Neutron Diffraction and Electrochemical Study of FeNb11O29

Mar 5, 2014 - Data recorded on the two high-angle banks with best resolution (2θ .... Taking into account their site multiplicities, the chemical for...
2 downloads 0 Views 2MB Size
Article pubs.acs.org/cm

Neutron Diffraction and Electrochemical Study of FeNb11O29/ Li11FeNb11O29 for Lithium Battery Anode Applications Ilya Pinus, Michele Catti,* Riccardo Ruffo, Matteo M. Salamone, and Claudio M. Mari Dipartimento di Scienza dei Materiali, Università di Milano Bicocca, via Cozzi 55, 20125 Milano, Italy S Supporting Information *

ABSTRACT: Lithium was found to be inserted into FeNb11O29 by reaction with n-butyllithium, producing Li11FeNb11O29 samples with good crystallinity, and the Amma crystallographic-shear structures of both compounds (a = 28.7093(6)/28.4036(4), b = 3.8256(1)/4.08447(9), c = 20.6241(4)/20.7067(2) Å, respectively) were Rietveld-refined by high-resolution neutron powder diffraction. Lithium atoms of Li11FeNb11O29 were located both in 4-fold- and in 5-foldcoordinated positions, lying respectively inside and between the 4 × 3 perovskite blocks of Nb(Fe)O6 octahedra. Electrochemical measurements on a FeNb11O29/ LixFeNb11O29 electrode vs metallic Li showed that lithium can be intercalated/deintercalated reversibly in the 1.1−2.5 V range with a stable capacity of 185 mAh/g, which roughly corresponds to 11 Li atoms per f.u. as obtained by chemical lithiation. The rather uniform structural distribution found for Li atoms is consistent with the good electrode reversibility, which makes this material promising as anode in rechargeable lithium batteries. can be achieved, but in the initial cycles only, where some Nb5+ and/or Nb4+ atoms would be more deeply reduced to the Nb3+ state. It seemed then interesting to explore the response to electrochemical Li intercalation of a similar material, FeNb11O29. This oxide is observed with a monoclinic and an orthorhombic phase similar to those reported for Nb12O29.14,15 However, the valence state distribution of Nb atoms is slightly different, as all of them are formally Nb5+ according to Fe3+Nb115+O29; further, also iron can be involved in the redox reaction during electrode operation. Differences in the electrochemical lithium insertion behavior may thus ensue, with a possible increase of capacity up to 20% including the Nb5+/ Nb4+ and Fe3+/Fe2+ processes. From the structural point of view, the issue of order/disorder distribution of Fe/Nb in the metal atomic sites deserves particular attention, as it can affect the local metal-atom bonding and thus the hierarchy of interstitial Li sites during the insertion reaction. For clarifying such points, a neutron diffraction and electrochemical investigation of this system was undertaken, and its results are presented and discussed below.

1. INTRODUCTION In the search for anodes alternative to Li/graphite in lithium batteries, a number of oxide materials with electrode voltages in the range 1.0−2.0 V vs the Li/Li+ couple have been proposed and tested successfully. Their advantage is to avoid the risk of electrochemical reaction with the conventional organic carbonate electrolyte, and also the possible formation of lithium dendrites with the LiC6 anode.1 Many such materials belong to the family of titanates, such as spinel-like Li4Ti5O122−4 and LiTi2O4, with spinel4,5 and with anatase5 structures, which exploit the convenient Ti4+/Ti3+ redox potential. Also the Nb5+/Nb4+ couple was considered as suitable in this respect, and different polymorphs of Nb2O5 were proposed as good candidates for anode applications.6 Other niobium oxides where Nb is slightly reduced with respect to the +5 state are also observed:7 in particular, Nb12O29 with an average oxidation state of 4.83. This compound is reported with a monoclinic A2/m8 and an orthorhombic Cmcm9 modification, whose crystal structures are related to that of αNb2O5.10 A relevant interest in Nb12O29 was raised by the fact that it is a metallic conductor in both modifications and that its monoclinic phase orders antiferromagnetically below 12 K.11,12 Electron conductivity is related to the formal mixed valence state Nb24+Nb105+O29. A recent study13 reports about the LixNb12O29 vs Li electrochemical cell, showing that about 9.5 Li per formula unit can be inserted/extracted reversibly into/from monoclinic Nb12O29 at 6.6 mA g−1 current density. This corresponds to approximately all 10 Nb5+ atoms/f.u. involved into the Nb5+/ Nb4+ redox reaction. Even a larger capacity up to 14.3 Li/f.u. © 2014 American Chemical Society

