Review Article

# Advances in understanding mechanisms underpinning lithium–air batteries

• Nature Energy 1, Article number: 16128 (2016)
• doi:10.1038/nenergy.2016.128
Accepted:
Published online:

## Abstract

The rechargeable lithium–air battery has the highest theoretical specific energy of any rechargeable battery and could transform energy storage if a practical device could be realized. At the fundamental level, little was known about the reactions and processes that take place in the battery, representing a significant barrier to progress. Here, we review recent advances in understanding the chemistry and electrochemistry that govern the operation of the lithium–air battery, especially the reactions at the cathode. The mechanisms of O2 reduction to Li2O2 on discharge and the reverse process on charge are discussed in detail, as are their consequences for the rate and capacity of the battery. The various parasitic reactions involving the cathode and electrolyte during discharge and charge are also considered. We also provide views on understanding the stability of the cathode and electrolyte and examine design principles for better lithium–air batteries.

Society will need energy storage that exceeds the limits of current technologies if we are to significantly reduce CO2 emissions. The lithium-ion battery has been a spectacular success and will continue to improve incrementally for years to come. However, it is imperative that we investigate now alternatives that offer the possibility of going beyond the limits of Li-ion technology, if we are to have any hope of meeting the energy storage needs of future generations1,​2,​3,​4.

The Li–air battery, which uses O2 derived from air, has the highest theoretical specific energy (energy per unit mass) of any battery technology, 3,500 Wh kg−1 (refs 5,6). Estimates of practical energy storage are uncertain, as many factors are unknown, but values in the range 500 to 1,000 Wh kg−1 — sufficient to deliver significantly in excess of a 500 km driving range if employed as an electric vehicle battery — have been proposed (Fig. 1a)7. The fuel at the positive electrode, O2, is freely available. Operation of the aprotic Li–O2 battery, which uses a non-aqueous-based electrolyte, is illustrated in Fig. 1b. Although Li–air batteries have the energy storage advantage over many other rechargeable batteries, understanding of the underpinning chemistry and electrochemistry is lacking. Li–air combines two challenging electrodes, Li metal and O2. Li-metal electrodes have been investigated for many years and still do not deliver the necessary cycling efficiency (ratio of discharge/charge capacity) and related suppression of dendrites8. However, the prospect of a protected Li anode — in which a solid electrolyte is interposed between the Li metal and the liquid electrolyte — offers an important path of exploration towards a functional high-energy Li battery9. In contrast, when the aprotic Li–air battery began to be explored, there were only a handful of papers on O2 reduction in relevant electrolyte solutions, rendering understanding of the O2/Li2O2 redox reaction in aprotic solvents a high priority. Aside from the O2/Li2O2 redox couple, aprotic Li–O2 faces a number of other challenges, not least of which is the stability of the electrolyte solution and the cathode towards reduced oxygen species. All of these processes need to be understood and mastered if we are to achieve progress towards a practical battery. Even then, the issue of air handling and filtering would need to be addressed by new chemical engineering solutions. Nonetheless, only through a deeper understanding of the science and engineering underpinning the Li–air battery can we hope to make reliable, evidence-based predictions of whether it can be a viable technology.

In this Review, we focus on the topics that were least understood and on which there has been the greatest progress over the past few years, including the mechanism of O2 reduction to Li2O2 at the cathode on discharge and the oxidation of Li2O2 on charge, and the mechanisms of electrolyte and cathode degradation. As with all such potentially game-changing ideas, they begin with over-hyped expectation, followed by the inevitable disillusionment, and thereafter equilibrium is often established; Li–O2 is now in this third phase10. There has been real progress in understanding the fundamental science underpinning Li–air, progress that has explained why many previous results led to the belief that Li–O2 would inevitably be a low rate, low capacity, high polarization cell, whereas new work on the fundamentals is revealing this is not necessarily the case. No one yet knows whether Li–O2 will ever be a technology, but we owe it to society and our future to explore what might be possible.

## Mechanisms at the aprotic O2 electrode

Unlike O2 reduction in fuel cells, which produces H2O, the product at the positive electrode in the Li–O2 cell is the insulating solid Li2O2, and this has profound implications for the operation of the cell.

A typical plot of voltage as a function of Coulombs passed on the discharge and charge cycle of a Li–O2 cell is shown in Fig. 1c. The kinetics of the Li anode are relatively facile, so the deviations from the thermodynamic potential (open circuit potential) on discharge and charge (overpotentials) in Fig. 1c are dictated primarily by reactions at the cathode. The ideal cathode reaction is shown in equation (1).

$(1)2Li++2e−+O2⇌ChargeDischargeLi2O2$

Many factors can influence the overpotentials, such as the deposition of side-reaction products from the electrolyte and electrode degradation. These factors and how they may be mitigated are discussed in subsequent sections. However, the mechanism of the reversible reaction described by Equation (1) lies at the heart of operation of the Li–O2 battery and understanding it is key to addressing the technological challenges; hence, we begin with this below.

O2 reduction to Li2O2 on discharge. The processes that occur on discharge are highly dependent on competitive factors involving the effective current density and overpotential and whether the LiO2 intermediate is dissolved in solution or adsorbed on the electrode surface11,​12,​13. At high overpotentials and high current densities, O2 is reduced to Li2O2, which grows as a film on the electrode surface. However, batteries must operate at low overpotentials and relatively high currents to maximize energy density (and must be designed to do so). At low overpotentials, Li2O2 has been observed to grow as surface films or large particles from a solution process, depending on the solvent or salt from which the electrolyte solution is formed or depending on additives in the electrolyte solution (Fig. 2)11,​12,​13,​14,​15,​16. Bulk Li2O2 exhibits ionic conductivity via lithium vacancies, and electronic conductivity via electron holes, both being very low17.

At low overpotentials, whether Li2O2 forms as a film on the electrode or particles from solution is controlled by the solubility of the LiO2 intermediate13. The O2 reduction mechanism is summarized in Fig. 3. The first step is the one-electron reduction of O2 (dissolved in the electrolyte solution at the electrode surface) to form LiO2. An equilibrium exists between LiO2* adsorbed on the electrode and LiO2 dissolved in the electrolyte solution11, the position of which is governed by the competition between LiO2 solubility and the adsorption free energy of LiO2* on the electrode; LiO2 solubility is a significant factor (Fig. 3)13,14,16. As is generally the case for salts dissolved in aprotic solvents (no hydrogen bonding), the solubility depends primarily on solvation of the cations by the solvent molecules (Gutmann donor number (DN)18 and the ionic dissociation strength)19, although interactions between the O2, ion and high acceptor number additives were also observed to play a role in LiO2 solubility14,16. Where a solvent has a sufficiently high donor number to strongly solvate Li+ (for example, dimethyl sulfoxide (DMSO), DN = 30), then LiO2 is dissolved mainly in the electrolyte solution, where it disproportionates to Li2O2 that grows as micrometre-sized particles (Fig. 3)13,14.

