Research Article

Cycling Li-O2 batteries via LiOH formation and decomposition

See allHide authors and affiliations

Science  30 Oct 2015:
Vol. 350, Issue 6260, pp. 530-533
DOI: 10.1126/science.aac7730

Solving the problems with Li-air batteries

Li-air batteries come as close as possible to the theoretical limits for energy density in a battery. By weight, this is roughly 10 times higher than conventional lithium-ion batteries and would be sufficient to power cars with a range comparable to those with gasoline engines. But engineering a Li-air battery has been a challenge. Liu et al. managed to overcome the remaining challenges: They were able to avoid electrode passivation, turn limited solvent stability into an advantage, eliminate the fatal problems caused by superoxides, achieve high power with negligible degradation, and even circumvent the problems of removing atmospheric water.

Science, this issue p. 530


The rechargeable aprotic lithium-air (Li-O2) battery is a promising potential technology for next-generation energy storage, but its practical realization still faces many challenges. In contrast to the standard Li-O2 cells, which cycle via the formation of Li2O2, we used a reduced graphene oxide electrode, the additive LiI, and the solvent dimethoxyethane to reversibly form and remove crystalline LiOH with particle sizes larger than 15 micrometers during discharge and charge. This leads to high specific capacities, excellent energy efficiency (93.2%) with a voltage gap of only 0.2 volt, and impressive rechargeability. The cells tolerate high concentrations of water, water being the dominant proton source for the LiOH; together with LiI, it has a decisive impact on the chemical nature of the discharge product and on battery performance.

Rechargeable nonaqueous lithium-air (Li-O2) batteries have attracted considerable interest over the past decade, because of their much higher theoretical specific energy than conventional Li ion batteries (13). A typical Li-O2 cell is composed of a Li metal negative electrode, a nonaqueous Li+ electrolyte, and a porous positive electrode. During discharge, O2 is reduced and combines with Li+ at the positive electrode, forming insoluble discharge products (typically Li2O2) that fill up the porous electrode (46). The porous electrode is not the active material but rather a conductive stable framework that hosts the reaction products, lighter electrode materials providing higher specific energies. During charge, the previously formed discharge products must be thoroughly removed to prevent the cell from suffocating after a few discharge/charge cycles, the electrode pores becoming rapidly clogged with discharge products and products from unwanted side reactions (713).

Several fundamental challenges still limit the practical realization of Li-O2 batteries (13). The first one concerns the reversible capacity (and thus energy density). This is determined by the pore volume of the porous electrode, which limits both the total quantity of the discharge products and how large the discharge product crystals can grow. The ultimate capacity—currently far from being reached—is, in theory, achieved in the extreme case in which large single crystals of the discharge product grow to occupy the full geometric volume of the positive electrode. The commonly used mesoporous Super P (SP)/Ketjen carbon electrodes have relatively small pore sizes and volumes, with their crystalline discharge products typically less than 2 μm in size (4, 5, 14); this limits the capacity to <5000 milli–ampere⋅hour per gram of carbon (mA⋅h/gc) (typically <1.5 mAh based on 1 mg of carbon and binder) (5, 710, 12, 13). In addition, uses of smaller pores tend to lead to pore clogging, hindering the diffusion of O2 and Li+ and causing high overpotentials during cycling. Second, severe side reactions can occur on cycling, involving the electrode materials, electrolyte, and intermediate as well as final discharge products (713). Major causes of these decomposition reactions include the superoxide ion that forms as an intermediate on reduction of oxygen, which readily attacks most electrolytes (79, 15), and the large overpotential on charge, often required to remove the insulating discharge products, which results in the oxidation of some of the cell components such as the host electrode (1013). Previous studies (1013) suggest that 3.5 V (versus Li/Li+) represents the maximum voltage that carbon-based electrodes can tolerate without significant side reactions. Third, the large hysteresis seen between charge and discharge (up to 2 V) (413), results in extremely low energy efficiencies, limiting the use of this battery in practical applications. Finally, the cells are very sensitive to moisture and carbon dioxide (1619): The more stable LiOH and Li carbonate phases are formed, which gradually accumulate in the cell, resulting in battery failure. Moisture and CO2 also have deleterious effects on the Li-metal anode (13).

