Electrochemical Oxidation of Lithium Carbonate Generates Singlet Oxygen

Abstract

Solid alkali metal carbonates are universal passivation layer components of intercalation battery materials and common side products in metal‐O₂ batteries, and are believed to form and decompose reversibly in metal‐O₂/CO₂ cells. In these cathodes, Li₂CO₃ decomposes to CO₂ when exposed to potentials above 3.8 V vs. Li/Li⁺. However, O₂ evolution, as would be expected according to the decomposition reaction 2 Li₂CO₃→4 Li⁺+4 e⁻+2 CO₂+O₂, is not detected. O atoms are thus unaccounted for, which was previously ascribed to unidentified parasitic reactions. Here, we show that highly reactive singlet oxygen (¹O₂) forms upon oxidizing Li₂CO₃ in an aprotic electrolyte and therefore does not evolve as O₂. These results have substantial implications for the long‐term cyclability of batteries: they underpin the importance of avoiding ¹O₂ in metal‐O₂ batteries, question the possibility of a reversible metal‐O₂/CO₂ battery based on a carbonate discharge product, and help explain the interfacial reactivity of transition‐metal cathodes with residual Li₂CO₃.

Energy storage in Li‐based batteries is limited by the cathode, which has triggered intense research efforts to increase cathode capacity and/or voltage.1 Candidate approaches include Li‐stoichiometric2 and Li‐rich3 transition‐metal oxide (TMO) intercalation cathodes, which have higher voltage and capacity than currently used cathodes, and metal‐O₂ or metal‐O₂/CO₂ cathodes,1, 4 which have lower voltage but substantially higher theoretical capacity. Making high‐voltage TMOs viable requires increasing the reversible potential window through understanding the high‐voltage instabilities of intercalation materials and electrolytes.1 Much recent work has revealed an intimate interdependence of electrolyte decomposition, surface species formation/decomposition, and TMO bulk and surface reconstruction.2d, 3d, 5 In particular, it was recently found that the outgassing of CO₂ during the first cycle in Li‐ion batteries is mostly governed by residual Li₂CO₃, which in turn affects O₂ evolution from the TMO lattice.5b With respect to Li‐O₂ batteries, Li₂CO₃ is an unwanted parasitic product, which hampers rechargeability, accumulates on cycling, and hence causes poor energy efficiency and cycle life.1, 4a–4f The burden of Li₂CO₃ formation was seemingly made use of in rechargeable metal‐O₂/CO₂ batteries based on the observation that Li₂CO₃ can be electrochemically decomposed.4f–4j, 6

Thus Li₂CO₃, be it a trace or main component, plays a central role in considerations of cyclability and stability for a large fraction of future Li battery systems, and understanding its electrochemical oxidation is paramount for further development. While it is clear that Li₂CO₃ decomposition evolves CO₂, the fate of the third O atom in CO₃ ²⁻ has been an enduring open question since no O₂ evolves, although this would be expected from the formal oxidation reaction:3e, 4c,4f–4h,4j, 5b

$${\displaystyle 2\mathrm{Li}{}_2\mathrm{CO}{}_3\to 4\mathrm{Li}{}^{+}+4\mathrm{e}{}^{-}+2\mathrm{CO}{}_2+\mathrm{O}{}_{2}E{}^{\circ }=3.82\mathrm{V}\mathrm{vs}.\mathrm{Li}/\mathrm{Li}{}^{+}}$$ (1)

Previous explanations have proposed the formation of superoxide or “nascent oxygen”, which could react with cell components in a reaction path involving carbon,4f, 6 without, however, definite proof for these mechanisms. Herein, we provide compelling evidence that the electrochemical oxidation of Li₂CO₃ forms highly reactive ¹O₂, which, through a parasitic reaction of ¹O₂ with battery components, explains the absence of O₂ evolution. Given its exceptional reactivity, the formation of ¹O₂ has far‐reaching implications for TMO surface reactivity and coupled parasitic reactions upon recharging metal‐O₂ and metal‐O₂/CO₂ batteries.

