High-energy anode-free Li metal batteries with in-built surface-fluorinated Li-rich Mn-based cathodes

Abstract

Anode-free lithium metal batteries (AF-LMBs) have been recognized as a promising approach for realizing highly safe and cost-effective lithium metal batteries with high energy density but suffer from low Coulombic efficiency (CE) and short cycle life. Here, we demonstrate high-energy and long-life AF-LMBs by leveraging the high-capacity and Li replenishment effect of surface-stabilized Li-rich Mn-based oxide (LRMO) cathodes. The irreversible capacity loss during LRMO formation, typically detrimental, is harnessed to release extra Li that deposits on the current collector, compensating for Li inventory loss and extending cell lifetime. Simultaneously, Li replenishment strategy during the formation cycle activates surface lattice oxygen, which reacts with hydrofluoroether cosolvents in a localized high-concentration electrolyte to enable a fluorinated cathode-electrolyte interphase and fluorination of the LRMO surface. As a result, a 2–ampere hour anode-free pouch cell retains 80% capacity after 260 cycles, and a 30–ampere hour pouch cell delivers 350 watt-hours per kilogram and 1200 watt-hours per liter with 84.4% retention after 180 cycles, alongside impressive low-temperature performance and rate capability. The demonstrated commercial viability of our AF-LMB showcases the essence of coherent cathode and electrolyte design for achieving high-energy and long-life AF-LMBs.

In situ fluorinated surfaces via lattice oxygen-cosolvent chemistry enable long-life anode-free batteries with LRMO cathodes.

INTRODUCTION

Lithium metal batteries (LMBs) have emerged as one of the most promising next-generation energy storage devices, owing to the high theoretical gravimetric and volumetric capacity [3860 milliampere hours (mA·hour) g⁻¹ and 2061 mA·hour cm⁻³] of lithium metal anodes (1–3). Despite these theoretical promises, the utilization of lithium metal foils as the anode poses formidable challenges including the safety concerns related to dendrite formation, the high cost of lithium metal foils, the operational challenges associated with moisture-free environments, and inefficiencies arising from the excessive utilization of lithium (4).

The concept of anode-free lithium metal batteries (AF-LMBs) presents an intriguing possibility to address these challenges. In the AF-LMB configuration, in principle, all the active Li sources are stored in the cathode and the anode is a bare current collector (CC). Compared to traditional LMBs, the AF-LMB configuration eliminates excess use of lithium metal and thus maximizes the energy density and optimizes the cell assembly procedure (4, 5). However, the AF-LMB system typically exhibits very poor cyclability, primarily due to the absence of excess Li. This leads to failure to offset the irreversible loss of Li caused by parasitic side reactions and isolated dead Li at the anode (6, 7).

To alleviate lithium depletion in AF-LMBs, considerable efforts have focused on modulating interfacial chemistry and lithium deposition behavior. These efforts include tailoring electrolyte compositions to stabilize the solid electrolyte interphase (SEI) and suppress dead Li formation (8–15), as well as engineering the surface of the anode CC to regulate lithium plating (16–22). In parallel, given the finite lithium inventory inherent to anode-free configurations, strategies aimed at replenishing lithium have been explored (23–26). Typical examples often adopt asymmetric charge-discharge formation protocols or elevated lower cutoff voltages to retain additional lithium at the anode (table S1). In addition, Li-rich cathodes or sacrificial prelithiation agents can be used to compensate for lithium loss during initial charging in the formation process (table S2).

These approaches have achieved considerable success in the lab, yet their transition into manufacturable, high-performance batteries faces unresolved critical issues. Under practical testing conditions, most reported AF-LMBs still exhibit a cycle life of fewer than 150 cycles (4, 9–17, 20–23, 25, 26). Furthermore, these strategies often rely on additional processing steps, such as incorporating sacrificial agents, fabricating Li-rich cathodes, or modifying the anode CC. These requirements increase manufacturing complexity, diminish cost advantages, and reduce compatibility with established production lines, thereby impeding the scale-relevant validation of anode-free battery technologies (see detailed discussion in tables S1 and S2). To date, there remains a lack of a facile, cost-effective, and scalable strategy capable of enabling highly stable AF-LMBs under practical conditions.

In this study, we use Li-rich Mn-based oxide (LRMO) as the cathode for high-energy and stabilized AF-LMBs. In contrast to conventional layered metal oxide cathode materials, the unique Li─O─Li bonding in Li-rich cathode materials induces extra anionic redox reaction, resulting in high specific capacity (27). The anionic redox of LRMO cathode shows certain irreversibility, which we here propose can release a considerable amount of extra Li⁺ ions during the first charge, supplementing the Li loss on the anode and benefiting the anode-free cells. However, the irreversible oxygen anionic redox of LRMO at high voltage destabilizes the lattice oxygen through O─O dimer formation, leading to irreversible oxygen release and transition metal (TM) migration (28, 29). This phase transition causes structural degradation of LRMO, resulting in severe voltage decay and capacity fade of the cell (30, 31). Moreover, the high voltages required during activation of O redox, together with oxygen-related species released from the cathode, induce severe parasitic reactions (32), which are particularly detrimental to ether-based electrolytes with limited oxidative stability, thereby compromising overall cell stability. Hence, although the LRMO cathode is expected to resolve the Li loss problem in AF-LMBs, it remains challenging to achieve practical long-life AF-LMBs.

To address these challenges, we describe an in-built surface fluorination reconstruction (SFR) strategy to stabilize the LRMO cathode via electrolyte engineering with the ether-based localized high-concentration electrolyte strategically using the hydrofluoroether cosolvent [typically, 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropyl ether (TTE)] (Fig. 1) while enabling long-term stable Li plating/stripping at the anode. We deliberately overcharge the LRMO to 4.65 V during the formation cycle; this simple yet effective strategy not only supplements the Li loss during the cycling for AF-LMBs but also intentionally creates O vacancies and releases oxidized lattice oxygen in the vicinity of the LRMO surface to attack the TTE cosolvent, producing abundant inorganic F⁻, which consequently in situ forms a LiF-rich cathode-electrolyte interphase (CEI). The SFR strategy leads to a stable CEI, improving LRMO structural integrity and suppressing electrolyte decomposition. Furthermore, beyond conventional surface doping, the F substitutes the O site in the reconstructed disordered rock salt phase at the LRMO surface, further stabilizing the lattice oxygen, suppressing the void formation, and restraining structural degradation of LRMO during prolonged cycling.

Fig. 1. Schematic diagram of the design strategy of our AF-LMBs.

