A high-energy-density lithium-oxygen battery based on a reversible four-electron conversion to lithium oxide

See allHide authors and affiliations

Science  24 Aug 2018:
Vol. 361, Issue 6404, pp. 777-781
DOI: 10.1126/science.aas9343

An elevated lithium battery

Batteries based on lithium metal and oxygen could offer energy densities an order of magnitude larger than that of lithium ion cells. But, under normal operation conditions, the lithium oxidizes to form peroxide or superoxide. Xia et al. show that, at increased temperatures, the formation of lithium oxide is favored, through a process in which four electrons are transferred for each oxygen molecule (see the Perspective by Feng et al.). Reversible cycling is achieved through the use of a thermally stable inorganic electrolyte and a bifunctional catalyst for both oxygen reduction and evolution reactions.

Science, this issue p. 777; see also p. 758


Lithium-oxygen (Li-O2) batteries have attracted much attention owing to the high theoretical energy density afforded by the two-electron reduction of O2 to lithium peroxide (Li2O2). We report an inorganic-electrolyte Li-O2 cell that cycles at an elevated temperature via highly reversible four-electron redox to form crystalline lithium oxide (Li2O). It relies on a bifunctional metal oxide host that catalyzes O–O bond cleavage on discharge, yielding a high capacity of 11 milliampere-hours per square centimeter, and O2 evolution on charge with very low overpotential. Online mass spectrometry and chemical quantification confirm that oxidation of Li2O involves transfer of exactly 4 e/O2. This work shows that Li-O2 electrochemistry is not intrinsically limited once problems of electrolyte, superoxide, and cathode host are overcome and that coulombic efficiency close to 100% can be achieved.

Lithium-oxygen (Li-O2) batteries have attracted interest because of their energy density being at least one magnitude higher than that of conventional Li-ion batteries (1). A typical Li-O2 cell is composed of a Li anode and a porous carbon cathode, separated by a Li+-ion conducting organic electrolyte (2). During discharge, oxygen is reduced on the carbon cathode, forming insoluble lithium peroxide (Li2O2) (Eq. 1), and this product is oxidized upon charge.2Li + O2 → Li2O2 ΔrG° = –571 kJ mol–1(1)The commercialization of the Li-O2 battery and its Na-O2 battery cousin (3) have been impeded by the decomposition of organic electrolytes as well as by the corrosion of the porous carbon cathode hosts (4, 5). These drawbacks are directly correlated to the high reactivity of peroxide and superoxide in the Li-O2 and Na-O2 batteries, respectively, and superoxide has also been reported as the sole product of the Li-O2 cell (6). Although effort has been made to improve the cycling performance of Li-O2 cells—albeit by means of a complex and debated pathway (7, 8)—the issues remain.

In comparison with the peroxide and superoxide, lithium oxide (Li2O) is much less chemically reactive with organic solvents. Both peroxide (9) and superoxide (10) react with organic electrolytes and with carbon in the cell. Owing to its oxide anion, Li2O is benign as an oxidizing agent; for example, it does not react with dimethyl sulfoxide (DMSO), whereas Li2O2 oxidizes DMSO to form dimethyl sulfone (fig. S1). More importantly, Li-O2 cells based on Li2O as the discharge product (Eq. 2) can theoretically deliver a high specific energy and energy density of 5.2 kilowatt-hour (kWh) kg−1 and 10.5 kWh L−1, respectively, exceeding that of fossil fuels [gasoline (9.5 kWh L−1)] (2)2Li + 0.5O2 → Li2O ΔrG° = –561 kJ mol–1(2)Therefore, it is of much interest to seek a pathway for reversible oxygen reduction to the oxide. Consideration of the thermodynamics of the oxygen reduction reaction (ORR) shows that not only is the standard Gibbs reaction energy (ΔrG°) of Eq. 1 [–571 kJ mol−1 (Li2O2)] lower than that of Eq. 2 [–561 kJ mol−1 (Li2O)] (11), but the formation of oxide requires O–O bond cleavage of oxygen molecules, whereas peroxide does not. Therefore, the formation of Li2O2, not Li2O, is thermodynamically and kinetically favored at ambient conditions.