¹O₂ may be detected using chemical probes, which react specifically with ¹O₂ and can be detected spectroscopically by measuring the disappearance of the probe and/or the appearance of the adduct. Reported probes include fluorophores or spin traps, which may be detected by fluorescence “switch on/off” or by EPR spectroscopy.7 However, these probes are typically electrochemically unstable above 3.5–3.7 V vs. Li/Li⁺ and do not allow access to the relevant Li₂CO₃ oxidation potential range above 3.8 V. Previously, we have shown that 9,10‐dimethylanthracene (DMA) fulfills these requirements: it rapidly forms the endoperoxide (DMA‐O₂) in the presence of ¹O₂; both DMA and DMA‐O₂ are electrochemically stable beyond 4 V (Figure S1); and DMA is also stable against superoxide, another possible reactive oxygen species. In other words, exposing DMA to superoxide does not form DMA‐O₂, which otherwise would be falsely assigned to the presence of ¹O₂.8 To further confirm that DMA‐O₂ forms only with ¹O₂ but not with other possibly reactive O‐containing species, we exposed the electrolyte with DMA separately to Li₂CO₃, O₂, CO₂, and Li₂O₂ and did not observe DMA‐O₂ (Figure S2). The same holds true for DMA exposed to Li₂O₂ with CO₂, which forms a peroxodicarbonate, a possible intermediate of Li₂CO₃ oxidation.9 Together, these results confirm that DMA→DMA‐O₂ conversion is a sensitive and selective method to detect ¹O₂ in the cell environment.

To probe whether ¹O₂ forms upon oxidizing Li₂CO₃, we constructed electrochemical cells with Li₂CO₃‐packed working electrodes as detailed in the Methods section in the Supporting Information. Li₂CO₃ was ball‐milled with carbon black to ensure intimate contact between the two and the resulting powder was used to form composite electrodes using PTFE binder. To specifically probe reactions at the working electrode and to exclude unwanted reactions of the electrolyte with a Li metal anode, we used Li_1−xFePO₄ (E°=3.45 V vs. Li/Li⁺) as the counter and reference electrode. First, we established the onset potential of Li₂CO₃ oxidation using a potential sweep measurement in an online electrochemical mass spectrometry (OEMS) setup to follow the gases evolved. Figure 1 shows CO₂, O₂, CO, and H₂ evolution in comparison to the electron consumption rate. CO₂ evolution commences at around 3.8 V, with a ratio of approximately 2 e⁻/CO₂ observed at higher voltages. Note that capacitive current accounts for the initial electron consumption rate above open circuit and causes the electron consumption rate to remain slightly higher than the CO₂ evolution rate. The onset of CO₂ evolution at 3.8 V is in accordance with the equilibrium potential of Reaction 1 (E°=3.82 V vs. Li/Li⁺).4c, 6 Consistent with numerous studies, O₂ was not detected throughout charging.4c,4g,4h, 5b H₂ and CO evolution is observed above 4.2 V during the anodic scan of the Li₂CO₃‐packed electrodes, but no gas evolution is observed below 4.5 V from blank carbon black electrodes (Figure S3). Absence of CO₂ when a blank electrode is charged proves Li₂CO₃ oxidation to be the CO₂ source in Figure 1. The comparison of the blank and Li₂CO₃‐packed electrode also indicates that the H₂ evolution observed (Figure 1) has to originate from a parasitic electrolyte degradation reaction induced by Li₂CO₃ oxidation, since the electrolyte otherwise appears stable at Li₂CO₃‐free electrodes until at least 4.5 V.

Figure 1.

CO₂, O₂, CO, and H₂ evolution from carbon black/Li₂CO₃/PTFE (9:1:1, m:m) composite electrodes during a linear potential scan at 0.14 mV s⁻¹ in 0.1 m LiTFSI in TEGDME under an Ar atmosphere.

To examine whether the highly reactive ¹O₂ forms and could thus explain the absence of O₂ release, we constructed cells with the same Li₂CO₃ working electrodes and 0.1 m LiTFSI in dimethoxyethane (DME) containing 30 mm DMA as the electrolyte. Cells were held at various charging potentials until a capacity of 0.064 mAh was reached. The electrolyte was then extracted and subjected to HPLC and ¹H NMR analysis (Figure 2).

