The water catalysis at oxygen cathodes of lithium–oxygen cells

Lithium–oxygen cells have attracted extensive interests due to their high theoretical energy densities. The main challenges are the low round-trip efficiency and cycling instability over long time. However, even in the state-of-the-art lithium–oxygen cells the charge potentials are as high as 3.5 V that are higher by 0.70 V than the discharge potentials. Here we report a reaction mechanism at an oxygen cathode, ruthenium and manganese dioxide nanoparticles supported on carbon black Super P by applying a trace amount of water in electrolytes to catalyse the cathode reactions of lithium–oxygen cells during discharge and charge. This can significantly reduce the charge overpotential to 0.21 V, and results in a small discharge/charge potential gap of 0.32 V and superior cycling stability of 200 cycles. The overall reaction scheme will alleviate side reactions involving carbon and electrolytes, and shed light on the construction of practical, rechargeable lithium–oxygen cells.

The success of deploying Li-O 2 cells in the future will be crucially dependent on how small the overpotentials at the cathode side can be reduced and how many cycles they can work reversibly [15][16][17] . Currently, two strategies have been widely applied for these issues. One is to use carbon-free or carbon-alternative cathodes, like nanoporous gold 10 , TiC 11 and conductive metal oxide supported Ru 12,13 , which can circumvent side reactions involving carbon. The charge overpotentials ensuring the decomposition of Li 2 O 2 are reduced to B0.54 V, corresponding to charge potentials of B3.50 V (refs [10][11][12][13]. This direct electrochemical oxidation of Li 2 O 2 was revealed to involve multiple processes, in which the initial delithiation occurring at an overpotential of 0.44 V, possibly suggests the theoretical limit 18,19 . Alternatively, redox mediators have been introduced to chemically oxidize Li 2 O 2 and the apparent charge potentials are typically determined by the mediators. For instance, tetrathiafulvalene and LiI reduce the charge potential to B3.50 V, equal to a charge overpotential of 0.54 V (refs 20-23). In principle, Li 2 O 2 can be oxidized by mediators possessing higher redox potentials than its equilibrium potential; however, practical mediators are very limited. A more efficient and compatible one for further potential reduction is still lacking. By either the direct or indirect (mediators) Li 2 O 2 oxidation strategy, it will be a great challenge to reduce the charge potentials to feasible values. Exploring new reaction mechanisms at the oxygen cathodes to significantly improve the round-trip efficiency and cycling stability is essential and urgently necessary for the development of Li-O 2 cells.
Although water in electrolytes has been found to affect the morphologies of discharge products and increase the discharge capacity [24][25][26] , its presence in electrolytes or O 2 atmosphere resulted in rapid charge potential increase and hence cell death after several cycles [26][27][28] . In addition to the main discharge product Li 2 O 2 , a byproduct LiOH was also detected 27 . We found that the decomposition of LiOH is strongly related to the applied catalysts, such as Ru nanoparticles supported on Super P (Ru/SP). To reduce the charge overpotentials of Li-O 2 cells, a trace amount of water in electrolytes and electrolytic mangnese oxide (EMD, g-MnO 2 ) nanoparticles are both utilized to favour the transformation of the discharge product from Li 2 O 2 to LiOH and its following decomposition on charging. The MnO 2 incorporated in Ru/SP (Ru/MnO 2 /SP) also allows for the water regeneration at the cathode during the discharge/charge cycles. This enables the Li-O 2 cell to operate with a small discharge/ charge potential gap and superior cycling stability. A reaction mechanism by converting Li 2 O 2 to LiOH in the presence of both water and Ru/MnO 2 /SP is proposed.

Results
Low overpotentials at Ru/MnO 2 /SP. The Li-O 2 cells were constructed with Ru/MnO 2 /SP pressed onto a carbon paper as cathodes, 0.5 M LiClO 4 in DMSO containing 120 p.p.m. of H 2 O as electrolytes and LiFePO 4 in replacement of Li anodes. The LiFePO 4 is not a practical anode, but is able to avoid the reaction of Li metal and the trace amount of H 2 O and any contamination from formation of the solid electrolyte interface layer on the Li anode 20 . It has a stable potential of B3.45 V vs Li þ /Li regardless of the state of charge and favours the investigation on the discharge/charge behaviour in the presence of H 2 O and the underlying reaction mechanism specifically at the oxygen cathode. The discharge/charge potentials at the oxygen cathode of the Li-O 2 cells are converted to against Li for discussion.
