H. Bluhm,     Lawrence Berkeley National Laboratory, USA

4.1 Introduction

X-ray photoelectron spectroscopy (XPS) is one of the most commonly used surface science techniques.^1–4 It is based on the photoelectric effect,⁵ where an X-ray photon is absorbed by a core or valence electron. If the incident photon energy is larger than the binding energy of the electron, the electron will be emitted. The basic set-up of an XPS experiment is shown in Fig. 4.1. Monochromatic X-rays from a laboratory source (X-ray tube) or from a synchrotron irradiate the sample. The emitted electrons are collected by an electrostatic lens and energy-analyzed by a spectrometer, most commonly a hemispherical electron energy analyzer. From the known photon energy hv and the measured kinetic energy KE, the binding energy BE of the electrons can be determined according to BE = hv – KE Φ, with Φ as the spectrometer work function. Typical incident photon energies in an XPS experiment range from several 10s to well over 1000 eV, with kinetic energies of the same order.

While the penetration depth of the X-rays into the sample is of the order of hundreds of nanometers or more,⁶ elastic and inelastic interactions of photoelectrons with atoms in the sample limit the probed depth to a maximum of a few nanometers over the kinetic energy range used in conventional XPS studies. Figure 4.2 shows the inelastic mean free path (IMFP) of electrons as a function of their kinetic energy in Cu; the minimum path length is at about 100 eV kinetic energy (see Fig. 4.2).⁷ This surface sensitivity makes XPS particularly well suited for the study of ultrathin films with only a few monolayer or less in thickness. Figure 4.2 also shows that the IMFP increases with increasing kinetic energy, e.g., in the case of Cu up to about 10 nm at a kinetic energy of 10 keV. With the use of high-brilliance synchrotrons that provide X-rays in this energy range, the study of thick multilayer system and buried interfaces has become feasible, as will be discussed later in this chapter. It should be noted that the inelastic mean free path is not a precise description of the actual probing depth in an XPS experiment, since it neglects elastic scattering effects. A more accurate quantity for the measurement depth is the effective attenuation length (EAL), which includes elastic scattering effects.⁸

While the short mean free path of electrons in matter makes XPS an exquisitely surface-sensitive method, it also requires in general high vacuum conditions during the measurements, since the emitted photoelectrons are also scattered by gas phase molecules. For example, the mean free path of 100 eV electrons (i.e., those with the highest surface sensitivity) is about 1 mm in 1 torr of water vapor. Since the electrons travel many centimeters on their way to the electron detector, attenuation of the signal under elevated pressure conditions poses a limit on the background pressure in conventional XPS set-ups. Over the years, different schemes for in situ XPS measurements have been developed that now allow measurements at pressures in the torr range, as will be shown later in the chapter.

The advantages of XPS for the investigation of thin films are its surface sensitivity and chemical specificity. Since each element exhibits a characteristic set of core level peaks in XPS, the chemical composition of the surface and near-surface region of a sample can be determined quantitatively. Moreover, the binding energy of core and valence electrons shows subtle changes (of the order of tenths to several eV) depending on the chemical bonding (e.g., the oxidation state). This allows the discrimination, for instance, of Si atoms in an oxide layer (~Si^4 +) from Si atoms at the interface (~Si^1 +, Si^2 +, Si^3 +) and elemental Si in the substrate (Si⁰) of a Si wafer covered by a native oxide layer (see Fig. 4.3).⁹ When interpreting these so-called ‘chemical shifts’, one has to keep in mind that the binding energy depends both on the state of the atom before the photoemission event (‘initial state effects’) as well as contributions due to the reorganization of the electrons in the atom during the relaxation process, when the core hole is filled (‘final state effects’). It is, however, in many cases possible to assign chemical shifts to initial state effects. For instance, the observed chemical shifts for the different Si species in Fig. 4.3 can be rationalized by different degrees of charge transfer from Si to O (depending on the Si oxidation state), which leads to variations in the screening of the Si 2p electrons by Si 3s and 3p valence electrons. Thus, the 2p electrons in the case of Si^4 + experience a stronger Coulomb attraction to the nucleus than in Si^3 +, resulting in a higher binding energy in the case of Si^4 +. This sensitivity to the chemical state of the atoms means that valuable information can be gleaned from XPS measurements about, e.g., bonding within thin films, at the surface, as well as at the interface to the substrate (see Fig. 4.3).

