In solid state physics, the «surface» is regarded as the few last layers of a solid or a liquid that can interact with the external environment. The study of «clean» surfaces is conducted in ultra-high vacuum and gives important insights on how and why the last layers of matter rearrange their atomic bonding structure to cope with the physical and macroscopic truncation of the volume. The truncation of the volume also affects the dynamics of the surfaces, bringing new interesting phenomena such as the «surface melting» that is often completely different from the standard melting of the bulk. Surfaces play a fundamental role in technology, since a huge variety of processes are inherently linked to surface phenomena: MBE growth, catalysis, corrosion and electrochemistry are just a few examples.


X-ray reflection study of surface melting



Melting of a solid is the most popular example of a phase transition. Despite its presence in many processes of everyday life, the mechanisms by which a solid transforms into a liquid are poorly known.

The observation that a body can remain liquid below its solidification temperature («undercooling») whereas the reverse, namely a body remaining solid above its melting temperature, never occurs (whatever the material or its purity) indicates that something more general triggers the melting of a body.

It has been shown that surface is the key answer to this hysteretic behaviour. It can be understood easily: atoms on a surface or close to it have less neighbours than when embedded in the bulk so that they may start to move at lower T than others. This phenomenon is called surface melting: it manifests itself by the presence of a liquid layer at the solid/vapour interface. As temperature approaches from below the melting temperature, the thickness of the molten film usually goes to infinity.

Surface melting was studied for the first time using X-rays in the 1980s (Van der Veen) through experiments on lead single crystals. The temperature dependence of the thickness of the so-called «pre-melted» film is a function of a detailed balance between short- and long-range inter-atomic forces.

To better understand the melting of a solid and the wetting physics at the solid/vapour interface, it was decided to work on a system where modelling can be better performed. Rare gases are the ideal candidates since they are compounds where interatomic potentials are best known.

In order to build up a flat surface of solid argon, several layers were adsorbed on a MgO single crystal at low temperature. Increasing the temperature close to the argon triple point (T = 83.8K), variations in thickness of both solid and liquid parts of the film (Figure 37) could be directly observed using X-ray surface diffraction and reflection (Figure 38). Data show the divergence of the molten part of the surface and allow a measurement of the associated exponents and force constant coefficients (a power law with exponent 1/3 was found). In this case, a wetting theory description accounts for both amounts of adsorbed material and repartition between solid and liquid (surface melted) parts in this Van der Waals system.

In this system also, for the first time, the initial step of surface melting was found. At a temperature of 70 K (i.e. 14 K below Tm) the overall film thickness jumped by half an atomic layer. Theoretical arguments predict that this half-layer shift is the signature for the pre-roughening transition. Complementary experiments of surface scattering and non-specular diffuse scattering also confirm the absence of a direct roughening of the surface at this temperature and support the pre-roughening scenario.

The data show that the presence of a liquid film at the surface can be detected above this temperature only, so that this pre-roughening may appear as the onset of surface melting, supporting arguments from numerical simulations which show that pre-roughening is associated with an increase of mobility of the atoms.

Synchrotron grazing incidence surface scattering techniques are ideal tools to study the evolution and phase transformation of solid surfaces close to their melting temperature. The rather high vapour pressure of many compounds at their triple point makes the use of synchrotron X-rays a quite unique probe for this purpose. The experiments on rare-gases open up wide perspectives for the study of many open issues in the study of thermal disordering of surfaces as for example:

- what is the relation between roughening and surface melting?

- how can we describe the roughness of a (molten) surface? Are capillary waves or solid-on-solid descriptions better?




F. Rieutord (a), R. Simon (a), R. Conradt (b), P. Müller-Buschbaum (c), submitted

(a) ESRF and CEA-Grenoble

(b) Univ. of Konstanz (Germany)

(c) Institut für Experimentalphysik, Univ. of Kiel (Germany)



-Al2O3(0001)-(1x1) surface structure


The structure, composition and morphology of ceramic surfaces strongly influence their chemical, mechanical and electrical properties and thus have a dominant contribution in many technologically important processes such as corrosion, catalysis and sintering. They also affect the nature and strength of bonding at metal/ceramic interfaces used in composites or in electronic packaging. The (0001) surface of a-alumina is of major importance: it is one of the most widely-used substrate for thin film growth and its initial state is known to have a dominant effect on the overlayer properties. Despite this interest, little is known about the structure and energetics of oxide surfaces. In particular, the nature of the terminating plane (Al or O) and the relaxations of the -Al2O3(0001) unreconstructed surface was still an open topic, theoretical calculations yielding different answers.

