Investigation of the Chemical Disorder of LiNi0. 5Mn1. 5O4 Lattice by Means of Extended X-ray Absorption Fine Structure Spectroscopy

Investigation of the chemical disorder of

LiNi0.5Mn1.5O4 lattice by means of EXAFS

G. Greco,∗,† S. Brutti,¶ F. M. Vitucci,‡ L. Lombardo,† M. Köntje,‖ A. Savoini,⊥ A.

Paolone,‡ and S. Panero.†

Chemistry Department, Sapienza University of Rome, Piazzale A. Moro 5, I-00185 Roma, Italy,

Research Center Hydro-Eco, Sapienza University of Rome, Via A. Scarpa 14, I-00161 Roma,

Italy, Department of Science, University of Basilicata, V.le Ateneo Lucano 10, 85100 Potenza,

Italy, CNR-ISC, UOS Sapienza, P.le A. Moro 5, I-00185 Rome, Italy, ZSW âAS Zentrum fÃijr

Sonnenenergie und Wasserstoffforschung, Baden-WÃijrttemberg, Helmholtzstrasse 8, D-89081

Ulm, Germany, and Research Center for Non Conventional Energies, Istituto ENI Donegani, ENI

S.p.A. via Fauser 4, I-28100 Novara, Italy

E-mail: [email protected]

∗To whom correspondence should be addressed†Chemistry Department, Sapienza University of Rome, Piazzale A. Moro 5, I-00185 Roma, Italy‡Research Center Hydro-Eco, Sapienza University of Rome, Via A. Scarpa 14, I-00161 Roma, Italy¶Department of Science, University of Basilicata, V.le Ateneo Lucano 10, 85100 Potenza, Italy§CNR-ISC, UOS Sapienza, P.le A. Moro 5, I-00185 Rome, Italy‖ZSW âAS Zentrum fÃijr Sonnenenergie und Wasserstoffforschung, Baden-WÃijrttemberg, Helmholtzstrasse 8,

D-89081 Ulm, Germany⊥Research Center for Non Conventional Energies, Istituto ENI Donegani, ENI S.p.A. via Fauser 4, I-28100 Novara,

Italy

1

Abstract

This work reports on a characterization methodology to investigate the transition metal

site disorder in the LiNi0.5Mn1.5O4 spinel structure (LNMO,Fd3m space group,cF56 lat-

tice). This approach, previously adopted on simple intermetallic ordered solid solutions, is

here applied to a complex lattice. The LNMO material synthesized by us has been investigated

by advanced techniques such as XAS (X-ray Absorption Spectroscopy), XRD (X-ray Diffrac-

tion), SEM (Scanning Electron Microscopy) and high resolution TEM (Transmission Electron

Microscopy). High-quality XAS spectra measured at both theNi K-edge and the Mn K-edge

have been analyzed using double-edge multiple-scatteringdata analysis. Computer simula-

tions, based on structural lattice models describing substitutional disorder between Ni and Mn

in the 16d site of the LNMO lattice, have been carried out by varying theorder parametersof

the site occupancy from 0 (completely disordered) to 1 (completely ordered). These simula-

tions allow to predict the evolution of the two- and three-body structural parameters versus the

order parameters. The chemical ordering in the transition metal 16d atomic site of the prim-

itive LNMO cell has been then analyzed by coupling computer simulations and EXAFS data

analysis results. The EXAFS signals are sensitive to the surrounding substitutional disorder

and their intensities, especially those of the collinear three-atom configurations, can be used

to evaluated the ordering level of the transition metals in the lattice. The final refined order

parameter value (s= 0.0) suggests that only negligible ordering occurs in thislattice between

Ni and Mn and the structure is therefore completely disordered.

Introduction

Material science (sensu lato) literature is rich in papers dealing with the structural, physical and

chemical properties of phases belonging to the spinel structure type. End members and solid solu-

tions within these chemical systems are easily synthesized, and due to their physical and electronic

properties, they are very widely used as technological materials, for examples in magnetic record-

ing media, lithium ion batteries, catalysts and pigments.1

2

LiM xMn2−xO4 spinels phases with x= 0.5; M = Mg, Cr, Mn, Fe, Ni, Zn2–4 have recently

attracted huge interest as innovative cathodes of lithium batteries: they are environmental friendly,

low cost and have a good theoretical properties in lithium cells. In particular, LiNi0.5Mn1.5O4 is

able to reversibly cycle lithium in electrochemical cells for hundreds of times at high current rates,5

and use in full lithium ion cells has been demonstrated.8,9 It displays a high charge/discharge

potential of about 5V with minor irreversible losses upon cycling.6

The synthesis procedure deeply influences both the crystal structure and the performances of

this compound.5,6,10,14–17It is known that samples synthesized above 700oC have a face-centered

cubic spinel structure (Fd3mspace group,cF56 lattice). In addition, all materials produced above

