Role of precursor chemistry in the direct fluorination to form titanium based conversion anodes for lithium ion batteries

Abstract

A new synthetic route for the formation of titanium oxydifluoride (TiOF₂) through the process of direct fluorination via a fluidized bed reactor system and the associated electrochemical properties of the powders formed from this approach are reported. The flexibility of this synthetic route was demonstrated using precursor powders of titanium dioxide (TiO₂) nanoparticles, as well as a reduced TiO_xN_y. An advantage of this synthetic method is the ability to directly control the extent of fluorination as a function of reaction temperature and time. The TiOF₂ synthesized from TiO₂ and TiO_xN_y showed reversible capacities of 300 mA h g⁻¹ and 440 mA h g⁻¹, respectively, over 100 cycles. The higher reversible capacity of the TiOF₂ powders derived from TiO_xN_y likely relate to the partial reduction of the Ti in the fluorinated electrode material, highlighting a route to optimize the properties of conversion electrode materials.

---

1. Introduction

There are a number of chemistries which are attractive for application to lithium ion batteries beyond traditional intercalation compounds such as graphite and LiCoO₂. One such class of materials is the so-called conversion electrodes, such as the metal (oxy)fluorides FeF₃, FeOF, CuF₂, or TiOF₂. These materials have capacities in the range of 1000 mA h g⁻¹, almost three times that of the standard graphite electrode,¹ due to the ability to access multiple oxidation states of the transition metal and follow the generalized reaction:

M^z+X_y + zLi ⇌ M⁰ + yLi_z/yX (1)

where M is a cation with multiple oxidation states, and X is an anion such as oxygen or fluorine. Fluoride materials are an important class of conversion materials because of their relatively high potentials, which result in higher energy densities than the corresponding metal oxides or sulfides.² However, metal fluorides are electrically insulating due to large band gaps, and as electrode materials they typically exhibit a large voltage hysteresis between charge and discharge cycles that results in poor efficiency.^3–5 Various approaches have been used to improve the electrochemical activity of metal fluorides, which include reducing the crystal size, adding a highly conductive carbon to ensure electrical connection to the metal fluoride particles, and the addition of oxygen to improve the cycle stability while slightly lowering capacity.^4,6 Although conversion materials have been studied for decades the underlying mechanisms of these conversion reactions are still not well understood.^7–9

Fluorination reactions to form the electrode materials can be performed by a variety of methods including precipitation from a fluorinated solution, reaction with HF,¹⁰ direct reaction with F₂(g), solid-state reactions with a fluorinating agent such as NH₄F, CuF₂, ZnF₂, XeF₂, AgF₂ or NH₄HF₂,^11,12 electrochemical fluorination¹³ and reaction with poly(vinylidene fluoride) (PVDF).¹⁴ Experimental conditions including reaction time, temperature and fluorine concentration can be used to control the level of fluorine insertion and the homogeneity of the products. This control could be beneficial for cycling stability and lowered impedance which are believed to be caused by a passivating fluoride layer on the electrode surface.^15,16 The elevated temperature, fluidized bed reactor fluorination route is capable of modifying electrochemical properties by modifying composition (e.g., TiO_2−xF_x), and through a change in structure (e.g., from anatase TiO₂ to TiOF₂). Other methods to form TiOF₂ have included solvothermal, hydrothermal synthesis and direct fluorination in a standard tube furnace.^17–21 For the fluidized bed system, the degree of fluorination can be controlled by varying the time and temperature at which the fluorination reaction is performed.

The goal of this work is to explore the role precursor chemistry in the formation of fluoride electrodes. Specifically, we sought to explore whether a pre-reduced, more electrically conducting metal oxide precursor would influence the electrical transport and electrochemical behavior of the fluorinated product. Furthermore, we note that many reduced metal oxides have lower electronic reactivities than the fully oxidized materials, e.g. Ti₂O₃ vs. TiO₂, such that the fluorination process could be modified by pre-reduction of the metal oxide precursor. We present the direct fluorination of nanoscale titanium oxide precursors and the effect on the electrochemical behavior in lithium ion batteries. Through variation in temperature in a fluidized bed reactor system the overall capacity and cycle stability are shown to increase in response to higher levels of fluorination. Furthermore, the electrochemical properties of TiOF₂ were found to differ based upon whether a TiO₂ or TiO_xN_y precursor was used, with the partially reduced TiO_xN_y precursor providing a higher reversible capacity.

2. Experimental

2.1. Fluorination of TiO₂ nanoparticles to form TiOF₂

Nanoparticles of TiOF₂ were synthesized via the direct fluorination of high surface area anatase TiO₂ nanoparticles (Nanostructured and Amorphous Materials Inc., ∼15 nm APS, 99.7%) in a fluidized bed reactor (FBR) system shown schematically in Fig. S1.† Before fluorination, the precursors were dried in vacuo for 24 hours at 250 °C using a hot oil bath. Drying is essential remove any adsorbed water from the nanoparticles, which if present will react with fluorine gas to form hydrofluoric acid. Titania powder samples of 0.5–1.0 g size were loaded into the FBR within a nitrogen filled glovebox and sealed, then removed and weighed before being connected to the fluorination line. The FBR and sample were heated to 225 °C under flowing helium, at which point a flow of 2.0 sccm fluorine (Air Products) mixed with helium carrier gas (23 sccm) was initiated. After 24 h the fluorine flow was shut off, and the system was allowed to cool to room temperature under helium flow. The FBR was weighed again to determine any weight loss or gain that occurred during the reaction before collecting the sample in the glovebox. It is worth noting when using a direct fluorination that for higher temperatures or longer reaction times, there is an increase in the level of fluorination which may eventually lead to a fully fluorinated, possibly volatile product (e.g., TiF₄). These factors must be optimized for any new material and reactor design, as has been done in this study. Furthermore, higher flow rates of fluorine may increase the rate of reaction. In this study the flow rate and fluorine concentration were maintained at a level that was both safe and minimized waste of fluorine gas. The fluorinated sample produced from titanium oxide is labelled O–TiOF₂. Reaction at lower temperatures (175 °C and 200 °C) produced mixed phase (MP) TiO₂/TiOF₂ products (MP175-TiOF₂ and MP200-TiOF₂) according to analysis of X-ray diffraction data (Fig. S2†), as discussed further in the ESI.†

2.2. Fabrication of pre-reduced Ti–O–N precursor

Titanium oxynitride nanoparticles were synthesized via the direct nitridation of TiO₂ nanoparticles. The TiO₂ samples were placed in alumina crucibles and reacted under a 200 sccm ammonia flow (Air Gas, Electronic Grade, 99.9995%) at 860 °C over 19 h and cooled under flowing ammonia. Nitride powders were subsequently fluorinated using the procedure detailed in Section 2.1. Elemental analysis was carried out by Galbraith Laboratories, Knoxville, Tennessee, to determine the nitrogen (method GLI procedure E7-1) and titanium (method GLI procedure ME-70) weight percentages. The fluorinated sample produced from a nitrogen containing precursor is labelled N–TiOF₂.

