Third generation synchrotron sources allow a very wide range of investigations on the properties of condensed matter. X-ray elastic scattering is used at the ESRF to investigate not only the structural properties of new materials, but also to elucidate the nature of exotic phases, such as the charge density wave (CDW) state or the Spin-Peierls (SP) phases of quasi one-dimensional systems, or the orbitally ordered state in transition metal oxides.

Surface diffraction, performed by exploiting the grazing incidence geometry to enhance surface sensitivity, has been used to investigate the ordering transition of Pb deposited on a Ge(111) surface. Inelastic X-ray scattering is a versatile technique that is performed in rather different modes on different beamlines. Compton scattering has been used to investigate electronic properties of organic charge-transfer solids as well as of ice; ultra-high energy resolution scattering provides important information on the dynamics of systems as diverse as polymer glasses and dense fluids; resonant scattering is a powerful tool to disentangle the electronic properties of correlated electron systems such as mixed-valent Ce compounds. Of particular interest is the investigation of Eu compounds by resonant nuclear scattering on the 151Eu isotope, which opens up a new way to investigate a rich variety of electronic and magnetic phase transitions.




Superconductivity in alkaline earth fullerides


The interdisciplinary research on the solid state properties of the fullerenes and their derivatives that started with the isolation of solid C60 from arc-processed graphite continues unabated. The focus of solid state and materials science research has gradually shifted away from pristine samples towards fullerene derivatives which display magnetic and superconducting properties. Intercalation of solid C60 with electron donors has resulted in a wealth of fulleride salts. Prominent among these have been the alkali fullerides with stoichiometry A3C60 (A = alkali metal) which are superconducting with critical temperatures, Tc, as high as 33 K (at ambient pressure), surpassed only by the high-Tc cuprates. While the properties of these salts have been exhaustively investigated by a variety of experimental techniques, little is currently known from experiment about the analogous properties of alkaline-earth fullerides, AExC60, principally because of the difficulties associated with the preparation of phase-pure crystalline samples.

Recent advances at our laboratories have allowed us to synthesize single-phase alkaline earth fullerides with stoichiometries, AE3C60, AE4C60, and AE6C60 (AE = Ba, Sr) and initiate a systematic study of their structural, conducting and electronic properties. In brief, the AE3C60 (like AE6C60) salts are insulating, consistent with the doubly positive charge of the alkaline earth ions which results in the band derived from the t1u (LUMO) levels being fully occupied. Increasing the doping level then necessitates the population of t1g (LUMO+1)-derived band. Surprisingly, however, the body-centred cubic Sr6C60 and Ba6C60 salts (space group Im-3) are metallic (but not superconducting) with measured densities-of-states at the Fermi level, N(F), which decrease with increasing interfullerene separation, a trend opposite to that documented in metallic alkali fullerides. The origin of this behavior can be traced to the strong hybridisation between the d and s orbitals of the alkaline earths and the p orbitals of C60.

Superconductivity is encountered for alkaline-earth fullerides at the unusual doping level, x = 4. While the alkali fullerides with stoichiometry A4C60 are insulating, the Sr4C60 and Ba4C60 salts are superconducting with Tc = 4.4 and 6.7 K, respectively. In order to determine the crystal structure of these novel systems, we collected high-resolution powder diffraction data ( = 0.84884 Å) for a high-quality Ba4C60 sample on the BM16 beamline at ambient temperature. Rietveld refinement of the diffraction profile which was of excellent quality (Figure 36) revealed that the crystal structure of Ba4C60 is orthorhombic (space group Immm) with cell dimensions: a = 11.6101(2), b = 11.2349(2) and c = 10.8830(2) Å, making it the first known low-symmetry non-cubic fulleride superconductor. Apparently, low-symmetry distortions are not detrimental to the occurrence of superconductivity in alkaline-earth fullerides in which the conduction band is derived from the t1g levels. This is in sharp contrast to the situation universally encountered up to now for superconducting alkali fullerides. The crystal structure is shown as a projection down the orthorhombic c-axis in Figure 37. The fulleride anions are orientationally ordered in such a way that the hexagon-hexagon C-C fusions lie parallel to the orthorhombic b-axis and the Ba2+ cations occupy two distinct symmetry-inequivalent crystallographic sites.

The magnitude of the orthorhombic distortion and the orientational order of the fulleride units is sensitively controlled by the Ba2+-C608- close contacts, as the C608-ions prefer to expose towards the Ba2+ ions that part of the fullerene surface with the largest area ­ namely the hexagonal faces ­ leading to a highly anisotropic solid. This was revealed by complementary high-resolution diffraction experiments ( = 0.79930 Å) performed on the same sample on the BM1 beamline at temperatures of 5 and 10 K. While there is no change in symmetry on cooling, the structure contracts in a strongly anisotropic way, so that the Ba2+ ions continue to nest within the hexagonal faces of the fulleride units and repulsive forces with the pentagonal faces are minimised. The orthorhombic unit cell dimensions at 10 K are: a = 11.5997(5), b = 11.1827(5) and c = 10.8828(4) Å, remarkably showing no measurable contraction along the shortest orthorhombic c-axis. The structural results summarised here pose a stringent test for any model of the electronic and conducting properties of these novel alkaline earth fulleride superconductors.


C. M. Brown (a, b), S. Taga (c), B. Gogia (d), K. Kordatos (a), K. Prassides (a), Y. Iwasa (c), K. Tanigaki (d), A. N. Fitch (e) and P. Pattison (f), manuscript in preparation.

