The generation and use of polarized X-rays for the investigation of magnetic systems has always had a prominent place in the scientific program of the ESRF, with beamlines entirely devoted to magnetic scattering and dichroism. Nowadays, the basic principles and merits of the different techniques have been established, and this field of research has attained a stage of maturity. Nonetheless, a large variety of systems are being investigated, and very interesting results are still copiously obtained. Resonant magnetic scattering experiments, once thought to be of interest only for rare-earth or actinide based magnetic systems, are increasingly used to probe 3d transition metal systems in the photon energy range near the K edge. The interplay of quadrupolar resonances involving the localized d electrons, and of dipolar resonances to the delocalized p electrons, is shining new light on the electronic and magnetic properties of transition metals and their compounds, notably oxides, which encompass so many interesting phenomena, including high-temperature superconductivity. X-ray magnetic scattering and dichroism are also very useful in the study of surface and thin film magnetic phenomena, a field of growing fundamental as well as technological interest, where the interplay of structural and magnetic properties is very important. In multilayer or alloy systems, the element specificity of X-ray dichroism and resonant scattering is a definite advantage over competing techniques. It should also be mentioned that the application of polarized X-rays to the study of magnetism has been one of the main axes of theoretical research at the ESRF, and it continues to benefit from a strong collaboration and interplay between theory and experiments.


Resonant magnetic X-ray scattering in NiO

In the previous edition of the ESRF Highlights (p.14), it was shown that the polarization dependence of the non-resonant magnetic scattering allows a clear separation of spin and orbital contributions to the total magnetic density in NiO. Investigations of the magnetic scattering from NiO have been taken further by studying resonant effects.

Resonant magnetic X-ray scattering is interpreted in terms of electronic multipole transitions from a core level to an empty state above the Fermi level. Important enhancements of the diffracted signal near an absorption edge indicate that the electronic levels involved exhibit spin-orbit splitting and the intermediate states above the Fermi level are strongly spin-polarized. In the case of the 3d transition metal crystals only the K-edges can be exploited in diffraction experiments where two resonant channels are available. The first one leads an electron from the 1s level to the highly polarized 3d levels via weak quadrupolar transitions, and the second one (dipolar transition) brings a core level electron to the 4p levels, which are only weakly spin polarized. The study of magnetic resonances in 3d systems appears to be an important step for the understanding of the polarisation of p electrons bands. This point is further supported by the observation of orbital order.

In the case of NiO, the experiments on ID20 have made possible the observation of both resonances, whereas previous experiments had only given evidence of the quadrupolar transition [1]. Figures 55 and 56 show the integrated and corrected intensity of the magnetic Bragg peak (3/2, 3/2, 3/2) as a function of photon energy in the two polarisation channels ss and . Two resonances can be observed. The first resonance (8.344 keV) shows an enhancement by a factor two and is present in the two polarisation channels. This is typical for quadrupolar transitions. The enhancement of the second resonance (E = 8.358 keV), corresponding to the dipolar transition, is also of the order of two. This enhancement can only be observed in the rotated channel, in agreement with the existing model. The existence of a dipolar resonance indicates that the p levels of Ni are partially spin polarized. The structure of the dipolar resonance differs from that observed in other transition metal compounds: e.g. in RbMnF3 where a complex line shape had been observed. These results will be analyzed in terms of character mixing of p and d electrons. Complementary studies of resonant magnetic X-ray scattering in CoO and MnO are in progress. Preliminary results on these compounds also show resonant enhancements of dipolar and quadrupolar origin but with different energy profiles.

[1] J.P. Hill, C.-C. Kao, and D.F. McMorrow. Phys. Rev. B., 55, R8662, 1997.

W. Neubeck (a), C. Vettier (a), F. de Bergevin (b), F. Yakhou (a), D. Mannix (c) , A. Barbier (d), to be published.

(a) ESRF
(b) CNRS, Laboratoire de Cristallographie, Grenoble (France)
(c) EITU, Karlsruhe (Germany)
(d) DRFMC, CEA, Grenoble (France)




Incommensurate Magnetism in PrBa2Cu3O6.92

One strategy for uncovering the mechanism behind high-Tc superconductivity is to study compounds in which the superconductivity is anomalously suppressed. In this regard, PrBa2Cu3O6+x has attracted a great deal of attention as an anomalous member of the [RE]Ba2Cu3O6+x series, where RE = rare earth or Y. In addition to the absence of superconductivity it possesses other unique magneto-transport properties; it is an insulator, exhibiting Cu antiferromagnetism for all x and the Pr sublattice appears to order at unexpectedly high temperatures, TPr(x) = 10-20 K, relative to the other members of the series for which TRE = 0-2 K. This behavior is particularly puzzling in the light of the fact that the hole densities in the CuO planes and chains are very similar to the superconducting members of the series. The large value of TPr suggests that the Pr-Pr magnetic coupling is enhanced by electronic interactions with the CuO planes and interest has therefore focused on the Pr site magnetism. Here, X-ray resonant magnetic scattering can make unique contributions because one can selectively study the Pr moments without contributions from the Cu sublattice, by tuning the incident photon energy to a Pr absorption edge.

This element specificity is of particular importance because of a debate that has arisen over the presence of Pr order. While neutron scattering measurements have been interpreted in terms of an ordered Pr moment, these data cannot easily distinguish the Pr and Cu sublattices and NMR experiments find little or no ordered Pr moment.

