Long-lived Aqueous Rechargeable Lithium Batteries Using Mesoporous LiTi2(PO4)3@C Anode

The instability of anode materials during cycling has been greatly limiting the lifetime of aqueous rechargeable lithium batteries (ARLBs). Here, to tackle this issue, mesoporous LiTi2(PO4)3@C composites with a pore size of 4 nm and a large BET surface area of 165 m2 g−1 have been synthesized by a novel two-step approach. The ARLB with this type of LiTi2(PO4)3@C anode, commercial LiMn2O4 cathode and 2 M Li2(SO4) aqueous solution (oxygen was removed) exhibited superior cycling stability (a capacity retention of 88.9% after 1200 cycles at 150 mA g−1 and 82.7% over 5500 cycles at 750 mA g−1) and excellent rate capability (discharge capacities of 121, 110, 90, and 80 mAh g−1 based on the mass of LiTi2(PO4)3 at 30, 150, 1500, and 3000 mA g−1, respectively). As verified, the mesoporous structure, large surface area and high-quality carbon coating layer of the LiTi2(PO4)3@C composite contribute to the breakthrough in achieving excellent electrochemical properties for ARLB.

, and LiMn 2 O 4 11 have been well studied as the cathodes for ARLBs. The anode for ARLBs requires the electrode materials with a Li + intercalation potential of 2~3 V vs. Li + /Li 12 . There are only several kinds of suitable candidates, e.g., vanadates and LiTi 2 (PO 4 ) 3 . The first ARLB of VO 2 //LiMn 2 O 4 reported by Dahn et al. 13 can just cycle for 25 cycles. Since then, the ARLB systems such as LiV 3 4 and new aqueous battry systems have been constructed 2, [14][15][16][17][18][19] . Many of these systems, however, only displayed limited cycling stability due to the vanadium dissolution in aqueous solution and degradation of crystal structure, especially at a low current density 5 . LiTi 2 (PO 4 ) 3 /C has shown the potential as anode for ARLB with relatively high power density and good cycling stability. By eliminating the soluble oxygen in Li 2 SO 4 solution, LiTi 2 (PO 4 ) 3 //LiFePO 4 ARLB constructed by Xia et al. 20 demonstrated a 1000 cycle life at a high current rate of 6C. However, the cycling stability of such ARLB system at low rates was still insufficient (85% after 50 cycles at a current rate of 8 hrs for a full charge/discharge test). Hence, a breakthrough in cycling life, particularly at a lower current density is urgently required for further applications of ARLBs.
There has been a consensus that LiMn 2 O 4 and LiFePO 4 could be used as advanced cathodes for ARLBs. As reported by Wu et al. 21 , porous LiMn 2 O 4 nanograins could be stably cycled up to 10000 cycles with a capacity retention of 93% at a rate of 9C. In contrast, the instability of anode mainly caused by H 2 O attacking, dissolution of active materials and repetitive volume expansion has been remaining as a key issue for ARLBs. Pristine LiTi 2 (PO 4 ) 3 often shows a low electronic conductivity 22 , which could be greatly improved by reducing the particle size to nanoscale thanks to their shortened electron/Li ions diffusion paths 23 . Meanwhile, coating strategy with high quality carbon layer could not only significantly enhance the conductivity of materials but also protect active material from electrolyte corrosion, resulting in better cycling stability 24 . It is clear that the characteristics of carbon layer (e.g., content, thickness, uniformity and structure), which can significantly affect the coating quality, generally depend on the selected carbon sources and the coating methods 25,26 . In our previous work, LiTi 2 (PO 4 ) 3 with high-quality carbon coating has been obtained and it demonstrated excellent cycling life with a capacity retention of 90% after 300 cycles at 0.2C 27 .
