Page 1

Resonant x-ray scattering in 3d-transition-metal oxides:

Anisotropy and charge orderings

G. Subías1, J. García1, J. Blasco1, J. Herrero-Martín2 and M. C. Sánchez1

1 Instituto de Ciencia de Materiales de Aragón, Departamento de Física de la

Materia Condensada, CSIC-Universidad de Zaragoza, 50009-Zaragoza, Spain

2 European Synchrotron Radiation Facility, 38043-Grenoble-Cedex, France

E-mail: gloria@unizar.es

Abstract. The structural, magnetic and electronic properties of transition metal oxides

reflect in atomic charge, spin and orbital degrees of freedom. Resonant x-ray scattering

(RXS) allows us to perform an accurate investigation of all these electronic degrees.

RXS combines high-Q resolution x-ray diffraction with the properties of the resonance

providing information similar to that obtained by atomic spectroscopy (element

selectivity and a large enhancement of scattering amplitude for this particular element

and sensitivity to the symmetry of the electronic levels through the multipole electric

transitions). Since electronic states are coupled to the local symmetry, RXS reveals the

occurrence of symmetry breaking effects such as lattice distortions, onset of electronic

orbital ordering or ordering of electronic charge distributions. We shall discuss the

strength of RXS at the K absorption edge of 3d transition-metal oxides by describing

various applications in the observation of local anisotropy and charge

disproportionation. Examples of these resonant effects are (I) charge ordering

transitions in manganites, Fe3O4 and ferrites and (II) forbidden reflections and

anisotropy in Mn3+ perovskites, spinel ferrites and cobalt oxides. In all the studied cases,

the electronic (charge and/or anisotropy) orderings are determined by the structural

distortions.

1. Introduction

The discovery of new electronic properties, including high-TC superconductivity, colossal

magnetoresistance and multiferroic modifications, in transition-metal oxides has fuelled a

resurgence of interest in atomic charge, spin and orbital degrees of freedom in systems of highly

correlated electrons. X-ray spectroscopy techniques are highly sensitive to these electron

degrees of freedom. However, they probe short-distance correlations only. In contrast, x-ray

diffraction reveal long-range ordered static correlations. Resonant x-ray scattering (RXS)

combines absorption and diffraction as they have in common the x-ray atomic scattering factor

(ASF), f, which is usually written as: f = f0 + f ’+ if ” [1]. It contains an energy independent part,

f0, corresponding to the classical Thomson scattering and two energy-dependent terms, f ‘ and f

“, also known as the anomalous ASF. RXS occurs when the x-ray energy is tuned near the

absorption edge of an atom in the crystal. In this case, the anomalous ASF strongly depends on

the photon energy, which manifests in marked variations of the scattered intensity. This

dependence of the scattered intensity appears in any Bragg reflection when crossing the

absorption edge of a constituent atom.

The study of the energy-dependent modulation of the diffraction intensity of intense Bragg

peaks is the scope of the diffraction anomalous fine structure technique (DAFS) [2,3]. This

technique allows the determination of the local structural information around the anomalous

atom that is chemical and valence specific similar to that of Extended X-ray Absorption Fine

Page 2

Structure (EXAFS). The advantage of DAFS is that it is spatially and site-selective. Our interest

here is focused on the study of either weak-allowed or forbidden reflections that appear in a

phase transition. In the first case, the anomalous scattering contribution is comparable to the

Thomson one and the peculiar characteristics of RXS can be studied. The intensity of weak

superlattice reflections can be due to either a structural modulation, contributing to the structure

factor as a Thomson term or because the anomalous ASF of atoms, now in different

crystallographic sites, differ in some energy range. Normally, these differences are larger for

photon energies close to an absorption edge, showing an enhancement (or strong decrease) of

the scattered intensity just around the absorption threshold. In the case of symmetry forbidden

reflections, only the resonant term is present. Since RXS involves virtual transitions of core

electrons into empty states above the Fermi level, the excited electron is sensitive to any

anisotropy around the anomalous atom so that the anomalous ASF has a tensorial character.

Thus, a reflection can be observed on resonance if any of the components of the structure factor

tensor is different from zero. In the case of weak superlattice reflections, the resonant atoms

occupy different crystallographic sites and have different local structures. Thus, the anomalous

ASF are different at energies close to the absorption edge and the scattered intensity shows a

resonance reflecting the differences of the ASF between these non-equivalent atoms. RXS

intensity is then observed coming from the difference between the diagonal terms of the ASF. In

this case, there is a Thomson contribution and the analysis of the spectral shape must include it.

On the other hand, resonances observed in forbidden reflections are related to atoms that occupy

equivalent crystallographic sites. Symmetry elements with translation components (screw axes

and glide planes) of the crystal space group transform an atom into another equivalent in the

lattice, but with a differently oriented atomic surrounding. This makes that some of the off-

diagonal terms change sign after these symmetry operations, giving rise to structure factor

tensors that contain non-vanishing off-diagonal terms. Templeton & Templeton [4] first noticed

such reflections that show a local polarization anisotropy of the x-ray susceptibility and they are

now known as ATS reflections. We note that ATS reflections have strong polarization and

azimuth dependences.

