Enabling safe aqueous lithium ion open batteries by suppressing oxygen reduction reaction

Due to the non-flammable nature of water-based electrolytes, aqueous lithium-ion batteries are resistant to catching fire. However, they are not immune to the risk of explosion, since the sealing structure adopted by current batteries limits the dissipation of heat and pressure within the cells. Here, we report a safe aqueous lithium-ion battery with an open configuration using water-in-salt electrolytes and aluminum oxide coated anodes. The design can inhibit the self-discharge by substantially suppressing the oxygen reduction reaction on lithiated anodes and enable good cycle performance over 1000 times. Our study may open a pathway towards safer lithium-ion battery designs. Despite the non-flammable nature of water-based electrolytes, aqueous lithium-ion batteries still carry an explosion risk due to the sealing structure. Here the authors report a safe aqueous battery with an open configuration, utilizing highly concentrated electrolytes and Al2O3 coated anodes.

T he safety of lithium-ion batteries (LIBs) has raised significant concerns in recent years due to several fire-related incidents [1][2][3] . The fully charged LIB consists of a highly energetic transition metal oxide cathode and a lithiated graphite anode in intimate contact with a flammable organic electrolyte. Any abuse by overcharging, external short-circuiting, or crushing can trigger spontaneous heat-generation. This exothermic event can result in thermal runaway, or even explosion due to high internal cell pressure producing fire and toxic gases. The thermal stability of batteries depends on the electrolyte's flammability and its ability to dissipate the heat and pressure. Currently, all commercial LIBs are in sealed configurations to protect the highly reactive electrodes and liquid electrolytes from reacting with the moisture and the reactive gases in the air, inherently limiting the dissipation of heat and pressure (Fig. 1a). Although many safety devices, such as safety vents, thermal fuses, circuit breakers, positive temperature coefficient elements, and shutdown separators, have been applied to LIBs and battery packs 3 , the fire and explosion risks of LIBs are still present. To make LIBs intrinsically safe, the LIBs' chemistry and configuration have to be changed.
At the chemistry level, the volatile, flammable, and toxic organic electrolytes in commercial LIBs should be replaced by non-flammable aqueous electrolytes [4][5][6][7][8] . However, the traditional non-flammable aqueous LIBs suffer from a low energy density due to the narrow electrochemical stability window of aqueous electrolytes 9 . The performance of aqueous LIBs is significantly impacted by the dissolved O 2 in electrolytes from the air through the oxygen reduction reaction (ORR). Xia and coworkers 10 have demonstrated that any discharged negative electrode in an aqueous LIB would react with O 2 , resulting in the marked capacity fading. Sealing cell can help to eliminate the O 2 and can significantly improve the battery performance. But sealed aqueous LIBs still have the risk of thermal runaway and explosion when the internal cell pressure quickly rises, either by electrochemical water decomposition or uncontrolled high-temperature thermovaporization, as observed in sealed aqueous Ni-MH and VRLA (valve regulated lead acid) batteries. The battery's vents are not always able to prevent cell explosions because the pressure builds up from accelerated side reactions producing gas at a rate too fast for the safe operation of the safety valves, leading to the dangerous cell explosions. The sealed aqueous LIBs have to be replaced by an open configuration design due to their superior ability to dissipate heat and pressure. To our best knowledge, such open configuration aqueous LIBs have not been demonstrated yet. Recently, we developed a "water-in-salt electrolyte" (WiSE) that significantly reduced the O 2 solubility in the electrolyte and expanded the electrochemical stability window of aqueous electrolytes from 1.5 to >3.0 V. The WiSE enabled the energy density of aqueous LIBs to be significantly enhanced from 75 Wh kg −1 to over 300 Wh kg −1 (on the material level) [11][12][13][14][15][16] . The uniqueness of WiSE has been extensively reviewed by Eftekhari 9 .