We demonstrate that by increasing the operating temperature and exploiting stable inorganic electrolytes and ORR catalysts, the reversible formation of Li2O leads to a highly rechargeable Li-O2 cell with high capacity, low overpotential with transfer of 4 e/O2, and excellent cycling performance. Our cells were cycled at 150°C because the thermodynamic driving force favors Li2O as the product above this temperature rather than Li2O2 (Fig. 1A), as described by the Gibbs-Helmholtz equation. The cell design is depicted in Fig. 1B. A lithium nitrate/potassium nitrate (LiNO3/KNO3) eutectic molten salt operates as the liquid electrolyte because of its good chemical stability and high conductivity (12), and a solid electrolyte [Li1.5Al0.5Ge1.5(PO4)3 (LAGP)] membrane at the Li anode inhibits the crossover of soluble products (fig. S2). A noncarbonaceous composite cathode composed of nickel (Ni) nanoparticles coated in situ to form LixNiO2 supplies the vital electrocatalyst that reversibly catalyzes O–O bond cleavage and formation (fig. S3). Detailed material synthesis and characterization methods are provided in the supplementary materials.

Fig. 1 Thermodynamics and configuration of the Li-O2 cell.

(A) Gibbs reaction energy for formation of Li2O and Li2O2 as a function of temperature. The thermodynamic data were calculated according to the database of HSC chemistry version 5. (B) Configuration of the inorganic electrolyte Li-O2 cell and schematic illustration of Li2O formation during discharge.

