Oxygen (O2) reduction is one of the most studied reactions in chemistry.1 Widely investigated in aqueous media, O2 reduction in non-aqueous solvents, such as CH3CN, has been studied for several decades.2–7 Today, O2 reduction in non-aqueous Li+ electrolytes is receiving considerable attention because it is the reaction on which operation of the Li–air (O2) battery depends.8–29 The Li–O2 battery is generating a great deal of interest because theoretically its high energy density could transform energy storage.8, 9 As a result, it is crucial to understand the O2 reaction mechanisms in non-aqueous Li+ electrolytes. Important progress has been made using electrochemical measurements including recently by Laoire et al.29 No less than five different mechanisms for O2 reduction in Li+ electrolytes have been proposed over the last 40 years based on electrochemical measurements alone.25–29 The value of using spectroelectrochemical methods is that they can identify directly the species involved in the reaction. Here we present in situ spectroscopic data that provide direct evidence that LiO2 is indeed an intermediate on O2 reduction, which then disproportionates to the final product Li2O2. Spectroscopic studies of Li2O2 oxidation demonstrate that LiO2 is not an intermediate on oxidation, that is, oxidation does not follow the reverse pathway to reduction.
In this work CH3CN was used as the solvent because it has been used widely for O2 reduction and shown to be sufficiently stable towards reduced O2 species for the studies undertaken herein.4, 5 Au was chosen as the electrode because it is an excellent substrate for surface enhanced Raman spectroscopy (SERS).30 To confirm that the electrode/electrolyte combination used herein is sufficiently stable for our studies, cyclic voltammograms (CVs) were collected in 0.1 MnBu4NClO4-CH3CN saturated with O2 at an Au electrode (Figure S1 in the Supporting Information). These data show that QA/QC≈1 (Q is the charge passed on the anodic (A) and cathodic (C) sweeps, see Table S1) at all scan rates in Figure S1, in accord with the reversibility and hence stability of the electrolyte and electrode towards the reduced O2 species.
O2− has been detected previously in CH3CN.4 To confirm its formation on reduction of O2 in our experiments, in situ SERS data were collected on the Au electrode in 0.1 MnBu4NClO4-CH3CN at various potentials on reduction and oxidation, indicating respectively the formation of O2− and its disappearance (Figure 1). The peak assignments in Figure 1 were based on the vibrational spectrum for O2− [4a] and the SERS spectrum in Figure S2; the data for the latter were obtained by dissolving KO2, using a crown ether to complex the K+ and hence promote dissolution (see caption to Figure S2). The CVs of O2 reduction as a function of scan rate were analyzed using the DigiSim software.31 The heterogeneous rate constant, k0=2.1×10−4 cm s−1, was obtained from this analysis. The O2 concentration [O2]=6.8 mM and O2 diffusion coefficient 7.0×10−5 cm2 s−1 were obtained by fitting the current response to a potential step at an Au microelectrode (Figure S3) following the procedure described previously7 (see Experimental Section). It is known that O2− can form ion pairs with molecular cations such as organic ammonium ions, however, such interactions are weak compared with those involving Li+ ions.2, 26
The reaction between O2− and Li+ was investigated as a function of Li+ concentration (Figure 2). Addition of a 1 mM concentration of Li+ resulted in the appearance of a new reduction peak at higher potentials (2.35 V) compared with the original O2 reduction peak. The magnitude of the new peak grows with increasing Li+ concentration and at the expense of the area under the original O2 reduction peak. This behavior is consistent with an EC mechanism, that is, electrochemical reduction followed by a chemical step.32 Such following chemical reactions severely deplete the concentration of O2− thus shifting the potential to higher voltages, as observed here.32 When the concentration of Li+ ions is lower than O2, then there is insufficient Li+ to react with all the O2− that is generated, hence “unbond” O2− persists and two peaks are apparent. When the concentration of Li+ exceeds that of O2 (in this case the O2 concentration is 6.8 mM) then all the O2− is consumed by reacting with Li+. The low voltage reduction peak disappears leaving only one reduction peak. For a similar reason the peak at 2.75 V associated with O2− oxidation disappears when the Li+ concentration exceeds that of O2. The shift of the reduction potentials to lower voltages and the lowering of the reduction current with increasing Li+ concentration are consistent with partial blockage of the electrode surface by the insulating reduction products, which becomes more severe at higher Li+ concentrations. Such a phenomenon has been observed before.27 As stated above, Au was used because it permits SERS studies of the electrode surface. The same electrochemical reactions occur on glassy carbon electrodes, as shown in Figure S4.
