The nucleus of an atom is surrounded by electrons rotating on their own axis (spin) and orbiting around the nucleus as the earth orbits around the sun and spins on its axis. The magnetic moment of a material is the sum of the electron orbital and spin contributions.

For a long time magnetic materials were studied with neutrons, due to the strong coupling of the neutron spin to the magnetic moments in the sample.

In the Hamiltonian describing the interaction of X-ray photons with the electrons of a material, the main term corresponds to the interaction with the charge (charge X-ray scattering or Thomson scattering). However, there are also relativistic corrections corresponding to spin dependent terms but they are 3 or 4 orders of magnitude weaker. Discovered by De Bergevin and Brunel in 1972 this magnetic scattering was for long a curiosity until, in 1987, it was shown at the NSLS (Brookhaven Nat. Lab.) that this magnetic term could increase by a few orders of magnitude when the photon energy was tuned near an absorption edge. These resonance enhancements originate from electric multiple transitions between spin polarised electronic states, which are also at the origin of the X-ray magnetic dichroism.

At the ESRF, the intensity of the photon beam is such that magnetic scattering can be observed easily outside of resonances. Even surface magnetic scattering has been observed. 

The intensity of the magnetic scattering signal can be increased by using photons of 60 or 80 keV instead of 10 keV. Also, the K-space resolution for X-rays is much better than for neutrons.

Until 1987, there was essentially one synchrotron radiation technique to probe magnetism: spin polarised photo-emission. Today at the ESRF we have eight kinds of experiments dealing with magnetic studies:

  • magnetic scattering (resonant or non resonant)
  • surface magnetic scattering
  • dichroism (with or circular polarisation)
  • resonant inelastic scattering with circular polarisation
  • Compton scattering with circular polarisation
  • Mössbauer spectroscopy
  • photoemission microscopy with circular polarisation
  • spin polarised photoemission.




Photo: On the Magnetic Scattering beamline (ID20)





Among the X-ray methods, resonant magnetic scattering is a very powerful probe for microscopic magnetic properties because it offers element and site selectivity, polarisation dependance, high intensity and high Q resolution. This is illustrated by giving four examples on bulk, thin films and surface magnetic scattering.


Orbital and spin magnetisation densities in holmium



Orbital (L) and spin (S) magnetic moments are fundamental quantities for understanding the macroscopic magnetic properties of matter. Their determination remains a challenging task which can be adressed by both X-ray magnetic scattering (XMS) and X-ray magnetic circular dichroism (XMCD). While XMCD is applicable only to ferro- (ferri-) magnets with a net magnetic moment, XMS can also be used to investigate antiferromagnets. Controlling precisely the scattering geometry as well as the polarisation states of the incident and diffracted beams, XMS can be used to extract orbital and spin magnetisation densities out of the scattered intensities. For a quantitative determination of L and S a high degree of polarisation (linear or circular) of the incident beam as well as the possibility to determine the linear ( and ) polarised components of the diffracted beam is indispensable. The Troika team, in collaboration with the Magnetic Scattering Group, have perfomed an investigation of the spiral antiferromagnet holmium with incident linear () polarisation to determine the orbital to spin (L/S) magnetisation density ratio. The incident beam delivered by the undulator of ID10 (Troika) was highly polarised (97 %) and the degree of linear polarisation P' of the diffracted beam was determined with a graphite polarisation analyser. Figure 19 shows the degree of linear polarisation as a function of L (reciprocal c direction) as determined at the magnetic satellite positions of holmium. P' = -1 is equal to a completely polarised diffracted beam, while P' = +1 stands for a polarised beam. The lines represent calculated values for L/S ratios of 2, 3 and 4 and incident polarisation. The experimentally determined points are well described by the theoretical curve for an orbital-to-spin magnetisation ratio of L/S = 3. The result is in agreement with the expectation for a 5I8 Hund's rule ground state of holmium.




C. Sutter (a), G. Grübel (a), D. Gibbs (b), C. Vettier (a), F. de Bergevin (a), A. Stunault (a), to be published

(a) ESRF

(b) Brookhaven National Laboratory, Upton NY (USA)




Study of magnetism in DyFe4Al8



The site selectivity of resonance X-ray scattering represents a new tool in the study of magnetism in materials that have more than one magnetic atom. We illustrate this at ID20 with a study of the interactions between the Fe and Dy sublattices in the material DyFe4Al8. The material is an antiferromagnet and neutron experiments have shown that the Fe 3d moments order at ~ 170 K, and the Dy 4f moments order at a much lower temperature of 20 K. One major question is how the exchange develops between these two sublattices. Theory suggests that the Dy 5d electrons play a role, and synchrotron X-rays can give information on this important aspect.

The experiments use an X-ray energy corresponding to the Dy LIII edge (7.79 keV) (Figure 20), in which case the (dipole) transition 2p - 5d is allowed, and leads to an enhancement of the magnetic signal. Thus, the experiment is sensitive to Dy only, and it also selects the 5d electrons. In principle quadrupole transitions from 2p to 4f states are allowed but they may be distinguished by their energy and polarisation dependence. We have been unable to detect such transitions, so that all the scattering represents dipole transitions involving the 5d transitions. As mentioned above, the resonant magnetic scattering rotates the polarisation and all the scattering discussed here is rotated.