If, on the other hand, Li+ is weakly solvated (for example, acetonitrile CH3CN, DN = 14), then LiO2 is present primarily on the electrode surface (Fig. 3), where it undergoes a second electron reduction or disproportionation to form a Li2O2 film on the electrode.

It is interesting that the fate of Li2O2 in aprotic solvents (that is, without protic additives) depends primarily on the solvation of Li+ and not the species actually formed on reduction, O2, (ref. 16). Superoxide radicals can be stabilized indirectly by the electrolyte counter anions20,​21,​22. In solutions containing highly associated lithium electrolytes, the counter anions are strongly coordinated to the solvated lithium cations. Somewhat analogous to high donor number solvents strongly solvating the Li+ ions, this stabilizes the LiO2, lowering its free energy and favouring its presence in solution. The LiO2 species in solution can then go on to disproportionate to Li2O2. Hence, superoxide moieties formed by oxygen reduction can be stabilized even in solvents with a low donor number, such as glymes20.

Additives to the electrolyte solution can also influence the LiO2 solubility, through solvation of Li+ or by direct solvation of superoxide using Lewis acidic (electron accepting) electrolyte compositions. For example, when added at impurity (<1 vol%) levels, water, a strong Lewis acid, along with other protic compounds, can induce solution growth of Li2O2 through interactions with reduced oxygen species14,23,24. Water unfortunately also induces other parasitic reactions within the cell, such that a search is needed for other, more stable additives that provide a solution pathway to Li2O2 formation and hence increased cell capacity. Additive design criteria to provide both appropriate stability and oxygen species solvation are still not entirely understood.

Understanding the O2 reduction mechanism has important implications for practical Li–O2 batteries. Much work in the literature has employed ether-based electrolyte solutions. The above mechanistic understanding explains why these low donor number solvents (DN ≈ 20) result in significant Li2O2 film growth on the electrode. Such insulating films can grow only to 6–7 nm, resulting in electrode passivation25 and leading to low capacities, low rates and early cell death. In contrast, electrolyte solutions that dissolve LiO2 minimize surface film growth, resulting in high rates, high capacities and sustained discharge11,​12,​13. This changes thinking regarding the design of porous cathodes for Li–O2, away from high surface areas and towards electrodes with a high proportion of large pores to store the Li2O2 particles growing from solution. Li–O2 cells with such electrodes can deliver relatively high specific energies (gravimetric) and energy densities (volumetric) compared with those that rely on surface films of Li2O2 (ref. 13).

The LiO2 intermediate is a disadvantage, especially in its solubilized form. It is reactive, especially towards high donor number solvents that might be used to dissolve LiO2, it is reactive towards possible electrode materials and it pins the discharge potential at a value somewhat lower than the open circuit potential, even in the absence of mass transport limitations26. Very recently, an approach has been described that persuades reduction of O2 to Li2O2 along a different pathway, avoiding LiO2 and its disadvantages27. A molecule such as 2,5-di-tert-butyl-1,4-benzoquinone (DBBQ) is added to the electrolyte solution. On discharge, it is first reduced to DBBQ at the cathode, then binds Li+ and reduces O2 to O2, in the process forming the complex LiDBBQO2 in solution, which in turn disproportionates to Li2O2 and reforms DBBQ. The cycle repeats to sustain discharge27. By avoiding LiO2, relatively stable electrolyte solutions that do not dissolve LiO2 for example, lithium bis(trifluoromethane)sulfonimide (LiTFSI) in dimethoxyethane (DME), may be used while still forming Li2O2 in solution, resulting in significantly higher discharge rates and higher capacities (typically by 80-fold) than is the case in the absence of DBBQ in the same DME electrolyte (Fig. 4). The new intermediate, LiDBBQO2, is more stable (lower free energy) than LiO2, thus lowering the overpotential on discharge (Fig. 1c). Another benefit of avoiding reactive solution-soluble LiO2 is potentially reduced side reactions, evidenced by higher yields of Li2O2 (95% with DBBQ in DME compared with 87% without). This reduction mechanism may be distinguished from previous studies that used molecular shuttles, which reduce O2 by outer-sphere electron transfer at potentials lower than direct reduction, for example, viologen, or by molecules that bind O2 before and after reduction, for example, pthalocyanines28,​29,​30.

Other discharge mechanisms that result in different products have also very recently been reported. Lu et al. suggested that it may be possible to stabilize LiO2, which is known to be stable only at <50 K (refs 31,32), by adsorption on an iridium-based electrode such that this becomes the product of the reaction rather than an intermediate33. Such a cell reportedly shows much better reversibility than the O2/Li2O2 reaction. It should be noted that work continues on the aqueous Li–O2 battery and its related Zn–O2 cousin34,35. The fascinating prospect of a Li–O2 aprotic battery in which the reaction at the cathode involves the formation of LiOH has also been raised recently, although the mechanism is still not yet clear36.

Li2O2 oxidation to O2 on charge. Electrochemical oxidation of Li2O2 does not require a true electrocatalyst — as does oxidation of LiOH in non-aqueous media, for example — because the O–O bond is not cleaved on reduction. Oxidation of thin Li2O2 films on the electrode surface may be relatively facile, whereas oxidation of large particles of insulating Li2O2 is a challenge (see the following sections). Theoretical studies intriguingly suggested that the peroxide surface exhibits a half-metallic state37, but experimental proof of surface conductivity is lacking to date.

Many Li2O2 oxidation mechanisms have been invoked to provide a description of galvanostatic charge potential profiles. Elucidating the mechanistic origin of the charge profile is complicated by its dependence on two factors: (1) the discharge Li2O2 deposition mechanism, and hence the Li2O2 morphology; and (2) formation of products from parasitic reactions involving the cathode and electrolyte during discharge and charge. These variables are intrinsically linked to the cell composition (that is, the cathode and electrolyte employed), and numerous questions about the influence of these complexities on the mechanism and charge overpotential still remain to be answered. In theory, low overpotentials (<0.2 V) exist for O2 evolution from Li2O2 at many Li2O2 crystal facets, as calculated using density functional theory38. This is in agreement with observations of O2 evolution at low (3 V versus Li+/Li) potentials at early stages of charge from conformally deposited Li2O2. The ever-increasing overpotential observed in these cells was proposed to be related to the deposition of solid carbonates at the Li2O2/electrolyte interface39,40 (see below).

Studies on charging electrodes pre-loaded with crystalline lithium peroxide have helped disentangle the oxidation of Li2O2 from side reactions occurring on reduction related to Li2O2-induced decomposition of both the cathode and electrolyte. Such reactions deposit products on the peroxide that exhibit very high oxidation potentials, and thus inhibit Li2O2 oxidation to O2 (refs 24,41). First, the studies demonstrate that the surface chemistry of the cathode support is critical in determining the efficiency of electron transfer to the insulating Li2O2 (ref. 42). Studies using preloaded cathodes have also elucidated the role of passivating films formed on surfaces, such as nitrides and carbides, that can either shut down or facilitate charging, as further discussed below. Second, micrometre-sized crystalline peroxide particles exhibit a rather low charge overpotential (only 600 mV for passivated metallic TiC cathode surfaces), much lower than one might expect. This is in agreement with the above finding — namely that Li2O2 can be rather readily oxidized if decomposition products can be avoided to insure facile transport across interfaces.