Figure 2.

a) HPLC analysis of the electrolyte after polarizing carbon black/Li₂CO₃/PTFE (9:1:1 m:m) composite electrodes at the indicated potential to reach a capacity of 0.064 mAh in 0.1 m LiTFSI in DME that contained 30 mm DMA. ¹H NMR confirms DMA‐O₂ to elute at 2.6 min (Figures S2, S5). b) ¹H NMR spectra of the same electrolyte samples. Reference measurements are shown with the starting electrolyte (labeled as DMA) and electrolyte where the DMA was fully converted into DMA‐O₂ by in situ photogenerated ¹O₂ (labeled as DMA‐O₂) as described in the Supporting Information.

HPLC analysis showed that DMA‐O₂ formed at all charging voltages from 3.8 V onwards (Figure 2 a). Blank measurements, where electrodes without Li₂CO₃ were polarized analogously, did not yield DMA‐O₂ (Figure S4). ¹H NMR analysis of the samples confirmed the presence of DMA‐O₂ at these voltages (Figure 2 b, S6). The HPLC and NMR results confirm that electrochemical oxidation of Li₂CO₃ forms ¹O₂ from the onset of oxidation at 3.8 V.

Figure 3 relates the amount of ¹O₂ formed to the charge passed in the reaction:

$${\displaystyle 2\mathrm{Li}{}_2\mathrm{CO}{}_3\to 4\mathrm{Li}{}^{+}+4\mathrm{e}{}^{-}+2\mathrm{CO}{}_2+{}^1\mathrm{O}{}_2}$$ (2)

Figure 3.

Amount of ¹O₂ (as quantified by HPLC as DMA‐O₂) relative to the charge passed in Equation (2) at different charging potentials. Values represent lower bounds since not all ¹O₂ may react to DMA‐O₂ or the electrolyte may be incompletely extracted.

A maximum of one ¹O₂ could be produced per four electrons. ¹O₂ formed at all probed voltages to an extent well above 50 % of the 4 e⁻/¹O₂ maximum limit. The amount of ¹O₂ must, however, be inferred with caution from the measured amount of DMA‐O₂ and represents a lower bound of the true value. This is because not all ¹O₂ will react with DMA, but may decay along other routes. Furthermore, the electrolyte may be incompletely extracted and thus result in an artificially low ¹O₂ value. At higher voltages (e.g., 4.2 V), DMA‐O₂ could degrade to a minor extent, as shown in Figure S1 in the Supporting Information, which may explain the observed lower yield of DMA‐O₂ at 4.2 V compared to 4.05 V in Figure 3. Overall, the values in Figure 3 suggest that the majority, if not all, of the “missing O₂” from the electrochemical oxidation of Li₂CO₃ forms ¹O₂ and is thus not detected in the gas phase.

The complete lack of O₂ evolution during oxidation of Li₂CO₃ (Figure 1) implies that the formed ¹O₂ reacts with cell components rather than being, even in part, deactivated to ³O₂. We therefore investigated the use of a ¹O₂ quencher, which deactivates ¹O₂ to ³O₂,10 to possibly promote ³O₂ evolution. A variety of quenchers have been reported, including azides and aliphatic amines.10, 11 We have previously shown that 1,4‐diazabicyclo[2.2.2]octane (DABCO) is effective in non‐aqueous environments.8a For use during electrochemical oxidation of Li₂CO₃, however, the electrochemical stability of the quenchers is problematic, since DABCO and other quenchers (e.g., LiN₃) are electrochemically oxidized at approximately 3.5–3.6 V (Figure S7).12 Nevertheless, diffusion of fresh quencher from the separator may counterbalance quencher oxidation at the working electrode and thus may show some quenching efficiency. Figure 4 shows CO₂ and O₂ evolution during an OEMS measurement similar to Figure 1, but with an electrolyte that contained 30 mm DABCO. DABCO oxidation accounts for the anodic process that onsets at around 3.6 V and peaks at 3.9 V. As before, CO₂ evolution starts at around 3.8 V and reaches a rate close to 2 e⁻/CO₂. Intriguingly, O₂ evolution does start together with CO₂ evolution at around 3.8 V with a similarly growing rate as the voltage rises. This result further corroborates ¹O₂ formation and also shows that if a suitable quencher can be found, then Li₂CO₃ could be oxidized without the detrimental effects of ¹O₂.

