Scholarly article on topic 'High-Rate Intercalation without Nanostructuring in Metastable Nb2O5Bronze Phases'

High-Rate Intercalation without Nanostructuring in Metastable Nb2O5Bronze Phases Academic research paper on "Nano-technology"

Share paper
Academic journal
J. Am. Chem. Soc.
OECD Field of science

Academic research paper on topic "High-Rate Intercalation without Nanostructuring in Metastable Nb2O5Bronze Phases"



Subscriber access provided by UNIV OF NEBRASKA - LINCOLN


High-Rate Intercalation without Nanostructuring in Metastable NbO Bronze Phases

Kent J. Griffith, Alexander C. Forse, John M Griffin, and Clare P. Grey

J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/jacs.6b04345 • Publication Date (Web): 06 Jun 2016

Downloaded from on June 7, 2016

Just Accepted

"Just Accepted" manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides "Just Accepted" as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. "Just Accepted" manuscripts appear in full in PDF format accompanied by an HTML abstract. "Just Accepted" manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). "Just Accepted" is an optional service offered to authors. Therefore, the "Just Accepted" Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the "Just Accepted" Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these "Just Accepted" manuscripts.

Journal of the American Chemical Society is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

8 High-Rate Intercalation without Nanostructur-

90 ing in Metastable Nb2O5 Bronze Phases

11 12 13

15 Kent J. Griffith, Alexander C. Forse, John M. Griffin,

17 Clare P. Grey

21 Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, UK

24 Email:

1 Abstract

4 Nanostructuring and nanosizing have been widely employed to increase the rate capability in a variety

6 of energy storage materials. While nano processing is required for many materials, we show here that

8 both the capacity and rate performance of low temperature bronze-phase TT- and T- polymorphs of

11 Nb2O5 are inherent properties of the bulk crystal structure. Their unique "room-and-pillar" NbO6/NbO7

13 framework structure provides a stable host for lithium intercalation; bond valence sum mapping exposes

15 the degenerate diffusion pathways in the sites (rooms) surrounding the oxygen pillars of this complex

18 structure. Electrochemical analysis of thick films of micrometer-sized, insulating niobia particles indi-

20 cates that the capacity of the T-phase, measured over a fixed potential window, is limited only by the

23 Ohmic drop up to at least 60C (12.1 A g ), while the higher temperature (Wadsley-Roth, crystallo-

25 graphic shear structure) H-phase shows high intercalation capacity (> 200 mA h g ) but only at moder-

27 ate rates. High-resolution 6/7Li solid-state nuclear magnetic resonance (NMR) of T-Nb2O5 revealed two

30 distinct spin reservoirs, a small initial rigid population and a majority-component mobile distribution of

32 lithium. Variable temperature NMR showed lithium dynamics for the majority lithium characterized by

34 very low activation energies of 58(2) to 98(1) meV. The fast rate, high density, good gravimetric capac-

37 ity, excellent capacity retention, and safety features of bulk, insulating Nb2O5 synthesized in a single

39 step at relatively low temperatures suggest that this material is not only structurally and electronically

44 high rate performance without nanostructuring in a complex insulating oxide expands the field for bat-

46 tery material exploration beyond conventional strategies and structural motifs.

exceptional but merits consideration for a range of further applications. In addition, the realization of

10 11 12

20 21 22


There is a growing need for high-power, high-capacity energy storage materials for applications that require faster rate performance than traditional battery materials can offer, along with higher charge storage capability than can be achieved with supercapacitor systems. Unless nanosized, conversion materials and two-phase intercalation compounds commonly used in lithium-ion batteries generally do not deliver high power due to the kinetics associated with major structural transformations. Electric double-layer capacitors (EDLCs) can deliver high rate performance but are limited to relatively low volumetric

and areal energy densities as redox reactions offer the opportunity for 10-100 times greater charge stor-

age than the electrostatic mechanism of EDLCs. ' In lithium-ion batteries, realization of the maximum capacity of an electrode material in a given potential window is inherently dependent upon the ability of lithium to reach the particle interior. This has generally limited high rate performance to materials with short diffusion distances typically achieved via nanoscaling or nanostructuring of the particles.

Figure 1 - Crystal structures of (a-b) T-Nb2O5 (c) B-Nb2O5 and (d) H-Nb2O5. Oxygen and (partially occupied) niobium sites are represented by orange and (partially filled) blue spheres, respectively.

and capacity as intercalation electrodes. In this voltage range, safety and stability issues from SEI for-

5 The disadvantages of the synthesis and use of nanoparticles and nanoarchitectures for battery applica-

8 tions are well known: high surface area leading to increased dissolution and increased surface-

10 electrolyte interface (SEI) formation, low packing density, toxicity, high cost, chemical waste genera-

12 tion, scalability issues, and often many-step synthetic complexity.3-5 Preparation of energy-dense mate-

15 rials with high capacity and good rate performance through a simple and efficient synthetic route is

17 clearly desirable but evidence from e.g. Li4Ti5O12, LiFePO4, and TiO2 suggests that this is not generally

19 observed.

24 Niobium(V) oxides, in the potential window ca. +2.0 to +1.0 V vs. Li+/Li, have shown impressive rate

29 mation associated with electrolyte instability below +1.0 V vs. Li+/Li, as well as lithium dendrite for-

31 mation, can be avoided.6-8 Furthermore, as lithium does not alloy with aluminum until 300 mV vs.

33 Li+/Li9, copper foil can be substituted with significantly cheaper and lighter aluminum foil as the anode

36 current collector. The phase stability of Nb2O5 is complex and at ambient temperature and pressure

38 many metastable polymorphs exist that depend on heating conditions, precursors, and synthesis proce-

41 dures10. In this study, we examine four polymorphs: TT-Nb2O511-13, T-Nb2O5 (Pbam, Figure 1a-b)14,

42 15 16

43 B-Nb2O5 (C2/c, Figure 1c)15, and H-Nb2O5 (P2/m, Figure 1d)16. Regarding nomenclature, the T desig-

45 nation comes from the German tief, which means low, as in low temperature; the B polymorph was

48 named for its crystal habit (blatter or bladed); and the H designates high as it is the high temperature

50 phase.10 T-Nb2O5 is structurally similar to the tungsten bronzes but composed of primarily highly dis-

53 torted octahedral and pentagonal bipyramidal Nb environments rather than regular octahedra. Partially

55 occupied niobium sites (totaling 0.8 of the 16.8 niobium atoms per Nb168O42 unit cell) with high coordi-

nation and long Nb-O interatomic distances are proposed in the crystal structure to balance the charge

3 from 42 oxygen atoms per unit cell.14 The TT-phase is a metastable structure, whose structural details

5 are not fully understood but are apparently similar to T-Nb2O511. On the basis of diffraction data, TT-

10 lation of niobium atoms above and below the mirror plane at (x, 0.5, z) in T-Nb2O5 . Monoclinic B-

12 Nb2O5 possesses a TiO2(B)-like structure. H-Nb2O5, another monoclinic polymorph and the phase

15 which is not known to convert to any other polymorph as a function of temperature, fits into the Wads-

17 ley-Roth family of crystallographic shear structures with (3 x 4)1 and (3 x 5)OT ReO3-like blocks of oc-

19 tahedra. Shared octahedral edges along shear planes at the periphery of the blocks account for the oxy-21

22 gen:metal ratio of 2.5. The blocks are infinite parallel to b and the subscripts 1 and denote block con-

24 nectivity in the ac plane .

29 Recent studies on the TT- and T-polymorphs have shown excellent rate and cycle life performance on a

31 variety of nanoscaled and nanostructured morphologies. Examples include nanobelts18, nanofibers19,

33 nanosheets20'21, nanoparticles/nanocrystals, templated mesoporous nanoparticles22-27, nanocrystalline

28 29 30

36 Nb2O5/carbon nanotube nanocomposites , nanocrystalline Nb2O5/graphene nanocomposites , , nano-

38 crystalline Nb2O5/carbide-derived carbon nanocomposites31, and Nb2O5@carbon core-shell nanocrys-

41 tals . While some electrochemical and structural aspects of bulk TT-, T-, and H-Nb2O5, prepared from

43 solid-state methods have been investigated , , their rate behavior has evidently not been characterized.

45 In this work, we examine the electrochemical properties of a family of Nb2O5 polymorphs synthesized

48 via oxidation of niobium(IV) oxide. This simple solid-state route enables access to the low temperature

50 metastable phases and has been used to produce TT-, T-, B-, and H-Nb2O5 previously10; however, few

55 cedure has been reported; (ii) no electrochemistry of Nb2O5 phases from NbO2 has been reported; and

Nb2O5 has been described as a disordered modification of T-Nb2O5, the disorder being related to modu-

details were given in the earlier study and to the best of our knowledge, (i) no systematic synthetic pro-

na, and non-bulk-like local atomic structure environments, facilitating a study of the inherent properties

1 (iii) the electrochemistry of B-Nb2O5, irrespective of precursor, has not been previously discussed.

