Challenges and opportunities: Nonaqueous Li-O₂ batteries

Notable progress has been made in the development of rechargeable nonaqueous Li-O₂ batteries in the past five years. Work to date has enabled prototypical lab-scale demonstration of significant gravimetric energy density enhancements of Li-O₂ positive electrodes (Figure 1a) compared to state-of-the-art lithium-ion electrodes such as LiCoO₂,^{Reference Xia, Lu and Ceder15} LiFePO₄,^{Reference Zhu and Wang16} and LiNi_0.5Mn_0.5O₂.^{Reference Kang, Meng, Bréger, Grey and Ceder17} Reported gravimetric energy densities of cells based on oxygen electrodes^{Reference Christensen, Albertus, Sanchez-Carrera, Lohmann, Kozinsky, Liedtke, Ahmed and Kojic4,Reference Oh and Nazar9,Reference Wang, Xu, Xu, Zhang and Zhang10} upon first discharge are ∼3 times higher than those based on Li-ion positive electrodes (∼600 Wh/kg_cell),^{Reference Kang, Meng, Bréger, Grey and Ceder17,Reference Chen, Armand, Demailly, Dolhem, Poizot and Tarascon18} as shown in Figure 1b. It is important to note that the gravimetric energies for Li-O₂ systems only include the weight of carbon, Li₂O₂, and a metallic lithium anode, which are necessary to fairly compare the potential of the oxygen electrode relative to conventional Li-ion positive electrodes. Packaged Li-O₂ cell prototypes have not been widely developed, and the true gravimetric energy advantage of practical devices is not known. It is too early in the exploration of Li-O₂ battery technology to make accurate predictions of practical energy storage. Since the nature of the electrodes or electrolytes that might be used in a practical system are unknown, their tolerance to constituents in the air other than O₂ (e.g., CO₂) and the nature of any O₂ filtration approaches that might have to be taken to operate in air are also unknown. Additionally, if Li metal is used as the anode, the excess of Li that might be required in a new generation of protected Li anodes is as yet unknown. Such issues need to be investigated before hard-and-fast evidence-based estimates of practical performance can be made.

Figure 1.

(a) Working principles of an Li-O₂ cell and (b) gravimetric Ragone plot (power versus energy) of reported discharged Li-O₂ cells containing various carbon cathodes: carbon nanotubes (CNTs),^{Reference Gallant, Mitchell, Kwabi, Zhou, Zuin, Thompson and Shao-Horn8} carbon nanofibers (CNFs),^{Reference Mitchell, Gallant, Thompson and Shao-Horn20} and freestanding hierarchically porous carbon (“graphene”).^{Reference Wang, Xu, Xu, Zhang and Zhang10} Cells based on conventional Li-ion positive electrodes, LiCoO₂^{Reference Chen, Armand, Demailly, Dolhem, Poizot and Tarascon18} and LiNi_0.5Mn_0.5O₂,^{Reference Kang, Meng, Bréger, Grey and Ceder17} coupled with Li anodes for direct comparison are shown. Note that for Li-O₂ cells, the total mass of the cathode was used (carbon + Li₂O₂). Open orange circles are for CNTs discharged to a limited capacity of 1,000 mAh/g_C, while closed orange circles are from CNT electrodes discharged to 5,000 mAh/g_C.^{Reference Gallant, Kwabi, Mitchell, Zhou, Thompson and Shao-Horn19} The theoretical gravimetric energy of Li₂O₂ is calculated assuming a discharge voltage of 2.96 V versus Li and 1168 mAh/g Li₂O₂.

Unfortunately, rechargeable nonaqueous Li-O₂ batteries currently suffer from low round-trip efficiency (energy delivered on discharge divided by energy consumed on charge), cycle life, and power capability,^{Reference Lu, Gallant, Kwabi, Harding, Mitchell, Whittingham and Shao-Horn1,Reference Bruce, Freunberger, Hardwick and Tarascon3–Reference Bruce, Hardwick and Abraham5,Reference Gallant, Kwabi, Mitchell, Zhou, Thompson and Shao-Horn19} which limits their practical use. These performance challenges result, in part, from a lack of fundamental understanding of the mechanisms of oxygen reduction and evolution reactions in the oxygen electrode. Challenges also include the chemical instability of active and inactive components, including the dimensional stability of metallic lithium during repeated oxidation and plating, the chemical stability of conventional current collector materials for the oxygen electrode (such as carbon), and nonaqueous electrolytes in the oxygen electrode. In the first part of this article, we discuss recent research findings that highlight the current understanding of oxygen reduction and evolution reactions in nonaqueous electrolytes and the state of the search for chemically stable nonaqueous electrolytes and porous cathode substrates.