650oC display a lower Ni content than the nominal one and contain,as a impurity, a LiNi1−xO1−x/2

phase (Fd3m space group,cF8 lattice).10 The crystallization of this undesired contaminant has

been observed by many authors in the literature and it is commonly recognized as an apparently

unavoidable byproduct of all syntheses carried out above 700oC (see discussion in ref.19,27). How-

ever, despite this drawback, high temperature synthesis are usually used, because the growth of

the spinel phase at low temperatures is extremely slow.20,22,25There is a general consensus about

the random distribution of the manganese and nickel ions in the 16d sites in theFd3m lattice. On

the other hand prolongued annealing below 700oC, after the first calcination at higher temeprature,

leads to the ordering of Mn and Ni and the formation of a superstructure, thus transofrming the

Fd3m face centred structure (disordered) to the primitive cubiclattice P4332 (ordered).23,24 In-

frared spectroscopy is the more used method to distinguish these two structures and can be used

qualitatively to evaluate ordering percentage in spinel which contains both ordered and disordered

LiNi 0.5Mn1.5O4.25

The Li intercalation path is greatly influenced by the Ni and Mn ordering that form an electro-

static repulsion. This repulsion is is different in the two ordered and disordered phases, so the Li

migration activation barriers will be different from each other. The previous studies showed that

disordered LiNi0.5Mn1.5O4 exhibited better cycling performance than ordered LiNi0.5Mn1.5O4 at

high rates26,27 A synthesis procedure for LiNi0.5Mn1.5O4 (LNMO from here on) has been de-

3

veloped in our lab: the samples obtained by wet chemistry, starting from acetate precursors and

synthesized at 800oC,5,6 shows the lowest contamination, the highest crystallinityand the best

electrochemical response in term of stability and high currents rate in a Li-ion cell (100 mAh g1

for more than 400 cycles at 146 mA g1 (1C)).5

Combining different techniques including x-ray absorption spectroscopy (XAS), x-ray diffrac-

tion (XRD), scanning electron microscopy (SEM) and high-resolution transmission electron mi-

croscopy (TEM), in this paper we present a deep quantitativestructural characterization never

performed up to now in a complex structure such as LiNi0.5Mn1.5O4 spinel. In particular we ad-

dress the issue of chemical disorder in the 16d sites (0.6250, 0.6250, 0.6250), as both Ni and Mn

ions reside on these positions.

Our goal is to clarify once and for all the atomic structure and so the influence of chemical

disorder on the electrochemical performances of this complex material also giving an effective

quantitative characterization method that can be applied in principle to every active material. In-

strumental for this task was the use of EXAFS technique. In fact it has been shown that the EXAFS

signals due to multiple scattering between different atomsoccupying identical crystal sites (partial

occupancies) allow the estimate of the probability to find the same chemical species on the same

position in each unit cell.28–32 Double edge EXAFS analysis coupled with computer simulations

has been recently used in Ref.33 to study the chemical disorder of a PtxCo solid solution. This

approach is apparently capable to disclose unique quantitative information about the distribution

of different chemical species among equivalent crystallite sites. Here we extend this model to a

complex cubic lattice with 56 atoms (8 Li, 12 Mn, 4 Ni and 32 O) in the face-centered unit cell.

The paper is organized as follows: first section is dedicatedto sample characterization by using

TEM and XRD techniques. XAS experimental details and preliminary data are reported in sec-

ond section . Third section is focused on the methodology of the XAS data analysis using an

advanced technique basing on the GNXAS method34,35 and statistical computer simulation. In

fourth section the main conclusions of this work are given.

4

Experimental details and preliminary characterization by diffrac-

tion and electron microscopy

LiNi 0.5Mn1.5O4 spinel samples have been synthesized accordingly to ref.5 by our standard wet

chemistry route followed by high temperature annealing in air at 800oC degrees. Conventional

CuKα X-ray diffraction (Rigaku Ultima+ diffractometer) confirmed the formation of thecF56

structure with a minor contamination by LiNi1−xO1−x/2 (∼2.7 wt.%) due to the precipitation of

the expectedcF8 phase.5 The composition of the sample has been confirmed by elementalanaly-

sis using inductively coupled plasma mass spectrometry (ICP-MS) on solid pellets of the starting

material. A Li:Mn:Ni ratio of 1.01:1.50:0.51 with an estimated accuracy of 0.01 on each stoi-

chiometric coefficient. The structural refinement by Rietveld method carried out by the GSAS

program11,12 following a constrained approach proposed by as in ref.5 The overall stoichiometry

of the sample has been assumed from the ICP-MS data; similarly the Mn:Ni and Li:Ni ratios in

thecF56 andcF8 phases, respectively, have been postulated by a preliminary determination of the

cubic lattice parameter of the two phases on the basis of the benchmark data by Dahn and cowork-

ers10 and Toussaint.13 Having kept constant all these parameters with the additional assumption

of an unitary occupancy of the oxygen atoms in the atomic sites of both thecF8 andcF56 lat-

tices, the lithium stoichiometric coefficient in thecF56 lattice results by the sole optimization by