In the present work, LiTi 2 (PO 4 ) 3 @C composite with mesoporous structure and homogeneous carbon coating was successfully fabricated by a solvothermal process combined with an annealing treatment. It is suggested that mesoporous composite could not only well accommodate the volume change during cycling by abundant pores but also provide reduced lengths for both mass and charge transports thanks to this unique structure 23 . More importantly, high quality carbon coating was achieved by using phenolic resin as carbon source and an in situ coating strategy. The as-prepared LiTi 2 (PO 4 ) 3 @C composite exhibits superior electrochemical properties. The strategies proposed in this paper are absolutely template free and very facile for practical application.

Results
The X-ray diffraction (XRD) pattern (a) and transmission electron microscopy (TEM) image (b) of the precursor obtained by solvothermal process are shown in Figure S1, in which a very low crystallinity and an average particle size of 5 nm can be seen. The XRD pattern of as-prepared material after being sintered with carbon source is displayed in Fig. 1a. The diffraction peaks can be well indexed into LiTi 2 (PO 4 ) 3 phase with a rhombohedral NASICON type structure and a R3c space group (JCPDS#35-0754). The measured lattice parameters a = 0.8464 nm and c = 2.1442 nm are in good agreement with the previous reports 28, 29 . The calculated average crystal size of LiTi 2 (PO 4 ) 3 from XRD pattern based on the Debye-Scherrer equation is 32.7 nm. The high resolution X-ray photoelectron spectroscopy (XPS) spectrum of Ti ( Figure S2) confirms the existence of Ti 4+ in LiTi 2 (PO 4 ) 3 @C 30 . Figure 1a (inset) shows the crystal structure of LiTi 2 (PO 4 ) 3 . The three-dimensional (3D) anionic framework is formed by corner-sharing PO 4 tetrahedra and TiO 6 octahedra, leaving large interconnected channels which can be occupied by Li ions 31 . This rigid 3D crystal structure could be a promising host for Li + insertion/extraction. The TEM image in Fig. 1b indicates that LiTi 2 (PO 4 ) 3 @C is composed of individual particles with a size range of 30-50 nm, which is close to the primary crystalline grain size (32.7 nm) obtained from Debye-Scherrer equation. Note that slight agglomeration takes place. The high resolution TEM (HRTEM) image ( Fig. 1c) of the composite reveals clear lattice fringes with many cavities (as marked by red oval), indicating the existence of mesopores. The appearance of mesopores may originate from the assembly and recrystallization process of ultra-fine precursor nanoparticles (5 nm, Figure S1) and the decomposition of phenolic resin into carbon 32 . The study of detailed formation mechanism is still under way. The N 2 adsorption-desorption isotherm of as-prepared LiTi 2 (PO 4 ) 3 @C is shown in Fig. 1e, from which it can be seen that the LiTi 2 (PO 4 ) 3 @C shows a typical IV isotherm and has a large Brunauer-Emmett-Teller (BET) surface area of 165 m 2 g −1 . The large surface area may be due to its mesoporous morphology (inset in Fig. 1e) and the carbon coating. A narrow size distribution of 4 nm is observed, in good agreement with the HRTEM result (Fig. 1c). It should be noted that high surface area and mesoporous structure can significantly improve the electrode/electrolyte contact, facilitate the Li ions transport, and enhance the utilization efficiency of the material. More importantly, it is the mesoporous structure that enables is able to accommodate strain/stress during the Li ion insertion/extraction process 21 .
