Disordered Rocksalt Structural Properties

Li₂MnO₂F was prepared by mechanochemical ball-milling, and the oxidation state of Mn was subsequently confirmed as +3.00(5) from iodometric titration. Li₂MnO₂F possesses a cubic rock-salt structure where each cation (Li⁺ or Mn³⁺) is octahedrally coordinated to six anions (O^2– or F^–) and vice versa (illustrated in Figure 2). To investigate the possibility of local ordering, Mn K-edge extended X-ray absorption fine structure (EXAFS) and neutron pair distribution function (PDF) analysis were performed on pristine Li₂MnO₂F (the experimental methods are detailed in the Supporting Information (SI), Section S1).

Figure 2

Structure of Li₂MnO₂F. (a) Mn K-edge EXAFS for pristine Li₂MnO₂F compared with a MnO reference, each with cubic rocksalt crystal structure. The spectra are normalized to the area under the first peak corresponding to the occupancy of the first nearest neighbor anion site which is the same for both, 6. A good fit of the EXAFS data can also be obtained with a rocksalt model with 4 × 2nd nearest neighbor Mn. (b) Neutron PDF data fitted to structural models of distortion-free random cubic rocksalt and the Monte Carlo derived model for pristine Li₂MnO₂F. Li, Mn, O, and F atoms are indicated by green, purple, red, and gray spheres, respectively. There is very good agreement between the models and the PDF data showing Li₂MnO₂F exhibits a close to completely disordered rocksalt structure. The slight asymmetry of the first peak at around 1.9 Å may indicate some element-specific preference for shorter bond length. Refined cell parameters a = 4.117 Å and a = 12.152 Å, b = c = 8.336 Å, and U_iso values 0.027 and 0.014, respectively.

The EXAFS technique is an element-specific probe of local coordination environment, in this case around Mn, giving a plot of the nearest-neighbor (NN) atoms as a function of distance from the central atom. As shown in Figure 2a, the first two peaks in the EXAFS spectrum for Li₂MnO₂F are a close match in shape and relative position to that of MnO, which has a well-defined cubic rocksalt structure. These peaks correspond to the first and second NN shells of atoms, anions and cations, respectively. Normalizing each spectrum by the first peak area, since O^2– and F^– are indistinguishable with EXAFS, permits direct comparison of the second peak area. Here, Mn is a much stronger scatterer of the photoelectron wave than Li, which is a very weak scatterer, so Mn dominates the second peak intensity. The difference in peak area between the two materials indicates a much lower amount of second NN Mn for Li₂MnO₂F than MnO, consistent with the presence of Li on the cation sites in the former. Measuring the peak intensity relative to a baseline of zero scattering from 12 Li and maximum scattering from 12 Mn (as in MnO), shows an average of 3.6 Mn as second NN for Li₂MnO₂F. This is in line with that expected from a completely random distribution of Li and Mn in a 2:1 ratio, i.e., four Mn. As further confirmation, a good fit was obtained of the EXAFS data using a rocksalt structure with six degenerate first NN O and four degenerate second NN Mn atoms, Figure 2a and Table S1.

PDF is another powerful tool for probing local structure, giving a superposed plot of all atom–atom pairs throughout the structure resolved as a function of increasing separation. Unlike EXAFS, neutron PDF can probe over much longer correlation lengths and is much more sensitive to Li, since neutrons are more strongly scattered by Li than X-rays. The fitted PDF data (Figure 2b), show that the local structure of Li₂MnO₂F can be well-described by a disordered rocksalt model indicating minimal short-range order. Together with the EXAFS data, the experimental evidence supports a close to completely disordered rocksalt structure for Li₂MnO₂F.

To obtain a computationally tractable structural model for Li₂MnO₂F capturing this disorder, a Monte Carlo random sampling approach was employed to generate a 3 × 2 × 2 unit cell which possessed a representative distribution of different sites (computational methods applied to battery cathode materials are well established^9,11,53,54 and detailed in the SI, Section S1). The validity of this Monte Carlo-derived model was checked by fitting to the neutron PDF data. The quality of the fit was even better than the distortion-free one showing it is a closer match to the experimentally observed structure. Furthermore, the calculated mean lattice parameter, 4.146 Å, compare well with the experimental value (a = b = c = 4.118 Å) from X-ray diffraction studies.³⁶ Our ab initio simulations confirm that the disordered rocksalt structure of Li₂MnO₂F does not exhibit the cooperative Jahn–Teller distortion usually associated with Mn³⁺ in ordered structures, which often leads to poor cycling. The full structural data set for the pristine Li₂MnO₂F computational model is given in the SI, section S2.

