J. Phys. Chem. C 2010, 114, 16813–16817
16813
Structural and Electronic Properties of Li-Ion Battery Cathode Material FeF3 R. F. Li,† S. Q. Wu,*,† Y. Yang,‡ and Z. Z. Zhu*,† Department of Physics and Institute of Theoretical Physics and Astrophysics, Xiamen UniVersity, Xiamen 361005, China, and State Key Lab for Physical Chemistry of Solid Surfaces, Xiamen UniVersity, Xiamen 361005, China ReceiVed: June 2, 2010; ReVised Manuscript ReceiVed: August 15, 2010
A new type of cathode materials for Li-ion batteries has been explored recently based on FeF3 in an attempt to raise the energy density and discharge voltage. In this work, the structural and electronic properties of cathode materials FeF3 for lithium ion batteries have been studied by the first-principles calculations within both the generalized gradient approximation (GGA) and GGA+U frameworks. Our results show that the antiferromagnetic configuration of FeF3 is more stable than the ferromagnetic one, which is consistent with experiments. An analysis of the electronic density of states shows that FeF3 is a Mott-Hubbard insulator with a large d-d type band gap. Additionally, a small spin polarization was found on F, consistent with a fluorin-mediated superexchange mechanism for the Fe-Fe magnetic interaction. 1. Introduction Lithium ion batteries (LIBs) have undergone intensive scientific research and shown successful commercial applications in a variety of portable electronic devices, hybrid electric vehicles, and energy storage for renewable energy sources, such as solar and wind. At the present time, the layered intercalation compound LiCoO21 is utilized as a cathode in most commercial lithium ion batteries. Recently, more developments are based on cathodic insertion materials, such as LiMnO2,2 LiNiO2,3 and LiFePO4.4 The reversible reaction of these insertion materials gives excellent cycling performance but involves at most a single electron transfer per formula unit, corresponding to a limited capacity. What is more, these oxides (e.g., LixCoO2, LixNiO2, and LixMn2O4) (x < 1)) have safety issues when involved in electrode reaction, because they release oxygen gas at elevated temperatures, which could react with organic solvent exothermally.5-7 As opposed to intercalation reactions, the reversible conversion process can utilize all the oxidation states of the transition metal during the redox cycle, which expects to attain a large theoretical energy density. Recently, 3d-transition metal binary compounds MxXy (M ) Co, Fe, Ni, etc.; X ) F, O, N, etc.)8-18 have been extensively investigated for a large multielectron redox capacity through reversible electrochemical conversion reactions: MxXy + nyLi+ + yne- ) yLinX + xM. Of the MxXy materials, only the metal fluorides can be used as alternative positive electrode materials for LiBs owing to the high operating voltages caused by their ionic character. However, the highly ionic nature of M-F bond leads to a poor electronic conductivity with a wide energy gap. Among transition metal fluorides, iron trifluoride (FeF3) was selected as a model system because of its low cost and low toxicity. The electrochemical properties of FeF3 were first investigated by Arai et al.8 The reported discharge specific capacity of FeF3 (80 mAh g-1)8 is far below the theoretical 1e- transfer reaction * To whom correspondence should be addressed. Tel.: +86 592 2182248. Fax: +86 592 2189426. E-mail address:
[email protected] (Z.Z.Z.) and
[email protected] (S.Q.W.). † Department of Physics and Institute of Theoretical Physics and Astrophysics. ‡ State Key Lab for Physical Chemistry of Solid Surfaces.
