions instead of three. The first fraction is obtained after centrifugation and rinsing with water, and should contain the Ni–Co alloy and manganese carbonate (mixture of P2′ and P3′) if we refer to the Figure 1b. The second fraction is recovered by evaporating the rinsing water (P1) and should contain lithium carbonate. XRD characterization of these two fractions is shown in Figure 5.

As expected, the powder fraction (P2′ + P3′) contains the alloy and MnCO₃, but it also contains graphite, which obviously was not affected by the process (Figure 5a). Concerning the powder P1, it contains Li₂CO₃, but this phase is not the main one (Figure 5b). Another phase, identified as Li₃PO₄ (ICDD file No. 25-1030) is present with a higher content by regarding its diffraction peak intensities. But the main phase (by intensity) is a third compound, identified as LiF (ICDD file No. 45-1460). The formation of these two phases could be due to the presence of PVDF and/or the electrolyte in the BM before the treatment, although we did not identify these two compounds by the techniques we used. Concerning PVDF, its degradation seems unlikely at the temperature applied in the present study. Indeed, PVDF undergoes thermal decomposition only in the range 400–500 °C under atmospheric pressure [35], and under supercritical conditions in water it decomposes only above 290 °C [11]. Furthermore, it was shown that the PVDF present on cathodes was not degraded when treated in ethylene glycol at a temperature close to its boiling point (197 °C) [36]. A recent study shows that the conversion of LiPF₆ into LiF and Li₃PO₄ is possible by water extraction using a hydrothermal way [37]. In this study, the authors show that the interest of treating LiPF₆ lies in the fact that this compound is an environmental pollutant, and that producing LiF and Li₃PO₄ is economically profitable. The possibility that the process developed in the present study could be another way of treating LiPF₆ needs to be supported by additional characterisation, particularly spectroscopic, in order to clearly highlight the presence of the electrolyte before treatment.