These parameters were experimentally determined for the first time by a Grazing Incidence X-ray Scattering (GIXS) study using the ID3 surface diffraction set-up. This surface of very light material could only be studied with a high brillance undulator beam such as the one of ID3. It allowed us to avoid the large background intensity coming from bulk defects by using 10 µm slits perpendicular to the surface, at a very grazing incident angle of 0.1°, enabling measurements of the very weak surface specific reflection in-between Bragg peaks.

Figure 39 shows clearly that the O termination can be ruled out: a good fit cannot be achieved with this model, even with large atomic displacements. Single and double Al terminations yield very similar good fits. However, the latter yields an outward relaxation of the outermost Al plane of +93 %, which is very unlikely. Moreover, for this model, the "bump" at l ~ -3.5 on the (10l) reflection is not reproduced.

The most likely surface is therefore terminated with a single Al plane. This agrees with most theoretical results, based on the fact that there must be no net dipole moment at the surface. This is also the only termination which leaves the surface «autocompensated», i.e. both charge neutral and chemically stable.

Regarding the surface relaxations (Figure 40), theoretical results differ depending on the method used. However they all agree with a large negative relaxation of the topmost Al layer, which is also our conclusion, the first interplanar spacing being reduced by -51 %. However, the first Al-O bond length is only 4.5 % shorter than the bulk one because the underlying oxygen atoms shift mainly parallel to the surface plane as they are repelled from the top Al atoms, moving almost radially towards the second Al sites. This almost bond-length conservative relaxation has also been theoretically predicted.




P. Guenard (a), G. Renaud (a), A. Barbier (a), M. Gautier-Soyer (b), Applications of Synchrotron Radiation to Materials Science III, Mat. Res. Soc. Proc, to be published

(a) CEA, DRFMC/SP2M, Grenoble (France)

(b) CEA, DRECAM/SRSIM, Saclay (France)



Strain relaxation in Si/Si0.7Ge0.3 quantum dot structures upon fabrication by reactive ion etching



Monolithically integrating opto-electronic devices, e.g. light emitters and detectors with Si based high speed amplifier etc. using mature Si IC technology, have been a goal in micro-electronic industry for a long time. Due to the indirect bandgaps of Si and Ge, however, it has been impossible to obtain efficient luminescence from these materials in their bulk forms. In 1969, Esaki and Tsu proposed that the electronic properties of semiconductors could be strongly modified by making an artificial one-dimensional superlattice by sequentially depositing different semiconducting elements and alloys. Since then a vast amount of work has been devoted tailoring the properties of various semiconductor systems in this way with enhanced electrical and optical properties. However, the outcome for Si and other group-IV semiconductors is disappointing and does not quite match the expectations. For example, despite theoretical predictions, no light emitting diode (LED) based on ordinary Si/SiGe superlattices exists yet.

However, positive results have been reported recently for Si/SiGe superlattices with reduced lateral dimensions, so called «quantum dots» (QDs). A QD is a semiconductor structure where the charge carriers are confined to a very small (< ~ 10,000 nm3) volume limited in all three dimensions such that quantum effects can be expected (in analogy with a three-dimensional particle-in-a-box model). In 1995, Tang et al. reported that three-dimensional arrays of Si/SiGe QDs exhibit a luminescence intensity more than two orders of magnitude higher than that obtained from the original superlattice structure. The fact that substantial luminescence persists at room temperature indicates that the process cannot be described by a quantum phenomenon only. A couple of models proposed to explain the luminescence include strain relief through lattice relaxation as an essential part of the argumentation. The strain is caused by the lattice mismatch between Si and SiGe and is created during the coherent growth of the two-dimensional superlattice. When the two-dimensional structures are processed into quantum dots or wires, the removal of part of the material may relax the strain in the structures. Figure 41 shows the structure on which experiments were performed.

Figure 42a shows an overview low resolution reciprocal space map around the 224 peaks for the as-deposited superlattice and Figure 42b shows the corresponding map from the etched QD sample (data were collected on the Materials Science beamline, ID11). The structural parameters obtained by analyses of this type of map from the as-deposited sample are dramatically different from the quantum dot sample, in that all peaks are flanked by satellites in the qll-direction and the peaks related to the superlattice have all shifted towards lower qll-values values and higher q-values. The shifts show that the average lattice parameters in the [110]- and [0011]-directions have increased and decreased, respectively (Figure 43). This is the first direct evidence that strain relief indeed occurred through lattice relaxation in the small superlattice pillars upon reactive ion etching.