(A) The SFR strategy using the hydrofluoroether cosolvents enables a uniform LiF-rich CEI and fluorination of disordered rock salt phase at the LRMO surface, which stabilizes the high-energy AF-LMBs. (B) In the absence of the hydrofluoroether cosolvents, the activation of anionic redox reaction leads to appearance of voids in LRMO, oxidative decomposition of the electrolyte, and a nonuniform and thick CEI, which decrease the stability of the AF-LMB cell. The bubbles represent gas evolution from electrolyte decomposition in the vicinity of the cathode, which originates from the oxygen oxidation process of LRMO; the gray layer indicates the LiF-rich CEI, and the blue layer indicates organic CEI formed in AF-LMB cells without hydrofluoroether cosolvents. The porous lithium in the bottom panel indicates nonuniform lithium deposition and premature cell failure without hydrofluoroether cosolvents.

Distinct from previously reported formation protocols for LRMO cathodes that rely on carbonate-based electrolytes and thick lithium metal anodes (table S3), our strategy operates under anode-free conditions and fundamentally differs in formation chemistry. Under symmetric charge-discharge protocol (0.1 C/0.1 C), the formation process effectively compensates for lithium loss at the anode without the use of external sacrificial agents or additional processing steps. Moreover, it enables stable operation of LRMO cathodes at a high cutoff voltage of 4.65 V in an ether-based electrolyte for more than 50 cycles, overcoming the long-standing incompatibility between LRMO and Li-metal-friendly ether electrolytes. This strategy is fully compatible with existing manufacturing processes and readily scalable to pouch cell formats, providing a facile and low-cost strategy for fabricating a commercially viable 30-A·hour-scale AF-LMB with high energy density [350 watt-hours (Wh) kg⁻¹ and 1200 Wh liter⁻¹] and a long cycle life (84.4% capacity retention after 180 cycles). The utilization of the LRMO cathode and the SFR strategy demonstrated in our work represents a unique and promising solution for high-energy and long-life AF-LMBs.

RESULTS

Pouch cell electrochemical performances

We used the LRMO (Li_1.11Ni_0.34Mn_0.53Al_0.02O₂) cathode, a bare Cu CC anode, and a localized high-concentration ether-based electrolyte [6.9 M lithium bis(fluorosulfonyl)imide (LiFSI) in 1,2-dimethoxyethane (DME), with 70% TTE cosolvent; denoted as LiFSI-DME-TTE] to fabricate a 2-A·hour-scale anode-free pouch cell (Fig. 1A). A reference 2-A·hour-scale anode-free pouch cell, denoted as LiFSI-DME, was fabricated using a high-concentration ether-based electrolyte without TTE (6.9 M LiFSI in DME) (Fig. 1B). Our anode-free pouch cells were subjected to galvanostatic cycling at 0.1 C in the first three formation cycles, with subsequent cycles by charging at 0.2 C and discharging at 1 C. The voltage range for the second formation cycle was 3.0 to 4.65 V, whereas that of the first and third formation cycles was 3.0 to 4.3 V. The elevated voltage during the second formation cycle serves to activate the anionic redox (O) processes, leading to a certain irreversible capacity loss in the LRMO cathode (33). The irreversible Li extracted from LRMO could efficiently be plated on the Cu anode, serving as excess Li resources in the following cycles (Fig. 1A). This Li replenishment contributes to improving life span of our AF-LMBs. The formation process proves crucial to the electrochemical performances of our AF-LMBs.

Figure 2A illustrates the charge/discharge curves of the second formation cycle for LiFSI-DME-TTE and LiFSI-DME pouch cells. Both LiFSI-DME-TTE and LiFSI-DME pouch cells exhibited similar Coulombic efficiency (CE) and irreversible capacity loss, indicating comparable amounts of Li replenishment. To investigate the Li replenishment effect, the front and cross-sectional scanning electron microscopy (SEM) analyses were conducted to characterize the anode after the formation process. As depicted in Fig. 2 (B and C) and fig. S1, the morphology of Li metal plated in the LiFSI-DME-TTE and LiFSI-DME cells is comparable, and the thickness of Li metal is ~35 μm, aligning with the second formation cycle profiles of the two cells. As shown in fig. S2, we have calculated that the total amount of extra Li reservoir is 0.419 mg cm⁻², and the dead Li amount is 0.005 mg cm⁻² by titration-gas chromatography (TGC). These results strongly suggest that the irreversible extraction of Li from LRMO can serve as an effective Li replenishment source for our AF-LMB and that the prelithiation effect of LiFSI-DME-TTE and LiFSI-DME cells is rather identical.

Fig. 2. Electrochemical performances of 2-A·hour anode-free pouch cells and characterizations of the Li metal anode in different electrolytes.

(A) 0.1 C (~0.40 mA cm⁻²) charge/discharge curves during the second formation cycle at 3.0 to 4.65 V using 6.9 M LiFSI-DME-TTE and 6.9 M LiFSI-DME electrolytes. (B and C) Cross-sectional SEM images of the Li metal anodes in 6.9 M LiFSI-DME-TTE (B) and 6.9 M LiFSI-DME (C) electrolytes after the Li replenishment. (D) High-voltage test of the pouch cells using 6.9 M LiFSI-DME-TTE and 6.9 M LiFSI-DME electrolytes (constant current charge to 4.6 V, constant voltage charge at 4.6 V, and then constant current charge to 4.8 V). (E) Cell capacity and CE of the investigated electrolytes (0.2 C charge/1 C discharge at 3.0 to 4.3 V after the first three formation cycles, 0.1 C for the formation cycles, cutoff voltage for the second formation cycle is 4.65 V, and cutoff voltage for other formation cycles is 4.3 V). (F) Electrochemical performances of 2-A·hour anode-free pouch cells using 6.9 M LiFSI-DME with different hydrofluoroether cosolvents (1,1,2,2-tetrafluoroethyl 2,2,2-trifluoroethyl ether, 1,1,2,2-tetrafluoro-3-methoxypropane, and 1H,1H,5H-perfluoropentyl-1,1,2,2-tetrafluoroethyl ether). (G) Capacity and CE of cells using the investigated electrolytes within the voltage window of 3.0 to 4.65 V at 0.2 C charge/1 C discharge (formation was performed at 0.1 C with one cycle cutoff at 4.3 V and another cycle cutoff at 4.65 V; an additional formation cycle with a cutoff at 4.60 V was added in between for the LiFSI-DME cell to stabilize the CEI in the absence of TTE). (H and I) Charge/discharge curves of the pouch cells using 6.9 M LiFSI-DME-TTE (H) and 6.9 M LiFSI-DME (I) electrolytes at 3.0 to 4.65 V after the formation cycles. Scale bars, 20 μm [(B) and (C)].

