Coupling the structure evolution and the oxygen redox electrochemistry in layered transition metal oxides


 Lattice oxygen redox activity offers an unexplored way to access the latent superior electrochemical property of transition metal oxides for rechargeable batteries. However, the redox reaction of the lattice oxygen is often accompanied by unfavorable structural transformations and the corresponding degradation of electrochemical performances, precluding its practical application. Herein, we explore the close interplay between the local structural change during the dynamic intercalation process and the solid-state oxygen electrochemistry in the short- or long-term battery operation for layered transition metal oxides. By employing two model systems of the layered Na0.6(Li0.2TixMn0.8−x)O2 with the oxygen redox capability, it is demonstrated that the substantially distinct evolutions in the oxygen redox activity and reversibility are caused by different cation migration mechanisms available in the system during the de/intercalation (i.e. out-of-plane and in-plane migrations of transition metals (TMs)). We show that the π stabilization upon the oxygen oxidation initially aids in the reversibility of the oxygen redox and is predominant in the absence of TM migrations, however, the π-interacting oxygens are gradually replaced by the σ-interacting oxygens that trigger the formation of O–O dimers and the structural destabilization over cycles. More importantly, it is revealed that the distinct TM migration paths available in the respective layered materials govern the conversion from π to σ interactions and its kinetics. It infers that regulating the dynamics of TMs in the layered materials can play a key role in delaying or inhibiting the deterioration of the oxygen redox reversibility. These findings constitute a step forward in unraveling the correlation between the local structural evolution and the reversibility of solid-state oxygen electrochemistry, and provide a guidance for developing oxygen-redox layered electrode materials.

Despite the disputes regarding the origin of the lattice oxygen redox, a clear consensus has been reached that a speci c oxygen 2p state, which is generated by the TM-de cient environment around oxygen in Arich layered materials (A: Li, Na, Mg, and vacancy), plays a pivotal role in electrochemically activating the lattice oxygen. [12][13][14][15] In principle, such an ionic state is similar to a non-bonding state because of the weak hybridization of the O 2p orbital through the A-O-A con guration, and thus, the extraction of electrons from the oxygen band becomes energetically feasible, allowing the oxygen redox in the conventional electrochemical operation window.
Nonetheless, we note that such a local con guration of oxygen that enables the redox activity is not likely to be always preserved during the dynamic intercalation process, and various studies have experimentally reported substantial evidences of structural transitions involving signi cant cation migrations around oxygen after electrochemical cycles. [16][17][18][19][20] Chueh and co-workers demonstrated that a TM-oxygen decoordination process could occur inducing the change in the oxygen coordination, and discussed the possible effect of cation migrations on the oxygen redox in Li 1 21,22 In this case, the local coordination of oxygen undergoes noticeable changes via π or σ interactions with neighboring TM or oxygen. The substantial structural rearrangements were also witnessed in various kinetic studies of the anionic redox reaction. 18,23,24 It was emphasized that the anionic redox is accompanied by the local structural changes such as O-O dimerization or TM migration, causing the voltage hysteresis and sluggish kinetics simultaneously. 23,24 In addition, such massive TM migration was recently alleged to cause sluggish lithium diffusion rate during the electrochemical cycling. 18 These series of ndings indicate that the initial condition of the oxygen redox with respect to the coordination is not constantly maintained, and imply that the reversibility of the oxygen redox can be easily jeopardized. Although numerous attempts have been made to elucidate the general nature of the oxygen redox 15-17, 21,25-28 , the clear-cut explanation on the interplay among the structural stability/transition kinetics and the reversibility/kinetics of oxygen redox has yet to be established particularly for the 3d-metal-based layered oxides of practical interest.