Rietveld method of the relative amount ofcF8 phase in the sample. This indirect and constrained

approach is applied in order to circumvent the limitations of the XRD technique to correctly eval-

uate compositions in particular in the case of similar heavynuclei (like manganese and nickel) or

for light elements (like lithium). As a final point it is important to underline that in the refine-

ment the Debye-Waller factor have been assumed equal to B=1 Å2 for all the atomic species in all

lattices. Our constrained Rietveld refinement of the XRD pattern suggests for the LNMOcF56

lattice a cubic parameter of a=8.172±0.002 Å, a slight lithium deficiency, a final stoichiometry

Li0.98Mn1.54Ni0.46O4 with estimated s=0.02 inaccuracies on the stoichiometric coefficients. The

estimated manganese oxidation state estimated from this stoichiometry is 3.96 to be compared with

5

the 3.98 value derived by electrochemical analysis (see Refs.,5 6). Scanning electron microscopy

(SEM, JEOL JSM 7600F) investigation suggests an highly homogeneous morphology of the sam-

ple (see fig. Figure 1(a)): it is constituted by well formed octahedral-like crystallites with regular

shapes and sharp edges. The image analysis of 100 randomly selected quasioctahedron shaped

particles observed in several SEM images has been carried out by the ImageJ program36 in order

to derive the mean particle geometrical properties. Crystallites have been assumed to be regular

octahedral and the reported size distribution plot shows the diameters of the corresponding outer

sphere (Fig. Figure 1(b)). The crystallites size distribution is peaked at approximately 300nm:

as expected it is asymmetric and shows a tail extended to larger-sized particles. The apparent

mean crystallite size in about 400 nm. The size distributioncan be reproduced rather accurately

through a log-normal model.37,38The size distribution maximum and mean values (dmaxand dmean,

respectively) and the distribution broadening (σ) are summarized in Tab. Table 1. TEM investi-

gation (JM-3010 HR TEM) confirmed the long range crystallineordering of the particles (Fig.

Figure 1c). Fast Fourier Transform (FFT) analysis of several HR (high resolution) TEM images

confirmed the formation of a face centered cubic lattice witha cell parameter of approximately

7.9-8.0 Å (Tab. Table 1).

Table 1: The lattice parameter (a), the mean (dmean) and the maximum (dmax) diameter ofcrystallites, the broadening of the diameter distribution (σ); oxidation state of manganeseion (Os) and the ratio between Mn and Ni atoms (Mn:Ni).

LNMO a dmax dmean σ Os Mn:Ni×100 ×100 ×100

[Å] [nm] [nm] [nm]XRD 8.172(2) – – – 3.96 1.50 : 0.51SEM, TEM 7.9(5) 3(1) 4(1.5) 1.2(5) – –XAS – – – – 3.93 1.57:0.43

6

a

b

c

Figure 1: (b) metallic particles size distribution obtained from the analysis of the SEM images (a)of LiMn2xNixO4. (c) The high resolution TEM (HRTEM) image of the crystalline agglomeratesand in the inset the fast Fourier Transformation of HRTEM.

7

XAS experimental details and results

EXAFS measurements have been performed in transmission mode at room temperature at the

beamline XAFS of the ELETTRA Synchrotron Light Laboratory (Trieste, Italy) using a double

crystal monochromator equipped with a Si(111) crystal.39 The sample powder has been mixed

with graphite, in the proper ratio in order to give an optimized absorption step of∼1.1 at the Mn

K-edge and∼0.4 at the NiK-edge.34,35 The same pellet has been used for acquisition at both

absorption edges in order to avoid possible inhomogeneities of the powders.

The raw XAS data at the Mn and NiK-edges are reported in Fig. Figure 2. The XANES (X-

ray Absorption Near Edge Spectroscopy) signals intensity after removal of the pre-edge absorption

and without normalization (Fig. Figure 2(a)) are directly proportional to the Mn and Ni quanti-

ties in the sample. The MnK-edge XANES signal is typical of the spinel phase with a prepeak

features between 6540 eV and 6545 eV.7 This preedge features result from transitions to bound

final state in 3d orbital and are proportional to the Mn oxidation state:7 the obtained stoichiometry

and oxydation state is reported in Tab. Table 1, in good agreement with XRD and electrochemical

results.

