Temperature Dependence of Aliovalent-Vanadium Doping in LiFePO4

Article pubs.acs.org/cm

Temperature Dependence of Aliovalent-Vanadium Doping in LiFePO4 Cathodes Katharine L. Harrison,† Craig A. Bridges,‡ Mariappan Parans Paranthaman,‡ Carlo U. Segre,∥ John Katsoudas,∥ Victor A. Maroni,⊥ Juan Carlos Idrobo,§ John B. Goodenough,† and Arumugam Manthiram*,† †

Department of Mechanical Engineering, University of Texas at Austin, Austin, Texas 78712, United States Chemical Sciences Division and §Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States ∥ Center for Synchrotron Research and Instrumentation, Physics Department & CSRRI, Illinois Institute of Technology, Chicago, Illinois 60616, United States ⊥ Argonne National Laboratory, Argonne, Illinois 60439, United States ‡

S Supporting Information *

ABSTRACT: Vanadium-doped olivine LiFePO4 cathode materials have been synthesized by a low-temperature microwave-assisted solvothermal (MW-ST) method at ≤300 °C. The samples have been extensively characterized by neutron/ X-ray powder diffraction, infrared and Raman spectroscopy, elemental analysis, electron microscopy, and electrochemical techniques. The compositions of the assynthesized materials were found to be LiFe1−3x/2Vx□x/2PO4 (0 ≤ x ≤ 0.2) with the presence of a small number of lithium vacancies (□) charge-compensated by V4+, not Fe3+, leading to an average oxidation state of ∼3.2+ for vanadium. The vacancies on the Fe site likely provide an additional conduction pathway for Li+ ions to transfer between neighboring 1D conduction channels along the crystallographic b axis. Heating the pristine 15% V-doped sample in inert or reducing atmospheres led to a loss of vanadium from the olivine lattice with the concomitant formation of a Li3V2(PO4)3 impurity phase; after phase segregation, a partially V-doped olivine phase remained. For comparison, V-doped samples were also synthesized by conventional ball milling and heating, but only ∼10% V could be accommodated in the olivine lattice in agreement with previous studies. The higher degree of doping realized with the MW-ST samples demonstrates the temperature dependence of the aliovalent-vanadium doping in LiFePO4. KEYWORDS: olivine cathodes, LiFePO4, vanadium doping, lithium-ion batteries, cathodes

■

INTRODUCTION

LiFePO4 particles must be synthesized, and the material must be coated with conducting materials.11−20 Improvements in the electrical conductivity through aliovalent-cation doping have also been reported,21 but these results were later contested.22,23 Computational studies have suggested that aliovalent doping is not energetically favorable; moreover, the unbreakable PO43− bonding prevents charge balance by oxygen-vacancy formation.9,24 Despite these arguments, doping LiFePO4 with several different aliovalent cations is often claimed, and many studies have reported higher electronic conductivity and an increase in the FePO4/LiFePO4 solid solution range for doped samples.25−36 Doping has been suggested to occur on the Li or Fe sites for several different aliovalent dopants.25−36 Of particular relevance to this study is previous work pertaining to doping LiFePO4

Conventionally, layered LiCoO2 and/or LiNi1/3Mn1/3Co1/3O2 have been used as cathode materials in lithium-ion batteries, but the transportation and grid energy-storage sectors require new cathode materials that are safer and less expensive. Polyanion cathodes such as Fe2(SO4)3 and Fe2(MoO4)3 were first pursued by Manthiram and Goodenough in the late 1980s.1,2 It was recognized that the covalently bonded polyanion groups lower the energy of redox couples through the inductive effect, which shifts the redox energy of couples like Fe2+/3+ into a useful range. Of particular interest is olivinestructured LiFePO4, which has generated significant interest in the literature and in industry because of its high capacity and good cycling performance.3−6 Additionally, it is thermally stable, environmentally benign, and inexpensive. However, the material suffers from poor electronic and ionic conductivities as well as a power-limiting two-phase reaction during the charge− discharge process.7−10 To achieve high capacities, small © 2013 American Chemical Society

Received: December 7, 2012 Revised: February 7, 2013 Published: February 7, 2013 768

dx.doi.org/10.1021/cm303932m | Chem. Mater. 2013, 25, 768−781

Chemistry of Materials

Article

Conventional Synthesis. For direct solid-state synthesis of LiFe0.775V0.15□0.075PO4, the vanadium precursor was first made by the reaction

with vanadium, which has been shown to improve the electrochemical performance and electronic conductivity.27−36 A recent computational study suggests that doping LiFePO4 with V reduces the activation energy for Li-ion diffusion.37 Although there was initial disagreement regarding whether V substituted on the anion or cation site, there is now general agreement that V substitutes for Fe with the exception of one study that suggests that 1% doping can occur on the Li site.28,32,36 There are also two studies that suggest that V could not be doped into the olivine lattice.38,39 Overall, while there is still some disagreement in the literature regarding V doping, most studies report that ≤10% doping is possible and V exists as V3+ and/or V4+. We have recently reported the synthesis of cathodes with the formula LiFe1−x(VO)xPO4 with 0 ≤ x ≤ 0.25 by a facile, low temperature microwave-assisted solvothermal (MW-ST) synthesis method. 40 After investigating the structure and compositions more thoroughly, we find that the formula is consistent with LiFe1−3x/2Vx□x/2PO4 (where □ represents Fe vacancies) rather than LiFe1−x(VO)xPO4, with evidence for a small doping-dependent lithium deficiency (∼1−3%) and mixed-valent vanadium (V3+:V4+ = 4:1). The MW-ST synthesized LiFe0.775V0.15□0.075PO4 disproportionates on heating to 725 °C to form Li3V2(PO4)3 and an olivine phase with a lower V doping level than in the pristine MW-ST LiFe0.775V0.15□0.075PO4; heating at 625 °C led to only a very small amount of impurity. For a comparison, we attempted to prepare 15% V-doped samples by conventional ball milling and heating, but by this approach we can accommodate only a maximum of ∼10% V in the olivine lattice. Although previous studies have reported a maximum of 10% V doping,27−36 we are able to demonstrate at least 20% V in the olivine lattice due to the low-temperature synthesis method employed, which can stabilize metastable phases.

■

V2O5 + 3H 2C2O4 → 2VOC2 O4 + 3H 2O + 2CO2

(1)

Water was evaporated while stirring the solution on a hot plate, and the powder was dried in a vacuum oven at 100 °C. Inductively coupled plasma (ICP) analysis was used to determine the water content of VOC2O4. Subsequently, the VOC2O4 was ball milled in acetone for 5 days with stoichiometric amounts of iron oxalate (Fisher, 99%), ammonium dihydrogen phosphate (Fisher, 99%), and lithium hydroxide monohydrate (Fisher, 98%) to prepare LiFe0.775V0.15□0.075PO4. The products were dried, ground, and heated in flowing 5% H2−95% Ar or 100% Ar atmospheres for 3 h at 350 °C followed by 6 h at temperatures ranging from 525 to 725 °C, which is similar to known conventional synthesis methods for LiFePO4 and Vdoped LiFePO4. Additionally, LiFePO4 was also made by ball milling and heating with stoichiometric amounts of precursors. A naming scheme is employed to refer to the pristine and subsequently heated samples, defined by “synthesis method−sample−furnace heating atmosphere−furnace heating temperature−furnace heating time.” For example, MW-ST V-doped LiFePO4 that is then heated at 525 °C in 5% H2 and 95% Ar for 6 h will be referred to as MW-LFVP-H2-525-6h. The corresponding conventionally prepared sample will be referred to as CONV-LFVP-H2-525-6h. LiFePO4 samples will be described with “LFP” rather than “LFVP”. The relevant samples will be later summarized in Table 4. The samples referred to by this naming convention were prepared by mixing precursors in the appropriate proportions to form LiFe0.775V0.15□0.075PO4. For comparison, Li3V2(PO4)3 was prepared by a procedure similar to that for the conventionally prepared LiFe0.775V0.15□0.075PO4 samples described above, but with the appropriate stoichiometry for Li3V2(PO4)3. Finally, β-LiVOPO4 was prepared by a conventional sol−gel method, as described elsewhere.40,41 Materials Characterization. Neutron powder diffraction (NPD) data were collected at 295 K on the HB2A beamline at the High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. Isotopically enriched lithium (7Li; scattering length of −2.22 fm) was used to remove uncertainties in the bound coherent neutron scattering length needed for Rietveld refinements of the neutron diffraction data. Such uncertainties can exist due to variations in the isotopic composition of different lithium sources. The samples consisted of approximately 1 g of powder and were contained in 8mm vanadium cans. Data were collected for each sample over two ranges, 10−130° 2θ and 33−153° 2θ, followed by merging of the data to produce a final file over the range of 10−153° 2θ with a step size of 0.05° 2θ for refinement. Data were collected by alternating between each range. With 2 h data collection time per range and collection of multiple data sets per sample, the total time of data collection was 12 h for the 5% and 10% V-doped samples and 10 h for the 15% V-doped sample. Data files containing the standard deviation (ESD) for each point were used to ensure the appropriate calculation of errors during Rietveld refinement. The wavelength was selected using a Ge [115] monochromator, and the wavelength was determined to be 1.5395 Å by refinement against a silicon standard. NPD data were analyzed by Rietveld refinement42 with GSAS/EXPGUI43,44 and Fullprof/ WinPLOTR.45,46 X-ray powder diffraction (XRD) data were collected with a Rigaku Ultima IV instrument with Cu Kα radiation and analyzed by Rietveld refinement with Fullprof/WinPLOTR. X-ray absorption data were taken at the Fe and V K-edges in transmission at the MRCAT (Sector 10, Advanced Photon Source) bending magnet beamline at Argonne National Laboratory. Samples were prepared by grinding between 5 and 20 mg (depending on V and Fe composition) of finely powdered specimens with boron nitride and polyvinylpyrrolidone (PVP). The mixtures were then pressed into 7mm-diameter pellets of less than 1 mm overall thickness. For LiFePO4 samples, a 50/50 mix of (25 mg) boron nitride and (25 mg) PVP was used for aiding easy sample release from the press and to bind a better pellet (with 5 mg of sample). The X-ray energy was selected by a water cooled Si(111) monochromator with a 50% detuned second crystal for

