Competing Mechanisms Determine Oxygen Redox in Doped Ni–Mn Based Layered Oxides for Na‐Ion Batteries

Cation doping is an effective strategy for improving the cyclability of layered oxide cathode materials through suppression of phase transitions in the high voltage region. In this study, Mg and Sc are chosen as dopants in P2‐Na0.67Ni0.33Mn0.67O2, and both have found to positively impact the cycling stability, but influence the high voltage regime in different ways. Through a combination of synchrotron‐based methods and theoretical calculations it is shown that it is more than just suppression of the P2 to O2 phase transition that is critical for promoting the favorable properties, and that the interplay between Ni and O activity is also a critical aspect that dictates the performance. With Mg doping, the Ni activity can be enhanced while simultaneously suppressing the O activity. This is surprising because it is in contrast to what has been reported in other Mn‐based layered oxides where Mg is known to trigger oxygen redox. This contradiction is addressed by proposing a competing mechanism between Ni and Mg that impacts differences in O activity in Na0.67MgxNi0.33‐xMn0.67O2 (x < 0 < 0.33). These findings provide a new direction in understanding the effects of cation doping on the electrochemical behavior of layered oxides.


Introduction
Sodium-ion batteries (SIBs) are gaining importance as promising alternatives to lithium-ion batteries (LIBs) for grid and large-scale energy storage applications due to the high Earth abundance of the elements used in SIBs such as Na, Fe, and Mn. [1]1b] Stable and high-capacity positive electrodes are a prerequisite for increasing the energy density of SIBs.Among the various reported cathode materials, layered oxides are ideal candidates because they can have tunable compositions with reasonable capacity and the electrochemical performance can be easily tailored by introducing different elements into the host structures.
Ni-Mn based layered cathode materials have attracted much attention in both LIBs and SIBs due to the high operating voltage of the Ni redox and its multiple electron couples (Ni 2+ →Ni 3+ →Ni 4+ ).In these materials, Ni is not only an active element that provides capacity, but it can also trigger oxygen redox when the voltage is above 4.1 V.This has been found for avariety of different Ni/Mn ratios, for example, compositions of Na 0.78 Ni 0.23 Mn 0.69 O 2 , Na 0.67 Ni 0.2 Mn 0.8 O 2 , Na 0.5 Ni 0.25 Mn 0.75 O 2 , and Na 0.67 Ni 0.33 Mn 0.67 O 2 . [2]Among these materials, Na 0.67 Ni 0.33 Mn 0.67 O 2 is one of the most common compositions, and it is known to have a P2 phase (following Delmas' notation). [3]Although oxygen redox in the high voltage region is known to initially improve the storage capacity in layered oxides, it ultimately can result in capacity fade due to a combination of oxygen loss and cation migration. [4]Another process that causes capacity fade is the phase transition from P2 to O2 which leads to significant volume changes and has been reported to start when exceeding approximately 4.1 V. [5] Therefore, finding a balance between stability and high capacity is required for future battery applications.
Cation doping, or more specifically ion substitution, is an effective and practical strategy for achieving superior properties in cathode materials, which has been widely demonstrated in both laboratory experiments and commercial products.5a,6] Suppression of the P2 to O2 phase transition is considered to be the most important reason for the better electrochemical properties by doping with these cations.In these systems, instead of a flat voltage plateau, a sloping potential profile can be observed for voltages above 4.0 V.This behavior has also been found in many systems where Ni n+ and Li + coexist with Mg 2+ , Cu 2+ , or Zn 2+ . [7]The origins of the sloping electrochemical behavior are still unknown, but they may be a result of a solid-solution type storage mechanism, kinetic effects, or different electrochemical contributions, among others.Similarly, the relationship between the sloping potential profiles and the P2 to O2 phase transition have not yet been well described, making further investigations necessary.
In this work we study the impact of substituting Ni in P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 with Mg and Sc.Mg 2+ and Sc 3+ were chosen as dopants because they both have similar ionic radii to Ni 2+ , while having different valence states.Sc 3+ was also chosen as a dopant because it is a 3d cation with a +3 valence state that is expected to remain electrochemically inactive.To the best of our knowledge, this is the first report of Sc 3+ doping in P2 layered oxide materials.Through a combination of resonant inelastic Xray scattering (RIXS), operando X-ray diffraction, X-ray absorption spectroscopy (XAS), X-ray pair distribution function analysis (PDF), and density functional theory (DFT), we found that while Mg and Sc can both decrease the oxygen activity through reduction of the Ni content, they have very different structural responses during cycling.Where the undoped (largest volume change) and Mg-doped samples (intermediate volume change) exhibit structural reversibility, the Sc-doped sample (lowest volume change) shows irreversible structural changes.These results were surprising at first because they point toward redox behavior being more critical than structural changes for improved cycle life in these materials.To further understand the role of Mg in suppressing the oxygen redox activity, we varied the amount of Mg doping and found that there exists a competition between Ni and Mg for triggering oxygen redox in Na 0.67 Mg x Ni 0.33-x Mn 0.67 O 2 which is therefore dependent on the molar ratio of Mg/Ni.

