Local structure of vanadium in doped LiFePO₄

LiFePO₄ composites with 5 at.% vanadium doping are prepared by solid state reactions. X-ray absorption fine-structure spectroscopy is used as a novel technique to identify vanadium sites. Both experimental analyses and theoretical simulations show that vanadium does not enter into the LiFePO₄ crystal lattice. When the vanadium concentration is lower then 1 at.%, the dopant remains insoluble. Thus, a single-phase vanadium-doped LiFePO₄ cannot be formed and the improved electrochemical properties of vanadium-doped LiFePO₄ previously reported cannot be associated with crystal structure changes of the LiFePO₄ via vanadium doping.

1. Introduction

As a type of energy storage device characterized by the advantage of low maintenance costs, negligible self-discharge, high energy density and no memory effects, lithium-ion batteries are becoming more and more popular in technological applications from wireless communications to mobile computing. The key issue for the development of lithium-ion batteries is the electrode materials. Although the commercial use of LiCoO₂ has already gained great success, further large-scale applications of lithium-ion batteries in the future (e.g. in hybrid electric vehicles) demand for a cheaper cathode material. Lithium iron phosphate (LiFePO₄), proposed by Padhi et al. (1997), is one of the most promising candidates not only because of its potential low cost but also for its environmental benignancy and structural stability. However, in the almost perfect framework its bottleneck is represented by the low intrinsic electrical conductivity (Jugovic & Uskokovic, 2009). Several approaches have been explored to overcome the conductivity problem. These approaches include the reduction of the particle size, the achievement of a homogeneous particle size distribution (Gaberscek et al., 2007), particles coated with carbon (Franger et al., 2006; Song et al., 2007), and doping with supervalent cations to enhance the intrinsic conductivity (Chung et al., 2002; Ravet et al., 2003; Wagemaker et al., 2008; Meethong et al., 2009).

Recently, a composite of LiFePO₄ with Li₃V₂(PO₄)₃ addition was prepared (Wang et al., 2008; Choi & Kumta, 2007), and both electrical conductivity and high-rate discharge capacity were improved. Later, both Sun et al. (2009) and Hong et al. (2009) reported an enhancement of electrochemical performances of the single-phase vanadium-doped LiFePO₄. However, structural evidence for the vanadium substitution was still incomplete. X-ray diffraction (XRD) analysis is not sufficient for characterizing these materials because this technique is sensitive to crystalline systems and amorphous impurities such as Fe₂O₃ in LiFePO₄ are not detectable (Liu et al., 2009). In addition, XRD patterns of complex compounds show several intense peaks that may easily cover the contribution of impurities, particularly when their content is low. This is probably the inconsistency present in the contribution of Sun et al. (2009) where doped vanadium atoms replace iron atoms and in the contribution of Hong et al. (2009) where vanadium atoms occupy phosphor sites.

Synchrotron-radiation-based X-ray absorption fine-structure (XAFS) spectroscopy is a powerful spectroscopic tool that provides detailed information regarding the local geometry and the electronic structure of the investigated materials. Previous XAFS investigations have been performed on lithium-ion batteries (Wu et al., 1996; Deb et al., 2004; Kobayashi et al., 2004). Owing to the elemental specificity and detection sensitivity, XAFS is a unique tool for characterizing the aliovalent doping in LiFePO₄. In this contribution we apply both X-ray absorption near-edge spectroscopy (XANES) and extended X-ray absorption fine-structure (EXAFS) spectroscopy to identify whether vanadium atoms get inside the crystal lattice of LiFePO₄, and furthermore to investigate what is the existing form of vanadium in the doped LiFePO₄.

