Kosmotropic anion for improving cycling stability of aqueous lithium-ion batteries in salt-in-water electrolytes

The incompatibility between Li + -intercalated electrodes and water limits the practical feasibility of aqueous lithium-ion batteries (LIBs), which are economical and environmentally benign energy storage systems. Tremendous amounts of salts dissolved in water (water-in-salt) have been utilized to mitigate the access of water to the electrode/electrolyte interface and to extend the electrochemical potential window of aqueous LIBs. However, this approach has low viability owing to the expense of the salts. Here, we show that kosmotropic anions with moderate concentrations (0.5 ~ 3 mol kg − 1 ) protect the LiCoO 2 electrode by harnessing water molecules. The sulfates of kosmotropic anions develop rigid water-solvation shells and also form ion pairs with Li + . All-atomic-level multiscale simulation revealed that sulfates tied with Li + and the water shell are highly concentrated at the interface, thus decreasing the density of free water. The suppressed water activity explains the superior cell performance achieved with 0.5 mol kg − 1 sulfate relative to that in cells with 1 mol kg − 1 of chaotropic anions such as nitrate, perchlorate, and bis(triuoromethylsulfonyl)imide. The formation of a liquid-phase protective layer is a new concept for developing stable aqueous batteries without the requirement for a solid-state electrolyte or an articial protective layer on the electrode.


Introduction
With the increasing importance of safety, cost, and eco-friendliness, rechargeable batteries employing aqueous medium have received much attention as suitable for grid-scale energy storage systems (ESSs). [1][2][3] Aqueous lithium-ion batteries (LIBs) are one of the most feasible systems for industrialization as they can be promptly developed by transformation of well-established non-aqueous LIBs. However, most Li + -intercalating electrodes are incompatible with water, undergoing swift structural deformation, which results in poor cell performance. [4][5][6][7] Recently, invaluable studies on water-in-salt electrolytes (WiSEs) have been demonstrated using very large amounts of bulky imide anions, such as 21 m (molal concentration, mol kg -1 ) bis(tri uoromethylsulfonyl)imide (TFSI -) and additional bis(penta uoroethylsulfonyl)imide. 8,9,10 The limited free-water concentration and the formation of aggregated ion pairs in WiSEs alter the cell chemistry to establish a dense protective surface layer and widen the potential window. 8, [11][12][13] Nevertheless, utilizing WiSEs in practical aqueous LIBs is inexpedient because of the high cost of the large amount of salts and easy crystallization of the symmetrically structured anions. 10,14,15 Therefore, a more desirable approach is to maintain salt-in-water electrolytes and control the content of water and protons (H + ), which are much more abundant than the electrolyte salts, to stabilize the Li +intercalating electrodes. Water deforms the electrode surface chemically during cell operation. 4,16 Furthermore, the dissociation of water produces H + , which is irreversibly inserted into the electrode structures and expedites capacity fading over a wide pH range. [17][18][19][20][21] To curtail structural deformation of the electrodes, inorganic and graphitic protective layers have been introduced onto the electrode surface. [22][23][24] However, these solid-state layers may retard ion transport at the interface and generate cracks during cycling.
Instead, we herein introduce the new concept of a liquid-state protective layer formed with moderate concentrations (0.5 ~ 3 m) of anions. The key principle is to enhance the interaction of the anion with both water and Li + , and reduce the activity of free water molecules that are considered to be the source of H + . It is found that sulfate (SO 4 2-), as a kosmotropic anion, mitigates the detrimental deformation of the lithium cobalt oxide (LiCoO 2 , denoted as LCO) electrode. The rigid water structure formed adjacent to the kosmotropic anion aids in improving the cell cyclability. The structural details of the electrode-electrolyte interface are investigated using multiscale QM/MM simulations, revealing that the high charge density of the kosmotropic anion in the interfacial region induces localized concentration of the Li + ions and lowers the free water density. In stark contrast, diffuse chaotropic anions such as TFSI -, nitrate (NO 3 -), and perchlorate (ClO 4 -) neither strongly controlled the water shell nor formed ion pairs. The resulting capacity fading with the chaotropic anions is in stark contrast with that of the kosmotropic system.