2. EXPERIMENTAL SECTION Synthesis and X-ray Diffraction. The orthorhombic modification of FeNb11O29 was synthesized by solid state reaction techniques.16 Fe2O3 and orthorhombic T-Nb2O5 (Sigma-Aldrich) in stoichiometric ratio were mixed with ethanol and treated in a Retsch planetary ball mill for 1 h at 400 r.p.m. speed. The mixture was dried, pelletized and Received: February 6, 2014 Revised: March 4, 2014 Published: March 5, 2014 2203

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials

Article

annealed at 1300 °C for 6 h, and then naturally cooled down to room temperature. X-ray powder diffraction (Bruker D8 Advance diffractometer, CuKα radiation) proved the obtained sample to be pure Amma phase of FeNb11O29. Chemical lithiation of Amma-FeNb11O29 was performed by reaction with n-butyllithium in hexane, under N2 flow and with vigorous stirring. The reaction at 50 °C for 96 h produced a black powder, which was filtered and washed repeatedly with hexane in a glovebox under Ar atmosphere; the sample was finally dried under vacuum. By X-ray diffractometry (PANalytical equipment with a sample holder airprotected by Kapton foil) a single-phase pattern was obtained which could be indexed in the Amma space group. Chemical analysis by atomic emission flame photometry (Perkin-Elmer Analyst 400) gave a Li content of 13.7 Li per f.u., indicating a slight excess with respect to the formal Nb5+ → Nb4+ reduction of all niobium atoms. However, the Li11FeNb11O29 chemical formula was retained because consistent with the results of the structure determination by neutron diffraction (cf. the corresponding section below). All attempts to increase the amount of inserted lithium by raising temperature and/or time in the course of the reaction with n-butyllithium failed to produce samples with acceptable crystallinity for crystallographic characterization. Neutron Diffraction. Time-of-flight (T.O.F.) neutron diffraction intensity profiles were measured at the ISIS pulsed spallation source (Rutherford Appleton Laboratory, Chilton, Didcot, U.K.) on samples of FeNb11O29 and Li11FeNb11O29 in two different experiments. In both cases powder specimens of about 2 cm3, put in vanadium cans under vacuum, were employed, but for the lithiated compound the can was sealed with indium wire in an Ar atmosphere. The experiment on FeNb11O29 was carried out on the INES diffractometer, with use of nine counter banks located at 2θ from 163.2 to 19.0°. The corresponding Δd/d resolution range is 0.0012−0.013. Data recorded on the two high-angle banks with best resolution (2θ = 163.2° and 145.2°) were used for the Rietveld refinements, in the dhkl ranges 0.42−1.51 and 0.36−1.57 Å, respectively. Li11FeNb11O29 was measured on the high-resolution HRPD instrument with three banks; only the data from the high-angle bank at 2θ = 168.3° (Δd/ d resolution = 4 × 10−4; dhkl range 0.64−2.64 Å) were employed for the structure analysis. Preliminary data reductions included merging of outputs from single counters in the bank and correction for detector efficiency as a function of neutron wavelength. The Rietveld refinements of the crystal structures were carried out by the GSAS computer package.17 The intensity background was modeled by six Chebyshev polynomials, and the peak shape was represented by a convolution of a pseudoVoigt function (linear combination of Gaussian and Lorentzian components, with σ and γ half-widths, respectively: sample contribution) with two back-to-back exponentials (instrumental and moderator contributions).18 The σ and γ parameters were assumed to depend on dhkl according to σ = (σ1dhkl2 + σ2dhkl4)1/2 and γ = γ2dhkl2; the σ1, σ2, γ2 quantities were included in the refinement. The mixing coefficient and the full width of the pseudo-Voigt function depend on σ and γ according to equations given in the literature.19 For the lithiated compound, an anisotropic broadening contribution to peak shape20 had to be included in order to attain a satisfactory profile fit. The technique of difference Fourier maps was used to locate lithium atoms. Electrochemical Measurements. The FeNb11O29 sample was ball milled by a Retsch planetary equipment at 400 r.p.m. for 4 h. Active material electrodes were fabricated by mixing the milled powder (70%), PVDF binder (15% Solvay 6020) and conductive carbon black (15%, Super P MMM Carbon). The mixture was dispersed in nmethyl-pyrrolidone (NMP) to obtain a dense slurry which was casted on copper foil and dried at 90 °C for 2h. The active material load was around 3−5 mg/cm2. Two electrochemical cell setups were used in the functional characterization. The cyclic voltammetries were performed in three electrode T shaped Swagelock cells while the galvanostatic cycling were carried out using CR2032 coin cells. In both cases, metallic lithium foils were used as reference and counter electrode (all potentials are reported versus the couple Li+/Li). Polypropilene discs were used as separators and filled with the electrolyte, which was a

