Exclusive Solution Discharge in Li-O2 Batteries?

Capacity, rate performance, and cycle life of aprotic Li-O2 batteries critically depend on reversible electrodeposition of Li2O2. Current understanding states surface-adsorbed versus solvated LiO2 controls Li2O2 growth as surface film or as large particles. Herein, we show that Li2O2 forms across a wide range of electrolytes, carbons, and current densities as particles via solution-mediated LiO2 disproportionation, bringing into question the prevalence of any surface growth under practical conditions. We describe a unified O2 reduction mechanism, which can explain all found capacity relations and Li2O2 morphologies with exclusive solution discharge. Determining particle morphology and achievable capacities are species mobilities, true areal rate, and the degree of LiO2 association in solution. Capacity is conclusively limited by mass transport through the tortuous Li2O2 rather than electron transport through a passivating Li2O2 film. Provided that species mobilities and surface growth are high, high capacities are also achieved with weakly solvating electrolytes, which were previously considered prototypical for low capacity via surface growth.

Reducing the cost and ecological footprint of energy storage is mandatory and requires alternatives to Li-ion batteries with abundant, low-cost materials. Metal–air and metal–sulfur batteries show great potential because of the high theoretical capacities and the cheap and abundant materials.[1,2] In both systems, insulating solids, such as Li2O2 and Li2S, are reversibly deposited and stripped at the cathode upon cycling. Determining the high practical capacities and lifetime are large fractions of deposited material while avoiding parasitic reactions.[2−6] Capacity, deposit structure, and battery lifetime are intrinsically linked to the underlying physicochemical mechanisms.[5,7−10]

Current literature[2,8,11,12] states that Li–O2 batteries discharge in between two limiting cases after O2 reduction to superoxide: (i) solution discharge, where Li2O2 forms by solution-mediated LiO2 disproportionation, or (ii) surface discharge, where a thin film of Li2O2 forms via direct consecutive 2 e– electroreduction. Determining the predominance of a mechanism would be the current density and the electrolyte’s ability to dissociate and solvate the surface adsorbed superoxide. Solution discharge dominates in highly solvating electrolytes, enabling large (toroidal) Li2O2 particles of hundreds of nanometers and high capacities.[8,12−14] Surface discharge is considered to dominate in weakly solvating electrolytes and at high overpotentials, leading to a passivating film and low capacities.[15−17] Surface film growth is, in principle, self-limited by the tunneling thickness, often considered to be ∼7 nm.[16,18,19] To what extent the surface or solution mechanism prevails is still unclear; capacity would be limited by either electron transport through a Li2O2 film or mass transport (O2, LiO2, O2–, and Li+) through a porous particulate Li2O2 deposit.[16,17,20−23] In a recent study with operando small- and wide-angle X-ray scattering (SAXS/WAXS), we found that Li2O2 structures indicating surface growth are absent even in weakly LiO2-solvating electrolytes and at high overpotentials.[10] This is in line with large Li2O2 particles imaged via electron microscopy in weakly solvating electrolytes at practical current densities and raises questions about the surface mechanism occurring.[24−27] Consequently, truly capacity-limiting factors as well as measures and governing factors for Li2O2 packing density are still obscure.

Here we show that Li2O2 forms via solution-mediated LiO2 disproportionation across a wide range of relevant conditions: weakly to highly solvating electrolytes and a wide range of current densities and voltages. The obtained capacities contradict the currently accepted surface-versus-solution growth model. For instance, weakly solvating low-donor-number (DN) electrolytes, previously considered prototypical for exclusive surface growth, yield large particles and the highest capacities at low current densities. Rotating ring-disc electrode (RRDE) measurements and electron microscopy give evidence for soluble and mobile LiO2 even in low DN electrolytes. Supported by a numerical reaction model, we derive a Li2O2 growth mechanism that explains particle morphology and ordering across electrolytes. Capacity is limited by species (O2, LiO2, O2–, and Li+) transport through the porous particulate Li2O2 deposit rather than electron transport through a thin passivating Li2O2 film. The current Li–O2 discharge mechanism needs to be refined.

