Chapter 3.18. Thin and ultrathin films and multilayers

Early X-ray absorption fine-structure (XAFS) studies were performed on bulk samples using laboratory X-ray tubes. The advent of dedicated sources for synchrotron radiation (SR) drastically increased the sensitivity and discrimination capabilities of XAFS. Hence, the transition from bulk samples to thin films or even surface monolayers was enabled by monochromated SR. In the late 1970s, surface-sensitive surface extended X-ray absorption fine structure (SEXAFS; Citrin et al., 1978; Stöhr, 1978) was introduced to reveal the distances of atoms located on the surface of a solid to neighbouring atoms by means of energy-dependent oscillations of the photoelectric cross section. For the XAFS measurement procedure, either partial (Auger) or total electron yields (EY) or element-specific X-ray fluorescence yield (FY), i.e. fluorescence radiation intensity, have been used. For near-surface atoms contributing to XAFS signals, the information depths of a few nanometres in the case of electron detection allows for efficient discrimination from the same kind of atoms that are part of the bulk sample below the surface region. When using X-ray fluorescence emission the detection sensitivity of near-surface atoms can be considerably higher, but the penetration and information depths are at least one order of magnitude higher than in the case of electrons, so that specific excitation schemes, such as total reflection of the incident X-ray beam at a very flat substrate, are required to ensure a similar discrimination capability between surface and substrate bulk atoms of the same kind. In the early 1970s it was shown that the energy dependence of the electron yield is proportional to the bulk absorption coefficient, thus qualifying EY for SEXAFS (Gudat & Kunz, 1972).

The anisotropy in X-ray absorption has been studied using the layered transition metal dichalcogenide compounds 2H-WSe₂ and 1T-TaS₂. They form a quasi-two-dimensional structure, i.e. a layer of transition-metal atoms covalently bonded between two layers of chalcogen atoms (Heald & Stern, 1977). The covalently bonded sandwiches form a three-dimensional structure due to van der Waals forces. Thus, one may expect intralayer interactions to be considerably stronger than interlayer interactions, resulting in anisotropic properties that are experimentally confirmed in the white line of selenium.

One of the first ultrathin-layer XAFS experiments was conducted on the approximately 3 nm thick native oxide layer on top of an aluminium substrate (Stöhr et al., 1978). Recording the partial electron yield (PEY) emitted from the aluminium oxide layer by means of a cylindrical mirror analyzer corresponds to the dynamics of a transmittance measurement on a free-standing 3 nm aluminium oxide film which could not be prepared in order to cover at least the area corresponding to the incident X-ray beam profile. The PEY recording across the O K absorption edge took about 20 min on the `grasshopper' design beamline at Stanford Synchrotron Radiation Lightsource. This PEY spectrum revealed three EXAFS oscillations, the Fourier transform of which is in line with the characteristic distances of O atoms in Al₂O₃.

With respect to effective measurement times and detection sensitivities, one has to consider that measurements in fluorescence detection mode suffer from decreasing values of the atomic fundamental parameter of the fluorescence yield, as opposed to the Auger emission probability, with decreasing atomic number Z, while PEY measurements are restricted by an Auger peak being located on a relevant inelastic electron background. Regarding the probing depth in a thin layer, one should note that the PEY energy degrades due to losses with increasing depth, thus allowing escape-depth discriminations when applying appropriate detection grid voltages (Genzer et al., 2002).

One of the first examples of EXAFS investigation of a thin film in the low-nanometre range was performed using copper samples deposited on glass microscope slides by sputtering. The experiments (del Cueto & Shevchik, 1978) were conducted using a monochromated rotating-anode laboratory X-ray source, with typical measurement times of about one hour per sample. The monochromatization was achieved by using entrance and exit slits in conjunction with a curved LiF crystal as a dispersive element. It turned out that the thinnest 30 Å copper depositions showed structures in line with CuO, while all thicker depositions were nearly in line with copper.

The initial studies by X-ray absorption near-edge structure (XANES) spectroscopy were aimed at revealing information about the chemical binding state of the element which had been probed across an absorption edge in the tender to hard X-ray range, while the far-edge investigations by EXAFS were used to derive coordination information. One of the first studies probing the coordination sensitivity of XANES (Farges et al., 1997) employed both experimental investigations and theoretical calculations based upon the ab initio multiple-scattering code FEFF7. Different titanium compounds with titanium located in sites coordinated by four, five or six O atoms were studied. The extent to which coordination information can be revealed from Ti K pre-edge features of the XANES structure has successfully been demonstrated.

