λ-MnO2 with a spinel structure was chosen from the reported Li-ion sieve materials. λ-MnO2 has three-dimensional paths for Li diffusion in the lattice (Supplementary Fig. 2), whose size is small enough to prevent the intercalation of other metals that abundantly exist in seawater (e.g., Na+, K+, and Ca2+) 5,9. In addition, manganese oxides are biocompatible electron acceptors for MR-1 and other EET-capable bacteria 27,32,33.
We examined Li recovery from LiCl solutions dissolved in phosphate-buffered saline (PBS) at pH 7.4 and at high concentrations of Na+ and K+. The λ-MnO2 nanoparticles (Supplementary Fig. 3) and MR-1 were anaerobically incubated with 50 mM LiCl/PBS in the presence of lactate as the electron donor for 24 h at 30°C. The nanoparticles formed sheet-like agglomerates at the bottom of the vial (Fig. 1a). The resuspension of the precipitate yielded 10–50-µm particles (Fig. 1b), which were approximately 100 times larger than those in the initial state (Supplementary Fig. 4). No such agglomerates were seen in the absence of LiCl (insets in Figs. 1a and b). Tunnel-structured MnO2 polymorphs did not agglomerate even in the presence of LiCl, and the dry weight decreased in relation to that of λ-MnO2 (Supplementary Fig. 5) with negligible structure change (Supplementary Fig. 6). Contrary to the common knowledge that EET dissolves MnO2 in aqueous solutions 27, λ-MnO2 remained in the solid state and agglomerated specifically in the presence of LiCl.
Scanning electron microscopy (SEM) combined with energy-dispersive X-ray spectroscopy (EDS) was conducted for analysing the agglomerates; the results indicated that microbes with a size of 1–2 µm were attached to the nanoparticles (Figs. 1c and d). The presence of a lower volume of bacteria, along with the nanoparticles, was further confirmed through fluorescence microscopy of DNA-stained MR-1 (Fig. 1e and Supplementary Fig. 7), indicating that MR-1 cells were physically associated with only a minor part of nanoparticles. However, almost all λ-MnO2 was electrochemically reduced in the precipitate. X-ray diffraction (XRD) peaks shifted clearly without losing their sharpness (Fig. 2a), whereas such a shift was not observed when LiCl was omitted (Supplementary Fig. 8). Furthermore, X-ray absorption near-edge structure (XANES) analysis and inductively coupled plasma optical emission spectroscopy (ICP-OES) results indicated Li intercalation to form Li1 − xMn2O4 (Fig. 2b and Supplementary Fig. 9).
The Li intercalation was associated with microbial respiration. CO2 production for Li intercalation (Supplementary Fig. 10) indicates an NADH-dependent carbon catabolic pathway coupled with lactate oxidation 34. In the absence of lactate, the XRD peaks did not shift in the presence of MR-1 (Fig. 2c). Although partial reduction occurred even without MR-1 because lactate can act as a mild reductive agent, the peak shift was insufficient, and the peak intensity significantly decreased (Supplementary Fig. 11). This suggests that accelerating the Li intercalation kinetics using MR-1 is critical for maintaining crystallinity during the reaction.
The selectivity of Li intercalation was further confirmed using Li–Mg mixed solutions. Although λ-MnO2 prevents the intercalation of other alkali metals, this was not the case for Mg2+ because its ionic radius is similar to that of Li+ 35. The molar ratio of Li/Mg was 200–1000 times higher in the agglomerates than that in the solution (Fig. 2d). This order is consistent with the difference in diffusion distances estimated from their activation energies (Supplementary Fig. 12). Such slow kinetics of Mg2+ diffusion in spinel-oxide materials have also been reported in previous studies on Mg batteries 36–38. Furthermore, the properties of intermediate states formed by limiting the amount of Li in the reactor (Supplementary Figs. 13–16) are consistent with reported electrochemical charge/discharge of LiMn2O4/λ-MnO2 5,39,40. The homogeneous and efficient electrochemical reactions exhibited by our approach closely resemble those observed in electrode-driven systems. This similarity offers a significant advantage for scaling up Li recovery, as our method circumvents the common challenges associated with electrode-based systems, such as mass loading limitations and increased ohmic drop during scale-up 4–6.