The carbon paper has been demonstrated to have negligible contribution to cell performance (Supplementary Fig. 1; Supplementary Note 1). The Li-O 2 cell with Ru/MnO 2 /SP as cathode and the DMSO-based electrolyte containing p.p.m.-leveled H 2 O is discharged and charged at 500 mA g À 1 (corresponding to 0.71 mA cm À 2 based on the Brunauer-Emmett-Teller, (BET) surface area of Ru/MnO 2 /SP) as shown in Fig. 1a. A charge potential plateau as low as B3.20 V, corresponding to an overpotential of B0.24 V, can be clearly observed. Its corresponding dQ/dV curve in Fig. 1b shows sharp oxygen reduction reaction and oxygen evolution reaction peaks in the discharging and charging process, respectively, which are consistent with the flat discharge and charge plateaus. In addition, a small plateau at B3.60 V is also obtained in the charge profile in The corresponding dQ/dV curves and the contact angle of the electrolyte on the cathode. The discharge and charge cutoffs are 1,000 mAh g À 1 (5,099 mC cm À 2 based on the BET surface area of Ru/MnO 2 /SP) and 4.0 V, respectively. The potentials against Li þ /Li are converted from LiFePO 4 . Rate: 500 mA g À 1 -based on the total weight of Ru, MnO 2 and SP, corresponding to 0.71 mA cm À 2 ; loading: B0.5 mg cm À 2 .
Ru/MnO 2 /SP cathode that has a large contact angle of 62°, compared with the discharge and charge profiles of the Li-O 2 cell with a dimethoxyethane-(DME-)-based electrolyte and the good wettability of DME on the cathode in Supplementary Fig. 2. This may enable the oxygen evolved in a charging process to accumulate in the voids between the residual discharge product and the cathode, as schematically described in Supplementary  Fig. 3 and Supplementary Note 2. It would induce slow diffusion of electrolytes into the voids for the decomposition of the remaining discharge products and result in the relatively large overpotential at B3.60 V, which is not observed in the affinitive DME-based electrolyte ( Supplementary Fig. 2). Similar observations are also obtained in the other kinds of electrolytes, such as triglyme (G3)-and tetraglyme (G4)-based electrolytes ( Supplementary Fig. 4). The trace amount of H 2 O in electrolytes is demonstrated to play a crucial role on reducing charge overpotentials of Li-O 2 cells in all the widely used electrolytes, which was empirically considered as a negative factor in a Li-ion cell 29 .
Analysis on the discharged/charged cathode. The discharged and charged Ru/MnO 2 /SP cathodes in the DMSO-based electrolyte containing water have been characterized by X-ray diffraction (XRD). As shown in Fig. 2a, the discharge products are identified as a mixture of LiOH and Li 2 O 2 , referring to the standard powder diffraction files of 01-085-0777 and 00-009-0355, respectively. The diffraction peaks of LiOH become sharper and stronger at the cathode with a discharge capacity of 4,000 mAh g À 1 and addition of more electrolyte ( Supplementary  Fig. 5), suggesting more LiOH converted from Li 2 O 2 . To avoid the effect of MnO 2 , quantification of LiOH and Li 2 O 2 was conducted on the Ru/SP cathode with the same discharge capacity as in Fig. 1 via iodometric titration. The discharge product Li 2 30,31 , where H 2 O 2 further oxidizes iodide to iodine, the titrated target. The LiOH and Li 2 O 2 in the discharged cathode were titrated via two steps (Supplementary Figs 6 and 7; titration processes in Supplementary Methods) and estimated to 16.02 and 1.48 mmol, respectively, which are in agreement with the XRD patterns of the discharged cathode in Fig. 2a. The majority of LiOH formed at a discharged cathode is revealed by the characteristic absorbance peak in the infrared (IR) spectra in Fig. 2b. Based on the electrons passing through the cathode and the oxygen derived from the discharge products of LiOH and Li 2 O 2 by iodometric titration, the discharging process is a 1.97 e À /O 2 process, quite close to the theoretical value of 2.00 e À /O 2 . It is confirmed that in other electrolytes, like DME-, G3-and G4-based electrolytes, the discharging process is also a B2.00 e À /O 2 process, as listed in Supplementary Table 1. This suggests that LiOH is converted from Li 2 O 2 via a chemical, not an electrochemical, process in the discharging processes.