capacity of the Fe3+/Fe2+ couple (237 mAh g-1). On the other hand, the fact that electronic/ionic conductivity limitations are no longer insurmountable in the design of high performance electrode materials has led to a renewed interest in the lowcost and abundant Li-based insulating compounds such as phosphates (LiMPO4), silicates (Li2MSiO4), borates (LiMBO3), and fluorophosphates (LiMPO4F). As for FeF3, the limitation of poor conductivity had been overcome through the use of carbon-metal fluoride nanocomposites (CMFNCs).9,10 The CMFNCs also enabled a three-electron reversible process resulting in specific capacities as high as 600 mAh g-1 at 70 °C (note that the theoretical capacity of FeF3 is 712 mAh g-1), which is the highest energy density value in iron-based cathodes. This performance is roughly four times the specific capacity and nearly three times the gravimetric energy density of LiCoO2, respectively. It is also found that FeF3 shows good thermal stability even at elevated temperatures12 as a cathode for Liion batteries. Moreover, iron trifluoride composites (FeF3-C) prepared by planetary ball milling with carbon showed reversible charge/discharge behavior not only for Li, but also for Na.13 These previous studies9-13 suggest that FeF3 could be a strong contender as a potential alternative cathode material. Consequently, it is of great interest to explore and understand the intrinsic structural properties and bonding characteristics of the FeF3 material. In this paper, we focus on the study of ground state properties of FeF3, by analyzing the structural, magnetic, and electronic properties from first-principles calculations. 2. Computational Method The present calculations have been performed by using the Vienna ab initio simulation package (VASP),19,20 which is based on the density functional theory, the plane-wave basis, and the projector augmented wave (PAW) representation.21 The exchangecorrelation potentials are approximated by the generalized gradient-corrected function (GGA).22 To take into account the strong correlated character of the d-electrons of iron, a Hubbardlike correction (GGA+U)23 is included. Within the GGA+U approach, the on-site columb term U and the exchange term J can always be grouped together into a single effective interaction parameter Ueff ) U - J. The exchange parameter J is believed
10.1021/jp1050518 2010 American Chemical Society Published on Web 09/03/2010
16814
J. Phys. Chem. C, Vol. 114, No. 39, 2010
Figure 1. The crystal structures of FeF3 (a) in the hexagonal representation and (b) in the trigonal unit cell. The (012), (024), (111), and (211) planes are shown by pink shadows. and Fe and F atoms are denoted by dark red and blue gray, respectively. The spin directions in the AF state are shown by blue arrows.
to be close to its atomic value J ≈ 1 eV. The value of U is generally related to the valence state and the structure. Because the calculations on LaFeO3 with the same valence state of Fe and a similar structure (Perovskite) with FeF3 agree with experiment closely when U is set as 6 eV and Ueff ) U - J ) 5 eV,24 we, therefore, adopt the value of U as 6 eV in the present calculations, which should be a reasonably good approximation. The wave functions were expanded in plane-wave basis up to a kinetic energy cutoff of 500 eV. Brillouin-zone integrations were performed by using special k-point sampling of the Monkhorst-Pack scheme25 with a 4 × 4 × 2 grid. The convergence of total energy with respect to the kinetic energy cutoff and the k-point sampling has been carefully examined. The atomic geometry of FeF3 obtained was fully relaxed until the Hellmann-Feynman forces on all atoms were less than 0.01 eV/Å. Iron trifluoride was known to show antiferromagnetic (AF) behavior below the Ne´el temperature of 394 K.26 In the corresponding magnetic structure, Fe ions are coupled antiferromagneically via the intervening F ions to all six nearest neighbor Fe ions. The spin direction had been shown to be parallel to the (111) plane in the bimolecular unit cell,26 as shown in Figure 1b. For comparison, in the present work, calculations (collinear) are performed not only for the antiferromagnetic configurations but also for the ferromagnetic (FM) ones. The noncollinear magnetism of FeF3, in the (1j01) spin direction (which is parallel to the (111) plane), in the AF configuration with the GGA+U method is also calculated. The energy differences between (1j01) and other spin directions which are also parallel to the (111) plane are all minor. We therefore only present the results in the (1j01) spin direction. 