EXPERIMENTAL SECTION

Microwave Synthesis. The vanadium-doped samples were prepared at ≤300 °C as described in more detail previously by a microwave-assisted solvothermal (MW-ST) process. It involves a mixing of iron acetate (STREM, 97%) and lithium hydroxide monohydrate (Fisher, 98%) into a solution of phosphoric acid (Fisher, 85%) in tetraethylene glycol (Alfa Aesar, 99%) and then adding vanadium triisopropoxide oxide (Alfa Aesar, 96%).40 The precursor solutions were prepared according to the formulas LiFe1−x(VO)xPO4 with 0 ≤ x ≤ 0.25, LiFe1−2xVx□xPO4 with 0 ≤ x ≤ 0.15, and LiFe1−3x/2Vx□x/2PO4 with 0 ≤ x ≤ 0.15 in order to take into account different expectations for the vanadium oxidation state and coordination environment. Fe3O4 forms as an impurity for the samples prepared according to LiFe1−x(VO)xPO4 with 0 ≤ x ≤ 0.25, which can be subsequently removed by stirring the product solution repeatedly with a magnet.40 To determine the stability of the V-doped samples at elevated temperatures, the pristine MW-ST sample prepared according to LiFe0.775V0.15□0.075PO4 was ground and heated in flowing 5% H2−95% Ar and 100% Ar environments at various temperatures (525−725 °C) and for various times (6 and 15 h), similar to the conditions used in the conventional high-temperature synthesis of undoped and V-doped LiFePO4.9,11−13,20−22,25−36 Due to the large amounts of material needed for all of the heating conditions, several batches of the pristine microwave-synthesized material were prepared. After checking XRD and ICP for consistency, the batches were mixed together, and the large mixture was used for all post-heating experiments to ensure that the pristine microwave-synthesized material was the same for each of the heating conditions. A similar protocol was used to prepare the 1 g batches of powder for neutron diffraction experiments. 769



Article

synthesize ‘LiFe1−x(VO)xPO4’ resulted in Fe3O4 impurity,40 further suggesting that ‘LiFe1−x(VO)xPO4’ is not the correct formula and that iron vacancies are present in the structure. If V4+ (rather than (VO)2+) substitution occurs in these samples, ‘LiFe1−2xVx□xPO4’ should be the expected formula. Note that a vanadyl bond could still occur from local V displacement toward one of the oxide ions in the VO6 octahedra in a manner similar to that in LiVOPO4.50 To examine whether ‘LiFe1−2xVx□xPO4’ (V4+) cathodes form, we mixed precursors stoichiometrically to obtain ‘LiFe1−2xVx□xPO4’ with x = 0.05, 0.10, and 0.15 by our previously described MW-ST method.40 The ‘LiFe0.80V0.10□0.10PO4’ and ‘LiFe0.90V0.05□0.05PO4’ materials formed phase-pure olivine samples according to XRD data, but attempts to prepare ‘LiFe0.70V0.15□0.15PO4’ resulted in Li3PO4 impurity (see Supporting Information Figure S1). Because this material was clearly deficient in Fe, the amount of Fe precursor was then increased with constant V:P:Li ratios until the Li3PO4 impurity disappeared. A phase-pure sample formed (with no Li3PO4 or Fe3O4) when the Fe/P ratio was ∼0.78−0.79. Furthermore, ICP data, shown in Table 1, indicate

the elimination of harmonics. Data were taken in transmission mode with metal (Fe and V) reference foil downstream of the sample. The 20-cm-long ion chambers contained flowing gas mixtures tuned to obtain 10% absorption in Io and 80% absorption in It and Iref. X-ray absorption near edge spectroscopy (XANES) data were processed with Athena47,48 by first aligning the reference spectra for all data sets and then adjusting the normalization parameters so as to have all spectra match before the edge and approximately 150 eV above the edge. Least-squares fitting was performed by Athena, with all fractions constrained to add up to 1 and requiring any energy shift to be identical for all standards. The Raman measurements were made with a Renishaw inVia Raman Microscope at Argonne National Laboratory equipped with a set of four lasers that provide excitation wavelengths at 785, 633, 514, 442, and 325 nm. The samples (in powder form) were loaded into a threaded Teflon holder equipped with a rubber O-ring seal to prevent contact of the sample with ambient atmosphere. In this embodiment, the powder is pressed between a stainless backing disc and a BaF2 window (10 mm in diameter and 2 mm thick). Raman spectra were recorded through the BaF2 window with a 50-X focusing/collection optic with an NA of 0.5. The excitation laser was brought to a focus at the sample/BaF2 interface, and the beam was spread to a circular area approximately 10 μm in diameter. The effective laser power density used in these measurements did not exceed 1 mW/μm2. Scanning transmission electron microscopy (STEM) images were obtained on a Hitachi S5500 SEM/STEM microscope with energy dispersive spectroscopic (EDS) capability at the University of Texas at Austin. Fourier transform infrared (FTIR) spectra data were collected with a PerkinElmer BX FTIR spectrometer at the University of Texas at Austin. Pellets for FTIR analysis were prepared by grinding and pressing samples with dried KBr powder. To make electrodes, 75 wt % active material was ground with 12.5 wt % each conductive carbon and teflonated acetylene black (TAB). Composite electrodes were made by rolling the ground material into thin sheets and cutting out 0.64 cm2 circular disks with a punch. Each electrode contained ∼5 mg of active material. The electrodes were dried overnight in a vacuum oven at 115 °C. CR2032 coin cells were then assembled in an Ar-filled glovebox with metallic lithium as the anode, Celgard polypropylene separator, and 1 M LiPF6 in 1:1 diethyl carbonate/ethylene carbonate as the electrolyte. A Radiometer Analytical Voltalab PGZ402 potentiostat was used to collect cyclic voltammetry (CV) data with a two-electrode coin cell assembly at a rate of 0.1 mV/s. Coin cells were tested with an Arbin battery cycler, and first charge−discharge curves at a rate of C/10 are reported.

Table 1. ICP Data of the V-Doped Samples Prepared According to ‘LiFe1−x(VO)xPO4’, ‘LiFe1‑2xVx□xPO4’, and LiFe1‑3x/2Vx□x/2PO4 intended sample

impuritya

Li/Pb

Fe/Pb

V/Pb

‘LiFe0.90V0.05□0.05PO4’ ‘LiFe0.80V0.10□0.10PO4’ ‘LiFe0.70V0.15□0.15PO4’ LiFe0.775V0.15□0.075PO4 LiFe0.775V0.15□0.075PO4c ‘LiFe0.85(VO)0.15PO4’ ‘LiFe0.75(VO)0.25PO4’

none none Li3PO4 none none Fe3O4 (removed) Fe3O4 (removed)

0.99 1.00 1.00 1.00 0.97 1.02 0.99

0.92 0.83 0.74 0.79 0.79 0.77 0.70

0.05 0.10 0.15 0.14 0.15 0.16 0.24

“none” means no crystalline impurities were detected by XRD or by stirring with a magnet. bErrors in ICP ratios are estimated to be around 2−3%. cWashed with isopropyl alcohol.

a

that a ll of the sa mples p repared according t o ‘LiFe1−2xVx□xPO4’ had higher iron contents than expected from the precursor stoichiometry. These results suggest an oxidation state closer to V3+ than V4+; a V3+ oxidation state implies LiFe1−3x/2Vx□x/2PO4. We also found that if we washed the samples with isopropyl alcohol (in addition to normal acetone washing), the lithium content found by ICP apparently decreased a few percent (although the decrease was within error of the ICP measurements), which may indicate slight lithium deficiency in the olivine structure. The lithium hydroxide precursor is not very soluble in the synthesis solvent (TEG) at room temperature or in the acetone used to wash the samples, although it has limited solubility in isopropyl alcohol. We also synthesized LiFePO4 with excess Li in the precursor solution and the ICP data revealed excess Li in the product. Since LiFePO4 accommodates only one Li+, these results indicate that excess Li precursor does not wash out completely in acetone after a reaction. Therefore, lithium deficiency in the V-doped samples with an acetone wash may not be adequately detected by elemental analysis. ICP (Table 1) data and synthesis results (Supporting Information Figure S1) suggest a formula of LiFe 1 − 3 x / 2 V x □ x / 2 PO 4 (implying V 3 + ) rather than ‘LiFe1−2xVx□xPO4’ or ‘LiFe1−x(VO)xPO4’ (implying V4+). Again, we use single quotes around these two formulas to

■

RESULTS AND DISCUSSION Characterization of Pristine MW-ST V-Doped LiFePO4 Samples. The present work clarifies and expands upon our previous report of V-doped LiFePO4 synthesized by the MWST method.40 Synthesis of ‘LiFe1−x(VO)xPO4’ (0 ≤ x ≤ 0.25) cathodes was initially reported since a vanadyl precursor was used in the synthesis and FTIR spectra showed bands at the same location as the VO bond in LiVOPO4.40,49 We use single quotation marks around ‘LiFe1−x(VO)xPO4’ to indicate that this was the intended composition in the sense that precursors were mixed in ratios according to ‘LiFe1−x(VO)xPO4.’ However, as was discussed in our previous work and as we will briefly describe here,40 ‘LiFe1−x(VO)xPO4’ is not the composition that actually formed. Electrochemical data for ‘LiFe1−x(VO)xPO4’ also showed activity in the same voltage ranges as LiVOPO4, and XPS analysis was consistent with an oxidation state of V4+.40 However, upon closer examination of the olivine structure, we recognized that the hexagonal, close packed oxygen array prevents accommodation of large (VO)2+ ions in the octahedral sites due to size constraints and the steric hindrance associated with adding an extra oxide ion into the lattice. Moreover, attempts to 770