A number of strategies have been proposed to reduce the voltage hysteresis, involving the use of electrocatalysts (2027), porous electrode structures (2830), and redox mediators (3135). Soluble redox mediators, such as tetrathiafuvalene (TTF) (32) and LiI (34), have been used to reduce the overpotential of the charge process, resulting in the overall voltage hysteresis dropping to around 0.5 V (34). Their operation relies on the electrochemical oxidation of the mediator, which itself then chemically decomposes the Li2O2. The charge voltage is thus tuned close to the redox potential of the mediator. For discharge, the ethyl viologen redox couple (31) has also been used to reduce O2 in the liquid electrolyte rather than on the solid electrode surface, again to help prevent rapid blocking of the solid electrode surface by Li2O2. We used the redox mediator LiI and report a Li-O2 battery with an extremely high efficiency, large capacity, and a very low overpotential. This battery cycles via LiOH formation, not Li2O2, and is able to tolerate large quantities of water. This current work directly addresses a number of critical issues associated with this battery technology.

A Li-O2 battery was prepared by using a Li metal anode, a 0.25 M lithium bis(trifluoromethyl) sulfonylimide/dimethoxyethane (LiTFSI/DME) electrolyte with 0.05 M LiI additive, and a variety of different electrode structures (fig. S1). Hierarchically macroporous reduced graphene oxide (rGO) electrodes (binder-free) are used because they are light, conductive, and have a large pore volume that can potentially lead to large capacities. Mesoporous SP carbon and mesoporous titanium carbide (TiC) (36) electrodes were studied for comparison. Cyclic voltammetry (CV) measurements confirmed that rGO, SP, and TiC electrodes all exhibit good electrochemical stability within a voltage window of 2.4 to 3.5 V in a LiTFSI/DME electrolyte and can be used to reversibly cycle LiI (I3+2e↔3I) (37) (fig. S2).

In the absence of LiI, cells using either mesoporous TiC or macroporous rGO showed much smaller overpotentials during charge, in comparison to that obtained with the SP electrode (Fig. 1A). These decreases in overpotential are tentatively ascribed to the higher electrocatalytic activity of TiC (38) and the faster diffusion of Li+ and solvated O2 within the micrometer-sized pores of the rGO electrodes (fig. S1). The addition of LiI to the SP electrode led to a noticeable drop in the overpotential over that seen with SP only, suggesting that the polarization during charge is largely caused by the insulating nature of the discharge products. The charge voltage profile is not, however, flat, but gradually increases as the charge proceeds to above 3.5 V. In contrast, when LiI is used with hierarchically macroporous rGO electrodes, a remarkably flat process is observed at 2.95 V, representing a further reduction in overpotential by ~0.5 V over that seen for SP. This reduction is ascribed, at least in part, to the interconnecting macroporous network of rGO, which allows for much more efficient mediator diffusion than in the mesoporous SP electrode, even when the macropores are filled with insoluble discharged products.

Fig. 1 Electrochemical profiles of cells with different electrode/electrolyte combinations.

(A) Discharge-charge curves for Li-O2 cells using mesoporous SP and TiC, and macroporous rGO electrodes, with capacities limited to 500 mA⋅h/g (based on the mass of carbon or TiC) and a 0.25 M LiTFSI/DME electrolyte. For SP and rGO electrodes, 0.05 M LiI was added to the LiTFSI/DME electrolyte in a second set of electrodes (purple and red curves). All cells in (A) were cycled at 0.02 mA/cm2. The horizontal dashed line represents the position (2.96 V) of the thermodynamic voltage of a Li-O2 cell. (B) Galvanostatic charge-discharge curves of cells containing 0.05 M LiI and 0.25 M LiTFSI, cycled under an Ar atmosphere with different electrode/electrolyte solvent combinations with a current of 0.2 mA/cm2. The crossing points (with appropriate voltages labeled) of the charge/discharge curves indicate the positions of the redox potential of I/I3 in the specific electrode/electrolyte system. A direct comparison of capacities between LiI in Ar and Li-O2 cells is given in fig. S3.