3 Along with the ease of synthesis relative to nanostructuring, solid-state methods produce larger particles

5 with decreased contributions from electric double-layer capacitance, surface and near-surface phenome-

10 of Nb2O5. Lithium environments and dynamics from Li solid-state NMR and bond valence sum maps

12 performed as part of this study aid the understanding of the structural origin of the anomalous electro-

15 chemical performance in micrometer-sized metal oxides. These tools show degenerate diffusion path-

17 ways with a very low activation energy for Li motion on the order of a few kBT (ca. 60-100 meV).

19 Through this investigation, we find that the accessible capacity and rate performance of the low temper-21

22 ature Nb2O5 polymorphs is a general feature of the structure type and not a function of nanoscaling or

24 nanostructuring.

29 Experimental Methods

32 Synthesis. Nb2O5 polymorphs were synthesized by heating separate aliquots of NbO2 (Alfa Aesar,

34 99.5+%) in alumina crucibles in air for 24 hours at 19 temperature points between 200 and 1100 °C

39 temperature was attained; the temperature was then held for 24 hours before ambient cooling. Thermal

44 ples were placed in a tared 100 ^L alumina crucible and the mass was recorded from 50 °C to 1000 °C

45 -1 -1

46 in steps of 1 °C min-1 under constant air flow (50 mLmin-1). A blank, with an empty crucible, was

51 ically differentiated to obtain differential thermogravimetry (DTA) curves.

56 ing of Nb2O5, Super P carbon (Timcal), and Kynar® polyvinylidene difluoride (PVDF, Arkema) in an

spaced at 50 °C increments. The samples were heated at a ramp rate of 10 °C-min 1 until the desired

gravimetric analysis (TGA) was performed on a Mettler Toledo TGA/SDTA 851 thermobalance. Sam-

recorded under the same heating conditions and subtracted from the sample data. The data were numer-

Electrochemistry. Cathodes (in half-cell configuration) were prepared by agate mortar and pestle grind-

ostat/galvanostat instrument running EC-Lab® software. In each coin cell, the niobium ox-

1 8:1:1 mass ratio with #-methyl-2-pyrrolidone (NMP, Sigma-Aldrich, 99.5%, anhydrous) to form a vis-

3 cous slurry. High-carbon proof-of-concept electrodes were prepared in a similar manner but with an

5 active material:carbon:binder ratio of 3:6:1. The slurry was tape cast onto an aluminum foil current col-

8 lector with a 150-200 pm doctor blade. After drying for at least 12 h at 60 °C, 1 cm circular cathodes

10 were cut via a punch press. Unless otherwise noted, mass loadings were 4-6 mg Nb2O5-cm . Coin cell

12 batteries were constructed in an argon-filled glove box with O2 and H2O levels < 1 ppm. The electro-

15 lyte-solvent system consisted of 1 M LiPF6 dissolved in a 1:1 volume ratio of ethylene car-

17 bonate/dimethyl carbonate (EC:DMC, Merck). The aforementioned cathode, a glass microfiber (What-

19 man) separator saturated with electrolyte-solvent, a lithium metal foil anode (Sigma-Aldrich, 99.9%) on 21

22 a stainless steel current collector, and a wave spring were compressed within a standard 2032-type coin

24 cell casing. All electrochemical measurements were performed with a Bio-Logic potenti-

29 ide/carbon/binder film served as the working electrode and the lithium metal as both the counter elec-

31 trode and reference electrode. Discharge/charge took place within the range +3.0 V to +1.2 V or +1.0 V

34 with respect to Li+/Li. Galvanostatic charge/discharge data were numerically differentiated to give dif-

36 ferential voltage curves. Specific cycling conditions are denoted in the text for each experiment. For

38 the preparation of samples for NMR analysis, thick (150-750 mg) pellets of pure Nb2O5 powder were

41 cold pressed at 5 MPa and assembled into coin cells as above. These pellets were discharged at 10042 2

43 500 pAcm until a desired degree of lithiation was reached. For clarity, in this article C rate refers to

45 inverse hours required to reach the theoretical capacity of 201.7

47 -1 -1

48 mA h gNb2o5 (e.g. C/5 implies a current of 40.34 mAgNb2O5 ) and discharge refers to lithium insertion

49 5+ 4+

50 into the Nb2O5 structure. Theoretical capacity for a one electron reduction from Nb5+ to Nb4+ is 201.7

52 mA h g based on the mass of Nb2O5.

Morphological Analysis. Brunauer-Emmett-Teller (BET) surface areas were obtained from nitrogen

1 X-Ray Diffraction (XRD). Laboratory powder x-ray diffraction (PXRD) patterns were recorded at

3 room temperature on a Panalytical Empyrean diffractometer emitting Cu Ka (1.540598 A + 1.544426

5 A) radiation. Patterns were recorded from 5-80° 20 in steps of 0.017° 20. Rietveld analysis was per-

7 35 36 37

8 formed in GSAS and GSAS-II with the aid of the CMPR toolkit . Crystal structures and isosurfaces

10 (with bond valence energy level cut-off set to 2.0 eV, see Discussion) were visualized in VESTA 3.0 .

15 adsorption isotherms at 77 K. BET data were collected on a TriStar 3000 gas adsorption analyzer (Mi-

17 cromeritics Instrument Corp., V6.08). Scanning electron microscopy (SEM) images were taken with a

20 Eigma VP microscope (Zeiss). Tap density was recorded on an AutoTap (Quantachrome Instruments)

22 instrument operating at 257 taps min . The tap densities were measured according to ASTM interna-

24 tional standard B527-15 modified to accommodate a 10 cm3 graduated cylinder. 26

27 Li Solid-State Nuclear Magnetic Resonance. One- and two-dimensional solid-state NMR experi-

29 ments were performed in a 4.0 mm probehead on a 200 MHz Bruker Avance III spectrometer at the 7Li

32 Larmor frequency of 77.7 MHz and on a 700 MHz Bruker Avance III spectrometer at the Li Larmor

34 frequency of 103.0 MHz and Li Larmor frequency of 272.0 MHz. Magic angle spinning (MAS) up to

37 14 kHz was applied; the specific rotational frequency is denoted for each experiment. One-dimensional

39 spectra were recorded with a single n/2 pulse or Hahn-echo pulse sequence; the applied n/2 pulse

40 6 7

41 lengths for 6Li and 7Li were 7.00 p,s and 2.60-3.75 p,s, respectively. Two-dimensional exchange spec-

46 t2 pulse sequence where tmixing is a variable mixing period during which the nuclei are allowed to interact

48 via chemical and/or spin diffusion. Spectra were collected as a function of temperature for 1D, 2D,

51 spin-lattice (r1), and spin-spin (T2) relaxation measurements of selected samples. A saturation recovery

53 pulse sequence was employed for T1 measurements: a series of n/2 pulses was first applied to eliminate

55 bulk magnetization (M0) along z and then the z-magnetization was allowed to relax during a recovery

troscopy (EXSY) measurements were performed with a rotor-synchronized n/2 - t1 - n/2 - tmixing - n/2 -

time (t) before recording the magnetization (Mt) over a range of recovery times. A variable-delay spin echo sequence was used to measure T2. As the sample temperature cannot be measured directly with sufficient accuracy under MAS conditions, a temperature calibration was performed based on the tem-

10 11 12

20 21 22

207 39

perature-dependent shift of Pb in Pb(NO3)2 (Supplementary Figure S1). All samples were ground with an agate mortar and pestle and packed into 4.0 mm ZrO2 rotors in an Ar-filled glovebox with < 1 ppm O2 and < 1 ppm H2O; lithiated samples were washed with dimethyl carbonate (3 x 3 mL) to remove any residual LiPF6 and dried in vacuo before being ground and packed. The 6 7Li spectra were referenced with a secondary reference, (6/7Li 1:1 at./at.) Li2CO3, at +1.1 ppm versus the 1.0 M LiCl (aq.) primary reference at 0.0 ppm.