2.3. Characterization of TiOF₂ nanoparticles

The nanoparticles were characterized using a Panalytical Empyrian powder X-ray diffractometer employing Cu K_α radiation. Data was collected within the 2θ range of 10–90°, using a spinning reflection sample stage and Si zero background sample holder. Crystal structures were refined using the Rietveld method in FullProf Suite.^22,23 X-ray diffraction (XRD) data collected on silicon powder standard under identical run conditions were used to obtain instrument profile parameters for microstructural analysis. The specific surface areas were evaluated with a Quantachrome analyzer from the results of N₂ physisorption at 77 K.

2.4. XPS measurements

The surface chemistry of the precursor materials, electrode powders and cycled electrodes were investigated using a PHI 3056 XPS spectrometer equipped with an Al K_α source (1486.6) at a measurement pressure below 10⁻⁸ Torr. After electrochemical cycling the electrodes were immediately disassembled in an Ar filled glovebox and rinsed with anhydrous DMC (less than 0.5 mL – Aldrich). Samples were transferred to the XPS chamber using an air-tight vacuum transfer system. High resolution scans were acquired at 350 W with 23.5 eV pass energy and 0.05 eV energy step. Survey scans were measured at 350 W with 93.9 eV pass energy and 0.3 eV energy step. The binding energies were shifted by setting the adventitious carbon signal to 284.8 eV to account for charging (0.1–0.3 eV). The intensities of the presented spectra are not normalized and are simply shifted vertically for clarity. The spectra were deconvoluted using Gaussian–Lorentzian functions and a Shirley-type background. Powders of TiO₂ (Sigma Aldrich) were measured as references.

2.5. Electrochemical measurements

The electrochemical measurements were carried out at room temperature using 2 electrode CR2032 type coin cells. The working electrodes were prepared from a slurry of active material (TiOF₂ nanoparticles), conductive agent (carbon black), and a binder (polyvinylidene difluoride) at a ratio of 90 : 5 : 5 percent by weight. The mixture was cast onto copper foil using the slurry method and dried to form the electrode. All electrodes contained a mass loading of between 1.3 and 1.6 mg cm⁻² of active material. All coin cells were assembled in an argon filled glove box using pure lithium foil as the counter electrode and polypropylene as the separator. The electrolyte solution consisted of 1 M LiPF₆ in a 1 : 1 : 1 volume ratio of ethylene carbonate EC, dimethyl carbonate DMC, and diethyl carbonate DEC from Novolyte with water impurity <20 ppm. The assembled cells were cycled galvanostatically using a LAND CT2001A battery testing system. The battery cells were tested galvanostatically, with a current constant of 33.5 mA g⁻¹ while cycling between 3.0 and 0.05 V, which corresponds to a C/10 rate relative to the maximum theoretical capacity of 335 mA h g⁻¹ for LiTiO₂. To provide a direct comparison, other electrode materials in this study were also tested at 33.5 mA g⁻¹ between 3.0 and 0.05 V.

3. Results and discussion

3.1. Precursor chemistry

The X-ray diffraction patterns TiO₂ and TiO_xN_y starting materials are shown in Fig. S3† and 1, respectively. The anatase TiO₂ nanopowder consists of a mixture of anatase (ICSD 9852; tetragonal I4₁/amd structure) and a small amount of rutile (ICSD 9161) crystal structures, with a surface area of 163.2 ± 1.1 m² g⁻¹ and estimated crystalline domain size of 11 nm based upon X-ray diffraction data analysis. After ammonolysis at 860 °C, the rock salt structure (ICSD 152807) of TiN was observed. The Ti–N has a refined lattice parameter of 4.2184(2) Å and domain size of 36 nm calculated using the Scherrer formula via microstructural analysis with FullProf Suite. The lattice parameter of TiN is expected to be 4.235 Å, where as the refined value 4.218 Å is, which is intermediate between that of TiN and stoichiometric TiO (a = 4.1766 Å; ICSD 77692). Bulk elemental analysis of the Ti–N precursor showed the concentration of titanium and nitrogen in the material to be 67.0% Ti and 20.76% N. The balance is most likely O (12.24 at%) given the known reactivity of TiN in air, or from residual oxygen during ammonolysis, indicating the Ti–N precursor is actually an oxynitride, i.e. TiO_yN_x. The lattice parameter derived from Rietveld refinement is similar to other TiO_xN_y materials seen in the literature, which has been reported to vary between the lattice parameters of the TiN and TiO based on the ratio of oxygen to nitrogen.^24,25

Fig. 1 Rietveld refined X-ray diffraction data for TiN which indicates that the sample is the oxynitride TiO_xN_y. The red line is the calculated pattern, black is the experimental diffraction pattern, and blue is the difference between the calculated and the observed data. Reflection positions for TiOF₂ are shown in green. The χ² is 2.14 and the R_wp is 13.8. The remaining difference is likely due to slight compositional inhomogeneity.