In Figure 21 is shown the T dependence of the magnetic intensity at the positions of the magnetic satellites. The large value at low temperature corresponds to the polarisation of the 5d band electrons by the ordered Dy 4f electrons. The rapid decrease on warming to 20 K corresponds to the effect of disordering the Dy 4f moments, which is seen very clearly in neutron experiments. However, unlike the Dy 4f moments, the long-time averaged polarisation of the 5d band does not go to zero when the Dy 4f moments disorder. Instead, it continues at a constant value until the iron sublattice disorders at 175 K. Thus the 5d and 3d bands are intimately connected - as has been predicted by theory. By estimating (from neutron and magnetisation measurements) the 5d polarisation at ~ 0.6 µB at low temperature, it gives an approximate value of 0.1 µB for the 5d moment when the Fe moment only is ordered.

X-rays can also give valuable, and perhaps unique, information about the arrangements of moments in the unit cell. In Figure 22 are shown intensities for the first harmonics in two different orientations of the crystal as a function of the scattering angle . The upper line gives a large intensity which is independent of . This arises only when the moments are rotating in the scattering plane. However, for the second orientation, the moments are rotating in a plane perpendicular to the scattering plane, and the cross-section then goes to zero at = 0, and follows a sin 2 dependence, which is shown by the lower line.

These experiments give new information on the role of the 5d conduction electrons, which are essentially invisible to neutron scattering, in mediating the interaction between the Dy 4f and Fe 3d electrons. In addition, using polarisation analysis, we have shown that reliable intensity measurements can give unique information about the arrangements of the magnetic moments in the unit cell.





S. Langridge (a), J. A. Paixão (b), S. Aa. Sørensen (c), G. H. Lander (d) and C. Vettier (a), to be published

(a) ESRF

(b) Univ. of Coimbra (Portugal)

(c) European Commission, Karlsruhe (Germany)

(d) Risø National Laboratory (Denmark)




Resonance magnetic scattering as a probe of thin magnetic films



The magnetic properties of magnetic thin films have attracted much interest because of potential technological applications. The influence of film thickness and the role of non-magnetic layers are among the current fields of research. Again resonant magnetic X-ray scattering offers several advantages: the site selectivity allows the separation of magnetic elements and the inherent surface sensitivity of X-rays makes this method well suited to the study of thin films and surfaces. We make this point clearer with the experiments performed at ID20 on the magnetic behaviour of dysprosium in thin films and in multilayers Dy/Er grown along the [0001] direction of their hcp structure by molecular beam epitaxy. The substrate is a sapphire plate with niobium and yttrium buffers. Bulk dysprosium exhibits helicoidal magnetic order below Tn ~ 176 K. It becomes ferromagnetic at ~ 83 K. Erbium orders with a c-axis modulated structure at Tn ~ 84 K. The spiral magnetic structure of Dy is evidenced in a non-resonant scattering experiment by the occurrence of first order magnetic harmonics. In the dipole approximation, the resonant magnetic scattering amplitude gives rise to first and second order diffraction harmonics near the resonance. First order satellites were easily observed (Figure 23). Second order satellites are ~ 40 times weaker. Quadrupole transitions at the resonance would lead to third and fourth order diffraction harmonics.

Figure 24 shows the energy dependence of first and second order harmonics in the ordered magnetic phase of a 2500 Å Dy film. The first order harmonics have a resonant energy at the Dy LIII edge, while the second harmonics show two resonant energies, indicating possible quadrupole transitions. No higher order harmonics could be found. The observed magnetic intensities are such that films with thickness down to below 100 Å can be studied.

The interest of the Dy/Er superlattice arises from the presence of paramagnetic layers of Er when the Dy layers are magnetically ordered. The site selectivity of resonant scattering makes it possible to distinguish magnetic moments on Dy and Er. At the Dy LIII edge, we detect magnetisation on the Dy sites. The first harmonic of the pure Dy are replaced by several peaks due to the modulation of the multilayer periodicity. The presence of these resolution limited peaks is a fingerprint for the coherence of the helicoidal magnetic orders that propagate through the paramagnetic Er layers. The temperature dependence of the line width of the magnetic peaks show that long range order develops through the multilayer at Tn. A further study will involve the observation of induced moment on the Er sites.These experiments demonstrate that resonant magnetic X-ray scattering techniques are an excellent probe for the microscopic magnetic properties of very thin films and superlattices.





C. Vettier (a), A. Stunault (a), K. Dumesnil (b), Ph. Mangin (b), D. Wermeille (a), S. Langridge (a) and N. Bernhoeft (a)

(a) ESRF

(b) Univ. of Nancy (France)




Resonant surface magnetic X-ray diffraction from Co3Pt (111)



The possibility of detecting magnetic X-ray scattering from surfaces presents the difficulty of measuring intensities around 11 or 12 orders of magnitude weaker than the normal bulk Bragg intensities. This is very demanding even for third generation synchrotron sources such as the ESRF. However, the discovery in 1988 of the resonance exchange scattering which results in a large (two orders of magnitude or more) increase of the magnetic signal when the photon energy is tuned to an absorption edge, opens the possibility of performing surface magnetic X-ray diffraction experiments. This has been confirmed in previous experimental work by detecting magnetic signals in X-ray beams scattered from surface regions only a few nm thick. We have been able to make a step forward since we have measured the magnetism in the topmost atomic plane of the surface of a ferromagnetic alloy.