Article

Li3V2(PO4)3 has much more symmetric octahedra. For Li3V2(PO4)3, there is a double peak, which reflects the crystal-field splitting of the V 3d orbitals into t2g and eg sets.56−58 The V-doped LiFePO4 samples show larger preedge peaks than Li3V2(PO4)3, but smaller pre-edge peaks than LiVOPO4, indicating that the VO6 octahedra are less distorted than in LiVOPO4, but more distorted than in Li3V2(PO4)3. It is interesting to note that there is a slight decrease in pre-edge intensity with increasing V doping levels, which is consistent with previous results on V-doped LiFePO4 samples.28 The V oxidation state can be estimated by least-squares fitting of the standards. The estimated V oxidation states for the samples are presented in Supporting Information Table S1. The fit (detailed in Supporting Information Figure S2) suggests a mixed oxidation state consisting of 77−88% V3+ and 12−23% V4+. Although this is only an estimate, the XANES (Figure 1 and Supporting Information S2), ICP (Table 1), and synthesis results (Supporting Information Figure S1) all suggest that the oxidation state of V is much closer to V3+ than V4+ or (VO)2+ as was originally reported. The edge position of the V-doped sample does not exactly match that of the V3+ standard and corresponds to an estimated oxidation state of V3.2+. The XANES data show that there is no systematic difference in oxidation state between the samples prepared according to different formulas, indicating a preferred stoichiometry near LiFe1−3x/2Vx□x/2PO4, with Fe2+ and V3.2+. There is a Li3PO4 impurity when mixing precursors according to ‘LiFe1−2xVx□xPO4’ and Fe3O4 forms when mixing precursors according to ‘LiFe1−x(VO)xPO4’, reflecting the fact that these are not the preferred stoichiometry. XANES data (Supporting Information Figure S3) were also taken on the Fe edges and were compared to commercial FePO4·xH2O, MW-ST LiFePO4, and conventionally prepared LiFePO4. The edge positions and pre-edge regions for the Vdoped samples and LiFePO4 (Fe2+ standards) are almost identical and are distinctly different from the spectrum for the FePO4 (Fe3+) standard. The edge position shows no change from the LiFePO4 standards, and there is little change in the pre-edge feature, indicating that the FeO6 octahedra are not significantly changed by the V doping, in agreement with the literature.28 In our initial publication on V-doped LiFePO4,40 we discussed a shoulder in FTIR measurements around the same location as the VO bond in LiVOPO4. Since V3+ does not form a VO bond and because we have only a small amount of V4+ in the V-doped samples, it is unlikely that the FTIR shoulder around 900 cm−1 (discussed in more detail in our previous publication40) is indicative of VO. To better understand the FTIR shoulder, we performed the Raman spectroscopy measurements shown in Figure 2. Intense fluorescence obscured some details of the Raman spectral patterns from the Fe-containing samples. Through the use of low laser excitation power it was possible to avoid burning/ decomposition of the samples, such that spectra were obtained for LiFePO4, 15% V-doped LiFePO4, and LiVOPO4. It is clear that, unlike LiVOPO4, there is no peak around 882 cm−1 corresponding to a VO bond for the V-doped LiFePO4 sample.49 Thus, the shoulder around 900 cm−1 seen in the FTIR patterns40 is not indicative of a VO bond in these samples. In our previous study,40 we presented lattice parameters indicating that the unit-cell volume decreases with increasing doping levels, which is consistent with the formation of cation

remind the reader that precursors were mixed in the ratios corresponding to these formulas. However, the actual formula is LiFe1−3x/2Vx□x/2PO4. As shown in Table 1, mixing precursors to form ‘LiFe1−x(VO)xPO4’ or LiFe1−3x/2Vx□x/2PO4 essentially leads to the same product after the Fe3O4 impurity is removed. Note that the Fe/P ratio for ‘LiFe0.70V0.15□0.15PO4’ differs, due to the fact that the ICP ratios consist of contributions from both the olivine and Li3PO4 impurity phases and the Li3PO4 impurity cannot simply be removed as in the case of the Fe 3 O 4 impurity. The formula LiFe1−3x/2Vx□x/2PO4 implies V3+, which disagreed with our previous analysis of XPS data showing V peaks at locations similar to those in LiVOPO4,40 indicating a V4+ oxidation state. A V3+ standard was not analyzed for comparison in our previous publication.40 New analysis revealed that we could not distinguish between Li3V2(PO4)3 (V3+ standard) and βLiVOPO 4 (V 4+ standard) with XPS. XPS analysis is complicated by many factors. It is a surface sensitive technique, and the charge compensation used during data collection for insulating samples like LiFePO4 necessitates calibration of the data to a common C or O peak location,51 which can complicate the analysis. We were unable to distinguish between our V3+ and V4+ standards with XPS regardless of whether we calibrated the data to C 1s or O 1s peak locations. The splitting between V3+ and V4+ has been previously demonstrated to be small (0.1−0.5 eV), and there is comparatively large variation in the literature for the V 2p3/2 peak location for Li3V2(PO4)3 (∼1.9 eV).51−54 To determine the oxidation state without the difficulties associated with XPS and to take advantage of bulk rather than surface sensitivity, we collected XANES data (see Supporting Information Figure S2) on samples prepared according to the formulas ‘LiFe 1−x (VO) x PO 4 ’ (x = 0.05, 0.10, 0.15), ‘LiFe1−2xVx□xPO4’ (x = 0.05 and 0.10), and LiFe1−3x/2Vx□x/2PO4 (x = 0.15). A subset of these samples are compared to LiVOPO4 and Li3V2(PO4)3 standards in Figure 1. The standards show spectra consistent with the literature.55 The pre-edge feature is indicative of VO6 distortion and arises because of the 1s → 3d transition for V which is possible due to V 3d/V 4p and O 2p mixing. LiVOPO4 has a much larger pre-edge peak than Li3V2(PO4)3 because the VO bond in LiVOPO4 leads to a significant distortion, whereas

Figure 1. V edge XANES data of V-doped LiFePO4 samples and standards, with the inset showing an expanded region of the V preedge. With the exception of the sample LiFe0.775V0.15□0.075PO4, the formulas for the V-doped LiFePO4 samples represent the intended sample compositions, and the precursor ratios that were used for synthesis, rather than the actual compositions. 771



Article

diffraction was performed for the 5, 10, and 15% V-doped samples, and Table 3 shows a summary of the results. It should be noted that the 25% V sample is difficult to synthesize in large amounts required for neutron diffraction due to the high pressure generated by the low-boiling-point V precursor used in the MW-ST synthesis. Therefore, the 25% V-doped sample was not included in the neutron analysis. The lattice parameters found by neutron diffraction are similar to those found from XRD, as is shown in Table 3. Fits to the neutron data with no chemical constraints and the V occupancy set at the nominal value are physically reasonable, with the implication that the occupancies are consistent with Fe2+ and V3+ within error. A large Li displacement parameter (Uiso) may relate, in part, to static disorder, but it is difficult to model adequately by Rietveld refinement using the available data. Attempts to refine with the vanadium oxidation state constrained to V4+ produced unphysical results. It is also worth noting that the neutron refinement results suggest a small amount of Li deficiency in the samples with higher V doping levels, which is consistent with the presence of a small amount of V4+ to give charge balance on the Fe site (in agreement with XANES oxidation state analysis). A representative neutron refinement fit is shown in Supporting Information Figure S7. It is also worth noting the possibility that the vacancies on the Li site suggested by the NPD results for higher V doping levels could represent a small amount of doping on the Li site, given that NPD is not sensitive to V. Attempts to refine Fe on the Li site resulted in ∼0% Fe occupancy within error for all V doping levels. While 1% V doping on the Li site has been reported,36 it is unlikely that large amounts of V would be present on the Li site since our other experiments (ICP, synthesis results in Supporting Information Figure S1, and XRD refinement) all are consistent with Fe site doping. Also, V doping on the Li site is reported to cause an increase in unit cell volume,36 while we observe a significant decrease in the unit cell volume with V doping. Our volume decrease is consistent with the many studies also showing V doping on the Fe site. Furthermore, at the high doping levels presented in this work, there would be significant capacity loss if the majority of the doping occurred on the Li site, which is not consistent with our electrochemical results. Therefore, we believe that the majority of the doping occurs on the Fe site, but there is some evidence from the neutron diffraction data that suggests a small amount of V may be present on the Li site for the higher V-doped samples. Further insight into structure and oxidation state can be provided by STEM and EELS, shown in Supporting Information Figure S4. EELS data confirm an oxidation state close to V3+, as demonstrated by the vanadium L3/L2 ratio of 1.61 ± 0.08.59 High-angle annular dark-field STEM images clearly show the presence of defects in the V-doped LiFePO4 samples, which likely represent a small amount of Fe or V on

Figure 2. Raman spectra of LiFePO4, V-doped LiFePO4 prepared according to LiFe0.775V0.15□0.075PO4, and LiVOPO4.

