Highly Reversible Lithiation of Additive Free T‐Nb2O5 for a Quarter of a Million Cycles

Fast energy storage via intercalation requires quick ionic diffusion and often results in pseudocapacitive behavior. The cycling stability of such energy storage materials remains understudied despite the relevance to lifetime cost. Orthorhombic niobium oxide (T‐Nb2O5) is a rapid ion intercalation material with a theoretical capacity of 201.7 mAh g−1 (Li2Nb2O5) and good cycling stability due to the minimal unit cell strain during (de)intercalation. Prior reports of T‐Nb2O5 cycling between 1.3–3.1 V versus Li/Li+ noted a 50% loss in capacity after 10 000 cycles. Here, cyclic voltammetry is used to identify the role of the voltage window, state of charge, and potentiostatic holds on the cycling stability of mesoporous T‐Nb2O5 thin films. Films cycled between 1.2–3.0 V versus Li/Li+ without voltage holds (Li1.1Nb2O5) exhibited extreme cycling stability with 90.8% capacity retention after 0.25 million cycles without detectable morphological/crystallographic changes. In contrast, the inclusion of 60 s voltage holds (Li2.18Nb2O5) led to rapid capacity loss with 61.6% retention after 10 000 cycles with corresponding X‐ray diffraction evidence of amorphization. Cycling with other limited voltage windows identifies that most crystallographic degradation occurs at higher extents of lithiation. These results reveal remarkable stability over limited conditions and suggest that T‐Nb2O5 amorphization is associated with high extents of lithiation.


Introduction
The development of lithium intercalation electrode materials with high cycling stability is key to reducing the total consumer cost of electrical energy storage.[3][4][5][6][7] These operating costs are proportionally reduced by improvements to the cycling stability.The DOE Long Storage Shot aims to reduce energy storage costs by 90% which could be feasible by extending the durability of DOI: 10.1002/adfm.202312839known materials by 10×.The degradation of lithium intercalation materials from extended charging and discharging is typically manifested as a loss of power or gradual capacity fade.The processes leading to these losses are often convolved but can generally be summarized into the following: 1) loss of lithium inventory through solid electrolyte interphase (SEI) growth, lithium plating, interfacial instabilities, and other negative side reactions, 2) disconnection of active material from the current collector via either delamination/crack formation, or material dissolution caused by large volume changes during lithiation, 3) degradation of the active material, e.g.14][15] The rapid intercalation of lithium into T-Nb 2 O 5 on the time scale of seconds to minutes led to much excitement associated with the new concept of intercalation pseudocapacitance. [16,17][20][21][22][23][24] The lithiation of Nb 2 O 5 to Li 2 Nb 2 O 5 (1 lithium per Nb metal) corresponds to a theoretical capacity of 201.7 mAh g −1 .The cyclic voltammetry (de)lithiation peaks for T-Nb 2 O 5 occur between 1.5-1.7 V versus Li/Li + ; voltages that are not associated with the decomposition of carbonate electrolytes which occur predominantly at voltages <1.0 V and >4.7 V versus Li/Li + . [3,25]The lithiation of T-Nb 2 O 5 in the voltage window of 1.2V-3.0V versus Li/Li + occurs with minor lattice strain, resulting in only a 3% increase of the c-axis parameter without the presence of a first-order phase transition, thus avoiding abrupt changes in density which could promote material delamination and lead to premature device failure. [26]T-Nb 2 O 5 also exhibits fast lithium diffusion which is accommodated by its "room and pillar" crystal structure consisting of spacious channels between loosely packed oxygen atoms bridging between sheets of densely packed, highly distorted NbO 6 /NbO 7 octahedra and bipyramids. [27,28][31] To date, additive-free, mesoporous T-Nb 2 O 5 thin films enabled the fastest kinetics, where 95% of the overall capacity of 160.4 mAh g −1 was retained with short 2.25 s delithiation times. [32]Avoiding the use of additives (conductive additives and binder) helps avoid capacity loss mechanisms associated with those components, thus better elucidating the degradation of active materials. [33]rior reports of the cycling stability in T-Nb 2 O 5 electrodes are limited to <10 000 cycles; a target that would generally be considered highly durable.