Using vertically aligned carbon nanotubes (CNTs) as the oxygen electrode has provided valuable insights into the mechanism of oxygen reduction and evolution reactions, specifically Li₂O₂ growth and oxidation mechanisms and the relationship between Li₂O₂ morphology/chemistry and the voltage penalty, or overpotential, associated with Li₂O₂ growth and oxidation. Toroidal structures of Li₂O₂ (Figure 2a), with sizes of hundreds of nanometers, form upon discharge at high discharge potentials (greater than 2.7 V_Li). The aspect ratio of the toroids are scaled with the toroidal diameter, while small Li₂O₂ particles (Figure 2b, <20 nm) are formed conformally on CNTs at discharge voltages lower than 2.7 V_Li.^{Reference Wang, Xu, Xu, Zhang and Zhang10,Reference Mitchell, Gallant, Thompson and Shao-Horn20–Reference Xu, Wang, Xu, Zhang and Zhang22} Toroidal morphologies are characteristic of electrochemically formed Li₂O₂, and they have been observed in a broad range of non-carbonate electrolytes (ethers and dimethyl sulfoxide [DMSO]).^{Reference Wang, Xu, Xu, Zhang and Zhang10,Reference Xu, Wang, Xu, Zhang and Zhang22,Reference Mitchell, Gallant, Shao-Horn and Thompson23} Mitchell et al.^{Reference Mitchell, Gallant, Shao-Horn and Thompson23} have shown that these toroids are composed of thin plates of epitaxially oriented Li₂O₂, ∼10 nm in thickness with the $\left( {001} \right)_{{\rm{Li}}_2 {\rm{O}}_2 } $ direction normal to the plate surface (Figure 2c–d). Recent surface-sensitive x-ray absorption near edge structure (XANES) findings suggest that although Li₂O₂ is invariably formed upon battery discharge, the electronic structure of the surfaces of the toroids have oxygen-rich LiO₂-like species (Figure 2f), while the conformally coated particles retain a more stoichiometric Li₂O₂-related configuration. These findings are in agreement with recent theoretical calculations that reported the thermodynamically stable $\left( {001} \right)_{{\rm{Li}}_2 {\rm{O}}_2 } $ surface of Li₂O₂, which is the dominant surface termination of toroidal Li₂O₂ particles, should possess LiO₂-like character.^{Reference Hummelshøj, Luntz and Nørskov24}

Figure 2.

Transmission electron microscopy (TEM) images of carbon nanotube (CNT) electrodes discharged at (a) 10 mA/g to 4000 mAh g_C^–1 showing Li₂O₂ toroids, (b) 2.4 V to 5,000 mAh g_C^–1, (c) electron diffraction patterns of toroidal Li₂O₂, and (d) high-resolution TEM imaging of [001]-oriented Li₂O₂ plates, with lower magnification in inset. Reproduced with permission from References 19 and 23. © 2013 American Chemical Society. (e) In situ surface-enhanced Raman spectra during oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) on an Au electrode in O₂-saturated electrolyte. The peaks are assigned as follows: (1) C–C stretch of CH₃CN at 918 cm^–1, (2) O–O stretch of LiO₂ at 1137 cm^–1, (3) O–O stretch of Li₂O₂ at 808 cm^–1, and (4) Cl–O stretch of ClO₄ at 931 cm^–1. Adapted with permission from Reference 25. (f) Surface-sensitive x-ray absorption near-edge structure O–K edge data of toroids formed at 10 mA/g to 4,000 mAh g_C^–1 and particles formed at 2.4 V to 5,000 mAh g_C^–1 showing that particles closely resemble stoichiometric Li₂O₂, while toroids exhibit LiO₂-like features. (g) Constant-current charging of Li₂O₂ discharged at 2.0, 2.4, 2.5, and 2.76 V versus Li/Li⁺ at 100 mA/g_C, showing the evolution of charge potential (E) with charge capacity (Q), with particle charging occurring at a lower average voltage penalty than toroid oxidation. Reprinted with permission from References 19 and 23. © 2013 American Chemical Society.