When moving from ultrathin layers to thicker layers or even bulk samples, various challenges need to be addressed when obtaining reliable information by X-ray absorption spectroscopy in different configurations. While the information depths of total or partial electron detection is limited to a few nanometres, thus limiting information to only ultrathin layers or the surface-proximal region of thicker specimens, the penetration depth of incident X-rays can range from several hundred nanometres to even a few millimetres depending on the sample composition and the photon energy of the incident radiation. The X-ray information depth depends equally on the sample composition and the photon energies of the emitted fluorescence radiation. For very shallow incidence and very flat sample substrates, X-ray standing-wave (XSW) fields (Bedzyk et al., 1989) can occur that allow the information depth to be driven by means of the incident angle-dependent penetration depth, for example in the case of nanolayers deposited onto such a substrate (Pollakowski et al., 2008). In the case of significant penetration of the incident X-rays into a thick layer or bulk sample, the energy dependence of the photoelectric cross sections of the probed orbitals X_i may cause drastic distortions of the emitted fluorescence intensity due to self-absorption effects. In the case of good a priori knowledge of both the main matrix element composition of the layer or the bulk sample and of the attenuation coefficients, the main contributors to which are in general the photoelectric cross sections of the main matrix elements, the XAFS distortions due to self-absorption can be corrected for. This procedure has been demonstrated for distortions caused by self-absorption effects affecting the oxygen EXAFS spectrum of an NiO single crystal (Tröger et al., 1992), which was recorded using fluorescence radiation detection. In a similar procedure, self-absorption effects have been quantitatively investigated for concentrated samples (Eisebitt et al., 1993). In both of these pioneering studies the self-absorption effect was systematically investigated by varying the experimental geometry of both the incident and observation (detection) directions. In the following, further details of the origin of these distortions and the underlying excitation and emission relationships, as well as their correction scheme, will be given.

This detection of fluorescence radiation is, in general, called X-ray fluorescence (XRF), but in the XAFS community the expression `recorded in fluorescence yield' (FY) is often used. However, outside the XAFS community the expression fluorescence yield is used for the probability of the emission of a fluorescence photon as opposed to that of an Auger electron following an inner-shell excitation by X-rays.

When exciting an element of interest j in a thin layer or bulk sample with monochromatic X-ray radiation, the photon energy E₀ of which is above the absorption edge X_i of the element, an inner-shell ionization takes place that can decay either by the emission of a fluorescence photon or an Auger electron. The intensity of the fluorescence radiation emitted is proportional to the photoelectric cross section (E₀) and the fluorescence yield .

However, both the experimental beam geometry (the angles of incidence α and observation β) and the mass deposition (number of atoms per unit area) of all elements in the sample considerably affect the emitted fluorescence intensity. On the way into the sample the X-ray absorption, which is calculable using the total absorption coefficient μ_tot(E₀) of all matrix elements, defines the number of atoms j per unit area that are actually excited while the emitted fluorescence radiation is being reduced by X-ray absorption at the photon energy E_f of the fluorescence radiation. The absorption coefficient μ contains the sum of all (E₀) for all shells X_i with absorption-edge energies below the excitation energy E₀ as well as the total scattering cross sections for elastic and inelastic scattering of the excitation energy E₀. The integration over the penetration depths of the excitation radiation into the sample including the absorption of the induced fluorescence radiation leads to an absorption correction factor depending on the angles of incidence α and observation β and the absorption coefficients μ_tot(E₀) and μ_tot(E_f). As the incident photon energy E₀ is being varied during an EXAFS scan, the energy dependence of (E₀) and its increasing contribution to μ_tot(E₀) with increasing matrix concentration¹ of element j in the matrix is the source of the damping of EXAFS resonances proportional to . The absorption correction factor is part of the so-called Sherman equation (Sherman, 1955) for the emission of element-specific fluorescence radiation in single- or multi-elemental samples, which was proposed in the mid-1950s. For the sake of completeness one should mention that the detected fluorescence count rate in an XAFS or XRF experiment is proportional to the emitted fluorescence intensity divided by the effective solid angle of detection and the energy-dependent efficiency of the detector (Mantler, 2006). For the quantitative description of X-ray fluorescence analysis of thin-layer samples the Sherman equation has been modified to include XSW modulations of the incident beam intensity under grazing-incidence conditions and contains the respective expressions for the absorption correction for thin films (de Boer, 1991).