The vector potential of the electromagnetic field in the matter-radiation interaction term can

be developed in a multipolar expansion in such a way that the symmetry of the excited levels

can be chosen. Thus, we can speak on dipolar-dipolar transitions, dipolar-quadrupolar

transitions, etc. Consequently, resonant methods are also sensitive to the symmetry of the

electronic shells, which compose the intermediate states. For electric dipole-dipole (E1E1)

transitions the wave vectors ki and kf of the incident and scattered x-rays do not enter the

scattering amplitude. The strength of the x-ray resonances associated with electric quadrupole-

quadrupole (E2E2) and electric dipole-quadrupole transitions (E1E2) are less intense than for

E1E1 process but must content the ki and kf dependence. A complete polarization and azimuthal

analysis of the RXS experiment is needed to assign the multipolar origin of the RXS signal.

In this contribution, after a recall of the RXS amplitude, we will discuss several examples

that correlate with the charge and orbital ordering concepts. We restrict to the metal K edges

and mainly to the dipolar-dipolar channel, which describes most of the phenomenology. Two

types of reflections are observed. The first one originates from the different anomalous

scattering factors of non-equivalent atoms in the crystal. The occurrence of this type of

resonance is correlated to the energy shift of the absorption edge (chemical shift) and it has been

considered as a proof of charge ordering (CO). The second one arises from the anisotropy of the

anomalous scattering factor of an atom at crystallographic equivalent sites. These ATS

reflections have been assigned to orbital ordering (OO). However, the first type of reflections

can also exhibit anisotropic behaviour. In all the studied cases, we conclude two main results: (i)

the charge segregation is much smaller than one electron and consequently, it would be better

described as a charge modulation and (ii) the anisotropy is present in the p-empty density of

states and it seems not to be correlated with a real d-orbital ordering.

2. Resonant x-ray scattering tensor

In RXS, the global process of photon absorption, virtual photoelectron excitation and photon re-

emission, is coherent through the crystal, giving rise to the usual Bragg diffraction condition

Page 3

)(

"'

0

ifff

e

j

F

jjj

j

RQi

++

∑

=

⋅r

r

(1)

where Rj

r

is the position of the j-th scattering atom in the unit cell,

r

f

k

if

kkQ

rr

r

−=

is the

scattering vector (

is the Thomson scattering part of the atomic scattering factor. The resonant part,

is given by the expression [5]

and are the wave vectors of the incident and scattered beams) and

,

i

k

r

j

f0

"'

iff +

∑

n

Γ

2

−−−

〉 〉〈〈

ψ

−

=+

n

gn

g

i

nn

f

ggn

e

2

iEE

O

ˆ

O

ˆ

EE

m

f

i

f

)(

)(

1

ω

h

*3

"'

ω

h

ψψψ

h

(2)

In this expression, ω

h

is the photon energy, is the electron mass,

E

n

Γ are the energy and inverse

ψ . The interaction of the electromagnetic radiation

Oˆ

ˆf

O

e

m

g

ψ describes the

initial and final electronic state with energy and and

g

E

n

lifetime of the intermediate excited states

n

with matter is expressed by the operators

operators up to the electric quadrupole term, we have [6]:

ˆ

(

O

and . By multipole expansion of these

i

*

)

2

1

1 (

r

)()()

r k i

fififi

r

r

r

r

⋅−⋅= ε

(3)

Here rr is the electron position measured from the absorbing atom and

polarization of the incident(scattered) beam. Correspondingly, we get that there are three

contributions to the resonant scattering factor: dipole-dipole (dd), dipole-quadrupole (dq) and

quadrupole-quadrupole (qq).

Using the Cartesian reference coordinate system defined in figure 1 for the photon

polarization and wave vector, we can develop the scalar product of equation (3) and the resonant

scattering amplitude of equation (2) can be written in the form:

−

=+

ω

)(

EE

gn

h

is the

)( fi

εr

⎥⎦

⎤

⎢⎣

⎡∑

α

∑

αβγ

∑

+−−

∑

n

Γ

2

−−−

β

,

αβγδ

αβγδ

Q

δγβαβαγγαβγ

I

γβααββα

εεεεεε

ω

h

****

3

2

"'

4

1

)(

2

)(

1

kkIkk

i

D

i

EE

m

f

i

f

ififfiifif

n

gn

e

h

(4)

where α, β, γ, δ are indexes that vary independently over the three Cartesian directions x, y, z,

and the transition matrix elements ,

contributions are characterized by the following Cartesian tensors of second, third and fourth

rank, respectively:

∑〈

=

n

ψψ

α αβγ

and associated to dd, dq and qq

αβ

D

αβγ

I

αβγδ

Q

〉〉〈

∑〈

n

=

〉〉〈

∑〈

n

=

〉〉〈

gnng

gnng

gnng

rrrrQ

rrrI

rrD

ψψψψ

ψ

γ

ψ

ψψψψ

δγβααβγδ

β

βα αβ

(5)

Page 4

Figure 1. Scheme of a general

RXS experiment

scattering plane)

(vertical

It is important to determine the symmetry properties of the D, I and Q tensors [7] that depend

on two factors: (i) the transformation properties of a space group itself and (ii) the local

symmetry of the resonant atoms position. Referring to non-magnetic samples, the structure

factor tensor is a second rank symmetric tensor (parity-even) with six independent components

in the dd approximation. This number reduces upon taking into account the local site symmetry

of the atom. The qq term is also symmetrical under the inversion symmetry, whereas the dq

contribution is only antisymmetrical. As a result, if an atom sits in an inversion centre, the I

tensor must be zero and the dipole-quadrupole transition is only allowed for atoms breaking the

inversion symmetry.