Noteworthy is that recent works identi ed series of layered oxides that could maintain the structural integrity remarkably well with the reversible anionic redox (e.g. P-type layered structures). [29][30][31] It was proposed that in the P-type structure 32 , unlike conventional lithium layered oxides, TM ions do not migrate to the interlayer space because of the instability of TM ions in the prismatic sites in the layer, thus the structural change is effectively minimized. It implies that these layered materials can serve as an ideal platform to inspect the intrinsic oxygen-redox chemistry excluding the effect of TM migrations, or, in other words, to systematically investigate the effect of TM migrations, if can be intentionally induced, on the lattice oxygen redox. In this work, we selected two model P-type layered sodium oxides; one that has been previously reported to be structurally stable with the anionic redox, Na 0.6 (Li 0.2 Mn 0.8 )O 2 (NLMO), and the other that undergoes a noticeable structural evolution with the anionic redox, Na 0.6 (Li 0.2 Ti 0.2 Mn 0.6 )O 2 (NLTMO), while both being in the same P-type layered structure with the identical transition metal/lithium ordering (ribbon-like ordering) in the lithium-containing TM layer. We show that the substantially distinct evolutions in the oxygen redox activity and reversibility are observed for the two samples during the cycles, and reveal that the evidently different behaviors arise from the two dissimilar TM-migration mechanisms; (i) rapid out-of-plane disordering (NLTMO) and (ii) slowly-occurring in-plane disordering (NLMO). It is further veri ed that the TM-migration modes are closely coupled with the change in the oxygen stabilization mechanism from the π stabilization to the σ-interacting oxygens that trigger the formation of O-O dimers and the structural destabilization over cycles. The correlation among the distinct TM migration path, the formation of the O-O dimers and the oxygen redox stability will be extensively discussed in the comparative analysis of the two model materials.
Redox activity of the lattice oxygen in layered structures NLMO and NLTMO both adopt a P3-type layered structure with the alternative stackings of sodium layer and TM layer (partially containing lithium) 33 , as schematically illustrated in Fig. 1a. Because of the unique 1:4 ratio between the lithium and TM ions, the ribbon-like ordering is present in the TM layer, which is con rmed by the superstructure peaks observed for both materials in high-resolution powder diffraction (HRPD) measurements (see dotted boxes and the crystal structure in Fig. 1a and Supplementary Fig. 1-2). 34,35 The initial charging of the two electrodes reveals that the lattice oxygen is the main redox reservoir in the electrochemical system, while the oxidation states of Mn and Ti remain + 4 ( Supplementary Fig. 3). 36 Scanning transmission X-ray microscopy (STXM) in Fig. 1b clearly visualizes that, after the charge up to 4.5 V of both materials, a new peak at ~ 531 eV evolves (highlighted region in the gure), providing a solid evidence for the oxidation of the lattice oxygen. 37,38 Despite these similarities in the structure and the initial oxygen oxidation, the electrochemical behavior during the discharge process differs markedly for the two electrodes as presented in Fig. 1c (see Supplementary  Fig. 4 together). The high-voltage plateau that was observed for the both electrodes during charge is maintained during discharge only for the NLMO electrode, implying the reversible oxygen redox reaction, whereas it irreversibly transforms into a slanted curve in the NLTMO electrode, exhibiting a large voltage hysteresis of the oxygen redox (see more details in Supplementary Fig. 5). We also found that the change in the oxygen redox makes the kinetic properties signi cantly altered. Figure 1d and e comparatively illustrate that the oxygen redox in the NLTMO displays notably inferior kinetic performance to that of the NLMO after the initial charging process. While the NLMO electrode retains a respectable rate capability with increasing current densities, the discharge capacities of the NLTMO electrode rapidly decrease presenting increasingly large polarizations with higher current rates. These results suggest that the change in the oxygen redox activity is also accompanied by the kinetic performance degradation of the electrode, in accordance with previous reports. 23,24 Interestingly, we observed that such an irreversible transition also appears gradually in the charge/discharge pro le of the NLMO electrode with the extended cycle numbers, in contrast to the previous ndings. 29,34 Fig. 1f displays how the electrochemical pro le of the NLMO electrode evolves with the cycle number, which systematically changes from the at plateau to the slanted curve that is similar to that of the NLTMO electrode. Eventually, the two pro les of the NLMO and NLTMO electrodes become nearly indistinguishable with the similar average voltage after 30 cycles, as plotted together in Fig. 1g. It implies that the nature of the oxygen redox becomes comparable for the two electrodes after su cient cycle numbers in contrast to the previous belief regarding the stable oxygen redox activity in the NLMO. 29,34 It also infers that the dissimilar behaviors of the oxygen redox in the NLMO and NLTMO electrodes may be simply rooted from the different rates of the transition or degradation that occurs in the two.