After charge, the discharge products LiOH and Li 2 O 2 were decomposed, as evidenced by the disappearance of their characteristic diffraction peaks in Fig. 2a. Further, the discharge products were directly observed by scanning electron microscope (SEM) in Fig. 2c. All the products are toroidal particles with obvious layering. These are quite similar to the toroidal aggregates in a Li 2 O 2 -only discharged cathode 32,33 , which may suggest that the LiOH was derived from Li 2 O 2 in the presence of H 2 O in electrolytes by persisting the similar shape. The charged cathode becomes more porous after decomposition of the discharge products in Fig. 2d in comparison to the fresh cathode in Fig. 2e. These results indicate that the low charge overpotentials demonstrated in Fig. 1 Fig. 4a. The polarization is obviously increased with the current density from 250 to 500 and 1,000 mA g À 1 . The overpotential in the discharging and charging process at 250 mA g À 1 is 0.11 V and 0.21 V, respectively, leading to a small discharge/ charge potential gap of 0.32 V. The Li-O 2 cell was continuously discharged and charged at 500 mA g À 1 for 200 cycles and B800 h, and the selected runs of discharge and charge are shown in Fig. 4b. The discharge and charge curves during the first 150 cycles are almost overlapped except the first charge. A slight charge potential increase beyond the 150 cycles can also be observed, which could be related to the electrolyte instability during many cycles ( Supplementary Fig. 11) 37,38 . This indicates superior cycling stability of the Li-O 2 cell, which is in sharp contrast to that without MnO 2 in Fig. 3. The discharge and charge capacities in the 200 cycles are almost constant, and the corresponding coulombic efficiency in each run is approaching 98% (Fig. 4c), indicative of good reversibility. This may be partially benefited from the conversion of chemically active Li 2 O 2 to LiOH. The reversible formation and decomposition of LiOH and Li 2 O 2 during the 200 cycles are further confirmed by ex situ XRD patterns in Supplementary Fig. 12 and SEM images in Supplementary Fig. 13 (Supplementary Note 3). The low charge potentials sustained for so many cycles, to the best of our knowledge, have never been achieved before. The small discharge/ charge potential gap and good cycling stability of the Li-O 2 cell are rewarded by the 'water catalysis' at the Ru/MnO 2 /SP cathode. C (mAh g -1 ) Cycle number it can be largely promoted to move forward by the conversion of one of the products H 2 O 2 to H 2 O over MnO 2 via Step (ii) 36 . The two Steps (i and ii) occur sequentially, and quickly transform Li 2 O 2 to LiOH as long as H 2 O remains in the electrolyte. This has been confirmed by the presence of substantial LiOH in discharged cathodes as revealed by both of the XRD patterns and the IR spectra in Fig. 2. This is consistent with the observations of Aetukuri et al. 25 , and Schwenke et al. 26 , where the H 2 O in electrolytes is possibly consumed by the employed Li anode or saturated by the product LiOH, and the lack of a promoter like MnO 2 for Step (ii) made the equilibrium reaction in Step (i) to move backward and result in the major discharge product Li 2 O 2 as detected.