Various mechanisms have been proposed for oxidation of Li2O2 deposited as toroids or incorporated into the cathode as a macroscopic particle-containing powder. Ganapathy et al. used operando X-ray diffraction to show that a Li-deficient component (that is, Li2xO2) is formed during the charging process, presumably as a result of a one-electron Li+ de-insertion43. In fact, this mechanism was first proposed by theoretical studies that showed that topotactic delithiation based on Li2xO2 is rendered accessible at relatively small overpotentials of 0.3–0.4 V (ref. 44). Li2O2 formed electrochemically via the solution mechanism described above often crystallizes in flat platelets, as determined from X-ray broadening, and as also observed by imaging45. In ref. 43, these are embedded in an amorphous lithium-bearing component (Fig. 5a), which may be comprised of both Li2O2 and side products arising from electrolyte degradation, such as formate, that can be oxidized at relatively low potentials without a catalyst41. Such products, formed on discharge, probably account for the less than 100% yield of peroxide (with respect to electrons passed in the cell) as determined by detailed assay studies46. The non-crystalline phase is removed first (Fig. 5a,b) and at higher potentials, the underlying crystalline peroxide is proposed to charge via the Li-deficient solid solution (Li2xO2) phase. The small actively oxidizing fraction results in a gradual reduction of the Li2O2 crystallites until their complete disappearance (Fig. 5c,d). Although the mechanism of oxygen release is not yet clear, it is probable that Li-deficient Li2xO2 exists at the surface of Li2O2 in the cathode and immediately undergoes disproportionation as it forms to evolve O2.

We note that some studies have suggested that the initial charging regime at low potentials is due to oxidation of a LiO2/Li2O2-type material formed on discharge based on observation of paramagnetic superoxide domains based on Raman and magnetic measurements47. While Li-deficient species are undoubtedly present at the peroxide surface, the Raman signature of the superoxide species is very close to that of the product formed by decomposition of the binder polyvinylidene fluoride (PVDF)48, which has been well established to occur49, thus complicating the assignment. In fact, importantly, the signature peak does not appear when using Teflon as a binder in the cathode48. Further work is needed to resolve this open question.

Soluble oxidation mediators for improved charging efficiency. Although solution growth of Li2O2 is advantageous on discharge, as described above, it makes the charging of Li–O2 cells more challenging, because of the need to oxidize relatively large Li2O2 particles in the pores of the cathode somewhat remote from the electrode surface. Oxidation mediators can address this problem. They are molecules dissolved in the electrolyte that are oxidized at a potential slightly above the equilibrium potential of Li2O2 formation. Once oxidized at the electrode surface, they diffuse to and oxidize Li2O2 particles. The substantially reduced charging overpotential is evident in the presence of an oxidation mediator, as shown in Fig. 1c. Important properties include a high diffusion coefficient, fast charge transfer kinetics (particularly the charge transfer associated with Li2O2 oxidation) at voltages approaching the Li2O2 formation potential, and high stability. Many redox mediators have been proposed and explored50. Substantial interest in oxidation mediators was spurred by Chen et al.'s report of tetrathiafulvalene as an efficient mediator51. While the oxidation potential of tetrathiafulvalene is higher than desired, recent reports have demonstrated other oxidation mediators with potentials closer to that for Li2O2 oxidation, including tris(4-(diethyl amino)phenyl)amine52, 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO)53,54, LiI (refs 55,​56,​57) and a variety of quinone analogues and other redox-active molecules58,59, that lower charge overpotentials without severely compromising electrolyte stability. Nevertheless, the search for a perfectly stable mediator with an appropriate operating voltage is still an important challenge. We note that the mediator approach requires the Li anode to be protected by a solid electrolyte membrane to avoid reaction of any mediator that diffuses to the anode.

## Li–O2 battery electrolytes

Perhaps the largest scientific challenge facing Li–O2 batteries is long-term stability of the electrolyte. Li–O2 cell configurations employ a liquid electrolyte comprised of a Li salt dissolved in an aprotic organic solvent. These electrolytes encounter highly reactive conditions. There are a plethora of critical reactions that the highly reactive reduced oxygen species (RROS; for example, O2, O22−, HOO and HO) can undergo with various polar aprotic solvents. Over the past half-century, in-depth studies have been carried out on the reactivity of these reduced oxygen species in general, and superoxide in particular, towards a wide-range of organic substrates60,61. While Li cations are highly electrophilic, in aprotic media, superoxide can function as a supernucleophile (equations (2) and (3)), as an efficient base (equation (4)) and as a good electron-transfer agent (equation (5)).

$(2)RX+O2·−→X−+ROO·$$(3)RCOX+O2·−→RCOO−+XO−$$(4)RCH2CH2X+O2·−→RCH=CH2+HO2−$$(5)A+O2·−→A·−+O2$

Polar aprotic solvents (for example, propylene carbonate, DMSO, dimethylacetamide (DMA), N,N-dimethylformamide (DMF), all contain heteroatoms and, hence, polarized bonds. These polar solvents can undergo thermodynamically favourable reactions with RROS in the presence of Li ions60. A good example are alkyl carbonates, which undergo nucleophilic attack that is affected either by the electrochemically formed superoxide anion radicals (O2•−; equation (3)) or by the peroxide dianion (O22−) (refs 62,​63,​64). Polyethers, such as glymes, are not highly polar. Hence, they are generally more inert solvents, although they too can undergo limited attack on α-H and β-H (different possible locations in the polyether molecules for nucleophilic attack) by superoxide and peroxide species formed by oxygen reduction49,65,​66,​67,​68 (equations (6) and (7)).

$(6)(CH3OCH2CH2)2O+Li+ −OO−Li+→CH3OO−Li++CH3O(CH2)2O(CH2)2O−Li+$$(7)(CH3OCH2CH2)2O+Li+ −OO−Li+→CH3O−Li++CH3O(CH2)2O(CH2)2OO−Li+$

Reactivity is attributable to two factors. One is that not only the superoxide, but also the peroxide, is considered a supernucleophile owing to the adjacent pairs of non-bonding electrons69. The actual nucleophilicity is much higher than would have been expected based on its basicity. The second factor is the omnipresence of the highly electrophilic lithium cations. The lithium cations bond strongly to the lone pairs of the oxygen atoms of the ether molecules, which are hard Lewis bases. This promotes conversion of the alkoxy groups of polyethers into much better leaving groups, facilitating their loss upon nucleophilic attack.

