Lithium titanate hydrates with superfast and stable cycling in lithium ion batteries

Lithium titanate and titanium dioxide are two best-known high-performance electrodes that can cycle around 10,000 times in aprotic lithium ion electrolytes. Here we show there exists more lithium titanate hydrates with superfast and stable cycling. That is, water promotes structural diversity and nanostructuring of compounds, but does not necessarily degrade electrochemical cycling stability or performance in aprotic electrolytes. As a lithium ion battery anode, our multi-phase lithium titanate hydrates show a specific capacity of about 130 mA h g−1 at ~35 C (fully charged within ~100 s) and sustain more than 10,000 cycles with capacity fade of only 0.001% per cycle. In situ synchrotron diffraction reveals no 2-phase transformations, but a single solid-solution behavior during battery cycling. So instead of just a nanostructured intermediate to be calcined, lithium titanate hydrates can be the desirable final destination.

LTO-400 and LTO-500 materials was illustrated in Supplementary Fig. 13b. LTO-400 can deliver more specific capacity (125 mAh g -1 ) in the initial cycle compared to LS and HN, however, its capacity drops from 118 mAh g -1 to 66 mAh g -1 in the first three thousand cycles and then tapers off to 36 mAh g -1 in the following several thousand cycles. Similarly, the capacity of LTO-500 plummets from 148 mAh g -1 to 74 mAh g -1 in the first three thousand cycles and maintains only 51 mAh g -1 after 10,000 cycles. The reason is that Li4Ti5O12 and TiO2 with mainly diffusioncontrolled redox reaction can usually storage more Li-ions than LS with mainly surface-controlled redox reaction, but the Li-ion diffusion coefficient of Li4Ti5O12 and TiO2 is lower than the 2D materials of LS and HN, resulting in the relatively unsatisfactory cycling performances of LTO-400 and LTO-500 electrodes. In the comparison of C 1s spectrum ( Supplementary Fig. 15a), three main peaks at 289.8, 286.5 and 284.8 eV can be assigned to carbon atoms in a three-oxygen environment (CO3-like), one-oxygen environment (CO-like), carbon atom bound only to C or H atoms, respectively. These three carbon species could demonstrate the presence of Li2CO3 and lithium alkyl carbonate (ROCOOLi) well. 24,25 The possible SEI formation mechanisms are as follows: i) intrinsic catalysis by Ti 4+ species during the discharge process, Ti 3+ is oxidized to Ti 4+ , whereby the latter species may have a catalytic influence on the electrolyte decomposition (Ti 4+ + EC/PC/DMC→Ti 3+ + radicals + CO2); ii) electrolyte (EC) reduction from lithiated LTO (2Li + + EC → Li2CO3 / (CH2CH2OLi)2 + C2H4 ). 26,27 In the comparison of F 1s spectrum ( Supplementary Fig. 15b), two peaks at 687.1 and 684.6 eV can be assigned to PVDF and LiF, respectively. For DN which is completely water-free, a small amount of the LiF was observed which could be caused by some possibilities as follows: a) the hydrolysis of LiPF6 (LiPF6+H2O→LiF+POF3+2HF), as water unavoidably exists in a very low concentration (ppm) in the electrolyte; b) LiPF6 salt would decompose by itself during reduction in the charge/discharge cycles (LiPF6→LiF+PF5); c) salt reaction with products such as Li2CO3, resulting from reaction (PF5+Li2CO3→2LiF+POF3+CO2, LiPF6+Li2CO3→3LiF +POF3+CO2, et al). 28,29 The intensity of LiF peak gradually increased from HN to LS, which implied that more structural water might be broken and released into the electrolyte during long-term and super-fast cycling process, resulting in slight hydrolysis of LiPF6. This could be the reason of the slight capacity fading of LS and HN. As a contrast, there shows much more LiF at the surface of LTHs electrodes due to the strong peak of LiF. This result demonstrate that the electrodes with more loosely bound water (such as crystallographic water) would be more likely to cause the decomposition of LiPF6, leading to electrochemical performances worsened rapidly in aprotic electrolyte. charging/discharging process. 26,27 We have designed the following experiments using LiTFSI as electrolyte salts (1M LiTFSI in EC and DMC (1:1 by volume)) to support the opinion mentioned above. As LiTFSI salt is not sensitive to water, therefore if the side reactions will still happen between the electrode and electrolyte, there could be some else decomposition mechanism resulting in unsatisfactory "Coulombic inefficiency" (CI). 30 We further tested some electrochemical performances of HN when using different lithium salts (LiTFSI and LiPF6). The CI of HN electrode with LiTFSI electrolyte (hereinafter refer as HN-LiTFSI, Supplementary Fig. 16) is quite similar with that of HN electrode with LiPF6 electrolyte (hereinafter refer as HN-LiPF6), especially for the first 10 cycles, implying that the main decomposition mechanism may not be the hydrolysis of LiPF6 with trace amount of water.