A characteristic feature of surface X-ray scattering is that the diffracted intensities along straight lines in reciprocal space, normal to the surface of the crystal and passing through reciprocal lattice points where Bragg conditions are fulfilled, are not zero along the segments joining neighbouring Bragg reflections. In these segments, the relatively weak intensities are sensitive to the structure of the surface. These 'one-dimensional' distributions of intensity are designed as diffraction rods. We measured the diffracted intensities along several rods in order to determine the structure and stoichiometry of the surface of our alloy that was found to be Pt enriched. Next, we made similar measurements isolating the part of the diffracted intensity sensitive to the magnetisation of the crystal. The structural results were used as input for the analysis of the magnetic measurements that revealed a depressed magnetic moment of the Pt atoms in the topmost atomic plane.

In order to isolate the magnetic part of the diffracted intensity, a magnet, originating a field along the vertical direction, was situated in the vicinity of the surface . The field could be directed upwards () or downwards () by rotating the magnet. It can be shown that, if the photon energy is tuned to the L absoption edge of the Pt atoms, under a determined scattering geometry, the quantity R = (I- I)/(I+ I), which measures the relative difference in diffracted intensities for both field directions, is non-zero and it depends in a simple manner on the magnetic moment of the Pt atom.

In order to obtain information on the magnetic structure of the surface we measured R along the diffraction rods. The experiment was carried out on the Surface Diffraction beamline (ID3) and the results are shown in Figure 25.





S. Ferrer (a), P. Fajardo (a), F. de Bergevin (a), J. Alvarez (a), X. Torrelles (a), H. A. van der Vegt (b) and V. H. Etgens (c), Phys. Rev. Lett. 77, p. 747 (1996)

(a) ESRF

(b) FOM Institute for Atomic and Molecular Physics, Amsterdam (Netherlands)

(c) Laboratoire de Minéralogie-Cristallographie, CNRS, Univ. of Paris VI (France)








X-ray Magnetic Circular Dichroism (XMCD) is a technique for the study of ferromagnetic materials (but also paramagnetic systems polarised by an intense external magnetic field). It consists on the measurement, as a function of photon energy, of the difference in the absorption coefficient for X-rays of opposite circular polarisation, which in many cases of interest can be relatively large and easily observable in the vicinity of an absorption edge (see Figure 26 for Fe L2,3 edge).

XMCD has rapidly gained considerable popularity because of its attractive features, including atomic selectivity, through the choice of the absorption edge, and high sensitivity, as the technique allows the detection of the magnetisation in just a few atomic layers of magnetic material. A further attractive feature of this novel technique is the possibility to extract precise quantitative information on the atomic magnetic moments, thanks to theoretical results known as «sum rules».

Sum rules: These were introduced a few years ago by the ESRF Theory Group. They allow the orbital magnetic moment <Lz> and the spin magnetic moment <Sz> to be obtained from the dichroism measurements. As for the bulk, they give very good results for the transition metals (Fe, Co, Ni...). It has been shown that, in an atomic model:

1. <Lz> = B [AL3 + AL2]

2. <Sz> + C <Tz> = D [AL3 - 2AL2]

where AL3 and AL2 represent the dichroic signal at the L2 and L3 edges; B, C, D are simple constants, and <Tz> is an additional term which vanishes for bulk transition metals.


The third «sum rule»



The first sum rule, often referred to as the «orbital» sum rule, relates very directly a (properly normalised) integral of the dichroic signal, e.g. at the L2,3 edge of a 3d transition metal system, with the orbital part of the magnetic moment per atom, denoted by <Lz>. The second one, on the other hand, relates another integral of the signal to a linear combination of two objects, <Sz> and <Tz>. The former, <Sz>, is the expectation value of the spin component along the magnetisation and is therefore the spin contribution to the magnetic moment. The latter object, <Tz>, provides somewhat more complicated, combined information on both orbital and spin components.

Therefore, while the «orbital» sum rule provides a direct determination of the orbital magnetic moment, it is in general not possible to disentangle <Sz> and <Tz> in the second sum rule (often referred to as the «spin» sum rule), and to achieve an analogous determination of the spin moment. Applications of the «spin» sum rule have been limited so far to cases in which it can be argued that <Tz> is small. This is the case for the bulk cubic or hcp transition metals, where the <Tz> term is less than 10 % of the spin term, and it was shown that, if it is simply ignored, the sum rules provide accurate values for the orbital to spin moment ratio in bulk Fe and Co.

However, bulk transition metals have been thoroughly investigated for many decades, and it is much more interesting to study more novel and technologically relevant systems, such as magnetic surfaces, thin films, multilayers, etc. In such systems the problem of separating <Sz> and <Tz> cannot be eluded. Recent theoretical advances allow this problem to be overcome, and allow one to proceed to a separate determination of <Lz>, <Sz> and <Tz> for 3d-metal based heterostructures.