Structural evolutions with distinct TM migration paths
In order to elucidate the changes in the oxygen redox reactions, we comparatively examined the structural evolutions of the NLMO and NLTMO electrodes as a function of the cycles. Figure 2a and b present exsitu HRPD patterns of the NLMO and NLTMO electrodes, respectively, during the rst cycle in which each exhibited the similar charge behavior followed by the dissimilar discharge pro le. Figure 2a depicts that the NLMO electrode undergoes a notably reversible phase transition during the initial cycle, where the (003) peak at 2 = 15.7° is reversibly recovered, and the ribbon-like ordering with the superstructure patterns (red shaded) was well preserved after the cycle. A closer inspection of the pattern discloses that the (003) peak at 2 = 15.7° gradually weakens and a new peak appears at 2 = 17.3° (denoted with asterisk) with charging, which is characteristic of a two-phase reaction, in accordance with the previous report. 29 Both phases could be indexed as the P3-type layered structure with the R3m space group, implying that the layers maintain the prismatic sodium sites even in the desodiated state. 29 In contrast, the patterns of the NLTMO electrode in Fig. 2b illustrate that the (003) diffraction peak gradually shifts to a higher angle of 17.9° without an emergence of a new peak during charge, which is reversed during the discharge. It was previously discussed that the de-/sodiation of the Ti or Fe-doped P-type sodium layered materials involves the reversible stacking change, where prismatic sites glide into the octahedral sites, leading to the continuous formation of O-P hybrid layers. 36,39 It suggests that the gradual peak shifts of the NLTMO electrode is attributable to the solid-solution-like behavior arising from the formation of O-P hybrid layered structures upon desodiation. Nevertheless, we noticed that the diffraction patterns of the discharged NLTMO electrode did not fully match with those of the pristine electrode. Speci cally, the (003) diffraction peak exhibited a signi cant peak broadening after a cycle of charge and discharge, showing that the cycled NLTMO is in a highly strained state (see more discussion in Supplementary  Fig. 6). More importantly, the ribbon-like ordering of NLTMO in TM layer was lost and converted into a honeycomb-like ordering at 2 = 22.8° (blue shaded) after the rst cycle. 40 It differs from the case of the NLMO as comparatively shown in the magni ed views of Fig. 2a and b, which implies that a substantial TM migration has occurred during the dynamic intercalation process of the NLTMO electrode. It is also noteworthy that it is in contrast to the previous nding that the ribbon-like ordering in TM layer generally prohibits the migration of TMs in the structure, aiding in the reversible oxygen redox. 35 Raman spectroscopy analysis was further performed to verify the TM migrations and the corresponding TM-O bonding characters of the NLMO and NLTMO electrodes in Fig. 2c and e, respectively. It has been θ θ θ θ widely known that the peak at 636.5 cm − 1 is a signature of symmetric stretching (A 1g ) of the TM-O bond in layered structures, whereas the peak at 580 cm − 1 arises from asymmetric stretching (E 2g ) of the Na-O bond. 41 Fig. 2c depicts that the A 1g TM-O vibration mode denoted with circle is preserved without any additional peaks for the rst cycle of the NLMO electrode, which is indicative of the unaltered TM-O nature, while the Na-O E 2g peak changes its intensity simply with de-/sodiation. Notably, this nding contrasts to the case of the NLTMO in Fig. 2e, where a new shoulder peak at 688 cm − 1 (denoted with rhombus) appears and disappears reversibly during the successive charge and discharge. According to previous studies, the new peak with a higher wavenumber of TM-O vibrations is attributed to the substantial out-of-plane cation disorder with a new TM coordinating environment. 20,42 It suggests that the charge/discharge of the NLTMO electrodes involves the out-of-plane migration of TMs in the P-type layered structure, which is not observed in the similarly structured NLMO electrode.