The data of Figure 57, showing a resonance at the Pr LII edge, unambiguously demonstrate the presence of an ordered moment on the Pr site, thus resolving the debate. Further, scans through the antiferromagnetic position (Figure 58) reveal that the peak is split, i.e. that the Pr site magnetism is incommensurate, with a modulation wavelength of 600 Å. This splitting is too small to have been detected in the neutron scattering experiments. The modulation was found to be highly ordered, with a correlation length in the CuO2planes in excess of 900 Å, and ~300 Å perpendicular to the planes. Interestingly, the data of Figure 58 also show that the incommensurability increases by approximately 30% on warming from T = 5 K to TPr = 19 K.

An important open question is whether the Cu site magnetism is also driven incommensurate. Here X-ray magnetic scattering in the vicinity of the Cu K-edge would be invaluable. Preliminary experiments performed at ID20 measured scattering at the Cu antiferromagnetic Bragg position. Count rates of 10 counts per sec were obtained. If confirmed, this would indicate that at least one component of the Cu magnetic order remains commensurate, and these results would represent the first observation of X-ray magnetic scattering from the copper spin in a high-Tc system, pointing the way to further such studies at third generation sources.

In summary the high resolution and element specificity of resonant X-ray magnetic scattering have revealed new features in the magnetism of this intriguing compound. While it remains to be understood whether or not the incommensurate modulation is associated with the suppression of superconductivity, perhaps in an analogous manner to the static stripes observed in the "214" materials, it is clear that this technique can provide new insights into the problem of high-Tc superconductivity.

J.P. Hill (a), A.T. Boothroyd (b), D.F. McMorrow (c), C. Vettier (d), A. Stunault (d), A. Markvardsen (b), D. Gibbs (a), N. Andersen (c), E. Brecht (e) and Th. Wolf (e), to be published.

(a) Brookhaven National Laboratory (USA)
(b) Oxford University (UK)
(c) Risf National Laboratory (Denmark)
(d) ESRF
(e) Forschungszentrum Karlsruhe (Germany)




First observation of X-ray quadrupolar diffraction peaks


With the development of high-flux X-ray sources such as the ESRF, scattering experiments revealing details of the deep electronic charge distribution of atoms may be envisaged. For instance, it should become possible, with reasonable counting times, to measure Bragg reflections due to a periodical aspherical 4f density in rare-earth compounds.

In such compounds, the pair interactions between the 4f shells drive orderings in which their electronic density no longer respects the point symmetries of the disordered phase: the dominant modification of the asphericity is described by the emergence of additional quadrupolar components. If the lattice translational symmetries are preserved, the arrangement is called ferroquadrupolar, otherwise the ordering, in some way reminiscent of the antiferromagnetic order, is called antiferroquadrupolar. Such types of antiferroquadrupolar arrangements are supposed to stabilize even in the absence of magnetic order. However, until now, due to the lack of an unambiguous macroscopic or microscopic technique, the possible cases of antiferroquadrupolar order are still open to debate.

Although the multipolar scattering amplitudes involved are very weak, the Thompson scattering appears as the most efficient technique for solving these questions. Moreover, it is at present the only one allowing a quantitative interpretation of the phenomena [1, 2]. In contrast to the spherical scattering amplitude, the multipolar ones are maximum for non-zero scattering angle, the optimum being reached for sinq/l 0.5 Å-1 for the quadrupolar scattering. For this optimum, one may then estimate a quadrupolar reflection to be five orders of magnitude less than a lattice reflection. This is indeed a weak diffraction signal, but substantially larger than, for instance, a X-ray non-resonant magnetic one.

In order to validate this technique, we have applied it to a rare-earth compound presenting a multiaxial magnetic structure, NdMg. The advantage of using such a magnetic structure is that it coexists with a well-defined antiferroquadrupolar arrangement, unambiguously deduced from the magnetic structure itself.



NdMg is a cubic compound (CsCl-type) which orders antiferromagnetically at TN = 61 K and displays a second magnetic transition at TR = 35 K. From neutron diffraction, it has been established that the transition at TR corresponds to a change from a collinear structure, at high temperature to a multiaxial one, at low temperature, both based on wave vectors from the (1/2, 0, 0) star. NdMg's magnetic multiaxial structure consists of a double-k structure, with moments along twofold axes [3]. In the plane defined by the moments' directions, this structure corresponds to an antiferroquadrupolar arrangement with a (1/2, 1/2, 0) wave-vector. This magnetic structure is then suitable for the observation of X-ray antiferroquadrupolar diffraction peaks, at reciproqual space nodes Q = (h k l) + (1/2, 1/2, 0), where (h k l) is a reciprocal lattice vector.



This experiment has been performed using the 7-circle goniometer of the D2AM beamline and a closed-cycle helium refrigator to cool the sample below TR = 35 K. As very weak signals had to be drastically optimised, a short X-ray wavelength = 0.8943 Å, well below the Neodymium L absorption edges, was used. In order to further decrease the background level, a Ge(111) analyser was mounted in front of the detector.

The illuminated face of the crystal, obtained by cleaving, was unaffected by mechanical stress or any other damage associated with spark or mechanical cutting. This face, perpendicular to the (1 0 0) direction, was tilted in order to have the axis confused with the (1 1 0) direction. All measurements were performed in reflection conditions. With this sample setting, all reflections, at small enough Q, within the (h k 0) positive plane of the crystal were accessible.