Figure 27 shows schematically a magnetic thin film deposited on a substrate, with three orthogonal directions labelled by , and . Let us assume that three circular dichroism experiments are performed, aligning in each case the photon propagation direction to the external field directed along , and , and that, each time, the integrals appearing in the right-hand-side of the spin sum rule are evaluated. It was then shown by Stöhr and König that, by adding the three results, one can determine <Sz>, because the <Tz> contribution is averaged out in the three measurements (third sum rule).




J. Stöhr (a), H. König (b), Phys. Rev. Lett. 75, 3748 (1995)

(a) IBM, Almaden (USA)

(b) ESRF




Departure from the independent electron behaviour in itinerant magnets



Metallic iron is generally assumed to be a prototype of a system whose valence electrons are nearly independent. We report on a novel way to determine the spin polarisation of the core hole by using Auger decay spectroscopy with light circularly polarised perpendicular to the magnetisation direction. This method does not suffer from the inherently low count rates associated with spin detection. It demonstrates the existence of strong 3d electron correlation even in itinerant materials and the breakdown of the one-electron approach.

Circularly polarised radiation in the soft X-ray region is generated by a helical undulator. Either linear or circular polarisation is obtained by setting the phase between the horizontal and vertical magnetic field in this device. A Dragon monochromator at ID12B delivers a monochromatised beam with an optimum energy resolution of 180 meV and a degree of circular polarisation of 85 % of either helicity, which is switched by changing the phase of the two magnet structures in the insertion device. Fe ultra-thin films were prepared by evaporating 10 monolayers (ML) onto a Cu(110) surface under ultra-high vacuum conditions. The depositions and subsequent measurements were done at room temperature. A current pulse through a nearby coil was used to magnetise remanently the films in the surface plane. The magnetic order of the films was verified in situ by comparing the magnetic dichroism, i.e. the difference in X-ray absorption for left and right circularly polarised light, at the 2p absorption edges to their known magnitude. All photoemission measurements were performed with the light polarisation vector perpendicular to the sample magnetisation. The photoelectrons were collected at an angle of 60° with the surface normal in the plane spanned by the sample magnetisation and the light polarisation vectors using a hemispherical multichannel analyser with an acceptance angle of 40°. This large angle ensures an averaging over diffraction effects which can give variations on a scale of a few degrees.

In X-ray photoemission an electron is excited from a core level into a continuum state far above the Fermi level. Figure 28 gives the Fe photoemission spectra measured with left and right circularly polarised light. The large core spin-orbit coupling divides the spectrum into a 2p3/2 and a 2p1/2 structure. These structures have narrow asymmetric line shapes, typical for metallic systems. Also in Figure 28 the magnetic dichroism is shown, which is the difference between the two polarised signals.

The 2p3/2 edge exhibits a leading sharp negative peak followed by a very broad positive structure with a pronounced structure. In the 2p1/2 edge the leading peak is positive followed by a negative structure. The dichroism signal is proportional to the orbital polarisation of the magnetic sublevels of the core hole induced by interaction with the magnetically aligned valence band orbital and spin moments. In the 2p3/2 (2p1/2) level the spin and orbit are coupled (anti)parallel. As shown by the dichroism signal, the photoemission final state at both edges consists of high-spin states at the low binding energy side and low-spin states at the high binding energy side. For these high (low) spin states the spin direction of the core hole is parallel to the majority (minority) spin of the valence electrons. The shape of the magnetic dichroism suggests that the core hole state contains more majority than minority spin. This can be due to an alignment of the core hole spin parallel to that of the valence states. The spectrum calculated using an independent particle model is included in Figure 28. In this approach the core level is split by an effective spin field, resulting in a symmetric line shape that is distinctly different from the experimental result.

We use the Auger decay to quantify the spin polarisation of the core hole. In the L3M2,3M2,3 Auger process, 2p63dn --> 2p53dn--> 2p63p4', the intermediate 2p core hole state, created after photoelectron emission, decays by Coulomb interaction into a two-hole final state 3p4 under emission of an Auger electron ' into the continuum. Since the magnetic 3d is a spectator shell for an Auger final state with two core holes, the integrated dichroism signal, which measures the difference between the number of holes created with left and right circularly polarised light, should be zero. We have shown previously that, in the case of resonant photoemission, spin conservation in the initial

X-ray absorption process where a 2p electron is excited into an empty 3d state causes an alignment of the intermediate 2p core hole. This results in a non-zero dichroism signal in the geometry employed in our experiments. For the Auger process, however, this mechanism will not be present and dichroism should be zero.