The presence of the out-of-plane TM migrations with the electrochemical cycle could be more closely investigated by comparatively conducting spherical-aberration-corrected scanning transmission electron microscopy (Cs-STEM) analysis on the NLMO and NLTMO samples. Figure 2d and f present the highangle annular dark-eld (HAADF) images of the samples in the fully charged states exploiting the Zcontrast imaging. 43,44 For the desodiated NLMO ( Fig. 2d), alternating bright TM layers and dark Na layers are clearly observed with an interlayer spacing of 5.55 Å, con rming that the spacing of the prismatic sites is well preserved. 45,46 It supports the observation from the Raman spectroscopy that the out-ofplane cation disorder does not take place in the NLMO electrode after the initial charging. On the other hand, a noticeable amount of TM ions was found to occupy the vacant Na sites in the desodiated NLTMO (denoted with the arrows in Fig. 2f). Moreover, some of the interlayer spacings were estimated to be 4.61 Å, which is markedly smaller than 5.55 Å and corresponds to the typical values observed for the Otype layers with the octahedral environments. 26,47 It indicates that the charged state of the NLTMO adopts the O-P hybrid layered structures 36 , where the occupancy of TM in the Na layer is likely to induce the O-type layers. We also speculate that the loss of the ribbon-like ordering in TM layer observed in Fig. 2b is partly attributable to the substantial out-of-plane migrations of the TM ions. The extra vacancies in the TM layers owing to the out-of-plane migrations would provide su cient vacant sites for the in-plane cation (Li/TM) redistributions upon the initial charging, causing the loss of the initial ribbonlike orderings and the formation of the honeycomb-like orderings.
Effect of TM migrations on the oxygen redox Inspired by the distinct behaviors of TM migrations observed for the two electrodes, we attempted to gain further insights in the in uence of the TM migrations on the oxygen redox by the theoretical considerations of the electronic and magnetic structures. When the lattice oxygen gets oxidized, the holes are created, yielding the unpaired electrons in the orphaned oxygen 2p states. The change in the electronic/magnetic structure can subsequently occur via various types of interactions with neighboring oxygens and TMs, which involves dynamic electron transfer. 22 We deduce that the oxidized oxygen in lattice can be in a state of (i) non-stabilized oxygen, (ii) π-stabilized oxygen, or (iii) σ-stabilized oxygen, which would accompany localized holes, delocalized holes by TM-O coupling, or delocalized holes by O-O dimerization, respectively, as schematically illustrated in Fig. 3a. According to previous works 14,30,48 , the π-stabilization is readily formed between the oxygen and the neighboring TM in the pristine layered structure due to the intrinsic directional coordination involving oxygen 2p and TM 3d orbitals. However, it is noteworthy that, when the out-of-plane TM migrations occur, density functional theory (DFT) calculations revealed the TM-O decoordination that is usually accompanied by the TM migration strongly promotes the σ-stabilized O-O dimerization. Figure 3b illustrates that the dangling oxygen ion is generated near the vacant TM defect sites, weakening the potential π-stabilization between the oxygen and TMs. Moreover, it was found that short oxygen dimers are subsequently induced with the lengths of 1.35 and 1.42 Å in the vicinity of the out-of-plane TM migration. Considering the signi cant TM migration observed in the desodiated NLTMO electrode, a substantial number of single-or even non-coordinated oxygens are expected to evolve during charging, which would likely form the short oxygen dimers with the adjacent oxygen ions for the stabilization.
