Defective Ti2Nb10O27.1: an advanced anode material for lithium-ion batteries

To explore anode materials with large capacities and high rate performances for the lithium-ion batteries of electric vehicles, defective Ti2Nb10O27.1 has been prepared through a facile solid-state reaction in argon. X-ray diffractions combined with Rietveld refinements indicate that Ti2Nb10O27.1 has the same crystal structure with stoichiometric Ti2Nb10O29 (Wadsley-Roth shear structure with A2/m space group) but larger lattice parameters and 6.6% O2– vacancies (vs. all O2– ions). The electronic conductivity and Li+ion diffusion coefficient of Ti2Nb10O27.1 are at least six orders of magnitude and ~2.5 times larger than those of Ti2Nb10O29, respectively. First-principles calculations reveal that the significantly enhanced electronic conductivity is attributed to the formation of impurity bands in Ti2Nb10O29–x and its conductor characteristic. As a result of the improvements in the electronic and ionic conductivities, Ti2Nb10O27.1 exhibits not only a large initial discharge capacity of 329 mAh g–1 and charge capacity of 286 mAh g–1 at 0.1 C but also an outstanding rate performance and cyclability. At 5 C, its charge capacity remains 180 mAh g–1 with large capacity retention of 91.0% after 100 cycles, whereas those of Ti2Nb10O29 are only 90 mAh g–1 and 74.7%.

there is a one-electron transfer between Ti 4+ and Ti 3+ ions and two-electron transfer between Nb 5+ and Nb 3+ ions. As a result, TiNb 2 O 7 has a large theoretical capacity of 388 mAh g -1 based on the five-electron transfer per formula unit, and that of Ti 2 Nb 10 O 29 is 396 mAh g -1 based on its 22-electron transfer. These theoretical capacities are ~1.2 times larger than that of Li 4 Ti 5 O 12 and even surpass that of graphite. TiNb 2 O 7 and Ti 2 Nb 10 O 29 exhibit Wadsley-Roth shear structures constructed by m × n × ∞ (m = n = 3 for TiNb 2 O 7 , Fig. 1a; m = 4 and n = 3 for Ti 2 Nb 10 O 29 , Fig. 1b) ReO 3 -type blocks, where m and n are respectively the length and width of the blocks in numbers of octahedral 10 . All cations (Ti 4+ and Nb 5+ ions with molar ratios of 1 : 2 for TiNb 2 O 7 and 1 : 5 for Ti 2 Nb 10 O 29 ) randomly occupy the octahedral sites connected by edges and corners. Since no cations are resided at the tetrahedral sites, the Wadsley-Roth shear structure is more open than the spinel structure, inferring its larger Li + ion diffusion coefficient. In spite of the above advantages, TiNb 2 O 7 and Ti 2 Nb 10 O 29 suffer from their intrinsically low electronic conductivities and Li + ion diffusion coefficients, which significantly limit their rate performances.
To fulfil the requirement of high power density, it is highly necessary to modify TiNb 2 O 7 and Ti 2 Nb 10 O 29 . However, very limited studies have followed the reports from Han et al. and Wu et al. so far [11][12][13][14][15][16][17] . Only TiNb 2 O 7 nanoparticles, nanofibers and nanoporous particles were prepared and exhibited good rate performances due to the short transport distance for electrons and Li + ions within the TiNb 2 O 7 particles [11][12][13][14][15] . Nevertheless, their fabrications are rather complex and their tap densities are rather small, limiting their practical applications. Crystal structure modification (including doping) has been demonstrated as an effective and facile strategy to improve the rate performance of intercalation-type electrode materials due to the resultant improvements of the electronic conductivity and/or Li + ion diffusion coefficient 5,6 . In this study, for the first time, we have employed the strategy of crystal structure modification to improve the rate performance of Ti 2 Nb 10 O 29 . Using a facile solid-state reaction method, defective Ti 2 Nb 10 O 29-x containing O 2vacancies has been successfully fabricated. Its crystal structure, material properties and electrochemical performances has been intensively studied and compared with the stoichiometric Ti 2 Nb 10 O 29 . The results based on the experiments, Rietveld refinements and first-principle calculations reveal that Ti 2 Nb 10 O 29-x has a larger unit cell volume, enhanced electronic conductivity, improved Li + ion diffusion coefficient, large capacity, high rate performance and good cyclability. Therefore, Ti 2 Nb 10 O 29-x is able to fulfil the two key requirements of high power density and high energy density for EVs.

