Introductory guide to the application of XPS to epitaxial films and heterostructures

Epitaxial films and heterostructures are important in many areas of technology and are the subject of widespread research activity. X-ray photoelectron spectroscopy (XPS) is an important tool for obtaining information about these films and is especially powerful when used as an in situ probe. In this context, in situ means that the XPS capability is in the same ultrahigh vacuum environment as the deposition tool but not necessarily in the same chamber. While the most definitive studies done to date involve in situ application of XPS, much useful information can also be gleaned by using XPS as an ex situ probe.

Many important uses of XPS for analysis of epitaxial thin films are common to those of other materials systems as well. These include determining the chemical states, identifying the presence of impurities on or near the surface, and measuring layer thicknesses. However, the epitaxial and very thin nature of many of these films, along with the presence of one or more solid/solid interface, impacts the nature of the XPS signals and limits the usefulness of some of the standard analysis approaches, which usually assume a homogenous material within the analysis depth and do not consider diffraction effects.^1,2 Nevertheless, the ways in which the detailed nature of these films impact the signals can be turned around to access the richness in XPS data that are generally ignored during routine measurements and analyses. These include direct measurement of the built-in electrostatic potentials and band offsets that result from the heterojunction formation during film growth, as well as the determination of local structural environments of specific elements by means of x-ray photoelectron diffraction.

This article is part of a collection of guides and tutorials intended to provide a basic understanding of important topics involving the application of XPS.³ The intent of this guide is to highlight some of the ways in which the properties of heterostructures influence XPS spectra along with the challenges they present to the analysis. Also highlighted are the opportunities these challenges provide for gaining insights into these structures that are often missed when using standard analysis methods. Theoretical understanding and computer models of the complex physics of photoemission and photoelectron propagation in solids enable XPS to provide a level of information of which many practitioners are unaware. Although extracting the highest level of information may require experimental care and detailed modeling, knowledge of the effects, how they impact the data, and what information might be obtained are useful for all XPS analysts. More details will be found in the referenced literature and in a chapter on XPS in the book Materials Characterization Methods of Epitaxial Films and Heterostructures, which examines these topics in much greater detail than attempted in this short guide.⁴ 

In this guide, we describe how the nature of epitaxial thin films and heterostructures impact XPS spectra and their interpretation in three areas: (i) signal intensity and compositional analysis, (ii) spectral shape and measured binding energies, and (iii) the importance of built-in potentials at interfaces. Although the topics discussed are highly relevant to epitaxial films and heterostructures, aspects of the discussion are also important for XPS of single crystals, advanced complex materials, and other types of thin films.

The standard equations for quantification of material composition using XPS effectively assume that the elemental distribution is uniform over the analysis volume. Thus, in effect, the atom number density for a given element is assumed to be constant over the probe depth. If a structure is known to consist of a single film on a substrate with an interface at a depth less than the XPS probe depth, other analysis equations may be used as summarized by Shard.² However, neither approach is accurate for multilayer epitaxial heterostructures, which may have many layers within the XPS sampling depth as well as electron scattering effects related to the highly order structures.⁵ Thus, there are two important ways in which the constant density and the thin overlayer assumptions fail for epitaxial films, both related to specific impacts of the structural forms and elemental distributions therein. Section II A describes the effect of crystallographic order, which introduces orientation effects on signal intensities. Section II B summarizes an approach to determining the composition of complex materials, which avoids the use of sensitivity factors and deals with epitaxial samples for which multiple layers my exist within the XPS probe depth.

The XPS probe depth is typically defined as some multiple of the electron attenuation length (λ) for the peak in question. In the simplest model that takes into account only isotropic inelastic scattering, the intensity falls off as exp(−t/λ cos θ), where t is the film thickness and θ is the electron exit angle relative to the surface normal. Since 99% of the intensity originates over a depth of 4.6λ, and most XPS systems with monochromatic x-ray sources are sensitive to the ∼1% level, 4.6λ is a useful working definition of the probe depth. However, for spectrometers without x-ray monochromators which thus exhibit inherently higher background levels, 3λ, which encompasses 95% of the total intensity, may be more appropriate.