The analysis of the XAS signals has been performed by means ofthe GNXAS program.34,35

The k weighted EXAFS oscillations, kχ (k) have been extracted from the experimental data and

are shown in Figure Figure 2(b). Good signals have been obtained up to 15 and 13.5 Å−1 for the

Mn and Ni K-edge, respectively. The k weightedχ (k) data have been Fourier transformed (FT)

in the k range between 3.5 and 14.9 Å−1 for the Mn K-edge and between 3.5 and 13.4 Å−1 for

the Ni K-edge. The Fourier transform obtained shows many correlation peaks up to R = 6 Å,

as expected for a crystalline structure. The FT signals are reported in Figure Figure 2(c). In the

following analysis we do not take in to account the secondaryLi xNi1−xO1−x/2 (cF8) phase. Indeed

the results obtained on the Mn edge is not affected by this impurity phase because thecF8 does

not contain manganese. Moreover the content of the secondary phase is so low (≤4%) that also

the NiK-edge is only very weakly influenced by the presence of the this contamination. It is likely

that an increase of the Debye-Waller factor could occur, as in the case of manganites with a low

8

concentration (≤2%) of spurious Mn3O4.40

(a) (b)

(c)

Figure 2: (a) Comparison of Mn K and NiK near edge XAS spectra of LNMO powderEo - edgeenergy. Experimentalχ(k) signals and its FT of LNMO powder (b) and (c) respectively, for - MnK-edge, and - Ni K-edge

Therefore a good starting point for the analysis of the EXAFSdata is the bulk structure of the

spinel LiNi0.5Mn1.5O4, (cF56 cell, cubic lattice parameter a=8.16 Å and space groupFd3m, #227)

in which the lithium ions are located at the 8a sites(0,0,0), the oxygen atoms reside on the 32e

positions(0.3792,0.3792,0.3792), and the manganese and nickel ions are distributed in the 16d

sites(0.6250,0.6250,0.6250).

9

Computer simulation to model EXAFS signal of disordered struc-

ture

Recently, it has been proven that the analysis of the scattering of triangular configurations is a

valuable tool to evaluate the ordering of simple cubic alloys, such as Pt3Co.33 In particular the

presence of chemical disorder affects the geometrical parameters such as the coordination numbers

of neighboring atoms Here we discuss the extension of such analysis method to a more complex

structure compared to a simple cubic alloy in order to investigate capabilities and limits of this

approach. In order to evaluate the effect on the EXAFS signals of the chemical disorder on the

16d crystal site of the spinel structure, occupied alternatively by the two transition metals (Mn

and Ni), we focused our attention only on the Mn/Ni sublattice in the conventional unit cell (cF56

lattice, cubic lattice parameter a=8.16 Å , space groupFd3m, metal atoms in the 16d atomic site

in (0.625,0.625,0.625)). In Fig. Figure 3(a) the hypothetical ordered cell made by 16 atoms of Mn

and Ni is shown, the atomic ratio Mn:Ni is 3:1. Substitutional disorder is simulated introducing

a finite probabilityp for the occupancy of selected lattice sites with Ni atoms. Inparticular,p

is taken as the probability that a Ni atom occupies the same site position in each unit cell, see

Fig. Figure 3(a). The range 0.25≤ p≤1 describes all possible levels of chemical disorder, spanning

from a completely ordered structure (p=1) to a completely disordered one (p=0.25, for which no

preferential position can be defined for Ni). We calculate the multiplicity and coordination numbers

of the main two- and three- body contributions to the EXAFS signals as a function of chemical

disorder, by means of a computer simulation scheme, previously discussed in Ref.33 The effect of

the finite size of the calculation have been minimized by considering a supercell 4×4×4 replica

of the conventional spinel cell, which contained 64 cubic structures and a total number of 1024

atoms. For each supercell we preserve the stoichiometry, maintaining the ratio Mn:Ni 3:1. In the

following the results of the computer simulations are presented introducing the order parameter

sfor the disorder, defined as

s=p−Ca

1−Ca, (1)

10

where p, is the occupation probability defined above andCa is the atomic concentration of the

selected species (Ca =1/4 for Ni). Therefores= 0 corresponds to the random distribution of Mn

and Ni atoms in the 16d sites, whiles= 1 describes a perfectly ordered crystal. We calculate

by computer simulations the coordination numbers and degeneracy of the two- (g2) and three-

body (g3) configurations on 3000 different supercells, which corresponds to 16 differents values

spanning from 0≤ s≤1, with steps ofs= 0.0625. In Fig. Figure 3(b), (c) and (d) we report

three supercells, which are representative of three chemical ordering: s= 1.00,0.45, and 0.00.

Moreover, in Table Table 2 and Table 3 we report the parameters of the two- and three-body

configurations for the ideal ordered structure.

Table 2: Two-atom (g2) configurations in a hypothetical Mn-Ni ordered structure (only 16dpositions in a LNMO Fd3m spinel structure have been considered, see Fig. and Figure 3aand b): R1 is the mean distance between atoms Mn-M (Mn or Ni) andR2 is the mean distancebetween atoms Ni-M (Mn or Ni), NC is the number of neighbors M.