Although these and previous electrochemical studies are very valuable, they cannot identify directly the species formed on reduction. This is illustrated by the fact that different authors have proposed different mechanisms for O2 reduction based on electrochemical measurements;25–29 two examples are given here:((1a)), ((1b))
Spectroscopic methods can identify directly the reaction products and their intermediates, and therefore are invaluable in investigating the O2 reduction mechanism. The results of in situ SERS measurements are presented in Figure 3. A background spectrum was collected before application of a potential to the cell (OCV; open circuit voltage). The spectrum is consistent with that expected for CH3CN; the peak (1) at 918 cm−1 is assigned to the CC symmetric stretch in CH3CN. Data were then collected at a potential of 2.2 V, that is, within the reduction peak in Figure 2. Spectra are shown at this potential for successive time intervals. Within a short time, two new peaks (2 and 3) appear that were not present at OCV. The most prominent occurs at 1137 cm−1 and is associated with the OO stretch of LiO2.33, 34 The smaller peak at 808 cm−1 corresponds to the OO stretch of adsorbed Li2O2.35, 36 With the passage of time the LiO2 peak diminishes until only the Li2O2 peak remains. The LiO2 peak occurs some 28 cm−1 higher than O2− in nBu4N+ solution, in accord with the stronger binding of O2− to Li+ in LiO2, resulting in the transfer of anti-bonding electron density from OO− to Li+. The Raman spectra provided direct spectroscopic evidence that reduction of O2 in the presence of Li+ ions in a non-aqueous electrolyte first forms O2− that then binds to Li+ forming LiO2 on the surface of the electrode. They further demonstrate that LiO2 is unstable and disproportionates to the more stable Li2O2, that is, 2 LiO2→Li2O2 + O2. In other words the Raman spectra have shown that the process of O2 reduction in the presence of Li+ follows Equations (2a–c) described above, and does not proceed through disproportionation of superoxide to peroxide ions [Equations (1a,b)] followed by the formation of Li2O2 without passing through LiO2 as an intermediate.25–29 It also shows that the LiO2 intermediate is present on the electrode surface. It is known that on extending the voltage range to much more cathodic potentials further reduction process occurs that may be assigned to O2− reduction to O22− in the absence of Li+.2, 3 In the presence of Li+, LiO2 reduction to Li2O2 occurs (Figure S5). However, this is at a significantly more negative (lower) potential than the SERS data (0.8 V lower), thus direct reduction of LiO2 to Li2O2 is unlikely to make a major contribution to the transformation of LiO2 to Li2O2 at the voltages used in our experiments, which instead are dominated by disproportionation.
Returning to the CVs in Figure 2, at high Li+ concentrations two oxidation peaks are apparent at 3.55 and 3.75 V. To investigate the oxidation in more detail, a series of CVs was collected in 0.1 M LiClO4-CH3CN (Figure 4). Each CV was collected by first sweeping the potential from 3.2 V (OCV) to 2.2 V and then keeping the potential at OCV for various dwell times, before completing the oxidation sweep. Sweeping the potential to 2.2 V results in the formation of LiO2 and Li2O2. Thereafter, short dwell times lead to the presence of the lower voltage oxidation process, whereas this peak diminishes as the high voltage peak grows with increasing dwell time. The lower voltage oxidation peak is associated with the oxidation of LiO2, whereas longer dwell times result in a proportionately greater quantity of Li2O2, the decomposition of which is associated with the high voltage oxidation peak. The purpose of the dwell time experiments was to investigate the kinetics of disproportionation: by employing various dwell times and deconvoluting the areas under the two oxidation peaks (see Experimental Section), the first-order rate constant for the disproportionation reaction from LiO2 to Li2O2 + O2 was calculated, k=2.9×10−3 s−1.