In the following we shall describe some significant RXS experiments following the two

types of resonant reflections, weak-allowed and forbidden. In the first case, we shall relate them

to the charge modulation, intimately joined to the transition metal-ligand bond distances and in

the second case we shall relate them to the anisotropy of the geometrical local structure around

the resonant atoms. Since the charge is a time-reversal invariant quantity, we shall deal either

with pure electric dd and qq transitions that are even under parity or with parity-odd electric dq

transitions. Time-reversal odd events that are related to the magnetic properties of the system

will be not considered.

3. Resonant effects due to different crystallographic sites (charge) ordering

Understanding the charge state in the high and low temperature phases of mixed valence

transition-metal oxides is of fundamental interest in the context of metal-insulator transitions

that are assumed to be driven by CO. The sensitivity of RXS to the CO relies on the fact the

energy values of the absorption edge for the two different valence states of the transition-metal

atom are slightly different (known as chemical shift). If long-range order of these valence states

exists, superlattice reflections due to the contrast between the atomic scattering factors of the

two valence states will exhibit a resonance enhancement. The point is that CO is intimately

correlated with the associated crystal distortions coming from the structural transitions that

accompanied the metal-insulator ones. Thus, a question arises whether the electronic CO is the

cause or the effect of the lattice distortions.

The concept of CO in solids was first applied by E. J. W. Verwey to the metal to insulator

transition that occurs in magnetite (Fe3O4) at Tv ∼ 120 K, now known as the Verwey transition

[8]. Above Tv, Fe3O4 has the inverse spinel cubic AB2O4 structure, where A and B are the

tetrahedral and octahedral Fe sites, respectively. Verwey originally proposed that the hopping of

valence electrons on the octahedral B-site sublattice is responsible for metallic conductivity. In

the insulating phase, spatial localization of the valence electrons on these B-sites gives rise to an

ordered pattern of Fe3+ and Fe2+ ions in successive [001] planes (cubic notation). The B-site

sublattice, shown in the inset of figure 2, can be regarded as a diamond lattice of tetrahedra of

nearest-neighbour Fe atoms sharing alternate corners. This simple model was implying the

observation of (0,k,l) reflections with k+l=4n+2 in the low temperature phase. These reflections

Page 5

are forbidden above Tv because of the diamond glide plane of the spinel structure. The first RXS

experiments in magnetite showed that (002) and (006) reflections belonging to this type of cubic

forbidden reflections originates from the local structural anisotropy of the Fe atoms at the B-

sites that have a trigonal point symmetry (3m) [9-11]. These works discarded the Verwey’s CO

model but they did not guarantee the lack of CO with other periodicities of the cubic unit cell.

In order to investigate other possible CO periodicities, we need to start from the low

temperature crystallographic structure. The symmetry lowering

16 non-equivalent Fe sites at tetrahedral and octahedral positions, respectively; each one can

have its own local atomic charge. However, the exact structure is not yet perfectly known [12-

14]. A good approach to the real structure consists of a

aaa

22/2/

××≈

, ac being the cubic cell parameter [12,14]. Complexity is greatly

reduced because there are only 6 non-equivalent Fe atoms, two in tetrahedral sites (A1 and A2)

and four in octahedral sites (B1, B2, B3 and B4). It can be noticed that atomic displacements in

this low temperature structure result in two main types of superlattice reflections, which are

indexed in the cubic notation: (I) (h,k,l) reflections such that h+k=even resulting form the loss of

the translation that give rise to charge modulations with wave vector

fcc

half-integer (h,k,l+1/2) reflections arising from the doubling of the cell along the c axis that

corresponds to charge modulation with

q

/ 1 , 0 , 0 (

=

possible charge segregation over the octahedral atoms along the c axis given by the sensitivity

of the RXS technique in a highly stoichiometric single crystal of Fe3O4 (TV=123.5 K) [15].

Figure 2 (left panel) shows the energy dependence of the intensities for some characteristic

Bragg and forbidden reflections at the Fe K-edge and at 60 K compared to the fluorescence

spectrum. Experimental data have been corrected for absorption.

CcmFd

→

3

generates 8 and

cell with lattice parameters

cP /2

ccc

cq

) 1 , 0 , 0 (

=

r

and (II)

c

) 2

r

. We examine now the limit for the

7,107,117,12

Energy (keV)

7,137,14 7,15

0

20

40

60

80

100

(0 0 7/2)σ-π

(0 0 1)σ-σ

(1 1 0)σ-σ

(4 4 3/2)σ-σ

x 10

3

Intensity (arb. units)

0

0.2

0.4

0.6

0.8

1

1.2

116118120122124126128

(4,0,1/2)

(0,0,7/2)

(0,0,1)

I (T) / I (115 K)

T (K)

Figure 2. The left panel shows some of the experimental RXS spectra around the Fe K-edge in

Fe3O4, corrected for absorption. The right panel shows the temperature dependence of the

integrated intensities, on and off-resonance, normalized to the low-temperature value.

