Electrochemical Oxidation of Lithium Carbonate Generates Singlet Oxygen.

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, Li2 CO3 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 Li2 CO3 →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 (1 O2 ) forms upon oxidizing Li2 CO3 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 1 O2 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 Li2 CO3 .

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

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 Li1−FePO4 (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:

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.

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.