vacancies to balance the higher charge from the V dopant. To obtain more accurate values for the lattice parameters, a Si internal standard was used and Rietveld refinement was employed. The lattice parameter values presented in Table 2 confirm a decreasing unit-cell volume with increasing V doping. The occupancies of V and Fe were also estimated (Table 2). Convergence could not be reached through attempts to restrain the total cation charge to be equal to 3+ (to balance with PO43−). Since X-rays are not sensitive to Li, it would be difficult to determine the lithium content through the refinement using XRD data, so the Li occupancy was assumed to be unity and all of the V was assumed to be on the Fe site. A restraint was placed such that the total charge on the iron site had to add up to 2+ (including Fe, V, and vacancies). The oxidation state of V was assumed to be 3.2+ for this analysis based on the average oxidation state found from the XANES data. Although 3.2+ is only an estimate for the V oxidation state, this assumption leads to good agreement between the ICP data and the XRD-refined V occupancies. However, if V3+ is assumed instead, the results are still reasonable within error. For example, structural refinement of the 15% V-doped structure assuming V3.2+ results in an occupancy of 0.149(11) and assuming V3+ results in an occupancy of 0.172(13). The error bars for these values overlap. Attempts to add analogous restraints on the Li site led to nonphysical results. More discussion regarding the assumptions made during the refinement is given in the Supporting Information. Rietveld refinement using the XRD data support V and Fe occupancies similar to the expected values from ICP in these highly doped samples. However, the V occupancy of the 25% V-doped sample was found to be slightly lower than expected, indicating a solubility limit below 25% V from the 300 °C synthesis. As neutron diffraction is sensitive to Fe and Li, but not V, it can be used to estimate the Fe and Li occupancies. Neutron

Table 2. Rietveld Refinement Results Using the X-ray Diffraction Data of Undoped and V-Doped LiFePO4 (V-Doped Samples Prepared According to LiFe1‑3x/2Vx□x/2PO4) sample

a (Å)

b (Å)

c (Å)

V (Å3)

Feocc

Vocc

χ2

LiFePO4 5% V 10% V 15% V 25% V

10.32093(20) 10.31193(21) 10.30262(22) 10.27589(25) 10.24307(34)

6.00127(12) 5.99441(12) 5.98984(13) 5.97117(14) 5.95064(19)

4.69262(9) 4.69345(9) 4.69563(10) 4.69939(11) 4.70406(14)

290.655(10) 290.121(10) 289.772(11) 288.350(12) 286.726(16)

0.933(18) 0.823(17) 0.764(18) 0.689(17)

0.042(11) 0.111(11) 0.149(11) 0.195(11)

1.55 1.92 1.70 1.40 1.44

772



Article

Table 3. Rietveld Refinement Results for Samples Prepared According to ‘LiFe1−x(VO)xPO4’ Using Neutron Diffraction Data Collected at Room Temperaturea samples [Li at (0,0,0)] a (Å) b (Å) c (Å) V (Å3) UisoLi Liocc Feocc Vocc Fen+ Vn+ χ2 wRp

samples (Li off-site)

5% V

10% V

15% V

5% V

10% V

15% V

10.3122(7) 5.9963(4) 4.6982(4) 290.51(6) 2.4(3) 1.04(3) 0.916(5) 0.05 1.97 2.52 2.24 0.0334

10.2986(7) 5.9840(4) 4.7003(3) 289.67(6) 2.4(3) 0.98(3) 0.855(5) 0.10 2.01 3.12 2.98 0.0313

10.2826(8) 5.9714(5) 4.7018(4) 288.70(7) 2.1(4) 0.94(3) 0.801(6) 0.15 2.01 3.05 3.43 0.0374

10.3115(7) 5.9959(4) 4.6979(4) 290.46(6) 1.0(3) 1.04(3) 0.923(5) 0.05 1.96 2.28 2.17 0.0329

10.2980(7) 5.9838(4) 4.7001(3) 289.63(6) 0.3(4) 0.92(3) 0.860(5) 0.10 2.02 3.20 2.92 0.0309

10.2820(8) 5.9712(4) 4.7016(4) 288.66(6) −0.6(4)b 0.92(3)c 0.808(6) 0.15 2.02 3.12 3.30 0.0367

a

For the Fe oxidation state calculation, V3+ was assumed. For the V oxidation state calculation, Fe2+ was assumed. bThe Uiso values have been multiplied by 100, and have units of Å2. cFor the Li off-position refinements, the refined occupancy has been multiplied by a factor of 2 for comparison with the refinements with Li at (0,0,0). The multiplicity of Li off-position is 8, rather than 4 at (0,0,0).

Table 4. Summary of As-Prepared and Post-Heated Samples of Undoped and V-Doped LiFePO4a

a

sample

intended product

synthesis method

MW temp.

MW time

furnace temp.

furnace time

furnace atmosphere

MW-LFP-unheated MW-LFPdef-unheated MW-LFP-H2-725-6h MW-LFVP-unheated MW-LFVP-H2-525-6h MW-LFVP-H2-625-6h MW-LFVP-H2-725-6h MW-LFVP-Ar-525-6h MW-LFVP-Ar-625-6h MW-LFVP-Ar-725-6h MW-LFVP-Ar-525-15h MW-LFVP-Ar-625-15h MW-LFVP-Ar-725-15h CONV-LFP-Ar-725-6h CONV-LFPdef-Ar-725-6h CONV-LFP-H2-525-6h CONV-LFP-H2-625-6h CONV-LFP-H2-725-6h CONV-LFVP-H2-525-6h CONV-LFVP-H2-625-6h CONV-LFVP-H2-725-6h CONV-LFVP-Ar-525-6h CONV-LFVP-Ar-625-6h CONV-LFVP-Ar-725-6h Li3V2(PO4)3 LiVOPO4

LiFePO4 LiFe0.85PO4 LiFePO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFePO4 LiFe0.85PO4 LiFePO4 LiFePO4 LiFePO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 LiFe0.775V0.15PO4 Li3V2(PO4)3 LiVOPO4

MW MW MW MW MW, heat MW, heat MW, heat MW, heat MW, heat MW, heat MW, heat MW, heat MW, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat ball mill, heat sol−gel, heat

300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C 300 °C N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min 30 min N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A N/A

N/A N/A 725 °C N/A 525 °C 625 °C 725 °C 525 °C 625 °C 725 °C 525 °C 625 °C 725 °C 725 °C 725 °C 525 °C 625 °C 725 °C 525 °C 625 °C 725 °C 525 °C 625 °C 725 °C 725 °C 500 °C

N/A N/A 6h N/A 6h 6h 6h 6h 6h 6h 15 h 15 h 15 h 6h 6h 6h 6h 6h 6h 6h 6h 6h 6h 6h 6h 4h

N/A N/A 5% H2 N/A 5% H2 5% H2 5% H2 Ar Ar Ar Ar Ar Ar Ar Ar Ar Ar Ar 5% H2 5% H2 5% H2 Ar Ar Ar 5% H2 Air

Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4.

and 725 °C in Ar and 5% H2−95% Ar environments. We compare these results to V-doped LiFePO4 prepared by conventional ball milling and high-temperature heating. Samples are identified by the naming scheme given in the Experimental Section and summarized in Table 4. XRD patterns are shown in Figure 3 for the as-prepared MW-LFP-unheated and MW-LFVP-unheated samples as well as the MW-ST samples subsequently heated in 5% H2−95% Ar at various temperatures. Results for the samples heated in Ar and 5% H2−95% Ar were found to be very similar, so only the samples heated in 5% H2−95% Ar are shown here. The XRD

the Li site, though this may not be representative of the bulk. This possibility is discussed further in the Supporting Information. Effect of Temperature on V-Doped LiFePO4 Samples. Since we are able to incorporate at least 20% V into the olivine lattice while other studies have been limited to 10% or less,27−36 we investigated whether the high-doping levels are due to the low-temperature MW-ST synthesis method. Accordingly, we prepared the 15% V sample by mixing precursors in the appropriate ratios to form LiFe0.775V0.15□0.075PO4, and we have heated it at 525, 625, 773



Article

number of impurity peaks corresponding to monoclinic Li3V2(PO4)3. Conversely, the MW-LFVP-H2-725-6h sample has obvious impurity peaks that can be identified easily as Li3V2(PO4)3. Several studies28−30,32,33,60 have shown that a Li3V2(PO4)3 impurity forms with high-temperature synthesis when substituting V into the olivine lattice at levels above 5− 10%, so this result was anticipated. XRD patterns for the post-heated microwave-synthesized samples can be compared to the LiFe0.775V0.15□0.075PO4 samples synthesized by conventional ball milling and heating (CONV-LFVP in Figure 3). Li3V2(PO4)3 peaks are detectable in the samples at all three heating temperatures in contrast to the MW-ST synthesized and heated samples which showed major impurity peaks at 725 °C and slight impurity peaks at 625 °C. To eliminate any possibility that different heating conditions were responsible for the difference in the Li3V2(PO4)3 formation, we heated the MW-ST sample and the precursors for the conventionally doped samples in the same tube furnace in adjacent crucibles. This result was repeatable, which confirms that the MW-ST synthesis provides a product with higher doping levels that is stable at elevated temperatures. If the V had not been present within the olivine lattice, the V should have formed Li3V2(PO4)3, as it did in the conventionally prepared samples. Even after 15 h of heating in Ar, the microwave-synthesized LFVP sample showed a phasepure pattern at 525 and 625 °C (Supporting Information Figure S5). As shown in Tables 2 and 3, in addition to an increase in the length of the c axis, there is a systematic decrease in unit-cell volume (and a and b axes lengths) with increasing V doping. Therefore, an increase in unit-cell volume and decrease in c axis length are expected if heating results in V being leached out

Figure 3. XRD patterns of the as-prepared and post-heated samples of undoped and V-doped LiFePO4 with an internal Si standard. Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4.

patterns for MW-LFVP-H2-525-6h and MW-LFVP-H2-625-6h closely resemble that for MW-LFVP-unheated with no obvious impurity phases. Closer inspection of the MW-LFVP-H2-6256h sample (Supporting Information Figure S5) reveals a small

Table 5. Summary of Lattice Parameters of As-Prepared and Post-Heated Samples of Undoped and V-Doped LiFePO4 Obtained by Rietveld Refinement Using XRD Dataa

a

sample

a (Å)

b (Å)

c (Å)

V (Å3)

