Nonpolarizing oxygen redox in Na₂Mn₃O₇

Na₂Mn₃O₇ was synthesized from a solid-state reaction following a reported procedure^30,31. The powder X-ray diffraction pattern for the resulting compound (Supplementary Fig. 1) is indexed to triclinic P$$ \overline{1}$$, confirming the successful synthesis of Na₂Mn₃O₇. The selected-area electron diffraction (SAED) pattern (Supplementary Fig. 2a) reveals the ordered arrangement of Mn atoms and vacancies in the Mn slab ($$ \sqrt{7}\times \sqrt{7}$$ superlattice). The charge/discharge curves at a rate of C/20 show a reversible oxygen-redox capacity of approximately 70 mAh/g (Supplementary Fig. 3), which is consistent with a previous report³¹. Fig. 2a shows the dQ/dV plot for Na_2-xMn₃O₇ between 3.0 and 4.7 V vs. Na/Na⁺ during the second charge/discharge cycle. As the nominal valence state of Mn in Na₂Mn₃O₇ is tetravalent (maximum valence of octahedral Mn), the redox center in Na_2-xMn₃O₇ is a priori attributable to oxygen; the Mn L-edge X-ray absorption spectra in our previous work confirmed the negligible contribution of Mn redox to the capacity³¹. Importantly, the dQ/dV peaks of the charge/discharge processes have an extremely small voltage hysteresis of merely 0.03–0.04 V, implicating the reversible redox reaction of oxygen (O²⁻/O^−•) (Fig. 1b) without peroxide-like O₂²⁻ formation (Fig. 1a).

O K-edge X-ray absorption spectra (Fig. 2b inset) show the emergence of a new absorption peak at 531.5 eV after the charge process, which corresponds to the excitation of an O 1 s core electron to an O 2p hole, as commonly observed for other oxygen-redox electrodes^35–38. To monitor the detailed valence partial density of states (pDOS) of oxygen, resonant inelastic X-ray scattering (RIXS) spectra were measured using an incident photon of 531.5 eV. The emergence of a new emission peak at 523 eV in the RIXS spectrum for charged Na_2-xMn₃O₇ is typical for charged oxygen-redox electrodes^{5,24,28,39,40}. Although O^−• formation is implicated by the small voltage hysteresis in the charge/discharge processes (Fig. 2a), the new inelastic scattering can be explained by either (1) energy loss from a π(Mn-O) → π^*(Mn-O) transition (O^−• formation), or (2) energy loss from a σ(O-O) → σ^*(O-O) transition (peroxide-like O₂²⁻ formation)⁴⁰. While observation of the new RIXS peak at 523 eV confirms the occurrence of the oxygen-redox reaction itself in Na_2-xMn₃O₇, but this evidence alone does not definitely identify the reaction mechanism, or the chemical state of the oxidized oxygen species.

Magnetic elucidation of O⁻

To identify the oxidized species, we measured the magnetic susceptibility $$ \left(\chi \right)$$ of Na_2-xMn₃O₇ with respect to its sensitivity to local coordination structures and electronic configurations. Figure 3a shows the temperature dependence of $$ {\chi }^{\mathrm{-1}}$$ during the second charge/discharge cycle. The $$ {\chi }^{\mathrm{-1}}$$ vs. T plot follows the Curie-Weiss law ($$ \chi =\frac{C}{T-\mathrm{\Theta }}$$, $$ C$$ = Curie constant, $$ \mathrm{\Theta }$$ = Weiss temperature) at high temperature (> 200 K), providing the spin-state information of Mn and O in Na_2-xMn₃O₇. Because the ordered arrangement of Mn atoms and vacancies in the Mn slab is maintained after the charge process, as confirmed from the SAED pattern (Supplementary Fig. 2b), in-plane and out-of-plane Mn migrations do not occur in Na_2-xMn₃O₇. Therefore, the magnetic property changes are ascribed solely to the spin-state changes of Mn and O. Figure 3b shows the values of $$ C$$ and $$ \mathrm{\Theta }$$ as a function of x in Na_2-xMn₃O₇. Before the charge process, the value of $$ C$$ (5.77 cm³ K mol⁻¹) agrees with the calculated value (5.63 cm³ K mol⁻¹ for g = 2.0) for three Mn⁴⁺ (S = 3/2) in Na₂Mn₃O₇. The negative $$ \mathrm{\Theta }$$ of −192 K indicates an in-plane antiferromagnetic interaction between Mn⁴⁺ cations⁴¹, which contradicts the prediction of ferromagnetic superexchange from the Goodenough-Kanamori rule⁴². This magnetic behavior is ascribed to the deviation of Mn-O-Mn bond angles from 90°, which most likely induces the antiferromagnetic superexchange; an antiferromagnetic d-d direct exchange interaction may also exist, as reported for Li₂MnO₃ ⁴³.