The uniformity and structure features of carbon coating were also investigated. As seen from Fig. 1d, a uniform carbon layer with a thickness of ca. 5 nm can be observed on the edge of particle. The Fast Fourier Transform Algorithm (FFT) image further confirms its amorphous nature. The carbon coating with a thickness of 4-8 nm on electrode materials could reach a good balance of e − conductivity and Li + diffusion, promising superior electrochemical properties. The regular lattice fringe and its corresponding FFT image validate the crystal nature of as-prepared LiTi 2 (PO 4 ) 3 . The inter-planar spacing deduced from the Fig. 1d is 0.42 nm, agreeing well with the d-spacing of the (104) plane of rhombohedral LiTi 2 (PO 4 ) 3 . The carbon content of LiTi 2 (PO 4 ) 3 @C is measured to be 12.3wt% by DSC/TG curve (Fig. 1f). To gain an insight into the structure of the carbon layer, Raman spectroscopy was performed ( Figure S3). The two strong bands around 1330 and 1600 cm −1 could be attributed to the inplane vibrations of disordered amorphous carbon (D band) and crystalline graphic carbon (G band), respectively. The relatively low intensity of D-band to G-band (I D /I G = 0.86) value confirms a certain degree of graphitization of carbon, which is beneficial to the improvement of electrochemical properties for carbon coated composites 33 . Furthermore, the scanning transmission electron microscope-energy dispersive spectrometer (STEM-EDS) elemental mapping (Fig. 2) demonstrates that the Ti, P, O and C atoms are uniformly distributed, which unambiguously reveals the uniformity of carbon coating. That is, a homogeneous and high-quality carbon layer was successfully coated on the surface of LiTi 2 (PO 4 ) 3 .
Lithium intercalation and deintercalation behavior of LiTi 2 (PO 4 ) 3 @C and LiMn 2 O 4 electrodes in aqueous electrolyte were investigated by CV measurement (Fig. 3a). The LiTi 2 (PO 4 ) 3 @C demonstrates four reduction peaks (ca.  work as the cathode in consideration of its relatively high intercalated potential, low cost and excellent cycling stability in aqueous electrolyte. Good lithium insertion/extraction behavior is also observed in Fig. 3a,b depicts the first two CV curves of LiTi 2 (PO 4 ) 3 @C//LiMn 2 O 4 ARLB. In the first cycle, there are six main oxidation peaks (ca. 0.95 V, 1.09 V, 1.46 V, 1.52 V, 1.60 V and 1.70 V, respectively) and four reduction peaks (ca. 1.03 V, 1.37 V, 1.47 V and 1.58 V, respectively). While the oxidation peaks at 0.95 V and 1.60 V disappear in the second cycle, which may have a relationship with the structure rearrangement of LiTi 2 (PO 4 ) 3 @C anode and the details will be discussed in Figure S6. Note that there is no obvious peak corresponding to the evolution of hydrogen or oxygen, consistent with the high Coulombic efficiency ( Fig. 3d-f). The rate performance of LiTi 2 (PO 4 ) 3 @C is shown in Fig. 3c. It exhibits a discharge capacity of 121 mAh g −1 (based on the mass of LiTi 2 (PO 4 ) 3 ) at 30 mA g −1 and 90 mAh g −1 at 1500 mA g −1 , respectively. When the current density is increased to 3000 mA g −1 , a discharge capacity of 80 mAh g −1 is retained with apparent charge/discharge plateaus. The excellent rate capability may originate from the large surface area, abundant mesostructure and high quality carbon coating, which contribute to significantly improved electrode/electrolyte contact area and enhanced conductivity 24,34,35 . Long-term cycling stability of ARLB at various rates were performed to evaluate the cycling stability of the mesoporous LiTi 2 (PO 4 ) 3 @C composite. As seen in Fig. 3f, the LiTi 2 (PO 4 ) 3 @C delivers an ultralong cycling life of 5500 cycles with a capacity retention of 82.7% at a current density of 750 mA g −1 . More importantly, the as-prepared material also shows superior cycling stability at relatively low current densities. At 30 mA g −1 , the electrode delivers a discharge capacity of 118 mAh g −1 , and no capacity fading is observed after 100 cycles (Fig. 3d). An initial discharge capacity of 108 mAh g −1 and a capacity retention of 88.9% after 1200 cycles are also illustrated at 300 mA g −1 in Fig. 3e. The superior cycling stability is further confirmed by the performance at extreme high current density (1500 mA g −1 , Figure S4). These results demonstrate clearly that LiTi 2 (PO 4 ) 3 @C//LiMn 2 O 4 can be tolerant to various charge/discharge current densities. Poor cycling stability at low current densities is still a critical challenge for ARLB and there are no very clear explanations available so far. It is speculated that the crystal deterioration of electrode, reaction between electrode materials and water or O 2 and decomposition of water may be the main causes 2,12 . The details will be discussed later.