The sample was then cooled down to 19 K. Scanning the lattice reflections, it appeared that the tetragonal striction was large enough to separate, in Q-space, the peaks associated with the three magnetic domains (Figure 59). As only one domain could result in quadrupolar diffraction peaks in the accessible (h k 0) plane, it was necessary to focus on the corresponding reciprocal lattice. Unfortunately, this domain appeared to be the least represented one, its associated Bragg reflections being approximately ten times less than for the two other domains.

Attention was first focused on the (5/2, 5/2, 0) quadrupolar reflection which was expected to be the most intense. First, the background counting level was estimated about the (5/2, 5/2, 0) node, its value being a little less than 2 counts/s. Considering the order of magnitude of the expected quadrupolar reflection (some tenth of count per second), counting times of at least ten minutes per point were necessary for the peak to emerge from the background. Due to the w extension of the Bragg reflections, a diagonal scan in the (h k 0) plane appeared to be the most appropriate process. Another advantage of this h,k scan was to measure the equivalent (5/2, 5/2, 0) nodes of the two other domains, which allows one to check the systematic emergence of peaks at such positions, in relation, for instance, with the /2 harmonic of the monochromator. As a result of these scans, with total counting time of 20 minutes per point (108 monitor counts), only one peak was clearly defined and located at the expected position for the quadrupolar scattering (Figure 60, upper part), whereas no /2 peaks were identifiable. The maximum of the peak reached more than 200 counts above the background, that is about 0.2 count per second, which is typically the expected order of magnitude.

Equivalent measurements were performed about the (3/2, 5/2, 0) reflection. This scan also revealed a peak at the expected position, even better defined than the previous one (Figure 61, upper part). In both cases, the peaks' Full-Width-at-Half-Maximum are comparable to the lattice reflections close to these reciproqual space positions.

A check of these reflections at high temperature shows a background level increased by about 40%, due to the thermal incoherent scattering of the lattice, but peaks could no longer be identified at the (5/2, 5/2, 0) and (3/2, 5/2, 0) positions (Figure 60 and 61, lower parts).

In summary, the existence of diffraction peaks at the positions and with the order of magnitude expected for quadrupolar reflections has been established in NdMg. These peaks disappear as the temperature is raised into the paramagnetic range and cannot be ascribed to /2 pollution. They have been observed despite an unfavourable domain partition of the crystal and using a Neodymium based sample, which does not correspond to the maximum quadrupolar scattering amplitude among the rare-earths. This experiment demonstrates that the tiny X-ray reflections associated with 4f orbital orderings can be efficiently measured thanks to the high flux available from synchrotron radiation sources.

[1] M. Blume, A.J. Freeman and R.E. Watson, The Journal of Chemical Physics 37 (1962) 1245.
[2] M. Amara and P. Morin, Submitted to Journal of Physics Condensed Matter (1998).
[3] M. Deldem, M. Amara, R.M. Galéra, P. Morin, D. Schmitt and B. Ouladdiaf, Journal of Physics: Condensed Matter 10 (1) (1998) 165.

M. Amara (a), R.M. Galéra (a), P. Morin (a) and J.F. Bérar (b), to be published.

(a) Laboratoire Louis-Néel*, CNRS, Grenoble (France)
(b) Laboratoire de Cristallographie*, CNRS, Grenoble (France)
* Associated with the University Joseph-Fourier of Grenoble.




FexMn1-x alloy phases strained in thin films.

Depth analysis of FexMn1-x/Ir(001) multilayers by X-ray diffraction

3d metals, such as Fe, Ni and Mn, and their alloys are intensively studied for their moment-volume instabilities. In the cases of the Fe and Mn g-phases (fcc structure), the ground states are predicted to be antiferromagnetic (AFM) but, if the volume is expanded beyond a critical value, a ferromagnetic (FM) order takes place, with a magnetic atomic moment increasing with the volume. Therefore, much attention has been paid to stabilizing the -phase at room temperature with an expanded volume.

The structural features of FexMn1-x alloy strained in thin films and in superlattices (SL) were measured with the 7-circle diffractometer of the D2AM French CRG beamline. The study focused on the high Fe concentration range where the phase diagram of the bulk alloys is rather complex. As the Fe content increases, the alloy undergoes a first martensitic transition at x = 0.7 (fcc to hcp) and a second one at x = 0.9 (hcp to bcc). The alloy magnetic state is AFM up to x = 0.9 and FM above. The transitions can be suppressed, by adding 4% of carbon and the fcc alloy is AFM even beyond x = 0.9.

Another way to modify the equilibrium phase of the FexMn1-x binary alloy is to prepare a thin film via epitaxy. Actually, the buffer stress is strongly anisotropic since it is applied parallel to the surface. As a consequence, the stabilized structures deviate significantly from the cubic symmetry and are tetragonally distorted. This tetragonalization is however very interesting since it allows the change in the magnetic moment amplitude to be studied during the continuous fcc to bcc transformation, obtained by varying the tetragonal ratio c/a between 1 (bcc) and 2 (fcc), along the so-called Bain deformation path. The bcc phase being FM (High Spin) and the stabilized fcc one AFM, somewhere along the deformation path, ferromagnetism must disappear. The case of FeMn on Ir(001) is particularly attractive since the two fcc and bcc phases exist in the bulk alloys and the value of the Ir lattice constant is just in between those of the two cubic structures (aIr = 2.715 Å and for x = 0.9, afcc = 2.53 Å and abcc = 2.87 Å).

