Extended X-ray Absorption Fine Structure (EXAFS)
(H. Wende, C. Sorg, K. Baberschke)




• Surface EXAFS and Near-edge EXAFS from adsorbed atoms and molecules on metallic surfaces

 
EXAFS (extended x-ray absorption fine structure spectroscopy) is a well established tool for analyzing the local structure and the different kinds of disorder (for example due to thermal vibrations) for various systems. The technique is used in the broad field of structural and chemical analysis. It may give us the opportunity to study the local structure around atoms or molecules adsorbed on metallic surfaces and may reveal the vibrational fine structure of resonance modes. Furthermore the local dynamics can be studied by means of temperature-dependent measurements. Modelling this temperature dependence with anharmonic pair potentials the adsorbate substrate bonding strengths can be characterized [Ref. 181] . Here we inspect three characteristic examples studied at BESSY. In the first one [Ref. 226] we deal with the p4g(2x2) reconstruction of C on Ni(100) and on Ni/Cu(001) films. Figure 1 presents SEXAFS oscillations for the p4g(2x2) C/Ni(100) system for normal (90^o) and grazing (20 ^o) x-ray incidence as a function of the sample temperature for the Ni single crystal substrate (solid line) and the 4 monolayer Ni film evaporated on a Cu(100) single crystal (dotted line). The sample temperatures for C on the Ni film were 30 K and 300 K. Analyzing the data it is possible to determine the distances between the C and Ni atoms and to conclude that the structural and dynamic results are the same for the Ni(100) crystal and the Ni film substrates.This shows that the Ni fct film also reconstructs in the same way as it is observed for the Ni single crystal indicating that for this system the fct distortion of the clean Ni film hardly changes the adsorption geometry for the C/Ni system. In addition the analysis of the local dynamics allows to describe the reconstruction strengths of the C/Ni system. It turns out that the reconstruction of C/Ni is stronger compared e.g. to the N/Ni system [Ref. 226].


Fig. 1

In the second example [Ref. 232] we deal with the vibrational fine structure in the N 1s  pi* resonance of the N₂ molecule physisorbed on the Cu(100) surface. In Fig. 2(a) we plot the vibrationally resolved photoabsorption spectrum in the N 1s  pi* resonance region of 0.15 ML N₂ physisorbed at T=25 K on theCu(100) surface (dots) and gas-phase spectrum calculated using the vibrational constants, FWHM, relative intensities for each vibrational peak (by C.T.Chen et al., Phys. Rev. A40, 6737 (1989)) and assuming a Gauss broadening (spectral resolution) as in our measurements (solid line). The inset shows the discrepancy in the peak position for higher vibrational levels. Fig.2(b) depicts the fitted spectrum. The dots and the solid line represent the data and the fit, respectively. The dashed curves are the individual Voigt peaks. Our results for the low N₂ coverage show that noangular dependence in the spectroscopic parameters are present indicating, consequently, an isotropic interaction of the  pi_g orbitals with the substrate. On the other hand, for the multilayer absorption there is a preferential orientation.

In a third example [Refs. 225, 240] we study the shape resonances of oriented molecules by ab initio theory and NEXAFS experiment at the C K-edge of hydrocarbon molecules. We demonstrate that the sigma-shape resonance originates from the multiple scattering of the photoelectron within the molecule. The angular dependence of the electric dipole transition in the calculations, as well as the angular dependent experiments for the oriented molecules give a good opportunity to compare both. The resonance can be assigned to a sigma* shape resonance. The multiple scattering formalism and the experiment agree well and thereby support the existence of such features in the spectra. The results of the calculated C 1s NEXAFS spectra for different orientations of the E-vector of the bottom beam with respect to the intramolecular C-C bond axis that illustrate the above ideas are presented in Fig. 3. The vertical lines represent the calculated IP values. The atomic background µ₀(E) is indicated by the yellow dotted line. For all three hydrocarbons a broad feature in the continuum regime is visible, which shows an asymmetry with a tail on the high energy side in full agreement with general considerations in textbook scattering theory. The angular dependence of this structure is in correspondence to predictions for the angular dependence of sigma* shape resonance.
 