(a) Energy scans of the permitted (0,0,1) and (1,1,0) reflections show three resonant peaks, a

first one at the Fe K threshold (7118 eV) and the other two around 7124-7129 eV, which

corresponds to the white line in the fluorescence spectrum. The observed strong resonant effect

is a consequence of electronic and structural differences among the Fe atoms at B1 and B2

octahedral sites. This can be parameterised in terms of valence as a charge segregation δ=0.23e.

This result confirms the lack of ionic CO in terms of Fe2+ and Fe3+ ions, in agreement with

previous RXS [16-18] and synchrotron powder diffraction [14] studies.

Page 6

(b) Resonant intensity is only observed in the σ−π’ channel for the (0,0,7/2) reflection,

indicating that this is a forbidden reflection in the low temperature phase. The energy scan

shows a three-peak structured resonance nearly at the same energies as the (0,0,1) and (1,1,0)

reflections. In this case, the electronic anisotropy comes from interference among equivalent

crystallographic sites. Six different sites are present for the Fe atoms so up to six different terms

could contribute to the resonant signal [19]. The superlattice (4,4,3/2) corresponds to the same

periodicity along c as the forbidden (0,0,l/2) reflections. However, it displays hardly any

resonant effect opposite to what is expected to occur at those very weak reflections. Therefore,

we can conclude that no charge segregation exists with (0,0,1/2) periodicity and the ATS

(0,0,l/2) reflections have its origin in the loss of octahedral and tetrahedral symmetry originated

by the structural phase transition.

In order to establish the correlation between the lattice distortion and the charge segregation

and anisotropy orderings, we have measured the temperature dependence of the intensity of the

following superlattice reflections: (4,0,1/2) off-resonance (E=7.1 keV) and (0,0,1) and (0,0,7/2)

on resonance (E=7.125 keV), which is reported in figure 2 (right panel). The resonant and non-

resonant signals simultaneously disappear at Tv (±0.5 K) and the intensity of all these reflections

is zero at temperatures above 125 K. This result shows that RXS in Fe3O4 comes from the

ordering of local distortions at the structural transition, which leads to an ordered formal charge

segregation and electronic anisotropy at the Fe atoms [15].

We will comment now on the half-doped manganites such as Nd0.5Sr0.5MnO3 and

Bi0.5Sr0.5MnO3. It was proposed the ordering of an alternating pattern of Mn3+ and Mn4+ ions

was predicted leading to the onset of superlattice reflections doubling the b axis of the

orthorhombic Pbnm (Ibmm) cell [20]. The observed modulation wave vectors are (0,k,0) and

(0,k/2,0) with k odd for the ordering of charge and orbital degrees of freedom, respectively. The

occurrence of orbital order (OO) is also predicted independently from the CO because Mn3+ are

Jahn-Teller distorted ions. We have measured the energy dependence of the intensity at the Mn

K edge at Q=(0,3,0) and Q=(0,5/2,0) for both, Nd:Sr [21] and Bi:Sr [22] single crystals. Figure

3 summarizes the results obtained.

6,54 6,556,56 6,57

0,0

0,4

0,8

1,2

1,6

2,0

Intensity (arb. units)

E (keV)

Bi0.5Sr0.5MnO3

Nd0.5Sr0.5MnO3

6,546,556,566,57

0,0

0,2

0,4

0,6

0,8

1,0

1,2

Bi0.5Sr0.5MnO3

Nd0.5Sr0.5MnO3

Intensity (arb. units)

E (keV)

Figure 3. The left panel shows the normalized intensity for the non-rotated polarization channel

at the (0,3,0) peak associated to CO whereas the right panel shows the rotated polarization

channel at the (0,5/2,0) peak associated to OO in Nd0.5Sr0.5MnO3 and Bi0.5Sr0.5MnO3.

Non-resonant intensity can be observed away from the K-edge of Mn and a broad resonance

is found at the absorption edge at (0,3,0) reflections in the non-rotated (σ-σ’) channel. On the

other hand, a strong Gaussian-shaped resonance is observed at energies close to the K-edge of

Mn at (0,5/2,0) reflections only in the rotated (σ-π’) channel, which identifies these half-integer

reflections as structurally forbidden ones. The variation with azimuth of these resonances shows

a characteristic oscillation with π periodicity. We note that in this case, the two kinds of

Page 7

behaviour mixed since the marked different anisotropy of the two types of atoms. The resonant

scattering at either (0,k,0) or (0,k/2,0) reflections does not depend on the

arises from E1 (1s→ np) transition. The temperature dependence of the resonant intensities

shows that the two types of reflections disappear at the metal-insulator phase transition.

The complete analysis of the energy line-shape and the azimuth and polarization dependence

of the resonant intensities is carried out using a semi-empirical structural model [23]. The

checkerboard arrangement of two types of crystal Mn sites gives an excellent agreement with

the experimental data. These two sites differ in their local geometric structure: one site is

anisotropic (tetragonal distorted oxygen octahedron) and the other site is isotropic (nearly

undistorted oxygen octahedron). Estimates of the valence modulation between the two Mn

atoms are slightly variable depending on the manganite but all fall below the ideal charge

segregation of ±0.5e. As shown in [21] and [22], ±0.08 for Nd0.5Sr0.5MnO3 and ±0.07 for

Bi0.5Sr0.5MnO3, respectively.