[36] Additive-free thin films of T-Nb 2 O 5 fabricated via electrophoretic deposition demonstrated ≈50% capacity retention after 10 000 cycles with the loss of capacity being attributed to a convolution of material loss, amorphization, and thin-film delamination. [31]Although T-Nb 2 O 5 is the most widely studied Nb 2 O 5 polymorph, the niobia phase diagram has several other phases including, pseudohexagonal (TT-Nb 2 O 5 ), bronze (b-Nb 2 O 5 ), tetragonal (M-Nb 2 O 5 ), and monoclinic (H-Nb 2 O 5 ). [37,38]he TT-Nb 2 O 5 polymorph is most similar to T-Nb 2 O 5 in terms of electrochemistry but has additional structural disorder. [28,38,39]ecent reports noted phase changes in niobia as a result of electrochemical lithiation, most notably, the transformation of amorphous niobia into a rock salt structure capable of intercalating 1.5 Li per Nb atom. [40]Similarly, single crystal T-Nb 2 O 5 thin films were reported to transform reversibly into a monoclinic version and then irreversibly into an amorphous version with progressively higher extents of lithiation. [41]Such electrochemically induced crystallographic changes are important to understand for the realization of long-term cycling stability in T-Nb 2 O 5 electrodes.Herein, the effects of the electrochemical cycling conditions on T-Nb 2 O 5 capacity retention are examined using CV.The effects of voltage window and hold time are specifically examined where the state of charge (SOC) and capacity with cycling are examined with respect to changes observed using grazing incidence wide-angle X-ray scattering (GI-WAXS), scanning electron microscopy (SEM), and cyclic voltammetry (CV).

Synthesis and Characterization of Mesoporous T-Nb 2 O 5 Thin Films
Here, a model study is presented using additive-free electrodes to focus on the degradation mechanisms that are inherent to T-Nb 2 O 5 itself.Prior additive-free studies of this material noted material loss, amorphization, and delamination, all of which could be induced by the strain that occurs upon intercalation.The introduction noted examples of reported crystallographic changes occurring at extents of lithiation between 0.9 < x < 1.8 and also at x > 1.8 in Li x Nb 2 O 5 .Furthermore, to expedite the time needed for cycling, this study utilizes nanostructured materials with short diffusion lengths and correspondingly more rapid intercalation time scales. [17]Such nanostructure strategies can result in additional SEI formation which can expedite degradation.4] The studied mesoporous thin films were prepared using polymer templates and their as-made characteristics were measured.Previous studies on closely related mesoporous samples identified electrolyte transport was not rate-limiting. [29]Persistent micelle templates composed of poly(styrene-b-ethylene oxide-bstyrene) (Figure S1, Supporting Information) were used as structure directing agents for niobia precursors (niobium ethoxide, HCl(aq)) which were coated onto a variety of substrates for different measurements. [45,46]The subsequent removal of the polymer by calcination in air at 600 °C produces nanoscale pores within continuous niobia as shown by SEM in Figure 1a.This electron micrograph shows disordered packing of 45.5 ± 0.8 nm average diameter spherical pores (dark) with 13.8 ± 0.4 nm average niobia wall thickness (light).This uniform and disordered arrangement was consistent with the SAXS measurements (Figure 1b) which had a prominent singular peak with d-spacing (2 q −1 ) of 60.8 nm, corresponding well to the sum of the average pore and wall dimension from SEM.The calcination treatment also induced crystallization of the niobia which was examined with GI-WAXS.These diffraction patterns (Figure 1c) match well with T-Nb 2 O 5 (JCPDS 30-873) and Scherrer analysis of the (001) peak (2 = 22.6°) corresponded to an average crystallite size of 13.0 nm.Spectral reflectance was used to estimate the film thickness as ≈69 nm.Finally, the electrochemical characteristics of the uncycled samples were characterized with a 1 mV s −1 voltage sweep rate to present the pseudo-equilibrium capacity versus voltage characteristics (Figure 1d).][49] Also apparent in this plot is that most of the capacity is stored between 1.2-2.0V versus Li/Li + , so it is expected that most of the SOC-related degradation would occur between these voltages.From these characterizations, the as-prepared electrodes were consistent with uniform nanostructured T-Nb 2 O 5 with typical electrochemical characteristics.

