## Introduction

Lithium carbonate (Li2CO3) is involved in many electrochemical systems, such as lithium-oxygen (Li-O2) batteries1,2,3,4,5,6,7,8,9,10,11,12,13,14,15, lithium-carbon dioxide (Li-CO2) batteries16,17,18,19,20,21,22,23,24,25,26,27,28,29,30, and lithium-ion (Li-ion) batteries31,32,33,34,35,36,37,38,39,40,41,42,43. Li2CO3 has extremely low ionic and electronic conductivity due to its wide bandgap44,45. In Li-O2 batteries, Li2CO3 mainly results from the side reactions of the reduced oxygen species attacking the electrolytes1,2,3. It not only passivates the electrode surface and polarizes the cell but also consumes the electrolyte, leading to electrolyte depletion and premature cell death. Therefore, Peng et al. referred to Li2CO3 as the “Achilles’ Heel” because it dominates the electrochemical performance of cells3. The accumulation of Li2CO3 during cycling has to be well-addressed and resolved in the pursuit of high-performance Li-O2 batteries. In Li-CO2 batteries, Li2CO3 is the main desirable discharge product, but the Li2CO3 decomposition during charging has sluggish kinetics and requires a large overpotential. Many efforts have been devoted to designing highly efficient catalysts to reduce the large overpotential18,19,20,21,22,23,24,25,26,27,28,29. In both Li-O2 batteries and Li-CO2 batteries, Li2CO3 in the composite electrode needs to be oxidatively decomposed during the charging process, otherwise, it passivates the electrode surface and kills the cells. However, the mechanism of Li2CO3 decomposition is still unclear and this would seriously hinder the research progress. For instance, the role of carbon in the charging process is still under debate. In Li-ion batteries, Li2CO3 is one of the main components of the solid electrolyte interphase of the anode and exists as a surface contaminant present on lithium transition metal oxides used in the cathode thus it influences the cell performance32. For instance, the lithium transition metal oxides cathode materials are usually covered with a layer of Li2CO3 due to the residual lithium precursors reacting with CO2 from the ambient atmosphere33,34,35. Very recently, McCloskey and co-workers have studied the Li2CO3 decomposition mechanism on Li-ion cathodes. Using isotopic labeling, they found that when Li2CO3 is present at the cathode surface, organic fragments containing diatomic oxygen are formed on the cathode surface during the charging process above 4.2 V versus Li+/Li and the diatomic oxygen within these fragments mainly originates from the lithium transition metal oxides lattice and only a minor fraction originates from the Li2CO3 itself33,34,35. In summary, the decomposition of Li2CO3 is so important that it determines the electrochemical performance in many systems. However, its mechanisms are still controversial and not yet well understood, even in the Li-CO2 cells. Li2CO3 decomposition during the charging process is generally divided into two types, chemical routes, and electrochemical routes41,42. Recently, Freiberg et al claimed that Li2CO3 decomposition in lithium hexafluorophosphate (LiPF6)-ethylene carbonate (EC)-ethylmethyl carbonate (EMC) electrolyte follows a chemical route reacting with H+, which is induced by electrolyte oxidation at >4.6 V41. In contrast, Mahne et al. claimed that Li2CO3 decomposition is an electrochemical process33,46. Here, our results show that the electrolyte salt affects the route and the chemical reaction is likely caused by LiPF6, which will be discussed later.

So far, several electrochemical mechanisms have been proposed and four possible reaction pathways are summarized below:

$$\begin{array}{cc}{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}-2{{{{{{\rm{e}}}}}}}^{-}\,\to \,2{{{{{{\rm{Li}}}}}}}^{+}+{{{{{{\rm{CO}}}}}}}_{2}+1/2\,{{{{{{\rm{O}}}}}}}_{2} & {{{{{{\rm{E}}}}}}}_{{{{{{\rm{rev}}}}}}}=3.8\,{{{{{\rm{V}}}}}}\end{array}$$
(1)
$$\begin{array}{cc}{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}-2{{{{{{\rm{e}}}}}}}^{-}\to 2{{{{{{\rm{Li}}}}}}}^{+}+{{{{{{\rm{CO}}}}}}}_{2}+1/2{}^{1}{{{{{\rm{O}}}}}}_{2} & \,{{{{{{\rm{E}}}}}}}_{{{{{{\rm{rev}}}}}}}=4.1\,{{{{{\rm{V}}}}}}\end{array}$$
(1b)
$$\begin{array}{cc}2{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}+{{{{{\rm{C}}}}}}-4{{{{{{\rm{e}}}}}}}^{-}\to 4{{{{{{\rm{Li}}}}}}}^{+}+3{{{{{{\rm{CO}}}}}}}_{2} & {{{{{{\rm{E}}}}}}}_{{{{{{\rm{rev}}}}}}}=2.8\,{{{{{\rm{V}}}}}}\end{array}$$
(2)
$$2{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}-3{{{{{{\rm{e}}}}}}}^{-}\,\to \,4{{{{{{\rm{Li}}}}}}}^{+}+2{{{{{{\rm{CO}}}}}}}_{2}+{{{{{{{\rm{O}}}}}}}_{2}}^{\cdot -}$$
(3)

In Eq. (1a and b), both CO2 and O2 formed and it takes 2e per CO2 molecule16,17. The only difference is that O2 forms as a triplet O2 in Eq. (1a) and as a singlet O2 (1O2) in Eq. (1b), respectively. 1O2 has been detected at a charging voltage above 3.8 V in the oxidation of Li2CO3, by using high-performance liquid chromatography (HPLC) and nuclear magnetic resonance spectrometry (1H NMR) analysis46. In Eq. (2), carbon and Li2CO3 were oxidized together to form CO2 via a 4e process. That is a common mechanism proposed for the charging process of Li-CO2 cells. In fact, Eq. (2) is unlikely to be an elemental reaction, which will be discussed in the text later. In Eq. (3), oxygen is released in the form of the superoxide radical. Qiao et al.27 used in situ surface-enhanced Raman spectroscopy to observe the dimethyl sulfone during the charging process and they explained that dimethyl sulfone is attributed to the nucleophilic attack on DMSO solvent from reduced oxygen species (superoxide radicals etc.).

Despite the above progress, researchers mainly focus on Li2CO3 and little attention has been paid to the carbon in the charging process. Carbon is always added to the composite electrodes as the conductive additives, however, it is always neglected and its role in Li2CO3 decomposition has not been considered yet. Overall, the mechanism of Li2CO3 decomposition is still under debate and the role of carbon is mysterious.

Here, we labeled the Li2CO3 and carbon substrate with 13C-isotope and qualitatively analyzed the gas products from the decomposition of Li2CO3, carbon, and electrolyte, respectively. We quantified the gas evolution, particularly CO, during the charging process using an in situ differential electrochemical mass spectroscopy-gas chromatography (DEMS-GC) coupling system. We found that Li2CO3 decomposition is mainly an electrochemical process rather than a chemical process induced by electrolyte oxidation. Li2CO3 decomposes to CO2 only, but no CO nor O2. The oxygen from Li2CO3 is released as highly reactive 1O2, which further attacks the electrolyte and carbon substrate in the composite electrodes to form CO2 and CO.

