Measuring contaminants

Epitaxial thin film research has a very low tolerance for elemental contamination. Despite engineering controls, contaminants from unknown sources sometimes make their way into samples and can either hinder data acquisition in experiments or change the physical properties of the material entirely. Surface-sensitive XPS is an excellent way to detect any unwanted element contamination, with an elemental sensitivity of <1 atomic percent. The most efficient way to measure contamination is by acquiring wide-range surveys of samples from maximum binding energy to zero. In general, it is good practice to perform surveys on every sample before performing a longer-duration, high-resolution measurement over core level and valence band regions. XPS surveys allow for a quick qualitative depiction of a sample's surface stoichiometry in addition to indicating any unexpected elemental contaminants.

As an example, Fig. 5 is a survey of a CoMn₂O₄ spinel thin film sample grown in our MBE system. There are many peak features in the survey representing cobalt, manganese, and oxygen signals, but additional peaks indicate a contaminating element. Based on the position and shape of these extra features, we determined the presence of sodium on the sample surface. By inspecting surveys of samples grown at different conditions, we pinpointed the source of the sodium to the radio-frequency oxygen plasma source. After discussions with the manufacturer, the Na-contaminated quartz discharge tube was replaced with an alumina discharge tube, which has eliminated contamination. We are aware of at least one other plasma source that has also been a source of Na contamination and suggest that groups should be careful to test new plasma sources upon purchase. Given that the design employed by the manufacturer (Mantis Deposition) has been ubiquitous in the field for many years, we suspect that other groups without access to in situ XPS may have Na contamination in films without realizing. Other contaminants can be detected by XPS, such as F due to surface treatments performed on SrTiO₃ with a buffered oxide etchant [Reference Chambers, Droubay, Capan and Sun79]. In this sense, the availability of in situ XPS can accelerate the calibration process for ideal growth conditions and pristine materials, making it valuable as a day-to-day diagnostic tool even when the data will not be used for publishable research.

Figure 5: XPS survey of CoMn₂O₄ thin film showing Na contamination from oxygen plasma source.

Film stoichiometry and surface termination

In practice, the depth sensitivity of XPS can be both a blessing and a curse. The angle-resolved XPS technique can be used to measure the surface termination of a sample, cation intermixing across an interface, and determine a depth profile of specific chemical features, providing rapid new depth-resolved insights into synthesized materials without destroying the sample through sputter etching or focused ion beam liftout for electron microscopy measurements. Conversely, the surface sensitivity of XPS means that a change as small as flipping the surface termination from TiO₂ to SrO on a SrTiO₃ single crystal can reduce the measured Ti:Sr peak area ratios by ~10%. This makes stoichiometry quantification for epitaxial thin films by XPS very difficult, with conventional sensitivity factors such as those used in XPS data analysis software providing very little value. Further consideration must be given to the effects of X-ray photoelectron diffraction, which can change relative peak intensities due to forward scattering of photoemitted electrons from the single crystal surface. It is generally best to benchmark the peak area ratios with a bulk-sensitive technique such as Rutherford back scattering on a film grown to produce a known surface termination. Alternatively, one can measure a single crystal reference with a known surface termination such as a treated SrTiO₃ substrate [Reference Kawasaki, Takahashi, Maeda, Tsuchiya, Shinohara, Ishiyama, Yonezawa, Yoshimoto and Koinuma80]. After these initial tests, is it then possible to accurately compare peak area ratios for subsequent films to the measured area ratios from the calibrated reference sample while taking care to only compare samples with the same surface terminations.

However, despite the challenges of absolute stoichiometry quantification via XPS, we have found a convenient trick that employs in situ capabilities to accelerate the calibration process for oxide MBE growth. Given the layered nature of the perovskite crystal structure and the propensity for excess cations to reside on the film surface [Reference Lee, Luo, Tung, Chang, Luo, Malshe, Gadre, Bhattacharya, Nakhmanson, Eastman, Hong, Jellinek, Morgan, Fong and Freeland81, Reference Nie, Zhu, Lee, Kourkoutis, Mundy, Junquera, Ghosez, Baek, Sung, Xi, Shen, Muller and Schlom82], using angle-resolved XPS to determine the surface termination is generally an effective means to determine whether a perovskite oxide film has excess A-site or B-site cations even without an RBS or single crystal reference standard. The segregation is shown schematically in Figs. 6(a) and 6(b), with excess TiO₂ and SrO residing on the film surface. Note that this figure is meant to schematically portray clustering of excess atoms on the film surface and not a particular surface reconstruction that might occur in such a case. The angle-resolved XPS measurement process involves comparing peak area ratios for the A-site ion to the B-site ion in a bulk-sensitive configuration near-normal emission (θ < 45°) and in a surface-sensitive configuration (θ > 70°). Such measurements can generally be completed in under an hour, making it possible to estimate whether a sample is A-site or B-site rich and recalibrate prior to additional depositions. For example, in the case of SrTiO₃ films, we have seen that the ratio of the Ti 2p to the Sr 3d peak areas increases for larger values of θ if the sample is Ti-rich [Reference Thapa, Provence, Jin, Lapano, Sadowski, Brahlek and Comes83]. A comparison of one such sample is shown in Fig. 6(c). In this case, the bulk and surface-sensitive Sr 3d and Ti 2p peaks have been normalized so that the Ti 2p peaks show the same total peak area, which highlights the smaller peak area for the surface-sensitive Sr 3d peak relative to its bulk counterpart. An additional source of information is the presence of surface core-level shifts that can be measured due to incomplete coordination or OH adsorption of surface A-site ions, as has been shown for Ba-terminated BaTiO₃, which shows a distinct peak at higher binding energy for the Ba 3d peak [Reference Berlich, Strauss, Langheinrich, Chassé and Morgner84]. Using these approaches while understanding the kinetics of the growth process can make in situ XPS particularly valuable for quick tuning of growth parameters sample-to-sample.

Figure 6: (a) Model of excess TiO₂ on the film surface of TiO₂-terminated SrTiO₃, (b) model of excess SrO on the film surface of SrO-terminated SrTiO₃, and (c) angle-resolved XPS measurement of off-stoichiometric SrTiO₃ film with excess Ti cations at the film surface.

Beyond the convenience of angle-resolved XPS in the synthesis process, it has helped to explain surface-dependent phenomena in a variety of systems. We have employed angle-resolved XPS to determine for the first time that an SrO termination is more stable for stoichiometric SrTiO₃ films grown by hybrid MBE [Reference Thapa, Provence, Jin, Lapano, Sadowski, Brahlek and Comes83]. Furthermore, in the case of a LaAlO₃/SrTiO₃ heterostructure grown on SrO-terminated SrTiO₃, an unexpected AlO₂ termination was observed by angle-resolved XPS, which helped to explain the absence of conductivity at the interface [Reference Singh-Bhalla, Rossen, Pálsson, Mecklenburg, Orvis, Das, Tang, Suresha, Yi, Dasgupta, Doenning, Ruiz, Yadav, Trassin, Heron, Fadley, Pentcheva, Ravichandran and Ramesh85]. Angle-resolved measurements of LaFeO₃ film surfaces have also helped to explain the role of surface termination on chemical reactivity for catalytic water splitting [Reference Stoerzinger, Comes, Spurgeon, Thevuthasan, Ihm, Crumlin and Chambers86]. Despite the simplicity of angle-resolved XPS and the relatively mundane knowledge gleaned from knowing the surface termination of a thin film, there is a great deal of new physical understanding that can arise from such day-to-day measurements through in situ XPS.