Two Ir(001)/FexMn1-x/Ir sandwiches and two FexMn1-x/ Ir(001) SL have been studied for Fe concentrations x = 0.7 and x = 0.9. They were prepared by Molecular Beam epitaxy (MBE) and investigated by combining different X-ray scattering geometries. The average out-of-plane atomic spacing of the two FeMn and Ir sublattices were determined by anomalous diffraction. The homogeneity of the 3D structure throughout the multilayers stacking were probed by recording reciprocal space maps (RSM) around the (111) Ir Bragg peak, with in- and out-of-plane components for the scattering vector. The depth analysis of the structure was performed by studying these maps as a function of the grazing incident angle , the emergent angle being large.

In this communication, the case of the [Fe0.9(20Å) Mn0.1(25Å)/Ir(001)]20 SL is highlighted. The RSM around (111) exhibits four diffuse spots, regularly spaced along the L axis, which are provided by the SL modulation (Figure 62). Each SL spot is divided into two intense areas, one centered at h1 = k1 = 0.995 and the other at h2 = k2= 0.9775. The latter is predominant at very low incidence angles (Figure 62a). When increasing a, which means when probing the SL deeper, the intensity area around h1 becomes stronger (Figure 62c). The evolution of the diffracted intensity allows two areas in each satellite to be distinguished, as schematically represented in Figure 62d, and suggests the existence of two phases, labeled SR1 and SR2.

The SR1 phase is homogeneously strained by the Ir buffer and the alloy is in a tetragonal structure close to that observed in the Ir/Fe0.9 Mn0.1/Ir(001) sandwich. In the SR2 phase, h2 gives an in-plane lattice parameter, corresponding to the equilibrium value calculated for one free bilayer with the two FeMn and Ir sublayers in mutual elastic interaction. The analysis of the diffused intensity of SR2 shows that it is due to a correlated distribution of the tetragonal parameters a and c. This means that, when a coherent SL domain exhibits a a fluctuation of the in-plane parameter, this induces, according to the elastic theory, a c variation in each of the Ir and FeMn sublattices. The average out-of-plane parameter also varies, providing a global shift L of the satellites. Furthermore, the spots are always observed centered on h - h2, whatever the value of . It follows that the diffuse intensity originates from domains laterally distributed instead of a gradient of the SR2 lattice parameters along the growth axis, when the Ir buffer stress is progressively relaxed.

The localization and the knowledge of the proportion of the two phases are essential to the quantitative analysis of the magnetic measurements. The variation of the diffracted intensity of both phases as a function of the incident angle have been modelled. The kinematic theory can be applied, in the involved geometry, and the diffracted intensity is proportional to the absorbed intensity [1]. The experimental intensities are obtained by a 2D integration of each spot of the respective phase in the RSM. The simulations lead to a very good agreement between the calculated and the experimental values for the model represented in Figure 63.

The two SL phases, evidenced in the Fe0.9 Mn0.1/Ir(001) SL, arise from the two structures existing in the bulk alloy with 90% Fe atoms. The relaxation of the thick buffer stress occurs for an SL thickness of about 300 Å; afterwards the SL behaves as if it were free of buffer. The coexistence of these two phases in this SL is interesting since it allows one to look at the influence of the structure on the magnetic properties for the same Fe and Mn concentrations in the alloy.

Finally, the tetragonal ratio of the strained alloy has been found to lie between 1.06 and 1.14 for SR2, c/a = 1.17 for SR1 and c/a = 1.23 in the Fe0.7 Mn0.3/Ir(001) SL. These structural features of the bct alloy have been used to analyze the magnetic properties which were measured by XRMS at room temperature. They show that the magnetic moment of Fe is in a High Spin (HS) FM state and carries ~2.1 µB in both phases of the Ir/Fe0.9Mn0.1/Ir(001) SL, like the pure bulk bcc Fe. In the Ir/Fe0.7Mn0.1/Ir(001) SL the Fe atoms are in a low-spin (LS) FM state with a magnetic moment of ~0.27 µB.

These results can be consistently understood in the framework of tetragonal Fe, by including other data on bulk alloys and on Fe/Ir(001) SL. They show that a transition between an FM HS state for the bcc structure and LS state occurs at around c/a = 1.2, when Fe atoms are in an AFM state for c/a = 2. They are in a good agreement with the theoretical work of Krasko and Olson who studied the usual bcc (HS-FM) to fcc (NM) martensitic transition along the continuous transformation. They predicted that the FM-HS state is first stable and that a transition to an NM state occurs for the c/a = 1.2. It is also pointed out that the succession of magnetic states observed along this deformation, in other words from an HS-FM state to an AFM state through an LS-FM state, is rather similar to the occurrance of magnetic states when increasing the atomic volume as predicted for fcc Fe by Moruzzi et al [2].

[1] M. Brunel and F. De Bergevin, Acta Cryst. A42 (1986) 299.
[2] V.L. Moruzzi, P.M. Marcus and J. Kübler, Phys. Rev B 39 (1989) 6957.

A. Déchelette (a), M.C. Saint-Lager (a), J.M. Tonnerre (a), G. Patrat, D. Raoux, H. Fischer (b), S. Andrieu, M. Piecuch (b), accepted by PRB.