Results and Discussion

1. Structural Characterization of Nb2O5 Polymorphs. Thermal gravimetric analysis of NbO2 and Nb (Supplementary Figure S2) showed that the onset of oxidation for NbO2 occurs at a temperature significantly below that for oxidation of Nb metal—290 °C versus 420 °C—allowing a greater range of meta-

stable phases to be prepared with this starting material. A systematic X-ray diffraction investigation of thermal oxidation of NbO2 revealed that four from the range of polymorphs with a nominal composition of Nb2O5 could be observed upon elevated-temperature oxidation of NbO2 (Figure 2). While some changes are apparent after 24 h at 250 °C, NbO2 does not oxidize to Nb2O5 until ca. 300 °C in

Figure 2 - XRD patterns of phases observed air' тт-^05 being the first oxidized phase that is upon heating NbO2 in air. The patterns dominated by Nb02, TT-Nb205, T-Nb205, B-Nb205, and H-Nb205 are shown in orange, black, red, blue, and green, respectively. Lines with more than one color indicate a temperature where a significant ACSParagon Plus Environment more than one phase was present.

10 11 12

20 21 22

observed. Three further irreversible phase transitions are observed at higher temperature: nearly phasepure T-Nb2O5 is found from approximately 550-600°C, B-Nb2O5 is observed from 700-850 °C, and H-Nb2O5 is observed at and above 900 °C. Rietveld analysis (Supplementary Figure S3) revealed minor impurities from the reactant or other polymorphs in all but the H-phase. A color change is observed from deep blue, d1, niobium(IV) to white, d0, niobium(V) for all samples heated above 400 °C. This is an indication that the oxidation is complete for the T-, B-, and H-phases. The TT-phase is a pale-grey blue and thus may retain a small (ca. <0.5%)10 fraction of niobium(IV). The observed color changes are in accordance with the color change from white, through grey-blue, toward dark blue observed as lithium is inserted into any of the Nb2O5 structures in this study and elsewhere40.

Unlike the step changes in crystal structure from NbO2 to orthorhombic and then through two distinct monoclinic Nb2O5 polymorphs, the surface and particle morphology, as viewed by SEM (Figure 3 and Supplementary Figure S4), transformed more smoothly upon heating. After treatment at

300°C in the TT-phase, anhedral particles (i.e. Figure 3 - SEM images showing the particle sizes and morphologies of the Nb205 polymorphs without well-formed crystal faces) with cracked-obtained from oxidation of Nb02; phase and

synthesis temperature are denoted. mud-like topology dominated which, at 600 °C in

the T-phase, had partially annealed. The samples heated to 850 °C yielded rounded steps on interconnected subhedral particles of the B-phase, which, after 1100 °C treatment crystallized further to capped euhedral particles of several micrometers with distinctly striated edges in the H-phase. In addition to the smaller primary particle features, a larger secondary particle size (more visible in Supplementary Figure S4) can be considered which corresponds to discrete particles composed of interconnected primary particles; this secondary particle size is of the order of tens of micrometers. BET surface area 10

2 —1

measurements revealed a quite small surface area, on the order of 1—2 m •g , for all phases. For further particle characterization, tap density was measured and reported with BET surface area in Table 1. As volumetric, rather than gravimetric, energy density dictates many applications, the tap density of Nb2O5 from solid-state synthesis is noteworthy at 1.2 — 1.8 gcm . By comparison, commercial TiO2 na-nopowders of 5 — 20 nm have almost an order of magnitude smaller tap density (0.12 to 0.24 gcm-3)41. Thus, the Nb2O5 phases examined in this study differ in structure and morphology but all are ^m-scale, dense, and extremely low surface area.

10 11 12

20 21 22

Phase Synthesis Temperature (°C) BET Surface Area (m2g-1) Tap Density (g^cm-3)

TT-Nb2Os 300 1.8 ± 0.1 1.4 ± 0.1

T-Nb2Os 600 1.9 ± 0.1 1.4 ± 0.1

B-Nb2O5 850 1.5 ± 0.1 1.2 ± 0.1

H-Nb2Os 1100 0.7 ± 0.1 1.8 ± 0.1

Table 1. Physical properties of micrometer-sized Nb2O5. BET surface area, as determined from N2 adsorption isotherms at 77 K, and tap density of four polymorphs of Nb2O5 synthesized from NbO2 via thermal oxidation.

2. Electrochemical Properties of Nb2O5 Polymorphs. Electrochemical lithiation of NbO2 and the four

Nb2O5 polymorphs revealed a range of structure-driven mechanisms. Chronopotentiometric discharge

and charge were performed at a rate of C/10 where kinetic limitations should be minor. The results

(Figure 4a), demonstrate (i) a high intercalation capacity and three distinct regions for H-Nb2O5; (ii) a

close-to-linear sloping voltage profile for T- and TT-Nb2O5; (iii) a very small capacity for B-Nb2O5 and

NbO2 and (iv) a low overpotential between charge and discharge for all phases in the electrochemical

window 3.0 to 1.2 V vs. Li+/Li. Cycling studies (Figure 4b) were performed with galvanostatic dis-11

mAhg 1 after 100 cycles. Conversely, B-Nb2O5 stored only 20 mAhg 1 on first discharge but re-

1 charge and charge at 1C for T- and TT-Nb2O5 and C/10 for B- and H-Nb2O5 with an additional constant

3 voltage charge step (CCCV charging) at 3.0 V that is widely employed to optimize charging while pre-

5 venting overcharge. H-Nb2O5 showed a first cycle capacity of 235 mAhg-1 but dropped to 175

10 tained that capacity with no diminution over the recorded cycle range. After a first cycle loss of ~25

12 mAhg-1, TT- and T-Nb2O5 exhibited reversible capacities of 165 and 160 mAhg-1, respectively, for

15 100 cycles with the capacity of the TT-phase edging slightly upward after the initial capacity loss. Dif-

20 21 22

10 11 12

20 21 22

ferential capacity plots (Figure 4c) reveal the (de)lithiation behavior in more detail. H-Nb2O5 exhibits reversible peaks centered at 2.05 V, 1.67 V, 1.42 V, and 1.22 V with a significant amount of charge stored at intermediate potentials between the peaks. B-Nb2O5 and NbO2 exhibit only one small peak at 1.68 V and 1.71 V, respectively; however, the cathodic-anodic peak separation in NbO2 is much greater, indicating a kinetic limitation even at C/10. To the best of our knowledge, bulk NbO2 intercalation

250 225 200

t 150 E

> 125 -1—|

TO 100

u 75 50 25 0

50 75 100125 150175 200 225 Capacity (mA-h-g"1) r

▼ H-Nb205 -

- A B-Nb205 -

_ • T-Nb205 .

- ■ TT-Nb205 J


400 200 0 -200 -400 400 200 0

-200 -400 400 200 0

-200 -400 400 200 0

-200 -400 400 200 0 -200 -400

40 60 Cycle (Number)

- 1 1 1 II 1 1 1 __ /fiV . 1 1 1 - H-Nb O 2 5 _

~ r Vf 1 , IU 1 i 1 1 . 1 ~

k B-Nb O _ 2 5

-1 1,1,1,1 1,1-

T-Nb O _ 25

-1 I.I.I.I 1 . 1 -

TT-Nb O ~ 25

-1 1,1,1,1 1,1-

-1 ....... 1,1-

1.2 1.4 1.6 1.8 2.0 Potential (V) 2.2 2.4

Figure 4 - Galvanostatic discharge-charge curves for the Nb2O5 polymorphs from 3.0 to 1.2 V (a) Electrochemical discharge/charge profiles obtained at a rate corresponding to C/10 (b) Cycle tests at either 1C (T, TT) or C/10 discharge/charge with a constant voltage charge step (B, H). (c) Differential capacity plots derived from the discharge/charge profiles in (a); solid lines are on the same scale while the dashed green line is shown at 1/20 scale to reveal the peak positions in H-Nb2O5

1 has not been reported but is evidently not significant between 3.0 and 1.2 V vs. Li+/Li. NbO2 electro-

3 chemical behavior is unlike that seen for the 2H layered transition metal dichalcogenide phase

5 LiNbO242, though both involve the Nb4+/Nb3+ redox couple. In LiNbO2 the reversible delithiation from

0 LiNbO2 to Li05NbO2 occurs from 2.5-3.0 V43. TT- and T-Nb2O5 store charge nearly evenly across the

10 potential range 1.9 V to 1.2 V with minor peaks at 1.68 V. Accounting for the minor quantity of B- and

12 H-Nb2O5 present in the lower temperature phases, these small dQ/dV peaks may be related to one of the

15 higher temperature polymorphs. Note that although the capacity of the TT- and T-phases in Figure 4

17 corresponds to ca. 0.8 Li per Nb, one full equivalent of lithium can be intercalated if the potential cutoff

19 is moved below 1.0 V. The limit of 1.2 V was chosen because a lower cutoff voltage resulted in de-21

22 creased cycle life at least in part due to SEI formation. The discharge profile features, cycle life, and

24 capacity for TT-, T-, and H-Nb2O5 at low current density (slow rate) are in agreement with previous

26 work from solid-state33 and nanostructured19'25'44 synthesis methods.