3.2. Fluorinated electrode materials

After fluorination of the TiO₂ and TiO_xN_y precursors at 225 °C the resulting materials were white powders. XRD data indicated the formation TiOF₂ (ICSD 160661; a = 3.8019 Å), which has a cubic ReO₃-type structure in space group Pm3̄m. This is surprising given the different reaction precursors. However during fluorination of the TiN_xO_y precursor, there is a significant weight loss of ∼80% of the mass of the starting material, which is associated with sublimation of a TiF₄. XPS results, detailed below, indicate that no measurable N is left on or near the surface of the material. Weight loss similarly occurs during fluorination of the oxide precursor, but only ∼20% of the mass of starting material is lost. These results suggest that the N containing precursor is much more reactive, and the nitrogen is likely easily displaced forming N₂ while the Ti bound to N readily reacts to form the volatile TiF₄, which accounts for the weight loss. BET analysis of nitrogen adsorption–desorption isotherms finds surface areas of 46.7 m² g⁻¹ for O–TiOF₂ and 52.9 m² g⁻¹ for N–TiOF₂, indicating similar surface areas for both samples of aggregated nanopowders (Fig. S4†). Rietveld refinements of X-ray diffraction data for TiOF₂ synthesized from anatase TiO₂ nanoparticles and TiO_xN_y nanoparticles are shown in Fig. 2 and 3, respectively, and indicate that at 225 °C the fluorination route produces largely phase pure powders. It is notable that a TiOF₂ phase prepared through a low temperature metathetic reaction with SiO₂ has also been reported to crystallize in a rhombohedral structure with a = 5.333 Å and c = 13.232 Å, and to undergo a reversible transition to a high temperature cubic structure near 60–65 °C.²⁶ The clearest distinction between the diffraction patterns of cubic TiOF₂ and the rhombohedral structure can be made with peaks near to 2θ = 40°; the (113)_H reflection of the R3̄c TiOF₂ structure is reported at 39.4° 2θ, as compared to 38.5° 2θ for the (113)_H of TiF₃, and there is no peak from the cubic structure present in this 2θ range. There is no strong (113)_H present, whereas there is a (111)_C reflection of the cubic Pm3̄m TiOF₂ structure present near 41° 2θ, and Rietveld refinements of TiOF₂ powders prepared at 225 °C with direct fluorination clearly favor a cubic structure. It should be noted that a very weak, broad peak near 38.2° 2θ is observed in some samples that can be assigned either to a partially fluorinated TiO₂ anatase structure, or to a rhombohedral TiF₃ or TiO_1−xF_2+x structure. The Rietveld refinements in the cubic structure give lattice parameters of a = 3.8052(3) Å and a = 3.81594(6) Å, as well as average crystalline domain sizes of 21 nm and 30 nm, for the TiOF₂ synthesized from the TiO₂ and the TiO_xN_y precursors, respectively. This indicates that a slight increase in domain size during conversion to the fluoride from the TiO₂ precursor, but a slight decrease in domain size during the conversion to the fluoride from the TiN precursor, though the results are similar within the error of the method. The approximately 0.01 Å higher lattice parameter of the N–TiOF₂ may indicate a slight difference in composition, possibly due partial reduction of the sample to form the slightly larger Ti³⁺ cation as implied by the TiO_1−xF_2+x formula. Therefore, the two fluorinated materials exhibit the same structure and similar crystallinity, with slight differences in composition on the basis of which precursor was used.

Fig. 2 Rietveld refined X-ray diffraction data for O–TiOF₂. The red line is the calculated pattern, black is the experimental diffraction pattern, and blue is the difference between the calculated and the observed. Reflection positions for TiOF₂ are shown in green. The statistical goodness-of-fit parameters are χ² of 0.807 and R_wp of 25.8.

Fig. 3 Rietveld refined X-ray diffraction data for pure N–TiOF₂. The statistical goodness-of-fit parameters are χ² of 1.97 and the R_wp of 13.8.

3.3. Surface analysis

Ti2p, O1s, F1s, and N1s XPS data collected for the TiO_xN_y precursor material and the TiOF₂ compounds derived from the TiO₂ and TiO_xN_y precursors are presented in Fig. 4. Data collected for the TiO_xN_y precursor (bottom row) are consistent with the presence of both oxygen and nitrogen. The data show the formation of a complex mixture of Ti species on the surface of the TiO_xN_y. From the curve fits of the Ti2p3/2 peak we see evidence for the presence of three Ti species. The low energy peak is centered around 454.7 eV which is due to Ti³⁺–N bonding.^24,27–43 A second, higher energy species is located at 455.9 eV which is attributed to Ti^3+/4+–O/N bonding or satellite peaks for the Ti³⁺–N species.^{24,27,28,34,37–52} Given the clear evidence of O on the surface (Fig. 4, lower middle), it is likely this species due to the Ti–O/N species. Finally, there is a high energy feature at 458.2 eV due to Ti⁴⁺–O bonds. The intensity of this peak, and the apparent corresponding concentration, maybe a bit misleading given the overlap between the Ti2p1/2 peak originating from the Ti³⁺–N which overlaps with the Ti2p3/2 peak of Ti⁴⁺–O.^{24,27,28,30,33,34,37–41,43–49,51–56} The O1s data collected for the TiO_xN_y precursor shows clear evidence for the presence of two O species. The first O species has a binding energy of ∼530 eV and is due to the Ti–O bonds evident in the Ti2p data.^{24,27,30,34,39,41,42,47,48,50–53} The second species is located at a higher binding energy (∼532 eV) and can be attributed to either an oxynitride species, adsorbed water or oxygen on the surface, or metal hydroxy species.^{24,27,34,39,41,47,52} We believe that this peak is attributed to either the oxynitride species or the adsorbed water/oxygen. The N1s data also shows the presence of two N containing moieties. The main N species has a binding energy of 396.4 eV and is due to the Ti–N bonds discussed for the Ti2p XPS data.^{24,27–30,34,39,41–43,47,48,50–54} The second N species has a binding energy of 399 eV and is due to Ti–O/N species.^{27,28,39,41–43,47,51,54} However, it has similarly been debated as to whether the peaks in this region are intrinsic to TiN and other metal nitrides, and can be attributed to shake up peaks.^30,39 Together this data indicates the surface chemistry of the TiO_xN_y is a complex mixture of N, O/N and O species. This is not surprising given the reactivity of metal nitrides in air, however, the data clearly shows the presence of reduced Ti in the top few nanometers of the TiO_xN_y precursor.

Fig. 4 XPS data for O–TiOF₂ and N–TiOF₂ in the Ti2p, O1s and F1s regions, are shown in the upper and middle rows, respectively. Data show similar binding energy (BE) peaks for both TiOF₂ materials, but there are additional peaks in the N–TiOF₂ spectra at lower BE indicating the presence of reduced species. N–TiOF₂ showed no nitrogen peaks in XPS data. For comparison, XPS of TiO_xN_y, in the Ti2p, O1s and N1s regions are shown in the bottom row, indicating a large contribution of nitrogen and reduced titanium species.