The observation that the LiI/DME Li-O2 cell charges at 2.95 V is of note, because it is slightly below the thermodynamic voltage of 2.96 V of the Li-O2 reaction. During charge, the redox mediator is thought to be first electrochemically oxidized on the electrode (32); this oxidized form then helps to chemically decompose the discharge product. The charge voltage then reflects the redox potential (versus Li/Li+) of the I/I3 redox mediator in the electrode/electrolyte system rather than the redox potential associated with the oxidation of the solid discharge product. A low redox potential of a mediator is important for the long-term stability of the Li-O2 cell.

To investigate factors affecting the redox potential, LiI was cycled galvanostatically in an Ar atmosphere with different electrode/electrolyte combinations (Fig. 1B). The electrolyte solvent has a larger effect on the redox potential of the I/I3 couple than the electrode material, with the DME electrolyte consistently exhibiting lower charge voltages than TEGDME (tetraethylene glycol dimethyl ether) for all three electrodes. In addition, the voltage gaps between the charge and discharge plateaus are smaller for DME than for TEGDME electrolytes, which is consistent with the smaller voltage separations seen between the redox peaks in their respective CV curves (fig. S2). The discharge capacity is always smaller than the previous charge capacity for all cells (Fig. 1B), indicating that some mediators, after being oxidized, have diffused into the bulk electrolyte. This observation is more prominent with DME, suggesting faster mediator diffusion in DME than in TEGDME.

The discharge overpotential for rGO-based Li-O2 cells also decreases by 0.15 V (marked by arrows in Fig. 1A), from 2.6 (SP/TiC) to 2.75 V (rGO), regardless of the use of LiI. Overall, the voltage gap becomes only 0.2 V (indicated by arrows), representing an ultrahigh energy efficiency of 93.2%.

X-ray diffraction (XRD) patterns (Fig. 2A) for the rGO electrodes cycled with LiI show that LiOH is the only observed crystalline discharge product; LiOH is then removed after a full charge. This is confirmed in the solid-state nuclear magnetic resonance (ssNMR) measurements (Fig. 2B), where a single resonance due to LiOH is observed at –1.5 ppm and at 1.0 ppm in the 1H and 7Li magic angle spinning ssNMR spectra (13, 39), respectively (further corroborated by the 7Li static NMR spectrum in fig. S4). After charge, the 1H and 7Li LiOH resonances are no longer visible. We emphasize that without added LiI, the predominant discharge product for rGO electrodes is Li2O2 (fig. S5), the chemistry radically changing when 0.05 M LiI is added to the DME electrolyte.

Fig. 2 Characterization of pristine, discharged and charged electrodes.

XRD patterns (A) and 1H and 7Li ssNMR spectra (B) comparing a pristine rGO electrode to electrodes at the end of discharge and charge in a 0.05 M LiI/0.25 M LiTFSI/DME electrolyte (the electrochemistry of the cells is in fig. S10). The spectra are scaled according to the mass of the pristine electrode and the number of scans. Asterisks in (A) represent diffraction peaks from the stainless steel mesh. 1H resonances of proton-containing functional groups in the pristine rGO electrode are not visible in the 1H ssNMR spectrum in (B), because they are very weak in comparison to the LiOH signal. The weaker signals at 3.5 and 0.7 ppm are due to DME and grease/background impurity signals, respectively. Optical (C) and SEM (D) images of pristine, fully discharged and charged rGO electrodes obtained with a 0.05 M LiI/0.25 M LiTFSI/DME electrolyte in the first cycle. The scale bars are all 5 mm and 20 μm in the optical and SEM images, respectively.