10 11 12

20 21 22

JE G 100

ro Q. ro U

c/io c/s : ^


-Nb205 x B-Nb205 • T-Nb205 ■ TT-Nb205

Xxxxx^xxxxx _l_

5 10 15 20 25 30 35 Cycle (Number)

3. Charge Storage Kinetics of Micrometer-Sized

Nb2O5. Charge storage kinetics, i.e. rate dependences, are contingent upon a range of interrelated factors such as atomic structure, electronic structure, particle size, particle geometry, and intra- and inter-particle porosity. Numerous studies have focused on the latter, microstructural, aspects of Nb2O5 by designing

creative synthetic strategies to attain nanocrystals ' ,

1 1 1 10C C/10

■ ■■■■



Figure 5 - Relationship between discharge nanowires45, nanosheets20'21, and even hierarchical rate and capacity for thick films (4-6

mg-crrf2) of bulk H-, B-, T-, and TT-Nb205 structures of nanoparticles containing nano-scale po-between 3.0 and 1.2 V vs. Li+/Li.

rosity22'46. To determine whether the high rate properties discussed in the literature for Nb2O5 are limited to nanoscale or nanostructured analogues, a rate

performance study (Figure 5) was conducted on thick (4-6 mg cm ) electrodes of the micrometer-sized Nb2O5 particles. Under these conditions, the capacity of H-Nb2O5 is highly correlated with discharge rate; a change from C/10 to 1C causes a 50% decrease in capacity and no charge storage is observed beyond 10C. The capacity of B-Nb2O5 is only weakly rate dependent but a maximum value of 20 mAhg-1 means that only 10% of the niobium sites are reduced; this suggests that a surface or near-surface redox process is occurring and thus a weak structure-rate relationship might be expected. Significantly, T- and TT-Nb2O5 retain much of their initial capacity under fast discharge conditions: 150

10 11 12

a 3.0 2.8 2.6

I 1 I 1 I 1 I 1 I 1 I 1 I 1 I 1 I 1

T-Nb2O5 -C/50

• - - ■ C/10

21 3.0

22 2.8

24 2.6

25 2.4

27 2.2

29 c (LI 2.0

30 O CL 1.8

32 1.6

33 1.4

35 1.2

■ i i i i i i i i i 1 1 1 1 1 1 1 1

r H-Nb2O5 -C/50 ~

— - - C/10 "

---C/3 J

- i -----10C -

\\ ■

J \ -- —

\ Nx _

■ i ■ ' ...... \ . 1 . 1 . 1 . 1

0 25 50 75 100125 150175 200 225 Capacity (mA-h-g1)

Figure 6 - Discharge behavior of (a) T-Nb2O5 and (b) H-Nb2O5 over ca. three orders of magnitude of current density. After five cycles, cells were cycled between 3.0 and 1.2 V vs. Li+/Li at the rate shown. The charging step is rate limiting for T-Nb2O5; discharge was examined after CCCV charging so as to always

Journal of the American Chemical Society Page 16 of 35

mA-h-g-1 at 5C and 100-120 mA-h-g-1 at 10C (3.5 minute discharge for a capacity of 120 mA-h-g-1). A plot of discharge potential vs. capacity as a function of current (Figure 6) for (a) T-Nb2O5 and (b) H-Nb2O5 displays fundamental differences between the lithiation of these polymorphs. At the initial discharge current pulse, both phases show a potential drop down to the niobium redox potential with a superimposed IR drop associated with the total resistance from the electrode and cell components. From this stage, under a small current with a consequently small Ohmic contribution, T-Nb2O5 shows an approximately linear relationship between potential and charge while H-Nb2O5 contains distinct shoulders and a plateau. At higher rates, the initial intercalation voltage of T-Nb2O5 decreases, as expected due to the increased Ohmic contribution; however, not only does the Q-V relationship remain linear but the slope of that line is nearly parallel across more than three orders of magnitude of current. Thus, it seems that, over a very wide range

examine the discharge behavior of the fully of current? ^ amount of charge stored in ^^ delithiated material.

is determined by the Ohmic potential loss that effectively narrows the potential window for charge storage. Conversely, intercalation into H-Nb2O5 was 16

10 11 12

20 21 22

severely rate limited even at relatively low currents. This evidence suggests a structural (i.e., Li+ transport) limitation such that, even when Ohmic effects are not significant, H-Nb2O5 cannot intercalate lithium rapidly. For example, at C/3 the initial intercalation potential of H-Nb2O5 was approximately

2.1 V but the shoulder normally at 2.05 V and plateau at 1.67 V were still dampened. In order to test the hypothesis that the electrode/cell resistance is capacity limiting in T-Nb205, proof-of-concept 1-2 mg electrodes with additional conductive carbon were constructed and tested under high rate conditions (Figure 7). When the initial potential drop is decreased, the trend in linear intercalation with constant slope is

0 20 40 60 80 100 120 140 160 180 found to extend to even higher rates Capacities of Capacity (mA-h-g

Figure 7 - Discharge and charge curves of T- 119 mA'h'^ and 85 mA'h ^ were realized in 71 s

Nb205 at current densities ranging from 1C „ _ A , , _ „„ „ A _i

A , mr- Mn , A -u A u- u at 6.05 Ag 1 30C and 25 s at 12.1 Ag 1 (60C),

0.2 A-e to 60C 12.1 A-e . A high con- & v ' & v h

ductive carbon ratio (2:1) was used to study .. , A ,

v ' 1 respectively. Again, the discharge capacity appears to

the discharge profiles at extremely high

be limited by IR drop rather than particle size to at

rates where Ohmic losses are significant. After five cycles, cells were cycled between

3.0 and 1.2 V vs. Li+/Li at the rate shown. ieast 60C despite the fact that is a bulk material The charging step is rate limiting for T-

Nb205; discharge was examined after CCCV governed by solid-state diffusion and not surface/near-charging for uniform comparison.

surface reactions. At the high rates and currents required (>10 mA), coin cell resistance, Li+ transport in the electrolyte (particularly between the particles within the electrode film), and lithium anode kinetics all become significant; it is clear that advanced cell design and decreasing the total resistance of the cell/electrode configuration, through e.g., carbon coating or calendaring is important to reach the limits of Nb2O5.

10 11 12

20 21 22

Further characterization of the high rate phases to study the effects of charge-limitation on reversible capacity at different potential windows (Supplementary Figure S5a) and rates (Supplementary Figure S5b) are presented in the supporting information. Upon cycling in a wider potential window (3.0 V -1.0 V), a tradeoff is observed between capacity and long-term capacity retention (S5a). Electrochemical cycling without CCCV charging is unable to remove as much lithium from the Nb2O5 framework, resulting in a ca. 25% lower reversible capacity (S5a-b). Even without a potentiostatic charging step, the capacity retention is current-independent and stable throughout 300 cycles (S5b).

T-Li Nb O

1.78 2 5

T-Li Nb O

1.56 2 5

T-Li Nb O

1.28 2 5

T-Li Nb O

0.96 2 5


0 -10 -20 -30 6 'Li (ppm)

1.86 2 S

1.60 2 5

1.28 2 S

0.96 2 5

0.64 2 5

| I I I I | I I I I | I I I I | I I I I | I I I I |

20 10 0 -10 -20 -30

6 'Li (ppm)

4. Lithium Dynamics. In order to quantify the lithium transport that appears to be inherent to T-Nb2O5, 6Li and Li MAS NMR

were investigated. Paramagnetic compounds can exhibit lithium resonances shifted over

a range of many hundred ppm , , due to the Fermi contact interaction, yet, despite containing d1 niobi-um(IV), LixNb2O5 shows 'Li NMR

resonances (Figure 8) within a small frequency range for all values of x and longer T1 relaxation values (~ 1s) than expected for 7Li in paramagnetic solids. These values are, however, similar to those observed for Ti3+-containing paramagnetic materials49. The spectra shown in Figure 8 and Supplementary Figure S6 show the presence of two discrete environments for intercalated lithium. The sharp feature in Figure 8a comes from residual

Figure 8 - 6/7Li MAS NMR of LixNb205. (a) 7Li spectra at 9 kHz MAS and 4.7 T (b) 6Li spectra at 9 kHz MAS and 16.4 T. The high-resolution 6Li spectra suggest a distribution of similar lithium local ACSParagon Plus Environment ments.

