A high-capacity cathode for rechargeable K-metal battery based on reversible superoxide-peroxide conversion

Abstract As a promising low-cost energy storage device, the development of a rechargeable potassium-ion battery (KIB) is severely hindered by the limited capacity of cathode candidates. Regarded as an attractive capacity-boosting strategy, triggering the O-related anionic redox activity has not been achieved within a sealed KIB system. Herein, in contrast to the typical gaseous open K-O2 battery (O2/KO2 redox), we originally realize the reversible superoxide/peroxide (KO2/K2O2) interconversion on a KO2-based cathode. Controlled within a sealed cell environment, the irreversible O2 evolution and electrolyte decomposition (induced by superoxide anion (O2−) formation) are effectively restrained. Rationally controlling the reversible depth-of-charge at 300 mAh/g (based on the mass of KO2), no obvious cell degradation can be observed during 900 cycles. Moreover, benefitting from electrolyte modification, the KO2-based cathode is coupled with a limited amount of K-metal anode (merely 2.5 times excess), harvesting a K-metal full-cell with high energy efficiency (∼90%) and long-term cycling stability (over 300 cycles).


INTRODUCTION
Enlarging the energy density is a universal and eternal topic for energy storage devices, especially for low-cost potassium-ion battery (KIB) technology, in which the limited specific capacities of the cathode seriously hinder its further development [1][2][3][4][5][6]. Due to the confined choices of cathode candidates, sluggish progress has been made on the cathode side of KIBs [7][8][9].
Introducing the O-related anionic redox activity into the cathode reaction has been regarded as a promising way to boost the specific capacity, e.g. Li/Na-O 2 batteries and Li/Na-rich layered oxide cathodes [10][11][12][13]. Actually, as for the potassiumbased system, the non-aqueous K-O 2 battery relying on O-related pure anionic redox (the conversion between gaseous O 2 and KO 2 ) was originally introduced by Wu's team in 2013 [14], and has made remarkable progress with regard to the aspects of electrolyte stability and KO 2 /electrolyte interface in the past seven years [14][15][16][17][18][19][20]. However, the K-O 2 battery would inevitably suffer from the intrinsic challenges of the typical gaseous open battery system [21,22]. For instance, during the practical evolution from K-O 2 to K-air battery, the moisture and CO 2 from the air would essentially influence both the cathode reaction pathway and anode stability [23], while carrying the cumbersome supplementary facility (either air purifier devices or an O 2 storage cylinder) would add huge burden on improving energy density. In this case, there is still a long way to go to evolve the gaseous open battery system into a practical energy storage device, while controlling the redox among various solid phases within a typical sealed cell environment seems to be more practical [21,22,24,25], e.g. interconversion among superoxide, peroxide and oxide for KIB technology. Unfortunately, so far no attempt has been reported to successfully confine the high-energy-density O-related anionic redox reaction within a practical sealed cell environment in a KIB. The pioneering realization of the K-O 2 battery has demonstrated that KO 2 can be regarded as a thermodynamically stable compound for K-ion storage [14,26]. Furthermore, enlightened by the recently-arisen oxide/peroxidebased cathode reaction in lithium-related battery systems (Li 2 O 2 /Li 2 O interconversion) [24,25], we propose a cathode reaction that operates via superoxide/peroxide (KO 2 /K 2 O 2 ) interconversion, in which a theoretical specific capacity of 377 mAh/g (based on the mass of preloaded KO 2 ) can be expected at a thermodynamic equilibrium potential of 1.92 V vs. K/K + [27]. In this proposed system, not only the traditional electrolyte stability issue (against superoxide anion, O 2 − ) should be considered, but also the charge overpotential should be rationally controlled to restrain O 2 evolution (corresponding thermodynamic equilibrium potential of OER is summarized in Supporting Data).