Results and Discussion
Crystal structure analysis. The observed, calculated, error XRD patterns for the two prepared Ti-Nb-O samples are plotted in Fig. 2. The sharp diffraction peaks are indicative of their good crystallinity rooted in the high-temperature calcination at 1200 °C. The pattern of the Ti-Nb-O sample calcined in air (Fig. 2a)    These increases can be due to the existence of the O 2vacancies, Ti 3+ ions with a larger size (0.67 Å) than that of Ti 4+ ions (0.605 Å) and Nb 4+ ions with a larger size (0.68 Å) than that of Nb 5+ ions (0.64 Å) 19 .
The crystalline characteristics of Ti 2 Nb 10 O 29 and Ti 2 Nb 10 O 27.1 were further examined by the HRTEM tests (Fig. 3). As can be seen in Fig. 3a,b, the atomic layers in Ti 2 Nb 10 O 29 are positioned in orderly and repeated patterns. When combining the three white stripes indicated by the arrows as a unit in Fig. 3a, the layers can be considered as an equidistant arrangement of the unit. The interval between the units is 1.03 nm, which is equal to a half of the lattice parameter c ( Table 1). As shown in Fig. 1b, c is parallel to the width direction of the block (i.e., n direction). Thus, the unit can reflect the width characteristic of the block. Similarly, in Fig. 3b, the interval between the units is 1.55 nm, corresponding to the lattice parameter a. This unit can reflect the length and width characteristics of the block since a is not parallel to or perpendicular to the length/width direction of the block (i.e., m/n direction). When comparing Fig. 3a,c as well as Fig. 3b Particle morphology and size. Figure 4a,b illustrate the particle morphologies and sizes of Ti 2 Nb 10 O 29 and Ti 2 Nb 10 O 27.1 , respectively. Both samples exhibit similar morphologies with wide primary particle-size distributions ranging from less than 100 nm to more than 1 μ m. Aggregates exist in both samples. Therefore, the morphologies show the common features of the powders from solid-state reaction. The BET specific surface area of Ti 2 Nb 10 O 27.1 is 1.30 m 2 g -1 , which is very similar to that of its stoichiometric counterpart (1.26 m 2 g -1 ). This comparison suggests that the different calcination atmosphere negligibly affects the particle size.
Electronic conductivity. The electronic conductivities of Ti 2 Nb 10 O 29 and Ti 2 Nb 10 O 27.1 were determined based on a two-probe direct current method. Ti 2 Nb 10 O 29 has an electronic conductivity which is so low that it cannot be accurately measured using the electrochemical workstation. Since the electrochemical workstation has a current limit of 1 nA, it can be deduced that its electronic conductivity is below 1 × 10 -9 S cm -1 , inferring its insulator characteristic. This result can be due to the fact that there are no free electrons within its cations (Ti 4+ and Nb 5+ ions) in their highest valence states. In sharp contrast, the electronic conductivity of Ti 2 Nb 10 O 27.1 is increased by at least six orders of magnitude to a large value   5,6 . Therefore, compared with doping, the crystal structure modification based on the production of O 2vacancies in this work is a much more effective strategy to enhance the electronic conductivity of metal oxides.