Undoped and Mg-, Sc-Doped (10%) Samples
The crystal structures of Na 0.67 Ni 0.33 Mn 0.67 O 2 (designated as P2-NaNMO), Na 0.67 Mg 0.1 Ni 0.23 Mn 0.67 O 2 (designated as P2-NaM 0.10 NMO), and Na 0.67 Sc 0.1 Ni 0.23 Mn 0.67 O 2 (designated as P2-NaS 0.10 NMO) were analyzed by XRD and Rietveld refinement.Each material was found to have a layered structure with space group P6 3 /mmc without any impurities (see Figure S1, Tables S1-S3, Supporting Information).The sodium storage performance of the three cathode materials was evaluated at a current density of 20 mA g −1 in a voltage range of 2.5-4.4V, as shown in Figure 1.Due to a higher Ni content in the undoped material (P2-NaNMO), the charge/discharge capacities of P2-NaNMO are 150/136 mAh g −1 , which is higher than that of P2-NaM 0.10 NMO (128/118 mAh g −1 ) and P2-NaS 0.10 NMO (136/122 mAh g −1 ).Compared to the undoped material, the Mg-and Sc-doped materials exhibit much smoother charge/discharge curves below 4.0 V.In the high voltage region up to 4.4 V, significant differences between different materials were observed.There is a long and flat high voltage plateau for the P2-NaNMO sample, which can deliver a capacity of about 70 mAh g −1 (out of 150 mAh g −1 for the full charge capacity in the 1st cycle).P2-NaS 0.10 NMO has a similar high voltage plateau, but it is shorter than that of the undoped P2-NaNMO, and only delivers a capacity of about 45 mAh g −1 (out of 136 mAh g −1 for the full charge capacity in the 1st cycle).In contrast, the Mg-doped material shows a sloping potential profile in the high voltage region, and the corresponding dQ/dV curve shows the lowest signal intensity above 4.0 V among the investigated materials, indicating that Mg doping suppresses the high voltage plateau more completely.The Sc-doped material shows electrochemical behavior similar to that of the undoped P2-NaNMO material, which also shows a high intensity peak above 4.0 V in the dQ/dV curve.These observations demonstrate that the two dopants have different roles in the electrochemical reaction.During long-term cycling, it can be seen that P2-NaM 0.10 NMO shows the best stability, delivering a reversible capacity of 58 mAh g −1 after 500 cycles and 160 days.Under the same conditions, P2-NaS 0.10 NMO and undoped P2-NaNMO only deliver capacities of 38 and 3 mAh g −1 , respectively (see Figure 1d,e; Figure S3, Supporting Information).Based on the excellent Na storage of P2-NaM 0.10 NMO, it can be concluded that suppressing the high voltage plateau is a promising strategy for enhancing the properties of P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 .
Oxygen redox can be triggered at high voltages, which has been shown for many layered oxides above 4.0 V. [8] Here, we used RIXS to identify differences in the oxygen redox behavior for the different compositions.Figure 2a-c shows the results of RIXS measurements at the O K-edge for P2-NaNMO, P2-NaM 0.10 NMO, and P2-NaS 0.10 NMO at open-circuit voltage (OCV), when fully charged to 4.4 V, and when fully discharged to 2.5 V during the first cycle.The spectra for the OCV and the fully discharged state were similar for all samples.However, at the fully charged state, a new energy loss peak at 8 eV can be observed, which is consistent with a signature of O─O bond formation that has been associated with oxygen redox, as reported by Bruce et al. and Yang et al. [9] While a strict quantification of oxygen redox is not appropriate due to issues with normalization and sample-to-sample variability when probing ex situ, we can nevertheless observe trends in the relative ratio of the main oxide signal at ≈6 eV in Figure 2a-c and the new energy loss feature at ≈8 eV to report the following qualitative trend in oxygen redox: P2-NaNMO > P2-NaS 0.10 0NMO > P2-NaM 0.10 NMO.This assignment is supported in Figure 2d, which highlights the region near 0 eV of the RIXS spectra of the three states for each material.A progression of sharp peaks can be observed only in the fully charged state, which are associated with vibrations arising from molecular O 2 units.4a,9a,b] Among the three materials, the intensity of the vibration signals for molecular O 2 in the P2-NaM 0.10 NMO material are the lowest and exhibit the fastest attenuation within the energy loss window.This is consistent with a lower intensity of the oxygen redox signal at an energy loss of 8 eV.The findings confirm that oxygen redox occurs for all samples at varying degrees and that the Mg-doped sample shows the smallest oxygen redox activity.The related mechanism is discussed in more detail below.
