Changes in Electronic Structure upon Li Insertion Reaction of

The electrochemical lithium insertion reaction of monoclinic Li3Fe2(PO4)3 .... Feng Yu , Lili Zhang , Mingyuan Zhu , Yongxin An , Lili Xia , Xugen Wan...
0 downloads 0 Views 802KB Size
Download PDF
J. Phys. Chem. B 2006, 110, 17743-17750

17743

Changes in Electronic Structure upon Li Insertion Reaction of Monoclinic Li3Fe2(PO4)3 Junichi Shirakawa,† Masanobu Nakayama,† Masataka Wakihara,*,† and Yoshiharu Uchimoto‡ Department of Applied Chemistry, Tokyo Institute of Technology, 2-12-1, Ookayama Megro-ku, Tokyo 152-8552, Japan, and Department of Interdisciplinary EnVironment, Graduate School of Human and EnVironmental Studies, Kyoto UniVersity, Yoshida-nihonmatsu-cho, Sakyo-ku, Kyoto 606-8501, Japan ReceiVed: April 10, 2006; In Final Form: July 14, 2006

The electrochemical lithium insertion reaction of monoclinic Li3Fe2(PO4)3 as cathode materials of lithiumion batteries was investigated from the viewpoint of the electronic structure around Fe and the polyanion unit (PO4). Fe K-edge and LIII,II-edge XAS measurements revealed that Fe3+ was reduced to Fe2+ upon Li insertion. In addition, O K-edge and P K-edge XAS also showed spectral changes upon Li insertion, which corresponded to changes in the electronic structure of the PO4 polyanion unit. The ab initio density functional calculation was performed within the GGA and LDA+U methods. The LDA+U method reproduced well the cell potential upon lithium intercalation into Li3Fe2(PO4)3, whereas the GGA method underestimated the intercalation. The calculated electronic structure of Li3Fe2(PO4)3 described strong P 3p-O 2p covalent bonding, while weak hybridization was indicated in Fe 3d-O 2p. Moreover, the difference in electronic density between Li3Fe2(PO4)3 and the lithiated model indicated that the polarization effect between inserted Li and oxygen induced the changes in the electronic structure around the polyanion unit.

Introduction Development of rechargeable lithium ion battery meets the crucial demands of our modern society, acting as the power source of portable devices, and future usage in electric vehicles, and so on. The higher energy, higher density, lower cost, and environmentally friendly electrode materials are urgently requested today. Among several transition-metal compounds, iron compounds are the most desirable cathode materials from both their economical and environmental viewpoint. Goodenough et al. reported that several iron compounds with NASICON-type or NASICON-like structure, Fe2(XO4)3 (X ) S, Mo, W) and Li3Fe2(PO4)3, showed much higher redox potential compared with that of a normal oxide such as Fe2O3.1-3 For example, olivine-related-type LiFePO4, which has been investigated as one of the most promising candidates for practical usage, also shows higher voltage.4-6 These materials were classified as “polyanion” compounds containing XO4 (X ) S, Mo, P, etc.) compact tetrahedral units. It is known that the conduction band of these polyanion units is placed far from the Fermi level where electronic exchange mainly occurs because of its closed shell electronic configuration.4-8 In addition, these XO4 polyanion units form strong covalent bonding. On the other hand, the valence d band of transition metals lies around the Fermi level and is isolated from the band of polyanions. Therefore, the electronic exchange arising from Li removal/uptake mainly occurs at transition metal ions. This simple electronic description of polyanion compounds could account for the observed high voltage properties (Table 1).4-8 From this viewpoint, we previously reported the changes in the electronic structure of olivine Li1-xCoPO4 as 4.8 V cathode material using the X-ray absorption spectroscopy (XAS) technique. It was shown that the electronic structure changes around PO4 units as well as * Address correspondence to this author. E-mail: o.cc.titech.ac.jp. † Tokyo Institute of Technology. ‡ Kyoto University.