(a) School of Chemistry, Physics and Environmental Studies, University of Sussex (UK)
(b) ILL, Grenoble (France)
(c) Japan Advanced Institute of Science and Technology, Tatsunokuchi (Japan)
(d) NEC Fundamental Research Laboratories, Tsukuba (Japan)
(e) ESRF
(f) University of Lausanne (Switzerland)




Electronic structure of the organic conductor TTF-TCNQ

TTF-TCNQ is an organic conductor, an example of a class of materials with interesting physical properties, ranging from metal/insulator transitions to superconductivity. Most of these compounds have complex crystal structures. For example, TTF-TCNQ, an archetypal and widely investigated member of the class contains 68 atoms per unit cell, forming linear chains of the TTF and TCNQ molecules, with a total of 416 electrons per unit cell. It is therefore quite difficult to investigate its electronic structure in detail and, as a consequence, little experimental information is available on the electronic wavefunctions. These wavefunctions are usually determined indirectly by comparing bond lengths, excitation energies or vibration frequencies with calculations. Compton profiles provide a direct way to measure the electronic wavefunctions in momentum space, making quantitative comparisons with calculated the electronic structure possible.

Figure 38 shows the difference between the two measured profiles (squares) in different crystalline directions (anisotropy) compared with two complementary theories. The first (green line) is a molecular orbital (MO) calculation performed with the Gaussian 94 package within the Hartree-Fock approximation. The second (red line) is an ab-initio pseudo-potential LDA band structure calculation. The general agreement is satisfactory, all the observed features being reproduced by the two models. Small differences are nevertheless noticed: the position of the peaks and dips are more accurately reproduced by the ab-initio calculation. This indicates relaxation effects of the molecular orbitals in the crystalline state. The slight discrepancies (less than 1% at the peak of the profile) in the amplitudes of the features are ascribed to the approximations introduced in the descriptions of the electron-electron correlation.

A drastic dependence of the MO result with the different wavefunctions available in the widely-used Gaussian 94 package has been noticed. Figure 38 shows, the result obtained with the 6-31G** basis set; it gives the best agreement among those tested. Other basis sets, like STO-3G for example, give results which compare poorly with the measured Compton profiles and have to be used with some caution. This work opens a new way to investigate the electronic structure of complex compounds like organic metals thanks to the information provided by Compton profiles concerning electronic wavefunctions in momentum space. It has been found that ab-initio LDA calculations produce a precise description of TTF-TCNQ. The molecular orbital approach misses some crystal effects but its physical transparency and its low computational cost are valuable qualities. It requires nevertheless a careful choice of the analytical wavefunctions from among those routinely used.

S. Ishibashi (a,b), A.A. Manuel (a), D. Vasumathi (a), A. Shukla (c), P. Suortti (c) M. Kohyama (d) and K. Bechgaard (e), submitted to Phys. Rev. Lett.

(a) University of Geneva (Switzerland)
(b) Electrotechnical Laboratory (Japan)
(c) ESRF
(d) Osaka National Research Institute, Japan
(e) Riso National Laboratory, Denmark


Charge fluctuations and magnetism studied with nuclear forward scattering at the 151Eu resonance

Most of the nuclear resonance scattering experiments were performed up to now at the 14.4 keV resonance of 57Fe. One of the most challenging nuclei is 151Eu ­ with a resonance energy of 21.5417 keV ­ observed in nuclear forward scattering (NFS) at the ESRF in 1995 [1]. The numerous spectroscopic applications of this resonance benefit from the two valence states of europium, Eu2+ ion with the magnetic 4f7(8S7/2) and Eu3+ ion with the non-magnetic 4f6(7F0) configurations of the 4f shell. The Eu valence states, valence transitions and the phenomenon of valence fluctuations, as well as magnetic properties, can be easily monitored with the 151Eu resonance.

This article reports on 151Eu NFS studies of valence transitions and magnetism including first applications at high pressure carried out at ID18. NFS is a local probe which is sensitive to the splitting or/and shift of nuclear energy levels (hyperfine interaction) that are caused by the interaction of the nuclear moments with electric charge distributions and/or magnetic moments of the sample under investigation. The pulsed SR with its excellent time structure coherently excites the nuclei. In the case of hyperfine interactions the de-excitation amplitudes interfere, giving rise to oscillations in the scattered intensity with time (quantum beats). The experimental set-up for NFS at the 151Eu resonance was introduced in an earlier publication [1]. The high resolution monochromator, two Si channel cut crystals - Si(8 0 0)/Si(12 4 4) - in nested configuration, delivered an energy band of 7 meV width, and a fast detector system served a stack of four avalanche photo diodes with ~1.4ns time resolution.

Charge fluctuations: The valence states in Eu systems determine the isomer shift, which is quite large between Eu2+ and Eu3+. The isomer shift shows up as quantum beats in NFS studies, if both charge states coexist in the same sample [1] or if a reference sample with a well-defined isomer shift is measured together with the sample under investigation. Here we apply this technique for studying a pressure-induced valence transition. A specially designed high-pressure cell with B4C anvils was used with a sample diameter of about 1 mm. EuNi2Ge2, an intermetallic compound with divalent Eu at ambient pressure [2], was studied at pressures up to 10 GPa (cf. Figure 39). EuF3 (left column) and EuS (right column) were used as isomer shift references. The spectra around 5 GPa indicate ­ from a strong variation of the spectral shape ­ a mixed-valent behavior with a distribution of valence states. At higher pressures the trivalent state is approached. Note that spectra at 5.4 GPa were taken with both references, demonstrating that large differences in isomer shifts can be better determined than small shifts. The isomer shift diagrams in the middle of Figure 39 illustrate how the beat frequencies correspond to the isomer shifts between EuNi2Ge2 and the EuF3 and EuS references, a large energy difference corresponds to a fast beating. The valence transition, being already completed at 10 GPa, is depicted in the isomer-shift vs. pressure diagram at the bottom of Figure 39.

EuNi2P2 is a homogeneous intermediate valent compound, in which electrons fluctuate (rate ~1015 s-1) between a localized 4f level and the conduction band. Figure 40d shows the temperature behavior of the isomer shift when measured against the divalent EuS reference. The observed beating frequencies indicate that Eu in EuNi2P2 is in an intermediate valence state with a weak temperature dependence as visualized by the dashed line in Figure 40. The shift of the quantum beats to shorter times with decreasing temperature ­ equivalent to a larger isomer shift relative to EuS ­ is consistent with a valence shift closer to the Eu3+ state [3].

