B5 Scattering Techniques II:

Magnetic X-Ray Scattering

Th. Brückel
Institut für Festkörperforschung
Forschungszentrum Jülich GmbH

Contents

1 Introduction ............................................................................................... 2

2 The cross section for magnetic x-ray scattering..................................... 3

3 Experimental consideration ................................................................... 10

4 Resonance exchange scattering.............................................................. 14

5 Non-resonant magnetic x-ray diffraction ............................................. 20

6 Summary; comparison to magnetic neutron scattering ...................... 31

References .......................................................................................................... 33

Index 35
B5.2 Th. Brückel

1 Introduction
Most of our present knowledge of the atomic structure of condensed matter results from x-ray
diffraction studies, which probe the interaction of the electric field with the electric charge of
the electron. However, since x-rays represent an electromagnetic radiation and since in mag-
netic materials some electrons carry a magnetic moment due to spin- and angular momentum,
we naturally would expect a magnetic interaction in addition to the pure charge interaction.
Even so this interaction was well established in theory [1] since Klein-Nishina 1929, the first
magnetic diffraction effect was demonstrated only 1972 by de Bergevin and Brunel [2] with a
commercial x-ray tube. The same authors gave a classical picture of the interaction process,
deduced the detailed polarisation dependence and presented measurements on ferromagnetic
compounds in a subsequent paper [3]. However, since the magnetic interaction gives just a
relativistic correction to the cross section, the amplitudes of magnetic diffraction are down by
approximately three orders of magnitude as compared to charge scattering, resulting in an
intensity ratio of about 10-6. Therefore magnetic x-ray scattering was considered an exotic
topic until the experiments on Ho by Gibbs et al. [4], which took advantage of the high bril-
liance of a synchrotron radiation x-ray source thus compensating by a high photon flux at the
sample position for the weak magnetic scattered intensities. The polarisation properties and
the tunability of synchrotron radiation offered new perspectives for magnetic x-ray investiga-
tions. This was again demonstrated on Ho [5] by an attempt to separate spin- and angular
momentum with polarisation analysis and by the observation of a resonance enhancement of
the magnetic signal at the absorption edges.

Nowadays, synchrotron radiation techniques for the study of properties of magnetic materials
are well established. Very widespread is the application of incoherent probes which measure a
macroscopic ensemble average of local magnetic properties. Among these we mention Kerr-
microscopy, measurements of the Faraday effect and the linear or circular x-ray magnetic di-
chroism. The Kerr- and Faraday effect measure the rotation of the plane of polarisation of an
electromagnetic wave as it is reflected from or transmitted through a magnetic material, re-
spectively. Magnetic circular dichroism describes the difference in the absorption of right-
and left circularly polarised x-rays by magnetic materials. It measures essentially the same
quantities as the Kerr- and Faraday effect, namely the orbital and spin contributions to the
magnetic moments with element and certain site specifities. Kerr microscopy and x-ray to-
pography are used for magnetic domain imaging. Absorption techniques become local micro-
scopic probes when the spin resolved x-ray absorption fine structure is observed. In analogy
to classical EXAFS experiments, such measurements provide information about the local en-
vironment, but are explicitly sensitive to the magnetic neighbours only. True microscopic
spatial resolution is obtained with the coherent probes, namely magnetic x-ray diffraction (as
well non-resonant as resonance exchange scattering) and nuclear resonant scattering. Mag-
netic scattering provides a wealth of information on magnetic correlation lengths, the local
magnetic moments and environment, the magnetic structure and phase transitions. Magnetic
x-ray reflectivity is the corresponding probe for the investigation of magnetic thin films. Nu-
clear resonant scattering yields information on hyperfine fields and might eventually become
important for the measurement of magnetic excitations. Resonant diffraction and absorption
techniques are intimately related by the optical theorem, which states that the attenuation
coefficient is proportional to the imaginary part of the forward scattering amplitude. In this
sense, diffraction experiments comprise absorption techniques, but in addition they provide
true atomic resolution. In this impressive list of synchrotron radiation techniques we want to
finally mention magnetic Compton scattering for the determination of the spin resolved elec-
Scattering Techniques B5.3

tron momentum density and angular- and spin resolved photoemission, which gives the spin
resolved band structure.

Many of these topics are discussed in detail in a textbook [6]. In what follows I will concen-
trate on magnetic x-ray diffraction. I will introduce the cross section for non-resonant and
resonant magnetic x-ray scattering in section 2 and discuss experimental considerations in
section 3. Examples of resonance exchange scattering experiments on bulk antiferromagnets,
ferromagnets and thin film systems are given in section 4. Section 5 gives some examples for
non-resonant x-ray scattering, before we summarise in section 6 some important features of
magnetic x-ray scattering and compare it to magnetic neutron diffraction.

2 The cross section for magnetic x-ray scattering

A calculation of the cross-section for x-ray scattering including the magnetic terms from a
quasi-relativistic Hamiltonian for electrons in a quantised electromagnetic field within sec-
ond-order perturbation theory was done by Blume [7] and Blume and Gibbs [8]. Platzman and
Tzoar [1] and de Bergevin and Brunel [3] started from the Dirac equation and reduced this
relativistic ansatz using a Foldy-Wouthuysen transformation to a quasi-non-relativistic form
analogous to that obtained from the non-relativistic Hamiltonian. The expansion of this quasi-
non-relativistic Hamiltonian in dependence of photon energy over electron rest mass hω / mc2
allows the description of the magnetic scattering process. Grotch, Kazes, Bhatt and Owen [9]
extended the Foldy-Wouthuysen transformation to second order in hω / mc2 . Here we follow
a presentation given by Blume [7] and Blume and Gibbs [8] based on a non-relativistic treat-
ment in second order perturbation theory. We start with the Hamiltonian for electrons in a
quantised electromagnetic field:
eh
H=∑ (P j − A(r j )) 2 + ∑ V(rij ) −
mc ∑
1 e
s j ⋅ ∇ × A(r j )
j
2m c ji j
(1)
eh
∑j j ∑
e 1
− s ⋅ E( r ) × ( P j − A ( r j )) + hω k (c +
( k λ)c(k λ) + )
2(mc) 2 j c kλ
2
Here, the first term corresponds to the kinetic energy of the electrons in the electromagnetic
field, represented by the vector potential A(r), the second term corresponds to the Coulomb
interaction between the electrons, the third term to the Zeeman energy -µ⋅H of the electrons
with spin sj, the fourth term to the spin-orbit coupling and the final term to the self energy of
the electromagnetic field. From the form of (1), we can immediately guess that the cross-sec-
tion and polarisation dependence of the scattering of an electromagnetic wave from magnetic
materials is more complex than the corresponding cross-section for neutron scattering - at
least if we only consider the two main interaction potentials for nuclear scattering and mag-
netic dipole scattering. In the case of neutron scattering, only the magnetic dipole interaction
of the neutron spin with the magnetic field of the electrons gives rise to magnetic scattering.
In the case of x-rays, we have several interaction terms as well between the spin of the elec-
trons and the electromagnetic field as between the orbital momentum and the magnetic field.
In addition, photons are spin 1 particles as compared to spin 1/2 neutrons. Therefore we can
expect a much more complex polarisation dependence.

The vector potential A(r) in (1) is linear in photon creation and annihilation operators, c + (kλ)
and c(kλ) and is given in a plane wave expansion by:
B5.4 Th. Brückel
1
⎛ 2πhc 2 ⎞ 2
A(r ) = ∑ ⎜⎜
i q⋅r − i q⋅r
⎟⎟ × [ε (qσ )c(qσ)e + ε * (qσ )c + (qσ)e ] (2)
qσ ⎝
Vω q ⎠
Here V is a quantisation volume and ε(qσ) is the unit polarisation vector corresponding to a
wave with wavevector q of polarisation state σ. Two polarisation states σ = 1,2 of the photons
have to be distinguished. As a basis, we can either use linear polarisation in two perpendicular
directions or left and right circular polarisation. Since A(r) is linear in the c+ and c-operators,
scattering occurs in second order for terms linear in A and in first order for quadratic terms.
We do not want to reproduce the calculation given in [7] in detail. The Hamiltonian (1) is
written as a sum
H = H 0 + H r + H' (3)
where H0 contains only the degrees of freedom of the electron system, Hr is the Hamiltonian
for the quantised electromagnetic field and H' corresponds to the interaction between the
electrons and the radiation field. Scattering cross-sections are calculated by assuming that ini-
tially the solid is in a quantum state |a>, which is an eigenstate of H0 with energy Ea, and that
there is a single photon present. We then calculate the probability of a transition induced by
the interaction Hamiltonian H' to a state |b> with photon k'λ'. For elastic scattering |b> = |a>.
The transition probability per unit time can be calculated by the golden rule to second order
perturbation theory. The fact that we have to go to second order perturbation theory for terms
linear in A immediately implies that besides the so-called non-resonant magnetic x-ray scat-
tering, resonance phenomena will appear due to the energy denominator found in second or-
der perturbation theory (compare the Breit-Wigner-formula for resonant scattering of the
neutron from a nucleus). Here we will just quote the final result of this calculation: at
moderately high x-ray energies and far away from all absorption edges of the elements in the
sample, the elastic cross-section for scattering of photons with incident polarisation ε into a
state of final polarisation ε' can written as:
λC
2 2
dσ ⎡ e2 ⎤
= ⎢ 2 ⎥ ⋅ fC +i fM (4)
dΩ ⎣ mc ⎦
ε 'ε d ε 'ε
ε →ε '

Here re=e2/mc2 = 2.818 fm denotes the classical electron radius, λC = h/mc = 2.426 pm the
Compton length of an electron. The scattering amplitudes <fC> and <fM> are given as matri-
ces which describe the polarisation dependencies of charge and magnetic scattering, respec-
tively. Here we discuss the case of linear polarisation, described by unit vectors perpendicular
to the wave vectors of incident and scattered photons, k and k'. σ-polarisation corresponds to
the basis vector perpendicular to the scattering plane, π-polarisation corresponds to the vec-
tors in the k, k' plane. The basis vectors for the components of the magnetic moment of the
sample and for the polarisation states are defined as follows, see figure 1:
uˆ1 = (k + k ' ) k + k'
uˆ2 = (k ' ×k ) k' ×k ≡ σ ≡ σ '
(5)
uˆ3 = (k' −k ) k' −k = Q Q
π = kˆ × σ ; π ' = kˆ' × σ '
Scattering Techniques B5.5

Fig. 1: Illustration of the definition of the co-ordinate system and the basis vectors used to
describe the polarisation dependence of x-ray scattering.