The differences in the morphology and surface chemistry of Li₂O₂ particles may be explained by the following oxygen reduction elementary reaction steps^{Reference Peng, Freunberger, Hardwick, Chen, Giordani, Bardé, Novák, Graham, Tarascon and Bruce25} in nonaqueous electrolytes:

$${{\rm{O}}_2} + {{\rm{e}}^ - } \to {{\rm{O}}_2}^ -$$

$${\rm{L}}{{\rm{i}}^ + } + {{\rm{O}}_2}^ - \to {\rm{Li}}{{\rm{O}}_2}$$

$$2{\rm{Li}}{{\rm{O}}_2} \to {\rm{L}}{{\rm{i}}_2}{{\rm{O}}_2} + {{\rm{O}}_2}$$

or

$${\rm{L}}{{\rm{i}}^ + } + {\rm{Li}}{{\rm{O}}_2} + {{\rm{e}}^ - } \to {\rm{L}}{{\rm{i}}_2}{{\rm{O}}_2}$$

The appearance of LiO₂ as an intermediate (Figure 2e), followed by a disproportionation or a second electron transfer reaction to form Li₂O₂, is supported in work by Peng et al.^{Reference Peng, Freunberger, Hardwick, Chen, Giordani, Bardé, Novák, Graham, Tarascon and Bruce25} using in situ surface-enhanced Raman spectroscopy (SERS). At high discharge potentials (low overpotentials), the growth of toroidal Li₂O₂ particles may involve the formation of soluble reaction intermediate species such as O₂^– and LiO₂ and the growth of Li₂O₂ by chemical disproportion of LiO₂.^{Reference Adams, Radtke, Black, Trudeau, Zaghib and Nazar26,Reference Zhai, Wang, Yang, Lau, Li, Amine and Curtiss27} These findings suggest that the increasing surface area associated with the increasing disc diameter during discharge promotes Li₂O₂ growth onto the disc surfaces, leading to the development of thick, toroidal particles. Lowering the discharge potential (increasing the overpotential) promotes the kinetics of the second electron transfer step such that soluble reaction intermediates are reduced to form Li₂O₂ conformally coated on CNTs before they diffuse.

Differences in the morphology and surface chemistry of Li₂O₂ toroids and conformally coated particles on CNTs appear to influence Li₂O₂ oxidation kinetics.^{Reference Gallant, Kwabi, Mitchell, Zhou, Thompson and Shao-Horn19} Upon charging at low rates, the oxidation of conformally coated Li₂O₂ particles on CNTs exhibit a solid-solution-like voltage profile, while oxidation of Li₂O₂ toroids yields two-phase oxidation with a higher voltage penalty (Figure 2g). Although the physical origin of this observed difference is not fully understood, it is hypothesized^{Reference Gallant, Kwabi, Mitchell, Zhou, Thompson and Shao-Horn19} that the oxidation kinetics of Li₂O₂ are dependent on surface termination. Thus, the oxidation of Li₂O₂ particles coated on CNTs with multiple facets can occur at different voltages, giving rise to the sloping voltage in contrast to toroidal particles with predominantly ${\rm{Li}}_{\rm{2}} {\rm{O}}_{\rm{2}}$ surface terminations. This work highlights the need to understand how the surface chemistry and microstructure of discharged Li₂O₂ influence oxygen evolution reaction (OER) kinetics and opens up potential opportunities to explore nucleation and growth mechanisms of Li₂O₂ particles for high-energy and/or high-power Li-O₂ batteries.

More interestingly, electron rather than lithium-ion transport limits the electrochemical oxidation kinetics of Li₂O₂ when the oxidation current is not limited by charge transfer kinetics under large overpotentials. This was shown in recent in situ transmission electron microscopy (TEM) imaging experiments with solid-state microbatteries^{Reference Zhong, Mitchell, Liu, Gallant, Thompson, Huang, Mao and Shao-Horn28} consisting of Si nanowires (NWs) coated with Li-ion-conducting solid electrolyte contacting Li₂O₂ particles. In particular, as shown in Figure 3, the oxidation of toroidal Li₂O₂ particles at very high overpotentials is initiated preferentially at the CNT/Li₂O₂ interface, as evidenced by the development of a light-contrast stripe beginning along the CNT axes (highlighted by the dashed red lines) of the CNT bundle contacting the Li₂O₂ particle. The formation of the light-contrast stripe in TEM can be explained by the thinning of Li₂O₂ particles at the Li₂O₂/CNT interface associated with electrochemical oxidation of Li₂O₂ (Li₂O₂ → 2Li⁺ + O₂ + 2e^–) and the concurrent release of O₂, while lithium ions migrate through the solid electrolyte coating the Si NW, and electrons flow through CNTs to the Au substrate.

Figure 3.

In situ transmission electron microscope (TEM) images showing oxidation of Li₂O₂ particles. (a) Schematic illustration of the in situ TEM microbattery superimposed over a low magnification TEM image of a solid electrolyte (SE)-coated Si nanowire contacting a single Li₂O₂ particle. (b) Higher magnification TEM image of the particles in (a) showing a multiwalled carbon nanotube (MWCNT) bundle contacting two physically separated Li₂O₂ particles labeled as Particle 1 and Particle 2, respectively. (c–g) Oxidation of Particles 1 and 2 during application of a 10 V_cell potential to the MWCNT/Li₂O₂ positive electrode against the Si nanowire (NW) negative electrode. CNT axes are delineated by the red, dashed lines. Reprinted with permission from Reference 28. © 2013 American Chemical Society.