## Results and discussion

To study the oxidative decomposition process of Li2CO3, a cell with a Li2CO3-carbon composite electrode was constructed and charged. Li2CO3 was electro-oxidized and the gas evolution was quantified. Because both CO and N2 have the same mass-to-charge ratio of 28 (m/z = 28), the mass spectrometer typically used in a DEMS system lacks sufficient mass resolution to distinguish the contribution from CO (m/z = 28.0104) and N2 (m/z = 28.0140). Although GC could separate the CO and N2, it cannot distinguish and quantify the 13CO2/13CO2 and 12CO/13CO. Therefore, an in situ DEMS-GC coupling system (Supplementary Fig. 1) was used to quantify the evolution of 12CO, 12CO2, 13CO, and 13CO2. The details of the experiments are described in Methods. As shown in Supplementary Fig. 2, after calibration, the CO evolution signals from DEMS and GC experiments are consistent, which provides a reliable amount of CO in the following experiments.

Super P carbon (Timcal) was ball milled with Li2CO3 with a mass ratio of 1:1. A Li2CO3-Super P (1:1) composite electrode was prepared to construct a cell with 1 M lithium bis(trifluoromethane-sulphonyl)imide (LiTFSI)—tetraglyme electrolyte as stated in Methods and the cells were charged by linear sweep voltammetry (LSV). As shown in Fig. 1a, the anodic current for oxidation reaction and gas evolution started at 3.9 V (vs. Li+/Li, all potentials in the text below are versus Li+/Li), which is consistent with the thermodynamic decomposition potential of Li2CO3 (3.82 V according to Eq. 1). A control experiment without Li2CO3 was carried out (Supplementary Fig. 3). The onset potential of electrolyte decomposition is at 4.3 V. The background current of carbon/electrolyte oxidation at 3.8 V is 10-fold smaller than Li2CO3 decomposition (Fig. 1a) and there is no CO2 or O2 evolution. In Fig. 1a, a large amount of CO2 and CO were identified as the gas products, which confirms the decomposition of Li2CO3, consistent with the literature16,17. According to the mass loading of the electrode (see Methods), ~60% of the preloaded Li2CO3 decomposed eventually. The ratio e-/CO2 for Fig. 1a is 2.1, close to 2e per CO2 (Supplementary Table 1), suggesting it is an electrochemical process. The deviation is from the electrolyte electro-oxidation without producing CO2 (Supplementary Fig. 3), which is consistent with the literature41. The molar flux of both CO2 (denoted as CO2) and CO (denoted as CO) follow the trend of the current during charging (equivalent to electron) but they could derive from different processes because of the multisource of the CO2 and CO evolution. Therefore, we mainly focus on the ratio CO2/CO because it helps us to determine the mechanisms (Eqs. 13). The ratio CO2/CO exhibits the comparison between Li2CO3 decomposition and other reactions because Li2CO3 decomposition does not form CO, as discussed later. The ratio CO2/CO in Fig. 1a is 7.06 (Supplementary Table 2), which does not fit any reaction pathways proposed above. Due to the lack of O2 evolution, Eq. (1a) could be excluded. The oxygen appears as singlet O2 (1O2) instead of low-energy triplet O2 (3O2), which will be discussed in detail later.

To further study the influence of electrode composition on the gas products, the same experiment was conducted using a Li2CO3-Super P (4:1) composite electrode. As shown in Fig. 1b, similar gas species without O2 were identified. Because the Li2CO3-Super P (4:1) composite electrode contains more insulating Li2CO3 solid than that in Li2CO3-Super P (1:1) electrode, leading to poor solid-solid contact, the charging current and consequent gas evolution in Li2CO3-Super P (4:1) is one order of magnitude lower than those in Li2CO3-Super P (1:1) electrode. If the CO2 and CO came from a certain intermediate or a simple one-step reaction, the same ratio CO2/CO would be observed. However, the ratio CO2/CO varies with the electrode composition, from 7.06 to 9.82 (Supplementary Table 2). This result suggests that the Li2CO3 decomposition is a complicated multistep process, instead of a simple one-step reaction as proposed previously16,17. Because Li2CO3, the carbon substrate, and the electrolyte all might contribute to CO2 and CO evolution, it brings up the confusion about the source of CO2 and CO, which is the key to addressing the reaction mechanisms.

### Gas products of Li213CO3 decomposition

The isotope-labeled Li213CO3 was used to identify the decomposition mechanisms of Li2CO3. The Raman spectrum and XRD pattern of the Li213CO3 confirm its composition (Supplementary Fig. 4). The 12C-impurity in Li213CO3 is determined by using a mass spectrometer (MS). The details are described in Methods—Isotope impurities. The Li213CO3 contains 15% isotope impurity of Li212CO3 (Supplementary Fig. 5), which would be subtracted from the MS results in the following experiments.

Figure 1c exhibits the gas evolution during the charging process of the cell with a Li213CO3-Super P composite electrode. Li213CO3 was electro-oxidized and a large amount of 13CO2 is identified, which is definitely from the Li213CO3 decomposition because both Super P carbon and electrolyte are unlabeled. No O2 evolution is identified again. 13CO was not observed as well, indicating that the Li213CO3 decomposed to merely 13CO2 without 13CO. The decomposition of Li213CO3 (releasing 13CO2) contributes 79% of the overall CO2 and CO evolution and it is the dominant process during charging. Meanwhile, some CO2 and CO from the inevitable decomposition of electrolyte and carbon substrate were identified during charging. The electrolyte/carbon decomposition contributes to approximately one-fifth of total gas evolution (Supplementary Table 3), which is a high ratio of side reactions.

In the Li-CO2 chemistry with lithium carbonate isotopically labeled with 13C on an unlabeled carbon substrate (which we write here as 12C for simplicity), Li213CO3 is posited to decompose together with the carbon substrate in the following reaction:

$$2{{{{{{\rm{Li}}}}}}}_{2}{\,\!}^{13}{{{{{\rm{C}}}}}}{{{{{{\rm{O}}}}}}}_{3}+{\,\!}^{12}{{{{{\rm{C}}}}}}-4{{{{{{\rm{e}}}}}}}^{-}\to 4{{{{{{\rm{Li}}}}}}}^{+}+{2}{\,\!}^{13}{{{{{{\rm{CO}}}}}}}_{2}+{\,\!}^{12}{{{{{\rm{C}}}}}}{{{{{{\rm{O}}}}}}}_{2}$$
(4)

If it was the case, the ratio between 13CO2 and 12CO2 is expected to be about 2/1, but we find a ratio of 6.1/1 (Supplementary Table 4), much higher than 2/1 because 12C does not transform to sufficient 12CO2. This result suggests that the decomposition of the Li2CO3-C electrode is a complicated reaction instead of a simple reaction with known stoichiometric numbers like Eq. (4). Therefore, the contribution of the C substrate during the charging process is the key to demystifying the reaction mechanisms.