The presence of oxidation peaks for both LiO2 and Li2O2 in the CVs only occurs because of the relatively short time scales. In practice, Li–O2 cells are charged and discharged over much longer times, and hence all the LiO2 will have disproportionated to Li2O2 by the end of discharge. Therefore, in the context of the Li–O2 battery, it is interesting to examine the oxidation (charging) of pure Li2O2. As for reduction, different mechanisms may be proposed for oxidation of Li2O2; for example one mechanism involves the oxidation of Li2O2 to LiO2 then to O2 [Equations (3a,b)], but others are also possible, such as Equation (4), which does not involve LiO2 as an intermediate.((3a)), ((3b))
As noted above, the oxidation peak for Li2O2 occurs at 3.75 V, which is above the oxidation potential for LiO2 at 3.5 V, leading us not to expect LiO2 as an intermediate on oxidizing Li2O2 since LiO2 would be unstable at 3.75 V. However, to explore this directly, in situ SERS and in situ differential electrochemical mass spectroscopy (DEMS) data were obtained. Considering first the Raman data: after applying a reducing potential of 2 V until only Li2O2 was present, the potential was switched to 3.75 V (the oxidation potential of Li2O2) then to 4.4 V. The SERS spectra collected at these oxidation potentials are shown in Figure 3, and there is no evidence of LiO2, consistent with Li2O2 decomposing directly without passing through LiO2 as an intermediate. It has recently been confirmed that O2− reacts strongly with propylene carbonate electrolytes to form various decomposition products including CO2.37 As a result we can now use this electrolyte as a chemical probe for superoxide: if oxidation of Li2O2 formed LiO2 or O2− then in PC electrolyte CO2 would be observed. An electrode was constructed in the discharged state, that is, containing Li2O2 (Aldrich), then placed in contact with a 0.1 M solution of LiPF6 in propylene carbonate. The electrode was charged in successive current steps, the results are presented in Figure S6. As the current is increased there is an immediate increase in the cell potential and a corresponding increasing in the m/z=32 signal due to O2 evolution on decomposing Li2O2 upon charging. No other gases are evolved at a level greater than 1 % of the O2. In particular, the absence of a mass signal at m/z=44, corresponding to CO2, indicates that Li2O2 decomposes directly in a one-step process to Li++e−+O2, that is, through Reaction (4). O2 evolution upon charging Li2O2 has been reported earlier,11 however, the sensitivity of the former DEMS apparatus did not allow the detection of minority gases. Here we have used a new generation of DEMS apparatus (developed by one of us; P.N.), with gas detection limits improved to low ppm values thus making possible detection of CO2, and hence superoxide (if present), down to a level of 0.1 % of O2. Together, the SERS and DEMS results demonstrate that oxidation occurs by direct decomposition according to the reaction Li2O2→2 Li+ + 2 e− + O2. In other words, the pathways followed on reduction and oxidation are different. This is in accord with the different voltages for charge and discharge, the separation of which persists even at low charge/discharge rates as confirmed by the CV with low scan rate in Figure S7. The different pathways for reduction and oxidation do not violate the principle of microscopic reversibility. The discharge reaction is in three steps and may be viewed as an ECC mechanism [Equations (2a–c)]. For microscopic reversibility, the reaction on charging would have to reverse along the same path, that is, CCE. As reduction occurs at 2.2 V and oxidation of Li2O2 at approximately 3.7 V, even at low rates, as discussed above, this implies that at least one of the forward reaction steps, most likely reaction step (2c), is irreversible or at least the reverse chemical reaction is very slow. As a result, direct electrochemical oxidation of Li2O2 occurs more readily than reversing along the same pathway as reduction. This phenomenon has been discussed before, all be it in quite different systems such as reductive decomposition of alkylhalides.38
In conclusion, in situ spectroscopic studies of O2 reduction in non-aqueous solvent, in the presence and absence of Li+ ions, have provided direct evidence of LiO2 as an intermediate on O2 reduction, which then disproportionates to Li2O2 (kdispr.=2.9×10−3 s−1). On charging, in situ spectroscopic studies reveal that Li2O2 decomposes directly, in a one-step reaction to evolve O2 and does not pass through LiO2 as an intermediate.