In order to more clearly observe differences in the oxygen redox of the three samples, RIXS mapping was conducted on samples in the fully charged state with an excitation energy between 526 and 534 eV.2d,9,10] The undoped material and Sc-doped material both show a bright spot in this region, but a similar feature is not found in the case of P2-NaM 0.10 NMO.This further supports that there is less oxygen redox in the Mg-doped material.Based on the Ni content in the three compositions, it is also expected that the intensity of the oxygen redox peak for the doped materials should be lower than that of the undoped P2-NaNMO material and also that the doped materials should demonstrate similar intensities for the oxygen redox.In contrast, according to the observed potential profiles and RIXS analysis, we find that the Mg-doped material has the lowest oxygen activity.These results are surprising, because Mg has been a popular element used to trigger oxygen redox in many Mn-based layered oxides such as Na 0.67 Mg 0.28 Mn 0.72 O 2 [9a,11] and Na 0.67 Mg 0.33 Mn 0.67 O 2 , [9a,11,12] and one would therefore expect Mg doping to enhance the oxygen activity also in Ni/Mn-based layered oxides.The above analysis, however, shows that Mg-doping in Ni-Mn based cathode materials may also suppress the oxygen redox activity.The RIXS data is Each material shows a unique structural evolution in the high voltage region, which can be attributed to the influence of the different dopants.It can be observed that the intensity of the main (002) reflection of P2-NaNMO gradually decreased upon charging in the high voltage region, and a new phase appears, as indicated by the appearance of a new reflection at 2.7°(2) (see Figure S7, Supporting Information, for enlarged view).5a,b,6c,9a] Unlike the clear phase change exhibited by P2-NaNMO, the doped materials show more continuous structural evolution in the high voltage region, as can be seen in Figure 3b,c.6b] The OP4 phase contains layers with alternating octahedral and prismatic stacking of Na in the c-axis, leading to the presence of oxygen in four different environments.The OP4 phase can be formed when full slab gliding is prohibited and is related to the reported "Z-phase," which consists of a continuously evolving intergrowth structure arising from P2 and O2 domains that go through an intermediate OP4 structure. [13]Materials that undergo a P2 to OP4 phase transition will have a much smaller volume change than the P2 to O2 phase transition of the parent material.In the case of P2-NaS 0.10 NMO, no clear phase change is observed in the high voltage region, although there is a change in lattice parameters which indicate a solid solution-type reaction. [14]hile the O2 phase of P2-NaNMO and the OP4 phase of P2-NaM 0.10 NMO return to the P2 phase, the structural changes for all samples are not fully reversible, where the lattice parameters are found to not fully recover to their initial state at the beginning of charge.
The corresponding c and a lattice parameters obtained from the refinement of the XRD patterns are shown next to the contour plots in Figure 3.For all samples, similar trends in the a lattice parameter can be observed, with this parameter decreasing during charge and increasing during discharge, although the reversibility upon discharge differs for each sample.In contrast, all three samples exhibit significant differences in the behavior of the c lattice parameter.The c lattice parameters in the investigated materials decrease by 21.8% and 4.6% for P2-NaNMO and P2-NaM 0.10 NMO, respectively, while P2-NaS 0.10 NMO demonstrated a 0.3% expansion.Although P2-NaS 0.10 NMO shows the smallest changes in lattice parameters, it also shows the highest structural irreversibility.This is surprising at first because it indicates that it is not the structural changes that dominate cycle life in these materials, but instead the redox behavior is more important.These parameter changes are in line with the observation from PDF measurements (see Figure S8, Supporting Information).Overall, Mg is found to be the better dopant for Na 0.67 Ni 0.33 Mn 0.67 O 2 compared Sc.While Sc doping is more effective in minimizing the lattice shrinkage during charging, it is less effective in mitigating oxygen redox and shows only a limited structural reversibility during subsequent charging.As a consequence, cycle life is better compared to the parent material but poorer compared to the Mg-doped sample.This shows that the capacity fade is not linked to only volume and structural changes, but instead redox chemistry becomes dominant for dictating cyclability in this case.X-ray absorption near edge structure (XANES) analysis was used to investigate the changes in oxidation states of Ni and Mn for all samples.Figure 4a-c and Figure S9a, Supporting Information, shows the XANES data for the Ni K-edge at different charge/discharge states.When charging to 4.0 V (see Figure S10, Supporting Information), P2-NaNMO and P2-NaM 0.10 NMO show a small shift toward higher energies relative to the OCV state, whereas P2-NaS 0.10 NMO shows a stronger shift, indicating that more Ni redox is involved in the electrochemical reaction for P2-NaS 0.10 NMO at this stage than in the other compositions.In the high voltage region at 4.4 V, all three samples exhibit distinct energy shifts, indicating that Ni redox also takes place in the high voltage regime.Upon discharge, P2-NaNMO and P2-NaM 0.10 NMO show a high degree of reversibility, almost perfectly overlapping the XAS spectra of the OCV state.For the P2-NaS 0.10 NMO sample, however, the Ni redox is clearly only partially reversible with Ni remaining in a higher oxidation state compared to what was found at OCV, as can be observed by the jump remaining at higher energy for the fully discharged state.This indicates that P2-NaNMO and P2-NaM 0.10 NMO have a better Ni redox reversibility compared to P2-NaS 0.10 NMO in the first cycle, which is consistent with the XRD results.In addition, the Mn K-edge XANES at the different charge/discharge states was also collected (see Figure S11a-c, Supporting Information).These results show that Mn is only electrochemically active in the Sc-doped material.In this case, an Mn 3+/4+ redox is most probable because the high valence Sc 3+ replacing the low valence Ni 2+ requires lowering of the Mn valence for charge compensation.This has the additional consequence of increasing distortion in the MnO 6 octahedra, as evidenced in the comparison of the Mn K-edge EXAFS of the samples at OCV in Figure S9d, Supporting Information.Here is it shown that despite an overall average octahedral coordination in the samples, the intensity of the Mn─O signal is reduced and shifted slightly to lower radial distances in the case of Sc doping.This suggests greater disorder in the structure which would be consistent with Mn and O existing in multiple sites.This is strongly related to the reported pseudo-Jahn Teller Effect which is known to impact the structural reversibility in Ni-free Mn layered oxides, [15] and can help explain the observed structural irreversibility found during cycling for the Scdoped material.