XPS can also be used to determine the thickness of deposited layers on a substrate, assuming ideal layer-by-layer growth and smooth interfaces. For a single layer with homogeneous chemical composition and thickness d_m, the detected intensity I_m, using the straight-line approximation, is given by^2,10

4.1

with λ_m^′ = λ_m cos(α) as the effective attenuation length at angle a between the surface normal and detector position (see Fig. 4.4), λ_m the attenuation length of electrons for layer m at a given kinetic energy, and z the depth measured from the layer surface. The factor S contains a number of experimental parameters and is given for layer m and inner shell orbital j by:

4.2

with Φ as the X-ray flux at a given X-ray energy (hv), σ the photoemission cross-section¹¹ at a given hv, β the orbital specific asymmetry parameter, D the spectrometer efficiency for a given kinetic energy, and N the number of atoms per unit volume. For an infinitely thick layer with d_m → ∞ (i.e., a bulk substrate), the intensity is I_subs = S_subs^jλ_subs^′.

For a multilayer system consisting of n homogeneous layers above a bulk substrate (see Fig. 4.4), where the order of stacking from bottom to top is given by layer 1, layer 2, …, layer n, the XPS intensity from the substrate is attenuated by each layer according to the Beer– Lambert law:¹²

4.3

Likewise, the intensity of layer m in a system of n layers is being attenuated by (n – m) layers, and is thus given by:

4.4

There are two principal methods to obtain depth-dependent elemental and chemical information using XPS: (1) depth-profiling by etching (for instance with argon ions), where surface layers are gradually removed by sputtering, while XPS spectra are obtained at various stages during the removal of surface layers, and (2) non-destructive depth profiling through the variation of the escape-depth of the electrons, either by varying the detection angle relative to the sample surface or by measuring photoelectrons with different kinetic energies. The advantage of depth profiling by argon etching is that it is not limited by the escape depth of the electrons, so that information about the sample from depths in excess of several nm can be obtained. The disadvantage is that the sample surface is destroyed in the process, and that the chemical state of the surface may be altered due to the impact of Ar ions.¹³

Non-destructive depth profiling in an XPS experiment that uses an X-ray source with a fixed energy (i.e., a lab-based experiment) relies on the variation of the take-off angle α of the photoelectrons (see Fig. 4.4), which determines (at a fixed kinetic energy) the information depth in the experiment.^2,14 An example for this technique is shown in Fig. 4.5 for the case of a 4 nm thick SiO_xN_y film on a Si substrate.¹⁵ The relative intensity of the Si⁴+ peak is clearly enhanced at shallow detection angle. The schematic representation of the electron trajectories in Fig. 4.5 also illustrates that the depth resolution in angle-resolved XPS depends on the angular acceptance of the electrostatic analyzer, which should be ideally as narrow as possible. Since the effective probing depth λ′ varies with the take-off angle α according to λcos(α) (with λ as attenuation length), the most sensitive determination of the film thickness is at the most grazing detection angles. Measurements at these angles, however, are hampered by surface roughness effects, which need to be taken into account.¹⁴