CO for x different from 0.5 is in general not well defined. We have also explored further the

Bi1-xSrxMnO3 system toward the Mn3+-rich side, that is for x=0.37 [24]. Superlattice reflections

of (0,k,0) and (0,k/2,0) types with k odd were observed at the Mn K-edge. This is to be

compared with the observation of the K-edge resonances in Bi0.5Sr0.5MnO3. Figure 4 shows an

example of the resonant enhancement as observed at (0,3,0) and (0,7/2,0) reflections in

Bi0.63Sr0.37MnO3.

4 1.5

(0,3,0)

Intensity (arb. units)

wave vector and it

k

1.5

2

2.5

3

3.5

0

0.5

1

6.53 6.546.55

E (keV)

6.56 6.576.58

Intensity σ-σ'

Intensity σ-π'

x 200

0

-60 -30 0

0.2

0.4

0.6

0.8

1

1.2

30 60 90 120 150 180

ϕ (º)

E=6552.6 eV

σ-σ'

6540655065606570

0,0

0,2

0,4

0,6

0,8

1,0

-60-40 -20 0 20 40 60 80 100120140160180

ϕ (deg)

0,0

0,2

0,4

0,6

0,8

1,0

E=6551.6 eV

Intensity (arb. units)

(0, 7/2 ,0)

Intensity (arb. units)

E (keV)

Figure 4. RXS results in Bi0.63Sr0.37MnO3 at the Mn K-edge. (a) Polarization analysis of the

(0,3,0) reflection as a function of photon energy. (b) energy dependence of the forbidden

(0,7/2,0) reflection for the σ-π’ channel. The insets show the respective azimuthal angle

dependence on resonance.

The energy dependence of the intensity for both, weak-allowed (0,k,0) and forbidden

(0,k/2,0) reflections is identical to that observed in the half-doped bismuth manganite (figure 3).

The dependencies of the resonant intensities with azimuthal angle reveals identical twofold

symmetry too. We note that the σ-σ’ intensity of the (0,3,0) peak approaches the non-resonant

intensity at the minimum opposite to the constant evolution expected for a pure CO reflection.

These results indicate that the checkerboard ordering of two types of Mn atoms in terms of the

local structure in the ab plane in a ratio 1:1 is strongly stable and extends to doping

concentrations x<0.5. Intermediate valence states lower than +3.5 are deduced for

Bi0.63Sr0.37MnO3, where the charge disproportion is found to be ±0.07.

The last example is provided by La0.33Sr0.67FeO3, which shows a charge modulation that is

commensurate with the carrier concentration (n=1-x) and corresponds to a wave vector

q=(2π/ap) (1/3,1/3,1/3), being ap the primitive cubic lattice parameter. Ordered layers of Fe3+

and Fe5+ ions in a sequence of …Fe3+Fe3+Fe5+… along the cubic [111] direction were originally

Page 8

proposed to explain the metal-paramagnetic to insulator-antiferromagnetic transition at 200 K

[25]. We have investigated this three-fold CO using the RXS technique [26]. Superlattice

(h/3,h/3,h/3) reflections in the pseudocubic cell notation were observed in the non-rotated σ-σ’

polarization channel for h=2, 4 and 5. These reflections are forbidden in both cubic (Pm-3m)

and rombohedral (R-3c) symmetries. However, they exhibit a resonant behavior at energies

close to the Fe K edge on top of a non-resonant signal, as shown in figure 5.

7,107,117,127,137,147,15

0

5

10

15

20

25

30

35

40

45

50

55

60

-90-60-30

Azimuthal angle (°)

0306090

0

1

2

3

(4/3,4/3,4/3)

σ-σ'

Integr. Intens. (a.u.)

(5/3,5/3,5/3)

(2/3,2/3,2/3)

(4/3,4/3,4/3)

Intensity (electron units)

E (keV)

Figure

scans of the σ-σ’

intensity

(h/3,h/3,h/3)

reflections measured

at 10 K and corrected

for absorption. The

inset shows

dependence of the

resonant

(7129.5 eV) on the

azimuth angle

(4/3,4/3,4/3)

reflection.

5.

Energy

for the

no

intensity

for

The non-resonant intensities are explained due to a periodic structural modulation of both Fe

and Sr(La)O3 atomic planes along the cubic [111] direction. To account for the strong variation

in the Thomson scattering among the different satellite reflections, the Fe and Sr(La)O3 atomic

planes must move in opposite senses, keeping the inversion symmetry of the supercell. These

shifts differentiate two crystallographic sites for the Fe atoms and produce an ordered sequence

of two compressed and one expanded FeO6 octahedra. The compressed octahedron is

asymmetric with three short and three long Fe-O bonds whereas the expanded one is regular.

Concerning the local charges of the two types of Fe cations, the chemical shift between the

expanded and compressed Fe atoms was found to be 0.7±0.1 eV. This is to be compared to the

energy shift of 1.26 eV between Fe3+ and Fe4+ [27]. Assuming a linear relationship between

energy shift of the K absorption edge and local charge, we determined that the amount of charge

segregation is 0.6e. Since the resonant intensity is constant as a function of the azimuth angle

over a range of 180 º, as shown in the inset of figure 5, the Fe atoms have not a local anisotropy

in this case. This result indicates the presence of a pure charge density wave ordering in the

mixed-valent perovskite La0.33Sr0.67FeO3.