Charge-compensation on Lithium Ion Extraction

To investigate the redox processes occurring over the first cycle in Li₂MnO₂F, operando Mn K-edge XANES was performed. Operando experiments allow the intercalation reaction in the cathode to be followed under operating conditions, eliminating the effect of any relaxation phenomena. As shown in Figure 3, a continuous shift in the Mn K-edge is observed during charge and discharge in line with the expected oxidation and reduction from Mn³⁺ toward Mn⁴⁺. Near the top of charge, around x ≈ 1.2 in Li_xMnO₂F, the changes become less pronounced as oxygen oxidation starts to dominate the redox process; this corresponds to a slight inflection in the electrochemical load curve. Note that previous operando electrochemical mass spectrometry studies³⁶ indicate that there is negligible oxygen loss. Given that the edge continues to evolve and the voltage profile remains sloped throughout this region (Figure 1), there must be a significant degree of overlap between the Mn and O redox couples. Close analysis of the Mn K-edge pre-edge (Figure 3d), which is generally considered to be a better measure of oxidation state than the main edge,⁵⁵ reveals a similar trend.

Figure 3

Operando Mn K-edge XANES and EXAFS on Li₂MnO₂F. (a) Load curve for the cell charged at a rate of 50 mA g^–1 between voltage limits of 2 and 4.8 V vs Li⁺/Li. The XANES spectra (b) and (c) show a continuous shift in edge energy as Mn is oxidized which slows toward the top of charge as O oxidation starts to dominate. There is no substantial increase in pre-edge intensity (inset) which would be expected from tetrahedral Mn⁵⁺ or Mn⁷⁺. The very slight increase that is observed is characteristic of slight distortions to the octahedra which allows mixing between the Mn 3d and 4p states. (d) Variation in energy of the weighted center of the pre-edge. (e) and (f) EXAFS data for charge and discharge, respectively. The shape of the first two EXAFS peaks showing the octahedral geometry is maintained throughout the first cycle.

The pre-edge shape (insets of Figure 3b,c) does not appear to change much, but there is evidence of a slight increase and decrease in intensity of the twin peaks. These peaks arise from the quadrupole-allowed transition from the Mn 1s to the Mn 3d states, which are subdivided by crystal field splitting, and are weak due to the centro-symmetry of octahedral coordination. The intensity gain can be attributed to a slight distortion of this centro-symmetry allowing mixing between the Mn 3d and 4p states. In contrast, pre-edge features for tetrahedral geometries tend to be of significantly larger intensity, often of comparable height to the main edge, due to the complete lack of centro-symmetry.⁵⁶ The absence of a substantial increase in pre-edge intensity here means that the presence of Mn⁵⁺ or Mn⁷⁺, each of which are only known to occupy tetrahedral coordination environments, can be ruled out in Li₂MnO₂F. It has been recently suggested that tetrahedral Mn⁷⁺ forms as an intermediate state in Li-rich manganese oxides before O–O dimerization.⁵⁷ Our operando data, which should capture intermediate species formed during battery operation, does not suggest this is the case here for Li₂MnO₂F. Using X-ray spectroscopy techniques, Rana et al.⁵⁸ also rule out oxidation beyond Mn⁴⁺ in Li-rich Mn-oxides.

Interestingly, recent reports for the vanadium-based disordered rocksalt systems, Li_1.25Nb_0.25V_0.5O₂³⁴ and Li₂VO₂F,^59,60 show XANES data indicating a strong increase in pre-edge intensity, characteristic of vanadium in a noncentrosymmetric coordination environment such as tetrahedral V⁵⁺. Baur et al.⁴⁷ recently confirmed the presence of tetrahedral vanadium by PDF in Li₂VO₂F, and Chang et al.⁴⁸ report superoxide formation in this oxyfluoride from computational and EPR studies. We note that Lun et al.⁴¹ studied the Li–Mn–O−F chemical space to derive a capacity map of Li percolation and redox properties. In the context of Li-rich oxide structures, Hong et al.¹⁹ report that oxygen redox may be stabilized in the local coordination environments created through cation vacancies, and Gent et al.²⁵ find that the defect formation energy landscape is a key factor controlling the electrochemical reversibility of high valent redox. Our previous simulation study⁹ on Li-rich layered Li₂MnO₃ suggests that delithiation leads to oxygen dimerization and eventually to the formation of molecular O₂.