χ2

MW-LFP-unheated MW-LFPdef-unheated MW-LFP-H2-700-6h MW-LFVP-unheated MW-LFVP-H2-525-6h MW-LFVP-H2-625-6h MW-LFVP-H2-725-6h MW-LFVP-Ar-525-6h MW-LFVP-Ar-625-6h MW-LFVP-Ar-725-6h MW-LFVP-Ar-525-15h MW-LFVP-Ar-625-15h MW-LFVP-Ar-725-15h CONV-LFVP-H2-525-6h CONV-LFVP-H2-625-6h CONV-LFVP-H2-725-6h CONV-LFVP-Ar-525-6h CONV-LFVP-Ar-625-6h CONV-LFVP-Ar-725-6h CONV-LFP-Ar-725-6h CONV-LFPdef-Ar-725-6h CONV-LFP-H2-525-6h CONV-LFP-H2-625-6h CONV-LFP-H2-725-6h

10.32093(20) 10.32392(14) 10.32653(8) 10.27589(25) 10.28668(14) 10.29490(14) 10.31397(10) 10.28145(15) 10.28643(14) 10.31407(10) 10.28433(16) 10.28678(14) 10.31685(8) 10.31085(30) 10.30902(25) 10.30636(21) 10.3097(3) 10.30809(27) 10.31044(20) 10.32494(8) 10.32681(11) 10.32535(15) 10.32441(13) 10.32507(9)

6.00127(12) 5.99971(8) 6.00370(5) 5.97117(14) 5.97441(8) 5.98145(9) 5.99741(6) 5.97428(9) 5.97664(9) 5.99812(6) 5.97423(9) 5.97686(8) 6.00016(5) 6.00156(16) 5.99882(14) 5.99633(12) 6.00242(17) 6.00045(15) 5.99990(12) 6.00591(5) 6.00602(6) 6.00610(9) 6.00614(7) 6.00606(5)

4.69262(9) 4.69249(6) 4.69084(4) 4.69939(11) 4.69636(6) 4.69423(6) 4.69195(5) 4.69722(7) 4.69590(6) 4.69239(5) 4.69697(7) 4.69540(6) 4.69230(4) 4.69548(13) 4.69585(11) 4.69407(10) 4.69601(14) 4.69596(12) 4.69530(9) 4.69161(4) 4.69043(5) 4.69367(7) 4.69241(6) 4.69162(4)

290.655(10) 290.655(7) 290.820(4) 288.350(12) 288.623(7) 289.063(7) 290.230(5) 288.523(7) 288.696(7) 290.295(5) 288.586(8) 288.686(7) 290.466(4) 290.562(14) 290.400(12) 290.095(10) 290.604(14) 290.461(13) 290.459(10) 290.930(4) 290.915(5) 291.079(7) 290.975(6) 290.941(4)

1.55 2.66 2.44 1.40 1.94 2.21 2.67 1.88 2.24 2.52 1.85 2.11 2.70 1.68 1.74 2.54 1.84 1.85 1.95 2.40 3.64 1.80 1.95 2.27

Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4. 774



Article

from the olivine lattice. Table 5 (and Supporting Information Figure S6) shows that the MW-LFVP-H2-525-6h, MW-LFVPAr-525-6h, and MW-LFVP-Ar-525-15h samples have similar unit-cell volumes and all show an increase relative to the pristine MW-LFVP-unheated that cannot be accounted for by the errors in refinement. At higher temperatures, there is a systematic increase in unit-cell volume with increasing temperature, with the most significant increase in the unit-cell volume present for the samples heated at 725 °C. In addition, there is a systematic decrease in the length of the c axis. However, the lattice parameters did not shift completely back to those for MW-LFP-unheated; a phase with ∼10% V doping remains after segregation of the Li3V2(PO4)3 phase. The unit cell volumes for the conventionally heated samples are shown to be similar for all heating conditions, and the unit-cell volumes are all lower than that for the conventional LiFePO4 samples, indicating V doping. The unit-cell volumes are similar to those for the microwave-synthesized samples heated at 725 °C and have a volume similar to our pristine MW-ST LiFePO4 doped with ∼5% V (with no post-heating step). Since the precursors for these doped samples all have deficient amounts of Fe relative to Li and P, we also synthesized samples with Fe deficiency, but without V, i.e., “LiFe0.85PO4” as the intended formula, both conventionally and by the MW-ST process. These samples are referred to as CONV-LFPdef-Ar725-6h and MW-LFPdef-unheated, respectively, and the lattice parameters are shown in Table 5. In both cases, olivine and impurity phases formed; the deficient iron samples have similar unit-cell volumes for the olivine phase as their stoichiometric counterparts. This result shows that Fe vacancies do not occur in the MW-ST process described here without V doping, and the lower unit cell volume of the doped samples are due to the presence of V in the olivine lattice. It is worth noting that the MW-LFP-unheated sample (prepared at 300 °C) has a slightly lower unit-cell volume than the conventionally prepared LiFePO4 samples (prepared at 725 °C). However, upon heating in a 5% H2−95% Ar atmosphere, the lattice parameters are similar, as indicated in Table 5 by the MW-LFP-H2-725-6h sample. Because XANES data shows no evidence of Fe3+ in our pristine MW-ST LiFePO4 samples (i.e., no shift in the edge position from the conventionally prepared LiFePO4), it is unclear why the lattice-parameter change occurs with heating temperature. Several studies have suggested an increase in unit cell volume with increasing heating temperature and/or particle size.12,61−64 The change in unit cell volume with temperature may be due to defects in the material, as has been discussed in more detail elsewhere for MW-ST LiFePO4 samples.61 To estimate the amount of vanadium remaining in the lattice after heating, we refined the XRD patterns for V and Fe occupancies with a restraint that the total charge on the Fe site must be 2+ and assuming Fe2+ and V3+. The MW-LFVPunheated sample was found to have a slightly higher oxidation state for V, but the heated samples were assumed to consist of V3+ since the conventional V-doped samples have Li3V2(PO4)3 impurity, indicating V3+ is the stable oxidation state. A representative refinement fit is shown in Supporting Information Figure S7, where the refinement conditions are discussed in more detail. As shown in Table 6, the samples heated at 525 and 625 °C still have around 15% V doping, but the samples heated at 725 °C lose V from the olivine lattice until doping levels of ∼10% or less are achieved.27−36 To corroborate these results, the amount of V remaining in the lattice was also estimated by assuming V loss from the olivine

Table 6. Fractional Occupancy and Phase Fraction Results for As-Prepared and Post-Heated Samples of Undoped and V-Doped LiFePO4 from Rietveld Refinementa Vocc from % LVPb

%LVP

0.0883(16)

9.50(33)

0.0802(18)

10.61(36)

0.070(10)

0.0668(21)

12.39(41)

0.843(35)

0.098(40)

0.1168(15)

5.36(31)

0.839(17)

0.107(12)

0.1144(15)

5.72(30)

0.805(19)

0.130(13)

0.0975(19)

8.24(39)

0.842(20)

0.105(13)

0.1175(14)

5.26(30)

0.824(18)

0.117(12)

0.1171(14)

5.33(29)

0.830(17)

0.113(11)

0.1035(17)

7.36(35)

sample

Feocc

Vocc

MW-LFVPunheated MW-LFVP-H2525-6h MW-LFVP-H2625-6h MW-LFVP-H2725-6h MW-LFVP-Ar525-6h MW-LFVP-Ar625-6h MW-LFVP-Ar725-6h MW-LFVP-Ar525-15h MW-LFVP-Ar625-15h MW-LFVP-Ar725-15h CONV-LFVP-H2525-6h CONV-LFVP-H2625-6h CONV-LFVP-H2725-6h CONV-LFVP-Ar525-6h CONV-LFVP-Ar625-6h CONV-LFVP-Ar725-6h

0.764(18)

0.149(11)

0.765(14)

0.156(9)

0.798(14)

0.135(10)

0.875(14)

0.084(9)

0.777(13)

0.148(9)

0.763(14)

0.158(10)

0.878(14)

0.082(10)

0.767(14)

0.155(9)

0.769(14)

0.154(9)

0.895(15)

a

Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4. bErrors for “Vocc from %LVP” estimated by sequential perturbation because an iterative solution was required. “LVP” refers to Li3V2(PO4)3. Note that the site occupancies as presented are double the value refined in Fullprof, such that a value of 1 would indicate full occupancy. Fullprof defines site occupancy as the site multiplicity divided by the multiplicity of the general position, necessitating the correction factor to get the occupancies on a scale of 1. For the Li and Fe sites, full occupancy of the site is given as “0.5” in the program.

phase forms Li3V2(PO4)3. This indirect analysis generally produced similar results (Table 6). The maximum V doping demonstrated in the olivine lattice is summarized as a function of temperature in Figure 4 and compared to other studies. The results indicate the overall trend that the maximum achievable doping levels before impurity formation decreases with increasing temperature. To confirm the assumption made in the Rietveld analysis that the monoclinic Li3M2(PO4)3 phase consists of M = V rather than M = Fe, STEM data were collected with elemental mapping analysis. Figure 5a shows a MW-LFVP-unheated nanorod and corresponding elemental dot maps for Fe, V, and P. It is clear that the V and Fe are evenly distributed along the nanorod imaged on the length scale of the image. Since the morphology of this sample includes nanorods as well as more spherical particles, we imaged a range of different particle sizes and shapes and were unable to find evidence of any V-rich or Fe-poor regions. Figure 5b shows an STEM image of MWLFVP-H2-725-6h with corresponding elemental dot maps. It is 775



Article

agrees with our refinement results. At lower heating temperature (MW-LFVP-H2-525-6h) we were unable to find any Vrich or Fe-poor particles. In the MW-ST LFVP samples, these microscopy results corroborate the XRD results showing no impurities for the MW-LFVP-H2-525-6h sample, which illustrates that phase segregation to form Li3V2(PO4)3 occurs at some point above 525 °C. Microscopy was not performed for MW-LFVP-H2-625-6h, but XRD and electrochemical results (shown later) suggest that there is a small Li3V2(PO4)3 impurity present in the MW-LFVP-H2-625-6h sample. In contrast, for conventionally prepared samples CONV-LFVP-H2-525-6h and CONV-LFVP-H2-725-6h there is clear evidence of phase segregation even at 525 °C (Figure 6a,b). It is also worth noting that there is still a significant Figure 4. Maximum V doping as a function of synthesis/heating temperature. MW-ST V-doped LiFePO4 samples after heating are compared to conventionally prepared V-doped LiFePO4 samples from this study (red triangles) and from literature (black symbols). The legends indicate the heating environment and heating time. Note that the V occupancies used here are “Vocc from %LVP” where applicable. The MW-ST data point is V occupancy value on the Fe site for the “25%” V-doped sample, which is the maximum V doping that has been demonstrated. Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4.