The preferential oxidation at the MWCNT (multiwalled CNT)/Li₂O₂ interface, but not at the interface between Li₂O₂/solid electrolyte/Si, suggests that electrochemical oxidation kinetics of Li₂O₂ is electron-transport-limited instead of lithium-ion-transport-limited at very high overpotentials. These findings suggest that electrodes with large specific surface area to maximize the Li₂O₂/electrode interfacial area and high electronic conductivity to provide facile electron transport to reaction sites are desirable for the design of high rate rechargeable Li-O₂ batteries.

Stability of inactive components in the oxygen electrode

Understanding the stability of inactive electrode components in Li-O₂ batteries is an intensive area of research since their stability is a major factor determining rechargeability and cycle life.^{Reference McCloskey, Speidel, Scheffler, Miller, Viswanathan, Hummelshøj, Nørskov and Luntz29–Reference Mizuno, Nakanishi, Kotani, Yokoishi and Iba35} Differential electrochemical mass spectrometry (DEMS) experiments on isotopically labeled ¹³C cathodes have suggested that after discharge, a thin layer of Li₂CO₃ is present between Li₂O₂ and carbon.^{Reference McCloskey, Speidel, Scheffler, Miller, Viswanathan, Hummelshøj, Nørskov and Luntz29} This hypothesis is supported by XANES studies^{Reference Gallant, Mitchell, Kwabi, Zhou, Zuin, Thompson and Shao-Horn8} on discharged CNTs in Li-O₂ batteries, which show that Li₂CO₃ is present at the interface between Li₂O₂ and the CNT walls on first discharge, and suggest that carbon is not chemically stable during the oxygen reduction reaction (ORR) and is potentially reactive against Li₂O₂ and ORR intermediates such as O₂^– and LiO₂.

Recent in situ ambient pressure x-ray photoelectron spectroscopy (XPS) measurements further demonstrate the instability of carbon during ORR^{Reference Itkis, Semenenko, Kataev, Belova, Neudachina, Sirotina, Hävecker, Teschner, Knop-Gericke, Dudin, Barinov, Goodilin, Shao-Horn and Yashina30} at the interface between graphene and a solid-state lithium ion conductor (with no nonaqueous electrolyte present). The formation of epoxy groups on the carbon surface and their conversion to carbonates in the presence of LiO₂ was observed and provides direct evidence for carbon instability during battery discharge. Interestingly, studies by Thotiyl et al.^{Reference Ottakam Thotiyl, Freunberger, Peng and Bruce32} using acid treatment and Fenton’s reagent, which quantify the amounts of side-reaction products formed from decomposition of the electrolyte and the carbon electrode respectively, revealed that on discharge, it is the decomposition of the electrolyte not the carbon electrode that dominates, forming Li₂CO₃ and carboxylate species. The degree of electrolyte decomposition depends on the type of carbon used. This finding was further confirmed in a recent study by McCloskey et al.^{Reference McCloskey, Valery, Luntz, Gowda, Wallraff, Garcia, Mori and Krupp31} In addition to significant electrolyte decomposition during ORR, they also showed that the degree of parasitic reactions is sensitive to discharge rate (Figure 4b).^{Reference McCloskey, Valery, Luntz, Gowda, Wallraff, Garcia, Mori and Krupp31} Further studies are required to elucidate the reaction mechanism responsible for electrolyte decomposition with and without carbon as a function of discharge rate, carbon chemistry, and electrolyte solvent and salt.

Figure 4.

(a) Discharge-charge curve showing evolution of reaction potential (U) with charge passed (Q) of a carbon paper-based cathode in 1 M LiTFSI (Li(CF₃SO₂)₂N) in dimethoxyethane, along with reaction product quantification during oxygen reduction reaction and oxygen evolution reaction using iodometric titration and nuclear magnetic resonance techniques. Reaction product quantification shows that Li₂O₂ is the dominant product on discharge (along with carbonate species, R = Li or CH₃), while CO₂ and H₂O are evolved during charge as a result of electrolyte decomposition. (b) Yield of Li₂O₂ ($Y_{{\rm{Li}}_2 {\rm{O}}_{\rm{2}} }$) as a function of Li-O₂ battery discharge rate (i) for the same cathode, showing greater chemical reactivity between Li₂O₂ and the electrolyte at lower rates. Reprinted with permission from Reference 31. © 2013 American Chemical Society.