In this work, the proposed non-O 2 reversible KO 2 /K 2 O 2 interconversion has been successfully achieved by embedding KO 2 into a catalytic ruthenium dioxide (RuO 2 ) nanoparticles-loaded reduced-graphene-oxide (rGO) matrix. Demonstrated as the culprit for irreversible O 2 evolution and electrolyte decomposition, the formation of a superoxide anion (O 2 − ) has been restrained by the fabrication of a unique K-deficient K 1-x O 2 -Ru intermediate state on the electrode/electrolyte interface during charging. As a result, the half-cell presents merely 0.2 V average round-trip overpotential and provides a reversible capacity of 300 mAh/g (based on the mass of preloaded KO 2 ) during 900 cycles. Moreover, benefitting from electrolyte modification, the corresponding full-cells deliver superior cycling stability after being assembled with a limited amount of excess K-metal anode, which is remarkably competitive with other state-of-the-art KIB systems.

RESULTS AND DISCUSSION
In order to make fair comparisons with regard to cathode energy density in KIBs, the cathode candidates should be grouped into two categories (based on different pristine states): K-contained and K-deficient states ( Fig. 1a and Table S1). As for the proposed KO 2 /K 2 O 2 redox process, the pristine KO 2 state is the K-deficient state, which has a prominent advantage over other candidates (Prussian blue cathode, etc.) [8,9,[28][29][30][31]. Meanwhile, compared to other K-contained cathodes (layered oxides, etc.) [2,32,33], the corresponding K 2 O 2 (K-contained discharged state) also presents large superiority on the aspect of energy density. As the cathode matrix, RuO 2 @rGO nano-composite is prepared via a typical microwave-hydrothermal method [34], in which the RuO 2 nanoparticles (NPs, 2-3 nm in size, RuO 2 : 75 wt%) are homogeneously dispersed onto the rGO sheet (Fig. S1). After sufficient ball-milling procedure, KO 2 is mixed with the RuO 2 @rGO matrix (KO 2 : 60 wt%), and KO 2 -RuO 2 @rGO cathode composite can be harvested, which is identified by X-ray diffraction (XRD) characterization (Fig. 1b). Visual transmission electron microscopy (TEM) images clearly show that KO 2 NPs (10-15 nm in size) are well dispersed and embedded within the RuO 2 @rGO matrix (inset Fig. 1b and Fig. S2). In the bulk cathode, the stacking of 2D rGO sheets would finally lead to an interlaced 3D network, in which the conductive pathway can be well guaranteed by the enfolding mode of KO 2 and rGO substrate, while the tiny catalytic RuO 2 NPs intimately surround the KO 2 active materials, and sufficiently enrich their contact sites. The role of RuO 2 will be further interpreted subsequently.
Within a half-cell, the as-prepared KO 2 -based cathode is coupled with a huge excess amount of K-metal (merely for cathode assessment); the electrolyte issue will be discussed in the following full-cell section. As for RuO 2 -free KO 2 -rGO cathode (gray trace, Fig. 1c), the discharge process is initially conduced for KO 2 reduction, while the subsequent oxidization/charging potential rapid climbs and then surpasses the thermodynamic oxygen evolution reaction (OER) potentials of both K 2 O 2 and KO 2 (2.2 V and 2.48 V, respectively). A very different electrochemical behavior can be observed on the KO 2 -RuO 2 @rGO cathode (blue trace, Fig. 1c), in which a long flat plateau can be obtained with an average potential of 1.88 V vs. K/K + (0.06 V overpotental vs. thermodynamic potential) and a specific capacity of 355.6 mAh/g (94.3% depth-of-discharge vs. theoretical capacity). In order to control the reduction process, the cutoff voltage has been set at 1.5 V, at which the inactive RuO 2 -rGO substrate would not participate in the electrochemical redox process (Fig. S3). During subsequent charging, a plateau can be harvested with tiny polarization during the initial depth of 300 mAh/g. While, at the end of charging, the polarization phenomenon can be ascribed to the overcharge and OER. Turning to cycling performance (Fig. 1d), after a pre-discharging procedure, the KO 2 -RuO 2 @rGO cathode presents long-term cycling stability with a cutoff chargedepth at 300 mAh/g (to avoid undesired OER). After 750 cycles, obvious potential polarization can be observed at the end of charging, and the final charge potential touches the 'OER redline' (K 2 O 2 OER potential: 2.2 V) at the end of the 750th cycle. During long-term cycling, the discharge/charge overpotential has been effectively restrained (Fig. 1d, inset), resulting in an average round-trip efficiency of 88.9% at a large current rate of 300 mA/g. Taking comprehensive assessment aspects into consideration (current density, cycling stability, specific capacity, energy/coulombic efficiency, etc.), the performance of the currently-introduced KO 2 -based cathode delivers considerable improvements beyond other reported cathode candidates in KIBs (Table S2). Moreover, the mass loading of KO 2 is controlled around 3.0-4.0 mg/cm 2 , thus, the areal specific capacity of the current cathode can reach 0.9-1.2 mAh/cm 2 , which is indeed competitive even compared with typical K-O 2 open battery systems.
