Origin of Aging of a P2-Na x Mn 3/4 Ni 1/4 O 2 Cathode Active Material for Sodium-Ion Batteries

: Sodium-ion batteries (SIB) are currently being developed and commercialized as a promising new technology for cost-effective and powerful electrical energy storage. In this study, we investigate the origin of capacity fading in P2-type layered sodium cathode materials for SIBs using a micron-sized single-crystalline P2-Na x Mn 3/4 Ni 1/4 O 2 model cathode active material. Using various electrochemical techniques, we identify the following aging effects upon cycling: (i) a state of charge (SOC)- independent increase in polarization, (ii) a SOC-dependent increase in polarization at high voltage, and (iii) a loss of active material due to electronic disconnection after prolonged cycling. With high-resolution transmission electron microscopy (HRTEM) and energy-dispersive X-ray (EDX) spectroscopy, we identify surface densification, resulting in 5 − 10 nm thick surface layers on cycled cathode active materials as the origin for SOC-independent increase of polarization. The corresponding oxygen loss is in accordance with gas evolution in differential electrochemical mass spectrometry (DEMS) measurements. Furthermore, with scanning electron microscopy (SEM) electrode cross sections, we identify (partly) reversible cracking at a high SOC as the cause for increased polarization depending on SOC. Operando X-ray diffraction (XRD) identifies significant anisotropic volume change, which suggests mechanical stress as the cause for cracking at a high SOC and loss of active material after prolonged cycling. We believe that the herein provided understanding on the aging of this highly


■ INTRODUCTION
Electrical energy generation from renewables and storage thereof lies at the heart of the worldwide transition toward climate neutrality.−6 Sodium-ion batteries (SIB) utilize the same working principle as LIBs 7 but are designed to operate without (or with minimal) critical raw materials. 8,9Therefore, SIBs are currently developed and industrialized as a new, sustainable technology to complement current LIBs. 8,10,11In a typical rocking-chair type SIB, sodium ions replace lithium ions as ionic charge carriers.−14 On the cathode side, sodium-containing intercalation compounds replace lithium cobalt manganese nickelates (NCM) or lithium iron phosphate (LFP). 8,15−17 Furthermore, SIBs may offer desirable advantages against LIBs in terms of lowtemperature performance, 10 0 V storage and transport, 18 as well as possibly enhanced battery safety. 19,20−24 Additionally, sodium-ion batteries can be produced based on the same production processes and manufacturing lines as current lithium-ion batteries (drop-in technology), which minimizes necessary investments and potentially accelerates industrializa-tion, making SIBs a promising complementary technology to LIBs. 15 It must be noted that current sodium-ion batteries still come with a slight cost disadvantage based on the amount of stored energy ($/kWh). 17,21,25However, limited availability and projected price increase for lithium and other critical raw materials or the development of materials with higher energy density for SIBs might reverse this trend in the near future. 6,21he cathode active material remains one of the main bottlenecks toward cost-effective, powerful, and sustainable SIBs.−28 These compounds offer a wide structural variety and can be synthesized as O3, P3, and P2 types, following Delmas' notation for alkali metal layered oxides. 29,30Among these, P2-type materials are reported to be beneficial in terms of structural stability during electrochemical (de)sodiation and fast solid-state diffusion (e.g., rate capability). 312-type sodium manganese nickel layered oxides offer a good mix of high energy density, power capability, and sustainability.−34 During electrochemical (de)sodiation at room temperature, P2-Na x Ni 1/3 Mn 2/3 O 2 maintains the P2 structure over a wide range of stoichiometry (x ≥ 1/3), following solid solution-like behavior. 33At sodium stoichiometries x < 1/3, a two-phase r e a c t i o n w i t h t h e r e s p e c t i v e e n d m e m b e r s P 2 -Na 1/3 Ni 1/3 Mn 2/3 O 2 and O2-Na 0 Ni 1/3 Mn 2/3 O 2 takes place. 33he potential profile of this material typically shows distinct voltage jumps at sodium stoichiometries of x = 2/3, 1/2, and 1/3, which are associated with Na + /vacancy orderings. 33,35In sodium cells, this material is capable of delivering specific discharge capacities of approximately 160 mAh g −1 at an average potential of approximately 3.6 V when cycled between 2.0−4.5 V vs Na/Na + . 33,36,37When the potential window is expanded to 1.5−4.5 V, discharge capacities of ≥ 220 mAh g −1 at a mean potential of approximately 3.1 V are reached. 38,39−42 At high potentials (≥4 V), charge is either compensated by oxygen redox, following a reductive coupling mechanism 38,43,44 or by Ni 3+/4+ cationic redox. 33,37At medium potentials (approximately 2.5−4 V), charge is compensated by Ni 2+/3+ redox couple. 38At low potentials (approximately <2.5 V), Mn 3+/4+ redox couple is active. 38o far, the main drawback of these materials is their low cycling stability.Typically, for P2-type Na x Ni y Mn 1−y O 2 , the capacity fades quickly during the first ten to twenty cycles when the upper cutoff exceeds 4 V vs Na/Na + . 45In the literature, the pronounced capacity fade has been assigned to exfoliation, 46,47 side reactions with the electrolyte, 38 partial irreversibility of the P2/O2 phase transition, 36 surface densification, 44 or oxygen redox. 43Yet, despite considerable efforts, the underlying aging mechanism for this highly promising material class remains elusive.
In this work, we have synthesized micron-sized single-crystal P2-Na x Mn 3/4 Ni 1/4 O 2 as a model material to investigate the origin of the capacity fade of this promising class of cathode active materials for sodium-ion batteries.The chemical, physical, and electrochemical properties of the obtained material are thoroughly characterized.By combining several characterization techniques, we provide evidence for two aging mechanisms, namely, surface densification and intracrystallite cracking.Based on these results, we postulate and discuss an aging mechanism for this material class.