Despite their relatively inactive nature, glyme molecules can be also attacked by O2•− or O22− moieties, which act as bases. While peroxide and superoxide species are not necessarily strong bases (for example, the pKa of HOO is 4.7), the presence of Li+ ions — which are strong Lewis acids — in solutions may facilitate concerted elimination reactions of glymes. The Li ions interact with the oxygen atoms of the ethers, thus enabling formation of Li–alkoxide leaving groups (for example, CH3OLi) simultaneous with proton abstraction by superoxide or peroxide moieties.

Another important possible route of reactions may relate to the radical nature of superoxide moieties. While the superoxide anion radical is a ‘super’ nucleophile, and it is not considered a strongly active radical moiety, evidence of its (or the hydroperoxy radical's) deleterious effect exists70.

Finally, it was also suggested that superoxide can induce the oxidation of various substrates via initial hydrogen atom abstraction71. However, thermochemical calculations and experimental data may suggest that superoxide is not a strong enough oxidizer. Any oxidation process observed during oxygen reduction reactions (ORR) may result from base-catalysed autoxidation (rather than from oxidation by superoxide). This suggestion certainly warrants further study. Consequently, it can be concluded that the partial instability of polyether solvents during oxygen reduction in the presence of Li ions stems from a combination of nucleophilic substitution and elimination reactions as major possible side reactions38.

In previous work, Li2O2 was found to be reactive with glymes as it was electrochemically formed, and prolonged exposure of Li2O2 at low currents resulted in lower Li2O2 yields during discharge46. Based on extensive work and observation, the side reactions of these solvents in Li–O2 cells are most pronounced and fast when the oxygen reduction products are present in the solution phase, before final precipitation71. To overcome the instability of the ethereal solutions, Nazar et al. protected the ethylene oxide backbone of DME by methylation72. These analogues demonstrated better cyclability of Li–O2 cells and less accumulation of side products compared with unprotected DME.

DMSO-based electrolyte solutions and carbon-free electrodes, such as Au (ref. 73), and TiC (ref. 74), have demonstrated impressive cycling behaviour. However, as with all solvents studied so far, DMSO reacts with superoxide radicals75, and also peroxide76,77, to generate the dimsyl anion and a hydroperoxy radical nucleophile78. The hydroperoxy radical can attack the sulfur atom of the sulfoxide moiety, yielding the corresponding dimethyl sulfone ((CH3)2SO2) and lithium hydroxide. DMSO also reacts with lithium metal, and thus lightly protected anodes and/or a combination of the abovementioned reactivity may underlie variability in reported results for this electrolyte solvent.

Given the particularly aggressive nature of RROS species towards organic molecules, exploration of purely inorganic electrolytes may be an interesting direction of research. To this end, solid inorganic ceramics have been employed to serve as both a Li–O2 electrolyte and an oxygen and water-impermeable Li-metal protective layer34,79,​80,​81,​82. Engineering a low-impedance porous cathode/electrolyte interface with long-term stability will be a critical challenge facing this cell architecture. Recently, an intermediate temperature (150 °C) molten salt electrolyte, based on the LiNO3–KNO3 eutectic, was also found to be stable in the presence of the Li–O2 cathode electrochemistry83. Identifying lower-temperature inorganic molten salt alternatives and cathodes with improved stability (discussed in a later section) are useful directions to explore when employing these electrolytes.

In summary, it appears that all ‘commercial’ off-the-shelf electrolytes examined to date exhibit instability to superoxide and peroxide moieties formed during oxygen reduction, in the presence of Li ions. Some modifications, for example, using fluorinated solvents, may enhance stability, but severely reduces Li-salt solubility. Hence, a compromise must be found, which could arise from designer electrolytes.

## Designing better cathodes

Many important properties have to be considered when designing a practical Li–air battery cathode. Li2O2 is an electronic insulator, insoluble (or very sparingly soluble) in all known organic liquid-based electrolytes and a strong oxidizer. As a result, a cathode has to be designed to provide sufficient electronic conductivity and Li+/O2 transport, be resilient towards Li2O2-induced oxidation and not promote parasitic processes associated with the electrolyte and Li2O2 (Fig. 6). Typical Li–O2 battery cathodes usually comprise a mixture of high surface area carbon or catalyst powders bound to a porous current collector using a polymer binder. A large number of materials have now been studied for each of these cathode components, and general guidelines for material development are still emerging. This section will highlight useful cathode material properties and the effect of cathode composition on the performance of a Li–air battery.

Heterogeneous catalysis at the cathode. Substantial effort in the Li–air field has been devoted to identifying heterogeneous electrocatalysts to reduce overpotentials of both the oxygen reduction (discharge) and, in particular, oxygen evolution (charge) reactions. However, given the problematic stability issues observed with the electrolyte and cathode carbon, careful quantitative analysis of product formation must always be performed to confirm any claims of true Li2O2 formation and oxidation electrocatalytic performance84. In fact, when such quantitative measures are taken, parasitic electrolyte decomposition has been observed as the primary reaction being catalysed when using efficient aqueous ORR electrocatalysts85. For example, the best known ORR catalyst, Pt, was observed to most efficiently catalyse these parasitic reactions, allowing electrolyte oxidation to occur at potentials (<3.5 V) that, by themselves, would lead to erroneous conclusions about efficient Li2O2 oxidation85. In addition, in cases where Li2O2 forms as a film on the electrode surface, the presence of a heterogeneous electrocatalyst on the electrode would be of little value beyond the first monolayer. It has been shown by Viswanathan et al. that electrochemical formation and oxidation of Li2O2 films is actually quite facile on glassy carbon surfaces, with constructed Tafel plots for this redox reaction being similar to what would be expected of a fast, reversible reaction with an exchange current density of 1 mA cm−2 (Fig. 7a)86. In fact, Viswanathan et al. make a compelling argument that reducing overall cell impedance is substantially more important to eliminate overpotentials than improved ORR electrocatalysis.

On charge, the apparently large overpotential has been linked to deposition of solid electrolyte decomposition products at the O2 evolving surface and the Li2O2/electrolyte interface (Fig. 7b), and not to poor Li2O2 oxidation kinetics39,86. As they are continuously formed, these carbonaceous products cover an increasingly larger portion of the Li2O2/electrolyte interface, therefore necessarily driving the charging potential to higher values over the course of the charging process. At the end of charge, the decomposition products oxidize at high (>4.2 V versus Li+/Li) potentials. The identification of stable electrolytes that are not susceptible to oxidation, even with a carbon-only cathode, should result in improved energy efficiency. For example, the molten nitrate eutectic electrolyte mentioned previously exhibits high stability and measurable Li2O2 solubility, allowing Li2O2 to diffuse back to electronically accessible cathode surfaces83. As a result, extremely low overpotentials (50 mV at 0.25 mA cm−2) throughout discharge and most of charge, with analytically confirmed 2e ORR and oxygen evolution reactions (OER), were observed when employing this electrolyte in a carbon cathode-based cell.