Cycling Stability Experiments
CV measurements were used with a variety of cycling conditions to track degradation.The conditions studied either varied the potentiostatic hold time at the end of each voltage sweep or varied the voltage window.The conditions examined are summarized and named in Table 1.The conditions used for extensive cycling all had voltage sweep rates of 360 mV s −1 .At the end of each high-rate cycling run, an additional slower diagnostic CV measurement ("Full-Slow") probed the full voltage range at a slower 1 mV s −1 rate to check for changes to the pseudo-equilibrium lithiation characteristics.All reported voltages herein are with respect to a lithium reference electrode.The results of a cycling condition exhibiting remarkable endurance are elaborated first in detail before comparing the effects of various cycling parameters in the following sections.

Cycling Conditions for 0.25 Million Cycles (Full-t0)
Condition Full-t0 involved cycling between the "full" voltage window of 1.2 and 3.0 V without potentiostatic holds and resulted in remarkable electrode endurance.The CV scans during condition Full-t0 are shown in Figure 2a every 50 000 cycles up to 250 000 where negligible changes were observed in the currentvoltage characteristics during rapid cycling.Likewise, the capacity was found to be remarkably stable over this period where the six electrodes tested started with 112 ± 4 mAh g −1 and ended with 102 ± 5 mAh g −1 these cycling conditions, corresponding to 90.8 ± 2.1% capacity retention (Figure 2b).This corresponds to a negligible loss rate of capacity of 4.12 −5 mAh g −1 per cycle.The capacity during cycling corresponds to an intermediate SOC of Li 1.11 Nb 2 O 5 which is kinetically limited by the rapid testing protocol.[50] This strategy is rational from an economic perspective if the reduced capacity is more than offset by increases in longevity.Diagnostic ("Full-Slow") CV measurements are shown before and after 0.25 million cycles of rapid Full-t0 cycling (Figure 2c).The diagnostic sweeps before cycling included broad peaks at 1.80 V (anodic) and 1.78/1.49V (cathodic) corresponding to the Nb 4+ /Nb 5+ oxidation state change during (de)intercalation.After the 0.25 million Full-t0 cycles, these plots included sharper peaks appearing at 1.75 (anodic) and 1.80 V (cathodic) and the corresponding capacity was 180 ± 8 mAh g −1 (not kinetically limited during this measurement).The sharpening of the peaks and a slight increase of CV peak separation from 0.02 V (uncycled) to 0.05 V (Full-t0) suggests changes to the atomic structure.The crystal  Samples were extensively cycled under condition "Full-t0" (see Table 1 for details).a) The current-voltage data is presented every 50 000 cycles up to 0.25 million where b) the corresponding initial capacity of 112 ± 4 mAh g −1 was 90.8 ± 2.1% retained after 0.25 million cycles.The cycling data corresponds to the average and standard error of the mean for six samples.c) After cycling, a slow analytical measurement ("Full-slow") was used to identify changes in the pseudo-equilibrium voltage-current behavior.
structure before and after cycling is compared with GI-WAXS measurements in Figure 3a.A grazing incidence geometry (incidence angle 0.15°) was used to increase the diffraction contribution from the thin film of active material (i.e., reduce the substrate intensity).Here, a reference pattern for the FTO substrate is included where the peaks attributed to T-Nb 2 O 5 (centered at 22.6°and 28.9°) before and after cycling did not exhibit apparent changes.Electron diffraction data (Figure S2, Supporting Information) similarly showed reflections corresponding to T-Nb 2 O 5 , including (001), (180), (181), and (002) planes in the uncycled samples and reflections corresponding to (180), (181), and (002) in the Full-t0 samples which had been cycled 0.25 million times.Thus, the CV, GI-WAXS, and electron diffraction data are consistent with invariant morphology and crystal structure, with the possibility of minor changes to the local atomic structure after extensive cycling.Understanding these very stable cycling conditions is aided in the next sections by examining less stable cycling conditions.