Figure 4.

CO₂ and O₂ evolution from Super P/Li₂CO₃/PTFE (9:1:1 m:m) composite electrodes during a linear potential scan at 0.14 mV s⁻¹ in 0.1 m LiTFSI in TEGDME that contained 30 mm DABCO.

Detection of ¹O₂, and ³O₂ when a quencher is present, implies that a mechanism of Li₂CO₃ oxidation involves the formation of O−O bonds. In analogy to carbonate oxidation in aqueous media,13 it has been suggested that Li peroxodicarbonate (LiO₂COOCO₂Li) forms as an intermediate.4h Such an intermediate has been questioned on the basis that 1) CO₃ ²⁻ is poorly soluble and would thus lack mobility to combine to peroxodicarbonate and 2) the high charge density of the peroxodicarbonate anion (⁻O₂COOCO₂ ⁻) would not allow O−O bond formation or would lead to immediate bond cleavage.4c, 14 However, neither large carbonate mobility nor dissociation are required and a mechanism via a peroxodicarbonate intermediate can be proposed (Figure S8a) and rationalized based on previous reports.4i, 15 Formally, peroxodicarbonate can form through a 1 e⁻ oxidation/Li⁺ extraction of two Li₂CO₃ to form two LiO₂CO^. moieties (2), which combine to LiO₂COOCO₂Li (3). Within the Li₂CO₃ crystal structure (Figure S8b), adjacent carbonate moieties appear to be sufficiently close to form O−O bonds once an e⁻ and a Li⁺ is extracted in each. Mobility of the intermediates or even dissociation from the crystal lattice is thus not required. A DFT study on the oxidation of Li₂CO₃ surfaces has shown that after first oxidation/Li⁺ extraction, further Li⁺ extractions are energetically most favorable at adjacent carbonate moieties, which makes their recombination likely.15 Such recombination within the crystal lattice is also supported by DFT work on the formation of Li₂CO₃ via peroxodicarbonate, which yields adjacent carbonate moieties within the Li₂CO₃ lattice.4i According to the same work, the O−O bond in LiO₂COOCO₂Li is strongly stabilized by coordination with Li⁺ ions in comparison to ⁻O₂COOCO₂ ⁻, which is unlikely to form in a nonaqueous environment. A possible ongoing pathway to form ¹O₂ is shown in Figure S8a. Further oxidation and decarboxylation could yield LiCO₄ (4; Figure S8a), which then in turn could yield ¹O₂. Clarification of the exact mechanism, however, will need further computational or/and experimental work.

In conclusion, by using a selective ¹O₂ trap and online mass spectrometry, we have shown that electrochemical oxidation of Li₂CO₃ in a nonaqueous environment yields up to stoichiometric amounts of ¹O₂ according to the reaction 2 Li₂CO₃→4 Li⁺+4 e⁻+2 CO₂+¹O₂. This explains the absence of O₂ evolution, which has been a long‐standing conundrum and a cause for much speculation regarding potential reactive oxygen species. The reaction proceeds from an onset potential of approximately 3.8 V, which is close to its thermodynamic value of 3.82 V. When a ¹O₂ quencher is present, part of the formed ¹O₂ could be evolved as ³O₂. Li₂CO₃ is a universal passivating agent in Li‐ion battery cathodes and decisive in interfacial reactivity. Li₂CO₃ is also a common side product in Li‐O₂ cathodes, as well as the targeted discharge product in Li‐O₂/CO₂ batteries, where it then needs to be oxidized on charge to form a reversible system. Our results thus strongly suggest that Li₂CO₃ formation, even at impurity levels, will have a deleterious affect on the stability of all Li batteries where electrodes operate beyond 3.8 V vs. Li/Li⁺, which includes most currently studied cathodes. Strategies to avoid ¹O₂ formation or the presence of Li₂CO₃ during battery operation are therefore warranted.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

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

N. Mahne, S. E. Renfrew, B. D. McCloskey, S. A. Freunberger, Angew. Chem. Int. Ed. 2018, 57, 5529.