To our best knowledge, the cycling performance of ARLB here has been advanced to a new level, which is much superior to all the reported ARLBs using vanadium oxides, vanadates or LiTi 2 (PO 4 ) 3 /C as anode materials to date (see Table S1) 2,12,14,15 . This is a breakthrough for ARLB in term of the cycling life, particularly at a low current density. Note that the ARLB with such superior electrochemical performance can certainly meet the demands of various practical applications. Low Coulombic efficiency due to the decomposition of water and the interaction between aqueous electrolyte and electrode surface 2,12,20 , is considered as an important origin of capacity fading for ARLB. As demonstrated, the ARLB here can deliver very high Coulombic efficiency at various current densities (~94% at 30 mA g −1 , > 99% at 150 mA g −1 , 750 mA g −1 , 1500 mA g −1 ) as shown in Fig. 3(d-f), and Figure S4, in good accordance with the superior cycling performance.

Discussion
Structure deterioration, electrode pulverization, and detachment of active material from the conducting environment resulting from the repetitive volume expansion/shrinkage during the Li insertion/extraction are considered as great challenges for long-lifetime battery 36 . For ARLB, H 2 O attacking which leads to the decrease of electrode surface integrity and dissolution of surface active materials is also a fatal cause for capacity fading 28,37 . Wang et al. 38 confirmed that the crystalline structure of Li x V 2 O 5 became nearly amorphous after 40 cycles in ARLB. The formation of new compounds was also considered to be the cause for capacity fading of TiP 2 O 7 by Chen and his co-workers 39 . Caballero et al. 37 considered the dissolution of electrode material as the origin of capacity fading for ARLB. Therefore, the structures of LiMn 2 O 4 cathodes ( Figure S5) and LiTi 2 (PO 4 ) 3 @C anodes (Fig. 4a) after different cycles (1, 2, 100, 3000 and 5000) were examined by XRD. Same as reported in the references, the LiMn 2 O 4 cathodes used here show good structure stability in ARLB ( Figure S5). To find out the reasons for such good structure stability, the structure and surface morphology evolution of LiTi 2 (PO 4 ) 3 @C anodes were investigated in details. As the cycling proceeds, the intensities of diffraction peaks located at 2θ = 28.7°, 29.5°, 31.0° and 32.6° decrease gradually. However, there are no new impurity peaks for the electrodes after different cycles in comparison with that after 1 cycle, implying excellent structure stability of LiTi 2 (PO 4 ) 3 @C anode. Note that there is a slight difference for the XRD patterns of cycled LiTi 2 (PO 4 ) 3 @C electrode in comparison with the LiTi 2 (PO 4 ) 3 @C powder 40 . With this regard, XRD patterns of LiTi 2 (PO 4 ) 3 @C electrodes at different states (marked as a-n) in the first two cycles are given in Fig. 5. It should be noted that the diffraction lines at 28.7°, 29.5°, 31.0° and 32.6° in mark a are much weaker than that at 24.4°. During the first charge process (a-f), the intensities of these lines increase obviously with the rising of cell voltage and gradually become the main ones. In the following process (g-n), their diffraction intensities change slightly. Compared with those in the first charge process, the intensities of the diffraction lines at 28.7°, 29.5°, 31.0° and 32.6° in the second charge process are different, implying a structure rearrangement of LiTi 2 (PO 4 ) 3 @C after the first cycle. Structure information is consistent with the first two CV curves in Fig. 3b. The CV curves ( Figure S6) of LiTi 2 (PO 4 ) 3 @C//LiMn 2 O 4 after 5000 cycles further confirms good Li ion insertion/extraction kinetics after long-term cycling. The surface microstructures of LiTi 2 (PO 4 ) 3 @C Before disassembling, each cell was charged to 1.6 V and then kept at that voltage for 2 h; SEM images of LiTi 2 (PO 4 ) 3 /C electrode after 5 cycles (b), 100 cycles (c), 1000 cycles (d) and 5000 cycles (e) at 750 mA g −1 ; (f) Schematic illustration of the tentative Li ion insertion mechanism in mesoporous LiTi 2 (PO 4 ) 3 with homogeneous carbon coating