Singlet Oxygen during Cycling of the Aprotic Sodium-O2 Battery.

Aprotic sodium-O2 batteries require the reversible formation/dissolution of sodium superoxide (NaO2 ) on cycling. Poor cycle life has been associated with parasitic chemistry caused by the reactivity of electrolyte and electrode with NaO2 , a strong nucleophile and base. Its reactivity can, however, not consistently explain the side reactions and irreversibility. Herein we show that singlet oxygen (1 O2 ) forms at all stages of cycling and that it is a main driver for parasitic chemistry. It was detected in- and ex-situ via a 1 O2 trap that selectively and rapidly forms a stable adduct with 1 O2 . The 1 O2 formation mechanism involves proton-mediated superoxide disproportionation on discharge, rest, and charge below ca. 3.3 V, and direct electrochemical 1 O2 evolution above ca. 3.3 V. Trace water, which is needed for high capacities also drives parasitic chemistry. Controlling the highly reactive singlet oxygen is thus crucial for achieving highly reversible cell operation.

The need to advance batteries beyond the limits of current technology in terms of energy, sustainability, and cost has generated immense interest in rechargeable aprotic metal–O2 batteries.1 They store charge at the cathode by reversibly forming/decomposing Li2O2 or NaO2 in the Li–O2 or Na–O2 cell, respectively. Despite the lower theoretical specific capacity of 488 mAh g−1 of NaO2 and lower voltage of 2.27 V (vs. 1168 mAh g−1 at 2.96 V for Li–O2), the Na–O2 cell has been reported to have significant advantages over Li–O2 with respect to rechargeability and energy efficiency.2 Realizing the Na–O2 cell, however, still faces many challenges in practice, including the Na‐metal anode, lower than theoretical cathode capacity, and perhaps most importantly, insufficient cycle life associated with parasitic chemistry (that is, side reactions) at the cathode.2, 3

Since the very first papers published on Na–O2 batteries, superoxide has been perceived responsible for parasitic chemistry with electrolyte and electrode.1c, 3, 4 A key measure for parasitic chemistry is the ratio of e− passed to O2 consumed/evolved. During discharge, this ratio is typically at the ideal value of one despite approximately 5 % of the expected NaO2 being missing and replaced by side products, such as Na2CO3, Na acetate, and Na formate. During subsequent resting and charge, more of these side products form and typically the e−/O2 ratio deviates by several percent from one.2a, 3c, 4b, 5 Although less side products form than in the Li–O2 cell, cyclability is similarly poor: restricted capacity can often be maintained for up to hundreds of cycles albeit at the expense of true energy, but at full discharge cells fail within some 10 cycles and capacity fading becomes significantly worse with rising charge cut‐off voltage.1c, 2a,2c,2d, 3a,3d, 6

Superoxide's potential reactivity towards organic substrates stems from its nucleophilicity, basicity, and radical nature, which may cause nucleophilic substitutions, H+ and H‐atom abstraction.7 Theoretical work, however, has revealed that all these reactions are unlikely with commonly used ether electrolytes owing to the high activation energies and strong endothermicity.8 Also, the extent of parasitic chemistry at the various stages of cycling does not match the abundance of superoxide. Overall, the reactivity of superoxide cannot consistently explain the observed parasitic chemistry, which thus may only be inhibited with better knowledge about reactive species and their formation mechanism.

Herein we show that singlet oxygen (1Δg or 1O2), the first excited state of ground state triplet oxygen (3Σg − or 3O2), is the reactive species, which drives parasitic reactions. It is generated at all stages of cycling in relative quantities resembling the occurrence of side reactions: relatively little during discharge, rest, and low charging voltages, and strongly increasing amounts at higher charging voltages.

Methods to sensitively detect 1O2 rely on chemical probes, which selectively react with 1O2. Probes include spin traps and fluorophores, which become EPR active or fluorescing upon reaction with 1O2.9 However, these probes are not electrochemically inert in the relevant potential range between approximately 2 and 3.6 V versus Na/Na+ and may react with superoxide. Previously, we have shown that 9,10‐dimethylanthracene (DMA) fulfills all the requirements;10 it is stable in contact with superoxide, reacts rapidly with 1O2 to its endoperoxide (DMA‐O2), and has a sufficiently wide potential window (Figure S1 and Scheme S1 in the Supporting Information). 1O2 can either be monitored by DMA consumption via its absorbance or fluorescence between 300 and 500 nm (Figure S2), or by detecting DMA and DMA‐O2 via HPLC.