The change in the oxygen nature could be further supported by the comparative magnetic measurements of the charged NLMO and NLTMO electrodes using a superconducting quantum interference device (SQUID). 49,50 Supplementary Fig. 7 shows the temperature dependence of the reciprocal magnetic susceptibility in the NLMO and NLTMO electrodes. Through Curie-Weiss tting of the susceptibility curves between 180 K and 250 K, the effective magnetic moments of two electrodes were extracted as a function of the state of charge (SOC) and plotted in Fig. 3c and d. The magnetic moments of the pristine NLMO and NLTMO electrodes were measured to be 3.94 and 3.92 µ B , respectively; these values are close to the expected value of the high-spin Mn 4+ ions (3.87 µ B ), con rming that the oxidation states of Mn ions are close to + 4 in both pristine electrodes. 51 Upon charging (i.e. oxygen oxidation), however, two electrodes exhibited the opposite behavior in the change of the effective magnetic moments with the SOCs. Noteworthy is that the magnetic moment of the NLTMO increases with two positive slopes along with the oxygen oxidation, whereas that of the NLMO decreases with a constant slope, indicative of the distinguishing nature in the electronic structure evolutions of the two electrodes with charging. It clearly manifests the contrasting oxygen stabilization mechanisms in the two systems, considering the relationship between the electronic structure expected for the possible oxygen states and the magnetic moments of the materials (see detailed explanations on the expected magnetic moments with different types of oxygen states in the caption of Fig. 3a). Moreover, the quantitative analysis of the slopes in Fig. 3c and d could offer more detailed picture of the oxygen stabilization process during the charge. Based on the slopes observed in the plot, the fraction of the aforementioned types of oxidized oxygens could be quantitatively deconvoluted for the respective regions of Region 1, Region 2 and Region 3 in Fig. 3c and d, which is depicted comparatively in Fig. 3e. Details on the deconvolution process and the calculations of the fractions are provided in Supplementary Note 1. It reveals that the most prominent interaction in Region 1 is the π-type interaction, which accounts for approximately 62.6% of the oxidized oxygen, suggesting that the majority of oxygen holes in the charged NLMO electrode are likely to be delocalized with the neighboring Mn ions. On the other hand, oxygen holes in the charged NLTMO electrode tend to be delocalized through the σ-interaction of O-O pairing, as indicated with the Region 2 and 3 in Fig. 3e. In particular, during the later charging process of the NLTMO electrode, the fraction of the σ-stabilized oxygen is as high as 95.7%, signifying that almost all the oxygen holes were delocalized through the O-O pairing. This quantitative result is remarkably consistent with our theoretical calculations, which suggested that the out-of-plane TM migration in the NLTMO induces the subsequent oxygen dimerization between the adjacent lattice oxygens. The results here imply the close correlation between the out-of-plane TM migration and the stabilization mechanism of the oxidized lattice oxygen in the layered materials.
We further veri ed that the stabilized lattice oxygen displays the redox potential that is markedly distinct from that of the pristine non-stabilized oxygen. The partial density of states (pDOS) of oxygen 2p orbitals in Fig. 3f shows that the short oxygen dimerization occurring with the TM migration signi cantly redistributes the electronic structure of the oxygen owing to the strong σ-type O-O interaction. The pristine non-bonding O 2p states are reorganized into sharp and discrete states corresponding to σ-type O-O bond, thereby lifting the overall oxygen states from below to far above the Fermi level. Considering that the shifted O-O states are located at far higher energy states than the minimum of empty (Mn-O) e g * states, as highlighted in the DOS of Fig. 3g, a partial manganese reduction prior to complete oxygen reduction is expected upon discharge process. Notably, this speculation could be experimentally con rmed through our X-ray absorption near edge spectroscopy (XANES) analysis. Figure 3h and i present the Mn K-edge absorbance spectra of the NLMO and NLTMO electrodes, respectively, during the successive charge and discharge process. It shows that the Mn edge position of the NLTMO is maintained initially, however, upon discharge to 2.0 V, shifts toward a lower energy close to that of the Mn 3 + 2 O 3 reference, accounting for the partial reduction of Mn 4+ to Mn 3+ in the low-voltage region, while that of the NLMO does not shift from the initial oxidation state of Mn 4+ over the entire voltage range. It is a remarkable agreement with the results of the DOS in Fig. 3g, and is attributed to the oxygen stabilization mechanism in the presence of the TM migration, i.e. O-O pairing. It is supposed that the reversed electronic states of the fully charged states of NLTMO eventually induces the detour of the thermodynamic path in the discharging process, triggering the voltage hysteresis. Moreover, since the dimer formation entails structural rearrangements involving TM migrations, the kinetic properties would become comparatively sluggish, elucidating the inferior rate performance of the NLTMO electrode in Fig. 1e. It suggests that the oxygen stabilization mechanism has a crucial in uence on the thermodynamic and kinetic properties of the oxygen redox in the layered materials.
