Anionic Redox and Electrochemical Kinetics of the Na 2 Mn 3 O 7 Cathode Material for Sodium-Ion Batteries

: Manganese-based layered oxides have gained wide attention as cathode materials for sodium-ion batteries due to their cost-e ﬀ ectiveness and nontoxicity. Among them, Na 2 Mn 3 O 7 , which shows promising electrochemical properties as a host material for sodium ions, has been extensively investigated recently. However, the charge compensation mechanisms during battery operation are still ambiguous. Herein, we investigate the electronic structure of Na 2 Mn 3 O 7 using X-ray absorption spectroscopy (XAS) and resonant inelastic X-ray scattering techniques. Mn L II,III -edge XAS spectra show that manganese ions do not undergo any oxidation reaction during the ﬁ rst charge process, suggesting that sodium removal is instead charge compensated by oxygen-ion redox reactions. This, in turn, has an impact on the cycling performances delivered by the material, especially the capacity retention over cycles and also the electrochemical kinetics of sodium ions in Na 2 Mn 3 O 7 .


INTRODUCTION
The intrinsic properties of cathode materials largely determine the lifetime of sodium-ion batteries (SIBs), 1−3 one of the costefficient electrochemical energy storage technologies.Therefore, finding suitable cathode materials with good cycling stability, rapid sodium-ion diffusion, ease of synthesis as well as cost-effectiveness is the key point for developing SIB systems, especially for grid energy storage where the cost of the total system is a comparatively important criterion for consideration. 4To date, a variety of SIB cathode materials have been explored.−9 Second, there are polyanionic compounds, where the structural stability is the main advantage, but the complex synthesis route as well as the need of carbon coating to increase the electronic conductivity make these materials commercially nonviable. 10,11Third, prussian blue analogues offer a rigid open framework, which enables fast sodium transportation, while often suffering from inferior rate capability owing to structural deficiencies and low electronic conductivity. 12,13Moreover, there exist organic electrode compounds with rich structural diversity and flexibility, but they often suffer from inferior electrochemical performance, especially in SIB systems. 14,15Among all of these categories, layered oxide cathode materials with the general formula Na x TMO 2 (TM = transition metals) are considered to be among the most promising electrode materials for future largescale stationary applications.Especially, manganese-based layered oxides, which have gained significant interest due to their advantageous properties including low cost and nontoxicity, constitute a strategy to get rid of other complicated transition metals such as cobalt and nickel, which are more expensive than lithium, and thereby correlate well with the main benefits of sodium-ion batteries. 16,17a x MnO 2 was first investigated by Hagenmuller et al. in 1971, where synthesis and structures of different sodium− manganese−oxygen systems from the tunnel to layered structures were reported. 18Later, many studies have explored the sodium insertion/disinsertion in this system including lowtemperature α-NaMnO 2 , 19 high-temperature β-NaMnO 2 , 20 P2-Na x MnO 2 , 21,22 and tunnel-Na x MnO 2 . 23,24Recently, Na 2 Mn 3 O 7 has attracted great attention as a potential cathode material for SIBs.This compound was first described by Chang  et al. in 1985,  25 where they reported a synthesis method as well as an in-depth investigation of the structure using the Rietveld refinement of X-ray diffraction (XRD) data.In 2001, Weller and co-workers investigated the electrochemical properties of Li 2 Mn 3 O 7 synthesized by ion exchange from Na 2 Mn 3 O 7 in nonaqueous media, which delivered a capacity of 160 mAh g −1 . 26It is noteworthy that the direct synthesis of Li 2 Mn 3 O 7 is complicated due to the formation of several metastable phases during the process.Later, Zhou and co-workers explored the structural and electronic properties of Na 2 Mn 3 O 7 using firstprinciples computational studies and predicted a theoretical capacity of 124 mAh g −1 in the voltage range of 3.1−3.6V. 27 These findings encouraged other researchers to investigate the insertion/disinsertion of sodium in this material.This was first done by Adamczyk and Pralong in 2017, 28 where outstanding electrochemical properties of Na 2 Mn 3 O 7 were suggested to occur based on a two-sodium reaction that is charge compensated by Mn 3+ /Mn 4+ redox reaction.This cathode material delivered a reversible capacity of 160 mAh g −1 in the voltage range of 1.5−3.0V, opening the door for others to more deeply explore the intrinsic properties of this material, including the structural stability during cycling, the electronic structure, 29−31 and the insertion of other alkali metals such as potassium. 32,33n this study, we explore the sodium diffusion kinetics and anionic redox reactions of Na 2 Mn 3 O 7 in the wider voltage range of 1.5−4.5 V using combined soft X-ray absorption spectroscopy (XAS) and resonant inelastic X-ray scattering (RIXS).These techniques provide direct information about the local electronic structure and elucidate the charge compensation mechanisms in Na 2 Mn 3 O 7 during the insertion/disinsertion of sodium.To the best of our knowledge, this is the first exploration of anionic redox reactions in Na 2 Mn 3 O 7 using the RIXS technique.

