Enhancement in Lithium Recovery from Spent Lithium Batteries by Nanofiltration Membranes

Abstract

The recovery of lithium from extracts obtained from a black mass of spent lithium-ion batteries treated with a ternary solvent system at acidic pH was investigated using flat-sheet nanofiltration (NF) membranes operated according to a dead-end configuration. Specifically, four samples obtained at different pH values (2.5 and 5) and extraction times (48, 96 and 168 h) were treated in selected operating conditions by using two commercial polymeric membranes (denoted DK and HL, with an approximate molecular weight cut-off of 150–300 Da) up to a volume reduction factor (VRF) of 4. Membrane performance was assessed in terms of productivity and selectivity towards specific ions, including lithium. For most treated samples, the HL membrane exhibited higher permeate fluxes in comparison to the DK membrane. However, the DK membrane performed better in terms of lithium rejection than the HL membrane, with a negative rejection at VRF 4 observed for all treated samples. More than 90% of multivalent ions were rejected by both membranes independently of the VRF. The membrane ability to retain multivalent ions led to their progressive concentration in the retentate as the VRF increased. The extraction time did not impact the NF performance of both membranes in terms of ion rejection. For the DK membrane conditions of extraction of 96 h and pH 5 represented the best trade-off between flux, ion rejection, and total lithium recovery.

1. Introduction

The world is currently facing two big challenges: meeting the ever-increasing demand for energy and doing so without relying on fossil resources in order to avoid climate changes problems. The adoption of renewable energies has become a feasible solution to overcome both these issues. Electric vehicles (EVs) are good example of how the scientific community has managed to overcome the use of conventional fossil fuels [1]. The expansion of this market, along with the widespread use of electronic devices, has strongly contributed to the huge spreading of lithium-ions batteries (LIBs). These batteries play a key role in the electrification transition in different industries [2]. Nevertheless, the increase in lithium demand has intensified the utilization of lithium natural resources. Lithium resources are located in a few specific geographical areas (Figure 1), and they are not sufficient to fulfil the present and future needs. This leads to the necessity to recover this valuable metal from other sources, such as exhausted batteries [3]. In addition, if spent LIBs are discarded without proper treatment, heavy metals like Co, Ni, and Mn would contaminate soil and underground waters, leading to related environmental and safety problems [4]. Therefore, the recycling process of LIBs, within the logic of the circular economy, is particularly beneficial as it reduces the demand for the extraction of new resources [5]. Hydrometallurgy, pyrometallurgy, and direct recycling are the three main pathways used to recycle spent LIBs [6].

LIBs are mainly made of four parts: anode, cathode, separator, and electrolyte. All of them are closed up in an aluminum case covered by a plastic label. Their recycling process starts with the discharge that is carried out by exploiting an electrolyte solution [7] to avoid short-circuit current flows when the spent LIBs are dismantled. The following step is dismantling, which is carried out manually, with knives and saw, or automatically for certain kinds of LIBs, to separate the individual spent cells. Separation steps include mechanical, physical, thermal, chemical, and mechano-chemical processes. These pre-treatment steps aim at enriching the metallic fraction, lowering scrap volumes and energy consumption, enhancing the recovery rate [4]. They are also useful before hydrometallurgy because the leaching process is hindered by the presence of impurities [6]. Pyrometallurgy employs high temperatures to recover and purify valuable metals through both physical and chemical transformations [6,8,9]. The main goal of this kind of process is to convert the battery elements into useful phases for the further process of hydrometallurgy. Hydrometallurgy, instead, is beneficial to reclaim high-value elements from spent lithium-ion batteries (LIBs), such as cobalt, manganese, nickel, and lithium. Then, these elements can be transformed into individual metals, new cathodic substances, or precursors. Hydrometallurgical processes mainly consist of two steps: leaching LIBs and recovering valuable metals from the leachate. Different techniques are used to carry out both steps. The first one can exploit bacteria, alkaline or acid (including organic acid and inorganic acid) solutions, while the latter can employ a solvent extraction, an electrolysis, a precipitation, or an ion exchange process [10].

Nanofiltration (NF) is also a promising process for the selective recovery of lithium from the leachate of end-of-life lithium-ion batteries. It is a pressure-driven membrane operation typically used for the separation of multivalent ions and neutral components from monovalent ions in aqueous solutions. NF membranes are characterized by a pore-size range between 1 and 10 nm (nominal cut-off between 200 and 1000 Da) with intermediate separation capabilities between ultrafiltration (UF) and reverse osmosis (RO) membranes [11,12]. These membranes have received much attention from the scientific community in different fields, including food [13], textile [14], and pulp and paper industries [15], as well for the separation of lithium from salt-lake brines and aqueous resources [16,17,18,19,20,21,22,23,24,25].