Before systematically analyzing the electrochemical redox behavior, the difference between the superoxide (KO 2 ) and superoxide anion (O 2 − ) should be emphasized in advance. Typically, within non-aqueous metal-O 2 batteries, the O 2 molecule is initially reduced to an O 2 − anion (oxygen reduction reaction, ORR), which would carry out a strong nucleophilic attack on organic electrolyte components, resulting in electrolyte decomposition [35,36]. After being combined with K-ion and deposit as a solid state (KO 2 superoxide compound), the nucleophilic character can be essentially restrained [17,20,26]. Based on the nuclear magnetic resonance (NMR) observations on ether solvent (tetraethylene glycol dimethyl ether, TEGDME, G4), after aging with KO 2 powder for one week, no obvious decomposition can be observed (blue trace, Fig. 2a). Once trace 18-crown-6 is added to extract the K-ion, the superoxide (KO 2 ) would dissolve and Natl Sci Rev, 2021, Vol. 8, nwaa287   convert into the relative free anion state (O 2 − ) [17]. Without the stabilization of K-ion, the O 2 − (radical) induced electrolyte decomposition, resulting in the observation of formate and acetate (8.39 and 1.92 ppm), respectively (red trace, Fig. 2a) [36]. In this case, as for the superoxide-involved battery systems, the solubility of superoxide in the electrolyte system should be strictly controlled (Fig. S4). In other words, as an 'Achilles' heel', the superoxide anion (O 2 − ) is an unescapable discharge intermediate product in typical metal-O 2 batteries during ORR [37]. However, upon KO 2 /K 2 O 2 interconversion process, the production of O 2 − can be reasonably eradicated, if we protect KO 2 against further oxidation during charging (limit the redox between KO 2 /K 2 O 2 , but do not extend to O 2 /KO 2 ). In summary, as the chief culprit for both electrolyte decomposition and irreversible O 2 evolution, the formation of O 2 − should be critically restrained, which is the core mission for the achievement of reversible KO 2 /K 2 O 2 interconversion.
Operando Raman spectroscopy is firstly employed to reveal the electrolyte redox process of KO 2 -based cathodes (Fig. 2b), in which the skin depth of typical Raman (0.1-5.0 μm related to electronic conductivity and magnetic permeability) can be regarded as a bulk measurement as the particle size of KO 2 lies between 10-15 nm. As assigned in Fig. S5, the sharp peak located at 1146 cm −1 indicates the O-O stretch within the KO 2 -preloaded cathode at the open circuit voltage (OCV) state [38]. The spectra collected from the KO 2 -rGO and KO 2 -RuO 2 @rGO cathodes share a similar variation trend upon the initial discharging processes (around 175 mAh/g), during which the reduction of superoxide peak is accompanied by the growth of the peroxo-related one (at 759 cm −1 ) [39]. However, upon subsequent charging on KO 2 -rGO (left group, Fig. 2b), the peroxide has not been totally oxidized, nor the superoxide restored, which indicates an irreversible redox process. As a comparison, on the KO 2 -RuO 2 @rGO cathode (right group, Fig. 2b), symmetrically reversible variation trends can be clearly observed for both peroxide consumption and superoxide restoration, which indicates the achievement of reversible KO 2 /K 2 O 2 interconversion. The corresponding capacity-dependence of Raman peak intensities and offset Raman spectrum are shown in Figs S6 and S7 for clarity. Strictly speaking, due to the relationship between skin depth and absorbance, operando Raman intensity variation can be regarded as a general quantification analysis, but not a rigorous one. In this case, by the combination of 'double-check' titrations and a gas chromatographmass spectrometer (GC-MS) characterization, we develop a more precise quantitative method [25], and the results reprove the reversible interconversion between KO 2 and K 2 O 2 during cycling on the KO 2 -RuO 2 @rGO cathode (Figs S8 and S9).