XPS data for the fluorinated TiO_xN_y samples in the Ti2p, O1s, and F1s regions are dominated by the peaks expected for TiOF₂. From the curve fits of the Ti2p3/2 peak we see evidence for the presence of two Ti species shown in Fig. 4 (middle left). The high energy peak at 459.8 eV is attributed to Ti⁴⁺2p3/2 peak for TiOF₂.^57,58 This +0.6 eV shift relative to Ti⁴⁺ in the precursor is due to the more electronegative F added to the surface. The small peak at lower binding energy (457.6 eV) indicates the presence of a reduced Ti species, likely Ti³⁺.^47,55,59 This species is similar to the peak reported at 457.6 eV for “TiOFN” due to Ti³⁺.⁶⁰ The O1s data collected for N–TiOF₂ shows clear evidence for the presence of three O species as shown in Fig. 4 (middle center). The first peak is located ∼531 eV and is attributed to a Ti–O species.^58,59,61,62 The next peak is seen at ∼533 eV and is attributed to adsorbed water.^58,59 The third smaller peak is located at 528.5 eV, which is attributed to oxygen bound to a reduced Ti³⁺ species shown to be present Ti2p spectrum of N–TiOF₂. The F1s data also shows the presence of two F containing moieties shown in Fig. 4 (middle right). The main species has a binding energy of 685.2 eV and is due to the Ti–F bond in TiOF₂.^{58,61,63–67} Two other possibilities exist for the presence of an F1s peak near this binding energy, which are adsorbed fluorine (683.9 to 684.3 eV) or fluorine bound in the TiO₂ lattice (687.2 to 688.6 eV); based upon the XRD data, the main contribution is likely to be from Ti–F in TiOF₂.^63–68 We also see a small peak centered at 682.9, which again we believe to be evidence of some binding to a reduced Ti³⁺ species. An N1s spectrum was taken for N–TiOF₂ but shows no peaks, indicating no nitrogen was left near the surface the material after fluorination. However, the presence of Ti³⁺ in N–TiOF₂ is a direct result of the reduced oxynitride precursor.

XPS data for the fluorinated TiO₂ samples in the Ti2p, O1s, and F1s regions are solely peaks expected for TiOF₂. From the curve fits of the Ti2p3/2 peak we see evidence for the presence of one Ti species shown in Fig. 4 (top left). The high energy peak at 459.9 eV is attributed to Ti⁴⁺2p3/2 peak for TiOF₂.^57,58 The O1s data collected for O–TiOF₂ shows clear evidence for the presence of two O species as shown in Fig. 4 (top center). The first peak is located ∼531 eV and is attributed to a Ti–O species.^58,59,61,62 The next peak is seen at ∼533 eV and is attributed to adsorbed water.^58,59 The F1s data also shows the presence of only one F containing species shown in Fig. 4 (top right). This peak shows a binding energy of 685.6 eV and is due to the Ti–F bond in TiOF₂.^{58,61,63–67} The O–TiOF₂ showed no peaks at lower binding energies as seen in the N–TiOF₂ material. Therefore, there is no reduced titanium species present in the fluorinated TiO₂ sample, which is expected since there was no reduced titanium species in the precursor. The main difference in surface chemistry between N–TiOF₂ and O–TiOF₂ is the presence of Ti³⁺ in N–TiOF₂, which may enhance the electrical conductivity of the material through doping into a conduction band.

3.4. Electrochemistry of TiOF₂ formed through direct fluorination

The electrochemical performance of the precursor powders was characterized for comparison with the fluorinated products. The electrochemical performance of TiO₂ and TiO_xN_y between 3.0 and 0.05 V at a rate of C/10 are shown in Fig. 5. The theoretical capacity of TiO₂ as an intercalation compound is 335 mA h g⁻¹, for the intercalation of 1 mole of Li per formula unit. For these commercial materials we measure a capacity of ∼138 mA h g⁻¹ as compared to values of up to 235 mA h g⁻¹ (or a composition near Li_0.7TiO₂) for more optimized nanophase electrodes.⁶⁰ A plateau is observed at approximately 1.75 V vs. Li/Li⁺, as expected for the intercalation reaction of Li into TiO₂ in nanosized anatase (the majority phase) particles.⁶⁹ This intercalation plateau corresponds to a partial reduction of titanium from the +4 to +3 oxidation state, and to a phase transition to orthorhombic Imma for a composition near Li_0.5TiO₂.⁷⁰ For the high surface area TiO₂ nanoparticles employed in this study, the capacity quickly fades with cycling from 520 mA h g⁻¹ in the first cycle, to under 250 mA h g⁻¹ in subsequent cycles. The high initial capacity is due in part to SEI formation, and the subsequent capacity fade is due to a pulverization of the material from the expansion and contraction of the lattice as lithium is intercalated and deintercalated. The TiO_xN_y precursor displayed a reversible specific capacity of less than 50 mA h g⁻¹ during cycling versus lithium (Fig. 5, top right), consistent with a low level of solid electrolyte interphase (SEI) formation and the capacity of carbon present in the electrode formulation; the poor diffusion rate of lithium in the rock salt structure, combined with the fact that most of the titanium has already been reduced from Ti⁴⁺, is likely responsible for the lack of capacity.⁷¹ Due to contributions from non-optimal crystal structure and morphology, the capacities of the precursor TiO₂ and TiO_xN_y powders are relatively low.

Fig. 5 Galvanostatic cycling data of voltage versus specific capacity for O–TiOF₂ and N–TiOF₂, with their corresponding precursor materials shown in the upper row. All cells were cycled at a rate of 33.5 mA g⁻¹, equivalent to C/10 for TiO₂, and cycled between 3.0 and 0.05 V. The fluorinated materials show improved reversible capacity and cycle stability relative to their precursor.

The electrochemical profiles of the N– and O–TiOF₂ materials are shown in Fig. 5 and S5,† for data collected from 3.0–0.05 V at a rate of C/10. There are significant differences between the TiOF₂ samples over all cycles due to the different precursor chemistry: during the first discharge O–TiOF₂ has a capacity near 650 mA h g⁻¹ and N–TiOF₂ has a capacity of over 1100 mA h g⁻¹, such that the initial capacity of the N–TiOF₂ is over 30% higher than that of O–TiOF₂. There is a large irreversible capacity loss between the first and second discharge cycles; O–TiOF₂ gives a second discharge capacity of 418 mA h g⁻¹ and N–TiOF₂ a second discharge capacity of 587 mA h g⁻¹, which suggests an irreversible capacity loss of 47% and 51%, respectively. This compares well with results for TiOF₂ materials produced by other methods (with morphologies such as nanotubes, nanocubes and nanospheres), which give initial discharge capacities between 860 and 1045 mA h g⁻¹, and show irreversible capacity losses between the first and second discharge cycles of 44% to 48%.^17–19,72 The large irreversible capacity loss, along with the first discharge capacity above the theoretical maximum for N–TiOF₂, clearly indicates that some Li is being consumed in the formation of an SEI layer.

Fluorination of the precursors has resulted in a large increase in reversible capacity beyond the initial cycles. After 10 cycles, where the SEI should presumably be relatively constant, the capacities for N– and O–TiOF₂ are 470 and 325 mA h g⁻¹, respectively, whereas the TiO₂ capacity is near 140 mA h g⁻¹. Electrochemical cells for each material were run in triplicate and cycling capacities were within 10% of the given value. In the literature reversible capacities of TiOF₂ materials were seen to range from 377 to 510 mA h g⁻¹ at a current density of 30 mA g⁻¹, which is in the range of materials reported here.^18,19,65 This compares with theoretical capacities of 263 mA g⁻¹ for 1 Li per TiOF₂ and 1052 mA g⁻¹ for complete reduction to titanium metal. The results are consistent with a maximum of close to 2Li per formula unit of reversible capacity, due to the difficulty in reducing Ti below an oxidation state of +2.