EXPERIMENTAL METHODS
2.1.Sample Preparation.Na 2.16 Mn 3 O 7 was synthesized using a conventional solid-state reaction of Na 2 CO 3 and MnCO 3 in their respective ratios (all of the precursors were from Aldrich ≥ 99%).A 5 wt % excess of Na 2 CO 3 was added to compensate the sodium volatility during the calcination process.The precursors were mixed with a SPEX 8000 M MIXER/MILL using two 20 mm stainless steel balls for 30 min.The obtained mixture was heated at 600 °C at a rate of 5 °C min −1 for 4 h under air, followed by slow cooling to room temperature.The obtained dark brown powders were stored in an argon-filled glovebox.
2.2.Material Characterization.The chemical composition of the N 2.16 Mn 3 O 7 material was measured using inductively coupled plasma optical emission spectrometry (ICP-OES) (Avio 500 Scott/ Cross-Flow configuration); nitric acid (65%, EMSURE) was used to digest the sample, which was then diluted 100 times with Milli-Q water (ASTM Type I, Fisher Scientific), in 15 mL falcon tubes (VWR).The diluted solution was then filtered by syringe filters with Supor membrane.
The structure of the synthesized material was characterized by XRD using a Bruker twin-twin instrument [Cu Kα radiation, λ = 1.54056Å] by measuring the diffraction angle (2θ) between 10 and 80°in a continuous scanning mode with a step size of 0.01°.XRD data were analyzed by the Rietveld method using FullProf software.The morphology of the pristine material and the size distribution of the particles were measured by scanning electron microscopy (SEM) using a Zeiss Leo 1550 scanning electron microscope.
For electrochemical tests, cathode electrodes were prepared by mixing 60% of the as-prepared Na 2 Mn 3 O 7 powder with 30% of carbon black (Super P) and 10% of carboxymethyl cellulose using deionized water as a solvent.The black slurry formed was then cast onto an aluminum foil, punched into 13 mm diameter electrode disks, and then dried under vacuum overnight at 120 °C to remove residues of solvent and moisture.The loading density of electrodes was around 1 mg cm −2 .Electrochemical tests were performed using a coin-type cell (CR2032), in which sodium metal was used as a counter electrode, glass fiber Whatman (type D) as a separator, and 1 M NaPF 6 dissolved in a mixture of ethylene carbonate (EC) and diethyl carbonate (DEC) with a volume ratio of (1:1) as an electrolyte.The electrodes were cycled within the voltage window of 1.5−4.5 V. Sodium diffusion kinetics during the first cycle were also studied using the galvanostatic intermittent titration technique (GITT; charge and discharge at a rate of C/20 for 1 h followed by rest for 10 h) and cyclic voltammetry (CV) with various sweep rates from 0.01 to 0.3 mV s −1 .
