Lithium ion batteries (LIBs) based on rocking lithium between cathode and anode electrodes are used in a wide range of applications from portable electronics to vehicle propulsion. However, there are cost and resource problems since most LIBs use not only expensive lithium but also less abundant metals such as cobalt. In addition, use of flammable organic electrolyte solutions may cause a safety problem. Those problems are critical especially for large-scale batteries used for electric vehicles and electric grids. Some alternatives have recently been proposed by the replacement of lithium and cobalt-based oxides with sodium and iron-based compounds, respectively: For example, Ellis et al.1 have shown the viability of sodium-ion batteries utilizing a sodium iron fluorophosphate (Na2FePO4F) as a cathode material, where reversible redox of iron (Fe2+/Fe3+) proceeds with topotactic extraction and reinsertion of sodium in organic electrolyte solutions. More recently, Wessells et al.2 have reported that aqueous sodium ion batteries employing a Prussian blue analogue of a nickel hexacyanoferrate AxNiyFe(CN)6·nH2O (A: alkali metal ions) as an intercalation cathode are attractive for grid-scale power storage. Batteries using aqueous electrolytes have advantages in terms of safety and cost in addition to lithium-free, while they do not show high cell voltages due to potential window of water. In addition, large-scale energy storage is needed more and more today. Therefore, it is urgent to develop novel batteries satisfying not only cost and safety but also “element strategy3” requirements.

It is well known that some kinds of perovskite-type oxides of A1−xSrxCoO3 (A = La or Nd) show good oxygen diffusivity4,5 and the oxygen atoms of such compounds can be extracted and reinserted topotactically by electrochemical reduction and reoxidation in an aqueous alkaline solution6,7. Longo et al. have reported that CaMnO3 electrochemically undergoes extraction and reinsertion of oxygen in a strong basic solution and that this reaction can be applied to cathode materials8.

Electrochemical oxidation and reduction (Fe4+/Fe3+) of Sr1−xLaxFeOz in an alkaline solution reversibly take place at least on the surface in the range 2.5 + x/2 < z < 39, while the iron valences can possibly be changed from 3 − x to 3 in the oxygen composition range 2.5 < z < 2.5 + x/2. Therefore, we have reached an idea that compounds A1−xLaxFeOz(A = Ca, Sr) can work as not only a cathode but also an anode utilizing different oxygen composition ranges of the Fe3+/Fe4+and Fe2+/Fe3+ couples, respectively.

In this report, we show the principle of novel lithium-free secondary oxygen rocking batteries, in which oxygen shuttles between the cathode and anode composed of iron-based perovskite-related oxides Ca0.5La0.5FeOz(2.5 ≤ z ≤ 2.75 and 2.75 ≤ z ≤ 3.0) by utilizing topotactic oxygen extraction and reinsertion during discharge and charge processes (Fig. 1).

Figure 1
figure 1

Operating principle of oxygen rocking batteries using CLFO.

This is an example of oxygen rocking batteries. The directions of arrows indicate discharging. In discharging, cathode and anode materials undergo oxygen extraction and insertion, respectively and oxygen is transferred from the cathode to the anode in an electrolyte.


Electrochemical behaviors of Ca0.5La0.5FeOz

Figure 2 illustrates the potential variation of Ca0.5La0.5FeOz(CLFO) at a constant current of 1.40 mA g−1. Since the composition of the as-synthesized sample was determined to be Ca0.5La0.5FeO2.863, the electrochemical measurement was started from z = 2.863 as denoted by the initial state. In the oxidation process, the potential gradually increased and reached a plateau (ca. 0.5 V) below z = 3 due to oxygen evolution. On the other hand, in the reduction process, the potential gradually decreased around 0 V, steeply decreased from −0.1 V to −0.8 V around z = 2.75, gradually decreased around −0.9 V and then leveled off around −1.1 V due to hydrogen evolution.

Figure 2
figure 2

Potential variation of Ca0.5La0.5FeO2.863 in the oxidation and reduction at a constant current of 1.40 mA g−1.

The upper transversal axis indicates the oxygen contents (z) electrochemically estimated with integrated currents assuming the following reaction, Ca0.5La0.5FeO2.863 + δH2O + 2δe Ca0.5La0.5FeO2.863−δ + 2δOH.

Changes in crystal structures and valence states of iron with electrochemical reduction

Figure 3 shows X-ray diffraction (XRD) patterns of the CLFO samples prepared by electrochemical-reduction shown by the circles in Fig. 2. All peaks of Ca0.5La0.5FeO2.863 were assignable to the GdFeO3-type structure, which is slightly-distorted cubic. The Rietveld analysis successfully provided the cell parameters and atomic positions with small value of goodness-of-fit of 1.54 (see Fig. S1 and the refined parameters were shown in Supplementary information). The unit cell could be regarded as pseudo-cubic (Fig. 3). Upon the reduction of Ca0.5La0.5FeO2.863, neither new peak nor peak-splitting was observed except for lower angle shifts of peak positions, showing the expansion of the cell parameter of a*cubic. The a*cubic values are plotted as a function of oxygen contents in Fig. 4. They increased with decrease in z from 2.863 to 2.744 and did not much change below z = 2.744.