mwakihar@

TABLE 1: Various Iron Compounds as Cathode Materials for Li Ion Batteries3-8 compd

crystal structure

Fe2(SO4)3 LiFePO4 Fe2(MoO4)3 Fe2(WO4)3 Li3Fe2(PO4)3 Fe2O3

NASICON (space group R3h) Olivine (space group Pnma ) monoclinic (space group P21) monoclinic (space group P21) monoclinic (space group P2/n)

voltage vs Li+/Li 3.63,6,8 3.54,5,8 3.03,8 3.03,8 2.85,6,8 1.07

Co ion upon electrochemical reaction, although the electrons of the PO4 polyanion are generally considered not to contribute to the oxidation process.9,10 This result indicates that the electron transfer occurs via a hybridized orbital between Co 3d and O 2p accompanying a slight change of the electrons around P.9,10 Recently, an ab initio study based on density functional theory within the LDA+U approximation was investigated to understand the electrochemical behavior and magnetic structure for olivine-type LiMPO4 and MPO4 (M ) Mn, Fe, Co, Ni).11-15 In these previous reports, the LDA+U method showed good agreement with the experimental observations, whereas the LDA and GGA fail to reproduce these observations because of a strong electron correlation effect.12,13 In addition, Bacq et al. revealed in their computation of LiCoPO4 and CoPO4 that Li induced strong polarization with the PO4 covalent bond, Li removal enhances the electronic polarization around P, and the wide Co-band structure affects the (Li)PO4 band. That is to say, the PO4 polyanion as well as the Co ion change their electronic structure via the polarization effect upon the Li removal reaction. Thus, the computational study agrees with our experimental observation in XAS. In the present study, we focused on monoclinic iron compounds, Li3+xFe2(PO4)3. The change in electronic structure before and after Li insertion for monoclinic Li3+xFe2(PO4)3 was investigated by using a combination of X-ray absorption spectroscopy (XAS) and ab initio density functional theory

10.1021/jp0622379 CCC: $33.50 © 2006 American Chemical Society Published on Web 08/18/2006

17744 J. Phys. Chem. B, Vol. 110, No. 36, 2006

Figure 1. Crystal structure of Li3Fe2(PO4)3 with monoclinic structure (space group; P21/n).

(DFT) calculations in this paper. The crystal structure of monoclinic Li3Fe2(PO4)3 belongs to the space group P21/n as shown in Figure 1.5,16,17 This structure consists of a threedimensional framework of FeO6 octahedra and PO4 tetrahedra sharing oxygen vertexes and contains relatively large interstitial spaces that accept up to five Li+ ions at the maximum associated with the electrochemical reduction of Fe3+ to Fe2+.

Shirakawa et al. absolute energy scale was determined by comparison with data of the standard material Fe2O3. For the samples after electrochemical treatment, all installation operations were performed under N2 or Ar atmosphere. B. Computational Part. The ab initio calculations based on DFT within the generalized gradient approximation (GGA) and local density approximation including U parameter (LDA+U) for both Li3Fe2(PO4)3 and lithiated Li5Fe2(PO4)3 were performed with the Vienna ab initio simulation package (VASP)18 program utilized with the projector-augmented wave (PAW) method.19,20 In the case of LDA+U approximation, the effects due to the localization of the d electrons were reproduced by taking a Hubbard-like on-site coulomb parameter U into account for LDA. In this study, the U parameter for Fe 3d electrons was fixed to 3.5 eV from the literature.11 An energy cutoff of 360 eV was set, and the only k-point used was the Γ point due to the large size of the unit cell. The structural energy minimization was carried for the structure shown in Figure 1 in the case of Li3Fe2(PO4)3. The positions of insertion of the two Li into Li3Fe2(PO4)3 were decided by using the bond-valence sum, which is widely used in crystallography to evaluate the plausibility of structure models or diffusion pass of ions.21,22 The bond-valence sum V of Li+ was expressed as

V)

(1)

where individual bond-valence sLi-O values for bonds to oxide ion are calculated as follows,

