Superionic Lithium Intercalation through 2 nm ´ 2 nm Columns in the Crystallographic Shear Phase Nb 18 W

Nb 18 W is the tungsten-rich end-member of the Wadsley–Roth crystallographic shear ( cs ) structures within the Nb 2 O 5 –WO 3 series. It has the largest block size of any known, stable Wadsley–Roth phase, comprising 5 ´ 5 units of corner-shared MO 6 octahedra between the shear planes, giving rise to 2 nm ´ 2 nm blocks. Rapid lithium intercalation is observed in this new candidate battery material and 7 Li pulsed field gradient nuclear magnetic resonance spectroscopy – measured in a battery electrode for the first time at room temperature – reveals superionic lithium conductivity with Li diffusivities at 298 K predominantly between 10 –10 and 10 –12 m 2 ·s –1 . In addition to its promising rate capability, Nb 18 W 8 O 69 adds a piece to the larger picture of our understanding of high-performance Wadsley–Roth complex metal oxides.


Introduction
Safe, durable, high power batteries are necessary for vehicular energy storage in, e.g., passenger vehicles, buses, and warehouse vehicles, where they can assist with acceleration, regenerative braking, and rapid charging. Fast recharge, without accelerated cell degradation, is also important for portable electronics, e-bikes and e-scooters, and frequency-stabilization at the grid scale.
Lithium titanate spinel (Li4Ti5O12) is used as an anode in state-of-the-art, commercial high-rate batteries because it reacts with lithium at 1.55 V vs. Li metal, thus avoiding the propensity for coulombically inefficient and potentially dangerous Li metal deposition on the anode under high charging current densities. However, Li4Ti5O12 has a low gravimetric and volumetric energy density due to its partial utilization of Ti 4+ redox centers, cycling between Li4Ti5O12 and Li7Ti5O12, and the necessity of nanosizing to enable high-rate operation. The nanoparticle morphology also presents challenges for material scale-up and processing. Alternatives to Li4Ti5O12 that retain the excellent safety properties but overcome some of the disadvantages could enable wider penetration of high-power battery energy storage.
Niobium-based complex oxides have emerged as promising candidates for high-rate anode materials due to the inherent stability and fast lithium diffusion within some common Nb-oxide structure types as well as the multi-redox capability of niobium (i.e. Nb 5+ to Nb 3+ ). 1 In this work, we explore Nb18W8O69 (9Nb2O5·8WO3) 2 , the tungsten-rich end member of the xNb2O5·yWO3 homologous series. This phase represents the last (i.e., highest tungsten content) of the series to adopt a Wadsley-Roth crystallographic shear (cs) structure ( Figure 1). With increased tungsten content, the niobium tungsten oxides adopt a bronze-like structure, starting with Nb2WO8. 3,4 Up to this point, niobium oxide acts a structure-directing agent, in the sense that Nb-rich compounds form cs phases with various block sizes and inter-block connectivities when alloyed with other metal oxides at high temperature. Nb18W8O69 is comprised of the largest (m × n)p blocks of octahedra in any known cs structure with (5 × 5)1 blocks. The subscript p refers to the connectivity of the (m × n) blocks of octahedra where individual blocks isolated by tetrahedra correspond to p = 1, pairs of blocks correspond to p = 2, and chains of blocks without tetrahedra correspond to p = ∞. Many cs structures (and substituted derivatives) have been investigated including binary H-Nb2O5, 5 niobium titanium oxides-TiNb2O7, [6][7][8] Ti2Nb10O29, [9][10][11] TiNb24O62, 12,13 and niobium tungsten oxides-Nb12WO33, 6,14,15 Nb26W4O77, 16,16 Nb14W3O44, 6,[17][18][19] and Nb16W5O55 1 . These phases vary in block size from (3 × 3) for TiNb2O7 20,21 to (4 × 5) for Nb16W5O55 22 .