Atoms Shell Nc R1 [Å] Atoms Shell Nc R2 [Å]Mn-Mn I 4 2.86 Ni-Mn I 5 2.86Mn-Ni I 2 2.86 Ni-Ni I 1 2.86Mn-Mn II 10 4.96 Ni-Mn II 10 4.96Mn-Ni II 2 4.96 Ni-Ni II 2 4.96Mn-Ni III 4 5.72 Ni-Mn III 8 5.72Mn-Mn III 8 5.72 Ni-Ni III 4 5.72Mn-Ni IV 4 6.40 Ni-Mn IV 10 6.40Mn-Mn IV 8 6.40 Ni-Ni IV 2 6.40

In Fig. Figure 4 the calculated coordination numbers (Nc) for the first (c=1), second (c=2),

third (c=3) and fourth (c=4) shellsg2 are reported as a function ofs. In Fig.Figure 5 we report

the calculatedg3 coordination numbers (N3(s)) which corresponds to the scattering of three atoms

connected by two bonds forming an angle of 120o or 180o. These simulations highlight that the

chemical disorder affects the coordination numbers of the transition metal sublattice. In particular

at s<0.4 (order structures) it is possible to find marked differences in theg2 Nc(s), even for little

variation of s, whereas the coordination numberN3(s) of the three body configurationsg3 can

probe even the limited ordering ats<0.4. In Fig. Figure 6 the calculated FT both at the Mn and

11

a

b

c

d

Figure 3: Structural representation of: (a) Mn-Ni primitive cell (only 16d sites are shown); (b),(c)and (d) three different Mn (blue) Ni (yellow) supercells with a cubic structure and with a differentdegree of chemical disorder,s=1.00 ordered,s=0.45 ands=0.00 random structure, respectively.

12

Table 3: Three-atom (g3) configurations in an hypothetical Mn-Ni ordered structure (only 16d posi-tions in a LNMO Fd3m spinel structure have been considered, see Fig. Figure 3 (a)and (b) a for Mnand Ni K-edges: Deg. – degeneracy specified for each configuration, with the photoabsorber placedat the vertex between two shortest bonds (e.g. Mn-Ni-Ni: Mn atom is at the vertex); R01, R02 – themean distances of two shortest bonds with angleΘ in between; R03 – the corresponding atoms andbelonging shell; Freq. (frequency)– half of the length of the triangle perimeter, corresponding to thepeak leading frequency in Fourier spectrum. phN – the photoabsorber is positioned in position N(side of the triangle, not the triangle vertex).

Atoms types Deg. R01 = R02 R03 Θ Freq.[Å] [Å] [deg] [Å]

Mn KNi-Mn-Mn(ph2) 4 2.86 Mn-Mn I shell 60 4.29Mn-Mn-Mn 2 2.86 - 60 4.29Mn-Ni-Ni 1 2.86 - 120 5.34Ni-Ni-Mn(ph2) 1 2.86 Mn-Ni II shell 120 5.34Mn-Ni-Mn(ph2) 2 2.86 Mn-Ni II shell 120 5.34Mn-Mn-Mn 3 2.86 - 120 5.34Mn-Mn-Mn(ph2) 7 2.86 Mn-Mn II shell 120 5.34Ni-Ni-Mn(ph2) 1 2.86 Mn-Ni III shell 180 5.72Ni-Mn-Mn(ph2) 1 2.86 Mn-Mn III shell 180 5.72Mn-Ni-Mn 2 2.86 Mn-Ni III shell 180 5.72Mn-Ni-Mn 1 2.86 Mn-Ni III shell 180 5.72Mn-Mn-Mn 1 2.86 - 180 5.72Mn-Mn-Mn(ph2) 1 2.86 Mn-Mn III shell 180 5.72Ni KNi-Ni-Mn 2 2.86 - 60 4.29Ni-Mn-Mn 4 2.86 - 60 5.34Ni-Ni-Mn 2 2.86 Ni-Mn II shell 120 5.34Mn-Ni-Ni(ph2) 2 2.86 Ni-Ni II shell 120 5.34Ni-Mn-Mn 4 2.86 - 120 5.34Mn-Ni-Mn(ph2) 8 2.86 Ni-Mn II shell 120 5.34Ni-Ni-Mn 1 2.86 - 180 5.72Ni-Ni-Mn(ph2) 1 2.86 Ni-Mn III shell 180 5.72Mn-Ni-Ni(ph2) 2 2.86 Ni-Ni III shell 180 5.72Ni-Mn-Mn 2 2.86 - 180 5.72Mn-Ni-Mn(ph2) 3 2.86 Ni-Mn III shell 180 5.72

13

Figure 4: Two-body coordination numbers as a function ofs (chemical order) parameter for theLNMO phase, calculated in this work. Top-left – first, top-right – second, bottom-left – third andbottom-right – fourthg2 shells.

Ni edges are shown for four different values of the order parameters. It can be noted that the peak

related to the Mn-M path (M = Mn or Ni) centered around 2.9 Å andthe peak around 5.6 Å which

is due to the three body scattering are the features more affected by the chemical disorder.