3. Results and Discussions 3.1. Crystalline Structure and Structural Parameters. Several experimental studies27-29 have shown that FeF3 is in a trigonal structure with space group R3cj. The hexagonal representation of the lattice, as shown in Figure 1a, is related to a collapsed ReO3 perovskite structure. The Fe-F-Fe bond angle has deviated from the ideal 180° (experimental value of 153° 27). FeF3 has a layer structure and comprises corner-sharing FeF6/2 octahedra. In this structure, planes of Fe atoms and planes of fluorine atoms, perpendicular to the [001] direction, lie alternately and regularly. The fluorine atoms arrangement in FeF3 may be regarded as a considerably distorted hexagonal closepacking. In addition, it is notable to remark that the Fe3+ ions, which are in the center of a regular octahedron of equidistant fluorine, lie on the (012) planes of the trigonal structure. It is
Li et al. reasonable to expect that the (024) vacant plane between Fe (012) planes is available for Li+ insertion. In Table 1, we present the calculated lattice parameters, the bond length of Fe-F, the bond angle of Fe-F-Fe, and the spin magnetic moments of Fe and F ions in FeF3. Both of our GGA and GGA+U calculations predict that the AF configuration is more stable than the FM one. The differences of cohesive energies (per molecular formula) between AF and FM configurations are 0.291 eV for GGA and 0.130 eV for GGA+U calculations. It is noted that the optimized lattice parameters for AF configuration are in good agreement with experiment results27 (with an error less than 2%). Overall, our GGA+U calculations slightly overestimate the equilibrium unit cell volume with an error of 3.8%. Such an overestimation is in line with that found in the common GGA calculations. As we can see, the calculated Fe-F bond lengths show slight overestimation with respect to experimental data to within 1.0% for the AF and GGA+U calculations. The structure of FeF3 is related to a collapsed ReO3 perovskite structure, and the distortion of the ReO3 perovskite structure can be considered as a tilting of the rigid FeF6/2 octahedron. To better understand the distorted structures, the bimolecular unit cell of FeF3 is shown in Figure 1b. Such a rhombohedrally distorted structure can be characterized by three parameters: (1) the octahedral tilt angle w, which denotes the rigid rotation of a FeF6/2 octahedron; (2) the change of unit cell volume; and (3) a c/a ratio, with c and a shown in the Figure 1b. For the present FeF3 with an AF configuration, the octahedral tilt angle is w ) 12.9°, which is obtained from the bond angle of Fe-F-Fe, i.e., (180° 154.2°)/2 ) 12.9°. Due to the octahedral tilt, the unit cell volume shrinks, leading to a decrease in the volume of ∆V/V ) -6.88% and an increase of ratio c/a ) 1.027. Both values are very close to the calculated values of ∆V/V ) -sin2 w ) -5% and c/a ) 1/cos w ) 1.026, respectively. 3.2. Electronic and Magnetic Properties. The band structure of FeF3, together with the corresponding total density of states (TDOS), in the antiferromagnetic configuration calculated by the GGA+U scheme has been shown in Figure 2. To comprehend the electronic properties more clearly, the ldecomposed and projection over atomic contributions of the density of states for FeF3 predicted by GGA and GGA+U methods are also calculated for both the FM and AF configurations, and the data are presented in Figure 3 for comparison. By considering data outlined in Figure 3d, we can interpret the band structure and the sequence of partial DOS peaks appearing from low to high energies as follows. The band structure is shown to be divided into several regions. Two narrow bands located at around -7 and -6 eV, corresponding to two sharp peaks in the DOS plot (Figure 3d), are of mainly Fe-3d character and with small F-2p contributions. In the energy region from -4.5 eV to the top of the valence band, according to the DOS plot in Figure 3, the bands show significant admixture of fluorine 2p and iron 3d states (i.e., showing a significant hybridization of F-2p and Fe-3d atomic orbitals). For the conduction band located at energy larger than 4 eV, the contribution is almost fully from the Fe-3d orbit (Figure 3d). Figure 3a-d suggests that the electronic structures predicted by the GGA+U method are quite different from those by the GGA scheme. The GGA calculations predicted FeF3 to be a semiconductor with band gaps of 0.267 eV for FM and 1.800 eV for AF, respectively, while GGA+U calculations indicated that FeF3 is an insulator with band gaps of 3.019 eV for FM and 4.138 eV for AF configurations, respectively. As the valence conduction band gap separates Fe-d states, FeF3 is a classic Mott-Hubbard