commercial solution of 1 M LiPF6 in ethylenecarbonate:diethylcarbonate (EC:DEC) 1:1. Cells were assembled in an argon filled glovebox ([O2] < 1 ppm). The measurements were carried out at room temperature using a Biologic VMP3 multichannel battery tester.

3. RESULTS AND DISCUSSION Structural Properties. The X-ray diffraction patterns of FeNb11O29 (a = 28.719(3), b = 3.8272(4), c = 20.630(3) Å) and Li11FeNb11O29 (a = 28.401(7), b = 4.081(2), c = 20.724(6) Å) are shown in Figure 1. By inspection of reflection indexes

Figure 1. X-ray diffraction patterns (CuKα radiation) of orthorhombic FeNb11O29 and Li11FeNb11O29, with some Bragg peak indexes outlined.

the hkl peaks with k > 0 appear to be strongly displaced (cf. in particular the 020 and 011 cases), whereas all h0l peaks do not move very much, as a consequence of the lithium insertion reaction. This corresponds to the large increase of the b cell parameter and to the minor changes of a and c on lithiation, as will be discussed further below. The Rietveld refinement of FeNb11O29 neutron data was started from atomic coordinates of orthorhombic Nb12O29,9 neglecting the presence of Fe. However, the nonstandard setting Amma of the symmetry space group was used, rather than the standard Cmcm one as in ref 9, for consistency with most of the literature on these compounds.8,10−12 Two different structural models including Fe were tried: (i) full disordered distribution, with occupancies 1/12 and 11/12 for Fe and Nb, respectively, on all six independent Nb sites; (ii) six different ordered models, with occupancy 1/2 for both Fe and Nb on one of the Nb sites, and full Nb occupation on all the remaining ones. At convergence, all ordered models of type (ii) gave agreement factors slightly worse than those for case (i), even on relaxing the site occupancies. It was then concluded that Fe is completely disordered over all sites, and the results of model (i) were retained. The obtained structure of the unlithiated compound was used as starting model for refining data of Li11FeNb11O29 in the same space group Amma. By use of Fourier difference maps, 2204

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials

Article

seven Li atoms in the asymmetric unit were located as negative peaks, and their positions could be successfully included in the refinement. Taking into account their site multiplicities, the chemical formula with 11 Li atoms per f.u. ensues, and this was retained as definite composition. The final unit-cell constants and agreement factors at convergence are reported for both compounds in Table 1, Table 1. Unit-Cell Constantsa and other Rietveld Refinement Results of Neutron Diffraction Data for FeNb11O29 and Li11FeNb11O29 in Space Group Amma (Z = 4) a (Å) b (Å) c (Å) V (Å3) data variables wRp Rp R(F2) red. χ2 a

FeNb11O29

Li11FeNb11O29

28.7093(6) 3.8256(1) 20.6241(4) 2265.17(8) 2547 80 0.0502 0.0544 0.1158 2.014

28.4036(4) 4.08447(9) 20.7067(2) 2402.26(6) 4679 89 0.0314 0.0315 0.0917 4.960

Figure 3. T.O.F. neutron powder diffraction data (HRPD, 2θ = 168.33°) of Li11FeNb11O29; observed, Rietveld calculated, background, and difference intensity profiles are shown.