Figure 2, C and D, shows optical and scanning electron microscopy (SEM) images of electrodes during the first cycle. After discharge, the electrode surface is completely covered by LiOH agglomerates, tens of microns in size, and the color of the electrode has changed from black to white. When the interior of the electrode was investigated, many crystalline “flowerlike” agglomerated LiOH particles were observed within the graphene macropores. Although these particles are more than 15 μm in diameter (fig. S6), much bigger than the Li2O2 toroids (fig. S5), they are in fact formed from thin-sheet primary building blocks, resulting in a more open (porous) structure. The large LiOH agglomerates efficiently fill up the pore volume available in the hierarchical macroporous electrode, leading to much larger capacities (fig. S6). When TEGDME was used as the electrolyte solvent, the discharge product, although still LiOH, now forms a thin film on the rGO electrode surface (fig. S7). After charge in DME, the hierarchically macroporous structure reappeared and the electrode turned black again (Fig. 2C). Higher-magnification SEM images revealed very small traces of residual LiOH on the electrode surface (fig. S8). We found that the reversible formation and removal of LiOH with the LiI mediator are not restricted to rGO electrodes, because mesoporous SP electrodes show similar results (fig. S9) but with larger overpotentials and lower capacities.

Kang and co-workers (34) previously reported a highly rechargeable Li-O2 cell using carbon nanotubes and the mediator LiI (0.05 M) in a TEGDME-based electrolyte, ascribing the electrochemistry to the formation and decomposition of Li2O2. Sun et al. (35) recently pointed out, however, that LiOH, rather than Li2O2, was the dominant discharge product when 0.05 M LiI mediator was added to the TEGDME electrolyte; LiOH was still present in the SP electrode used in their study after charge, and they suggested that LiOH could not be decomposed by the mediator. In our work, we have seen clear evidence that the discharge product is overwhelmingly LiOH and that it can be removed at low potentials of around 3 V.

In a redox-mediated Li-O2 system, the effective removal of the insulating discharge products is affected by a few factors: (i) availability of bare electrode surfaces to oxidize the mediator during charge, (ii) whether the discharge product is uniformly distributed throughout the electrode, and (iii) efficient diffusion of the oxidized mediator from electrode surfaces (that supply and remove electrons) to the discharge product. TEGDME, being a more viscous solvent than DME, will lead to more sluggish I3 and O2 diffusion. When it is used with mesoporous (rather than macroporous) electrodes, the discharge product tends to concentrate on the electrode surface facing the gaseous O2 reservoir, with its concentration dropping noticeably in the electrode interior [the reaction zone problem for Li-air batteries (40)]. The much more soluble LiI, however, is likely to be uniformly oxidized across the whole thickness of the electrode during charge. Consequently, for equal capacities for discharge and charge, the oxidized mediator (I3) formed during charge may remain in excess in the electrode interior regions, where the discharge product is scarce; similarly, discharge product may be left unreacted at regions close to the O2/electrolyte interface, where the discharge product is abundant. The remaining mediator in the oxidized form (I3) will then be reduced during the next discharge, resulting in a voltage plateau at its redox potential in addition to that due to oxygen reduction. This unbalanced distribution of the mediator LiI and the discharge product LiOH across the thickness of the mesoporous electrode may be a cause of the unreacted LiOH and iodine-dominated electrochemistry observed in the work by Sun and co-workers (33). Furthermore, the thin-film morphology of the discharge product formed in the TEGDME-based electrolyte [fig. S7 and (35)] effectively passivates the electrode surface. As a result, triiodide anions may first form on bare electrode surfaces that are distant from the discharge products. They then need to diffuse to the passivated regions to remove LiOH, reducing the efficiency of LiOH removal and providing another explanation for the observed residual LiOH. We used a macroporous rGO electrode and DME, which provides faster mediator and O2 diffusion, and higher LiO2 solubility, leading to a more uniform Li-O2 reaction during discharge and larger reversible capacities.

Figure 3 shows the electrochemical performance of the Li-O2 battery. When limiting the specific capacity to 1000, 5000, and 8000 mA⋅h/gc, the cells show no capacity fade, with little increase in voltage polarization after 2000, 300, and 100 cycles, respectively (Fig. 3, A to C). Higher capacities >20,000 mA⋅h/gc have also been demonstrated (figs. S5 and S11). When cycled at 1 A/gc (Fig. 3, A and C), the voltage gap is only ~0.2 V; at higher rates, the gaps widen (Fig. 3D), increasing to 0.7 V at 8 A/gc. Furthermore, at this higher rate, the cell is polarized each cycle (fig. S11), and after 40 cycles the electrode surface is covered by a large number of particles (with morphologies unlike those of LiOH observed at lower currents), which do not seem to be readily removed during charge at these voltages. At these higher overpotentials, more substantial parasitic reactions probably occur, rapidly polarizing the cell by increasing its resistance and impeding the electron transfer across the electrode/electrolyte interface. A narrower operating electrochemical window within 2.96 ± 0.5 V is key for the prolonged stability of the rGO electrodes.