4. Resonant effects due to anisotropy (orbital) ordering

The direct observation of OO is a difficult task because it is accompanied by other effects such

as lattice distortions and charge segregations. The study of the RXS and the anisotropic

character of the x-ray scattering were developed in crystallography more than 20 years ago [4].

In particular, Dmitrienko [28,29] developed a general theoretical treatment of the “forbidden”

ATS reflections, deriving new extinction rules valid near the absorption edge. In the

microscopic point of view, these ATS reflections are due to the presence of the absorbing atom

in an anisotropic chemical environment, which brings about the orientation of unoccupied

electronic levels.

The experimental application of RXS to the study of OO started in 1998, when Murakami

and co-workers [30] investigated the CO and OO in La0.5Sr1.5MnO4 at the Mn K-edge. It was

Page 9

concluded the successful observation of OO of the

type orbital) on Mn

angular dependence around the scattering vector of the intensity (azimuthal dependence)

confirms that they result from the anisotropic character of the anomalous part of the atomic

scattering factor. However, it was later demonstrated that the Jahn-Teller distortion of the

oxygen octahedra surrounding the Mn atoms is sufficient to reproduce the experimental results

at the K edge without invoking any contribution of the 3d-OO [31].After this first observation,

RXS is widely used to study the orbital degree of freedom in several manganites [32-36] and

other transition-metal oxides [37-40]. In particular, the following experiments provide key

information for the mechanism of RXS and OO.

(1) The parent compound of colossal magnetoresistive manganites, LaMnO3, was expected

to show an alternating ordering of and

780 K due to the greatly distorted MnO

[41]. The observation of resonance in the σ-π’ scattered intensity of the (3,0,0) forbidden

reflection at the Mn K-edge was initially interpreted as a direct probe of this OO [32].

Moreover, the increase of the resonant signal with decreasing temperature as TN was

approached from above was interpreted as a direct correlation of the magnetic order with OO.

Recently, we have revisited the RXS at the Mn K-edge of LaMnO3, both experimentally and

theoretically [36]. We have observed two independent forbidden reflections, (0,3,0) [or,

equivalently (3,0,0)] and (0,0,3), which are related to two different nonzero off-diagonal

elements of the second-rank atomic scattering factor tensor. The different energy dependence of

the RXS spectra for the two types of forbidden reflections has been explained within the

multiple scattering theory in terms of long-range ordered structural distortions around Mn atoms

using a cluster that includes up to 63 atoms beyond the first oxygen neighbors [42], as

demonstrated in figure 6(a). Since no change in either the intensity or its azimuthal dependence

was observed when crossing TN at 140 K (see figure 6(b)), these forbidden reflections are

ascribed as ATS reflections in agreement with a structural origin and opposite to the OO

interpretation.

2.1

1.2

absorption, theory

(030), theory

(003), theory

electrons (- and -

g e

)3 (

22

rz −

)(

22

yx −

3+ sites by detecting the (2n+1/4,2n+1/4,0) forbidden reflections. The

orbitals in the ab plane below

)3 (

22

rx −

)3 (

22

ry −

6 octahedron and the A-type antiferromagnetic order

0

6.54

0.5

1.1

1.6

6.556.56

E (keV)

6.576.58

Intensity (arb. units)

(003)(030)

0

5

10

15

20

25

0.0

1.2

0.4

0.8

absorption, theory

(030), theory

(003), theory

0

5

10

15

20

25

0.0

0.4

0.8

(030)

(003)

E-E0 (eV)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

050100150200 250300

(003)

(030)

Intensity (arb. units)

Temperature (K)

TN

0

-180-160-140-120-100 -80 -60 -40 -20 0

Azim uth angle (degree)

0.2

0.4

0.6

0.8

1

1.2

Calculation

(030)

(003)

Intensity (arb. units)

Figure 6. (a) Experimental (circles) RXS of the (0,3,0) and (0,0,3) forbidden reflections in

LaMnO3. The inset shows the MXAN calculations done using the 63-atoms cluster (upper panel)

and the 6-atoms cluster (lower panel). (b) Temperature dependence of the integrated intensity of

the (0,3,0) and (0,0,3) reflections on resonance, normalized for comparison. The inset shows the

respective azimuthal angle dependence on resonance.

(2) Another example of the excitement of forbidden reflections on a cubic crystal structure

that deserves discussion are the (0,0,l) (l=4n+2) reflections in Fe3O4. Once the Verwey model of

Page 10

CO was discarded, these reflections offer the possibility to study RXS excited by a dq-transition

which is only allowed on the tetrahedral site whereas the octahedral site allows, on the contrary,

only for dd and qq transitions. Two distinct resonant lines are observed in the RXS spectra of

(0,0,2) and (0,0,6) forbidden reflections in Fe3O4 at the pre-edge and at the main edge energies

of the Fe K-edge [10]. The comparison of the energy and azimuthal dependence of these

reflections at the Fe and Co K edges in Fe3O4, CoFe2O4 and MnFe2O4 spinels is reported in

figure 7 [38]. No pre-edge resonance is observed either at the Fe K-edge in MnFe2O4 or at the

Co K-edge in CoFe2O4, whereas the main-edge resonance is observed at the Fe K-edge in both,

CoFe2O4 and MnFe2O4, and at the Co K-edge in CoFe2O4.