Results And Discussion
The galvanostatic cyclability of the LCO electrodes was rst evaluated using two representative electrolyte salts, namely, 0.5 m Li 2 SO 4 and 1 m LiTFSI, which both provided 1 m Li + . The LCO electrode with 0.5 m Li 2 SO 4 exhibited a gradual increase in capacity at 0.5C over 30 cycles as the wetting of the electrode increased, 5 and the coulombic e ciency (CE) approached 97% (Figure 1a and 1c). By comparison, 1 m LiTFSI reduced the capacity from 110 to 30 mAh g -1 ( Figure 1b) and afforded an inferior CE of 94% ( Figure 1c). Rate-capability tests with 1 m LiTFSI displayed an ill-de ned capacity at various current rates ( Figure 1d). In contrast, complete recovery of the capacity was found for the LCO electrode with 0.5 m Li 2 SO 4 at the terminating 0.2C after 10C operation. These results reveal that the cyclability and rate capability of the LCO electrode are both anion-dependent.
To determine the origin of the capacity fading, the short-range order of the LCO structure was analyzed using soft X-ray absorption ne structure (XAFS) spectroscopy. 7 In surface-sensitive partial electron yield (PEY) mode (< 10 nm depth), the Co 3+ 3d-O 2p hybridization peak in the oxygen (O) K edge spectra (529.8 eV) was attenuated for both LCO electrodes after 30 cycles. In addition, a new signal emerged at 531.4 eV, attributed to partial oxidation of the lattice oxide of LCO; this signal is denoted as O1 (Figure 1e-f). [25][26][27] The O1 signal was particularly intense for LCO with 1 m LiTFSI, indicating more extensive surface oxidation than in the case with 0.5 m Li 2 SO 4 . In bulk-sensitive partial uorescence yield (PFY) mode (> 100 nm depth), the Co 4+ 3d (t 2g 5 e g 0 )-O 2p hybridization band was observed at 528 eV; [25][26][27] this band is denoted as O2 (Figure 1f). The signal was more pronounced with 1 m LiTFSI, consistent with a subtle shift of the main Co 3+ signals to higher energy than others in the Co L 2 and L 3 edge spectra ( Figure S1).
Because the O2 signal should disappear after Li + intercalation during the discharging process, its existence after 30 cycles indicates incomplete Li + intercalation into the LCO electrode. The remaining Li vacancy and Co 4+ cause local distortion of the octahedral CoO 6 unit. 7,27 Long-range structural transformation was also observed. Powder X-ray diffraction analysis of the LCO electrode after operation with 1 m LiTFSI displayed split peaks related to the 003 re ection ( Figure S2). The new peak appearing at 18.7° indicates widening of the interlayer distance due to the de ciency of Li + . 4,28 In contrast, a less profound O2 signal was observed with 0.5 m Li 2 SO 4 in the XAFS pro le, along with a single 003 re ection, indicating complete Li + intercalation into LCO after the same number of cycles.
Such inadequate Li + intercalation has been ascribed to the insertion of H + into the LCO electrode in previous studies. [17][18][19][20] H + can be produced through the dissociation of water at the electrode surface in neutral or weakly alkaline solutions. Electrochemical impedance spectroscopy (EIS) can be used to distinguish Li + intercalation and H + insertion in the LCO electrode. 19 Successive Nyquist plots were acquired at open-circuit potential (OCP, Figure S3), during the charging process (from 0.66 V to 0.8 V, red to yellow curves), and the discharging process (from 0.8 V to 0.63 V, light green to dark green curves) for three galvanostatic cycles ( Figure 2). The emerging semicircle at the beginning of the 1 st charge is attributed to the charge-transfer resistance of Li + (R ct,Li+ ). 19,20,29 The increased electronic conductivity of LCO decreased the semicircle size during the charging process, as a result of transformation from the semiconductor to the semi-metal by Li + extraction. 30 During the discharging process, the semicircle gradually increased again with 0.5 m Li 2 SO 4 . Enlargement of the semicircle at the end of discharge (0.63 V) re ects Li + lling at the topmost surface of the LCO electrode (Figure 2a-b). 31 This behavior was repeated for all three cycles, and R ct,Li+ was estimated as 27.8 Ω after 30 cycles ( Figure S4a and Table   S1). In comparison, the Nyquist plot of the LCO electrode with 1 m LiTFSI showed a larger semicircle corresponding to R ct,Li+ at the beginning of the 1 st charge (Figure 2c-d). More importantly, a new semicircle was observed in the low-frequency region at the terminal stage of the discharge process.