Cathode stability. Although carbon has many desirable characteristics (high electronic conductivity and surface area; cost effective) that make it widely used as a cathode material, it degrades during Li–O2 battery operation. For example, 13CO2 evolution was observed during charge from a Li–O2 cell employing a cathode composed of high purity 13C carbon black, indicating that a parasitic reaction between Li2O2 and C occurs during cell operation, particularly at oxidative potentials (Fig. 7b)39,87. Thotiyl et al. and Itikis et al. observed that this Li2O2/C reaction is influenced by the carbon surface composition, with a more hydrophilic and defective carbon surface resulting in faster carbon degradation87,88.

Although these studies hold hope that a resilient carbon cathode could eventually be engineered, the search for alternative stable materials is also underway. Au and TiC were initially identified to provide more stable cycling than carbon73,74. Theoretical work confirms that growth of Li2O2 on TiC occurs via a surface conduction mechanism89. Metallic nitrides are fully oxidized by Li2O2 when an anodic current is applied, completely inhibiting further electron transfer42. However, thermodynamically more stable metallic carbides, such as TiC, form a thin layer of conductive TiO2x, which acts to passivate the surface and inhibit further oxidation42,74. Li2O2 oxidation is facilitated with a greatly decreased overpotential. Not all carbides are so stable. Mo2C reacts to form a surface layer of MoO2 on discharge, which appears to result in a low charge overpotential90, but in fact forms soluble LixMoO3 and leads to electrode degradation91. Thus, practical cathode materials must be designed such that their surfaces inhibit excessive oxidation and/or are concealed by conductive oxide layers92,​93,​94. Deposition of thin metal-oxide coatings, such as Al2O3, has also been used as a strategy to protect carbon from degradation95. Furthermore, electrocatalytic activity towards electrolyte decomposition also has to be taken into consideration when designing a new electrode material (Fig. 6).

Also of note, the stability of certain polymer binders has been called into question, with polyvinylidene fluoride, a typical binder used in Li-ion batteries, being identified by Black et al. as unstable during battery operation49. Higher stability alternatives include polyethylene, polytetrafluoroethylene and Nafion, as shown from studies on chemical stability of polymers in the presence of Li2O2 by Nasybulin et al.96.

Cathode passivation limiting cell capacity. The primary advantage that Li–air batteries potentially have over current state-of-the-art Li-ion batteries is their high theoretical specific energy. Of course, to attain high energy densities, the electrochemical capacity of the cell should be very high, with the required minimum Li–O2 capacity ultimately being a function of the battery pack configuration and composition. The projected required capacities by Gallagher et al. and Christensen et al. (5–30 mAh cm−2) are relatively large compared with most reported capacity values found in the literature7,97.

Numerous potential causes have been conjectured for the observed poor capacity, all of which are related to processes occurring at the porous cathode (rather than the Li-metal anode). These include pore clogging induced by solid Li2O2 deposition, O2 transport limitations and charge transport limitations7. When the surface mechanism for Li2O2 formation is dominant, cell death has been clearly linked to a substantial increase in charge transfer resistance related to a growing conformal Li2O2 film13,14,25,98. Charge transfer through the growing Li2O2 film was probed by Viswanathan et al. to show the direct influence of charge transfer resistance and cell death at a flat, nonporous electrode, where pore clogging was not a possible cause for cell death98. A clear correlation therefore exists between cell capacity and the total cathode surface area when Li2O2 film growth is observed, as was reported by Meini et al.99. As was discussed earlier, the ideal cathode design in a cell with a dominant solution mechanism of Li2O2 is likely different, with low surface area and high porosity possibly required.

## Outlook

There has been real progress in understanding the fundamental chemistry and electrochemistry underpinning the aprotic Li–O2 battery. Work on the mechanism of the reaction at the positive electrode (equation (1)) has shown that electrolyte solutions that strongly solvate Li+ result in O2 reduction to Li2O2 occurring in solution, whereas solutions that weakly solvate Li+ result in Li2O2 films on the electrode surface. Much recent Li–O2 work has focused on ether-based electrolytes, because of their superior stability compared with alternatives (especially those with high donor numbers). We can now understand why the use of such low donor number ethers leads to Li2O2 surface films, passivating the electrode and resulting in low rates and early cell death, whereas the use of high donor number electrolyte solutions, dominated by Li2O2 growth in solution, can lead to relatively high rates and high capacities. Nevertheless, low donor number solvents are attractive because they are generally more stable, being less susceptible to nucleophilic attack by reduced oxygen species. Recent work has shown that the pathway by which O2 is reduced to form Li2O2 can be altered, such that the intermediate LiO2, reactive towards electrolyte solutions, can be avoided and O2 reduction to Li2O2 can take place in solution despite the use of low donor number ethers. As a result, Li–O2 cells using ether-based electrolyte solutions can deliver high rates and high capacities, avoiding early cell death. This is achieved using a solution soluble mediator, which also raises the discharge potential.

Understanding charge overpotentials after Li2O2 has been formed via the solution mechanism is an important direction of future research. Charge transport associated with toroids likely contributes to the overpotentials observed on charge, and hence the interest in developing stable oxidation redox mediators. However, other factors could also substantially contribute to the observed overpotential, such as electrolyte and cathode degradation. It is worth comparing the solution mechanism of Li2O2 formation to the chemistry of the Na–O2 battery, where NaO2 rather than Na2O2 is the product, formed as large crystals tens of micrometres in size100. NaO2 is solubilized via a phase transfer catalyst that operates both on discharge and charge, leading to little hysteresis and a low overpotential on charging despite the large crystallite size101. Thus, to take advantage of Li2O2 growth, it is necessary to utilize soluble oxidation mediators to oxidize the insoluble Li2O2 particles on charging. In effect, the reaction at the positive electrode becomes one in which Li2O2 acts as the energy storage medium, coupled electrochemically to the electrode surface via molecular mediators.

Work on the effect of H2O has shown that in contrast to earlier expectations, several 100s of ppm H2O can be tolerated with the overall reaction still being dominantly Li2O2 formation and oxidation However, there is evidence that H2O changes the reaction pathway. Importantly, the proton activity of H2O in an aprotic organic solvent can be much less than in water, explaining in part why addition of significant quantities of H2O does not radically change the overall reaction (that is, Li2O2 is still the dominant discharge product even in a water-contaminated cell).