Methods A. Experimental Part. Monoclinic Li3Fe2(PO4)3 was synthesized by a citric acid complex method with Fe(NO3)3‚9H2O, LiNO3, and NH4H2PO4 as starting materials. Each starting material mixed in a stoichiometric ratio was solved in pure water. And then, excessive citric acid, which is about 50% more than the stoichiometric amount of each metal (pH ∼3), was slowly added into the mixed solution with stirring on the hotplate. After evaporation of solvent, the precursor was sintered twice at 650 °C for 12 h, and the target products were obtained. The phase identification and the refinement of lattice parameters for the samples were performed by powder X-ray diffraction (XRD), using Cu KR radiation (RINT-2500V, Rigaku Co. Ltd). Electrochemical Li insertion into Li3Fe2(PO4)3 was performed with a three-electrode cell. The working electrode was prepared by a mixture of 70 wt % active materials, 25 wt % acetylene black as a current collector, and 5 wt % poly(vinylidene fluoride) (PVdF) binder. Li foil (Aldrich) was used as a counter and a reference electrode. The electrolyte was 1 M LiClO4 dissolved in ethylene carbonate/diethyl carbonate (volume ratio of 1:1) (Tomiyama Pure Chemical Industries, Co. Ltd.). The Li insertion reaction was carried out at 0.1 C, where 1 C corresponds to 64.2 mA g-1, up to reaching 2.0 V. The Fe K-edge XAS measurements were performed by using the transmission method with synchrotron radiation at the beam line BL-9C, Photon Factory (PF), High Energy Accelerator Research Organization, in Tsukuba, Japan. The absorption of Cu K-edge was used for calibration of the absolute energy scale. The XAS measurements within the soft X-ray region were carried out with synchrotron radiation at the beam lines BL8B1 for Fe L-edge and O K-edge and BL-1A for P K-edge, respectively, at UVSOR, Institute for Molecular Science in Okazaki, Japan. The absorption was determined by the totalelectron-yield method. The total yield was divided by the storage-ring current that was recorded simultaneously. The

sLi-O ∑ O

sLi-O ) exp

(

)

R0 - RLi-O b

(2)

where R0 and b are empirical bond-length and bond-valence parameters, respectively. In this study, we used empirical parameters given by R0 ) 1.17096 Å and b ) 0.516 Å for the Li-O bond.23 Averaged cell voltage (ACV) was also calculated by using both GGA and LDA+U approximations by following a wellestablished method,11,13,24

ACV )

E(Li5Fe2(PO4)3) - E(Li3Fe2(PO4)3) - nE(Li) (3) nF

where E(Li5Fe2(PO4)3), E(Li3Fe2(PO4)3), and E(Li) represent the total energy of Li5Fe2(PO4)3, Li3Fe2(PO4)3, and Li metal, n is the number of electron for the reaction, and F is the Faraday constant. Results and Discussions A. Phase Identification and Electrochemical Li Insertion. In Figure 2, the XRD pattern of synthesized Li3Fe2(PO4)3 is shown. All the Bragg peaks can be indexed to integral Miller indices with monoclinic symmetry (space group, P21/n), and no additional peaks due to impurity phase were observed. The result of refinement of lattice parameters (a ) 8.569 Å, b ) 12.0782 Å, c ) 8.609 Å, γ ) 90.27° and V/Z ) 222.7 Å3) showed good agreement with that of previous report (V/Z ) 222.0 Å3).26 The SEM image of synthesized Li3Fe2(PO4)3 is shown in Figure 3. The primary-particle size of ∼0.5 µm was obtained although some particles are condensed with each other, which was similar appearance to the sample synthesized by another solution method in a previous report.27 The discharge curve is shown in Figure 4. Observed potentials gradually decreased with lithium insertion from ∼2.75 to 2.0

Li Insertion Reaction of Monoclinic Li3Fe2(PO4)3

J. Phys. Chem. B, Vol. 110, No. 36, 2006 17745

Figure 2. XRD pattern of Li3Fe2(PO4)3: (a) experimental pattern and (b) simulated pattern (ref 25). Figure 5. (a) Fe K-edge XAS of Li3+xFe2(PO4)3 with references, Fe2O3 and FeO, and (b) variation of absorption energy as a function of composition x.

Figure 3. SEM image of Li3Fe2(PO4)3 synthesized by citric complex method.

Figure 6. Fe LIII,II-edge XAS of Li3+xFe2(PO4)3.

Figure 4. Discharge curve of Li3Fe2(PO4)3. Solid circles indicate the samples for XAS measurement. The discharge measurement was carried out at 0.1 C. 1 C corresponds to 64.2 mA g-1.

