## Introduction

The rechargeable aprotic lithium–O2 (air) battery operates by the reduction of O2 at the positive electrode forming Li2O2 on discharge, with oxidation of Li2O2 taking place on charge1,2,3,4,5,6,7,8,9,10. Li2O2 is an insulating and insoluble solid11,12,13,14,15,16. Ether-based electrolytes, such as dimethoxyethane (DME) and tetra ethylene glycol dimethyl ether (tetraglyme), have been used as the basis of electrolyte solutions in most Li–O2 cells, because of their relative stability towards reduced oxygen species. However, they cannot dissolve LiO2, the intermediate in the reduction of O2 to Li2O2,

$${\rm O}_2 + {\rm L}{\rm i}^ + + {\rm e}^- \to {\rm LiO}_2$$
(1)
$$2{\rm LiO}_2 \to {\rm Li}_2{\rm O}_2 + {\rm O}_2$$
(2)
$${\rm LiO}_2 + {\rm Li}^ + + {\rm e}^- \to {\rm Li}_2{\rm O}_2$$
(3)

resulting in LiO2 being adsorbed on the electrode surface, and resulting in the growth of Li2O2 films on the electrode, leading to low rates, low capacities and early cell death14,17. The problem is exacerbated by the formation of Li2CO3 between Li2O2 and carbon, the latter is usually employed as the material for the porous positive electrode18. Use of redox mediators (RMs) on discharge, such as 2,5-di-tert-butyl-1,4-benzoquinone (DBBQ), which are reduced at the electrode surface on discharge and then go on to reduce O2 to Li2O2 in solution, can help to mitigate these problems, but result in the formation of Li2O2 disconnected from the electrode surface and therefore electronically isolated during charging19. This introduces the need for a redox mediator to be employed on charging that can oxidise Li2O220,21,22,23,24,25,26,27,28,29,30,31,32,33. Such mediators are molecules capable of oxidation at the surface of the pores in the porous positive electrode on charging and then transfer of holes to the electronically isolated Li2O2 particles within the pores. As a result, Li2O2 is oxidised and O2 released, the mediator molecule being reduced in the process and returning to the electrode surface for the cycle to be repeated.

Suitable oxidation mediators must have a redox potential above that for O2/Li2O2: 2.96 V vs. Li+/Li, a sufficiently high heterogeneous rate constant for electron transfer at the electrode surface to support the required charging rates, a highly reversible redox process such that the cycle may be carried out many times, and of course not only be capable of oxidising Li2O2, but with a sufficiently high rate to sustain the required charging current34. Stability of the mediators, especially on long-term cycling, is also an important challenge and recent work has considered the design of more stable redox mediators for cycling22; however, very little is known about the factors affecting the reaction between oxidation mediators and Li2O2. It is often assumed that a mediator with a high redox potential has fast kinetics for the oxidation of Li2O2, but this is not necessarily so33. Alternatively, the kinetics of Li2O2 oxidation by a mediator could be linked to the kinetics of its own redox process, but this would only be the case if both were outer-sphere electron transfer processes. Importantly, little experimental evidence exists about the kinetics of Li2O2 oxidation by redox mediators, yet their use and such kinetics are crucial to the operation of the Li–O2 cell.

Here, we investigate the kinetics of Li2O2 oxidation by several classes of redox mediators, which differs in the Eo (standard redox potential) and ko (standard heterogeneous electron transfer rate constant) values, to ascertain the factors that control the rate of Li2O2 oxidation by the mediators.

## Results

### Apparent rate constants (kapp) of mediators

Apparent rate constants (kapp) for Li2O2 oxidation by the redox mediators were determined using scanning electrochemical microscopy (SECM). Details of the cell and procedures used are given in the Methods section. In brief, SECM feedback approach curves at a Li2O2 disk, composed of a pressed pellet of commercial Li2O2 with a diameter of 12 mm, were recorded and apparent rate constants, kapp, for Li2O2 oxidation were obtained by fitting to the theoretical feedback approach curves developed by Cornut et al.35,36,37. When recording an approach curve, the SECM tip was held at a sufficiently positive potential such that a steady-state current was obtained for the oxidation of the redox mediator. The tip approaches the Li2O2 disk and at small separation distances, the mediator oxidised at the tip diffuses to the Li2O2 disk where it oxidises Li2O2, regenerating itself and contributing to a feedback loop, while concurrently, diffusion of the mediator to the tip is blocked by the surface. The balance of the two alter the current at the SECM tip, iT, and the faster the kinetics of Li2O2 oxidation by the mediator the greater the current, see Supplementary Figure 1. As we do not know the mechanism by which the mediators oxidise the lithium peroxides surface, we can only obtain an apparent rate constant (kapp) based on the feedback response; however, this provides a comparison between the different mediators and indicates the overall rate capability.