Oxide surface chemistry

While understanding of the surface termination regarding perovskite AO or BO₂ layers is valuable for film synthesis, it only begins to address the subtleties of surface chemistry that occur in complex oxides. The measurement of the O 1s peak can also provide valuable information to understand how the surface is passivated, a process that occurs in different ways depending on the surface polarity. It is known that water will adsorb differently depending on the AO or BO₂ termination of a (001)-oriented surface [Reference Thapa, Provence, Jin, Lapano, Sadowski, Brahlek and Comes83, Reference Stoerzinger, Comes, Spurgeon, Thevuthasan, Ihm, Crumlin and Chambers86]. However, even in UHV after in situ growth, the adsorption of water on the surface can impact the electronic properties of the material. Studies that integrate XPS with atomically resolved STM are incredibly valuable in this regard to demystify some of the surface phenomena that drive emergent behavior.

As discussed above, the KTaO₃ surface 2DEG is of significant interest for the Rashba splitting that is observed [Reference King, He, Eknapakul, Buaphet, Mo, Kaneko, Harashima, Hikita, Bahramy, Bell, Hussain, Tokura, Shen, Hwang, Baumberger and Meevasana33]. Through STM and XPS studies, Setvin et al. demonstrated a 2 × 1 surface reconstruction of the surface with uniform K(OH)₂ coverage after water exposure [Reference Setvin, Reticcioli, Poelzleitner, Hulva, Schmid, Boatner, Franchini and Diebold38]. An OH peak on the O 1s core level at higher binding energy was reported for this case to confirm their results. Similar experiments have also been performed on layered ruthenates (Ca₃Ru₂O₇ [Reference Halwidl, Mayr-Schmölzer, Fobes, Peng, Mao, Schmid, Mittendorfer, Redinger and Diebold87] and Sr₂RuO₄ [Reference Halwidl, Stöger, Mayr-Schmölzer, Pavelec, Fobes, Peng, Mao, Parkinson, Schmid, Mittendorfer, Redinger and Diebold88]) and show that water can adsorb onto the surfaces of oxides in a variety of different configurations with different binding energies relative to the metal oxide O 1s peak that is typically at ~530.0 eV. It is generally observed that films with AO termination will more readily adsorb water [Reference Setvin, Reticcioli, Poelzleitner, Hulva, Schmid, Boatner, Franchini and Diebold38, Reference Thapa, Provence, Jin, Lapano, Sadowski, Brahlek and Comes83, Reference Stoerzinger, Comes, Spurgeon, Thevuthasan, Ihm, Crumlin and Chambers86, Reference Halwidl, Stöger, Mayr-Schmölzer, Pavelec, Fobes, Peng, Mao, Parkinson, Schmid, Mittendorfer, Redinger and Diebold88], though the exact surface structure is very difficult to decipher. The effect of the adsorbed water—even after in situ growth—should not be discounted when analyzing the observed properties. For example, it is common in the literature to associate a higher binding energy peak 2.3 eV above the metal oxide binding energy with oxygen vacancies, but the origin of this peak is the subject of some controversy and it can easily be due to other surface chemistry effects that will complicate the analysis [Reference Stoerzinger, Comes, Spurgeon, Thevuthasan, Ihm, Crumlin and Chambers86].

Interfacial chemistry and intermixing

The surface sensitivity of XPS through angle-resolved measurements can be extremely useful for the purposes of determining depth-dependent phenomena in oxide heterostructures. We have already showed the effects of surface variations and how they can be used to understand oxide film growth. However, the same techniques can be taken a step further to understand such behavior as interfacial intermixing in heterostructures and growth-induced defects such as oxygen vacancies. By varying the photoelectron emission angle to the detector and modeling the angular dependence of the relative peak areas based on the IMFP for the sample, one can construct a quantitative model of the cation or defect profile as a function of depth within a material.

For a buried interface, angle-resolved measurements can be used to model the depth of defects or dopants. The approach has been particularly valuable in studies of polar/non-polar interfaces, where differences between the idealized physical model of an interface and the actual synthesized heterostructure have led to confusion as to the physical origin of emergent behavior. In the case of LaAlO₃/SrTiO₃ interfaces, angle-resolved measurements have shown that cations from the underlying SrTiO₃ substrate will readily out-diffuse into the LaAlO₃ film, with the degree of diffusion dependent on the cation stoichiometry of the LaAlO₃ film [Reference Chambers, Engelhard, Shutthanandan, Zhu, Droubay, Qiao, Sushko, Feng, Lee, Gustafsson, Garfunkel, Shah, Zuo and Ramasse19, Reference Qiao, Droubay, Varga, Bowden, Shutthanandan, Zhu, Kaspar and Chambers89]. Through rigorous Monte Carlo modeling of the angle-resolved data, one can even estimate the cation concentration profile across the interface in a heterostructure, as Salvinelli and colleagues have demonstrated [Reference Salvinelli, Drera, Giampietri and Sangaletti90]. These measurements set the stage for further studies that confirmed that the 2DEG at these interfaces is dependent on the film stoichiometry [Reference Warusawithana, Richter, Mundy, Roy, Ludwig, Paetel, Heeg, Pawlicki, Kourkoutis, Zheng, Lee, Mulcahy, Zander, Zhu, Schubert, Eckstein, Muller, Hellberg, Mannhart and Schlom20, Reference Breckenfeld, Bronn, Karthik, Damodaran, Lee, Mason and Martin21] and thus driven in part by the formation of interfacial defects such as oxygen vacancies. Angle-resolved measurements can also be used to model the depth of distinct chemical and electronic states, such as the Ti³⁺ states at the LaAlO₃/SrTiO₃ interface. By probing the Ti 2p core level of samples using ex situ hard X-ray photoemission, Sing et al. developed a model of the thickness of the 2DEG in UCs to show that it was confined to a single UC at the surface of the SrTiO₃ substrate [Reference Sing, Berner, Goß, Müller, Ruff, Wetscherek, Thiel, Mannhart, Pauli, Schneider, Willmott, Gorgoi, Schäfers and Claessen91]. The results from this work are shown in Fig. 7. Other angle-resolved studies have shown that oxygen vacancies in SrTiO₃-based heterostructures are localized at the interface [Reference Chen, Bovet, Trier, Christensen, Qu, Andersen, Kasama, Zhang, Giraud, Dufouleur, Jespersen, Sun, Smith, Nygård, Lu, Büchner, Shen, Linderoth and Pryds92, Reference Saghayezhian, Wang, Guo, Zhu, Plummer and Zhang93].

Figure 7: Ti 2p core-level data showing Ti³⁺ intensity at the LaAlO₃/SrTiO₃ interface for two different samples (a and b). Variation with angle in (a) shows that Ti³⁺ is non-uniformly distributed in comparison to the uniform distribution in (b). (c) Schematic of angle-resolved XPS measurement of 2DEG location. Reprinted figure with permission from Sing et al. [Reference Sing, Berner, Goß, Müller, Ruff, Wetscherek, Thiel, Mannhart, Pauli, Schneider, Willmott, Gorgoi, Schäfers and Claessen91]. Copyright 2009 by the American Physical Society.