In the following charging process, the resultant LiOH can be directly oxidized via Step (iii) to regenerate H 2 O at low charge potentials, by which the residual Li 2 O 2 is then converted to LiOH via Steps (i and ii) and oxidized. To demonstrate the feasibility of Step (iii), commercial LiOH was ball-milled and then thoroughly mixed with Ru/SP with a ratio of 30:70 (wt/wt). The electrodes of Ru/SP with and without LiOH under Ar atmosphere were subjected to linear scanning voltammetry (LSV) at a low scan rate of 0.01 mV s À 1 as depicted in Fig. 5b. The Ru/SP electrode incorporated with LiOH presents significant oxidation currents and an oxidation onset potential of B3.27 V in Fig. 5b, which is in sharp contrast with no oxidation response on the electrode without LiOH. At a carbon electrode pre-filled with LiOH, a high charge potential of 44.0 V was reported 43 . This suggests that over Ru/SP LiOH was oxidized at low potentials, though it may be dependent on its morphology and facet orientation and the catalyst. The gas evolved in a charging process can be identified as O 2 by a gas chromatography (GC) mounted with a thermal conductivity detector, as evidenced by the sharp GC signal of O 2 in Fig. 5c. As shown in the whole discharging and charging   Fig. 5a, H 2 O is not consumed and can be circulated to behave like a catalyst. The total electrochemical reaction occurring at the oxygen cathode is 2Li þ þ O 2 þ 2e À 2Li 2 O 2 , consistent with the Li-O 2 cell chemistry.

Discussion
The dependence of the charge potentials on the concentration of H 2 O in electrolytes is shown in Fig. 6. In the dried DMSO-based electrolyte, one charge potential plateau can be obtained at B3.65 V, which is attributed to the oxidation of Li 2 O 2 and in good agreement with the literatrues [10][11][12][13][14][15] . When there is 120 p.p.m. of water in the electrolyte, the charge potential plateau is significantly reduced to B3.2 V. This has been confirmed to the oxidation of LiOH that can be quickly converted from the primary discharge product Li 2 O 2 via the sequential Steps (i and ii) in Fig. 5a over the catalyst of Ru/MnO 2 /SP in both the discharging and charging processes. However, when the concentration of water in the electrolyte is increased to 281 p.p.m., the charge potential plateau is shortened and increased. It may be induced by the LiOH on the surface of the discharge product, which is surrounded/adsorbed by H 2 O molecules in the electrolyte, and hence has high oxidation potentials following the Nernst equation 17,32,33 .
In addition, by controlling the discharge depth of the Li-O 2 cells, the discharge product can be ranged from LiOH only to a majority of Li 2 O 2 plus LiOH with the same amount of electrolytes used in cells. The cathode with a relatively small discharge capacity of 250 mAh g À 1 was covered with LiOH, on which no Li 2 O 2 was detected by iodometric titration. It is in the shape of thin disks as shown in Supplementary Fig. 14 (refs 25,33). In the following charging process, the potentials are increased to B3.65 V and similar to that with 281 p.p.m. of H 2 O in electrolytes in Fig. 6. The low charge potentials at B3.20 V can be observed again when the discharge capacity is increased to 500, 1,000 and 2,000 mAh g À 1 , as shown in Fig. 7. At large discharge capacities, the H 2 O molecules in the electrolyte can be consumed by Li 2 O 2 , producing surface-clean LiOH/Li 2 O 2 in the cathode. This suggests that the relative amounts of H 2 O, LiOH and Li 2 O 2 coexisting at a cathode side affect the charging overpotentials. The detailed formation/decomposition mechanism of LiOH and Li 2 O 2 in the presence of water in electrolytes and the dependence of their decomposition potentials on their morphologies deserve further investigation.
The 'water catalysis' at the oxygen cathode side has been demonstrated to reduce the charge overpotentials to B0.24 V, corresponding to B3.20 V vs Li þ /Li and provides a possible solution to the current challenges of Li-O 2 cells. With the Ru/MnO 2 /SP cathode and the DMSO-based electrolyte containing p.p.m.-leveled H 2 O, the Li-O 2 cell presents a small discharge/ charge potential gap of 0.32 V and superior cycling stability of 200 cycles B800 h. These have not been achieved before, and are rewarded by the proposed reaction mechanism. Although the LiFePO 4 applied in cells is not a practical anode of Li-O 2 cells, the reaction mechanism at the oxygen cathode may be extended to a practical Li-O 2 cell by using a Li þ ion-conducting ceramic membrane to separate the electrolyte with water and the Li anode 2,39 . This could also alleviate the carbon-related side reactions by converting the chemically active Li 2 O 2 to LiOH, and it will make the cheap and lightweight carbon possible as cathodes in Li-O 2 cells, which has been suggested to avoid. This investigation will enable the cell to operate in ambient air by eliminating CO 2 and advance the Li-O 2 /air cell technology.