### Decomposition of 13C-carbon substrate

To determine the contribution of carbon in this reaction, the same experiments were carried out again but by replacing the Super P carbon substrate with 13C-carbon and leaving the lithium carbonate and electrolyte unlabeled. The 13C-carbon contains 1.5% of 12C impurity as shown in Supplementary Fig. 6. As shown in Fig. 2 and Supplementary Fig. 7, both 13CO2 and 13CO evolved during the charging process and the rest of CO2 and CO evolution came from the decomposition of Li2CO3 and electrolyte. As the only source of 13C-isotope, 13C is oxidized to release both 13CO2 and 13CO during the charging process. The formation of 13CO exhibits that the 13C is oxidized incompletely like an incomplete combustion reaction. The 13C substrate could be oxidized either electrochemically by potential or chemically by oxidative agents that are formed in the previous steps. If 13C is oxidized electrochemically, the ratio 13CO2/13CO would depend on the potential (Fig. 2). If 13C is oxidized chemically by oxidative agents like 1O2 and superoxide species, which are generated during the charging process, the electrode composition affects the ratio between 13CO2 and 13CO.

### Impact of the potential on the decomposition of 13C-carbon substrate

Due to the high equilibrium potential of Li2CO3 decomposition (3.82 V), the cell potential could affect the decomposition reactions and thus affect the formation of 13CO2 and 13CO at different potentials during the charging process. Therefore, a cell with Li2CO3- Super P composite electrode was charged with LSV from 4.3 to 4.7 V and the gas evolution was measured. The cell was rested for 4 h after reaching 4.3, 4.4, 4.5, 4.6, and 4.7 V, respectively, to obtain a low background of gas evolution. The cumulative molar of gas evolution at various potential stages is shown in Fig. 2. The amount of CO2 from Li2CO3 decomposition increased when the potential increased. The higher the potential was, the more CO2 evolved, in accord with the accelerating decomposition of Li2CO3. Meanwhile, 13CO2 and 13CO evolved simultaneously and their amount increased together with the CO2 evolution when the potential increased. Supplementary Fig. 7 shows the gas evolution at various stages of the charging process from 4.4 to 4.7 V. The amount of 13CO2 and 13CO evolution increased with the rising potential up to 4.7 V. This increase is probably because more singlet oxygen forms at a higher potential, which leads to more severe carbon and electrolyte degradation. Although both 13CO2 and 13CO increased, the ratio 13CO2/13CO remains almost identical at these stages. This result suggests that the 13CO2 and 13CO likely originated from the same chemical reaction.

To check whether 13C is oxidized to any solid byproducts, the cell was dissembled at the end of charging and the composite electrode was collected to further quantify the remained solid by-product of inorganic/organic carbon as described in Methods—Quantification of the solid byproducts. Neither Li213CO3 (Supplementary Fig. 8a) nor organic 13C-carbonates (Supplementary Fig. 8b) were detected in the charged electrode, indicating that the 13C did not oxidize to any solid by-products, i.e., inorganic and organic carbonates. Only unlabeled inorganic Li2CO3 and organic carbonate were identified, which are from undecomposed Li2CO3 and electrolyte decomposition. The incomplete decomposition of Li2CO3 may be due to the poor solid-solid contact between the carbon substrate and Li2CO3 during the charging process. It is unlikely to decompose all Li2CO3 even up to 4.7 V, which is consistent with the literature41. Therefore, as only 13CO and 13CO2 and no 13C-containing solid products are detected when using 13C as the carbon substrate, the carbon substrate is only oxidized to gaseous carbon monoxide and carbon dioxide during charge in the presence of lithium carbonate and the 1 M LiTFSI-tetraglyme electrolyte.

Previous results showed that the functional groups at the carbon surface dominated the decomposition of the carbon substrate47. Therefore, the Raman and XPS spectra of 13C and Super P are recorded to show their surface condition (Supplementary Fig. 9). D-band and G-band of 13C were observed at 1289 and 1507 cm–1, respectively, in the Raman spectrum (Supplementary Fig. 9a). Both D-band and G-band drift in the negative direction due to the isotope effect. The intensity of the D-band and G-band (ID/IG) typically represents the disorderliness and amounts of defects in the carbon structure. Here, 13C and Super P carbon show the same ID/IG of 1.20, which indicates the same degree of disorderliness and a similar amount of defects from the surface of 13C and Super P. Similar XPS spectra are shown in Supplementary Fig. 9b, confirming the similar surface groups of these two carbon substrates. Therefore, 3C and Super P have similar surface functional groups and they are likely to exhibit similar behavior in this carbon decomposition on charge.

### Impact of the ratio of Li2CO3 and carbon on the gas evolution

Although the 13C carbon substrate oxidized to 13CO and 13CO2, its decomposition pathway is still ambiguous. It might be a pathway similar to the incomplete combustion reaction. Alternatively, 13CO and 13CO2 might be derived from a reaction intermediate with known structures like oxalate (C2O42−). To clarify the pathway, we studied the gas products of four Li2CO3-13C composite electrodes with various mass ratios of Li2CO3/13C from 2:1 to 1:4 (Fig. 3). Overall, the carbon decomposition contributes to about 10% of the total gas evolution and the exact contribution depends on the composition of the electrodes (Supplementary Table 3). When the composite electrode contains less Li2CO3, ratio 13CO2/13CO decreases from 1.25 to 0.51 (Fig. 3h), indicating the extent of oxidation of 13C is restricted by the amount of Li2CO3 in the composite electrodes. Otherwise, the 13C would be completely oxidized to 13CO2 instead of the mixture of 13CO2 and 13CO. This varying ratio 13CO2/13CO confirms that the overall reaction is a multistep reaction, rather than forming a complex intermediate like C2O42− which should give a certain ratio 13CO2/13CO independent of the ratio of Li2CO3/13C. The 13C is highly likely to be oxidized by the oxidative intermediates from Li2CO3 decomposition. Therefore, as ratio 13CO2/13CO varied with varying composition of the Li2CO3-13C composite electrode, we argue that the amount of Li2CO3 present in the electrode, and thus the amount of oxidative intermediates from Li2CO3 decomposition, influences the decomposition route of the carbon substrate.

### Oxidation of 13C-carbon by 1O2 and superoxide

Both 1O2 and superoxide are reactive intermediates formed during the discharge and charging process in Li-air batteries, which could result in the decomposition of the carbon substrate48,49,50,51,52. 1O2 could attack the electrolytes and the electrodes as a strong oxidative species. Here, 9,10-dimethylanthracene (DMA), a molecular trap for 1O2, is used to identify the 1O2 during electrochemical oxidation for the Li2CO3-C composite electrodes (see Methods—Identification of 1O2). When 1O2 forms, it rapidly reacts with DMA to form DMAO2, which could be identified in the 1H NMR spectrum. Here, at the end of the charging process, the electrolyte with DMA in the cell was extracted and its 1H NMR spectrum (Supplementary Fig. 10) shows that DMAO2 has formed, which confirms the formation of 1O2 during the charging process. Once 1O2 forms, 1O2 attacks carbon and releases 13CO and 13CO2.