Here in this study, we report intrinsically safe LIBs with an open configuration. It is found that the O 2 solubility markedly decreases in the electrolytes from 1.97 mg L −1 in 1 m "salt-inwater electrolytes" (SiWE) (

Results
WiSE enabled open configuration. The low O 2 content in the 28 m WiSE reduces the ORR kinetics at the electrodes. In order to build an open aqueous battery, the ORR kinetics at the electrodes must be reduced. The gas solubility in a solvent can be largely reduced by increasing the salt concentration in the solvent 17 . Figure  1 Where j is the measured current density; j k is the kinetic current density; F is the Faraday constant (96485 C mol −1 ); D O 2 is the O 2 diffusion coefficient in the electrolytes; C O 2 is the O 2 saturation concentration in the electrolytes (6.156 × 10 −5 mol L −1 for the 1 m SiWE and 2.968 × 10 −5 mol L −1 for the 28 m WiSE) (Fig. 1b); ν is the kinematic viscosity of the electrolyte that is obtained using the viscometer (1.396 mPa s −1 for the 1 m SiWE and 202.44 mPa s −1 for the 28 m WiSE); ω is the electrode rotation rate; and n is the electron transfer number (for the 1 m SiWE, n can be obtained via the RRDE test using Eq. 2).
Where j r and j d are the ring and the disk current, respectively, and N is the collection efficiency (0.37). From the disk currents of the RRDE in Fig. 1d The reaction mechanism suggests that n for the ORR at the carbon electrode in the 28 m WiSE should be 2. The D O 2 in the 28 m WiSE was determined from the K-L equation (Eq. 1) and the RDE data in Fig. 1e to be 1.54 × 10 −5 cm 2 s −1 . The D O 2 in the 28 m WiSE is less than the value (5.45 × 10 −5 cm 2 s −1 ) in the 1 m SiWE, which further reduces the ORR kinetics at the carbon black electrode in the 28 m WiSE. The formation of an insulating Li 2 O 2 and LiF nano passivation layer on the anodes further reduces the ORR kinetics. Finally, the highly insulating Li 2 O 2 at the carbon electrode does not dissolve in the 28 m WiSE but forms a nano passivation layer, which inhibits the ORR [20][21][22][23] . Moreover, an LiF-based solid-electrolyteinterphase (SEI) can also be formed on the anode from reduction of the LiTFSI in the WiSE at around −0.5 to −0.3 V (vs. Ag/ AgCl) 5 , further suppressing the ORR.
In addition to the capability of WiSE to reduce the ORR kinetics, it also prevents the water from evaporating. Fig. 1g demonstrates the mass retention of the 28 m WiSE during exposure to the air with a relative humidity of~68% at room temperature. The 28 m WiSE retains the water content for over 60 days with only a slight mass intake of 0.8%, indicating that the electrolyte actually absorbs moisture from the ambient air. Moreover, even when the 28 m WiSE is exposed to the air at a high temperature of 100°C (boiling point of water) for 30 min, the mass retention of this electrolyte is still over 99% (Fig. 1h), which indicates that only trace water will evaporate at this high operating temperature. The high stability of the TFSI anion in the aqueous electrolyte at different pH values for 100 days at various temperatures has been shown in a previous study 24  To suppress the ORR during the initial passivation layer formation process, an Al 2 O 3 nano-layer was coated on the electrode using ALD to generate an artificial SEI layer prior to cell construction.
The O 2 adsorption and the subsequent ORR on the Al 2 O 3 surface were investigated. Figure 3a, b delivers the different charge densities of the O 2 adsorbed on the Al 2 O 3 slab. The corresponding adsorption energies (E ad ) were calculated using Eq. 6: Where the E The RDE curves in Fig. 3c, d confirm that the ORR kinetics at the Al 2 O 3 surface are much slower than the ORR kinetics at the carbon back electrode. Figure 3c shows the LSV of the carbon black and the Al 2 O 3 in the O 2 -saturated 1 m SiWE, the ORR onset potential of the ORR at the Al 2 O 3 surface is 0.1 V lower than at the carbon black surface. The limiting current density at 1600 rpm at the Al 2 O 3 surface (0.107 mA cm −2 ) is only 1/8 of that (0.85 mA cm −2 ) at the carbon black surface. Similarly, the Al 2 O 3 also reduces the ORR current in the 28 m WiSE (Fig. 3d). The onset potential of the ORR at the Al 2 O 3 surface is also 0.2 V lower than that at the carbon black surface in the 28 m WiSE (Fig. 3d). In addition, the Al 2 O 3 has a fast Li-ion diffusivity 27 (Fig. 4b). The open-circuit potential changed from the fully lithiated potential of −0.7 V to the fully dilithiated potential of +0.2 V during 10 h (Fig. 4c). Therefore, the fully lithiated Li 3 Ti 2 (PO 4 ) 3 in the 1 m SiWE has completely self-discharged to be LiTi 2 (PO 4 ) 3 after 10 h rest in an open-cell configuration. Supplementary Fig. 4 and Supplementary Note 3 confirm the ORR at the LiTi 2 (PO 4 ) 3 electrode. Supplementary Fig. 5 and Supplementary Note 4 further confirm the effects of the dissolved O 2 in the electrolytes on the self-discharge performance of the lithiated Li 3 Ti 2 (PO 4 ) 3 , the open-circuit potential changed to +0.2 V after only a half hour in a pure O 2 atmosphere, while its lithiated potential can be maintained for 50 h without potential increase when the cell was in the absence of O 2 (in a N 2 atmosphere). Therefore, it can be concluded that the self-discharge of the fully lithiated Li 3 Ti 2 (PO 4 ) 3 in the 1 m SiWE open-cell is proceeds through the following ORR pathway: In an open-cell configuration, the dissolved O 2 in the electrolyte receives electrons from the Li 3 Ti 2 (PO 4 ) 3 and is reduced into OH − or H 2 O 2 . As expected, when the Al 2 O 3 nano-layer coated LiTi 2 (PO 4 ) 3 electrode ( Supplementary Figs. 6 and 7) is charged/ discharged in the 28 m WiSE, the Coulombic efficiency markedly increases to 99.9% (Fig. 4d). The capacity retention of the fully lithiated Li 3 Ti 2 (PO 4 ) 3 after 10 h rest in an open-cell configuration still reaches a high value of >97% (Fig. 4e). Also, the open-circuit potential is maintained at −0.5 V (Fig. 4f) (Fig. 5a) and displayed a high Coulombic efficiency of over 99.9% with a stable cycle life of >1000 cycles (Fig. 5b, c). In sharp contrast, the capacity of the LiMn 2 O 4 //LiTi 2 (PO 4 ) 3 opencell in the 1 m SiWE decreased markedly due to the poor Coulombic efficiency ( Supplementary Fig. 11, Fig. 5b, c). Selfdischarge is a key issue for all batteries, Supplementary Fig. 12 and  Fig. 13). The capacity retention is still larger than 90% after 15 days of rest ( Supplementary Fig. 14). Furthermore, the charging of the  Fig. 13b). Even at a low rate of 0.  Fig. 15).
To further enhance the cell energy density, a 2.  Fig. 5d). The cell achieved over 50 cycles with a Coulombic efficiency of 96.8% (Fig. 5e). Energy efficiency is another critical factor for new battery systems 28,29 . We have calculated the energy efficiency of the LiMn 2 O 4 //Al 2 O 3 @LiTi 2 (PO 4 ) 3 open-cell at the rate of 1 C and found it to be 90% (Supplementary Fig. 13C). The energy efficiency of the LiVPO 4 F//Al 2 O 3 @Li 4 Ti 5 O 12 open pouch cell, cycled at 0.2 C, (Fig. 5e) is as high as 92.8%, which is comparable with the value (in the range of 86-98%) of commercial LIBs 29,30 . Therefore, the open-cell configuration developed here can satisfy the demand of the market.
To address the potential electrolyte leakage issue in the opencell design, we added PVA (Poly(vinyl alcohol)) into the 28 m WiSE to form a gel electrolyte. The gel electrolyte can flow when the temperature is increased to 95°C in order to facilitate battery assembly. When the electrolyte is cooled back to room temperature it forms a solid phase that can eliminate any electrolyte leakage. Supplementary Fig. 19 and Supplementary Note 9 show that the 28 m gel WiSE do not flow at room temperature. As shown in Supplementary Fig. 19b This aqueous open-cell configuration is a universal design that can also be used in other aqueous batteries. For example, the 35 m WiSE also effectively suppressed the ORR at the discharged electrode, which enabled the intercalated Na 3 Ti 2 (PO 4 ) 3 electrode in an open-cell configuration to maintain over 98.5% of the initial capacity after resting at open-circuit for 10 h ( Supplementary  Fig. 21). In addition, the 30 m WiSE also effectively suppressed the ORR at the Zn anode, as demonstrated by the much higher Coulombic efficiency (87.5%) for the Zn plating and stripping in the 30 m WiSE than in the 5 m ZnCl 2 SiWE (53%) (Supplementary Fig. 22). The open-cell configuration has several other advantages: (1) some safety devices are not necessary within the cell such as an additional ventilation system; (2) they can be charged at a high voltage setting and high rate; and (3) they have a much lower up-front cost by removing the safety devices and sealing process.