The samples were also checked for SEI growth which would deplete the lithium reserves in a typical balanced two-electrode configuration.SEM measurements after 0.25 million Full-t0 cycles exhibited average pore diameters of 44.5 ± 0.9 nm which is within the error of the as-made sample pore dimensions, albeit with some roughening of the pore edges (Figure 3b).The corresponding pore size histograms before and after this cycling are also quite similar (Figure 3c,d), suggesting limited, if any, SEI growth.Lastly, energy dispersive X-ray spectroscopy (EDS) measurements were compared before and after this cycling (Figure S3, Supporting Information) which indicated no detectable change in the ratio of niobium relative to carbon or oxygen, within the error of the measurements.Thus, the feature sizes and film composition measurements did not show evidence of marked SEI deposition within the sensitivity limits of the instruments.

Effects of Varied Cycling Conditions
Next, capacity retention and SOC were examined with a variety of voltage windows and hold durations.Here, the SOC corresponds to the stoichiometry x in Li x Nb 2 O 5 during (de)lithiation which influences the equilibrium atomic arrangement.High SOCs lead to increased strain which can affect the material/current collector interface or could induce changes to the atomic structure.Different SOCs can be attained by varying the duration of hold periods (kinetic limitation) and by changing voltage windows (range of equilibrium states of charge).Several combinations and permutations of these different cycling conditions are listed in Table 1, examining both parameters separately.
The addition of voltage holds to the full 1.2-3.0V window led to progressive capacity loss.This effect is elucidated by comparing cycling conditions Full-t0, Full-t5, and Full-t60 which include 0, 5, and 60 s potentiostatic holds to reach SOCs of L i1.11 Nb 2 O 5 , Li 2.06 Nb 2 O 5 , and Li 2.18 Nb 2 O 5 , respectively.Figure 4a shows the effects of these rapid cycling conditions after 10 000 cycles upon the current-voltage behaviors when measured with the "Full-Slow" diagnostic scans.Here, increasing hold time gradually shifts the lithiation peaks (negative current) to more anodic voltages, shifting from 1.8 to 1.6 V.The shifting CV peaks and increasing peak separation of 0.07 V for Full-t5 and 0.15 V for Full-t60 suggest more sluggish charging kinetics and changes to the atomic structure as a result of the hold times. [51]Figure 4b shows how this increasing hold time also leads to progressive capacity loss (with respect to the first cycle capacities), with a 66.7% ± 3.0% retention of capacity in Full-t5 and a 61.6% ± 2.0% retention in Full-t60.Here, the longer hold times lead to larger extents of lithiation as the measurement time becomes longer than the diffusion time, with 208 ± 8 mAh g −1 stored for Full-t5 and 220 ± 8 mAh g −1 stored for Full-t60.Curiously, condition Full-t60 led to an initial capacity in slight excess of the assumed theoretical capacity as the electrodes continued to lithiate during the potentiostatic hold.Such excessive lithiation of Nb 2 O 5 has previously been reported to result in irreversible crystal structure changes. [27]The corresponding GI-WAXS patterns are consistent with amorphization where after 10 000 cycles of condition Full-t60 there were not any apparent T-Nb 2 O 5 diffraction peaks centered at 22.6/28.9= 2 (Figure 5).   1) after 10 000 rapid cycles.a) After this cycling, a slow analytical measurement ("Full-slow") was used to compare changes to the pseudo-equilibrium voltage-current behaviors.b) The changes to the lithiation capacities during rapid cycling (Table 1) were also compared before and after these 10 000 cycles of each treatment.The reported values in (b) correspond to average values and the standard error-of-the-mean values.