Unexpected Performance Relations

Electrolyte solvation and applied current densities are known to significantly alter Li2O2 morphologies and achievable discharge capacities. To investigate the critical role of solvation and current densities in conjunction, we conducted galvanostatic discharge measurements while varying the electrolyte and carbon cathode. We used 1 M lithium bis(trifluoromethane)sulfonimide (LiTFSI) in (i) acetonitrile (MeCN), (ii) dimethylacetamide (DMAc), and (iii) tetraethylene glycol dimethyl ether containing 4000 ppm H2O (TEGDME/H2O) as electrolyte. While MeCN is weakly solvating and considered as a prototype solvent to form Li2O2 as a conformal surface coating via direct electroreduction, TEGDME/H2O is strongly solvating and considered to form Li2O2 as large toroidal particles via solution-mediated LiO2 disproportionation.[8] The DMAc electrolyte shows intermediate solvation. We rigorously excluded H2O contamination since already small concentrations of H2O could alter product growth and discharge capacities in weakly solvating electrolytes[8] (see Methods in the Supporting Information). To vary the current density normalized by true surface area (and overpotential), we used porous electrodes made from carbons with widely varying BET areas: glassy carbon beads (GC, 1.3 m2 g–1), Super P carbon black (SP, 55 m2 g–1), and KetjenBlack carbon black (KB, 1398 m2 g–1). An overview of current densities used in this work and literature is given in Figure S1.

Figure presents full discharge capacities at 50 μA cmgeom–2 with combinations of these three electrolytes and electrodes (Figure S2 shows cell voltage vs capacity; Table S1 summarizes normalized discharge capacities, current density, and Li2O2 degree of pore filling). Data are expressed in terms of specific capacity (mAh gC–1) as a function of LiO2 solvation and true areal rate (current normalized by BET area, μA cmreal–2). The latter amount to 0.0027, 0.046, and 1.34 μA cmreal–2 for KB, SP, and GC electrodes, respectively. Specific capacities generally increase with increasing BET area (Figure a) and decreasing areal rate (Figure b). At low and intermediate rates (with KB and SP), capacities do not follow the order of highest capacity with the highest degree of LiO2 solvation; the weakly solvating MeCN electrolyte gives the highest capacities, and the highly solvating TEGDME/H2O gives the lowest. Transition from surface to solution routes fails to explain this, suggesting that LiO2 solvation is not the sole factor determining capacity order at any given rate. Only the low surface area GC electrodes show the lowest capacity with MeCN and could possibly be in accord with surface growth in MeCN and successive change to solution growth in the other electrolytes.[12] SEM images show that the Li2O2 formed at the high surface KB electrode in MeCN electrolyte to be individual, large particles of hundreds of nanometers (Figure S3). Overall, the current understanding of discharge via solution or surface routes cannot consistently explain these Li2O2 morphologies and performance relations. Solution Li2O2 growth in weakly solvating electrolytes must be considered.

Figure 1

Unexpected performance relations. (a) Specific capacity versus degree of LiO2 solvation (governed by the electrolyte) for galvanostatic discharge at 50 μA cmgeom–2. Three different carbon cathodes, KetjenBlack (KB), SuperP (SP), and glassy carbon (GC), were measured in three different electrolytes, 1 M LiTFSI in MeCN, DMAc, and TEGDME + 4000 ppm H2O. (b) Specific capacity versus real areal current density. Note that the order of capacity values changes systematically when going from high to low surface area carbon (KB and GC). (c) Capacity with DMAc (light gray) and TEGDME/H2O (blue) relative to the MeCN electrolyte (gray). Capacities have a standard deviation of ∼10% (see Figure S4).

Solution Discharge in Weakly Solvating Electrolytes

Associated LiO2 clearly dominates in weakly solvating electrolytes, such as MeCN. Hence, solution discharge in weakly solvating electrolytes contradicts the previous understanding that associated LiO2 would be insoluble. To probe for soluble LiO2 in MeCN, we conducted RRDE measurements at true areal current densities close to those relevant for porous electrodes (discussed in Figure S1). The electrode was immersed in O2-flushed 0.1 M LiTFSI/MeCN electrolyte and rotated at rates ranging from 600 to 6000 min–1, and the ring was held at a potential where superoxide is oxidized at a transport-limited rate (Figure a). A constant reducing current was then applied to the GC disc in a range between 0.025 and 10 μA cmreal–2. The ring current was then corrected for collection efficiency (jR = −iR/N0) to arrive at the ring-to-disc current fraction (jR/jD), which indicates the fraction of the formed superoxide that has reached the ring electrode. The measurements go beyond previous RRDE data[10] in that a different setup was used that allowed for higher rotation rates, improved RRDE geometry, and lower currents. Experimental details are given in Supplementary Note 1 and Figure S5.