To further probe the changes in local environment around the Mn during the first cycle, the Mn K-edge EXAFS were analyzed. As shown in Figure 3e,f, the first and second neighbor atoms to Mn do not change as a function of state of charge. This clearly shows Mn remains octahedrally coordinated throughout the charge/discharge cycle, with no evidence of tetrahedral Mn.

To complement our XANES and EXAFS work, we used DFT methods to examine local structures and to quantify the redox chemistry on lithium extraction from Li_xMnO₂F. As in previous studies,^9,11 we stress that high level hybrid functionals were employed as they are found to be important in reproducing accurately the electronic structure of oxygen states (further details in the SI, S1 Methods). Figure 4a illustrates the overall contribution of Mn vs O redox as a function of Li content in Li_xMnO₂F derived from the ab initio calculations (Figure S2 shows the change in the average oxidation states of all the component elements of Li_xMnO₂F as Li is removed).

Figure 4

Manganese/oxygen redox activity and impact of local coordination environments in Li_xMnO₂F. (a) Relative contribution (%) of the Mn versus O redox processes to the overall redox activity at a given Li content (x). The data show significant overlap between Mn and O redox couples. (b) Oxidation potentials of Mn and O as a function of their coordination environment in Li₂MnO₂F; here, the Li content (x) is related to the experimentally measured voltage at that state of charge derived from Figure 1. OLi₃Mn₃ is omitted for clarity as these O atoms did not show significant oxidation during Li extraction. (c) Change in the average oxidation state of octahedral Mn atoms in Li₂MnO₂F that are coordinated by three or more F atoms and those that are coordinated by fewer than three F atoms. (d) Change in the average oxidation state of O atoms with coordination environment in Li₂MnO₂F.

The results clearly show significant overlap between Mn and O redox couples with O redox activity starting from about x ≈ 1.5, well before all of the Mn has been oxidized and in accord with the XANES results; this is attributed to the disordered structure with a range of local ion environments (which we return to below). This behavior contrasts with the redox activity found in layered Li-rich ordered oxides in which there is no significant overlap between Mn and O states.^11,14,45 The projected density of states for Li_xMnO₂F (x = 2.0, 1.5, 0.75) (Figure 5) also show the strong hybridization between the Mn 3d and O 2p states, with the energy of the O-2p states raised in the charged systems, which promotes O redox activity.

Figure 5

Projected density of states (pDOS) for Li_xMnO₂F. (a) Li₂MnO₂F, (b) Li_1.5MnO₂F, and (c) Li_0.75MnO₂F. The blue and red lines correspond to the Mn 3d and O 2p pDOS, respectively. For the charged structure, Li_0.75MnO₂F, the black bands represent the electron holes localized on molecular O₂. With respect to Li₂MnO₂F, the occupied states close to the Fermi level (which is set to zero) are composed of a mix of O 2p and Mn 3d states. For the charged structures, the energy of the O-2p states are raised, which promotes O redox activity.

Regarding charge compensation on Li extraction, Mn undergoes oxidation from Mn³⁺ at x = 2.0 toward Mn⁴⁺ at x = 0.75, with no evidence of any change from octahedral to tetrahedral coordination. Our DFT structural analysis also indicate that tetrahedral Mn⁵⁺ is not formed during delithiation of Li_xMnO₂F, which agrees with the operando Mn K-edge XANES and EXAFS results. As expected, the oxidation states of Li and F do not change on delithiation. When considering the overall Li deintercalation process (x = 2.00 to 0.75 in Figure 4a), Mn redox activity accounts for approximately 75% of the total capacity, while O redox accounts for the remaining 25%.

Role of Local Environment

To gain an atomistic understanding of the overlapping nature of Mn and O redox, the local coordination environment of each atom was investigated as a function of lithium content (x). Figure 4b shows the oxidation potentials of Mn and O as a function of their coordination environment in Li₂MnO₂F; the value of x for Li_xMnO₂F in the plot is related to the experimentally measured voltage at that state of charge (from Figure 1). Figure 4c compares the calculated change in the average oxidation state of Mn atoms that are initially coordinated by three or more F atoms with those that are coordinated by fewer than three F atoms. Figure 4d compares the change in the average oxidation state of O atoms with their coordination environments.