LiPF6 from the electrolyte, and as seen in Figure S6 disappears entirely upon rinsing with dimethyl car-

3 bonate; no observable effect is seen on the peaks from intercalated lithium. The 6Li spectra at high

5 magnetic field strength (Figure 8b) should in principle give the best resolution. This is a result of re-

8 duced dipolar broadening from the smaller natural abundance and gyromagnetic ratio of Li vs. Li as

10 well as reduced quadrupolar broadening due to both the smaller nuclear quadrupole moment of Li and

12 the higher magnetic field. Despite all this, only a moderate reduction in line width is observed between

14 7 6

15 the Li spectra at a low magnetic field and Li spectra at high magnetic field, which suggests that the

17 peak widths are dominated by the range of shifts caused by a distribution of local environments. Upon

19 discharge, the first lithium ions occupy a site that gives rise to a resonance at 2 ppm. Above Li02Nb2O5, 21

22 the lithium begin to occupy one or multiple sites with resonances at approximately -5 ppm. As lithia-

10 11 12

20 21 22

Figure 9 - Variable temperature 7Li MAS NMR of LixNb2O5 at 12.5 kHz at 16.4 T. The central line-shape and rotational sideband manifold (marked by asterisks) are shown for x = 0.44 and x = 1.86 from 306 K to 435 K. As the ions become more mobile, sideband intensity decreases and the central line narrows. T-Li0.44Nb2O5 spectra were normalised to account for T2 magnetization loss during the echo pulse sequence.

tion increases, site(s) with negative shifts are further populated and a positive shift is observed for all resonances. Given the structural complexity of T-Nb2O5 with multiple reasonable lithium positions and the lack of crystal structure data for the lithiated compound, there are not presently known crystallographic sites to compare with the local environments derived from NMR.

The collection of high signal-to-noise spectra of 7Li and natural abundance 6Li in a matter of

minutes was facilitated by the use of thick pellet electrodes. Despite the size (ca. 400-500 mg) and lack of conductive carbon or binder in the pellet design, the structure and electrochemistry upon discharge is identical between the pellets and a conventional film as confirmed by chronopotentiometry (Supplementary Figure S7). Furthermore, lithiated T-Nb2O5 appears to be quite stable; no changes were observed in the NMR spectra over the course of several months even after intentional exposure to air for a period of 24 h.

10 11 12

20 21 22

Variable temperature 6,7Li MAS NMR experiments were used to investigate timescales for lithium motion and the electronic structure of lithiated Nb2O5. T1 (spin-lattice) and T2 (spin-spin) relaxation rates, as well as 1D lineshape and 2D spin exchange measurements were recorded over a range of temperatures at low and high magnetic field for selected samples. Figure 9 displays the effect of temperature on the lineshape and shift of Li in LixNb2O5 at x = 0.44 and 1.86. Motional line narrowing and loss of rotational sideband intensity is observed which is ascribed to lithium mobility, which averages the chemical shift anisotropy (CSA) and quadrupolar

Figure 10 - The trend in correlation times derived from the BPP model display Arrhe-nius behavior. Lithium hopping was found to occur in electrochemically lithiated T-

interaction50. As the temperature is increased, the

LixNb205 with Ea = 98 1 meV, 91 6 meV, ^ . , „ , . ^

, , , , ' lithium resonance at ca. 1 ppm gradually begins to

83(5) meV, 68(4) meV, and 58(2) meV for x FF & J &

= 0.64, 0.96, 1.28, 1.56, and 1.78, respectively.

participate in the rapid process at lower frequency.

Above 400 K, the average Li shift moves 1-2 ppm lower in frequency for highly lithiated samples at both 16.4 T (Figure 9) and 4.7 T (not shown). Wagemaker et al5 observed this effect in TiO2 and attributed the increased shielding to a lithium nucleus-electron interaction. Through similar arguments about a small shift range and long T1s, they concluded that conduction, rather than paramagnetism, was the dominant interaction, accounting for the small observed Li shifts in their system. In addition to electronic changes, the sampling of environments via chemical exchange is temperature-dependent; thus, the nature of the shift may be some combination of conduction and diffusion effects in both the 21

ductive additives.

1 niobium and titanium oxides. We note that an increase in conductivity upon lithiation does suggest de-

3 localization of electrons through the interconnected Nb-O-Nb network in T-Nb2O5 and offers an expla-

5 nation for the high-rate and complete lithiation of insulating pellets of T-Nb2O5 in the absence of con-

12 The spin-lattice relaxation time of 7Li was on the order of 1 s for all lithium concentrations (Figure S8).

15 Spin-spin relaxation was on the order of 1 ms but not strongly temperature dependent in the explored

17 region. The aforementioned weak dependence of shift of the resonances on temperature suggests that

19 magnetic effects do not dominate the T1 changes. Employing the Bloembergen, Purcell, and Pound 21

22 (BPP) model for random jump diffusion and assuming relaxation dominated by the Li quadrupolar

24 interaction (see Supplementary Information for further details)53'54 at the lower magnetic field strength

27 of 4.7 T, correlation times and trends for lithium motion were extracted from the variable temperature T1

29 data. Single exponential fitting of the T1 relaxation curves was achieved for samples with intermediate

31 to high lithium content; samples below ca. x = 0.5 in LixNb2O5 showed a complex multiexponential re-

36 resonances, which becomes less significant as the lithium content increases and the mobile lithium sub-

38 lattice increasingly comprises the overall signal. Strictly, the isotropic BPP model applies to 3D diffu-

41 sion; however, in the low temperature regime of the BPP model where T1 decreases with increasing

42 55 56

43 temperature, 2D diffusion—expected in T-Nb2O5—is theoretically55 and experimentally56 shown to

45 yield indistinguishable results. Determination of activation energies via the BPP model (Figure 10)

48 yields 98(1) meV, 91(6) meV, 83(5) meV, 68(4) meV, and 58(2) meV for T-LixNb2O5 with x = 0.64,

50 0.96, 1.28, 1.56, and 1.78, respectively.

laxation curve. The multiexponential behavior is likely due to the overlapping rigid and mobile lithium

10 11 12

20 21 22

1111111111111111ii11111111111 5 0 -5-10 F2 [ppm]

I I I I | I I II | I I I I| I I I I | I I I I | I I I I 1 S 0 -5-10 F2 [ppm]

I I I I | I I I I | I I I I | I I I I | I I I I | I I I I 5 0 -5-10 F2 [ppm]

Figure 11 - 7Li EXSY spectra of T-Li0.44Nb2O5 with mixing periods of (a) 100 ^s (b) 1 ms (c) 5 ms (d) 10 ms (e) 20 ms and (f) 100 ms. Spectra were collected at 12.5 kHz MAS, 306 K, and 16.4 T.

Rotor-synchronized MAS two-dimensional exchange spectroscopy (EXSY) data were also recorded for Li. These experiments have proven to be a useful tool to probe dynamics on a microsecond to milli-

57 58 7

second timescale. , Figure 11 shows the Li exchange spectra of T-Li044Nb2O5 as a function of mixing time at 16.4 T and 12.5 kHz MAS. The presence of a narrow diagonal in Figure 11a provides direct evidence that the peaks in the 1D spectra are not true singular broad components but comprised of distributions of individual lithium environments. Exchange between these local environments occurs when the spins are allowed to interact for 1 ms or longer. On the other hand, the intensity pattern and cross sections through the spectra indicate that exchange on this timescale is not occurring between the ca. 1

ppm resonance and the lower frequency distribution of sites. Thus, the lithium in T-Nb2O5 can be more 23

ture EXSY spectra showed a strong temperature dependence, which is indicative that chemical ex-

tallographic properties. All Nb2O5 polymorphs presented herein may be viewed as being derived from

1 precisely described as two reservoirs: a weakly bound distribution of sites with resonances centered

3 around -5 ppm and a more rigid lithium sublattice with resonances at 1 ppm. Note that spin exchange

5 in solids can occur through chemical exchange or dipole-mediated spin interactions; variable tempera-

10 change dominates spin diffusion in these spectra. Qualitatively similar results were obtained from

12 EXSY measurements at higher lithium contents (Supplementary Figure S9); the increased off-diagonal

15 intensity observed at 100 p,s in Li128Nb2O5 and Li186Nb2O5 indicates that exchange occurs on a shorter

17 timescale at higher lithium concentrations.