In order to understand more about the physical essence of the pristine Ti 2 Nb 10 O 29 and its defective material, DFT calculations were carried out, and their calculated total density of states (TDOS) and partial density of states (PDOS) are illustrated in Fig. 5. As can be seen in Fig. 5a, for the valence band of Ti 2 Nb 10 O 29 , O 2p, Ti 3d and Nb 4d states fill up the relative lower electronic states from − 5.4 eV to 0.3 eV. The clear overlaps of these states indicate significant hybridization between Ti 3d and O 2p orbitals and between Nb 4d and O 2p orbitals (i.e., ionic Ti-O and Nb-O bonds are formed.). The conduction band mainly comprises unoccupied Ti and Nb states, located from 1.9 eV to 6 eV. The resultant band gap is ~1.6 eV, which is large enough to confirm the insulator characteristic of Ti 2 Nb 2 O 29 since it is well known that the standard DFT calculation underestimates the band gap. Therefore, the electronic conductivity of the pristine Ti 2 Nb 10 O 29 is very low. In contrast, as shown in Fig. 5b, the Fermi level in Ti 2 Nb 10 O 27 is inside of some bands, suggesting that Ti 2 Nb 10 O 27 is no longer an insulator but changed to be a conductor. In addition, impurity bands appear close to but below the Fermi level. These impurity bands are composed of the Ti 3d and Nb 4d states with slightly hybridization with the O 2p states. After the calcination in argon, the production of O 2vacancies in the defective material alters the electron configurations of Ti and Nb ions, leading to the formation of the impurity bands, the conductor characteristic and thus the huge improvement of the electronic conductivity.
Li + ion diffusion coefficient. The Li + diffusion coefficients of Ti 2 Nb 10 O 29 and Ti 2 Nb 10 O 27.1 were determined using the CV technique. The Ti 2 Nb 10 O 29 /Li and Ti 2 Nb 10 O 27.1 /Li cells were tested at a scanning rate of 0.1 mV s -1 for four cycles and then successively at 0.3, 0.5 and 0.7 mV s -1 for one cycle each between 3.0 and 0.8 V vs. Li/Li + at room temperature, as displayed in Fig. 6a,b. For each cell at 0.1 mV s -1 , the intensive cathodic peak shifts to a larger potential after the first cycle. This shift may be attributed to the variation of the electronic structure of Ti 2 Nb 10 O 29 /Ti 2 Nb 10 O 27.1 rooted in the irreversible lithiation process in the first cycle (note that the initial Coulombic efficiency for each cell is not 100% as presented below) 12,17 . In fact, such shift can also be observed in other intercalation-type anode materials, such as Li 4 Ti 5 O 12 21 . Each cycle of the Ti 2 Nb 10 O 29 /Li cell shows two cathodic peaks and three anodic peaks, and that of the Ti 2 Nb 10 O 27.1 /Li cell also exhibits five peaks at the similar positions but with larger intensities. These peaks can be ascribed to the Ti 3+ /Ti 4+ , Nb 4+ /Nb 5+ and Nb 3+ /Nb 4+ redox couples. A previous report reveals that the Ti 4+ and Nb 5+ ions in TiNb 2 O 7 are simultaneously and continuously reduced during the discharge (lithiation) process 15 . Thus, each of the five peaks may be assigned to two or more redox couples. For instance, the cathodic peaks centered at ~1.88 V vs. Li/Li + (at 0.1 mV s -1 ) may correspond to Ti 3+ /Ti 4+ and Nb 4+ /Nb 5+ redox couples. Among all the peaks at 0.1 mV s -1 , the cathodic one centered at ~1.61 V vs. Li/Li + and the anodic one centered at ~1.73 V vs. Li/Li + are relatively intensive. Taking the middle points between these two peaks, The average working potentials of both cells were determined to be ~1.71 V vs. Li/Li + , which is similar to those of the TiNb 2 O 7 /Li cell (~1.64 V vs. Li/Li + ) 17  In addition, there is a linear relationship between the peak current of the intensive cathodic/anodic reaction I p and the square root of the sweep rate v 0.5 , as illustrated in Fig. 6c. As a result, the Li + ion diffusion coefficient D can be calculated based on the Randles-Sevcik equation 22   Li/Li + , which match well with the two intensive peaks in the CV curves (Fig. 6). These two plateaus can correspond to a two-phase reaction 9 . The sloping curves before and after the plateau region are indicative of two different solid-solution reactions. During the first cycle at 0.  and between Nb 3+ and Nb 4+ ions. Since there are considerable amounts of Ti 3+ and Nb 4+ ions in the defective Ti 2 Nb 10 O 27.1 , its contents of Ti 4+ and Nb 5+ ions are smaller than those in its stoichiometric counterpart, leading to its lower initial discharge capacity. However, the Ti 2 Nb 10 O 27.1 /Li cell exhibits a larger initial Coulumbic efficiency (86.9%) than that of its stoichiometric counterpart (82.8%) probably due to its better electrochemical kinetics. Thus, its initial charge capacity (286 mAh g -1 ) can approach that of its stoichiometric counterpart (294 mAh g -1 ).