Kumar et al. [26] studied, for the first time, the performance of a membrane-hybrid integrated process for the recovery of lithium from liquors obtained by the leaching of black mass (BM). The pH of these liquors was adjusted to precipitate iron, aluminum and phosphate. The liquid phase was then treated by UF to reduce the turbidity of the solution and the UF permeate was then treated by NF to separate Li from other metals. The NF tested membranes (VNF1, VNF2) showed a Li rejection close to 10% when operated according to a cross-flow configuration. The dead-end configuration was also exploited to evaluate the effect of ion concentration on metal rejection. In this last case, a slight decrease in Li rejection was observed for both membranes, along with a slight decrease in the rejection of other ions for the VNF1 membrane and a stronger decrease for the VNF2 type. Finally, Li was precipitated by reactive crystallization to obtain Li₂CO₃.

Gao et al. [27] evaluated the performance of other two NF membranes (DK and NF270) in separating Li from spent LIBs leaching solutions. Tests were performed at pH 1 using virgin membranes and membranes soaked in HCl for various durations. According to the experimental results, the DK membrane exhibited better performance than the NF270 membrane, with higher rejection capabilities for multivalent ions (over 99.0% rejection), while allowing a remarkable lithium passage rate (of the order of 40%). In addition, the DK membrane showed better behavior in the experiments performed after treatment with HCl.

In light of the above considerations, this work aimed at evaluating the performance of a green pathway for the recovery of Li from the black mass (BM) of exhausted batteries using NF membranes. Specifically, the productivity and the Li separation capability of two commercial polymeric NF membranes in flat-sheet configuration (DK and HL) were evaluated in the treatment of extracts obtained from BM treated with a ternary solvent system at acidic pH in selected operating conditions. The effect of influencing factors, such as pH and time extraction on cation rejection, especially the Li one, was also evaluated and discussed.

2. Materials and Methods

2.1. Black Mass and NF Samples Preparation

The black mass was produced from a mixture of spent lithium-ion batteries composed of 90% 18,650 Nickel Cobalt Manganese Oxide (NCM) cylindrical cells (commonly used in power tools and notebook computers) and 10% lithium cobalt oxide (LCO) batteries from smartphones. The black mass (BM) was treated with a ternary solvent system at acidic pH, under magnetic stirring, at room temperature. Samples were taken at different extraction times (48, 96, and 168 h). The liquid fraction of each sample was filtered, and the organic phase was removed under vacuum. The aqueous phase was treated with 1 M Ba(OH)₂ until the desired pH was reached. Specifically, the pH was set to 5 for samples 1 and 3, and to 2.5 for samples 2 and 4.

2.2. Nanofiltration Set-Up

Nanofiltration (NF) experiments were performed on Li-based liquors obtained from the black mass (BM) leaching. Tests were carried out according to a dead-end configuration using a high-pressure stirred cell (HP 4750 from Sterlitech, Kent, WA, USA) able to accommodate a flat-sheet membrane with a membrane surface area of 13.85 cm² and a processing capacity of 300 mL (Figure 2). Nitrogen gas was entered into the cell from the top in order to pressurize the cell so providing the driving force for filtration. The prevailing operating pressure was monitored by a manometer connected at the cell inlet. Stirring inside the cell was accomplished by using a magnetic stirrer. Two different flat-sheet polymeric membranes, named DK and HL (both from GE Water & Process Technologies, Trevose, PA, USA), were used for NF experiments. Their technical specifications are reported in

Table 1. Characteristics of NF flat-sheet membranes (PA-TFC, polyamide–thin film composite; MWCO, molecular weight cut-off).

Membrane Type	DK	HL
Manufacturer	GE Water & Process Technologies	GE Water & Process Technologies
Membrane material	PA-TFC	PA-TFC
Configuration	flat-sheet	flat-sheet
Nominal MWCO (Da)	150–300	150–300
pH operating range	2–10	3–9
Max. operating temperature (°C)	80	50
Max. operating pressure (bar)	40	40
MgSO4 rejection * (%)	98	95
Water permeability (L m−2 h−1 bar−1)	5.4	9.6
Contact angle (°)	41 a	41 b

* Test conditions according to membrane supplier information: 2000 mg/L inlet solution at 110 psi (760 kPa) operating pressure, isothermal process conditions at 25 °C, tests at 15% permeate recovery after 24 h of filtration. Data from ^a Benitez et al. [28]; ^b Verliefde et al. [29].

