scopic characterization indicates a more nuanced structural speciation. Infrared, Raman, electron paramagnetic resonance, diffusereflectance UV-vis and X-ray absorption spectroscopy reveal that the lithium salt of the benzoquinone radical anion forms on the surface of the lithium peroxide, indicating the occurrence of electron and lithium ion transfer in the solid state. As a result, obligate lithium superoxide is formed and encapsulated in a shell of Li[p-C 6 H 4 O 2 ] with a core of Li 2 O 2 . Lithium superoxide has been proposed as a critical intermediate in the charge/discharge cycle of Li-air batteries but has yet to be isolated owing to instability. The results reported herein provide a snapshot of lithium peroxide/superoxide chemistry in the solid state with redox mediation.

Lithium superoxide | Lithium peroxide oxidation | Li-air battery T he advent of metal-air batteries has provided impetus for understanding the structure, spectroscopic properties and chemical reactivity of various metal oxides. Lithium-air batteries, which possess a theoretical energy density approaching that of liquid fuels, have emerged as potential candidates to replace lithium-ion batteries (1-4). Lithium-air batteries operate by electron transfer from a high surface area cathode to oxygen gas during discharge, generating lithium peroxide deposits. Upon charging, the lithium peroxide is oxidized back to oxygen gas. Despite demonstrating promise as a replacement for lithium-ion batteries, this electrochemical energy storage system suffers from numerous challenges that must be overcome (5)(6)(7)(8), the most important of which is reversible charging.

Lithium superoxide, LiO2, is an important intermediate in both the reduction of oxygen to lithium peroxide and oxidation of lithium peroxide back to oxygen (9)(10)(11)(12)(13). Lithium superoxide, via disproportionation, is thought to be responsible for the growth of large lithium peroxide toroids commonly observed during discharge of nonaqueous lithium-air cells, but is also implicated in numerous studies (14)(15)(16)(17)(18) as being responsible, either directly or through the intermediacy of 1 O2 (19,20), for the degradation of the organic solvent and electrolyte in the battery. Furthermore, 'superoxide like' sites on the surface of lithium peroxide are thought to be responsible for both enhanced reactivity with electrolytes in lithium-air batteries and enhanced conductivity (21,22).

While the superoxide salts of cesium, rubidium, potassium

Lithium superoxide (LiO2) is an important intermediate in lithium-air batteries, a promising next-generation energy storage platform. The conductivity, stability and reactivity profiles of LiO2 are thought to play a crucial role in the cyclability of lithium-air batteries. We demonstrate that physical encapsulation of Li2O2 with an appropriate redox-active molecule may be a viable strategy to access and stabilize LiO2 at room temperature while simultaneously protecting the solvent and electrolyte from deleterious reactivity derived from LiO2. Encapsulation with a redox mediator does not impede interfacial electron and lithium-ion transport and provides researchers with a model system that recapitulates the charging of a lithium-air cell.

Synthesis. Lithium peroxide was oxidized by p-benzoquinone (p-C6H4O2) using a setup similar to that of growing crystals with two organic solvents (Fig. 1). Vapor diffusion of p-C6H4O2 (vapor pressure = 0.1 mm Hg at 25 • C) (36) onto solid lithium peroxide resulted in a gradual color change initially to very faint blue, followed by considerable darkening over the course of several weeks to furnish a sample that ultimately appeared black. This material, designated as compound 1, was thermodynamically unstable and detonated upon scratching with a metal spatula presumably releasing oxygen (caution!); for this reason use of a plastic spatula was preferred.

To accelerate the diffusion of p-C6H4O2 onto Li2O2, these reactants were placed together as solids into a sealed ampule and heated at 70 • C under a slight vacuum overnight to produce the identical black color as observed for the material prepared via the vapor diffusion method. Comparison of the spectroscopic data for 1 (vide infra) produced by these two different methods established the materials to be identical. Sampling by gas chromatography the headspace gases generated, if any, when using the accelerated ampule synthesis method showed that oxygen gas was not evolved.

Phenylboronic acid is easily oxidized to phenol by hydrogen peroxide generated in situ by hydrolysis of either peroxide or superoxide salts (37). Aqueous titration of 1 with phenylboronic acid gave nearly quantitative conversion to phenol, as verified by nuclear magnetic resonance (NMR) spectroscopy (SI Appendix, Fig. S1). This result confirms that the O-O bonds remains intact in 1 and reinforces the conclusion that a negligible amount of O2 gas is released during the formation of 1. Running the synthesis of 1 using an excess of p-C6H4O2 also does not result in fractional compositions of p-C6H4O2 exceeding 0.7, as the excess p-C6H4O2 is recovered following its sublimation from the black solid sample of 1.