Fig. 3 Electrochemical performance of the Li-O2 battery.

Discharge/charge curves for Li-O2 batteries using rGO electrodes and a 0.05 M LiI/0.25 M LiTFSI/DME electrolyte with capacity limits of 1000 mA⋅h/gc (A), 5000 mA⋅h/gc (B), and 8000 mA⋅h/gc (C), as a function of rate (D); three cycles were performed for each rate in (D). The cell cycle rate is based on the mass of rGO; i.e., 5 A/gc is equivalent to 0.1 mA/cm2.

The sensitivity of the cell to water was explored by either deliberately adding ~45,000 parts per million (ppm) of H2O (37 mg per 783 mg of DME) to the electrolyte or cycling cells under a humid O2 atmosphere. In both cases, no appreciable change in the electrochemical profile was observed (fig. S12), compared to that using a nominally dry electrolyte. Furthermore, the added water was found to promote the growth of even larger LiOH crystals >30 μm (fig. S13).

Although a certain level of scattering in the total capacity is observed, probably due to variations in the electrode structure, the cell capacity is typically within 25,000 to 40,000 mA⋅h/gc (i.e., 2.5 to 4.0 mA⋅h) range for an rGO electrode of 0.1 mg and 200 μm thick. After discharge, the weight of an electrode removed from the stainless steel mesh was about 1.5 mg (2.7 V, 3.2 mA⋅h), giving a specific energy of 5760 W⋅h/kg (see section 13 of the supplementary materials).

The mechanism of O2 reduction in aprotic Li-O2 batteries has been extensively discussed. Many authors (5, 41, 42) have shown that the ability of an electrolyte to solvate O2 (characterized by the Guttman donor number, DN) is important in governing the discharge reaction mechanism. Higher LiO2 solubility favors a solution precipitation mechanism leading to large toroidal Li2O2 crystals and thus higher discharge capacity; lower LiO2 solubility tends to drive a surface mechanism where Li2O2 forms thin films on the electrode surface and a lower capacity. Because of its intermediate DN, solution precipitation and surface reduction mechanisms can occur simultaneously in DME (41).

With added LiI, although LiOH rather than Li2O2 is the prevailing discharge product, many parallel phenomena are observed: The similar discharge voltages (2.75 V, Fig. 1A) observed with and without the added LiI suggest that the first step on discharge is an electrochemical reaction, where O2 is reduced on the electrode surface to form LiO2. It is unlikely that O2 is reduced to O22– via two-electron electrochemical steps or even dissociatively reduced to O2­– (or LiOH) via four-electron electrochemical steps. Subsequent conversion of LiO2 to LiOH is proposed to be a chemical process that occurs via a solution mechanism. Strong support for a solution mechanism comes from the observation that LiOH grows on both the electrode and the insulating glass fiber separators (fig. S6), the latter not being electrically connected to the current collector. This process must involve the iodide redox mediator, because in its absence Li2O2 is formed, even in cells with high moisture contents.

A key question is the origin of the H+ in the formed LiOH, potential sources being the DME electrolyte, surface functional groups of rGO, and moisture in the cell (fig. S14). To investigate this, NMR measurements were conducted on two sets of discharged electrodes from cells that were prepared using either deuterated DME or deuterated water (figs. S15 to S18). When deuterated DME was used, only a very small quantity of LiOD was detected in the 2H NMR spectra (figs. S15, S16, and S18), the dominant product being LiOH. In contrast, when D2O was added to the protonated DME electrolyte, a significant amount of LiOD was observed (fig. S17), LiOH only being a minor signal in the corresponding 1H NMR spectrum. In summary, these experiments clearly demonstrate that (i) although DME can be a potential H+ source for LiOH, it is by no means the dominant one; (ii) the added water preferentially supplies H+ to form LiOH, substantially minimizing DME decomposition (figs. S17 and S18); and (iii) even our nominally dry cells contain sufficient water to promote LiOH formation. This latter statement is consistent with the formation of large toroidal Li2O2 particles in the absence of LiI (fig. S5); earlier work shows that this requires water levels of >500 ppm (42).