XAS and RIXS measurements were performed using the beamline BL27SU of SPring-8, Japan.XAS spectra were collected for the Mn L II,III -edge and the O K-edge by collecting energy-resolved fluorescence X-rays (500−1000 eV) and by measuring the sample drain current simultaneously.Bulk representative XAS spectra for the Mn L II,III -edge and O K-edges were obtained using the fluorescence yield (FY), the partial FY (PFY) mode was used for O K-XAS, and the inverse partial FY (IPFY) mode (O K_PFY was monitored) was used for Mn L_XAS.The surface-related drain current produced XAS spectra equivalent to the total electron yield (TEY) mode.The FY was collected using an energy-resolving, liquid-nitrogen-cooled, solidstate detector.The TEY mode XAS spectra and the incident intensity were measured by recording the sample drain current and focusing the mirror current (for normalization) using pico-ammeters.FY XAS spectra are bulk sensitive with an average information depth of about 100 nm, while TEY mode XAS spectra are more surface sensitive with an average information depth of about 5 nm.The spectral resolutions were set to ∼0.1 eV (XAS) and ∼0.2 eV (RIXS), respectively.Quantitative fitting of Mn L II,III -edge XAS at TFY and TEY mode spectra was done using an optimization algorithm to minimize the absolute difference between the spectra and the sum of the reference spectra (Mn 2+ , Mn 3+ , Mn 4+ ).

RESULTS
3.1.Structure, Morphology, and Electrochemical Properties of Na 2 Mn 3 O 7 .The elemental composition of the sample is measured by ICP.The atomic ratio of Na/Mn is 0.723 leading to the formula Na 2.16 Mn 3 O 7 , which is consistent with the expected stoichiometry.The crystal structure of the as-prepared Na 2 Mn 3 O 7 material was determined using X-ray diffraction (Figure 1a).The Rietveld refinement of the corresponding pattern confirms the formation of the Na 2 Mn 3 O 7 phase within the triclinic P1̅ space group.The obtained unit cell parameters from profile matching are: a = 6.61002(1)Å; b = 6.84928(5)Å; c = 7.43913(7) Å, α = 106.29 (7); β = 108.36(9); γ = 111.39(1); and V = 266.184(6)Å 3 , in good agreement with previously reported results. 25owever, the relatively high Rietveld factors (Bragg R-factor = 22.6% and R f factor = 17.4%) reveal the existence of some deficiencies in the profile fit.The broad and asymmetric peaks arise from structural defects associated with local manganese vacancies in transition-metal stacking layers.These vacancies induce the formation of unhybridized oxygen 2p bands, which affect the alignment of Mn 3 O 7 2− layers throughout the crystal structure. 26,29he Na 2 Mn 3 O 7 crystal structure is built up of Mn 3 O 7 2− layers with manganese coordinated octahedrally by oxygen atoms and separated by sodium atoms located at two different sites: half of the sodium atoms (Na1) occupy the prismatic sites above and below the vacant manganese sites, and the other half (Na2) occupy the octahedral sites.This makes Na 2 Mn 3 O 7 different from conventional layered oxides as two different environments exist for the involved sodium atoms (Figure 1b).SEM images reveal the formation of agglomerated particles with an inhomogeneous particle size distribution and the presence of thin plates of a few microns.Elemental mapping and energy-dispersive X-ray spectroscopy (EDS) denote the uniform distribution of sodium, manganese, and oxygen in Energy & Fuels pristine material particles, which largely confirm the elemental composition of the material (Figure 1c).
The electrochemical performances of Na 2 Mn 3 O 7 were investigated at room temperature (25 °C) using sodium halfcells at C/20 (1 Na + in 10h) in the voltage range of 1.5−4.5 V as displayed in Figure 2a.
The open-circuit voltage (OCV) of the half-cell was about 2.8 V.The Na 2 Mn 3 O 7 electrodes deliver a first charge capacity of 151 mAh g −1 , corresponding to the removal of 1.87 of sodium ions.During the subsequent discharge, the capacity exceeds 195 mAh g −1 corresponding to the re-insertion of 2.37 of sodium ions, leading to the formula Na 2.5 Mn 3 O 7 at a discharged state (1.5 V) and a first Coulombic efficiency exceeding 100%.This increase results from the reduction of manganese ions below 2.5 V, as will be discussed later.The potential vs capacity curves exhibit several slopes resulting from the ordering between Na ions and Na vacancies and/or new reactions involving oxygen redox.These slopes partially disappear with increasing cycle number, indicating that the associated reaction is partially reversible before it disappears completely from the system.