In this basis the matrices in (4) can be written as

- <fM> for the magnetic part:

to \ from σ π
σ' S2 ⋅ cos θ [( L1 + S1 ) ⋅ cos θ + S3 ⋅ sin θ ]⋅ sin θ
[ ]
(6)
π' [−( L1 + S1 ) ⋅ cos θ + S3 ⋅ sin θ ]⋅ sin θ 2 L2 ⋅ sin 2 θ + S2 ⋅ cos θ

- <fC> for charge scattering:

from
to \ σ π
σ' ρ (Q) 0 (7)
π' 0 ρ (Q)(cos2 θ )

Here Si = Si(Q) and Li = Li(Q) (i=1, 2, 3) denote the components of the Fourier transform of
the magnetisation density due to the spin and orbital angular momentum, respectively. ρ(Q)
denotes the Fourier transform of the electronic charge density distribution.

As can be seen from (4), magnetic scattering is a relativistic correction to charge scattering.
For coherent elastic Bragg scattering, the ratio between the magnetic and the charge ampli-
tude is determined by the momentum transfer and therefore we have written the pre-factor for
the magnetic amplitude in the cross-section (4) as λC/d which emphasises that for a given
Bragg reflection the ratio between magnetic and charge scattering is virtually independent of
photon energy, at least to within the approximations leading to (4).

(4) contains three terms: pure Thomson-scattering, purely magnetic scattering and an interfer-
ence term. The latter becomes important if charge- and magnetic scattering occur at the same
position in reciprocal space, which is the case for ferromagnets. Note, however, that the
prefactor "i" in front of the magnetic scattering amplitude means that magnetic scattering is
shifted in phase by π/2 as compared to charge scattering. Therefore if both amplitudes, <fC>
and <fM> are real, the interference term vanishes. The interference can only be observed, if
one of the amplitudes contains an imaginary part (e. g. non centrosymmetric structures or
photon energy close to an absorption edge for charge scattering) or if circular polarised radia-
B5.6 Th. Brückel

tion is used. The importance of the interference term for ferromagnets becomes evident, if we
consider the ratio between magnetic and charge scattering amplitudes. An estimate for this
ratio can be given as:
< f M > λC N M f M
~ ⋅ ⋅ S (8)
< fC > d N⋅ f
Here, N(NM) and f(fM) denote the number and the form factor of all (the magnetic) electrons,
S the expectation value of the spin quantum number. Using appropriate values for the pa-
rameters in (8), one finds that the amplitude for magnetic scattering is typically three orders of
magnitude smaller than the amplitude of charge scattering, resulting in an intensity ratio of
10-6 between pure magnetic and pure charge scattering. It is not practical to measure a 10-6
effect in intensities. Therefore for ferromagnets, where charge and magnetic scattering coin-
cide in reciprocal space, the interference term between charge and magnetic scattering is the
leading term after charge scattering. In a very similar way to flipping-ratio measurements in
neutron scattering, the direction of the magnetisation (or the incident photon polarisation) is
changed periodically to change the sign of the interference term and thus extract this term
from the pure charge scattering.

(6) and (7) show that magnetic scattering can be discriminated from charge scattering by a
polarisation analysis experiment, where the off-diagonal terms σ→π' or π→σ' are being
measured. Finally, (6) shows that the spin and orbital contributions have different angular-
and polarisation dependencies and can therefore be distinguished in principle.

We have sketched a derivation of the non-resonant magnetic scattering cross sections starting
from non-relativistic quantum mechanics and applying perturbation theory up to second order.
It should be noted that the scattering cross-section can also be derived in a purely classical
theory [10]. It turns out that the classical calculation reproduces the quantum mechanical
cross-section for the spin part, but not for the orbital part. De Bergevin and Brunel [3] have
drawn a simple diagram, representing the various interaction processes in such a classical
model. This diagram is reproduced as figure 2.

The first process shown in figure 2 is the classical charge or Thompson scattering: an electro-
magnetic wave is incident on a free electron and due to the Coulomb force between the elec-
tric field vector and the charge of the electron, the electron is accelerated into a harmonic os-
cillation and re-radiates electric dipole radiation. The three other processes only appear if the
electron carries a spin momentum, i. e. these processes give rise to magnetic x-ray scattering.
The second process in figure 2 arises from the same Coulomb interaction with the incident
electromagnetic wave. The accelerated spin moment gives rise to re-radiation of magnetic
quadrupole radiation. In the third and fourth process of figure 2, the interaction with the inci-
dent electromagnetic field is between the spin moment and the magnetic field vector.

From figure 2, the polarisation dependence of charge and magnetic scattering becomes imme-
diately evident. In charge scattering, the polarisation of the incident wave is conserved. From
our simple classical pictures, it is immediately evident that the matrix (7) has to be diagonal.
The cos 2θ factor for π→π’-scattering is simply explained by the projection of the accelera-
tion vector onto a plane perpendicular to the observation direction. Figure 2 shows that in
contrast to charge scattering, the polarisation can indeed change for magnetic x-ray scattering.
Therefore the existence of off-diagonal terms in the matrix (6) can easily be motivated from
the classical picture figure 2. Polarisation analysis allows us to clearly distinguish charge and
magnetic scattering.
Scattering Techniques B5.7

interaction re-radiation
E
-e E-dipole
force
E
σ -eE σ

H
-eE -e H-quadr.
E µ
σ π,σ

E
-e
grad(µ H) E-dipole
µ
σ π
H
µ H
torque H-dipole
Hxµ
σ π
H

Fig. 2: Illustration of the processes leading to scattering of x-rays by the charge (top) and
the spin moment (bottom three) of the electron in a classical picture (from [3]).

For what follows it is of interest to examine the high energy limit of the purely magnetic dif-
fraction cross section. It can be easily calculated from (6). In this limit, the cross-section be-
comes virtually independent of polarisation and is sensitive only to the component of the Fou-
rier transform of the spin density distribution perpendicular to the scattering plane:
⎛λ ⎞
2
dσ 2
= r02 ⎜ C ⎟ S2 (Q) (9)
dΩ magnetic ⎝ d ⎠
Thus at high photon energies around 100 keV, the pure spin density distribution becomes ac-
cessible without polarisation analysis, while in neutron diffraction always the sum L+2S is
being measured.

If the x-ray energy is tuned to the absorption edge of magnetic elements, resonance phenom-
ena occur due to second order perturbation theory [11]:
dσ < a | O + (k ' ) | c >< c | O(k ) | a >
~∑ (10)
dΩ c Ea − Ec + hω − iΓ / 2
Here |c> denotes an intermediate excited state with energy Ec, hω the photon energy and Γ
the level width of the excited state due to the finite lifetime (Γ ⋅ τ ≈ h) . The operator O(k) is
given by the expression:
O(k ) = ∑ e i k ⋅r i (P i − ih(k × s i )) (11)
i
(10) gives rise to anomalous dispersion, i. e. an energy dependence of the charge scattering, as
well as to resonant magnetic scattering. The operator (11) can be expanded in a multipole se-
ries. It turns out that in the x-ray regime, the spin and orbital contributions can be neglected in
most cases, and only the electric multipole terms have to be retained. These electric multipole
B5.8 Th. Brückel

(predominantly dipole and quadrupole) operators induce virtual transitions between core
levels and unoccupied states above the Fermi energy with subsequent reemission of a photon.
These processes become sensitive to the magnetic state in exchange split bands due to the
difference in occupation of minority and majority bands leading to so called resonance ex-
change scattering XRES [12] as illustrated schematically in figure 3.

Due to the resonance denominator in (10), resonance enhancements occur at the absorption
edges of the magnetic elements. The strengths of these enhancements for XRES depend
mainly on three factors:

Fig. 3: Schematic illustration of the second order perturbation process leading to XRES in
the case of a lanthanide metal, e. g. a Gd3+ - ion.

1. The magnitude of the transition matrix element. Dipole transitions between states |a> and
|c> differing in orbital angular momentum quantum number by ∆L = 1 are generally
stronger than quadrupolar transitions with ∆L = 2. A large overlap of the wave functions
|a> and |c> favours large transition matrix elements. In contrast, transitions from "s" core
levels to "p" or "d" excited states do not show large resonance enhancements due to the
small overlap of the wave functions.
2. The difference in the density of empty states above the Fermi level for minority and ma-
jority spin states. To give an example: in lanthanide metals, the 5d bands are spin po-
larised due to the magnetic 4f states. However, the exchange splitting in the 5d is much
weaker as compared to the 4f states and dipolar transitions 2p → 5d are sometimes not
much stronger than quadrupolar transitions 2p → 4f.
3. The strength of the spin-orbit coupling in the ground- and excited states. Only due to this
coupling do the electric multipole transitions become sensitive to the spin magnetism.

Using these criteria, we can qualitatively categorise the possible transitions according to the
magnitude of the resonance enhancement, see Tab. 1. Here we define the term "resonance
enhancement" as the ratio between the intensity of magnetic Bragg peaks in the maximum of
the resonance relative to the intensity for non-resonant magnetic scattering.
Scattering Techniques B5.9
elements edge transition energy range [keV] resonance comment
strength
3d K 1s → 4p 5-9 weak small overlap
3d LI 2s → 3d 0.5 - 1.2 weak small overlap
3d LII, LIII 2p → 3d 0.4 - 1.0 strong dipolar, large overlap,
high spin polarisation of 3d
4f K 1s → 5p 40 - 63 weak small overlap
4f LI 2s → 5d 6.5 - 11 weak small overlap
4f LII, LIII 2p → 5d 6 - 10 medium dipolar
2p → 4f quadrupolar
4f MI 3s → 5p 1.4 - 2.5 weak small overlap
4f MII, MIII 3p → 5d 1.3 - 2.2 medium dipolar
3p → 4f to strong quadrupolar
4f MIV, MV 3d → 4f 0.9 - 1.6 strong dipolar, large overlap,
high spin polarisation of 4f
5f MIV, MII 3d → 5f 3.3 - 3.9 strong dipolar, large overlap,
high spin polarisation of 5f

Tab. 1: Magnitude of the resonance enhancements for XRES for some elements relevant for
magnetism. Only order of magnitude estimates are given with "weak" corresponding
to a factor of about "100", "medium" to about "102" and "strong" to ">103".