The use of variable incident photon energies (and thus kinetic energies) for depth profiling is performed using a fixed sample–detector geometry and is thus not influenced by surface geometry effects. This method requires a photon source with variable energy, such as a synchrotron. The technique makes use of the dependence of the mean free path of electrons in matter on their kinetic energy (see Fig. 4.2). An example is shown in Fig. 4.6, where the growth of praseodymium-SiO₂ layers on Si was monitored using XPS.¹⁶ Figure 4.6(a) shows Si 2p spectra after each preparation step: first after the growth of a 2.6 nm thick SiO₂/Si film, then after the deposition of 1 nm of metallic Pr on top of SiO₂, and finally after annealing the sample to 600 °C, which leads to incorporation of Pr into SiO₂ in a solid state reaction. under each condition Si 2p spectra were taken with kinetic energies of 440, 660 and 920 eV, with the most surface-sensitive spectrum at 440 eV and the most bulk sensitive at 920 eV. The spectra after annealing show a strong peak at a binding energy of ~102 eV in addition to the Si⁰ and Si^4 + peaks due to the substrate and the SiO₂ film, respectively. The peak at 102 eV is attributed to the formation of Pr silicate, which is located predominantly at the sample surface, as the depth profiling spectra after post-deposition annealing (or PDA, in Fig. 4.6a) prove. This scenario is confirmed by cross-sectional transmission electron microscopy images that were taken after the reaction (Fig. 4.6b).

To summarize so far, XPS is a quantitative method for the determination of thin film elemental composition and chemistry and can also be used under favorable circumstances to determine the film thickness, as well as to distinguish surface from bulk species. The following section discusses how XPS can be used to in situ monitor thin film growth, the reaction of thin film surfaces with adsorbates and gases, as well as the chemistry of solid/solid interfaces in multilayer systems. Some of these investigations required the development of new XPS instrumentation, which will be discussed alongside the experimental results.

4.2 In situ monitoring of thin film growth

Studying the elemental and chemical composition as well as the thickness of thin films in situ during film growth using XPS requires an expansion of the technique to pressures in the mtorr range. The first dedicated XPS instrument for monitoring thin film growth was described by Kelly et al. in 2001.¹⁷ A schematic of the instrument is shown in Fig. 4.7. This system is capable of operating at pressures in the 10⁻³ torr range, about three orders of magnitude higher than the pressure limit in conventional XPS systems. The X-ray source (Mg Kα anode) is separated from the gas atmosphere in the measurement chamber by a 2 μm thick aluminum foil, which is about 70% transparent for Mg Kα radiation (1254 eV).⁶ The electrostatic lens system of a commercial hemispherical analyzer was modified by introducing a 1 mm diameter differentially pumped aperture in the lens column. Two spherical stainless steel meshes are mounted in front of the standard input lens to collect and focus electrons from the sample onto the differentially pumped aperture (see Fig. 4.8). After passing through the meshes the electrons are accelerated to energies between 1000 to 1300 eV, which decreases the lens magnification to about 1 and also decreases the scattering of electrons by gas molecules due to the high electron kinetic energy. At the given lens magnification the electron analyzer measures a sample area of about 1 mm diameter.

The pressure differential between the measurement/thin film growth chamber and the hemispherical analyzer was 3–4 orders of magnitude, providing the necessary pressure difference for operation at mtorr pressures in the experimental chamber. Other considerations in the design of the spectrometer included high-speed acquisition of spectra to follow film growth with high-time resolution, a large working distance (~3 to 5 cm) between the electrostatic analyzer input lens and the sample to avoid interference by the electrostatic lens system with the reactant flow to the sample, as well as the suppression of stray electrons and ions that may be generated in plasmas near the growing surface.¹⁷ The latter is achieved by biasing the meshes in the input lens positively at 20 V (thus rejecting slow ions) and operating the lens elements in a manner that reduces the transmission of secondary electrons. High data acquisition speed was achieved by a large collection angle of the electrostatic lens (30° total) and the use of an imaging electron detector that measures a range of electron energies simultaneously.

A case study for the application of this instrument is shown in Fig. 4.9. The growth of a thin film of tungsten oxide on a Si substrate was monitored by measuring the W 4f, O 1s and Si 2p core level peaks (see upper panel in Fig. 4.9). The acquisition time for each spectrum is 30 s. The initial substrate surface shows the characteristic Si 2p peaks for a native oxide layer on a Si substrate, where the peak in the O 1s spectrum is purely due to the native silicon oxide layer. Tungsten is evaporated onto the substrate from a hot filament in a background of 1 mtorr of oxygen. The evolution of the W 4f, O 1s and Si 2p peaks during deposition shows a decrease in both the Si and SiO_X peaks due to attenuation of Si photoelectrons by the deposited layer. As expected, the W 4f signal increases with increasing deposition time. The same holds for the O 1s intensity, which indicates that oxygen is incorporated into the growing film, since the initial oxygen intensity was only due to oxygen in SiO₂, and the Si 2p peak of SiO₂ decreases with time. The integrated intensities for Si 2p, O 1s and W 4f as a function of deposition time are plotted in the bottom panel of Fig. 4.9.