Figure 3
figure 3

XRD profiles of electrochemically reduced CLFO.

The hkl indices assigned to the cubic cells with the reduced-cell parameter, a*cubic (a ≈ 21/2×a*cubic, b ≈ 2×a*cubic, c ≈ 21/2×a*cubic; a, b, c: cell parameters based on GdFeO3-type structure) were shown. z; see the caption note of Fig. 2.

Figure 4
figure 4

The variation in cell parameters of CLFO on the basis of the pseudo-cubic cell.

The estimated standard deviations were shown by lengths of bars. z; see caption note of Fig. 2.

Mössbauer parameters of CLFO samples (z = 2.744, 2.701 and 2.618) are summarized in Table 1.

Table 1 Mössbauer parameters and oxygen contents estimated from valence and coordination states of iron ions in Ca0.5La0.5FeOz

Repeatability of Electrochemical reaction of CLFO

As shown in Fig. 2, the electrochemical reduction (viz. oxygen extraction) of CLFO proceeded stepwise in two distinct potential regions around 0 V (due to Fe4+/Fe3+) and −0.8 V (Fe3+/Fe2+). To construct a rechargeable battery using CLFO, reoxidation reactions must proceed reversibly at these potential levels. Figure 5 shows the results of repeated redox tests under a constant current density of 5.60 mA g−1. The electrochemical reactions in potential regions around both 0 V and −0.8 V were substantially reversible and repeatable. Especially in the higher potential region, the shapes of the reduction and oxidation curves remained almost unchanged with repetition of the cycles and the potential difference between the reduction and oxidation processes was small.

Figure 5
figure 5

The potential profiles of CLFO of redox repetition tests in the two different potential ranges at a constant current of 5.60 mA g−1.

The maximum capacities of the reduction and reoxidation processes in the potential region around −0.8 V were 20 and 15 mA h g−1, respectively and different from each other. This difference does not result from side reactions, but from the slow diffusion of oxygen, which would be overcome by fabricating a well-designed nano-structured electrode.

When particles of Ca0.5La0.5FeO2.863 were treated in a strong basic aqueous solution at pH 14 for 24 h, no change in the XRD pattern was observed, showing that Ca0.5La0.5FeO2.863 is stable in a strong basic aqueous solution. In addition, the XRD pattern of Ca0.5La0.5FeO2.863 was hardly changed and only the perovskite phase was observed even after discharge and charge cycles. We also confirmed that the electrolyte solution was clear and no unloading of the electrode materials occurred after discharge and charge cycles.


The oxygen contents obtained in different two ways (from (i) valence of iron and (ii) coordination number of oxygen) with Mössbauer data are also shown in Table 1. These oxygen contents were close to each other, showing that the oxygen contents in CLFO are appropriately determined. In addition, these oxygen contents agreed with those estimated electrochemically, indicating that CLFO can electrochemically be reduced with high Faradaic efficiency and that the oxygen contents are electrochemically controlled by the integrated currents.

Since the valence of the all iron ions is +3 at z = 2.75, Fe4+ ions in Ca0.5La0.5FeO2.863 are probably reduced to Fe3+ ones around 0 V. The increase in the a*cubic values is explained by the reduction of Fe4+ ions to larger Fe3+ ones. Further electrochemical reduction would proceed by the reduction of Fe3+ ions to Fe2+ ones. The a*cubic values did not much change beyond z = 2.744, which is likely explained by the reduction of high-spin Fe3+ ions (ionic radius10: 79 pm) to similarly-sized low-spin Fe2+ ones (75 pm). The Mössbauer analysis showed that most of Fe2+ ions were in the low spin state S = 0, supporting the reduction of high spin Fe3+ ions to low spin Fe2+ ones beyond z = 2.744.

The small potential difference between the reduction and oxidation processes in the higher potential region around 0 V in Fig. 5 would result from fast oxygen diffusion in CLFO at the corresponding composition range, because the kinetics of electrode reactions involving insertion and extraction of a substance is usually controlled by the transport in a host material. Therefore, the chemical diffusion coefficient of oxygen was estimated by using a simple plane sheet diffusion model to be 1.2×10−13 cm2 s−1 (see Fig. S3 in Supplementary Information). In comparison with the chemical diffusion coefficients of Li+ ions in the typical LIB materials, the value is approximately a few to several orders of magnitude lower than those of LiCoO2 (typically, 10−8 – 10−10 cm2 s−1)11,12,13 and LiMn2O4 (typically, 10−10 – 10−13 cm2 s−1)14,15,16 and one order of magnitude higher than that of LiFePO4 (1.8×10−14 cm2 s−1)17. Since the size of the oxygen ion is much larger than those of lithium, sodium and magnesium ions, the diffusivity of the oxygen ion is generally lower than those of lithium, sodium and magnesium ions. This low diffusivity would result in the operation of oxygen rocking batteries at lower currents.