Li11FeNb11O29 is shown in Figure 4. Four blocks of 4 × 3 Nb(Fe)O6 octahedra are contained in the unit-cell, with each block corresponding to one f.u.; intrablock connection of octahedra is provided by corner-sharing, whereas interblock connection occurs by edge-sharing. Blocks are linked along y by octahedral edge-sharing, forming infinite [010] columns. As a first important result, the structural arrangement of the octahedral framework, similar to that observed in Nb12O29,9 is proved to be substantially preserved on lithiation. On this basis, a good operation reversibility should be expected for the electrochemical reaction of Li insertion/extraction in the LixFeNb11O29 system. However, some significant changes of the Nb(Fe)−O bonding geometry can be observed as a result of Li insertion (Table 3). The most important effect is a systematic volume expansion of all coordination octahedra, as shown by the trend of average bond lengths, which is consistent with the unit-cell behavior. Further, on lithium insertion most octahedra become less distorted, displaying a smaller range of Nb(Fe)-O distances; a substantial increase of the displacement factors of Nb(Fe) atoms is also generally observed, and it should be ascribed to disorder induced by lithiation rather than to thermal effects. The bonding environments of lithium atoms and Li−O bond lengths of Li11FeNb11O29 are reported in Table 4. Three Li atoms (Li1, Li6, Li7) in the asymmetric unit are 5-fold- and the other four are 4-fold-coordinated (cf. also Figure 4). Indeed the Li4 site should be normally 5-fold-coordinated, but because of Li displacement the fifth Li4−O distance attains 2.68 Å and becomes nonbonding. As discussed previously in particular with reference to Li1.714Nb2O5,22 in niobate shear structures the possible Li sites with coordination number (C.N.) =4 (called henceforth Li(IV)) lye close to the square windows delimiting the 12-corners oxygen cages, with a flat square pyramidal bonding environment, just as in standard perovskites. Li sites with C.N.=5 (denoted as Li(V)) occur near windows adjacent to edge-sharing octahedra (“crystallographic shearing”), so that the lithium C.N. is raised from 4 to 5 by an extra contact with an O atom belonging to the sheared plane. Therefore the latter sites should be preferred on pure energy grounds with respect to those with lower C.N., and this has been confirmed by accurate first-principles calculations.22

E.s.d.’s are given in parentheses.

and the graphs of observed, calculated and difference intensities are plotted vs dhkl in Figure 2 (higher angle bank) and Figure 3

Figure 2. T.O.F. neutron powder diffraction data (INES, 2θ = 163.20°) of FeNb11O29; observed, Rietveld calculated, background, and difference intensity profiles are shown.

for FeNb11O29 and Li11FeNb11O29, respectively. The quality of the refinements is satisfactory in both cases, but it is slightly better for the lithiated compound measured on the higher resolution instrument, as expected. On comparing the two unitcell geometries, a volume increase of 6.1% appears to be caused by lithium insertion; this is yet highly anisotropic and mostly due to the large b edge expansion (6.8%), whereas c expands by only 0.4% and a even contracts by −1.1%. Such a behavior is consistent with what previously observed for other lithium niobates,21,22 and it can be explained on the basis of the structural analysis discussed below. The refined atomic fractional coordinates and displacement factors are given in Table 2 for both phases, and the structure of 2205

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials

Article

Table 2. Refined Atomic Fractional Coordinates and Displacement Factorsa of FeNb11O29 (Upper Lines) and of Li11FeNb11O29 (Lower or Single Lines) site

X

y

z

U (10−2 Å2)

Nb(Fe)1

8f

Nb(Fe)2

8f

Nb(Fe)3

8f

Nb(Fe)4

8f

Nb(Fe)5

8f

Nb(Fe)6

8f

O1

4c

O2

4c

O3

4c

O4

8f

O5

8f

O6

8f

O7

8f

O8

8f

O9

8f

O10

8f

O11

8f

O12

8f

O13

8f

O14

8f

O15

8f

O16

8f

Li1 Li2 Li3 Li4 Li5 Li6 Li7

4c 4c 4c 8f 8f 8f 8f

0.0466(2) 0.0384(3) 0.0496(2) 0.0421(5) 0.0486(2) 0.0346(6) 0.1890(2) 0.1754(7) 0.1820(3) 0.1779(4) 0.1808(4) 0.1817(5) 0.25 0.25 0.25 0.25 0.25 0.25 0.0441(3) 0.0382(4) 0.0356(3) 0.0360(5) 0.0379(4) 0.0350(5) 0.0328(3) 0.0391(5) 0.0276(3) 0.0364(5) 0.1122(4) 0.1099(4) 0.1113(4) 0.1116(5) 0.1086(3) 0.1087(6) 0.1790(3) 0.1810(6) 0.1812(4) 0.1810(5) 0.1771(3) 0.1819(8) 0.1796(5) 0.1794(6) 0.1844(3) 0.1838(6) 0.25 0.25 0.25 0.0549(9) 0.1190(12) 0.0227(24) 0.1199(12)