Here, to study the contribution of 1O2 and KO2 to carbon decomposition, 1O2 and KO2 were respectively used to react with 13C as ex situ chemical experiments. 1O2 is produced from the disproportionation of superoxide species47 (The reaction between KO2 and Li+ in this work). More details are described in Methods—Chemical experiments between reactive oxygen species and carbon. The gas evolutions of 13CO2, 13CO, 12CO2, and O2 in the reaction between 1O2 and 13C are shown in Fig. 3e–g and Supplementary Fig. 11. Figure 3e–g show both 13CO and 13CO2 evolution, which confirms the incomplete oxidation of 13C-carbon by 1O2. The formation efficiency of 1O2 by disproportionation in solution is low, therefore, most O2 is released as 3O2 (Supplementary Fig. 11). Because 1O2 is formed in the electrolyte solution rather than at the surface of 13C, fresh 1O2 is more likely to attack the electrolyte than the 13C, forming a large amount of CO2. In addition, because a large amount of carbon makes the suspension viscous, the total amount of 13CO and 13CO2 slightly decreased with the increase of 13C. When the amount of 13C increased, more 13CO formed and the ratio 13CO2/13CO decreased (Fig. 3i and Supplementary Table 5). Both the in situ electrochemical charging experiments (Fig. 3h) and ex situ chemical experiments (Fig. 3i) show the same decreasing trend of ratio 13CO2/13CO with the increase of the amount of 13C relative to 1O2 or Li2CO3.

On the other hand, to further exclude the contribution from superoxide species, the same chemical experiments were carried out but replaced 1O2 with O2(sol). KO2 was dissolved in tetraglyme and crown ether was added to maximize the concentration of the O2(sol) in the solution. As shown in Supplementary Fig. 12, only 13CO but no 13CO2 was identified, which suggests that O2(sol) is incapable to oxidize 13C to 13CO2. In summary, the detected 13CO and 13CO2 come from the side-reaction of 1O2 attacking the 13C substrate (Eq. 6). Herein, we simulate the chemical reaction between 1O2 and 13C without applying potential, however, the potential applied in the charging process could make the real side-reaction more complicated. More 1O2 might be produced at a higher potential. It is noted that oxygenated byproducts from chemical reactions with 1O2 could first be produced and then electro-oxidatively decomposed48,49.

### The CO and CO2 evolution of electrolyte decomposition

The electrolyte decomposition during the oxidation of Li2CO3 on charge is inevitable and should not be ignored. A cell with Li213CO3-13C composite electrode was charged to quantify the contribution from electrolyte degradation (Fig. 4). Because the entire composite electrode is labeled with 13C isotope, the 12CO2 must come from the decomposition of the tetraglyme electrolyte. Figure 4 shows that 12CO2 and 12CO from the electrolyte decomposition contribute ~12% to the total gas evolution (Supplementary Table 6). The ratio between carbon dioxide species (12CO2 + 13CO2) and carbon monoxide species (12CO + 13CO) is 7.76, which is in accord with the ratio CO2/CO of 7.06 in the Li2CO3- Super P (1:1) composite electrode, as stated in Supplementary Table 2.

As discussed above, the decomposition of Li2CO3 and carbon substrate contributes to 79% and ~10%, respectively, of the total gas evolution. The side-reactions of carbon depend on the electrode composition because this charging process highly depends on the solid-solid contact between Li2CO3 and carbon substrate. The exact ratios of decomposition of individual components were not completely identical in each experiment because the nature of Li2CO3 decomposition is based on the solid-solid contact. At some contacting points, the charging process will stop due to the large overpotential caused by the restricted solid-solid contact. The sum of gas evolution from Li2CO3, carbon, and electrolyte is ~100%, which confirms the reliability of individual measurements.

Although an important consideration, full understanding of electrolyte decomposition is not our focus in this work. Therefore, to ensure that it does not affect the major gas evolution from Li2CO3 and C, we quantified the gas by-products and solid by-products from the electrolyte decomposition without further analyzing the liquid by-products dissolved in the solvent. The solid by-products, like Li2CO3 and organic carbon residuals in the composite electrode at the end of the charge, were quantified respectively. As shown in Supplementary Fig. 13, approximately one-third of Li213CO3 is unreacted and remained in the electrode and a trace amount (<0.1 μmol) of organic carbonate were identified. This result indicates that there is no exchange of 12C/13C between the electrolyte and Li2CO3/C.

### Electrochemical vs. chemical pathways of Li2CO3 decomposition

Here our results show that Li2CO3 decomposition in the ether-based electrolyte is mainly an electrochemical process, which is in good agreement with Kaufman et al.33, but Freiberg et al claimed that Li2CO3 decomposition was a chemical process in LiPF6 carbonate-based electrolyte and it was induced by water41. They argued that the electrolyte was oxidized and decomposed on high potential and it formed some proton-derivatives which chemically react with Li2CO3 to evolve CO2. To study the reasons of these contradicting results, we conducted a series of comparative experiments. Firstly, to avoid the direct participation of water in Li2CO3 decomposition, we carried out a chemical reaction between the commercial Li2CO3 and 1 M LiTFSI tetraglyme electrolyte with 1000 ppm and 5000 ppm H2O, respectively, and the gas evolution was recorded. The result in Supplementary Fig. 14 indicates that no CO2 and O2 were detected even if there were lots of protons (from water) in the electrolyte. Then, to check the impact of potential, a cell with blank carbon (Super P or 13C) electrode without Li2CO3 but with 1000 ppm of H2O in the electrolyte and the gas evolution was recorded, Supplementary Fig. 15. No O2 and CO2 evolved during the charging process below 4.7 V. Interestingly, in the presence of 1000 ppm H2O, the onset potentials of electrolyte decomposition drifted to a lower potential of 3.8 V, compared with 4.1 V without H2O. This result implies that H2O encourages the electro-oxidation of the electrolyte but there is still no gas evolution below 4.5 V.

Learning from Freiberg’s work, we decomposed Li2CO3 without direct electric contact to confirm the chemical mechanism41. Briefly, an extra Celgrad membrane was placed between Li2CO3 powder and the blank carbon composite electrode to separate them. On one hand, the Li2CO3 cannot be electro-oxidized due to the lack of direct electronic contact with the carbon electrode. On the other hand, the porous Celgard membrane allows the electrolyte decomposition byproducts that form at the carbon composite electrode to diffuse across the Celgard and react with Li2CO3.