At a given Li content, Figure 4c indicates that Mn atoms with low coordination to F (e.g., Mn(O₄F₂)) are more oxidized than those Mn atoms coordinated to three or more F atoms (e.g., Mn(O₃F₃)) and helps to explain why O redox is predicted before all the Mn are oxidized. In other words, the redox potential of the Mn^3+/4+ couple is raised by an increasing number of F in its coordination shell, shown in Figure 4b. These results suggest that the substitution of O by F leads to greater Mn redox overlap with oxygen.

With respect to the O environments, Figure 4d shows greater oxidation for O atoms with five Li nearest-neighbors (O(Li₅Mn)) at the fully lithiated state than those with four or three Li ions. This trend is consistent with that found by Seo et al.,¹¹ who showed that the anion redox chemistry in a variety of Li-intercalation oxide cathodes, including LiNiO₂ and Li₂MnO₃, is dependent on the anion nearest-neighbor coordination environment. They focused on how the presence of linear Li–O–Li configurations promote labile oxygen electrons in the adjoining 2p orbital that effectively pin them at a set energy above the bonding electrons. With the extension of this model, our results show a more continuous variation in the O-redox potential dependent on the number of Li coordinated to a given O^2– ion. We note that recent computational screening work on layered oxide cathodes⁶¹ report trends in O-redox activity associated with the electrostatic (Madelung) energy at oxygen sites. In general, our findings indicate that a greater number of coordinating Li can promote O oxidation at a lower potential (as shown in Figure 4b). This highlights the more general role of ionicity of the coordination environment around O in tuning its oxidation potential.

In summary, the effect of the local coordination environment has a major effect on the redox potentials of both Mn and O leading to increased competition between the Mn and O redox couples across the voltage range. The results presented here indicate that the O–Li environments in Li₂MnO₂F encourage O oxidation at lower potentials than the typically observed 4.6 V plateaus for layered Li-rich oxide materials. Meanwhile, the presence of Mn–F bonds increases the voltage of Mn oxidation, leading to increased overlap between Mn and O redox processes.

Trapped Molecular O₂ in the Bulk Structure

Detailed analysis of the ab initio simulated structures of Li_xMnO₂F on Li removal indicate that the possibility of dimerization of oxidized O species is heavily dependent upon the local environment. At high degrees of Li deintercalation (high states of charge) corresponding to the Li-deficient composition Li_0.75MnO₂F, some of the oxygen ions will be undercoordinated (fewer than three cations). The calculations indicate that such undercoordination results in molecular O₂ formation in the bulk structure (shown in Figure 6a); for this delithiated composition Li_0.75MnO₂F, when the cations are arranged in such a way that the cation vacancies are clustered together, O₂ molecules form during energy minimization to give a structure with a significantly lower total energy (>320 meV/formula unit) than other configurations. Hence, bulk O₂ formation is triggered by driving the oxyfluoride electrode to high degrees of Li deintercalation, such that the local coordination around oxidized O ions is reduced to below three with local Li vacancies, e.g., OMn₂□₄. For instances where the oxidized O species are bonded to three or more cations, dimerization does not occur and the O–O separation remains at 2.3–2.8 Å.

Figure 6

Structures and dynamics of molecular O₂ trapped in the bulk. (a) Calculated structures of Li₂MnO₂F and Li_0.75MnO₂F, and the local environment of molecular O₂ in the bulk of Li_0.75MnO₂F. Li, Mn, O and F are indicated by green, purple, red, and gray, respectively. The weak Mn–O₂ and Li–O₂ interactions are indicated with dotted lines. (b) Ab initio molecular dynamics simulations of the molecular O₂ in the bulk cavity at 300 K showing the variation in O–O and Mn–O separations vs simulation time. The calculated O–O distances of the trapped molecular O₂, and the nearest-neighbor Mn–O₂ distances are shown in green and blue, respectively (derived at intervals of 1 ps). (c) Mean square displacements (MSDs) vs simulation time for molecular O₂ in the bulk cavity in comparison with free gaseous O₂ molecules in a very similar nanosized volume.

The formation of molecular O₂ inside the particles is only possible due to the loss of lithium ions at high states of Li deintercalation to create vacancy clusters that accommodate the O₂ molecules. From local analysis of the simulated relaxed structure of Li_0.75MnO₂F, the diameter of the vacancy cluster is approximately 6.5 Å. The undercoordination of oxygen results in O 2p orbitals that no longer interact with any cations, and are at a high energy (shown in the pDOS plots, Figure 5c), which are where the oxygen holes are concentrated. Such undercoordinated oxygen will be unstable in the lattice and leads to molecular O₂ formation; this accords with our previous suggestions that undercoordinated oxygen more easily form O–O species,²³ and now shown explicitly in disordered rocksalt oxyfluorides.