Thus, one may state that the valuable absorption correction procedures indicated in the pioneering XAFS studies (Tröger et al., 1992; Eisebitt et al., 1993) rely on the same type of absorption-correction algorithms as developed earlier in the XRF (Sherman, 1955) to GIXRF (de Boer, 1991) quantification schemes.

Other work following the pioneering XAFS work described similar correction algorithms for hard X-ray EXAFS spectra recorded from flat specimens of known composition in varying detection beam geometries (Pfalzer et al., 1999) for different sample thicknesses (Booth & Bridges, 2005), or in the case of angle-dependent studies of the magnetic circular dichroism of Co/Cu(100) in total electron yield (TEY) detection (Hunter Dunn et al., 1995).

The beam geometry of the X-ray excitation and the flatness of the layered structure determine absolute detection limits based upon fluorescence detection for XAFS spectroscopy. In the case of very flat substrates and thin layers deposited on them, total reflection and grazing-incidence regimes can successfully be arranged for (see Fig. 1). In contrast to the analysis of bulk samples, where beam geometries with a normal incidence for the excitation channel and a grazing emission (GE) with respect to the sample surface in the detection channel allow self-absorption effects and related spectral distortions to be substantially reduced, the speciation and coordination of thin layers by XAFS does not require normal incidence of the excitation radiation but rather grazing-incidence (GI) conditions. For very thin layers, no relevant self-absorption distortions occur even under GI conditions. In the soft X-ray regime roughness values of about 5 nm r.m.s. and in the hard X-ray range roughness values of about 1 nm r.m.s. and below are required to fulfil the requirements of these GI and GE regimes. Tuning the angle of incidence from very shallow angles across the angle of total reflection of the cap layer, the detection sensitivity decreases by about one order of magnitude. A further increase in the incident angle by a factor of about four or more results in a decrease in the detection sensitivity by another order of magnitude. When comparing the detection sensitivity of X-ray fluorescence emission as opposed to electron emission, with the latter originating only from the top few nanometres of a layered sample, one may notice the detection sensitivity for photon detection being a few orders of magnitude higher. Of course, the photon-related detection sensitivity scales with the square root of the effective solid angle of detection depending on the projected footprint size, the area of the energy-dispersive detector employed and its distance from the sample surface.

Figure 1 Beam geometries for XAFS of layered samples in fluorescence-detection mode (detection of element-specific X-ray fluorescence, XRF, often also labelled as fluorescence yield, FY, in the XAFS community) ranging from total reflection (TXRF) over grazing incidence (GIXRF) to conventional beam geometry (XRF) as depicted to the left. When working at shallow angles of incidence, a reliable beam monitor for both the beam position and incident radiation intensity is crucial, in particular for the EXAFS regime. An alternative to using the fluorescence radiation emitted would be to use the radiation reflected at a flat sample (as shown) or the emitted electron yield, EY, depending on the experimental strategies. In order to reduce the (self-)absorption effects that reduce XAFS resonance intensities, one may use a combined GI excitation and grazing-emission (GE) observation scheme as depicted to the right. The beam geometries depicted here differ from the conventional geometries used for XAFS of bulk samples, which is due to the short penetration length of the incident beam into a thin layer even under GI conditions.

Last, but not least, one should mention the strict requirements on the spatial stability and repeatability of a monochromator beamline when performing an energy variation across the near-edge XAFS (NEXAFS or XANES) region to the extended EXAFS region. In the case of laterally heterogeneous samples or under grazing-incidence conditions, it is crucial that the exciting beam remains spatially stable within only a few micrometres during the variation of the photon energy determined by the beamline. The spatial beam stability is crucial if the beam-intensity monitor is not perfectly homogenous across the beam profile, as may be the case for thin photodiodes (which in the case of a thickness of 6 µm can be employed above about 2.4 keV photon energy) as opposed to a gas ionization chamber. The beam intensity monitor recording the radiant power of the monochromated radiation serves to improve the repeatability of XAFS scans, as tiny variations of either the electron beam orbit in the storage ring or the local heat load on the dispersive elements of the beamline may affect the beamline transmittance, thus affecting the intensity of either the reference energy scan or the XAFS scan itself.

In the case of thin multilayered depositions on very flat substrates, another detection scheme of XAFS based upon modulation of the reflectance has been investigated (Wang et al. (2000). The energy-dependent variation of the reflectivity of a multilayered sample is due to the energy variation of the optical constants or atomic scattering factors of the layer materials that are related to the photoelectric cross section of the atoms of the materials: the imaginary part of the atomic scattering factor is proportional to the product of the photoelectric cross section and the photon energy, while its real part is related to the imaginary part by the Kramers–Kronig dispersion relation.