EXAFS data analysis and discussion

The fitting of the EXAFSχ(k) signals has been performed considering the two- and three-body

main contribution listed below. The other possible scattering paths give much weaker signals and

have been therefore neglected:

• First and second neighbors: in the range of distances between 1.5 and 3.2 Å the two-body

scattering due to the Mn-O or Ni-O (I shell) and to the Mn-Mn, Mn-Ni, Ni-Mn and Ni-Ni (II

shell) paths are modeled by usual gaussian distributions. The coordination numbers (Nc) for

14

Figure 5: Calculated degeneracy (Nc) for three-body configurationsg3 with 120o 180o vertex angleas a function ofs parameter.

Figure 6: (Fourier Transforms of the Mn and Ni K theoretical XAS signals calculated for differentvalues of the disorder parameters.

the four contributions of the II shell have been allowed to vary according to the chemical or-

der parameter s, imposing, however, the constraint that thesum of the coordination numbers

of the Mn-Mn (Ni-Ni) and Mn-Ni (Ni-Mn) shells equals 6.

• Third neighbors: we consider both the two-body and three-body signals due to the scattering

of the Mn-O-O and Ni-O-O triangular configurations formed bytwo first neighbor bonds.

The value of the angle formed by the bonds was imposed to be 180o. Also the coordination

number and the degeneracy of these paths have been fixed in theanalysis.

• Fourth and fifth neighbors: we include both the two-body and the three-body signals using

15

Ni-Ni-Ni, Ni-Ni-Mn, Ni-Mn-Mn, and Mn-Mn-Mn triangular configurations formed only by

two first-neighbor bonds with bond angles of 120o and 180o (fixed values), as the signals

related to 60o configuration are negligible.Nc and degeneracy of the triangular configura-

tions have been varied according to the chemical order parameters (see Figs Figure 4 and

Figure 5). Diagonal covariance matrices for the thermal average have been used and only

the bond-angle-varianceσ(Θ) was introduces as an additional fit parameter.35,41

The best EXAFSχ(k) fitting results are shown in the Fig. Figure 7 and Figure 8 the best fitting

parameters are reported in the Table Table 4.

Figure 7: The best-fit results of GNXAS analysis performed for LiNi 0.5Mn1.5O4 at Mn (left) andNi K-edge (right) (k3 weighted). Dashed line – the experimental signal, solid line – sum of alltheoretical components of the signal. Upper curves represent components of the model signal,vertically shifted for clarity; solid line – model signal, dashed line – experimental signal. Thescale is indicated within the figure.

The values of the distances Mn-O and Ni-O of the first shell andof the Mn-Mn and Mn-Ni

distances of the second shell are slightly lower (∼ 0.02 Å) than the values obtained in a previous

study.42 Bond-angles standard deviations have been found in a range between 0-1o whereas the

bond-bond correlation of the collinear configuration is negativeρr, r ′ ∼ −0.26 as found in other

systems.32,33The bond variance,σ2, around Mn atoms is in good agreement with previous works,

16

Table 4: Best fit structural parameters of the two-body configurations of the LiNi 0.5Mn1.5O4spinel structure. Nc is the coordination number, R is the average interatomic distance,σ2 isthe distance variance. Estimated errors are shown in the brackets on the last digit.

LiNi 0.5Mn1.5O4

Atoms shell Nc R [Å] σ2 [Å2]Mn-O I 6.00 1.900(5) 0.0030(5)Mn-Mn II 4.48 2.887(5) 0.0034(2)Mn-Ni II 1.50 2.890(5) 0.0042(2)Mn-O IIIa 3.00 3.797(2) 0.0064(2)Mn-Mn IVa 8.93 5.000(2) 0.0085(2)Mn-Ni IV a 3.05 5.002(3) 0.0091(1)Mn-Mn Va 9.09 5.774(2) 0.0050(2)Mn-Ni Va 2.90 5.773(4) 0.0083(1)Ni-O I 6.00 2.030(1) 0.0050(5)Ni-Mn II 4.40 2.890(5) 0.0042(2)Ni-Ni II 1.56 2.89(1) 0.0058(3)Ni-O III a 3.00 4.062(5) 0.0100(5)Ni-Mn IV a 9.00 5.002(3) 0.0091(1)Ni-Ni IV a 2.98 5.004(2) 0.0098(5)Ni-Mn Va 8.59 5.774(2) 0.0050(2)Ni-Ni V a 2.90 5.776(6) 0.0092(2)

a Parameter derived from the three-atom configuration.

17

Figure 8: Fourier Transforms of the EXAFS experimental signals collected at Mn (left) and Ni K-edge (right) compared with theoretical ones calculated with chemical disorder parameters= 0.0.

i.e. σ2=0.0034-0.0050 Å2 for Mn-O and Mn-Mn first and second shell signals.7,42 Around the Ni

atom, the variance isσ2=0.005-0.006 Å2 for both Ni-O and Ni-Ni bond. These values are larger

compared to literature (σ2(Ni-O)=0.0039 Å2 andσ2(Ni-Ni)=0.0023 Å2),43 probably due to the

presence of the secondary LixNi1−xO1−x/2 phase described in the first section . The same holds

for the second and higher shells where the distance variances approach the uncorrelated limit of

2u2 = 0.01 Å2, where u2 is the fluctuation around the equilibrium positions. As discussed before

the increase of the Debye-Waller factor was already observed in other compounds with secondary

phases, such as in manganites with a low concentration (<2%) of spurious Mn3O4.40

Turning to the chemical disorder the best fit results are reported in Tab. Table 4, the results

obtained are compatible with the order parameters=0.0±0.4, completely disordered, as expected.