2. Experimental section

2.1. Materials and experimental conditions

Vanadium-doped LiFePO₄ samples were synthesized by solid state reactions. The raw materials, Li₂CO₃, FeC₂O₄·2H₂O, NH₄H₂PO₄ and NH₄VO₃, were thoroughly mixed together in stoichiometric ratios for the target compounds (e.g. Li:Fe:P:V = 1.00:1.00:0.95:0.05 for LiFeP_0.95V_0.05O₄) and then ball-milled for 3 h in acetone. After being dried, the mixture was heated at 623 K for 3 h to be pre-decomposed, and subsequently sintered at 973 K for 9 h under a flowing gas mixture (5% H₂ in Ar). Crystalline phases of the obtained materials were analyzed by XRD using a D8 Advance X-ray diffractometer working with Cu Kα radiation over an angular 2θ range from 10 to 90°. V₂O₃ and VO₂ standard samples were bought from Fisher Scientific (USA), V₂O₅ from Tianjin Kermel Chemical Reagent company (China), while Li₃V₂(PO₄)₃ was prepared by methods described elsewhere (Rui et al., 2009).

2.2. XAFS measurements

2.2.1. Fe K-edge XAFS measurements

X-ray absorption spectra at the Fe K-edge were collected in transmission mode at the U7C beamline of the National Synchrotron Radiation Laboratory (NSRL) from 200 eV below up to 800 eV above the Fe K absorption edge (7112 eV). The storage ring was working at 800 MeV and experiments were performed at room temperature with an electron current decreasing from 200 to 100 mA during a time span of about 12 h. Si (111) crystals were used to monochromatize the X-rays and to reduce higher harmonics; the second crystal was detuned by about 30%. The intensities of the incident and the transmitted X-rays were monitored by ionization chambers filled with nitrogen and helium gas mixtures.

2.2.2. V K-edge XAFS measurements

X-ray absorption spectra at the V K-edge were obtained at the X-ray absorption station of the 1W1B beamline at the Beijing Synchrotron Radiation Facility (BSRF) in the same relative energy range as the Fe K-edge. Standard samples of V₂O₃, VO₂, V₂O₅ and Li₃V₂(PO₄)₃ were recorded in transmission mode while the prepared doped samples were measured in fluorescence mode. The typical energy of the storage ring was 2.5 GeV with the current decreasing from 250 to 150 mA during a time span of about 8 h. Also a Si (111) double-crystal monochromator was used with a detuning of about 50% at the V K-edge.

3. Results and discussion

3.1. Crystalline phases of LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄

For both Fe and P sites, 5 at.% vanadium atoms were considered in LiFePO₄. XRD patterns of doped LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄ compounds together with that of the undoped LiFePO₄ are compared in Fig. 1. For the pure LiFePO₄ sample all diffraction peaks can be indexed assuming an orthorhombic structure and space group Pnma. For LiFe_0.95V_0.05PO₄, as also found by Sun et al. (2009), XRD patterns do not show any impurity phase peaks. On the contrary, in the spectrum of LiFeP_0.95V_0.05O₄, a clear peak from an impurity phase appears at about 45° (see the dashed rectangle in Fig. 1). This impurity phase cannot be identified owing to the high intensity of the main peaks.

Figure 1 X-ray diffraction patterns of LiFeP0.95V0.05O4, LiFe0.95V0.05PO4 and pure LiFePO4 samples.

3.2. XANES and EXAFS analysis of LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄

Fe K-edge XANES data for LiFeP_0.95V_0.05O₄, LiFe_0.95V_0.05PO₄ and pure LiFePO₄ are compared in Fig. 2. Within the tolerance of the signal-to-noise ratio, the spectra of doped LiFe_0.95V_0.05PO₄ and pure LiFePO₄ appear identical, indicating a similar local structure around the Fe sites. On the contrary, visible differences exist in the main peak and in the pre-peak between the data of LiFeP_0.95V_0.05O₄ and pure LiFePO₄. Considering that the XRD patterns of LiFeP_0.95V_0.05O₄ show the presence of an impurity phase, we may claim that in the case of LiFeP_0.95V_0.05O₄ vanadium atoms do not enter inside the crystal lattice of LiFePO₄ leading to a relative excess of iron, while forming another iron compound compatible with the differences observed in the XANES spectra at the Fe K-edge.

Figure 2 Comparison of normalized Fe K-edge XANES spectra of LiFeP0.95V0.05O4, LiFe0.95V0.05PO4 and pure LiFePO4 samples.