■ EXPERIMENTAL SECTION
Material Synthesis.We have synthesized micron-sized singlecrystal P2-Na x Mn 3/4 Ni 1/4 O 2 in a multistep process starting from a spherical Mn 3/4 Ni 1/4 (OH) 2 precursor, which was obtained by coprecipitation in a continuously stirred tank reactor as described elsewhere. 42These dense spherical particles were disintegrated by ball milling.The resulting powder was mixed with Na 2 CO 3 (Sigma-Aldrich) and NaCl (Sigma-Aldrich) in molar ratios of Na 2 CO 3 /M = 0.44 and NaCl/M = 0.55, respectively, and then calcined for 24 h in air at 1000 °C in a batch furnace (Nabertherm).The obtained powder was washed with deionized water (5 mS cm −1 ), filtered, dried, and subsequently mixed with Na 2 CO 3 and Na 2 SO 4 (Alfa Aesar) in molar rations of Na 2 CO 3 /M ≈ 0.10 and Na 2 SO 4 /M ≈ 0.55, respectively.The obtained mixture was calcined in the same batch furnace for 10 h in air at 1000 °C, subsequently washed in deionized water, filtered, dried, and ground with a mortar and pestle.Final calcination in CO 2 -free synthetic air (20 vol % O 2 in Ar) for 5 h at 900 °C was performed in a box furnace (Carbolite Gero).After calcination, the obtained powder was transferred in a Buchi glass oven with minimal time at ambient conditions, dried at 200 °C under dynamic vacuum overnight, and subsequently stored in an Ar-filled glovebox (MBraun, O 2 < 0.1 ppm, H 2 O < 0.1 ppm).
Material Characterization.The chemical composition was analyzed by ICP-OES (Spectro Arcos SOP) using a diluted aqua regia solution.ATR-FTIR was performed inside an Ar-filled glovebox using an Alpha spectrometer (Bruker) with an Alpha-P (Bruker) ATR unit.Powder X-ray diffraction (XRD) was performed on a D8 Advance (Bruker) in Bragg-Brentano geometry using a Cu X-ray tube and a LynxEyeXE-T detector.Rietveld refinement based on the resulting diffraction pattern was performed using TOPAS V6 software.Throughout this work, all depictions of crystal structure were generated using VESTA software. 48The morphology of the asprepared powder was characterized by scanning electron microscopy (SEM) with a Leo 1530 VP (Zeiss) at 5 kV acceleration voltage using an Everhart−Thornley SE detector.
Electrodes, Cells, and Electrochemistry.Slurry preparation, electrode coating, punching, and cell assembly were performed inside an Ar-filled glovebox avoiding any contact with the environment to minimize any effect due to the formation of surface impurities. 49By dispersing the active material, conductive carbon (SuperP-Li, Imerys Graphite & Carbon), and polyvinylidene difluoride binder (Solef P5130, Solvay) with the respective weight ratio of 84:8:8 in an appropriate amount of N-methyl-2-pyrrolidone solvent (NMP, anhydrous, Sigma-Aldrich), a homogeneous slurry was obtained.This slurry was cast on an aluminum current collector foil using the doctor blade technique.After the coating was dried, electrodes with 12 mm diameter were punched and dried at 130 °C under dynamic vacuum (2 × 10 −2 hPa) in a Buchi glass oven overnight.Aluminaprotected CR2032 coin cells (Hohsen) were built using these electrodes as the working electrode, two layers of glass fiber separator (Ø 16 mm, GFA, Whatman), 150 μL of 1 M NaPF 6 in a propylene carbonate (PC) electrolyte, and sodium metal foil serving as the counter electrode (Ø 16 mm, Acros Organics).Cycling with various cutoff voltages (1.5−4.3,2.0−4.3,1.5−3.8,2.0−3.8V) was performed on a multichannel galvanostatic cell test system (CTS, Basytec).In any case, cycling procedures commenced with discharge.GITT measurements were performed on a VMP3 potentiostat (Biologic) within the voltage window of 4.3−1.5 V using specific current pulses of 17.3 mA g −1 (C/10) for 10 min.Equilibrium was assumed when the voltage change was <0.0001 V h −1 .For impedance study, PAT cells (EL-Cell) are constructed with the same working electrodes, glass fiber separators with a sodium ring reference electrode (PAT core, EL-Cell), 150 μL of 1 M NaPF 6 in PC + 5% FEC electrolyte, and sodium metal foil counter electrodes (Ø 18 mm).A VMP2 potentiostat (Biologic) was used.The temperature was controlled using a climate chamber (CTS GmbH) at 25 °C.Prior to every potentiostatic electrochemical impedance spectroscopy (PEIS) measurement, the cell was allowed to rest for 1 h.PEIS measurements were performed in the frequency range 200 kHz to 5 mHz with 10 mV of amplitude.
Operando Characterization.Operando X-ray synchrotron diffraction measurements were performed at PETRA III, beamline P02.1 (DESY, Hamburg, Germany). 50An eight-fold coin cell holder, connected to a Biologic Instruments potentiostat, was applied as described elsewhere. 51A wavelength of 0.20734(1) Å was determined by analysis of positions of reflections of the LaB 6 reference material (NIST 660C).For the characterization of the pristine material, the first diffraction pattern was recorded before starting the electrochemical measurement.Subsequently, the cell was first discharged and then charged at a constant current.All diffraction patterns were analyzed by the Pawley method with TOPAS V6 software.Reflections of the Al current collector served as an internal standard during the measurements, thus providing independent control of the reliability of the obtained model parameters.A differential electrochemical mass spectrometry (DEMS) cell was assembled under an argon atmosphere.Before assembly, all cell parts were dried at 60 °C for at least 12 h.As separators, Whatman GF/A glass microfiber filters were used after drying at 110 °C under vacuum overnight.1.2 mL of 1 M NaPF 6 in PC was used as the electrolyte.The same electrode sheet as described above served as the working electrode.The DEMS cell was constantly flushed with a stream of helium to transfer evolving gases to the mass spectrometer (Pfeiffer Vacuum OMNISTAR GSD320).Before the inlet to the mass spectrometer, the gas stream is passed through a cold trap (−30 °C) to reduce the background from the evaporated electrolyte solvent.A resting time of 24 h was applied before cycling to equilibrate the background signals.The cycling was performed at a constant current of 17.3 mA g −1 between 1.5−4.3V vs a sodium-metal quasi-reference electrode.The full mass spectrum (m/ z = 1−100) was measured every 100 s.The gases were quantified from the full spectrum against a calibration matrix, which was measured from calibration gas mixtures of single gases in helium.A detailed description of the setup and the data evaluation will be given in an upcoming publication.
Ex Situ Characterization.For ex situ TEM and XRD investigations, electrodes were cycled in CR2032 coin cells as described above for 3 or 10 cycles, galvanostatically discharged to 2.68 V (equals OCV of pristine cells), followed by a constant voltage step for 5 h and then carefully opened inside an Ar-filled glovebox.The electrode composite was scratched off the aluminum current collector and carefully ground with the mortar and pestle.By discharging to 2.68 V, the sodium stoichiometry of the as-synthesized material was approximately restored, leading to a stable and relatively moisture-insensitive sample.Lamellae for TEM investigations were prepared with a Zeiss NVision 40 Ar (Zeiss Microscope) Ga-ion focused ion beam (FIB) device.The cross-sectional lamellae were prepared along the [112̅ 0] crystal axis to allow imaging of the crystal stacking in the subsequent TEM investigations.A few micrometer carbon layer was deposited by the gas injection system (GIS) to protect the sample surface.The roughly cut lamella was transferred via a micromanipulator to a standard copper lift-out grid and thinned with decreasing beam currents.End thinning was performed in the low voltage 5 kV mode.Bright-field imaging, selected area electron diffraction (SAED), and aberration-corrected high-resolution transmission electron microscopy (HRTEM) imaging were carried out using an image side C S -corrected FEI Titan 80−300 operated at 300 kV.High-angle annular dark-field (HAADF) scanning TEM images and drift-corrected energy-dispersive X-ray (EDX) elemental maps were acquired using a Thermo Fisher Talos 200× microscope operated in STEM mode at 200 kV.The intensity of the elemental maps shows the counts according to the background corrected and fitted model.Fast Fourier transformations (FFTs) were performed with Digital Micrograph software (Gatan).The bulk crystal structure after cycling was characterized ex situ using capillary transmission XRD on a D8 Advance (Bruker) diffractometer equipped with a Cu X-ray source and a LynxEyeXE-T detector.For this purpose, the scratched-off electrode composite was filled in 0.3 mm quartz glass capillaries and sealed inside the glovebox.The bulk morphology of the cycled cathode active material was investigated with air-tight cross sections of the cycled electrodes using SEM.Cross sections were polished using Ar-ion beam milling in an IM4000 (Hitachi).SEM investigations were performed on a Leo 1530 VP (Zeiss) with 4 kV acceleration voltage and an Everhart−Thornley SE detector.Any contact with the environment is avoided by the use of air-tight transfer chambers (Kammrath & Weiss).An overview of the electrochemical history of samples for ex situ investigations is presented in Supporting Figure S1.