Local structural variations around Ni and Mn were monitored through analysis of the extended X-ray absorption fine structure (EXAFS) at the Ni and Mn K-edges (see Figure 4d-f; Figures S9  and S10, Supporting Information, respectively) for all samples.9a,14b] In Figure 4d,e, the Fourier transformed (FT) EX-AFS spectra at the Ni K-edge are shown, where it can be seen that the signal intensities of the Ni─O shell for P2-NaNMO and P2-NaS 0.10 NMO both decrease between the OCV and charged to 4.0 V states, indicating a decrease in coordination number.No change in signal intensity is observed between these two voltages in the case of P2-NaM 0.10 NMO, supporting a higher structure stability in the Mg-doped material.When the voltage is increased to 4.4 V, all materials show a significant reduction in both the radial distance and intensity related to the Ni─O shell, implying that the bond length of Ni─O and the coordination number both decrease.However, the Mn─O bond length in the Mn K-edge EX-AFS does not appear to change appreciably (see Figure S11d-f, Supporting Information).
Combined with the RIXS measurements and the above EX-AFS data, it can be concluded that the Ni-O bond length is directly related to the oxygen activity.Based on the operando XRD, P2-NaS 0.10 NMO follows a solid solution reaction in the high voltage region.However, the material still shows significant changes in the Ni─O and Ni─Me radial distances and the coordination numbers when charging from 4.0 V to 4.4 V, indicating that the local structural distortion caused by the oxygen redox was much higher.Similar to P2-NaNMO, P2-NaS 0.10 NMO also showed a large decrease in the radial distance and coordination number.However, P2-NaS 0.10 NMO exhibited a larger change in the radial distance compared to the P2-NaNMO material.The P2 to O2 phase transition experienced by the P2-NaNMO material has some small benefits related to limiting the change in radial distance between TM-O, but overall, the phase change is not beneficial due to the global change in lattice parameters.As for P2-NaM 0.10 NMO, it showed the smallest changes among the three samples in the high voltage region due to a less pronounced oxygen redox and moderate phase transition.In the shell of Ni-Me, P2-NaNMO, and P2-NaS 0.10 NMO also showed a clear change in the high voltage region, but these changes were quite small, which can be related to the inactive Mn in these compositions.Moreover, the curves for P2-NaNMO and P2-NaM 0.10 NMO in the discharged state can coincide with the OCV state, further demonstrating that the reversibility is higher than that of P2-NaS 0.10 NMO in the first cycle.
In order to understand the above findings, DFT calculations were used to study the effect of Mg and Sc dop-ing on the lattice parameters and the redox mechanism of P2-Na x Ni 0.33 Mn 0.67 O 2 .The atomistic structures of the three materials with Na contents of x = 0.67 and 0.22 (corresponding to the OCV and fully charged states, respectively) were determined (see Figure 5a; Table S5, Supporting Information) by performing an extensive set of Coulomb energy and DFT calculations as discussed in the method section.The computed lattice parameters using DFT-PBE+U and DFT-PBE+U+D3 agree with the experimental observations and confirm the oxidation of TMs with desodiation (for further details, see Table S5, Supporting Information).
To characterize the redox mechanism, the magnetic moment (represented by the number of unpaired electrons (N unp )) was calculated on the elements and the spin density (SD, Figure 5b) using DFT-HSE06.The redox mechanism of the elements in each composition was first briefly summarized and compared in Figure 5b-h and Figure S11, Supporting Information.According to our calculated results, the Ni and Mn have the same oxidation state of ≈2 + and ≈4 + for P2-NaNMO and P2-NaM 0.10 NMO in the OCV state (see Figure 5f-h; Figure S12, Supporting Information).Upon desodiation from x = 0.67 to 0.22, Ni is oxidized to a higher oxidation state and Mn remains almost unchanged.These results are consistent with the XAS results that Ni is the only cation involved in the electrochemical reaction of P2-NaNMO and P2-NaM 0.10 NMO.For the Sc-doped material, the Ni is also ≈2 + in the OCV state, while Mn shows mixed oxidation states of ≈3 + and ≈4 + .This is supported by a linear combination analysis of the Mn K-edge XAS data which indicates a significant increase in the presence of Mn 3 + with Sc doping.As a result, both Ni and Mn are expected to be electrochemically active during the desodiation process which is in line with the XANES measurements (for further details, see Figure S12, Supporting Information).