Properties of Li-Ion Battery Cathode Material FeF3
J. Phys. Chem. C, Vol. 114, No. 39, 2010 16815
TABLE 1: The Optimized Lattice Parameters, the Fe-F Bond Length, the Bond Angle of Fe-F-Fe, and the Magnetic Moments of Fe and F Ions of FeF3 method
a, Å
c, Å
V, Å3
RFe-F, Å (error)
∠Fe-F-Fe, deg
m (Fe), µB
m (F), µB
AF: Exp 31 FM: GGA FM: GGA+U AF: GGA AF: GGA+U
5.198 5.199 5.177 5.278 5.277
13.331 13.573 13.484 13.495 13.430
311.93 317.52 312.84 325.44 323.90
1.92 1.959 (2.0%) 1.946 (1.4%) 1.943 (1.2%) 1.939 (1.0%)
153 147.2 147.9 154.2 154.2
4.528 (Exp 37) 4.230 4.445 4.047 4.367
0.216 0.146 0 0
insulator like FeF2,30 rather than a charge-transfer insulator such as Fe2O3.31 It is reported that the energy gap in FeF3 thin films is about 5.96 eV,32 which suggests that the present AF calculations is a better prediction of the band gap than FM ones. This is reasonable when taking into account the strong correlated character of the Fe d-electrons. The wide energy gap resulting in poor electronic conductivity seems likely to restrict the application of FeF3 as rechargeable cathode materials. However, electronic limitations are no longer an insurmountable issue in the design of high-performance electrode material. It can be overcome through various materials processing approaches, including the use of carbon coatings, mechanical grinding, or mixing,9 and low-temperature synthesis routes11 to obtain tailored particles. Therefore, low-cost and abundant Li-based insulating compounds, such as FeF3, have gained renewed interests in order to obtain high voltage and large capacity. From Figure 3, it is also shown that antiferromagnetic DOS’s are much more localized than those of ferromagnetic ones (i.e., the band structures of AF FeF3 are much more localized than those of FM FeF3, for both the GGA and GGA+U calculations). For all the FM and AF calculations, Fe ions are found to have high-spin ground states with the up-spin Fe-3d bands completely filled while the down-spin Fe-3d bands are nearly unoccupied. The calculated magnetic moments of Fe, as shown in Table 1, are all close to the experimental value of 4.528 µB,33 which are also quite near the saturation value of 5 µB according to Hund’s rule. We have also performed calculations on the noncollinear magnetism of FeF3 in the AF configuration with the GGA+U method. The magnetic moment from noncollinear magnetism calculation is 4.429 µB, while the magnetic moment is 4.367 µB for the collinear magnetism calculations. The spin direction was suggested to be parallel to the (111) plane in the
Figure 2. Band structure and the corresponding total density of states (TDOS) for FeF3 in the antiferromagnetic configuration from GGA+U calculations. As for DOS, only the spin-up results are given. The Fermi level is at zero energy.
bimolecular unit cell from the experiment,26 we therefore performed the noncollinear calculation in the (1j01) spin direction (which is parallel to the (111) plane), as shown in Figure 1b. Orbital moment is rather small, about (-0.007, 0.007) for the neighboring Fe atoms. The difference of cohesive energies per molecular formula for the noncollinear and collinear magnetism is 0.002 eV, while noncollinear calculation turns out to be a little bit more stable than the collinear calculation. The spin configuration of Fe-ions in the material is vital to the bond angle of Fe-F-Fe in the structure. The high-spin configuration of Fe3+ (3d5 in t2g3eg2) with three holes in the t2g shell contributes to the distorted structure of FeF3 due to the PJT coupling (pseudo Jahn-Teller effect) between the empty 3d orbit of Fe and the full 2p orbit of the F.34 The deformation charge density plot for the plane containing the Fe-F-Fe bond of FeF3 is shown in Figure 4a. The deformation charge density, ∆F(r b), is defined as the difference between the total charge density in the solid and the superposi-
Figure 3. The density of states for FeF3 predicted by GGA and GGA+U methods for ferromagnetic and antiferromagnetic states, respectively. The positive (negative) value is for the majority (minority) spin contribution. The total DOS and the Fe-3d and F-2p partial DOS’s are presented by black, red, and blue lines, respectively. The Fermi level is at zero energy.