By comparing the charge and discharge curves of Na 2 Mn 3 O 7 electrodes at the 1st, 5th, 10th, 20th, 30th, and 38th cycles (Figure 2a), the profiles clearly show that the capacity drops continuously with increasing cycle number, exhibiting a capacity retention of 85.3% after 38 cycles (Figure S1), suggesting the occurrence of an irreversible reaction, which can be the reason behind the capacity fade.Besides these observations, we can clearly see that the profile of the first charge has a different feature compared to the other cycles.Therefore, in the following parts, the charge compensation mechanism in Na 2 Mn 3 O 7 during the desodiation−sodiation process will be elucidated for the first and second cycles by performing soft XAS as well as advanced RIXS at different states of charge as indicated in Figure 2b.
3.2.Manganese Evolution during the Electrochemical Process.The electrochemical behavior as well as the sodium insertion/extraction processes are highly associated with intrinsic redox reaction mechanisms of the active material.
To get an insight into manganese oxidation states at the surface and the bulk of the electrodes, soft XAS was performed at different states of charge using two different modes: TEY and FY.TEY measurement probes the surface of the electrodes around 5 nm, but it is unsuitable for investigating the redox reaction in the cathode material, whereas TFY can probe the bulk structure as deep as 100 nm, but it usually suffers from a large distortion due to extreme self-absorption. 34It is important to probe the surface and bulk separately to achieve depth profiling of the electronic structure of the investigated material.
XAS provides measurements with high sensitivity that renders straightforward information of unoccupied orbitals related to oxidation states, especially L 2−3 -edge XAS, which is more sensitive to the 3d electronic states of transition metals than K-edge XAS.L-edge XAS represents transition of core electrons from TM 2p levels into 3d levels.The L 3,2 edge (2p 6 3d n → 2p 5 3d n+1 ) transitions generate 2p 5 core configurations that due to spin−orbit coupling splits into two different states.This splitting results in two different excitation edges, one at the lower energy 2p 3/2 referred to as the L 3 edge and one at the higher energy 2p 1/2 known as the L 2 edge. 35igure 3a shows the Mn L II,III -edge XAS spectra recorded during the first cycle and the second subsequent charge in the IPFY mode in the voltage range of 1.5−4.5 V.For comparison, reference spectra of MnO (Mn 2+ ), Mn 2 O 3 (Mn 3+ ), and MnO 2 (Mn 4+ ) collected from the same beamtime in the IPFY are also included.In the bulk of the electrodes, the pristine spectrum has a main peak around 643.5 eV and a shoulder around 641 eV, which coincide well with those of Mn 4+ , thereby revealing that the manganese in Na 2 Mn 3 O 7 is in its tetravalent state as expected from the stoichiometry of the material (Figure 3a).During the whole battery charge process (desodiation), the spectra show no obvious change, meaning that the manganese ions remain at Mn 4+ and do not participate in the charge compensation process.During the subsequent sodiation, the spectrum at the fully discharged state (Dis 1.5 V) shows a different shape compared to the spectra recorded during the desodiation process.Indeed, an increase of the peak intensity at 641.5 eV characteristic of Mn 3+ can be observed, indicating that Mn 4+ undergoes a reduction to Mn 3+ upon discharge and then reoxidizes during the following charge up to 4.5 V (Ch 4.5 V 2nd spectrum).Since mixed oxidation states occur, a linear combination fitting of the Mn L II,III -edge XAS spectra was employed especially for Dis 4.17 V and Dis 1.5 V spectra (Figure 3b) to obtain Mn valence percentages as a function of state of charge.−40 As shown in Figure 3b, the fitted spectra (dashed lines) are in good agreement with TFY spectra (solid lines) revealing that 82% of Mn 4+ ions had been reduced to Mn 3+ upon discharge to 1.5 V (Figure 3c).
Figure 3d represents Mn L II−III XAS spectra collected at the surface of the electrode using the TEY mode.Mn 2+ , Mn 3+ , and Mn 4+ spectra at the TEY mode are also shown as references.In the surface layer, the Mn valence shows a different trend compared to the bulk: Mn 2+ (640 eV) is more involved in the surface of the electrode compared to Mn 3+ and Mn 4+ , revealing that the Mn oxidation states in the surface and in the bulk are different.A similar cation oxidation gradient between the bulk and the surface has been reported in another study. 41sing the same fitting method, a quantitative fitting of TEY spectra is plotted in Figure 3e.The Mn oxidation states are plotted in Figure 3f as a function of state of charge.In the pristine material, an amount of 19% of Mn 2+ in addition to 28% of Mn 3+ and 53% of Mn 4+ is found.However, upon charging, the percentages of Mn 2+ and Mn 3+ are increasing at some points instead of getting reduced.This is frequently observed in cathode materials where oxygen redox reactions are involved, 36 suggesting that the surface reactions in these cathode materials probably are related to oxidized oxygen ions during the charge process.At the end of the discharge process, the percentages of Mn 3+ and Mn 4+ are abruptly reduced to 0%, resulting in a corresponding rapid increase of Mn 2+ at the discharged state.