Direct information regarding vanadium in LiFeP_0.95V_0.05O₄ and in LiFe_0.95V_0.05PO₄ is given by the V K-edge XANES data shown in Fig. 3. XANES structures originating from multiple-scattering processes of the photoelectrons with neighbouring atoms return details about the electronic structure and the local arrangement of atoms around the V absorbing atom. Looking at Fig. 3, the XANES of V₂O₃, VO₂, V₂O₅ and Li₃V₂(PO₄)₃ show clear differences owing to their different chemical constitutions, and the XANES of the well known chemical constituents provide useful standards for recognizing unknown chemical forms such as V atoms in the doped LiFePO₄. In the literature, the vanadium oxidation state is determined through the position of the absorption near edge, which shifts to higher energies with increasing valence state. In our study the absorption edge increases from about 5454.5 eV to 5458.2 eV as the valence state of vanadium shifts from +3 in V₂O₃ to +5 in V₂O₅. Because the absorption edge is at 5454.3 eV, the valence state of vanadium in LiFe_0.95V_0.05PO₄ may be assigned to +3, while vanadium in LiFeP_0.95V_0.05O₄ has a lower valence state for the absorption edge at 5453.5 eV. The XANES features of LiFe_0.95V_0.05PO₄ and LiFeP_0.95V_0.05O₄ also show large differences. As a consequence, different stoichiometric ratios largely affect the existing form of vanadium. The pre-edge peak corresponding to the 1s → 3d transition is a clear fingerprint of the symmetry and can be used to evaluate qualitatively changes of the vanadium local symmetry. Symmetrical vanadium–ligand coordinations with inversion symmetry, such as a regular `VO<sub>6</sub>' octahedral environment in the VO compound, have negligible pre-edge intensities (Silversmit et al., 2006). In contrast, coordinations without inversion symmetry, such as distorted `VO₆' units in V₂O₃ and in VO₂ or the `VO₄' tetrahedral coordinations in NH₄VO₃, show significant to large pre-edge intensities (Silversmit et al., 2005; Wu & Xian, 2001). In the case of LiFeP_0.95V_0.05O₄, if vanadium atoms are located at P sites a large pre-edge intensity may appear because phosphorus atoms share a tetrahedron geometry with the surrounding oxygen atoms in the LiFePO₄ crystal. However, XANES data of LiFeP_0.95V_0.05O₄ in Fig. 3 show only a small pre-edge peak, not in agreement with the vanadium substitution at P sites.

Figure 3 V K-edge XANES spectra of LiFeP0.95V0.05O4 and LiFe0.95V0.05PO4 samples and of reference materials such as Li3V2(PO4)3, V2O3, VO2 and V2O5.

To extract detailed structural information around the absorbing vanadium atoms, XANES simulations were performed in the framework of the multiple-scattering theory (Lee & Pendry, 1975; Natoli et al., 1990) using the FEFF8.2 code (Ankudinov et al., 1998). V K-edge XANES calculations were performed with the structural model and lattice parameters of the host structure, i.e. olivine LiFePO₄ (Garcia-Moreno et al., 2001), in which a V atom replaces P and Fe in LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄, respectively. The FEFF input file was generated by the ATOMS package and the cluster potential was approximated by a set of spherically averaged muffin-tin potentials. For the calculations a model cluster including all atoms within 5 Å of the central V atom was constructed for the self-consistent field potential calculations, and a cluster with a radius of 7 Å was used for full multiple-scattering calculations. Calculation results and experimental data are compared in Fig. 4. Clear discrepancies can be recognized in the relative peak intensities and peak positions for both LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄. Data suggest that vanadium atoms do not enter the selected sites in both LiFeP_0.95V_0.05O₄ and LiFe_0.95V_0.05PO₄.

Figure 4 Comparison of theoretical calculations (CAL) and experimental spectra (EXP) of LiFeP0.95V0.05O4 (a) and LiFe0.95V0.05PO4 (b).