Rate performance and power density.
With increasing the rate, the plateaus become inconspicuous; the discharge curves monotonically drop and the corresponding charge curves monotonically rise, indicating the increasing polarization. The polarization Δ E vs. rate of both cells is plotted in Fig. 7b. The value of Δ E in this study is defined as the difference between the two potentials at the SOC of 50% With further increase of the rate to 5 C (~2 A g -1 ), the charge capacity of the Ti 2 Nb 10 O 27.1 /Li cell still reaches 180 mAh g -1 , which is twice of that of the Ti 2 Nb 10 O 29 /Li cell (90 mAh g -1 ) and even exceeds the theoretical capacity of the popular Li 4 Ti 5 O 12 /Li cell (175 mAh g -1 ). During the electrochemical reaction of an LIB, electrons and Li + ions simultaneously transport in active material particles. Since both samples in this work have similar particle sizes, their rate performances are determined by their electronic conductivities and Li + ion diffusion coefficients. As demonstrated previously, the defective Ti 2 Nb 10 O 27.1 exhibits improved electronic conductivity and Li + ion diffusion coefficient, which are at least six orders of magnitude and ~2.5 times larger than those of the stoichiometric Ti 2 Nb 10 O 29 , respectively. These two improvements can facilitate the transport of electrons and Li + ions in the Ti 2 Nb 10 O 27.1 particles, resulting in the better rate performance and thus higher power density. All these discharge-charge results are well consistent with the previous CV analysis.
Cyclability. The two cells were further subjected to cyclability evaluation at 5 C, as displayed in Fig. 7c.
The Ti 2 Nb 10 O 27.1 /Li cell remains a large charge capacity of 164 mAh g -1 after 100 cycles, which keeps 91.0% of its initial charge capacity. In sharp contrast, the corresponding values for the Ti 2 Nb 10 O 29 /Li cell are only 67 mAh g -1 and 74.7%. Besides the good cyclability, the Ti 2 Nb 10 O 27.1 /Li cell also exhibits excellent Coulombic efficiency of ~100% throughout the cycling (Fig. 7c). These results demonstrate the highly reversible characteristic, outstanding structural stability as well as fast electronic and ionic transport in the Ti 2 Nb 10 O 27.1 electrode. Its good cyclability is further supported by its ex situ XRD result. Figure 8 exhibits the XRD patterns of the Ti 2 Nb 10 O 27.1 electrodes after as-fabricated, first-discharged to 0.8 V vs. Li/Li + , first-charged to 3 V vs. Li/Li + , and charged to 3 V vs. Li/Li + in the 10 th cycle. As can be seen, the four patterns are very similar. There are little diffraction peak shift and no new peaks at different SOC in spite of some variations in peak intensity, which confirms that basic monoclinic crystal structure of Ti 2 Nb 10 O 27.1 was maintained during the repeated discharge and charge processes. Therefore, Ti 2 Nb 10 O 27.1 is an intercalation-type anode material, similar to Ti 2 Nb 10 O 29 (see Supplementary Fig. S1 online), TiNb 2 O 7 and Li 4 Ti 5 O 12 . The desirable intercalation/deintercalation characteristic and good structural reversibility of Ti 2 Nb 10 O 27.1 can also be ascribed to its good cyclability.