Tab. 1 only lists some of the most prominent examples. It demonstrates that the resonance
enhancements for 3d transition metal ions is negligible in the hard x-ray regime (e. g. [13]),
while it can be strong for soft x-rays. Unfortunately, at wavelengths of 12 to 30 Å, atomic
resolution cannot be obtained under normal conditions. However, the transition metal LII and
LIII edges turn out to be extremely important for the investigation of magnetic thin films and
nanostructures (e. g. [14]). For the 4f elements, resonance enhancements of about two orders
of magnitude are observed in the hard x-ray range at the LII and LIII edges (e. g. [15]). At
these edges, dipolar transitions are in general dominant, but quadrupolar transitions can be
significant. The so-called "branching ratio", i. e. the ratio between resonance enhancement at
the LII edge and the LIII edge has a tendency for a systematic variation along the rare earth
series. While it is close to 1 for rare earth ions with seven 4f electrons, the LIII resonance is
generally stronger for ions with more than seven 4f electrons while the LII resonance tends to
be stronger for less than half filling of the 4f shell. As in the case of the 3d transition metals,
the soft x-ray range with the MIV and MV resonances is of importance for magnetic
nanostructures [16]. At the MIV edge of actinides, the intensity gain due to XRES can be as
high as seven orders of magnitude [17]. Finally, we have not listed the 4d and 5d transition
metal elements in Tab. 1, even so resonance enhancements at the LII and LIII edges can be so
large that surface magnetic x-ray diffraction becomes possible, e. g. in Co3Pt (111)
( LPtIII ≈ 11.5 keV : 2p3 / 2 → 5d ) [18]. We can conclude that XRES can provide large intensity
gains for magnetic x-ray scattering, allows a spectroscopy of the exchange split empty states
above the Fermi level and renders magnetic diffraction sensitive to the magnetic species.

Let us come back to the explicit form of the cross-section, including resonant magnetic scat-
tering. We start from (10), which gives the general form of the cross-section for anomalous
scattering. In what follows, we will neglect the spin dependent part and limit ourselves to
electric dipole transitions. Detailed derivations are given in [11], [12] and the polarisation
dependence, also for the case of electric quadrupole transitions, is discussed in [19]. Anoma-
lous scattering becomes relevant close to the absorption edges of the elements. Then, an
B5.10 Th. Brückel

energy dependent amplitude has to be added to the expression (4) for the scattering cross-sec-
tion. In dipole approximation, this amplitude reads:
f resE1 ( E ) = f o ( E ) + f circ ( E ) + f lin ( E ) (12)
with
f 0 ( E ) = (ε '⋅ε ) [ F+11 + F−11 ]
f circ ( E ) = i (ε '×ε ) ⋅ m [ F−11 − F+11 ] (13)
f lin ( E ) = (ε '⋅m)(ε ⋅ m)[2 F − F − F ] 0
1 1
+1
1
−1

f0 is independent of the magnetic state (i. e. the conventional anomalous charge scattering),
while fcirc and flin are the amplitudes connected for the special case of forward scattering with
circular and linear dichroism, respectively. All three amplitudes have different polarisation
properties. fcirc depends linear on the magnetic moment m, while flin depends quadratic on m.
Therefore for antiferromagnets, only fcirc gives a contribution at positions in reciprocal space
separated from the main charge reflections by the magnetic propagation vector. Finally, for a
simple excitation into one atomic-like level, the energy dependence of the amplitudes is con-
tained in the oscillator strengths
αM
FM1 =
(ω − ωres ) − i Γ 2h (14)

Here ω denotes the photon energy, ω res the position of the absorption edge and Γ the reso-
nance width. The phenomenological parameter αM gives a measure for the amplitude of the
resonance and stands for the product of the transition matrix elements.

After this discussion of the cross section for non-resonant magnetic x-ray diffraction and for
resonance exchange scattering, we will now demonstrate the possibilities of these techniques
by some illustrative examples. Just for convenience most of these examples are from our own
research, while we are very well aware of the beautiful work done by other groups. However,
before presenting these examples, we have to discuss the experimental conditions for obser-
vation of magnetic x-ray scattering.

3 Experimental consideration
Figure 4 shows the sketch of a synchrotron x-ray source, indicating the special properties of
the radiation.
small source
size
wiggler clean ultra-high
properties vacuum source
calculable

time structure
undulators
50 m

highly collimated
intense
continuous
spectrum

.
polarised
.
5 mm

Fig. 4: Schematic sketch of a synchrotron x-ray source.

Scattering Techniques B5.11

In a synchrotron x-ray source, electrons or positrons are circulating in bunches in a vacuum

chamber at a velocity close to the speed of light. They are kept in the curved storage ring by
magnetic fields that accelerate the charged particles with the Lorentz-force in a direction
transverse to their propagation direction. The radially accelerated particles emit electromag-
netic radiation. Due to their relativistic movement, this radiation is very much collimated in
forward direction with an opening angle θ ≈ 2 / γ , where γ = E / m0 c 2 with E = particle
energy, m0c2 ≈ 511 keV = energy equivalent of the rest mass of an electron. This is a conse-
quence of the so-called relativistic Doppler effect, which results from the transformation of
the emitted electromagnetic field distribution from the moving eigenframe of the charged
particles into the laboratory system. The high collimation of the beam is essential for the ob-
servation of magnetic scattering, since one can conduct the experiment in such a way that
only the desired scattering channel is excited, which leads to a significant better peak-to-
background ratio as compared to conventional x-ray tubes. Moreover, the beam is highly po-
larised with a linear component in the orbital plane of the storage ring and elliptical polarisa-
tion of opposite sense above and below the orbital plane. Again, this polarisation of the beam
is essential for magnetic scattering, since it allows one to separate experimentally charge-,
spin- and orbital contributions to the scattered intensity, see chapter 2. Synchrotron radiation
exhibits a continuous "white spectrum". This property makes possible investigations, where
the x-ray energy has to be tuned to the absorption edge of an element in the compound under
investigation. While at x-ray tubes intense radiation is emitted only for the characteristic lines
of the anode material, one can choose a narrow energy band width of the synchrotron radia-
tion, just by applying a grating (for soft x-rays) or crystal-Bragg (for hard x-rays) monochro-
mator. Modern synchrotron radiation sources do not just use the radiation of the so-called
bending magnets, but mainly employ insertion devices, such as wigglers and undulators as
even more powerful radiation sources. Both types of insertion devices are permanent magnet
structures, which create an alternating field, perpendicular to the orbital plane of the storage
ring. This field leads to a sinusoidal movement of the charged particles within the insertion
devices. At each reversal point, radiation is emitted similar to the radiation from the dipolar
magnets. The difference between wigglers and undulators lies in the amplitude of the sinusoi-
dal movement. This amplitude is large for wigglers, giving rise to an incoherent superposition
of the radiation from the various magnet poles. Therefore, the wiggler spectrum resembles the
spectrum of a bending magnet, but the intensity is amplified by a factor corresponding to the
number of magnet poles. In an undulator, where the amplitude of the sinusoidal movement is
smaller, the radiation from the various poles superimposes coherently in forward direction.
This leads to a line-spectrum in energy with odd harmonics on axis. The intensity within these
lines is proportional to the square of the number of magnet poles in the undulator. The posi-
tion of the undulator harmonics can be tuned by the strength of the magnetic field or the un-
dulator period. In practical devices, the magnetic structure is realised by permanent magnets
and the magnetic field strength can be varied by opening or closing the undulator gap. In an
undulator, coherence is achieved for the following condition:
λ ⎛ 2
⎞
λ = u2 ⎜1 + K + γ 2θ 02 ⎟ (15)
2γ ⎝ 2 ⎠
Here λ u denotes the undulator period, γ the particle beam energy, expressed in units of the
energy equivalent of the electron restmass mec2 ≈ 511 keV, θ0 is the angle of emittance and K
the so-called wiggler or undulator parameter
λ eB
K= u = γ ⋅θ w (16)
2πme c
B5.12 Th. Brückel

Here, B is the amplitude of the magnetic field on axis and θw the maximal angle of deviation
of the particle against the nominal orbit. In figure 5, the spectra of bending magnet, wiggler
and undulator are compared. Since the undulator radiation is quasi-monochromatic with a
energy width inverse proportional to the number of magnetic poles and has a high collimation
with an opening angle of 2 / γ (about 0.1 mrad for a 5 GeV machine) in both directions,
undulator radiation resembles already very much a laser beam known from optics. Its unique
properties make the undulator the x-ray source of the choice for magnetic x-ray scattering
applications.

Fig. 5: left: comparison of the emittance characteristics for a bending magnet, a wiggler
and an undulator. right: Frequency spectrum of undulator radiation. The dashed line
in d corresponds to a wiggler spectrum. (from Saile, 23. IFF-Ferienkurs 1992)

Worldwide, there exist three so-called "third generation synchrotron sources" for the hard x-
ray regime: the European Synchrotron Radiation Facility ESRF in Grenoble, SPRING-8 in
Japan and the Advanced Photon Source (APS) in Argonne, USA. All three facilities have
storage rings with a circumference of about 1 km and feature beamlines dedicated to magnetic
scattering. The typical set-up of such a beamline is as follows: the beam from an undulator
source is tailored by slit systems. Mirrors are used to suppress higher harmonics in the x-ray
beam and/or to focus the beam onto the sample in the experimental hutch. The x-ray energy is
selected by a double crystal Bragg monochromator. To handle the high heatload of several
hundred W/mm2, liquid nitrogen cooled silicon crystals are employed for monochromatisa-
tion. Usually the 111 reflection is chosen, since the second harmonic is largely suppressed for
the Si crystal structure by the diamond glide planes. This beamline optics is situated in a so-
called optics hutch with lead walls serving as biological radiation shielding. In the optics
hutch, the x-ray beam is prepared with a desired properties and then enters the experimental
hutch, where the actual scattering experiment is situated. The polarisation of the incident
beam is monitored with two orthogonal scintillation detectors, which use the emittance
characteristic of charge scattering, according to (7). The incident photon flux is being moni-
tored with ionisation chambers. The sample is situated in the center of an Eulerian cradle,
which allows arbitrary orientation of the sample in space. On the detector arm, an analyser
unit is mounted, which serves to suppress background, enhance the resolution in reciprocal
space, but also allows an analysis of the polarisation of the scattered beam. Such a polarisa-
tion analyser is shown in some more detail in figure 6.
Scattering Techniques B5.13

Fig. 6: Polarisation analyser according to [37] and measurement of the polarisation of a

wiggler beam. 1: motorized double slit system; 2: evacuated flight tube; 3: base plate
with three manual degrees of freedom for alignment; 4: goniometer for the α move-
ment (see text); 5: entrance slit; 6: goniometer for the ω movement of the analyser
crystal; 7: tilt of the analyser crystal; 8: aluminium housing of the analyser; 9: go-
niometer for the analyser 2θ movement; 10: detector aperture slits; 11: NaI detector.