In the case of a 6 UC SrTiO₃–3 UC LaCrO₃ superlattice, in situ angle-resolved XPS measurements were used to determine the degree of Cr out-diffusion into the topmost SrTiO₃ layer [Reference Comes, Spurgeon, Kepaptsoglou, Engelhard, Perea, Kaspar, Ramasse, Sushko and Chambers94]. The resulting data were compared with a model with differing degrees of Cr intermixing to estimate the Ti–Cr concentration profile across the B-site of the superlattice. Interfacial models were constructed to represent differing percentages of B-site cation intermixing and contributions to the overall Cr 2p and Ti 2p peak intensities were estimated by accounting for the IMFP in the material for each layer. The resulting estimates showed that there is significant intermixing of Ti across the interface, with approximately 25% of the middle B-site ions in the LaCrO₃ layer occupied by Ti ions. Critically, the model agreed very well with extracted concentration profiles from STEM-EELS measurements, as shown in Fig. 8 [Reference Comes, Spurgeon, Kepaptsoglou, Engelhard, Perea, Kaspar, Ramasse, Sushko and Chambers94].

Figure 8: (a) Angle-resolved XPS Cr 2p:Ti 2p peak ratio with models assuming various degrees of intermixing (0° is normal to the film surface), (b) concentration profile for the 35%-intermixing model, and (c) STEM-EELS integrated signal profile throughout superlattice determined using MLLS fitting of the Cr L₂₃ edge and the background-subtracted peak area of the Ti L₂₃ edge. The signal has been normalized to the substrate. Reprinted with permission from Comes et al. [Reference Comes, Spurgeon, Kepaptsoglou, Engelhard, Perea, Kaspar, Ramasse, Sushko and Chambers94]. Copyright 2017 American Chemical Society.

Band alignment and potential gradients

Semiconductor heterostructures are key to modern-day electronics, and numerous oxide heterojunctions have been pursued for their potential applications. The performance of these heterostructures can be engineered by manipulating valence and conduction band offset at the interface. It is crucial to understand this band offset to accurately predict and manipulate the behavior of thin film heterostructures.

There are different schemes to determine the band offset experimentally such as ultra-violet spectroscopy, internal photoemission spectroscopy, and XPS. Band offset determination via XPS has been used extensively. Unlike other methods, XPS is sensitive to details at the interface such as changes in valence and chemical intermixing and thus is often preferred over other methods. The XPS approach to valence band alignment determination was first introduced by Kraut et al. [Reference Kraut, Grant, Waldrop and Kowalczyk95]. It is based on the premise that the energy difference between a core level and valence band maximum (VBM) is an intrinsic property and remains constant independent of any formation of a heterostructure. Any change in the binding energy of the VBM will change the core level equally, making the energy of the core-level peak a good proxy for the VBM. This becomes important when measuring a thin film heterostructure, as the valence band region will now be the convolution of two or more distinct materials, while appropriately chosen core levels can still be measured easily and fit repeatably. The precision of this method lies in the measurement of core level and valence band maxima binding energies from reference samples for each material in a heterojunction. This can be accomplished using a single crystal reference or carefully grown film that is sufficiently thick such that the underlying substrate does not contribute any signal to the data. With the two reference samples and one more heterojunctions under examination, at least three samples are needed to apply this approach.

Using their data, Kraut et al. [Reference Kraut, Grant, Waldrop and Kowalczyk95] determined the VBM by fitting valence band spectra with the theoretical valence band density of states. However, fits involving a linear extrapolation of the leading edge of the valence band spectra to the zero-level backgrounds have been found to be as accurate and require much less effort. Various refinements to this approach over the years for oxide heterostructures have been proposed [Reference Chambers, Du, Comes, Spurgeon and Sushko96, Reference Chambers, Droubay, Kaspar and Gutowski97]. A diagram illustrating Kraut's method is shown in Fig. 9. The valence band and conduction band offset are calculated as

$$\eqalign{\Delta E_{\rm v} &= \lpar {E_{{\rm CL}}^{\rm A} -E_{\rm V}^{\rm A} } \rpar -\lpar {E_{{\rm CL}}^{\rm B} -E_{\rm V}^{\rm B} } \rpar -\lpar {E_{{\rm CL}}^{\rm A} -E_{{\rm CL}}^{\rm B}} \rpar, \cr \Delta E_{\rm C} &= E_{\rm g}^{\rm A} -E_{\rm g}^{\rm B} + \Delta E_{\rm V}.}$$

Figure 9: Schematic rendering of method to measure band alignment in XPS via core-level binding energies. E _CL refers to core-level binding energy, E _V refers to the VBM, and E _C refers to the conduction band minimum for materials A and B. By measuring these values for thick films or single crystal references, one can then determine the value of ΔE _V for a heterostructure by measuring the difference E _CL^A – E _CL^B.

The core-level peaks and valence band reference spectra should be acquired with high energy resolution, as the uncertainty of the measurement of the core-level peak position will propagate into the band offset calculation. The choice of peaks that have a narrow intrinsic width and are easier to fit repeatably is thus very important.

The LaAlO₃/SrTiO₃ interface has been widely studied for its unusual transport properties. The properties can be attributed to an effective 2DEG at the interface. The origin of this 2DEG has been the source of great controversy, with some attributing it to the polar discontinuity [Reference Ohtomo and Hwang1, Reference Sing, Berner, Goß, Müller, Ruff, Wetscherek, Thiel, Mannhart, Pauli, Schneider, Willmott, Gorgoi, Schäfers and Claessen91] and others to defect-driven phenomena [Reference Chambers, Engelhard, Shutthanandan, Zhu, Droubay, Qiao, Sushko, Feng, Lee, Gustafsson, Garfunkel, Shah, Zuo and Ramasse19, Reference Warusawithana, Richter, Mundy, Roy, Ludwig, Paetel, Heeg, Pawlicki, Kourkoutis, Zheng, Lee, Mulcahy, Zander, Zhu, Schubert, Eckstein, Muller, Hellberg, Mannhart and Schlom20, Reference Breckenfeld, Bronn, Karthik, Damodaran, Lee, Mason and Martin21, Reference Chambers23]. Owing to its high precision in measuring core levels and VBMs, and subsequently the valence band and conduction band offsets, XPS has been a powerful tool in studying this heterostructure interface. Additionally, XPS data can be modeled to estimate the potential gradient across the film.

Multiple groups have used XPS to study the LaAlO₃/SrTiO₃ interface grown by both MBE [Reference Segal, Ngai, Reiner, Walker and Ahn22] and PLD [Reference Chambers, Engelhard, Shutthanandan, Zhu, Droubay, Qiao, Sushko, Feng, Lee, Gustafsson, Garfunkel, Shah, Zuo and Ramasse19, Reference Qiao, Droubay, Shutthanandan, Zhu, Sushko and Chambers98, Reference Drera, Salvinelli, Brinkman, Huijben, Koster, Hilgenkamp, Rijnders, Visentin and Sangaletti99] with different film thicknesses and terminations for possible explanations for the origin of the interface's 2DEG. In one such work, Segal et al. confirmed the metal–insulator interface transition with 4 UC-thick LaAlO₃ on a TiO₂-terminated surface [Reference Segal, Ngai, Reiner, Walker and Ahn22]. Many studies suggest that the 2DEG at the interface stems from an electronic reconstruction due to polar discontinuity (LAO has alternating positive and negative atomic layers), known as the “polar catastrophe” [Reference Sing, Berner, Goß, Müller, Ruff, Wetscherek, Thiel, Mannhart, Pauli, Schneider, Willmott, Gorgoi, Schäfers and Claessen91]. From the experimental band offset and band gap, Segal et al. calculated that a potential gradient of 1.15 ± 0.06 at a 4 UC thickness is required for the polar catastrophe to occur, whereas the measured potential gradient at 4 UCs was much less than the required value. This means that other phenomena such as oxygen defects, lattice distortion, and cation mixing at the interface must be investigated to explain this unique behavior at the interface. In this way, XPS has helped to further elucidate the origin of 2DEG at the LAO-STO interface, which was later shown to be strongly dependent on LaAlO₃ cation off-stoichiometry [Reference Warusawithana, Richter, Mundy, Roy, Ludwig, Paetel, Heeg, Pawlicki, Kourkoutis, Zheng, Lee, Mulcahy, Zander, Zhu, Schubert, Eckstein, Muller, Hellberg, Mannhart and Schlom20, Reference Breckenfeld, Bronn, Karthik, Damodaran, Lee, Mason and Martin21].