To compare XANES spectral features we set three vertical dashed lines in Fig. 3 as guides for the eye. From the comparison the spectral profile of LiFe_0.95V_0.05PO₄ appears quite similar to that of Li₃V₂(PO₄)₃, a behaviour consistent with the presence of the vanadium inside LiFe_0.95V_0.05PO₄ and in agreement with EXAFS data. Looking at Fig. 5, the EXAFS signal of LiFe_0.95V_0.05PO₄ shows great differences with that of the pure LiFePO₄, whereas a good correspondence occurs between the structures of LiFe_0.95V_0.05PO₄ and Li₃V₂(PO₄)₃. In the case of LiFeP_0.95V_0.05O₄ the presence of vanadium cannot be recognized by a qualitative comparison of XAFS data via a linear combination of reference spectra owing to the lack of reference systems.

Figure 5 (a) k^2-extracted EXAFS signals in the range 2–10 Å^−1 at the Fe K-edge for pure LiFePO4 and at the V K-edge for LiFe0.95V0.05PO4, LiFeP0.95V0.05O4 and Li3V2(PO4)3 samples; (b) the corresponding radial structure functions in r-space.

3.3. XANES and EXAFS analysis of LiFeP_0.99V_0.01O₄ and LiFe_0.99V_0.01PO₄

The insolubility of 5 at.% vanadium doped in LiFePO₄ at both Fe and P sites may be due to the high doping level. Studies were then focused on samples with a lower vanadium concentration such as LiFeP_0.99V_0.01O₄ and LiFe_0.99V_0.01PO₄ samples. XRD patterns of these samples display a single LiFePO₄ phase (Fig. 6) and may barely provide any useful information about the local environment of vanadium. Moreover, XANES data in Fig. 7 show that the spectrum of LiFeP_0.99V_0.01O₄ as well as that of LiFe_0.99V_0.01PO₄ has a weak pre-edge peak that suggests a `VO₆' octahedral environment. They show a great similarity with Li₃V₂(PO₄)₃. The data show that in these two low-level doped samples vanadium forms a Li₃V₂(PO₄)₃ phase, a hypothesis supported by the analogy of the EXAFS spectra shown in Fig. 8.

Figure 6 X-ray diffraction patterns of LiFeP0.99V0.01O4, LiFe0.99V0.01PO4 and pure LiFePO4 samples.

Figure 7 V K-edge XANES spectra of LiFeP0.99V0.01O4, LiFe0.99V0.01PO4 and Li3V2(PO4)3 samples.

Figure 8 Comparison of k^2-extracted V K-edge EXAFS signals in k-space (a) and r-space (b) for LiFeP0.99V0.01O4, LiFe0.99V0.01PO4 and Li3V2(PO4)3 samples.

4. Conclusions

In this study, LiFePO₄ samples were synthesized via solid state reactions to achieve vanadium doping at both Fe and P sites. The solubility of vanadium was investigated by the XAFS technique. XANES theoretical simulations at the V K-edge show large deviations from experimental spectra, providing reliable evidence of the insolubility of vanadium in LiFePO₄. XANES and EXAFS experiments confirm that vanadium in doped samples of Fe sites is consistent with the Li₃V₂(PO₄)₃ composition. On the contrary, that in doped samples of P sites appears to show filling a VO₆ octahedral environment although a definite chemical configuration has not been determined. Also, at lower dopant concentrations, for example, at 1 at.%, vanadium does not select Fe and P sites in the crystal lattice of LiFePO₄ while forming a Li₃V₂(PO₄)₃ phase. Actually, this study provides a new reliable method of investigating doped LiFePO₄ materials. Structural information associated with a dopant barely distinguished in an XRD pattern may be revealed by the XAFS analysis. However, further work is necessary to build up a clear framework regarding the doping of transition metal elements in LiFePO₄. The results achieved also point out that the method can be applied to investigate other doped electrode systems.

The above analysis clearly shows that enhanced electrochemical performances of the LiFePO₄ composite following vanadium addition cannot be due to a crystal structural change of the LiFePO₄. Alternatively, mechanisms suggested by Wang et al. (2008) and Choi & Kumta (2007) should be considered. In addition, the deviation from the stoichiometry of LiFePO₄ that may induce the presence of secondary phases such as Fe₂P/Fe₃P may also contribute to the observed enhanced conductivity (Subramanya Herle et al., 2004; Kang & Ceder, 2009). An accurate investigation with other technical means is under way to give the exact mechanism.