Looking at Figs. Figure 4 and Figure 5 the coordination numbers are weakly dependent on the

order parameter at low disorder level except for the tree body configuration 180o at Ni K-edge,

unfortunately weak signal (see Fig. Figure 7). In Table Table 5 theResvalues for constrained

fittings with fixed order parameters s are shown, beingResthe sum of the residuals calculated

between models and experiments.44 Again the best fitting is observed fors= 0.0. In summary the

16d atomic site in the LNMO unit cell is apparently occupied randomly by Ni or Mn atoms without

even small ordering in surrounding unit cells. In the Fig. Figure 8 the FT of the EXAFS signals

collected at the Mn and Ni K-edges are compared with the theoretical signal fors=0.0: theoretical

18

and experimental data are in very good agreement.

Table 5: Resis the residual expected value obtained for different orderparameterss.

s Res× 10−7

1.00 1.840.45 1.740.20 1.720.00 1.63

Conclusions

In this work, the complex local structure of nickel doped cubic manganese spinel have been inves-

tigated using XAS double-edge multiple-scattering GNXAS refinement and considering pair and

triplet contributions associated with distances and bond angles up to the fifth coordination shell.

Crystalline structure, size distribution and stoichiometry of the crystallites has been obtained by

combining SEM, TEM, XRD and XAS results.

The EXAFS spectrum of LiNi0.5Mn1.5O4 phase has been analyzed giving particular relevance

to the Mn and Ni atoms distribution in the spinel structure (16d sites). Computer simulations

based on a simple model accounting for substitutional disorder have been carried out calculating

the variation of the intensity of the individual pair and triplet contributions to the XAS spectra

taking in account only for the Mn and Ni atoms in the primitivecell, as a function of an order

parameters (s= 0 indicates random occupancy of lattice sites ands= 1 perfect ordering). The

Ni and Mn in the LiNi0.5Mn1.5O4 phase has been found to be chemically disordered (s = 0.0±

0.4), in agreement with previous estimates. Chemical disorder has been evaluated with very good

accuracy by the XAS signals associated with collinear triplet configurations including Ni atoms.

In summary here we demonstrate the extension to a complex structure of the model proposed in

the Ref.33 for the evaluation of the chemical disorder between different metals occupying the same

atomic crystal site. This approach is based on double edge EXAFS data coupled with computer

19

simulations of model structures from completely ordered tocompletely disordered. Our analysis

suggests that such robust approach is suitable for the understanding of atomic disorder in crystals

also in complex oxide structures such as the LNMO lattice, thus going beyond intermetallic solid

solutions.

Acknowledgement

The results of this work have been obtained thanks to the financial support of the European Com-

munity within the Seventh Framework Programme APPLES (Advanced, High Performance, Poly-

mer Lithium Batteries for Electrochemical Storage) Project (contract number 265644). We grate-

fully acknowledge the support of the Synchrotron Light Laboratory ELETTRA and the entire

beamline staff Giuliana Aquilanti, Luca Olivi (XAFS 11.1 station).

References

(1) C. M. B. Henderson, J. M. Charnock, D. A. Plant,J. Phys.: Condens. Matter2007, 19,

076214

.

(2) S. Mandal, R. M. Rojas, J. M. Amarilla, P. Calle, N. V. Kosova, V. F. Anufrienko and

J. KRojo,M. Chem. Mater.200214, 1598

.

(3) H. Shigemura, M. Tabuchi, M. Kobayashi, H. Sakaebe, A. Hirano and H. Kageyama,J. Mater.

Chem.200212, 1882

.

(4) R. Alcantara, R. Jaraba, M. Lavela and J. L. Tirado,J. Mater. Chem.200315, 1210

.

(5) L. H. Chi, N. N. Din, S. Brutti and B. Scrosati,Elecrochimica acta201055, 5110

20

.

(6) S. Brutti, G. Greco, P. Reale and S. PaneroElectrochem. Acta2013106, 483

.

(7) B. Ammundsen, D. Jones, J. Rozier, G. R. Burns,Chem. Mater.1996, 8, 2799-2808

.

(8) S. Brutti, V. Gentili, P. Reale, L. Carbone and S. PaneroJ.Power Sources2011196, 9792

.

(9) J. Hassoun, K-S. Lee, Y.-K. Sun and B. ScrosatiJ. Am.Chem.Soc.2011) 133, 3139

.