Fig. 3

Magnetic elucidation of the existence of O⁻ in Na_2-xMn₃O₇.

a The inverse of a magnetic susceptibility (χ) as a function of temperature, and b Curie constant (C) and Weiss temperature (Θ) for Na_2-xMn₃O₇ during the second cycle. Filled and empty circles correspond to data points for the charge and discharge processes, respectively.

The value of $$ C$$ monotonically decreases to 4.22 cm³ K mol⁻¹ upon charge, which is close to the calculated value (4.75 cm³ K mol⁻¹ for g = 2.0) under a simple assumption that Mn⁴⁺-O^−• behaves as S = S_Mn + S_O = 1 (Zhang-Rice triplet)^44–46. Note that the Zhang-Rice triplet in Na_2-xMn₃O₇ arises from π-type direct exchange while the Zhang-Rice singlet originally proposed for superconducting cupper oxides is generated via σ-type direct exchange⁴⁴. Diamagnetic peroxide-like O₂²⁻ formation, which should yield a constant C, is clearly ruled out by the observed decrease in C, as speculated from the very small hysteresis in the dQ/dV plot. The increase in $$ \mathrm{\Theta }$$ upon charge suggests the increment of ferromagnetic superexchange interaction between Mn spins through bridging O^−• over the effects of antiferromagnetic superexchange or d-d direct exchange. After discharge, both $$ C$$ and $$ \mathrm{\Theta }$$ return to their initial values, supporting the reversibility of the charge/discharge processes without structural or electronic degradations. Overall, the magnetic measurements support the reversible oxygen-redox reaction of O²⁻/O^−• (Fig. 1b) in Na_2-xMn₃O₇.

O^−• vs. O₂²⁻

Why does Na_2-xMn₃O₇ specifically exhibit the reversible oxygen-redox reaction of O²⁻/O^−•? To answer this question, we compared the thermodynamic stability of O^−• and peroxide-like O₂²⁻ in Na_2-xMn₃O₇ using DFT calculations. Figure 4a shows the calculated magnetic moments of Mn and O in optimized structures without peroxide-like O₂²⁻ (Supplementary Fig. 4a). For the experimentally accessible desodiation range (0 ≤ x ≤ 1 in Na_2-xMn₃O₇), a magnetic moment gradually emerges on the O atoms upon desodiation, which indicates O^−• formation. The slight decrease in the average magnetic moment of Mn is explained by the initiation of metal-to-ligand charge transfer (π back-donation) from the formation of a ligand hole (O^−•)⁴⁰. The calculated voltage profile under the O^−• formation model (Fig. 5 and Supplementary Fig. 5) accurately reproduces the experimentally observed voltage plateaus (4.30 and 4.53 V), further supporting O^−• formation in Na_2-xMn₃O₇.