(a) Laboratoire de Cristallographie, CNRS, Grenoble (France)
(b) Laboratoire de Physique des Matériaux, CNRS / Université Henri Poincaré, Nancy (France)




Growth and magnetism of Co/NiO(111) thin films

The Co/NiO(111) system is promising in the framework of devices based on the spin-valve geometry. NiO(111) is an antiferromagnetic polar oxide insulator highly-resistant to oxidation and with a high Néel temperature (523 K). It is a good candidate for providing the magnetic pinning mandatory in spin-valve structures.

This study was carried out on the BM32 (CRG-IF) beamline, by grazing incidence X-ray diffraction. In previous work, it has been shown that the p(2x2) stabilized (111) surface of the NiO(111) single crystal exists. Here, the interface was investigated during its formation, starting from the very early stages of deposition up to a fairly thick film (~20 nm of Co) with different preparation conditions: room temperature (RT) and high temperature (HT) deposition (703 K and 773 K).

The NiO(111) surface was indexed using a hexagonal surface unit cell. From the symmetry of the system, HCP, FCC (the stacking which continues the NiO one) and twined-FCC Co (rotated by 60°) could be expected. In-plane scans along the [h,h,0] direction, passing through the Co Bragg peak common to all these stackings, show a great difference between the two situations. For the RT deposit, the signal of crystalline Co is found to appear only after ~1nm of Co deposition, whereas for the HT deposit, the signal is found in the very early stages of growth. Moreover, the intensity is 200 times larger for the HT deposit compared to the intensity of the RT one, both for 20 nm thick films. Thus the RT deposit is poorly ordered whereas the HT deposit is epitaxic and well ordered.

Out-of-plane scans along the [1,0,l]Co rod allow a distinction to be made between different Co stackings because they cross successive Bragg peaks of HCP, FCC and twined-FCC stackings (Figure 64). At the end of the RT deposit, the [1,0,l]Co rod shows no peaks at the expected Co positions. It is only after recrystallization (by annealing the film for 45 minutes at 973 K) that FCC, twined-FCC and HCP Co peaks appear. Integrated and corrected intensities of the peaks show the presence of all stackings in almost equal amounts in the Co layer. For the HT deposits, the intensity of all Co peaks increases with Co thickness. The FCC stacking is strongly dominant for all thickness, but residual contributions of all other stacking are present for the deposit at 703 K. At 773 K, no HCP Co is found and 98-99% of the 20 nm thick Co film is FCC. The metastable Co FCC phase has thus been stabilized for a very large deposited thickness. At the end of the deposition, the samples were brought to a temperature higher than the Néel temperature of NiO and left to cool to RT under a magnetic field strong enough to ensure the saturation of the ferromagnetic layer, in order to magnetically couple the Co layer and the NiO substrate.

Magneto-Optical Kerr Effect (MOKE) measurements are shown in Figure 65. The annealed RT deposited film presents a large and distorted hysteresis loop while the HT sample shows a much squarer one. For the 703 K deposit, a study of the evolution of the coercive field (Hc) and of the shift of the hysteresis loop (Hxc) with respect to the temperature was achieved. Hc increases with decreasing temperature, passing through a maximum value of 855 Oe at 50 K. Below 150 K, the exchange field appears. It has a similar behavior to Hc, with a maximum of 135 Oe at 50 K. No exchange field was observed at RT for this sample. The values of Hc obtained on NiO(111) single crystals are close to the largest values measured on spin-valves. The thickness of the Co layer in these samples is however ten times larger than the usual thickness of the pinned layer in spin-valves (of the order of 30 Å). Since the coercive field results here from a balance between a volume energy (the Zeeman coupling of the Co magnetization to the external field) and the interfacial coupling energy between NiO and Co (i.e. Hc is inversely proportional to the thickness of the Co film), there is the implication that the coupling energy in these samples is ten times larger than in sputtered spin-valves.

Although Hxc only appears at low temperature, it has been shown that the magnetic properties of interest exist when the substrate is a single crystal. Moreover, they seem to be dominated by the crystalline quality of the Co layer. In conclusion, the NiO substrate temperature strongly influences the Co film structure. A clear correlation between the crystallographic structure of the Co film and its magnetic properties is shown. It is also shown that characteristic magnetic properties (Hc, Hxc) exist when the substrate is a single crystal and that the magnetic properties are strongly improved by increasing the structural quality of the layer, a feature which is obtained at HT.

A. Barbier (a), G. Renaud (a), C. Mocuta (a) and A. Stierle (b), Surface Science, in press (1998).

(a) SP2M, CEA-Grenoble (France)
(b) ESRF




Spin and orbital magnetism in self-assembled Co/Au(1 1 1) clusters

Supported nanoscale particles are of fundamental interest as a link between the localized atomic and itinerant metallic description of the solid state and are of growing importance in magnetic information storage devices. The reduced dimensionality of the particles dramatically influences their magnetic properties through changes in 3d electron localization and hybridization. These effects have been probed with magnetic circular X-ray dichroism (MCXD) which measures the orbital magnetic moment, mL, a direct link between electronic and magnetic properties. It is shown that an increase of the magnetic anisotropy energy with decreasing cluster size is caused by an enhancement of the orbital moment per spin due to electron confinement.