According to the published phase diagram, Nb18W8O69 is obtained only in a narrow temperature range from ca. 1265-1375 ºC. 23 Like all of the metastable niobium tungsten oxides, it requires a reconstructive phase transition and mobility of highly charged niobium and tungsten cations to convert to the neighboring phases and is thus kinetically stable at room temperature. However, WO3 sublimes via the gaseous molecules W3O9 and W4O12 at these temperatures. 24 There is competition between (i) reaction kinetics and product stability favored at higher temperatures and (ii) reactant stability favored at lower temperatures. Roth and Waring noted in their study of Nb2O5-WO3 phase equilibria that "[T]he binary equilibrium stability regions of this structural group of compounds tend to decrease with increasing W 6+ content. This phenomenon is probably due to the increasing size of the basic "building block" unit, and therefore to the greater energy needed to maintain long range ordering." 23 Nb18W8O69 presents synthetic challenges but its relation to Nb16W5O55 and its status as the largest known block phase motivate further study as a new electrochemical energy storage material.
Based on lithiation to one lithium per transition metal, the gravimetric capacity of Nb18W8O69 is 164.1 mA×h×g -1 . A two electron 'multiredox' process, as observed in other niobium tungsten oxides 1 , would lead to 328.2 mA×h×g -1 . The high crystal density of 5.37 g×cm -3 should lead to electrodes with high volumetric density. Thus, (5 × 5) Nb18W8O69 is interesting as new anode candidate in its own right, but also to help understand the fundamental interplay of block size, composition, and electrochemical properties in the cs phases.

Results and Discussion
Nb18W8O69 was synthesized from niobium pentoxide and tungsten trioxide according to the following heat treatments that were all followed by quenching on a metal plate: 1. heating the ground starting materials in the furnace to 800 ºC at 10 ºC×min -1 and holding for 16 h followed by heating to 1280 ºC at 3 ºC×min -1 and holding for 2.5 h (Nb18W8O69-I); 2. placing the starting materials into the furnace directly at 1280 ºC for 50 min (Nb18W8O69-II); 3. placing the starting materials into the furnace directly at 1280 ºC for 8 h (Nb18W8O69-III) (Supplementary Figure S1).
These three conditions were chosen to balance the competition between reactant instability and product stability at high temperature: after 24 h at 1200 ºC, mostly Nb2WO8 was found by powder X-ray diffraction (PXRD) but with some shear structure reflections. After 24 h at 1300 ºC, mostly Nb16W5O55 was found by PXRD, which is consistent with the loss of WO3 via sublimation. The stability range of 1265-1375 ºC for Nb18W8O69 from Roth and Waring 23 was based on an annealing time of 1-3 h in a sealed Pt tube and they always observed Nb16W5O55 as a secondary phase.  (2) adjusted to accommodate disorder on the tetrahedral site (see Supplementary Table  S1 for cross-referenced site labels).   Figure S3). Rietveld refinement of the Nb18W8O69 structural model was performed with the X-ray diffraction data from Nb18W8O69-II ( Figure 2c). It was immediately apparent that the peak shapes are almost purely Lorentzian. The atomic positions, isotropic atomic displacement parameters, and cation site fractions were refined in GSAS-II 25 . The light atom (i.e. oxygen) parameters are not reliable from the laboratory X-ray data but some tentative conclusions can be drawn regarding niobium and tungsten occupancies (Supplementary Table S1). If Nb 5+ and W 6+ were distributed randomly across all cation sites, the expected tungsten partial occupancy would be 4/13 » 0.31. However, the data suggest that tungsten prefers sites M2-4 at the center of the blocks while niobium is enriched at M5-8 along the crystallographic shear planes. When niobium and tungsten are allowed to mix on the tetrahedral M1 site during the Rietveld refinement (Supplementary Table S1), the tungsten occupancy moves away from unity, in contrast to the assumption made by Roth and Wadsley. 2 Cheetham and Allen observed a strong preference for Nb 5+ at the edges of the (4 ´ 4)1 blocks in Nb14W3O44. 26  The low-angle diffraction region is significant in cs phases because the long d-spacing inter-block arrangement differentiates various cs phases with nearly identical short-range structure motifs. The unindexed peaks in Figure 2, that are increasingly pronounced from Nb18W8O69-I to III, do not correspond to expected compounds from the phase diagram such as Nb16W5O55, Nb2WO8, Nb14W3O44, Nb12WO33, H-Nb2O5, or WO3. Another hypothesis is that these peaks relate to defects in Nb18W8O69. Wadsley defects-coherent intergrowths of different, usually adjacent, cs phasesare characteristic defects in niobium tungsten oxide shear phases. Allpress and Roth studied Nb18W8O69 with electron microscopy and observed both (4 ´ 5) blocks of octahedra with the nominal composition Nb16W5O55 and (5 ´ 6) blocks of octahedra with the theoretical composition Nb20W11O83. 30 We postulate that the unidentified peaks observed here may be related to extended defect regions within Nb18W8O69.