■ RESULTS AND DISCUSSION
Elemental analysis of the as-prepared cathode powder was performed by ICP-OES.From the obtained results, the chemical formula Na 0.585 Mn 0.734 Ni 0.266 O 2 is calculated, which is in good accordance with our targeted stoichiometry of Na 0.58 Mn 3/4 Ni 1/4 O 2 .The considerably low sodium content was targeted to minimize surface impurities, such as Na 2 CO 3 , NaHCO 3 , or NaOH, 49,52 which will later allow us to correlate transferred charge in electrochemical measurements with the sodium content of the layered oxide.The absence of surface impurities is confirmed by ATR-FTIR measurements as presented in Supporting Figure S2.The bulk crystal structure is analyzed using XRD as presented in Figure 1A and Supporting Figure S3.The diffraction pattern exhibits narrow  S1.We evaluated the volume-weighted strain and anisotropic crystallite size under the assumption of hexagonal morphology 53−55 by the double Voigt approach.We found a mean anisotropic crystallite size with a diameter of 876 nm in the a−b plane and 170 nm in the c-direction.The strain is as low as e 0 = 0.000261.The powder morphology was investigated with a secondary-electron SEM as presented in Figure 1B.The powder consists of micron-sized particles with a hexagonal platelet-like morphology.The morphology of the particles clearly reflects the hexagonal nature of the layered crystal structure.The crystal size observed in the SEM is in good accordance with the anisotropic crystallite size in the a−b plane as derived from XRD.In the c-direction, on the other hand, crystals appear thicker in SEM images compared to the values obtained from refinement of the XRD pattern, which hints to the fact that a few crystals might be stacked in the cdirection.In fact, we have found grain boundaries within the a−b plane using TEM (compare Supporting Figure S4).during electrochemical (de)sodiation at room temperature (Figure 2 and Supporting Figure S5).During the initial sodiation (discharge), the c-lattice parameter decreases and the a-lattice parameter increases.At a high degree of sodiation, a two-phase reaction into an orthorhombic distorted P′2 phase takes place.The distortion most likely results from the presence of Jahn−Teller active Mn 3+ . 40In the subsequent desodiation (charge), the system reversibly recedes to the initial P2 structure.The c-lattice and a-lattice parameters continuously increase and decrease, respectively, indicating a solid solution-like behavior of the P2 structure.At sodium contents x < 1/3 (≳4 V), a voltage plateau and the emergence of the second phase (O2) with a decreased c-lattice and similar a-lattice parameter are detected.Electrochemistry indicates nearly full desodiation of the cathode active material at the end of charge, indicating that the O2 structure most likely is completely desodiated (O2-Na 0 Mn 3/4 Ni 1/4 O 2 ). 33During subsequent discharge (sodiation), this P2−O2 two-phase reaction is reversible 42 (not shown here).Please note the P2−O2 phase transition results in an anisotropic reduction of the unit cell volume of approximately 22% from P2-Na 1/3 Mn 3/4 Ni 1/4 O 2 to O2-Na 0 Mn 3/4 Ni 1/4 O 2 .
Based on this crystallographic evaluation, we can distinguish two different effects, which might contribute to aging of P2-Na x Mn 3/4 Ni 1/4 O 2 , namely, Jahn−Teller distortion due to the presence of Mn 3+ at low voltage and anisotropic volume change of the P2−O2 two-phase reaction at high voltage.To distinguish these effects, we have cycled cells with various cutoff voltages as presented in Figure 3 and Supporting Figure S6.Independent of the voltage window, cycling procedures commenced with an initial discharge to the respective voltage.For ease of comparison, only cycles at C/10 (17.3 mA g −1 ) are presented in Figure 3C.The full cycling data is provided in Supporting Figure S6.If the upper cutoff voltage exceeds 3.8 V (utilizing the P2−O2 phase transition), a pronounced capacity fade during the first twenty cycles is observed.If the low cutoff voltage is set below 2.0 V, a gradual capacity decay from cycle 20 onward is detected.If both critical voltage regimes are avoided and the voltage window is limited to 2.0−3.8V, cycling is relatively stable but with limited specific capacity.Independent of the voltage window, an increasing difference between the specific capacity at slow (C/10 = 17.3 mA g −1 ) and faster (C/3 = 57.7 mA g −1 ) cycles is apparent, hinting toward decreasing cell kinetics upon cycling (Supporting Figure S6).At high specific currents, the sodium metal counter electrode can be significantly polarized. 57To avoid any misinterpretations due to any potential influence of the sodium metal counter electrode, only cycles at C/10 are discussed throughout the manuscript.Additionally, the improved resolution of the voltage steps at low specific currents (C/ 10) allows differential capacity analysis.dQ/dV analysis for cycle 1, cycle 10, and cycle 100 is presented in Supporting   To distinguish between nonkinetic aging effects (e.g., loss of capacity, which remains inaccessible even under infinitely low current, such as mechanical/electrical isolation of active material or transformation toward electrochemically inactive phases) and kinetic aging effects (e.g., cell polarization caused by CEI growth and/or surface densification), we have performed GITT measurements in cycle 1 and cycle 10 (Figure 4A,B).Between these cycles, the cells followed the same cycling procedure as described above.The voltages after relaxation (OCV) are highlighted with a blue and red hull curve for cycle 1 and cycle 10, respectively.For the GITT measurement in cycle 10, we used the voltage shoulder of the large-zig-zag Na + /vacancy ordering (LLZ) at x = 2/3 to align the OCV hull curves of both GITT measurements at this point (x-offset).Both OCV hull curves are in good agreement, indicating that the structure of the pristine material is preserved, and no additional electrochemically active phases arise over the first ten cycles.However, utilization of the (de)intercalation capability is limited in cycle 10, which is supported by the differential specific capacity plot derived from the relaxed voltages (dQ/dV OCV ), as presented in Figure 4C.Comparing the on-load voltage profile of the GITT measurements (C/10 = 17.3 mA g −1 ), we find significant state of charge (SOC)-dependent differences between cycle 1 and cycle 10.At a high SOC (≥ 3.8 V), the initially small polarization within the P2−O2 two-phase region is signifi-cantly increased in cycle 10.During charge of cycle 10, polarization is continuously increasing with advancing transformation from P2 to O2, leading to a sloping voltage profile and limited desodiation of the electrode.During the subsequent discharge of cycle 10, polarization is significantly higher during discharge than in the previous charge, leading to a sloping voltage profile under the current load.A hysteresis in the x-direction between charge and subsequent discharge is

