Reactor design and operation. The SMER consisted of an AEM sandwiched by two symmetric cathode and anode compartments (Fig. 1b). Brine and recovery solutions were pumped to the flow channels inside the cathode and anode compartments, respectively (Supplementary Fig. 1). We selected a 0.1 M aqueous tributylammonium chloride solution as the recovery solution to allow ion transport while avoiding additional sources of existing cations in brine in order to better elucidate the selectivity of Li+ against impurity cations. For future up-scaled systems, low-cost NaCl can replace tributylammonium chloride in the recovery solution because alkali metal ions do not complicate the final product separation process.16 Spiral microstructures were rationally designed on the flow channel walls (zoomed-in inset of Fig. 1b) to create forced vortex flows (see Supplementary Fig. 2 for design dimensions). FePO4 and LiFePO4 were used as the cathode and anode, respectively (Supplementary Fig. 3). When constant voltages were applied to the reactor (Fig. 1c), the FePO4 cathode can preferentially adsorb Li+, instead of other cations (e.g., Na+, K+, Mg2+, Ca2+, as detailed in Supplementary Figs. 4 and 5), from the brine and form LixFePO4 (Eq. 1), whereas the LixFePO4 anode releases Li+ into the recovery solution and becomes FePO4 (Eq. 2). During the course, anions (mainly Cl-) simultaneously migrate across the AEM from the brine to the recovery solution to maintain the charge balance. The applied full-cell voltage was set at < 1.00 V to avoid the side reactions (i.e., H2O electrolysis and Cl- oxidation).2
As a result, we can collect a Li-rich recovery solution with a low impurity level at the anode channel outlet and a de-lithiated brine effluent from the cathode channel. Upon the complete transition of electrodes to their paired forms (i.e., the FePO4 cathode reduces to LixFePO4 and the LixFePO4 anode is oxidized to FePO4), the cathode and anode are swapped to complete one cycle and are reused to repeat the Li adsorption/desorption operation with fresh brine and recovery solutions. The overall stoichiometric reaction (Eq. 3) can be deemed a one-way Li+ delivery from the brine to the recovery solution under applied voltages:
Cathode reaction: FePO4 + x Li+ + x e- → LixFePO4, (1)
Anode reaction: LixFePO4 → FePO4 + x Li+ + x e-, (2)
Overall reaction: Li+ (Brine) → Li+ (Recovery solution). (3)
Table 1
Major cation concentrations in the collected recovery solution after cyclic operation
Cycle | Li (ppm) | Na (ppm) | K (ppm) | Mg (ppm) | Ca (ppm) |
Fresh brine | 308.45 ± 0.43 | 56329.62 ± 6.83 | 473.98 ± 0.22 | 20324.56 ± 2.48 | 214.73 ± 0.33 |
1st | 53.66 ± 0.21 | 43.88 ± 0.21 | 3.02 ± 0.12 | 7.47 ± 0.07 | 1.71 ± 0.21 |
2nd | 51.06 ± 0.23 | 47.89 ± 0.31 | 3.70 ± 0.15 | 9.34 ± 0.14 | 1.65 ± 0.23 |
3rd | 46.44 ± 0.17 | 42.39 ± 0.25 | 3.34 ± 0.09 | 9.56 ± 0.09 | 1.98 ± 0.18 |
4th | 45.26 ± 0.14 | 44.21 ± 0.27 | 3.74 ± 0.11 | 8.38 ± 0.12 | 1.48 ± 0.20 |
5th | 44.54 ± 0.18 | 41.91 ± 0.24 | 3.20 ± 0.07 | 7.56 ± 0.08 | 1.82 ± 0.15 |
6th | 42.91 ± 0.14 | 40.70 ± 0.28 | 2.93 ± 0.09 | 7.15 ± 0.11 | 1.38 ± 0.12 |
7th | 44.54 ± 0.17 | 41.25 ± 0.31 | 4.10 ± 0.15 | 7.37 ± 0.11 | 1.20 ± 0.15 |
8th | 45.15 ± 0.21 | 43.08 ± 0.23 | 3.02 ± 0.12 | 8.09 ± 0.12 | 1.95 ± 0.17 |
Lithium extraction performance. We performed Li extraction on the SMER using brine from Taijinar Lake, which is more difficult to mine compared to most low-grade salt lake brine, as it possesses a relatively low Li concentration of 308.45 ppm and a high Mg/Li ratio of 65.89.46,47 The applied voltage was kept constant at 0.70 V and the cut-off time was set to 6,000 s for one cycle. Inductively coupled plasma (ICP) measurements were conducted to identify the solution composition. As listed in Table 1, the concentrations of major cations in the collected recovery solution, except for Li+, dropped significantly after one cycle compared to those in the fresh recovery solution. In particular, Mg2+ and Na+ reduced from 20,324.56 to 7.47 ppm and from 56,329.62 to 43.88 ppm, respectively. Consequently, the Mg/Li ratio was lowered from 65.89 to 0.14, indicating that the as-obtained recovery solution was qualified for preparing battery-grade Li2CO3.16,23
We then continued the Li extraction for another seven cycles by reusing the electrodes and controlling the applied voltage at 0.70 V. As displayed in Fig. 2a, the current density decayed over time during each cycle, likely because of the continuous consumption of active materials on electrodes. Meanwhile, the current density profiles of different cycles were largely similar from the third cycle: the weighted average current density was approximately 2.98 mA/cm2 in the first cycle, then gradually decreased to 2.50 mA/cm2 in the third cycle, and remained relatively steady thereafter. Correspondingly, the Li extraction rate was 150.14 µmol/cm2/h in the first cycle and became stable at around 81.00 µmol/cm2/h from the third cycle onwards (Fig. 2b). Normalized by the active material loading of the electrode (i.e., 25.59 mg/cm2), the Li extraction rate of 81.00 µmol/cm2/h corresponds to 21.96 milligram of Li per gram of LiFePO4 per hour (mg/g/h).