Supplementary
Besides, the rate performance as well as cyclability of HN-LiTFSI are also similar with HN-LiPF6.
This phenomenon could demonstrate that the SEI formed after several cycles on the surface of HN with LiTFSI electrolyte is as stable as the one formed on the surface of HN with LiPF6 electrolyte, thus both two HN electrodes with different lithium salts could exhibit excellent high-rate capacity and cycling stability. In the same way, we also illustrated electrochemical performances of LS when using LiTFSI as lithium salts (Supplementary Fig. 17). The result and trend are the same as the condition of HN discussed above. Figure 18): The contribution of pseudocapacitity and diffusion-controlled capacity was further revealed by analyzing the cyclic voltammetry data (in Supplementary Figure 18) and can be quantified according to the following Equation (1-2): 31,32

Supplementary Note 9 (Supplementary
where k1v represents pseudocapacitive current (Ic) contributed by interfacial storage and bulk faradic pseudocapacitance; k2v 1/2 represents the (bulk) diffusion current (Id). The coefficients k1 and k2 can be determined by linear fitting to jp v -1/2 vs. v 1/2 in Supplementary Fig. 19a. Therefore, it is possible to calculate the relative contribution of Ic and Id at specific scan rate (v). Figure 19): where Vm is the molar volume (we defined Vm for all three materials as 45.73 cm 3 mol -1 (Li4Ti5O12)

Supplementary Note 10 (Supplementary
for simplification), F is the Faraday constant, A is the total contact area between the electrolyte and the electrode, and w is the Warburg coefficient which was obtained from the Warburg region of impedance response. The w values at different discharge depths (different potentials) can be obtained from the slope of lines in ˊ vs. −1/2 plots (ω is the angular frequency) for the Warburg region. As shown in (Supplementary Fig. 21), the ˊ vs. −1/2 plot for the low frequency Warburg region can be summarized as: The dE/dx obtained from discharge curve illustrates that the slopes of the discharge curve at 25%, 50%, 75% and 100% respectively ( Supplementary Fig. 22). Each sample was activated for 10 cycles between 1.0-2.5V at 50 mA g -1 before test and stopped according to the percentage of the total discharge capacity (as shown in Supplementary Fig. 22). The EIS measurement was conducted right after the cell was stopped, and then it was charged/discharged for 2 complete cycles between 1.0-2.5V at 50 mA g -1 before the next measurement.
The reason why the Li-ion diffusion coefficients at 25%, 50% and 75% are lower than 100% is that 25%-75% of the HN and DN electrodes are associated the two-phase region, in which the total Li-ion insertion process is limited by Li-ion diffusion across the interface. 36 Therefore the Liion diffusion within the two-phase region is much lower than the Li-ion diffusion within the singlephase regions. However, there is a little change of DLi for LS (8.21×10 -9 -2.46×10 -8 cm 2 s -1 ), which might due to the surface-controlled feature of LS (here "surface" is taken to mean in the broad sense, that may include spacing between two adjacent LS monolayers). Figure 25): Note that the LTO in our system reveal much larger lattice changes and we expected that this phenomenon might be ascribed to the size effect 37 Figure 28): Herein, we will evaluate the possibility of in full batteries of HN material by using the concept of "Coulombic inefficiency". 30 In half cells, as the "live Li-ion" in pure lithium counter electrode is infinite compared to the working electrode, the electrochemical performance of working electrode is not affected by the CE or CI at all.

Supplementary Note 14 (Supplementary
However, when it comes to full cells, almost all the "live Li-ion" are from the cathode (i.e. LiFePO4, LiCoO2, etc.), if the CE (half cell) of the anode cannot be higher than 99.5% in less than 10 cycles, one solution is to use the slightly excess cathode as the supplement of "live Li-ion". However, if the CEstabilized is lower than 99.9% after 10 or more cycles, the "live Li-ion" would be consumed rapidly even if the cathode is several hundreds percent excess to the anode. In this situation, it is impossible to get a satisfactory cycling performance in full cells.
Supplementary Figure 28a illustrates calculated CI for HN electrode at 4,000 mA g −1 . The CI value for the first cycle is 10 -1 , and then it dropped to ＜10 -2 after 10 th cycle, followed by fluctuating values distributed between 10 -2 and 10 -4 for the following 10,000 cycles. We note that there are more number of negative (green) CIs than positive (red) CIs, which means if a running-  28b). For the cycling performance at 200 mA g -1 in Supplementary Fig. 28c, 30% of the initial capacity is lost in the first 50 cycles due to SEI growth and some side reactions for the electrodes.
However, the cyclability tends to be quite stable for the next 2,000 cycles with only 10% of the initial capacity lost, indicating that HN is a very promising material as a LIB anode for the reallife energy storage application. It is true that the output voltage of the LiFePO4 vs. HN full battery is somewhat low due to the relatively high potential (vs. Li) of HN compared with graphite anode.
However, the advantages of improved safety over graphite as well as outstanding high-rate and long-cycling performances for lithium titanate hydrates are still obvious in some applications like electric buses, electric vans, smart grids, large-scale storage grids, et al. Table 1): It is noted that for some diffusion-controlled materials (like LiFePO4), the ionic transport through the electrode and electrolyte is rate limiting when charged/discharged at high current density. 42,43 Therefore, if one electrode is thin or with less mass loading (say, less than 0.5 mg cm -2 ), it could exhibit better high-rate performance than the thicker one or with higher mass loading. As a result, it is recommended to deal with the comparison above correctly by considering the loading density of each electrode.