Turning to charging, DEMS experiments with isotopic carbon labeling show that carbon is unstable during Li₂O₂ oxidation above 3.5 V versus Li/Li⁺: CO₂ is evolved, with the carbon originating from the carbon support in the oxygen electrode (Figure 5a–b).^{Reference McCloskey, Speidel, Scheffler, Miller, Viswanathan, Hummelshøj, Nørskov and Luntz29} The analysis of ¹³C labeled cathodes at different states of charge with acid and Fenton’s reagent has revealed that the process is complex. First, while the electrolyte continues to decompose, carbon electrode decomposition to form Li₂CO₃ becomes significant and grows to dominate at higher states of charge. The fact that carbon decomposition commences at 3.5 V, where Li₂O₂ starts to oxidize significantly, and not at 4 V, where carbon oxidation is expected, indicates that the process of carbon decomposition involves Li₂O₂ or its intermediates of oxidation. In situ DEMS shows that Li₂CO₃ simultaneously forms by decomposition of both the carbon cathode and the electrolyte and decomposes on charging. As a result, the amount of O₂ evolved on charge is not necessarily expected to be equal to that predicted by the amount of Li₂O₂ present at the end of discharge. An iodometric titration method has been developed to quantify the amount of Li₂O₂ present in the oxygen electrode and has confirmed this picture. When applied at different stages of discharge and charge^{Reference McCloskey, Valery, Luntz, Gowda, Wallraff, Garcia, Mori and Krupp31} and combined with DEMS measurements of O₂ evolution and consumption, it has been shown the Li₂O₂ yield during charge is considerably less than 100%. This finding shows that not all Li₂O₂ present upon discharge evolves O₂ on charge (i.e., it partially participates in decomposition reactions), and confirms that carbon is relatively more unstable during charge than discharge.

Figure 5.

Quantification of carbon-containing by-products in carbon and TiC cathodes on discharge and charge, respectively. Data demonstrate significant side reactions both on discharge and charge at carbon electrodes. These are greatly suppressed at TiC cathodes. (a) Discharge-charge curve of a carbon cathode in 0.1 M LiClO₄/dimethyl sulfoxide electrolyte on the first cycle. (b) Moles of CO₂ evolved from the cathode stopped at certain points of discharge/charge indicated in (a) and then treated to decompose Li₂CO₃ and lithium carboxylates.^{Reference Ottakam Thotiyl, Freunberger, Peng and Bruce32} (c) The same procedure applied to TiC electrodes in the same electrolyte at the indicated numbers of discharge (1D, 5D, . . . ) and charge (1C, 5C, . . . ); shown are inorganic and organic carbonaceous products. (d) Evolution of percentage of carbonate products with cycle number at a carbon and TiC electrode, respectively.^{Reference Ottakam Thotiyl, Freunberger, Peng, Chen, Liu and Bruce34} Reprinted with permission from Reference 32. © 2013 American Chemical Society and Reference 34. © 2013 Nature Publishing Group.

In situ ambient pressure XPS measurements have shown highly reversible Li₂O₂ formation on a Li_xV₂O₅ surface in a solid-state cell,^{Reference Lu, Crumlin, Veith, Harding, Mutoro, Baggetto, Dudney, Liu and Shao-Horn33} thus highlighting the importance of the design of oxygen electrodes that are stable for Li-O₂ batteries to allow for long cycle life. In light of the observed instability of carbon, replacing it with a nanoporous gold electrode was found to provide stable cycling of Li-O₂ batteries.^{Reference Peng, Freunberger, Chen and Bruce7} As Au is heavy and expensive, TiC, TiN, Fe/N/C, N-doped carbon, and stainless steel have all been investigated as alternative materials to replace carbon.^{Reference Ottakam Thotiyl, Freunberger, Peng, Chen, Liu and Bruce34} Titration of each candidate cathode after each discharge/charge with acid and Fenton’s reagent revealed the amount of inorganic carbonates such as Li₂CO₃ and organic carboxylate species, respectively, that were formed (Figure 5c–d). In comparison to carbon, TiC showed encouraging performance and stability, which is attributed to a layer of TiOC formed on its surface.^{Reference Ottakam Thotiyl, Freunberger, Peng, Chen, Liu and Bruce34}