In addition to the cation redox, oxygen redox was also found in the three materials, in particular for O anions that are close to Ni cations.It can be observed that the calculated value of N unp on O is almost zero for these samples in the OCV states, indicating an average charge of 2 − on this anion (see Figure 5c-e; Table S6, Supporting Information).Upon desodiation from x = 0.67 to 0.22, the computed N unp for the O anions show different changes amongst the different materials.For the undoped material, the O anions experience an obvious oxidation during charging, as shown by the large differences in N unp in the OCV and fully charged states (for example, selected O atoms are highlighted in Figure 5b and their corresponding changes in N unp are highlighted by labeling numbers 2, 4, 8, 13, 15, and 16 in Figure 5c).For the Mg and Sc doped materials, a distinctly different oxidation behavior of the O anions can be observed depending on their nearest neighbor metal ions.O anions that are close to Mg or Sc are inactive, as shown by the very limited change in N unp (see O atoms numbered 5, 6, 7, 13, 15, and 16 for the Mg-doped material in Figure 5b,d and O atoms numbered 2, 6, 12, 13, 14, and 18 in the Sc-doped material in Figure 5c,e).This is in contrast to the oxygen anions surrounding Ni, which show significant changes in N unp for all samples.These differences are greatest in the undoped material, followed by the Sc-doped material, and the Mg-doped material shows the smallest change.This indicates that while all three materials undergo oxygen redox, Mg doping can significantly reduce the oxygen redox activity.These calculations are consistent with the experimental results.

Mg-Doping with Varying Concentrations (5-20%)
The experimental results and the computational study both show that Mg doping (10%) decreases the oxygen activity for P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 .7c,9a,11] To further understand the role of Mg in suppressing the oxygen redox activity, a series of materials with different Mg contents (5-20%, replacing Ni) were synthesized and inves-tigated by XRD, RIXS, EXAFS, and PDF analysis.The XRD patterns in Figure S13, Supporting Information, indicate that all materials have a P2 layered structure with a space group of P6 3 /mmc, but they exhibit different electrochemical behavior, especially in the high voltage region (grey region in Figure 6a).With increasing Mg-content, the charge/discharge curves become smoother and the corresponding capacity was found to first decrease before increasing again.It was also found that this resulted in increased long-term cycling stability for the 10% Mg-doped sample relative to the other Mg doping concentrations, as shown in Figure S15, Supporting Information.The high voltage plateau gradually becomes sloping and then a new flat high voltage plateau reappears as the doping content reaches 20%.The corresponding dQ/dV curves in Figure S14, Supporting Information, show that the intensity of the peak in the high voltage region also gradually decreases, followed by the appearance of a new and high intensity peak with higher potential.9a,11,12] The results of the RIXS measurements in Figure 6b and Figure S14, Supporting Information, are in line with the electrochemical data.The ratio of the signal associated with oxygen redox at an energy loss of ≈8 eV and that of the main oxide signal at ≈6 eV gradually decreases, then increases.The P2-NaM 0.15 NMO sample exhibits the weakest oxygen redox signal, which is consistent with the lowest intensity of the peak in the dQ/dV plot (see Figure S14d, Supporting Information).The Ni and Mg K-edge EXAFS at the fully charged states were also collected and are shown in Figure 6c,d, respec In the Mg K-edge EXAFS, the radial distance of the Mg─O shell shows a similar value for P2-NaM 0.10 NMO and P2-NaM 0.15 NMO, and a smaller value for P2-NaM 0.20 NMO.This result is consistent with what is observed in PDF analysis which probes the global Me─O distances (where Me is Ni, Mn, and/or Mg) and shows that this bond distance increases with Mg doping content from 0% to 15% and then decreases again for P2-NaM 0.20 NMO (see Figure S16b, Supporting Information).All of the above results reveal that the Mg─O bond length becomes shortened with increasing Mg doping content, which can therefore activate the oxygen redox in P2-NaM 0.20 NMO.9a]

Discussion
From the above analysis, we know that both the Ni─O and Mg─O bond lengths will shorten during charging, thereby activating oxygen redox.This relationship can be clearly observed by comparing the normalized Ni─O and Mg─O bond lengths to the normalized oxygen activity, as shown in Figure 7a.At the same time, we expect that Ni and Mg share the same lattice sites within the crystal structure, which means that Ni and Mg will surround similar O atoms.Under these circumstances, Ni and Mg in Na 0.67 Ni x Mg 0.33-x Mn 0.67 O 2 will compete for oxygen polarization (Ni─O vs Mg─O).With no Mg doping (i.e., in the parent P2-NaNMO material), there is nothing to impact the Ni─O bond length, and a 0.06 Å decrease of the Ni─O radial distance from the OCV state to the fully charged state can be observed in the Ni EXAFS.As the Mg doping content increases, the O atoms gradually move closer to the Mg atoms, and their distance from the Ni atoms elongates.This corresponds to the observed increase in the Ni─O bond length, and induces the uncommon phenomenon that Mg suppresses the oxygen redox for P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 .These measurements clarify that the oxygen activity is directly related to the ratio of Mg/Ni, which dictates the degree of oxygen redox for this class of materials.