To probe whether 1O2 would form during discharge and charge we constructed Na–O2 cells as detailed in the Methods Section of the Supporting Information with 0.5 m NaSO3CF3 (NaOTf) in diglyme containing 40 ppm H2O and 30 mm DMA as the electrolyte. Water was used because it is required as a phase‐transfer catalyst to allow large NaO2 particles to grow.2c To specifically probe reactions at the cathode and to exclude unwanted reactions of electrolyte components with a Na‐metal anode, we used the Na intercalation material Na3−V2(PO4)3 as the counter electrode. It operates at approximately 3.4 V versus Na/Na+ and is thus well within the stability window of the electrolyte.11 Cells were run at constant current to various states of discharge or charge, then stopped, and the electrolyte extracted and analyzed using HPLC (Figure 1).

Figure 1

a) Load curve for galvanostatic cycling of a carbon‐paper cathode at 90 μA cm−2 in 0.5 m NaOTf in diglyme containing 40 ppm H2O and 30 mm DMA. b) HPLC runs of electrolyte samples taken at the points ②, ③, and ④ in (a), showing 1O2 to have formed as indicated by the conversion of DMA into DMA‐O2. The blank sample was extracted from a cell that rested for 5 h.

The cycle curve shows flat discharge and charge plateaus at approximately 2.2 and 2.3 V, respectively, with a sharp rise in voltage when recharge approaches about 55 % (Figure 1 a). X‐ray diffraction confirms that the cube‐shaped discharge product is NaO2 (Figure S3, S4) in accord with many previous reports on similar cells.1b,1c, 2a,2b, 3a After discharge or recharge to 2.8 and 3.65 V, 4.1, 4.3, and 7.2 %, respectively, of the DMA at the sample points ②, ③, and ④ was converted into DMA‐O2 (Figure 1 b). The amounts of DMA‐O2 equate to 1.4, 1.3, and 2.1 % of the O2 involved based on the charge at these stages of cycling to have turned into 1O2. 1O2 is thus generated both on discharge and charge, and charging to higher voltage yields significantly more 1O2.

Operando spectroscopy is well suited to probe in detail onset potentials and reaction rates. An operando fluorescence set‐up as detailed in the Supporting Information was constructed. Briefly, it consisted of a gas‐tight quartz cuvette with a carbon paper working electrode and Na3−V2(PO4)3 as counter and reference electrode and an O2 filled head space. The O2‐saturated electrolyte was the same as before except for a lower DMA concentration (1.6×10−5  m), which best suits fluorescence detection (see the Supporting Information). Excitation and emission wavelengths were chosen according to the maxima in the respective spectra (Figure S2).

While cells as in Figure 1 comprising a porous separator sandwiched between working and counter electrode typically yield NaO2, cycling the cathode in the operando setup yielded Na2O2⋅H2O as can be inferred from the high charging voltage (Figure S5). Rather similar thermodynamics of NaO2 and Na2O2 ( =2.27 V, =2.33 V) were used to explain that the Na–O2 cell could yield both as a discharge product.1b, 3b Proton sources decisively influence the discharge product, yet the precise governing factors are unknown.1b The reason for peroxide rather than superoxide as a discharge product in the operando setup may be the large electrolyte‐to‐electrode volume ratio and the stirred electrolyte, both of which are essential for the method. The electrolyte after discharge in the operando setup also contains DMA‐O2, revealing that discharge to Na2O2 also forms 1O2 (Figure S6).

To investigate charging of NaO2, we first discharged a cathode in the standard cell and then placed it in the operando cell. Figure 2 shows the DMA concentration and consumption rate for stepwise potentiostatic charging up to 3.7 V. As soon as the cell was polarized to charging potentials, DMA was consumed at an initially moderate rate, which sharply increased beyond a voltage of about 3.3 V. Thermodynamically, 1O2 can be expected to form upon electrochemically oxidizing NaO2 above . With 0.97 eV energy difference between 1O2 and 3O2 the thermodynamic threshold to evolve 1O2 can be estimated to be 3.24 V. 1O2 forming at high rate above approximately 3.3 V can thus be explained by the reaction NaO2→Na++e−+1O2 beyond the thermodynamic threshold of 3.24 V. Up to 3.3 V the 1O2 fraction is around 0.75 to 1.1 % of the expected O2 (Figure S7). Beyond 3.3 V this fraction increases steadily to around 4 %.

Figure 2

Operando fluorescence detection of 1O2 during potentiostatic charging of a Na–O2 cathode. The carbon‐paper cathode was first discharged in a Swagelok type cell to 75 mAh cm−2 in 0.5 m NaOTf in diglyme containing 40 ppm H2O and then introduced into the operando setup containing the same electrolyte and additionally 1.6×10−5  m DMA. a) Voltage and current profile. b) DMA concentration. c) DMA consumption rate.