Cells were sealed with oxygen and cycled between 2.6 and 3.5 V at an applied current of 0.1 mA cm−2 (Fig. 2). The inorganic electrolyte Li-O2 cell with a Ni-nitrate composite cathode exhibits a very high discharge capacity of 11 mA·hour cm−2 (Fig. 2A, red solid curve). This value is more than 20-fold higher than achieved in a cell using an aprotic organic electrolyte and a carbon electrode (0.5 mA·hour cm−2) (Fig. 2A, black dashed curve) and infinitely higher compared with a Ni electrode in an aprotic cell (Fig. 2A, red dashed line). After fully discharging the cell to 2.6 V, the x-ray diffraction (XRD) pattern of the composite cathode shows two peaks at 34° and 56° assigned to the (111) and (022) reflections of Li2O [Joint Committee on Powder Diffraction Standards (JCPDS) 01-077-2144] (Fig. 2B). A Raman band definitive of Li2O at 523 cm−1 further supports formation of the oxide (Fig. 2C). The scanning electron microscopy (SEM) image in Fig. 2E reveals that the discharged cathode is covered with large ~5-μm octahedral crystals, a morphology characteristic of the Li2O antifluorite structure. Because the solubility of Li2O in molten nitrate is 27 mM at 150°C (fig. S5), we speculate that solution-mediated Li2O transport—as reported in aprotic Li-O2 and Na-O2 cells (13, 14)—is responsible for crystal nucleation and growth. A high initial coulombic efficiency (CE) of 96% was achieved after recharging the cell to 3.5 V, accompanied by a very low polarization of 0.2 V. The disappearance of Li2O in the XRD pattern (Fig. 2B) and Raman spectrum (Fig. 2C) of the recharged cathode indicates Li2O is fully removed by oxidation. Furthermore, the charged cathode is bare (Fig. 2F, SEM), identical to before discharge (Fig. 2D), indicating excellent electrochemical reversibility. There is a gravimetric energy penalty owing to the higher mass of Ni compared with that of carbon, but this can be improved by optimizing the cathode microstructure and mass. By contrast, the inorganic electrolyte Li-O2 cell by using a cathode composed of Super P carbon—run at the same temperature of 150°C—shows about half the discharge capacity (6.5 mA·hour cm−2) (Fig. 2A, black solid line). Li2O2 is identified as the main discharge product in the carbon electrode cell according to XRD and Raman analysis (Fig. 2, B and C, black curves) and is present as hexagonal-shaped Li2O2 crystal agglomerates ~8 μm in dimension (Fig. 2G). We assume that the quinone groups on carbon are responsible for peroxide formation, as reported (15). Crystalline Li2CO3 and amorphous Li2O were also identified with XRD and x-ray photoelectron spectroscopy (XPS) analysis, respectively (fig. S4). These by-products are attributed to carbon corrosion by Li2O2 as reported in aprotic Li-O2 cells (Eq. 3) (16) and to the reaction of Li2O2 with nitrite (Eqs. 4 and 5). These electrochemically inert by-products lower the cell rechargeability, evidenced by a low CE of only 75% on the first cycle. 2Li2O2 + C → Li2CO3 + Li2O(3)Li2O2 + C + LiNO3 → Li2CO3 + LiNO2(4)Li2O2 + LiNO2 → Li2O + LiNO3(5)To further examine the electrochemical reversibility, quantitative compositional changes of the cathodic products at the 1st and 10th cycle were analyzed (fig. S6). Cells with either carbon or the Ni-nitrate composite cathode were discharged to 2 mA·hour cm−2 and recharged to 3.5 V. Li2O is the product in the composite cathode on discharge (36.1 μmol, as quantified by means of acid-base titration) (Fig. 3A and supplementary materials, section S6-3), along with a tiny fraction of Li2O2 (1.1 μmol). The total amount of Li2O and Li2O2 (37.2 μmol) is nearly identical to the theoretical value of 37.3 μmol, assuming a 2e transfer per mole of products (Eqs. 6 and 7) Embedded Image(6)2Li+ + O2 + 2e → Li2O2(7)After recharging the cell to 3.5 V, the amount of Li2O on the cathode is reduced to 2.6 μmol, whereas no Li2O2 is observed. The residual Li2O likely arises from its low solubility in the molten nitrate electrolyte and crosses over to the electrically insulating LAGP membrane, rendering it electrochemically inaccessible. There are no notable changes in the product quantity at the 10th cycle, indicating a reversible cathodic reaction. By contrast, the Li-O2 cell using a carbon cathode presents poor chemical reversibility (Fig. 3B). On discharge, in addition to the major product (Li2O2; 22.1 μmol), Li2CO3 (4.37 μmol) and Li2O (12.4 μmol) also form. After recharging the carbon cathode to 3.5 V, Li2O2 is fully oxidized, whereas Li2CO3 and Li2O remain. On the basis of mass spectrometry analysis of model cathodes, formation of Li2CO3 and Li2O is likely electrochemically irreversible below 4 V in the absence of an oxygen evolution reaction (OER) catalyst (fig. S7). These by-products accumulate during cycling, leading to cathode passivation (fig. S8). Online electrochemical mass spectrometry (OEMS) monitoring of the gaseous products formed during the 1st and 10th charge showed no other signals aside from O2 (such as CO2, NO, and H2O) below 3.5 V. The first charge profile of the Li-O2 cell with the Ni-nitrate cathode exhibits two plateaus at 3.0 and 3.3 V (Fig. 3C). The rate of O2 evolution on either plateau is exactly equal to the theoretical value based on 4 e/O2, indicating the electrochemical oxidation of Li2O (Eq. 8)2Li2O → 4Li+ + O2 + 4e(8)We speculate that the first plateau at 3.0 V is due to the oxidation of Li2O crystallites that are deposited near or on the Ni catalyst at the cathode, whereas the higher plateau at 3.3 V (whose disappearance at the 10th charge is not well understood at present) corresponds to the oxidation of Li2O deposited on LAGP. In the latter case, the soluble oxide must diffuse back to the Ni particles for OER, creating a kinetic overpotential as previously noted in Na-O2 cells (17). Moreover, rechargeability is improved on cycling as indicated by more oxygen evolution and a more prolonged charge plateau at 3.0 V, with a CE of 100%.

Fig. 2 Characteristics of Li-O2 cells using carbon and Ni-nitrate composite cathodes.

(A) First discharge and charge curves of Li-O2 cells with a carbon cathode (black) and a Ni-based cathode (red). The cells using aprotic electrolyte [0.5 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) in tetraethylene glycol dimethyl ether (TEGDME)] were examined at 25°C (dashed lines), whereas the cells using the molten nitrate electrolyte were measured at 150°C (solid lines). The current density is 0.1 mA cm−2, and cutoff voltages are 2.6 and 3.5 V. (B) XRD patterns, (C) Raman spectra, and (D to G) SEM images of (D) pristine Ni cathode, (E) Ni cathode discharged to 2.6 V then (F) recharged to 3.5 V, and (G) carbon cathode discharged to 2.6 V. Scale bars, 2 μm.