The Auger spectra measured at a photon energy of 900 eV are shown in Figure 29. Similar to the 2p photoemission spectrum, the L3M2,3M2,3 Auger spectrum has high-spin states at the low binding energy side and low-spin states at the high binding energy side. For the magnetic dichroism we find a value of 0.06 ± 0.02 with respect to the background-corrected peak intensity with unpolarised light. From this we estimate that the spin polarisation of the core hole is equal to 0.19 ± 0.06. The identical spectral lineshapes of the dichroism in the Auger (solid symbols) and resonant photoemission (solid line) spectra in Figure 29b indicate a similar origin, viz. the orientation of the intermediate 2p core hole. The non-zero dichroism in the Auger decay for Fe shows the existence of a strong mechanism which aligns the core hole and valence spins. We can presently only speculate about its origin but the absence of this effect in Ni seems to indicate a connection with the (de)localised nature of the spin-polarised valence electrons.

Spin-dependent properties can be obtained in spin-integrated measurements because of the spin-orbit coupling of the 2p core hole. The detection of core hole polarisation in both resonant photoemission and Auger decay using circularly polarised X-rays adds considerably to the possibility of studying the magnetic properties of materials.

The localised and element-specific nature of these probes enables them to investigate surface and interface magnetism in solids, thin films and nanostructures.





H.A. Dürr (a), G. van der Laan (a), D. Spanke (b), F.U. Hillebrecht (b) and N.B. Brookes (c), submitted to Physical Review

(a) SRS Daresbury Laboratory (UK)

(b) Institut für Angewandte Physik, Heinrich-Heine-Universität, Düsseldorf (Germany)

(c) ESRF




New light on the XMCD of the L2,3 edges of rare-earth paramagnetic insulators



One of the key issues in the magnetism of rare-earth materials is to understand how the strongly localised 4f moments of the rare-earth atoms couple to the surroundings. It is most often assumed that one should consider two types of exchange interactions:

(i) intra atomic 4f-5d exchange,

(ii) inter atomic coupling involving the 5d electrons and the surrounding atoms (see Figure 20).

The latter (super) exchange interaction is, however, poorly understood mainly because the magnetic moments of the 5d electrons are difficult to study, since the 5d moments are typically one order of magnitude smaller than the 4f moments. Shortly after the discovery of XMCD, it was realised that the element, but also the shell specificity of this technique, could help considerably in understanding the role of the 5d orbitals since one may probe the contribution of the 4f and 5d moments separately. While the 3d --> 4f XMCD spectra (accessible on the Dragon beamline ID12B) are relatively well understood and can be reasonably well described within the atomic multiplet theory, the 2p --> 5d XMCD spectra (which are accessible on the higher energy beamline ID12A) are now stimulating the interest of the theoreticians because of the interplay between band structure and atomic effects.

Recently, we recorded on ID12A the L-edge XMCD spectra of several rare-earth insulators: Gd3Ga5O12, TbF3 and EuS. Under conditions of high magnetic field (7 T) and low temperature (<3 K), we obtained high quality spectra that display very strong dichroic signals (Figure 30) by flipping the photon helicity. This was the first time that XMCD spectra could be recorded in the paramagnetic phase with hard X-rays. As apparent from Figure 30, maximum amplitudes of the XMCD signals can be as high as 2 % for the L3 edge and 7.5 % for the L2 edge. In all cases, we obtained dispersive line shapes that look like the derivatives of the absorption spectra. Indeed, it is easy to check (e.g. by reversing the magnetic field) that the XMCD signals were not contaminated by artefactual derivative signals. On the other hand, dispersive lineshapes are precisely what one would expect if the XMCD signals are controlled by exchange interactions. Notable differences still exist between the measured XMCD signal and a true derivative lineshape, especially in the pre-edge region of the L3 edge spectra where we found a signature of a weak 2p -> 4f quadrupolar transition.

We also compared the XMCD spectrum of the paramagnetic sample Gd3Ga5O12 with that of the isostructural ferrimagnet Gd3Fe5O12 which has a Curie temperature of 566 K. It can be seen from Figure 30 that the XMCD spectra are almost identical. This indicates that the XMCD spectrum is in both cases dominated by the intra atomic 4f-5d exchange interaction and is nearly insensitive to the inter atomic (super) exchange. Furthermore, the XMCD lineshape of Gd3Fe5O12 was found to be temperature independent, while the magnitude of the XMCD follows the 4f sublattice magnetisation. This tends to confirm that the

XMCD signature is basically a very local property, at least for these compounds.

These results shed new light on the detailed interpretation of the XMCD spectra of rare-earth metals or intermetallic compounds that were measured earlier. It is indeed expected that the exchange interaction between the localised 4f shell and a photoelectron featuring the d-type symmetry should also be present in metallic or intermetallic systems.

These results have obviously attracted the interest of the theoreticians (M. Van Veenendaal, J.B. Goedkoop and B.T. Thole, to be published): they checked that the measured XMCD lineshapes could still be well described by an atomic multiplet model for the transition from the initial state 4fN 5d0 to the 2p 4fN 5d1 final state.




J.B. Goedkoop (a), A. Rogalev (a), C. Neumann (a) and J. Goulon (a), Proceedings of XAFS-IX, in print

(a) ESRF




Structure and magnetism of Pd in Pd/Fe multilayers studied by XMCD at the Pd L2,3 edges



We have measured X-ray Magnetic Circular Dichroism (XMCD) at the L2,3 absorption edges of Pd in Pd/Fe multilayers with different Pd interlayer thicknesses. The high quality of the data has allowed us to obtain information about the atomic and electronic structures as well as the magnetic properties of Pd in these multilayers. Changes in the position, shape and branching ratio of the L3 and L2 edges give information about changes in the electronic structure of the Pd atoms. The hybridisation of the Pd 4d states with the Fe 3d band leads to a small increase in the number of 4d electrons and to a narrowing of the Pd 4d band.

