Analysis of angle-resolved electron energy loss in XPS spectra of Ag, Au, Co, Cu, Fe and Si
Introduction
Photoemission and in particular X-ray photoelectron spectroscopy (XPS) has become an important technique for the investigation of surfaces and interfaces. In XPS as well as in other electron spectroscopies, the final shape of the measured energy spectrum is highly influenced by the inelastic scattering events taking place during electron transport out of the solid. Quantitative investigations using XPS/AES are usually focused on determining the chemical composition and morphology of nanostructures in the surface region. To establish this information, it is required that the data analysis procedure contains an appropriate and realistic correction for the inelastic electron scattering effects. That is, a good procedure for background subtraction is required.
A photoelectron loses energy when the charge density, induced in the medium by the moving electron and the stationary core hole, interacts with the charge of the electron. The interaction is conveniently described in terms of the inelastic scattering cross-section K(T) giving the probability density per unit path length of an energy loss T. In photoemission, one distinguishes between intrinsic and extrinsic energy losses, both contributing to the measured peak shape. Intrinsic excitations result from the sudden initial creation of the photoelectron and core hole, while the extrinsic losses are excited during the subsequent electron transport from the point of origin to the analyzer. In addition, the interaction of the electron with the induced field in the surface region is different from the bulk case. When excitations in the surface region become important, surface losses will appear at a reduced energy. In free-electron-like materials with a narrow plasmon structure, the effect of the surface on K(T) is to subtract intensity from the bulk plasmon at ωp and to add intensity at the surface plasmon at [1]. However, in transition metals with a broader loss structure, the effect is smeared out [2], [3].
In the past, various models have been proposed to account for the background of inelastically scattered electrons in photoemission. Among these, the straight-line and the Shirley [4] methods have been widely used [5]. A method involving a physical description of the extrinsic energy loss was developed by Tougaard and co-workers [6], [7]. The method treats the background in terms of the multiple inelastic scattering events taking place during electron transport. It has been found to be superior to the straight-line and Shirley methods in a systematic study of metals and alloys [8].
The accuracy of the Tougaard method relies on an accurate description of the inelastic scattering properties of the medium, given by K(T) and on an accurate value of the inelastic mean free path (IMFP) λ. Surface loss effects will be subtracted to the extent that they are represented in K(T).
One of the simplest cases of energy loss is for stationary transport in an infinite medium. In this case the bulk cross-section is given as [9]:where a0=ℏ2/me2 is the Bohr radius, E0 is the kinetic energy and ε(k, ω) is the complex dielectric function dependent on momentum and energy transfer. The validity of this expression in the case of reflection electron energy loss (REELS) experiments has previously been investigated [3].
Detailed dielectric response models for K have recently been developed, describing electrons traveling in general XPS and REELS geometries [2], [10], [11]. The models determined inelastic cross-sections from the trajectory of the electron in the surface region, and from the dielectric function of the medium. The REELS cross-sections showed good agreement with experimentally deconvoluted cross-sections for varying incidence and exit angles and for varying energy. Generated XPS model spectra showed good agreement with experimental spectra in the low energy loss region. These expressions, however, are rather complex, and for practical analysis more simple approximations are desirable.
As a simple tool in background subtraction and peak shape analysis of metals, the inelastic cross-section is frequently described by a universal function [12], [13]. This is because, for noble and transition metals, λ(E)K(E, T) depends only weakly on E and on which element is considered. It is approximated by the formula:where B=2866 eV2 and C=1643 eV2 were determined by fitting to cross-sections calculated from optical data of noble metals. Fine structure in the true cross-section is not reproduced by λK(T), but fortunately this is not important for the final spectrum, since detailed features will tend to be smeared out by multiple scattering.
Several attempts have been made to modify the background subtraction procedure, starting with the Tougaard algorithm. In the approach by Tokutaka et al. [14], the cross-section is approximated by Eq. (2) and certain criteria regarding the peak area are used to determine modified B and C values. For the same XPS line, but with different instruments, very large variations are found in the C parameter. These variations seem physically unjustified. In the method of Jo [15], the background subtraction is performed using an optimization routine involving two functionals. This method determines additionally an inelastic cross-section corresponding to the background removal. As input, the procedure depends on values of the photoelectron excitation probability. Recently Seah [16] has evaluated the consequences for quantification when C and B are varied. It is found that peak intensity ratios are unaffected by variations in C, provided that B and C are chosen so that the background matches the data far below the main peak. However, no comments are made on the physical significance of varying C.