The Si 2p and W 4f spectra at deposition times of ~11 min (‘A’) and ~61 min (‘B’) are displayed in Fig. 4.10, upper panels.¹⁷ The Si 2p spectra show a decrease in intensity in both the SiO₂ and the elemental Si 2p peaks, as expected for the growth of a layer at the surface. The W 4f spectra initially show the characteristic binding energy for a WO₃ species (‘A’), while at later deposition times some elemental W is also observed. By relating the Si 2p intensity of the pristine substrate to the Si 2p intensity at various stages during deposition, the attenuation of the Si 2p signal can be used to determine the thickness of the WO_x film (see Eq. [4.3]). With the detection angle at 25° from the sample normal, and assuming an attenuation length of 1.8 nm for the Si 2p electrons (kinetic energy ~1150 eV), the escape depth of the electrons in the experiment is 1.8 nm × cos(25°) ~1.6 nm. Using these values, a model for the thickness of the SiO₂ and WO_X layers at different stages during the deposition can be postulated (bottom panel of Fig. 4.10).

Apart from the experiments shown in Figs 4.9 and 4.10, this instrument has also been used to monitor the deposition of Al₂O₃ on InGaAs,^18,19 and the oxidation of Ge(100).²⁰ While the performance of the current set-up used in the case studies described above provided hitherto unavailable opportunities for in situ film growth monitoring, Kelly et al. estimated that the performance of the instrument can be increased by more than an order of magnitude through two measures: (a) increasing the acceptance angle of the spectrometer to 40°, and (b) by use of a state-of-the-art focused X-ray source instead of the unfocused one which is used in the current set-up. A focused X-ray source would increase the X-ray flux in the 1 mm² field of view of the detector by more than a factor of 10. This would then allow the acquisition of the order of one spectrum per second, an improvement of a factor > 10 over the current speed. If in addition the lens design were modified so that the acceptance area of the lens is increased to 4 mm², another factor of 10 in acquisition speed could be gained, making it possible to collect 10 spectra per second. Even though these improvements are challenging, they are merely technical in nature; monitoring film growth with a time resolution of 0.1 s should therefore be possible in the future.

4.3 Measuring the reaction of thin films with gases using ambient pressure X-ray photoelectron spectroscopy (XPS)

In the following we will focus on the reaction of gases with thin films. Ultrathin oxide, metal, organic as well as semiconductor films play an ever-increasing role in technological applications, including industrial catalysis and data storage devices. The interaction of these systems with reactants as well as environmental gases (such as water vapor) has great influence on their performance and longevity. In particular in the field of heterogeneous catalysis the correlation between the yield and conversion in a catalytic reaction (measured by the composition of the gas phase) with the chemical nature of the active catalyst surface is of great importance for a better understanding of the basic, atomic scale processes in a catalytic reaction, and may lead to a rational design of more efficient catalytic materials.

On the other hand, thin film systems can also be used to study surface reactions, in particular phase transitions and volatilization processes²¹ that are difficult to quantify in a bulk system. This is illustrated in Fig. 4.11. The upper panel shows the scenario for a bulk sample that interacts with the gas environment and forms a reacted layer at its surface. Photoelectrons with a certain escape depth (symbolized by the length of the arrow) are used to monitor the reaction. If the reaction also involves volatilization, this process is not detectable in an experiment on a bulk sample: the sample on the left will give an identical signal in an XPS experiment as the sample on the right in the upper panel. However, if one uses a thin film sample instead (lower panel, Fig. 4.11) where the film thickness is of the order of the escape depth of the electrons, the attenuation of the substrate electrons can be used to gauge the degree of volatilization of the film, or of its partial conversion into another phase. In addition, since the sample under investigation has a finite thickness of the order of the escape depth of the electrons, the signal of the sample itself can be used to measure the volatilization or conversion of the material into another phase. In essence, thin film samples are, under certain circumstances, superior to bulk samples to monitor gas/surface interactions. It is, however, necessary to point out two caveats. It has been shown that thin film systems can show markedly different properties from their bulk counterparts; this is in part their appeal for the tuning of reaction properties in catalysis.²² In addition, all the above considerations hold only true in the absence of morphological changes (i.e. deviations from a strictly two-dimensional model) to the film and substrate–film interface; such changes would make the quantitative analysis of thickness changes challenging.