While there is an increase in reversible capacity for both fluorinated samples, there are clear differences between the two samples in the voltage profile during the first cycle lithiation. For example there is a second plateau at higher voltage evident at 1.4 V during lithiation for the N–TiOF₂ sample (∼80 mA g⁻¹) that is absent the O–TiOF₂ sample. Secondly, the large plateau at 0.7 V is much larger for N–TiOF₂ (∼410 mA g⁻¹) as compared to O–TiOF₂ (∼300 mA g⁻¹). XPS results have shown that the surface chemistries and compositions of the electrodes are similar, with the exception of partial reduction in the case of N–TiOF₂. Although N–TiOF₂ and O–TiOF₂ appear to have the same structure from XRD analysis, N–TiOF₂ exhibits a higher reversible capacity in addition to having a second plateau at 1.4 V in the first discharge cycle. The differences between the N– and O–TiOF₂ raises questions about the role of structure and composition in the reaction mechanism of TiOF₂.

There has been a significant discussion in the literature regarding the details of the reaction mechanism. Li-ion half-cell batteries containing TiOF₂ electrodes have been reported to show a plateau at 1.3 V by Zeng et al. and Dambournet et al., while others do not show any plateau in this region.^17–19,72 The studies that contain the plateau at 1.3 V attribute it to a phase transition of TiOF₂ from a cubic ReO₃-type structure to a rhombohedral structure upon lithiation, a structure which has been independently reported in a sample prepared by chemical lithiation.⁷³ The chemical lithiation study was carried out using n-butyl lithium, which corresponds to a potential of 1 V vs. Li⁰, meaning that only reactions that occur above that potential can be observed.^72,73 This provides support for the possibility that the plateau at 1.3 V is from an intercalation reaction and the associated structural change of TiOF₂. It has not been clearly explained in the literature why TiOF₂ prepared through some routes show this small plateau at 1.3 V, while others do not. Given that amorphization at lower voltages appears to remove this plateau in subsequent cycles, the presence of this peak may be related to the initial crystallinity of the TiOF₂ sample, or subtle structural and compositional changes due to precursor chemistry.

It is worth further examining the role that structure and composition may play in the reaction mechanism, through consideration of the relationship between the electrochemistry TiOF₂ and that of similarly structured MF₃ compounds (M = Ti, Fe, V, Co, Cr). Metal fluorides such as these crystallize in a VF₃-type rhombohedral structure which is derived from the ReO₃ cubic structure type.⁷⁴ Comparing the electrochemical data of TiF₃ to N–TiOF₂ when cycled versus lithium, a similar voltage profile is observed. TiF₃ shows two plateaus: the first is a smaller plateau near 1.4 V, and the second is a larger plateau at 0.9 V.^75,76 The plateau near 1.4 V is attributed to intercalation to form a Li_xMF₃ (M = Fe, Ti, V) phase, where x can be up to 0.5, and the second plateau to the conversion reaction to form LiF and Ti.^75–77 TiF₃, Li_xTiF₃ and Li_xTiOF₂ have rhombohedral ReO₃-type structures, and consequently have very similar diffraction patterns. Chemical lithiation of the oxyfluoride was carried out by Dambournet et al. with a Ti/Li ratio of 3 : 1 and was refined using R3̄c space group and achieved a fit with refined lattice parameters of a = 5.143 Å and c = 13.928 Å.⁷² This compares to the chemical lithiation of TiOF₂ by Murphy et al., who for a structure of Li_0.43TiOF₂ gave the lattice parameters of a = 5.12 Å and c = 13.615 Å.⁷³ Murphy et al. also gave lattice parameters for a Li_1.5TiOF₂ phase of a = 5.131 Å and c = 13.998 Å which more closely matches the lattice parameters given by Dambournet et al., despite a much higher level of lithiation than was reported in the Dambournet et al. experiments.⁷³ Furthermore, a non-lithiated R3̄c rhombohedral phase of TiF₃ is reported to have lattice parameters of a = 5.144 Å and c = 13.845 Å with reflections in the same positions as is seen in ICSD 52160.⁷⁸ It should be noted that due to the structural similarity between the various possible lithiated and non-lithiated rhombohedral phases, and the variations in lattice parameter reported in the literature, it is difficult to definitively state the composition of lithiated materials on the basis of XRD data. The presence of a plateau near 1.4 V in N–TiOF₂ is a sign of a lithium intercalation mechanism, which by comparison with rhombohedral TiF₃ may reflect the presence of Ti³⁺ in a fluorine-rich TiO_1−xF_2+x. The presence of Ti³⁺ near the particle surface is supported by XPS data, and the possibility of fluorine-rich TiO_1−xF_2+x by the slightly larger lattice parameter of N–TiOF₂. The lack of the 1.4 V plateau in O–TiOF₂ may relate to the lack of Ti³⁺ in this structure. Furthermore, the partial reduction may contribute to the improved reversible capacity of the N–TiOF₂, as this could lead to a slight increase in electronic conductivity resulting in better electrode utilization.

As the crystallinity of the electrodes decreases upon cycling to the point at which XRD data are no longer useful in identifying the phases present, alternative spectroscopic techniques must be used to gain further information on the mechanism of reversible lithiation. Such studies have been reported on materials related to TiOF₂. Li et al. performed ex situ micro-Raman spectroscopy on TiF₃ at various stages of Li insertion during cycling that indicated the formation of Li_xTiF₃ for x ≤ 0.5, beyond which a conversion reaction is initiated to form LiF and Ti⁰; traces of TiF₃ can be seen in the spectra up to an insertion of 2.5 Li.⁷⁵ Upon Li extraction the reappearance of TiF₃ peaks in the spectra are apparent below 2.5 Li, which confirms that TiF₃ can be cycled reversibly. A reversible capacity of 400 mA h g⁻¹ at a current density of 50 mA g⁻¹ has been reported for a related titanium hydroxyfluoride material.⁷² Dambournet et al. have performed ⁷Li NMR on cycled titanium hydroxyfluoride material from discharged and charged cells which gave spectra that were consistent with the presence of a Li_xTiO₂ phase. By combining the NMR data with the pair distribution function (PDF) approach and XPS data, Dambournet et al. proposed a mechanism for the reversible capacity of nanoscale titanium hydroxyfluoride that involved parallel conversion and intercalation reactions written as⁷²

Ti⁰ + LiF/Li_xTiO₂ ↔ TiF₃/Li_yTiO₂. (2)

It was further suggested that a similar mechanism may be active in reversible cycling of TiOF₂. Essentially, after the first discharge cycle TiOF₂, a composite would form of fluoride and oxide phases that undergoes reversible cycling. More detailed investigations into the reaction of TiOF₂ and Li at higher potentials need to be carried out in order to determine whether an Li_xTiOF₂ phase can be formed after the first cycle, or whether the formation of a titanium fluoride and titanium oxide composite is the only available mechanism for reversibly cycling TiOF₂.