Examination of 1 by scanning electron microscopy (SEM, Fig. 2) revealed distinct morphological changes as compared with the Li2O2 starting material. Commercial Li2O2, as purchased, is composed of particles several hundred nanometers in diameter. Upon exposure of Li2O2 to p-C6H4O2, fusing of particles was observed, suggesting that the black material that formed on the outer surface of the peroxide causes the particles to coalesce (Fig. 2 and SI Appendix, Fig. S2).

The blue color of quinone monoanions (38,39) prompted us to independently prepare Li[p-C6H4O2] (2), to determine whether the benzoquinone anion radical is responsible for the black color of 1. As the lithium salt of the dianion, Li2[p-C6H4O2], is known, a comproportionation strategy was pursued. Using the vapor-diffusion method, it was found that 2 could be prepared from Li2[p-C6H4O2] and p-C6H4O2 as an intense blue powder. Salt 2 could also be prepared via mechan- profiles of 1 and 2 is that the former spectrum exhibits a more pronounced shoulder at λ = 250 nm; superoxide exhibits an absorption at this wavelength (42).

EPR spectra of 1 and 2 (Fig. 5) display a single broad signal centered at g = 2.008, a feature consistent with the presence of a spin one-half organic radical. Hyperfine coupling could not be resolved at 77 K or by dilution of the samples with sodium sulfate due to the close intermolecular contact of the spin bearing species composing both samples (vide infra).

EPR spin quantification, a method used to determine the number of radicals present in a bulk sample, was performed by blue gradient developed across the Li2O2 (SI Appendix, Fig. 222 S10). Raman spectral analysis along this gradient revealed 223 the presence of a weak band at 1139 cm -1 (inset, Fig. 7) 224 superimposed on a highly fluorescent background. This band 225 was not attributable to p-benzoquinone or its radical anion 226 (SI Appendix, Fig. S11), is in a region typically associated 227 with superoxide O-O stretches (32), and may be that of LiO2 228 present in the sample.

Oxygen K-edge (1s → valence) X-ray absorption spec-230 troscopy (XAS) was used to differentiate the natures of oxy-231 gen present in 1. Oxygen K-edge XAS of 1, 2, Li2O2 and 232 Li2[p-C6H4O2] were collected and the corresponding data are 233 presented in Fig. 8; the fitted peak energies are shown in SI 234 Appendix, Fig. S12 and summarized in SI Appendix, Table S4. 235 In general, the oxygen K-edge π * features of C=O are at lower 236 energy than the σ * features of O-H (48). A detailed report of 237 the electron energy loss spectra (EELS) of p-benzoquinone, 238 hydroquinone, and phenol gives assignments of these features 239 based on molecular orbital theory (49).

The π * feature of p-C6H4O2 is at 529. 85 color change of p-C6H4O2 from yellow to blue (and ultimately, black), which is indicative of reduction of p-C6H4O2. Similarly, reactions 4 and 5 may be dismissed as analysis of the reaction headspace by GC did not reveal any oxygen production. Moreover, results from the titration of the product with PhB(OH)2 to produce PhOH quantitatively suggest that no O2 was lost from the sample; the oxygen speciation has an O-O bond of peroxide or superoxide, and peroxide can directly oxidize boronic acids while superoxide may convert to peroxide upon disproportionation in water (37,(51)(52)(53). Additionally, the cyclic voltammograms in Fig. 3 show that it is thermodynamically unfavorable for the radical anion of p-C6H4O2 to oxidize superoxide to produce O2. Reactions 2 and 3 depict electron transfer and lithium ion diffusion from lithium peroxide to benzoquinone, to generate lithium superoxide and either the benzoquinone radical monoanion or dianion, respectively. The spectroscopic properties of 2 rule out reaction 3 and point to reaction 2 as being operative.