Fig. 4

Theoretical derivation of stable existence of O^−• in Na_2-xMn₃O₇.

a Calculated magnetic moments of Mn and O as a function of x in Na_2-xMn₃O₇, and b calculated formation energy of peroxide-like O₂²⁻ as a function of the average magnetic moment (i.e., degree of oxidation) of O in Na_2-xMn₃O₇. For the plots of the magnetic moments, black crosses are a magnetic moment of each atom, while blue and red circles are average magnetic moments. For the plot of the formation energy, black crosses represent the formation energy of various O-O pairs, while red circles represent the lowest value at each desodiated state. Gray shaded area is experimentally inaccessible due to too high desodiation potential (see Supplementary Fig. S5).

Fig. 5

Predicted voltage hysteresis of Na_2-xMn₃O₇ with hypothetical peroxide-like O₂²⁻ dimers.

DFT calculated convex hull and voltage profile of Na_2-xMn₃O₇ with O^−• and with O₂²⁻. The formation energies of both pristine structures and peroxide phases were calculated relative to pristine Na₂Mn₃O₇ and NaMn₃O₇ phases. Black and red crosses in the convex hull are formation energies of Na_2-xMn₃O₇ with O^−• and with O₂²⁻, respectively. Black and red circles are the lowest states at various desodiated states. The black solid line of the voltage profile represents the charge/discharge curves without O₂²⁻ formation, while the red solid line represents the discharge curve for hypothetical Na_2-xMn₃O₇ with O₂²⁻.

The formation energy of peroxide-like O₂²⁻ in Na_2-xMn₃O₇ was calculated for various O-O pairs, where calculated O-O bond lengths for the lowest energy structures are in the peroxide O₂²⁻ range of 1.46-1.47 Å (Supplementary Fig. 4b)⁴⁷. The average magnetic moment of the O-O dimers is small (-0.094(6) μ_B), also suggesting the formation of peroxide-like O₂²⁻ rather than superoxide-like O₂⁻ when O-O dimerization occurs. The formation energy of O₂²⁻ (Fig. 4b) was calculated as the difference between the total energies of Na_2-xMn₃O₇ with O^−• and with O₂²⁻ as a function of the average magnetic moment of O (i.e., the degree of O oxidation). During the experimentally accessible charge process (0 ≤ x ≤ 1 in Na_2-xMn₃O₇), the formation energy of O₂²⁻ is positive relative to that of O^−•, indicating that Na_2-xMn₃O₇ with O^−• is thermodynamically favorable compared to Na_2-xMn₃O₇ with O₂²⁻. Presumably, a hole on O^−• in Na_2-xMn₃O₇ is stabilized through a (σ + π) multiorbital Mn-O bond⁴⁰, as demonstrated using a crystal orbital overlap populations (COOP) analysis in our previous works³¹. Note that, in parallel with a localized feature, the oxygen hole has an itinerant feature through the (σ + π) multiorbital interaction. Indeed, the experimentally observed value of C for Na₁Mn₃O₇ (4.22 cm³ K mol^-1) is slightly lower than that calculated for a localized model (4.75 cm³ K mol⁻¹) (Fig. 3b).

Figure 4a also indicates that further oxygen oxidation leads to negative O₂²⁻ formation energy (O₂²⁻ favorable): the oxygen oxidation lowers the energy level of O 2p well below that of Mn t_2g, which renders Mn-O interaction weak⁴⁸. In the balance of competing stabilization mechanisms via (σ + π) multiorbital Mn-O bond versus peroxide O-O bond, the adequate oxygen-redox capacity (approximately 70 mAh/g, Supplementary Fig. 3) keeps the (σ + π) multiorbital stabilization dominant, making O^−• stable in Na_2-xMn₃O₇. It should also be emphasized that the calculated voltage profile for the sodiation of Na_2-xMn₃O₇ with O₂²⁻ shows a voltage hysteresis of 0.3–0.5 V relative to the charge process of Na_2-xMn₃O₇ with O^−• (Fig. 5), which clearly contradicts the experimental observations. O₂²⁻ formation is therefore inhibited in Na_2-xMn₃O₇, leading to an O²⁻/O^−• based reversible and nonpolarizing oxygen-redox capacity. Excessive oxidation of oxygen (e.g., x ≥ 1 in Na_2-xMn₃O₇), which drives O^−• to dimerization, is an absolute taboo for a nonpolarizing oxygen-redox capacity^48,49. Based on this criterion, the oxygen-redox reaction should be employed as an auxiliary charge-compensation mechanism under a prudent control rather than as a main charge-compensation mechanism.