Eu3S4 is ­ at low temperatures ­ an example of an inhomogeneous mixed valence compound, where the Eu2+ and Eu3+ states coexist (intensity ratio 1:2) [4] leading to a pronounced quantum beat pattern even without the reference sample (Figure 40g). At temperatures above ~150 K electron hopping between the Eu2+ and Eu3+ states sets in drastically changing the NFS spectra (cf. Figure 40e, f). The time scale of these hopping processes is comparable to the time window accessible by NFS at the 151Eu resonance (lifetime 14.1 ns). Note the strong acceleration of the nuclear decay, which is well reproduced by simulations assuming temperature-dependent hopping rates between the two charge states as denoted in Figure 40e, f. The "smearing'' of the beats at 4.2 K (Figure 40h) is probably due to the onset of magnetic ordering.

Magnetism: The Eu(II)-chalcogenides are considered as model compounds for pure spin-magnetism. Within the EuX (X = O, S, Se, Te) series, ferro- or antiferromagnetic ordering, as well as pressure-induced structural phase transitions from the NaCl to the CsCl phase occur.

EuS, which is ferromagnetic below TC = 16.5 K in the NaCl structure, was measured at room temperature (Figure 41a) and at 4.2 K in zero field (Figure 41b) as well as with an external magnetic field applied parallel to the synchrotron radiation beam (Figure 41c). The fits clearly reveal an unsplit resonance line at room temperature (no magnetic ordering), at 4.2 K a ferromagnetic hyperfine field of 31.17(6) T with random orientation of magnetization (without external field) and ­ with external magnetic field ­ a complete orientation of the magnetic moments corresponding to an orientation of the hyperfine fields antiparallel to the external field.

In addition studies were carried out on EuTe ­ being antiferromagnetically ordered in the NaCl structure (TN = 10.6 K) at ambient pressure ­ at pressures up to 20 GPa in a diamond-anvil cell. At 20 GPa the NaCl phase is completely converted to the CsCl structure. Preliminary NFS studies at 18 GPa and variable temperatures point to a (probably ferromagnetic) ordering temperature of 42(5) K in the CsCl phase.

[1] O. Leupold, J. Pollmann, E. Gerdau, H.D. Rüter, G. Faigel, M. Tegze, G. Bortel, R. Rüffer, A.I. Chumakov, A.Q.R. Baron. Europhysics Letters 35, 671 (1996) (and references therein).
[2] H.-J. Hesse, R. Lübbers, M. Winzenick, H.W. Neuling, G. Wortmann, J. Alloys Compounds 246, 220 (1997).
[3] R. Nagarajan, G.K. Shenoy, L.C. Gupta, E.V. Sampathkumaran, J. Magn. Magn. Mat. 47&48, 443 (1985).
[4] O. Berkooz, M. Malamud, S. Shtrikman, Solid State Comm. 6, 185 (1968).

O. Leupold (a, *), E. Gerdau (a), M. Gerken (a), H.D. Rüter (a), R. Lübbers (b), M. Pleines (b), M. Strecker (b), G. Wortmann (b), M.M. Abd-Elmeguid (c), H. Winkelmann (c), J. Plessel (c), H. Micklitz (c), A.I. Chumakov (d), J. Metge (d), R. Rüffer (d), G. Faigel (e), M. Tegze (e), to be published.

(a) II. Institut für Experimentalphysik, Universität Hamburg (Germany)
(b) Fachbereich Physik, Universität Paderborn (Germany)
(c) II. Physikalisches Institut der Universität zu Köln (Germany)
(d) ESRF
(e) Research Institute for Solid State Physics, Budapest (Hungary)
* present address: ESRF, Grenoble




Deformation profile in the sliding charge density wave state of NbSe3

Phase slippage is a general phenomenon in condensed matter systems with complex order parameters. When external forces impose different order parameter phase velocities 1 and 2 in two regions 1 and 2 of the same system, the phase conflict at the boundary between the two regions is released by the formation of vortices at a rate given by = 1 ­ 2. Phase slippage has been intensively studied in narrow superconducting channels, in superfluids and more recently in quasi one-dimensional conductors with a charge density wave (CDW) ground state.

In the latter case, below the Peierls transition temperature Tp, the system is driven into a modulated state of the conduction electron density:

(x) = o [1 + cos (Qox + )]

accompanied by a periodic lattice distortion of the same wavelength, 2/Qo, where Qo = 2kF is the (generally incommensurate) modulation wave vector (kF = electronic Fermi momentum). The new periodicity opens a Fermi surface gap in the electron density of states and leads to the appearence of new satellite Bragg reflections. Pinning of the local CDW phase f by defects or impurities breaks the translation invariance of the CDW ground state. However, when an electric field E higher than a threshold value ET = RIT (R: the ohmic resistance) is applied, the CDW is set in motion. This "sliding" motion gives rise to collective electron transport, a mechanism originally proposed by Fröhlich (1954), as an early (and unsuccessful) attempt to account for the phenomenon of superconductivity.

Phase slippage is required at the current electrodes for the conversion from free to condensed carriers: CDW wave fronts must be created near one electrode and destroyed near the other (Figure 42). This process is mediated by phase-dislocation loops which climb to the sample surface, each dislocation loop allowing the CDW to progress by one wavelength.

We have performed high resolution X-ray scattering measurements of the variation q(x) of the CDW wave vector Q(x) = Qo + q(x) along a thin NbSe3 whisker of cross-section 10 µm x 2 µm and 4.1 mm length between the electrodes. There are several challenging aspects to this type of experiment arising from the combined requirement of a) micron-size sample cross-sections in order to optimize the crystallographic quality and homogeneity of the sample as well as to minimize ohmic heating under dc current ; b) high spatial resolution achieved by reducing the lateral dimension of the X-ray beam down to 100 µm (30 µm in the near-contact regions); c) high momentum-space resolution, compatible with the small amplitude of the observed shifts (typically a few 10-4 Å-1).