Co clusters were chosen, supported on an Au(111) single crystal surface as a model system. The clusters were self-assembled at defect sites of the reconstructed Au(111) surface during molecular beam epitaxy under ultra-high vacuum conditions [1]. Their morphology is very well known from scanning tunneling microscopy studies and is characterized by nearly circular pancake-shaped clusters of two atomic layers (AL) height for a Co coverage below an equivalent of = 1 AL [1, 2]. Above that coverage the individual clusters start to percolate [1, 2] leading to structures with more fractal-like boundaries.

Magnetic characterization of the in-situ grown clusters was performed using magnetic circular X-ray dichroism (MCXD) at soft X-ray circular polarised beamline ID12B. The experimental set-up is shown schematically in the inset of Figure 66b. When the X-ray energy is swept across the spin-orbit split Co L2,3 absorption edges 2p core-electrons are excited into unoccupied 3d valence states. The spin conservation in the absorption process aligns the spin of the 2p shell with that of the magnetic 3d orbitals. Strong spin-orbit coupling in the core shell leads to different X-ray absorption spectroscopy (XAS) intensities depending on the (anti)parallel alignment of photon spin and sample magnetization. This can be clearly seen in Figure 66a, where typical XAS spectra are shown for a Co coverage of 1.6 AL. The spectra were obtained with 85% circularly polarized photons by reversing the magnetic field direction H = ±4 T. The X-ray absorption cross-section was measured by monitoring the fluorescence yield of the X-rays emitted during the decay of the 2p core hole with a photo-diode. The MCXD spectrum (Figure 66b) is the difference between the two XAS spectra and contains information about the magnetic ground-state moments that will be utilized below.

For a precise determination of the cluster size (average number of atoms/cluster) measurements were made of the sample magnetization monitored by the XAS intensity of the L3 absorption edge as a function of applied magnetic field in the superparamagnetic regime and fitted the data to a Langevin function [2]. The result for Q = 1.6 AL at a sample temperature of T = 300 K is shown in the inset of Figure 66a (symbols). From the fit to the experiment (line in the inset of Figure 66a) one obtains N = 8000 ± 300 atoms/cluster assuming the bulk value of the total magnetic moment of m = 1.7 µB [2].

If the sample is cooled down in a small magnetic field the onset of remanent magnetization is observed at the so-called blocking temperature, TB. This is when the anisotropy barrier between opposite magnetization directions can no longer be overcome by thermal spin fluctuations and a preferred spin direction is selected by the external field leading to a non-zero macroscopic sample magnetization. As shown in the inset of Figure 67a, TB can be determined from the ratio of the average remanent magnetization M(H = 0T) normalized to the saturation magnetization, MS, at high magnetic fields. A variation of TB from 224 ± 10 K to 76 ± 10 K for N = 8000 ± 300 and 800 ± 60 atoms/cluster, respectively, was found. Although these values demonstrate a reduction of the absolute height of the anisotropy barrier, the anisotropy energy per atom EA kTB/N actually increases when the cluster size is reduced (see Figure 67a).

The electronic origin of this phenomenon was studied by determining the orbital moment per Co spin r = mL/mS. Two typical spectra, normalized to constant L3 intensity are shown in Figure 66b for 8000 (dots) and 300 atoms/cluster (line). From the integrated MCXD intensities at the corresponding edges the orbital moment per spin can be obtained using sum rules [3].

The results are summarized in Figure 67b where the orbital magnetic moment per spin is shown as a function of cluster size for temperatures TB (solid symbols). While the value for the largest measured cluster size is identical to those obtained for ultra-thin Au/Co/Au(111) sandwich films [4], there is a clear increase in the values of r with decreasing cluster size. A qualitatively similar tendency has been reported for the surface layer of epitaxial Co/Cu(100) films [5] and reflects the 3d band narrowing due to the loss of atomic coordination. In the case of clusters, the perimeter Co atoms have less atomic neighbors compared to the center atoms. Interatomic 3d hybridization is then less pronounced leading to an increase of the orbital moment.

In conclusion, it has been demonstrated that there is a direct way of correlating the magnetic and electronic properties of magnetic clusters and it has been shown that there is a clear relation between the orbital magnetic moment and the anisotropy energy. There is also evidence for a reduction of 3d hybridization from the MCXD line shapes, which will be discussed elsewhere [6].

[1] B. Voigtlander et al. Phys. Rev. B 44, 10354 (1991).
[2] H. Takeshita et al. J. Magn. Magn. Mater. 165, 38 (1997).
[3] B. T. Thole et al. Phys. Rev. Lett. 68, 1943 (1992).
[4] D. Weller et al. Phys. Rev. Lett. 75, 3752 (1995).
[5] M. Tischer et al.Phys. Rev. Lett. 75, 1602 (1995).
[6] H. Durr et al. (submitted to Physical Review).

H.A. Dürr (a), S.S. Dhesi (a), E. Dudzik (a), D. Knabben (b), G. van der Laan (a), J.B. Goedkoop (c), F.U. Hillebrecht (b), accepted.

(a) Daresbury Laboratory, Warrington (United Kingdom)
(b) Institut für Angewandte Physik, Heinrich-Heine- Universität Düsseldorf (Germany)
(c) ESRF





Nanosecond-resolved X-ray magnetic circular dichroism



Time-resolved XMCD experiments have been carried out using a pump-probe approach, thanks to the ESRF single-bunch time structure associated with a pulsed magnetic field generated by a 50 µm micro-coil. Last year the feasibility of such a time-resolved XMCD experiment, with the energy-dispersive spectrometer of ID24 was reported. A diamond quarter wave plate (QWP) was used to tune the helicity of circularly polarized X-rays (report HE-004, 31/aug/97).