Anod-t5 initially stored 164 ± 5 mAh g −1 (Li 1.63 Nb 2 O 5 ), while the Cath-t5 stored 122 ± 10 mAh g −1 (Li 1.21 Nb 2 O 5 ).The effects of 10 000 such cycles upon the pseudo-equilibrium current-voltage behaviors are shown in Figure 4a.Here, the Anod-t5 condition leads to both more anodic peak shift and more peak separation (0.15 vs 0.05 V), suggesting potential crystallographic changes with higher SOCs during cycling.Furthermore, the relative capacity retention after 10 000 cycles also correlated with the SOC where condition Anod-t5 exhibited 71.3 ± 1.6% capacity retention as compared to condition Cath-t5 which exhibited a capacity retention of 88.7 ± 2.1% .After 10 000 cycles, both these conditions exhibited similar total capacities despite Anod-t5 starting with a higher capacity (Figure 4b).The crystal structures of Anod-t5, and Cath-t5 were compared in Figure 5 to associate these capacity losses with changes to the crystal structure.After 10 000 cycles the Anod-t5 samples show no peaks centered at 22.6°or 28.9°= 2 corresponding to the (100) and ( 180) planes of T-Nb 2 O 5 suggesting that the higher SOC led to amorphization.In contrast, after 10 000 cycles the Cath-t5 samples exhibited a peak corresponding to the (100) peak of T-Nb 2 O 5 at 22.6°= 2 and a weakly defined (180) peak at 28.9°= 2 suggesting limited crystal structure changes.
The pseudo-equilibrium capacities ("Full-slow" measurement) after extensive cycling (Figure 4a) were also compared (Table 1) where all cycling conditions led to a drop in the specific capacity after high-rate cycling, except for Anod-t5.This is surprising since Anod-t5 high-rate capacities progressively degraded in terms of the kinetically accessible capacity.With the "Full-Slow" conditions, however, Anod-t5 cycled samples stored 264 ± 9 mAh g −1 which corresponds to Li 2.61 Nb 2 O 5 .This is beyond the reversible capacity of T-Nb 2 O 5 (assuming Nb 4+/5+ ) and is closer to the capacity of rock-salt Nb 2 O 5 which can reversibly intercalate 3 lithium ions per formula unit (Nb 3+/4+/5+ ).These observations, the change of the CV characteristics, and the lack of apparent GI-WAXS peaks corresponding to T-Nb 2 O 5 suggest The overall collection of cycling stability measurements here indicates that greater SOCs (lower voltages or longer hold times) can irreversibly alter T-Nb 2 O 5 , leading to shifting CV peaks, increasing peak separation, and less high-rate capacity.In contrast, cycling conditions with lower SOCs (more cathodic V-limits or shorter hold times) enabled more stable capacity during rapid cycling without apparent changes to crystal structure or drastic changes to the CV characteristics, suggesting minimal changes to the atomic structure.The conditions with the highest retention of capacity all used limited SOCs during cycling, either achieved with kinetically limited cycling (non-equilibrium) or with a voltage window that limited the SOC.

Conclusion
We report new insights into the effects of cycling conditions upon the degradation of T-Nb 2 O 5 thin films during high-rate lithiation using CV, GI-WAXS, and SEM/TEM.Samples cycled without a potentiostatic holds (Full-t0) achieved 0.25 million cycles with a 90.8% ± 1.8% capacity retention (−4.12E-5 mAh g −1 per cycle) while maintaining constant crystal structure and morphology without apparent SEI growth.In contrast, the addition of potentiostatic holds led to progressive capacity loss upon cycling that correlated with amorphization apparent by GI-WAXS.Cycling conditions with limited voltage windows revealed that degradation was most prominent with lower voltages (higher SOC), correlating the SOC to degradation.These results reveal how the lithiation cycling conditions have pronounced effects on the longevity of T-Nb 2 O 5 .Unraveling the causal factors of active material degradation is essential to enable extreme longevity in future devices.