In this way, we utilize the molar ratio of Mg/Ni in Na 0.67 Ni x Mg 0.33-x Mn 0.67 O 2 (0 ≤ x ≤ 0.33) to clearly demonstrate changes in the oxygen activity, as shown schematically in Figure 7c.When the molar ratio of Mg/Ni is close to 1 (P2-NaM 0.15 NMO), the contributions from Ni and Mg toward oxygen polarization become balanced, thereby limiting oxygen redox.When the ratio of Mg/Ni is less than 1 (P2-NaNMO, P2-NaM 0.05 NMO, and P2-NaM 0.10 NMO), the Ni is responsible for triggering oxygen redox, but Mg is able to stabilize the oxygen environment due to its higher covalency and shorter bond length, thereby decreasing the overall oxygen activity.When the ratio of Mg/Ni is greater than 1 (P2-NaM 0.20 NMO in this study and P2-Na 0.67 Mg 0.28 Mn 0.72 O 2 from the literature [9a,10] ), Mg becomes responsible for activating oxygen redox, where the Mg─O bond length distortion becomes greater than that of Ni─O.This finding is well supported by DFT results that suggest a higher redox activity of the oxygen for the sample with the highest Mg content (Na 0.22 Mg 0.22 Ni 0.13 Mn 0.67 O 2 ), see Figure 7b.Due to this competing mechanism, the intensity of the oxygen redox becomes a function of the Mg content with a minimum at P2-NaMg 0.15 NMO, see Figure 7c.Li, Cu, and Zn are also known to trigger oxygen redox in Ni free Mn-based layered oxides, and it is likely that these doping elements have similar concentrationdependent properties to that of Mg in Na 0.67 Ni 0.33 Mn 0.67 O 2 .P2-Na 0.67 (Li, Mg, Zn, Cu) 0.10 Ni 0.23 Mn 0.67 O 2 samples were prepared with these dopants (10%), as shown in the Supporting Information (Figures S17 and S18, Supporting Information), and as predicted, the oxygen redox was also suppressed with the introduction of these dopants (see Figure S19, Supporting Information).This demonstrates that the changes in oxygen redox for doping with these elements are also suitable, and we can express this with the equation Na 0.67 Ni x (Li, Mg, Zn, Cu) 0.33-x Mn 0.67 O 2 .It is expected that other compositions containing the P2 phase (e.g., with changes in Na concentration) will share these features of inducing a sloping electrochemical behavior when the molar ratio of Ni/Me (Me = Li, Mg, Zn, Cu, or a combination of dopants) is close to 1.

Conclusion
In summary, doping with both Mg and Sc significantly improves the cycling stability of Na 0.67 Ni 0.33 Mn 0.67 O 2 , but two different behaviors are induced in the high voltage region with the different dopants.Table 1 provides a comparison of the electrochemical, structural, and redox behavior observed for the samples with different dopants.While the Sc-doped sample showed the smallest change in lattice parameters during charging (best structural stability), it also presented the highest structural irreversibility during subsequent sodiation.Overall, the capacity of the Sc-doped material fades faster relative to the Mgdoped material.This demonstrates that it is the redox behavior and not the structural stability, as normally understood, that dominates the cycle life in these materials.Through RIXS measurements, we found that both dopants successfully decrease the de-gree of oxygen redox, but the suppression is strongest in the case of Mg.This result seems to contradict other studies as Mg is wellknown to trigger oxygen redox for Ni-free Mn-based layered oxides.In order to determine the origin of the reduced oxygen redox in the Mg-doped materials, a series with different Mg doping concentrations was investigated by RIXS, PDF analysis, and EXAFS.The results demonstrated that the presence or disappearance of oxygen redox is dependent on the molar ratio of Mg/Ni in Na 0.67 Mg x Ni 0.33-x Mn 0.67 O 2 .When the Mg/Ni ratio was less than (greater than) 1, the Ni (Mg) will play the role of triggering the oxygen redox.A balance is achieved for a Mg/Ni ratio close to 1, where the oxygen redox could be prevented to the greatest degree.This indicates that a balance between Ni─O and Mg─O covalency is the decisive factor for controlling oxygen redox in doped Ni-Mn based layered oxides.This competition is identified as the origin of the lower oxygen activity of P2-NaM 0.10 NMO, which we similarly demonstrated for other dopants that show similar electrochemical behavior in the high voltage region, including Li, Zn, and Cu.Furthermore, it is well-known that oxygen redox is the main bottleneck for cycle life beyond the well-known P2 to O2 phase transition.These findings provide a new direction for understanding the effects of cation doping on the electrochemical behavior of layered oxides.

Experimental Section
Synthesis Method: A sol-gel method was used to synthesize the precursor followed by high-temperature calcination. [16]NaNO 3 (>99.0%,Sigma-Aldrich), Me(CH 3 COO) 2 (Me = Mg, Zn, Cu, CH 3 COOLi, >99%, Alfa Aesar), and Sc(NO 3 ) 3 •9H 2 0 (>99.99%,abcr Gute Chemie) were used as raw materials.An excess of 5 mol% NaNO 3 was added for all reactions, and 0.03 mol of product was synthesized in each experiment.All raw materials were dissolved in 80 mL of deionized water in a stoichiometric ratio, mixed for 2 h at 50 °C before adding the corresponding amount of citric acid (>99.0%,Sigma-Aldrich) (metal:acid = 1:1) and mixing for a further 3 h.The temperature was then increased to 120 °C to evaporate the water, forming a gel after around 3 h.The gel was then transferred to a crucible and heated in a furnace at 500 °C for 5 h.After cooling down to room temperature, the materials were ground to powder, and a pellet was pressed under 15 MPa.The pellets were calcined at 900 °C for 10 h and at 950 °C for another 10 h in air.The pellets were then transferred to an argon-filled glovebox after the temperature cooled down to 200 °C.