Electrochemical cells were prepared to analyze the lithium intercalation behavior of bulk Nb18W8O69. The C/2 discharge curves of the three versions of Nb18W8O69 were nominally identical ( Figure 3a), but the irreversible capacity and cycling performance differed ( Figure 3). Nb18W8O69 exhibits similar rate performance to its neighboring cs structure Nb16W5O55 up to 20C with similar Nb14W3O44, 19 and Nb12WO33, 14,15 . Cells were cycled between 3.0 V and a lower voltage cutoff to suppress the proclivity of lithium cation ordering that can lead to phase transitions between lowenergy ordered compositions, which can be detrimental for high-rate performance. Another feature that is evident in the dQ/dV plots is the asymmetry in lithium insertion vs. extraction (Figure 4b,d).
Cycling within a narrower voltage window (Figure 4b), combined with the overpotential, result in a spike at the start of charge at moderate C-rates. On the other hand, the major redox peaks of Nb18W8O69 (1.6-1.7 V) shift minimally up until ≥20C rate even as the initial overpotential increases. Overpotential scales with current density and resistance; the asymmetry shows that this effect is more pronounced on lithium extraction. For clarity we note that, due to the large particle sizes and small accessible surface areas, we do not ascribe the high-rate performance to capacitive or 'pseudocapacitive' effects. Overall, bulk morphologies of Nb18W8O69 appear to (de)intercalate lithium with rapid kinetics like in the Nb16W5O55 cs phase.
To investigate the capacity fade over cycling, voltage and dQ/dV profiles were examined over 100 cycles at C/2 rate (Supplementary Figure S4). The onset of lithiation in the 2 nd cycle occurs at a higher voltage than the first. This phenomenon was also observed in Nb16W5O55 and attributed to V, the oxidative peak shift over 100 cycles is reduced to only 50 mV. In both cases, the second charging peak is unaffected. Notably, the capacity fade in this niobium tungsten oxide differs from the titanium niobium oxide TiNb24O62, which shows increasing polarization between its more pronounced reduction and oxidation peaks. 13 The mechanistic difference in degradation between decreasing dQ/dV intensity vs. peak shift is consistent with active material loss vs. polarization, respectively. 7 Li pulsed field gradient nuclear magnetic resonance (PFG NMR) spectroscopy is commonly applied to liquid and superionic solid electrolytes where lithium diffusion is sufficiently fast and nuclear relaxation is sufficiently slow that it is possible to measure lithium motion over micrometer distances. [35][36][37][38] In principle, PFG NMR should be extremely useful for measuring ionic transport properties in mixed ionic-electronic conductors such as battery electrode materials but prior to Nb16W5O55, this had proven impossible due to relatively slow Li + diffusion and electronic effects.  Non-monoexponential diffusion appears to be a real feature of the niobium tungsten oxide crystallographic shear phases; the fit to the data with a single-component fit was poor and it is clear from the decay curves that a biexponential fit is a far better representation of the data (Supplementary Figure S6). This phenomenon was also observed in LixNb16W5O55 [1] whereas the bronze phase LixNb18W16O93 was well-described by a single-component fit. The biexponential may represent two diffusion processes or it may capture a distribution. The data are insufficient to explore more complex models; however, the current fit does capture the distribution of correlations and rapid transport for all populations of ions -even the slow ones. The structural origin of this complex diffusion may be related to the diversity of tunnels and of sites within the tunnels: there are three classes of tunnels in the material, four per block that border two perpendicular shear planes (i.e., at the corners of the blocks), eight on the edge that border one shear plane, and four in the center of the block that do not border a shear plane. Each tunnel may contain multiple Li local minima. 