Other studies of interfaces where both sides of the heterojunction are comprised of band insulators with the VBM comprised of O 2p-derived electronic states, have generally shown small valence band offsets of ~0.5 eV or less [Reference Giampietri, Drera and Sangaletti100]. This includes the SrTiO₃/(La,Sr)(Al,Ta)O₃ interface [Reference Comes, Xu, Jalan and Chambers101], the SrZrO₃/SrTiO₃ interface [Reference Schafranek, Baniecki, Ishii, Kotaka, Yamanka and Kurihara102], and the BaSnO₃ interface with both SrTiO₃ and LaAlO₃ [Reference Chambers, Kaspar, Prakash, Haugstad and Jalan103]. The small valence band offset can generally be attributed to the condition that the O 2p energy level is continuous across the heterojunction [Reference Zhong and Hansmann104], which will be discussed in more detail shortly.

Unlike the isoelectronic interfaces where O 2p states form the top of the valence band on both sides of the heterojunction, interfaces between Mott–Hubbard and band insulators offer additional degrees of freedom. For example, the LaFeO₃/n-SrTiO₃ system also creates a polar non-polar interface and is studied for its photocatalytic applications [Reference Nakamura, Kagawa, Tanigaki, Park, Matsuda, Shindo, Tokura and Kawasaki105, Reference Nakamura, Mashiko, Yoshimatsu and Ohtomo106, Reference Comes and Chambers107, Reference Xu, Han, Rice, Jeong, Samant, Mohseni, Meyerheim, Ostanin, Maznichenko, Mertig, Gross, Ernst and Parkin108]. Nakamura et al. [Reference Nakamura, Kagawa, Tanigaki, Park, Matsuda, Shindo, Tokura and Kawasaki105] studied the interface-induced polarization of LaFeO₃/Nb-doped SrTiO₃ heterostructures and observed novel shift current that suggested the polarization direction switches for different substrate terminations (SrO and TiO₂). This would mean that potential gradients took on opposite signs for different terminations and suggested that changes in polarization are responsible for observed photoconductivity. Subsequent studies of LaFeO₃/Nb-doped SrTiO₃ junctions with varying LaFeO₃ film thicknesses (3,6 and 9 UCs) and different substrate terminations (SrO and TiO₂) grown using MBE and characterized by in situ XPS helped to address these questions [Reference Comes and Chambers107]. These studies found an increase in separation between core levels (La 4d and Sr 3d) with increases in film thickness, clearly indicating that band offset and built-in potential change with thickness. In addition, by modeling the broadening of the La 4d peaks with different film thicknesses and terminations, it was possible to estimate the potential gradient across the film. Contrary to the reported interface-induced polarization [Reference Nakamura, Kagawa, Tanigaki, Park, Matsuda, Shindo, Tokura and Kawasaki105, Reference Nakamura, Mashiko, Yoshimatsu and Ohtomo106], the results indicated that the interface termination had only a small impact on the band alignment and the built-in potential gradient [Reference Comes and Chambers107]. The models are shown in Fig. 10. The chemical instability of the SrO/FeO₂ interface has been proposed as a likely explanation for these results [Reference Spurgeon, Sushko, Chambers and Comes109]. These results were subsequently matched by first-principles band alignment models of the interface [Reference Xu, Han, Rice, Jeong, Samant, Mohseni, Meyerheim, Ostanin, Maznichenko, Mertig, Gross, Ernst and Parkin108].

Figure 10: (a) Sr 3d and La 4d core-level spectra for the family of heterostructures, shifted to align the Sr 3d peaks; (b) model of La 4d peak broadening in the 6 u.c. films; (c) Ti 2p core-level spectra for each film and substrate, with the inset showing the peak shifts; (d) valence band offsets determined from the core-level spectra for each heterojunction. Reprinted figure with permission from Comes and Chambers (2016) [Reference Comes and Chambers107]. Copyright 2016 by the American Physical Society.

Interfacial charge transfer

As in modulation-doped semiconductor heterostructures, interfacial charge transfer is an important tool in complex oxides to induce novel magnetic and electronic functionalities that do not occur in equilibrium. One of the crucial parameters to determine the charge transfer probability is the charge transfer gap. In most of the cases, the charge transfer gap is determined by band splitting of B-site cations or the band gap between B-site cations and fully occupied O 2p bands, while A-site cations have a less significant role. Predicting charge transfer is not straight forward in oxide heterostructures since the classic band alignment principle is not strictly followed as in the case of semiconductors. A modified rule is based on the continuity of states of the O 2p band across the interface and allows for a qualitative prediction of band alignments and charge transfer in complex oxide heterostructures [Reference Zhong and Hansmann104]. In cubic oxides, the local energy of O 2p states relative to the Fermi level, ɛ _p−E _F, has been proposed to be the deterministic factor to tune the direction of charge transfer from one complex oxide to another. Generally, electron transfer favors electron flow from lower (more negative) ɛ _p−E _F to higher. The classic semiconductor rule is shown in Fig. 11(a), while the modified rule for oxides is shown in Figs. 11(b)–11(e). This rule led to the development of alternative computational methods to predict charge transfer that have been shown to be effective in recent years.

Figure 11: (a) Conventional model for semiconductor band alignment based on aligning the vacuum level for constituent materials; (b) preliminary band alignment based on the alignment of O 2p states prior to charge transfer; (c) reconstructed band alignment after charge transfer equilibration of the Fermi energy level; (d) schematic model of ABO₃/AB'O₃ interface emphasizing continuity of O 2p electronic states; (e) summary of bulk ɛ_p (filled symbols) and ɛ_d (empty symbols) with respect to the Fermi level (E _F = 0) for different SrBO₃ (solid line) materials, with 3d (black), 4d (red), and 5d (blue) elements. The simple criterion for the direction of the charge transfer at the ABO₃/AB′O₃ interface is that the component with lower (more negative) ɛ_p−E _F will donate electrons across the interface to the opposite material. Also plotted is LaBO₃ (dashed line) for B=3d, for estimates for ABO₃/A′BO₃ interfaces. Adapted under Creative Commons Attribution 3.0 License from Zhong and Hansmann [Reference Zhong and Hansmann104].

Density functional theory (DFT) is one of the most widely used theoretical modeling techniques to predict the charge transfer mechanism in the metal oxides interface. Different metal oxides interfaces follow different charge transfer mechanisms. Particularly in Mott–Hubbard insulators, the O 2p band alignment and the competition between crystal field and correlation energy of d electrons are crucial to determine the electronic rearrangement [Reference Zhong and Hansmann104]. Generally, the apical oxygen atom at the interface [see Fig. 11(d)] is shared by the materials at the heterostructure interface. This results in the alignment of O 2p bands at the interface. However, considering the O 2p band alignment alone disregards the creation of the internal electric field that balances the electrochemical potential between two B-sites and prevents further charge transfer. Hence, interfacial charge transfer between two materials transfer cannot completely be understood relying only on O 2p band alignment. An extra factor that comes into play is the rearrangement of the d- bands on B-site cations. Specifically, the density of states of B-site cations of each material reflects a clear picture of possible electronic rearrangement.