Polymer Synthesis and Characterization: Poly(styrene-b-ethylene oxideb-styrene), PS-b-PEO-b-PS dichelic block polymer was synthesized by a two-step method.First, a dichelic macroinitiator was synthesized using a Steglich esterification followed by activators regenerated by reversibledeactivation radical polymerization (RDRP). [52]The conditions for the Steglich esterification are listed in detail elsewhere. [53]The following conditions were used for the RDRP: Br-PEO-Br: PMDETA: Cu(II)Br : Ascorbic Acid: Styrene of 1:1:1:0.4:1100.Inhibitor-free styrene (7.4 mL, 63 mmol) was added to 35 000 g mol −1 Br-PEO-Br macroinitiator (2.00 g, 0.057 mol) of in a 100 mL round-bottom flask along with of anisole (1.5 mL).An aliquot of a Cu(II) 30 mg mL −1 solution (0.43 mL, 0.057 mmol) and an aliquot of a 12.5 mg mL −1 ascorbic acid solution (0.32 mL, 0.023 mmol) was added to the reaction flask along with PMDETA (11.9 μL).Then, the solution flask was sealed with a rubber septa and copper wire and then nitrogen sparged for 30 min before it was added to a 110 °C oil bath while stirring for 17 h, at which point the reaction mixture became sufficiently viscous.The reaction vessel was placed in the freezer for an hour to cool and then the polymer was dissolved in DCM so the mixture could be passed over a basic alumina column to remove the remaining copper catalyst.The polymer/DCM mixture was precipitated into 1 L of methanol and dried in air.The molar mass of the PS block was characterized by comparison to the known PEO (M n = 35 000 g mol −1 ) using a Bruker Avance III HD 300 1 H NMR and then the molar mass dispersity (Ð) was measured using gel permeation chromatography (GPC) (Figure S1a-c, Supporting Information).NMR samples were prepared in deuterated chloroform (CDCl 3 ) at a concentration of 10 mg mL −1 .The GPC data was collected using a Waters instrument equipped with a 515 HPLC pump, a 2410 refractive index detector, and three styragel columns eclipsing the effective molecular weight range of 0.1-600 kg mol −1 .The eluent used was tetrahydrofuran at a temperature of 30 °C and a flow rate of 1 mL min −1 .GPC samples were prepared by dissolving the PS-b-PEO-b-PS in THF with a concentration of 10 mg mL −1 and filtering through a 0.2 μm syringe filter prior to injection.
Kinetically Trapped Micelle Stock Preparation For Micelle Templating: To prepare kinetically trapped micelle templates ("persistent micelle templates") solutions with the previously synthesized PS-b-PEO-b-PS dichelic block polymer, the block polymer (200 mg) was dissolved in DCM (5 mL).Then, anhydrous ethanol was added incrementally following a previously published method. [41]After micellization, the DCM was removed via rotary evaporation and the resulting kinetically trapped micelle stock had a concentration of 8.53 g mL −1 .
Preparation of Mesoporous T-Nb 2 O 5 Thin Films: The kinetically trapped micelle stock (5.86 mL) was acidified with concentrated HCl (128.4 mg, 1.303 mmol) to reach a water content of 1.8%.Then, niobium ethoxide (237.6 mg, 0.7466 mmol) precursor was added.The solutions were used immediately to spin coat borosilicate glass slides for SAXS characterization, 1 × 1 cm (100) cut silicon wafers for SEM and GI-WAXS, and on FTO for electrochemical characterization and GI-WAXS.Spin coating was done with 10 μL of solution for smaller glass and silicon substrates and with 150 μL of solution for FTO waters.Spin coating conditions were 1000 rpm for 30 s with 15% relative humidity, using a homemade set-up described in detail elsewhere. [40]Samples were immediately aged for 12 h at 110 °C and then calcined at 600 °C for 12 h with a ramp rate of 5 °C min −1 from RT-200 °C and 10 °C min −1 from 200-600 °C to remove the block polymer and crystallize the metal oxide.