Fig. 3 High reversibility of molten salt electrolyte Li-O2 cells using a Ni-nitrate composite cathode.

(A and B) Quantitative analysis of cathodic products. Shown are Li2O (red), Li2O2 (yellow), and Li2CO3 (gray), at the 1st and 10th cycle using (A) Ni-nitrate composite cathode and (B) carbon cathode. The discharge limit is 2 mA∙hour cm−2, and the charge limit is 3.5 V. The current density is 0.1 mA cm–2. (C and D) Online mass spectrometry analysis of gaseous oxygen evolution upon (C) 1st charge and (D) 10th charge using a carbon cathode (black) and a Ni-based composite cathode (red). Cells were predischarged to 2 mA∙hour cm–2 at 0.1 mA cm–2. The charging rate was 0.2 mA cm–2 in order to enhance the O2 mass spectrometry signal.

By contrast, the carbon cathode Li-O2 cell exhibits a voltage plateau at ~2.9 V on the 1st charge, where OEMS shows that oxygen is initially evolved at a rate close to theoretical for Li2O2 oxidation based on 2e/O2 (Eq. 9). The voltage is close to expected at 150°C (2.84 V)Li2O2 → 2Li+ + O2 + 2e(9)Subsequently, the rate of oxygen evolution greatly diminishes and is followed by a false “peak” as a consequence of the voltage cutoff (3.5 V). This behavior is indicative of substantial parasitic reactions that hinder the OER process (18). Passivation of the carbon cathode results in an increase of cell overpotential at the 10th charge (Fig. 3B).

The Ni-nitrate cathode Li-O2 cell was cycled with a limited capacity of 0.5 mA·hour cm–2 at 0.2 mA cm–2 for 150 cycles (Fig. 4A), with low overpotential. At a lower current density of 0.1 mA cm−2 (fig. S9), the polarization is further reduced (0.16 V; comparable with Fig. 2A). Shallow cycling was necessary to limit the amount of Li transfer at the negative electrode in order to provide proof of concept of the rechargeability at the cathode. The cell exhibits two discharge plateaus, the first at ~2.9 V corresponding to a very low capacity (~0.1 mA·hour cm−2), followed by a much longer discharge plateau at ~2.7 V. The Ni-nitrate cathode cell at a much deeper state of discharge of 11 mA·hour cm−2 exhibits a similar electrochemical profile, although the first short plateau is masked by the large discharge capacity (Fig. 2A). We ascribe the first plateau in Fig. 4A to initial formation of the Li2O species in solution, which has a different formation energy than that of solid Li2O. Phase field simulations of Li-O2 electrochemistry report that a higher potential results from nucleation of a supersaturated solution of solvated species before growth of solid products (19). The longer plateau at 2.7 V corresponds to oxygen reduction at the surface of the catalyst to form Li2O. Upon charge, the cell also exhibits two charge plateaus at ~3.0 and ~3.3 V, which is in agreement with the OEMS measurement (Fig. 3D) and the deep-cycled Ni-nitrate cathode in Fig. 2A. The CE of the cell increases rapidly from 80 to 100% over four cycles and is subsequently stable (Fig. 4B). This performance is superior to that of aprotic organic electrolyte Li-O2 cells and their cousins (Na-O2 and K-O2) (14, 20), where parasitic reactions cause poor cycling performance: A monolayer of carbonate created at the C-Li2O2 interface causes an increase of interfacial resistance (16), and decomposition of organic electrolyte is triggered by a superoxide attack that forms carbon-centered radicals (21). By using inorganic electrolytes, electrolyte degradation is avoided. A layer of lithiated nickel (III) oxide and Ni2O3 on the surface of the Ni particles (because of the oxidation of Ni by LiNO3) (fig. S3) provides a protective passivation layer; meanwhile, lithiation improves the electronic conductivity of the oxide layer (22). Hence, the chemically stable inorganic electrolyte and cathode play critical and synergistic roles. Although we used a Li+-ion conducting interface between the Li and LAGP in order to prevent direct reduction of the LAGP (23), degradation of the LAGP membrane over 150 cycles—arising from its probable poor stability in molten nitrate and some localized reduction (fig. S10, A and B)—led to increased impedance from the membrane (fig. S10C). This resulted in an increase of the overpotential by 0.15 V over 150 cycles. More effort is needed to address the challenges on the negative electrode side, such as the development of robust polymer-inorganic membranes.