The results give support for recombination models based on biaxially strained layers.




J. Birch (a), W.-X. Ni (a), Y.S. Tang (b), K.B. Joelsson (a), C. Sotomayor-Torres (b), Å. Kvick (c) and G.V. Hansson (a)

(a) Department of Physics, Linköping Univ. (Sweden)

(b) Nanoelectronics Res. Centre, Univ. of Glasgow (UK)

(c) ESRF




In-situ XAFS study of the CO oxidation over platinum catalysts - breathing platinum clusters



ID24, the energy-dispersive X-ray absorption spectroscopy beamline, has become operational in February 1996. The whole spectral range of the X-ray absorption fine structure (XAFS) at K or L edges is acquired simultaneously with a CCD camera, which allows to take spectra with a frequency of 0.1 kHz. A mode is currently tested to go up to 10 kHz spectrum acquisition frequency for special applications (Figure 44).

These characteristics make this new tool especially adapted for time-resolving studies of chemical reactions, where metal atoms are directly involved in the process such as in the catalysis example which follows.

The oxidation of carbon monoxide over platinum has been studied thoroughly in the past due to the intriguing kinetics of the reaction. A high and a low reaction rate branch as well as self-sustained chemical oscillations have been observed on heterogeneous catalysts for the first time in 1972. Since then many research groups have investigated the system in order to determine the origin of the oscillations as recently reviewed by R. Imbihl and G. Ertl.

The CO molecule adsorbs on a single site on the Pt surface, whereas two free sites are necessary for dissociative oxygen adsorption. The low reaction rate branch is observed at high CO partial pressure, with adsorbed CO molecules poisoning the catalyst. Self- sustained chemical oscillations have been predicted with these three basic reaction steps alone, but the slow rate oscillations which are observed experimentally require an additional buffer step. This step could be a reversible surface phase transformation or a reduction-oxidation of the metal clusters.

In order to elucidate the structural variations during the chemical oscillations, a Pt zeolite catalyst has been studied using in-situ XAFS spectroscopy during the CO oxidation. It has been one of the first experiments performed on ID24. Due to the relatively slow variations of the reaction mechanisms studied here, single spectra have been acquired every 10 seconds. The Pt catalyst sample has been installed in a flow-through cell operated at atmospheric pressure. The gas flow is remotely controlled. Product gases are measured with a residual gas analyser.

Four different reaction states (T = 115 °C) have been set and can be clearly distinguished in the intensity of the white line at the Pt L3 edge (Figure 45). The variation as a function of time is shown in Figure 46. The intensity of the white line (upper curve) and a shift of the distance of the nearest neighbour distance (lower curve) is observed. A short distance corresponds to predominantly oxygen backscattering.

The initial state (step 1) is characterised by an excess of CO partial pressure in the air/CO gas mixture.

The reaction is in the low rate branch, inhibited by a strong adsorption of CO molecules on the metal surface. Then the CO flow has been stopped (step 2), followed by a strong increase of the white line intensity and a shift of the first peak to smaller distances. Both signals indicate a strong oxidation of the metal cluster. A CO flow leading to a CO partial pressure of 3.8 % started the catalytic CO oxidation (step 3). A small reduction of the white line intensity indicates the adsorption of CO molecules on the Pt cluster, which remains nevertheless oxidised. The instability of the reaction rate manifests itself in fluctuations of the white line intensity. A slight increase in CO partial pressure to 4.2 % (step 4) poisons the catalyst and leads to a strong reduction of the intensity of the white line and to a shift of the maximum to longer distances. The poisoned Pt clusters are in the reduced state.

Self-sustained chemical oscillations have been observed over several hours under steady state gas flow. The temperature variations did not exceed 2 °C which indicates that the rate oscillations are controlled by the reaction kinetics only. Similar processes have been observed on facetted Pt(110) surfaces using ultra-high vacuum techniques. But the measurements presented here open the possibility to study a wide variety of processes on real catalysts.

On the dispersive EXAFS beamline (ID24)




M. Hagelstein (a), T. Ressler (b), U. Hatje (b), H. Förster (b), to be published

(a) ESRF

(b) Institut für Physikalische Chemie, Univ. of Hamburg (Germany)