.

The water permeability of each membrane before experiments was calculated as the slope of the straight line fitted in the plot reporting the relationship between transmembrane water flux (J) and transmembrane pressure (TMP). After each test, the membranes were washed with distilled water at 14 bar and 20 °C for 1 h, and the water permeability was measured again.

NF experiments were performed at constant temperature (23 ± 1 °C) and pressure (20 bar). Permeate flux (J_p) was monitored continuously by measuring the volume of the permeate in a laboratory graduated cylinder, and it was normalized using water permeability values in order to obtain data that are independent from membrane fouling.

$$J_P^{Norm}={\displaystyle \frac{L_{P0}}{L_{P1}}}{J}_P$$

(1)

where L_P0 is the water permeability of the virgin membrane and L_P1 is the one of the cleaned membrane.

An initial extract volume of 200 mL was used and the permeate was collected separately until reaching a final volume of 50 mL corresponding to a volume reduction factor (VRF) of 4.

VRF is dimensionless and defined as

$$VRF={\displaystyle \frac{V_f}{V_r}}=\left(1+{\displaystyle \frac{V_p}{V_r}}\right)$$

(2)

where V_f, V_p, and V_r are respectively feed, permeate and retentate volumes.

Rejection was calculated at each VRF value according to the following equation

$$R=\left(1-{\displaystyle \frac{C_{pi}}{C_{ri}}}\right)·100$$

(3)

where C_pi and C_ri are respectively the permeate and retentate concentration of i species.

2.3. ICP-MS Analysis

The determination of lithium, cobalt, manganese, and nickel was carried out utilizing an Elan DRC-e ICP-MS instrument (Perkin-Elmer SCIEX, Concord, ON, Canada) equipped with a dynamic reaction cell (DRC) for suppressing or reducing polyatomic interferences, operating with CH4 (99.996% purity) as reaction gas. The sample delivery system consisted of a PerkinElmer autosampler model AS-93 Plus with a peristaltic pump and a cross-flow nebulizer with a Scott type spray chamber. The ICP torch was a standard torch (Fassel type torch) with a platinum injector. A multielement solution containing the target elements (100 mg/L, Merck, Darmstadt, Germany) was used for the preparation of aqueous calibration standard solutions after appropriate dilution. Aqueous solutions were prepared using ultrapure water, with a resistivity of 18.2 MΩ cm, obtained from a Milli-Q plus system (Millipore, Bedford, MA, USA).

To assess possible polyatomic isobaric interferences, metals were monitored in both standard and DRC modes after carrying out the optimization of methane flow rate and RPq value.

The recovery of each metal in both permeate and retentate samples of the NF process was calculated according to the following equation

$$Recover{y}_i\%={\displaystyle \frac{{\left(m_i^{cum}\right)}_j}{m_i^{feed}}}·100$$

(4)

where j refers to a specific stream (either permeate or retentate), i denotes the ion for which the recovery is being evaluated, $m_i^{cum}$ represents the cumulative mass in the selected stream regardless of the VRF, and $m_i^{feed}$ is the mass of the selected ion in the feed.

3. Results and Discussion

3.1. Analyses of NF Feed

The concentration of Li, Co, Mn, and Ni in samples obtained in specific conditions of pH and extraction times are reported in

Table 2. Concentration of Li, Co, Mn and Ni in samples obtained under different pH values and extraction times.

N. Sample	Extraction Conditions	Li (mg L−1)	Co (mg L−1)	Mn (mg L−1)	Ni (mg L−1)
1	48 h, pH 5	71 ± 1	228 ± 1	100 ± 1	159 ± 1
2	96 h, pH 2.5	72 ± 3	232 ± 5	107 ± 6	162 ± 2
3	96 h, pH 5	69 ± 2	231 ± 5	98 ± 3	154 ± 3
4	168 h, pH 2.5	76 ± 3	265 ± 5	141 ± 6	167 ± 3

. The obtained results indicate that for all detected elements, the highest concentration was reached at pH 2.5 and for an extraction time of 168 h (sample 4). In order to evaluate the effect of both extraction time and pH on the NF performance of selected membranes in terms of both productivity and separation capabilities, all samples were processed in defined operating conditions.