layer, which can efficiently buffer the volume expansion and avoid H 2 O attacking during repetitive Li + insertion/extraction. electrodes after 5, 100, 1000 and 5000 cycles at 750 mA g −1 are compared in Fig. 4. The LiTi 2 (PO 4 ) 3 @C electrode surface after 5000 cycles still remains intact in comparison with that after 5 cycles, implying the negligible effect of H 2 O attacking. That is, a relatively stable electrode surface and effective suppression of active materials dissolution have been achieved for the as-prepared LiTi 2 (PO 4 ) 3 @C. Figure S7 demonstrates a slight increase of R ct after 5000 cycles, which is in agreement with the capacity fading for ARLB. Generally, the possible capacity fading mechanisms of bare LiTi 2 (PO 4 ) 3 and LiTi 2 (PO 4 ) 3 with heterogeneous carbon coating are illustrated in Figure S8, demonstrating the inferior cycling stability during the cycling process. Hence, according to the discussions above, a tentative Li ion insertion mechanism in mesoporous LiTi 2 (PO 4 ) 3 with homogeneous carbon coating layer is proposed in Fig. 4f. That is, the mesoporous structure confirmed to be more stable than the more common ones (e.g., bulk, nanoparticle) 23,35,41 , could provide enough void space to accommodate the volume expansion during cycling and the outer high-quality carbon coating layer could withstand the attacking of H 2 O, thus resulting in stable crystal structure and electrode surface during the long term cycling (Fig. 4f).
In summary, we have developed a solvothermal method accompanied with an advanced carbon coating strategy to synthesize mesoporous LiTi 2 (PO 4 ) 3 @C composite. When used as an anode for ARLB, the electrode delivered an ultra-long cycling life up to 5500 cycles at 750 mA g −1 . Even at a relatively low current density of 30 mA g −1 , no obvious capacity fading was observed after 100 cycles. This is a breakthrough in the cycling stability of ARLB at both high and low current densities, which should be mainly ascribed to the high performance LiTi 2 (PO 4 ) 3 @C anode. The mesoporous structure, large surface area, high-quality carbon coating layer and the stable 3D crystal structure have been verified as important factors. By virtues of its superior electrochemical performance, the mesoporous LiTi 2 (PO 4 )@ C composite prepared in the present study could be considered as a very promising candidate as an anode for ARLB.

Methods
Synthesis of LiTi 2 (PO 4 ) 3 @ C composite. All the starting materials were analytically pure grade and directly used without any purification. A novel two-step strategy involving a solvothermal method and a following carbon coating process was employed. Certain amounts of lithium hydroxide, titanium sulfate and ammonia phosphate (NH 4 H 2 PO 4 ) with the molar ratio of 2.7: 2: 4.5 were dissolved in ethylene glycol in advance, respectively. The lithium hydroxide and ammonia phosphate solution were first mixed quickly and stirred for 3 h. The titanium sulfate ethylene glycol solution was then gradually added into the mixed solution. After stirring for 0.5 h, the suspension was transferred into a 100 ml Teflon lined stainless steel autoclave. The autoclave was sealed and heated at 160 °C for 10 h and then cooled to room temperature naturally. The white precipitates were collected by centrifugation, and washed with distilled water several times and then dried at 80 °C overnight. Following that, 0.3 g of the white precursor was dispersed well in distilled water in the presence of sodium dodecylsulfonate (10 g/L) and 0.2 g of phenolic resin powder was dissolved in absolute ethanol. The phenolic resin solution was dropwise added into the suspension of LiTi 2 (PO 4 ) 3 precursor. The mixed suspension was heated to 50 °C on a hotplate with stirring till the ethanol evaporated. In this way, the phenolic resin could be covered well on the surface of LiTi 2 (PO 4 ) 3 precursor. After the centrifugation and washing several times by distilled water, the obtained solid was dried at 80 °C overnight and then calcined at 700 °C for 5 h with a ramping rate of 5 °C/min in a mixed flow of H 2 /Ar (5:95, v/v).