7,107,117,12

Energy (KeV)

7,13 7,14 7,15

0

1

2

3

4

0

0.2

0.4

0.6

0.8

1

1.2

7.77.727.74

Intensity (arb. units)

E (keV)

CoFe2O4

Co K-edge

CoFe2O4

MnFe2O4

Fe3O4

Fe K-edge

Intensity (arb. units)

Figure 7. Energy scans of the (0,0,2)

(closed symbols) and (0,0,6) (open

symbols) forbidden reflections at the Fe K

edge in the spinel ferrites. The inset

compares the Co K-edge RXS of the

(0,0,2) and (0,0,6) reflections in CoFe2O4.

These results confirm that the pre-peak resonance originates from dq transitions at the

tetrahedral Fe atoms and the main-edge resonance is due to the anisotropy of the trigonal (

point symmetry of octahedral B sites of the spinel structure. The azimuthal dependence

corroborates the dd character of the main-edge resonance, which does not depend on either the

type or the formal valency of the transition-metal atom that occupies the octahedral site. We

note here that we have observed anisotropic RXS, mainly of a structural origin, in a system

where the local atom environment is nearly isotropic since there is not a significant distortion of

the ligand configuration [43].

(3) Recently, the correlation of OO and magnetic ordering has been studied on layered cobalt

oxides, RBaCo2O5.5 with R=rare-earth [39,40]. Resonances have been observed at the Co K

edge for (0,k,0) reflections with k odd in TbBaCo2O5.5 whereas (h,0,0) reflections with h odd

were not detected either on or off resonance in any phase transition [39]. The cusp of the

resonant scattering is either up –(0,3,0) and (0,7,0)- or down –(0,1,0) and (0,5,0)- depending on

the k value. This behavior arises from displacements of the Tb and Ba ions from the ideal

tetragonal positions while the occurrence of RXS and its azimuthal dependence comes form the

presence of two different environments for Co atoms (octhedron and pyramid), ordered along

the b axis. Therefore, resonant (0,k,0) reflections with k odd corresponds to ATS reflections and

no further contribution due to OO has been deduced from this experiment [40]. On the other

hand, superlattice (1,0,l) with l even reflections were reported occurring at the metal-insulator

transition in GdBaCo2O5.5-x [40]. RXS was observed at these reflections, which is sensitive to

the difference in anisotropy of adjacent Co pyramids and/or Co octahedral and it has been

associated to the presence of OO.

)

Page 11

5. Conclusions

RXS has demonstrated its potential to solve old and new problems on strong correlated

transition-metal oxides. The results on the low temperature insulating phase of magnetite has re-

opened the discussion on the origin of the Verwey transition. The classical CO model has been

overcome and many recent theoretical papers gives a more realistic interpretation without

invoking to ionic ordering. In general, all these RXS results discard the description of mixed-

valence transition-metal oxides as a bimodal distribution of integer valence states and the so-

called CO phase implies the segregation in different crsytallographic sites whose electronic

differences are mainly determined by the structural distortions. These structural distortions

induces a different charge density on different crystallographic sites and in some cases, an

electronic anisotropy by lowering the local symmetry. Finally, RXS has been also fundamental

to determine the type of ordering in many cases, such as Bi0.67Sr0.33MnO3 and La0.33Sr0.67FeO3.

Acknowledgement

The authors thank financial support from the Spanish MICINN (FIS08-03951 project) and DGA

(Camrads). We also acknowledge ESRF for granting beam time and technical support.

References

[1]

Materlik G, Spark C J and Fisher K 1994 Resonant Anomalous X-ray Scattering, Theory

and Applications (Amsterdam: North-Holland/Elsevier Science B. V.)

Stragier H, Cross J O, Rehr J J, Sorensen L B, Bouldin C E and Woicik J C 1992 Phys.

Rev. Lett. 21 3064

Proietti M G, Renevier H, Hodeau J L, García J, Bérar J F and Wolfers P 1999 Phys. Rev.

B. 59 5479

Templeton D H and Templeton L K 1980 Acta Cryst. A 36 237

Blume M 1994 Resonant Anomalous X-ray Scattering, Theory and Applications ed

Materlik G, Spark C J and Fisher K (Amsterdam: North-Holland/Elsevier Science B.

V.) pp 495-515

Hannon J P, Trammel G T, Blume M and Gibbs D 1988 Phys. Rev. Lett. 61 1245

Di Matteo S, Joly Y and Natoli C R 2005 Phys. Rev. B 72 144406

Verwey E J W 1939 Nature 144 327

Hagiwara K, Kanazawa M, Horie K, Kokubun J and Ishida K 1999 J. Phys. Soc. Jpn. 68

1592

[10] García J, Subías G, Proietti M G, Renevier H, Joly Y, Hodeau J L, Blasco J, Sánchez M

C and Bérar J F 2000 Phys. Rev. Lett. 85 578

[11] García J, Subías G, Proietti M G, Blasco J, Renevier H, Hodeau J L and Joly Y 2001

Phys. Rev. B 63 054110

[12] Izimu M, Koetzle T F, Shirane G, Chikazumi S, Matsui M and Todo S 1982 Acta Cryst.