Figure 6a illustrates the local environment of the O₂ molecule with a calculated O–O bond length of 1.23 Å, directly comparable to that of molecular O₂, 1.21 Å, and in accord with RIXS results (Figure 7a) discussed in detail below. Analysis of the calculated charge density of O₂ within this cavity (Figure S4) indicate the electron density is heavily localized on the O₂ molecule as expected from the strong, covalent O=O bonding, with only weak chemical interaction with its neighboring environment. Once O₂ is formed, some Mn octahedra around the vacancy cluster may experience greater distortion to accommodate the defect. However, the amount of O₂ that is expected to form is small (around 5% of all the oxide ions). Hence, the impact on the distortion to Mn octahedra overall is likely to be limited, consistent with the slight increase in the pre-edge intensity of the Mn K-edge on charge.

Figure 7

High resolution RIXS data showing molecular O₂ trapped in the bulk. (a) Ex situ high resolution O K-edge RIXS data collected at 531 eV excitation energy showing the presence of molecular O₂ trapped in the bulk of the charged cathode particles (Li_0.8MnO₂F), which is reversibly reduced back to O^2– on discharge. Both spectral features labeled are also observed in a reference spectrum of pure molecular O₂.⁶³ (b) The peak spacing of the vibrational progression decreases linearly with increasing energy loss (Birge–Sponer plot, upper panel) consistent with an anharmonically oscillating O₂ diatomic. The initial peak spacing, equivalent to the fundamental vibration frequency (v_f), is 0.192 eV (1550 cm^–1) in charged Li₂MnO₂F corresponding closely to molecular O₂ indicating negligible interaction with the cathode structure. A reference spectrum for Li₂O₂ is also included at 531 eV showing a fundamental vibrational frequency of 0.098 eV (790 cm^–1) consistent with a peroxide O₂^2– moiety.

In addition, ab initio molecular dynamics simulations of the trapped molecular O₂ in the oxyfluoride structure were performed for the first time. We stress that our focus here was to probe the degree of oxygen mobility within the vacancy cluster, rather than long-range diffusion. The variation in O–O bond distance and Mn–O₂ separation over the simulation time was analyzed (Figure 6b); we also examined oxygen mobility in the cavity through the mean square displacements (MSDs), which were directly compared with MSDs from simulations of free gaseous O₂ molecules within a volume of the same size (Figure 6c).

Two key features emerge. First, the trapped O₂ molecule has an O–O bond length that remains directly comparable to that of gaseous molecular O₂ (1.21 Å) whereas the nearest-neighbor Mn–O₂ distance (mean value of 2.33 Å) is always longer than the Mn–O bond (2.0 Å) in the solid lattice, again confirming weak O₂ interactions with the host lattice. Even when the O₂ molecule makes very close approach to its nearest-neighbor Mn atom, the O–O distance remains around 1.2 Å.

Second, the increase in the mean square displacement with time for free gaseous O₂ (Figure 6c) clearly indicates significant molecular diffusion as expected, whereas this is not the case for the O₂ in the solid particle. The results therefore indicate that while the O–O bond length might be similar between the two, the trapped molecular O₂ is different from free, gaseous O₂ in exhibiting substantially reduced freedom of mobility, making it more characteristic of a solid-like environment (Figure 6c); this result is in line with recent solid state ¹⁷O NMR measurements for O₂ in Li_1.2Ni_0.13Mn_0.54Co_0.13O₂.²⁶ The rigid trapping of O₂ within close proximity to cation centers in this cavity also helps to rationalize how it could be reduced to O^2– with ease on discharge.