Characterizations. All X-ray diffraction (XRD) data were examined by the X-ray diffractometer (Dandong Haoyuan, DX-2700) utilizing a Cu-Kα 1 source with a step of 0.02°. XRD measurement of electrodes was different from the examination of the powder. After being washed with distilled water and dried for several hours, the whole electrode consisting of active material, Super P carbon and polytetrafluoroethylene (PTFE) was directly used to perform the XRD measurements and it is worthy to note that no signal of stainless steel mesh was observed probably due to the thick electrode film. Each cell was charged to 1.6 V and kept at that voltage for 2 h before disassembling. The XPS patterns were collected using Al Kα radiation at a voltage of 12 kV and current of 6 mA. Charging effect was corrected by adjusting the binding energy of C1s peak from carbon contamination to 284.5 eV. Microstructural studies of electrodes after different cycles were conducted using a Nova NanoSEM 230 SEM. TEM, high resolution TEM (HRTEM) images and STEM-EDS elemental mapping of as-prepared LiTi 2 (PO 4 ) 3 @C powders were obtained using a FEI Tecnai G2 F20 S-TWIX TEM. The BET surface area of the samples was detected by nitrogen adsorption/desorption at − 196 °C using a Builder SSA-4200 apparatus. The pore size distributions for LiTi 2 (PO 4 ) 3 @C were obtained by the Barrett-Joyner-Halenda (BJH) method. The XPS fitting was performed using XPSPEAK software, and the crystal structure of LiTi 2 (PO 4 ) 3 was drawn by Diamond 3.2.
Electrochemical measurements. The used LiMn 2 O 4 was provided by Hunan Reshine New Material Co., Ltd. The LiTi 2 (PO 4 ) 3 @C and LiMn 2 O 4 electrodes were made in a similar way. Tested electrodes were obtained by pressing a mixture of the active material, Super P carbon and PTFE in a weight ratio of 80:10:10 using distilled water as solvent on a stainless steel mesh and then dried at 110 °C for 8 h. Cyclic voltammetry (CV) of LiTi 2 (PO 4 ) 3 @C anode and LiMn 2 O 4 cathode was performed using a three electrode system, respectively, where the tested electrode was used as working electrode, platinum sheet electrode as the counter electrode and saturated calomel electrode (SCE, 0.242 V vs. SHE: standard hydrogen electrode) as reference electrode. CV test was investigated at room temperature using an electrochemical station (CHI660D). The CR2016 coin-type cells were constructed by using LiMn 2 O 4 electrode as cathode, LiTi 2 (PO 4 ) 3 @C electrode as anode, 2 mol L −1 Li 2 SO 4 as electrolyte. Excessive LiMn 2 O 4 , with cathode/ anode mass ratio of (1.5~2.0)/1 was designed for exactly evaluating the electrochemical properties of LiTi 2 (PO 4 ) 3 @C. The Li 2 SO 4 electrolyte was pre-treated by the flowing argon injection into the solution to eliminate the soluble oxygen. Charge and discharge tests were conducted under a desired current density by a Neware battery testing system (CT-3008W) at room temperature. Electrochemical impedance spectroscopy (EIS) was recorded by a Princeton workstation (PARSTAT2273, EG&G, US) over the frequency range from 100 kHz to 10 mHz with an amplitude of 5 mV. Before testing, the measured cell was charged to 1.6 V at 30 mA g −1 , and then kept for a period of time to reach a stable state.