Previously it was reported that most parasitic chemistry occurs during discharge and less on charge.2a, 3a Charging above around 3.5 V caused the number of O2 per e− evolved to deviate more significantly from one than at lower voltages. Correlating these findings with the results in Figure 1, Figure 2, and Figure S7 suggests that parasitic chemistry in the Na–O2 cathode is closely related to 1O2 formation since the extent of side reactions follows the occurrence of 1O2: 1O2 is formed at a higher rate during discharge than at charging below 3.3 V and at much higher rate above 3.3 V. In quantitative terms the 1.4 % 1O2 during discharge and an approximately 95 % yield of NaO2 suggests a significant fraction of the side products stem from 1O2. Measured 1O2 amounts represents a lower bound since at the low DMA concentration not necessarily all 1O2 will react with DMA and may undergo other decay routes, such as reactions with cell components.

While thermodynamics directly explains how 1O2 forms when NaO2 is oxidized above 3.3 V, its formation during discharge and recharge below 3.3 V is more surprising. With the Li–O2 cell, disproportionation of the LiO2 intermediate to Li2O2 was suggested as a major 1O2 source during discharge and low charging voltages.10 This pathway is inactive in the Na–O2 cell since discharge stops at NaO2. When, however, water or other proton sources are present, oxygen reduction will lead to HOO..12 This soluble species enables proton‐assisted phase‐transfer catalysis, which allows the typical micron‐sized NaO2 cubes to grow on discharge and to be oxidized on charge; HOO. is soluble and mobile in the electrolyte and precipitates NaO2 via the metathesis reaction HOO.+Na+ NaO2+H+.2c, 13 HOO., however, can be reduced by superoxide or disproportionate according to Equation (1), (2),12 which both have been shown to release 1O2.14

That water is driving these reactions is supported by significantly less 1O2 being formed with dry electrolyte as compared to wet electrolyte (Figure S8).

These reactions may not only take place during discharge, where new superoxide is generated, but also at rest since the metathesis reaction is an equilibrium and allows NaO2 to redissolve into HOO.. A number of studies have shown instability of NaO2 upon prolonged rest.3c, 4, 5 Although details vary, they generally report gradual conversion of NaO2 into Na2O2⋅H2O accompanied by electrolyte degradation to form NaOH, Na2CO3, and organic compounds including Na formate and acetate. Explanations included nucleophilic attack, H+ and H‐atom abstraction by superoxide. To probe how 1O2 is involved in forming the parasitic products we first discharged electrodes in DMA‐free electrolyte and then placed them for various times in electrolyte containing 30 mm DMA. X‐ray diffraction after storage shows decreased NaO2 and side products to have formed in accord with previous studies (Figure S9). The electrolyte was then analyzed for DMA‐O2 and the electrodes were analyzed for Na2CO3 content by immersing them in 0.1 m H2SO4 and monitoring the CO2 evolved by mass spectrometry (Figure 3). 1O2 as indicated by the presence of DMA‐O2 continuously increases with resting time as does the Na2CO3 content, which serves as an indication for the extent of parasitic products.

Figure 3

1O2 generation and parasitic chemistry during prolonged contact of the discharge product NaO2 with the electrolyte. Cells were first discharged in DMA‐free electrolyte (0.5 m NaOTf in diglyme containing 40 ppm H2O), then the cathodes washed and immersed in the same electrolyte containing additionally 30 mm DMA. After the times indicated the electrolyte was analyzed for DMA‐O2 and the electrodes for Na2CO3. The dotted lines are linear fits versus the square root of rest time.

DMA‐O2 and Na2CO3 roughly increase proportional to the square root of the resting time, which means that the formation rate decreases with time. The formed parasitic products thus partly passivate the NaO2 surface and slow down further decomposition. A larger amount of DMA‐O2 than additional Na2CO3 formed, which can be explained by two phenomena. First, 1O2 trapped by DMA does not react with cell components and would thus have led to even more parasitic products without DMA. Second, other compounds than Na2CO3 will form which are not expelled by acid. Overall, Figure 3 shows that the previously reported electrolyte degradation upon resting is significantly driven by 1O2 formation.

Pathways to 1O2 during discharge, rest, and charge are summarized in Scheme 1. We suggest proton mediated superoxide disproportionation or reduction to interfere with the metathesis equilibrium HOO.+Na+ NaO2+H+ as a universal path to 1O2 during all stages of cycling. Upon charge at voltages exceeding the thermodynamic threshold of 3.24 V 1O2 evolves by direct electrochemical oxidation according to NaO2→Na++e−+1O2.

Scheme 1

Pathways leading to 1O2 during discharge, rest, and charge. “Disp”=disproportionation and “red”=reduction.

In conclusion we demonstrate that singlet oxygen forms at the cathode of the Na–O2 cell during all stages of cycling and during rest, and accounts for a significant fraction of the side‐products formed. 1O2 is thus a major hurdle in cycling the Na–O2 cell via the reversible formation/decomposition of NaO2. Water takes up an ambivalent role by being required for high capacity via solution‐mediated growth and dissolution of large NaO2 deposits, and at the same time appearing to be the main driver for 1O2 formation during discharge and at low charging voltages. Future work should thus focus on finding ways to either prevent 1O2 formation or to efficiently eliminate it.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.