In general, X-ray reflectivity can provide dimensional information on layer thicknesses and roughnesses and diffusion and interfacial profiles, as well as on the layer density. The data evaluation is based upon the Fresnel equations used to calculate the XSW field as in grazing-incidence X-ray fluorescence techniques. Fig. 2 shows the nodes and antinodes of XSWs at a stratified sample for grazing-incidence XAFS techniques. Soft X-ray resonant reflectivity (Wang et al., 2000) has been applied to low-Z materials such as a bilayer polymer thin film, offering enhanced and selective sensitivity to specific chemical species near the absorption edges of constituent elements (NEXAFS). It can be shown that the strongly oscillatory energy dependence of the reflectivity for the various interfaces provides tunable, high sensitivity to a particular interface.

Figure 2 Beam geometry and XSW in a layered specimen for XAFS investigations under grazing-incidence conditions. The local XSW field intensity varies with the depth from the surface of the layered specimen and can show drastic changes at interfaces.

Other work (Lützenkirchen-Hecht & Frahm, 2000, 2005) has addressed the extended X-ray absorption fine-structure technique (EXAFS) in reflection mode for the ex situ investigation of sputter-deposited thin films of silver, gold and Ta₂O₅ on float glass substrates, where reactively sputtered Ta₂O₅ thin films reveal a high degree of disorder compared with crystalline β-Ta₂O₅. A detailed analysis of the reflectivity fine structure enabled the extraction of short-range order structural information such as bond distances, coordination numbers and Debye–Waller factors (Lützenkirchen-Hecht & Frahm, 2000). The surface roughness and the density of the thin films were determined from specular and nonspecular X-ray scattering experiments, with silver showing the largest differences from the other depositions, with an r.m.s. difference of below 1 nm. Using grazing-incidence X-ray absorption spectroscopy (GIXAFS) in reflection mode, in situ investigation of solid–liquid interfaces (Lützenkirchen-Hecht & Frahm, 2005) was performed during the active dissolution of metals. For silver in weakly acidic Na₂SO₄ solutions, the formation of an AgO species at the surface of a silver electrode was found, i.e. the active dissolution proceeds by means of a non­protecting surface layer of about 5 nm in thickness. The respective atomic short-range order differs from polycrystalline silver oxides (Ag₂O and AgO) and relates to a more disordered or amorphous Ag⁺ oxide. The XAFS techniques in reflection mode also led to dedicated experimental stations such as the GILDA CRG beamline at ESRF, France (D'Acapito et al., 2003). Here, the sensitivity of the approach allowed the detection of very small signals from very thin samples of just a few monolayers, for example for depositing thin films under controlled conditions and thermally treating the samples in order to study dynamical processes.

The field of probing a minute amount of material deposition on flat substrate and thin-layered systems using the various XAFS methods has been developed in many different applications. NEXAFS measurements were performed on a variety of carbon materials covering a range of hybrid bonding characters from pure sp³ (diamond) to pure sp² (graphitic) (Coffman et al., 1996), including diamond and chemical vapour-deposited (CVD) diamond films of different qualities. The NEXAFS investigations demonstrated clear qualitative bonding information for a full range of sp²/sp³ bonding ratios and, in fact, NEXAFS provided a more sensitive quality assessment of the diamond bonding than the Raman technique used for comparison. NEXAFS has been used to study the defect content and the bonding modifications induced in boron nitride thin films by ion implantation (Jiménez et al., 1996). The initial films were hexagonal-like boron nitride grown on Si(100) by pulsed laser deposition (PLD) that were modified by means of a high-energy implantation which induced the formation of a significant proportion of sp³ bonding (cubic-like) and the formation of nitrogen void defects in the remaining sp² boron nitride. These bonding modifications lacked long-range order and clearly demonstrated the significant discrimination capabilities of NEXAFS. In order to better understand the microscopic origin of magnetic exchange anisotropy, soft X-ray XAFS has been employed to confirm and to study chemical reactions at as-grown metal/oxide interfaces, which could be described as an oxidized-metal region adjacent to a reduced-oxide region (Regan et al., 2001). These quantitative investigations allowed the chemical changes to be attributed to one or two atomic monolayers at the interfaces. Another field of application of XAFS concerns self-assembled monolayers and films. Here, not only the bonding state is of interest but also the molecular orientation, which can be probed by means of two orthogonal polarizations of the incident radiation. A first example of polarization-dependent NEXAFS was to determine the ordering of octadecyltrichlorosilane (OTS) molecules in self-assembled (SA) films on Si/SiO_x (Peters et al., 2002). For different sample-preparation modes, the strong polarization dependencies of the C K NEXAFS spectra allowed both different molecular tilt angles against the substrate surface and different SA coverages (or mass depositions) to be derived.