Figure 6. STEM images and corresponding elemental dot maps of (a) CONV-LFVP-H2-525-6h and (b) CONV-LFVP-H2-725-6h. Note that the samples shown here are older samples which were actually heated at 500 and 700 °C rather than 525 and 725 °C, respectively, but this is not expected to affect the results.

amount of V in the Fe-rich particles at 725 °C. Clearly, for an initial stoichiometry of >10% V the conventional synthesis produces V-rich impurities at all synthesis temperatures, while the MW-ST samples show no XRD or microscopy evidence of decomposition until above at least 525 °C. In addition to diffraction and imaging techniques, FTIR spectroscopy can also provide insight regarding the phases formed upon heating the samples. Figure 7 shows FTIR spectra for several samples. The FTIR spectrum for the MW-LFVPunheated sample shares several features with the MW-LFPunheated spectrum. As discussed in more detail in our previous publication,40 a major distinction between the spectra for the doped and undoped LiFePO4 samples prepared by MW-ST is the shoulder at around 900 cm−1, which agrees with the previously documented position of the VO bond in

Figure 5. STEM images and corresponding elemental dot maps of (a) MW-LFVP-unheated and (b) MW-LFVP-H2-725-6h. Note that the MW-LFVP-H2-725-6h sample shown here is an older sample which was actually heated at 700 °C rather than 725 °C, but this is not expected to affect the results.

clear that after post-heating there are V-rich and Fe-poor particles present in this sample, indicating clear phase separation into a V-rich Li3M2(PO4)3 impurity with little Fe incorporated. There is still a significant amount of V in the Ferich particle indicating it remains significantly doped, which 776



Article

dashed-line box, as expected for Li3V2(PO4)3. These features are not present in the pristine doped or undoped MW-ST samples, or the MW-LFVP-H2-525-6h and MW-LFVP-H2-6256h samples. However, the features for Li3V2(PO4)3 are clearly present for the MW-LFVP-H2-725-6h and all of the conventionally heated samples. Similar IR spectra suggest the MW-ST LFVP samples are similar chemically at 525 and 625 °C, while a significant change in the spectrum is noted at 725 °C. This is in agreement with XRD results that also show little evidence for impurities until 725 °C. Electrochemical Characterization. Electrochemical data can be instructive in terms of understanding redox couples and for examining the presence of multiple phases. Figure 8 shows first charge−discharge curves for various samples in the voltage range from 1.5 to 4.8 V. Although it is common to test LiFePO4 between the limits of 2.0 and 4.3 V (which we have shown in Supporting Information Figure S9 for reference), we chose to

Figure 7. FTIR absorbance spectra of the as-prepared and post-heated samples of undoped and V-doped LiFePO4, conventionally prepared V-doped LiFePO4, and reference samples. Arrows pointing up indicate the presence of features of interest. Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4.

LiVOPO449 outlined in a dashed-line pink box. Although the VO bond peaks generally occur at higher wavenumbers, near 1000 cm−1, there are many examples of vanadium-based phosphates in the literature that exhibit VO bonds near 900 cm−1, as is the case for LiVOPO4.49,65−69 Reference spectra show that this peak is not from remnant reaction solvent. The MW-ST doped samples show a decreasing shoulder in this location with increasing heating temperature. Since Raman spectroscopy showed no evidence of a vanadyl bond, this peak cannot correspond to VO. However, it remains a noteworthy feature because it is present in the MW-ST doped samples, but not in the MW-LFP-unheated sample or in the conventionally doped samples. More data and discussion regarding the FTIR is given in Supporting Information Figure S8. The black dashed-line boxes in Figure 7 indicate another feature related to PO bonds that changes with V doping. The peaks for the ν3 PO stretching modes70,71 around 1050, 1100, and 1140 cm−1 are distinct for the MW-LFP-unheated sample. Only the peak around 1050 cm−1 remains sharp for the MW-LFVP-unheated and MW-LFVP-H2-525-6h samples. The peak at 1140 cm−1 becomes visible for the MW-LFVP-H2-6256h sample and is quite apparent for the MW-LFVP-H2-725-6h sample. The peak at around 1100 cm−1 appears more like a shoulder attached to the larger peak at around 1050 cm−1 for the MW-LFVP-H2-725-6h sample rather than two distinct peaks as are clear in the MW-LFP-unheated sample. The peak around 1100 cm−1 is barely visible in the MW-LFVP samples at the lower heating temperatures. The FTIR spectra for the conventionally prepared V-doped samples all appear similar to the MW-LFVP-H2-725-6h sample, regardless of temperature. In addition, the spectra of the heated materials in some cases contain extra features around 1220 cm−1,72 outlined by a gray

Figure 8. First charge−discharge curves of the MW-ST LiFePO4, heated and unheated MW-ST V-doped LiFePO4 samples, and conventionally prepared V-doped LiFePO4 samples, as well as Li3V2(PO4)3 and LiVOPO4 for a comparison. Capacities in the 1.5− 2.5 and 3.8−4.8 V regions are noted for each sample. Note that all Vdoped samples were prepared according to LiFe0.775V0.15□0.075PO4. 777



Article

expand the range to look for additional redox couples that could be assigned to different vanadium plateaus. The pristine LiFePO4 material has little capacity in the 3.8−4.8 V and 1.5− 2.5 V regions (the Fe2+/3+ couple operates at 3.45 V vs Li), but the 15% V-doped sample has significant capacity in both of these regions that can be assigned to V activity. In our previous publication, we compared V-doped samples to LiVOPO4 because the V-doped sample has activity around 4.1 V and close to 2 V, which are consistent with the V4+/5+ and V3+/4+ plateaus in LiVOPO4, respectively.41 However, in Li3V2(PO4)3 the first Li+ extraction (V3+/4+) occurs in two steps from 3.5 and 3.8 V.73 There are additional charge plateaus around 4.1 and 4.6 V, which correspond to the second (V3+/4+) and third (V4+/5+) Li+ extraction, respectively. The V3+/4+ reaction in Li3V2(PO4)3 and the V4+/5+ reaction in LiVOPO4 occur at similar potentials (near 4.1 V) because of the strong VO bond in LiVOPO4 and the different structures. Given that our samples have been proven to consist of mostly V3+, the charge plateau at around 4.1 V should be assigned to V3+/4+. This corresponds with the V3+/4+ couple in Li3V2(PO4)3. Besides the plateau around 4.1 V (dotted line box) of the MW-LFVP-pristine sample that can be assigned to the V3+/4+ couple, there is no evidence of a second higher-voltage plateau corresponding to V4+/5+ in this sample. This result is demonstrated more clearly by the cyclic voltammetry shown in Figure 9, which upon repeated cycling shows significant V activity only between 4.1 and 4.3 V. In the lower-voltage range, it has recently been shown that two more Li+ can be reversibly inserted into Li3V2(PO4)3 upon discharge between 1.5 and 2.0 V for the V2+/3+ couple to form Li5V2(PO4)3.74 As the discharge curve for LFVP also shows significant capacity between 1.5 and 2.5 V, as shown in the dashed line boxes of Figure 8, this plateau can be assigned to a V2+/3+ couple for LFVP. The ability to access the V2+ oxidation state may relate to the presence of Li and Fe vacancies (total occupancy on Fe site is 0.925) in the V-doped olivine structure, as the vacancies could be filled by Li+ ions. We obtain an additional ∼18 mAh/g (∼10% of theoretical capacity) in the 1.5−2.5 V range for our pristine V-doped sample when compared to LiFePO4 in the same range. This corresponds well with the additional capacity that can theoretically be obtained by filling the empty Fe sites which constitute 8.5% of the total Fe sites with extra Li which would correspond to V2+/3+ activity. Furthermore, displacement of Li to Fe vacancies can increase the vacancies in the 1D Li+ channels, and increased Li diffusion has been demonstrated previously for aliovalently doped samples.26,37 The overall contribution of V to the electrochemical behavior of the MWLFVP-pristine sample can be described by high voltage (∼ 4.1 V) and low voltage (∼1.5−2.5 V) plateaus corresponding to the V3+/4+ and V2+/3+ couples, respectively, with the vacancies on the Fe/V site likely playing a role in lithium diffusion and redox behavior. Further evidence of phase separation into LiFePO4 and Li3V2(PO4)3 can be established through electrochemical measurements owing to the different redox energies of the Fe and V redox couples. After heating the MW-LFVP-pristine sample at 525 and 625 °C, the capacity between 3.8 and 4.8 V (mainly V3+/4+) decreases slightly, but the capacity between 1.5 and 2.5 V (V2+/3+) decreases more significantly. This decrease is particularly obvious for the sample heated at 625 °C, indicating that the V2+/3+ couple is no longer accessible in this sample. Cyclic voltammetry data show that small peaks arise in the 3.8 to 4.8 V region that were not present in the pristine sample,

Figure 9. Second cyclic voltammetry curves of MW-ST LiFePO4, heated and unheated MW-ST V-doped LiFePO4 samples, and conventionally prepared V-doped LiFePO4 samples, as well as Li3V2(PO4)3 and LiVOPO4 for a comparison. Note that all V-doped samples were prepared according to LiFe0.775V0.15□0.075PO4.