Fig. 4 Cycling performance of a molten salt electrolyte Li-O2 cell with a Ni-nitrate composite cathode.

(A) Discharge and charge curves over 150 cycles at a constant current of 0.2 mA cm–2. (B) The corresponding changes of discharge capacity (red), charge capacity (black), and CE (blue) during cycling. The CE ultimately reaches 100.8% because a 3.5 V cutoff was chosen as the charge limit, whereas O2 evolution stops at a charge voltage of 3.48 V (Fig. 3D); the slight additional charge capacity can be ascribed to a trace of nitrate decomposition.

To explain the mechanism, we propose a peroxide-mediated ORR pathway, illustrated in Fig. 5A and outlined below.

Fig. 5 Mechanistic studies of the oxygen reduction reaction over the Ni-based composite catalyst.

(A) Schematic illustration of the pathway of the oxygen reduction reaction. (B and C) Effects of (B) operating temperature and (C) discharge rates on the composition of oxygen reduction products, Li2O (red) and Li2O2 (yellow). A discharge rate of 0.1 mA cm–2 was applied to study the temperature effects, and 150°C was chosen to investigate the effects of discharge rate. All cells were discharged to 2 mA·hour cm−2 before quantitative analysis. (D) Oxygen mass spectrometry signal response on heating a mixture composed of commercial Li2O2 powder with Ni-nitrate composite cathode (red), Ni (blue), and molten nitrate (black) at 150°C, respectively. (E) Their corresponding XRD patterns were obtained after heating the mixtures for 1 week.

Diffusion: Upon discharge, oxygen adsorbs on the surface of the cathode (Eq. 10)O2 → O2,ad(10)Reduction: Oxygen is electrochemically reduced to form lithium peroxide (Li2O2,ad), via a two electron transfer, on the surface of the Ni/LixNiO2 catalyst (Eq. 11)O2,ad + 2e + Li+ → Li2O2,ad(11)Desorption: A small amount of Li2O2,ad slowly desorbs from the catalyst surface, governed by its low solubility and diffusibility in the molten nitrate electrolyte (Eq. 12)Li2O2,ad → Li2O2,electrolyte(12)Disproportionation: The major remaining Li2O2 is converted to Li2O by the catalyst through disproportionation (Eq. 13)Embedded Image(13)Transport: Once formed, Li2O is soluble in the electrolyte (Eq. 14), and on supersaturation formation of Li2O nuclei triggers nucleation and growth to result in micrometer-sized Li2O crystals following (24)Li2O,ad → Li2O,electrolyte(14)Experimental and computational studies suggest that the ORR pathway via peroxide is operative over a variety of metal catalysts (such as platinum, mercury, and silver) (25, 26). We quantified the amount of Li2O2 formed at different discharge conditions after discharge to 2 mA·hour cm−2. As shown in Fig. 5, B and C, the amount of Li2O2 increases from 0 μmol at 0.05 mA cm−2 to 4 μmol at 0.2 mA cm−2, whereas it decreases from 3.5 μmol at 135°C to 0 μmol at 170°C. According to the proposed ORR pathway, the fast formation of Li2O2 at higher discharge rates likely results in some Li2O2 remaining owing to relatively slow disproportionation (rate-determining step). However, the elevated temperature accelerates disproportionation, rapidly converting Li2O2 to Li2O. Evidence confirming the catalytic disproportionation of Li2O2 is shown in Fig. 5, D and E. When a mixture composed of commercial Li2O2 powder and the Ni-nitrate cathode was heated at 150°C, oxygen evolution was detected with mass spectrometry, accompanied by diffraction peaks of Li2O in the XRD pattern of the mixture. Neither features are observed in the absence of either component.