3.3.Oxygen Evolution during the Electrochemical Process.As shown by the Mn L II,III -edge XAS spectra, the manganese ions do not participate in the charge compensation mechanisms in Na 2 Mn 3 O 7 electrode material during the entire charge process.To elucidate the oxygen state evolution upon cycling, the O-K edge XAS spectrum was measured at different state of charge, as shown in Figure S2a,b.The pre-edge peaks below 535 eV represent the local density of empty oxygen 2p states that hybridize mainly with transition-metal 3d-orbitals, which here is manganese.More precisely, the large peak around 530 eV is related to the t 2g state and the small one around 532.5 eV is related to the e g state. 42,43pon desodiation, the O K-edge XAS pre-edge peaks show an intensity increase and a broadening, for both the bulk-sensitive PFY mode and the surface-sensitive TEY mode.This increase in intensity reveals the creation of hole states in the TM 3d−O 2p level, which was observed for other battery electrodes without oxygen redox and has been related to the oxidation of transition metals during cycling. 40However, since manganese remains at its tetravalent state during the entire charge process in the bulk of the material, this increase of hole state density is probably related to the removal of electrons from the oxygen 2p orbitals.During discharge, the peaks tend to narrow because of the reduction of Mn 4+ ions to Mn 3+ .This shrinking is more visual for the surface of the electrode (TEY spectra), indicating that manganese is in a lower valence state (Mn 2+ ) in the surface layer, consistent with the Mn L II,III -edge XAS results discussed above.
To achieve more specific oxygen redox signals, the pre-edge peaks (528−533 eV) are further resolved using the RIXS analysis for Na 2 Mn 3 O 7 electrodes at the same states of charge as mentioned above during the first cycle and the second charge.Figure 4a displays RIXS maps providing intensity as a function of incident energy (y axis) and emission energy (x axis).The intense features around 525 eV emission energy (red circles) are characteristics of the O 2− state in conventional Energy & Fuels transition-metal oxides in battery electrodes.Upon charging, these broad emission features get intense and broadened in line shape, which is in good agreement with the changes observed in FY-spectra of the O K-edge XAS, but as stated above, this change is not relevant for oxygen redox reactions.However, the feature around 523.7 eV emission energy and 530 eV incidence energy, which represents the fingerprint of oxidized oxygen reported in many research papers of Li-rich cathode materials, 44 can be seen at the beginning of the charge process (Ch 3.5 V; black rectangle) and becomes stronger with further charging up to 4.5 V confirmed also by RIXS spectra of O K-edge at 530 and 530.5 eV excitation energies (Figure S3a,b).During discharge, this feature gets weaker until it disappears completely at Dis 1.5 V.Then, it appears again during the second charge to 4.5 V but not as intense as the one observed during the first charge to 4.5 V (Figure 4b), revealing that the oxygen redox reaction leading to the creation of oxidized oxygen ions is not totally reversible, which leads to the capacity loss observed in the electrochemical profile.
3.4.Impact of Anionic Redox on the Electrochemical Kinetics in Na 2 Mn 3 O 7 .To evaluate the electrochemical kinetics of Na + in Na 2 Mn 3 O 7 within the potential window of 1.5−4.5 V, galvanostatic intermittent titration technique (GITT) and cyclic voltammetry (CV) at varied scan rates were performed to measure diffusivity of ions in electrodes, in particular during the first cycle.For GITT measurements, the electrodes were cycled at 0.1C in the voltage range of 1.5−4.5 V for 1 h followed by 10 h relaxation, as shown in Figure 5a.A demonstration of a single titration at 3.5 V during the GITT measurement is shown as an inset in Figure S4.