Recent measurements on beamline ID10A (TROIKA I) have succeeded in monitoring the CDW deformation profile in the sliding state of NbSe3. The satellite shift, q, measured 100 µm away from the injection contact sets in at the threshold current value IT and increases rapidly for applied dc currents I > IT, reaching a saturation value of 8.5 10-4 b* for I > 2 IT (Figure 43).

For a fixed dc current value |I|/IT = 2.1, the shift

q± = Q(+I) - Q(-I) between satellite positions measured with positive and negative current polarities, vanishes below the electrode and rises abruptly to a maximum value at the electrode boundary (Figure 44). With increasing distance x from the contact, q± (x) decays exponentially. For x > 0.5 mm a cross-over to a linear decrease is observed with q± (x) = 0 at the midpoint between electrodes (x = 2 mm ; beyond x > 2 mm the sign of q± (x) is reversed).

The CDW deformation profile mirrors the variation of local free carrier density along the sample length. The profile shown in Figure 44 above, including the two deformation regimes, is in excellent agreement with the prediction of a recent semi-microscopic model describing the normal to condensed carrier conversion process via nucleation and growth of phase-dislocation loops.

Preliminary pulsed current measurements of the shift q have revealed important differences between pulsed and dc current data, suggesting a spatially-dependent relaxational behavior of the CDW deformations. Further studies along these lines are in progress.


H. Requardt (a,b), F. Ya Nad (a,c), P. Monceau (a), R. Currat (b), J.L. Lorenzo (d), S. Brazovski (b,e,f,g), N. Kirova (g), G. Grübel (f), C. Vettier (f), Phys. Rev. Lett. 80, (1998), 5631.

(a) Centre de Recherches sur les Très Basses Températures, CNRS, Grenoble (France)
(b) Institut Laue-Langevin, Grenoble (France)
(c) Institute of Radio-Engineering and Electronics, Moscow (Russia)
(d) Laboratoire de Cristallographie, CNRS, Grenoble (France)
(e) Institut Landau, Moscow (Russia)
(f) ESRF
(g) Los Alamos National Laboratory (USA)




Origin of the 3x3R30° <=> 3 x 3 phase transition in Pb/Ge(111)


The investigation of phase transitions in low-dimensional systems, such as surfaces and interfaces, provides insights into several fundamental fields of condensed matter physics. This is due to the wide variety of distinct mechanisms which may act as the driving force behind the transition. They include electron-phonon coupling, electron-electron interactions, entropy and disorder effects, and thus they are at the cutting edge of research in condensed matter physics.

A one-third of monolayer of Pb atoms deposited on the Ge(111) surface at room temperature form a well-ordered surface structure denoted as 3x3R30°. This structure undergoes a temperature-driven phase transition (starting at approx. 200 K) to a 3 x 3 low-temperature phase. Carpinelli and co-workers [1] were the first to report this phase transition. Their STM results showed that in the 3 x 3 phase, an asymmetry is produced in one out of three surface atoms. They interpreted the 3 x 3 phase as the observation of the stabilization of a surface charge-density-wave (CDW), the two-dimensional equivalent to better-known three-dimensional CDW's. However, the origin of the asymmetry (structural, electronic) was not clarified from the experimental point of view. This is crucial in order to be able to establish a model for the phase transition, which later on was also observed in the parent system Sn/Ge(111). It is also critical if one is to perform reliable theoretical calculations which can shed light on the nature of the phenomenon.

Since then, a lot of experimental work has been done on the system, because the possibility of observing a two-dimensional CDW is extremely attractive, both from the experimental and the theoretical point of view. However, a precise knowledge of the structural parameters of both the 3 x 3 and 3x3R30° remains the key information needed to understand the Physics involved in the phase transition. To this end, surface X-ray diffraction (SXRD) is certainly the technique of choice, because it provides unsurpassed accuracy of information on the atomic geometry of a surface reconstruction. In fact, 3x3R30°-Pb/Ge(111) was one of the first reconstructions to be analyzed using this technique several years ago. Nevertheless, several experimental features of the 3 x 3 phase prevent a straightforward application of the technique:

- the phase transition takes place very slowly in a broad temperature range. As-low-as-possible temperatures are required to achieve the true ground state, and they must be maintained over several hours during the measurement.

- the structural distortion related to the 3 x 3 phase is expected to be very small (in the range of a few tenths of Å at most). In addition to this, the 3 x 3 surface unit cell is very large. In order to obtain enough information, ample reciprocal space access is required.

- the 3 x 3 domain size (several unit cells only), whilst sufficient for electron-diffraction measurements, is much lower than the usual values needed for SXRD. Peaks related to the 3 x 3 superstructure are expected to be rather broad.

All these difficulties have been easily overcome at the ESRF, where the unique experimental facilities available allow measurements at liquid He temperatures with accurate control of the temperature. An outstanding access to reciprocal space provides enough information to obtain a reliable model. Finally, the high flux available on ID3 permits the observation of true peaks (see Figure 45) from the 3 x 3 phase in many different reciprocal space locations, in spite of their weakness. A selected rod from the reconstruction is shown in Figure 45 together with a representative scan showing a 3 x 3-related peak. Although the data are still being analyzed, the observation of intensity for large l values, and the rather negligible value for low l values, indicates a significant vertical distortion in the Pb layer. This crucial information is obtained only by reaching l values above approximately 1.5, which incidentally was the maximum l value achieved in a previous study [2].

[1] Carpinelli et al., Nature 381 (1996) 398.
[2] Baddorf et al., PRB 57 (1998) 4579.