In the lower potential region around −0.8 V, reduction and oxidation capacities are ca. 20 and 15 mA h g−1, respectively and much smaller than 70.2 mA h g−1 estimated on the assumption that the oxygen content ranges from 2.75 to 2.50 (Δz = 0.25). In the same potential region in Fig. 2, the oxygen content ranged from 2.75 to 2.53 (Δz = 0.22) and the Δz of 0.22 corresponded to 88% of the Δz = 0.25. Since the electrochemical reduction in Figs. 2 and 5 were conducted at 5.6 and 1.4 mA g−1, respectively, the capacities around −0.8 V depended on current densities (see Fig. S3 in Supplementary information). This is likely explained by much slower oxygen diffusion (the estimated chemical diffusion coefficient; 1×10−15 cm2 s−1, see eq. S5 and discussion in Supplementary information). Such slow oxygen diffusion in this content range would be related to ordering of oxygen vacancies such as one dimensional arrangement of oxygen vacancies observed in the brownmillerite structure, since the formation of microdomains upon the appearance of the brownmillerite CaFeO2.5 phase in CaxLa1−xFeO3−x/2 has been reported18.

In conclusion, the electrochemical reduction of Fe4+ to Fe3+ and Fe3+ to Fe2+ and their reoxidation in CLFO proceed reversibly and repeatedly via extraction and reinsertion of oxygen in two potential regions around 0 V and − 0.8 V. This shows that CLFO can serve as both cathode and anode materials in an oxygen rocking battery. The CLFO is, however, one of model electrode materials of the oxygen rocking batteries and the developments of more suitable materials for each of a cathode and an anode will enable the new practicable oxygen rocking batteries.


Powders of Ca(CH3COO)2·H2O (0.881 g, 5.0 mmol), La(CH3COO)3·1.5H2O (1.715 g, 5.0 mmol), Fe(NO3)3·9H2O (4.040 g, 10.0 mmol) and citric acid monohydrate (4.203 g, 20.0 mmol) were dissolved in ultrapure water (80 mL). The solution was evaporated to dryness at 60°C followed by the evaculation at 100°C. The powder was ground with an agate mortar and calcined at 300°C in air followed by grounding. The powder was heated to 1000°C at 10°C min−1, kept at 1000°C for 10 min and cooled to room temperature in a furnace. The average valences of iron ions was determined to be 3.226(2) by the iodometric titration and the empirical formula was Ca0.5La0.5FeO2.863(1).

The CLFO samples (z = 2.827 − 2.618) were prepared by electrochemical reduction of Ca0.5La0.5FeO2.863 with three electrode glass cells.

The electrochemically-reduced CLFO samples were removed from the glass cells, mounted in air-tight sample holder in an inert gas system and were structurally analyzed by X-ray diffractometry using monochromated Cu-Kα1. The local conditions of iron ions were investigated by Mössbauer spectroscopy using 57Co at room temperature. The parameters (the isomer shift relative to α-Fe (IS), the quadrupole splitting (QS), the internal magnetic field (B) and the relative intensity of the component (I)) were refined by profile-fitting (see Fig. S2 and discussion in Supplementary information). The oxygen contents in CLFO could be determined with the Mössbauer data in two ways: (1) The average valences of iron ions were calculated with the valences and relative intensities of the components and the oxygen contents were determined. (2) The oxygen contents were calculated with the coordination numbers of the oxygen surrounding iron ions (tetrahedral, square-pyramidal and octahedral iron sites possess 2, 2.5 and 3 oxide ions, respectively.)

Electrochemical measurements were performed using a three electrode beaker cell with a reference electrode (Hg/HgO) and an electrolyte solution (1 M NaOH aq). A platinum mesh was used as a counter electrode. A mixture of CLFO, conducting additive (KS-6L, Timcal) and binder (PTFE powder) in the ratio 77/19/3 (w/w) was ground, pressed on a current collector and used as a working electrode. A platinum mesh was employed as the current collector for oxidation and a gold one for reduction, because they have high overpotential of oxygen evolution and hydrogen evolution, respectively. The oxidation or reduction at a constant current ranging from 1.40 to 28.0 mA g−1 was performed to investigate the electrochemical behavior of Ca0.5La0.5FeO2.863. The reversibility and repeatability were investigated by the redox tests at a constant current of 5.60 mA g−1 in the potential ranges 0.5 − −0.5 V and −0.4 − −1.1 V.