Continuously, we employ in situ shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS) [40], a surface-sensitive operando observation (probe depth: ∼1.5 nm), to investigate the behaviors on the electrode surface/interface, which are essentially helpful to deeply reveal the electrochemical redox processes (Fig. 2c). Regardless of the broad peak around 1150 cm −1 assigned to the electrolyte (consistently existing during discharge/charge), a sharp peak at 1146 cm −1 (O-O stretch in KO 2 ) can be observed at the pristine state of each KO 2 -preloaded cathode, which rapidly disappears at the initial stage of discharging. This indicates the surface of KO 2 NPs has been reduced to K 2 O 2 , which shields the signal from the bulk KO 2 core. While, from the beginning of charging, a sharp difference can be observed between KO 2 -rGO (left pattern, Fig. 2c) and KO 2 -RuO 2 @rGO (right pattern, Fig. 2c) cathodes. On the surface of the KO 2 -rGO cathode, the peak located at 1101 cm −1 gradually increases during charging, which is assigned to the adsorbed superoxide anion species (Fig. S5) [37]. Simulta-neously, the appearance of the carbonate-related peak (at 1063 cm −1 ) can be rationally ascribed to the superoxo-induced electrolyte degradation [41]. While, on the KO 2 -RuO 2 @rGO cathode surface, a new peak located at 1137 cm −1 appears and rapidly becomes saturated at the initial stage of charging. In principle, at the very beginning of charging, the K 2 O 2 (formed upon discharging) on the surface lose a K + and are oxidized to the form of superoxo species. Without RuO 2 , surface K 2 O 2 directly evolved into free superoxide anion, which presents strongly nucleophilic properties and can easily be further decomposed. In the RuO 2 -contained condition, the surface species can be rationally assigned to superoxide rather than superoxide anion, since it demonstrates a more combining state than a free anion state. Explained by Raman-related language, the peak located at 1101 cm −1 (free anion state, RuO 2 -free condition) indicates a larger O-O bond distance than the 1137 cm −1 peak observed in the RuO 2 -contained condition [38,39,42,43]. In this case, the interaction between K + cation and superoxo-anion presents much more strongly at 1137 cm −1 (RuO 2 -contained condition), and close to the fully-compound state (KO 2 , 1146 cm −1 ) [37]. In other words, rather than oxidize to a free anion state upon charging, the KO 2 -RuO 2 @rGO cathode can be protected by a moderate K-deficient superoxide compound state (K 1-x O 2 state), a more stable and mild intermediate state than nucleophilic anion (O 2 − ). Moreover, the electrolyte decomposition and carbonate formation cannot be observed on the KO 2 -RuO 2 @rGO cathode during charging, which is consistent with the moderate property of K-deficient K 1-x O 2 state.