A. Mascaraque (a), E.G. Michel (a), J. Avila (b), M.C. Asensio (b), J. Alvarez (c) S. Ferrer (c), to be published.

(a) Dto. Física Materia Condensada, Universidad Autónoma de Madrid, Madrid, Spain
(b) Instituto de Ciencia de Materiales de Madrid (Spain) and LURE, Orsay (France)
(c) ESRF




Resonant inelastic X-ray scattering from Ce


High-energy spectroscopic studies have been at the forefront in the elucidation of cerium's mixed valent behavior. Elemental cerium has a -phase in which the f electron is localized, but also an isotructural lower-temperature a-phase where the f electron plays some part in cohesion. The explanation given by Zachariasen and Pauling was a change in valence from 3 to 4. The Anderson impurity model has been used to refine this interpretation [1], the ground state being described as | G > = a | f1 > + b/f0 >, meaning a combination of configurations where the 4f electron is either totally localized or totally transferred to the conduction band. Variations in electron correlation stem from the fact that in Ce 4f states are spatially localized with a smaller radial extent than the 5s, 5p wave functions, but have an energy comparable to the valence states which makes them very sensitive to the local environment. The relative values of coefficients a and b depend critically on alloying, for instance, which in turn leads to interesting temperature-dependent magnetic properties in a wide variety of intermetallics [2].

In an L3 X-ray absorption experiment a 2p core electron is excited into the 5d conduction state (dipole transition); the core hole will either be screened by a localized f electron (plus the 5d conduction electron) or less effectively by more delocalized 5d-4f states. There is also the small probability of a quadrupole transition leading to a very well screened 2p54f2 final state. Analysis of the L3 edge has been in use for a decade and a half [3] to establish a purely empirical "spectroscopic valence" based on the relative intensities of f1- and f0-related features of the absorption edge. Excitation to these three final state configurations in the X-ray absorption process are intermediate states of resonant inelastic X-ray scattering (RIXS) involving the 3d 2p de-excitation channel. (Details of the first experiment of this kind performed on a rare earth by Krisch et al. may be found in Ref. 4).

Experiments performed on resonance are ideally suited to dealing with such materials as they spotlight specific aspects of the electronic structure. Being all-photon experiments they are also very much bulk-sensitive. To illustrate the potential of the method we show measurements for CeF3 taken as a "textbook example" (Figure 46).

The RIXS measurements were performed using a UHV-compatible cylindrically bent crystal spectrometer designed to take full advantage of the exceptional qualities of beamlines ID12A and ID12B over a 0.8 ­ 6 keV energy region. These qualities include not only brightness, resolution, and stability but also the possibility to focus the beam down to 20 ­ 40 µm in the vertical plane which defines the resolving power of the RIXS experiment.

The inelastically scattered photon spectrum, produced as a 3d electron fills the 2p core-hole left behind by the primary excitation, changes as a function excitation energy. A double structure situated in the region of 880 eV transfer energy (transfer energy = incident ­ scattered photon energy) marks the 3d94f2 final state involving a quadrupole transition in the primary excitation. At 887 eV we observe a very highly localized 5d state identified as such by a constant transfer energy (Raman process) both below threshold and above threshold excitation. A further peak is seen to disperse as the excitation energy is increased. It corresponds to normal fluorescence, meaning that the 3d 2p transition takes place after ionization.

Applied to Ce intermetallics, this technique is able to identify the 5d state delocalization resulting from both hybridization with neighbouring atoms and correlation effects.

[1] O. Gunnarsson and K. Schönhammer, Phys. Rev. B 48, 4315 (1983).
[2] Y. Baer and W.-D. Schneider, in Handbook on the Physics and Chemistry of Rare Earths, Vol. 10 (Elsevier, 1987).
[3] D. Wohlleben and J. Röhler, J. Appl. Phys. 55, 1904 (1984).
[4] M. Krisch et al., Phys. Rev. Letters 74, 4931 (1995).

J.-J. Gallet (a), L. Journel (a), J.-M. Mariot (a), A. Rogalev (b), J. Goulon (b), and C. F. Hague (a, c), to be published.

(a) Laboratoire de Chimie Physique (UMR 7614), Université P. et M. Curie, Paris (France)
(b) ESRF
(c) LURE, Orsay (France)




A new inorganic spin-Peierls system NaV2O5



Since the discovery of high temperature superconductivity in doped antiferromagnetic cuprate materials, there has been a renewed interest in low-dimensional quantum antiferromagnets. A linear S = 1/2 chain with antiferromagnetic interaction along the chain interacting with the three-dimensional phonons can lead to a spin-Peierls (SP) phase transition. A dimerization of the spin chain below the transition temperature TSP leads to the formation of a non-magnetic singlet ground state. The transition is called a spin-Peierls transition because it is a magnetic analog of a Peierls transition in quasi-one-dimensional conductors. This transition was initially observed only in a few organic compounds. The discovery of an SP-state in the inorganic compound CuGeO3 has renewed strong interest in this phenomenon. Recently a new inorganic spin-Peierls system -NaV2O5 has been discovered with the highest critical temperature so far known TSP ~ 34 K.

Here the results of X-ray diffraction investigations on a single crystal of a-NaV2O5 are reported. The diffraction experiment was performed with the helium cryostat mounted on the triple crystal diffractometer of the high energy beamline ID15A. An X-ray energy of 114.3 keV was used. Superlattice reflections corresponding to the cell doubling along a and b and quadrupling along c were detected below TSP ~ 34 K. Figure 47 shows the temperature variation of the integrated intensity of the (3/2, 1/2, 15/4) reflection which decreases continuously and becomes zero at TSP showing that the phase transition is of second order. In addition the relative temperature variation of the lattice parameter d/d of the 005 reflection was measured in high resolution mode. d/d shows an anomalous decrease with increasing temperature up to the phase transition and then increases normally by the phonon contribution, the continuous line fitted to the data in Figure 48. We interpret this anomalous expansion as being due to the magnetoelestic coupling. The deviation of the lattice expansion to the phonon contribution delivers a critical exponent of 2ß = 0.673(3).