22 Structure-Property Relationships. In light of the observations on micrometer-sized TT-, T-, B-, and

24 H-Nb2O5, their respective electrochemical performance appears to be strongly dictated by inherent crys-

29 layered structures; however the nature of these layers differs in all cases and all are distinct from the O3

31 and P2 type59 layered structures found in LiCoO2 and Na07CoO2, respectively. Herein, we describe T-

36 and O-Nb polygons; the interlayer distance is 3.93 A. Unlike classical layered structures with structural

38 lithium slabs, the room-and-pillar framework is self-supporting and allows lithium to act as a true guest

40 26 33

41 atom over the whole range of 0 < x < 2 in LixNb2O5. In situ laboratory XRD from several groups ,

43 has suggested that intercalation into T-Nb2O5 is accompanied by an expansion of the layers in a solid-

45 solution reaction. B-Nb2O5 is structurally similar to TiO2(B) with bilayers of edge-sharing octahedra;

48 the interlayer distance is 3.51 A. The layers of H-Nb2O5 comprise ReO3-type blocks of dimension (3 x

50 4)1 and (3 x 5)OT NbO6 octahedra, exhibiting crystallographic shear in the ac plane; the layers are offset

52 from one another by b with an interlayer distance of 3.83 A. Given the complexity of these layered

55 oxides and the distortion of the individual polyhedra due to the second-order Jahn-Teller effect of d

Nb2O5 as a "room-and-pillar" framework with alternating layers (along c) of bridging oxygen "pillars"

structures during crystal structure solution and refinement; however, it may be extended to help predict

Nb5+, the sites and diffusion pathways for lithiation may not be immediately discernable. In order to

3 overcome this challenge, we applied the bond valence sum mapping method to Nb2O5 and related struc-

5 tures (Figure 12). The bond valence sum (BVS) method is a commonly employed tool to validate trial

10 theoretical sites and diffusion pathways through bond valence sum mapping. This method has proven to

12 be an accurate and useful tool for the qualitative evaluation of atomic sites and diffusion pathways in a

15 range of materials.60 It should be especially relevant where solid-solution or only minor rearrangement

17 occur during lithiation; the use of this approach to study two-phase reactions should be treated with care

19 as diffusion through the structure and lithium positions within the structure are not representative of a

10 11 12

20 21 22

Journal of the American Chemical Society Page 26 of 35

reaction across a boundary to a distinct phase. In this study, the program 3DBVSMapper61 was used to calculate the BVS of a theoretical lithium at each position on a fine grid over a unit cell. Starting with T-Nb2O5 (Figure 12a-b), one can visualize the nearly degenerate 2D diffusion pathways throughout the structure as represented by the continuous 2 eV isosurface. The isosurface energy is a relative value and does not correspond trivially to a measurable physical quantity; it is most instructive to use an energy corresponding to the onset of sites/connected pathways to see those relatively low in energy60'61. The high degree of favorable sites on interconnected pathways without obvious minima to trap lithium offers

Figure 12 - Bond valence sum maps of (a-b) T-Nb2O5 (c-d) H-Nb2O5 (e) ReO3 (f) B-Nb2O5 (g) TiO2(B). Bond valence energy level calculations were performed for a theoretical grid of Li+ and 2.0 eV isosurfaces are shown in purple.

to find an off-center local minimum. In a true ReO3-type structure such as NbO2F, topotactic lithiation

1 some qualitative insight into the fast kinetics of the system. The BVS map of H-Nb2O5 (Figure 12c-d),

3 reveals that within the ReO3 (Figure 12e)-type blocks, there are large spaces where lithium may sit;

5 however, these large vacancies with 12 nearest neighbor oxygens are too big for lithium, which is likely

10 would theoretically lead to an ABX3 perovskite but the size-mismatch instability instead leads to tilting

12 of the BX3 octahedra to reduce the size of the A site, and a distortion toward the LiNbO3 structure62.

15 The crystallographic shear in H-Nb2O5 and related compounds such as TiNb2O763'64 precludes these tilt-

17 ing and distortion modes that are common in perovskites, which will have implications for Li+ mobility

19 within the perovskite blocks. The BVS map of H-Nb2O5 indicates, via breaks in the isosurface, that 21

22 there is a relatively high energy barrier to diffusion between the 3 x 4 and 3 x 5 blocks. The abundance

24 of lithium sites, which have been observed via neutron diffraction for H-Lii2/ 7Nb2O565, but local minima

29 mance of H-Nb2O5. On the other hand, the 2 eV isosurface for B-Nb2O5 (Figure 12f) shows only dis-

31 crete sites between the layers. Given the structural similarity to TiO2(B) (Figure 12g), one might expect

36 ing (3.51 A vs. 3.66 A) and higher oxidation state (Nb5+ vs. Ti4+) apparently preclude facile lithiation.

45 Discussion

48 In stark contrast to the well-known positive effects of nanoscaling in Li4Ti5O1266'67, LiCoO268,

and poor connectivity of diffusion pathways helps to explain the high capacity but poor rate perfor-

significant and perhaps high-rate lithiation into B-Nb2O5; however, the slightly smaller interlayer spac-

50 LiFePO46 , MnO2,1>,z, and polymorphs of TÍO2 (e.g. anatase, rutile, TiO2(B), brookite)z, both the total

18 20 23

55 Nb2O5 are comparable to the best nanostructured Nb2O5 particle geometries ' ' . It is important to

charge storage capacity and the rate performance of thick electrodes of bulk TT- and particularly T-

1 stress that care must be taken when comparing high-rate electrochemical data in the literature; in partic-

3 ular, cyclic voltammetric (CV) and chronopotentiometric (i.e. galvanostatic) methods are not necessarily

5 directly comparable. For Nb2O5 with a theoretical capacity of 201.7 mAhg-1, a rate of 1C corresponds

8 to 201.7 mA g-1 and a time of one hour for one electron/one lithium discharge. The corresponding ex-

10 periment via CV is conducted at a sweep rate determined by the potential window divided by the de-

12 sired discharge time (e.g. 3600 s). Thus, a change in the potential window will translate to a change in

15 the sweep rate without affecting the C rate. More significantly, the discharge time from CV will always

17 be longer than the time for a high-rate galvanostatic experiment that does not reach theoretical capacity

19 (and fast discharge rarely ever reaches theoretical capacity). A sweep of 1200 mV at 20 mVs-1 requires 21

22 60 s and is therefore called 60C while T-Nb2O5 reaches only circa half of theoretical capacity at this rate

24 and thus requires only a 30 s galvanostatic discharge. The applied galvanostatic current at 60C for T-

27 Nb2O5 is by definition 12.1 Ag while the average current in a CV experiment would be scaled by the

29 percent of theoretical capacity attained (i.e. 6.05 A g in this example). Since ohmic (IR) losses be-

31 come significant at high currents, galvanostatic cycling further suffers as the enforcement of a single

36 noted 30C and 60C in Figure 7 were acquired in 71 s and 25 s, respectively, rather than 120 s and 60 s if

38 recorded via a cyclic voltammogram. This difference is significant for high-rate materials and high-

41 power applications.

45 The activation barriers in T-LixNb2O5 at 98(1)-58(2) meV are significantly lower than those found in

47 56 73

48 other lithium-ion battery materials: Li07TiS2 at 370-410 meV , Li6Ti5O12 at 430 meV , Li4Ti5O12 at

49 73 49 74

50 760 meV73, Li0 74HO2 at 370 meV49, and LiFePO4 at 270-500 meV74. The dynamics of T-LixNb2O5 are

52 more similar to several recently reported Li-rich solid electrolyte materials75,76; Li10GeP2S12, Li7GePS8,

55 and Li11Si2PS12 show activation barriers for lithium motion of 200-250 meV. NMR relaxometry is a

initial current means the full IR drop is realized at the start of discharge. For reference, the curves de-

microscopic probe of lithium diffusion and it is known that, in certain cases, the thermally activated

3 process was not one of bulk diffusion but, for example, local librational motion of protons or restricted

5 lithium hopping in a system where diffusion is controlled by phase or grain boundaries51. From the

10 apparent lack of phase transition in T-Nb2O5 suggest that, analogous to the layered compound

12 Li0.7TiS256, the microscopic processes translate to the macroscale.

17 The high rate capability observed for micrometer-sized particles of bronze-phase Nb2O5 is somewhat

19 remarkable. To place context for this result, it is useful to compare TiO2 more explicitly as titanium(IV) 21

structural arguments presented above, the large particles, degenerate atomic diffusion pathways, and

22 oxide is chemically similar to niobium(V) oxide in that both are insulators with a d0 cation with similar

24 ionic radii and several known polymorphs. The maximum capacity of all aforementioned TiO2 poly-

29 ed that full lithiation to LiTiO2 occurred only for particles <7-15 nm while bulk anatase reached a max-

31 imum lithium content of Li0 5-0 6TiO2 and the term "bulk" was applied to anything over ca. 120 nm78. In

36 generally accepted that both the maximum capacity and rate performance are size-dependent properties

38 in anatase TiO2. A first-principles thermodynamic and kinetic study79 by the Van der Ven group

4° showed that, as TiO2 is lithiated and Ti4+ is reduced to Ti3+, the distortions in octahedral lithium sites are

43 removed, which causes the minimal energy migration paths to increase from 0.50 eV to 0.78 eV to 1.37

45 eV for dilute LixTiO2, Li0 5TiO2 and LiTiO2, respectively. While polyhedra in niobium(V) and titani-