Ether electrolytes

Recent extensive investigation of nonaqueous electrolytes for Li-O₂ cells, based on ethers, amides, ionic liquids (ILs), and DMSO, have shown that they all predominantly support the formation of Li₂O₂ upon first discharge and its removal on recharge.^{Reference Ottakam Thotiyl, Freunberger, Peng and Bruce32,Reference Freunberger, Chen, Drewett, Hardwick, Bardé and Bruce44–Reference Leskes, Drewett, Hardwick, Bruce, Goward and Grey46} Ethers are attractive for the Li–O₂ battery because they are safe, cheap, compatible with lithium metal anodes, more stable toward reduced O₂ species compared with organic carbonates, and have relatively high oxidation windows (up to 4 V versus Li/Li⁺), and low volatilities (for longer-chain ethers). However, there is growing experimental and theoretical evidence that ether-based electrolytes are not suitable for Li-O₂ cells.^{Reference McCloskey, Bethune, Shelby, Girishkumar and Luntz13,Reference McCloskey, Speidel, Scheffler, Miller, Viswanathan, Hummelshøj, Nørskov and Luntz29,Reference McCloskey, Valery, Luntz, Gowda, Wallraff, Garcia, Mori and Krupp31,Reference Freunberger, Chen, Drewett, Hardwick, Bardé and Bruce44,Reference Younesi, Hahlin, Treskow, Scheers, Johansson and Edstrom60–Reference Ryan, Trahey, Ingram and Burrell62} Ethers are known to form peroxides under ambient conditions by auto oxidation reactions.^{Reference Clover63} This instability toward auto oxidation has been identified as a possible pathway of initial attack of the solvent that gives rise to further decomposition reactions Figure 6b).^{Reference Bryantsev and Faglioni61}

Figure 6.

Possible reaction schemes for ether decomposition. (a) Proposed reaction scheme on discharge to explain decomposition of ether: electron reduction of O₂ to superoxide (1), binding with Li⁺ (2), and disproportionation to Li₂O₂ (3) as the ideal sequence. Alternatively, superoxide may abstract a proton from the solvent (4) to give hydroperoxide (5), which may further react in a series of oxidative decomposition steps (6) or yield esters or polymeric compounds (7). Superoxide and CO₂ from step (6) react to form Li₂CO₃ (8 and 9). (b) Alternative possible pathway with initial attack by autoxidation to give hydroperoxyl and alkyl radicals (I) and chain propagation by O₂ addition and further proton abstraction (II and III). Note: The dot (^•) denotes a free radical. (a) Reprinted with permission from Reference 44. © 2012 American Chemical Society. (b) Reprinted with permission from Reference 61. © 2011 American Chemical Society.

Challenges and opportunities: Aqueous lithium-oxygen batteries

In the second part of this article, we will discuss recent efforts in protected lithium metal as the negative electrode in aqueous Li-O₂ batteries. Aqueous batteries have received significantly less attention than nonaqueous Li-O₂ batteries due to the practical challenges involved with preventing a violent reaction between lithium and water. Developing lithium metal as the negative electrode that can cycle safely and stably has been a challenge for 40 years, owing to the dendrites that form on repeated lithium dissolution and lithium deposition. Dendrites can be fatal to battery performance since they pose the risk of growing into the cathode and causing a short circuit and battery failure. There are still no rechargeable lithium anode-based batteries on the market; all rechargeable Li-ion batteries have thus far contained the lithium inside another material such as aluminum (LiAl),^{Reference Jang, Neudecker and Dudney67,Reference Levi and Aurbach68} carbon (LiC₆),^{Reference Xia, Lu and Ceder15,Reference Zhu and Wang16,Reference Visco, Katz, Chu and De Jonghe69} and, more recently, tin^{Reference Xia, Lu and Ceder15,Reference Zhang, Li, Khan and Fedkiw70,Reference Rao, Francis and Christopher71} and silicon.^{Reference Megahed and Scrosati72}

Zhang et al.^{Reference Zhang, Li, Khan and Fedkiw70} have reported lithium dendrite formation in a typical nonaqueous electrolyte containing polyethylene glycol dimethyl ether (PEGDME, Mw. 250) with Li(CF₃SO₂)₂N (LiTFSI) for lithium-air rechargeable Li-O₂ batteries. The Li/PEGDME-LiTFSI/Li cell showed short circuits after 21.4 h of polarization at 0.2 mA cm^–2 and 4.0 h of polarization at 1.0 mA cm^–2. A lithium metal alloy with aluminum solved the dendrite issue,^{Reference Rao, Francis and Christopher71} but alloy electrodes survive only a limited number of cycles owing to extreme changes in volume during operation.