Figure 2

Evidence for soluble superoxide in weakly solvating MeCN electrolyte. (a–c) RRDE data with 0.1 M LiTFSI/MeCN and galvanostatic disc current. The ring was held at ∼3.6 V vs Li/Li+; the disc current jD was varied between 0.025 and 10.2 μA cmreal–2; the rotation rate was between 600 and 6000 min–1 (corresponding to ω–1/2 = 0.126 and 0.039 s1/2, respectively). The ring current, jR, is corrected for collection efficiency (jR = −iR/N0). (a) Sketch of the RRDE and the processes. (b) The collected fraction jR/jD as a function of rotation rate for three different disc currents jD. The solid lines are exponential fits to guide the eye. (c) The collected fraction jR/jD as a function of disc current jD at 3000 and 6000 min–1. The solid lines are power law fits to guide the eye. (d–g) SEM images of a discharged RRDE in 0.1 M LiTFSI/MeCN with jD = 0.14 μA cmreal–2 for 18 h (discharge capacity of 2.56 μAh cmD–2) at 800 min–1. Li2O2 particles are deposited on the GC disc and on the insulating PTFE with decaying density with growing distance from the disc edge. (h) The EDX line profile that shows that the particles on the PTFE substrate are most likely Li2O2.

Results in Figure b,c show significant ring fractions and prove that LiO2 is soluble in MeCN. The ring fraction increases significantly with increasing rotation rate and decreasing current and points toward a value of 1 at high rotation rates and practical current densities. The ring fraction pointing toward 1 as the transit time between disc and ring tends to zero (angular frequency ) is in accord with the solution species undergoing a chemical (C-step) but not an electrochemical reaction (E-step) during its passage from the disc to the ring. Hence, it is in accord with an EC mechanism.[28] Ring fractions <1 cannot be explained by the partition between surface and solution mechanism as any share of the surface mechanism would be largely independent of the rotation rate. Growing ring fractions with decreasing disc current density (Figure c) refine the picture: while a purely homogeneous C-step would result in current-independent ring fractions, its dependence suggests a nucleation step, which is driven by high local LiO2 concentrations (high currents). Scanning electron microscopy of the discharged RRDE in Figure d–h shows that neither nucleation nor growth requires direct electroreduction (i.e., the surface mechanism) as an explanation. Particles with similar morphology as on the disc were also found on the insulating PTFE spacer. Energy dispersive X-ray measurements (EDX Figure h) identify them as Li2O2. RRDE and SEM data in Figure give evidence for soluble LiO2 and solution discharge in weakly solvating electrolytes.

High ring fractions in MeCN require small transit times between disc and ring (, Figure b). This suggests that the disproportionation kinetics is faster than in strongly solvating electrolytes, where soluble superoxide has already previously been identified by RRDE.[12,29] We probed the disproportionation kinetics of KO2 in the three electrolytes by measuring the pressure evolution after mixing the electrolytes with KO2 in a custom-built pressure cell (see Methods in the Supporting Information). KO2 in contact with Li+ electrolyte disproportionates and liberates O2. The results in Figure a show that superoxide disproportionates fastest in MeCN electrolyte and slowest in the strongly solvating TEGDME/H2O electrolyte. This is in line with findings for NaO2 DISP in Na–O2 batteries[30] and kinetic measurements in DMSO, MeCN, or DMF by stopped-flow UV–vis spectroscopy or SECM,[29,31,32] but contrary to what the previous O2 reduction mechanism suggests: gradual shift from the surface to solution mechanism as LiO2 solvation decreases would imply slowing DISP in low DN electrolytes. The increased DISP kinetics in weakly solvating electrolytes show that the lower RRDE ring fractions stem from a larger fraction of the soluble LiO2 disproportioning to Li2O2 before it can reach the ring rather than a larger fraction of Li2O2 formed via the surface mechanism (as indicated in the sketch in Figure a).

Figure 3

DISP kinetics in MeCN, DMAc, and TEGDME/H2O. (a) Pressure evolution versus time for three different electrolytes (0.1 M LiTFSI in MeCN, DMAc, or TEGDME + 4000 ppm H2O) upon mixing them with KO2 (10 mM KO2 in the final solution). The pressure rise stems from 2 KO2 + 2 Li+ → O2 + Li2O2 + 2 K+; its time constant is proportional to the DISP rate constant. (b) Sketch of possible free energy levels in differently solvating electrolytes during the DISP reaction in weakly solvating (low DN) and highly solvating (high DN electrolytes). The activation barrier for association in high DN electrolytes is much higher, resulting in lower DISP rate constants, lower nucleation rates, and finally fewer and larger Li2O2 particles.