Electrochemical Characterization: The working electrodes were prepared using the active material, Super P (SP, MTI Corp.) and polyvinylidene difluoride (PVDF, MTI Corp) in a weight ratio of 80:10:10.The PVDF binder was dissolved in N-methyl-2-pyrrolidone (NMP, Sigma-Aldrich) with a concentration of 0.04 g mL −1 .The active material and Super P were mixed in a mortar for 10 min.The PVDF solution was then added, and mixed for another 5 min to form a homogenous slurry, which was then cast onto carbon-coated Al foil (MTI Corp.) with a doctor blade at a thickness of 300 μm.All the above-mentioned steps were carried out in an argon-filled glove box (with H 2 O and O 2 concentrations <0.1 ppm).The electrodes were dried on a hot plate and then punched into 12 mm diameter electrodes which were then dried in a Büchi at 120 °C for 12 h. 1 m NaClO 4 dissolved in PC:FEC (propylene carbonate, fluoroethylene carbonate, Sigma-Aldrich) with a volume ratio of 98:2 was used as the electrolyte and glass fiber (Whatman) as the separator.2032-type coin cells were assembled in an argon-filled glove box with sodium metal (BASF Corp.) as the counter electrode.All cells were tested in the voltage range between 2.5 and 4.4 V (vs Na + /Na) using a constant-current protocol at various current densities with a Neware battery test system ( CT-4008Tn-5V10mA-164, Shenzhen, China).
Structural Characterization: Powder XRD was performed with a Bruker D2 Phaser XRD using Cu K radiation ( = 1.5406Å, 30 kV, 10 mA) between 10°and 70°(2) with a step of 0.02°/min.An air-tight Si-based sample holder was used for all measurements.The operando XRD measurements were carried out at beamline P02.1, PETRA-III at DESY (Deutsches Elektronensynchrotron) in Hamburg, Germany, using synchrotron radiation with an energy of ≈60 keV ( = 0.20733 Å). [17] The operando cells were designed in-house and assembled in an argon-filled glovebox (H 2 O and O 2 concentrations <5 ppm, provided by beamline). [18]The electrochemical data were collected using a Biologic VMP-3 potentiostat with an EC-Lab software, which version was 11.34.All the cells were rest for 20 min, then charged/discharged at 0.1 C at a voltage range of 2.5-4.4V.The operando XRD data were collected as an ensemble of up to 5 cells simultaneously through the cells with an acquisition time of 1 min per cell before moving to the next cell using a motorized X-Y stage with a travel range of 350 mm for both directions.The movement of the stages and triggering of the acquisition were controlled by the python3 script.Therefore, for each operando cell a 1-min measurement was performed every 6 minutes.All the Rietveld refinements of the XRD patterns were performed using GSAS I.
Pair Distribution Function Analysis: Total scattering were performed at room temperature using the instrument at beamline P02.1, PETRA III at DESY, Germany.The electrodes were washed with dimethyl carbonate (DMC, Alfa Aesar) and the electrode material was then scraped off the current collector and filled into glass tubes with a diameter of 5 mm.A corner configuration was applied to the all the materials, the acquisition time was set to 6 min, and pdfget3 was used to transform the data.
Resonant Inelastic X-Ray Scattering: RIXS measurements were carried out at room temperature and under 10 −8 mbar at U41-PEAXIS at BESSY II, Germany. [19]The spectrometer was positioned at specular conditions relative to 60 degree scattering angle and was optimized to a combined resolution of 90 meV using a carbon tape.The excitation energy of each 1D curve was at the O K-edge at 531 eV, and the acquisition time was 20 min.The excitation energy of each mapping was at the O K-edge at 526-534 eV and the acquisition time was 9 h.A vacuum suitcase shown in Figure S20, Supporting Information, was used to transfer the sample from an N 2 -filled glovebox to the test chamber.The coin cells used for the RIXS measurement were charged/discharged to the target voltage, then disassembled and the electrodes were washed with DMC before drying and spreading onto Cu tape.All steps were carried out in an argon-filled glovebox.Data were processed using the beamline software, Adler-4.0.
Spectroscopic Characterization: Ex situ Ni and Mn K-edge XAS were collected at room temperature at KMC-2 at BESSY II, Germany in both transmission and fluorescence modes. [20]Reference spectra for energy calibration were collected before and after each sample measurement.XAS spectra were obtained by subtracting the pre-edge background from the overall absorption and normalizing to the spline fit using Athena software.The k 2weighted EXAFS was Fourier transformed over the limited range of k from 3 to 11 Å −1 .Ex situ Mg K-edge XAS was collected at room temperature in a UHV chamber at PIRX beamline [21] at National Synchrotron Radiation Centre Solaris, [22] Poland in the partial (PFY) fluorescence detection mode.The data was processed using Athena software.The k 2 -weighted EXAFS was Fourier transformed over the limited range of k from 3 to 8 Å −1 .