B 38 2121

[13] Zuo J M, Spence C H and Petuskey W 1990 Phys. Rev. B 42 8451

[14] Wright J P, Attfield J P and Radaelli P G 2001 Phys. Rev. Lett. 87 266401; 2002 Phys.

Rev. B 66 214422

[15] García J, Subías G, Herrero-Martín J, Blasco J, Cuartero V, Sánchez M C, Mazzoli C and

Yakhou F 2009 Phys. Rev. Lett. 102 176405

[16] Subías G, García J, Blasco J, Proietti M G, Renevier H and Sánchez M C 2004 Phys. Rev.

Lett. 93 156408

[17] Nazarenko E, Lorenzo J E, Joly Y, Hodeau J L, Mannix D and Marin C 2006 Phys. Rev.

Lett. 97 056403

[18] Joly J, Lorenzo J E, Nazarenko E, Hodeau J L, Mannix D and Marin C 2008 Phys. Rev.B

78 134110

[19] Wilkins S B, Di Matteo S, Beale T A W, Joly Y, Mazzoli C, Hatton P D, Bencok P,

Yakhou F and Brabers V A M 2009 Phys. Rev. B 79 201102(R)

[20] Radaelli P G, Cox D E, Marezio M and Cheong S W 1997 Phys. Rev. B 55 3015

[21] Herrero-Martín J, García J, Subías G, Blasco J and Sánchez M C 2004 Phys. Rev. B 70

024408

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

Page 12

[22] Subías G, García J, Beran P, Nevriva M, Sánchez M C and García-Muñoz J L 2006 Phys.

Rev. B 73 205107

[23] García J, Sánchez M C, Blasco J, Subías G and Proietti M G 2001 J. Phys.: Condens.

Matter 13 3243

[24] Subías G, Sánchez M C, García J, Blasco J, Herrero-Martín J, Mazzoli C, Beran P,

Nevriva M, and García-Muñoz J L 2008 J. Phys.: Condens. Matter 20 235211

[25] Park S K, Ishikawa T, Tokura Y, Li J Q and Matsui Y 1999 Phys. Rev. B 60 10788

[26] Herrero-Martín J, Subías G, García J, Blasco J and Sánchez M C 2009 Phys. Rev. B 79

045121

[27] Blasco J, Aznar B, García J, Subías G, Herrero-Martín J and Stankiewicz J 2008 Phys.

Rev. B 77 054107

[28] Dmitrienko V E 1983 Acta Cryst. A 39 29

[29] Dmitrienko V E 1984 Acta Cryst. A 40 89

[30] Murakami Y, Kawada H, Kawata H, Tanaka M, Arima T, Moritomo Y and Tokura Y

1998 Phys. Rev. Lett. 80 1932

[31] Benfatto M, Joly Y and Natoli C R 1999 Phys. Rev. Lett. 83 636

[32] Murakami Y, Hill J P, Gibbs D, Blume M, Koyama I, Tanaka M, Kawata H, Arima T,

Tokura Y, Hirota K and Endoh Y 1998 Phys. Rev. Lett. 81 582

[33] Zimmermann M, Hill J P, Gibbs D, Blume M, Casa D, Keimer B, Murakami Y, Tomioka

Y and Tokura Y 1999 Phys. Rev. Lett. 83 4872

[34] Endoh Y, Hirota K, Ishihara S, Okamoto S, Murakami Y, Nishizawa A, Fukuda T,

Kimura H, Nojiri H, Kaneko K and Maekawa S 1999 Phys. Rev. Lett. 82 4328

[35] Di Matteo S, Chatterji T, Joly Y, Stunault A, Paixao J A, Suryanarayanan R, Dhalenne G

and Revcolevschi A 2003 Phys. Rev. B 68 024414

[36] Subías G, Herrero-Martín J, García J, Blasco J, Mazzoli C, Hatada K, Di Matteo S and

Natoli C R 2007 Phys. Rev. B 75 235101

[37] Paolasini L, Caciuffo R, Sollier A, Ghigna P and Altarelli M 2002 Phys. Rev. Lett. 88

106403

[38] Subías G, García J, Proietti M G, Blasco J, Renevier H, Hodeau J L and Sánchez M C

2004 Phys. Rev. B 70 2155105

[39] Blasco J, García J, Subías G, Renevier H, Stingaciu M, Conder K and Herrero-Martín J

2008 Phys. Rev. B 78 054123

[40] García-Fernández M, Scagnoli V, Staub U, Mulders A M, Janousch M, Bodenthin Y,

Meister D, Patterson B D, Mirone A, Tanaka Y, Nakamura T, Grenier S, Huang Y and

Conder K 2008 Phys. Rev. B 78 054424

[41] Goodenough J B 1963 Magnetism and Chemical Bond (New York:Interscience)

[42] Benfatto M, Della Longa S and Natoli C R 2003 J. Synchrotron Radiat. 10 51

[43] Subías G, García J and Blasco J 2005 Phys. Rev. B 71 155103