Nevertheless, the CE of the LiFSI-DME cell (69.9%) is slightly lower than that of the LiFSI-DME-TTE cell (72.9%) (table S4). In addition, compared to the LiFSI-DME-TTE cell, an anomalous response peak appears near 4.65 V in the dQ/dV curve of the LiFSI-DME cell (fig. S3), indicating potential side reactions of the electrolyte under high voltage. As the formation process in the voltage range of 3 to 4.65 V plays a key role in our AF-LMB, the electrochemical stability window of the investigated electrolytes was assessed. As shown in fig. S4, in a Li || Al configuration, both electrolytes showed the same oxidation stability up to 6 V, which indicates that both could withstand the cutoff voltage of 4.65 V used in this work. However, when LRMO was used as the working electrode, the LiFSI-DME electrolyte in Li || LRMO exhibited increasing leakage current after the activation of O redox (4.56 mA cm⁻² at 5.5 V) whereas the LiFSI-DME-TTE electrolyte exhibited very small leakage current (0.902 mA cm⁻² at 5.5 V) after the activation of O redox. We further explored the high-voltage stability of both electrolytes by the high-voltage charging test. As shown in Fig. 2D, the LiFSI-DME-TTE cell could easily charge to 4.8 V, whereas the LiFSI-DME cell exhibited overcharging and oxidative decomposition of the electrolyte. The corresponding dQ/dV curves of the two pouch cells further revealed the difference (figs. S5 and S6), with the LiFSI-DME cell showing a strong response peak at 4.66 V, indicating the severe side reactions. In contrast, in the LiFSI-DME-TTE cell, a weak response peak appeared at 4.72 V. This result indicates that the difference in oxidative stability mainly results from the different interphases formed by the reaction of the two electrolytes with the oxidizing lattice oxygen released from the LRMO cathode.

To further evaluate our AF-LMB designs, the electrochemical performances of LiFSI-DME-TTE and LiFSI-DME cells were examined. Both the LiFSI-DME-TTE and LiFSI-DME cells exhibit a similar Li replenishment effect and highly stable Li plating/stripping as confirmed by the high CEs of Li || Cu coin cells (figs. S7 and S8). Moreover, the two electrolytes show high stability for the Al CC (no corrosion observed) and excellent wettability to the Al₂O₃-coated polyethylene separators (fig. S9 and table S5). However, their electrochemical performances differed substantially in the AF-LMB configuration. The LiFSI-DME cell showed less than 80% capacity retention after 50 cycles and a low average CE of 99.3% within the voltage window of 3.0 to 4.3 V. In contrast, the LiFSI-DME-TTE cell achieved 80% capacity retention after 260 cycles with an average CE of 99.8% (Fig. 2E and figs. S10 and S11). Even when the cutoff voltage was set to 4.4 V (versus Li/Li⁺) during the cycling, the LiFSI-DME-TTE–based AF-LMB still maintained 80% capacity retention after 170 cycles with an average CE of 99.8%, whereas the LiFSI-DME cell decayed rapidly (figs. S12 to S14). Moreover, pouch cells using LiFSI-DME-TTE exhibited no obvious increase in gas volume over 20 cycles at 3.0 to 4.4 V, indicating effectively suppressed parasitic reactions at the electrode-electrolyte interphase. In contrast, cells using LiFSI-DME display continuous gas evolution within 20 cycles (fig. S15).

We show that other hydrofluoroethers (herein 1,1,2,2-tetrafluoroethyl 2,2,2-trifluoroethyl ether, 1,1,2,2-tetrafluoro-3-methoxypropane, and 1H,1H,5H-perfluoropentyl-1,1,2,2-tetrafluoroethyl ether) exhibit a similar stabilizing effect when used as the cosolvent, supporting steady cycling of 2-A·hour pouch cells for over 150 cycles (Fig. 2F). Furthermore, with a higher cutoff voltage of 4.65 V, the LiFSI-DME-TTE cell could still exhibit impressive cycling stability with negligible voltage decay within 50 cycles, whereas the hydrofluoroether-free LiFSI-DME cell experienced fast capacity and voltage fade (Fig. 2, G to I, and fig. S16). Note that, when the cathode is NCM811, although the anode is lithium metal, the battery still suffers fast capacity decay with a high cutoff voltage of 3 to 4.65 V using the LiFSI-DME-TTE electrolyte (average CE: 95.76%) (fig. S17). These results agree well with the oxidative stability of LiFSI-DME-TTE and LiFSI-DME electrolytes when the LRMO cathode was used (Fig. 2D), indicating that merely replenishing the Li resource could not effectively improve the life span of AF-LMB; the stability of the formed CEI during the O oxidation process of LRMO is also essential for the electrochemical performance of the AF-LMB.

To gain deeper insights into our AF-LMB system, we tested the electrochemical performances of 2-A·hour-scale LiFSI-DME pouch cells that are operated under different cutoff voltages in the second formation cycle. As depicted in fig. S18, when the cutoff voltage in the second formation cycle was set lower to 4.4 V (using a 6.5-μm thin Li on the CC to offset the prelithiation effect, the voltage range of the first formation cycle and subsequent cycles was set to 3.0 to 4.4 V), the pouch cell using LiFSI-DME exhibited excellent cyclability with an average CE of 99.7% and maintained 80% capacity retention after 200 cycles. However, despite using a 6.5-μm thin Li foil on the CC to offset Li loss, when the cutoff voltage in the second formation cycle was set to 4.65 V, the cell using LiFSI-DME exhibited severe decay, indicating the importance of TTE in stabilizing the cell. Moreover, as shown in fig. S19, when the cutoff voltage was 4.5 V in the second formation cycle (cycling at 3 to 4.3 V), the LiFSI-DME-TTE cell exhibited better cyclability compared to the LiFSI-DME cell. These electrochemical performances suggest that the formation process in the range of 4.4 to 4.65 V at the cathode side is the primary cause of the electrochemical performance enhancement of LiFSI-DME-TTE over LiFSI-DME pouch cells. We propose that the activation of anionic redox reaction over 4.5 V with the presence of TTE plays an essential role in our AF-LMB system, which is also supported by the galvanostatic intermittent titration technique (GITT) results and the electrochemical impedance spectroscopy (EIS) analysis at 100% state of charge (SOC) (figs. S20 and S21). This observation guides us to examine the side reactions at the cathode surface, components of the CEI (32, 34), and the phase transition of our LRMO cathode (31, 34–37).

To benchmark performance, we further evaluated the electrochemical performance of 2-A·hour anode-free pouch cells using representative carbonate-based electrolytes. As shown in fig. S22, all carbonate-based systems exhibit pronounced capacity fading after ~40 cycles, demonstrating that carbonate electrolytes are not optimal for AF-LMBs under the tested conditions, consistent with previous reports (table S3). Collectively, these results underscore the advantage of localized high-concentration electrolytes incorporating hydrofluoroethers for enabling stable cycling in LRMO-based AF-LMBs.