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.5 0 0 0 0 0

0.0387(3) 0.0495(4) 0.6677(4) 0.6591(7) 0.8518(5) 0.8577(7) 0.0369(4) 0.0408(6) 0.6682(3) 0.6625(3) 0.8531(6) 0.8555(7) 0.0416(8) 0.0540(7) 0.6673(6) 0.6467(9) 0.8510(12) 0.8449(13) 0.5618(4) 0.5463(5) 0.1487(6) 0.1460(6) 0.7565(7) 0.7537(6) 0.3463(6) 0.3478(7) 0.9488(6) 0.9512(6) 0.0336(5) 0.0426(5) 0.6615(5) 0.6587(7) 0.8589(5) 0.8477(8) 0.5603(4) 0.5489(7) 0.1445(6) 0.1496(7) 0.7570(5) 0.7503(8) 0.3514(7) 0.3572(6) 0.9475(4) 0.9485(6) 0.540(7) 0.4366(20) 0.353(4) 0.4331(11) 0.3254(15) 0.239(4) 0.1490(20)

0.38(9) 0.4(2) 0.61(9) 4.2(4) 1.1(1) 4.6(4) 0.53(9) 5.3(4) 0.9(1) 0.5(2) 2.3(2) 2.8(3) 1.1(3) 0.1 0.4(2) 1.1(3) 1.9(3) 4.2(7) 0.5(1) 0.3(2) 1.0(1) 1.4(3) 1.3(2) 1.1(2) 1.2(2) 2.0(3) 0.76(1) 0.8(3) 1.3(2) 0.2(2) 1.2(2) 2.5(3) 0.9(1) 2.8(3) 0.3(1) 1.8(3) 1.0(1) 1.6(3) 1.0(2) 3.8(4) 1.6(2) 2.0(3) 0.5(1) 2.0(3) 10(4) 0.7(9) 2.5(1.4) 0.1 1.6(7) 8(2) 2.0(8)

Figure 4. Crystal structure of Li11FeNb11O29 projected onto (010). Li atoms are shown as small dark circles and numbered as in Table 2.

as due to deep reduction of all niobium atoms from Nb5+ to Nb3+ state, plus an additional Fe3+ to Fe2+ reduction of the iron atom. This is the ideal ‘crystallographic limit’ for Li insertion into FeNb11O29. On the other hand, structural results indicate that the chemical lithiation reaction in well crystalline conditions ended at the 11 Li/f.u. stage, corresponding to a formal Nb5+ to Nb4+ reduction of Nb atoms and to an electrochemical capacity of 191 mAh/g. Taking into account the site multiplicities (Table 2), there are 5 Li(V) atoms (contributed by Li1, Li6 and Li7) and 6 Li(IV) (contributed by Li2, Li3, Li4, Li5) out of 11 Li per f.u.. The observed occurrence fraction of 5/11=0.455 for Li(V) is hardly larger than the probability fraction of 10/23=0.435. This would indicate that entropic and/or kinetic factors play a more important role than bare energy in driving Li atoms into the available sites during the insertion reaction. Further, the occupied Li sites do not lie exclusively on the periphery of the 3 × 4 octahedra blocks (Li(V)), as previously assumed,21,13 but they also lie within the blocks (Li(IV)). Similar principles should apply also for the electrochemical lithiation discussed below. Electrochemical Properties. A SEM (Scanning Electron Microscopy) picture of the iron niobium oxide powder used for electrode preparation is shown in Figure 5. The specimen displays a fairly homogeneous distribution of submicrometer sized particles with platelet morphology. Results of cyclic voltammetry (CV) measurements on the FeNb11O29 electrode are reported in Figure 6. The first cathodic branch shows three current peaks at 1.55, 0.99, and 0.73 V, respectively. The current/time integration reveals a charge of 555 mAh/g inserted into the electrode in the explored potential range. This charge value is much larger compared to the theoretical capacity calculated by the reduction of 11 Nb5+ to Nb4+ ions with the insertion of 11 Li+ ions in the octahedral framework structure (191 mAh/g). It is also larger than full saturation of the available 23 Li crystallographic sites per f.u. (400 mAh/g), so that further reduction of all Nb4+ ions to lower oxidation states in the process is very likely. The observed excess capacity of 155 mAh/g may be actually due to electrolyte decomposition, as it is well-known that EC based electrolytes are not stable below 1.2 V vs metallic Li, though this phenomenon is limited by