In contrast, the extraction rates of Mg+ and Ca2+ were below 5.01 and 0.65 µmol/cm2/h, respectively, throughout the cyclic operation (Fig. 2b). Cyclic voltammetry (CV) in Supplementary Figs. 4 and 5 suggests that it is difficult to insert Mg+ and Ca2+ into the FePO4 electrode. The small amounts of Mg+ and Ca2+ detected in the recovery solution were ascribed to cation penetration through the AEM from the brine due to the high osmotic pressure. As a result, we achieved a Li/Mg and Li/Na selectivity of 393.80 and 198.96, respectively (Fig. 2c), comparable with that in the literature and practical applications.16,23,44,45 Scanning electron microscopy (SEM) images and x‑ray diffraction (XRD) patterns of FePO4 and LiFePO4 electrodes before and after the eight-cycle operation validated the good material stability without obvious structural and phase changes (Supplementary Figs. 6 and 7).
For comparison, we extracted Li using a traditional H-shape reactor (Supplementary Fig. 8) under the same operating conditions. In sharp contrast to the SMER, the H-shape reactor only allowed a weighted average current density of 0.43 mA/cm2 (Fig. 2d) and a Li extraction rate of 13.00 µmol/cm2/h after three cycles (Fig. 2e), although with similar Li/Mg and Li/Na selectivity. In other words, the SMER drastically improved the Li extraction rate by over 6 times when benchmarked against H-shape reactor (Supplementary Fig. 9). With reference to the literature, the SMER also substantially outperformed other state-of-art Li extraction devices aiming at low-grade brine while maintaining high Li selectivity under inferior brine conditions: the Li extraction rate reported in this study surpassed that of existing technologies by about 5.64 to 236.13 times (Fig. 2f and Supplementary Table 1).
Solar-driven up-scaled system and techno-economic analysis. In practical operations, high-throughput and independently powered Li extraction is desired because most low-grade salt lake brine is located in remote areas, where municipal power supply is not feasible but intense sunlight is usually secured. We up-scaled the SMER by integrating two reactors in series (Series I and Series II, as highlighted in Fig. 3a), which were then arrayed in parallel to form a module (Figs. 3a, 3b, and Supplementary Fig. 10). A recently developed monolithic perovskite/silicon tandem solar cell (active area: 1 cm × 6 cm)48 was employed to power the SMER module and lower its carbon footprint. Under natural sunlight, the off-grid solar cell could stably output a voltage of 1.75 V, sufficient to drive the series-connected SMERs (Supplementary Fig. 11). As recorded in Fig. 3c and 3d, the weighted average current densities in each cycle for both reactor series were largely similar at approximately 3.00 mA/cm2 despite the different shapes of the chronoamperograms. The weighted average current density of the series was slightly higher than that of a single SMER due to the higher distributed voltage of 0.88 V from the tandem solar cell. The ICP analysis suggested that the SMER module achieved a Li extraction rate of 82.00 µmol/cm2/h (Fig. 3e), close to that of a single SMER. Meanwhile, the extraction rate of Na+ and Mg2+ in the module was 40.00 and 4.60 µmol/cm2/h, respectively, slightly higher than a single SMER (Fig. 3e). We again attributed the faster Na+ and Mg2+ extraction to the higher applied voltage, which forced the insertion of Na+ and Mg2+ into the lattice of FePO4. After collecting the Li-rich recovery solution, we precipitated Li out in the form of Li2CO3 through the lime soda evaporation process25 and obtained refined Li2CO3 sediment from the module at a yield of 3.41 mg/h (Fig. 3f). XRD and ICP analyses verified that the impurity levels of Na+, K+, Mg+ and Ca+ in the Li2CO3 product (Figs. 3f and 3g) were in good compliance with the standard for commercial battery-grade Li2CO3 (YS/T582-2013).
Furthermore, we demonstrated a scale-out strategy for the SMER module and conducted a comprehensive TEA to validate its commercial viability. The recently growing lab-on-a-chip technology49–51 was proposed to scale our Li extraction system from the microgram to kilogram level. We envisioned integrating 100,000 SMERs in series and parallel as a stack while maintaining precise control of the microstructures and flows (inset of Fig. 3h). In this way, the Li extraction yield is proportional to the number of reactors, but the operating conditions of individual reactors remain unchanged. Based on the Li2CO3 output from the SMER module, we anticipated that the 100,000 reactors assembly would achieve a Li2CO3 throughput of 279.07 kg/year (Supplementary Table 2).
The TEA model revealed that this scale-out strategy could significantly reduce the capital expenditure (CapEx) and operating expenses (OPEX) of SMERs. The calculated plant-gate levelized cost of production (LCOP) for the up-scaled SMER assembly was $28.55/kg Li2CO3 (Fig. 3h, see supplementary material for the detailed method and analysis), corresponding to a remarkable gross profit rate as high as 149% based on the average market price of Li2CO3 in 2022, at about $70.99/kg.52 It should be noted that we could employ NaCl as the solvent of the recovery solution in practical scenarios, further lowering the cost. Underpinned by the reasonable assumption that the market price of Li2CO3 will continue to rise with the increasing Li-ion battery demand and that the solar electricity cost will continue to decrease, 53,54 the up-scaled SMER assembly could reach the break-even point and start making a profit within the first year of operation (Supplementary Fig. 12).