In the present work, we evaluate the importance of the emission angle on the magnitude of surface losses in XPS. It is considered how these effects may be taken into account in the usual procedure for background subtraction, by varying C and B in a systematic way. The physical significance of C is demonstrated by correlating the shape of the universal function with the shape of exact cross-sections in REELS.
Section snippets
Experimental
Angle-resolved XPS spectra were measured with a VG ESCALAB 210 spectrometer. The system was equipped with a hemispherical energy analyzer with a five-channel multiplier, a dual (Al/Mg) X-ray source and an ion sputter gun. The polycrystalline samples of Fe, Co, Cu, Ag and Au were cut from high purity sheets and mechanically polished before mounting in the UHV chamber with base pressure of 3×10-10 Torr. The Si samples were cut from a single crystal n-doped Si(111) wafer (resistivity 3–6 Ω cm). All
Results
From all measured spectra a straight line fitted on the high energy side of the peak was subtracted. Secondly, the intensities of the spectra within a series were normalized at the position of the main photoemission peak. The exact energy-dependent transmission was not known for the present analyzer settings and no corrections were made. We do not expect the transmission function to influence the angular dependence of energy losses in XPS spectra because it will have the same relative effect
Data analysis
The algorithm used in background subtraction of XPS and AES spectra was described in a number of previous papers [6], [18], [19]. We will only mention that, when analyzing homogenous materials, the in-depth distribution function of electron emitters f(x) is constant. In this case the primary spectrum F(E) is found from the measured spectrum as:where j(E) is the measured spectrum, L is a characteristic path length for elastic scattering and normally L≫λ. The
Conclusions
Surface losses in the XPS spectra from a series of homogenous transition metal samples and one Si sample, were investigated using a small spot analyzer. Angle-resolved measurements with exit angles 0, 45, 60, 75 and 82° were considered. For the transition metals we find an overall increase in the background level with increasing exit angle. We have analyzed the angle dependence of the background using the Tougaard method and the universal cross-section. For the transition metals, we find that a
References (25)
J. Electron Spectrosc. Relat. Phenom.
(1990)Solid State Commun.
(1987)Surf. Sci.
(1994)Surf. Sci.
(1999)Surf. Sci.
(1989)- et al.
Solid State Commun.
(1986) Phys. Rev.
(1957)- et al.
Phys. Rev. B
(1996) - et al.
Phys. Rev. B
(1991) Phys. Rev. B
(1972)
Surf. Interface Anal.
Cited by (13)
Improved peak-fit procedure for XPS measurements of inhomogeneous samples - Development of the advanced Tougaard background method
2015, Journal of Electron Spectroscopy and Related PhenomenaCitation Excerpt :Therefore, we had to use two peaks for each line with different PIESCS but equivalent in all parameters except for the peak intensities. The complexity of the loss structure of Si was already observed by various groups [16–18] using REELS and other methods. The main loss peaks of the loss structure (bulk plasmon at ∼17 eV and surface plasmon at ∼12 eV) were simulated in a good agreement using the PIESCS of insulators.
Dielectric description of the angular dependence of the loss structure in core level photoemission
2012, Journal of Electron Spectroscopy and Related PhenomenaCitation Excerpt :For the spectrum acquired at 82° emission angle, it seems that this loss feature is composed of two peaks at about ∼4 and ∼7 eV, respectively. Note that background subtraction of the angular effects seen in the XPS spectra here was previously achieved [19] by empirically introducing a surface and a bulk cross section and the shape and relative intensities of these were parameterized by fitting to experiments. This gave a good practical method to account for both surface and bulk contributions to the background signals.
Attenuation lengths of high energy photoelectrons in compact and mesoporous SiO <inf>2</inf> films
2012, Surface ScienceCitation Excerpt :Besides, the Drude plasmon excited at the Si substrate is also clearly visible at 17.2 ± 0.1 eV lower kinetic energies than the Si0 signal. Bulk plasmon in Si0 is reported with 16.8 eV energy loss both from reflection electron energy loss [23] and standard photoemission experiments [24]. The difference (about 0.4 eV) can be ascribed to recoil effect due to the high energy of the emitted electrons, in agreement with theoretical predictions [7].
Bulk and surface plasmon excitations in amorphous carbon measured by core-level photoelectron spectroscopy
2009, Applied Surface ScienceBackground subtraction III - the application of REELS data to background removal in AES and XPS
2001, Surface ScienceCitation Excerpt :Although other workers [18] in REELS, have considered the scattering angle variation of the elastic scattering, there seems no data published for the angular dependence of the loss spectra. Tougaard and coworkers [26,27] consider the emission angle dependence of the spectra, but at fixed scattering angle. The relative effects in Fig. 13 may be judged by dividing each spectrum by the average spectrum.