The F-rich CEI

The CEI of LRMO particles after 5 cycles was studied using high-resolution transmission electron microscopy (HRTEM) (Fig. 3, A and B, and figs. S23 and S24). The LRMO cycled in the LiFSI-DME-TTE cell was covered by a thin and uniform CEI with a thickness of about 3 nm. The crystalline components in the CEI showed a well-defined layered structure, indicating limited TM migration to the rock salt NiO-like structure. In comparison, LRMO in LiFSI-DME exhibited a relatively thick layer of rock salt phase, exceeding 50 nm in thickness (Fig. 3B). In addition, a nonuniform and thick CEI was observed, implying severe cathode-electrolyte side reactions in LiFSI-DME. Notably, a large-scale spinel-like phase with distortion was found on the cathode surface cycled in LiFSI-DME (fig. S25), which is consistent with the x-ray diffraction (XRD) results (discussed in figs. S26 and S27). This finding suggests that oxygen release and TM migration in the cathode cycled in LiFSI-DME may lead to diverse phase transitions, despite the known surface phase transition from a layered to resistive structure that degrades cathode performance (31, 34–38).

Fig. 3. Investigation of structural changes of LRMO and CEI after 5 cycles using different electrolytes.

(A and B) HRTEM images and corresponding FFT (insets) of formed CEI layers and the LRMO cathode after 5 cycles using LiFSI-DME-TTE (A) and LiFSI-DME (B) electrolytes. (C) O 1s XPS spectra of LRMO after 5 cycles using LiFSI-DME-TTE (top) and LiFSI-DME (bottom) electrolytes. (D) O 1s XPS spectra of LRMO after 5 cycles using LiFSI-DME-TTE (top) and LiFSI-DME (bottom) electrolytes after 2-min Ar⁺ etching. (E) C 1s XPS spectra of LRMO after 5 cycles using LiFSI-DME-TTE (top) and LiFSI-DME (bottom) electrolytes. (F) Enlarged C 1s XPS spectra of LRMO after 5 cycles using the LiFSI-DME-TTE electrolyte. (G) F 1s XPS spectra of LRMO after 5 cycles using the LiFSI-DME electrolyte. (H) F 1s XPS spectra collected from the LRMO cathode after 5 cycles using the LiFSI-DME-TTE electrolyte. Ar⁺ etching was conducted to collect information at various depths. (I) Quantified atomic ratios of different elements by XPS conducted on the cathode after 5 cycles in different electrolytes. Scale bars, 20 nm [(A) and (B)].

To further elucidate the structural origin of the unprecedented electrochemical performances of the AF-LMB, we performed x-ray photoelectron spectroscopy (XPS) and Ar⁺ sputtering depth analysis on the cycled cathodes in LiFSI-DME-TTE and LiFSI-DME cells. Before XPS measurements, both the cathodes of LiFSI-DME-TTE and LiFSI-DME cells were sonicated twice in N-methylpyrrolidone (NMP) for 30 min to eliminate the influence of polyvinylidene fluoride (PVDF) binders and residual Li salt. XPS was first used to characterize oxygen redox behavior during the formation process. We observed that the signals in the O 1s spectra contained electrolyte oxidation species (533.0 eV) (20.35% in LiFSI-DME-TTE and 25.54% in LiFSI-DME), oxygenated deposited species (532.0 eV) (42.23% in LiFSI-DME-TTE and 44.53% in LiFSI-DME), and lattice oxygen (529.5 eV) (24.72% in LiFSI-DME-TTE and 14.39% in LiFSI-DME). In addition, the signal of oxidized lattice oxygen species (Oⁿ⁻, n < 2) appeared at 530.9 eV (Fig. 3, C and D) (12.71% in LiFSI-DME-TTE and 15.54% in LiFSI-DME) (28, 39–41). In contrast to LiFSI-DME, the O 1s spectra of LiFSI-DME-TTE displayed a markedly lower signal of oxidized lattice oxygen. In contrast, the oxygen redox occurred substantially in LiFSI-DME, resulting in severe structural degradation and oxygen loss. Moreover, compared to the weaker lattice oxygen signal and stronger electrolyte oxidation species signal for the cathode in LiFSI-DME, the cathode in LiFSI-DME-TTE had fewer electrolyte oxidation species (39, 40), confirming that electrolyte decomposition and side reactions on the LRMO surface were successfully suppressed. This finding aligns well with the HRTEM and SEM studies (figs. S28 and S29).

The C 1s spectra in Fig. 3E revealed the organic species [peaks attributed to C─C/C─H (284.8 eV), C─O (286.3 eV), and C═O (288.7 eV)] in the CEI cycled in LiFSI-DME-TTE (42). In contrast to the C 1s spectra of LiFSI-DME, an extra signal of C─F (290.6 eV) was detected in the CEI cycled in LiFSI-DME-TTE (Fig. 3F). Besides, the F 1s spectra of the CEI in LiFSI-DME-TTE (Fig. 3, G and H) showed a strong signal of both C─F and inorganic F⁻, whereas that of LiFSI-DME exhibited no signal, which was consistent with the result of C 1s spectra. These results strongly indicate that an extra F-rich species was formed in the surface CEI of LiFSI-DME-TTE, and the component of C─F species probably originated from the TTE rather than PVDF binders or LiFSI as no F signal was detected in the cathode in LiFSI-DME. Moreover, after Ar⁺ sputtering, the inorganic F⁻ signal maintained its intensity throughout the whole sputtering process for more than 30 min, indicating that the inorganic F species may have diffused into the interior of the particle. On the other hand, the C─F signal markedly dropped to the noise level after 240-s sputtering, suggesting that the organic C─F was mainly distributed in the CEI (Fig. 3H).

Figure 3I shows the atomic ratios of Li, C, O, F, Mn, and Ni on the LRMO surface in LiFSI-DME-TTE and LiFSI-DME during Ar⁺ sputtering. In the LiFSI-DME-TTE cathode, along with the increasing fraction of Mn, Ni, and O, the F content gradually decayed, which can be ascribed to the signal from the surface to the bulk of LRMO. The Li signal in LiFSI-DME-TTE also remained constant, demonstrating that the phase transition to a resistive Li-poor rock salt structure and other Li-poor phase was limited (37, 43). In contrast, the F and Li signals were not detected for the cathode in LiFSI-DME. Besides, the stronger C signal and weaker O signal for the cathode in LiFSI-DME also revealed the release of unstable lattice oxygen, side reactions at the cathode surface, and the overgrowth of CEI. In conclusion, the XPS results strongly support the in-built SFR scheme.