The large plateau at 0.7 V has been attributed to three different possibilities. First it could be a reaction with two different lithiated phases in equilibrium, expressed as Li_xTiOF₂ and Li_yTiOF₂ where y > x.^17,19,58 This would imply a phase transition that has not been directly observed due to the low crystallinity at higher levels of lithiation. The second possibility proposed is the direct conversion reaction^18,72

2TiOF₂ + (4 + x)Li⁺ + (4 + x)e⁻ → Ti⁰ + 4LiF + Li_xTiO₂. (3)

The third possibility is that all or part of this plateau around 0.7 V is caused by SEI formation, which for a cell with LiPF₆ in EC:DEC electrolyte typically occurs near this potential.⁷⁹ Given the possible mechanisms for the 0.7 V plateau, the disappearance of this plateau in later cycles suggests either that an irreversible structural change has occurred or that a stable SEI has formed. Given the large irreversible capacity in the first discharge cycle, it is likely that SEI formation contributes significantly to this plateau. Furthermore, ex situ XPS data collected seen in Fig. S6† confirms the presence of reduced titanium species after discharging to 0.05 V, as expected for both intercalation and conversion mechanisms. However, because of the large SEI layer formation that occurs when cycling to this voltage the noise in the spectrum is quite high and does not allow a concrete assignment of fully reduced or partially reduced Ti species. In consideration of the evidence for intercalation at higher voltages, and the apparently irreversible conversion reaction to a mixed phase fluoride/oxide in the related titanium hydroxyfluoride, the lithiation mechanism in TiOF₂ is likely a combination of the reported possible mechanisms. This would involve lithium intercalation at higher voltage, followed by SEI formation and conversion reaction at lower voltage during the initial discharge, forming a mixed composite electrode that exhibits reversible cycling (as in reaction (1)).

A rate capability experiment was also performed on the O–TiOF₂ to determine the stable capacity of the electrode material at current densities of C/10, C/2, 1C, 5C, 10C, 20C and 50C. The capacity quickly fades to below 70% of the initial stable capacity as the current is increased by a factor of ten, as can be seen in Fig. 6. The capacity returns to the initial capacity when the current is reduced, meaning that no irreversible reactions occurred from the use of increased current, and the capacity loss is largely due to polarization and diffusion limitations. Due to partial reduction of N–TiOF₂ this material displays a higher capacity at lower rates, but at a rate of 1C or above the capacity fades more rapidly than O–TiOF₂. This suggests that a second factor is leading to polarization of the electrode at higher rates, which is likely a limitation in the rate of Li⁺ diffusion. The presence of nitrogen defects may play a role in limiting the diffusion of Li⁺ across the surface, but further work will be required to fully discern the factors behind the capacity reduction at high rates in N–TiOF₂. The results suggest that reduction of the TiO₂ prior to fluorination is necessary to obtain Ti³⁺ in the final fluorinated product. To enhance the level of Ti³⁺ was in the final product it may be useful to minimize the loss of nitrogen, perhaps through co-doping with a cation (e.g., Cr³⁺/N³⁻ co-doping) to stabilize the nitrogen in the lattice. Only one prior report of rate capability for TiOF₂ was found in the literature, which was for TiOF₂ nanotubes prepared by Zeng et al. Similar results were obtained in this study and for the nanotubes, with a reversible decrease in capacity as current density was raised.¹⁸ The rate capability was found to be similar, though slightly higher in the nanotubes reported by Zeng et al. for a similar current density, due to the lower diffusion length of the nanotube wall.

Fig. 6 Rate capability test showing specific capacity of TiOF₂ formed from TiO₂ and TiO_xN_y at increasing charge–discharge current densities of (A) 33.5 (C/10), (B) 167.5 (C/2), (C) 335 (C), (D) 1675 (5C), (E) 3350 (10C), (F) 6700 (20C), and (G) 16 750 (50C) mA g⁻¹. All measurements taken within the voltage window of 3.0–0.05 V. Capacity fades as current density is increased, but the capacity recovers when the current density is decreased.

Fluorination of TiO₂ nanoparticles produces an enhancement of electrochemical performance, and the fluidized bed method provides the capability to readily form mixed phase or single phase powders from any given precursor powder. The method allows for the possible control over anion composition by varying the reaction conditions (i.e., time, temperature, etc.). The enhanced performance of N–TiOF₂ suggests that starting from reduced or partially reduced precursors may result in higher performance, and more generally that a useful focus for research is to optimize the anion composition of precursor materials.

4. Conclusions

A new method for the synthesis of TiOF₂ nanoparticles has been demonstrated through the direct fluorination of TiO₂ or TiO_xN_y precursors in an FBR reactor system. Using the new method for the synthesis of TiOF₂ we have been able to synthesize nanoparticles of approximately 20–30 nm domain size that show battery cycling data comparable to that of other nanostructured TiOF₂ based materials in the literature. Fluorination of the TiO_xN_y precursor yields a single phase product of TiOF₂, and shows no residual nitrogen by XPS. The N–TiOF₂ shows a second plateau in the initial discharge at 1.4 V which is not seen in the O–TiOF₂ material. This plateau is related to an intercalation-based lithiation reaction, and there is evidence to suggest a slightly reduced, fluorine rich TiO_1−xF_2+2x phase may be involved in the N–TiOF₂ electrode material. The N–TiOF₂ also shows an increased capacity over the O–TiOF₂ by about 30%, which is likely related to partial reduction in the N–TiOF₂ leading to improved electrical conductivity and electrode utilization. Mixtures of TiO₂/TiOF₂ were synthesized using the FBR reactor system simply by adjusting the reaction temperature of the fluorination. The fluorination of TiO₂ nanoparticles to form mixed phase powders was shown to increase the capacity by over 50% and greatly improve the cycle stability of the starting TiO₂ material. Our results highlight the role that anion modification of precursor materials can play in producing high performance electrodes, rather than the typical focus on post-processing of electrode powders.

Acknowledgements

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Division of Materials Sciences and Engineering.