indicating that a small amount of a new phase has formed. Although the peaks are shifted from their locations in Li3V2(PO4)3, Supporting Information Figure S10 shows that these peaks align well with those for Li3V2(PO4)3 when cycled from 2.0 to 4.3 V in agreement with the XRD results which suggest a small amount of Li3V2(PO4)3 in the MW-ST V-doped sample heated at 625 °C. Heating the pristine sample at 725 °C led to clear peaks in the cyclic voltammogram for Li3V2(PO4)3 (Figure 9) and a plateau around 4.6 V corresponding to V4+/5+ in Li3V2(PO4)3. Similarly, distinct peaks arise in the 1.5−2.5 V region for V2+/3+ in Li3V2(PO4)3, as is expected since Li3V2(PO4)3 was detected by XRD and FTIR for this sample. There are also distinct fingerprint plateaus and peaks for Li3V2(PO4)3 in all of the conventionally prepared samples, which also show higher capacities. In fact, the total capacities for the conventionally prepared samples are higher than the theoretical capacity of LiFePO 4 . However, since the theoretical capacity of Li3V2(PO4)3 is ∼329 mAh/g when cycled in this range with 778



■

5 Li+ inserted/extracted, even 5−10% Li3V2(PO4)3 impurity can approximately account for the increased capacity. It also may be possible that some Li ions can be inserted into vacancies on the Fe site. The higher capacities shown for the conventionally prepared samples are associated with carbon coating from the organic oxalate precursors used in the synthesis. The MW-ST samples are a light gray color with a pink tint. Contrastingly, the conventional samples are all dark gray in color, consistent with carbon-coated samples. What Is the Impact of Vanadium Doping on the Structure and Ionic Conduction Mechanism? This question is stimulated by the STEM images of Supporting Information Figure S4 showing defects in the vicinity of the Li site and the evidence for Li displacements in the NPD refinement. What distinguishes vanadium oxides from other transition-metal oxides, in addition to vanadyl VO formation with formal valence V5+ or V4+, is the formation at temperatures above or near 300 K of V−V homopolar bonds across shared octahedral-site edges or faces.75−78 The possibility of V−V bonding for V defects on the Li site leads to the question of whether V clustering may occur in LFVP. Any model of V clustering must retain local charge neutrality and account for evidence of V displacements within the octahedral sites that decrease, on average, with larger V concentrations as is suggested by the decrease in XANES pre-edge peak intensity with increasing doping levels. Although it is tempting to speculate that V−V bonding may stabilize clusters of vanadium, our data provide no evidence that significant V−V bonds are formed. Rather, the data are more consistent with random doping of V on Fe sites, perhaps in small V−O−□−O−Fe or V−O−□−O−V clusters, in a nonequilibrium (metastable) low-temperature phase with Fe-site vacancies (□) compensating for the larger formal valence states of the V atoms. The apparent displacement of V from the center of symmetry of its octahedral site (evidenced by the XANES pre-edge feature) may be partly attributed to stronger bonding to an oxygen neighboring an Fe-site vacancy in V−O−□−O−V, Fe clusters, this displacement stabilizing V4+ in the presence of Fe2+, which is generally not stable as demonstrated by FeVO3.79 The structural flexibility provided by the Fe vacancies may therefore allow the olivine structure to accommodate a small fraction of V4+, and, additionally, allow the majority of V3+O6 octahedra to become more regular than the corresponding V4+O6 octahedra. This could account for the intensity decrease in the XANES pre-edge feature with increasing V doping. The incorporation of Fe-site vacancies appears to be critical to the improved kinetics and cyclability (evident through cyclic voltammetry and cycle data in our previous publication)40 of a metastable V-doped LiFePO4 phase. Facile Li+ transfer to the Fe-site vacancies may retain Li+ mobility even in the presence of some V or Fe on Li sites that would block Li+ motion, normally restricted to 1D channels along the crystallographic baxis. There are six iron sites near each Li site with distances ranging from ∼3.3 to 3.7 Å. This distance is only slightly longer than the ∼3.0 Å distance between Li sites along the [010] direction but significantly closer than the ∼4.7 Å distance between Li sites along the [001] direction, making the vacant Fe sites a feasible route for Li ion migration. Moreover, improved electron conduction by Fe2+ + V4+ = Fe3+ + V3+ electron transfer must require only a small activation energy given the observation73 of Fe3+ + V3+ in FeVO3.

Article

CONCLUSIONS Although we originally reported ‘LiFe1−x(VO)xPO4’ by a lowtemperature, microwave-assisted solvothermal process, we have shown here that the resulting phases are consistent with LiFe1−3x/2Vx□x/2PO4 (x ≤ 0.2). The oxidation state of V was found consistently by XANES to be ∼ V3.2+ for different doping levels. Rietveld refinements of neutron and X-ray diffraction data show clear evidence of Fe vacancies in these materials, with the XRD data confirming that at least 20% V can be doped into the Fe site. There is some evidence to support the presence of a small fraction of vacancies on the Li site, or possibly V doping, for higher doping levels. This finding is in sharp contrast to other studies that employed a conventional high-temperature synthetic approach and demonstrated a maximum V doping of 10% without the formation of impurity phases. By providing an additional pathway for ionic conduction, the vacancies induced by doping are expected to provide an enhancement in the kinetics of Li-ion conduction. The stability of the samples has been examined by heating the pristine microwave-synthesized 15% V-doped LiFePO4 sample in 100% Ar or 5% H2−95% Ar at various temperatures and times. Heating the samples led to the formation of significant levels of Li3V2(PO4)3 impurities at 725 °C and potentially a very small amount of impurity at 625 °C; the samples heated at 525 °C exhibited no impurity in the XRD data, but subtle changes were seen in the electrochemical behavior and FTIR spectra. STEM/EDS data showed no evidence of Li3V2(PO4)3 for the MW-ST sample heated at 525 °C or the pristine sample. Contrastingly, attempts to synthesize 15% V-doped LiFePO4 samples by a conventional method resulted in significant impurities at all temperatures. Therefore, the pristine, microwave-synthesized V-doped LiFePO4 samples with >10% doping are metastable phases that are only accessible at low temperatures and exhibit moderate stability up to 625 °C.

■

ASSOCIATED CONTENT

* Supporting Information S

More synthesis details, XANES, X-ray and neutron diffraction, STEM, EELS, FTIR, and electrochemical data are available. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: (512) 471-1791. Fax: 512-471-7681. E-mail: manth@ austin.utexas.edu. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Research was supported by the U.S. Department of Energy, Basic Energy Sciences, Materials Sciences and Engineering Division, at the University of Texas at Austin (under Contract DE-SC0005397) and at Oak Ridge National Laboratory. We thank Clarina de la Cruz for assistance with collection of neutron powder diffraction data at the High Flux Isotope Reactor (HFIR), which is sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S. Department of Energy. Use of Raman instrumentation at the Argonne National Laboratory Center for Nanoscale Materials was supported by the Office of Basic Energy Sciences, U.S. 779