Gradual formation of oxygen dimers in LNMO and its direct visualization Despite the seemingly intact structural stability, the NLMO electrode also eventually exhibited the degradation of the electrochemical pro le with the voltage fades as depicted in Fig. 1f and g. In this regard, we carefully re-examined the structural evolution of the NLMO electrode over the extended cycles. Figure 4a presents ex-situ HRPD patterns and Raman spectra of the 1, 10 and 30 cycled NLMO electrodes at the fully charged states, where signi cant out-of-plane TM migrations have been observed in the NLTMO counterpart. It shows that the characteristic two phase reaction of the NLMO is consistently retained, clearly exhibiting the two (003) peaks at 2 = 15.7 and 17.3° at the charged state, over the 30 θ cycles, which differs from the O-P hybrid structure formation of the NLTMO involving the signi cant TM migrations. Moreover, the signature of the out-of-plane TM migrations such as a shoulder peak at 688 cm −1 does not appear even after the 30 times of the charging in the Raman spectra. Cs-STEM characterization in Supplementary Fig. 12 also supported that the P-type layered structure was well preserved for the NLMO electrode even after 30 cycles, implying the absence of the out-of-plane TM migration either at charged or discharged states for the 30 cycles. Nevertheless, we observed that the (003) peak intensities notably decrease with the extended cycling from the HRPD patterns, exhibiting the gradual loss of the crystallinity. Supplementary Fig. 13 illustrates that the full width at half maximum (FWHM) of the 30-cycled electrode is noticeably greater than that of the initial electrode. More importantly, it was found that the superstructure peaks corresponding to ribbon-like ordering in TM layer completely disappeared after 30 cycles, as shown in Fig. 4b. These results suggest that substantial TM migrations other than the out-of-plane migration have occurred and caused the partial structural inhomogeneity, strongly implying the presence of the in-plane TM migrations in the TM layer. It is also noteworthy that the lithium contents in the TM layer progressively decreased upon cycling, as presented in Fig. 4c (see more details in Supplementary Tables 2-3). It indicates that su cient vacant sites in the TM layer were created especially at the later cycles, which would facilitate the in-plane TM migration and the subsequent disordering of the TM layer. This speculation could be supported by the additional experiments provided in Supplementary Note 2, in which we found that in-plane migration is critically dependent on the lithium/vacancy (Li/V) concentrations in the TM layer.
To further probe the TM layer rearrangements in the NLMO, the extended X-ray absorption ne structure (EXAFS) analysis of the 1-, 10-, 20-, and 30-cycled electrodes was additionally conducted ( Supplementary  Fig. 19). Characteristic peaks of the rst and second shells in the Mn K-edge data, which correspond to the closest manganese-oxygen (Mn-O 6 ) and manganese-metal (Mn-M 6 ; M: Li or Mn in the same layers) bonds, respectively, were tted with the P3-type model structure (Supplementary Table 5). Figure 4d depicts the tted values of the Debye-Waller factor, which is indicative of the structural disorder, for every 10 cycles. The results indicate that the Debye-Waller factor of the Mn-M 6 shell continuously increases with cycling, supporting that the in-plane disordering in the TM layer takes place gradually. The factor of the Mn-O 6 shell, on the other hand, remained nearly constant for the electrochemical cycling. It signi es that the overall P-type layered structure is relatively well preserved over the cycling, in consistent with the HRPD results, even when the substantial disordering occurs within the TM layer. The gradual in-plane disordering in the TM layer was found to accompany a noticeable rearrangement of the electronic structure of the NLMO, displaying a continuous shift of the Mn K-edge spectra in Fig. 4e. The apparent reduction of the Mn in the NLMO at later cycles is reminiscent of the case of the NLTMO wherein the oxidized oxygens were stabilized by O-O pairing, causing the reversed electronic states at the fully charged states as elucidated in Fig. 3i and g of NLTMO. It implies that the inplane anti-site defects in the TM layer has also induced the gradual oxygen dimerization in the NLMO, as will be discussed further in the following.