Impact of the in-built fluorinated LRMO surface

To examine the origin of the high stability of LiFSI-DME-TTE cell, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) imaging was carried out (Fig. 4, A to D, and fig. S30) on LRMO particles after 5 cycles in the two electrolytes. The low-magnification HAADF-STEM images clearly confirm the stabilization of LRMO in the fluorinated cosolvent. As demonstrated by the previous reports (44, 45), the accumulation of inactive O₂ in the particles and release of O₂ from the opening of voids near the surface accounts for the capacity loss and voltage fade. In our system, the LRMO particles after cycling in LiFSI-DME-TTE present very few voids (Fig. 4, A and C, and fig. S30, A and B) whereas those in LiFSI-DME show formation of abundant voids (Fig. 4, B and D, and fig. S30, C and D), suggesting that marginal release of O² occurred for the LRMO cycled in LiFSI-DME-TTE, which is correlated with the higher cycling stability in LiFSI-DME-TTE. Furthermore, electron energy-loss spectroscopy (EELS) line scan was performed on the surface region of the LRMO cathodes (the marked regions shown in Fig. 4, C and D). The F K-edge EELS data confirm the presence of a unique fluorinated layer in LiFSI-DME-TTE (Fig. 4, E and F, and fig. S31). Furthermore, the Mn L-edge for the LRMO cycled in LiFSI-DME-TTE is located at 642.3 eV (at the surface; Fig. 4E), whereas the LRMO cycled in LiFSI-DME is lower at 641.1 eV (Fig. 4F). This is also true for the bulk of LRMO (Mn L-edge: 642.6 eV for cycling in LiFSI-DME versus 643.6 eV in LiFSI-DME-TTE) (Fig. 4, E and F). It is thus clear that the lower Mn oxidation state and hence more severe TM reduction (44) suggest severe TM migration, oxygen release, and Mn dissolution in the LiFSI-DME when a fluorinated cosolvent is absent.

Fig. 4. Impact of the in-built fluorinated LRMO surface structure.

(A and B) Low-magnification HAADF-STEM image of LRMO cathode particles after 5 cycles in 6.9 M LiFSI-DME-TTE (A) and 6.9 M LiFSI-DME (B) electrolytes. (C and D) Enlarged HAADF-STEM images from orange regions in A (C) and B (D). a.u., arbitrary units. (E and F) Mn and F edge of the corresponding EELS line scan of C (E) and D (F). (G) STEM-EDX mapping of F, O, Mn, and Ni elements of the cathode surface after 5 cycles in the LiFSI-DME-TTE electrolyte. (H) HAADF-STEM image along the [100] zone axis of the LRMO cathode after 5 cycles in the LiFSI-DME-TTE electrolyte. Scale bars, 200 nm [(A) and (B)], 20 nm (C), (D), and (G)], and 1 nm (H).

To further investigate the fluorinated LRMO structure, high-resolution energy-dispersive x-ray (EDX) mapping and EELS mapping were also performed on the LRMO cycled in LiFSI-DME-TTE (Fig. 4G and fig. S32), which offers direct insights into the spatial distribution of the F-rich layer at the LRMO cathode surface. The EDX mapping explicitly illustrates the existence of this fluorine-rich phase layer (3 to 5 nm), which homogeneously envelopes the surface of LRMO, consistent with the HRTEM images above [when the cathode material was replaced with NCM811, nonuniform fluorinated CEI was detected (fig. S33)]. This spatial configuration of F-rich layers is further evidenced by the elemental atomic ratio profiling along the surface to the bulk LRMO (fig. S34). The profiles provide a clear illustration of an O-deficient but F-rich area inside the LRMO particle, providing evidence of F substitution of the O sites. Moreover, the EELS F K-edge of the F-rich layer lies at a higher energy (~694.2 eV) compared to that of the LiF reference (fig. S35), which is owed to fluorine-doped cation disordered rock salt phase (46). This is consistent with previous findings where the limited phase transition to disordered rock salt phase and the loss of oxygen contribute to the fluorination of the LRMO cathode (47, 48). In contrast, it has been challenging to fluorinate layered metal oxide cathode materials, which generally results in the formation of LiF instead (fig. S33) (49). The atomic-resolution HAADF-STEM image at the surface (Fig. 4H) confirmed the presence of a thin (3 to 5 nm) disordered rock salt phase along the ($1\overline{2}1$) direction intertwined with the well-defined internal layer phase, in line with the HRTEM image (Fig. 3A). The STEM image inside LRMO demonstrates the two-phase composite of Li₂MnO₃ (C2/m) and LiTMO₂ (TM = Mn, Ni, Co; $R\overline{3}m$). As shown in Fig. 4H, in the Li₂MnO₃ structure, one-third of the TM sites are occupied by Li⁺ ions, manifesting the dumbbell-like superstructure in the TM slabs. This is consistent with the XRD results (fig. S27) as the long-range superstructure was not maintained in the cathode after cycling. This is owed to the nature of our strategy wherein the release of oxidized lattice oxygen (and TM migration) was not completely avoided but intentionally designed to be activated in vicinity of the LRMO surface to attack the fluorinated cosolvent in the formation cycles, which serves to stabilize the LRMO while providing excess Li to the anode in the following cycles (24). In contrast, the EELS of LRMO cycled in LiFSI-DME electrolytes shows no signal of the F element (Fig. 4F). To further investigate the influence of this unique fluorinated interphase, we conducted the XPS analysis and inductively coupled plasma mass spectrometry (ICP-MS) to investigate the components of SEI and electrolytes using LiFSI-DME-TTE and LiFSI-DME electrolytes. The results show that the F-rich layers suppressed the TM dissolution and cathode-to-anode crossover effect, which is important for the stabilization of the SEI film (fig. S36 and table S6). These observations demonstrate the stabilizing effect of the F-rich layers at the LRMO cathode surface.