The electron transfer limited at the bacteria/λ-MnO2 interface, however, does not fully explain the homogeneous and efficient electrochemical reactions because the amount of MR-1 in the agglomerate was much less than the amount of λ-MnO2 (Figs. 1d and e). This indicates the biological formation of a long-distance electronic network in the agglomerate with the semiconductive nanoparticles and the extracellular or cell-surface cytochromes 28. To analyse the cellular and material interface formed during agglomeration, we performed electron energy loss spectroscopy (EELS) mapping with scanning transmission electron microscopy (STEM). Subsequent to agglomeration, the Li-containing aggregated material was dispersed by ultrasonication and mounted on a TEM grid. The nanoparticles attached to the cells exhibited no obvious changes in morphology (Fig. 3a), and Mn valency evaluated by EELS decreased from + 4 (initial state) to typically around + 3.5 (after Li intercalation) even if Mn was not directly attached to the cell (Fig. 3b), suggesting the formation of a long-distance electron pathway. Mn was not detected inside the cell through either EELS or EDS (Supplementary Fig. 17), indicating that reduced Mn was not originated from dissolved Mn ions reduced in the cells. EELS and cross-sectional TEM results revealed the presence of a small amount of Mn2+ in the vicinity of the microbe (Supplementary Fig. 18), which can likely be attributed to the formation of hydrated Mn phosphate 41.
Focused ion beam (FIB) milling with cross-sectional SEM–EDS was employed for the direct observation of the cellular morphology inside the agglomerates. Note that the agglomerate was subjected to 3,3′-diaminobenzidine (DAB)-H2O2 and OsO4 staining in advance for labelling redox-active enzymes, including cytochromes 42. We selected a small piece with a diameter of 10–20 µm (Supplementary Fig. 19) and subjected it to FIB sputtering to expose a cross-section (Figs. 3c and d). Os was concentrated near microbes but observed with Mn in EDS profiles (Figs. 3e–g and Supplementary Fig. 20). Carbon incorporated in λ-MnO2 was also detected through EDS; the results suggested the expression of the extracellular and cell-surface cytochromes to physically and electrically support the MR-1–nanoparticle composite. In addition, we observed a nanowire-like structure that potentially facilitates electrical conduction via cytochromes 23, bridging microbes and nanoparticles (Fig. 3h), and suggesting a possible mechanism for electron transfer in this system. Given that genes for outer-membrane cytochromes (OM c-Cyts) support long-distance electron transport via bridging semiconductive nanoparticles 43, we used gene deletion mutants lacking representative OM c-Cyts (∆mtrC/omcA). In contrast to the wild-type MR-1 (Fig. 3i), little or no reduction of λ-MnO2 was seen even after 24 h (Fig. 3j). Furthermore, we added riboflavin to increase the rate of direct EET via OM c-Cyts as a bound cofactor 44,45. The presence of only 2 µM riboflavin accelerated the Li recovery kinetics by more than eight times (Fig. 3k). The estimated rate of Li recovery with 50 mg λ-MnO2 was approximately 660 µg/h in the presence of riboflavin. In general, cathode materials in LIBs can be charged/discharged within several hours 5,7. Therefore, once riboflavin enhances the rate of Li intercalation, the Li recovery kinetics is probably limited by the Li-ion diffusion kinetics in λ-MnO2. Given that the kinetics of our method closely align with such atomic-level phenomena, the comparable Li recovery efficiency to electrode-based technologies is well-justified. These results suggest the primary role of the extracellular and cell-surface cytochromes in Li intercalation and the formation of an electrically conductive network in the agglomerates.
Such microbial interaction is likely induced by the electrochemical properties of λ-MnO₂, given that MR-1 is sensitive to the potential of its extracellular environment 34. Among the MnO₂ polymorphs, λ-MnO₂ exhibits a particularly high potential for Li intercalation (Supplementary Fig. 21) 46. Consistent with this, CO₂ production from lactate oxidation was observed in the presence of λ-MnO₂, a process typically triggered when the anode potential is highly positive in a microbial fuel cell system 34. The elevated potential also facilitates electron transfer from MR-1 owing to the high driving force. Moreover, the lithiation of λ-MnO₂ enhances its electronic conductivity by shifting the Fermi level closer to the conduction band 47,48, potentially further promoting electron transfer within microbial agglomerates 43. Notably, the formation of over-lithiated Li₂Mn₂O₄ was prevented, likely due to the limiting reductive potential of MR-1.