The investigation of the reaction of surfaces with gas phase species needs to bridge the so-called ‘pressure gap’ in surfaces science. In the case of XPS this is hampered by the strong interaction of electrons with gas phase molecules, as pointed out in the section above. The differentially pumped electrostatic lens designed by Kelly et al. afforded measurements at pressures in the mtorr range.¹⁷ Many reactions, in particular in environmental science, require higher pressures: in order to measure, e.g., the surface of neat liquid water the water vapor pressure in the experimental chamber has to be at least 4.6 torr, which is the equilibrium water vapor pressure at the triple point.

To achieve higher pressures in an XPS experiment, the path length of the electrons through the high-pressure region has to be kept as short as possible. In addition, several differential apertures are necessary to keep the electron analyzer in a high-vacuum environment. This basic concept (see Fig. 4.12a and b) was developed more than 30 years ago in the original designs by Hans & Kai Siegbahn and collaborators, which allowed experiments of up to 1 torr.^23,24 Several other groups built instruments based on this concept.^25–27 To overcome the trade-off between an increase in detection efficiency through larger apertures on one hand, and better differential pumping through smaller apertures on the other hand, the latest generation of these instruments uses electrostatic lenses that are placed between the apertures, raising the pressure limit to more than 5 torr.^28–31 The increase in the pressure limit is also partly due to the use of synchrotron radiation, which offers higher photon flux and tighter focused X-ray beams. Since the instruments operate at realistic environmental humidities, the technique is often called ambient pressure XPS (APXPS). Review articles on the basics and applications of APXPS can be found in Refs. 29, 32, 33 and 34.

The application of APXPS to the study of the reaction of thin films is now illustrated using the example of the interaction of a 4 monolayer (ML) thick MgO(100) film grown on a Ag(100) substrate with water vapor. Due to its simple rock salt structure, MgO(100) is an ideal model metal oxide surface for studying the metal oxide–water interface, both theoretically and experimentally.³⁵ Of special interest is the nature of the interaction of water vapor with MgO, in particular molecular vs dissociative adsorption. In this case study, MgO(100) was grown on Ag(100) by vapor deposition following the method reported by Wollschläger et al.³⁶ Magnesium was deposited at a rate of ~0.1 nm min⁻¹ in the presence of 10⁻⁶ torr O₂ while maintaining the Ag(100) substrate at 300 °C. The interaction of water with the MgO(100) film surface was then studied by monitoring the O 1s and Ag 3d spectra in isobar experiments, where the water vapor pressure was kept constant while the sample temperature was decreased from well over 300 °C to below room temperature.³⁷ Figure 4.13 shows O 1s, Ag 3d, as well as C 1s spectra taken in a 0.15 torr isobar experiment. The O 1s spectrum exhibits four peaks assigned to (from low to high binding energy) oxide, hydroxide, adsorbed molecular water, and water vapor (the gas phase in front of the sample is also partly detected in the experiment). A comparison of the O 1s spectrum at a relative humidity of 0.008% with the one at 6% shows major changes in the MgO film: there is a strong increase of hydroxide uptake as well as water adsorption at higher humidity. The Ag 3d spectra do not show any change (apart from attenuation), which indicates that the MgO film is covering the whole of the Ag(100) substrate, thus preventing a reaction between water vapor and Ag substrate. In addition, C 1s spectra do show a slight increase in carbonaceous species, which is expected under ambient measurement conditions.³⁴