To probe experimentally the nature of oxidized oxygen in charged Li₂MnO₂F, O K-edge RIXS was performed at a higher resolution than previously achieved for this material.³⁶ Previous RIXS data for charged electrodes revealed a more prominent elastic peak when exciting at ∼531 eV and a new energy loss feature, attributed to the formation of localized electron holes on oxygen. The new high-resolution data in Figure 7a show that the broad elastic peak can in fact be resolved into a progression of sharp peaks, as also observed for the layered O-redox material Li_1.2Ni_0.13Mn_0.54Co_0.13O₂²⁶ This peak progression arises from the molecular vibrations of an O–O diatomic with well-defined frequency matching that of molecular O₂ (1550 cm^–1) and clearly distinguishable from superoxide O₂^– and peroxide O₂^2– which have vibrational frequencies of around 1100 and 790 cm^–1, respectively.⁶² The peak spacing decreases linearly with increasing energy loss (Birge–Sponer plot Figure 7b) consistent with an anharmonically oscillating diatomic. A reference RIXS spectrum for Li₂O₂ is also included showing a peak spacing consistent with peroxide O₂^2– as expected, demonstrating the ability of RIXS to distinguish different O–O bond orders.

The other energy loss feature at 8 eV also belongs to molecular O₂ and can be assigned to the filled π molecular orbitals. These results show that the localized electron hole states appearing at 531 eV reside on O₂ molecules, which, since the RIXS experiment was performed under ultrahigh vacuum conditions, must be trapped within the bulk of the cathode as found from our ab initio simulations. The similarity of the peak spacing to that of gaseous molecular O₂ suggests that there is minimal bonding interaction with the host lattice in agreement with the ab initio simulations, which would accord with the expectation for strong localization of electron density in the heavily hybridized O=O bond. However, as has been noted previously,^64,65 the excitation energy at which these O-redox RIXS features appear is slightly higher relative to O₂ in the gas phase by about 0.5 eV. This is consistent with the O₂ molecules being trapped in a solid-like environment.

On discharge, the signal is no longer evident indicating that the O₂ that formed in the fully charged samples is no longer present in the bulk material. The lack of O₂ gas evolution at the surface during discharge from differential electrochemical mass spectrometry measurements³⁶ which, coupled with the disappearance of the O₂ signal from the RIXS data, indicates that the trapped O₂ is reduced back to O^2–. It is worth mentioning that O^2– ions and molecular O₂ are the most stable forms of oxygen.

To investigate the possibility of the RIXS features being beam-induced, we undertook low temperature measurements (20 K) to suppress sample heating by the beam. The data, Figure S5, show negligible difference between the spectra indicating no such effect. Taken together, the full outgassing of the electrode under UHV conditions and the reversible reduction of O₂ rule out O₂ being trapped anywhere other than in the particle bulk, where it can still be reversibly reincorporated back into the structure as O^2–. Overall, the reversible O-redox process involves O^2– being oxidized to form bulk molecular O₂ on charge, followed by its reduction on discharge to reform O^2–.

The observation of molecular O₂ in layered O-redox cathode materials and here, for the first time, in disordered rocksalts, suggests the two systems share the same O-redox mechanism, (i.e., 2O^2– ⇋ O₂+ 4e^–). However, Li₂MnO₂F does not exhibit the commonly observed O-redox charging plateau at 4.6 V vs Li⁺/Li, nor such large first cycle voltage hysteresis (Figure 1). Both phenomena have been recently linked with the irreversible loss of highly ordered honeycomb superstructures belonging to the layered cathodes. In Li_1.2Ni_0.13Mn_0.54Co_0.13O₂,²⁶ all oxide ions in the honeycomb lattice will be coordinated by at least two transition metal (TM) ions (O(Li₄TM₂)) which, during charge, are oxidized to O^n– at a high potential of 4.6 V. However, this honeycomb arrangement of O^n– is highly unstable. In-plane TM migration to form vacancy clusters occurs, causing some O to become coordinated by fewer than two TM ions which then dimerize to form stable O₂ molecules. On discharge, these vacancy clusters are repopulated by Li leading to O(Li₅TM) and O(Li₆) configurations which remain in the structure explaining the lack of further voltage plateaus. In contrast, Li₂MnO₂F is already intrinsically disordered in the pristine state, possessing a range of coordination environments including some O(Li₅Mn) and O(Li₆) regions. Therefore, O⁻ redox can occur without such severe structural rearrangement and hence less pronounced voltage hysteresis. After the first cycle, the load curves for both compounds (Figure S2) exhibit a similar degree of voltage hysteresis in line with the presence of preformed sites for O₂ formation in both materials after disordering of the TM ions within the TM layer of Li_1.2Co_0.13Ni_0.13Mn_0.54O₂.

Our observations on Li₂MnO₂F are unexpected in the context of previous work, which cannot be simply translated from ordered Li-rich layered compounds to disordered rocksalt oxyfluorides, and are important in future strategies to develop new high capacity cathodes.