Suntivich et al. have shown that high ORR activity for transition metal oxide catalysts primarily correlates to σ*-orbital (eg) occupation and the extent of transition-metal–oxygen covalency (27). Optimal activity correlates to an eg occupancy is close to unity. Consistent with this design principle, the surface LixNiO2 species on the Ni particle contains Ni3+ with an electron configuration of t2g6eg1 (28). Furthermore, the Ni3+/Ni2+ redox couple promotes charge transfer between surface cations and adsorbates. Both factors give rise to a high ORR activity of LixNiO2 (29). Although the thermodynamic driving force for the disproportionation of Li2O2 is very small (only –0.063 kJ mol−1 at 150°C), the removal of Li2O from the catalyst surface via solution-mediated transport will shift the equilibrium (Eq. 13) toward the right, exposing the active catalyst surface. The spontaneous disproportionation reaction on oxygen reduction dictates that OER must follow a different pathway, however. Indeed, no Li2O2 intermediate is observed when charging a Li-O2 cell by using a Li2O-prefilled cathode (fig. S11). We speculate that Li2O is solubilized and diffuses to the surface of LixNiO2 for electrocatalytic oxidation via a direct 4 e pathway. Suntivich et al. also concluded that the OER activity of metal oxides, similar to ORR, is dependent on the occupancy of 3d electron states with eg symmetry (30). Thus, although LixNiO2 is an effective ORR catalyst, it also reversibly catalyzes OER (31), leading to a low charge overpotential.

By tuning the operating temperature and using a single bifunctional ORR/OER catalyst, our Li-O2 battery overcomes the barriers of thermodynamics and kinetics, leading to the electrochemically reversible formation of Li2O instead of Li2O2. The in situ–generated LixNiO2 electrocatalyst—applicable to other metal oxygen electrochemistries—catalyzes both O–O cleavage to form Li2O on discharge, and the reverse process that releases oxygen upon charge in a 4e/O2 process with excellent CE and low polarization. The latter is aided by electrolyte-solubilized Li2O transfer. The fact that the Li-O cell is more reversible when Li2O is the product is a consequence of the less reactive chemical nature of oxide versus superoxide or peroxide. Moreover, the use of chemically stable inorganic electrolytes and a noncarbonaceous cathode circumvents the degradation of organic electrolyte and carbon corrosion, which form the main failure mechanisms for nonaqueous Li-O2 cells. The “Li2O” cell presented here is akin to both fuel cells and electrolyzers—which also operate on the basis of a 4 e electrocatalyzed reaction—in which Li2O replaces H2O, and the combination forms a simple reversible energy storage system. Whereas elevated operating temperatures can limit battery applications, commercialized sodium nickel chloride (ZEBRA) cells and Na-S batteries run at much higher temperatures (~325°C), and proton-exchange membrane (PEM) fuel cells are recently trending to temperatures between 120° and 180°C (32). More fundamentally though, our work directly addresses a number of issues associated with Li-O2 chemistry, showing that it is not intrinsically limited and that four-electron transfer from Li2O is possible, reversible, and operates with almost theoretical CE.

Supplementary Materials

Materials and Methods

Figs. S1 to S11

Table S1

References (3340)

References and Notes

Acknowledgments: The authors thank V. Goodfellow and R. Smith at the University of Waterloo Mass Spectrometry Facility for their scientific input in the gas chromatography–mass spectrometry measurements. We also thank S. H. Vajargah for performing the TEM measurements. Funding: Research was supported by the Natural Sciences and Engineering Council of Canada through their Discovery and Canada Research Chair programs (L.F.N.), and a doctoral scholarship to C.Y.K. Partial funding for this work (C.Y.K.) was also provided by the Joint Center for Energy Storage Research, an Energy Innovation Hub funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences. Author contributions: C.X. led the design of the study, and all the authors contributed to the implementation and writing of the manuscript; data collection and analysis were conducted by C.X. and C.Y.K. Competing interests: The authors have no competing interests. Data and materials availability: All data are available in the manuscript or in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article