16816
J. Phys. Chem. C, Vol. 114, No. 39, 2010
Li et al. that conforms to the Fe1-Fe2 superexchange interaction via the diamagnetic ligand F. 4. Conclusion
Figure 4. Contour plots of (a) deformation charge density and (b) spin-density for AF with GGA+U calculations (collinear) on the (211) plane of trigonal unit cell containing the Fe-F-Fe bond. The atomic positions in the plots have been shown in Figure 1b. Isodensity curves are separated by (a) 0.025 and (b) 0.065 e Å-3. Solid, dash, and dotdash lines correspond to positive, negative, and zero values, respectively.
tion of independent atomic charge densities placed at the atomic sites of the same solid, i.e.,
∆F(b) r ) F(b) r -
∑ Fatom(br - bRµ) µ
where b Rµ is the atomic coordinate. For the present deformation charge densities, all the calculated results give similar maps. Therefore, only the result from the GGA+U approach in the AF configuration is presented in Figure 4a. According to Figure 4a, the bonding between Fe and F shows both conspicuous ionic characteristic and slight covalency, in line with the high electronegativity of the F atom. This is confirmed by the covalency analysis in which covalency of the Fe-F bonds in FeF3 was indicated to be quite low compared with that of a similar oxide environment for Fe3+.33 Remarkable charge polarization of F ion is also found, which is a consequence of interactions between the neighboring F-Fe ions. The anion polarizability and dipole-dipole interactions35 have been proved as the major influences about the stabilization of tilts of cornercoupled rigid octahedral. To further understand the magnetic properties, the calculated spin density plots (spin-up minus spin-down) for the AF configuration within the GGA+U scheme are given in Figure 4b. The magnetism of FeF3 is mainly expressed by Fe as expected. Fe shows remarkable high-spin magnetism corresponding to the qualitative analysis of both spin density plots in Figure 4b and density of states in Figure 3. It is also interesting to remark that fluorine exhibits magnetism. In the AF case, we define the atom with a net excess of spin-down over spin-up electrons as Fe1, the neighboring Fe ions with different spin as Fe2, as shown in Figure 4b. On Fe1 and Fe2, the spin-down and spin-up electrons are almost spherically distributed over all d orbits. On the site of fluorine, a small spin polarization is clearly visible in Figure 4b. This is because Fe1 repulses spin-down electrons but attracts spin-up ones of F, leaving an excess of spin-down electrons on F in the direction of Fe1-F. The opposite occurs in the Fe2-F case. A similar effect of spin electrons transfer was discussed before for FeF232 and Fe2O3.33 The spin polarization on F is convincing evidence
In summary, first-principles calculations based on the density functional theory and the GGA and GGA+U schemes have been used to predict the structural properties and electronic structures of FeF3 in the ferromagnetic and antiferromagnetic states. Our results show that GGA+U approximation provided more reasonable predictions on the structural and electronic properties of FeF3, considering the strongly correlated 3d electrons of Fe. The optimized structural parameters of antiferromagnetic FeF3 from GGA+U are 1-2% in excess in comparison with the experimental ones. By comparing the cohesive energies of FeF3 in the AF and FM states, the antiferromagnetic FeF3 also turns out to be more stable, consistent with experiments. The electronic structure calculations indicate that FeF3 is a Mott-Hubbard insulator with strong ionic character and a large bandgap (with then a poor conductivity). Actually, the limitation of poor conductivity has been overcome through the use of carbon-metal fluoride nanocomposites. Further theoretical and experiment work on the doping of transitional