We conclude that the disproportionation kinetics is related to the dissociation/association equilibrium. It defines the rate at which associated LiO2(sol) feeds into the disproportionation reaction.Overall, the free energy profile of association and disproportionation may look as indicated in Figure b. Low barriers for growth are in accord with DFT calculations,[29] showing that the activation barrier for disproportionation of associated LiO2 is low, such that its kinetics can be very fast.

A Reconsidered Oxygen Reduction Mechanism

Previously, the partition between surface adsorbed LiO2* and solvated LiO2(sol) (free ions, ion pairs, and clusters) has been invoked to explain a seeming shift between surface and solution growth. LiO2 solvation is governed by effective Lewis acidity and basicity of the electrolyte as determined by the solvent’s Gutmann donor and acceptor number (DN and AN); the salt; and, for example, protic additives.[2,11,12,14,33−35] However, given the above presented evidence for soluble, mobile superoxide and the absence of surface growth even in weakly dissociating MeCN, the currently accepted ORR model ought to be reconsidered. Here, we describe Li2O2 formation from solution by O2 reduction in aprotic Li+ electrolytes as a function of LiO2 dissociation in conjunction with current density and LiO2 mobility.

In line with previous understanding, the electrolyte’s ability to solvate LiO2 is central for determining the Li2O2 morphology and capacity limitation. However, two modifications need to be introduced. First, LiO2 solvation energy comes into effect by changing the dissociation/association equilibrium in solution and thus the rate to form associated rather than the desorption/adsorption equilibrium between solution and surface . Second, current density and LiO2 mobility in the electrolyte need to be accounted for. Importantly, the new model does not contradict recent key experimental findings but revises the interpretation based on new insights. Key experimental observations are the following: (i) Capacities do not simply follow the order of highest capacity with the highest degree of LiO2 dissociation at all current densities (Figure ). (ii) LiO2 is soluble and mobile even in weakly solvating electrolytes (Figure ). (iii) Li2O2 forms to the widest extent via solution-mediated DISP (Figure and a recent operando SAXS/WAXS study[10]). (iv) Li2O2 particles become smaller and more numerous with increasing current (operando SAXS/WAXS[10] and refs (8),[14], (36), and (37)). (v) Li2O2 particles become larger and less numerous with increasing LiO2 dissociation (operando SAXS/WAXS[10] and refs (8), (12), and (14)). (vi) Weakly solvating electrolytes accelerate superoxide disproportionation rather than slowing it down (Figure , refs (29) and (32)).

Deciding for Li2O2 formation is the association of solvated LiO2 according to the equilibrium . LiO2(sol) denotes associated species such as contact ion pairs or clusters as typical for ionic species in aprotic media.[12,33,38] This equilibrium defines the rate at which associated LiO2(sol) feeds into the disproportionation reaction with the overall sequenceNote that eq may involve an additional LiO2(sol) adsorption step prior to disproportionation, as physi- or chemisorbed LiO2 on existing Li2O2 crystallites has been ascertained experimentally.[39] The actual disproportionation step (k3) of chemisorbed LiO2 might even occur in the solid state. Electrolyte and current density dependence of the process in eq and capacity limitations are illustrated in Figure and discussed in the following.

Figure 4

Li2O2 growth model and governing factors for morphology and pore filling. (a) Sketch of oxygen reduction and Li2O2 formation mechanism and morphology. (b) Li2O2 formation rate and O2– concentration versus normal distance from the carbon surface as obtained from a numerical model. The example shows the impact of electrolyte solvation and thus the association kinetics k2. Fast association (high k2, dark blue curve) causes fast Li2O2 formation close to the surface and steep O2– concentration gradients, leading to high near-surface nucleation rates and a large number of small particles. Slow association (low k2, light blue curve) results in few, larger particles up to larger distances. (c) Li2O2 formation rate profiles for different O2– diffusivities. Lower diffusivities result in high rates of Li2O2 formation close to the surface and a high density of small, near-surface Li2O2 particles. The impact of current densities and the time dependency is explored in Supplementary Note 2. (d) Degree of pore filling with Li2O2 calculated from capacities in Figure . Note that the apparent high degree of pore filling (close to one) can be explained only by significant electrode swelling, as discussed in Supplementary Note 4. Arrows indicate factors influencing the Li2O2 morphology, pore filling and discharge capacity.