As an example, the Ti atom would be expected to donate electrons to the Fe atom at the LaTiO₃/LaFeO₃ interface, implying a formal charge of 4+ for the interfacial Ti ion and the shifting of Fe from a 3+ state to a 2+ state near the interface [Reference Zhong and Hansmann104]. A theoretical prediction of charge transfer in the metal oxides interface is supported by XPS analysis both qualitatively and quantitatively to resolve the valence band structure. XPS is well known for its high sensitivity to the variations in valence states of transition metal ions. Kleibuker et al. [Reference Kleibeuker, Zhong, Nishikawa, Gabel, Müller, Pfaff, Sing, Held, Claessen, Koster and Rijnders110] showed an extra peak at ~2 eV lower binding energy is observed for the LaFeO₃/LaTiO₃ heterointerface in reference to the bulk LaFeO₃ for Fe 2p XPS spectra, as shown in Fig. 12. This extra peak in Fig. 12(c) at slightly lower binding energy, denoted by an open circle, represents the presence of the Fe²⁺ state along with the typical Fe³⁺ state at higher binding energy, denoted by a solid circle. Interestingly, the extra peak gets more pronounced with the decrease in thickness of the LaFeO₃ layer which indicates that the charge transfer occurs near or at the interface. To understand the rearrangement mechanism for transfer of charge, a valence band spectrum in XPS is analyzed experimentally. The evolution of an extra peak around 1 eV in valence band spectra represents the completely filled t _2g band of Fe²⁺ in a low-spin configuration. Separate measurements of the Ti 2p spectrum confirmed 4+ formal charge of Ti, indicative of charge transfer from Ti to Fe of 1 e⁻/UC at the interface.

Figure 12: (a) Sketch of the LaTiO₃/LaFeO₃ sample geometry. (b) A typical 1 × 1 μm AFM height image of a LaTiO₃/LaFeO₃ heterostructure. (c) Fe 2p XPS spectra of LaTiO₃/LaFeO₃ heterostructures for various thicknesses of LaFeO₃, as well as of a 30 u.c. LaFeO₃ film and a (2/2) LaAlO₃/LaFeO₃ heterostructure. The solid and open circles mark the Fe³⁺ and Fe²⁺ peaks, respectively. (d) Valence band XPS spectra of LaTiO₃/LaFeO₃ heterostructures for various thicknesses of LaFeO₃. All spectra were taken near normal emission (θ = 3°). Reprinted figure with permission from J.E. Kleibeuker et al. [Reference Kleibeuker, Zhong, Nishikawa, Gabel, Müller, Pfaff, Sing, Held, Claessen, Koster and Rijnders110]. Copyright 2014 by the American Physical Society.

For other instances, the preparation of complex superlattices that are different from regular heterojunctions helps to enlighten the process of charge transfer. One such example is charge transfer in LaTiO₃/LaNiO₃/LaAlO₃ heterostructures [Reference Disa, Kumah, Malashevich, Chen, Arena, Specht, Ismail-Beigi, Walker and Ahn5]. In the bulk structure, B-site cations of each material have 3+ oxidation states. However, because the d ⁰ Ti⁴⁺ is the more stable oxidation state, one electron can move from the Ti t _2g band to Ni e _g band resulting in Ni transitioning to a 2+ state from a 3+ state that is also more energetically favorable [Reference Disa, Kumah, Malashevich, Chen, Arena, Specht, Ismail-Beigi, Walker and Ahn5]. Interestingly, XAS confirmed this charge transfer and showed a preferential orbital occupation of the Ni 3d _z2 e _g orbital, producing the orbital occupation that makes LaNiO₃ resemble the superconducting cuprates. This type of approach to interface-engineered symmetry breaking is an alternative approach to the reduced (Nd,Sr)NiO₂ films that have recently demonstrated superconductivity [Reference Li, Lee, Wang, Osada, Crossley, Lee, Cui, Hikita and Hwang56]. Similar measurements for a LaNiO₃/LaMnO₃ heterostructure indicated electron transfer from Mn to Ni at the interface [Reference Hoffman, Tung, Nelson-Cheeseman, Liu, Freeland and Bhattacharya4]. A combined XPS and XAS study of LaCoO₃/LaTiO₃ heterostructures also indicated charge transfer to produce Co²⁺ and Ti⁴⁺ valences at the interface [Reference Araizi-Kanoutas, Geessinck, Gauquelin, Smit, Verbeek, Mishra, Bencok, Schlueter, Lee, Krishnan, Fatermans, Verbeeck, Rijnders, Koster and Golden111]. Charge transfer in SrMnO₃/SrIrO₃ to produce Mn³⁺ and Ir⁵⁺ [Reference Nichols, Gao, Lee, Meyer, Freeland, Lauter, Yi, Liu, Haskel, Petrie, Guo, Herklotz, Lee, Ward, Eres, Fitzsimmons and Lee46, Reference Okamoto, Nichols, Sohn, Kim, Noh and Lee47] and LaMnO₃/SrIrO₃ to produce Mn²⁺ and Ir⁵⁺ [Reference Suraj, Omar, Jani, Juvaid, Hooda, Chaudhuri, Rusydi, Sethupathi, Venkatesan, Ariando and Rao48] heterostructures has also been demonstrated. Each of these examples agrees with the model put forward by Zhong and Hansmann [Reference Zhong and Hansmann104], indicating the value of their approach in predicting interfacial charge transfer and band alignment based on initially aligning the O 2p energy bands. Future studies are likely to benefit from moving beyond 3d transition metal systems into the 4d and 5d blocks of the periodic table to generate novel interfacial phenomena.

Experimental design considerations

We have shown above several examples of the use of in situ XPS studies to probe interfacial phenomena in oxide heterostructures. The availability of an XPS appended to a growth chamber makes these studies very convenient and generates important physical insights as soon as the synthesis is complete. However, the knowledge gleaned from such studies is constrained by both the physical limitations of XPS and by the researcher's chosen design for the synthesized heterostructure. We have already discussed the information depth within the sample for an Al K_α source, which is limited to ~5 nm for photoemission normal to the film surface. Thus, band alignment and charge transfer studies are impractical for interfaces that are more than 5 nm below the surface.

One must also consider which interfaces in a multilayer will exhibit emergent properties, the chemical stability of the surface in vacuum and in atmosphere, and potential interference due to overlapping core-level peak energies. In the latter case, LaNiO₃ is a good example of a material that cannot be easily examined via XPS due to the overlap of the La 3d _3/2 and the Ni 2p _3/2. The LaNiO₃/LaTiO₃ charge transfer described above would be very difficult to study in XPS, though the use of the Ni 3p peak does provide one way to probe changes in Ni valence [Reference Qiao and Bi112]. Furthermore, the depth sensitivity of XPS means that if two interfaces are present in a multilayer, as in the case of the LaTiO₃/LaFeO₃/LaTiO₃ trilayer in Fig. 12(a), then the top interface will generate a majority of the signal and any phenomena at the bottom interface may be difficult or impossible to measure. Finally, while in situ measurements protect against changes due to atmospheric exposure, some surfaces may be modified even during the sample cooldown process [Reference Aeschlimann, Preziosi, Scheiderer, Sing, Valencia, Santamaria, Luo, Ryll, Radu, Claessen, Piamonteze and Bibes32] or will have intrinsic oxygen vacancies [Reference Golalikhani, Lei, Chandrasena, Kasaei, Park, Bai, Orgiani, Ciston, Sterbinsky, Arena, Shafer, Arenholz, Davidson, Millis, Gray and Xi55]. The deposition of a protective capping layer during the growth process that will not induce any chemical changes in the underlying materials is thus valuable even for in situ studies if a surface is unstable.