The mitigated oxygen loss at the cathode

To computationally represent the physicochemical processes observed in experiments as accurately as possible and elucidate the underlying mechanisms involved, including the role of fluorination and the enhancement of cycle stability when TTE is added to the system, it is essential to understand the mechanism from the theoretical point of view by density functional theory (DFT) calculations. First, we calculated the oxygen release energy defined as follows (50)

$$E_{\mathrm{O}}^{\text{rele}}=E_{\mathrm{O}-x^{\prime }}^{\text{slab}}-E^{\text{slab}}+\mathrm{\Delta }{\mathrm{\mu }}_{\mathrm{O}}$$

Here, $E_{\mathrm{O}-x^{\prime }}^{\text{slab}}$ refers to the $x^{\prime }$ oxygen-deficient slab energy and $\mathrm{\Delta }{\mathrm{\mu }}_{\mathrm{O}}$ is defined as the oxygen chemical potential corrected in accordance with previous works to calibrate the formation enthalpies with experiments. Similar to numerous reports in the literature (51, 52), we found that theoretical calculations demonstrated an oxygen release energy of ~−0.72 eV for the (100) LRMO surface, indicating that the loss of lattice oxygen is thermodynamically favorable. Furthermore, with lithium deintercalation during the charging and discharging processes, the thermodynamic spontaneity of oxygen vacancy formation increased, with an oxygen release energy of about −1.64 eV. Given experimental observations that released lattice oxygen might attack the TTE, we subsequently calculated that F from the TTE is thermodynamically capable of substituting the O site with a formation energy of −2.07 eV, providing evidence for the thermodynamic advantage of SFR. The fluorinated structure reveals that the release of O becomes nonspontaneous with a much higher formation energy of about 4.70 eV, providing thermodynamic support for the enhancement of cycle stability. Building on the thermodynamic data, we conducted a detailed investigation of the “oxygen-attack-TTE” process through ab initio molecular dynamics (AIMD). Snapshots from AIMD simulations illustrate the spontaneous loss of lattice oxygen on the LRMO surface, where lattice oxygen approaches TTE clearly. As shown in Fig. 5A and fig. S37, TTE makes sufficient contact with the LRMO cathode surface, and F infiltration into the O sites is prominently observed. This simultaneous occurrence of oxygen loss and fluorine infiltration is considered a crucial process in the surface fluorination of LRMO. A relevant video illustrating this process can also be found in the Supplementary Materials. Therefore, by combining thermodynamic and AIMD analyses, we can provide strong theoretical support for experimental results, assisting in elucidating the LRMO SFR process.

Fig. 5. Schematic diagram of mechanism explanation combined with DFT calculations and AIMD simulations.

(A) Initial state (left) and crucial process (right), which demonstrate the trend of oxygen release and fluorination. (B) Crystal structure model. (C) DOS without fluorination (left) and with fluorination (right). (D) COHP analysis without fluorination (left) and with fluorination (right).

The above results provide the thermodynamic advantage of F entering the O site and some intuitive AIMD animation process; however, it is still imperative to engage in a more in-depth physicochemical discussion on the reason why the introduction of F notably improves cycle stability. Therefore, we further calculated the electronic structures and conducted a crystal orbital Hamiltonian population (COHP) orbital analysis before and after LRMO SFR (Fig. 5B). According to a previous theoretical study (53), anionic redox in LRMO materials is associated with the relative energy position between the filled lower Hubbard band (LHB) caused by Mott-Hubbard splitting. In this theory, the charge transfer term Δ and the d-d Coulomb repulsion term U are two key parameters, which are elaborately described in the Supplementary Materials and illustrated in fig. S38.

In our computational results shown in Fig. 5C, it is evident that, before fluorination, the O 2p orbitals are closer to the Fermi level, and electrons on these orbitals are highly active and prone to participating in reactions, which is an inherent reason for the spontaneous loss of lattice oxygen. When the SFR occurs, the LHB quickly approaches and crosses the Fermi level, reducing the U term and decreasing the priority of O 2p orbital electron transfer. Instead, electrons are provided through TMs, increasing the Δ term and thus suppressing the loss of oxygen and enhancing the overall cycle stability. In general, our results showed an increase in Δ term and a decrease in U term, providing robust evidence for the improvement in cycle stability according to the Mott-Hubbard splitting theory and previous theoretical investigations.

Furthermore, we also analyzed some changes during the fluorination process from the physicochemical perspective. As depicted in Fig. 5D, it is observed that fluorination led to a decrease in the Mn─O bonding states relative to the Fermi level, whereas the antibonding states shifted upward. These changes correspond to a reduction in U term and an increase in Δ term according to Mott-Hubbard splitting theory, which are considered crucial contributions to the enhancement of cycle stability. Moreover, there are notable differences observed in the LRMO crystal structure before and after fluorination, particularly in the differences of the Mn─O octahedra. Before fluorination, the Mn─O octahedra exhibited characteristic Jahn-Teller distortions with Mn³⁺ oxidation states and uneven bond lengths. In contrast, the fluorinated structure revealed Mn─O octahedra with nearly equal bond lengths corresponding to Mn⁴⁺ oxidation states and devoid of Jahn-Teller distortions. Additional statistics data regarding the bond lengths of Mn─O octahedra and the oxidation states of Mn before and after fluorination are presented in table S7. These structural changes were subsequently quantitatively confirmed by the integrated −COHP (−ICOHP) values. The results of -ICOHP values indicate that the introduction of F leads to a notable increase in the Mn─O bond strength from 1.38 to 2.52, which quantitatively illustrates that the introduction of F contributes to enhancement in the strength of Mn─O bonds. From the perspective of electronegativity, the high electronegativity of fluorine (F, with an electronegativity of 3.98) causes the electron cloud to shift toward it when introduced into a system with Mn─O bonds, hence affecting the electron cloud density of the Mn─O bond. The induced effect increases the bond polarity, enhancing the iconicity of the Mn─O bond. Furthermore, the redistribution of electron cloud leads to a change in the charges around the atoms, strengthening the Coulomb attraction between Mn and O atoms. The spatial hindrance and electrostatic balance can also be regulated in a way for a stronger Mn─O bond interaction. We also propose that the introduction of F may also change the orbital hybridization of Mn and O atoms, which increases the bond order and enhances the bond strength. On the basis of the above discussion, a fundamental understanding at the physicochemical level is provided to suppress the loss of oxygen, thereby offering microscopic insights into the improvement of cycle stability.

Electrochemical performances of 30-A·hour-scale AF-LMBs

To further demonstrate the in-built SFR in stabilizing AF-LMBs, we fabricated a 30-A·hour-scale anode-free pouch cell using the LRMO cathode and the LiFSI-DME-TTE electrolyte. The parameters for the cell assembly are listed in table S8. The 30-A·hour cell exhibited well-defined charge/discharge curves within the 3 to 4.4 V at 0.1 C (~0.40 mA cm⁻²) after the formation process (Fig. 6A) and was calculated to exhibit a high gravimetric energy density of 350 Wh kg⁻¹ and a substantial volumetric energy density of 1200 Wh liter⁻¹ (table S8). Impressively, the high-energy cell showed excellent cycling stability, maintaining 84.4% capacity retention after 180 cycles (Fig. 6B and fig. S39). To further assess the cell under more practical conditions, we tested the rate capability and performance within a wide temperature range. The results showed that the 30-A·hour cell had a high-rate discharge capability, with 29.7, 28.5, 27.8, 27.1, 26.5, 26.4, and 26.3 A·hour at 0.1 C, 0.2 C, 0.3 C, 0.5 C, 1 C, 2 C, and 3 C (voltage range of 3.0 to 4.4 V), respectively (Fig. 6C and fig. S40). Similarly, this cell has a good adaptability to both high and low temperatures (Fig. 6D and fig. S41). At the low temperature of 0°, −10°, and −20°C, a reversible capacity of 26.2 A·hour (88.5%), 23.4 A·hour (78.8%), and 20.5 A·hour (69.0%) was achieved, respectively. Even at −30°C, the cell still maintained 15.3 A·hour (51.5% of the capacity at 25°C). Furthermore, safety assessment experiments were conducted under harsh testing conditions (table S9). After the free-fall, flat crush, low pressure, vibration, forced discharge, and external short tests, the pouch cells did not show any leakage, fire, or explosion, demonstrating the exceptional safety of the 30-A·hour-scale AF-LMB.