1H NMR spectroscopy was used to quantify the number of moles of LiOH formed on discharge (with added LiI mediator). The (molar) ratio of electrons consumed on discharge to LiOH formed was close to 1:1 within the errors of the measurements (fig. S19), supporting the proposal that LiOH is formed in stoichiometric quantities (i.e., is not a minor side product).

During charge, an iodine-mediated LiOH decomposition reaction is observed at ~3 V, a clear distinction of our work from that of others (34, 35). Given that this charge voltage overlaps with that for I/I3 itself (measured with Ar gas in Fig. 1B), the first step should involve the direct electrochemical oxidation of I to I3. We suggest that the next step involves the chemical oxidation of LiOH by I3 to form O2 and H2O. To confirm this hypothesis, a pre-discharged Li-O2 cell (with LiI) was charged in a pure Ar atmosphere, and the gas atmosphere in the cell after charge was monitored by mass spectrometry. A clear O2 signal (fig. S21) was detected, consistent with the proposal that LiOH decomposition occurs via an O2 evolution reaction. The discharge and charge reactions are schematically represented in Fig. 4. We stress, however, that the equilibria that occur in the presence of oxygen, water, and iodine are complex, often involving a series of polyanions (including IO and its protonated form); further mechanistic studies are required to understand the role of these complex equilibria in the redox processes.

Fig. 4 Schematic mechanisms for the formation and removal of LiOH in iodide redox-mediated Li-O2 cells in the presence of water.

The electron/LiOH molar ratios during discharge and charge are both equal to 1.

By using an rGO electrode and the redox mediator LiI, in a DME-based electrolyte, we have demonstrated a highly efficient, rechargeable Li-O2 battery with extremely large capacities. Its operation involves the reversible formation and removal of LiOH crystals. The role of the additive, LiI, is threefold. First, it operates as a redox mediator whose redox potential can be tuned by using different electrolyte solvents and electrode structures; this redox potential guides the charge voltage and thus affects the cycling stability of the cell. Second, LiI, together with another additive H2O, affects the chemical nature and physical morphology of the discharge products, inducing the growth of large LiOH crystals that efficiently take up the pore volume of macroporous rGO electrodes; this is the origin of the observed large capacity. Finally, it enables a chemical pathway toremove LiOH at low overpotentials. Consequently, the cell becomes insensitive to relatively high levels of water contamination. The hierarchically macroporous rGO electrode is also an important factor for the high efficiency and capacity. Not only does the rGO framework provide efficient diffusion pathways for all redox active species in the electrolyte and hence, a reduced cell polarization and flatter electrochemical profile, it also permits the growth of LiOH crystals of tens of micrometers in size, resulting in a capacity that is much closer to the theoretical value of Li-O2 batteries. These desirable features were not observed for Li-O2 cells with mesoporous SP electrodes, even when the same electrolyte was used. The combination of electrolyte additives, the porous electrode structure, and the electrolyte solvent, synergistically, not only determines the chemical nature of the discharge product but also governs the physical size and morphology of it, playing a decisive factor in the capacity and rechargeability of the resulting Li-O2 battery. In a broader sense, this work can inspire ways to remove other stable, detrimental chemicals, such as Li2CO3, which is relevant to cycling Li-air batteries in real practical conditions.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S23

References (4347)


  1. ACKNOWLEDGMENTS: This work was partially supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Vehicle Technologies of the U.S. Department of Energy, under contract no. DE-AC02-05CH11231, under the Batteries for Advanced Transportation Technologies Program subcontract 7057154 (W.Y., M.J.L., P.M.B.); the Engineering and Physical Sciences Research Council (T.L.); Johnson Matthey (A.M.); Marie Curie Actions (P.M.B. and M.L.); and the EU Graphene Flagship under contract no. 604391 (G.K.). M. T. L. Casford and members of the Cambridge Graphene Centre are thanked for many useful discussions.
View Abstract

Stay Connected to Science

Navigate This Article