The electron attenuation length and, therefore, the probe depth vary with electron kinetic energy (E_k) as $E k n$ where n is material specific but typically ranges from ∼0.7 to ∼0.8 for most inorganic materials and E_k values in excess of a few hundred eV. Attenuation lengths can be estimated from fundamental material properties using databases generated and maintained by NIST.^6,7 More accurate values for a particular material can be measured by depositing epitaxial films of that material with different thickness on a substrate not containing the elements in the film, as described in detail elsewhere.⁸ 

To provide some sense of the analysis depth and relation to layered structures, the ordered layered structure of SrTiO₃ (STO) is shown in Fig. 1. The spacing between the Ti and Sr layers is roughly 0.2 nm (or a unit cell of ≈0.4 nm), and the electron attenuation length λ for the Ti 2p electron generated using an Al_kα x-ray source is roughly 1.5 nm.⁸ Using an analysis depth of 4.6λ, an XPS signal would come from roughly 1.5 nm * 4.6/0.2 nm ≈ 40 layers, half containing Sr and half containing Ti. As noted above, the signal from the deepest layers would be very small. Differences between the quantification of such materials using the “uniform” composition and considering actual elemental distribution are discussed in Secs. II A and II B.

In addition to the unknowns associated with several aspects of standard quantitative analysis,^1,2 the assumption that the distribution of elements can be described by the same density distribution for all elements is not appropriate or accurate for complex epitaxial materials. For instance, consider materials consisting of sublattices containing different elements that alternate with the film depth. Cubic perovskites oriented in the (001) direction, such as STO(001) as shown in Fig. 1, are one example of such materials. For STO(001), the terminating layer can be either TiO₂ or SrO. As pointed out by Chambers and Sushko⁹ with results shown in Fig. 2(a), equations appropriate for materials with a uniform distribution of elements may be replaced by sums over sublattices for the different elements in each layer of the film that enables differences in signal intensities for TiO₂ and SrO terminated films to be calculated.

However, simple layer-specific sums that take into account isotropic inelastic scattering do not account for photoelectron diffraction effects that occur in epitaxial and single-crystal materials, as seen in Fig. 2(b). As Chambers notes,⁴ there is a significant modulation of the photoelectron intensities with the exit angle due to elastic scattering and interference of outgoing photoelectron and Auger electron waves for epitaxial films and bulk single crystals.^5,10–12 Experimental data for the various elements in an STO(001) crystal are shown as a function of the polar angle in Fig. 2(b) to illustrate this point. As can be seen, diffraction effects can be significant. A useful feature is the so-called “forward focusing” peak, which results from Coulombic attraction between outgoing photoelectrons and ion cores along the exit path. This effect leads to large scattering amplitudes and small scattering phase shifts for low scattering angles, resulting in zeroth-order intensity maxima (see the inset in Fig. 2). Such features are particularly notable along close-packed, low-index directions such as [001] and [011] in the perovskite lattice and serve as a convenient means of orienting the crystal. Higher-order diffraction features are also generated, with constructive interference occurring at emission angles not corresponding to low-index directions (see the inset). Most of these data were collected with a full angular acceptance angle of 14° in the analyzer. However, in normal operation, many modern analyzers have angular acceptances ranging from 30° to 60° and these partially smooth out diffraction effects, as seen by comparing the O 1s scans for acceptance angles of 14° and 30° in Fig. 2(b). Diffraction effects can impact the quantitative analysis and orientation effects and need to be considered in data acquisition and analysis. Using larger acceptance angles and consistently using a well-defined emission direction can yield consistent measurements. See the study by Chambers and Du⁸ for a more detailed discussion of the interplay between elastic and inelastic scattering and its impact on the quantitative XPS analysis. Although diffraction modulation can be a complication for some types of analyses, it can also be used to extract important structural information. Fitting angular distributions of Auger and photoelectron signals with model scattering calculations enables surface and epitaxial film structures to be determined at the atomistic level.^5,10–12

By using this method, it is essential that the spectra be measured with the same analyzer and sample orientation conditions and that the peak areas be determined in a consistent manner including background removal. Based on the experience with multiple materials and growth conditions, it is estimated that this approach can be used to verify film stoichiometry to well within ∼10%.

Often the objective of growing epitaxial films is to synthesize new materials or material combinations. It is important to recognize that photoelectron peak shapes for such materials may deviate from the patterns that many XPS analysts have come to expect for more common materials, particularly when they involve transition metal cations with unpaired d electrons in the valence band. It is most useful for analysts to recognize that many-body and final-state effects can be important and have a significant impact on peak structure and shape.^4,15 Too often in the literature, unexpected peak features have been incorrectly identified as new or additional chemical states.¹⁶ The series of Ti 2p spectra shown in Fig. 3 illustrates this effect. Figures 3(a), 3(c), and 3(d) show a progression of measured binding energies that might be expected when increasing the oxidation state from 0 to 3+ to 4+.^17,18 However, the Ti formal charge for the material shown in Fig. 3(b) is also 3+, but this material has fundamentally different crystallographic and electronic structures than that for the material in Fig. 3(c).^17,19 Such examples highlight the importance of reference materials and careful examination of relevant spectra from the literature.