With the increasing complexity of advanced nanoscaled materials, the discrimination capabilities of XAFS methodologies have been improved in order to reveal depth-dependent information on buried nanolayers and interfaces. The respective XAFS techniques are based on grazing-incidence approaches, where the angular-dependent penetration depth can determine the information depth. As the penetration depth varies when tuning the photon energy of the incident radiation across the absorption edge of an element, a corresponding adaption of the angle of incidence has to be foreseen during the measurements in order to keep the intensity of the excitation radiation constant either within a nanolayer or an interface.

The corresponding measurement strategies (Pollakowski et al., 2013) have to be adapted to nanoscale layered samples, enabling the combination of NEXAFS spectroscopy and a grazing-incidence X-ray fluorescence (GIXRF) detection mode. This approach allows chemical nanometrology of internal material interfaces. Their validation has successfully been performed with nanolayered model structures consisting of a silicon substrate, a physically vapour-deposited nickel metal layer and, on top, a chemically vapour-deposited B_xC_yN_z light-element layer. In such a case of internal material interfaces, the matrix elements of the two adjacent layers are involved. By analyzing the elements of the deeper positioned layer, identification of the interface species can be achieved by means of comparing two NEXAFS spectra recorded at different angles of incidence. For an angle of incidence below the critical angle of total reflection of the lower layer, the NEXAFS spectrum is mainly influenced by the chemical bond of the interface, with only a small fraction originating from the whole lower layer. A second NEXAFS measurement carried out at a large incident angle allows identification of the species of the entire lower layer. Similar NEXAFS measurements of the upper layer are required at very shallow angles of incidence. A further development of this combined NEXAFS–GIXRF method (Pollakowski & Beckhoff, 2015) allows a depth-resolved analysis of multilayered nanoscaled thin-film structures which could be validated by means of appropriate model systems consisting of a carbon cap layer, two titanium layers differing in their oxidation states and separated by a thin carbon layer, and a silicon substrate covered with molybdenum and a carbon layer. The corresponding depth profile of the XSW field intensity is shown in Fig. 3, while Fig. 4 shows the respective NEXAFS spectra recorded under varying grazing-incidence conditions. A similar differential approach as in the case of the interfacial speciation has been set up to reveal the oxidation state of each of the titanium layers.

Figure 3 Angular-dependent depth distribution of the XSW field intensity modifying the excitation in XRF detection for soft X-ray NEXAFS at the Ti L absorption edge for two sample systems with a reversed sequence of titanium and Ti2O3 nanolayers.

Figure 4 Ti L NEXAFS spectra at shallow and steep angles of incidence of the two sample systems with reversed sequence of titanium and Ti2O3 nanolayers.

Complementary to the grazing-incidence techniques just described, grazing-emission X-ray fluorescence (GEXRF) can also be successfully combined with NEXAFS (Kayser et al., 2015) for elemental and species depth profiling of nano­scale sample systems. Here, the observation angle of the X-ray fluorescence radiation emitted is varied at a very small grazing-emission angle, thus tuning the probing depth from a few to several hundred nanometres. Two different detection modes have successfully been demonstrated: sequentially scanning a slit in conjunction with an energy-dispersive detector and using a scanning-free approach based on a position-sensitive charged-coupled device (CCD). The combination of scanning-free GEXRF and XAS allows depth-resolved chemical speciation with nanometre-scale accuracy. While the conventional GEXRF approach is advantageous to minimize self-absorption effects, as the emitted fluorescence lines have photon energies below the absorption edge of the matrix element probed, the use of a scanning-free GEXRF arrangement allows for larger solid angles of detection and reduced measurement times, paving the way towards time-resolved depth-sensitive XAS measurements. Initial GEXRF–XAFS studies addressed the surface oxidation of a iron layer on the top of bulk silicon and of a germanium bulk sample. The GEXRF–XAFS experimental approach has been demonstrated to be well suited for in situ sample surface studies in the nanometre regime.