27 High activation barriers are reported, at least for related phases 34 , for hops between blocks, so that transport is essentially one-dimensional along the c axis. Additional distributions in correlation times could also result from differences in the potential energy maps due to tungsten vs. niobium substitution in the metal octahedra that form the walls of the tunnels. A characteristic quadrupolar lineshape is observed in the static spectra, even at the highest temperature measured (453K) which reflects the anisotropic environment sampled by the Li ions as they move rapidly down the tunnels (Supplementary Figure S8). At lower temperatures, the signal broadens indicating a slowing down of the motion; furthermore the line shape does not appear to arise from a single environment (i.e., is not consistent with a single second-rank tensor) consistent with the multi-exponential behavior discussed above. The relatively sharp and featureless central transition resonance indicates that exchange between the lithium environments is rapid on the timescale determined by the 7 Li chemical shift. A single resonance was also observed in high-resolution magic-angle-spinning 6 Li measurements. Non-monoexponential PFG behavior with sharp 1D spectra suggest that lithium ions sample many environments within a block (local hopping) but diffuse down the different tunnels (net long-range diffusion) with a distribution of diffusion coefficients. While a more detailed analysis of the effect of different types of distributions of correlation times on both the static lineshapes, and on the PFG measurements, will be a subject of future work, the analysis presented here clearly indicates Li diffusivities of between 10 -10 and 10 -12 m 2 ·s -1 .
To the best of our knowledge, these are the first room temperature 7 Li PFG NMR diffusion measurements on battery electrodes. The room-temperature lithium self-diffusion coefficients in concentrations, T1 and T2 relaxation were too fast to allow sufficient diffusion distances or the application of a sufficient gradient pulse, respectively. We note that elevated temperatures were also required to overcome rapid T2 relaxation in our previous study demonstrating 7 Li PFG NMR diffusion measurements on LixNb16W5O55 at x = 6.3 and 8.4. It is expected that Nb16W5O55 and Nb18W8O69 become metallic at higher lithium contents and the nuclear relaxation is enhanced by conduction electrons. 27 The transition from wide band gap insulator to conductor is supported by optical measurements.
Prior to lithiation, Nb18W8O69 is off-white (light yellow-green) with an optical bandgap of 2.84 eV (Supplementary Figure S9). Upon lithium insertion, the color becomes dark blue. Experimental and theoretical work on related cs phases suggests that lithiation reduces d 0 Nb 5+ and W 6+ , 1 ndoping the structure into the conduction band and causing an increase in electronic conductivity, 27,34,41 which motivates detailed future work to understand subtle differences between compositionally and structurally similar cs compounds.