Fig. 1 shows the oxidation mediators studied. They are in three classes, amines, nitroxy and thiol compounds, chosen because they are classes of compounds known to exhibit reversible redox processes and include several of the compounds that have been used as oxidation mediators in Li–O2 cells, such as tris[4-(diethylamino)phenyl]amine (TDPA), 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO) and 10-methylphenothiazine (MPT)20,21,22. kapp for Li2O2 oxidation by the mediators are presented in Supplementary Table 1. The standard redox potential, Eo, and standard heterogeneous electron transfer rate constant, ko, were measured for each mediator using cyclic voltammetry, as described in the Methods section. The diffusion coefficients, D, were obtained from the steady-state current at an ultramicroelectrode (UME), also as described in the Methods section. The values for each of the three parameters are also given in Supplementary Table 1. Three additional mediators, tetrathiafulvalene (TTF), ferrocene (FC) and 5,10-dimethylphenazine (DMPZ), which do not belong to the above three classed, but have been commonly used as oxidation mediators, were also studied and are listed in Supplementary Table 123,24,38. The standard redox potentials are all positive for the O2/Li2O2 reaction. The diffusion coefficients vary by no more than a factor of 3. The ko for the mediators themselves are all relatively high, ranging from 0.007 to 0.078 cm s–1, sufficiently so to support an areal current density over 200 mA cm−2 at an overpotential of 60 mV, based on the true surface area of the pores and therefore more than sufficient to sustain an areal current density suitable for a Li–O2 cell. Of course, this does not take into account the kinetics of Li2O2 oxidation required to sustain the current, which will be discussed below after the presentation of the rates of Li2O2 oxidation. The assumptions regarding the porous cathode structure and the approach used to make this estimate are described in the Supplementary Note.

Before considering the kinetics of the mediator oxidation in more detail, we first determine the surface composition of the disk and the possibility of passivation with, for example, Li2CO3. A disk of Li2O2 was immersed in 1 M LiTFSI in tetraglyme for 3 h and then examined by time of flight secondary ion mass spectrometry (TOF-SIMS), alongside a disk that was not exposed to the electrolyte solution. As shown in Fig. 2, for both disks, the major peaks are from Li2O2+, with the secondary peaks being ascribed to Li2CO3. These results show that although there is some Li2CO3, even on the surface of the pristine disk, a significant proportion of the surface remains as Li2O2 even after 3 h of exposure to the electrolyte, confirming that the disk is suitable for the SECM measurements. Note that the sensitivity of TOF-SIMS to different species varies, consequently it is not possible to quantify the relative amounts of Li2O2 and Li2CO3 by simply comparing the areas under the peaks. Instead, the disk was etched until the signal from Li2O2 was constant, therefore corresponding to the bulk peroxide, i.e., 100% Li2O2. Comparing this signal with that for Li2O2 at the surface indicated that approximately 35% of the disk surface was Li2O2.

A disk of Li2CO3 was investigated with SECM using TEMPO as the oxidation mediator, as it has a sufficiently high potential to oxidise Li2CO3 and shows fast kinetics with the Li2O2 disk. The results are shown in Supplementary Figure 2. The kapp for oxidation of Li2CO3 by TEMPO is four orders of magnitude lower than the data collected on the Li2O2 disk, indicating that even for mediators with sufficiently high potentials the contribution of Li2CO3 oxidation to the kapp is very small. The dominant reaction for the range of mediators studied here, even taking account of partial coverage by Li2CO3, is oxidation of Li2O2.

It has been reported previously by us and by others that several of the redox mediators used in Li–O2 cells to date exhibit some degree of decomposition24,39,40,41. Assembling a cell with commercial Li2O2 and the oxidation mediators TTF and AZO, and then charging to a capacity of 1 mAh results in notable decomposition of TTF and AZO as seen by 1HNMR of the electrolyte, see Supplementary Figure 3. In the SECM experiments, only a small amount of charge, 1 nAh, is passed, therefore the fraction of mediator that is decomposed is negligible.