References

1. L. Taberna, S. Mitra, P. Poizot, P. Simon and J. M. Tarascon, Nat. Mater., 2006, 5, 567–573.
2. G. G. Amatucci and N. Pereira, J. Fluorine Chem., 2007, 128, 243–262.
3. M. R. Palacin, Chem. Soc. Rev., 2009, 38, 2565–2575.
4. R. Malini, U. Uma, T. Sheela, M. Ganesan and N. G. Renganathan, Ionics, 2009, 15, 301–307.
5. I. Plitz, F. Badway, J. Al-Sharab, A. DuPasquier, F. Cosandey and G. G. Amatucci, J. Electrochem. Soc., 2005, 152, A307–A315.
6. N. Pereira, F. Badway, M. Wartelsky, S. Gunn and G. G. Amatucci, J. Electrochem. Soc., 2009, 156, A407–A416.
7. F. Wang, R. Robert, N. A. Chernova, N. Pereira, F. Omenya, F. Badway, X. Hua, M. Ruotolo, R. G. Zhang, L. J. Wu, V. Volkov, D. Su, B. Key, M. S. Whittingharn, C. P. Grey, G. G. Amatucci, Y. M. Zhu and J. Graetz, J. Am. Chem. Soc., 2011, 133, 18828–18836.
8. J. M. Tarascon, Philos. Trans. R. Soc., A, 2010, 368, 3227–3241.
9. J. Cabana, L. Monconduit, D. Larcher and M. R. Palacin, Adv. Mater., 2010, 22, E170–E192.
10. Z. F. Wang, J. Q. Wang, Z. P. Li, P. W. Gong, X. H. Liu, L. B. Zhang, J. F. Ren, H. G. Wang and S. R. Yang, Carbon, 2012, 50, 5403–5410.
11. C. Greaves and M. G. Francesconi, Curr. Opin. Solid State Mater. Sci., 1998, 3, 132–136.
12. C. Greaves, J. L. Kissick, M. G. Francesconi, L. D. Aikens and L. J. Gillie, J. Mater. Chem., 1999, 9, 111–116.
13. M. H. Delville, D. Barbut, A. Wattiaux, J. M. Bassat, M. Menetrier, C. Labrugere, J. C. Grenier and J. Etourneau, Inorg. Chem., 2009, 48, 7962–7969.
14. E. E. McCabe and C. Greaves, J. Fluorine Chem., 2007, 128, 448–458.
15. S. H. Kang and K. Amine, J. Power Sources, 2005, 146, 654–657.
16. S. H. Kang and M. M. Thackeray, J. Electrochem. Soc., 2008, 155, A269–A275.
17. L. Chen, L. Shen, P. Nie, X. Zhang and H. Li, Electrochim. Acta, 2012, 62, 408–415.
18. Y. Zeng, W. Zhang, C. Xu, N. Xiao, Y. Huang, D. Y. W. Yu, H. H. Hng and Q. Yan, Chem.–Eur. J., 2012, 18, 4026–4030.
19. M. V. Reddy, S. Madhavi, G. V. S. Rao and B. V. R. Chowdari, J. Power Sources, 2006, 162, 1312–1321.
20. N. Louvain, Z. Karkar, M. El-Ghozzi, P. Bonnet, K. Guerin and P. Willmann, J. Mater. Chem. A, 2014, 2, 15308–15315.
21. B. Li, D. Wang, Y. Wang, B. Zhu, Z. Gao, Q. Hao, Y. Wang and K. Tang, Electrochim. Acta, 2015, 180, 894–901.
22. J. Rodriguezcarvajal, Phys. B, 1993, 192, 55–69.
23. T. Roisnel and J. Rodriguez-Carvajal, WinPLOTR: A Windows tool for powder diffraction pattern analysis, in Epdic 7: European Powder Diffraction, Pts 1 and 2, ed. R. Delhez and E. J. Mittemeijer, 2001, pp. 118–123.
24. A. Trenczek-Zajac, M. Radecka, K. Zakrzewska, A. Brudnik, E. Kusior, S. Bourgeois, M. C. M. de Lucas and L. Imhoff, J. Power Sources, 2009, 194, 93–103.
25. J. Guillot, A. Jouaiti, L. Imhoff, B. Domenichini, O. Heintz, S. Zerkout, A. Mosser and S. Bourgeois, Surf. Interface Anal., 2002, 33, 577–582.
26. S. Shian and K. H. Sandhage, J. Appl. Crystallogr., 2010, 43, 757–761.
27. K. S. Robinson and P. M. A. Sherwood, Surf. Interface Anal., 1984, 6, 261–266.
28. M. Krawczyk, W. Lisowski, J. W. Sobczak, A. Kosinski and A. Jablonski, J. Alloys Compd., 2013, 546, 280–285.
29. E. Galvanetto, F. P. Galliano, F. Borgioli, U. Bardi and A. Lavacchi, Thin Solid Films, 2001, 384, 223–229.
30. I. Bertoti, M. Mohai, J. L. Sullivan and S. O. Saied, Appl. Surf. Sci., 1995, 84, 357–371.
31. H. Hochst, R. D. Bringans, P. Steiner and T. Wolf, Phys. Rev. B: Condens. Matter Mater. Phys., 1982, 25, 7183–7191.
32. L. Ramqvist, K. Hamrin, G. Johansso, A. Fahlman and C. Nordling, J. Phys. Chem. Solids, 1969, 30, 1835–1847.
33. L. Porte, L. Roux and J. Hanus, Phys. Rev. B: Condens. Matter Mater. Phys., 1983, 28, 3214–3224.
34. N. Kaufherr and D. Lichtman, J. Vac. Sci. Technol., A, 1985, 3, 1969–1972.
35. B. J. Burrow, A. E. Morgan and R. C. Ellwanger, J. Vac. Sci. Technol., A, 1986, 4, 2463–2469.
36. A. Ermolieff, P. Bernard, S. Marthon and P. Wittmer, Surf. Interface Anal., 1988, 11, 563–568.
37. T. Brat, N. Parikh, N. S. Tsai, A. K. Sinha, J. Poole and C. Wickersham, J. Vac. Sci. Technol., B: Microelectron. Process. Phenom., 1987, 5, 1741–1747.
38. I. L. Strydom and S. Hofmann, J. Electron Spectrosc. Relat. Phenom., 1991, 56, 85–103.
39. J. F. Marco, J. R. Gancedo, M. A. Auger, O. Sanchez and J. M. Albella, Surf. Interface Anal., 2005, 37, 1082–1091.
40. P. Padmavathy, R. Ananthakumar, B. Subramanian, C. Ravidhas and M. Jayachandran, J. Appl. Electrochem., 2011, 41, 751–756.