The tunnels within the blocks of Wadsley-Roth crystallographic shear structures enable lithium transport and the 5 ´ 5 blocks of octaheda in Nb18W8O69 are the largest that have been synthesized as an isolated phase. These 2 nm ´ 2 nm blocks give rise to sixteen parallel lithium tunnels with twelve that are adjacent to shear planes (i.e. 'edge' tunnels) and four that are central and ReO3-like (i.e. 'central' tunnels). Bond valence sum maps ( Figure 6) and calculations on related systems indicate that lithium moves through both types of tunnel. 42,34 It is not known how the dynamics differ at central vs. edge tunnels; this is an interesting question requiring computational insights, though it is non-trivial as diffusion barriers are highly sensitive to the specific transition metal cation ordering in Wadsley-Roth TiNb2O7. 34 Regardless, the large blocks offer a maximal number of tunnels and rapid Li + conductivity. That the rate performance of Nb18W8O69 is not higher than common to unstabilized ReO3 and WO3 upon lithiation. [43][44][45][46] An appropriate balance between conducting tunnels and stabilizing shear planes is thus desired for high-rate battery performance but the optimal shear structure is not established. From the small but growing body of literature on Wadsley-Roth phases, there is no clear trend between long-term stability and shear plane density or shear plane arrangement, though many variables stand between the idealized atomic structure and battery performance. In addition to device fabrication and microstructural considerations, an unavoidable complication is that the variation in the blocks is accompanied by changes in composition. Though Wadsley defects are generally expected, a further challenge in understanding this family is that the exact type, concentration, arrangement, and role of defects is not well-known nor easy to determine. Future work could consider other subtle fundamental aspects of the interplay between synthesis-structure-defects-electrochemistry, the relationship between lithium content and reversibility, and practical strategies to increase the cycle stability of superionic Nb18W8O69 and enable its use in ultrafast charging and high power energy storage.

Conclusion
Nb18W8O69, with a crystallographic shear structure, hosts superionic lithium ion conductivity that leads to battery rate performance on par with recently reported members of the Nb2O5-WO3 phase diagram including shear phase Nb16W5O55 and bronze-like phase Nb18W16O93. PFG NMR was able to directly capture room temperature lithium diffusion of up to 10 -11 m 2 ·s -1 for the majority 7 Li signal contribution in this mixed ionic-electronic conductor. Electrode active materials mass loadings were 2.0 ± 0.2 mg·cm -2 . Electrochemistry was performed in a temperature-controlled room at 293 K with a Biologic galvanostat/potentiostat running EC-Lab software. Accessible capacity and, therefore, C-rate are not well-defined in the Wadsley-Roth phases due to the sloping voltage profiles that lead to a strong dependence of capacity on the voltage window; multielectron redox of both Nb 5+ to Nb 3+ and W 6+ to W 4+ are known. 1 In these phases, we believe the most convenient method is to defined C-rate relative to one electron transfer per transition metal: for Nb18W8O69 (to Li26Nb18W8O69): 1C = 164.1 mA×g -1 , 20C = 3282 mA×g -1 .
PFG NMR 7 Li diffusion measurements were carried out as described in detail in reference [1].
Lithium niobium tungsten oxide samples were prepared by electrochemically lithiating 210±20 mg pellets of Nb18W8O69 over the course of 60-100 h in coin cells following the procedures previously described. After lithiation, the cells were brought into the inert atmosphere of the glovebox where the electrodes were extracted, washed with DMC (3 ´ 2 mL), dried under the vacuum of the prechamber, ground in an agate mortar and pestle, and packed into 4.0 mm zirconia rotors that were then sealed into J. Young NMR tubes. Signal-to-noise (S/N) is particularly important for PFG NMR because it measures signal decay so the initial (g = 0) S/N ratios are given for all compositions and temperatures (Supplementary Table S2). All diffusion measurements were performed on Nb18W8O69-I due to the similarity of the electrochemical rate performance of the three samples and to the time-consuming nature of variable temperature and variable composition PFG NMR experiments on partially discharged battery electrode materials.

Associated Content
Supporting Information PDF containing synthesis heating profiles, SEM images, diffraction patterns, electrochemical data, PFG NMR decay curves and fitting models, 1D NMR spectra, diffuse reflectance spectra, crystal structure data, and quantification of the attained signals for each PFG NMR measurement.

Corresponding Author
Email: cpg27@cam.ac.uk Notes K.J.G. and C.P.G. are major shareholders in a start-up company developing fast charging batteries based on high-rate anode materials.