Scanning Electron Microscopy: SEM images were taken with a Phenom Pharos Desktop SEM from Phenom world using an accelerating voltage of 10 kV and a secondary electron detector.The corresponding energydispersive X-ray spectroscopy (EDX) mapping images were also collected for each sample, and the acquisition time was 20 min.
Computational Methods: Spin-polarized DFT calculations were performed using the projector augmented wave (PAW) potential method [23] implemented in the Vienna Ab Initio Simulation Package (VASP) code. [24]Generalized gradient approximation (GGA) within the scheme of Perdew-Burke-Ernzerhof (PBE) [25] was used as the basis of the exchange-correlation (XC) functional.The Hubbard U correction proposed by Dudarev et al. [26] with the following U values was applied, respectively, for Sc, Mn, and Ni to compute the atomistic structures: 5.0 eV, 5.2 eV, and 6.8.The dispersion D3 correction [27] was added for the low Na concentration.We applied the Heyd-Scuseria-Ernzerhof (HSE06) [28] functional to compute the electronic structure (i.e., number of unpaired electrons (N unp ) on elements and spin density).The discharged and charged systems were modeled by 3 × 3 × 1 supercells with the following number of atoms: Na A Gamma-centered k-point mesh of 2 × 2 × 2 and an energy cut off of 550 eV as well as an electronic and a force convergence criterion of 10 −4 eV and −2.00 × 10 −2 eV Å −1 , respectively, were used for DFT calculations.Magnetic moments (number of unpaired electrons) were calculated using a tighter electronic and force convergence criterion of 10 −6 eV and 10 −5 eV Å −1 , respectively.Total Coulomb energy calculations on possible combinations were carried out using the so-called supercell code. [29]Atomistic structures and spin density (SD) were visualized with VESTA program. [30]To find the most favorable structures of P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 with different x values, the total Coulomb energy (E c ) of a large number of likely structures was modeled and calculated.By performing DFT-PBE calculation on the electrostatically most favorable configuration, the lowest total energy structure was determined and that structure was used to study charged case as well as Mg and Sc doped system.The PBE optimized geometries were used for further DFT-PBE+U, -PBE+U+D3 (for Na 0.22 ), and -HSE06 calculations (see Supporting Information for details).

Figure 3 .
Figure 3. Contour plots of synchrotron-based operando XRD measurements during the first charge/discharge process and the corresponding changes in the c and a lattice parameter for a) P2-NaNMO, b) P2-NaM 0.10 NMO, and c) P2-NaS 0.10 NMO.Reflections from the aluminum current collector and sodium metal counter electrode are indicated with * and ♦, respectively.

Figure 4 .
Figure 4.The Ni K-edge XANES spectra and corresponding EXAFS spectra at various charge/discharge states for a,d) P2-NaNMO, b,e) P2-NaM 0.10 NMO, and c,f) P2-NaS 0.10 NMO.Insets in (a-c) magnify the region between 8340-8345 eV for the OCV and discharge to 2.5 V states in order to highlight the relative irreversibility of P2-NaS 0.10 NMO.

Figure 5 .
Figure 5. DFT calculations for P2-NaNMO, P2-NaM 0.10 NMO, and P2-NaS 0.10 NMO at OCV (n Na+ = 0.67 mol) and fully charged state (n Na+ = 0.22 mol).a) Atomistic structures by using DFT-PBE+U.b) Computed spin density plots (an isosurface of 0.006 eV Å −3 ) using DFT-HSE06, the up-and down-spin electrons are in yellow and blue, respectively (The atom colors are the same as in (a), but the isosurfaces distort the colors when overlayed).The number of unpaired electrons N unp on c-e) O and f,g) Ni for the three materials (black points are for the OCV state and red points are for the fully charged state).The labeled O atoms in (c-e) correspond to the labeled atoms in (b).
tively.The radial distance of the Ni─O shell increases with Mg doping from P2-NaNMO to P2-NaM 0.15 NMO, and then it shows nearly no change with further increase in the Mg doping content.This indicates that the radial distance of the Ni─O bond length is constant when the Mg doping content is over 15%.Based on the Ni K-edge EXAFS results, a shorter Ni─O bond length has a stronger oxygen redox in Na 0.67 Ni x Mg 0.33-x Mn 0.67 O 2 (x ≤ 0.15).

Figure 7 .
Figure 7. a) The relationship between the normalized bond length (Ni─O and Mg─O) and the oxygen activity.Bond lengths were normalized to the longest bond lengths taken from EXAFS measurements, while the oxygen activity was normalized to the intensity of the dQ/dV curves in the high voltage range.b) DFT calculations of the number of unpaired electrons N unp on anions for P2-NaNMO, P2-NaM 0.10 NMO, and P2-NaMg 0.22 NMO at the fully charged state (n Na+ = 0.22 mol).c) Schematic illustration of an oxygen environment and oxygen activity change with the different Mg doping contents in P2-Na 0.67 Ni 0.33 Mn 0.67 O 2 at the fully charged states.