The spectral features after the L3 and L2 white lines for the thinnest Pd interlayer thickness of 2 atomic layers (AL) are different from Pd metal or other samples where Pd has a fcc local structure (see Figure 31). This indicates that the interface Pd atoms have a local structure, imposed by the epitaxial growth on Fe, which is different from fcc. The spectra also show that these changes in atomic and electronic structures take place only for Pd layers directly at the interface, indicating a very good quality of the interfaces in our multilayers.

The XMCD results give a direct measurement of the Pd magnetic moments. The magnetic moment as a function of the distance to the interface is shown in Figure 32. The first four monolayers of Pd close to the interface are polarised, with a maximum moment of 0.4 µB for the first layer. The magnetic interface is therefore larger than the crystallographic one, which means that the interface atoms are polarised by direct Fe-Pd hybridisation, but that the adjacent layers are polarised by Pd-Pd interactions. The contribution of the orbital moment to the Pd magnetism is small, in contrast to what was found for the transition metal atoms in Pd/Co multilayers (see Table 1). The Pd moments therefore do not contribute directly to the in-plane anisotropy of these multilayers.




J. Vogel (a), A. Fontaine (a), V. Cros (b), F. Petroff (b), J.P. Kappler (c), G. Krill (d), A. Rogalev (e) and J. Goulon (e), submitted to Phys. Rev. B

(a) Laboratoire de Magnétisme Louis Néel, CNRS, Grenoble (France)

(b) Unité mixte CNRS-Thomson CSF, Orsay, France.

(c) IPCMS-GEMME, Strasbourg (France)

(d) Lure, Orsay (France)

(e) ESRF




Fluorescence detection mode of X-ray magnetic dichroism - Can it be trusted?



X-ray absorption spectroscopies are based on the measurement of the absorption coefficient. Ideally, this quantity should be measured in a transmission experiment, by measuring the ratio of the beam intensities after and before traversal of a thin specimen, as a function of the photon energy. Such a direct measurement is very seldom possible in reality, especially in the soft-X-ray range, where the high absorption would require sample thicknesses well below 1 µm. It is therefore usual to measure the absorption in a less direct way, by monitoring the decay products of the absorption process, i.e. the electrons or the photons produced in the de-excitation of the system, when the hole created in an atomic core by the absorbed photon is filled again by Auger or by fluorescence processes. In the total electron yield mode all electrons leaving the sample are collected, irrespective of their energy, and there is reasonable evidence showing that in many instances this electron count is proportional to the number of core holes created by the incoming beam, i.e. to the absorption coefficient.

Electron yield is a widespread technique, but it has nonetheless practical limitations. One of these is particularly severe when performing X-ray Magnetic Circular Dichroism (XMCD) measurements; in such experiments, the differences in absorption coefficient between left and right circularly polarised photons are measured. These differences are non-vanishing in systems displaying a macroscopic magnetisation, like ferromagnets (in an external saturating field) or even paramagnets if the external magnetic field is large enough. The problem is that the path of electrons is deflected in a magnetic field so that the whole collection efficiency is drastically affected. It is therefore often better to turn to the detection of fluorescence photons, which are not affected by magnetic fields.

XMCD is becoming an important technique because it is a sensitive and atom-specific probe of the magnetic moments, suitable to investigate novel materials such as magnetic multilayers, thin films as well as magnetic alloys. A few years ago, it was shown by the ESRF Theory group that suitable integrals of the dichroic signal from a given absorption edge of an atomic species provide a direct determination of the orbital and in some cases also of the spin magnetic moment of that atom. In order to apply these «sum rules» to experimental results obtained by monitoring the fluorescence yield, it is essential to establish whether the fluorescence signal for different polarisations, integrated over the incoming photon energy in the whole absorption edge region, is strictly proportional to the corresponding integrated absorption signal.

M. van Veenendaal and B.T. Thole, of the Theory Group, together with J. B. Goedkoop of the XAS Group have investigated this problem in detail for the L2,3 edges of the 3d transition metal series and the M4,5 edges of the 4f rare-earth series. They have computed both the absorption and the fluorescence yield spectrum, for three different polarisations, for magnetically polarised ions in a crystal field environment.