48 um(IV) oxides experience similar second-order Jahn-Teller distortions80, the relaxation of these distor-

50 tions from niobium(V) to niobium(IV) in the bronze-phase Nb2O5 structure does not cause the same

55 crease lattice flexibility is not as applicable to Nb2O5. The results presented here suggest that optimal

morphs is higher for nanoparticles than the bulk phase, e.g. in anatase TiO2, a careful study demonstrat-

terms of rate, it was shown that full capacity at 5C to 10C rates required sub-10 nm particles. Thus, it is

clamping of diffusion paths. Thus, the impetus to nanosize TiO2 to suppress phase transitions and in-

1 performance of Nb2O5 is observed even on the scale of micrometers rather than only at a few nanome-

3 ters. Furthermore, despite the insulating nature of Nb2O5, the performance herein is comparable to

29 30 28

5 nanocrystalline Nb2O5/carbon nanocomposites with graphene ' , carbon nanotubes , carbide-derived

7 31 32

8 carbon , and carbon core-shell nanocrystals , which, along with the immediate color change, suggests

10 that Nb2O5 must undergo significant conductivity changes even at low lithium content as electrons are

12 introduced into the transition metal d-orbitals. This rate behavior for solid-state-derived Nb2O5 is ob-

15 served in the absence of thin electrodes, lithium perchlorate salts, and carbon counter electrodes that are

17 frequently reported to improve Nb2O5 rate capability26'27.


25 Via solid-state synthesis and electrochemical characterization, we have demonstrated that the complex

27 oxide structure of T-Nb2O5 facilitates high-rate lithium intercalation into large particles on par with the

30 best nanostructured electrodes. Lithium dynamics were investigated with variable temperature NMR

32 relaxation and exchange measurements on electrochemically lithiated T-Nb2O5, which revealed an acti-

34 vation barrier for microscopic lithium diffusion on the order of a few kBT (ca. 60-100 meV). Intercala-

37 tion of lithium into the room-and-pillar layered structure of T-Nb2O5 results in high ionic mobility and

39 minimal strain, which negates the usual requirement of short diffusion pathways and phase transition-

44 tent suggest delocalized conduction electrons and clarify the electronic aspect of the observed high-rate

46 performance in this originally wide-bandgap oxide. Atomic origins of the vast differences in electro-

51 T-Nb2O5, low-capacity B-Nb2O5, and high-capacity but low-rate H-Nb2O5 can be understood in terms of

53 their structure and relationships to other compounds aided by lithium bond valence sum maps. The vol-

56 umetric power and energy density, safety, stability, and ease of synthesis make Nb2O5 an interesting

suppressing nanoparticles for rapid discharge. Temperature-dependent NMR shifts at high lithium con-

chemical behavior among Nb2O5 polymorphs were elucidated. The bulk properties of high-rate TT- and

unique advantages with minimal synthetic or post-synthetic processing. Forthcoming work will exam-

spectroscopy measurements can be found in the supporting information. This material is available free

1 candidate for energy storage applications demanding combined high rate and high capacity in a small

3 cell. We believe the structural principles and techniques presented here will aid the exploration of mate-

5 rials space for future high-rate electrode materials, especially for complex structures that may offer

10 ine the structural transitions, lithium dynamics, and lithium sites in H-Nb2O5, which are beyond the

12 scope of this study.


20 Rietveld refinements, additional SEM micrographs, electrochemical cycling data, 7Li MAS NMR spec-

22 tral deconvolutions, details of the data fitting and temperature calibration for VT NMR, chronopotenti-

24 ometric comparison of film and pellet electrodes, and additional data from relaxometry and exchange

29 of charge via the Internet at Data supporting this work are available from



38 Corresponding Author

40 *E-mail:

43 Notes

45 The authors declare no competing financial interest.


51 The authors thank Dr. Gunwoo Kim, University of Cambridge for discussions on variable temperature

54 NMR; Professor Bruce Dunn, University of California, Los Angeles for discussions on nanostructured

56 Nb2O5; Professors Siegbert Schmid and Christopher Ling, University of Sydney for structural discus-

10 11 12

20 21 22

sions on T-Nb2O5; Dr. Maxim Avdeev, Bragg Institute for his bond valence sum mapping program; Zlatko Saracevic, University of Cambridge for performing the BET measurements; and Dr. Pritesh Hi-ralal, University of Cambridge for assistance acquiring SEM images.

K.J.G. gratefully acknowledges funding from The Winston Churchill Foundation of the United States and the Herchel Smith Scholarship. A.C.F. and J.M.G thank the EPSRC, via the Supergen consortium, for funding. A.C.F. is also thankful to the Sims Scholarship for support.


10 11 12

20 21 22

Conway, B. E. Electrochemical Supercapacitors: Scientific Fundamentals and Technological Applications; Springer: New York, NY, 1999.

Dylla, A. G.; Henkelman, G.; Stevenson, K. J. Acc. Chem. Res. 2013, 46, 1104-1112. Wagemaker, M.; Mulder, F. M. Acc. Chem. Res. 2013, 46, 1206-1215. Buzea, C.; Pacheco, I. I.; Robbie, K. Biointerphases 2007, 2, 17-71. Palacin, M. R.; Simon, P.; Tarascon, J. M. Acta Chim. Slov. 2016, 63, 1-7. Reddy, M. A.; Varadaraju, U. V. J. Phys. Chem. C 2011, 115, 25121-25124. Han, J.-T.; Liu, D.-Q.; Song, S.-H.; Kim, Y.; Goodenough, J. B. Chem. Mater. 2009, 21, 47534755.

Han, J.-T.; Goodenough, J. B. Chem. Mater. 2011, 23, 3404-3407.

Wen, C. J.; Boukamp, B. A.; Huggins, R. A.; Weppner, W. J. Electrochem. Soc. 1979, 126, 2258-2266.

Schäfer, H.; Gruehn, R.; Schulte, F. Angew. Chem. Int. Ed. Engl. 1966, 5, 40-52.

Frevel, L. K.; Rinn, H. W. Anal. Chem. 1955, 27, 1329-1330.

Weissman, J. G.; Ko, E. I.; Wynblatt, P.; Howe, J. M. Chem. Mater. 1989, 1, 187-193.

Ko, E. I.; Weissman, J. G. Catal. Today 1990, 8, 27-36.

Kato, K.; Tamura, S. Acta Crystallogr. Sect. B 1975, 31, 673-677.

Laves, F.; Petter, W.; Wulf, H. Naturwissenschaften 1964, 51, 633-634.

Kato, K. Acta Crystallogr. B 1976, 32, 764-767.

Anderson, J. S.; Tilley, R. J. D. In Surface and Defect Properties of Solids; The Chemical Society: London, 1974; Vol. 3, pp 1-56.

Wei, M.; Wei, K.; Ichihara, M.; Zhou, H. Electrochem. Commun. 2008, 10, 980-983.

Viet, A. L.; Reddy, M. V.; Jose, R.; Chowdari, B. V. R.; Ramakrishna, S. J. Phys. Chem. C 2010,

114, 664-671.

Liu, M.; Yan, C.; Zhang, Y. Sci. Rep. 2015, 5, 8326.

Luo, H.; Wei, M.; Wei, K. Mater. Chem. Phys. 2010, 120, 6-9.

Brezesinski, K.; Wang, J.; Haetge, J.; Reitz, C.; Steinmueller, S. O.; Tolbert, S. H.; Dunn, B. J. Am. Chem. Soc. 2010, 132, 6982-6990.

Kong, L.; Zhang, C.; Wang, J.; Long, D.; Qiao, W.; Ling, L. Mater. Chem. Phys. 2015, 149-150, 495-504.

10 11 12

20 21 22

(25 (26

(36 (37 (38 (39 (40 (41

(42 (43

(45 (46 (47

(50 (51

(52) 33

Lim, E.; Kim, H.; Jo, C.; Chun, J.; Ku, K.; Kim, S.; Lee, H. I.; Nam, I.-S.; Yoon, S.; Kang, K.; Lee, J. ACSNano 2014, 5, 8968-8978.

Kim, J. W.; Augustyn, V.; Dunn, B. Adv. Energy Mater. 2012, 2, 141-148.

Come, J.; Augustyn, V.; Kim, J. W.; Rozier, P.; Taberna, P.-L.; Gogotsi, P.; Long, J. W.; Dunn,

B.; Simon, P. J. Electrochem. Soc. 2014, 161, A718-A725.

Augustyn, V.; Come, J.; Lowe, M. A.; Kim, J. W.; Taberna, P.-L.; Tolbert, S. H.; Abruna, H. D.; Simon, P.; Dunn, B. Nat. Mater. 2013, 12, 518-522.