Chemistry of Materials
apparent, which is more pronounced in cycle 10 than in cycle 1, hinting toward more pronounced side reactions, such as electrolyte oxidation in cycle 10.At an intermediate SOC (2.0−3.8V), only a slight increase in polarization from cycle 1 to cycle 10 is apparent.At a low SOC (<2.0 V), polarization in cycle 10 is significantly higher than polarization in cycle 1, limiting sodiation of the electrode during discharge.Overall, the comparison of GITT in cycle 1 and cycle 10 provides strong evidence for a SOC-dependent increase of polarization, which contributes to the pronounced capacity fade during the first 10 cycles of P2-Na x Mn 3/4 Ni 1/4 O 2 .No indication of irreversible loss of active material within the first 10 cycles is observed.
To further investigate the kinetic effects, impedance measurements (PEIS) in a three-electrode setup were applied.PEIS measurements at various SOC during the first charge half-cycle are presented in Supporting Figure S9.During cycling, the impedance of the P2-Na x Mn 3/4 Ni 1/4 O 2 working electrode was measured every tenth cycle after charge to 4.3 V, after discharge to 2.5 V, and after discharge to 1.5 V, all at C/ 10 (17.3 mA g −1 ), as presented in Supporting Figure S8.The obtained Nyquist and Bode plots of a representative cell are presented in Figure 5A−F.Nyquist plots consist of a small arc at a high frequency and a larger one at a low frequency.The respective Bode plots show a bimodal peak of the phase shift with a simultaneous increase of the absolute impedance in the high-frequency regime (200 kHz to 10 Hz) as well as a peak of the phase shift with a simultaneous increase of absolute impedance at low frequencies (1 Hz to 5 mHz).−60 The submerged electrode exhibits a capacitive behavior in parallel with a charge transfer resistance, which can be modeled using a simplified Randles circuit consisting of a capacitor (C) in parallel with a resistor (R). 58,60,61−66 Based on our findings, we apply an equivalent circuit, as presented in Supporting Figure S10, to fit the obtained impedance spectra.The equivalent circuit consists of a high-frequency resistance (R HF ) to account for electronic contacts and ionic transport resistance in series with three modified Randles circuits (R i || CPE i ).Additionally, in the first cycles after discharge to 2.5 V, measurements show an approximately 45°slope in the Nyquist plot as well as increasing absolute impedance and a constant phase shift at very low frequencies, which is typically fitted using a Warburg element (W diff ) to account for diffusion processes. 58,60,67This Warburg element is only applied to the fitting, where clear signs of Warburg diffusion are found in impedance spectra.Based on the characteristic time constants, the three Randles circuits are assigned to (i) contact impedance (τ ≈ 10 −5 s), 68−71 (ii) transport through CEI (τ ≈ 10 −3 s), 70,71 and (iii) charge transfer (τ ≈ 0.1−10 s). 70,72,73pon cycling, impedance measurements after discharge to 1.5 V indicate an additional process at low frequency (R CT2 || CPE CT2 ).So far, the physical interpretation of this process remains elusive and the corresponding discussion is provided in the Supporting Information.Fitting results are presented in Figure 5G,H and Supporting Figure S11.The actual potentials after relaxation, at which PEIS measurements are performed, are plotted in Figure 5I.After charge to 4.3 V, the potentials at which PEIS is measured remain almost constant throughout cycling, in accordance with the two-phase P2−O2 voltage plateau (compare Figure 4).After discharge to 2.5 V and 1.5 V, the rest potentials at which PEIS is measured increase within the first 20 cycles in accordance with increasing polarization and remain almost constant for the following cycles.Charge transfer resistance (R CT ) derived from the fitting of impedance spectra is presented in Figure 5G.R CT is the highest at a high SOC, which is in accordance with polarization at high voltage in GITT measurements and reports in the literature. 57For high and intermediate SOCs, R CT increases rapidly (approximately triples) during the first 20 cycles and then remains approximately constant at a high level.The increase in charge transfer resistance is in accordance with increasing polarization upon cycling as observed in GITT measurements (Figure 4) and dQ/dV analysis (Supporting Figure S7).The parameter Q of a constant phase element (CPE) can serve as an indicator for the double-layer capacitance 57,74 and is plotted in Figure 5H.After charge to 4.3 V, Q CT approximately triples within the first 10 cycles, followed by a slight decrease in the subsequent cycles, indicating a significant increase of double-layer capacitance after charge to 4.3 V.A similar trend is observed after discharge to 2.5 V, but to a much lesser extent.After discharge to 1.5 V, Q CT remains almost constant during cycling, indicating little to no change of the double-layer capacitance.Additional information regarding PEIS evaluation is provided in the Supporting Information and in Supporting Figure S11.
Overall, our impedance study provides further evidence for (i) a continuous SOC-independent increase of polarization, (ii) a SOC-dependent drastic increase of charge transfer resistance at high SOC during the first 20 cycles, and (iii) an SOC-dependent change of double-layer capacitance at high voltage (high SOC).
Ex situ aberration-corrected (AC) HRTEM was performed to investigate local changes on the pristine and cycled material (after 10 cycles), respectively.The hexagonal morphology of the single-crystal particles (see Figure 1B), which reflects the hexagonal crystal structure of the P2-Na x Mn 3/4 Ni 1/4 O 2 material, allowed us to prepare TEM lamellae with an orientation along the [112̅ 0] zone axis.Representative AC-HRTEM images of the pristine material and the material after 10 cycles are presented in Figure 6A,B,F, respectively.For the pristine material, lattice fringes of the layered structure are clearly visible in the AC-HRTEM images.In Figure 6B, a depiction of the crystal structure as derived from XRD is included to illustrate the orientation of the [112̅ 0] zone axis.Analysis of the line intensity profile in the c-direction is in good accordance with the interlayer distance as derived from XRD (Figure 6C).Fast Fourier transformation (FFT) within the area marked by the green box is presented in Figure 6D, showing the crystal structure.Selected area electron diffraction (SAED) of a different TEM lamella is presented in Figure 6E, where the spatial frequencies of the reflections are in good agreement with the lattice fringes derived from the FFT shown in Figure 6D.The intensity of the SAED pattern is in good accordance with the calculated kinematic diffraction pattern for the hexagonal structure (Figure 6I), confirming the bulk structure on a local level.The layered structure is maintained from bulk close to the surface of the particle (Figure 6B).After 10 cycles, HRTEM images show lattice fringes of the layered structure in the bulk and an approximately 5−10 nm thick surface film consisting of nanometer-sized, crystalline areas with different lattice fringes (Figure 6F).The FFT in the bulk area (blue box) shows the same lattice fringe pattern as in Figure 6D, confirming the layered structure in the bulk of the cycled material (Figure 6G,I), in accordance with ex situ capillary XRD patterns as presented in Supporting Figure S13.The FFT in the surface near region (red box) results in a pattern, which matches very well with the calculated diffraction pattern of cubic NiO along the [110] zone axis (Figure 6H,J).After 10 cycles, we have found this surface film on all surfaces exposed to the electrolyte on several TEM lamellae.The cubic nature of nanometer-sized particles similar to NiO or MnO is confirmed by FFTs for several areas and on various TEM lamellae.Additionally, we observed severe cracking in the a−b plane of the active material (Supporting Figure S14).The surfaces of these cracks exhibit similar surface films.
Energy-dispersive X-ray (EDX) mappings, using scanning (S)TEM imaging of the pristine and cycled material (after 10 cycles), are presented in Figure 7A,B, respectively.As described in the Experimental Section and illustrated in Supporting Figure S1, the cycled material was harvested from electrodes after discharge to 2.68 V, resulting in similar sodium content in the layered oxide compared to the pristine material.For the pristine material, a homogeneous distribution of the elements Na, Mn, Ni, and O is observed.After 10 cycles, an increased contrast in the surface near region is observed in the high-angle annular dark-field (HAADF) STEM image, which indicates either heavier elements or a denser structure in the surface near region. 75This surface near region is enriched in Mn and Ni and poor in Na.In the bulk of the cycled material, Na, Mn, Ni, and O are distributed evenly.A Na-and Ccontaining film sits on top of this transition metal-enriched dense surface film, which might be related to the carbon binder domain of the electrode composite or originate from oxidative electrolyte decomposition (CEI products).A second set of EDX mappings for the pristine material after 3 cycles and after 10 cycles is provided in Supporting Figure S12.
Overall, HRTEM images and STEM-EDX mappings suggest the formation of a dense, 5−10 nm thick film on the surfaces exposed to the electrolyte during cycling.EDX mappings show that this surface film is rich in transition metals and contains no (or very little) Na.FFTs indicate that the surface film exhibits nanoparticles with a cubic structure similar to NiO or MnO.Additionally, EDX mappings provide evidence for the formation of a CEI.Please note that the density of NiO (ρ = 6.807 g cm −3 ICDD00-047-1049) and MnO (ρ = 5.361 g cm −3 ICDD01-075-6876) is considerably higher compared to P2-Na 0.58 Mn 3/4 Ni 1/4 O 2 (ρ = 4.141 g cm −3 ).All of these findings are consistent with surface densification on layered