For comparison, in the LiPF6-ethylene carbonate (EC)/ethyl methyl carbonate (DMC) electrolyte, the same setup was used but the gas evolution was analyzed at the end of the charging process due to the volatility of the EMC. Because the Li2CO3 is separated from the carbon electrode, it cannot be electro-oxidized. However, a large amount of CO2 evolution was identified (Fig. 5a) above 4.3 V, suggesting the chemical decomposition of Li2CO3 by the electrolyte decomposition byproducts at high potential. This result is consistent with Freiberg’s work41. The ratio e/CO2 is 2.20, in good agreement with the literature41 but it does not mean that Li2CO3 is electro-oxidized via a 2-e process. Instead, the electrolyte decomposition is likely to produce active H+ at a ratio e/H+ of 1 and the H+ reacts with Li2CO3 to release CO2 at a ratio H+/CO2 of 241. Therefore, here the ratio e/CO2 is just used for comparison between these four cells. Here, the chemical pathway does not take place until a high potential over 4.3 V when the electrolyte was electro-oxidized first. For comparison, the CO2 evolution in the charging process starts from ~3.9 V (Figs. 13), which is too low to electro-oxidize the electrolyte. Therefore, the Li2CO3 decomposition at low potential <3.9 V is dominated by the electrochemical pathway, instead of by a chemical pathway induced by electrolyte degradation by-products.

On the contrary, the same experiment but with LiTFSI-tetraglyme electrolyte shows little CO2 evolution (Fig. 5b), which makes a sharp comparison to the LiPF6-EC/EMC electrolyte. In LiTFSI-tetraglyme electrolyte. CO2 evolved from 4.3 V, indicating the decomposition of electrolytes at high potential could induce the chemical decomposition of Li2CO3. However, the ratio e/CO2 is 12.29, much higher than that in LiPF6-EC/EMC. In addition, the total amount of CO2 is as low as 0.4 μmol, <2% of the CO2 evolution in the Super P-Li2CO3 electrode, Fig. 1a. This distinct comparison indicates that the reaction mechanisms of Li2CO3 decomposition are completely different in LiPF6-EC/EMC and LiTFSI-tetraglyme electrolytes. The chemical pathway proposed by Freiberg is not applicable for tetraglyme-based electrolytes. The difference is due to either the Li salts or the solvents.

To find out whether the Li salts or the solvents make the difference, the salts were exchanged in these electrolytes and LiTFSI-EC/EMC and LiFP6-tetraglyme were used, Fig. 5c and d. Interestingly, the LiPF6-tetraglyme shows much CO2 evolution similar to LiPF6-EC/EMC. Meanwhile, LiTFSI-EC/EMC shows little CO2 evolution similar to LiTFSI-tetraglyme. The cells using LiPF6 salt show a large amount of CO2 evolution. The electrolytes used in this work were all dried and the water concentrations were below 4 ppm (Karl-Fischer titration). Therefore, this chemical mechanism could be due to the LiPF6 salt itself or its impurities because LiPF6 is difficult to purify53. Freiberg et al. also proposed that the chemical pathway involving LiPF6 salt41. In the rest of this work, we used LiTFSI in tetraglyme as the electrolyte, thus excluding any possible effects of LiPF6 on our results. As little CO2 was evolved in the Celgard-separated experiment with LiTFSI-based electrolytes, our results provide evidence that Li2CO3 decomposition in LiTFSI-tetraglyme electrolytes is dominated by an electrochemical rather than chemical process.

We propose the following reactions for the electrochemical decomposition of Li2CO3 on carbon substrates:

$${{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}-2{{{{{{\rm{e}}}}}}}^{-}\to {{{{{{\rm{CO}}}}}}}_{2}+2{{{{{{\rm{Li}}}}}}}^{+}+1/{2}{\,\!}^{1}{{{{{{\rm{O}}}}}}}_{2}$$
(5)
$$(1-{{{{{\rm{x}}}}}}/2){\,\!}^{1}{{{{{\rm{O}}}}}}_{2}+{{{{{\rm{C}}}}}}\to (1-{{{{{\rm{x}}}}}}){{{{{{\rm{CO}}}}}}}_{2}+{{{{{\rm{x}}}}}}\,{{{{{\rm{CO}}}}}}$$
(6)
$${\,\!}^{1}{{{{{\rm{O}}}}}}_{2}+{{{{{\rm{electrolyte}}}}}}\to {{{{{{\rm{CO}}}}}}}_{2}+{{{{{\rm{CO}}}}}}+{{{{{\rm{byproducts}}}}}}$$
(7)

Li2CO3 is oxidized at high potentials to form CO2 and aggressive 1O2 (Eq. 5), Fig. 6. 1O2 attacks carbon substrate (Eq. 6) and electrolyte (Eq. 7) simultaneously like incomplete combustion reactions, leading to further side reactions forming CO2 and CO. It should be noted here, Eq. (7) is just a schematics equation instead of a proper stoichiometric equation for an elementary reaction because we do not completely know all byproducts from electrolyte decomposition. The exact ratio CO2/CO depends on the ratio between formed 1O2, the carbon substrate/electrolyte, namely the electrode composition, but not on the cell potentials. Although the carbon substrate is eventually oxidized to CO2 and CO, the carbon does not directly participate in the electro-oxidation of Li2CO3 as has been proposed in the literature28. As we mentioned above, the Eq. (2) could be recognized as a combination of Eq. (5) and Eq. (6) when x is equal to zero in Eq. (6). In this case, the reactive intermediate 1O2 reacts with carbon to produce only CO2 instead of a mixture of CO2 and CO. However, our chemical reaction between 1O2 and 13C shows both 13CO2 and 13CO evolution (Supplementary Fig. 11). If a highly reactive intermediate other than 1O2 forms in the Li2CO3 decomposition and rapidly attacks carbon to just CO2, Eq. (2) could establish and contribute to a parallel pathway with Eq. (1) in Li2CO3 decomposition. However, there is so far no evidence of such a highly reactive intermediate.

Overall, during the charging process of the cell with a Li2CO3-carbon composite electrode in LiTFSI-tetraglyme electrolyte, Li2CO3 decomposition contributes the major CO2 evolution (~80%), and the carbon substrate and electrolyte decomposition share the rest ~20% of gas evolution. Therefore, in the charging process of Li-air batteries and Li-CO2 batteries, the decomposition of carbon substrate and the electrolyte is inevitable and contributes to a significant amount of CO2 and CO unless the 1O2 can be suppressed.