The radical monoanion of p-C6H4O2 has been studied in great detail and can be prepared by a variety of methods, such as pulse radiolysis of the neutral quinone in matrices (54) or frozen solutions (55), or direct reduction of p-benzoquinone with potassium in the presence of Kryptofix® 222 or crown ethers in THF (56). The Li + salt of p-C6H4O2 radical monoanion is less studied, and only a handful of instances detailing the preparation of this compound are known (57)(58)(59)(60). We therefore sought to prepare Li[p-C6H4O2] (2) independently. We pursued the solid-state comproportionation reaction of Li2[p-C6H4O2], prepared by known methods (50), with p-C6H4O2 to deliver 2, as supported by a host of spectroscopic techniques: (i) the S=1/2 EPR signal of Fig. 5 The UV-vis absorption profile of 1 is strikingly similar to that of 2. The dark blue color of the two materials is a result of identical absorption bands centered at 825 nm (Fig. 4). This absorbance is strikingly similar to semiquinone radical anions prepared in anhydrous t-butanol (57) and reminiscent of that arising from the π dimer exciplex formed upon the reduction of 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) (38,39). In the cases of 1 and 2, π-stacking between quinone radical anions may be facilitated by lithium counterions bridging the oxygens of neighboring quinone radical anions. The absorption band of weaker intensity at 460 nm in 1 and 2 has been ascribed to the 2 B2g → 2 B3u HOMO-LUMO electronic transition of benzoquinone radical anion (47,61). This transition exhibits a strong dependence on the solvent environment, with values of the absorption maximum ranging from 427 nm in water to 454 nm in pyridine. The UV-vis spectrum of both powdered potassium superoxide (45) and superoxide in solution have been reported (42,62,63). Superoxide in solution has an absorption max at 250 nm while solid potassium superoxide's absorption max is 350 nm and tails out to 600 nm. Strong The EPR spectra of 1 and 2 are dominated by a broad featureless absorption at g = 2.008; however, the spectrum of 1 displays a significantly narrower signal. The superposition of a broad and narrow absorption has been noted for samples of potassium superoxide, with the broad line ascribed to the majority of the strong exchange-coupled superoxide anions present in the sample and the narrow line attributed to a very small population (approximately 1 in every 10 4 spins) of superoxide anions that are able to freely rotate in the sample and have poor electronic coupling with their environment (45).

The EPR spectrum of 1 is dominated by features arising from Li[p-C6H4O2], but the absence of the narrow absorption in the spectrum of 2 suggests that this feature is unique to 1 and may be tentatively assigned as arising from the superoxide ion.

The presence of this feature is not observed uniformly across all preparations of 1, presumably due to the extremely small percentage of spins contributing to this narrow line. The EPR spectrum of Li[p-C6H4O2] has been observed in prior work, (57,60) with the notable observation of hyperfine structure when EPR spectra were recorded with dilute samples. Loss of hyperfine structure in spectra of Li[p-C6H4O2] is seen for concentrated samples (60). The latter observation is consonant with the lack of hyperfine structure seen for solid samples of 1 and 2.

PXRD analysis lends further support to the assignment

Li + e -Fig. 10. Schematic representation of 1 (top) assuming the material starts from a 1 µm sphere of Li2O2. The analogy of 1 to a charging electrode is represented in the bottom panel.

of Li[p-C6H4O2] as a component of 1. The PXRD pattern 385 of 1 shown in Fig. 9 has peaks coincident with those of 2 386 prepared by mechanical comproportionation of Li2[p-C6H4O2] 387 and p-C6H4O2. Remaining peaks present in the pattern of 388 1 are assigned either to Li2O2 or unknown phase(s). Several 389 computational studies have predicted that the lowest energy 390 structure of LiO2 is the orthorhombic phase (64)(65)(66). Although 391 possible correspondence of the observed pattern with that of 392 a simulated pattern derived from ab initio calculations(64-66) 393 occurs near 2θ = 35 • , and the peaks labeled by asterisks 394 in Fig. 9 are coincident with peaks assigned as LiO2 (34), 395 definitive assignment of the unknown peaks cannot be made 396 nor can it be determined whether LiO2, if present in 1, is 397 amorphous.

O K-edge XAS of 1, 2, Li2[p-C6H4O2], and Li2O2 demon-399 strate marked differences between the samples. Differentiating 400 features are observed in the spectra of 2 and Li2[p-C6H4O2], 401 providing additional evidence for disproportionation of Li2[p-402 C6H4O2] and p-C6H4O2. The presence of 2 and Li2O2 as a 403 component of 1 obscures the pre-edge region of the O K-edge 404 XAS spectrum of 1 where LiO2 has been previously measured 405 in matrices (67), precluding its definitive identification.