To reduce lithium dendrite formation, solid electrolytes have been developed for electric vehicle applications.^{Reference Megahed and Scrosati72} Theoretical calculations suggest that a homogenous solid-state electrolyte with a modulus of 6 GPa would drastically reduce dendrite formation.^{Reference Monroe and Newman73} As the moduli of ceramic materials are generally higher than 6 GPa, little or no lithium dendrite formation between a ceramic lithium-ion conductor and lithium metal is expected. Visco et al.^{Reference Visco, Katz, Chu and De Jonghe69} and Stevens et al.^{Reference Stevens, Toussaint, Caillon, Viaud, Vinatier, Cantau, Fichet, Sarrazin and Mallouki74} first reported experimental evidence of the feasibility of protected lithium metal electrodes, with Li/Li₃N/Li_1+xAl_xTi_2–x(PO₄)₃ (LATP) and Li/Li_3–xPO_4–yN_y/LATP, respectively. Li₃N and Li_3–xPO_4–yN_y are used to prevent direct contact of lithium metal with LATP, since LATP is electrochemically unstable at the potential of metallic lithium.^{Reference Hasegawa, Imanishi, Zhang, Xie, Hirano, Takeda and Yamamoto11,Reference Imanishi, Hasegawa, Zhang, Hirano, Takeda and Yamamoto12,Reference Capsoni, Bini, Ferrari, Quartarone and Mustarelli75} Aqueous Li-O₂ batteries assembled with these lithium electrodes could be cycled at 1 mA cm^–2 and room temperature without degradation of the cell performance for a short period of 10 min.^{Reference Visco, Katz, Chu and De Jonghe69,Reference Monroe and Newman73} Subsequently, other protected lithium anodes based on garnet-type solid electrolytes^{Reference Murugan, Thangadurai and Weppner76} and polymer-ceramic composite membranes^{Reference Kumar, Kumar, Leese, Fellner, Rodrigues and Abraham77} have been reported.

While ceramic conductors are stiff enough to prevent lithium dendrite growth, they do not easily make low-resistance contacts with other solid components and thus have high charge transfer resistance. Recent research efforts have been focused on using polymer-ceramic composite electrolytes to form a conformal interface and reduce interfacial resistance. Rosso et al. investigated the mechanism of lithium dendrite growth in Li/PEO-based-electrolyte/Li (PEO = polyethylene oxide),^{Reference Rosso, Gobron, Brissot, Chazalviel and Lascaud78,Reference Brissot, Rosso, Chazalviel and Lascaud79} where the onset of dendrite formation and the time to reach short-circuit times can be studied by direct in situ observation and simultaneous cell voltage monitoring. Table I summarizes the dendrite onset and short-circuit time for different compositions of PEO-based electrolytes. No dendrite formation was observed in a Li/PEO₂₀LiTFSI/Li cell at a low current density of 0.03 mA cm^–2 for 100 h at 80°C, while at a high current density of 0.7 mA cm^–2, dendrites were observed after less than 1 h of deposition.^{Reference Brissot, Rosso, Chazalviel and Lascaud79} Such lithium dendrite formation onset times at high current density are too short to be used in practical applications. The use of ceramic or IL additives in polymer electrolytes comprises another strategy to suppress dendrite formation. The addition of nano-SiO₂ and/or IL of N-methyl-N-propylpiperidinium TFSI (PP13TFSI)^{Reference Liu, Imanishi, Zhang, Hirano, Takeda, Yamamoto and Yang80,Reference Liu, Wang, Imanishi, Zhang, Hirano, Takeda, Yamamoto and Yang81} to polymer electrolytes has been shown to increase the onset time of dendrite formation from 10 h for Li/PEO₁₈TFSI/Li to 21 h at 1.0 mA cm^–2 at 60°C. Figure 7a shows the cycling performance of the lithium deposition and dissolution process in a Li/PEO₁₈LiTFSI-SiO₂-PP13TFSI/Li cell at 60°C, where the cell voltage showed no significant change during cycling, and no dendrite formation was detected at a current density of 0.3 mA cm^–2 for 30 h.^{Reference Liu, Wang, Imanishi, Zhang, Hirano, Takeda, Yamamoto and Yang81}

Table I.

Dependence of lithium dendrite formation onset time (t _o) and short-circuit time (t _s) on electrolyte formulation, temperature, and current density.

Note: PVdF, polyvinylidene fluoride; HFP, hexafluorophosphate; EC, ethylene carbonate; PC, propylene carbonate; PEO, polyethylene oxide; LiTFSI, Li(CF₃SO₂)₂N; PEGDME, polyethylene glycol dimethyl ether; PP13, N-methyl-N-propylpiperidinium.

Figure 7.