V accompanying solid solution reaction. Its discharge capacity was about 126 mAh g-1, which was close to the theoretical capacity (128.4 mAh g-1) where all the Fe3+ ions reduce to Fe2+, leading to Li5Fe2(PO4)3. In a previous report of electrochemical measurement for Li3Fe2(PO4)3 synthesized by the solid-state reaction or solution method, it was revealed that smaller particle sizes led to superior electrochemical behavior such as larger capacity.27 In addition, it was reported that the sample ball-milled for longer time showed larger capacity because of a decrease of particle size.27 In this study, it was revealed that an appropriate particle-size sample was synthesized by the solution method with use of citric acid, since discharge capacity corresponded to the theoretical one.

B. XAS Measurements. In Figure 5a, the Fe K-edge XAS spectra of electrochemically prepared Li3+xFe2(PO4)3 with various inserted Li amounts (x) were shown. Since the energy of the absorption edge for the sample before Li insertion was close to that of Fe2O3, Fe ions exist as Fe3+ in Li3Fe2(PO4)3. As shown in Figure 5b, the edge energy was linearly decreased with an increase of Li insertion. The absorption energy of fully lithiated Li5Fe2(PO4)3 (x ) 2) was consistent with that of FeO. Thus, the oxidation state of Fe reduced from +3 to +2 during the electrochemical lithium insertion. Fe LIII- and LII-edge XAS spectra of Li3+xFe2(PO4)3 are shown in Figure 6. In the spectra of Li3Fe2(PO4)3, Fe LIII- and LII-edge absorptions were located ranging from 709 to 715 eV and from 717 to 720 eV, respectively. This corresponds to dipole transitions from Fe 2p3/2 (LIII) and 2p1/2 (LII) core electrons to an unoccupied 3d state (2p63dn f 2p53dn+1), respectively. As is the case with Fe K-edge spectra, energy shift toward lower side can be observed in both LIII and LII absorption with increasing Li content, which corresponds to the effective nuclear charge of the Fe ion decreases. Both LIII- and LII-edge absorption consist of two peaks, ascribed to the t2g and eg orbitals which are split by the crystal field theory with octahedral symmetry. The intensity ratio in either LIII- or LII-edge was gradually changed with increasing Li content, which would be due to a

17746 J. Phys. Chem. B, Vol. 110, No. 36, 2006

Figure 7. (a) O K-edge and (b) P K-edge XAS of Li3+xFe2(PO4)3.

decrease of strength for the crystal field or electronic interactions caused by an increase of the number of Fe d-electrons.28 According to the literature,29 the LIII absorption edge of Fe2+ in an octahedral crystal field typically describes a strong peak at lower energy and a weaker peak or a shoulder at higher energy, whereas the tendency of peak intensity was vice versa in the case of Fe3+. Therefore, this supports that the present results in Figure 6 indicate the reduction of iron proceeded from Fe3+ to Fe2+ as Li was inserted. Then, the changes in electronic structure of the polyanion unit upon Li insertion are presented hereunder. Variation of O and P K-edge XAS spectra of Li3+xFe2(PO4)3 with various Li content are shown in Figure 7, parts a and b, respectively. In O K-edge XAS of Li3Fe2(PO4)3, preedge peaks ranging from 525 to 535 eV arising from Fe 3d/O 2p hybridized orbital were observed. Above 535 eV, this region can be assigned to the transitions to O 2p orbital hybridized with the s and/or p orbital of P and Fe ions. These preedge peaks for initial Li3Fe2(PO4)3 (x ) 0) are quite small compared with those observed in iron oxides, Fe2O3 and FeO. Thus, the hybridization between Fe 3d and O 2p is small due to the strong polarization in the polyanion as mentioned in Introduction. However, large peaks A and B (see Figure 7a) appeared after the first stage of Li uptake (Li3.5Fe2(PO4)3). Thus, the changes in O K-edge XAS could be explained by two effects as follows: (1) Changes in the interaction of Fe 3d-O 2p, or reduction of Fe3+ to Fe2+, affected the electronic orbital of oxide ions. (2) The electrostatic interaction between inserted Li and O ions causes the changes in electronic structure around oxide ions as reported in previous papers.30-32 As seen in Figure 5b, the oxidation state of Fe decreased linearly with Li uptake, so that the second effect would contribute to the change in O K-edge spectra for early stage of the Li insertion. In addition, the absorption energy shifted toward the high energy region during lithium insertion from Li4.5Fe2(PO4)3 to Li5Fe2(PO4)3. As above, the electronic structure around oxide ions was varied by Li uptake as well as iron ions, although it was thought that the charge compensation mainly took place around Fe because of strong electronic polarization of the PO4 polyanion unit. On the other hand, no marked change was observed in P K-edge spectra upon Li insertion (Figure 7b). Two peaks were observed, (1) the main absorption peak B due to transition from P 1s to P 3p and (2) a weak preedge peak A, in the spectrum of Li3Fe2(PO4)3. In previous reports, the preedge peak of S K-edge XAS for CuSO4 was observed and would arise from the interaction between the

Shirakawa et al.