 

Fig. 2

Fig. 3


 
• Atomic EXAFS: Evidence for photoelectron backscattering by interstitial charge densities

 
It is surprising that the backscattering of the photoelectron after core-hole excitation at interstitial charge densities predicted in 1978 by Holland et al. was ignored over decades. In the latter work it was shown that the atomic x-ray absorption coefficient µ₀(E) of an embedded atom can also exhibit an oscillatory fine structure chi_e(E) due to scattering at an interstitial charge between the atoms - the atomic EXAFS (AXAFS). In Fig. 4 we plot the Fourier transform (the imaginary part in the inset) of the experimental data for the ( sqrt(2)x2 sqrt(2))R45^o O/Cu(100) system measured at T=35 K (black line) and the (2x3)N/Cu(100) (sample temperature 300 K, red line). Note that the AXAFS features at distances below the nearest neighbor distance show up in the experimental data because a stiff spline function was used to extract the oscillations from the experimental µ(E). In comparison to the reconstructed (2x3)N/Cu(100) system, the AXAFS peak is located at a 0.2Å higher distance in the Fourier transform of the experimental data for the O/Cu(100) system. The blue line represents the Fourier transform of the calculated AXAFS oscillations of µ ₀(E) only. Therefore, all the other peaks due to scattering at neighbor atoms do not show up in the calculation. The comparison to ab initio calculations allows us to address this effect to a higher charge transfer from the Cu atoms to the O atom in the O/Cu(100) system [Ref. 218] .
 
 
Fig. 4

A systematic analysis of the embedded atom EXAFS (AXAFS) effect is presented in Figs.5 and 6. This effect is explained by the backscattering of the photoelectron at interstitial charge densities. The reconstructed (2x1)O/Cu(110) system is an ideal prototype system in which to study the angular dependence of the AXAFS because oxygen-copper rows are formed resulting in a C₂ symmetry. The scattering potential is non-spherical because of the high directionality of the O-Cu bonds. The high signal-to-noise ratio of the experimental data enables us to clearly identify the AXAFS contribution. Here we present the definite angular dependence of the experimental AXAFS for the first time [Ref. 279], giving a unique opportunity to measure the anisotropy of the local embedded atom potential. The angular dependence demonstrates that the AXAFS effect cannot be mimicked by multi-electron excitations or experimental artefacts for this system. We compare our experiments to theoretical calculations within the muffin-.tin approximation and show that future full-potential calculations are needed to model the angular dependence determined.

 

Fig. 5

Fig. 6

• Magnetic EXAFS: A probe of the local spin-pair distribution function

 
Magnetic EXAFS (MEXAFS) is the spin-dependent counterpart of EXAFS. Recently, the temperature dependence of MEXAFS at the L edges of Fe and Gd received some attention in order to separate the magnetic from the structural static disorder. In case of Fe we studied up to 0.3 in a reduced temperature scale (t=T/T _C) [Ref. 214] while the much smaller T_C of a bulk Gd single crystal compared to Fe allowed us to reach t=0.85 [Refs.236, 246] .

Fig. 7

In Fig. 7 one may see the x-ray absorption coefficient at the Gd L₃ , L₂ and L₁ edges for right µ ⁺(E) and left µ^-(E) circularly polarized x-rays (top) at T=10 K. The magnetic signal is calculated from the difference µ _M(E) = µ⁺(E) - µ^-(E) (bottom). The data were recorded at the ID12A beamline at the E.S.R.F. using the gap scan technique. As seen in Fig. 7 the EXAFS oscillations are of about 4% of the edge jump, while MEXAFS is about a factor of 10 less. This shows that the data have to be recorded with an excellent signal-to-noise ratio. In Fig. 8 we plot together the experimental EXAFS (top) and MEXAFS (bottom) for the Gd single crystal at T=10 K compared to ab initio calculations using the FEFF7 code. The calculations show good agreement for the enveloping amplitude and phase in k-space as well as in R-space for the splitting of the main peak (RT resonance) and the peaks at larger distances. In Fig. 9 ab initio calculations of the MEXAFS for the bcc structure (Fe) are plotted with the help of the FEFF7 code. The single scattering contributions (dotted line) have been separated from the combined multiple and single scattering contributions (solid line). The peaks are assigned to the different scattering paths which are labeled according to the inset. The separation of the paths allows for the unambiguous detection of strong multiple scattering contributions for the MEXAFS which are not present for the normal EXAFS (see also [Ref. 192] ). Temperature-dependent measurements for both Fe and Gd [Refs. 214, 236, 246] have revealed a more pronounced decay for the MEXAFS than the regular EXAFS and this effect was attributed to contributions originating from the local spin disorder.
 
 

Fig. 8

Fig. 9