a

E.s.d.’s are given in parentheses. Site occupancies are 0.9167 and 0.0833 for Nb and Fe atoms, respectively.

In the present structure there are 13 Li(IV) and 10 Li(V) possible sites per f.u., corresponding to a maximum Li insertion capacity of Li23FeNb11O29. This would correspond to an electrochemical capacity of about 400 mAh/g, to be interpreted 2206

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials

Article

Table 3. Interatomic Nb(Fe)−O Distances (Å) of FeNb11O29 and Li11FeNb11O29a Nb(Fe)1−O9 −O8 −O4 −O8′ −O5 Average Nb(Fe)2−O10 −O6 −O5 −O4 −O7 Average Nb(Fe)3−O11 −O7 −O6 −O8 −O5 Average a

FeNb11O29

Li11FeNb11O29

1.886(13) 1.932(14) 1.973(3) × 2 2.145(11) 2.291(14) 2.033 1.775(13) 1.862(17) 1.994(4) × 2 2.190(12) 2.384(12) 2.033 1.728(12) 1.969(3) × 2 1.989(18) 2.090(16) 2.417(11) 2.027

2.035(17) 2.037(15) 2.0433(5) × 2 2.124(19) 2.000(16) 2.047 1.972(21) 1.969(19) 2.068(3) × 2 2.339(16) 2.311(21) 2.121 2.116(28) 2.056(3) × 2 2.153(17) 1.937(17) 2.006(26) 2.054

Nb(Fe)4−O1 −O16 −O12 −O13 −O9 Average Nb(Fe)5−O14 −O2 −O13 −O10 −O12 Average Nb(Fe)6−O15 −O16 −O3 −O14 −O11 Average

FeNb11O29

Li11FeNb11O29

1.755(7) 1.848(12) 1.993(3) × 2 2.230(15) 2.205(13) 2.004 1.836(13) 1.952(8) 1.975(4) × 2 2.035(14) 2.228(11) 2.000 1.9134(5) × 2 1.9497(15) 1.987(12) 1.985(16) 2.077(15) 1.971

2.136(20) 1.926(19) 2.055(3) × 2 2.259(19) 1.862(25) 2.049 1.821(17) 2.074(10) 2.062(2) × 2 1.886(17) 2.354(15) 2.043 2.0436(7) × 2 1.925(16) 1.951(16) 2.179(21) 2.081(26) 2.037

E.s.d.’s are given in parentheses.

Table 4. Interatomic Li−O Distances (Å) of Li11FeNb11O29a Li1−O12 −O1 −O2 Average Li2−O16 −O3 −O1 Average Li3−O15 −O3 Average Li4−O7 −O8 −O4 Average a

1.969(22) × 2 2.063(20) × 2 2.21(14) 2.055 1.897(18) × 2 1.90(5) 2.43(4) 2.031 2.007(16) × 2 2.049(8) × 2 2.028 1.823(24) 2.142(9) × 2 2.391(26) 2.124

Li5−O15 −O11 −O7 Average Li6−O6 −O5 −O6′ −O7 Average Li7−O13 −O10 −O9 −O5 Average

1.84(4) 2.114(9) × 2 2.317(33) 2.096 1.65(7) 1.95(8) 2.095(16) × 2 2.31(8) 2.020 1.74(4) 2.066(6) × 2 2.22(4) 2.38(4) 2.094

Figure 6. CV current potential profiles of the LixFeNb11O29 electrode obtained in three electrode cell at 0.1 mV/s.

E.s.d.’s are given in parentheses.