Methods
Materials preparation. Electrolytic MnO 2 (EMD, g-MnO 2 ) was received from TOSOH, Japan and ball-milled to nanometre scale. Ru/SP was synthesized as follows: 85 mg of SP was stirred into a solution of ethylene glycol (EG, purity of 499.5%, Wako Chemicals) containing 7.5 mg of Ru in RuCl 3 xH 2 O (purity of 36B42 wt% based on Ru, Wako Chemicals), and its pH was adjusted to 13 with 0.1 M of NaOH in EG. The suspension was then heated to 160°C for 3 h with flowing N 2 . After cooling to 80°C, its pH was adjusted to 3 using 0.1 M HCl and the resulting mixture was further stirred for 12 h. The final product Ru/SP was centrifuged, washed with de-ionized water until the solution pH reached B7 and dried at 80°C in a vacuum oven for 12 h. For the preparation of MnO 2 /Ru/SP, the as-synthesized Ru/SP and MnO 2 nanoparticles with a weight ratio Ru/MnO 2 /SP ¼ 7.5:7.5:85 were sonicated into an ethanol aqueous solution for 3 h and stirred overnight. Then, the product MnO 2 /Ru/SP was collected and dried at 80°C under vacuum for 12 h.
DMSO, DME, G3 and G4 were firstly dried over 4 Å molecular sieves and then using Li metal. The Li salts of LiClO 4 and lithium bis(fluorosulfonyl)imide were used as received from Wako Chemicals. The water in the prepared electrolytes was from the Li salts and measured on a desk-top Karl-Fisher Titration instrument.
Cell assembly. The electrode film composed of Ru/MnO 2 /SP or Ru/SP and PTFE (A dispersion of 60 wt%, Du Pont-Mitsui Fluorochemicals Co. Ltd.) with a ratio of 85:15 wt% was rolled with a glass rod. The mass loading of Ru/MnO 2 /SP or Ru/SP is B0.5 mg cm À 2 . The electrode film was pressed onto a hydrophobic carbon paper (SIGRACET Gas Diffusion Media, Type GDL 35BA) to work as a cathode. The anode was consisted of LiFePO 4 (Sumitomo Osaka Cement), SP and PTFE (70/20/10) pressed on an Al mesh. The amount of DMSO-and DME-based electrolytes in a coin cell was 100 ml and G3-and G4-based electrolyte was 50 ml, considering their different vapour pressures. The employed separator was a glass microfibre filter paper (GF/A, Whatman). The Li-O 2 cell assembly in 2032 coin cells was conducted in an Ar-filled glove box that has a dew point of around À 90°C (B0.1 p.p.m. of H 2 O) and O 2 content below 5 p.p.m.. The cells stored in a glass chamber with a volume capacity of 650 ml were purged with O 2 (99.999%) before electrochemical tests.
Characterization and measurements. The contact angles of electrolytes on the cathode were examined on AST Products with a model of Optima. The BET surface area of Ru/MnO 2 /SP was measured to be 70.6 m 2 g À 1 on Belsorp 18 via nitrogen adsorption-desorption. XRD was performed on a Bruker D8 Advanced diffractometer with Cu Ka (l ¼ 1.5406 Å) radiation with a scan rate of 0.016°per s. Galvanostatic discharge/charge was conducted on a Hokuto discharging/charging system. All the electrochemical measurements were conducted at 25°C. The specific capacities and currents are based on the mass of Ru/MnO 2 /SP or Ru/SP in cathodes. The discharged and charged cathodes were extracted in the glove box and washed with dried dimethoxyethane. SEM was obtained on Hitachi S4800. The Ru/MnO 2 /SP was mixed with KBr and then pressed to a pellet for Fourier transform infrared spectroscope (FTIR) characterization on a JASCO instrument of FT/IR-6,200 from 2,000 to 400 cm À 1 with a resolution of 2 cm À 1 . The electrolytes during cycles were collected by washing the glass fibre filter separators with acetone-d 6 , and subjected to 1 H NMR (Bruker, 500 MHz).