Figure 8. Isosurfaces of constant Li bond valence sum mismatch with ∆BVS(Li) ) 0.5 as models for the Li inserted site.

TABLE 2: Averaged Cell Voltages (ACV) Calculated by GGA and LDA+U compd

method

ACV/V

Li3Fe2(PO4)3 Li5Fe2(PO4)3

GGA

1.975

Li3Fe2(PO4)3 Li5Fe2(PO4)3

LDA+U

2.536

Cu2+ 3d orbital and the oxygen 2p based orbital in SO4.2,33,34 Therefore, it was suggested that similar interaction between Fe3+ and PO43- caused weak preedge peak A. Upon Li insertion, preedge peak A disappeared and main absorption peak B slightly shifted toward the lower energy region. Thus, it was suggested that the electronic structure of the PO4 polyanion unit changed as well as Fe ions upon Li insertion. That is to say, disappearance of preedge peak in P K-edge XAS indicated a decrease of interaction between the Fe 3d and the O 2p based orbital in the polyanion, PO4. On the other hand, in ab initio calculation10 and XAS study of LiCoPO4,12 it was suggested that the polarization effect induced by the strong ionic character of Li led to the modification of the electronic structure of the PO4 unit as well as oxidation of Co ions. Therefore, observed changes in XAS would be ascribed to the perturbation of covalent bonding in the PO4 unit by strong electrostatic interaction arising from inserted Li ions. C. Ab Initio Calculation. The comparison of electronic structure before and after Li insertion into Li3Fe2(PO4)3 was described by using two kinds of approximation, GGA and LDA+U. In case of the lithiated model for Li5Fe2(PO4)3, the sites for inserted Li in the framework of the parent Li3Fe2(PO4)3 were estimated by the BVS method. The BVS of Li-O interaction calculated for the entire space of the unit cell (Figure 1) and isosurface with 0.5 e BVS e 1 in Li3Fe2(PO4)3 was displayed in Figure 8. There were eight sites where BVS is close to +1 in the unit cell, which corresponds to two Li sites per Li3Fe2(PO4)3 formula unit. Thus, we set these sites as the Li insertion site in Li5Fe2(PO4)3, and then structural relaxation of the lithiated model, Li5Fe2(PO4)3, was carried out. The calculated ACVs are summarized in Table 2. A remarkable difference in ACVs between GGA and LDA+U was obtained, and LDA+U led to good agreement with experimental results rather than GGA, which suggests that the introduction of the U parameter reproduced well the changes in electronic structure from Li3Fe2(PO4)3 to Li5Fe2(PO4)3.

Li Insertion Reaction of Monoclinic Li3Fe2(PO4)3

J. Phys. Chem. B, Vol. 110, No. 36, 2006 17747

Figure 9. DOS for Li3Fe2(PO4)3 calculated within (a) GGA and (b) LDA+U. The Fermi energy is set as 0 eV.

Figure 10. DOS of Li5Fe2(PO4)3 calculated within (a) GGA and (b) LDA+U. The Fermi energy is set as 0 eV.

Figure 9 represents the comparison of density of state (DOS) for Li3Fe2(PO4)3 calculated by GGA and LDA+U. The energy gap between the top of the valence band and the bottom of the conduction band for the Fe 3d orbital was formed within LDA+U (Figure 9b), whereas the energy gap was not indicated within GGA (Figure 9a). In addition, LDA+U represented a sharp band of the unoccupied Fe 3d orbital compared with that in GGA, indicating the localization of the unoccupied Fe 3d

orbital. Both approximation methods showed high spin state for Fe3+. In addition, it was clearly shown that O 2s and 2p bands interacted with P 3s and 3p ranging from -10 to -5 eV and from 7 to 10 eV corresponding to the covalent bonding for the PO4 polyanion. The O 2p valence band formed right below the Fermi level where electronic exchange occured during Li insertion and seemed to interact partly with the Fe 3d orbital. Above the Fermi level from 0 to 3 eV, a quite narrow band of

17748 J. Phys. Chem. B, Vol. 110, No. 36, 2006

Figure 11. The calculated electron density for (a) Li3Fe2(PO4)3 and (b) Li5Fe2(PO4)3 with LDA+U.