To analyse the polarisation of the scattered beam, one has to find a crystal with a reflection at
a lattice d-spacing corresponding to 1/2 of the wavelength of the radiation d = λ / 2 , allowing
for 90° charge reflection. At 2 θ A = 90°, the formfactor and the Debye-Waller-factor strongly
reduce the reflected intensity. Therefore, the analyser crystal should have a very high peak
reflectivity. On the other hand, it should accept the whole divergence of the beam, diffracted
from the sample. Therefore, the best choice is a mosaic crystal, which compromises between
high peak reflectivity and a sufficient width of the rocking curve. Often, pyrolytic graphite
(PG002 and higher harmonics) is chosen as analyser, which, however, gives only very limited
choice of wavelengths. Under the conditions of 90° charge scattering, only σ → σ scattering
occurs. By rotating, the analyser crystal and detector around the beam axis ( α -movement 4 in
figure 6), the degree of polarisation of the x-ray beam can be determined. As an example,
figure 6 shows the measurement of the polarisation of the monochromatic primary beam from
a wiggler device at the DORIS storage ring in Hamburg. For an x-ray energy of 10.46 keV, a
PG008 analyser crystal fulfills the 90° condition. The integrated intensities scattered from this
analyser as a function of the angle α between the vertical and the scattering plane of the ana-
lyser, are being plotted together with a refinement according to
I (α ) = H cos 2 (α − α 0 ) + V sin 2 (α − α 0 ) (17)
Here, H and V denote the intensities corresponding to horizontal and vertical polarisation,
respectively. α0 describes the possible tilt. The degree of horizontal polarisation is given by
P = ( H − V ) /( H + V ) (18)
In the case of this second generation synchrotron radiation source and for complete open en-
trance slit, a polarisation P of about 84 (1 %) was determined. Undulator radiation from mod-
ern 3rd generation storage ring sources can have a degree of linear polarisation in the orbital
plane above 99.5 %. At these modern beamlines, the incident flux can exceed 1013 pho-
tons/mm2s.

According to (6), polarisation analysis in σ → π geometry is the most efficient way to distin-
guish between magnetic and charge scattering and thus also to suppress possible background
B5.14 Th. Brückel

from the sample. If no polarisation analyser is employed, background from Compton scat-
tering or sample fluorescence has to be suppressed to enhance the peak-to-background ratio.
A very efficient way to do this is by employing energy dispersive solid state detectors. They
allow one to set an electronic window for the spectro-amplifier and thus to discriminate the
elastic scattered photons from the background of inelastic scattering.

4 Resonance exchange scattering

As an example of the effect of resonance exchange scattering, we show in figure 7 raw data
taken at the W1 beam line at HASYLAB for the LII resonance of GdS [15]. Due to the high
absorption cross section of Gd for thermal neutrons, no detailed neutron diffraction studies
exist for GdS. With x-rays a comfortable count rate of about 3000 photons per second was
obtained for the 9/2 1/2 1/2 reflection on resonance and we could verify the assumed type II
antiferromagnetic ordering on the fcc lattice. While resonant exchange scattering can in prin-
ciple give information about the density of unoccupied states above the Fermi level, most
resonance line shapes can well be approximated with the simple two level model of equation
(14). This is also true for the resonance shown in figure 7, where only a small asymmetry re-
mains after absorption correction [15]. There are, however, examples of much more structured
resonances with double and multiple peaks as a function of energy. Examples are the K-edge
resonance of manganese in the perovskite type compound RbMnF3 [13] or the L-edge reso-
nances in the rare-earth metal Tb [28].
7935 eV
250
140 GdS
7933 eV 9/2 1/2 1/2

flourescence yield [a.u.]

120 L II edge 200
3000 7938 eV
peak intensity [a.u.]

100
150
2000 energy 80
photons/sec

7930 eV 7942 eV 60 100

40
1000 50
20
7924 eV 0 0
0 7920 7925 7930 7935 7940 7945 7950
5 5.2 5.4 5.6 5.8
energy [eV]
ω

Fig. 7: The resonance enhancement of the magnetic signal of the 9/2 1/2 1/2 superstructure
peak of GdS at the Gd LII edge measured at 4.2 K. The left diagram shows raw data
of rocking curve scans at various photon energies. The right diagram shows as a
function of photon energy the peak intensities together with the structure of the ab-
sorption edge.

The strength of resonance exchange scattering that makes it unique compared to all other
techniques is that it combines spatial resolution (in reciprocal Fourier space) with element
specificity. By choosing a resonance, where an enhancement of several orders of magnitude
for a specific magnetic element is obtained, non-resonant scattering becomes negligible and
only the magnetic pair correlation functions for this element in question are observed. This
can be nicely demonstrated for the mixed crystal series Gd1-xEuxS. While GdS is an antifer-
romagnetic metal, EuS is a ferromagnetic insulator. For some intermediate concentration, a
metal-insulator transition occurs. Moreover, since the system exhibits competing magnetic
interactions (ferromagnetic versus antiferromagnetic), frustration occurs, i. e. not all magnetic
bonds can be satisfied simultaneously. Frustration combined with disorder typical for a sto-
Scattering Techniques B5.15

chastic occupancy of the rare earth site in the solid solution, leads to a spin glass phase with-
out magnetic long range order separating the ferromagnetic from the antiferromagnetic phase.
Some questions that can be tackled with XRES are: can we observe the frustration mechanism
and what is the magnetic microstructure of the long range ordered and of the spin glass phase?
The answer to these questions lies in the study of the magnetic correlations with element
specificity [29, 30].

Fig. 8: Absorption corrected energy dependence of the intensity of the 1/2 1/2 9/2 magnetic
Bragg reflection at the Eu and Gd LII and LIII edges for a Gd0.8Eu0.2S sample at 4 K.
The solid line is a fit with (14), the dashed line shows the absorption coefficient.

Figure 8 shows the resonance behaviour for a Gd0.8Eu0.2S sample at all 4 LII and LIII edges to-
gether with a fit assuming a simple atomic-like two level dipolar transition [29,30] (compare
eq. (14)). The resonances for the two different elements are well separated in photon energy
and the enhancement amounts to between one and more than two orders of magnitude com-
pared to non-resonant scattering. This is the reason for the element specificity, as is illustrated
in figure 9.
non resonant Gd resonance Eu resonance

: Gd3+ ion : Eu2+ ion

Fig. 9: Illustration of the effect of XRES for GdxEu1-xS: In the case of non-resonant x-ray
diffraction or neutron scattering, the magnetic order is observed independent of the
magnetic species. In the maximum of the resonance of Gd, only the ordering of the
Gd moments is visible due to the large resonance enhancement. At the Eu reso-
nances, only the order of the Eu moments is visible.
B5.16 Th. Brückel

1 Gd
3+ Gd0.8Eu0.2S single Eu spin
JGG<0 JGE<0 (5.2% for x=0.8)
LII - edges
0.8 2+ Eu-spin pair
Eu (7.9% for x=0.8)
?
neutrons JEE>0
0.6
m ? ? ?
2+
Eu pairs JEE>0 Eu-spin triplet
? (4.9% for x=0.8)
0.4
single Eu2+
0.2

0
0.0 10 20 30 40 50 60
temperature [K]

Fig. 10: Temperature dependence of the element specific sublattice magnetisation for
Gd0.8Eu0.2S together with neutron data and a model refinement (left). The refinement
is based on the frustration model illustrated on a 2d lattice on the right.

Figure 10 shows the temperature dependence of the sublattice magnetisation determined with
neutron diffraction (filled triangles), with XRES at the Gd LII edge (small circles) and with
XRES at the Eu LII edge (open squares) [29, 30]. The sublattice magnetisation was obtained
from rocking curves as the normalised square root of the integrated intensities. This is the
correct procedure for the case of neutron diffraction. However, in the case of XRES, the in-
tensity depends on transition matrix elements (compare (14)), which can, in principle, change
with temperature. Scattering in second order perturbation theory is a priori not directly related
to the order parameter. Therefore the neutron data were taken as a cross check: it turns out
that the weighted sum of the XRES curves for Eu and Gd matches nicely the neutron curve,
indicating that with XRES we measure indeed the sublattice magnetisation for each species
Eu2+ and Gd3+ individually.