The formulation of 1 may be addressed having established 407 that 1 is composed of Li[p-C6H4O2] and Li2O2 and inferred 408 the presence of LiO2 through EPR, PXRD, and titration 409 methods. Elemental analysis of 1 consistently indicates a 410 formula of [Li2O2]•[p-C6H4O2]0.7 yet spectroscopic analysis 411 indicates that the redox level of p-C6H4O2 is that of a monoan-412 ion. To accommodate the redox level of p-benzoquinone, a 413 commensurate number of lithium ions and electrons must 414 be drawn from Li2O2. In the context of reaction 2, a more 415 detailed formula of 1 including speciation is proposed to be 416 [Li2O2]0.3•[LiO2]0.7•{Li[p-C6H4O2]}0.7. Why Li2O2 stops ab-417 sorbing p-C6H4O2 after 0.7 equivalents is currently unknown 418 but may be related to particle size.

A schematic representation of 1 is provided in Fig. 10. 420

Commercial lithium peroxide is composed of particles with a diameter on the order of several hundred nanometers (Fig. 2). (43). Deposition of thick films of Li[p-C6H4O2] highlight the excellent conductivity of this material, however, the counterbalancing lithium cation need only be deposited on the surface of the growing film from solution. In contrast, in forming the core-shell structure of 1, lithium cations must diffuse from the core to the outer surface of the shell. Thus, termination of 1 with a core-shell structure suggests that the lithium-ion mobility of the shell is kinetically limiting. The schematic presented in Figure 10 along with qualitative observations of the electrical and ionic conductivity of the constituents of 1 may also explain why 1 does not absorb more than 0.7 equivalents of p-C6H4O2.

At the early stages of the reaction in which 1 is formed from Li2O2 and p-C6H4O2, the area of the interface between LiO2 and Li2O2 is large allowing for lithium and electron transfer from Li2O2 despite its poor conductivity. However, as the 482 reaction progresses, this interface drastically shrinks in size 483 and ultimately the rate of the reaction becomes limited by the 484 poor conductivity of Li2O2. Additionally, disproportionation 485 of LiO2 at the Li[p-C6H4O2]/Li2O2 interface may place a 486 poorly conducting barrier between the quinone and original 487 particle.

Several recent publications have used benzoquinone and an-489 thraquinone derivatives as soluble redox mediators in lithium-490 air batteries (76). 2,5-di-tert-butyl-1,4-benzoquinone was 491 found to enhance the rate of ORR and drastically increase 492 the capacity of a lithium-air cells (77,78). Another study 493 found that benzoquinone exhibited the best performance as 494 a redox mediator based on cathodic chronopotentiometry, al-495 though detection of lithium peroxide via XPS was the sole 496 physical characterization method (79), while another study has 497 suggested an explicit interaction between anthraquinone and 498 LiO2 in solution (80). Importantly, in the present study, the 499 solvent-free conditions employed enforce physical confinement 500 of the lithium peroxide/superoxide layer which may be crucial 501 in stabilizing thermodynamically unstable lithium superoxide. 502 This strategy of physical confinement may also be crucial in 503 ameliorating lithium superoxide induced solvent degradation, 504 a key challenge to overcome if higher cycling numbers are to 505 be achieved in lithium-air batteries (15, 29, 81-83).

506 Conclusions 507 Exposure of solid Li2O2 to p-C6H4O2 results in the formation 508 of a dark black material. This material has been investigated 509 by a variety of spectroscopic methods and is best described as 510 a coating of Li[p-C6H4O2] on LiO2 and Li2O2. This reaction 511 is unique in that electron transfer from Li2O2 occurs with p-512 C6H4O2 resulting in a comproportionation-like reaction. This 513 reaction methodology can be extended to the preparation of 514 the quinone radical anion from Li2[p-C6H4O2] and p-C6H4O2. 515 The preparation of 1 and compounds similar to 1 allow for 516 controlled "snapshots" of lithium peroxide during the electron 517 transfer from Li2O2 to electron acceptors (redox shuttles). 518 Indeed, we show here that LiO2 may be stabilized on Li2O2 519 surfaces in the presence of the electron accepting p-C6H4O2, 520 which as highlighted in Fig. 10 is a surrogate for the anode of 521 a lithium-air battery. By careful control of the potential of the 522 electron acceptor and concentration, it may be possible to inti-523 mately study the properties of electron deficient Li2O2, LiO2 524 and the 'superoxide like' sites, which crucially contribute to the 525 conductivity of Li2O2. Furthermore, a strategy of molecular 526 encapsulation of Li2O2 with a conductive layer may serve as 527 a promising method to protect cell components, including the 528 electrolyte, from deleterious degradation reactions initiated by 529 LiO2 and by extension improve the performance of metal-air 530 batteries.