These results are consistent with those of chemical lithiation, showing that electrochemical reversibility of the Li intercalation/deintercalation process is closely connected to preserving the crystallographic long-range order in the reaction with nbutyllithium. Taking in account the CV results, we decided to cycle the coin cells in the 1.1/2.5 V range limiting the cathodic capacity to 191 mAh/g. To test the material rate capability, the cycling was started at very low current density (7.7 mA/g, which corresponds to C/25 with respect to the final Li11FeNb11O29 stoichiometry) increasing the rate up to the highest value of 77.4 mA/g (C/2.5). The last 5 cycles were performed again at low current (9.7 mA/g, C/20). Charge−discharge potential capacity profiles at different current rates are reported in Figure 7, and they are in agreement with the first cycle behavior observed in the CV results. In fact, the capacity shows a pseudoplateau at about 1.6 V which corresponds to the main CV peak at 1.55 V, and then it slopes down to the cutoff value. The electrochemical behavior of the LixFeNb11O29 electrode can be better appreciated by plotting the differential capacity as function of the measured voltage (Figure 8). The first discharge process at the lowest current rate (black curve) shows a peak current which matches the first CV reduction scan (1.57 vs 1.55

Figure 5. Scanning electron micrograph of ball milled FeNb11O29 used for electrode fabrication in electrochemical cells.

formation of stable thin electrolyte layers on the electrode surface (solid electrolyte interface, or SEI). Such reactions underlying the large CV capacity do not seem to be fully reversible. During the first reverse anodic scan, indeed, only two peaks (at 1.28 and 1.70 V) are present with a charge of 245 mAh/g, pointing out the low reversibility of the material in the large potential range. The further cycling of the material reveals again irreversible changes in the electrochemical behavior. 2207

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials

Article

Figure 9. Cycling results of the LixFeNb11O29 electrode obtained in coin cell at different current rates between 2.5 and 1.1 V.

the first cycle and 96/97% for the following four cycles, pointing out the absence of SEI formation and the good reversibility of the intercalation process in this potential range. The specific capacity decreases as the current density increases; however, the Coulomb efficiency also increases, with an effect probably due to the presence of a parasite cathodic reaction. The analysis of differential capacity curves (Figure 8) reveals that the capacity is lost in the lowest potential range, between 1.4 and 1.1 V. In any case, the stable capacities values of 168, 155, and 140 mAh/g are observed at the C/10, C/5, and C/2.5 rates, with charge efficiencies of 98, 99, and almost 100%, respectively. On just considering the pristine materials, without taking into account any additional carbon coating treatments, the FeNb11O29 electrode shows better results in terms of both rate capability and capacity retention compared to the iron free Nb12O29 system.13

Figure 7. Charge/discharge curves in the 2.5−1.1 V range for the LixFeNb11O29 electrode, obtained in coin cell at different current rates.

4. CONCLUSIONS Lithium can be inserted into FeNb11O29 both (i) chemically, preserving a nice crystalline quality, and (ii) electrochemically, with very good reaction reversibility, up at least to the Li11FeNb11O29 composition. Further lithiation can be accomplished to a much larger extent, but with (i) loss of crystallinity or (ii) deterioration of reversibility. The stable reversible electrochemical capacity of 185 mAh/g obtained improves the 160 mAh/g value of Nb12O2913 by 16%, even more than what was expected on the basis of the 11/10 ratio of Nb5+ atoms per f.u. Better performances are shown also in terms of current rates and cyclability, confirming that FeNb11O29 is an interesting candidate as anode material in lithium ion batteries. An accurate study by neutron diffraction of the chemically lithiated Li11FeNb11O29 phase has allowed us to understand the detailed Li atom distribution in its crystallographic-shear structure. Six Li(IV) and five Li(V) atoms per f.u., with coordination numbers 4 and 5, respectively, have been revealed, indicating a very slight preference for the higher coordination number with respect to the average distribution. The occupied Li sites do not lie exclusively on the periphery of the 3 × 4 octahedra blocks (Li(V)), as previously assumed,21,13 but they also lie within the blocks (Li(IV)). Thus, a tendency of Li atoms to distribute rather uniformly over the available sites appears clearly.

Figure 8. Differential capacity plots of the of the LixFeNb11O29 electrode derived from the charge discharge curves at different current rates: C/25 (black), C/10 (red), C/5 (green), C/2.5 (blue), and C/20 (magenta).