Figure 33 displays the results for the M5 edge of a Dy3+ (4f9) ion. Differences between the absorption spectrum, on the left, and the fluorescence yield, on the right, are easily visible, both for the individual polarisations and for their sum. The XMCD spectrum is the difference of the q = +1 and q = -1 curves. What can be said about the integral of the XMCD spectrum, which is the relevant quantity for the application of sum rules? From the results of calculations like those displayed in Figure 33, it is possible to define a proportionality coefficient < V>q between integrated absorption spectra for polarisation q, (IAbs)q, and the corresponding fluorescence spectra, (IFluor)q, so that (IFluor)q = (IAbs)q <V>q. Clearly, if <V>q is dependent on q, i.e. if it is different for q = 1 and q = -1, then the dichroic signal is strongly distorted, and the sum rules would give a wrong result when applied to the fluorescence spectra. In Figure 34, the results for <V>q are shown for the complete 3dn and the 4fn series, where n denotes the occupation of the magnetic shell. It is obvious from the figures that the assumption of a polarisation-independent <V>q is quite good for the transition metals, as well as for the early part of the rare-earth series, but that a large discrepancy is found in the later part of the series.

How are these results to be understood? What is the origin of the different behaviours of the two series? A careful analysis shows that the crucially different parameter is the sensitivity to the crystal field, that is to say the extent to which the electronic wavefunctions in the crystal environment are perturbed with respect to those in an isolated ion. In fact, one can show that, for isolated ions, a strong polarisation dependence of <V>q can be expected in the more than half-filled shells because of the stringent selection rules for optical transitions imposed by the spherical symmetry. In the case of the compact, core-like 4f wavefunctions, which are hardly affected from the crystalline environment, this feature is preserved in the solid state. For the less localised 3d transition metal orbitals, on the other hand, which in the crystals are sufficiently mixed and hybridised to acquire band-like features, the atomic selection rules are washed out to a large extent and the polarisation-independence of <V>q is obtained.

In conclusion, the final verdict on the reliability of the fluorescence detection mode for the magnetic dichroism sum rules is that it can be safely applied to the 3d transition metals and to the early part of the rare-earth series, but that is should be used with extreme caution in the late rare-earths.





M. van Veenendaal (a), J. B. Goedkoop (a) and

B. T. Thole (a, b), Phys. Rev. Lett. 77, 1508 (1996)

(a) ESRF

(b) Laboratory of Applied and Solid State Physics, Groningen (Netherlands)




XMCD in fluorescence emission spectra of 4d transition metal alloys



The detection of XMCD in fluorescence emission spectra is a new technique which was first pioneered by Hague and co-workers at LURE in 1993. The information content of such experiments is slightly different from what can be unravelled using conventional XMCD spectra recorded either in the transmission mode or in the fluorescence excitation mode. In fact, the two techniques have a remarkable complementarity:

  • in conventional XMCD experiments, the excited photoelectron is probing the exchange interaction in empty spin polarised excited states;
  • in XMCD fluorescence emission spectra, one is directly probing the exchange interaction in the (relaxed) filled valence band, taking advantage of the presence of a core hole created by the primary absorption process.

    So far, XMCD fluorescence emission spectra were recorded only in the soft X-ray range and concerned the valence states of 3d ferromagnets. For the first time, we have shown that this technique can be extended at much higher energies: we have recorded on ID12A fluorescence emission spectra of 4d transition metals alloyed to a ferromagnet (e.g. Co25Rh25). Two experimental options were initially envisaged:

  • Use directly the wide band, circularly polarised emission of the helical undulator Helios-II on ID12A to eject one 2p1/2, 3/2 core electron of the 4d transition metal into the continuum;
  • Use the monochromatic radiation to resonantly excite one 2p1/2, 3/2 core electron of the 4d transition metal into a spin polarised empty state just above the Fermi level.

The latter approach (i.e. resonant photon-in / photon-out process ) is expected to yield a larger dichroic signal but it turned out to be more difficult from the experimental point of view because the polarisation transfer function of the Si (111) monochromator is very poor at the L2,3 edge of Rhodium (at around 3 keV). This led us to prefer the former approach which certainly suffers from the handicap that it generates smaller XMCD signals but has still the great advantage that it measures unambiguously differences in the filled valence band while a resonant excitation may convolve the dichroism of filled and empty states.

We have reproduced in Figure 35 the Rh Lb2,15 fluorescence line excited either with right or left circularly polarised X-rays and, with a suitable scale, the corresponding difference in the emitted intensities. For this project, we have inserted a high-resolution fluorescence spectrometer in the so-called «white beam» experimental station located upstream with respect to the monochromator. We used one single channel of the four-mirror device to limit the bandpass of the emission spectrum of the undulator source: in practice, we were left only with the first harmonics emission featuring a measured circular polarisation rate of more than 97 %. The sample was cooled down to liquid nitrogen temperature and a permanent magnet was used to align the aimantation practically along the direction of the incident beam. The fluorescence analyser combined a quartz crystal (1011) bent to a radius of 0.8 m plus a position sensitive detector. The vacuum was kept of the order of a few 10-8 mbar. The data acquisition time was comparable to the data acquisition time of conventional XMCD experiments.

It should be born in mind that this XMCD experiment addresses a quite fundamental question in magnetism: can we access selectively to the spin dependent difference in the electron scattering of each individual component in alloyed ferromagnets? Due to the element specificity of X-ray emission spectra, this technique may now be considered seriously as a useful tool to learn more about the magnetism in such materials. A spin-polarised relativistic band structure calculation of this hcp alloy is underway in order to make our interpretation more quantitative.