Fig. 6. Electrochemical performances of 30-A·hour anode-free pouch cells using the 6.9 M LiFSI-DME-TTE electrolyte.

(A) Charge/discharge curve of the 30-A·hour AF-LMB at 0.1 C (25°C). (B) Capacity retention and CE of the 30-A·hour pouch cells at 0.2 C charge/0.5 C discharge (0.1 C for first three formation cycles) and photograph of the 30-A·hour AF-LMB pouch cell (left inset) and 30-A·hour AF-LMB pouch cell battery module (right inset). (C) Charge/discharge profiles of the 30-A·hour AF-LMB pouch cells with various discharging currents ranging from 0.1 C to 3 C (charging current: 0.1 C). (D) Temperature-dependent discharge profiles of the 30-A·hour AF-LMB pouch cells. (E) Radar plot showing the comparison of our 30-A·hour pouch cells and advanced ampere hour–level AF-LMBs in recent years. (F) Summary of advanced electrochemical performances of AF-LMBs in recent years (excluding the weight of the packaging).

Moreover, as shown in Fig. 6 (E and F) and tables S10 and S11, the key parameters of our high-energy-density anode-free pouch cells are compared to the previous reports of anode-free pouch cells. Our work stands out for its exceptional cost-effectiveness, high gravimetric and volumetric energy density, and impressive cyclability (8, 10–14, 16, 17, 23–25). Notably, our work represents the demonstration of long-term high-capacity anode-free pouch cells that show commercial viability. In addition, we successfully assembled a 30-A·hour-scale anode-free battery pack to achieve higher voltage output (22.8 V and 684 Wh) (Fig. 6B, right inset). Furthermore, we analyze how the energy density of lithium metal cells is affected by excess lithium (fig. S42). Compared to lithium metal batteries with excess Li and lithium-ion batteries, our anode-free cells stand out for the high energy density, environmental benignity, and simplified related manufacturing process. The success achieved at the pouch cell level shows great promises for advancing the commercial viability of a more practical anode-free battery.

DISCUSSION

AF-LMBs are acknowledged as a viable solution for achieving highly safe and cost-effective Li metal batteries with high energy density. In this work, a 2-A·hour anode-free pouch cell has been demonstrated with a capacity retention of more than 80% over 260 reversible cycles. In addition, a 30-A·hour-scale anode-free pouch cell with a capacity retention of 84.4% after 180 reversible cycles was realized to show the promising commercial viability of our technology. The key achievement lies in the realization of Li replenishment and in situ SFR of the LRMO cathode in our AF-LMBs. Different from previously reported approaches of fluorination to enable high-voltage CEI (see detailed discussion in table S12), our work introduces an in-built SFR mechanism, in which fluorination occurs in situ through the controlled reaction between oxidized lattice oxygen and a fluorinated cosolvent during the formation process. This strategy enables uniform and self-limited fluorination localized at the LRMO surface, leading to a thin (~3 nm) LiF-rich CEI and fluorination of the LRMO surface, which facilitate more reversible oxygen redox, stabilize lattice oxygen, suppress overgrowth of CEI, and notably enhance the high-voltage cycling stability of our AF-LMB during subsequent cycling. The effectiveness of this interphase is directly supported by atomic-scale characterization. As a result, our AF-LMBs with LRMO can operate at a high cutoff voltage of 4.65 V for over 50 cycles with negligible capacity loss and voltage decay in the ether-based LiFSI-DME-TTE electrolyte system.

The proposed methodology provides fundamental insights for the design of AF-LMBs with prolonged life and high energy density and represents a substantial advancement in terms of stabilization of LRMO by low-cost and facile fluorination methods. By integrating DFT calculations and AIMD analyses, we demonstrate changes induced by fluorination in both the electronic and crystal structures, comprehending the SFR on the LRMO surface. Fluorination is found to cause a reduction in the Mn─O bonding states relative to the Fermi level, whereas the antibonding states shift upward for electronic structures; also, the introduction of F leads to a substantial increase in the Mn─O bond strength to stabilize the crystalline structures. These changes are considered crucial contributions to the enhancement of cycle stability. Moreover, the SFR mechanism can be extended to other types of hydrofluoroether cosolvents, demonstrating the universality of our strategy. Therefore, this study not only provides theoretical guidance and valuable insights into stabilization of LRMO cathodes and the Li metal anodes but also paves the way for the commercialization of a practical high-energy anode-free battery.

MATERIALS AND METHODS

Electrolyte preparation

The DME solvent (Sigma-Aldrich, purity > 99%) and TTE (Sigma-Aldrich, purity > 99%) were dried using freshly activated 4-Å molecular sieves for several days. LiFSI (purity > 99%, Wako Chemicals) was dried by subjecting it to vacuum heating at 80°C in an oven overnight. The high-concentration electrolyte, referred to as LiFSI-DME, was prepared by dissolving 6.9 M LiFSI into DME at room temperature within an Ar-filled glove box with both H₂O and O₂ concentrations below 0.01 parts per million (ppm). To create a localized high-concentration electrolyte, additional TTE was introduced with a volume fraction of 70%. The water content in the electrolyte, determined via Karl Fischer titration, was ~10 ppm.

Preparation of pouch cells

Ampere hour–level battery cells were mass produced on the production line of Zhejiang Jinyu New Energy Technology Co. Ltd. Specifically, the cathode active material (Li_1.11Ni_0.34Mn_0.53Al_0.02O₂; produced by Jiangxi Hanyao Rich Lithium Technology Co. Ltd.), carbon black powder (SUPER P Li; produced by TIMCAL), and binder (PVDF; manufactured by SOLVAY) were mixed in a ratio of 97.3:1.5:1.2 in the NMP solvent within a 30-liter mixing tank. The slurry had a solid content of 74 ± 1% and a viscosity of 5500 to 8500 mPa·s. This slurry was then uniformly coated on both sides of 12-μm-thick aluminum foil (with a single-side loading of 25 ± 0.38 mg). During coating, the drying temperature was controlled at 95 ± 5°C. The coated electrodes were subsequently calendared to achieve a compacted density of 2.8 g cm⁻³.