In summary, the defective Ti 2 Nb 10 O 27.1 has been fabricated through a facile solid-state reaction in argon. It shows a Wadsley-Roth shear structure with A2/m space group, the same as that of the stoichiometric Ti 2 Nb 10 O 29 . In comparison with Ti 2 Nb 10 O 29 , Ti 2 Nb 10 O 27.1 not only exhibits larger lattice parameters and a larger unit cell volume but also contains Ti 3+ ions, Nb 4+ ions and 6.6% O 2vacancies (vs. all O 2ions). As a result of this advanced crystal structure, Ti 2 Nb 10 O 27.1 has improved electronic conductivity (1.06 × 10 -3 S cm -1 ) and Li + ion diffusion coefficients (averagely 2.11 × 10 -14 cm 2 s -1 ), which are respectively at least six orders of magnitude and ~2.5 times larger than those of Ti 2 Nb 10 O 29 . Consequently, Ti 2 Nb 10 O 27.1 presents outstanding electrochemical performances in terms of the capacity, rate performance and cyclability. At 0.1 C, it delivers a large initial discharge capacity of 329 mAh g -1 and charge capacity of 286 mAh g -1 . At 5 C, it still remains a large charge capacity of 180 mAh g -1 with large capacity retention of 91.0% over 100 cycles, in sharp contrast to the corresponding values of only 90 mAh g -1 and 74.7% from Ti 2 Nb 10 O 29 . Clearly, this intercalation-type Ti 2 Nb 10 O 27.1 possesses the same advantages of Li 4 Ti 5 O 12 but significantly larger capacities. Therefore, it is able to fulfil the two requirements of high power density and high energy density and thus may be a superior and practical anode material for the LIBs of EVs.

Method
Material preparations. The defective Ti 2 Nb 10 O 29-x was fabricated through a one-step solid-state reaction using TiO 2 (Sigma-Aldrich, 99.9%) and Nb 2 O 5 (Sigma-Aldrich, 99.9%) with a predetermined molar ratio of TiO 2 : Nb 2 O 5 = 2 : 5. These precursors were mixed and milled by a ball-milling machine (SPEX 8000M) for 4 h, and finally calcined at 1200 °C for 4 h in a tube furnace in an argon atmosphere. As a comparison, the stoichiometric Ti 2 Nb 10 O 29 was also synthesized by the same process except for the calcination in an air atmosphere.
To prepare the Ti 2 Nb 10 O 29-x and Ti 2 Nb 10 O 29 samples for electronic conductivity measurements, the above precursors were uni-axially pressed into pellets with a diameter of 10.25 mm at a pressure of 1000 kg cm -2 . The pressed pellets were calcined at 850 °C for 5 h and then at 1200 °C for 48 h in argon (for Ti 2 Nb 10 O 29-x ) or air (for Ti 2 Nb 10 O 29 ). After polishing the two sides of the calcined pellets, gold films were evaporated onto both sides, forming Au/Ti 2 Nb 10 O 29-x /Au and Au/Ti 2 Nb 10 O 29 /Au symmetric ion blocking cells.

Materials characterizations.