V, respectively). From the second cycle on, and at higher current rates, the cathodic peak is slightly shifted to higher potential. No additional dQ/dV peaks are observed in the FeNb11O29 electrode; this is a remarkable difference compared to the behavior of the iron free Nb12O29 material, showing sharp peaks at 1.18, 1.33, and 1.45 V.13 Finally, in the anodic branch of the curve a secondary peak is observed at high potential (around 2.1 and 2.2 V) whose intensity and position depend on the current rate. The cycling results are summarized in Figure 9. At the lowest current value, the theoretical specific capacity of 191 mAh/g was reached during the cathodic scans, while during the anodic ones the electrode delivered a slightly lower stable capacity of about 185 mAh/g. The resulting Coulomb efficiency is 99% for 2208

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209

Chemistry of Materials



ASSOCIATED CONTENT



AUTHOR INFORMATION

Article

S Supporting Information *

CIF files. This material is available free of charge via the Internet at http://pubs.acs.org. Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank Antonella Scherillo and Kevin Knight (ISIS Facility, Rutherford Appleton Laboratory, U.K.) for help with the neutron diffraction measurements.



REFERENCES

(1) Han, J.-T.; Goodenough, J.B. J.B. Chem. Mater. 2011, 23, 3404− 3407. (2) Ozhuku, T.; Ueda, A.; Yamamoto, N. J. Electrochem. Soc. 1995, 142, 1431−1435. (3) Kavan, L.; Procházka, J.; Spitler, T. M.; Kalbác, M.; Zukalová, M.; Drezen, T.; Grätzel, M. J. Electrochem. Soc. 2003, 150, A1000−A1007. (4) Colbow, K. M.; Dahn, J. R.; Haering, R. R. J. Power Sources 1989, 26, 397−402. (5) Cava, R. J.; Murphy, D. W.; Zahurak, S. M.; Santoro, A.; Roth, R. S. J. Solid State Chem. 1984, 53, 64−75. (6) Kumagai, N.; Koishikawa, Y.; Komaba, S.; Koshiba, N. J. Electrochem. Soc. 1999, 146, 3203−3210. (7) Sayagués, M. J.; Hutchison, J. L. J. Solid State Chem. 1999, 146, 202−210. (8) Waldron, J. E. L.; Green, M. A.; Neumann, D. A. J. Phys. Chem. Solids 2004, 65, 79−86. (9) McQueen, T.; Xu, Q.; Andersen, E. N.; Zandbergen, H. W.; Cava, R. J. J. Solid State Chem. 2007, 180, 2864−2870. (10) Kato, K. Acta Crystallogr., B 1976, 32, 764−767. (11) Andersen, E. N.; Klimczuk, T.; Miller, V. L.; Zandbergen, H. W.; Cava, R. J. Phys. Rev. B 2005, 72, 033413. (12) Cheng, J.-G.; Zhou, J.-S.; Goodenough, J. B.; Zhou, H. D.; Wiebe, C. R.; Takami, T.; Fujii, T. Phys. Rev. B 2009, 80, 134428. (13) Li, Y.; Sun, C.; Goodenough, J. B. Chem. Mater. 2011, 23, 2292−2294. (14) Trunov, V. K.; Kovba, L. M.; Pol’shchikova, Z. J. Zh. Neorg. Khim. 1968, 13, 1494−1499. (15) Brunner, H.; Gruehn, R.; Mertin, W. Z. Naturforsch., B 1976, 31, 549−553. (16) Tabero, P. Ceram.-Silik. 2005, 49, 126−131. (17) Larson, A. C.; Von Dreele, R. B. General Structure Analysis System (GSAS), Los Alamos National Laboratory Report LAUR 2004; 86−748. (18) Von Dreele, R. B.; Jorgensen, J. D.; Windsor, C. G. J. Appl. Crystallogr. 1982, 15, 581−589. (19) Thompson, P.; Cox, D. E.; Hastings, J. B. J. Appl. Crystallogr. 1987, 20, 79−83. (20) Stephens, P. W. J. Appl. Crystallogr. 1999, 32, 281−289. (21) Cava, R. J.; Murphy, D. W.; Zahurak, S. M. J. Electrochem. Soc. 1983, 130, 2345−2351. (22) Catti, M.; Ghaani, M. R. Phys. Chem. Chem. Phys. 2014, 16, 1385−1392.

2209

dx.doi.org/10.1021/cm500442j | Chem. Mater. 2014, 26, 2203−2209