Interfaces play a dominant role in the properties of many epitaxial heterostructure materials systems. In addition to the important uses of XPS to determine the composition and chemical state changes in epitaxial films and heterostructures, it is possible to obtain information about electronic structure at interfaces, which is critical for many potential applications. Significantly, XPS has been fruitfully used to monitor variations in the electrostatic potential with depth, making it possible to measure things such as band bending, heterojunction band discontinuities, and Schottky barrier heights.

From the 1980s, several groups used soft x-ray photoemission at synchrotron radiation facilities to investigate changes in band bending that occur at compound semiconductor surfaces upon submonolayer deposition of metals.^20–22 At the same time, Schottky barrier height formation and interface chemistry were being explored as metal films in the one-to-several monolayer range were deposited and probed with both synchrotron and lab-based x-ray sources.^23–52 Although most of these studies involved disordered metal films deposited at room temperature, others utilized metals that were lattice matched to the semiconductor at some level and deposited at elevated temperature, leading to heteroepitaxial growth.^44,53–58 Other groups combined III–V compound semiconductor MBE capability with in situ XPS to directly measure valence and conduction band offsets for a wide range of semiconductor heterojunctions.^59–74 In what follows, we present these methods and illustrate with a recent example from the world of complex oxide heterojunctions.

Electrostatic potentials present at interfaces include built-in potentials such as those that occur at p-n junctions and band bending due to charge accumulation near surfaces or interfaces. The series of energy diagrams in Fig. 4 show separated materials [panel (a)], materials in contact but without Fermi-level equilibration [panel (b)], and materials in contact with Fermi-level equilibration and the establishment of a space-charge region at the interface [panel (c)]. The situation that occurs when one of the materials is a thin layer is illustrated in Fig. 4(d). In this last case, there is charge accumulation and band bending at the interface and at the surface of the film. The variation in potentials caused by charge accumulation at interfaces and surfaces can alter XPS peak shapes and shift peak energies. When such potential gradients exist, photoelectrons emitted in each layer originate from a slightly different potential and thus appear at shifted binding energies and with asymmetrically broadened line shapes.

To provide specific examples of these effects, we consider data from measurements for p-La_0.88Sr_0.12FeO₃ on n-SrTi_0.99Nb_0.01O₃(001) [LSFO/Nb:STO] by Wang et al.¹⁴ These authors investigated the built-in potential, band alignment, and electrocatalytic activity of this heterojunction. The built-in potential that forms at a p-n junction would be useful for diverting photogenerated electrons into Nb:STO and photogenerated holes into the LSFO thin film to drive water oxidation. These interface potentials are “built-in” as a result of the material structure and are not externally applied. Epitaxial films of 3, 5, 9, and 35 unit cell (UC) thicknesses of LSFO were grown on Nb:STO by oxygen plasma assisted molecular beam epitaxy. This work demonstrates that the core-level binding energy differences across the buried interface, obtained by fitting the spectra in the usual way as has traditionally been used to determine valence band offsets (VBOs), lead to incorrect results because of the built-in potential in the film.

This material system is particularly interesting in that Sr is present in both the film and substrate, and some Ti diffuses out into the film from the substrate. Significantly, the differences in binding energy between the film and substrate species for both Sr 3d and Ti 2p are due at least in part to the difference in electrostatic potential between the two materials. Spectra for three elements are shown in Fig. 5 for three different LSFO thicknesses. There are several interesting features that emerge as the film thickness increases: (i) the Ti 2p peak shape remains essentially constant, but is different than that of pure Ti⁴⁺ in STO; (ii) a second set of Sr 3d peaks appears at a lower binding energy; and (iii) the width and position of the La 4d peak changes with film thickness.

As discussed by Wang et al.¹⁴ the Sr 3d spectra were fit with two pairs of spin–orbit doublet peaks. The most intense signals were assigned to Sr⁺² in the substrate and the smaller peaks to Sr⁺² in the LSFO. The binding energy differences are attributed to the band discontinuity at the interface and built-in potential in the film. Although the Ti 2p spectrum seems little changed with LSFO thickness, some Ti diffuses into the LSFO producing a second small peak at an energy consistent with the presence of Ti³⁺ in the film, which, in turn, has a different potential than Nb:STO. The absence of broadening in the Ti 2p⁴⁺ spectrum suggests that no built-in potential of any significant magnitude is present in Nb:STO. The variation in La 4d peak width is attributed to the presence of a large built-in potential in the thin film as discussed in additional detail below. This dataset from the LSFO/STO system demonstrates the impact of electrostatic potential at interfaces to shift photoelectron spectra and provides the basis for a method to extract useful information about the built-in potential.

Here, the terms (E_CL–E_V) are the binding energy differences between the chosen core level and the valence band maximum in each of the two materials as pure phases, and ΔE_CL is the core-level binding energy difference across the heterojunction (see Fig. 4).