Detailed crystal structures of Ti 2 Nb 10 O 29-x and Ti 2 Nb 10 O 29 were identified using X-ray diffractions (XRD) combined with Rietveld refinements. XRD patterns for the refinements were recorded in an angle interval of 5-130° (2θ) with a step width of 0.03° and a counting time of 8 s per step using an X-ray diffractometer (Bruker D8, Germany) with a monochromatic Cu Kα radiation (λ = 0.1506 nm). Ex situ XRD patterns were collected between 15° to 70° (2θ) at a scanning speed of 1° min -1 . The refinements were carried out using the GSAS program with the EXPGUI interface 23,24 . During these refinements, the following instrumental and structural parameters were refined: background parameters, zero-shift, unit cell parameters, profile parameters, atomic fractional coordinates, atomic isotropic displacement parameters and atomic occupancies. The site occupancies were constrained to the designed chemical formulas. Morphologies, particle sizes and microstructures were examined using a field emission scanning electron microscopy (FESEM, Hitachi S-4800, Japan) and a high-resolution transmission electron microscopy (HRTEM, JEOL JEM-2100, Japan). Nitrogen adsorption-desorption isotherms at 77 K were obtained in a surface area analyser (Quantachrome NOVA 2200e, USA). Specific surface areas were derived based on the Brunauer-Emmett-Teller (BET) model. Electronic conductivity tests were performed on the ion blocking cells using an electrochemical workstation (Zahner zennium, Kronach, Germany) under a small voltage of 50 mV until the corresponding currents stabilized.
Electrochemical tests. Electrochemical performances were examined by the tests of CR2016 coin cells assembled in an argon-filled glove box (Mbraum, Unilab, Germany). In these cells, pure Li foils were used as counter and reference electrodes, microporous polypropylene films (Celgard 2400, Celgard LLC., USA) as separators, and a mixture of ethylene carbonate, dimethyl carbonate and diethylene carbonate (1 : 1 : 1 by weight) containing 1 M LiPF 6 (DAN VEC) as electrolyte. Their corresponding working electrodes were fabricated by a slurry-coating procedure. The slurries contained 65 wt% active materials (Ti 2 Nb 10 O 29-x or Ti 2 Nb 10 O 29 ), 25 wt% super P ® conductive carbon (TIMCAL Ltd.) and 10 wt% polyvinylidene fluoride (PVDF, Sigma-Aldrich) in N-methylpyrrolidone (NMP, Sigma-Aldrich). After homogeneously blended, these slurries were uniformly coated on Cu plates. The coated plates were then dried in a vacuum oven at 120 °C for 10 h and finally roller-pressed by a rolling machine to form the working electrodes.
Galvanostatic discharge-charge tests were conducted using a multi-channel battery testing system (LANHE CT2001A, China) with a cut-off potential of 3.0-0.8 V vs. Li/Li + . All discharge/charge rates were denoted using C-rate where 396 mA g -1 was assigned to the current density of 1 C based on the theoretical capacity of Ti 2 Nb 10 O 29 (396 mAh g -1 ). To prepare the electrodes for the ex situ XRD experiments, the coin cells at different states of charge (SOC) were disassembled, and then the working electrodes were washed by dimethyl carbonate for three times and dried at 80 °C. Cyclic voltammetry (CV) measurements were performed using the above electrochemical workstation.
Calculation methodology. All calculations were performed using the projector-augmented wave (PAW) method within the density functional theory (DFT), as implemented in the Vienna ab initio simulation package (VASP) [25][26][27] . Electronic exchange-correlation functional was treated within the spin-polarized generalized gradient approximation (GGA) parameterized by Perdew-Burke-Ernzerhof (PBE) 28 . To address the on-site Coulombic interactions in the localized d electrons of Nb ions, the GGA + U method with an additional Hubbard-type U term (U eff = U -J = 1.5 eV) was employed, which has been proved to be a good approximation for Nb in electrode materials of LIBs 29,30 . Wave functions are expanded in plane waves up to a kinetic energy cut-off of 500 eV. Brillouin-zone integrations were approximated using special k-point sampling of Monkhorst-Pack scheme 31 with a k-point mesh resolution of 2π × 0.05 Å −1 . Lattice vectors (both unit cell shape and size) are fully relaxed together with atomic coordinates until the Hellmann-Feynman force on each atom is less than 0.01 eV/Å.