Dissolution and ionization of sodium superoxide in sodium–oxygen batteries

With the demand for high-energy-storage devices, the rechargeable metal–oxygen battery has attracted attention recently. Sodium–oxygen batteries have been regarded as the most promising candidates because of their lower-charge overpotential compared with that of lithium–oxygen system. However, conflicting observations with different discharge products have inhibited the understanding of precise reactions in the battery. Here we demonstrate that the competition between the electrochemical and chemical reactions in sodium–oxygen batteries leads to the dissolution and ionization of sodium superoxide, liberating superoxide anion and triggering the formation of sodium peroxide dihydrate (Na2O2·2H2O). On the formation of Na2O2·2H2O, the charge overpotential of sodium–oxygen cells significantly increases. This verification addresses the origin of conflicting discharge products and overpotentials observed in sodium–oxygen systems. Our proposed model provides guidelines to help direct the reactions in sodium–oxygen batteries to achieve high efficiency and rechargeability.

. The introduction of the rest time between the discharge/charge enables us to assess the effects of a non-electrochemical process on the voltage profiles, as discussed in the main discussion section. Even the high charge currents without any rest resulted in low polarization of charge; the insertion of rest time led to an increase in the charge polarizations. Thus, the effects of the chemical reactions were reconfirmed by similar electrochemical tests with relatively high charge currents. Moreover, the voltage after the 24-h rest has no flat region at 2.4 V, which is closely related to the decomposition reaction of NaO2. Figure 2). We also determined that the charge behaviors of Na-O2 cells could be altered by the utilized capacities even with equally applied currents, as demonstrated in Supplementary Fig. 2. Unlike the capacity-limited cycling, the voltage profile without any capacity limit shows a low polarized charge potential over the entire capacity range, as observed in Supplementary Fig. 2b. This finding indicates that the NaO2 formed without limited capacity was relatively stable and could be electrochemically decomposed rather than chemically deformed. Even though further analyses might be required to attain a clear understanding, we believe that the size of the discharged crystallite is also importantly affected by the dissolution of NaO2. Thus, the longer lifetime of the large-sized NaO2 from the fully operated discharge was attributed to the slow kinetics of dissolution in the electrolytes resulting from its low interfacial surfaceto-volume ratio. Figure 3). We also investigated the cycle properties of Na-O2 cells depending on the shape of the charge profiles by simply controlling the charge currents. Supplementary Fig. 3 demonstrates the better reversibility achieved with the low charge potential compared with the highly polarized charge profile.

Supplementary Note 3 (Supplementary
The irreversible reactions with the lower charge current might be due to the chemical evolution of Na2O2·2H2O, which shows the 3-step charge profile.
In addition, the amount of H2O in the electrolyte used to assemble the cell (200 μL) can be estimated using the following calculation (density of electrolyte ~ 1 g/mL, H2O content of electrolyte ~ 10 ppm): 18 g mol -1 ) × (6.02 × 10 23 mol -1 ) = 6.69 × 10 16 14 The amount of H2O in the electrolyte was only approximately 0.3% of the amount of NaO2. Considering the given ratio of H2O/NaO2 and the expected reaction, the amount of H2O is insufficient for the complete transformation of NaO2 into Na2O2•2H2O. Thus, the sources of H + must originate from the electrolyte solvents and not the residual H2O molecules.  Fig. 5, firstly, we generated O2 − in the electrolyte composed with 10 mM TBAClO4 in DEGDME in the symmetric cell as shown in Supplementary Fig. 6. The simulated coulomb was 5 mAh, which is 5-fold excess amount of electrochemically formed NaO2 with the discharge. If we exclude the shuttle effect during the discharge, the concentration of O2 − is approximately 0.93 M compared to the volume of injected electrolyte (200 μl). Figure 7). After allowing the relaxation of O2 − generated in Supplementary Fig. 6 in the presence of the electrolyte, the electrolyte was examined by FTIR to identify the chemical reactions triggered by O2 − . As compared to the as-prepared electrolyte, it was observed that the broad peak at about 3400 cm −1 evolves, which corresponds to νO-H stretch. This indicates the formation of free OH − with the chemical reaction of O2 − with the electrolyte. Furthermore, it was found that a small δO-H band and νC=O stretch at around 1625 cm −1 and 1728 cm −1 , respectively. The former signal also corresponds to the formation of the free OH − , and the latter could be attributed to the trace amount of carboxylic functional groups (-COOH) in the byproduct. The presence of H2O2 was difficult to confirm from the FTIR due to the overlaps of signatures with DEGDME and will be discussed in Supplementary Fig. 8 with iodometric experiments.