In developing this method, it was implicitly assumed by Kraut et al. that there is no measurable band bending on either side of the interface and, therefore, that a single core-level binding energy is representative of all layers in each material. With this assumption, and the assumptions of an atomically abrupt junction and no interface chemistry (leading to a unique phase not found in the bulk spectra), the band alignment can be determined using Eqs. (2) and (3). However, the first of these assumptions is not valid for the LSFO/Nb:STO system because of the presence of the sizeable built-in potential within the LSFO film.

This potential drop across the LSFO film was detected not only by core-level broadening but also by changes in the x-ray excited valence band spectra, specifically the dependence of the valence band maximum (VBM, E_V) on the thickness. Chambers et al.⁷⁷ examined three approaches for accurately determining the VBM and found that extrapolation of the linear portion of the leading edge to the energy axis is valid for many systems. This method was applied to the VB spectra for the LSFO/Nb:STO heterojunctions in the Wang et al. study, as shown in Fig. 6. The E_V values from Nb:STO and the 35 UC film were used for materials 1 and 2 in Eq. (2) above. It is then possible to determine ΔE_V values for several different pairs of core photoelectron lines, and the results are quite consistent, as shown in Table I. However, as noted in Ref. 14, these values are incorrect because they are based on core-level binding energies averaged over all layers when, in fact, they need to be based on binding energies for the layers directly at the interface because of the presence of the built-in potential.

In order to find the binding energies for La in the interfacial layers, the composite La 4d spectra in Fig. 5 were modeled as sums over layers across which a built-in potential is present. Such models are shown in Fig. 7 for 3, 6, and 9 UC films. The nearly flatband spectrum measured for the 35 UC film was assigned as a “basis spectrum” for each layer in the thinner films. The relative intensities were attenuated based on the depths of the various layers, and the energy shifts between layers were optimized to yield the best fit to the measured heterojunction spectrum. The potential drop per layer is relatively small for the 3 UC film, is significantly larger for the 5 UC film, and then decreases in magnitude as the film gets thicker. The conduction band offsets are determined from Eq. (3) using values for the bandgaps obtained from other types of measurements.^14,78 Using the binding energies of the interfacial layers for all three heterojunctions, the resulting energy diagram is quite different than the flatband picture that emerges using the average La 4d_5/2 binding energies shown in Table I. The final energy diagram is shown in Fig. 8. Instead of a constant VBO of 1.8 eV for heterojunctions consisting of 3, 5, and 9 UC thick films with flatbands throughout, we see upward band bending with increasing film thickness, culminating in the VBM being quite close to the Fermi level at t ≥ 9 UC, consistent with the VB spectra shown in Fig. 6.

The above analysis was carried out by manually modeling linear potential drops with either one (3 and 5 UC) or two (9 UC) segments across the film. However, a sophisticated algorithm that performs a comprehensive search over all binding energies for the best fit set of layer-resolved spectra has been recently developed.^79,80 This approach is particularly useful for analyzing hard x-ray XPS data for which many more layers contribute to the total spectrum.

The use of XPS to obtain electronic structure information for heterojunctions is much less common than measurements of composition and chemical bonding but is nevertheless quite important. Indeed, XPS, when properly interpreted, yields electronic structure information not readily available by any other technique. The discussion here has focused on measurements of complex oxide heterostructures. A creative application of XPS has also been used to extract important information that is difficult to obtain information such as dielectric, piezoelectric, and ferroelectric properties of films, along with the electric potential of selected layers in inorganic films and potential profiles across organic films.^81–84 

XPS is an important analysis tool for epitaxial thin films. In addition to what might be identified as standard XPS analysis to yield information about contamination, average elemental composition, and chemical states present, XPS can also be used to extract important information related to the crystallographic structure, layer stoichiometry, and band edge profiles. The structural order in these films adds a complication to the analysis that requires appropriate experimental planning and deeper levels of data analysis.

The less commonly utilized advantages of XPS to obtain information about the electronic properties of films and interfaces have been highlighted. An example of obtaining information about built-in potentials and band offsets demonstrates how such information can be obtained.

Although each of the uses of XPS described in this guide is relevant to epitaxial films and heterostructures, aspects of the methods presented are relevant to other single-crystal materials, advanced complex materials and structures, and electronic characteristics of other types of thin overlayers.

This work was supported by the U.S. Department of Energy, Office of Science, Division of Materials Sciences and Engineering under Award No. 10122 and was performed in the Environmental Molecular Sciences Laboratory (Grid. No. 436923.9), a national scientific user facility sponsored by the Department of Energy's Office of Biological and Environmental Research and located at the PNNL.

Data sharing is not applicable to this article as no new data were created or analyzed in this study.