Hard X-ray photoelectron spectroscopy

Advances in spectroscopy over the past decade have largely been driven by the development of new techniques from synchrotron light sources. These have included the emergence of hard X-ray photoelectron spectroscopy (HAXPES). A variety of new HAXPES beamlines [Reference Reininger, Woicik, Hulbert and Fischer113, Reference Rueff, Rault, Ablett, Utsumi and Céolin114, Reference Schlueter, Gloskovskii, Ederer, Schostak, Piec, Sarkar, Matveyev, Lömker, Sing, Claessen, Wiemann, Schneider, Medjanik, Schönhense, Amann, Nilsson and Drube115, Reference Yasuno, Oji, Koganezawa and Watanabe116] have been constructed as new light sources are brought online, enabling XPS studies that are not limited by the surface sensitivity of lab sources. Performing XPS studies with X-ray photon energies >2 keV provides more bulk sensitivity and enables access to transition metal 1s core levels that cannot be examined in a conventional Al K_α laboratory source. As discussed previously, the IMFP for photoelectrons, λ, scales with the square root of electron kinetic energy for high kinetic energies. This means that electrons with kinetic energies ~10 keV will have IMFP approaching 10 nm. An IMFP of 10 nm produces depth sensitivities such that only half of the escaping photoelectrons will be from the top 6.5 nm of the sample and only 10% of the signal comes from the surface (top 1.0 nm). Conversely, for an IMFP of 1.5 nm that would be typical for a lab source, half of the signal is generated from the top 1.0 nm and only 3% of the signal comes from 5.0 nm below the surface. Thus, HAXPES measurements enable studies of deeply buried interfaces more than 5 nm below the film surface that would be impossible with a lab source. Numerous synchrotron studies have made great use of HAXPES to probe oxide interfaces, including the LaAlO₃/SrTiO₃ interface [Reference Sing, Berner, Goß, Müller, Ruff, Wetscherek, Thiel, Mannhart, Pauli, Schneider, Willmott, Gorgoi, Schäfers and Claessen91, Reference Gabel, Zapf, Scheiderer, Schütz, Dudy, Stübinger, Schlueter, Lee, Sing and Claessen117] and the LaNiO₃ surface [Reference Gray, Janotti, Son, LeBeau, Ueda, Yamashita, Kobayashi, Kaiser, Sutarto, Wadati, Sawatzky, Van de Walle, Stemmer and Fadley53] that we have already discussed. We refer the reader to several good reviews and examples of HAXPES studies, [Reference Fadley and Nemšák118, Reference Chambers and Woicik119, Reference Sing, Claessen, Cancellieri and Strocov120, Reference Gariglio, Cancellieri, Cancellieri and Strocov121] and will focus our attention in this section on the use of the technique to probe BaSnO₃.

BaSnO₃ is an exciting wide band gap oxide that has been shown to exhibit extremely high electron mobilities at room temperature (150–200 cm²/V-s) when grown by molecular beam epitaxy [Reference Paik, Chen, Lochocki, Seidner H, Verma, Tanen, Park, Uchida, Shang, Zhou, Brützam, Uecker, Liu, Jena, Shen, Muller and Schlom122, Reference Prakash, Xu, Faghaninia, Shukla, Ager Iii, Lo and Jalan123, Reference Raghavan, Schumann, Kim, Zhang, Cain and Stemmer124]. Unlike other perovskite oxides, the conduction band of BaSnO₃ is derived from Sn 5s orbitals, [Reference Sallis, Scanlon, Chae, Quackenbush, Fischer, Woicik, Guo, Cheong and Piper125, Reference Lebens-Higgins, Scanlon, Paik, Sallis, Nie, Uchida, Quackenbush, Wahila, Sterbinsky, Arena, Woicik, Schlom and Piper126] which enable a much greater degree of electronic dispersion and carrier mobility than, i.e., 3d orbitals in SrTiO₃. While this dramatic enhancement of mobility is of great benefit for potential future device applications, such as field-effect transistors [Reference Park, Kim, Ju, Park, Kim and Char127, Reference Park, Paik, Nomoto, Lee, Park, Grisafe, Wang, Salahuddin, Datta, Kim, Jena, Xing and Schlom128, Reference Krishnaswamy, Bjaalie, Himmetoglu, Janotti, Gordon and Van de Walle129], it also poses challenges to effective doping, both through the use of n-type donors [Reference Weston, Bjaalie, Krishnaswamy and Van de Walle130] and across interfaces through modulation doping and charge transfer [Reference Chambers, Kaspar, Prakash, Haugstad and Jalan103]. HAXPES has been particularly useful to elucidate the nature of the electronic transport in BaSnO₃ and in examining interfacial carrier profiles in stannate heterostructures.

Initial studies of La-doped BaSnO₃ via HAXPES focused on understanding the nature of the conduction band profile in the material, by probing electronic states near the Fermi level [Reference Sallis, Scanlon, Chae, Quackenbush, Fischer, Woicik, Guo, Cheong and Piper125, Reference Lebens-Higgins, Scanlon, Paik, Sallis, Nie, Uchida, Quackenbush, Wahila, Sterbinsky, Arena, Woicik, Schlom and Piper126]. Through ex situ synchrotron HAXPES studies with 4 keV photon energy, the first evidence of mobile Sn 5s electrons near the Fermi level was reported [Reference Sallis, Scanlon, Chae, Quackenbush, Fischer, Woicik, Guo, Cheong and Piper125]. Laboratory-source Al K_α studies of the same samples could not confirm the presence of carriers at the Fermi edge. The authors attributed this observation to the greater photoelectron cross section for Sn 5s with 4 keV photons in comparison to the 1487 eV lab source. However, given continued studies of the stability of dopants in BaSnO₃ [Reference Weston, Bjaalie, Krishnaswamy and Van de Walle130], it also seems possible that atmospheric exposure led to the depletion of the surface carriers, which would explain the absence of observed carriers in the lab-source experiments. Carriers in identically doped BaSnO₃ were later observed in a lab-source experiment [Reference Zhang, Han, Luo, Xiang, Zou, Oropeza, Gu and Zhang131], lending further credence to this possibility. It is also possible that the HAXPES experiments [Reference Sallis, Scanlon, Chae, Quackenbush, Fischer, Woicik, Guo, Cheong and Piper125] were successful due to the greater probe depth for hard X-rays, which overcomes the surface depletion. This explanation would provide another example of the effects of atmospheric exposure on metastable electronic states near the Fermi level. Further HAXPES studies of BaSnO₃ doped with differing La concentrations helped to explain the nature of the conduction band filling and band gap renormalization due to the Burstein–Moss effect [Reference Lebens-Higgins, Scanlon, Paik, Sallis, Nie, Uchida, Quackenbush, Wahila, Sterbinsky, Arena, Woicik, Schlom and Piper126]. These studies have paved the way for the development of stannate thin films and heterostructures, as research moves from understanding the nature of electronic conduction to confinement of carriers in a high-mobility 2DEG for device applications.