Discussion
The non-flammable aqueous batteries still carry the risk of explosion due to rapidly increasing internal pressure caused by side reaction gaseous products and thermal runaway. The WiSE can effectively suppress the electrolyte evaporation and the ORR reaction at the discharged anode. The ORR at the anode can be further suppressed by coating an formed when the water was evaporated after heating at 80°C with continuous stirring. The product was heated at 900°C for 10 h at a temperature increasing rate of 5°C min −1 under the protection of N 2 flow. During the heat-treatment process, it was in a porcelain boat, and the tube furnace was applied. Thermal decomposition vapor-deposition technology was applied to coat carbon on the surface of LiTi 2 (PO 4 ) 3 . The as-prepared LiTi 2 (PO 4 ) 3 was placed in a tube furnace and further heated at 700°C for 2 h, during which, the toluene vapor was carried by N 2 through the tube. The flow rate of N 2 is 1 L min −1 . After that, the product was further heated at 900°C for 2 h under the protection of N 2 without toluene.
Atomic layer deposition. The electrodes were coated with Al 2 O 3 nano-layer using the atomic layer deposition (ALD) equipment (Beneq TFS 500). The carrier gas was high-purity N 2 with a temperature of 150°C. 20 precursor pulse cycles of ALD-Al 2 O 3 were applied to form 2-nm-thickness layer on the surface of electrodes. During each cycle, alternating trimethylaluminum (4 s, Al precursor) and H 2 O (4 s, oxygen precursor) flows were separated by flows of N 2 (4 s for the carrier gas, 10 s for the cleaning gas).
Electrode preparation and electrochemical measurements. tests were performed using a three-electrodes cell, in which, the activated carbon was used as the counter electrode, the Ag/AgCl electrode was used as the reference electrode. The LSV measurements were carried out on a CHI660B electrochemical workstation. The galvanostatic charge/discharge tests of anodes were performed using a three-electrodes cell, in which, the activated carbon was used as the counter electrode, the Ag/AgCl electrode was used as the reference electrode. The galvanostatic charge/discharge of full cells were performed using the Arbin electrochemical working station. The carbon black, Al 2 O 3 (nano particles form Sigma) and carbon-coated LiTi 2 (PO 4 ) 3 inks were prepared as follows: (1) the active materials (5 mg) was dispersed in a mixture of isopropanol (2.5 mL) and Nafion solution (20 μL); (2) treat the mixture with ultra-sonication for 30 min. After that, the ink (10 μL) was deposited on a glassy carbon disk electrode (RDE or RRDE) and evaporated the solvent in air at room temperature. The electrochemical measurements were performed in a three-electrodes cell at ambient temperature. The RDE or RRDE was used as the working electrode, a platinum electrode as the counter electrode, and Ag/AgCl as the reference electrode. The data was recorded using a Gamry interface 1000.
Materials characterizations. Raman measurements were carried out by a Horiba Jobin Yvon Labram Aramis using a 532 nm diode-pumped solid-state laser, attenuated to give~900 μW power at the sample surface. Scanning electron microscopy (SEM) measurements were carried out by Hitachi SU-70 analytical SEM (Japan). Viscosity measurements were carried out using a CANNON-FENSKE viscometer. The surface chemistry of the electrodes after ALD coating was examined by XPS with a Kratos Axis 165 spectrometer. XPS data was collected using a monochromated Al Ka X-ray source (1486.7 eV). The working pressure of the chamber was lower than 6.6 × 10 −9 Pa. All reported binding energy values were caLIBrated to the C 1s peak at 284.8 eV.
Computational details. All density functional theory (DFT) calculations 31,32 were performed using a Vienna Ab Initio Simulation Package (VASP) 33 with projector augmented wave (PAW) method 34 . The exchange-correlation energy is described by the functional Perderw, Burke, and Ernzerhof (PBE) version of the generalized gradient approximation (GGA) 35 . And the energy cut-off for the plane wave basis is 520 eV. A vacuum layer of 12 Å was used for all calculated models. The energy of O 2 is obtained from the Materials Project 36 . Visualization of the structures are made by VESTA 37 .

Data availability
The data that support the plots within this paper and other findings of this study are available from the corresponding authors upon reasonable request.