Intensity comparisons of the X-ray results with neutron data of the superstructure reflections, obtained from the same sample lead to the conclusion that mainly the V atoms and less the O atoms move at the dimerization. A possible model for the low temperature structure can be obtained by "decorating" the zig-zag vanadium chains parallel to the crystallographic b axis (Figure 49). The doubling of the b axis can be easily achieved by modulating the V-V distance (short S and long L). In the high temperature phase they are equal (D) along the zig-zag chain. So if one has the sequence S-L-L-L-S-L in the low temperature phase instead of the sequence D-D-D-D-D at room temperature then the unit cell is doubled along b. If we also decorate the next parallel chains as we move along the a axis like L-S-L-L-L-S, L-L-S-L-L-L-S and L-L-L-S-L-L-L-S we get cell doubling along the a axis. Now it is easy to stack these chains in the a-b plane along c in order to get the cell quadrupling along the c axis, although there are several ways of doing this. Such a model has indeed given a reasonable fit to the low temperature X-ray diffraction data which are limited to one reciprocal layer.

[1] T. Chatterji (a,d), K.-D. Liß (b), G. J. McIntyre (a), M. Weiden (c,d), R. Hauptmann (c), C. Geibel (c,d), Solid State Communications, (1998), 108 (1): P. 23 -26.
[2] T. Chatterji (a,d), G. J. McIntyre (a), K.-D. Liß (b), "ILL Annual Report". Institut Laue-Langevin: Grenoble. p. 48-49, 1997.

(a) Institut Laue-Langevin, Grenoble (France)
(b) ESRF
(c) Technische Physik, Technische Hochschule Darmstadt (Germany)
(d) Max-Planck-Institut für CPfS, Dresden (Germany)




Pressure-induced in-glass structural transformation in the amorphous polymer poly(methyl methacrylate)



The dynamics of disordered solids shows very specific behaviors such as two-level systems, a thermal conductivity plateau and an excess of vibrational density of states, which contrast strikingly with the absence of any obvious feature in their static characteristics at the nanometer scale. Incidentally, the question of whether these dynamical anomalies are rooted in some specificity of the static structural disorder is the subject of much debat and a unifying picture relating the glassy structure to the collective dynamics still awaits settlement. Theoretical arguments, molecular dynamics simulations in both the supercooled and glassy states, and even diffraction measurements suggest that the medium-range order of disordered systems is characterized by a correlation length of a few nanometers. The physical significance of such a length-scale, and, more important, its relevance to collective excitations and their propagating nature is still unclear.

Recently, dynamic structure factor, S(Q,E), measurements, at momentum transfers Q and energies E relevant to density fluctuations approaching the inter-particle distances, have been performed on strong, intermediate and fragile glasses using the inelastic X-ray scattering (IXS) method. Most of the measurements point to the existence of acoustic-like excitations: the energy of these modes, (Q), has been found to disperse linearly with Q, while their broadening energy, (Q), increases proportionally to Q2. There exists, therefore, a value of Q, Qm, at which (Q) ~ (Q), and where it is no longer possible to describe the inelastic part of S(Q,E) in terms of propagating excitations. In these systems, a relation between Qm and the microscopic structure has not yet been established. It is possible, therefore, that Qm is a physical quantity connecting the high frequency dynamics to the mesoscopic structural properties.

Obviously, further studies aiming better to identify these issues are urgently needed. For example, a possible direction of investigation is the study of the pressure dependence of the high frequency glass dynamics at pressures that may start to modify the glass bonding network. The effect of hydrostatic pressure on the nanoscopic glassy structure and its inherent dynamics has already been investigated in various materials with different techniques. Low temperature measurements on a semi-crystalline polymer showed a considerable decrease of the excess specific heat on increasing pressure, thus suggesting a strong perturbation of the acoustic phonon pattern with pressure.

Brillouin visible light experiments on amorphous poly(methyl methacrylate), PMMA, evidenced at around 1.1 kbar an abrupt change in the shear elastic constant and in the intensity of the transverse phonon lines, thus indicating a weak glass transition within the glass. Recently, non-linear optical measurements on several amorphous polymers confirmed the existence of a marked change in the local structure at pressures in the 1.1 to 1.5 kbar range.

The present work performed at ID16 has allowed new evidence to be gathered on the possible existence of an in-glass transition in PMMA: the IXS data points towards a structural transformation that takes place over a length-scale which is larger than that of typical interparticle distances, and which reflects onto the longitudinal collective dynamics. The dynamic structure factor S(Q,E) of this polymeric glass shows an inelastic feature that disperses with Q. By increasing the hydrostatic pressure up to 4.5 kbar the energy position of this excitation linearly increases, with a change of slope around Po @ 1.6 kbar.

Concurrently, around Po, the pressure dependence of the elastic signal at Q = 2 nm-1 shows a comparatively stronger cusp, a behavior which is not observed in the higher Q region. This result is reported in Figure 50, where it is also shown that there are no characteristic features appearing in the static structure factor S(Q). These results, therefore, seem to unveil the occurrence of a pronounced structural rearrangement within the amorphous state at Po, affecting, most dramatically, the inter-particle correlation at a length-scale larger than ~1 nm. It is possible to speculate that the cusps observed in some of the parameters describing the S(Q,E) spectra of amorphous PMMA at pressures around Po arise from an in-glass structural transition. The analysis reported here indicates that this transformation affects the high-frequency collective dynamics, and the structure only down to a length-scale of ~3 nm, i.e. a length-scale much larger than the size of the monomers constituting this polymer. The investigation of these pressure effects as a function of temperature in the region around Tg, where the disordered system becomes ergodic, may provide further information on the relation between dynamics and structural non-homogeneity at the nanometer scale.