Modulation doping of BaSnO₃ through engineering of band alignment and charge transfer is a key requirement for a 2DEG structure. Impurity scattering from La donor ions is a key limitation to the further enhancements of mobility in electron-doped BaSnO₃ [Reference Prakash, Xu, Faghaninia, Shukla, Ager Iii, Lo and Jalan123]. To overcome this, engineering of a heterostructure with conduction band electrons at greater energies than the BaSnO₃ conduction band is necessary. In one such heterostructure, a 14 nm La-doped SrSnO₃ top layer was engineered such that free carriers in that layer could flow “downhill” to an undoped BaSnO₃ below the SrSnO₃ layer to form a 2DEG [Reference Prakash, Quackenbush, Yun, Held, Wang, Truttmann, Ablett, Weiland, Lee, Woicik, Mkhoyan and Jalan132]. Given the thickness of the SrSnO₃ top layer, synchrotron HAXPES studies were needed to probe the buried interface. Using a photon energy of 5.93 keV, the group was able to extract the interfacial band alignment and estimate the degree of charge transfer into the underlying BaSnO₃. These results are shown in Fig. 13. Interestingly, the extracted band profile in Fig. 13(d) could be explained by surface carrier depletion from atmospheric exposure, as discussed above regarding La-doped BaSnO₃ [Reference Sallis, Scanlon, Chae, Quackenbush, Fischer, Woicik, Guo, Cheong and Piper125]. These results show that the band alignment techniques that have been used for in situ studies for many years can be applied to more challenging structures using hard X-ray sources that probe more deeply into materials. Continued development of HAXPES will dramatically improve the ability to measure interfacial phenomena in heterostructures that are not limited to a few nm in thickness due to the lab-based X-ray energies.

Figure 13: Band alignment at La-doped SrSnO₃ (LSSO)/SrSnO₃₍SSO)/BaSnO₃ (BSO) interface. (a) VB spectra of the reference BSO (green) (56 nm BSO/SrTiO₃ (001)) and LSSO (blue) (41 nm LSSO/8 nm SSO/GdScO₃ (110)) films. Electronic states near the Fermi states are magnified. Inset shows the La 3d _5/2 core-level X-ray photoelectron spectra, (b) VB spectra of the SSO/BSO heterostructure (red) along with the fit (black) using linear combination of the reference VB spectra (dotted green and blue lines) to determine the VB offset. (c) An energy level flat-band diagram showing the measured band offsets between LSSO and BSO, and (d) conduction band minima (red) referenced to the Fermi level (top panel) and 3D carrier density, n _3D (blue) as a function of depth for the SSO/BSO (bottom panel). The shaded regions indicate 2D density in LSSO and BSO layers after the charge transfer. Reprinted with permission from A. Prakash et al. [Reference Prakash, Quackenbush, Yun, Held, Wang, Truttmann, Ablett, Weiland, Lee, Woicik, Mkhoyan and Jalan132]. Copyright 2019 American Chemical Society.

Standing-wave XPS

While HAXPES measurements are one approach to probe more deeply within a material and examine both bulk electronic structure and buried interfaces, the underlying physics of the photoemission process is unchanged. As we showed above, HAXPES measurements can still be affected by surface chemistry of the films. The standing-wave XPS (SW-XPS) approach has emerged over the past decade, specifically focusing on studies of superlattice structures that exhibit regularly repeating interfaces. By varying the incoming X-ray angle such that the Bragg diffraction condition is satisfied for the superlattice, the intensity of the electric field is modulated across the interface. A stronger Bragg reflection produces a greater standing-wave intensity and greater modulation of the electric field strength across the repeating superlattice formula unit. The physics of SW-XPS has been described in more detail in various references [Reference Fadley and Nemšák118, Reference Gray133, Reference Gray, Nemšák and Fadley134, Reference Nemšák, Gray, Fadley, Cancellieri and Strocov135]. Here, we aim to discuss the applications of SW-XPS to understanding interfacial phenomena and present opportunities to expand on this unique approach by designing interfacial materials that will be well suited for SW-XPS studies.

As an initial example, we present a study of layer-resolved band alignment in a superlattice achieved through careful design of a superlattice structure. The structure and electric field strength modulation is shown schematically in Fig. 14 for a study of a LaCrO₃–SrTiO₃ superlattice using ~830 eV photon energy near the La M ₅ resonance. [Reference Lin, Kuo, Comes, Rault, Rueff, Nemšák, Taleb, Kortright, Meyer-Ilse, Gullikson, Sushko, Spurgeon, Gehlmann, Bowden, Plucinski, Chambers and Fadley136] In this case, the formula of 10 UCs of SrTiO₃ followed by 5 UCs of LaCrO₃ over 10 repeating layers in the superlattice was chosen to maximize the Bragg reflection and thus the sensitivity to individual interfaces. By choosing the superlattice periodicity prior to growth and considering the subsequent SW-XPS experiment, the largest standing-wave response ever achieved was reported, with individual core-level intensities varying by ~50% as the X-ray angle was varied across the Bragg peak. This enabled the extraction of the band alignment across the entire 15-unit-cell structure through careful modeling of the core-level peak positions as a function of incoming angle, producing good agreement with DFT models [Reference Lin, Kuo, Comes, Rault, Rueff, Nemšák, Taleb, Kortright, Meyer-Ilse, Gullikson, Sushko, Spurgeon, Gehlmann, Bowden, Plucinski, Chambers and Fadley136, Reference Comes, Spurgeon, Heald, Kepaptsoglou, Jones, Ong, Bowden, Ramasse, Sushko and Chambers137].

Figure 14: (a) Schematic of the superlattice made up of 10 bilayers of LCO and STO, consisting of 5 UCs of LCO, 17.6 Å thick, and 10 UCs of STO, 39.2 Å thick, grown epitaxially on a Nb-doped STO(001) substrate. The two sources of the standing-wave structure in the rocking curves are indicated: Bragg reflection from the multilayer with period d _ML and Kiessig fringes associated with the full thickness of the multilayer stack D _ML. Experimental (open circles) and simulated (solid) rocking curves of representative elemental states at photon energies of (b) 829.7 eV and (c) 831.5 eV. The colored dash lines are the guides to the eye to indicate the phase of the rocking curves in (b and c) to show sensitivity to the interfacial termination (SrO/CrO₂ versus TiO₂/LaO). The electric field strength distribution derived from X-ray optics calculations at two energies near the La M₅ resonance, (d) 829.7 eV and (e) 831.5 eV as a function of sample depth and incidence angle. Note the significant shift in position between the two energies. The corresponding photoemission yields, (f and g), plotted on log10 scales. Adapted figure with permission from S.-C. Lin et al. [Reference Lin, Kuo, Comes, Rault, Rueff, Nemšák, Taleb, Kortright, Meyer-Ilse, Gullikson, Sushko, Spurgeon, Gehlmann, Bowden, Plucinski, Chambers and Fadley136] . Copyright 2018 by the American Physical Society.

The SW-XPS technique has been used to probe changes in interfacial chemistry in a variety of other superlattice systems. Evidence for Ni²⁺ valence near the interface with SrTiO₃ in 4 u.c. LaNiO₃–4 u.c. SrTiO₃ superlattices was reported [Reference Kaiser, Gray, Conti, Son, Greer, Perona, Rattanachata, Saw, Bostwick, Yang, Yang, Gullikson, Kortright, Stemmer and Fadley138]. Studies of LaNiO₃–CaMnO₃ superlattices showed similar effects, with Ni²⁺ states observed at the interface [Reference Chandrasena, Flint, Yang, Arab, Nemšák, Gehlmann, Özdöl, Bisti, Wijesekara, Meyer-Ilse, Gullikson, Arenholz, Ciston, Schneider, Strocov, Suzuki and Gray139]. Likewise, changes in the interfacial Mn valence in (La,Sr)MnO₃–SrTiO₃ superlattices were also observed [Reference Gray, Papp, Balke, Yang, Huijben, Rotenberg, Bostwick, Ueda, Yamashita, Kobayashi, Gullikson, Kortright, de Groot, Rijnders, Blank, Ramesh and Fadley140]. Collectively, these studies have shown that the technique provides additional sensitivity to changing interfacial chemical states at interfaces. The X-ray standing wave provides a novel way to isolate the interfacial contribution to the detected signal that goes beyond conventional angle-resolved XPS. However, the design of the superlattice and choice of materials system under examination plays a significant role in how much information can be extracted from the measurement. For example, superlattices with large differences in density between layers produce larger superlattice Bragg peaks, so that a superlattice comprised of, i.e., SrTiO₃ and LaTiO₃ layers will produce a greater peak than one comprised of, i.e., SrTiO₃ and SrVO₃ or LaTiO₃ and LaVO₃ layers. This means that the technique is currently only suitable for select combinations of materials.