3.2. Analyses of Permeate Flux

The water permeability of the virgin membranes, measured at a temperature of 20 ± 2 °C, was of 4.5 L m⁻² h⁻¹ bar⁻¹ for the DK membrane and 9.5 L m⁻² h⁻¹ bar⁻¹ for the HL membrane. These values were quite in agreement with data reported by the manufacturer (5.4 and 9.6 L m⁻² h⁻¹ bar⁻¹, respectively). At the end of each NF test, the membranes were washed with distilled water at 14 bar and 20 °C for 1 h: this procedure allowed them to recover in average 90% of the initial water permeability.

Figure 3 depicts the evolution of permeate flux with VRF in the treatment of sample 1 (extract obtained at pH 5, 48 h) with both selected membranes: a clear decline of permeate flux occurs with the increase in VRF due to the increase in solute concentration in the retentate, thereby increasing membrane fouling. It can be seen that HL membrane exhibited a higher flux at VRFs 1 and 2, resulting in greater productivity.

The evolution of permeate flux with VRF for an extraction time of 96 h at two different pH values (2.5 and 5, samples 2 and 3) is illustrated in Figure 4. Also, in this case the HL membrane exhibited a higher flux in comparison with the DK membrane and the permeate flux was negatively impacted by increasing VRF, leading to a reduction in membrane productivity. For sample 3, the HL membrane still showed a significant flux (of the order of 40 L m⁻² h⁻¹) at VRF 4 (Figure 4b) compared to the DK membrane.

Figure 5 shows the behavior of permeate flux as a function of VRF in the treatment of sample 4 (extraction for 168 h at pH 2.5) with both selected membranes. The trend is exactly the same with respect to the one obtained by nanofiltering with the extract at 96 h, exhibiting a higher flux with the HL membrane. Therefore, the membrane productivity was not affected by the increased concentration of metals in the extract obtained at pH 2.5 after 168 h.

3.3. Rejection Performance

The performance of the selected membranes was also evaluated in terms of rejection regarding specific elements.

Multivalent ions were almost completely retained, and their rejection was higher than 90% for both membranes in each VRF condition, indicating that they do not pass into the permeate side. This behavior has to be ascribed to both dimensions and charge matters. In fact, NF membranes are characterized by a superficial charge that changes with the operating conditions of pH. In particular, DK is negatively charged for values of pH higher than 4 and positively charged for pH values lower than 4 (at pH 2.5 Zeta Potential~25 V) [30], whilst HL has a zeta potential of almost 5 mV at pH 2.5 and, thus, is negatively charged for pH values higher than 3 [31]. Specifically, cobalt, manganese, and nickel have quite large values of Van der Waals radii (

Table 3. Atomic radii, Van der Waals radii, and charge of all species.

Species	Atomic Radius (Å)	Van der Waals Radius (Å)	Charge
Lithium	1.45	1.82	+1
Cobalt	1.35	2.00	+2
Manganese	1.40	2.20	+2
Nickel	1.35	1.63	+2

) resulting in a solvation condition that occupies much more space. On the other hand, lithium is smaller and has a lower charge, which results in lower rejection (Figure 6); consequently, lithium passes into the permeate and is collected. It is interesting to notice that as feed becomes more concentrated and VRF increases, lithium rejection decreases, whilst multivalent-ion rejection remains almost constant (Figure 6). This trend aligns with the findings of Kumar et al. [26] in their study on lithium recovery from LIBs liquor using NF. This indicates that feed concentration, which increases with VRF, significantly impacts lithium retention. Since multivalent ions do not pass through the membrane and the electroneutrality of both the feed and permeate must be maintained, lithium is increasingly driven to pass through to maintain Donnan equilibrium [32].

Data on ion rejection as a function of VRF for samples obtained after 96 h of extraction at pH 2.5 and 5 (samples 2 and 3) are reported in Figure 7 and Figure 8, respectively. For both pH conditions, multivalent ions were almost completely retained by both selected membranes; on the other hand, Li rejection showed a decreased trend with VRF. For both membranes, the values of rejection at VRF 2 for samples obtained at pH 2.5 agree with data reported by Gao et al. [27] which investigated the NF of battery liquors at pH 1 for the recovery of Li with the DK membrane in total recycle configuration (lithium rejection of almost 60%). This confirms that the increase in multivalent concentration in the solution does have a fundamental effect on lithium rejection.

It is noteworthy that although lithium rejection obtained using the HL membrane is not strongly influenced by pH conditions (Figure 7b, Figure 8b), rejection observed with the DK membrane shows a specific trend (Figure 7a, Figure 8a). In fact, at pH 2.5 (Figure 7a) membrane surface has a positive charge, resulting in higher lithium rejection under all VRF conditions compared to those at pH 5 (Figure 8a). This is due to the repulsive nature of the electrostatic interactions between the cations and the membrane. Instead, at pH 5 (Figure 8a) membrane surface has a negative charge, resulting in a reduced lithium retention, which decreased further at VRF 4 due to attractive electrostatic interactions. This means that for both membranes, VRF 4 could be an optimal operating condition to improve the selective separation of lithium from other ions.