Chemistry of Materials
oxide cathodes for LIBs (e.g., NCM, NCA), indicating that a high degree of sodium de-intercalation results in oxygen release, reduction of transition metals, and formation of dense surface films with a cubic rock salt structure (e.g., NiO).−80 For LiNiO 2 , the overall reaction for surface densification can be written as 81 Based on our findings, we propose a similar reaction to happen during cycling (repeated full desodiation) on the surface of the P2-Na x Mn 3/4 Ni 1/4 O 2 cathode active material To check for further signs of oxygen release during cycling, we have performed DEMS measurements (Figure 8).Released lattice oxygen is directly detected in some DEMS studies. 44,82,836][77][78]82,84 In the measurements presented in this work, no direct oxygen release was observed. Durin the first cycle of our DEMS experiment, no significant gas evolution is detected until the cathode potential exceeds 3.8 V.The small amounts of H 2 and CO 2 evolution detected before the cell reaches 3.8 V can be attributed to the reaction of the electrolyte with the sodium metal counter electrode.
With the onset of the P2−O2 phase transition (≥3.8 V), CO 2 evolution is detected, which is continuously increasing until the end of charge.During the subsequent discharge, the CO 2 evolution decreases.In the following, no gas evolution is detected until the cathode potential reaches approximately 3.3 V in the subsequent charge of cycle 2.−87 When the cathode potential exceeds 3.8 V in the second cycle, CO 2 and H 2 evolution are detected once more.The following cycles undergo a similar course of gas evolution with increasing amounts of gases.The CO 2 peak during charging at potentials ≥3.8 V is clearly bimodal in cycles three to five, with the first maximum located at the onset of the phase transition from P2 to O2 and the second maximum at the end of charge.Under the assumption that oxygen release from Na x Mn 3/4 Ni 1/4 O 2 most likely occurs at a high degree of desodiation, similar to layered lithium transition metal oxides (e.g., NCM, NCA), 77,78 the oxygen release must be related to the O2 phase (Na 0 Mn 3/4 Ni 1/4 O 2 ).Under constant current charging, it is reasonable for the P2−O2 phase transition to start on the surface of the cathode active material, potentially leading to the release of CO 2 arising from the reaction of released lattice oxygen species with the electrolyte and resulting densification of the cathode active material subsurface.At the end of charge, most of the cathode active material has undergone the P2−O2 phase transition, which results in a substantial anisotropic volume reduction of 22%.This volume change potentially causes cracking of the cathode active material, exposure of new surfaces to the electrolyte, and subsequently, new surfaces, which might be subject to the release of oxygen species and surface densification.Therefore, continuous increase of surface area due to cracking can rationalize the increasing gas evolution during cycling.The detection of released lattice oxygen in the form of CO 2 due to the chemical reaction of the reactive oxygen species with the electrolyte is supported by an increasing R CEI in impedance measurements (Supporting Figure S11) and would be in accordance with potential CEI products in EDX measurements (Figure 7B).
To further investigate cracking of our Na x Mn 3/4 Ni 1/4 O 2 cathode active material, we have performed air-tight SEM cross sections of electrodes in the pristine state, cycle 1, cycle 3, and cycle 10 after charging to 4.3 V and discharging to 2.68 V, respectively (Figure 9).The pristine electrode shows dense particles.A direct discharge of the pristine electrode to 1.5 V has no influence on the particle morphology (Supporting Figure S15).After direct charge to 4.3 V, clear cracks in the a− b plane of the active material are apparent.After three and ten cycles, cracking can be observed after charging to 4.3 V.After subsequent discharge to 2.68 V, particles appear dense, indicating reversibility of crack formation with respect to the resolution of SEM.However, TEM lamellae prepared from the same cycled samples exhibit distinct cracks in the a−b plane when investigated with the higher resolution of TEM (Supporting Figure S14).We therefore suggest that hundreds of nanometer-wide cracks formed in charged samples narrow during subsequent sodiation, resulting in a few nanometerwide cracks after discharge to 2.68 V.With decreasing width of the cracks, the contact area between particle fragments potentially increases, eventually resulting in electronic recontacting during discharge.Due to the observed reversibility of cracking, we assume that cracks do not fully penetrate the single-crystalline particles but instead remain partially connected allowing heal up of the cracks.The observation of SOCdependent cracking is in accordance with SOC-dependent increase of double-layer capacitance as indicated by PEIS.In the charged state (4.3 V), the degree of cracking observed in SEM electrode cross sections seems to decrease from cycle 1 to cycle 10, which is in accordance with the decreasing utilization of the P2−O2 phase transition during cycling (compare Figure 4).To quantify the extent of cracking, further experiments such as impedance measurements in blocking conditions or ex situ physisorption would be necessary. 74,88,89verall, P2-Na x Mn 3/4 Ni 1/4 O 2 cathode materials for sodiumion batteries suffer from capacity fading upon cycling in halfcells with the voltage range of 1.5−4.3V.The significant capacity fading in the first twenty cycles is mainly associated with decreasing utilization of the high voltage P2−O2 phase transition.From GITT and impedance measurements, we find evidence for a SOC-independent general increase of polarization and a SOC-dependent increase of polarization at a high SOC, both contributing to capacity fading.Upon prolonged cycling, some capacity is lost due to the loss of active material caused by mechanical stress arising from an orthorhombic distortion due to Jahn−Teller active Mn 3+ at low voltage.
After repeated charging to high voltage, HRTEM and EDX provide clear evidence of surface densification on cycled P2-Na x Mn 3/4 Ni 1/4 O 2 .The resulting rock salt structures (e.g., NiO, MnO) are poor electronic conductors 90 and lack the necessary diffusion paths for ionic transport of alkali metals, 91 resulting in an SOC-independent increase of polarization.Furthermore, the release of oxygen species during surface densification leads to electrolyte oxidation and gas formation in accordance with DEMS measurement.In O3-type layered LIB cathode active materials, such as Ni-rich NCMs, surface densification is reported to proceed via a spinel-type intermediate structure due to the migration of Ni ions from the transition metal layer to the lithium layer. 1,81For SIBs, similar surface densification resulting in NiO-like surface layers was reported for O3-type NaNi 1/2 Mn 1/2 O 2 cathode materials. 92For P2-type cathode active materials, the transformation to a spinel-type intermediate structure is unlikely from a structural point of view due to the particular oxygen stacking sequence of P2-type materials. 93−96 Therefore, we suggest that surface densification in P2-type sodium layered oxides might follow a different mechanism.Instead of the above-described structural pathway to surface densification, an electronic pathway via the oxidation of oxygen, 43,44,97 release thereof from the near-surface structure, and subsequent surface densification might be possible.Further investigations to clarify the reaction mechanism toward surface densification will be necessary in the future.
At a high SOC and for cycled cathode active material, cracks in the a−b plane are observed using TEM lamellae and SEM electrode cross sections.Such cracking upon charge of the cathode active material can cause partial electronic disconnection for active material particle fragments, resulting in poor electronic connection for these fragments.Under the power load, the resistance of the electronic pathway toward these particle fragments results in increased overpotentials.Macroscopically, the difference in electronic connection for a large number of particle fragments potentially causes the observed sloped galvanostatic potential profile for the P2−O2 phase transition as observed in cycle 10.The increasing resistance is in accordance with the SOC-dependent increase of polarization at high voltage as observed in GITT and impedance measurements.Upon repeated cracking, some parts of the cathode active material might be completely electronically disconnected, leading to a loss of active material.These findings are summarized in Figure 10.
Therefore, our findings can explain the SOC-independent increase of polarization due to surface densification and the SOC-dependent increase of polarization at high voltage due to cracking of the active material as well as the observed capacity loss after prolonged cycling due to the loss of active material.Further investigations to elucidate the SOC-dependent increase of polarization at low voltage are currently ongoing.