metal ions to acquire better conductivity and phase stability should also be helpful. Furthermore, the structure analysis on the layered structure of FeF3 in this paper provides some useful insights into the understanding of the Li insertion properties (LixFeF3) as well as the lithium migration. Acknowledgment. The present work is supported by the National Natural Science Foundation of China (Grant No.10774124) and the National 973 Program of China (Grant No. 2007CB209702). References and Notes (1) Mizushima, K.; Jones, P. C.; Wiseman, P. J.; Goodenough, J. B. Mater. Res. Bull. 1980, 15, 783. (2) Armstrong, A. R.; Robertson, A. D.; Bruce, P. G. Electrochim. Acta 1999, 45, 285. (3) Thomas, M. G. S. R.; David, W. I. F.; Goodenough, J. B.; Groves, P. Mater. Res. Bull. 1985, 20, 1137. (4) Padhi, A. K.; Nanujundaswamy, K. S.; Goodenough, J. B. J. Electrochem. Soc. 1997, 144, 1188. (5) MacNeil, D. D.; Christensen, L.; Landucci, J.; Paulsen, J. M.; Dahn, J. R. J. Electrochem. Soc. 2000, 147 (3), 970. (6) Arai, H.; Tsuda, M.; Saito, K.; Hayashi, M.; Sakurai, Y. J. Electrochem. Soc. 2002, 149, A401. (7) MacNeil, D. D.; Hatchard, T. D.; Dahn, J. R. J. Electrochem. Soc. 2001, 148 (7), A663. (8) Arai, H.; Okada, S.; Sakurai, Y.; Yamaki, J. J. Power Source 1997, 68, 716. (9) Badway, F.; Pereira, N.; Cosandey, F.; Amatucci, G. G. J. Electrochem. Soc. 2003, 150 (9), A1209. (10) Badway, F.; Cosandey, F.; Pereira, N.; Amatucci, G. G. J. Electrochem. Soc. 2003, 150 (10), A1318. (11) Li, T.; Li, L.; Cao, Y. L.; Ai, X. P.; Yang, H. X. J. Phys. Chem. 2010, 114, 3190. (12) Zhou, M.; Zhao, L.; Doi, T.; Okada, S.; Yamaki, J. J. Power Sources 2010, 195, 4952. (13) Nishijima, M.; Gocheva, I. D.; Okada, S.; Doi, T. J. Power Sources 2009, 190, 558. (14) Morcrette, M.; Rozier, P.; Dupont, L.; Mugnier, E.; Sannier, L.; Galy, J.; Tarascon, J. M. Nat. Mater. 2003, 2–755. (15) Poizot, P.; Laruelle, S.; Grugeon, S.; Dupont, L.; Tarascon, J. M. Nature 2000, 407, 496. (16) Poizot, P.; Laruelle, S.; Grugeon, S.; Tarascon, J. M. J. Electrochem. Soc. 2002, 149, A1212. (17) Pereira, N.; Klein, L. C.; Amatucci, G. G. J. Electrochem. Soc. 2002, 149, A262. (18) Pereira, N.; Balasubramanian, M.; Dupont, L.; McBreen, J.; Klein, L. C.; Amatucci, G. G. J. Electrochem. Soc. 2003, 150, A1118. (19) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169. (20) Kresse, G.; Furthmu¨ller, J. Comput. Mater. Sci. 1996, 6, 15.
Properties of Li-Ion Battery Cathode Material FeF3 (21) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (22) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson, M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671. (23) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. P. Phys. ReV. B 1998, 57, 1505. (24) Yang, Z. Q.; Huang, Z.; Ye, L.; Xie, X. Phys. ReV. B 1999, 60, 15674. (25) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188. (26) Wollan, E. O.; Child, H. R.; Koehler, W. C.; Wilkinson, M. K. Phys. ReV. 1958, 112, 1132. (27) Hepworth, M. A.; Jack, K. H.; Peacock, R. D.; Westland, G. J. Acta Crystallogr. 1957, 10, 63. (28) Leblanc, M.; Pannetier, J.; Ferey, G.; De Paper, R. ReV. Chim. Miner. 1985, 22, 107.
J. Phys. Chem. C, Vol. 114, No. 39, 2010 16817 (29) Daniel, P.; Bulou, A.; Rousseau, M.; Nouet, J. Phys. ReV. B 1990, 42, 10545. (30) Valerio, G.; Catti, M.; Dovesi, R.; Orlando, R. Phys. ReV. B 1995, 52, 2422. (31) Catti, M.; Valerio, G.; Dovesi, R. Phys. ReV. B 1995, 51, 7441. (32) Lachter, A.; Salardenne, J.; Barrie`re, A. S. Phys. Status Solidi B 1979, 90, 147. (33) Jacobson, A. J.; McBride, L.; Fender, B. E. F. J. Phys. C: Solid State Phys. 1974, 7, 783. (34) Garcia-Fernandez, P.; Aramburu, J. A.; Barriuso, M. T.; Moreno, M. J. Phys. Chem. Lett. 2010, 1, 647. (35) Allen, P. B.; Chen, Y. R.; Chaudhuri, S.; Grey, C. P. Phys. ReV. B 2006, 73, 172102.
JP1050518