### Implications for batteries

Although Li2CO3 can be decomposed during the charging process, it is accompanied by severe side reactions by 1O2 (~20% of the total gas evolution) and leads to deterioration of cell performance. Here, we focus on Li-air and Li-CO2 batteries. In Li-air batteries, Li2CO3 is a major by-product and it passivates the electrode surface and kills the cells, which brings us to a dilemma. On one hand, if Li2CO3 is not decomposed during the charging process, Li2CO3 will rapidly accumulate during cell cycling and thus completely passivate the electrode. On the other hand, if Li2CO3 is decomposed during the charging process, it results in more side-reactions and thus accelerates cell deterioration. This effect is more significant in Li-CO2 batteries because Li2CO3 and carbon form are supposed to be the main discharge products. Therefore, a reversible decomposition of Li2CO3 and carbon is desired in the following charging process16,17. However, during the charging process, the discharge reaction cannot be completely reversed because the carbon does not directly participate in the Li2CO3 decomposition reactions. The carbon substrate is oxidized by the 1O2 intermediate to CO2 and CO. Meanwhile, the aggressive 1O2 attacks the electrolyte, leading to the depletion of the electrolyte and consequently cell failure. If the electrolytes and salts in the cells are sufficiently stable, e.g., solid-state electrolyte, the molten-salt electrolyte of LiNO3-KNO3, etc, the side-reactions with electrolyte could be avoided and 1O2 quenches to 3O2. Recently, Zhang’s group observed the O2 evolution when charging Li2CO3 in the cell using a solid-state electrolyte54. Due to the stability of inorganic solid-state electrolyte, 1O2 had nothing to attack and eventually was quenched to 3O2. However, even in this case, the 1O2 would oxidize the carbon substrate and thus the cell deterioration cannot be completely inhibited.

Therefore, the key to pursuing a better charging process is to design a catalyst that suppresses the formation of 1O2 rather than simply facilitates the kinetics of Li2CO3 decomposition and thus decreases the overpotentials. For instance, some redox mediators (RM) with moderate O2 binding energy could be applied to inhibit forming 1O2 and encourage 3O2 evolution by replacing the 1O2 precursor (e.g. CO42−) with a low-energy RM-involved intermediate, just like the way 2,5-di-tert-butyl-1,4-benzoquinone does to O255. Alternatively, some solid catalysts with suitable O2 binding energy could bind 1O2 precursor (the key reaction intermediator of Li2CO3 decomposition) in order to lower the energy and stabilize the precursor before it transforms to 1O2. The ratio of 1O2 in the final product will decrease when the energy of the 1O2 precursor decreases.

Very recently, Hu’s group used operando electron paramagnetic resonance to show that Co3O4 inhibited the 1O2 formation during the charging process of Li-O2 cells56. Here, we added some Co3O4 nanoparticles to the Li2CO3-Super P composite electrode and then charged the electrode. As shown in Supplementary Fig. 16a, some O2 was identified during the charging process. While the control experiment of charging the Co3O4 electrode itself without Li2CO3 exhibits no O2 evolution, Supplementary Fig. 16b. This result confirms that the Co3O4 catalyst successfully interacts with the precursor of forming 1O2 during the Li2CO3 decomposition and thus suppresses 1O2 formation. On the other hand, the CO evolution with Co3O4 is only half of that without Co3O4, whereas similar CO2 evolutions were identified in both experiments. Because CO is from side-reactions of 1O2 attacks electrolyte/carbon substrate, decreased CO evolution confirms that 1O2 formation was partially inhibited and thus fewer side-reactions of 1O2 attacking the carbon substrate and the electrolyte were detected than the counterpart without Co3O4, Supplementary Fig. 17. Our results are in good agreement with Hu’s group results55. Although the 1O2 formation and CO evolution cannot be completely inhibited, the Co3O4 does make some effects by stabilizing the reaction intermediates of forming 1O2 and thus encouraging the evolution of 3O2. This example confirms the feasibility of this strategy to promote the Li2CO3 decomposition with less parasitic 1O2, however further studies are needed to look for more effective catalysts to avoid the side-reaction caused by 1O2.

To explore the decomposition mechanisms of the Li2CO3 and clarify the role of carbon substrate in the charging process of Li2CO3-carbon composite electrodes in LiTFSI-tetraglyme electrolyte, we did a set of in situ DEM-GC experiments with the 13C isotope-labeled composite electrodes to systematically isolate each component of the cell. The gas evolution during the charging process, including CO, CO2, 13CO, 13CO2, and O2 was quantified. Li2CO3 decomposed to release CO2 at an onset potential of 3.8 V mainly via an electrochemical mechanism. The chemical mechanism of Li2CO3 decomposition in literature could take place in the presence of LiPF6 due to LiPF6 itself or its impurities. Carbon substrate did not directly participate in the decomposition of Li2CO3, that is to say, carbon did not react with Li2CO3 in a single step to form CO2 as desired. On the contrary, this process is a multistep reaction. In the first step, Li2CO3 was oxidized to CO2 and 1O2. Then, the 1O2 simultaneously oxidized the carbon substrate and electrolyte to form CO2 and CO as gaseous side products. Approximately 80% of the net/cumulative CO2 evolution is contributed by the Li2CO3 decomposition and the rest ~20% is contributed by the decomposition of the carbon substrate and the electrolyte, which cannot be ignored in batteries, particularly in Li-CO2 batteries. In this work, we clarify the reaction mechanisms of the Li2CO3 decomposition during the charging process and exhibit the role of carbon in this process. This finding establishes a detailed picture of the decomposition pathways of Li2CO3 which enables strategy for the design of highly efficient cathode catalysts for Li-air and Li-CO2 batteries.

## Methods

### Materials

Lithium carbonate (Li2CO3), potassium superoxide (KO2), lithium bis(trifluoromethane)sulfonamide (LiTFSI), 9,10-dimethylanthracene (DMA), and Co3O4 were purchased from Sigma-Aldrich. Li213CO3 and 13C were purchased from the Cambridge Isotope Ltd. Tetraethylene glycol dimethyl ether (tetraglyme), ethylene carbonate (EC), and methyl ethyl carbonate (EMC) were purchased from the TCI Chemical. Tetraglyme was distilled under vacuum and dried with activated molecular sieves (4 Å). Lithium hexafluorophosphate (LiPF6), ferrous sulfate (FeSO4), phosphoric acid (H3PO4), acetic acid, and hydrogen peroxide (H2O2) were purchased from Aladdin. Lithium iron phosphate (LFP) was purchased from Shenzhen Betterui New Materials Group Co., Ltd. Dimethyl sulfoxide-d6 (DMSO-d6) and 18-crown-6 were purchased from the Shanghai Yuanye Bio-Technology. Argon (N5 grade) and10 % Ar-O2 (N5 grade) were obtained from Nanjing Special Gas Ltd. Polytetrafluoroethylene emulsion (PTFE) was purchased from Innochem. Celgard separator (25 μm thickness, Celgard), glass fiber separator (GF/F, Whatman), and Super P carbon (Timcal) were purchased from Duoduo Chemical Technology Co. Ltd.