Superoxide forms at a rate proportional to the current density ja and associates with Li+ with the rate constant k2 to LiO2(sol), which then disproportionates with the rate constant k3 to Li2O2 and O2 (Figure a). Since superoxide disproportionation passes via the (LiO2)2 dimer or higher aggregates,[29,38,40] its formation from 2 LiO2(sol) is strongly favored over formation from 2 Li+(sol) + 2 O2(sol)–. Disproportionation of associated LiO2(sol) is second order in LiO2(sol) concentration and very small activation barriers suggest k3 to be very large. Superoxide disproportionation all the way from O2– to Li2O2 can be regarded as a pseudo-first order reaction in O2– since the Li+ concentration is orders of magnitude higher than the O2– concentration.[29,31] Association is hence the rate-limiting step in eq and determines the overall rate to form Li2O2 via disproportionation. The association rate constant k2 depends on the solvation strength of the electrolyte and is connected with the dissociation/association equilibrium (see eq and Figure ).

Low solvation energies (weakly dissociating electrolytes) shift the dissociation equilibrium toward associated LiO2(sol), in turn increasing the association rate constant k2 (Figure ). Figure a and eq illustrate that the profile of O2– concentration versus distance from the electrode surface determines local Li2O2 nucleation and growth and hence particle density and size. The Li2O2 formation rate profile arises from solvation, current density, and species mobility in conjunction.

To better grasp the mutual sensitivity of electrolyte solvation (LiO2 association), true areal current densities, and species mobilities, we implemented a simple 1D numerical model taking into account O2– production at the electrode interface, diffusive transport away, and disproportionation as a sink with a rate governed by the O2– concentration profile. The model intends to identify the important trends rather than accurately accounting for (heterogeneous) nucleation and growth of Li2O2 particles or the real carbon electrode structure. Further details and results are given in Supplementary Note 2, Table S2, and Figures S6 and 4b,c.

The model is based on eq and a pseudo-first order DISP kinetics with respect to O2– concentration as revealed by stopped-flow UV–vis spectroscopy.[29,31] Considering eq and the fact that LiO2 association is rate-limiting (k2), DISP at a planar electrode can be modeled by the following process:Herein, is the second-order DISP rate constant with respect to and , which translates into the pseudo-first order DISP rate constant with respect to . The model calculates the concentration profile and as a function of distance x from a planar electrode surface and time t by solving the following partial differential equations numerically via a finite difference method[41] assuming constant currentEquations and 5 account for the diffusion of O2– and Li2O2 via Fick’s law. O2– consumption and Li2O2 generation are considered as a sink term (eq ) and source term (eq ). The sink term is expressed by the second-order reaction ν = , or the equivalent pseudo-first order reaction ν = .

The resulting Li2O2 concentration profile (Figure S6d–f) gives an estimate for the thickness of the particulate Li2O2 layer on the electrode surface and the local rate at which Li2O2 forms. A high means a large quantity of Li2O2 formed at a high rate. We calculate the local Li2O2 formation rate by dividing by the time step . A high local Li2O2 formation rate causes high nucleation rates. Li2O2 particles would be smaller and more numerous.

Considering first the effect of LiO2 solvation (Figure b), weakly dissociating electrolytes (i.e., large k2) cause high O2– concentration and Li2O2 formation rates close to the electrode surface, and both sharply decay as distance grows. The used reaction rates (k2) are in the range of experimental values (disproportionation rate constants of MeCN electrolyte,[31] 25 s–1; DMSO electrolyte,[29] 560 s–1). High near-surface Li2O2 formation enhances near-surface nucleation, causing a larger number of small particles closer to the surface. Highly dissociating electrolytes with slow association kinetics k2 cause low Li2O2 formation rates at all distances from the electrode surface and flat O2– concentration gradients. This leads to low nucleation rates and few large Li2O2 particles that reach out several 100 nm into solution, in line with literature and our recent operando SAXS study.[10] To confirm this with mainly varying association while leaving transport largely constant, we tested dimethoxyethane (DME) electrolytes and different LiTFSI/LiNO3 concentrations ratios (Supplementary Note 3 and Figures S7 and S8). Adding NO3– significantly increases LiO2 dissociation[42,17] but does not primarily affect Li+, O2, and LiO2(sol) diffusion coefficients.