(a) Cycling performance for the Li/PEO₁₈LiTFSI-SiO₂-PP13TFSI/Li cell at 0.3 mA cm^–2 and 60°C. The inset shows a comparison of the 11th cycle for Li/PEO₁₈LiTFSI-SiO₂-PP13TFSI/Li and Li/PEO₁₈LiTFSI-PP13TFSI/Li. Cycling performance of the Li/PEO₁₈LiTFSI-1.44PP13TFSI/LATP/aqueous 1 M LiCl-0.004 M LiOH/Pt-air cell at 0.3 mA cm^–2 and 60°C between (b) 0 h and 400 h. The water-stable lithium electrode (WSLE) voltage was measured by a Pt-air reference electrode. Discharge and charge period are both two hours. Cycle was repeated without interval. Microscopic observations of the cell indicated no dendrite formation after cycling. Note: PEO, polyethylene oxide; LiTFSI, Li(CF₃SO₂)₂N; PP13, N-methyl-N-propylpiperidinium; LATP, Li_1+xAl_xTi_2–x(PO₄)₃. (a) Reprinted with permission from Reference 81. © 2014 Elsevier. (b) Reprinted with permission from Reference 82. © 2014 The Electrochemical Society.

Based on these positive results, this composite polymer electrolyte of PEO₁₈LiTFSI-SiO₂-PP13TFSI was also assessed as the protected lithium-metal anode for rechargeable lithium-air batteries. Figure 7b shows reversible cycling of a Li/PEO₁₈LiTFSI-PP13TFSI/LATP/aqueous 1 M LiCl-0.004 M LiOH/Pt, air cell at 0.3 mA cm^–2 and 60°C,^{Reference Zhang, Imanishi, Hirano, Takeda and Yamamoto82} with LATP as the solid electrolyte. The inclusion of PP13TFSI IL is effective in reducing interfacial resistance between lithium and the electrolyte and facilitating highly reversible Li-O₂ cycling.

While the early onset of lithium dendrite formation and low output power remain yet to be completely solved, recent research has shown that these challenges are closely related to the physical properties of the solid electrolyte used. Further materials development for the protection layer of lithium is therefore essential to achieve high power density at several mA cm^–2 and extended deep cycling of the lithium electrode for Li-O₂ batteries.

Summary

The need for storage technologies with much greater gravimetric energy than Li-ion batteries makes Li-O₂ batteries an attractive energy storage option for applications that require high-energy density such as electric vehicles. Although notable progress has been made in the development of rechargeable, lab-scale nonaqueous Li-O₂ cells, further investigation of the oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) in nonaqueous electrolytes, and parasitic reaction mechanisms involving inactive components in the oxygen electrode is required in order to aid the rational design and development of kinetically facile and chemically stable electrode and electrolyte configurations.

As detailed in this review, researchers have just begun to connect the physical properties (morphology and surface versus bulk chemistry) of Li₂O₂ with the corresponding kinetics of Li₂O₂ formation and oxidation. Important topics warranting further investigation in this area include the detailed reaction and growth mechanisms of Li₂O₂ at high and low overpotentials; the electronic nature (surface and bulk) of electrochemically formed Li₂O₂; intrinsic Li₂O₂ oxidation pathways; and the influence of non-carbonate solvents on the nucleation and growth of Li₂O₂ and its oxidation.

As carbon is not stable in the oxygen electrode, improved morphological, chemical, and kinetic understanding of Li-O₂ reactions in non-carbon electrodes will become increasingly relevant for the development of rechargeable Li-O₂ batteries. The search for electrolytes that are chemically stable against the oxygen electrode is another major focus of research. The discovery that reduced oxygen species produced during ORR are also highly reactive against a broad range of electrolyte solvents has spurred detailed spectroscopic, spectrometric, and titration studies examining the interplay between electrolyte formulation and discharge reaction product. Combining these experimental approaches with theoretical calculations evaluating electrolyte susceptibility to attack by Li₂O₂ and/or ORR intermediates will be crucial to finding an electrolyte that is moderately stable during ORR, but also possesses a favorable synergy with the chosen electrode material such that parasitic reactions during OER are minimized.

Significant progress has been made toward finding materials that suppress lithium dendrite formation and also adequately protect the lithium metal anode from contact with reactive gases and water in both nonaqueous and aqueous Li-O₂ battery environments. Representing a compromise between the favorable transport properties of liquid and the high mechanical stability of solid electrolytes, the use of polymers has emerged as a particularly promising strategy to protect lithium metal anodes. Further research into the possibility of incorporating nanostructured additives to improve power capabilities while maintaining chemical stability against lithium metal in full Li-O₂ cells is a promising direction for future efforts in this field. The need for fundamentally understanding the relationship between electrode reaction mechanisms and parasitic processes occurring in rechargeable Li-O₂ batteries presents an opportunity to combine cutting edge experimental and computational techniques with materials innovation in the areas outlined previously to enable the development of practical rechargeable Li-O₂ batteries, which combine high rate capability with high round-trip efficiency and long cycle life.