(10) Q. Zhong, A. Bonakdarpour, M. Zhang, Y. Gao and Y. Dahn,Electrochem. Soc.1997144,

205

.

(11) A.C. Larson and R.B. Von Dreele, "General Structure Analysis System (GSAS)", Los Alamos

National Laboratory Report LAUR 86-7482000

.

(12) B. H. Toby, EXPGUI, a graphical user interface for GSAS,J. Appl. Cryst.200134, 210-213

.

(13) C.J. Toussaint,J. Appl. Cryst.19714 293

.

(14) D. Aurbach, D. Markovsky, G. Salitra, E. Markevich, Y. Talyossef, M. Koltypin, L. Nazar,

B. Ellis and D. Kovancheva,J. Power Sources2007165, 491

.

(15) R. Alcantara, M. Jaraba, P. Lavela and J. L. Tirado,Electrochim. Acta200247, 1829

.

21

(16) K. Takahashi, M. Saitoh, M. Sano, M. Fujita and K. Kifune, Electrochem. Soc.2004151, 173

.

(17) Y. Idemoto, H. Narai, H. KouraElectrochem. Soc.2004151, 173

.

(18) J.-H. Kim, S.-T. Myung, C.S. Yoon, S.G. Kang, Y.-K. Sun,Chem. Mater.200416 906

.

(19) R. Santhanam, B. Rambabu,Journal of Power Sources2010195, 5442

.

(20) K. M. Shaju, P. G Bruce,Dalton Trans.200840, 5471-5475

.

(21) M. Kunduraci, J.F. Al-Sharab and G.G. AmatucciJ. Electrochem. Soc.200618, 3585-3592

.

(22) A. Bhaskar N.N. Bramnik A. Senyshyn H. Fuess and H. EhrenbergJ. Electrochem. Soc.2010

157, A689-A695

.

(23) N. Amdouni, K. Zaghib, F. Gendron, A. Mauger, C.M. Julien, J Magn Magn Mater2007

309, 100.

.

(24) G. Q. Liu, L. Wen, Y. M. LiuJ Solid State Electrochem201014 2191âAS2202

.

(25) M. Kunduraci, J. F. Al-Sharab, G. G. Amatucci,Chem. Mater., 200618, 3585,

.

(26) J.H. Kim, C.S. Yoon, S.T. Myung, J. Prakash and Y.K. Sun,Electrochemical and Solid- State

Letters, 20047 (7), A216

22

.

(27) J.H. Kim, S.T. Myung, C.S. Yoon, S.G. Kang and Y.K. Sun,Chem. Mater., 200416, 906.

.

(28) X-Ray Absorption Fine Structure For Catalysts and Surfaces, edited by Y. IwasawaWorld

Scientific, Singapore, 1996, pp. 375–384

.

(29) A. Di Cicco, M. Berrettoni, S. Stizza, E. Bonetti, and G.Cocco,Phys. Rev. B199450, 12386

.

(30) F. Boscherini, S. de Panfilis, and J. Weissmüller,Phys. Rev. B199857, 3365

.

(31) A. I. Frenkel, C. W. Hills, and R. G. Nuzzo,J. Phys. Chem. B2001105, 1115

.

(32) A. Witkowska, A. Di Cicco, A. Witkowska and E. Principi,Phys. Rev. B200776, 104110

.

(33) G. Greco, A. Witkowszka, E. Principi, M. Minicucci, andA. Di Cicco,Phys. Rev. B201283,

134103

.

(34) A. Filipponi, A. Di Cicco, and C. R. Natoli,Phys. Rev. B199552, 15122

.

(35) A. Filipponi and A. Di Cicco,Phys. Rev. B199552, 15135

.

(36) ImageJ, Version 1.38x, Wayne Rasband, National Institutesof Health, USA.

.

23

(37) J. I. Langford, D. Louër, and P. Scardi,J. Appl. Cryst.200033, 964

.

(38) L. Ottaviano, A. Filipponi, and A. Di Cicco,Phys. Rev. B199449, 11749

.

(39) A. Di Cicco, G. Aquilanti, M. Minicucci, E. Principi, N.Novello, A. Cognigni, and L. Olivi,

Journal of Physics: Conference Series2009190, 012043

.

(40) C. Castellano, F. Cordero, R. Cantelli and M. Ferretti,J. Appl. Phys.200292, 7206

.

(41) A. Filipponi and A. Di Cicco,TASK Quart.20004, 575

.

(42) W.-S. Yoon, C. P. Grey, X.-Q. Yang, and M. Balasubramanian, J. McBreen,Chem. Mater.

200315, 3161-169

.

(43) K. I. Pandya, R. W. Hoffman, J. McBreen, and W. E. O’Grady, J. Electrochern. Soc.1990

137, 383 .

.

(44) F. James,MINUIT reference Manual, Function minimization and error analysis, edited

by Long Writeup D506CERN Program Library, 1998375–384.

.

24