■ CONCLUSIONS
In this work, we have investigated the origin of capacity fading of micron-sized single-crystalline P2-Na x Mn 3/4 Ni 1/4 O 2 model material as the cathode for sodium-ion batteries.Using HRTEM and EDX, we identified surface densification as the possible cause for increasing polarization (SOC-independent).The occurrence of surface densification is supported by DEMS.Furthermore, we identified partly reversible cracking due to the significant volume change of the P2−O2 phase transition as a cause for SOC-dependent increase of polarization at high voltage.The aging behavior observed in this study is in good accordance with many reported P2-type sodium layered oxides, in particular P2-Na x Mn 1-y Ni y O 2 materials, 38,39,41,42,44 which suggests that our findings are not limited to the particular stoichiometry, but are rather a common phenomenon in this material class.Even though a couple of publications have provided indirect evidence for surface densification upon prolonged cycling of P2-type sodium layered oxides, 38,82,97−99 direct proof of surface densification in the literature is scarce.Zhang et al. 44 provide an indication for oxygen redox in P2-Na x Ni 1/3 Mn 2/3 O 2 , observe small amounts of oxygen release in DEMS, and demonstrate a three-layer thick (<1 nm) homogeneous densified surface layer via TEM.Using computational methods, a structure for the densified surface is proposed (Ni 2 Mn 2 O 7 ).In this work, we provide direct proof of surface densification in P2-type Na x Ni y Mn 1−y O 2 cathodes for SIBs using HRTEM and EDX and demonstrate the cubic rock salt structure of the nanoparticles within this surface film.The 5−10 nm thick, dense surface film consists of NiO and MnO nanoparticles.To the best of our knowledge, this is the first report providing clear evidence for composition and structure of surface densification on P2-type sodium layered oxide cathode materials for SIBs.For polycrystalline O3-type NaNi 1/2 Mn 1/2 O 2 cathode materials for SIBs, similar cracking behavior and comparable surface densification resulting in NiO-like surface films are reported. 92−81 However, in contrast to Ni-rich NCMs, we do not find any spinel-type intermediate phase.This finding is discussed with respect to the oxygen stacking in P2-type layered oxides and the thermodynamic stability of spinel-type alkali metal oxides.We hope our findings will spark further mechanistic investigations into the formation of surface densification in P2-type alkali metal layered oxides.