Electrode Preparation: Fluorine-doped tin oxide glass sheets were cleaned by wiping them off with IPA and kimwipes to remove excess organics.The sheets were then cut into 2.5 × 6 cm sections and sonicated for 30 min in a soapy alkanox bath.The sheets were rinsed in DI water until squeaky clean.After this, the FTO sheets were stored in the oven at 400 °C prior to spin coating to ensure their cleanliness.Just before spin coating, the FTO was removed from the oven and each section was cut in half to make 2.5 × 3 cm rectangles, and a strip at the top was covered with high-temperature Kapton tape to allow for an uncoated area for electrical contact.After the FTO films were spin-coated, aged, and calcined, the edges of the electrodes were trimmed to remove any edge effects from the spin coating, leaving enough space for two electrodes with roughly 1 cm 2 area.The area for each electrode was measured using photography over a ruler grid and ImageJ software.This area was used in conjunction with ICP-MS to determine the specific mass loading of metal oxide for the thin films.
X-Ray Scattering Measurements: Wide-angle and small-angle X-ray scattering (WAXS/SAXS) measurements were performed at the South Carolina SAXS Collaborative using a SAXSLab Ganesha instrument.A Xenocs GeniX3D microfocus source was used with a Cu target to create a monochromatic beam with a wavelength of 0.154 nm.The instrument was calibrated with a NIST reference material 640d silicon powder, with a reference peak position of 2 = 28.44°,where 2 represents the total scattering angle.A Pilatus 300 K detector (Dectris) was used to collect the 2D scattering patterns.The detector exhibited a nominal pixel dimension of 172 × 172 μm 2 .The SAXS data was acquired with an X-ray flux of ≈4.1 million photons s −1 incident upon the sample and with a sample-todetector distance of 1040 mm.The 2D images were azimuthally integrated to yield the scattering vector intensity.Peak positions were fitted using custom MATLAB software.GI-WAXS measurements were conducted with an incident angle ( i ) of 0.15°relative to the incident beam.The GI-WAXS sample-to-detector distance was 112.1 mm with an X-ray flux of ≈36.3 mil-lion photons s −1 upon the sample.The 2D WAXS data were masked to remove diffuse reflectance before integration and analysis of the resulting 1D data.For Scherrer crystallite size analysis, the instrumental broadening factor was fit as a Gaussian point-spread function and taken into account to interpret scattering data from grain-size broadening per the Scherrer formula using the same Gaussian point-spread function. [54]EM Characterization: Images of calcined films on fluorine-doped tin oxide were acquired using a Zeiss Ultraplus thermal field emission SEM using an accelerating voltage of 3 keV and an in-lens secondary electron detector.An ideal working distance was found to be ≈3.2 mm and all images were acquired with 400 000x and 800 000x magnification.At least 300 measurements were made for each feature (pores and walls) to determine their statistical metrics using ImageJ software.
TEM Characterization: TEM images were acquired using a Hitachi 7800 TEM operated in bright field and diffraction mode with an accelerating voltage of 100 keV.Samples were prepared by placing a single drop of solution containing suspended mesoporous film, scraped from their substrates with glass, onto carbon-coated 300 mesh copper grids.The solution was allowed to evaporate to dryness.Samples were collected by first focusing on a single sample area using the selective area aperture in a bright field at a magnification of 120000×.
Electrochemical Analysis: Electrochemical measurements were made in an Argon filled glovebox, maintained at <1 ppm oxygen and water, using a BioLogic SP-150 potentiostat with a three-electrode setup utilizing lithium metal as the reference and counter electrodes.Lithium foil was scraped shiny before use and matched in area to the working electrode.The working electrodes were Nb 2 O 5 thin films on FTO substrates.The electrolyte solution was 1.0 m LiClO 4 in propylene carbonate.A suite of electrochemical techniques was applied to each working electrode consisting of diagnostic cyclic voltammetry and electrochemical impedance spectroscopy to verify a suitable external contact to the FTO electrode.Subsequently, cycling stability tests were performed using the parameters listed in Table 1.The reported capacities corresponded to the integral of current with time during the voltage sweep, excluding any hold period.During extended 360 mV s −1 cycling, the Coulombic efficiency remained at 96% throughout the course of cycling.