Fe 3d resided with a minor contribution of the O 2p band. Thus, the Fe 3d orbital was localized considerably compared with strong covalent bonding formation between P and O. Partial DOS for the lithiated model, Li5Fe2(PO4)3, is shown in Figure 10. Electrons partly accommodate a down spin state of Fe 3d in Li5Fe2(PO4)3, where no electrons were found in Li3Fe2(PO4)3. Therefore, this indicated that the reduction of Fe3+ took place via a down spin state of the Fe 3d band upon Li insertion. Note that no hybridization between this down spin state of Fe 3d

Shirakawa et al. and O 2p was observed around Fermi energy (from -1 to 0 eV) for Li5Fe2(PO4)3, whereas part of the contribution of O 2p could be seen in Li3Fe2(PO4)3. In other words, the change of bonding character for Fe 3d/O 2p occurred for the valence band upon Li insertion into Li3Fe2(PO4)3, which would contribute to the change in O K-edge XAS spectra. These results were similar to previous reports about ab initio calculations of LiFePO4 and FePO4.28 In Figure 11, the isosurfaces of the electron density were visualized for Li3Fe2(PO4)3 and Li5Fe2(PO4)3 within LDA+U approximation. As shown in the figures, the electrons strongly hybridized between P and O, whereas no marked crossover was observed between Li and O, which reflected the covalence bonding character of the PO4 polyanion and ionic character for the Li-O bond. In addition, weak Fe 3d-O 2p hybridization was observed for Li3Fe2(PO4)3, and this hybridization between Fe-O bonding decreased after Li insertion (see the white arrow in the figures). These tendencies were close to the difference of DOS before and after Li insertion. Parts a and b of Figure 12 represent the contour image of the difference in electron density between Li5Fe2(PO4)3 and Li3Fe2(PO4)3 for GGA and LDA+U. From Figure 12a,b, the increase in electrons was clearly observed around the Fe site, indicating Fe2+ reduced to Fe3+ by Li uptake. In addition, electrons around the oxygen site were also increased between the Fe and O bond, indicating that charge compensation took place around oxide ions as well as Fe. Parts c and d of Figure 12 describe the isosurface of difference in electron density, where yellow and blue indicate an increase and a decrease in electron density upon Li insertion, respectively. A decrease in electron density as well as an increase in it around Fe ions is shown, indicating that the rearrangement of electronic structure within the d orbital of Fe ions took place accompanying the reduction of Fe ions. Furthermore, this tendency was enhanced for the results of LDA+U rather than that of GGA. Such a rearrangement of d electrons around Fe ions was also obtained in ab initio study on LiCoPO4 and CoPO4 within LDA, GGA, and LDA+U.12

Figure 12. Difference of electron density between Li5Fe2(PO4)3 and Li3Fe2(PO4)3. Contour plot of calculated results within (a) GGA and (b) LDA+U, where red indicates a larger increase in electrons after Li insertion. Isosurfaces are also shown in (c) GGA and (d) LDA+U, where yellow and blue correspond to an increase and a decrease of electrons, respectively.

Li Insertion Reaction of Monoclinic Li3Fe2(PO4)3

J. Phys. Chem. B, Vol. 110, No. 36, 2006 17749

EI )

Figure 13. Difference in radial integration of electrons around Fe, O, and P between Li3Fe2(PO4)3 and Li5Fe2(PO4)3 calculated by (a) GGA and (b) LDA+U, respectively.