Ideally, SW-XPS could be used to measure not simply changes in interfacial chemistry but also to probe emergent interfacial electronic states near the Fermi level. There have also been some efforts to employ SW-XPS to generate ARPES information as a function of depth within the material, including studies of SrTiO₃–GdTiO₃ [Reference Nemšák, Conti, Gray, Palsson, Conlon, Eiteneer, Keqi, Rattanachata, Saw, Bostwick, Moreschini, Rotenberg, Strocov, Kobayashi, Schmitt, Stolte, Ueda, Kobayashi, Gloskovskii, Drube, Jackson, Moetakef, Janotti, Bjaalie, Himmetoglu, Van de Walle, Borek, Minar, Braun, Ebert, Plucinski, Kortright, Schneider, Balents, de Groot, Stemmer and Fadley141] and SrTiO₃–(La,Sr)MnO₃ [Reference Gray, Minár, Plucinski, Huijben, Bostwick, Rotenberg, Yang, Braun, Winkelmann, Conti, Eiteneer, Rattanachata, Greer, Ciston, Ophus, Rijnders, Blank, Doennig, Pentcheva, Kortright, Schneider, Ebert and Fadley142] superlattices. These results are promising demonstrations of the technique but provided limited insights into emergent interfacial physical phenomena due in large part to the relatively weak X-ray standing wave. The weakness of the standing-wave can be observed in small variation of the core-level peak intensities as a function of angle (~20–30%). This means that differences in the ARPES data as a function of angle are fairly small and it is difficult to extract clear information about the electronic structure in individual layers of the superlattice or at a particular interface. In most cases of SW-XPS studies to date, the challenge has not been the technique but the limitation of the samples themselves. We suggest that a somewhat different model could bear fruit.

Because the photon energies for most SW-XPS studies have been in the soft X-ray regime where it is possible to take advantage of elemental resonances to enhance reflectivity, the measurements are still limited to the top few nm of the sample. Thus, the underlying superlattice serves only to generate the X-ray standing wave that modulates the field strength across the topmost repeat layer of the sample and does not have a meaningful effect on the actual data being measured. It, therefore, makes sense to choose an underlying superlattice solely from the perspective of generating the strongest standing-wave intensity while not confusing the data interpretation from the relevant interface in other ways. In this sense, the superlattice serves as a novel substrate for the actual experimental interface, as has been applied in other cases for non-epitaxial films such as the Fe–MgO interface [Reference Döring, Schönbohm, Berges, Schreiber, Bürgler, Schneider, Gorgoi, Schäfers, Papp, Balke, Fadley and Westphal143] and ambient-pressure XPS measurements on Fe₂O₃ [Reference Nemšák, Shavorskiy, Karslioglu, Zegkinoglou, Rattanachata, Conlon, Keqi, Greene, Burks, Salmassi, Gullikson, Yang, Liu, Bluhm and Fadley144]. For example, a SrTiO₃–BaTiO₃ superlattice is relatively easy to synthesize and would provide strong standing-wave effects across the Ba M ₅ resonance (~780 eV). Predictions of superlattice Bragg peaks for various BTO_n–STO_n superlattice structures using the X-ray Interactions with Matter website [Reference Gullikson145, Reference Henke, Gullikson and Davis146] provided by the Center for X-ray Optics at Lawrence Berkeley National Laboratory are shown in Fig. 15(a). The electric field strength varies as the square root of reflectivity, meaning that values greater than 0.1 on the scale (dotted line) will produce modulations of XPS peak intensity of 30% or better. The superlattices could then be used as X-ray mirror substrates for any number of intriguing interfacial materials to produce much stronger standing-wave effects. A schematic of such a heterostructure is shown in Fig. 15(b), with a BTO₅–STO₅ superlattice with an arbitrary SrMO₃-SrM'O₃ interface highlighted in yellow that will be strongly probed by the standing wave from the underlying mirror. Given this opportunity to shift the approach to standing-wave measurements, groups should consider the SW-XPS experiment when designing both their superlattice mirror layer and the interface of interest. Others have proposed additional approaches to enhance standing-wave effects in these measurements [Reference Gray133]. Ideally, collaborations between synthesis groups and experts in SW-XPS early in the research process will produce samples that are particularly well suited to answer specific questions about the properties of an interface.

Figure 15: (a) Modeled X-ray reflectivity for three n-unit cell BaTiO₃ (BTO_n)/n-unit cell SrTiO₃ (STO_n) superlattices on SrTiO₃ substrates at the Ba M₅ energy resonance; (b) Schematic of BTO₅–STO₅ superlattice mirror to probe an arbitrary SrBO₃-SrB'O₃ interface (highlighted in yellow).

Future applications

In summary, synchrotron XPS techniques have generated significant new insights into buried oxide interfaces by taking advantage of the variable X-ray energies that are provided from a synchrotron source that cannot be replicated in a lab. Many of these approaches have only been pursued over the past decade, however, leaving significant opportunities to refine the practices to better extract novel physics from the experiments. For example, many beamlines have begun to integrate MBE or PLD growth systems in situ as has been done in a lab environment for many years [Reference Palomino, Stavitski, Waluyo, Chen-Wiegart, Abeykoon, Sadowski, Rodriguez, Frenkel and Senanayake147, Reference Eres, Rouleau, Lu, Zhang, Benda, Lee, Tischler and Fong148]. We have shown in this review that surface exposure must be considered when interpreting results, even for HAXPES studies. It thus makes sense to perform SW-XPS and HAXPES studies using in situ growth capabilities at the beamline whenever possible. Alternatively, the use of a vacuum suitcase to transport samples from a home synthesis laboratory to the user facility can enable synchrotron studies that replicate in situ conditions.

Another promising new opportunity has been enabled by the development of lab-based HAXPES instruments [Reference Regoutz, Mascheck, Wiell, Eriksson, Liljenberg, Tetzner, Williamson, Scanlon and Palmgren149]. Rather than traditional Al or Mg sources, these commercially available instruments employ a 3d transition metal or metalloid such as Cr or Ga to produce monochromatic Kα photons with energies greater than 5 keV. The recent development of a liquid-metal Ga source with a Rowland circle monochromator has produced fluxes that are within a factor of 200 of synchrotron sources, making it possible to perform measurements in reasonable times in a lab environment [Reference Regoutz, Mascheck, Wiell, Eriksson, Liljenberg, Tetzner, Williamson, Scanlon and Palmgren149]. To date, we know of no laboratory that has an oxide or other complex materials film growth chamber attached in situ to a HAXPES instrument, but this would be an excellent resource for future film synthesis user facilities. An instrument suite integrating film synthesis and the large probing depth of a hard X-ray source in situ would be an invaluable resource for studies of charge transfer and band alignment in future complex heterostructures like the one presented in Fig. 13 [Reference Prakash, Quackenbush, Yun, Held, Wang, Truttmann, Ablett, Weiland, Lee, Woicik, Mkhoyan and Jalan132].