For thin-film amorphous systems (GdCo) measured in a perpendicular geometry, when the magnetic anisotropy is in-plane, the dynamic response of the coherent rotation is faster than the actual 2.8ns time step (Figure 68).

Improvements in the sample mounting allow the microcoil field to be applied almost along the plane of the sample (the easy axis of magnetization). In addition implementation in a cryostat permits the detection of the typical two regimes of the magnetization reversal at 100K: for zero bias nucleation of the domains (100ns) is followed by propagation of the domain walls, whereas for H = -1.04 mT the nucleation time is reduced to almost zero. With H = -0.52 mT as bias, the regime is obviously intermediate.




Time-resolved XMCD for spring-magnets

The element selectivity of XMCD was directed to the time-dependent measurements of the two-phase NdFeB spring-magnets. These new magnetic systems consist basically of hard magnetic grains (Nd2Fe14B) embedded in a magnetically soft matrix (Fe3B, -Fe). Crystallite nucleation and grain-growth are controlled in order to tune the magnetic properties of these materials whose coercivities range from 0 to 0.36T.

A relatively low-coercivity sample (Hc = 30mT) was measured in a quasi-planar configuration (30°). Nd L2 selective quasi-static hysteresis curves were obtained first (Figure 69). Magnetic pulses of 300mT height and 25ns long were then applied to the sample and various delayed XMCD spectra collected. Figure 70 shows the relaxation time of the Nd L2 magnetization, which corresponds to a exponential decay with a time constant of 37ns. Starting from different points of the major cycle enables an evaluation to be made of the mechanism of dynamic coercivity of the hard grains.



A real time XMCD approach with a fast X-ray detector in the 2/3 ESRF filling mode

A real time approach was implemented with a fast X-ray detector in the 2/3 ESRF filling mode. This was done for two reasons: i) There are not so many single bunch shifts, ii) To overcome the limitations in pulse height and length (0.7T, 20ns) due to the thermal power generated in the coil itself and in the electronics for pulse generation which go with the fast single-bunch repetition rate (357kHz).

The use of a fast X-ray detector (an avalanche photodiode - Hamamatsu C5658) instead of the CCD camera used at ID24, associated with a fast acquisition system (a digital oscilloscope - LeCroy LC394) has made it possible to measure the whole time dependence, each X-ray pulse probing the sample with an increasing delay. A 300 µm horizontal slit was used in front of the photodiode in order to select an energy bandwidth of 5eV in the Gd L3 absorption edge (7243eV).

The amorphous film of GdCo, with in-plane easy magnetization and a coercive field of about 1mT, was mounted in the micro-coil in a quasi-planar geometry (30°). To probe the magnetization reversal dynamics of the Gd, starting from the saturated magnetic state, a bias field of -1.8 mT was used, applied opposite to the field pulses (~ 60mT).

For statistical reasons, an average of 7.105 measurements were made, allowing a signal-to-noise ratio of over 0.5% in 18 minutes, with a time-resolution of 1ns in a 500 ns window. Figure 71 shows the pulsed magnetic field and the Gd L3 XMCD signal as function of time, at room temperature. After the field pulse (100ns), the relaxation of the Gd magnetization is delayed by about 25 ns needed for nucleation of the domain before the wall propagation takes place with a decay time of approximately 30ns. The insert shows the time structure of the 2/3 ESRF filling mode: there are 8 series of bunches over 2/3 of the 2.8 µs period. Other measurements were performed with various conditions of bias and pulsed field, and are consistent with the thermally-activated nature of the magnetization reversal process.

The pulse repetition rate was 1kHz, i.e. 360 times smaller than that of the single-bunch condition. Considering the thermal constraints, this should allow an increase of the possible field pulse by 19, i.e. fields up to 10T in perpendicular and 5T in parallel micro-coil geometries, with a pulse width of about 50ns. A new micro-coil current driver capable of delivering this current is now under construction.

M. Bonfim (a), S. David (a), S. Pizzini (a), K. Mackay (a), A. Fontaine (a) S. Pascarelli (b), T. Neisius (b), to be published.

(a) Laboratoire Louis Néel ­ CNRS, Grenoble (France)
(b) ESRF




Surface magnetic X-ray diffraction from cobalt ultrathin films on a Pt(110) substrate: measurement of magnetic dichroism in the crystal truncation rods

Surface diffraction measurements are normally done by setting the momentum transfer vector in a scattering experiment to lie in a region of reciprocal space away from the reciprocal lattice points of the crystalline sample. In this manner, surface sensitivity is achieved. In general, the measurements of the diffracted intensity are made along straight lines perpendicular to the surface of the sample connecting consecutive reciprocal lattice points in that direction. These are the so called Crystal Truncation Rods.

In order to extract magnetic information from the diffracted intensity of a surface, we take advantage of the resonant scattering effect which causes an enhancement of the magnetic signal when the photon energy is precisely tuned to an absorption energy where strong dipolar transitions occur.