The relationship between the voltage and the duration of the current pulse τ 1/2 during titration has a straight-line behavior (Figure S4).Therefore, the diffusion coefficient of Na + in Na 2 Mn 3 O 7 electrodes can be calculated by the simplified equation based on Fick's second law of diffusion as 45,46 i k j j j j j y { z z z z z i k j j j j j y here, m a (1.08 mg) and M a (322.78g mol −1 ) are the mass and molecular weight of the active material, respectively; A (1.266 cm 2 ) is the total contact area between the electrode and the electrolyte; V M (82 cm 3 mol −1 ) is the molar volume of the electrode; τ (3600 s) is the duration of the current pulse; L is the average thickness of the electrode; ΔE τ is the voltage change during a single titration; and ΔE s is the steady-state voltage change. 47,48igure 5b,c shows the variation of the diffusion coefficient of Na + in Na 2 Mn 3 O 7 electrodes during the first charge and discharge as a function of the cell voltage.The values of D Na + during the first charge vary from 2.26 × 10 −13 to 1.14 × 10 −11 cm 2 s −1 The coefficient D Na + exhibits the same variation during the first discharge.
Moreover, the sample exhibits large polarization up to 0.46 V during charge, especially between 3.5 and 4.5 V, which is not observed in the reported work about the insertion of Li in Na 2 Mn 3 O 7 . 32−51 In the beginning of the discharge process, the polarization is lower than that observed during charge, which indicates that this possible anionic redox reaction is not totally reversible.Moreover, this polarization is lower around 2.5 V, where the electrochemistry is dominated by the cationic manganese reactions, and then increases to 0.5 V at the end of the discharge process.This can be correlated to the increasing Mn 3+ concentration in the sample, aggravated by the Jahn Teller effects, which are often seen in Mn-based electrodes. 52o investigate the impact of anionic redox reaction on the kinetics of the cationic redox process, Figure 5d illustrates a comparison of the overpotential between the initial and final states of the rest at different potentials.By comparing rest 1 (3.2V) with rest 2 (4.25 V), which represent the beginning and end of the anionic redox process, respectively, it can be noticed that the potential drop is more notable in rest 2 around 450 mV, while for rest 1, the potential reaches an equilibrium after a drop of just 100 mV.Moreover, rest 3 at 2.2 V during discharge, which is dominated by cationic redox, has a potential drop of about 180 mV, which is higher than the reported values for cationic redox reactions in the literature. 53hese findings show that the anionic redox reaction is the main reason behind the sluggish kinetics and that it has a negative effect on the kinetics of cationic manganese redox in the subsequent discharge process.The oxygen redox reactions are usually associated with the migration of transition-metal ions from unstable oxygen octahedron to Na layer, which is critical for sodium diffusion. 54,55V tests at varied scan rates were also performed to confirm the kinetics of Na + extraction and insertion in Na 2 Mn 3 O 7 .CV Energy & Fuels profiles, recorded during five cycles in the range of 4.5−1.5 V at scan rates of 0.01, 0.1, 0.15, 0.2, and 0.3 mV s −1 , are depicted in Figure 6a.The major redox reaction occurred at 2.3 and 2.6 V vs Na + /Na at 0.01 mV s −1 , which corresponds to the Mn 4+ / Mn 3+ redox reaction.The other small peaks in the CV might be assigned to Na ion and Na vacancy ordering or to oxygen redox reactions.
The evolution of the peak shape and current strength along with the sweep rate is correlated to the kinetics of Na + in Na 2 Mn 3 O 7 .The cathodic O1 and anodic R1 current peaks (Figure 6a) show a good linear dependence on the square root of the scan rates (Figure 6b).Therefore, the diffusion coefficient of Na + can be calculated using the Randles−Sevcik equation 56,57 where I p is the peak current, n is the number of electrons, A is the total contact area between the electrode and electrolyte, D Na + is the diffusion coefficient of Na ions, C is the concentration of Na + in the lattice, and ν is the scan rate.The calculated D Na + values at different scan rates are presented in Table S1.
The average D Na + values are 9.0 × 10 −13 cm 2 s −1 during the charge process and 2.8 × 10 −13 cm 2 s −1 during the following discharge process, which are more or less in the same order as the average D Na + values obtained from GITT tests.