∫F(atom, r) dr

(4)

where EI and r denote the electron integration and radii from center of the atom, respectively. Then, the integrated electron’s distribution for each atom of Li5Fe2(PO4)3 was subtracted from that of Li3Fe2(PO4)3, and represented in Figure 13. The results of LDA+U clearly showed the increase of electrons around Fe after Li insertion rather than those of GGA. In addition, electron transfer was also indicated for the PO4 polyanion, especially around oxide ions. Figure 14 presents the radial integration of electrons EI for various oxygen sites (note that Figure 13 represents the averaged EI for all oxygen sites). The oxide ions can be classified into several types which depend on the number of Li and/or Fe ions neighbored. Figure 14a compares the EI around oxide ions which linked 1 Fe ion and 0, 1, and 2 Li ions. The more Li ion is neighbored, the larger the increase of electron indicated. This tendency applies for the remaining oxide ions. On the other hand, Figure 14b compares the EI around oxide ions which linked 1 Li ion and 0, 1, and 2 Fe ions. Clearly, the more Fe ion is neighbored, the larger the increase of electron is shown as is the case of Figure 14a. Accordingly, it was revealed that the changes of electron configuration for oxide ions were strongly affected by both insertion of Li and reduction of Fe3+. Probably, the strong ionic character of inserted Li would polarize neighboring oxide ions.10 O K-edge XAS shown in Figure 7a would reflect the changes in electronic structure around oxide ions due to both the change in Fe 3d-O 2p hybridized orbital and polarization induced by inserted Li. Moreover, these two factors would also contribute to the modification of P K-edge XAS through Li-O-P or Fe-O-P linkages. From the results of XAS and ab initio calculation, it was suggested that two factors, (1) charge transfer and rearrangement of d-electrons for Fe ions and (2) interaction of inserted Li and oxygen, were important to understand the change in electronic structure upon Li insertion into Li3Fe2(PO4)3. Especially, the former factor was crucially connected to the high voltage property of Li3Fe2(PO4)3, as indicated by the comparison of GGA with LDA+U. Conclusions

Figure 14. Difference in radial integration of electrons around various oxide ion sites between Li3Fe2(PO4)3 and Li5Fe2(PO4)3: (a) the oxygen that is neighbored by 1 Fe ion and 0, 1, and 2 Li ions and (b) the oxygen that is neighbored by 1 Li ion and 0, 1, and 2 Fe ions.

Hence, the amount of electronic rearrangement in the d orbital of Fe ions would be strongly connected to the high voltage for lithium insertion reaction of Li3Fe2(PO4)3 as indicated in the difference of voltage between GGA and LDA+U. This was consistent with the large changes in Fe L-edge XAS spectra for Li3+xFe2(PO4)3 (Figure 6), which could be interpreted as rearrangement of d-electrons of Fe ions. The increase in electrons around oxide ions depends on the coordination environment of the oxide ion (for example, O0 and O1 shown in Figure 12d). On the other hand, P ions hardly contribute to charge compensation as seen in the figures. To discuss these behaviors quantitatively, the number of electrons are integrated from the position of each nucleus, averaged for each species, and plotted as a radial function using following equation

In this study, the electrochemical Li insertion reaction mechanism of Li3Fe2(PO4)3 was investigated by using XAS and ab initio calculations. The sample synthesized by a citric acid complex method was quite a fine particle (∼0.5 µm), and 2 Li ions were inserted into Li3Fe2(PO4)3. From Fe K-edge and L-edge XAS, it was confirmed that almost all Fe3+ were reduced to Fe2+ and varied their electronic structure of the d orbital. The electron uptake in Fe ions was accompanied by the rearrangement of electrons among the d orbital, which was revealed by the ab initio calculation. In addition, LDA+U showed accurate cell potential and larger electron transfer around Fe ions than GGA. The electronic structure of the polyanion unit PO4 also varied upon Li insertion from the results of O K-edge and P K-edge XAS. From ab initio calculations, it was suggested that the electronic structure of oxide ions was affected by the polarization effect due to the strong ionic character of inserted Li as well as the changes in electronic structure via the Fe 3d-O 2p hybridized orbital. Acknowledgment. J.S. would like to thank to the Japan Society for the Promotion of Science (JSPS) for financial support of this work. Figure 1, 8, 11, and 12 were described by using VENUS developed by R. A. Dilanian and F. Izumi.