The surprising observation is that the sublattice magnetisation has a different temperature de-
pendence for the two ions Eu2+ and Gd3+, even so both have the same 8S7/2 electronic ground
state and they both are embedded “in the same sea of conduction electrons”. An explanation
for this observation can be given, if we assume frustration effects to occur. In a very simple
model, we start from an isotropic Heisenberg Hamiltonian
H = −∑ J ij S i ⋅ S j (19)
with nearest neighbour interactions only, place the two ions at random on a simple cubic lat-
tice and assume the exchange interaction between pairs of Gd-Gd and Gd-Eu to be antiferro-
magnetic and between pairs of Eu-Eu to be ferromagnetic. This will lead to frustration for
pairs, triplets etc of Eu spins within the surrounding Gd matrix, see figure 10 (right). To cal-
culate these frustration effects for the case of small Eu concentrations, we rewrite the Hamil-
tonian (19) in the form of a sum with one term for the Gd subsystem, including single Eu
spins, and another term for the Eu “clusters” (pairs, triplets etc.). The size of the Eu “clusters”
follows a binomial distribution. The Gd subsystem is treated in a mean field theory, while the
Hamiltonian for the Eu pairs in the mean field of the surrounding Gd ions can be diagonalised
exactly [29,30]. The result is shown in figure 10 (left): we obtain a surprisingly good agree-
ment between theory and experiment, indicating that the abnormal temperature dependence of
the Eu subsystem is actually due to frustration effects. With Monte Carlo simulations we can
employ a more realistic interaction model with exchange up to second neighbours on the fcc
lattice and obtain similar results. We observe a change from a collinear magnetic structure for
GdS to a canted structure for the antiferromagnetic mixed crystals, see figure 11.
Scattering Techniques B5.17

Fig. 11: Antiferromagnetic structure of Gd1-xEuxS for small x as obtained from Monte Carlo
simulations. While GdS has a collinear antiferromagnetic structure of type II on the
fcc lattice (i. e. all spins on [111] planes are parallel and the magnetisations for
neighbouring [111] planes are antiparallel), the doped crystals exhibit a canted spin
arrangement.

Eu: FM Gd: AF

Fig. 12: Linear scan in reciprocal space from (0.3 0.3 4.3) to (0.7 0.7 4.7) around the an-
tiferromagnetic Bragg peak position for Gd0.67Eu0.33S at a temperature of 4 K. The
energy was tuned to the Gd LII edge. A polarisation analyser with σ→π´ geometry
was used to suppress the background from charge scattering. The inserts show the
energy and temperature dependencies. On the right a model for the spin glass state is
shown.

The combined occurrence of frustration and disorder can result in a spin glass phase for in-
termediate concentrations. In fact, we do not observe long range magnetic order for a x=0.33
sample. Instead, short range antiferromagnetic correlations are observed for the Gd subsystem
with correlation lengths of about 40 Å, see figure 12. We could not detect any antiferromag-
netic correlations at the Eu LII or LIII edges. This leads us to the conclusion that a cluster spin
glass state is formed for intermediate concentrations: while the Gd spins show antiferromag-
netic correlations in regions of typical sizes of 40 Å, the Eu spins develop ferromagnetic cor-
relations. These spin “clusters” freeze in into arbitrary directions, leading to the spin glass be-
haviour.

To conclude this section on Gd1-xEuxS, we have shown which detailed information can be
obtained with an element-specific probe: for this mixed crystal series we could reveal the
frustration mechanism and verify that the spin glass state at intermediate compositions con-
B5.18 Th. Brückel

sists of a frozen cluster glass state. Such detailed information is not accessible with any other
probe.

We can give one other example for the element specific information XRES offers. This
example concerns thin film magnetism, namely Er/Tb rare earth super-lattices [31]. Such su-
per-lattices, for which the single Er or Tb layers are just a few mono-layers thick, can be
grown epitaxially with high quality on sapphire substrates. They show a rich magnetic phase
diagram. Phase transition temperatures are altered compared to the bulk and some phases ap-
pear, which are not present in the bulk materials.

Here we want to discuss only one special feature for a multilayer consisting of 150 double
layers of 20 mono-layers of Er and 5 mono-layers of Tb: [Er20 / Tb5] x 150. In a temperature
range between 80 K and 130 K, magnetic satellite peaks appear, which indicate that a helical
magnetic structure with a propagation vector close to the one for bulk Er is formed throughout
the multilayer, despite the fact that the Tb layers show basal plane ferromagnetic order. The
satellite peaks are resolution limited, which shows that the phase information for the magnetic
helix is carried through the ferromagnetic Tb layers. Can XRES help us to understand the
coupling mechanism of the Er layers through ferromagnetic Tb layers? Indeed it can: if we
tune the x-ray energy to the LII or LIII absorption edges of Tb, we enhance its magnetic scat-
tering and can thus observe specifically what happens within the Tb layers. Figure 13 shows a
plot of a XRES measurement as a function of x-ray energy and a reciprocal space co-ordinate.
These data tell us that XRES peaks at an energy corresponding to the Tb LIII edge and at a Q-
space position corresponding to the propagation vector of the Er magnetic helix.

Fig. 13: XRES from an [Er20/Tb5]x150 rare earth super lattice measured as a function of x-
ray energy in linear Q-scans along the [000l] axis. The intensity peaks at an energy
corresponding to the Tb LIII resonance energy and a Q-space position close to the
magnetic propagation vector of bulk Er.

However, from neutron scattering, we know that the 4f moments of Tb are aligned ferromag-
netically. This apparent contradiction can be resolved, if we consider the XRES process de-
picted schematically in figure 3. Polarisation analysis tells us that the relevant transitions are
dipolar in nature. Therefore, the intermediate states in the XRES process are the 5d conduc-
tion band states and with XRES, we do not observe the 4f magnetic order, but the spin polari-
sation of the 5d conduction electrons in the Tb layer. Apparently, these conduction band elec-
Scattering Techniques B5.19

trons form a spin density wave within the Tb layers corresponding to the helical magnetic or-
der in the Er layers. They carry the phase information through the Tb interlayers and thus al-
low the Er layers to develop a coherent helical magnetic structure throughout the super lattice.

To conclude this section, we have shown that the information provided by XRES is not only
element specific, but also specific to the electronic state of this element (4f versus 5d elec-
trons). In the case of rare earth superlattices, XRES allows a direct observation of the inter-
layer coupling mechanism. Spin polarisation in the conduction band is observed, as predicted
by the RKKY exchange mechanism.

So far, we have only discussed XRES from antiferromagnetic structures, for which magnetic
and charge scattering are well separated in reciprocal space. Due to the high brilliance of the
modern synchrotron radiation sources, even a small magnetic signal can be readily observed.
Polarisation analysis helps to distinguish charge and magnetic scattering and allows to sup-
press the charge background.

For ferromagnetic samples, the situation is quite different. Here charge and magnetic
scattering coincide. Since magnetic scattering is typically orders of magnitude weaker as
compared to charge scattering, it becomes difficult to observe. Combining (4) and (12), the
cross section for magnetic scattering takes the form:
dσ
= f = f C + f R (E )
2 2
(20)
dΩ
Here, fC denotes the amplitude for charge scattering, fR the amplitude for resonance exchange
scattering, and we have neglected non-resonant magnetic scattering close to a relevant ab-
sorption edge. Let us now chose the scattering geometry depicted in figure 14.

Fig. 14: Experimental set-up for the measurement of XRES from ferromagnetic samples. The
primary beam is polarised horizontally, the magnetic field is applied perpendicular
to the horizontal scattering plane and a polarisation analyser is employed to sup-
press the vertically polarised component.

In this geometry, |fcirc| of eq. (13) is maximised, since in saturation m (ε'×ε ) . The resonant
amplitude fcirc is proportional to the expectation value of the z-component of the magnetic
moment m z . Then the form of the cross section (20) will lead to the following form of the
ferromagnetic Bragg intensities:
B5.20 Th. Brückel

I = I mag mz2 + I int mz + I ch arg e (21)

There are three contributions to the scattered intensity I: a pure charge term Icharge, a pure
XRES term Imag and an interference term between charge and XRES scattering Iint. By
switching the direction of the external magnetic field, an asymmetry ratio
I↑ − I↓
Ra ≡ ↑ (22)
I + I↓
can be measured.

In the case of EuS and with the scattering geometry of figure 14, this asymmetry ratio is as
large as 67 % [32]. In fact, magnetic scattering, interference term and charge scattering be-
come quite comparable in magnitude. Figure 15 shows the LII resonance of EuS at 4 K.

Fig. 15: Absorption-corrected energy dependence of the (115) reflection measured at the Eu-
LII edge. Two measurements with magnetic fields of +0.5 T and –0.5 T were per-
formed at 4 K and one measurement at 40K well above the Curie temperature of 17
K. The solid line represents a refinement, see text.

Such data can be used to obtain spectroscopic information about the electronic states at the
Fermi energy. In [32] a model has been refined to the data, which assumes that the empty
states available for spin-up and spin down electrons are narrow, but shifted by an exchange
energy ε. The value for this exchange splitting was determined to ε = 0.27(1) eV.

To conclude this section, we have shown that XRES can also be measured from ferromag-
netic samples. This allows one to measure element specific hysteresis loops. Spectroscopic
information about the intermediate states close to the Fermi energy can be obtained from the
shape of the resonance curves. In order to determine values such as the exchange splitting
unambiguously, it is however, not sufficient to just measure the asymmetry ratio (22). Instead,
several data sets with the energy dependencies for positive and negative field and for pure
charge scattering have to be refined.

5 Non-resonant magnetic x-ray diffraction

Resonant magnetic x-ray scattering has the convenience of easily detectable signals due to the
resonance enhancement and provides element specific information. Therefore many magnetic
x-ray diffraction studies nowadays deal with resonance exchange scattering. However, for
transition metal ions, only K-edges lie in the range of hard x-ray wavelengths, where atomic
Scattering Techniques B5.21

resolution is achievable. Due to the dipolar and quadrupolar selection rules and the small
overlap between core 1s-states and the magnetic sensitive 3d or 4p energy bands, resonance
enhancements are negligible at transition metal K-edges and one is left with neutron- or non
resonant magnetic x-ray scattering. These techniques have the advantage that they measure
directly the order parameter, in contrast to resonant scattering, where transition matrix ele-
ments are involved, which are not known a priori. Moreover, the form of the cross section (6)
suggests that spin S and orbital L angular momentum can be determined separately by means
of polarisation analysis. Take the example of a collinear antiferromagnet for which charge –
and magnetic reflections are well separated in reciprocal space and for which we can align the
moments along the û2 – axis, i.e. S(Q) = (0,S(Q),0) and L(Q) = (0,L(Q),0). At the antiferro-
magnetic reflections charge scattering vanishes and, according to (6), magnetic scattering only
occurs in the σ→σ’ and the π→π’ polarisation channels. In the former, scattering is sensitive
to S(Q) only, in the latter, it is sensitive to a linear combination of S(Q) and L(Q), thus al-
lowing a unique determination of the ratio between both contributions. While such a model
independent separation of S and L cannot be achieved with neutron scattering due to the fact
that the neutron cross section is proportional to L+2S, it provides very important information
e.g. to verify band structure calculations. Such a separation of S and L by means of polarisa-
tion analysis of non-resonant magnetic x-ray diffraction has for example been done in Ho [5],
NiO [20] or Cr [21].