DISCUSSION
The Mn L II,III -edge XAS results for the bulk of Na 2 Mn 3 O 7 (Figure 3a) revealed that manganese remains at its tetravalent state (Mn 4+ ) during the first charge process.Moreover, the removal of sodium ions from Na 2 Mn 3 O 7 during the first charge is charge compensated by an oxygen redox reaction, not only in the high-voltage region but also from the very beginning of the charge process.This is shown both by O K-edge XAS (Figure S1) and O K-edge RIXS results (Figure 4).This oxygen redox reaction in turn results in high polarization and sluggish kinetics of sodium diffusion, as seen in the GITT test.Moreover, it has a negative impact on the cationic redox reactions during the subsequent discharge, which might be correlated to the migration of Mn ions into the Na layers.It is also likely the reason behind the strong reactivity in the surface layer detected by Mn L II,III -edge XAS at the TEY mode, where the concentrations of Mn 2+ and Mn 3+ in the surface layer are increasing at some points during the charge process.This is uncommon in cathode materials without anionic redox.The slopes seen in the electrochemical profile (Figure 2a), as well as the unknown peaks observed in the CV profile (Figure 6a), are probably also associated with the oxygen redox.With further cycling, these slopes partially disappear, suggesting that the oxygen redox reaction is not fully reversible (Figure 2), which was also confirmed when comparing the RIXS maps at the first and the second charge.Here, the feature related to oxidized oxygen ions decreases in intensity in the second charge up to 4.5 V, and it can be speculated that it will disappear completely after extended cycling beyond what is investigated here.It is likely that this partial irreversibility of the oxygen redox is the cause of the relatively low capacity retention after cycling (compared to other layered oxides), and thus a problematic feature for Na 2 Mn 3 O 7 cathodes.

CONCLUSIONS
In summary, redox reaction processes in Na 2 Mn 3 O 7 have been deeply investigated using XAS and RIXS measurements, correlated to electrochemical analysis.This shows that the sodium disinsertion−insertion process is heavily charge compensated by oxygen-ion redox reactions during the first charge (desodiation), while manganese redox reactions appear during the subsequent discharge process (sodiation).The results show that the oxygen-ion redox reaction is not fully reversible, which could be the cause of the relatively low capacity retention.

Figure 1 .
Figure 1.(a) Observed and calculated XRD patterns of Na 2 Mn 3 O 7 using the Rietveld refinement.(b) Schematic illustration of the Na 2 Mn 3 O 7 crystal structure perpendicular to the c axis.(c) SEM images of the as-prepared Na 2 Mn 3 O 7 and elemental mapping images; presence of manganese indicated in green, sodium in purple, and oxygen in orange.

Figure 2 .
Figure 2. (a) Electrochemical profile of Na 2 Mn 3 O 7 for different cycles at C/20 in the voltage range of 1.5−4.5 V. (b) First charge, discharge, and the second charge of Na 2 Mn 3 O 7 electrodes at C/20.The points used for ex situ XAS and RIXS analyses are labeled with arrows.

Figure 3 .
Figure 3. (a, b) Mn L II,III edge XAS spectra of Na 2 Mn 3 O 7 during cycling at different states of charge.The reference spectra Mn 2+ , Mn 3+ , and Mn 4+ are plotted at the bottom of each figure.(c, d) Fitted spectra of the pre-edge peaks.(e, f) Mn valence percentages at FY and TEY modes.

Figure 4 .
Figure 4. (a) O K-edge RIXs maps at different states of charge during initial cycle and the second charge.Features corresponding to the O 2− state are indicated by a red circle, and features associated with O-redox reaction are marked by black rectangle and arrows.(b) Comparison of O K-edge RIXs maps of the fully charged state in the first and second cycles.

Figure 5 .
Figure 5. (a) GITT voltage profile of Na 2 Mn 3 O 7 electrodes cycled at C/20 in the range of 1.5−4.5 V. (b, c) Sodium diffusion versus cell voltage during charge and discharge, respectively.(d) Overpotential between the initial and final states of OCV rest at different states of charge (rest 1, rest 2, and rest 3) illustrated in (a).

Figure 6 .
Figure 6.(a) CV profiles at various scan rates.(b) Linear relation between the current peaks and the square root of the scan rate.