As visualized by the sketches in Figure b, the increasingly tortuous transport path self-accelerates tortuosity increase with growing depth of discharge, finally causing the end of discharge by mass transport limitation (O2, Li+) toward the electrode surface combined with some degree of surface blocking by Li2O2 particles touching the carbon.[21] Dominance of surface growth even in MeCN implies these factors are limiting in all electrolytes.

Considering further superoxide mobility and current density, both in conjunction determine the near-surface O2– concentration and Li2O2 formation profile, in turn Li2O2 particle density/size, and how far the layer of Li2O2 particles can reach out from the surface (Figure c). This layer thickness determines the achievable degree of pore filling and hence capacity. Growing currents in a certain electrolyte cause growing O2– concentrations and steeper gradients, as illustrated in Supplementary Note 2 and Figure S6. Higher Li2O2 formation rates close to the carbon surface enhance near-surface nucleation, causing a larger number of small particles closer to the surface compared to low current.

With these relations between LiO2 dissociation, species mobility, and true surface current density in mind, a map of achievable capacity can be drawn as illustrated in Figure d. It is in accord with the capacities in Figure from where the degree of pore filling is taken for the contour plot. Importantly, Li2O2 particle size alone determines discharge capacities only at planar/low surface area electrodes. In moderate to high-surface area cathodes (where the pore size ≈ Li2O2 particle size), pore filling does. Next to the (i) LiO2 association rate, the other main parameters are (ii) current (raising local superoxide concentration and hence nucleation) and (iii) superoxide and other species mobilities[43,44] (determining how far LiO2(sol) can diffuse before it disproportionates and how tortuous the Li2O2 deposit may be before causing mass transport limitations). High disproportionation rates in weakly dissociating electrolytes are not detrimental if (i) areal current densities are low and (ii) species mobilities are high. The highest capacity being achieved with MeCN electrolyte and KB electrode even at high geometric rates confirms this (additional data and discussion in Supplementary Note 5 and Figure S9). To give absolute numbers of required current densities or species mobilities, future model calculations need to consider the actual porous electrode structure and the increasing tortuousity caused by Li2O2 formation.

By combining galvanostatic discharge with RRDE measurements, electron microscopy, O2 pressure evolution measurements, and a 1D numerical model, we show that discharge of aprotic Li–O2 batteries proceeds to the widest extent via solution-mediated LiO2 disproportionation to form Li2O2 particles. Li2O2 forming a passivating film via direct electroreduction of surface adsorbed LiO2 can be largely excluded under practically relevant conditions. This is true even for low DN electrolytes previously considered prototypical for the surface mechanism. Species transport through the increasingly tortuous particulate Li2O2 deposit hence limits capacity rather than electron transport across Li2O2 films. We describe a unified O2 reaction mechanism that can explain Li2O2 particle size and number density, packing density, achievable rate, and capacity across a wide range of electrolytes and operating conditions. Deciding factors are the dissociation of solvated LiO2, species mobilities (Li+, O2, O2–, and LiO2), and areal current densities.

This mechanism suggests strategies on how research toward highly reversible, high-performance Li–O2 cells should proceed. First, low-donor-number (weakly LiO2 dissociating) electrolytes, previously thought to be prototypical for low capacity, can achieve the highest pore filling and capacity. High species mobility and high electrode surface area are, however, a requirement. Mediators, for example, make superoxide more mobile[45,46] and allow oxidizing large particles and suppressing side reactions,[47] but their impact on, for example, packing density and ordering remains to be studied. They also shift the O2 reduction from the surface to the electrolyte volume,[48] reduce high near-surface nucleation, and may hence allow for lower-surface electrodes to achieve large capacities. Second, the previous paradigm can be lifted that highly solvating electrolytes are required for high capacity despite them being more susceptible to decomposition. Disproportionation has, however, been shown to be the source of the highly reactive singlet oxygen, which in turn is the major source of parasitic reactions and requires careful consideration when judging electrolytes.[5] We further note that the here derived mechanism holds for relatively defect-free carbon surfaces as found with pristine GC, SP, and KB, where LiO2 adsorbs weakly.[49] Highly defective carbonaceous electrodes[49] or catalyst surfaces[50] could change LiO2 adsorption and rates and hence favor the surface mechanism to some extent.

The current picture of Li2O2 formation, proceeding in-between the two cases of surface and solution mechanism, ought to be reconsidered. Why the second consecutive electron transfer at the carbon surface mechanism is so unlikely compared to LiO2 disproportionation remains to be clarified.