Supplementary Note 7 (Supplementary
The identification of νO-H, δO-H, νC=O band in the electrolyte strongly support the Reaction 6 in Fig. 5. Figure 8). We carried out the iodometric determinations to supplement the FTIR results above and verify the presence of H2O2 as shown in Supplementary Fig. 8. When the electrolyte exposed to O2 − was added to the basis of 1 M KI aqueous solutions, which were initially transparent, the color of the solution was immediately changed to yellow. It indicates the oxidation of iodide ion (I − ) to triiodide ion (I3 − ), which was induced by the presence of H2O2 via the reacttion (2 I − + 2 H + + H2O2 → I3 − + 2 H2O). Rest of the chemical additives such as as-prepared electrolyte or CH3COOH

Supplementary Note 8 (Supplementary
do not change the color of the iodide solution. This observation supports that H2O2 was formed after the chemical reactions coupled with ORR, and validates Reaction 7 in Fig. 5. Figure 9). On the basis of byproducts from Reaction 6 and 7 in Fig. 5, we attempted to simulate the formation of Na2O2·2H2O as proposed in Reaction 8 (2 NaOH + H2O2 → Na2O2·2H2O). 0.5 M H2O2 aqueous solution was added dropwise to the anhydrous ethanol solution of 1 M NaOH. Because NaOH is insoluble in ether-based solvents, the solvent was used with the anhydrous ethanol to reproduce the effectively dissolved state of NaOH. After the mixing, it was found that the white precipitates were immediately formed. The retrieved precipitates were examined by XRD after drying under vacuum for 30 min. Supplementary Fig. 9 identifies that the main phase of precipitates were Na2O2·2H2O with a trace amount of other phases that might be formed during the process. The formation of Na2O2·2H2O strongly supports the proposed Reaction 8 in Fig. 5. Figure 10-11). In our galvanostatic cycling experiments of Na-O2 cells, the dendritic failure of the Na metal anode was frequently observed during the charge process, similar to previous reports 1- 3 . The SEM images of the separators collected after cycling with the direct currents are presented in Supplementary   Figs. 10a-c, which reveal that the Na metal clogged and penetrated the pores of the separators, resulting in the dendritic growth of the Na metal and failure of the cells. The Na dendrites were visually inspected, as shown in the photographs of the separators and anodes in Supplementary Fig. 10g. These dendrites were critically damaged during the cycling process, resulting in short circuits and potential failures. To suppress and avoid the dendritic growth of Na metal, we applied special operating protocols based on pulsecharging, which is a common methodology in electroplating and deposition [4][5][6] . Several researchers in the battery community have already reported on pulse-charging to suppress the dendritic growth of Li metal [7][8][9] . Under these conditions, the stable electrochemical cycling of Na-O2 cells without any voltage fluctuations or sudden drops was possible. The SEM images in Supplementary Fig. 10d-f also demonstrate the absence of dendritic Na metal penetrating or clogging the pores of the separators. In addition, the upper surface of the pulse-charged Na metal in Supplementary Fig. 10i is seemingly much cleaner than that of the direct-current-charged Na metal.

Supplementary Note 10 (Supplementary
Meanwhile, it was concerned that the rest time between the inter-current periods could affect our experimental observations, because one of the main key parameters for regulating the electrochemical and chemical reaction was the rest time between the discharge/charge. However, as shown in Supplementary Fig. 11, we found that the inserted resting as a pulse did not have any effect on the experimental responses. This strongly indicated that the continuously accumulated resting time during the electrochemical operation was only meaningful. The severe dendritic growth in Na metals and feasibility of pulse-charging for Na metal batteries will be discussed and reported in a separate paper.