Although advances in understanding the underlying mechanisms has opened up promising directions for Li–O2 cells exhibiting good rate and capacity with cyclability, challenges remain. In particular, in our opinion, these are the problems of electrode and electrolyte stability. Although there have been important advances in exploring cathode materials, including modified carbons, breakthroughs are still required to demonstrate a low-cost, low-mass, conductive and highly stable porous gas diffusion cathode. A deeper understanding of the decomposition reactions occurring in the electrolyte solution is providing a sound foundation of fundamental understanding that will define directions for the exploration of more stable electrolytes. It is important to place electrode and electrolyte stability in context. Most batteries are thermodynamically unstable and operate because of kinetic stability. Namely, there are always decomposition reactions, such as in the case of lead–acid batteries. It is not a question of whether the electrode or electrolyte is stable or unstable in Li–O2 but rather is the degree of instability sufficiently small to not decrease performance over the cycle and calendar life of the battery. The nature of the decomposition products is important. If these are benign and soluble, they may not degrade performance, whereas if they are insulating and deposited on the electrode surface, they strongly limit cell lifetime. Comparing Li–O2 with other so called ‘beyond Li-ion’ batteries, Li–S has been commercialized, but not in a form that takes full advantage of its high specific energy. Rechargeable divalent metal batteries, Mg, Ca, and Zn, still present formidable challenges. The jury is out on which, if any, beyond Li-ion batteries will make it to the market place.

In conclusion, disruptive technologies do not happen overnight. Generally, years of effort and fundamental understanding and its exploitation in addressing challenges are required. In fields outside batteries this has been well illustrated and well understood. The recent example of the commercialization of organic light-emitting diodes, which, some 20 years ago was identified as a potentially promising game-changing technology, illustrates this reality. It is not yet known whether the problems that remain a challenge for the aprotic Li–O2 battery will ultimately be solved, but the last few years have identified many promising pathways forward to a solution.

## References

1. 1.

, & The Lithium Air Battery: Fundamentals (Springer, 2014).

2. 2.

, , , & Aprotic and aqueous Li–O2 batteries. Chem. Rev. 114, 5611–5640 (2014).

3. 3.

Prospects and limits of energy storage in batteries. J. Phys. Chem. Lett. 6, 830–844 (2015).

4. 4.

et al. The lithium/air battery: still an emerging system or a practical reality? Adv. Mater. 27, 784–800 (2015).

5. 5.

& A polymer electrolyte-based rechargeable lithium/oxygen battery. J. Electrochem. Soc. 143, 1–5 (1996).

6. 6.

, , , & Rechargeable Li2O2 electrode for lithium batteries. J. Am. Chem. Soc. 128, 1390–1393 (2006).

7. 7.

et al. A critical review of Li/air batteries. J. Electrochem. Soc. 159, R1–R30 (2012).

8. 8.

et al. A review of solid electrolyte interphases on lithium metal anode. Adv. Sci. 3, 1500213 (2016).

9. 9.

, , & Li–O2 and Li–S batteries with high energy storage. Nat. Mater. 11, 19–29 (2012).

10. 10.

, , , & Quantifying the promise of ‘beyond’ Li-ion batteries. Transl. Mater. Res. 2, 045002 (2015).

11. 11.

et al. Current density dependence of peroxide formation in the Li–O2 battery and its effect on charge. Energy Environ. Sci. 6, 1772–1778 (2013).This paper and ref. 12 showed that Li2O2 can form as either a surface film or particles depending on the current density.

12. 12.

et al. Rate-dependent morphology of Li2O2 growth in Li–O2 batteries. J. Phys. Chem. Lett. 4, 4217–4222 (2013).

13. 13.

et al. The role of LiO2 solubility in O2 reduction in aprotic solvents and its consequences for Li–O2 batteries. Nat. Chem. 6, 1091–1099 (2014).This paper demonstrated that the solvation of the intermediate, LiO2, controls the mechanism of discharge.

14. 14.

et al. Solvating additives drive solution-mediated electrochemistry and enhance toroid growth in non-aqueous Li–O2 batteries. Nat. Chem. 7, 50–56 (2015).

15. 15.

et al. Toward a lithium–“air” battery: the effect of CO2 on the chemistry of a lithium–oxygen cell. J. Am. Chem. Soc. 135, 9733–9742 (2013).

16. 16.

et al. Experimental and computational analysis of the solvent-dependent O2/Li+-O2- redox couple: standard potentials, coupling strength, and implications for lithium–oxygen batteries. Angew. Chem. Int. Ed. 55, 3129–3134 (2016).

17. 17.

, & Electron and ion transport in Li2O2. Adv. Mater. 25, 3129–3133 (2013).

18. 18.

Solvent effects on reactivity of organometallic compounds. Coordin. Chem. Rev. 18, 225–255 (1976).

19. 19.

Glyme–lithium salt phase behavior. J. Phys. Chem. B 110, 13177–13183 (2006).

20. 20.

, , , & Enhancing electrochemical intermediate solvation through electrolyte anion selection to increase nonaqueous Li–O2 battery capacity. Proc. Natl Acad. Sci. USA 112, 9293–9298 (2015).This paper, along with refs 21 and 22, demonstrated that the salt anion can promote a solution-based mechanism during discharge.

21. 21.

et al. Mechanistic role of Li+ dissociation level in aprotic Li–O2 battery. ACS Appl. Mater. Interfaces 8, 5300–5307 (2016).

22. 22.

, , , & A study of the influence of lithium salt anions on oxygen reduction reactions in Li–air batteries. J. Electrochem. Soc. 162, A1055–A1066 (2015).

23. 23.

, , , & The influence of water and protons on Li2O2 crystal growth in aprotic Li–O2 cells. J. Electrochem. Soc. 162, A573–A584 (2015).

24. 24.

, , , & The effect of water on the discharge capacity of a non-catalyzed carbon cathode for Li–O2 batteries. Electrochem. Solid State Lett. 15 A45–A48 (2012).This paper showed that water as an impurity has a significant effect on discharge capacity.

25. 25.

et al. Tunneling and polaron charge transport through Li2O2 in Li–O2 batteries. J. Phys. Chem. Lett. 4, 3494–3499 (2013).

26. 26.

Thermodynamic theory of multi-electron transfer reactions: implications for electrocatalysis. J. Electroanal. Chem. 660, 254–260 (2011).

27. 27.

, , & Promoting solution phase discharge in Li–O2 batteries containing weakly solvating electrolyte solutions. Nat. Mater. 15, 882–888 (2016).This paper demonstrated a homogeneous catalyst for O2 reduction that results in a dominant solution mechanism during discharge and large capacities.

28. 28.

, & A redox shuttle to facilitate oxygen reduction in the lithium air battery. Electrochem. Commun. 26, 74–76 (2013).

29. 29.

, , & A new method to prevent degradation of lithium–oxygen batteries: reduction of superoxide by viologen. Chem. Commun. 51, 1705–1708 (2015).

30. 30.

et al. A solution-phase bifunctional catalyst for lithium–oxygen batteries. J. Am. Chem. Soc. 136, 8941–8946 (2014).

31. 31.

Infrared spectrum, structure, vibrational potential function, and bonding in the lithium superoxide molecule LiO2. J. Chem. Phys. 50, 4288–4299 (1969).

32. 32.

, & Stability of lithium superoxide LiO2 in the gas phase: computational study of dimerization and disproportionation reactions. J. Phys. Chem. A 114, 8165–8169 (2010).

33. 33.

et al. A lithium–oxygen battery based on lithium superoxide. Nature 529, 377–382 (2016).