When discussing non-resonant magnetic x-ray diffraction, one has to start with the beautiful
pioneering work by Gibbs et al. [4] and Bohr et al. [34] on Ho metal. Together with the fol-
lowing work [35], this is one of the most comprehensive studies of magnetic x-ray scattering
so far undertaken. The basic magnetic structures in the heavy rare-earth metals are known for
quite a long time from the pioneering experiments by Koehler et al. The results are illustrated
in figure 16.

Fig. 16: On the left, the hexagonal close packed hcp crystal structure realised in the heavy
rare-earth metals is illustrated. On the right, the magnetic structures of the heavy
rare-earth metals are sketched. From left to right: basal-plane ferromagnet, helix
structure, cone structure and the c-axis modulated CAM structure.

The moments in any particular hexagonal base-layer are parallel and only the relative align-
ment of different planes, when propagating along the hexagonal c-direction are illustrated. In
Holmium, a helix structure is realised below the Néel temperature of 132 K down to 20 K,
where a cone structure develops. In the high temperature region, the modulation period in-
B5.22 Th. Brückel

creases continuously with decreasing temperature, which indicates that the helix is incom-
mensurate with the crystal lattice, i. e. the ratio between magnetic period and lattice period
cannot be expressed as a simple rational number. At 20 K, a lock-in transition occurs and the
magnetic modulation becomes commensurate with the chemical lattice with a period of 6 unit
cells, i. e. 12 atomic layers, corresponding to τ = 1/6. The beauty of the x-ray experiment was
to show that much of what was originally thought to be an incommensurate phase in fact
contains a series of subtle lock-in transitions. These details could be revealed because of the
high wavevector resolution of 10-3 Å-1 for the synchrotron x-ray experiments. Subsequently,
these observations have led to a new model of magnetism in Holmium and other rare-earth
metals. In the experiments by Gibbs et al. the scattering vector Q was aligned parallel to the
axis of the helical magnetic structure. In this case, we expect that the main charge peaks along
c* are accompanied by a pair of satellites at positions Qhelix=Gn±τ, where the turn angle from
plane to plane of the helix is given by Φ = τ·d. Figure 17 shows the original measurement by
Gibbs et al. in the vicinity of the 0004 Bragg peak in Ho.

Fig. 17: Scattering in the vicinity of the 0004 Bragg peak in Holmium at 17 K. The open cir-
cles show the scattering without polarisation analysis, while the filled circles show
the scattering observed with a polarisation analyser aligned to observe only σ-π'
scattering (from [4]).

Without polarisation analysis two peaks are visible: one sharper satellite reflection and one,
which is significantly broadened. Both reflections occur at rational positions indicating a
commensurate structure. According to (6) and (7), magnetic scattering can be identified with
the help of a polarisation analyser, aligned to observe the off-diagonal σ-π' scattering. This
experiment is shown by the filled circles. Obviously, only the sharp peak is of magnetic ori-
gin, while the broader feature has to be assigned to charge scattering. For a pure spiral struc-
ture, no strain wave should occur due to magnetoelastic interactions. To resolve the puzzle,
we can look at the temperature dependence of the magnetic modulation vector in holmium,
deduced from x-ray and neutron data (see figure 18). With the high resolution x-ray data,
several lock-in transitions as well as a hysteresis behaviour can be observed at low tempera-
ture. Figure 19 shows several scans in the low temperature range.
Scattering Techniques B5.23

Fig. 18: The temperature-dependent magnetic modulation wavevector of holmium from x-ray
and neutron data according to [4].

Fig. 19: Magnetic and lattice modulation diffraction peaks in holmium at various tempera-
tures (taken from [34]). The diagram on the right illustrates the model for magnetic
ordering that can describe the observations.

A consistent model that can describe these observations assumes the existence of discommen-
surations. The spin slip structure depicted on the right of figure 19 results from a competition
between exchange energy and the sixfold basal plane anisotropy. According to this model, the
basal plane anisotropy determines the orientation of the magnetic moment of two adjacent
layers, which adopt a parallel magnetisation. The magnetisation of the next double layer is
rotated by 60° and lies again along an easy in-plane axis. This pattern repeats until at one
B5.24 Th. Brückel

point exchange energy wins and a spin slip is introduced with only one layer and not a double
layer pointing along the easy axis. The change in magnetoelastic coupling at spin slip sides
causes a small distortion of the chemical lattice, with a period equal to the spin slip period λS.
As figure 19 shows, the magnetic period λM and the spin slip period λS are different, giving
rise to the two peaks observed. While the magnetic period is long range ordered, the broad
charge peak indicates that the spin slip structure is short ranged with a finite correlation
length. With this spin slip model, Bohr et al. [34] could calculate the allowed wavevectors for
magnetic and spin slip scattering and there is a remarkable agreement with the experimental
data. We do not want to discuss this example in more detail, but conclude that the high reso-
lution of magnetic x-ray scattering compared to magnetic neutron scattering immediately
gave rise to a much deaper understanding of magnetic structures and interactions in rare-earth
metals.

The second example we want to discuss deals with the separation of spin and orbital angular
momentum in the monoxides of 3d transition metals [36] and we have selected in particular
the example of NiO [20]. Why are we interested in these very simple compounds? After all,
MnO was the first material for which the antiferromagnetic structure has been determined
with neutron scattering, a main corner-stone for the nobel prize of Cliff Shull. The reason is
that electron correlation effects become particularly obvious in the transition metal oxides.
Prominent examples of systems, which cannot be described in a simple Fermi liquid picture
are the high TC cuprates or the colossal magnetoresistance manganites, which show some very
interesting and unexpected behaviour, due to electron correlations. MnO, FeO, CuO, CoO and
NiO can thus be seen as test samples for band theory models. In the most simple electronic
band model, the oxygen p-states are fully occupied, while the metal s-states are empty and the
d-states are partially occupied. With such a picture, one would expect a partially filled band at
the Fermi level, i. e. the oxides would be expected to be conductors. In reality, they show in-
sulating behaviour, which cannot be reproduced by simple density functional theory. Moreo-
ver it is common believe that the strong crystal field on the 3d orbitals leads to a suppression
of the orbital momentum, the so-called quenching. For these reasons, it is of interest to per-
form a separation of spin- and orbital moment magnetisation density in NiO. We have already
mentioned that this is possible, when discussing the amplitude (6) of non-resonant magnetic
scattering.

NiO has the NaCl fcc-structure with a = 4.177 Å at room-temperature. The ground state con-
figuration of the Ni2+ ion has the 3d8 configuration. Below TN = 523 K, NiO orders in the type
II antiferromagnetic structure, where ferromagnetic planes are stacked antiferromagnetically
along the [111] axis with their magnetic moments, aligned in the [111] planes, along one of
the [112] directions. The magnetic structure of MnO is depicted in figure 20.

Ni

Fig. 20: Magnetic unit cell of NiO.

Scattering Techniques B5.25

The separation of spin and orbital contributions can be done by means of polarisation analy-
sis. Here, we assume that the radiation incident on the sample is fully polarised in the hori-
zontal plane. In the actual experiment, performed at the magnetic scattering beamline ID20 at
ESRF, the degree of polarisation in the horizontal plane was found to be 99.5 %. As men-
tioned in the section on experimental considerations, pyrolytic graphite is an ideal material to
be used as a polarisation analyser. In the actual experiment, PG006 was used and the incident
beam energy was tuned to 7.84 keV to fulfill the 90° reflection condition at the analyser crys-
tal. A crystal was prepared with a [111] face. Specular magnetic reflections (1/2 1/2 1/2), (3/2
3/2 3/2) and (5/2 5/2 5/2) could be measured as a function of the angle of rotation ψ around
the scattering vector. The scattered magnetic intensity depends on this angle ψ since during
rotation we change the components of S and L, which enter the scattering cross section. Be-
fore we proceed with the discussion of the results, we have to give some explanation about
the magnetic domain structure that can develop in such a crystal as it is cooled from the cubic
paramagnetic phase into the orthorhombic antiferromagnetically ordered low temperature
phase. Since in the cubic phase 4 symmetric equivalent [111] directions exist ([111], [ 1 11],
[1 1 1], [11 1 ]) the magnetic propagation vector can align along one of these directions. There-
fore during cooling down, the crystal will be spontaneously develop 4 so-called K-domains
corresponding to the symmetry equivalent [111] directions, which can be realised in different
parts of the crystal. In addition the magnetic moments are aligned in the [111] planes along
one of the [11 2 ] directions. Since in the cubic state, the [111] axis is a three-fold symmetry
axis, there exist three equivalent possible spin directions perpendicular to the [111] axis, the
so-called S-domains. Depending on crystal faults (surfaces, small angle grain boundaries, im-
purity atoms etc.), but also on random processes, an arrangement of these K- and S-domains
develops in the sample in the low temperature phase. During the experiment it turned out that
the sample just developed a single K-domain with propagation vector parallel to the surface
normal, at least in the near surface region of roughly 40 micrometer probed in the experiment.
However, when measuring the scattered intensities from a given K-domain, all contributions
from the associated S-domains add incoherently. The S-domain distribution within a K-do-
main can be studied by rotating the sample about the surface normal. Such an experiment is
shown in figure 21.

Fig. 21: Normalised integrated inten-

sities of the 3/2 3/2 3/2
reflection as a function of
Renninger angle Ψ for the
two polarisation states σ→π
and σ→σ and the total scat-
tering, respectively (from
20).

Ψ (deg.)
B5.26 Th. Brückel

From such experiments, the L/2S ratio can be determined. It is plotted in figure 22 as a func-
tion of sinθ/λ.

Fig. 22: Measured variation of L(Q) / 2 S(Q) as a function of sinθ/λ = Q/4π. The continuous
line is the dependence estimated from theory, adjusted to fit through the data with a
contraction of the wave function by 17 % (from 20).