Film Thickness Measurements: A Filmetrics F20 optical reflectometer was used to measure film thickness of the T-Nb2O5 thin films.The average thickness across 500 measurements was 68.5 ± 1.1 nm, well in line with previous, more comprehensive measurements of similar samples. [29]A Si reference wafer was selected for baseline and air as the background.
Inductively Coupled Plasma Mass Spectrometry: A series of films prepared on FTO substrates were cut to ≈1 cm 2 of the Nb 2 O 5 coating.Im-ageJ analysis was used to account for the specific substrate area.These films along with FTO blanks were heated in Teflon vessels containing 70% HNO 3 , 37% HCl, and 48% HF (1:3:0.5 mL) respectively at 180 °C for 12 h before solutions were diluted with water (18.2MΩ cm) to 50 mL volume and measured using a Thermo-Finnigan Element XR inductively coupled plasma mass spectrometry (ICP-MS).The instrument was calibrated using a range of concentrations spanning those of the measured samples in conjunction with FTO blanks.These data were used to calculate the Nb 2 O 5 mass-per-area metric for each sample.

Figure 1 .
Figure 1.Characterizations of uncycled starting materials including a) SEM of nanostructured T-Nb 2 O 5 (light) with pores (dark), b) SAXS data of the corresponding nanostructure, c) WAXS data of the crystal structure consistent with T-Nb 2 O 5 (JCPDS#30-873), and d) initial state CV data at 1 mV s −1 ("Full-slow") which corresponds to the pseudo-equilibrium voltage versus.capacity relationship.

Figure 2 .
Figure2.Samples were extensively cycled under condition "Full-t0" (see Table1for details).a) The current-voltage data is presented every 50 000 cycles up to 0.25 million where b) the corresponding initial capacity of 112 ± 4 mAh g −1 was 90.8 ± 2.1% retained after 0.25 million cycles.The cycling data corresponds to the average and standard error of the mean for six samples.c) After cycling, a slow analytical measurement ("Full-slow") was used to identify changes in the pseudo-equilibrium voltage-current behavior.

Figure 3 .
Figure 3.The extensively cycled Full-t0 samples were further characterized by a) GI-WAXS and b) SEM.The GIWAXS data after cycling were compared to uncycled samples, the blank FTO glass, and the T-Nb 2 O 5 reference.GI-WAXS patterns were offset vertically for clarity.The SEM images of the samples after 0.25M cycles were similar to the uncycled samples with (c/d) similar pore size distributions and statistically equivalent pore diameter distributions.

Figure 4 .
Figure 4.The electrochemical effects of alternative cycling conditions were examined (Table1) after 10 000 rapid cycles.a) After this cycling, a slow analytical measurement ("Full-slow") was used to compare changes to the pseudo-equilibrium voltage-current behaviors.b) The changes to the lithiation capacities during rapid cycling (Table1) were also compared before and after these 10 000 cycles of each treatment.The reported values in (b) correspond to average values and the standard error-of-the-mean values.

Figure 5 .
Figure 5.The crystal structures resulting from different cycling conditions were examined after 10 000 rapid cycles using a) GI-WAXS.Panels (b/c) show zoomed in sub-regions around T-Nb 2 O 5 peak positions.Data were offset for clarity where the presented sequence matches the legend sequence

Table 1 .
Summary of parameters and results from cycling stability experiments.
a) High-Rate capacities were measured using the voltage window and sweep rate specified for each cycle condition; b) Analytical CV conditions (Full Slow, 1.2-3.0V at 1 mV s −1 ) representing pseudo-equilibrium conditions.