Although this new semicircle vanished at the beginning of the following charging process over three cycles, it eventually grew irreversibly and accounted for a larger portion of the resistance (~147 Ω) after 30 cycles ( Figure S4b and Table S1). This semicircle, which is apparently distinguished from that associated with R ct,Li+ , is attributed to the resistance from H + insertion (R ct,H+ ). 19,20 The result of H + insertion appeared at the end of the discharging process, that is, at the terminal stage of Li + intercalation. This behavior was not observed with 0.5 m Li 2 SO 4 .
The intriguing question is which factor related to the electrolyte salt leads to the distinct degree of H + insertion. Three possible factors were postulated: the pH, surface protective layer, and anion effects. The 1 m LiTFSI solution had a pH of 8, which is lower than that (pH 9.6) of the 0.5 m Li 2 SO 4 solution ( Figure   S5). When the pH of the LiTFSI solution was adjusted to ~10 by adding LiOH, the capacity retention of LCO was still limited to ~34%, which is far inferior to the value of ~100% achieved with 0.5 m Li 2 SO 4 . The Nyquist plot for the LCO electrode with 1 m TFSI -(pH 10) revealed suppression of H + insertion in the 1 st cycle ( Figure S6). Nonetheless, the R ct,H+ signal appeared at the end of the 3 rd cycle ( Figure S6), and both semicircles designated to R ct,Li+ and R ct,H+ grew over the course of 10 cycles ( Figure S7), which is in line with the capacity fading. 32 In stark contrast, the R ct,H+ signal did not appear for LCO with 0.5 m Li 2 SO 4 at the 10 th cycle, elucidating negligible capacity decay. Therefore, the pH effect was not critical under mildly alkaline conditions.
The formation of a solid-state protective layer on the LCO electrode was then considered. In non-aqueous media, decomposition of the electrolyte solution is accompanied by the construction of an interphase layer that serves to mitigate undesired surface reactions. 33,34 However, both SO 4 2and TFSIare inert in the given potential range. The X-ray photoelectron spectra (XPS) did not reveal any SO 4 2-, TFSI -, or fragmental species of the electrolytes after 30 cycles ( Figure S8). In the O 1s binding energy region, the signal of lattice Co-O at 529.7 eV was less intense, while the peak at 530.8 eV indicating both the oxide defects (O -2+δ ) and the hydroxide became signi cant. This surface deformation was inevitably and chemically induced by the contact of the electrode with water, especially after Li + extraction ( Figure   S9). 4,6,16 Therefore, the above results indicate the absence of a solid-state protective layer over the LCO electrode. After 30 cycles, a thicker amorphous surface layer was formed on the LCO with 1 m LiTFSI, as observed in the transmission electron microscope images ( Figure S10), demonstrating more acute surface degradation than that with 0.5 m Li 2 SO 4 .
With the use of 1 m LiTFSI, the capacity decreased immediately after 30 cycles, and the CE of ~94% was lower than that (>99%) achieved with all other electrolytes. Consequently, the cell performance was very distinct and followed the order: SO 4 2-> NO 3 -> ClO 4 -> TFSI -, which corresponds to the reinforcing kosmotropic nature of the anion. Consistently, it appears to larger R ct,Li+ of LCO with NO 3 than that with SO 4 2- (Figure S11a). The presence of ClO 4 promotes the formation of R ct,H+ ( Figure S11b). All these evidences support the decisive correlation between the surface reaction and cyclability.
The anion interaction perturbs the O-H vibrations of the water molecules, which are sensitive to the strength of the hydrogen bond. Infrared (IR) spectroscopy allows distinction of the vibrational bands of the O-H stretching at ~3200 cm -1 for the strongest and ordered water structure versus that at ~3400 cm -1 for the weaker hydrogen bond. 41 When Li 2 SO 4 was added to the water in concentrations from 0.1 to 3 m, the O-H band at ~3200 cm -1 gradually intensi ed as the hydrogen-bond became stronger ( Figure  3c). This result may be attributed to the Li + -water interaction, because the hydrogen bond of the water network is negligibly perturbed by SO 4 2-. 42 In other words, the SO 4 2--water interaction is as strong as the water network, thus providing a featureless signal. In sharp contrast, the IR band of LiNO 3 , LiClO 4 , and LiTFSI at ~3200 cm -1 became less intense in the order: NO 3 -< ClO 4 -< TFSI -. 41,[43][44][45] This trend underpins the disordering of the solvation structure with increasing chaotropic nature of the anion.
Additionally, the signals at 3500 ~ 3650 cm -1 are related to both the free water and the chaotropic anionwater interaction. 42,46-48 More pronounced absorption bands appeared at higher electrolyte concentrations, supporting the latter phenomenon. Figure S12 shows  Figure S13). In addition, the surface charge density (s) was calculated using Eq. 1 ( Figure  4b): where U is the electrode potential, and U PZC is the potential at the point of zero charge (PZC) where C dl approaches the minimum value. Figure 4a shows that the C dl of 1 m LiTFSI is 1.5 ~ 3 times higher than the C dl of the other electrolytes in the given potential range of 0.3 ~ -0.3 V vs. Ag/AgCl. It is attributed to large van der Waals volume, high polarizability, and mild hydrophilicity of TFSI -, 52-54 affording the excellent adsorption for both LCO material and carbon additive part in the EDL region ( Figure S14). However, a higher C dl does not necessarily mean a higher concentration of ion present at the "interface", due to the possible formation of ion pairs. As the ion pairs do not contribute to the C dl associated with the net surface charge, it is suggested to the necessity of full elucidation on the interfacial structure.
The detailed atomic arrangement of the interfacial structure formed at the LCO electrode was explicated using our recently developed mean-eld QM/MM simulation. 55 To investigate the electrolyte response to the cathodic charging, two different surface charge densities of s = 0 and -11.5 mC cm -2 were compared. First, the local ion concentration along the surface normal direction (which was chosen as the z-direction) revealed the formation of two layers within ca. 6 Å from the topmost atoms of the LCO electrode ( Figure S15a-b). Additionally, the local water concentration also showed a layering tendency near the electrode surface, and then converged into the bulk value at z > 6 Å ( Figure S15c). Thus, it is reasonably assumed that the region at z < 6 Å is an interfacial region, that is, a part of the EDL (nanometer scale) in direct contact with the solid electrode, at which the ion concentration and free water density were analyzed. Figure 4c shows the number of anions located in the interfacial region when the aqueous electrolyte was interfaced with the LCO electrode. Notably, the local anion population near the electrode increased at s = 0 mC cm -2 as the ion became more kosmotropic; SO 4 2showed the largest population. A water adlayer was formed on the LCO surface ( Figure S16), which can attract ions to the interfacial region via ion-dipole interactions. Because the ion-dipole interaction can be maximized for ions with a high charge density, kosmotropic ions such as SO 4 2-can accumulate in large quantities near the electrode surface.
Surprisingly, such a trend still holds even when the LCO electrode is cathodically polarized (s = -11.5 mC cm -2 ), where the anions experience electrostatic repulsion from the electrode (compare triangles and squares in Figure 4c). The LCO surface attracts Li + , which is further stabilized by the water adlayer ( Figure   4d). In parallel, Li + electrostatically attracts anions to form ion pairs. Kosmotropic anions having high charge density are now preferentially associated with small-size Li + ions (Figure 4e), according to the "hard and soft acid and base (HSAB)" concept. The present simulation shows that contact ion pairs (CIPs) are more favorably formed when the anions have more kosmotropic character ( Figure S17).
Along with the large increase in the local ion concentration (including Li + ) with kosmotropic anions, the increased ionic strength near the electrode surface can minimize the number of uncoordinated water molecules, termed 'free water'. This demonstrates the inverse relationship between the density of free water and the kosmotropic propensity of the anions (Figure 4f). Considering that the electrolyte pH is in the mildly alkaline range where water is regarded as the main H + source, the lower free water density induced by the kosmotropic anions conceivably helps to suppress H + insertion and maintain the stability of the electrode.
When the potential rose above 0.2 V, that is, positive s, the C dl increased substantially for all electrolytes the Li + intercalation/extraction process occurs (0.5 ~ 0.8 V vs. Ag/AgCl). If Li + intercalation competes with H + insertion, a high Li + concentration near the LCO electrode is bene cial. Therefore, the ion pair formed by SO 4 2is more likely to reduce the detrimental effect of water.
Based on the insights gained from the above studies, we attempted to improve the cell performance by increasing the concentration of SO 4 2-. The electrochemical potential window in the anodic region was slightly expanded with 3 m Li 2 SO 4 (~130 mV, Figure 5a), implying that more SO 4 2- Figure S22). The 0.5 m Li 2 SO 4 and 1 m LiTFSI-based full-cells both experienced rapid capacity fading over 50 cycles (Figure 5c) owing to the deformation of LNMO and the hydrogen evolution reaction in the salt-in-water system. In comparison, 3 m Li 2 SO 4 delivered 74% capacity retention over 500 cycles, which is superior to that achieved with 3 m LiTFSI (38%) and 9 m LiNO 3 (34%) (Figures 5c-d, Figure S23, and Table S4). The outstanding cyclability attained using 3 m Li 2 SO 4 is attributed to the extensive SO 4 2--based liquid protective layer that harnesses the water molecules. Although the cells with 1 ~ 3 m LiTFSI exhibited higher initial capacity than those with the corresponding Li 2 SO 4 , prompt capacity fading was observed during cycling. In addition, the poor cyclability of 9 m LiNO 3 con rmed the indispensable role of the kosmotropic anions. Although the cell stability with the salt-in-water electrolyte is still inferior to that of WiSE (92% capacity retention with 21 m LiTFSI, Figure S24), this nding provides promise for developing aqueous LIBs with far lower electrolyte concentrations by tailoring the anion species.
In summary, we demonstrated the imperative role of anions in the salt-in-water electrolyte of aqueous LIBs. In the presence of chaotropic anions, severe capacity fading appeared due to deformation of the LCO surface by water and H + insertion. In contrast, sulfate, as a kosmotrope, mitigated these detrimental effects by establishing stronger sulfate-water and sulfate-Li + ion interactions. Atomic-scale multiscale simulations illuminated the higher local concentration of sulfate at the interfacial region of the LCO electrode. The stiff water structures and kosmotropic anion-Li + pairs generated a liquid-phase protective layer at the interface and alleviated the detrimental effect of free water molecules in deforming the electrode structure. This fundamental understanding sheds light on the mechanistics of the improved cell cyclability with 0.5 m Li 2 SO 4 , providing 0.20% capacity-fading rate per cycle for 200 cycles in half-cells and 0.06% fading rate per cycle with 3 m Li 2 SO 4 for 500 cycles in full-cells. The insight into the interfacial structure paves the way for the design of stable, inexpensive, and safe aqueous LIBs by tailoring the anion species and interfacial environments.

Declarations SUPPLEMENTARY INFORMATION
Method, electrochemical and chemical analyses, details of computational data.

NOTES
The authors declare no competing nancial interest.