The high redox potential of λ-MnO₂ for Li intercalation is sustained even in low-concentration Li environments, such as seawater (Supplementary Fig. 22). Seawater mining is attracting growing attention since the Li content in seawater (about 25 µM) is orders of magnitude higher than that in most freshwater systems (on the order of 0.01 µM) 11,22. However, the concentration of Li in seawater is still low enough to make its recovery challenging 9. We, therefore, next examined the feasibility of bioelectrochemical Li recovery from real seawater collected from Oarai, located on the east seaside of Japan. λ-MnO2 (5 mg) and MR-1 were placed in a bottle filled with lactate-spiked seawater (50 mL), and the Li concentration of the supernatant liquid was evaluated through ICP mass spectroscopy (ICP-MS) in the absence of riboflavin. As shown in Figs. 4a and b, Li was recovered from seawater within several hours, consistent with PBS-based experiment without riboflavin, and inserted Li was stably stored even after the microbial activity had settled (Supplementary Fig. 23). The molar ratios of Li/M (M = Na, Mg, K, Ca) electrochemically intercalated into the material were all greater than 100 (Supplementary Tables 2 and 3). In contrast, λ-MnO2 itself, without microbial reduction, was not effective because of the absence of the electrochemical driving force (inset in Fig. 4b). The small variation in the reaction rate was dependent on the seawater sample, suggesting that transient substrates, such as flavins, in seawater affect the rate of EET. Indeed, the Li recovery rate was enhanced approximately two times by spiking 2 µM riboflavin (Supplementary Fig. 24). We also confirmed the reproducible Li recovery performance when using different seawater batches, volumes, and/or the mixing ratios of seawater/λ-MnO2 (Fig. 4c). Self-agglomeration was observed even at a higher seawater/λ-MnO2 ratio (Fig. 4d). These results further support the scalability of our method utilizing the autonomous formation of the agglomerates.
Remarkably, our proposed method achieved the Li recovery efficiency comparable to, and even displaying superior selectivity than, the electrode-based processes for seawater treatment, all without the need for electrodes or electrochemical equipment 11 (Fig. 4a). Also, ion-exchange materials 20 and the electrical pumping method that uses a solid-state electrolyte membrane 22 used for Li recovery necessitate time-consuming purification stages, highlighting the benefits of our efficient and straightforward bioelectrochemical approach for Li recovery.
To further evaluate the advantages of our method, we conducted comprehensive techno-economic analysis (TEA) and life-cycle assessment (LCA), which revealed that our approach could significantly outperform conventional evaporative processes. Considering that electrogenic bacterial strains capable of reducing MnO₂ under brine-level salinity conditions have been isolated 49, we used data from LiCl/PBS solution experiments to estimate the potential for Li₂CO₃ production from brine sources. Developed Li recovery process (Fig. 5a) involved oxidizing Li1 − xMn₂O₄ with Na₂S₂O₈ to obtain a Li-containing solution while regenerating λ-MnO₂ (Supplementary Fig. 25), with only a 0.001% Mn loss (Supplementary Fig. 26). The Li-containing solution underwent Mn2+ removal, volume reduction, and Li₂CO₃ precipitation with the addition of Na₂CO₃, resulted in the purity of ∼98% (Supplementary Fig. 27), followed by filtration, washing, and drying. The estimated production cost of our method is 5.97 USD/kg Li₂CO₃ (Fig. 5b), comparable to the evaporation method for brines 50,51, and below the industrial Li₂CO₃ market price (Supplementary Table 4). Moreover, our production rate is orders of magnitude faster than that of the evaporation method, highlighting the efficiency and scalability of our approach.
Our LCA indicated a water consumption of 0.39 m³eq/kg Li₂CO₃ (Fig. 5c), representing brine water loss being reduced by 97–99% compared to currently used methods 52, thus highlighting the potential of our technology to alleviate local water conflicts 3,9,10. The climate change impact was calculated at 16.8 kg CO₂eq/kg Li₂CO₃ (Fig. 5d), higher than currently used methods 52, primarily due to the use of Na2S2O8 in the Li₂CO₃ precipitation process. Na2S2O8 is also a significant factor in the overall cost (Supplementary Fig. 28) and water consumption. Development of more cost-effective Li leaching methods would further improve the sustainability of our approach. Importantly, given the cost, water, and CO₂ contribution from Li intercalation rate is much smaller than the input of chemicals, the advantages of our method remain even if microbial Li intercalation turns out to be slower in brine. We have now built a prototype of an anaerobic culture bottle with a cation-exchange membrane, and successfully demonstrated the collection of Li from outside the membrane (Supplementary Fig. 29). Installing such a membrane to separate the microbial reactor and environmental water source serves as an example of a scalable method that can be implemented with further minimizing environmental impact 3,53.