The trend found during the NF test on the extract at 168 h and at pH 2.5 (Figure 9) was similar to that observed for an extraction time of 48 h. Therefore, the extraction time did not impact the NF performance in terms of ion rejection. Also, in this case, better performance of the DK membrane was observed in terms of lithium rejection compared to the HL membrane.

For the DK membrane, all tests revealed that lithium rejection at VRF 4 is negative. This is in agreement with literature data related to NF experiments on multi-ionic solutions, which confirm the possibility of negative rejection for monovalent ions. Negative rejection usually occurs for ions with decreased concentrations in the membrane phase. In electrolyte mixtures, this happens due to of the acceleration of these ions by the electric field generated by the diffusion potential, which arises from the strong rejection of other components in the mixture. This phenomenon is most pronounced for singly charged ions when there are predominant amounts of higher-charged ions of the same sign [33]. However, this phenomenon is observed only with the DK membrane, as it strongly depends on membrane pore diameter [34].

3.4. Effect of VRF on Ion Concentration

The membrane’s ability to retain multivalent ions leads to their progressive concentration in the retentate as the VRF increases. This occurs because multivalent ions are effectively rejected by the membrane, preventing them from passing into the permeate. Consequently, as the retentate volume decreases, the concentration of these ions increases. In contrast, lithium, being a monovalent ion, is much more permeable through the membrane. As a result, its concentration in the retentate remains relatively stable or may even decrease slightly, depending on the extent of its passage into the permeate. Furthermore, the concentration of multivalent ions in the permeate is significantly lower than in the retentate. This suggests that the membrane effectively blocks multivalent ions while allowing lithium to pass through, highlighting its selective separation capability. Since lithium is not retained to the same extent as multivalent ions, its concentration in the permeate increases with higher VRF values as more of it accumulates in the permeate over time.

The concentration variations appear to be nearly identical for both membranes under all tested conditions (Figure 10, Figure 11, Figure 12 and Figure 13). This aligns with the fact that the rejection rates of multivalent ions are also similar for both membranes. The main difference lies in the lithium concentration in the permeate, which is slightly higher when using the DK membrane. This observation is consistent with the fact that the DK membrane generally exhibits a lower lithium rejection rate. Another noteworthy point is that the concentrations measured in the final test (168 h at pH 5) are higher than those recorded in all other tests (Figure 13). This can likely be attributed to the slightly higher initial feed concentration in this particular experiment (Table 2).

3.5. Ion Recovery

Figure 14, Figure 15, Figure 16 and Figure 17 show the recovery of detected ions in both permeate and retentate side for each membrane and for different extraction conditions. It is possible to notice that DK membrane ensured a higher lithium recovery in the permeate side under all test conditions, consequently resulting in a lower lithium loss in the retentate side. In particular, the highest recovery was observed for the sample 3 under conditions of extraction of 96 h and pH 5, as shown in Figure 16. These conditions can be considered the best trade-off between flux, ion rejection, and total lithium recovery.

4. Conclusions

This study evaluated the performance of two NF membranes with typical polyamide attributes (denoted DK and HL) for lithium recovery from black mass extracts derived from spent lithium-ion batteries. The extraction was performed using a ternary solvent system under acidic conditions, at two different pH (2.5 and 5) and different times (48, 96 and 168 h). Under the selected operating conditions of pressure and temperature, the HL membrane generally exhibited higher permeate fluxes than the DK membrane, indicating superior productivity. Both membranes showed consistent and high rejection (>90%) of multivalent ions regardless of the volume reduction factor (VRF), confirming their effective retention of cobalt, manganese, and nickel. On the other hand, lithium rejection decreased by increasing the VRF of the NF process.

Specifically, the DK membrane exhibited negative rejection values for lithium at VRF 4 for all treated samples indicating a strong selective separation capability, potentially driven by Donnan effects and concentration polarization. Therefore, with increasing concentration ratio, lithium recovery as well separation efficiency increased while membrane flux decreased. For this membrane conditions of extraction of 96 h and pH 5 represent the best trade-off between flux, ion rejection and total lithium recovery.

The obtained results offer interesting perspectives for developing innovative approaches based on the combination of green extraction methodologies and NF operations for an efficient recycling of spent lithium-ion batteries within the circular economy strategy.