Article

(30) Zhang, L.-L.; Liang, G.; Ignatov, A.; Croft, M. C.; Xiong, X.-Q.; Hung, I.-M. H.; Huang, Y.-H.; Hu, X.-L.; Zhang, W.-X.; Peng, Y.-L. J. Phys. Chem. C 2011, 115, 13520. (31) Wen, Y.; Zenga, L.; Tong, Z.; Nong, L.; Wei, W. J. Alloys Comp. 2006, 416, 206. (32) Hong, J.; Wang, C. S.; Chen, X.; Upretim, S.; Whittingham, M. S. Electrochem. Solid State Lett. 2009, 12 (2), A33. (33) Ma, J.; Li, B.; Du, H.; Xu, C.; Kanga, F. J. Electrochem. Soc. 2011, 158 (1), A26. (34) Yang, G.; Jiang, C.; He, X.; Ying, J.; Cai, F. Ionics 2012, 18, 59. (35) Hua, N.; Wanga, C.; Kang, X.; Wumair, T.; Han, Y. J. Alloys Compd. 2010, 503, 204. (36) Chiang, C.-Y.; Su, H.-C.; Wu, P.-J.; Liu, H.-J.; Hu, C.-W.; Sharma, N.; Peterson, V. K.; Hsieh, H.-W.; Lin, Y.-F.; Chou, W.-C.; Lee, C.-H.; Lee, J.-F.; Shew, B.-Y. J. Phys. Chem. C 2012, 116, 24424. (37) Lin, H.; Wen, Y.; Zhang, C.; Zhang, L.; Huang, Y.; Shan, B.; Chen, R. Solid State Commun. 2012, 152, 999. (38) Jin, Y.; Yanga, C. P.; Ruia, X. H.; Chengb, T.; Chena, C. H. J. Power Sources 2011, 196, 5623. (39) Zhao, T.; Xu, W.; Ye, Q.; Cheng, J.; Zhao, H.; Wu, Z.; Xiac, D.; Chu, W. J. Synchron Radiat. 2011, 17, 584. (40) Harrison, K. L.; Manthiram, A. Inorg. Chem. 2011, 50, 3613. (41) Ren, M. M.; Zhou, Z.; Su, L. W.; Gao, X. P. J. Power Sources 2009, 189, 786. (42) Rietveld, H. M. J. Appl. Crystallogr. 1969, 2, 65. (43) Larson, A. C.; Von Dreele, R. B. GSAS, General Structure Analysis System; Los Alamos National Laboratory: Los Alamos, NM, 1994. (44) Toby, B. H. J. Appl. Cryst 2001, 34, 210. (45) Rodriguez-Carvaja, J. Physica 1993, 192B, 55. (46) Roisnel, T.; Rodriguez-Carvajal, J. Eur. Powder Diffr. 2001, 378− 3, 118. (47) Ravel, B.; Newville, M. J. Synchrotron Radiat. 2005, 12, 537. (48) Newville, M. J. Synchrotron Radiat. 2001, 8, 322. (49) Baran, E. J.; Vassallo, M. B. J. Raman Spectrosc. 1994, 25, 203. (50) Lii, K. H.; Li, C. H.; Cheng, C. Y.; Wang, S. L. J. Solid State Chem. 1991, 95 (2), 352. (51) Silversmit, G.; Depla, D.; Poelman, H.; Marin, G. B.; De Gryse, R. J. Electron Spectrosc. Relat. Phenom. 2004, 135, 167. (52) Liu, H.; Cheng, C.; Huang, X.; Li, J. Electrochim. Acta 2010, 55, 8461. (53) Huang, J. S.; Yang, L.; Liu, K. Y. Mater. Chem. Phys. 2011, 128 (3), 470. (54) Ren, M.; Zhou, Z.; Li, Y.; Gao, X. P.; Yan, J. J. Power Sources 2006, 162, 1357. (55) Allen, C. J.; Jia, Q.; Chinnasamy, C. N.; Mukerjee, S.; Abraham, K. M. J. Electrochem. Soc. 2011, 158 (12), A1250. (56) Kuo, H. T.; Bagkar, N. C.; Liu, R. S.; Shen, C. H.; Shy, D. S.; Xing, X. K.; Lee, J.-F.; Chen, J. M. J. Phys. Chem. B 2008, 112, 11250. (57) Wong, J.; Lytle, F. W.; Messmer, R. P.; Maylotte, D. H. Phys. Rev. B 1984, 30 (10), 5596. (58) Giorgetti, M.; Passerini, S.; Smyrl, W. H.; Mukerjee, S.; Yang, X. Q.; McBreen, J. J. Electrohchem. Soc. 1999, 146 (7), 2387. (59) Chi, M.; Mizoguchi, T.; Martin, L. W.; Bradley, J. P.; Ikeno, H.; Ramesh, R.; Tanaka, I.; Browning, N. J. Appl. Phys. 2011, 110, 0461041. (60) Yang, M.-R.; Ke, W.-h.; Wu, S.-h. J. Power Sources 2007, 165, 646. (61) Bridges, C. A.; Harrison, K. H.; Unocic, R.; Idrobo, J.-C.; Paranthaman, M. P.; Manthiram, A. J. Electrochem. Soc. 2013, submitted. (62) Jugovic, D.; Cvjeticanin, N.; Mitric, M.; Mentus, S. Mater. Sci. Forum 2007, 555, 225. (63) Arnold, G.; Garche, J.; Hemmer, R.; Strobele, S.; Vogler, C.; Wohlfahrt-Mehrens, M. J. Power Sources 2003, 119, 247. (64) Koleva, V.; Zhecheva, E.; Stoyanova, R. J. Alloys Compd. 2009, 476, 950. (65) Yang, Y.; Fang, H.; Zheng, J.; Li, L.; Li, G.; Yan, G. Solid State Sci. 2008, 10, 1292.

Department of Energy. The work at the Argonne National Laboratory was performed under Contract DE-AC0206CH11357 between UChicago Argonne, LLC, and the USDOE. MRCAT operations are supported by the Department of Energy and the MRCAT member institutions. J.C.I. acknowledges support from Oak Ridge National Laboratory’s Shared Research Equipment (ShaRE) User Facility, which is sponsored by the Office of Basic Energy Sciences, U.S. Department of Energy.

■

REFERENCES

(1) Manthiram, A.; Goodenough, J. B. J. Solid State Chem. 1987, 71, 349. (2) Manthiram, A.; Goodenough, J. B. J. Power Sources 1989, 26, 403. (3) Padhi, A. K.; Nanjundasawamy, K. S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144 (4), 1188. (4) Yamada, A.; Hosoyo, M.; Chung, S.-C.; Kudo, Y.; Hinokuma, K.; Liu, K.-Y.; Nishi, Y. J. Power Sources 2003, 119−121, 232. (5) Li, Z.; Zhang, D.; Yang, F. J. J. Mater. Sci. 2009, 44, 2435. (6) Jugovic, D.; Uskokovic, D. J. Power Sources 2009, 190, 538. (7) Ellis, B.; Herle, P. S.; Rho, Y.-H.; Nazar, L. F.; Dunlap, R.; Perry, L. K.; Ryan, D. H. Faraday Discuss. 2007, 134, 119. (8) Ellis, B.; Perry, L. K.; Ryan, D. H.; Nazar, L. F. J. Am. Chem. Soc. 2006, 128 (35), 11416. (9) Zaghib, K.; Mauger, A.; Goodenough, J. B.; Gendron, F.; Julien, C. M. Chem. Mater. 2007, 19, 2740. (10) Zhou, F. Phys. Rev. B 2004, 69, 201101-1. (11) Yamada, A.; Chung, S. C.; Hinokuma, K. J. Electrochem. Soc. 2001, 148 (3), A224. (12) Ellis, B.; Kan, W. H.; Makahnouk, W. R. M.; Nazar, L. F. J. Mater. Chem. 2007, 17, 3248. (13) Wang, Y.; Wang, J.; Yang, J.; Nuli, Y. Adv. Funct. Mater. 2006, 16, 2135. (14) Murugan, A. V.; Muraliganth, T.; Manthiram, A. J. Electrochem. Soc. 2009, 152 (2), A79. (15) Muraliganth, T.; Murugan, A. V.; Manthiram, A. J. J. Mater. Chem. 2008, 18, 5661. (16) Murugan, A. V.; Muraliganth, T.; Manthiram, A. J. Phys. Chem. C 2008, 112 (37), 14665. (17) Murugan, A. V.; Muraliganth, T.; Manthiram, A. Electrochem. Commun. 2008, 10, 903. (18) Murugan, A. V.; Muraliganth, T.; Ferreira, P. J.; Manthiram, A. Inorg. Chem. 2009, 48 (3), 946. (19) Yoon, S.; Liao, C.; Sun, X. G.; Bridges, C. A.; Unocic, R. R.; Nanda, J.; Dai, S.; Paranthaman, M. P. J. Mater. Chem. 2012, 22, 4611. (20) Meethong, N.; Kao, Y.-H.; Tang, M.; Huang, H.-Y.; Carter, W. C.; Chiang, Y.-M. Chem. Mater. 2008, 20 (19), 6189. (21) Chung, S.-Y.; Bloking, J. T.; Chiang, Y.-M. Nat. Mater. 2002, 1, 123. (22) Herle, P. S.; Ellis, B.; Coombs, N.; Nazar, L. F. Nat. Mater. 2004, 3, 147. (23) Ravet, N.; Abouimrane, A.; Armand, M. Nat. Mater. 2003, 2, 702. (24) Islam, M. S.; Driscoll, D. J.; Fisher, C. A. J.; Slater, P. R. Chem. Mater. 2005, 17, 5085. (25) Wagemaker, M.; Ellis, B. L.; Lutzenkirchen-Hecht, D.; Mulder, F. M.; Nazar, L. F. Chem. Mater. 2008, 20, 6313. (26) Meethong, N.; Kao, Y.-H.; Speakman, S. A.; Chiang, Y.-M. Adv. Funct. Mater. 2009, 19, 1060. (27) Sun, C. S.; Zhou, Z.; Xu, Z. G.; Wang, D. G.; Wei, J. P.; Bian, X. K.; Yan, J. J. Power Sources 2009, 193 (2), 841. (28) Omenya, F.; Chernova, N. A.; Upreti, S.; Zavalij, P. Y.; Nam, K.W.; Yang, X.-Q.; Whittingham, M. S. Chem. Mater. 2011, 23 (21), 4733. (29) Chen, X. J.; Cao, G. S.; Zhao, X. B.; Tu, J. P.; Zhu, T. J. J. Alloys Compd. 2008, 463, 385. 780



Article

(66) Wang, X.; Liu, L.; Jacobson, A. J. J. Am. Chem. Soc. 2002, 124, 7812. (67) Sauvage, F.; Quarez, E.; Tarascon, J.-M.; Baudrin, E. Solid State Sci. 2006, 8, 1215. (68) De, S.; Dey, A.; De, S. K. J. Phys. Chem. Solids 2007, 68, 66. (69) Wang, L.; Yang, L.; Gong, L.; Jiang, X.; Yuan, K.; Hu, Z. Electrochim. Acta 2011, 56, 6906. (70) Salah, A. A.; Jozwiak, P.; Zaghib, K.; Garbarczyk, J.; Gendron, F.; Maugerd, A.; Julien, C. M. Spectrochim. Acta, Part A 2006, 65, 1007. (71) Burba, C. M.; Frech, R. J. Electrochem. Soc. 2004, 151 (7), A1032. (72) Fu, P.; Zhao, Y.; Dong, Y.; An, X.; Shen, G. Electrochim. Acta 2006, 52, 1003. (73) Yang, G.; Liu, H.; Ji, H.; Chen, Z.; Jiang, X. Electrochim. Acta 2010, 55, 2951. (74) Rui, X. H.; Yesibolati, N.; Chen, C. H. J. Power Sources 2011, 196, 2279. (75) Goodenough, J. B. Annu. Rev. Mater. Sci. 1971, 1, 101. (76) Goodenough, J. B.; Dutta, G.; Manthiram, A. Phys. Rev. B 1991, 43, 10170. (77) Manthiram, A.; Goodenough, J. B. Canadian J. Phys. 1987, 65, 87. (78) Goodenough, J. B. J. Solid State Chem. 2011, 15, 285. (79) Cox, D. E.; Taxel, W. J.; Miller, R. C.; Shirane, G. J. Phys. Chem. Solids 1962, 23, 863.

781


Temperature Dependence of Aliovalent-Vanadium Doping in LiFePO4

Recommend Documents