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.chemmater.3c01499.After galvanostatic charging/discharging to the respective potential, the cells were kept allowed to relax for 1 h to guarantee equilibrium before PEIS measurements.Supporting Figure S9.Impedance measurements (PEIS) at various state of charge for a representative cell.(A) Nyquist plot and (B) Bode plot.Supporting Figure S10.Equivalent circuits as applied to fit the obtained impedance data.Supporting Figure S11.Additional information for fitting of impedance data (PEIS).The applied equivalent circuits are presented on the top.(A) Potential at which PEIS measurements were performed, (B) high-frequency resistance, (C) contact resistance, (D) CEI resistance, (E) charge transfer resistance, and (F) characteristic time constants as derived from fitting of impedance data with the respective equivalent circuit.Supporting Figure S12.Additional TEM-EDX mappings for the pristine material, material after 3 cycles and after 10 cycles.Na is denoted in yellow, Mn in red, Ni in green, oxygen in blue, and carbon in magenta color.Supporting Figure S13.Ex situ capillary XRD pattern of the pristine material, cathode composite after 3 and after 10 cycles.Supporting Figure S14.TEM images of active material particles at different cycling stages showing the formation of crystal defects.Supporting Figure S15.SEM cross section after direct discharge to 1.5 V (PDF)

Figure 1 .
Figure 1.Physical characterization of the P2-Na x Mn 3/4 Ni 1/4 O 2 cathode active material.(A) Analysis of the XRD pattern using the Rietveld method with depiction of crystal structure and refined anisotropic crystallite size as insets.(B) SEM image.
Figure1B.The powder consists of micron-sized particles with a hexagonal platelet-like morphology.The morphology of the particles clearly reflects the hexagonal nature of the layered crystal structure.The crystal size observed in the SEM is in good accordance with the anisotropic crystallite size in the a−b plane as derived from XRD.In the c-direction, on the other hand, crystals appear thicker in SEM images compared to the values obtained from refinement of the XRD pattern, which hints to the fact that a few crystals might be stacked in the cdirection.In fact, we have found grain boundaries within the a−b plane using TEM (compare Supporting FigureS4).Synchrotron operando XRD was performed to analyze the evolution of the crystal structure of P2-Na x Mn 3/4 Ni 1/4 O 2

Figure 2 .b b hex 1 3
Figure 2. Synchrotron operando XRD of the first cycle.(A) Measured voltage profile and diffraction patterns presented as heat map.(B) Voltage and refined lattice parameters plotted over the calculated sodium stoichiometry.For the P′2 structure, the orthombic b-lattice parameter was transformed into the hexagonal system via = • b b hex

Figure 3 .
Figure 3. Electrochemical cycling using various voltage cutoffs.Cycles 1, 2, 3, and every tenth cycle are performed at C/10 (17.3 mA g −1 ) and all other cycles at C/3 (57.7 mA g −1 ).(A) Voltage profiles for a representative cell cycled within the voltage window of 1.5−4.3V. (B) Graphic representation of the voltage windows.(C) Specific discharge capacity plotted over the cycle number.Only cycles at C/10 (17.3 mA g −1 ) are presented.The full cycling data is presented in Supporting Figure S6.Error bars represent the standard deviation of at least two cells.