### Inner-sphere process for mediator oxidising Li2O2

To explore the possible correlations between kapp and the electrochemical parameters of the redox mediators, Eo and ko, plots of kapp vs. ko and Eo and are presented in Figs. 3 and 4, respectively. There is no apparent dependence of kapp on ko, Fig. 3. The values of ko for the different redox mediators appear independent of the nature of the electrode used to measure them, as demonstrated by measuring these values at Au and glassy carbon electrodes, see Supplementary Figure 4 and Methods section, consistent with the RM+/RM reactions occurring by outer-sphere electron transfer. If the oxidation of Li2O2 was also an outer-sphere electron transfer reaction, then kapp would be proportional to ko of the redox mediator (and hence the reorganisation energy of the RM and surrounding solution), or the rate of the reaction Li2O2 → Li2O2+ + e. Since there is no dependence of kapp on ko, the former cannot be true. If the rate was limited by the electron transfer kinetics associated with the Li2O2, then kapp would be invariant, which again is not the case. We conclude that oxidation of Li2O2 by the redox mediators is mainly an inner-sphere process, i.e., involves adsorption of the mediator on the peroxide surface. The values for kapp are one order of magnitude smaller than the corresponding ko values, indicating that the reaction of mediators oxidising Li2O2 is most likely to be the rate determining step of the entire charge process. This will particularly be true towards the end of charge when the surface area of the remaining Li2O2 is low. We estimate that a kapp from 2.5 × 10–5 to 7.9 × 10–3 cm s–1 in a Li–O2 cell with a porous cathode filled with Li2O2 would provide an areal current density of 108 mA cm–2 to 1.9 A cm–2 using the same model for the porous cathode as above. The details are described in the Supplementary Note and Supplementary Figure 5. Although we note that this equivalent charging current varies with consumption of Li2O2, it is sufficient to sustain the charging process, even for some of the slowest oxidation mediators investigated here.

Turning to the plot of kapp vs. Eo, Fig. 4, it appears that the highest rates are observed for mediators with potentials above 3.6 V. However, potential per se is not the explanation for the high rate, as there are examples of mediators with a high potential but low rate, e.g., BPPT. From the experiment on the Li2CO3 disk using TEMPO, we know higher rates at high potentials are not due to the onset of Li2CO3 oxidation contributing to the overall surface oxidation kinetics. Different crystal facets of Li2O2 will have different oxidation potentials42. Mediators operating at higher potentials could oxidise these higher potential facets and hence access a greater Li2O2 surface area. However, the fact that the rates vary for different mediators above 3.6 V and several high potential mediators have relatively low kapp suggests that this alone cannot be the reason for high rate mediators having a relatively high potential. As discussed below, we believe an important factor controlling the rate of the mediators is the nature of the oxidising centre and the degree of its steric hindrance.

Considering the molecules presented in Fig. 1 and the kapp values shown in the figure, it is evident that the nitroxy radicals exhibit the fastest rates of Li2O2 oxidation. The thiol group also provides a high rate, in contrast to the amines that are all low rate. The chemistry of the redox centre appears to be an important factor for controlling the rate of oxidation, probably due to the interaction with Li2O2 surface. The oxidation rates decrease when the redox centre of the molecule is surrounded by bulky groups, Fig. 1. This suggests that a key factor influencing the kinetics of Li2O2 oxidation is the steric hindrance as the molecule approaches the surface of Li2O2. The fastest kinetics is exhibited by 2-azaadamantane-N-oxyl (AZO), 7.9 × 10–3 cm s–1, which has the most exposed redox centre of all the redox mediators studied here. This observation is in accord with the lack of evidence for an outer-sphere reaction and provides direct evidence for Li2O2 oxidation proceeding by an inner-sphere mechanism.

## Discussion

In conclusion, we have measured the rate constants for the oxidation of Li2O2 particles by a series of molecular mediators spanning standard redox potentials, Eo from 3.1 to 3.9 V and standard heterogeneous rate constants for electron transfer, ko from 0.007 to 0.078 × 10–3 cm s–1. The surface of Li2O2 particles in a typical electrolyte solution, LiTFSI in tetraglyme, is partially covered by Li2CO3, but the rate of Li2CO3 oxidation, a mediator that operates at 3.8 V, TEMPO, is four orders of magnitude lower than for Li2O2, therefore Li2O2 oxidation dominates. There is no correlation between the variation of ko, the standard heterogeneous rate constant at the electrode surface for the mediators, and the rate of Li2O2 oxidation by the mediators, indicative of this not being an outer-sphere electron transfer process at the Li2O2 surface. There is evidence of Li2O2 oxidation rates depending on the nature of the oxidising molecule. Nitroxy radicals, especially those with low steric hindrances of access to the Li2O2 surface, exhibit the highest rates. Nevertheless, the mechanism of Li2O2 oxidation by molecular oxidants is still not well understood, and such understanding will be important in order to inform the design of optimised oxidation mediators. All mediators studied display kinetics sufficient to enable relatively high rates within a battery, charging current density exceeding 100 mA cm–2. A mediator with a kapp of 7.9 × 10–3 cm s–1 can sustain an areal current density of up to 1.9 A cm–2, based on the same model. It is important to note that stability is still a challenge for the Li–O2 battery and here we observe significant mediator decomposition when passing large amounts of charge. More stable electrolytes and mediators are required to minimise side reactions and hence improve cycleability.