### Preparation of the composite electrodes

The blank carbon electrode, Li2CO3-carbon composite, and Li2CO3-Super P-Co3O4 composite (1:1:0.5) electrodes were prepared as described in literature9,28,56. Briefly, a certain amount of Super P, Li2CO3, and Super P carbon or Li2CO3, Super P and Co3O4, and PTFE were mixed and the mixture was ball milled overnight. The mass ratio of active material and binder PTFE is 10:1. A certain amount of well-mixed powder was weighed and absolute ethanol was added to obtain a slurry. The slurry was cast onto pre-washed stainless steel (SS) mesh (100 mesh) and the electrodes were dried under vacuum at 120 °C overnight. The mass loading is 5 mg per electrode. For instance, a Super P- Li2CO3 (1:1) composite electrode contains 2.27 mg Super P, 0.46 mg PTFE, and 2.27 mg Li2CO3 (equivalent to 30.7 μmol). For the 13C-isotope-labeled electrodes, the Li2CO3 and the Super P were replaced with the Li213CO3 and 13C carbon, respectively. The 13C-isotope-labeled composite is only obtained by grinding because the 13C-isotope-labeled substance is too expensive.

### Preparation of the LFP electrode

80 mg LFP powder, 10 mg Super P, and 100 mg PTFE suspension (10%) were prepared. Firstly, 80 mg LFP and Super P were ground in a mortar to ensure that the LFP and Super P were evenly mixed. 0.1 mL of ethanol was dropped into the mixed powder to wet it and then the PTFE suspension was added to the mixed powder. After LFP-Super P and PTFE were mixed evenly, it was rolled several times with a roller. The final thickness of the electrode is 0.8 mm. The LFP electrode was punched into 12–25 mm diameter and then soaked in a solution (7.2 mL 30% H2O2 and 3 mL acetic acid in 500 mL H2O) for 30 min to precharge the LFP. The precharged LFP electrodes were rinsed with water five times to completely remove the residual H2O2 and acetic acid. Finally, the treated LFP electrodes were dried in a vacuum oven overnight and transferred to a glove box for later use.

### Potential calculation

Ag wire was used as a pseudo-reference electrode in a three-electrode cell. We immersed the Ag wire in 1 M LITFSI tetraglyme electrolyte with O2 to stabilize the potential of the Ag wire. A three-electrode cell with a LFP working electrode was charged and discharged in 1 M LiTFSI-tetraglyme using this Ag wire as the reference electrode. The equilibrium potential of LFP was 0.5 V versus Ag wire. Considering the potential of LFP versus Li+/Li is 3.45 V, the potential of the Ag wire is 2.95 V versus Li+/Li. Therefore, the potentials versus Li+/Li of the experiments in 1 M LiTFSI-tetraglyme could be calculated. All potentials in this manuscript are versus Li+/Li without further notice.

### DEMS-GC setup

A differential electrochemical mass spectrometer (DEMS, Prima BT, Thermo Scientific Ltd.) was coupled with a gas chromatograph (GC, Hope Ltd.) in parallel (Supplementary Fig. 1). The GC is equipped with a TCD and an FID detector (including a CO2/CO converter). The DEMS cell is based on a customized Swagelok design. It was assembled and charged/discharged in the Ar-filled glove box. The Ar carrier gas carried the evolved gas in the cell into DEMS and GC simultaneously. The typical flow rate of Ag carrier gas is 0.5 mL min−1 and the sweep rate for LSV is 0.05 mV s−1. The time resolutions of DEMS and GC are 10 seconds and 5 min, respectively. The general gas evolution in chemical reactions was examined by the mass spectrometer (MS, Prima BT) itself.

### Experiments of Li2CO3 with 1000/5000 ppm H2O in 1 M LiTFSI-tetraglyme electrolyte

Firstly, 1000 ppm and 5000 ppm H2O were added to the 1 M LiTFSI-tetraglyme electrolyte, respectively and the water concentration in the electrolyte was quantified by a Karl Fischer titrator (Mettler Toledo). Then, the gas analysis was conducted in a MS. Generally, 100 mg commercial Li2CO3 was added to a vial that was connected to a MS. After passing in the carrier gas for a while to obtain a steady baseline, 1 mL electrolyte with H2O was added to the vial and it was stirred for several hours. The gas evolution was quantified by a MS.

### Quantification of gas evolution during the charging process

The composite electrodes were charged in a homemade differential electrochemical mass spectrometry cell. A piece of precharged LiFePO4 served as the counter electrode and a piece of silver wire served as the reference electrode. 0.2 mL of 1 M LiTFSI-tetraglyme served as the electrolyte. The water content of the electrolyte was <4 ppm tested by Karl Fischer moisture titrator. To study the impact of the potential on the charging process, the composite electrodes were charged using linear sweep voltammetry (LSV) with a sweep rate of 0.05 mV s−1. Due to the stability window (4.8 V) of the tetraglyme-based electrolyte, the cell potential was cut off at 4.7 V. Argon was served as the carrier gas and its flow is 0.5 ml/min. The gas evolution was examined by a magnet sector mass spectrometer (Prima BT, Thermo Scientific Ltd.).

For the experiments for the chemical route, the Li2CO3 was deliberately separated from the Super P-PTFE composite electrolyte to allow the electrolyte decomposition byproduct to decompose Li2CO3 chemically. Twenty milligrams of commercial Li2CO3 was weighed and dusted to the top of the separator. A piece of Celgard separator (25 μm thickness) was stacked on the Li2CO3 and then a Super P-PTFE composite electrode is stacked on the Celgard. The composite electrode was electronically isolated from Li2CO3 by the Celgard separator. For the 1 M LiTFSI-tetraglyme electrolyte, the cell was charged to 4.7 V and the gas evolution was recorded. For the control experiment with EC/EMC electrolyte, due to the volatile electrolyte, the cell was not purging with Ar carrier gas during the entire charging process. The gas evolution in the headspace was purged with Ar at the end of the charging process to quantify the CO2 evolution during the entire charging process. A homemade in-line cold trap (−25 °C) was applied to condense the evaporated EMC to minimize the background noise caused by EMC.

### Quantification of CO

The mass-to-charge (m/z) ratio of 28 could be contributed by CO, N2, and the fragment of CO2. To exclude the contribution of the N2, the gas inlet of the DEMS was placed inside an Ar-filled glovebox as shown in Supplementary Fig. 1 and the DEMS experiments were conducted inside the glove box at 25 ± 2 °C to minimize the impact of N2 leakage. The mz = 28 fragment from CO2 was calibrated with a calibration gas of 1% CO2–99% Ar using DEMS. The intensity of the mz = 28 signal is 6.5% of the mz = 44 signal, which is subtracted from the concentration of CO in the calculation. This ratio(mz = 28/mz = 44) depends on many factors, such as electron voltage, emission current of the filaments, etc. Therefore, this ratio needs to be calibrated for individual MS. A calibration gas of 0.1% CO–99.9% Ar was used for calibration.

The CO evolution during the charging process was parallelly quantified by a GC with a TCD and a FID detector (including a CO2/CO converter). A FID detector was used to quantify the CO and 1000 ppm CO equals an area of 130 mVs in GC. Thus we could use these results as a reference and qualify the amount of CO in all experiments.