The clear evidence of the oxygen dimers could be obtained from the annular bright-eld (ABF) STEM images taken for the charged NLMO electrodes after 30 cycles, in comparisons with that at the rst cycle as presented in Fig. 4f and g. 52,53 Along the zone axis, alternating MnO 6 layers and vacant Na layers could be visibly projected. The ABF image for the rst charged electrode in Fig. 4f reveals a highly ordered structure of the NLMO, where the manganese and oxygen atomic features are evidently spotted and well aligned along the axis. In contrast, the charged electrode after 30 cycles in Fig. 4g manifests that the oxygen signals around the TM columns are relatively weaker in the intensity than those in the rst charged sample, indicating that the oxygen sites are signi cantly distorted. According to our ndings and the previous observations of the lattice distortion of oxygen-redox materials 26,54 , it suggests that various oxygen bonding environments have been generated after the 30 cycles. The lattice distortion over the oxygen regions is more obviously con rmed by the signal pro les along the horizontal direction in the inset of Fig. 4f and g, displaying additional oxygen peaks for the sample after 30 cycles. Contrary to the charged NLMO at the rst cycle, where the single oxygen peak is clearly observed, the oxygen peaks could be indexed apparently by two Gaussian functions, indicating that the distorted signals can be attributed to the two oxygen atoms with the unusually short distance, i.e. oxygen dimers.
To further assign the features of the oxygen dimers, we carried out Raman analysis to estimate the bonding lengths between the oxygens from the vibrational frequency. 55.56 While the direct measurement of the short oxygen bond length (< 1.34 Å) has not yet been achieved for battery electrodes, to the best of our knowledge, previous Raman studies on a variety of metal complexes containing O-O collectively suggested that the presence of the short oxygen dimers display the characteristic vibration frequency in the range of 1100 and 1500 cm − 1 . 55 It is also noteworthy that its detection has been generally ambiguous because the strong D and G bands at 1350 and 1587 cm − 1 of the carbon additives in the electrode screen the signals from short oxygen dimers. 57 In this respect, we measured the electrode without carbon effects (see Supplementary Note 3 for details) and performed the analysis as presented in assists the oxygen dimerization upon repeated cycling, as supported by our further theoretical calculations (Supplementary Note 4).
Overall, the characterization of the NLMO electrode unequivocally con rms that the gradual transition of the oxygen chemistry is caused by the progressively occurring in-plane TM movements, while the rapid out-of-plane TM migration dominated the structural evolution of the NLMTO electrode. As discussed in Fig. 4c, the in-plane TM migration is coupled with the Li/V concentration and promoted by the su cient number of vacant sites in TM layer. Since the electrochemical cycles of the NLMO electrode would inevitably involve the partial extraction of lithium ions from the structure at every cycle, it explains why the in-plane TM migration is accelerated with the cycle number. It also implies that the Li substitution itself in the TM layer, which is essential to form the non-bonding state for the oxygen redox activity, can exhibit the adverse effects to some extent on the reversibility of the oxygen redox reaction, thus the tradeoff relationship needs to be carefully considered. The contrasting structural and electrochemical behaviors of the two model systems unveil how different cation disordering mechanisms are associated with the oxygen network reorganization and the corresponding evolution of the oxygen redox activity. Our ndings thus establish a clearer picture of the structure-property relationships for the solid-state oxygen electrochemistry in the layered transition metal oxide electrodes.
In summary, we demonstrated the strong interrelationship between the lattice oxygen redox and local structural change during the dynamic intercalation process in TM-oxide-based electrodes by comparing two analogous materials with substantially different oxygen redox activity and stability resulting from distinct structural behaviors. It was found that the irreversible electrochemistry of lattice oxygen in the NLMO and NLTMO electrodes exclusively originates from distinct TM migration paths (i.e. in-plane and out-of-plane cation migration, respectively), as con rmed by HRPD, STEM, Raman spectroscopy, and complementary EXAFS measurements. Such cation migration eventually leads to the formation of short O-O dimers (< 1.34 Å), thereby stabilizing the oxygen network through rearrangement of the electronic structures. Two types of stabilization mechanisms for the oxidized oxygen and their relevance to the electrochemical behaviors were elucidated through the theoretical investigations of the electronic and magnetic structures. Furthermore, because of the structural simplicity of NLMO even in the disordered state, the O-O dimers were directly observed using ABF-STEM, and their bonding lengths with TM-(O-O) con gurations were concomitantly identi ed using Raman spectroscopy. This work provides a clearer view of oxygen-redox chemistry in relation with the local structure/transition, and hints at robust design guidelines for resolving the issues of voltage hysteresis and sluggish kinetics in oxygen-redox materials.