Figure 4 .
Figure 4. Comparison of GITT measurements in cycle 1 and cycle 10.GITT profile plotted over calculated sodium stoichiometry for (A) cycle 1 and (B) cycle 10.Cycle 10 is aligned using the relaxed voltages based on the voltage shoulder of the "large-zig-zag (LLZ)" ordering at x = 2/3.Voltages after relaxation are highlighted in full blue and red colors for cycle 1 and cycle 10, respectively.(C) dQ/dV plot calculated based on the relaxed potentials.

Figure S7 .
Figure S7.When the charging voltage exceeds 3.8 V, the capacity fade during the first ten cycles is mainly associated with a loss of capacity at high voltage (P2−O2 phase transition).For cycle 100, the high voltage plateau completely vanishes and all signals in the dQ/dV plot are diminished and slightly shifted in voltage, indicating an increase of polarization and a loss of utilizable active material due to complete electronic disconnection of corresponding parts of the active material (in the following referred to as the "loss of active material").For the 1.5−3.8V voltage window, no significant change in the dQ/dV plot is observed for cycle 1 and cycle 10.For cycle 100, all signals are decreased in intensity and shifted in voltage, again hinting at a loss of active material and an increase in cell polarization.For the narrow voltage window of 2.0−3.8V, only a slight shift in voltage is observed in the dQ/ dV plot during 100 cycles, indicating a minor increase in cell polarization but no loss of active material.Overall, the pronounced capacity fade over the first 20 cycles is dominated by a decreasing utilization of the high voltage plateau (P2−O2 phase transition), whereas the subsequent cycles show a slower

Figure 5 .
Figure 5. Impedance study.(A−C) Nyquist plots after charge to 4.3 V, discharge to 2.5 V, and discharge to 1.5 V, respectively.(D−F) Bode plots after charge to 4.3 V, discharge to 2.5 V, and discharge to 1.5 V, respectively.(G) Course of charge transfer resistance (R CT ) and (H) Q CT over cycling as derived from the fitting of impedance spectra.(I) Potentials at which PEIS measurements are performed after charge to 4.3 V, discharge to 2.5 V, and discharge to 1.5 V (all followed by 1 h of relaxation before PEIS measurements).In Nyquist and Bode plots, the obtained measurement data is presented as dots, and corresponding fittings are presented as lines.In panels (G)−(I), error bars represent the standard deviation of the respective values for three independent cells.

Figure 6 .
Figure 6.TEM images of pristine P2-Na 0.58 Mn 3/4 Ni 1/4 O 2 material (A−E) and material after 10 cycles with similar sodium content (F−H): (A) Aberration-corrected HRTEM image of the pristine material along the [112̅ 0] zone axis, (B) Magnification of image A (green box), crystal structure as derived from XRD is included as overlay in panel (B), (C) line profile derived along the orange line in image B, (D) FFT of the area marked with a green box in panel (A), (E) selected area electron diffraction derived from a different lamella of the pristine material, (F) aberrationcorrected HRTEM image of the material after 10 cycles, (G) FFT of the area marked with a blue box in panel (F), (H) FFT of the area marked with a red box in panel (F), (I) calculated kinematic diffraction pattern of P2-Na 0.58 Mn 3/4 Ni 1/4 O 2 along [112̅ 0], and (J) calculated kinematic diffraction pattern of NiO along [110].

Figure 7 .
Figure 7. HAADF image and EDX mapping of the surface near regions in TEM lamellae of (A) pristine material and (B) material after 10 cycles.Sodium is depicted in yellow, oxygen in blue, manganese in red, nickel in green, and carbon in magenta color.

Figure 8 .
Figure 8. DEMS measurement over the first eight cycles in a voltage window of 1.5−4.3V.

Figure 9 .
Figure 9. Air-tight SEM cross-sectional images illustrating the particle cracking phenomena.

Figure 10 .
Figure 10.Graphic summary of aging phenomena detected in this work.
Additional information for impedance measurements (PEIS) Supporting FigureS1.Illustration of electrochemical history of samples for ex situ characterization.SEM denotes samples for air-tight SEM electrode cross sections, TEM denotes samples for (AC)HRTEM investigations, EDX denotes samples for EDX mappings of TEM lamellae, XRD denotes ex situ capillary XRD measurements.A detailed description of the electro-chemical procedure is entailed in the experimental section of the manuscript.Supporting FigureS2.ATR-FTIR spectra of cathode active material powder and NaOH, NaHCO 3 and Na 2 CO 3 reference materials.(A) 400−4000 cm −1 and (B) zoom-in at 400−2000 cm −1 .Supporting FigureS3.Zoom into Rietveld Refinement as presented in Figure1A.Supporting TableS1.Rietveld refinement results and Wyckhoff sites.B iso values are are based on ref 100.Supporting Figure S4.TEM image of lamella prepared from pristine material.A grain boundary in a−b plane between two single crystals is visible.The crystalline nature of both crystallites was confirmed with SAED (not shown here).Supporting Figure S5.Waterfall diagram of obtained diffraction patterns from synchrotron operando XRD with corresponding galvanostatic voltage profile.Diffraction patterns and voltage curve of initial discharge is presented in blue color, subsequent charge is presented in red color.Supporting Figure S6.Specific discharge capacity over course of cycling for various voltage regimes.Cycles 1, 2, 3 and every tenth cycle are performed at C/10 (17.3 mA g −1 ), all other cycles at C/ 3 (57.7 mA g −1 ).(A) 1.5−4.3V, (B) 2.0−4.3V, (C) 1.5−3.8Vand (D) 2.0−3.8V.The standard deviation of at least two cells is presented as error bars.Supporting Figure S7.dQ/dV plots for cells cycled at various cutoff voltages for cycle 1, cycle 10 and cycle 100 with voltages windows of (A) 1.5−4.3V, (B) 2.0−4.3V, (C) 1.5−3.8V and (D) 2.0−3.8V. Supporting Figure S8.Cycling procedure for potentiostatic electrochemical impedance spectroscopy (PEIS) study.Every tenth cycle is performed at 17.3 mA g −1 (C/10, blue line), all other cycles are performed at 57.6 mA g −1 (C/3, red line).