A. Mermet (a), A. Cunsolo (a), E. Duval (b), M. Krisch (a), C. Masciovecchio (a), S. Perghem (c) , G. Ruocco (d), F. Sette (a), R. Verbeni (a) and G. Viliani (c), Phys. Rev. Lett. 80, 4205 (1998).

(a) ESRF
(b) Université Lyon I, LPCML, UMR-CNRS 5620, Villeurbanne (France)
(c) Universitá di Trento and Istituto Nazionale di Fisica della Materia, Trento (Italy)
(d) Universitá di L'Aquila and Istituto Nazionale di Fisica della Materia, L'Aquila (Italy).




The dynamics of dense super-critical Neon at the transition from hydrodynamical to single particle regimes


During recent decades much effort has been devoted to the study of the dynamic structure factor, S(Q,E), of dense fluids in both liquid and gaseous phases. One of the aims has been to understand the transition from the hydrodynamic regime towards the so called kinetic one, as a function of the density, r, and of the exchanged momentum, Q. These two different regimes can be distinguished by a length-scale characteristic of the system. This is often chosen as the Enskog mean free path, lE, defined as lE = lB/g(r0), where lB is the Boltzmann mean free path, lB-1 = prr02/2, and g(r0) is the pair distribution function evaluated at the particle's radius, r0. For excitations with wavelength 2p/Q much larger than lE, the fluid appears as a continuum, and the Navier-Stokes equation can be used to derive the S(Q,E). At intermediate Q-values, QlE ~1, the breakdown of the hydrodynamic theory is expected to occur. For QlE >>1, the dynamics becomes that of a single free particle between two collisions with its neighbors. In the two limiting cases the dynamic structure factor has a well-known shape and a simple physical interpretation: i) In the small QlE limit there are three modes. These are respectively the Stokes and the anti-Stokes compression (sound) modes, which disperse linearly with a slope corresponding to the adiabatic velocity of sound, and the heat diffusion mode, which is centered at zero energy transfer, and has a width proportional to DTQ2, DT being the thermal diffusion coefficient. ii) In the high QlE limit, within the impulse approximation, the lineshape reflects the initial state momentum distribution, i.e. the Boltzmann distribution. Here, the S(Q,E) reduces to a Gaussian centered at the recoil energy

h2Q2/2M, where M is the particle mass.

The extension of a three modes description of S(Q,E) beyond the hydrodynamic limit where QlE approaches unity, has been suggested by the kinetic theory and by several molecular dynamics studies performed with both hard spheres and Lennard-Jones potentials. However, firm experimental data covering in detail the whole transition region up to Qm, the Q-value of the first maximum in the static structure factor, are not yet available in high-density fluids above the critical point, in spite of the speculative interest in having a complete overview of all dynamic processes at both macroscopic and microscopic scales in a fluid. Up to now only neutron Brillouin scattering techniques have provided the opportunity to study the dynamics in a Q region below and around Qm, where the S(Q,E) is deeply influenced by the microscopic structure. This has been done in liquid neon, and in liquid and gaseous argon.

The development of the inelastic X-ray scattering (IXS) technique with meV energy resolution opens up new opportunities to explore microscopic dynamic properties of dense fluids in Q-E regions of difficult access to neutron spectroscopy. Furthermore, the small X-ray beam size gives the opportunity to explore thermodynamic states which can be produced only in small volumes, as noble gases at temperatures well above the critical point and at densities either comparable to or higher than the one of the triple point.

In this work, using IXS and molecular dynamics (MD) simulations, the S(Q,E) lineshape of a deeply supercritical gas in the momentum transfer region bridging the hydrodynamical and single particle regimes was determined. The experiment was performed at ID16 on neon at 3 kbar and 295 K, corresponding to a particle density of 28,9 molecules/nm3 (970 kg/m3), i.e. comparable to that of liquid at ambient pressure, and to a temperature ~ eight times larger than the critical temperature. The momentum region, spanned in theexperiment, allowed to follow the evolution of the S(Q,E) from QlE = 0.2 to 4 to be followed.

As shown in Figure 51, this line-shape continuously evolves from a triplet to a complex broad feature. This shows that the transition from the hydrodynamic to a more complex behavior takes place at a Q value of ~10 nm-1, which is comparable to the inverse mean free path. Here, the line-shape becomes a featureless single peak, which, however, cannot yet be described as a single gaussian, characteristic of the single particle dynamics and expected in the Q limit. It is found, as shown in the row a) of Figure 51, that one can successfully model this lineshape extending to these Q values the three modes model derived by molecular hydrodynamics theory. This shows that at Q ~ Qm, the damping of the excitations becomes comparable to the excitation energy, and therefore the propagating character of the sound modes is lost. A broad minimum in the dispersion relation and the de Gennes narrowing at Qm has been found, and these results provide further evidence that the kinetic regime of the free particles is not reached at the highest Q value considered here. Finally, the dynamics of super critical neon gas at high density shows important differences with respect to the liquid phase in almost isocore conditions. In particular, at variance with the liquid, there is no positive dispersion of the sound velocity in the gas.

A. Cunsolo (a), G. Pratesi (b), G. Ruocco (c), M. Sampoli (d), F. Sette (a), R. Verbeni (a), F. Barocchi (d), M. Krisch (a), C. Masciovecchio (a) and M. Nardone (e), Phys. Rev. Lett. 80, 3515 (1998).

(a) ESRF
(b) Institut Laue-Langevin, Grenoble (France)
(c) Universitá di L'Aquila and Istituto Nazionale di Fisica della Materia, L'Aquila (Italy)
(d) Universitá di Firenze and Istituto Nazionale di Fisica della Materia, Firenze (Italy)
(e) Universitá di Roma III and Istituto Nazionale di Fisica della Materia, Roma (Italy).