### Quantification of the solid byproducts

The solid Li2CO3 and organic carbon in the composite electrodes were quantified as reported previously57. An electrode was placed in a vial that is connected to a MS and 0.5 ml of H3PO4 (2 M) was injected into the vial to react with Li2CO3 to form CO2. The CO2 evolution was quantified by MS. Afterward, Fenton solution (400 μL FeSO4 + 50 μL of H2O2 (30%)) was added to oxidize the organic carbonates to CO2. The released CO2 was again quantified by a MS.

### Isotope impurities in Li213CO3 and 13C-carbon

A certain amount of Li213CO3 reacts with H3PO4 solution in a vial that is connected to a MS. The gas evolution was quantified by MS and shown in Supplementary Fig. 5. 16.92 μmol of 13CO2 and 2.49 μmol 12CO2 are identified. Therefore, the ratio between 12CO2 and 13CO2 is 0.147/1. This result indicates that the commercial Li213CO3 sample contains c.a. 15% of 12C-impurity.

The 12C impurity in the 13C is determined by a MS as well. The carbon was combusted under a O2 flow (0.5 mL min−1) in a quartz tube that was connected to a MS. The formed 13CO2 and 12CO2 are quantified and they are 82.42 μmol and 1.27 μmol, respectively. The 13C sample contains 1.5% of 12C as an isotope impurity.

### Identification of 1O2

30 mM DMA was added into 1 M LiTFSI-tetraglyme electrolyte as a molecular trap to singlet O2. The composite electrode was electrochemical oxidation by linear sweep voltammetry (LSV) with a voltage cutoff of 4.2 V. The electrolyte was extracted from all cell components using DMSO-d6 for further experiment. 1H-NMR spectra were recorded on a Bruker Avance III 300 MHz FT NMR spectrometer with autosampler (300.36 MHz, DMSO-d6).

We lack an effective way to quantify the 1O2 during charging to such a high potential of 4.7 V. On one hand, DMA decomposes above 4.2 V and thus we cannot use DMA to detect 1O2 at 4.7 V. However, there is only a small portion (<5%) of Li2CO3 decomposes below 4.2 V and the amount of 1O2 at this stage is low, which could not represent the overall reaction up to 4.7 V. For the aspect of energy, more 1O2 would form at a higher potential. On the other hand, the NMR is only a semi-quantitative technique. Thus probing DMA-O2 using NMR could only confirm the formation of 1O2 but it could not provide a reliable quantification of such a low amount of 1O2. The NMR of the electrolyte in a cell with Co3O4 is shown in Supplementary Fig. 18. The cell was charged by LSV to 4.2 V. The electrolyte was extracted from all cell components using DMSO-d6. The signals of DMAO2 in both samples are too weak to compare quantitatively.

### Chemical experiments between reactive oxygen species and carbon

The purpose of this ex situ chemical experiment is to identify the products of reactive oxygen species attacking 13C-carbon. We try to prove the feasibility that 1O2 attacks 13C-carbon to form 13CO2 and 13CO, which is detected during the charging process of the cell with Li2CO3. Meanwhile, our results show that superoxide species could not oxidize 13C-carbon to form 13CO2 and 13CO.

Singlet oxygen was obtained by the disproportionation of superoxide species, which is KO2 and Li+ here. Briefly, a varying amount of 13C-carbon and 22 mg KO2 powder was added to a vial that is connected to a MS before the solution is added into the vial. The carrier gas will be passed for a period to eliminate the residual gas in the vial. Then, 1 mL high concentration lithium salt electrolyte (4 M LiTFSI here) was injected into the vial to react with KO2 to disproportionate to produce 1O2. 1O2 immediately oxidized 13C-carbon to evolve 13CO2 and 13CO. The produced gases were detected by a MS. An excess amount of KO2 was used here to produce a large amount of 1O2. So far, we could not quantify the amount of 1O2 produced in this disproportionation reaction of KO2 and Li+. Here, the experiments were carried out without applying a potential, namely at “OCV”. The reactivity of carbon and the ratio of CO2/CO may change when voltage is applied. The impact of potential in the side-reaction of 1O2 and 13C is not the focus of this work. Also, when a high potential is applied to the carbon substrate, electrode decomposition will take place, which makes the situation complicated to decouple these factors. Here, we mainly focus on Li2CO3 electro-oxidation. Although carbon oxidation accompanies Li2CO3 decomposition, it only contributes to ~10% of the total CO2 evolution.

A control experiment of superoxide attacking 13C-carbon was carried out to make a comparison. Superoxide was obtained as previously reported52. In brief, firstly, argon was used to exclude the dissolved gas such as oxygen in the tetraglyme solvent. Then, 71 mg KO2 and 264 mg 18-crown-6 were added to 10 mL tetraglyme electrolyte and stirred for four hours to maximize the concentration of the O2(sol). The electrolyte was centrifuged and the supernatant also was injected with argon to get out the dissolved oxygen produced by the KO2 attack electrolyte for further experiment. Finally, varying amounts of 13C-carbon powder were added to a vial and 1 mL supernatant was injected into the vial for the source of superoxide. The produced gases were detected by a MS.

In the charging process of the cell, because Li2CO3 has ultralow ionic and electronic conductivities, its decomposition takes place at the 13C|Li2CO3 solid-solid interface (namely the contact points between 13C particles and Li2CO3 particles). Therefore, the decomposition product, 1O2, is just formed at this interface, which is very close to 13C, and thus it is easy to attack 13C. Therefore, the ratio between 1O2 and 13C is high. On the contrary, in the chemical experiments described above, 1O2 originates from the disproportionation of superoxide in the solution phase. Although 13C was added to the solution and stirring was applied, the fresh 1O2 forms from the disproportionation process in the electrolyte and it first attacks the electrolyte. Only a small portion of 1O2 would diffuse to the surface of the suspended 13C in the electrolyte and react with 13C. In addition, the ratio of releasing 1O2 from disproportionation is still unknown, which is highly likely below 10%. Freunberger et al show that more 1O2 forms during the charging process than in the discharging process involving disproportionation.

### Quenching 1O2

Once the 1O2 forms in the electrolyte, it is too late to stabilize it and it rapidly attacks the electrolyte and carbon substrate. Therefore, our strategy is to stabilize this intermediate/precursor which further produces singlet oxygen, rather than to quench the 1O2 after its formation in the electrolyte. In this case, we hope that Co3O4 binds the intermediate (namely the precursor of 1O2, e.g. CO42−) into a complex such as Co3O4-CO4, before 1O2 is formed and dissolved in the electrolyte. Our experimental results show that Co3O4 makes some positive effects, but it is not good enough. If it works properly, the ratio(CO2/O2) should be 2 according to the equ 1a, but the ratio(CO2/O2)Co3O4 is 6.7, much higher than 2. It might be a feasible strategy to inhibit 1O2 formation but there are still lots of work that need to be done.