By normalising to the charge reflections, the spin and orbital formfactors can be determined.
They are depicted in figure 23.

Fig. 23: Spin and orbital moment form factors of NiO. Again the continous lines are the
calculated variations of S(Q) and L(Q) with an expansion of the Q-scale by 17 %
(from 20).

Let us summarise the most important results of this investigation: First of all, there is a rather
large contribution of orbital moment in NiO. In fact, the extrapolation to Q = 0 gives an ef-
fective ratio L/S = 0.34. This large contribution clearly contradicts the common believe that in
such transition metal compounds, the orbital angular moment is largely quenched due to the
crystal field. Furthermore, the results also indicate a contraction of the atomic wave function
Scattering Techniques B5.27

for Ni in NiO, if the experimental values are compared with the predictions of theory. The
atomic wave functions are contracted by 17 % as compared to the free ion. We can conclude
that the magnetisation density in a simple system like NiO is not yet fully understood and the
results of these studies have to be taken into account, when models for the electronic and
magnetic structure and properties of transition metal compounds are being made.

So far, we have discussed the magnetic scattering of x-rays in the "conventional energy
range" between 4 and 20 keV. We now want to discuss an alternative method, namely the
non-resonant magnetic scattering of very hard x-rays with energies above 80 keV [22-25].
From (9) it follows that with high energy x-ray diffraction one can determine the spin density
distribution independent of the polarisation of the incident beam and without analysis of the
final polarisation after scattering. While in neutron diffraction only the total magnetic mo-
ment, proportional to the sum L+2S is accessible, x-ray diffraction in the conventional energy
range requires polarisation analysis to separate the spin momentum S(Q) from the orbital an-
gular momentum density L(Q). The additional principal feature of high energy magnetic x-ray
diffraction is the drastic increase in penetration depth. For 3d transition metals, the absorption
length 1 µ increases from some µm at 8 keV to several mm at 80 keV. This leads to a volume
enhancement of the signal which is, however, partly compensated by the λ2 term for the re-
flectivity. Moreover, true bulk properties become accessible, a feature especially important
for studies of magnetic disorder phenomena. Magnetic x-ray scattering can be studied in
transmission geometry. Corrections for absorption, extinction, beam foot print etc. are simple
and therefore, by normalising the intensity of the magnetic reflections to the intensity of the
charge reflections, absolute values for the spin moment can be determined [25]. Neutron and
photon experiments of bulk properties from the same crystal become possible, where one ad-
vantage of the x-ray study is the high intrinsic resolution of about 10-3 Å-1 longitudinal and
2·10-4Å-1 transversal.

Fig. 24: Schematic sketch of the three crystal diffractometer for high energy x-ray scattering
BW5 at HASYLAB/Hamburg (http://www-hasylab.desy.de).

Figure 24 depicts a typical experimental set-up. The diffractometer of the beamline BW5 at
HASYLAB receives a white x-ray beam from a 2T high field wiggler. Inclined water cooled
Cu plates limit the beam dimensions to about 4 x 4 mm2. To reduce the heat load on the opti-
cal elements, a water cooled 1 mm thick Cu window absorbs all radiation of energy smaller
than 60 keV. For the experiments on MnF2 described below, we employed annealed Si 311
crystals with a mosaic width of 10´´ in Laue (transmission) geometry as monochromator and
analyser crystals. The analyser can be used to increase the momentum space resolution and to
reduce the background, but should not be employed for the collection of integral intensities.
Iron collimators with a quadratic cross section of 50 x 50 mm2 and a free bore of 10 x 10 mm2
are positioned between monochromator and sample, sample and analyser and analyser and
detector to reduce the background. With an energy sensitive Ge solid state detector, the ine-
B5.28 Th. Brückel

lastic background from Compton scattering and fluorescence can be efficiently suppressed.
The sample is mounted inside a cryostat with Al windows and large tails to avoid Al back-
ground scattering to enter the detector.

The potential of magnetic high energy x-ray diffraction can be demonstrated with experiments
on the antiferromagnetic model system MnF2 [24]. Its structure is shown in figure 25. With an
energy of 80 keV, a high peak count rate of 12.000 counts/sec, a good peak-to-background
ratio of 230:1 and an excellent Q space resolution can be obtained for the magnetic 300 Bragg
reflection. Figure 26 shows a measurement of the temperature dependence of the sublattice
magnetisation. In the critical region close to the Néel temperature TN, the reduced sublattice
magnetisation m = M (T ) M (T = 0) follows very accurately a power law behaviour
m(τ ) = D ⋅ τ β (23)
as a function of the reduced temperature τ = (TN − T) TN . The value of the critical exponent
of β=0.333(3) corresponds well to the predictions of the Ising model.

Fig. 25: The crystalline and magnetic unit cell of MnF2.

1 500
Intensity [a.u.]

400
300
200
100
0
0 20 40 60 80
T[K]
m

β = 0.333(3)
T N= 67.713(2) K
0.1

0.0001 0.001
τ 0.01 0.1

Fig. 26: Critical behaviour of the sublattice magnetisation of MnF2 in a double logarithmic
plot in reduced variables. The insert shows the temperature dependence of the inten-
sity of the magnetic 3 0 0 Bragg reflection of MnF2 measured with 80 keV photons.
Scattering Techniques B5.29

MnF2 is a classical model antiferromagnet with localised spin moments and therefore is an
ideal test material for any new technique in magnetism. The 3d metal Cr on the other hand
shows very intriguing magnetic properties. Chromium is an itinerant antiferromagnet ex-
hibiting an incommensurate spin density wave (SDW) below TN = 311 K [26]. Above the spin
flip transition at TSF = 123 K the spin density wave is transversally polarised, whereas below
TSF the polarisation becomes parallel to the modulation wave vector, which leads to a longitu-
dinally polarised SDW. The SDW gives rise to magnetic satellite peaks at positions corre-
sponding to the magnetic propagation vector qm. These satellites could be readily measured,
despite the small root-mean-squared moment of 0.43 µB, see figure 27. The temperature de-
pendence of the intensity and propagation vector for the 1- δ 0 0 satellite is plotted in figure
28. One can clearly see the spin flip transition at 123 K in a drastic drop of the satellite inten-
sity as the spin moment rotates from an orientation perpendicular to the scattering plane to an
orientation within the scattering plane. Therefore at the spin flip transition, the spin compo-
nent S2 - and according to (9) the Bragg intensity - vanishes.

70
T=124.5 K

T=121.6 K
60
T=112.0 K
Intensity [counts/s]

50

40

30

20

10
−120 −80 −40 0 40 80 120

ωs [arcsec]

Fig. 27: Diffraction geometry (left) and 1-δ00 magnetic satellite (right) of Cr. The measure-
ments were done with photons of energy 100 keV. While above the spin flip transition
in the transverse SDW, moments perpendicular to the diffraction plane exist, these
moments come to lie within the diffraction plane as the SDW becomes longitudinal
polarised below TSF.

The main aim of our study of chromium was, however, to determine the relative contribution
of spin and orbital angular momentum to the SDW. As discussed above, the spin momentum
is directly accessible with high energy x-ray diffraction, while neutrons measure the combi-
nation L+2S. Therefore by combining results of measurements of both techniques, spin S and
orbital L momentum densities can be determined separately. To this end, we measured 5
satellites of Cr as a function of momentum transfer. In figure 29 we compare these high en-
ergy x-ray data with neutron results of Moon et al. [27]. In addition, results of calculations
using a fully relativistic spin density functional theory for the form factors of spin and orbital
angular momentum are shown. Neutron and x-ray data coincide within the error bars indi-
cating a magnetic moment due to spin only. This interpretation is supported by the band
structure calculation. A more detailed analysis gives an orbital contribution of -4(8)%. It is
interesting to note that the orbital moment for antiferromagnetic chromium is about an order
B5.30 Th. Brückel

of magnitude smaller than that calculated for the pure ferromagnetic transition metals Fe, Co
and Ni.

0.963
Neutrons
4 1−δ 0 0 BW5
0.961
1−δ 0 0 PETRA II

Integrated Intensity [arb. units]

0.959
qm [r.l.u.]

0.957
2 0.4 1 0 −δ
0.955
0.2

0.953
0.0
0 40 80 120 160 200
0.951
100 150 200 250 300 100 150 200 250 300

T [K] Temperature [K]

Fig. 28: Temperature dependence of the magnetic propagation vector (left) and of the inten-
sity (right) of the 1- δ 0 0 magnetic satellite of Cr measured with photons of energy
100 keV.

Fig. 29: Magnetic form factor of Cr in the ordered state as a function of momentum transfer
(respectively 1/2d, where d denotes the lattice spacing). Full triangles are neutron
results on a commensurate Cr alloy [27]. Open circles correspond to averaged
values from magnetic satellites around the commensurate positions determined with
100 keV high energy x-rays.
Scattering Techniques B5.31

6 Summary; comparison to magnetic neutron scattering

The above examples clearly demonstrate that x-ray diffraction from magnetic materials has
become a microscopic probe of magnetism, complementary to the traditional probe of neutron
scattering. First, it allows the investigation of strongly neutron-absorbing materials, e. g. GdS
and EuS. Second it stands out by the high momentum-space resolution of typically 10-4Å-1,
about one to two orders of magnitude better than a standard neutron experiment. This is of
advantage for the investigation of incommensurate structures (e.g. Ho or Cr), magnetic disor-
der phenomena which involve a broadening of the magnetic Bragg reflection or critical scat-
tering. In critical scattering investigations of the sublattice magnetisation, the magnetic Bragg
peak can be well separated from critical diffuse scattering (see our example of MnF2). For
measurements of the critical diffuse scattering, the high Q-space resolution permits a detailed
line shape analysis and a closer approach of the phase transition. Moreover, due to the high
incident energy the full integration over the energy spectrum of magnetic fluctuations is
guaranteed. Third, x-ray diffraction is a perfect tool for the investigation of surface-near phe-
nomena and ultimately of surface magnetism [33, 18].