The cathodes and copper foils were cut to design-specific dimensions, which varied according to the cell model. For a 2-A·hour cell, the cathode and copper foil dimensions were 30 mm by 14 mm and 32.8 mm by 16 mm, respectively, whereas for a 30-A·hour cell, these dimensions were 156 mm by 69 mm and 159 mm by 71 mm. The 2-A·hour pouch cell consisted of 12 cathode layers and 13 anode (Cu CC) layers, whereas the 30-A·hour pouch cell consisted of 35 cathode layers and 36 anode layers. The bare cells were assembled on a stacking machine using a Z-shaped stacking process with a 14-μm-thick Al₂O₃-coated PE separator. Following tab welding, aluminum-laminated packaging, and top and side sealing, the dry cells were obtained.

The electrolyte was injected under an inert atmosphere, with injection volumes of 3 g A·hour⁻¹ for 2-A·hour cells and 2.8 g A·hour⁻¹ for 30-A·hour cells. The cells were then aged at 45°C for 24 hours and at 25°C for another 24 hours. Subsequently, they underwent the formation process as described in table S4. After resting at room temperature for 10 to 14 hours, the finished battery cells were obtained. Detailed cell parameters are provided in table S8.

Electrochemical measurements

After completing the formation process, the pouch cells were degassed, sealed, and then subjected to undergo galvanostatic cycling tests using battery testers (Wuhan Lanhe, China). The cells were charged at a rate of 0.2 C (~0.80 mA cm⁻²) up to 4.3 V and discharged at a rate of 1 C (~4.0 mA cm⁻²) down to 3.0 V. The GITT was used to calculate the Li⁺ ion diffusion coefficient (D_Li+) (54). This involved discharging/charging the electrode with a pulse current of 0.1 C for 10 min, followed by a 60-min relaxation at open circuit to reach equilibrium potentials. EIS was performed with a 5-mV amplitude over a frequency range from 100 kHz to 0.01 Hz. The EIS measurements shown in fig. S21 were conducted after the second full charging cycle (i.e., at the end of the second charge) when the cells reached 100% SOC at a cutoff voltage of 4.65 V. Linear sweep voltammetry (LSV) was conducted using an electrochemical station (BioLogic VMP3) at a scan rate of 1 mV s⁻¹, using a Li || Al configuration and at a scan rate of 0.1 mV s⁻¹, using a Li || LRMO configuration. Li || Cu half-cells were assembled by lithium chips (11 mm in diameter, 0.6 mm in thickness), Al₂O₃-coated PE separators, and Cu chips (16 mm in diameter, 0.6 mm in thickness). Li || Cu half-cells were tested by depositing 0.5 and 4 mA·hour cm⁻² (current density: 0.5 mA cm⁻²) of Li onto the Cu electrode followed by stripping to 1 V. All electrochemical tests were executed at a temperature of 25°C.

Materials characterization

XRD was performed on a Rigaku SmartLab X-ray diffractometer using Cu Kα radiation (λ = 1.5406 Å) at a scan rate of 5° min⁻¹. The SEM images were obtained by field-emission scanning electron microscope (SU8220, Hitachi, Japan). For TGC measurement, the LRMO || Cu pouch cells were first charged to 4.65 V and then fully discharged to 1.0 V in the second formation cycle to quantify the dead Li. The cells were charged to 4.65 V and discharged to 2.0 V in the second formation cycle to determine the total amount of supplemented Li reservoir. The difference between these two measurements was used to evaluate the amount of electrochemically active lithium formed during the formation process. After electrochemical cycling, the pouch cells were disassembled in an argon-filled glove box. Samples were punched from the copper anode together with the attached separator and subsequently sealed in 5-ml headspace vials. Outside the glove box, 1 ml of ethanol was injected into each vial to react with metallic lithium, generating hydrogen gas. A gastight syringe was then used to extract 200 μl of the evolved gas, which was injected into GC system for quantitative H₂ analysis. For XPS studies, cycled pouch cells were disassembled in the glove box with O₂ and H₂O level < 0.01 ppm and then washed with pure DME and sonicated twice with NMP for 30 min. XPS was conducted on an AXIS Supra X-ray photoelectron spectrometer (Kratos Analytical Ltd.) using Al Kα radiation. Binding energy values were referenced to the C 1s peak at 284.8 eV. XPS depth profiling was performed using Ar⁺ sputtering at a beam energy of 5 keV over an etching area of 2.5 mm by 2.5 mm. The Ar⁺ sputtering rate was calibrated using a standard TiO₂ reference sample and determined to be ~2 nm min⁻¹. The CasaXPS software was used to fit the XPS spectra. HRTEM, HAADF-STEM, EDX mapping, and EELS of the LRMO cathode were performed on an aberration-corrected JEM-ARM300F2 microscope, equipped with a cold field-emission gun operating at 300 kV. The obtained TEM images were converted to the fast Fourier transform (FFT) and inverse FFT images through the Gatan Microscopy Suite software.

DFT calculations.

DFT calculations were performed using the Vienna ab initio simulation package (VASP code) (55, 56). The relative data were obtained using the generalized gradient approximation (GGA) expressed with the formulation of Perdew-Burke-Ernzerhof (PBE) for the exchange-correlation term. The strong on-site Coulombic interaction of 3d TMs were corrected using the Hubbard U parameter. According to the previous literature, the U values are set to 3.9 and 6.2 for Mn and Ni, respectively (52, 57). Projector augmented wave (PAW) pseudopotentials were used to describe the interaction of the core electrons. The cutoff energy was set to be 520 eV for all calculations. The convergence tolerances for energy and force were 10⁻⁵ eV and 0.01 eV/Å, respectively. The calculations for electronic structures were performed with a Γ-centered k-points mesh 5 by 5 by 1 for surface systems.

The AIMD simulations were performed using the Vienna ab initio simulation package (VASP code). The initial structures were generated by the Packmol program to optimize the random atomic positions (58). The relative data were obtained using NVT (constant number, volume, and temperature, also known as the canonical ensemble) simulations at 300 K with the Nosé-Hoover thermostat. The timestep was set to be 1 fs with 20,000 simulation steps. The COHP orbital calculations were performed using the LOBSTER (Local Orbital Basis Suite Towards Electronic-Structure Reconstruction) program (59).

Supplementary Materials

The PDF file includes:

Figs. S1 to S42

Tables S1 to S12

Legend for movie S1

References

Other Supplementary Material for this manuscript includes the following:

Movie S1