34. 34.

et al. Aqueous and nonaqueous lithium–air batteries enabled by water-stable lithium metal electrodes. J. Solid State Electrochem. 18, 1443–1456 (2014).

35. 35.

et al. Metal–air batteries with high energy density: Li–air versus Zn–air. Adv. Energy Mater. 1, 34–50 (2011).

36. 36.

et al. Cycling Li–O2 batteries via LiOH formation and decomposition. Science 350, 530–533 (2015).

37. 37.

, , & Lithium peroxide surfaces are metallic, while lithium oxide surfaces are not. J. Am. Chem. Soc. 134, 1093–1103 (2012).

38. 38.

, & J. Chem. Phys. 138, 034703–034713 (2013).This paper provided theoretical evidence for low overpotentials for O2 evolution from Li2O2 surfaces.

39. 39.

et al. Twin problems of interfacial carbonate formation in nonaqueous Li–O2 batteries. J. Phys. Chem. Lett. 3, 997–1001 (2012).This paper identified the major side reactions between the discharge products and cell components.

40. 40.

, , , & Role of Li2O2@Li2CO3 interfaces on charge transport in nonaqueous Li–air batteries. J. Phys. Chem. C 119, 18066–18073 (2015).

41. 41.

, , & A direct, soft chemical route to mesoporous metallic lead ruthenium pyrochlore and exploration of its electrochemical properties. Chem. Mater. 27, 2322–2331 (2015).

42. 42.

et al. The importance of nanometric passivating films on cathodes for Li–air batteries. ACS Nano 8, 12483–12493 (2014).

43. 43.

et al. Nature of Li2O2 oxidation in a Li–O2 battery revealed by operando X-ray diffraction. J. Am. Chem. Soc. 136, 16335–16344 (2014).

44. 44.

, , & A facile mechanism for recharging Li2O2 in Li–O2 batteries. Chem. Mater. 25, 3328–3336 (2013).

45. 45.

, & Mechanisms of morphological evolution of Li2O2 particles during electrochemical growth. J. Phys. Chem. Lett. 4, 1060–1064 (2013).

46. 46.

et al. Combining accurate O2 and Li2O2 assays to separate discharge and charge stability limitations in nonaqueous Li–O2 batteries. J. Phys. Chem. Lett. 4, 2989–2993 (2013).

47. 47.

et al. Evidence for lithium superoxide-like species in the discharge product of a Li–O2 battery. Phys. Chem. Chem. Phys. 15, 3764–3771 (2013).

48. 48.

& Nonaqueous Li–Air batteries: a status report. Chem. Rev. 114, 11721–11750 (2014).

49. 49.

et al. Screening for superoxide reactivity in Li–O2 batteries: effect on Li2O2/LiOH crystallization. J. Am. Chem. Soc. 134, 2902–2905 (2012).This paper presented a study on the chemical reactivity of superoxide and Li2O2 with various cell components.

50. 50.

et al. Soluble oxygen evolving catalysts for rechargeable metal-air batteries. US patent app. 13/093,759 (2011).This patent and ref. 51 are the first studies to employ redox mediators in an attempt to suppress charge overpotentials.

51. 51.

, , , & Charging a Li–O2 battery using a redox mediator. Nat. Chem. 5, 489–494 (2013).

52. 52.

, , & A highly active low voltage redox mediator for enhanced rechargeability of lithium–oxygen batteries. ACS Cent. Sci. 1, 510–515 (2015).

53. 53.

, , , & TEMPO: a mobile catalyst for rechargeable Li–O2 batteries. J. Am. Chem. Soc. 136, 15054–15064 (2014).

54. 54.

et al. Understanding the fundamentals of redox mediators in Li–O2 batteries: a case study on nitroxides. Phys. Chem. Chem. Phys. 17, 31769–31779 (2015).

55. 55.

et al. Understanding the behavior of Li–oxygen cells containing LiI. J. Mater. Chem. A 3, 8855–8864 (2015).

56. 56.

, , & A self-defense redox mediator for efficient lithium–O2 batteries. Energy Environ. Sci. 9, 1024–1030 (2016).

57. 57.

et al. Superior rechargeability and efficiency of lithium–oxygen batteries: hierarchical air electrode architecture combined with a soluble catalyst. Angew. Chem. Int. Ed. 53, 3926–3931 (2014).

58. 58.

, & Enabling catalytic oxidation of Li2O2 at the liquid–solid interface: the evolution of an aprotic Li–O2 battery. ChemSusChem 8, 600–602 (2015).

59. 59.

et al. An organic catalyst for Li–O2 batteries: dilithium quinone-1,4-dicarboxylate. ChemSusChem 8, 2198–2203 (2015).

60. 60.

& Chemical reactions of superoxide anion radical in aprotic solvents. Photochem. Photobiol. 28, 711–717 (1978).

61. 61.

& How super is superoxide? Acc. Chem. Res. 14, 393–400 (1981).

62. 62.

, , & The electrochemistry of noble metal electrodes in aprotic organic solvents containing lithium salts. J. Electroanal. Chem. Interfacial Electrochem. 297, 225–244 (1991).This paper, along with refs 63–68, elucidated instabilities of organic Li+ electrolytes during oxygen reduction.

63. 63.

et al. Reactions in the rechargeable lithium–O2 battery with alkyl carbonate electrolytes. J. Am. Chem. Soc. 133, 8040–8047 (2011).

64. 64.

, , , & Rechargeable Li–air batteries with carbonate-based liquid electrolytes. Electrochemistry 78, 403–405 (2010).

65. 65.

et al. The lithium–oxygen battery with ether-based electrolytes. Angew. Chem. Int. Ed. 50, 8609–8613 (2011).

66. 66.

, , , & Solvents' critical role in nonaqueous lithium–oxygen battery electrochemistry. J. Phys. Chem. Lett. 2, 1161–1166 (2011).

67. 67.

, , , & Interactions of dimethoxy ethane with Li2O2 clusters and likely decomposition mechanisms for Li–O2 batteries. J. Phys. Chem. C 117, 8041–8049 (2013).

68. 68.

et al. On the challenge of electrolyte solutions for Li–Air batteries: monitoring oxygen reduction and related reactions in polyether solutions by spectroscopy and EQCM. J. Phys. Chem. Lett. 4, 127–131 (2013).

69. 69.

, , , & Li–O2 battery degradation by lithium peroxide (Li2O2): a model study. Chem. Mater. 25, 77–84 (2012).

70. 70.

et al. The mechanisms of oxygen reduction and evolution reactions in nonaqueous lithium–oxygen batteries. ChemSusChem, 7, 2436–2440 (2014).

71. 71.

, & Solvent degradation in nonaqueous Li–O2 batteries: oxidative stability versus H-abstraction. J. Phys. Chem. Lett. 5, 2419–2424 (2014).