The O 1s oxide, hydroxide and adsorbed water peak intensity can be converted into equivalent monolayer coverages using a procedure (see Newberg et al. 37) that analyzes both the decrease in Ag 3d signal due to the adsorption of OH and H₂O, as well as the relative O 1s signals from the film. The results are plotted in Fig. 4.14. There is a sharp onset of surface hydroxylation at a relative humidity of ~0.01%, accompanied by a reduction in oxide coverage. Water adsorption increases notably at this relative humidity, too. A detailed analysis of the data leads to a model for the reaction of water vapor with MgO(100) under ambient relative humidity, depicted in Fig. 4.15. At a critical relative humidity (~0.01%), water molecules dissociate on the MgO(100) surface, leading to the hydroxylation of the oxygen atoms in the topmost MgO layer, and to the addition of a monolayer of OH groups to the surface. The validity of this reaction mechanism is borne out in the reduction of the total MgO(100) film thickness and the increase in the total O 1s surface signal. This underscores the utility of the application of thin films for the study of surface reactions, as depicted in Fig. 4.11.

4.4 In situ measurements of buried interfaces using high kinetic energy XPS (HAXPES)

In the final part of this chapter we would like to illuminate another promising XPS technique for the in situ study of thin film properties, namely high kinetic energy XPS. This technique has been pioneered by Lindau et al. who demonstrated in 1974 the use of high-energy photons to measure the intrinsic line width of Au 4f core.³⁸ The attraction of high kinetic energy XPS lies in the ability to measure buried interfaces at much larger depths than in conventional soft X-ray XPS (close to 10 nm at 10 keV in the case of Cu, as opposed to close to 1 nm at 1 keV – see Fig. 4.2). However, the detrimental aspects of this technique are the low photoemission cross-sections at high kinetic energies (several orders of magnitude lower than at low kinetic energies) as well as the need for high voltages on the lens elements for sufficient deceleration of electrons for high energy resolution measurements. With the recent availability of high-brightness synchrotron sources the low cross-sections can be compensated for. Figure 4.16 shows recent results for buried SiO₂ layers on a Si substrate, which are covered by a NiGe overlayer. Using incident photons with an energy of 7935 eV the authors recorded Si 1s spectra (binding energy ~1840 eV) with electron kinetic energies of over 6000 eV, which allowed probing at a depth of more than 15 nm.³⁹ In another case study the smoothness of the solid/solid interface of buried Ni/Cu interfaces was probed using high kinetic energy XPS.⁴⁰ Figure 4.17 (upper panel) shows the layout of the sample. Alternate layers of nickel and copper (5 ML each) were deposited on a MgO(100) substrate that was covered by intermediate layers of Fe, Pt, Cu and Ni. This structure was covered by a 1 nm thick Pt cap layer. High kinetic energy (~1200 eV) XPS spectra with a resolution of ~0.25 eV were taken of the Cu 2p core level (bottom panel of Fig. 4.17). Upon heating of the sample from room temperature to 300 °C a marked shift in the Cu 3p_3/2 peak position to lower binding energy was observed, which was interpreted as a signature for the smoothing of the Cu/Ni interface.⁴¹ These results demonstrate that XPS is not only a valuable tool for the study of the sample surface or near-surface region, but also for investigation of solid/solid interfaces up to a depth of several nanometers.

4.5 Conclusions

In summary, X-ray photoelectron spectroscopy is an excellent tool for the investigation of in situ film growth, reaction, as well as the study of buried interfaces. Recent developments of in situ XPS spectrometers as well as high-brilliance synchrotron facilities have expanded the application of XPS to higher pressures as well as larger sampling depths. The challenge for the future lies in the adaption of XPS instruments to realistic environments of thin film growth as well as gas/solid interactions.

4.6 Acknowledgments

The contributions in particular of D. E. Starr, J. T. Newberg, E. R. Mysak, and K. R. Wilson are gratefully acknowledged. This research, the ALS, and the ALS-MES beamline 11.0.2 are supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences of the US Department of Energy at the Lawrence Berkeley National Laboratory under Contract No. DE-AC02-05CH11231.

4.7 References