Our ndings indicate that regulating the TM migration and its pathway is vital for tailoring the oxygen stabilization mechanism and subsequently achieving highly stable and reversible anionic redox chemistry.
Electrochemistry. A slurry of 80 wt% active materials, 10 wt% carbon black (Super P), and 10 wt% polyvinylidene uoride dissolved in N-methyl-2-pyrrolidone (NMP; 99.5%, Sigma-Aldrich) was cast onto aluminum foil to fabricate the electrodes. The mixture was dried in a 70 °C vacuum oven overnight to evaporate the NMP solution. Coin cells (CR2032, Hohsen) and Swagelok-type cells were assembled using the electrodes, a sodium counter electrode, a separator (GF/F lter, Whatman), and a 1 M solution of Raman spectroscopy. Raman spectra were recorded using a Raman spectrometer (LabRAM HV Evolution, HORIBA) with an Ar laser as the excitation light source ( =532 nm). The scattered light of the Raman signal was collected in a backscattering geometry using a 50× microscope objective lens. The data were collected using an acquisition time of 15 s with 10 accumulations. The spectra were deconvoluted using the XPS Peak program.
Cs-STEM. Cross-sectional TEM specimens of the electrode slurry lms were prepared using focused ion beam (FIB) milling (Helios 650, FEI). The prepared specimens were used for HAADF and ABF imaging under 200 keV using aberration-corrected STEM (JEM-ARM200F, JEOL) with a point-to-point resolution of 0.08 nm at the National Center for Inter-university Research Facilities (NCIRF) at Seoul National University.
SQUID. For the magnetic characterization, a powder cell using a Swagelok-type cell with an 8:2 ratio of active materials to carbon black was tested. The magnetic properties of the collected powder samples were recorded using a SQUID magnetometer (MPMS-VSM, Quantum Design) at the Korea Basic Science Institute (KBSI). The magnetic moments were measured using a zero-eld cooling process from 300 K to 5 K. The eld was xed at 0.1 T during the process.
DEMS. The real-time gas evolution was detected and analyzed with a mass spectrometer (MS; HPR-20, Hiden Analytical) during the cell testing. A controlled amount of Ar ow swept the evolved gas from the Swagelok-type cell to the MS chamber. During cell operation, the pressure inside the MS chamber was maintained at 9.9×10 −7 Torr. XANES. XANES spectra of the harvested electrodes were obtained at beamline 7D of the PLS using a double-crystal monochromator containing two sets of Si(111) crystals. All the measurements were performed at room temperature, and the Ti and Mn K-edge spectra were collected in transmission mode. To accurately calibrate the energy scale and any drift of the monochromator position, metal foils were placed in a third chamber as a reference. All of the spectra were normalized and compared using the Athena and Artemis programs.
STXM. STXM analysis was performed at beamline 10A of the PLS to obtain the O K-edge spectra. Primary particles were drop-cast onto carbon-coated Cu TEM grids for the measurements. By maintaining the focal position at the same particle, the two-dimensional transmitted photon intensity was recorded in pixel form at a xed energy. To obtain image stacks, the same measurements were repeated over different X-ray energy ranges. The image stacks were acquired in 0.2-eV steps with a 2-ms dwell time and were aligned using the aXis 2000 software package.
First-principles calculations. All the ab initio calculations were spin-polarized based on density functional theory (DFT), as implemented in the Vienna ab initio simulation package (VASP). 61 Structural relaxations were conducted within the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation. 62 The rotationally invariant Dudarev method was applied, and an effective Hubbard-U parameter (3.9 for 3d electrons of Mn) was selected based on a previous report. 63 All the structures were fully relaxed until the interatomic forces converged to less than 0.02 eV Å -1 . A plane-wave cutoff of 520 eV was applied along with a well-converged k-point grid for each system. Detailed information on the model construction and TM migration analysis is provided in Supplementary Note 5.