## Methods

### Materials preparation

Li2O2 and Li2CO3 disks were obtained by pressing Li2O2 powder (Aldrich) and Li2CO3 powder (Aldrich) with a die set in an Ar-filled glove box. Disks of 13 mm diameter and 1 mm of thickness were prepared and served as substrate. A Au microelectrode (diam. 25 μm, CHI) served as an SECM probe tip. Prior to measurement, the Au tip was polished with a microelectrode beveller (Sutter) and checked with a microscope. A silver wire reference electrode (RE) and a platinum counter electrode (CE) were used. 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO), 2-azaadamantane-N-oxyl (AZO), 1-methyl-2-azaadamantane-N-oxyl (MAZO), tris[4-(diethylamino)phenyl]amine (TDPA), 1,4-bis(diphenylamino)benzene (DPAB), N,N,N′,N′-tetramethyl-p-phenylenediamine (TMPD), 10-methylphenothiazine (MPT), 10-isopropylphenothiazine (PPT), 10-(4-biphenylyl)phenothiazine (BPPT), tetrathiafulvalene (TTF), ferrocene (FC) and 5,10-dimethylphenazine (DMPZ) are from Aldrich. 10 mM redox mediators are dissolved in 100 mM LiTFSI–tetraglyme electrolyte for electrolyte solution.

A Swagelok cell was assembled as reported previously40, using a piece of gas diffusion layer electrode (GDL) as the positive electrode. A lithium super ionic conductor disc (LiSICON, Ohara) was used to protect Li metal as the negative electrode. A Li2O2 disk was placed between the GDL and the LiSICON essentially placing the cell in a discharged state. TTF and AZO were chosen as the oxidation mediators. The cell was charged by holding at 3.4 V for TTF and 3.7 V for AZO until 1 mAh charge passed prior to further chemical characterisations. For NMR analysis, the electrodes and separators were rinsed with 0.7 ml of CDCl3, and measurements were recorded on a Bruker spectrometer (400 MHz).

### Electrochemical measurements

SECM experiments were performed with SECM bipotentiostat (CHI 920) in an Ar-filled glovebox. Prior to kinetics measurement, the NG factor of Au tip was determined by approaching a completely insulating surface and fitting the negative approach curve. The data processing and fitting process were described elsewhere35,36,37. A dimensionless rate constant, κ, was obtained by data fit, which equals to kapp ro/D, where r0 is the radius of tip and D is the diffusion coefficient of redox mediators. D of various mediators were determined by measure steady-state current of a Au microelectrode with known radius r0, according to iss=4nFDroC.

The redox potential and heterogeneous electron transfer rate constants ko of redox mediators itself were determined using cyclic voltammetry(CV) measurements. The redox potential is determined by the centre of two redox peaks, which is measured in a 100-mM LiTFSI–tetraglyme solution with 10 mM of various mediators at a Au electrode. Partially charged LiFeO4 (LFP) protected by a glass frit served as an RE and it gave a potential of 3.45 V vs. Li+/Li as reported previously. A platinum wire served as a CE. The details of ko measurement are described elsewhere43. Briefly, CVs were recorded at various scan rates, ranging from 0.05 to 10 V s–1. Ψ, a function of CV peaks separation, was plotted vs. root of scan rate and a linear fit was applied. ko was obtained from the slope of linear fit. The ko measurement was carried out at both Au and glassy carbon (GC) WEs. Due to the non-negligible resistance of ether-based electrolytes, an Ohmic overpotential correction was applied to account for the uncompensated resistence during  CV measurements and a silver wire RE was used.

### Characterisations

For the surface characterisations, the Li2O2 disk was immersed in 1 M LiTFSI–tetraglyme solution for 3 h prior to XPS and TOF-SIMS experiments. Both pristine disk and treated disk were characterised in an air-sensitive holder. To measure the TOF-SIMS of bulk Li2O2, the data were recorded after 2 min etching.

### Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. Background data has been deposited in the Oxford Research Archive (ORA) at: https://ora.ox.ac.uk/objects/uuid:c23a0cc0-55b5-455f-bb68-a14d8ea2e3bc.