C. Hague (a), J.J. Gallet (a), J.M. Mariot (a),

J.P. Kappler (b), K. Hricovini (c), G. Krill (c), A. Rogalev (d), J. Goulon (d), Proceedings of XAFS IX, in print

(a) Laboratoire de Chimie Physique, URA-CNRS, Univ. of Paris VI (France)

(b) IPCMS-GEMME, Strasbourg (France)

(c) LURE, Orsay (France)

(d) ESRF





Observation of magnetic circular dichroism in resonant inelastic X-ray scattering at the L3 edge of gadolinium metal


Magnetic circular dichroism is observed in the inelastic X-ray scattering from d core electrons in magnetically aligned gadolinium at incident photon energies resonant with the Gd 23/2 excitations. The dichroism is dominated by the magnetic interactions between the valence 4f electrons and the final state d core-hole. Differently from photoabsorption - and similarly to core photoelectron - spectroscopy, this study allows us to probe electrons correlation effects without changing the valence ground state population.



M.H. Krisch (a), F. Sette (a), U. Bergmann (a), C. Masciovecchio (a), R. Verbeni (a), J. Goulon (a), W. Caliebe (b), C.C. Kao (b), Physical Review B, in print

(a) ESRF

(b) NSLS Brookhaven National Lab., Upton N.Y. (USA)






Magnetic Compton profiles and cross-section of iron at photon energies from 100 keV to 1 MeV


Within the impulse approximation the spectrum of inelastically (Compton) scattered photons can be interpreted in terms of the ground state electron momentum distribution of the scatterer. In magnetic Compton scattering, using circular polarised photons and the external magnetic field to align the magnetic moments, it is possible to measure the spin momentum density arising from those electrons which have unpaired spins. The cross-section for magnetic inelastic scattering depends only on the spin magnetisation, which has been proven in experiments on material carrying large orbital moments.

All Compton scattering experiments up to now have been performed at rather small energies around 60 keV. There are at least two reasons to increase the energy at which magnetic Compton scattering experiments are performed. The first is to study the magnetic scattering cross-section and discover whether it deviates from the prediction which has been developed for photon energies E << mc2 and to order (E/mc2)2. The second objective is to determine an optimum energy for magnetic Compton scattering experiments. At a practical level the strength of the magnetic scattering relative to the charge scattering increases with energy. This is important because the magnetic contribution to Compton scattering is small compared to the charge term, magnetic effects being typically of the order of 1 % of the charge effects for pure Fe at 60 keV, but only 0.05 % for compounds like HoFe2, which has a much smaller ratio of unpaired to spin-paired electrons. In addition to an increase in the magnetic effect, higher photon energies lead to an improvement in resolution when semiconductor detectors are used. The resolution of magnetic Compton experiments is dominated by the contribution from the detector resolution and is typically 0.6 to 0.8 a.u. at 60 keV. The magnetic Compton profiles presented were measured at the best resolution (² 0.4 a.u.) achieved to date without the use of a crystal analyser spectrometer.

With the availability of the superconducting wavelength shifter (SCWS) at the High Energy beamline it is now possible to extend the energy range for this kind of measurements. The critical energy of 96 keV for this insertion device enables experiments at photon energies up to 1 MeV. Having an asymmetric magnetic structure, the insertion device provides circular polarised photons off the orbit plane of the storage ring. The experiments have been performed on an iron sample at energies in the range from 84.4 to 1000.6 keV. For energies up to 256 keV an asymmetric wiggler was used but, for the very high photon energies, the SCWS is clearly the best insertion device.

The measurements have been performed at eight different energies but only at some energies was it attempted to gain high statistical accuracy of the data. The measurement of the integral of the magnetic Compton profile will give a means to the cross-section behaviour in the further data evaluation. An analysis of the data taken with the asymmetric multipole wiggler in the first experiment showed that the magnetic effect increases up to 256 keV proportionally to the magnitude of the scattering vector as predicted by the low energy approximation.

The evaluation of the data showed that the increase in the instrumental resolution gives rise to features in the profiles which could not be obtained up to now, but had already been predicted in a FLAPW (Full Potential Linearised Augmented Plane Wave) calculation. Not only the well-known «volcano» structure of the magnetic Compton profile is observed, but also the deepness of the dip and shoulders at 0.5 and 4.0 a.u. fit very well with the calculation. (Figure 36) The data taken at high energies between 470 and 1000 keV have not been analysed yet carefully enough to give a conclusion whether the cross-section still holds at these energies. A comparison of the magnetic and charge profiles taken at energies far below, about and above the momentum transfer corresponding to the rest energy of the electron, will allow the cross-section behaviour to be concluded. Additionally pure magnetic scattering may be obtained which is very weak but increases proportionally to |K|2.



J.E. McCarthy (a, b), M.J. Cooper (b), P.K. Lawson (b), D.N. Timms (c), S.O. Manninen (d), K. Hämäläinen (d) and P. Suortti (a), J. Synchro. Rad.

(a) ESRF

(b) Department of Physics, Univ. of Warwick (UK)

(c) Division of Physics, Univ. of Portsmouth (UK)

(d) Department of Physics, Univ. of Helsinki (Finland)