Resonant X-ray scattering and the direct observation of orbital order


The oxides of the transition metals have puzzled condensed matter physicists for over half a century. In fact, besides magnetism, which finds many applications in modern technologies, they display other fascinating, but poorly understood properties, such as high-temperature superconductivity in the cuprates, or metal-insulator transitions (an abrupt and dramatic change in electrical conductivity as a function of temperature, as seen e.g. in V2O3). Some of them are also believed to display orbital ordering, an elusive property which until very recently has escaped direct experimental confirmation.

The origin of all these intriguing phenomena is in the partial occupation of the highly localized d-electron orbitals in the transition metal elements. One of the features of the d-electrons is their high orbital degeneracy, i.e. the fact that there are orbitals with the same energy but differing in the direction of the lobes of the electron charge density. It may happen in some compounds, that in each transition metal ion an electron has two degenerate orbitals at his disposal, and must "make a choice" of which to occupy. Since there are complex electric and magnetic interactions between electrons on adjacent ions, at sufficiently low temperature electrons on different ions do not "choose" their orbital independently of one another, but rather in a correlated way, giving rise to a definite, periodical arrangement of occupied orbitals, as schematised in the example of Figure 52.

There are good reasons to believe that this is indeed the case. In the manganites, like LaMnO3 , the presence of orbital order can be inferred from the crystallographically observable displacements of the oxygens surrounding each transition ion. In other cases, like V2O3, it is the magnetic structure that suggests the presence of orbital order, which was theoretically proposed some twenty years ago [1], but never confirmed experimentally in an unambiguous way.

Scientists in the ESRF Theory Group have addressed this problem, and concluded that indeed resonant elastic X-ray scattering should be directly coupled and especially sensitive to the orbital order parameter [2]. This is because in a resonant scattering process, the incoming photon promotes an electron from a core level to an intermediate state close to the Fermi energy, while the outgoing photon is produced in the inverse process, in which the virtually excited electron decays back to fill the core hole. For a suitable choice of the beam polarization and an optimal experimental geometry, the scattering amplitude is very sensitive to the availability of a suitable intermediate state, and therefore to the occupation state of a certain orbital.

The same conclusion was arrived at independently by a group at the Photon Factory in Japan, that published the observation of orbital order in La0.5Sr1.5MnO4, by resonant scattering measurements [3]. It is hoped that a settlement of the more controversial issue of the existence of orbital order in V2O3 may soon be possible.

[1] C. Castellani, C.R. Natoli and J. Ranninger, Phys. Rev. B 18, 4945 (1978).
[2] M. Fabrizio, M. Altarelli and M. Benfatto, Phys. Rev. Lett. 80, 3400 (1998).
[3] Y. Murakami et al., Phys. Rev. Lett. 80, 1932 (1998).

M. Fabrizio (a), M. Altarelli (b), M. Benfatto (c), Phys. Rev. Lett. 80, 3400 (1998)

(a) SISSA, Trieste (Italy)
(b) ESRF
(c) Laboratori Nazionali INFN, Frascati (Italy) and ESRF




Coherent charge transfer in the hydrogen bond in ice: a Compton scattering measurement


Hydrogen bonds play a crucial role in determining many of the distinctive properties of water and biological complexes. In particular, in ice (Figure 53), hydrogen forms two distinct types of bond with neighboring oxygen. The shorter (1.00 Å) covalent bond is a typical molecular s-bond between the oxygen and hydrogen with a binding energy of approximately 4.8 eV. It has been appreciated since Pauling that the longer (1.75 Å) so-called hydrogen bond, with a binding energy of 0.29 eV, is well-described as a classical, electrostatic interaction between &laqno;frozen» charge distributions. However, ab-initio calculations predict a small amount of charge transfer and a microscopic quantitative understanding of the hydrogen bond's remains experimentally untested and controversial. Compton scattering (inelastic X-ray scattering with high momentum transfer) is unique in giving information on the Fourier transformed electronic ground state, and in particular on the extended wave functions of electrons involved in the hydrogen bond. A Compton profile measures the electronic momentum density projected onto the scattering vector and contains information on wave function coherence in the shape of oscillations with period determined by the real-space bond-length. Bonds parallel to the scattering vector give rise to oscillations while those perpendicular to it do not influence the profile. Molecular orientations in ice are such that one hydrogen atom(bond) lies along each of the tetragonally coordinated neighboring oxygen atoms (see Figure 53). The Compton profile measured along the c-axis, which is parallel to the bonds along one of the tetrahedral axes and nearly perpendicular to the others, contains a strong signature of the hydrogen bonds. On the other hand in a Compton profile measured in the a/b plane the signature is weak because the bonds nearly parallel to this plane are aligned along three different directions. Anisotropies of Compton profiles or the differences between profiles along different cristalline directions are exceptionally sensitive to the hydrogen bond. Furthermore, anisotropies remove the contributions from the isotropic core states (e.g., O 1s). In Figure 54 is shown the experimental anisotropy (dots) obtained on ID15 and plotted as the difference between the Compton profiles with momentum transfer Q along, and perpendicular to, the hexagonal c-axis. The dominant feature is a periodic variation of intensity corresponding closely to the O-O distance (2.75 Å), indicating wave-function phase coherency between neighbouring molecules. There is a striking agreement between the data and a fully quantum mechanical bonding model for ice Ih (blue line) and striking disagreement with a purely electrostatic bonding model (red line). The only adjustable parameter in the theory is a down-scaling of the anisotropy by a factor of 40%, due to an isotropic contribution in the experimental data from polycrystalline parts of the sample. We interpret our results as providing direct evidence for coherent charge transfer in the hydrogen bond, an inherently quantum phenomenon.

E.D. Isaacs (a), A. Shukla (b), P.M. Platzman (a), D.R. Hamann (a), B. Barbiellini (a), C.Tulk (c), accepted for publication in Phy. Rev. Lett.

(a) Bell Laboratories, Lucent Technologies, NJ (USA)
(b) ESRF
(c) National Research Council of Canada, Steacie Institute for Molecular Sciences, Ottawa (Canada)