Based on the surface investigations and related comparison, the RuO 2 plays an important role for the as-obtained K-deficient superoxide state during charging. Hard X-ray absorption spectroscopy (XAS) is employed to further investigate the role of RuO 2 (Fig. 3a). Compared with the K-edge Ru spectrum collected from RuO 2 standard, the discharged state stays on Ru 4+ , since we cannot observe any obvious difference on near edge (XANES) nor extended region (EXAFS). However, at charged state, the Ru K-edge peaks shift towards higher energy, which can be attributed to the oxidation of Ru 4+ into a higher valence state. The first derivative plot of XANES region (inset Fig. 3a) clearly shows the peaks shift. Due to the negligible charge capacity contribution from Ru 4+ on KO 2 -free RuO 2 @rGO cathode (Fig. S3), the RuO 2 would not directly participate in the charge compensation, but there still exists an interaction between RuO 2 and surface KO 2 , resulting in the  formation of a surface K-deficient K 1-x O 2 phase. On the aspect of molecular orbital, the covalency state between 4d transition metal (e.g. Ru) and peroxolike (O 2 ) n− species has been reported to enhance the stability of the O-related anionic redox process via orbital hybridization (4d-σ * remixing) and reductive coupling formation [44,45]. Moreover, a similar intermediate phase has been reported within the Li-deficient Li 2-x O 2 phase, which is assumed to prevent peroxide against OER decomposition [25,46]. In this case, combining the systematical information harvested from Raman (formation of the K-deficient superoxide phase) and XAS (interaction with Ru 4+ ), we rationally conclude that the formation of a K-deficient K 1-x O 2 -Ru surface phase prevents the further de-potassiation and O 2 − formation, while, without the stabilization from RuO 2 , the unstable intermediate K-deficient superoxide phase would immediately turn to anion state during subsequent charging. The electrochemical redox process is schematically illustrated in Fig. S10. Moreover, without the formation of the K-deficient K 1-x O 2 -Ru surface phase, a free O 2 − anion would either suffer from further oxidation or attack the electrolyte, resulting in the release of both O 2 and CO 2 , which was proven by in situ differential electrochemical mass spectroscopy (DEMS, Fig. 3b). As a comparison, O 2 evolution cannot be observed on the KO 2 -RuO 2 @rGO cathode during five reversible cycles. This negative gas evolution coincides well with the reversible KO 2 /K 2 O 2 redox process, which was previously demonstrated by both electrochemical and spectroscopic evidence.
Finally, we want to further extend the currentlyintroduced cathode KO 2 /K 2 O 2 redox reaction from proof-of-concept half-cell stage into the full-cell level. As shown in Fig. 4a, after being coupled with different anode candidates (carbon and K-metal, etc.) [47][48][49][50], the full-cell energy density would suffer from an obvious drop versus corresponding cathode energy density. The related conversion and calculation procedures on full-cell energy densities are illustrated in Fig. S11. In order to boost the full-cell energy density, we employ the most challenging K-metal anode for full-cell fabrication, and assemble the full-cell with limited amount of excess K-metal as anode (Fig. 4b  and c). Relatively speaking, the K-contained cathodes (layered oxides, etc.) present as being more suitable to couple with K-deficient anodes (graphitic and phosphorus materials, etc.), in which the K-ion stored in the cathode would meet the ion transfer. While, as for K-deficient cathodes (Prussian blue, TiS 2 , organic cathodes, O 2 cathode and KO 2 herein, etc.), the K-contained metallic potassium would be the priority anode candidate. The electrolyte modification is remarkably important for enhancing the reversibility and efficiency of the K-metal anode. Benefitting from the positive role of classic bis(fluorosulfonyl) imide (FSI) anion [17], we further introduce a trace amount (5 wt%) of fluoro-ether additive to strengthen the stability of the K-metal surface (e.g. solid-state-interface) in the ether-based electrolyte system, resulting in a K-K symmetric cell with 1000 hours long-term plating/stripping stability (Fig. S12) and K-Cu half-cells performing high coulombic efficiency (CE% > 99%) before ∼250 cycles (Fig. S13). Herein, we do not carry out a deep investigation/characterization on the K-metal anode side, but provide the performance of fluorinated ether/carbonate candidates as a promising additive towards K-metal protection (Fig. S14). After introducing the highly-efficient K-metal system, we assemble the full-cell with merely 100% excess amount of K-metal anode (Fig. 4b).
Due to the utilization of capacity-cutoff cycling mode (same cutoff depth as half cell: 300  mAh/g∼KO 2 ), the cycle life and round-trip energy efficiency present as the most important parameters (Fig. S15). As a result, based on a fair calculation, the initial output energy density of the current full-cell reaches 270 Wh/kg (taking the mass of both KO 2 and K-metal into consideration). Meanwhile, due to the effective restraining of potential polarization, the average energy efficiency remains at around 90% during 147 reversible cycles. After enlarging the excess amount of K-metal anode to 250%, reversible cycling can be achieved over 300 cycles with considerable round-trip efficiency (Fig. 4c).