Co/Pt systems are very interesting practically due to the large magneto-optical effect induced by the Pt caused by its relatively large spin-orbit coupling and also due to the perpendicular magnetic anisotropy in low period Co/Pt multilayers. We have investigated the interface Pt magnetism in the Co/Pt(110) system by means of resonant surface magnetic diffraction. The photon energy was tuned to the L absorption lines of Pt in order to investigate the magnetism of the Pt atoms when they are in the vicinity of the ferromagnetic Co. To isolate the magnetic signal from the "normal" diffracted intensity coming from charge scattering, flipping ratio measurements were performed: at every location in reciprocal space, the normalised intensity difference upon applying an external magnetic field in two opposite senses was evaluated. The resulting asymmetry ratio R is of the order of 10-3.

Six atomic layers of Co were grown on a Pt(110) substrate at room temperature in ultra-high-vacuum conditions in the diffraction chamber.

Figure 72 shows magnetic measurements along the platinum Crystal Truncation Rod (H, K, L) = (0, -1, L) carried out on ID3. L denotes the perpendicular momentum transfer and varies continuously from 0.2 to 1.6. H and K are reciprocal space coordinates in the plane of the sample surface. At L = 1 a Bragg reflection of the Pt crystal occurs. The filled data points are measurements of the asymmetry ratio R when the photon energy is tuned at the L3 absorption edge of the Pt. The non-zero values of R(L) are indicative of a magnetisation of the interface Pt atoms. Once the atomic structure of the interface is precisely known, one can extract from the R(L) data the magnitudes of the induced magnetic moment of the Pt atoms in different atomic layers. The unfilled data points correspond to the same measurements of R(L) the only difference being the photon energy that was tuned to the L2 absorption edge. As can be seen, the values of R have an opposite sign to the case of the L3 absorption edge. This is reminiscent of the results usually obtained in circular magnetic dichroism experiments. The bottom figure shows the sum of the above two data sets and it shows the non-zero dichroic signal along the crystal truncation rod. The sum rules of Thole and Carra [1] indicate that the normalized dichroic signal is directly proportional to the expectation value <Lz> of the orbital momentum of the magnetized atoms. Therefore the data shown in the bottom part of the figure contain direct information on the depth dependence of the angular momentum <Lz> of the magnetized Pt atoms at the interface.

[1] P. Carra , Synchrotron Radiation News , 5 (1992) 21.

S. Ferrer (a), J. Alvarez (a), E. Lundgren (a), K. Peters (a), X. Torrelles (b) and H. Isern (c), to be published.

(a) ESRF
(b) Institut Ciencia Materials, Barcelona (Spain)
(c) Instituto Ciencia Materiales, Madrid (Spain).




Circular magnetic X-ray dichroism and band structure

Magnetic circular X-ray dichroism (MXCD), i.e. the difference in absorption between right- and left-circularly polarized photons in a system with a net magnetization, was first observed by Schütz and co-workers [1] at the K edge of iron metal using synchrotron radiation. The experiment was soon followed by many other studies on different systems and compounds, ranging from 3d transition metals to actinides. Parallel with this experimental activity, theoretical efforts were made to identify the crystalline microscopic properties reflected in the observed spectra. In this context, atomic theory has provided powerful sum rules relating the MXCD signal, integrated over the finite region of an absorption threshold, to the ground-state expectation value of spin and orbital multipoles. For the important case of electric-dipole excitations, the atomic result indicated the possibility of an element-specific determination of the orbital and spin contributions to the magnetic moment, a prediction that was promptly verified.

The atomic sum rules have been tested against ab initio (electronic structure) calculations for 3d systems. As expected discrepancies have emerged, signalling the presence of corrective terms brought about by electron delocalization and demanding a more general analysis of X-ray absorption in solids. It should be noticed that central symmetry is essential to the states of an isolated atom, and orbital angular momentum eigenstates represent a very natural basis for analysis. In crystals, anisotropy spoils angular momentum conservation and a more suitable basis would transform according to the irreducible representations of the appropriate space group. This basic difference has hindered the identification of a well-defined connection between atomic results and electronic-structure calculations for magnetic X-ray dichroism.

Recently, a solution to the problem has been put forward by scientists in the ESRF Theory Group and in the Max Planck Institute, Stuttgart. Implementing the Korringa-Kohn-Rostoker formalism (or, alternatively, the linear-muffin-tin-orbital approach) for single-particle band structure calculations, an MXCD sum rule for itinerant electrons was derived.

The new formulation is best understood by considering the following specific example. Atomic theory links the MXCD signal, integrated over the two partners of a spin-orbit split core level, to the ground-state expectation value of the orbital moment of valence electrons. In the framework of band theory, the same integrated spectrum is given by an expansion in energy moments of the valence band, and sufficient accuracy is obtained by truncating the expansion to second order. The zeroth moment reproduces the atomic orbital sum rule, while the higher moments represent deviations from the atomic results and are caused by band broadening. On the basis of order-of-magnitude estimates, these deviations are expected to be small (~10-15%), thus confirming the satisfactory agreement between the atomic sum rule and experiments.

Extensions of the derivation to cover the cases of XAS, linear dichroism, and resonant X-ray scattering can be readily carried out when needed for comparison with experiment.

[1] G. Schütz, W. Wagner, W. Wilhelm, P. Kienle, R. Zeller, R. Frahm, and G. Materlik, Phys. Rev. Lett. 58, 737 (1987).

R. Benoist (a), P. Carra (a) and O.K. Andersen (b), to be published.

(a) ESRF
(b) Max-Planck Institut, Stuttgart (Germany)