Wang, X.; Li, G.; Chen, Z.; Augustyn, V.; Ma, X.; Wang, G.; Dunn, B.; Lu, Y. Adv. Energy Mater. 2011, 1, 1089-1093.

Long, D.; Kong, L.; Zhang, C.; Zhang, S.; Wang, J.; Cai, R.; Lu, C.; Qiao, W. M.; Ling, L. J. Mater. Chem. A 2014, 2, 17962-17970.

Arunkumar, P.; Ashish, A. G.; Babu, B.; Sarang, S.; Suresh, A.; Sharma, C. H.; Thalakulam, M.; Shaijumon, M. M. RSC Adv. 2015, 5, 59997-60004.

Zhang, C. (John); Maloney, R.; Lukatskaya, M. R.; Beidaghi, M.; Dyatkin, B.; Perre, E.; Long,

D.; Qiao, W.; Dunn, B.; Gogotsi, Y. J. Power Sources 2015, 274, 121-129.

Lim, E.; Jo, C.; Kim, H.; Kim, M.-H.; Mun, Y.; Chun, J.; Ye, Y.; Hwang, J.; Ha, K.-S.; Roh, K.

C.; Kang, K.; Yoon, S.; Lee, J. ACS Nano 2015, 9, 7497-7505.

Kumagai, N.; Koishikawa, Y.; Komaba, S.; Koshiba, N. J. Electrochem. Soc. 1999, 146, 32033210.

Kodama, R.; Terada, Y.; Nakai, I.; Komaba, S.; Kumagai, N. J. Electrochem. Soc. 2006, 153, 583-588.

A. C, Larson; Von Dreele, R. B. General Structure Analysis System (GSAS); LAUR 86-748; Los Alamos National Laboratory, 2000.

Toby, B. H.; Von Dreele, R. B. J. Appl. Crystallogr. 2013, 46, 544-549. Toby, B. H. J. Appl. Crystallogr. 2005, 38, 1040-1041. Momma, K.; Izumi, F. J. Appl. Crystallogr. 2011, 44, 1272-1276. Bielecki, A.; Burum, D. P. J. Magn. Reson. A 1995, 116, 215-220. Maruyama, T.; Arai, S. Appl. Phys. Lett. 1993, 63, 869-870.

Titanium Oxide (TiO2) Nanoparticles / Nanopowder (TiO2, Anatase, 99.5% 5nm) (accessed Dec 8, 2015). Meyer, G.; Hoppe, R. J. Common Met. 1976, 46, 55-65.

Kumada, N.; Muramatu, S.; Muto, F.; Kinomura, N.; Kikkawa, S.; Koizumi, M. J. Solid State Chem. 1988, 73, 33-39.

Reddy, M. V.; Jose, R.; Le Viet, A.; Ozoemena, K. I.; Chowdari, B. V. R.; Ramakrishna, S. Elec-trochimica Acta 2014, 128, 198-202.

Saito, K.; Kudo, A. Bull. Chem. Soc. Jpn. 2009, 82, 1030-1034.

Rauda, I. E.; Augustyn, V.; Dunn, B.; Tolbert, S. H. Acc. Chem. Res. 2013, 46, 1113-1124. Middlemiss, D. S.; Ilott, A. J.; Clément, R. J.; Strobridge, F. C.; Grey, C. P. Chem. Mater. 2013, 25, 1723-1734.

Strobridge, F. C.; Clément, R. J.; Leskes, M.; Middlemiss, D. S.; Borkiewicz, O. J.; Wiaderek, K. M.; Chapman, K. W.; Chupas, P. J.; Grey, C. P. Chem. Mater. 2014, 26, 6193-6205. Bottke, P.; Ren, Y.; Hanzu, I.; Bruce, P. G.; Wilkening, M. Phys. Chem. Chem. Phys. 2014, 16, 1894-1901.

Schurko, R. W.; Wi, S.; Frydman, L. J. Phys. Chem. C 2002, 106, 51-62.

Wagemaker, M.; van de Krol, R.; Kentgens, A. P. M.; van Well, A. A.; Mulder, F. M. J. Am.

Chem. Soc. 2001, 123, 11454-11461.

Bloembergen, N.; Purcell, E. M.; Pound, R. V. Phys. Rev. 1948, 73 (7), 679-712.

10 11 12

20 21 22

(55 (56 (57 (58 (59

(60 (61 (62 (63

(64 (65 (66

(67 (68

(70 (71 (72 (73

(74 (75

(76 (77 (78 (79 (80

Abragam, A. Principles of Nuclear Magnetism; International Series of Monographs on Physics; Oxford Science Publications: Oxford, 1961.

Steigel, A.; Spiess, H. W. Dynamic NMR Spectroscopy; NMR Basic Principles and Progress; Springer: Berlin, 1978.

Diffusion in Condensed Matter; Heitjans, P., Kärger, J., Eds.; Springer: Berlin, 2005.

Wilkening, M.; Küchler, W.; Heitjans, P. Phys. Rev. Lett. 2006, 97, 65901.

Xu, Z.; Stebbins, J. F. Science 1995, 270, 1332-1334.

Davis, L. J. M.; Heinmaa, I.; Goward, G. R. Chem. Mater. 2010, 22, 769-775.

Yabuuchi, N.; Kawamoto, Y.; Hara, R.; Ishigaki, T.; Hoshikawa, A.; Yonemura, M.; Kamiyama,

T.; Komaba, S. Inorg. Chem. 2013, 52, 9131-9142.

Avdeev, M.; Sale, M.; Adams, S.; Rao, R. P. Solid State Ion. 2012, 225, 43-46.

Sale, M.; Avdeev, M. J. Appl. Crystallogr. 2012, 45, 1054-1056.

Bohnke, C.; Bohnke, O.; Fourquet, J. L. Mol. Cryst. Liq. Cryst. 1998, 311, 23-29.

Lu, X.; Jian, Z.; Fang, Z.; Gu, L.; Hu, Y.-S.; Chen, W.; Wang, Z.; Chen, L. Energy Environ. Sci.

2011, 4, 2638-2644.

Dreele, R. B. V.; Cheetham, A. K. Proc. R. Soc. Lond. Math. Phys. Sci. 1974, 338, 311-326. Catti, M.; Ghaani, M. R. Phys. Chem. Chem. Phys. 2013, 16, 1385-1392. Borghols, W. J. H.; Wagemaker, M.; Lafont, U.; Kelder, E. M.; Mulder, F. M. J. Am. Chem. Soc. 2009, 131, 17786-17792.

Liu, W.; Zhang, J.; Wang, Q.; Xie, X.; Lou, Y.; Xia, B. Ionics 2014, 20, 1553-1560.

Okubo, M.; Hosono, E.; Kim, J.; Enomoto, M.; Kojima, N.; Kudo, T.; Zhou, H.; Honma, I. J. Am.

Chem. Soc. 2007, 129, 7444-7452.

Delacourt, C.; Poizot, P.; Levasseur, S.; Masquelier, C. Electrochem. Solid-State Lett. 2006, 9, 352-355.

Huang, H.; Yin, S.-C.; Nazar, L. F. Electrochem. Solid-State Lett. 2001, 4, 170-172.

Devaraj, S.; Munichandraiah, N. J. Phys. Chem. C 2008, 112, 4406-4417.

Ren, Y.; Armstrong, A. R.; Jiao, F.; Bruce, P. G. J. Am. Chem. Soc. 2010, 132, 996-1004.

Wilkening, M.; Iwaniak, W.; Heine, J.; Epp, V.; Kleinert, A.; Behrens, M.; Nuspl, G.; Bensch,

W.; Heitjans, P. Phys. Chem. Chem. Phys. 2007, 9, 6199-6202.

Liu, Z.; Huang, X. Solid State Ion. 2010, 181, 907-913.

Kuhn, A.; Gerbig, O.; Zhu, C.; Falkenberg, F.; Maier, J.; Lotsch, B. V. Phys. Chem. Chem. Phys. 2014, 16, 14669-14674.

Kuhn, A.; Duppel, V.; Lotsch, B. V. Energy Environ. Sci. 2013, 6, 3548.

Kim, G.; Blanc, F.; Hu, Y.-Y.; Grey, C. P. J. Phys. Chem. C 2013, 117, 6504-6515.

Wagemaker, M.; Borghols, W. J. H.; Mulder, F. M. J. Am. Chem. Soc. 2007, 129, 4323-4327.

Belak, A. A.; Wang, Y.; Van der Ven, A. Chem. Mater. 2012, 24, 2894-2898.

Kunz, M.; Brown, I. D. J. Solid State Chem. 1995, 115, 395-406.

10 11 12

20 21 22

For Table of Contents Only