CONCLUSION
In conclusion, for the first time, we realized the reversible redox of superoxide/peroxide (KO 2 /K 2 O 2 ) in potassium-ion battery cathode systems with a considerable specific capacity (300 mAh/g∼KO 2 ), tiny potential polarization (∼0.2 V round-trip overpotential), high current rate (300 mA/g) and long-term cycling stability (around 900 cycles in the half-cell). By embedding the KO 2 into the RuO 2 @rGO matrix, the reversible KO 2 /K 2 O 2 interconversion was confined within O 2 − /O 2 -free region, which was proven by systematically in/ex situ quantitative/qualitative characterizations. Moreover, based on bulk/surfacesensitive spectroscopic observations, we originally demonstrated the formation of a TM-covalent K-deficient superoxide intermediate surface phase (K 1-x O 2 -Ru), which protected the cathode against irreversible O 2 − /O 2 evolution upon charging. In addition, benefitting from electrolyte modification, we coupled the KO 2 -based cathode with limited excess amount of K-metal anode, and the full-cell system also performed with high output energy density and superior cycling stability. Based on fair comparison with other cathode and full-cell candidates for KIBs, the improvements of the current system not only lie on high energy density, but also energy efficiency, high-rate performance and long-term cyclability. We believe the first successful achievement of reversible superoxide/peroxide interconversion cathode reaction in a KIB will stimulate the development of O-related anionic redox reactivity for the fabrication of high-energy-density rechargeable battery devices.
For the RuO 2 @rGO based matrix, graphene oxide (GO) was purchased from Nanjing XFNANO Materials Tech Co., Ltd, and the preparation procedure of rGO-supported ultrafine RuO 2 NPs was similar to the previously reported microwave-hydrothermal synthesis method [34]. The RuO 2 loading in the RuO 2 @rGO nano-composites was estimated to be 75 wt% by thermo-gravimetric analysis. The KO 2 -based cathode powder was prepared by high-energy planetary ball milling (Planetary Mono Mill PUL-VERISETTE 6 classic line, Fritsch). The mass ratio of KO 2 (Sigma Aldrich, >99%) and RuO 2 @rGO was fixed at 6 : 4 (KO 2 : 60 wt%). The mass ratio of grinding media (zirconia ball) to material (KO 2 + RuO 2 @rGO) was fixed ∼8 : 1. The precursors (KO 2 -based cathode composite) were filled into the zirconia ball milling pot and sealed in the Ar-filled glove box. The rotational speed was controlled at 400 rpm for 15 min with a rest for another 3 min. The total ball milling time was around 190-200 hours. Then the as-prepared KO 2 -based cathode composite was harvested from the gas-sealed pot within the glove box.

Cell assembly and electrochemical measurements
The electrodes were assembled into a 2032 coin cell (Hohsen Corp.). The half-cell was assembled by successively stacking a K-metal foil anode (thickness, ∼0.4 mm), the glassy fiber filter (GF/A, Whatman) with 40-45 μL of electrolyte, the Al 2 O 3 coated polypropylene and polyethylene, and the KO 2 -based cathodic plate. For coin cell, the galvanostatic electrochemical measurements were carried out under potential control using the battery tester system HJ1001SD8 (Hokuto Denko) at 25 • C. Typically, the characterizations of the cell were carried out under galvanostatic control at a current density of 300 mA g −1 (based on the load mass of KO 2 : 3-4 mg/cm 2 ) from the open-circuit potential unless otherwise noted. For the in situ Raman/SERS test, the electrochemical experiments were carried out under the control of a potentiostat (Potentiostat/Galvanostat PGSTAT30, Autolab Co. Ltd., Netherlands) at room temperature. The current and potential outputs from the potentiostat were recorded by a multifunction data acquisition module/amplifier (PGSTAT30 Differential Electrometer, Autolab), which was controlled by General Purpose Electrochemical Software. Cyclic voltammetry curves were collected using the HJ1001SD8 (Hokuto Denko) system. Before each electrochemical characterization, the cells were kept on open circuit for 6-8 hours. All of the potentials in this study were referenced to K/K + without further interpretation.

Characterizations
See Supplementary Data section for details.