Electrochemistry

Na₂Mn₃O₇ was synthesized by sintering the stoichiometric mixture of NaNO₃ and MnCO₃ at 600 °C under O₂ flow for 4 h^30,31. Electrochemical measurements were conducted using CR2032-type coin cells. Positive electrodes consisted of 80 wt% Na₂Mn₃O₇, 10 wt% acetylene black, and 10 wt% polyvinylidene difluoride (PVDF), which were coated on Al foil using N-methylpyrrolidone (NMP) as the solvent. Sodium was used as the negative electrode, with 1.0 M NaPF₆ in ethylene carbonate (EC)/diethyl carbonate (DEC) (1:1 v/v%) as the electrolyte. The cells were cycled at a charge/discharge rate of C/20.

X-ray absorption/emission

Ex situ X-ray absorption spectroscopy (XAS) and resonant inelastic X-ray scattering (RIXS) measurements were performed on samples without exposure to air at BL07LSU of SPring‐8. A bulk‐sensitive partial fluorescence yield (PFY) mode was employed for O K‐edge XAS. Extended X-ray absorption fine structure was conducted at the BL-9C Beamline of the Photon Factroy, KEK, Japan. SAED patterns were recorded using an electron microscope (Titan Cubed, FEI Co.) operated at 80 kV after transferring the samples without exposure to air.

Calculations

All structures were calculated using DFT, as implemented in the Vienna Ab Initio Simulation Package (VASP)^53,54. The projector-augmented wave pseudopotential and a plane-wave basis set with an energy cut-off of 520 eV were used⁵⁵. The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof functional describes the exchange-correlation energy^56,57. To remove the self-interaction error, the Hubbard U correction was applied to the d electrons of Mn atoms (U_eff = 3.9 eV)^58,59. The k-point was sampled on a 3 × 3 × 5 grid for all calculations. The Grimme scheme (DFT-D3) was applied to include van der Waals corrections⁶⁰. Crystal structures were visualized with VESTA software ⁶¹.

We constructed a 2 × 2 × 1 supercell (Na_16-iMn₂₄O₅₆) to calculate a series of charged-structures of Na_2-xMn₃O₇ (0 < x ≤ 2). All possible cation orderings (Na⁺) were searched in the Supercell program, with the exclusion of symmetrical duplicates⁶². Considering the large number of configurations in the initial search (e.g., 1670 for Na₈Mn₂₄O₅₆), we conducted a multi-step calculation of total energies to determine the stable cation ordering for each structure. First, we simply calculated the total energies without geometry optimization and selected the 100 lowest-energy configurations for each charged-structure. Second, we optimized both the lattice and atomic positions of the selected configurations under a force convergence of 0.03 eV Å⁻¹, where the energy cut-off and number of k points were reduced to 400 eV and 1 × 1 × 3, respectively. At this point, we screened out the 20 lowest-energy configurations for each structure. Finally, we reconducted the geometry optimization with a more reliable energy cut-off (520 eV) and k-point mesh (3 × 3 × 5) under a force convergence of 0.01 eV Å⁻¹, identifying all the lowest-energy structures in the charge process.

The convex hull was constructed to identify the stable phases among the determined structures, with the formation energies (at 0 K) calculated as:

The formation of peroxide-like O₂²⁻ in Na_2-xMn₃O₇ was calculated by creating a short O-O bond in Na_16-iMn₂₄O₅₆, as shown in Supplementary Figure 4b. After the atomic positions were fully relaxed, no other significant structural change was found except for the O-O dimer. The formation energy of O₂²⁻ was calculated by: