Raising the cycling stability of aqueous lithium-ion batteries

Raising the cycling stability of aqueous lithium-ion batteries by eliminating oxygen in the electrolyte Jia-Yan Luo, Wang-Jun Cui, Ping He and Yong-Yao Xia* Aqueous lithium-ion batteries may solve the safety problem associated with lithium-ion batteries that use highly toxic and flammable organic solvents, and the poor cycling life associated with commercialized aqueous rechargeable batteries such as lead-acid and nickel-metal hydride systems. But all reported aqueous lithium-ion battery systems have shown poor stability: the capacity retention is typically less than 50% after 100 cycles. Here, the stability of electrode materials in an aqueous electrolyte was extensively analysed. The negative electrodes of aqueous lithium-ion batteries in a discharged state can react with water and oxygen, resulting in capacity fading upon cycling. By eliminating oxygen, adjusting the pH values of the electrolyte and using carbon-coated electrode materials, LiTi2(PO4)3/Li2SO4/LiFePO4 aqueous lithium-ion batteries exhibited excellent stability with capacity retention over 90% after 1,000 cycles when being fully charged/ discharged in 10 minutes and 85% after 50 cycles even at a very low current rate of 8 hours for a full charge/discharge offering an energy storage system with high safety, low cost, long cycling life and appropriate energy density. A review of the working mechanisms of commercially available aqueous battery technologies, such as nickel metal hydride (Ni-MH), nickel–cadmium (Ni-Cd) and lead-acid (Pb-acid) batteries has shown that none of them could offer long cycling stab- ility. This should not be surprising because the alloy in Ni-MH is pulverized during the cycling process and Pb-acid and Ni-Cd bat- teries rely upon the dissolution/deposition process with Pb or Cd, which means that the electrodes are not fully reversible. The lithium-ion battery using two intercalated compounds (carbon anode and LiCoO2 cathode) in an organic solution electrolyte was commercialized by Sony in 1990 and is now widely used for cellular phones, notebook-size personal computers, video and digital cameras, and other electronics owing to its long cycling capability and high energy density. The lithium-ion battery has much better cycling stability than Ni-MH, Ni-Cd and Pb-acid batteries because the lithium ion can be reversibly intercalated into a lithium-accepting anode and deintercalated from a lithium-source cathode without destroying the structure of the electrode materials. However, despite the remarkable performance of these organic- based systems, they suffer from the use of highly toxic and flam- mable solvents, which can cause safety hazards if used improperly, such as overcharging or short-circuiting. Recently, numerous lithium-ion battery accidents causing fires and explosions have been reported. Furthermore, non-aqueous electrolytes generally have ion conductivities about two orders of magnitude lower than those of aqueous electrolytes, and the fabrication costs when using organic electrolytes are high. These drawbacks limit their application in large-scale batteries, which require low cost, high safety and long cycling life. An attractive approach to circumvent this problem is to use an aqueous electrolyte for lithium-ion batteries, which adopt a ‘rocking-chair’ concept similar to the organic lithium-ion battery. In 1994, Dahn’s group reported an aqueous lithium-ion battery based on the same technological concepts developed by Sony, in which VO2 was used as a negative electrode and LiMn2O4 as a posi- tive electrode. The cell could operate at an average voltage close to 1.5 V with a specific energy density of 75 W h kg21 based on the total weight of both electrode materials1. With this combination, the safety problem arising from the use of organic electrolytes is fun- damentally resolved, and the rigorous assembly conditions required for non-aqueous lithium-ion batteries can be avoided. However, the cycling life of the VO2 (B)/LiMn2O4 aqueous lithium-ion battery is poor (‘B’ designates a particular crystal form of VO2 (ref. 1)). Following this work, many aqueous lithium-ion batteries such as LiV3O8/LiNi0.81Co0.19O2, LiV3O8/LiCoO2, TiP2O7/LiMn2O4 and LiTi2(PO4)3/LiMn2O4 systems have also been reported 2–4. However all these systems have poor cycling stability similar to that of VO2 (B)/LiMn2O4: the capacity retention is typically less than 50% after 100 cycles. Studies addressing the mechanism of capacity fading during cycling in aqueous lithium-ion batteries have been very limited, and mostly focus on the dissolution of elec- trode materials into the bulk electrolyte5–7. Even the battery systems consisting of insoluble electrode materials, such as LiMn2O4 positive and LiTi2(PO4)3 negative electrodes, have shown fast capacity fading4. Until now, the mechanism responsible for this severe capacity fading upon cycling has not been clarified, and little progress has been made during the past two decades. In the present work, we analysed the stability of electrode materials in aqueous electrolytes extensively, and for the first time we found that, the discharged state of lithium-ion intercalated compounds (LICs) of all negative electrode materials suitable for aqueous lithium-ion batteries reacts with water and O2, with no dependence on the pH value of the electrolyte. The instability of the lithiated LIC is mainly responsible for the capacity fading of aqueous lithium-ion batteries during charge/discharge cycling. By eliminating O2 (using a sealed cell), adjusting the pH values of the electrolyte, and using carbon-coated electrode materials, a LiTi2(PO4)3/LiFePO4 aqueous lithium-ion battery in Li2SO4 aqueous electrolyte exhibits significant improvement in cycling stability. Results and discussion Figure 1 gives the intercalation potential of some electrode materials that could possibly be used for aqueous lithium-ion batteries. The chemical/electrochemical processes of LIC electrodes in aqueous Department of Chemistry, Shanghai Key Laboratory of Molecular Catalysis and Innovative Materials, Institute of New Energy, Fudan University, Shanghai 200433, China. *e-mail: yyxia@fudan.edu.cn ARTICLES PUBLISHED ONLINE: 8 AUGUST 2010 | DOI: 10.1038/NCHEM.763 NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry 1 © 2010 Macmillan Publishers Limited. All rights reserved. solutions are much more complicated than those in the organic electrolytes. Many side reactions are involved, such as electrode materials reacting with water or O2, proton co-intercalation into the electrode materials parallel to the intercalation of lithium ions, H2/O2 evolution reactions, and the dissolution of electrode materials in water. It has been demonstrated that materials with a voltage versus Li/Liþ of greater than 3.3 V are basically stable7–10. When acting as negative electrodes for aqueous lithium-ion batteries, the intercalation potential of lithium-ions is generally below 3.3 V versus Li/Liþ. First, we considered the stability of the electrode materials in the presence of both H2O and O2, because the aqueous battery customa- rily operates in air. The following reaction may occur: Li(intercalated)+ (1/4)O2 + (1/2)H2O⇔ Li+ +OH− The potential of a LIC, V(x), can be calculated with the following equation7,9,11: V(x) = − 1 e uintLi (x) − u0Li ( ) (1) where uintLi (x) is the chemical potential of intercalated Li in LIC, u0Li is the chemical potential of Li in Li metal and e is the magnitude of the electron charge. The potential of a LIC, V(x), in equilibrium with O2 and H2O at a particular pH can be calculated with equation 2. (The details for the calculation process are shown in the Supplementary Information.) V(x) = 4.268− 0.059 pH (V) (2) Notably in the presence of O2, no materials can be used as negative electrodes for aqueous lithium-ion batteries regardless of the pH of the electrolyte, according to equation 2. The lithium-ion intercala- tion potential of the negative electrodes for aqueous lithium-ion bat- teries is generally below 3.0 V versus Li/Liþ, whereas the equilibrium voltage is 3.85 V at pH 7 and 3.50 V at pH 13. This means that the reduction state of all negative electrode materials would theoretically be chemically oxidized by the O2 and H2O rather than undergoing the electrochemical redox process. We speculate that, in the presence of O2, the discharged state of the LICs of any negative electrode candidates for the aqueous lithium-ion batteries would react with H2O and O2 for any pH value of the electrolyte, resulting in capacity fading upon cycling. Therefore, aqueous lithium-ion batteries can not work sustainably in the presence of O2. In the absence of O2, the lithium-ion intercalated compounds may react with H2O as in following reaction, which is similar to the result presented by Dahn8: Li(intercalated)+H2O⇔ Li+ + OH− + (1/2)H2 The calculated potential of the LIC V(x) in equilibrium with H2O at a particular pH is as follows: V(x) = 3.039− 0.059 pH (V) (3) H2O may also chemically oxidize the reduction state of some nega- tive materials for aqueous lithium-ion batteries. Theoretically, we can determine whether or not the lithium-ion intercalation at a par- ticular pH of electrolyte is stable, and we can also adjust the pH of the electrolyte to guarantee the stability of the electrodes. For example, as the lithium-ion intercalation potential in LiTi2(PO4)3 is 2.45 V versus Li/Liþ, it is theoretically not stable in pH 7 aqueous solutions (2.626 V versus Li/Liþ of equilibrium voltage). Chemical stability can however be obtained in the aqueous solution in the absence of O2 by adjusting the pH of the aqueous solution electrolyte to more than 10—for example at pH 13, the equilibrium voltage is 2.272 V versus Li/Liþ (see Supplementary Fig. S1 and Table S1). Second, the positive electrode materials are generally stable in water. However, protons may be co-intercalated into the electrode materials parallel to the intercalation of lithium ions in the aqueous solution electrolyte. The proton intercalation depends on both the crystal structure and the pH of the electrolyte. It has been reported that spinel Li12xMn2O4, and olivine Li12xFePO4 do not encounter such proton insertion, whereas delithiated layered Li12xCoO2, Li12xNi1/3Mn1/3Co1/3O2, and so on, show a significant concentration of protons in the lattice during deep lithium extrac- tion with a low pH electrolyte12,13. However, this can be addressed easily by modulating the potential for proton intercalation through adjusting the pH of the electrolyte; for example, in LiCo1/3Ni1/3Mn1/3O2 stable lithium-ion intercalation is possible in a solution over pH 11, and in LiCoO2 over pH 9 (ref. 14). It should be noted that LiFePO4 decomposes in strong alkaline sol- utions, but the decomposition can be slowed down by carbon coating the material15. The LiFePO4 used here contains 15 wt% of coating carbon and shows excellent stability (see later discussion). Third, the H2/O2 evolution reaction in aqueous electrolyte is a basic factor that needs to be considered, because the capacity of the electrode materials should be used as much as possible before electrolyte decomposition. Thermodynamically, the aqueous elec- trolyte shows an electrochemical stability window of 1.23 V. Kinetic effects may expand the stability limit to 2 V. For instance, Pb-acid batteries can have an output voltage of 2.0 V. In principal, the use of most electrode materials depends on the pH value of the electrolyte. Lithium ions in LiMn2O4 can be fully extracted at pH 7, but only half of the lithium ions can be extracted at a pH greater than 9 before O2 evolution. LiFePO4 can be used over a broader pH range from 7 to 14. Fourth, the electrode materials should be insoluble in water. The dissolution scales mainly with surface area. For example, the vanadium oxides VO2, LiV3O8, LiV2O5, and so on, are typically prepared at low temperature with a relatively large surface area and hence they are not good choices as electrode materials in –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5 4.5 4.0 3.5 3.0 2.5 2.0 1.5 Potential (V versus Li +/Li) pH 147 Po te nt ia l (V ve rsu s N HE ) LiCoO2 LiNiO2 LiMn2O4 LiV3O8 LiFePO4 0 NASICON AC O 2 evolution H 2 evolution γ -FeOOH VO2 LiV2O5 LiNbO5 Figure 1 | The intercalation potential of some electrode materials that could possibly be used for aqueous lithium-ion batteries. Left: O2/H2 evolution potential versus NHE for different pH in 1 M Li2SO4 aqueous solution. Right: lithium-ion intercalation potential of various electrode materials versus NHE and Li/Liþ. Theoretically, an aqueous lithium-ion battery can be assembled by combining a lower potential lithium-accepting anode and a higher potential lithium-source cathode within the O2/H2 evolution potential range. AC, activated carbon; NASICON, materials with NASICON structure. ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.763 NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry2 © 2010 Macmillan Publishers Limited. All rights reserved. aqueous lithium-ion batteries; electrode materials with small surface areas are preferred. Considering all the factors above, we chose LiTi2(PO4)3 as the negative electrode in our aqueous lithium-ion battery because it has relatively satisfactory capacity, excellent operating potential in an aqueous electrolyte (about –0.5 V versus natural hydrogen elec- trode (NHE)) and a flat voltage plateau. Moreover, it has a relatively small surface area of about 0.7 m2 g21 because it is normally pre- pared at high temperature5. The cyclic voltammograms of LiTi2(PO4)3 at various operating conditions are shown in Fig. 2. One pair of redox peaks was observed for LiTi2(PO4)3 tested under all conditions, which agrees with the lithium-ion intercalation/de-intercalation process in an organic electrolyte. In the presence of O2, the fast capacity fading upon cycling was detected in both pH 7 and pH 13 aqueous sol- utions. However, in the absence of O2, LiTi2(PO4)3 undergoes a different process. Even though there is some irreversibility in the first cycle, which is mainly due to the residual O2 in the testing cells, the good overlap of the anodic and cathodic peaks in the subsequent cycles indicates that the LiTi2(PO4)3 electrode is stable. This is consistent with the calculations. The reason for the capacity fading of LiTi2(PO4)3 in the aqueous electrolyte in the presence of O2 is that the reduction state Li32xTi2(PO4)3 is chemically oxidized by the O2 instead of under- going the electrochemical oxidation process. This can be illustrated directly with the charge/discharge coulombic efficiency of LiTi2(PO4)3 in the aqueous electrolyte in the presence and absence of O2. It can be seen from Fig. 3 that when cycled at a rate of 4C (where 1C corresponds to complete discharge in 1 hour), the charge/discharge coulombic efficiency of LiTi2(PO4)3 in the aqueous electrolyte in the absence of O2 was 99%, which was higher than that in the presence of O2 (92%). This discrepancy became more obvious when LiTi2(PO4)3 was cycled at a rate of 1C: the coulombic efficiency was 98% versus 77%. This result confirmed that the reduction state Li32xTi2(PO4)3 can be chemically oxidized to LiTi2(PO4)3 and other impurities by O2 during the discharging of LiTi2(PO4)3. –6 –4 –2 0 2 4 –1.0 –0.8 –0.6 –0.4 a c b d Voltage (V versus SCE) –1.0 –0.8 –0.6 –0.4 Voltage (V versus SCE) –1.0 –0.8 –0.6 –0.4 Voltage (V versus SCE) –1.0 –0.8 –0.6 –0.4 Voltage (V versus SCE) Cu rre nt (m A) –6 –4 –2 0 2 4 Cu rre nt (m A) –6 –4 –2 0 2 4 Cu rre nt (m A) –6 –4 –2 0 2 4 Cu rre nt (m A) Figure 2 | Cyclic voltammograms of LiTi2(PO4)3 at the 1st, 2nd, 3rd, 5th and 10th cycles. a, pH 7 in the absence of O2. b, pH 13 in the absence of O2. c, pH 7 in the presence of O2. d, pH 13 in the presence of O2. Scan rate¼0.3 mVs21. Fast capacity fading upon cycling was detected in both pH 7 and pH 13 aqueous solutions in the presence of O2. However, in the absence of O2, good stability of the LiTi2(PO4)3 electrode was observed (the arrows indicate the direction of cycle 1 to cycle 10). 0 200 400 600 800 In the absence of O2 In the presence of O2 In the absence of O2 In the presence of O2 Time (s) 0 1,000 2,000 3,000 4,000 –0.8 –0.6 –0.4 –0.2 0.0 Time (s) Vo lta ge (V ve rsu s S CE ) b –0.8 –0.6 –0.4 –0.2 0.0 Vo lta ge (V ve rsu s S CE ) a Figure 3 | Typical charge/discharge curves of LiTi2(PO4)3 at 4C and 1C charge/discharge rates in the presence/absence of O2. a, At a 4C rate, the coulombic efficiency of LiTi2(PO4)3 in the aqueous electrolyte was 99% in the absence of O2 and 92% in the presence of O2. b, This discrepancy in coulombic efficiency became more obvious when cycled at a 1C rate—the coulombic efficiency was 98% in the absence of O2 versus 77% in the presence of O2. This confirms that the reduction state Li32xTi2(PO4)3 can be chemically oxidized. NATURE CHEMISTRY DOI: 10.1038/NCHEM.763 ARTICLES NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry 3 © 2010 Macmillan Publishers Limited. All rights reserved. These results were further confirmed by the open-circuit voltage test in Fig. 4. The process was carried out as follows. LiTi2(PO4)3 was first discharged to –0.85 V at a current rate of 1C. The open circuit voltage of Li32xTi2(PO4)3 was then measured as a function of time. It was clearly observed that if the Li32xTi2(PO4)3 was exposed to O2 the open circuit voltage could only be sustained at the equilibrium potential for about 15 hours. In the absence of O2, however, it could last for about 10 days. A similar phenomenon was also found in other electrode materials, such as LiV3O8 (see Supplementary Fig. S2). This result again strongly confirmed that the reduction state Li32xTi2(PO4)3 can be chemically oxidized by O2 during the discharge process of LiTi2(PO4)3. This was also con- firmed by the charge/discharge test of LiTi2(PO4)3/LiFePO4 in an open system at very small current rate (C/8, 8 h rate). It was found that the cell was not rechargeable in the presence of O2 as the speed of the oxidation process for the reduced state Li1þxTi2(PO4)3 by O2 and H2O is much faster than that of the lithium intercalation (see Supplementary Fig. S3). It is worthmentioning that even though the reduction potential of water at pH 7 (2.63 V versus Li/Liþ) is slightly higher that the potential of LiTi2(PO4)3 (2.45 V versus Li/Liþ), LiTi2(PO4)3 did not show obvious capacity fading because the potential discrepancy DE between the water reduction potential and the lithium intercalation potential LiTi2(PO4)3 is only about 0.2 V, which is much lower than the DE of 1.4 V between the O2 reduction potential (3.85 V versus Li/Liþ) and the lithium intercala- tion potential LiTi2(PO4)3 (2.45 V versus Li/Li þ) at pH 7.We specu- late that the reaction between the reduction state Li32xTi2(PO4)3 and O2 is the primary cause of the capacity fading, whereas the reaction between the reduction state Li32xTi2(PO4)3 and H2O is secondary. As for the positive electrodes for aqueous lithium-ion batteries, layered Li12xCoO2, Li12xNi1/3Mn1/3Co1/3O2, and so on, may encounter proton insertion at deep lithium extraction; spinel Li1-xMn2O4 cannot be used in an electrolyte with a high pH because of its relatively high potential. Olivine Li12xFePO4 is there- fore a comparatively suitable positive electrode material owing to its appropriate potential, high capacity, low cost, safety, low environ- mental impact and high reversibility in aqueous electrolytes. The cyclic voltammograms show that the carbon-coated LiFePO4 can have a long stability in aqueous electrolyte (see Supplementary Fig. S4). Thus an aqueous lithium-ion battery consisting of a LiFePO4 positive electrode and a LiTi2(PO4)3 negative electrode in a 1 M Li2SO4