17750 J. Phys. Chem. B, Vol. 110, No. 36, 2006 References and Notes (1) Nanjundaswamy, K. S.; Padhi, A. K.; Goodenough, J. B.; Okada, S.; Ohtsuka, H.; Arai, H.; Yamaki, J. Solid State Ionics 1996, 92, 1. (2) Manthiram, A.; Goodenough, J. B. J. Solid State Chem. 1987, 71, 349. (3) Manthiram, A.; Goodenough, J. B. J. Power Sources 1989, 26, 403. (4) Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 1188. (5) Padhi, A. K.; Nanjundaswamy, K. S.; Masquelier, C.; Okada, S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 1609. (6) Padhi, A. K.; Nanjundaswamy, K. S.; Masquelier, C.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 2581. (7) Abraham, K. M.; Pasquariello, D. M.; Willstaedt, E. B. J. Electrochem. Soc. 1990, 137, 743 (8) Goodenough, J. B.; Manivannan, V. Denki Kagaku 1998, 66, 1173. (9) Nakayama, M.; Goto, S.; Uchimoto, Y.; Wakihara, M.; Kitajima, Y. Chem. Mater. Commun. 2004, 16, 3399. (10) Nakayama, M.; Goto, S.; Uchimoto, Y.; Wakihara, M.; Kitajima, Y.; Miyanaga, T.; Watanabe, I. J. Phys. Chem. B 2005, 109, 11197. (11) Zhou, F.; Cococcioni, M.; Marianetti, C. A.; Morgan, D.; Ceder, G. Phys. ReV. B 2004, 70, 235121. (12) Bacq, O. L.; Pasturel, A.; Bengone, O. Phys. ReV. B 2004, 70, 245107. (13) Bacq, O. L.; Pasturel, A. Philos.l Mag. 2005, 85, 1747. (14) Zhou, F.; Cococcioni, M.; Kang, K.; Ceder, G. Electrochem. Commun. 2004, 6, 1144. (15) Osorio-Guillen, J. M.; Holm, B.; Ahuja, R.; Johansson, B. Solid State Ionics 2004, 167, 221. (16) Rousse, G.; Rodriguez-Carvajal, J.; Wurm, C.; Masquelier, C. Chem. Mater. 2001, 13, 4527. (17) Anderson, A. S.; Kalska, B.; Eyob, P.; Aernout, D.; Ha¨ggstro¨m, L.; Thomas, J. O. Solid State Ionics 2001, 140, 63.

Shirakawa et al. (18) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169. (19) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953. (20) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (21) Adams, S.; Swenson, J. Phys. ReV. Lett. 2000, 84, 4144. (22) Thangadurai, V.; Adams, S.; Weppner, W. Chem. Mater. 2004, 16, 2998. (23) Adams, S. Acta Crystallogr. 2001, B57, 278. URL: http:// kristall.uni-mki.gwdg.de/softbv/index.html. (24) Aydinol, M. K.; Kohan, A. F.; Ceder, G.; Cho, K.; Joannopoulos, J. Phys. ReV. B 1997, 56, 1354. (25) Kondratyuk, I. P.; Maksimov, B. A.; Muradyan, L. A. Dokl. Akade. Nauk SSSR 1987, 292, 1376. (26) Masquelier, C.; Padhi, A. K.; Nanjundaswamy, K. S.; Goodenough, J. B. J. Solid State Chem. 1998, 135, 228. (27) Morcrette, M.; Wurm, C.; Masquelier, C. Solid State Sci. 2002, 4, 239. (28) Augustsson, A.; Zhuang, G. V.; Butorin, S. M.; Osorio-Guillen, J. M.; Dong, C. L.; Ahuja, R.; Chang, C. L.; Ross, P. N.; Nordgren, J.; Guo, J.-H. J. Chem. Phys. 2005, 123, 184717. (29) Crocombette, J. P.; Pollak, M.; Jollet, F.; Thromat, N.; GautierSoyer, M. Phys. ReV. B 1995, 52, 3143. (30) Kuiper, P.; Kruizinga, G.; Ghijsen, J.; Sawatzky, G. A. Phys. ReV. Lett. 1989, 62, 221. (31) Nakayama, M.; Imaki, K.; Ra, W.; Ikuta, H.; Uchimoto, Y.; Wakihara, M. Chem. Mater. 2003, 15, 1728. (32) Nakayama, M.; Usui, T.; Uchimoto, Y.; Wakihara, M.; Yamamoto, M. J. Phys. Chem. B 2005, 109, 4135. (33) Szilagyi, R. K.; Frank, P.; George, S. D.; Hedman, B.; Hodgson, K. Inorg. Chem. 2004, 43, 8318. (34) Dathe, H.; Jentys, A.; Lercher, J. A. J. Phys. Chem. B 2005, 109, 21846.