72. 72.

et al. Towards a stable organic electrolyte for the lithium oxygen battery. Adv. Energy Mater. 5, 1400867 (2015).A method of stabilizing ethers was presented by replacement of the reactive protons with methyl groups.

73. 73.

, , & A reversible and higher-rate Li–O2 battery. Science 337, 563–566 (2012).

74. 74.

et al. A stable cathode for the aprotic Li–O2 battery. Nat. Mater. 12, 1050–1056 (2013).

75. 75.

& Reaction of tert-butyl hydroperoxide anion with dimethyl sulfoxide. On the pathway of the superoxide-alkyl halide reaction. J. Org. Chem. 41, 2500–2502 (1976).

76. 76.

et al. Chemical instability of dimethyl sulfoxide in lithium–air batteries. J. Phys. Chem. Lett. 5, 2850–2856 (2014).

77. 77.

, & A new look at the stability of dimethyl sulfoxide and acetonitrile in Li–O2 batteries. ECS Electrochem. Lett. 3, A15–A18 (2014).

78. 78.

et al. Oxidation of dimethyl sulfoxide solutions by electrochemical reduction of oxygen. J. Phys. Chem. Lett. 4, 3115–3119 (2013).

79. 79.

, , & Lithium–air batteries with hybrid electrolytes. J. Phys. Chem. Lett. 7, 1267–1280 (2016).

80. 80.

, & A lithium–air capacitor–battery based on a hybrid electrolyte. Energy Environ. Sci. 4, 4994–4999 (2011).This paper showed a lithium–air capacitor–battery system based on a hybrid electrolyte.

81. 81.

& Hybrid and aqueous lithium–air batteries. Adv. Energy Mater. 5, 1401302 (2015).

82. 82.

et al. Water-stable lithium anode with Li1.4Al0.4Ge1.6(PO4)3–TiO2 sheet prepared by tape casting method for lithium–air batteries. J. Power Sources 235, 117–121 (2013).

83. 83.

et al. A molten salt lithium–oxygen battery. J. Am. Chem. Soc. 138, 2656–2663 (2016).This paper demonstrated a stable molten-salt electrolyte for Li–O2 batteries.

84. 84.

, , & Mechanistic insights for the development of Li–O2 battery materials: addressing Li2O2 conductivity limitations and electrolyte and cathode instabilities. Chem. Commun. 51, 12701–12715 (2015).

85. 85.

et al. On the efficacy of electrocatalysis in nonaqueous Li–O2 batteries. J. Am. Chem. Soc. 133, 18038–18041 (2011).

86. 86.

et al. Li–O2 kinetic overpotentials: tafel plots from experiment and first-principles theory. J. Phys. Chem. Lett. 4, 556–560 (2013).

87. 87.

, , & The carbon electrode in non-aqueous Li–O2 cells. J. Am. Chem. Soc. 135, 494–500 (2013).

88. 88.

et al. Reactivity of carbon in lithium–oxygen battery positive electrodes. Nano Lett. 13, 4697–4701 (2013).

89. 89.

, , & Adsorption and deposition of Li2O2 on TiC{111} surface. J. Phys. Chem. Lett. 5, 3919–3923 (2014).

90. 90.

et al. A MO2C/carbon nanotube composite cathode for lithium–oxygen batteries with high energy efficiency and long cycle life. ACS Nano 9, 4129–4137 (2015).

91. 91.

et al. Nanostructured metal carbides for aprotic Li–O2 batteries: new insights into interfacial reactions and cathode stability. J. Phys. Chem. Lett. 6, 2252–2258 (2015).

92. 92.

, , & A highly active nanostructured metallic oxide cathode for aprotic Li–O2 batteries. Energy Environ. Sci. 8, 1292–1298 (2015).

93. 93.

, , , & The role of catalysts and peroxide oxidation in lithium–oxygen batteries. Angew. Chem. Int. Ed. 52, 392–396 (2013).

94. 94.

, & A free-standing-type design for cathodes of rechargeable Li–O2 batteries. Energy Environ. Sci. 4, 4727–4734 (2011).

95. 95.

et al. A nanostructured cathode architecture for low charge overpotential in lithium–oxygen batteries. Nat. Commun. 4, 2383 (2013).

96. 96.

et al. Stability of polymer binders in Li–O2 batteries. J. Power Sources 243, 899–907 (2013).

97. 97.

et al. Quantifying the promise of lithium–air batteries for electric vehicles. Energy Environ. Sci. 7, 1555–1563 (2014).

98. 98.

et al. Electrical conductivity in Li2O2 and its role in determining capacity limitations in non-aqueous Li–O2 batteries. J. Chem. Phys. 135, 214704 (2011).This paper reported that Li2O2-induced cathode passivation limits Li–O2 battery capacity.

99. 99.

, , , & Effect of carbon surface area on first discharge capacity of Li–O2 cathodes and cycle-life behavior in ether-based electrolytes. J. Electrochem. Soc. 159, A2135–A2142 (2012).

100. 100.

et al. A rechargeable room-temperature sodium superoxide (NaO2) battery. Nat. Mater. 12, 228–232 (2013).One of the first papers to demonstate a Na–O2 battery that is able to cycle at a high current density.

101. 101.

, , , & The critical role of phase-transfer catalysis in aprotic sodium oxygen batteries. Nat. Chem. 7, 496–501 (2015).This paper showed that protons catalyse a solution mechanism in Na–O2 batteries.

## Acknowledgements

P.G.B. is indebted to the Engineering and Physical Sciences Research Council (EPSRC), including the SUPREGEN programme, for financial support. L.F.N. gratefully acknowledges Natural Resources Canada, and also Natural Sciences and Engineering Research Council of Canada (NSERC) for funding through its Discovery and Research Chair programs. D.A. thanks A. Frimer and D. Sharon, BIU for helpful discussions and the Israel Science Foundation (ISF) for support in the framework on the INREP project. B.D.M. gratefully acknowledges financial support from the FY 2014 Vehicle Technologies Program Wide Funding Opportunity Announcement, under Award Number DE-FOA-0000991 (0991-1872), by the US Department of Energy (DOE) and National Energy Technology Laboratory (NETL) on behalf of the Office of Energy Efficiency and Renewable Energy (EERE).

## Affiliations

1. ### Department of Chemistry, Bar Ilan University, Ramat-Gan 52900, Israel.

• Doron Aurbach
2. ### Department of Chemical and Biomolecular Engineering, University of California, Berkeley, California 94720, USA.

• Bryan D. McCloskey
3. ### Energy Storage and Distributed Resources Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA.

• Bryan D. McCloskey
4. ### Department of Chemistry, The Waterloo Institute for Nanotechnology, University of Waterloo, Waterloo, Ontario N2L 3G1, Canada.

• Linda F. Nazar
5. ### Departments of Materials and Chemistry, Parks Road, University of Oxford, Oxford OX1 3PH, UK.

• Peter G. Bruce

## Authors

### Competing interests

The authors declare no competing financial interests.

## Corresponding author

Correspondence to Peter G. Bruce.