Considering non-resonant and resonant scattering, some specific applications are opened for
each technique. In the case of resonance-exchange scattering, most evidently the resonance-
enhancements at the absorption edges allow the investigation of samples with very small
magnetic moments. A „magnetic“ spectroscopy of the unoccupied levels above the Fermi
edge is possible (compare the discussion of the scattering from EuS). Finally the resonance
effects render magnetic scattering element specific. In alloys and mixed crystals it is now pos-
sible to investigate the magnetic order of the various magnetic elements separately. In addi-
tion to the sensitivity to the magnetic species, a sensitivity to the electronic states can be
achieved (compare the Er/Tb multilayer). The big draw back of resonant scattering is, how-
ever, that as for any second order perturbation process the scattered intensity is not easily in-
terpretable in a simple and fundamental quantity such as the magnetic moment. This is one
reason, why non-resonant scattering remains essential. It permits the separate determination
of the spin- and angular moment contributions to the form factor. The rich cross section of
non-resonant scattering gives complementary information to neutron diffraction in the case of
complicated magnetic structures. As a special case of non-resonant scattering we introduced
the diffraction of high-energy x-rays with energies around or above 100 keV. The large pene-
tration power makes this radiation a true volume probe just like neutron scattering. A simple
sample environment without specific x-ray transparent windows can be used and x-ray and
neutron diffraction experiments can be performed from the same bulk crystal. A volume en-
hancement of the signal is obtained, independent of composition. The short wavelength and
the small cross section eliminate extinction effects, which allows precision measurements of
structure factors within the first Born approximation. Finally, the simple form of the cross
section (9) makes possible a mapping of the spin momentum distribution alone without po-
larisation analysis and by combination with neutron diffraction, a determination of L(Q) and
S(Q) can be achieved.

The experimental techniques for neutron and magnetic x-ray scattering are similar in many
ways. For both techniques, polarisation analysis allows the separation of magnetic from non-
magnetic scattering. “Flipping ratio” measurements are employed for both probes to measure
the interference effects between magnetic and non-magnetic contributions.

There are, of course, clear differences between the three probes:

B5.32 Th. Brückel

• For neutron scattering, nuclear and magnetic scattering are comparable in magnitude,
while for x-ray scattering the magnetic contribution is generally significantly smaller.
Therefore, magnetic structure determination from powder samples will remain a typical
task for neutron scattering.
• The form factors are quite different for the three techniques: In XRES, the spatial exten-
sion of the core levels is relevant and therefore virtually no decrease of the scattering am-
plitude as a function of momentum transfer is observed. In non-resonant x-ray scattering,
the form factors of spin and angular momentum can be determined separately, while neu-
trons are sensitive to a combination of both.
• Angular- and polarisation dependencies are richer for magnetic x-ray scattering. To give
an example: magnetic neutron scattering is sensitive only to the magnetic moment perpen-
dicular to the scattering vector Q, while x-rays see various components of L and S, com-
pare (6).
• Due to small cross section for magnetic x-ray scattering, the study of magnetic excitations
and fluctuations remains a domain of inelastic neutron scattering

To summarise: both neutron and x-ray scattering techniques are important for the investiga-
tion of magnetic structures. They are largely complementary and it is wise for every con-
densed matter scientist to know the strengths and weaknesses of both methods and to chose
carefully which probe to use to solve a specific problem in magnetism.
Scattering Techniques B5.33

References
[1] P. M. Platzman and N. Tzoar, Phys. Rev. B9, 3556 (1970)
[2] F. de Bergevin and M. Brunel, Phys. Lett. A39 , 141 (1972)
[3] F. de Bergevin and M. Brunel, Acta Cryst. A37 , 314 (1981)
[4] D. Gibbs, D. E. Moncton, K. L. D’Amico, J. Bohr and B.H. Grier, Phys. Rev. Lett. 55,
234 (1985)
[5] D. Gibbs, D. R. Harshman, E. D. Isaacs, D. B. McWhan, D. Mills and C. Vettier, Phys.
Rev. Lett. 61, 1241 (1988)
[6] P.M. Platzman and N. Tzoar, Phys. Rev. B9, 3556 (1970)
[7] M. Blume, J. Appl. Phys. 57, 3615 (1985)
[8] M. Blume and D. Gibbs, Phys. Rev. B37, 1779 (1988)
[9] H. Grotch, E. Kazes, G. B. Matt and D. A. Owen, Phys. Rev. A27, 243 (1983)
[10] S. M. Durbin, Phys. Rev. B57 , 7595 (1998)
[11] M. Blume in G. Materlik, C. J. Sparks and K. Fischer, Resonant Anomalous X-Ray
Scattering, 495 (North-Holland, Amsterdam, 1994)
[12] J. P. Hannon, G. T. Trammell, M. Blume and D. Gibbs, Phys. Rev. Lett. 61(1988),
1245 and Erratum: Phys. Rev. Lett. 62, 2644 (1989)
[13] A. Stunault, F. de Bergevin, D. Wermeille, C. Vettier, Th. Brückel, N. Bernhoeft, G. J.
McIntyre and J. Y. Henry, Phys. Rev. B60 , 10170 (1999)
[14] H. A. Dürr, E. Dudzik, S. S. Dhesi, J. B. Goedkoop, G. van de Laan, M. Belakhovsky,
C. Mocuta, A. Marty and Y. Samson, Science 284 , 2166 (1999)
[15] Th. Brückel, D. Hupfeld, J. Strempfer, W. Caliebe, K. Mattenberger, A. Stunault, N.
Bernhoeft and G. J. McIntyre, Eur. Phys. J B19, 475 (2001)
[16] C. Schussler-Langeheine, E. Weschke, A. Y. Grigoriev, H. Oh, R. Meier, D. V. Vya-
likh, C. Mazumdar, C. Sutter, D. Abernathy, G. Grübel and G. Kaindl, J. Electron.
Spectrosc. 114, 953 (2001)
[17] E. D. Isaacs, D. B. McWhan, C. Peters, G. E. Ice, D. P. Siddons, J. B. Hastings, C.
Vettier and O. Vogt, Phys. Rev. Lett. 62, 1671 (1989)
[18] S. Ferrer, P. Fajardo, F. de Bergevin, J. Alvarez, X. Torrelles, H. A. von der Vegt and
V. H. Etgens, Phys. Rev. Lett. 77, 747 (1996)
[19] J. P. Hill and D. F. McMorrow, Acta Cryst. A52, 236 (1996)
[20] V. Fernandez, C. Vettier, F. de Bergevin, C. Giles and W. Neubeck, Phys. Rev. B57,
7870 (1998)
[21] D. Mannix, P.C. de Camargo, C. Giles, A.J.A. de Oliveira, F. Yokaichiya and C. Vet-
tier, Eur. Phys. J. B20, 19 (2001)
[22] Th. Brückel, M. Lippert, R. Bouchard, T. Schmidt, J.R. Schneider and W. Jauch, Acta
Cryst. A49, 679 (1993)
[23] M. Lippert, Th. Brückel, Th. Köhler and J.R. Schneider, Europhys. Lett. 27, 537 (1994)
[24] J. Strempfer, Th. Brückel, U. Rütt, J.R. Schneider, K.D. Liss, and T. Tschentscher,
Acta Cryst. A52, 438 (1996)
[25] J. Strempfer, Th. Brückel, D. Hupfeld, J. R. Schneider, K.-D. Liss, and T. Tschent-
scher, Europhys. Lett. 40, 569 (1997)
[26] J. Strempfer, Th. Brückel, W. Caliebe, A. Vernes, H. Ebert, W. Prandl, and J.R.
Schneider, Eur. Phys. J. B14, 63 (2000)
[27] R. M. Moon, W. C. Koehler and A. L. Trego, J. Appl. Phys. 37, 1036 (1966)
[28] S. C. Perry, M. M. R. Costa, W. G. Stirling, M. J. Longfield, D. Mannix and Th.
Brückel. J. Phys.: Condens. Matter 10, 1951 (1998)
B5.34 Th. Brückel

[29] D. Hupfeld, W. Schweika, J. Strempfer, K. Mattenberger, G.J. McIntyre and Th.

Brückel, Europhys. Lett. 49, 92 (2000)
[30] D. Hupfeld, W. Schweika, J. Strempfer, W.A. Caliebe, U. Köbler, K. Mattenberger, G.
J. McIntyre, F. Yakhou and Th. Brückel, Eur. Phys. J. B26, 273 (2002)
[31] J. Voigt, E. Kentzinger, U. Rücker, D. Wermeille, D. Hupfeld, W. Schweika, W.
Schmidt and Th. Brückel, Europhys. Lett. 65, 560 (2004)
[32] D. Hupfeld, O.H. Seeck, J. Voigt, J. Bos, K. Fischer and Th. Brückel, Europhys. Lett..
59, 284 (2002)
[33] G.M. Watson, D. Gibbs, G.H. Lander, B.D. Gaulin, L.E. Berman, H. Matzke and W.
Ellis, Phys. Rev. Lett. 77, 751 (1996)
[34] J. Bohr, D. Gibbs, D. E. Moncton, K. L. D'Amico, Physica 140a, 349 (1986)
[35] D. Gibbs, G. Grübel, D. R. Harshman, E. D. Isaacs, D. B. McWhan, D. Mills, C. Vet-
tier, Phys. Rev. B 43, 5663 (1991)
[36] W. Neubeck, C. Vettier, F. de Bergevin, F. Yakhou, D. Mannix, L. Ranno, T. Chatterji,
J. Phys. and Chem. of Solids 62, 2173 2001)
[37] Th. Brückel, M. Lippert, T. Köhler, J. R. Schneider, W. Prandl, V. Rilling, M. Schil-
ling, Acta Cryst. A52, 427 (1996)
[38] S. P. Collins, D. Laundy, G. Y. Guo, J. Phys.: Condens. Matter 5, L637 (1993)
Index
asymmetry ratio 20
critical scattering 28
cross section 4, 7
element specifity 14
frustration 14
high energy x-ray diffraction 27
interlayer coupling 18
metal-insulator transition 14
non-resonant magnetic x-ray diffraction 20
Oscillator strength 10
polarisation analyser 13
polarisation dependence 5
resonance enhancement 9
resonance exchange scattering 14
spin density wave 29
spin glass 15, 17
spin slip structure 23
spin-orbit separation 21, 26, 29
synchrotron radiation 11
synchrotron x-ray source 10
thin film superlattice 18
transition metal oxides 24
undulator 11
Wiggler 11