41. B. Avasarala and P. Haldar, Electrochim. Acta, 2010, 55, 9024–9034.
42. D. Jaeger and J. Patscheider, J. Electron Spectrosc. Relat. Phenom., 2012, 185, 523–534.
43. M. Chisaka, A. Ishihara, K. Ota and H. Muramoto, Electrochim. Acta, 2013, 113, 735–740.
44. C. W. Louw, I. L. Strydom, K. Vandenheever and M. J. Vanstaden, Surf. Coat. Technol., 1991, 49, 348–352.
45. S. Lee, O. El-bjeirami, S. S. Perry, S. V. Didziulis, P. Frantz and G. Radhakrishnan, J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct.--Process., Meas., Phenom., 2000, 18, 69–75.
46. S. V. Didziulis, J. R. Lince, T. B. Stewart and E. A. Eklund, Inorg. Chem., 1994, 33, 1979–1991.
47. M. Wolff, J. W. Schultze and H. H. Strehblow, Surf. Interface Anal., 1991, 17, 726–736.
48. C. Ernsberger, J. Nickerson, A. E. Miller and J. Moulder, J. Vac. Sci. Technol., A, 1985, 3, 2415–2418.
49. D. Martinez-Martinez, C. Lopez-Cartes, A. Fernandez and J. C. Sanchez-Lopez, Appl. Surf. Sci., 2013, 275, 121–126.
50. P. Prieto and R. E. Kirby, J. Vac. Sci. Technol., A, 1995, 13, 2819–2826.
51. Y. L. Jeyachandran, S. Venkatachalam, B. Karunagaran, S. K. Narayandass, D. Mangalaraj, C. Y. Bao and C. L. Zhang, Mater. Sci. Eng., C: Biomimetic Supramol. Syst., 2007, 27, 35–41.
52. F. Esaka, K. Furuya, H. Shimada, M. Imamura, N. Matsubayashi, H. Sato, A. Nishijima, A. Kawana, H. Ichimura and T. Kikuchi, J. Vac. Sci. Technol., A, 1997, 15, 2521–2528.
53. M. J. Vasile, A. B. Emerson and F. A. Baiocchi, J. Vac. Sci. Technol., A, 1990, 8, 99–105.
54. L. Wan, J. F. Li, J. Y. Feng, W. Sun and Z. Q. Mao, Appl. Surf. Sci., 2007, 253, 4764–4767.
55. A. F. Carley, J. C. Roberts and M. W. Roberts, Surf. Sci., 1990, 225, L39–L41.
56. F. Peng, L. F. Cai, L. Huang, H. Yu and H. J. Wang, J. Phys. Chem. Solids, 2008, 69, 1657–1664.
57. C. Z. Wen, Q. H. Hu, Y. N. Guo, X. Q. Gong, S. Z. Qiao and H. G. Yang, Chem. Commun., 2011, 47, 6138–6140.
58. S. V. Gnedenkov, D. P. Opra, S. L. Sinebryukhov, V. G. Kuryavyi, A. Y. Ustinov and V. I. Sergienko, J. Alloys Compd., 2015, 621, 364–370.
59. T. Choudhury, S. O. Saied, J. L. Sullivan and A. M. Abbot, J. Phys. D: Appl. Phys., 1989, 22, 1185–1195.
60. X. Zong, Z. Xing, H. Yu, Z. G. Chen, F. Q. Tang, J. Zou, G. Q. Lu and L. Z. Wang, Chem. Commun., 2011, 47, 11742–11744.
61. J. Zhu, F. Lv, S. Xiao, Z. Bian, G. Buntkowsky, C. Nuckolls and H. Li, Nanoscale, 2014, 6, 14648–14651.
62. K. L. Lv, J. G. Yu, L. Z. Cui, S. L. Chen and M. Li, J. Alloys Compd., 2011, 509, 4557–4562.
63. J. H. Kim, F. Nishimura, S. Yonezawa and M. Takashima, J. Fluorine Chem., 2012, 144, 165–170.
64. J. Zhu, D. Q. Zhang, Z. F. Bian, G. S. Li, Y. N. Huo, Y. F. Lu and H. X. Li, Chem. Commun., 2009, 5394–5396,  10.1039/b909692b.
65. D. M. Chen, Z. Y. Jiang, J. Q. Geng, J. H. Zhu and D. Yang, J. Nanopart. Res., 2009, 11, 303–313.
66. D. Li, H. Haneda, N. K. Labhsetwar, S. Hishita and N. Ohashi, Chem. Phys. Lett., 2005, 401, 579–584.
67. C. R. Xue, T. Narushima and T. Yonezawa, J. Inorg. Organomet. Polym. Mater., 2013, 23, 239–242.
68. J. H. Pan, Z. Y. Cai, Y. Yu and X. S. Zhao, J. Mater. Chem., 2011, 21, 11430–11438.
69. M. V. Koudriachova, N. M. Harrison and S. W. de Leeuw, Phys. Rev. Lett., 2001, 86, 1275–1278.
70. T. Ohzuku, T. Kodama and T. Hirai, J. Power Sources, 1985, 14, 153–166.
71. H. C. M. Knoops, L. Baggetto, E. Langereis, M. C. M. van de Sanden, J. H. Klootwijk, F. Roozeboom, R. A. H. Niessen, P. H. L. Notten and W. M. M. Kessels, J. Electrochem. Soc., 2008, 155, G287–G294.
72. D. Dambournet, K. W. Chapman, P. J. Chupas, R. E. Gerald, N. Penin, C. Labrugere, A. Demourgues, A. Tressaud and K. Amine, J. Am. Chem. Soc., 2011, 133, 13240–13243.
73. D. W. Murphy, M. Greenblatt, R. J. Cava and S. M. Zahurak, Solid State Ionics, 1981, 5, 327–330.
74. A. Mogusmilankovic, J. Ravez, J. P. Chaminade and P. Hagenmuller, Mater. Res. Bull., 1985, 20, 9–17.
75. H. Li, G. Richter and J. Maier, Adv. Mater., 2003, 15, 736–739.
76. H. Li, P. Balaya and J. Maier, J. Electrochem. Soc., 2004, 151, A1878–A1885.
77. H. Arai, S. Okada, Y. Sakurai and J. Yamaki, J. Power Sources, 1997, 68, 716–719.
78. H. Sowa and H. Ahsbahs, Acta Crystallogr., Sect. B: Struct. Sci., 1998, 54, 578–584.
79. E. Peled, D. Golodnitsky, C. Menachem and D. Bar-Tow, J. Electrochem. Soc., 1998, 145, 3482–3486.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra17258f

---