Introduction

The technologies necessary for the electrification of transportation, the generation of green hydrogen at scale1 for transport and manufacturing,2 and the grid-scale energy storage to accommodate the intermittency of renewable electricity generation3,4 all require new materials solutions.5 The performance and service lifetime of these materials are inherently related to their microstructure, chemistry, physics, and interaction with their environment. Understanding the subtle processing–structure–property relationships is key to guide the design of future generations of materials. Yet, the variety of scales involved, from individual atoms to nano- to micro- to millimeter scales, and nature of the information—(i.e., surface atomic arrangements, crystalline structure, composition, and chemistry)—is simply too broad for a single technique to provide all the necessary insights. Hence, correlative approaches must be developed to build upon the strength of each individual technique, as showcased in the case of microstructural features found in the bulk of materials,6 especially structural defects.7,8,9 Another challenge is that a material’s chemical (re)activity can be underpinned by metastable species at the surface of the electrodes and their analyses can only be performed by using in situ or in operando techniques.

Atom probe tomography (APT) provides three-dimensional compositional mapping of materials with subnanometer resolution10 and as such is poised to provide extremely valuable insights. APT has progressively increased in prominence in the battery of characterization techniques for bulk metallic materials,11,12,13 ceramics and semiconductors,14,15 phase-change materials,16 and with recent forays into nanostructures,17,18 including nanoparticles19,20,21,22,23 and 2D materials.24,25 There are also numerous potential application in catalysis, as recently reviewed by Barroo et al.26 There is a compatibility in scales between specimens for APT, shaped as sharp needles with a near-spherical cap at their end, and nanoparticles used across many catalysis applications (i.e., 10–200 nm in diameter). Historically, this had enabled field-electron emission microscopy (FEEM/FEM) to be extensively used to study surfaces in reaction conditions.27 The reacted surfaces were then imagined by field-ion microscopy (FIM) in search for morphological or atomic-scale topographical modifications. Early implementation of atom probes were then sometimes used to measure changes in the surface chemistry; however, they were limited to one-dimensional depth-profiling of the composition or to study a single set of reacted planes.27

The analysis of catalysts and electrocatalysts is, however, an area in which deploying the full potential of APT could be transformational. However, despite outstanding works performed in a limited number of groups,28,29,30,31 APT has not seen a fast spread in surface sciences studies, in part because of difficulties in specimen preparation, but also a lack of direct structural information available along with the chemical and bonding state of surface species. These are accessible through x-ray photoelectron spectroscopy (XPS), one of the key techniques used in surface sciences. XPS allows access to the bonding state within the first few atomic layers of the surface. For its compatibility in scale, APT is more often correlated with (scanning) transmission electron microscopy, in which the bonding state can be obtained from electron energy-loss spectroscopy ((S)TEM–EELS).32,33 Energy-dispersive x-ray spectroscopy is also accessible in the (S)TEM (STEM–EDS) to assess the local composition; we, however, herein focus on techniques enabling to access the bonding state. The complementarity between STEM–EDS and APT had previously been discussed.34 Preliminary work on thin films35,36 demonstrated the importance of combining XPS, STEM–EELS, and APT to get a holistic understanding of the microstructural origin or the materials performance and degradation mechanisms.

In this article, following a brief overview of the working principles of these techniques to help the reader get to speed, we discuss the complementarities of these techniques and showcase examples from the recent literature to provide a perspective on the strength of the correlative approaches.

Working principles and techniques’ complementarity

Working principles

We briefly introduce the working principles of the techniques discussed next for nonexperts. This should in any case not be seen as an exhaustive overview of these techniques, merely a segue enabling to position the perspective on the correlative approaches. It should be stressed that these techniques cannot be deployed on the exact same specimen, in the same instrument, and simultaneously, but must be used successively.

APT

In APT, as schematically depicted in Figure 1a, the atoms leave the surface of a needle-shaped specimen successively in the form of ions under the influence of an intense electric field. Following field evaporation, these ions fly along well-defined trajectories37,38 and are collected by a single-particle detector that records the impact position.39 The field evaporation is time-controlled by using either fast voltage40 or laser pulses41,42 superimposed to a DC high-voltage, thereby enabling the identification of the elemental nature of each ion by time-of-flight mass spectrometry.43 The results from an APT experiment take the form of a mass spectrum (i.e., a histogram of the number of ions detected at a certain mass-to-charge ratio), and a point cloud, built from the detector impact positions, providing the atomic distribution in three-dimensions,44 as summarized in Figure 1b.

Figure 1
figure 1

(a) Cartoon view of the specimen in an atom probe, (b) main results from atom probe tomography (APT), (c) schematic view of x-ray photoelectron spectroscopy performed on a similar sample, and (d) projected view of the APT specimen as obtained from (scanning) transmission electron microscopy with an electron energy-loss spectroscopy (EELS) spectrometer and possible high-resolution images and chemical maps (modified from Reference 45).

Full size image

X-ray photoelectron and electron energy-loss spectroscopies

In XPS, a micron-to-millimeter-sized beam of x-rays is focused onto the sample’s surface. Upon penetration, some electrons from the material itself absorb photons from the incoming beam and get ejected from the material (i.e., the photoelectric effect). By measuring the kinetic energy of the emitted electron, and knowing the energy of the incoming photon, the bonding energy can be readily determined.46 This is schematically summarized in Figure 1c. XPS hence provides a precise account of the chemistry of the first few atomic layers at the specimen’s surface, averaged over microns to millimeters of the surface depending on the setup that is used.46 There are ongoing efforts to perform spatially resolved XPS experiments with a lateral resolution in the range of tens to hundreds of nanometers.47 A clear strength of XPS is the possibility to be used in operando (i.e., at temperatures and gas pressures that can mimic service conditions) to provide a precise account of the evolution of the surface chemistry over the course of reactions.48

STEM–EELS can provide similar information: as the incoming electrons travel through the thin specimen, inelastic scattering causes the loss of amounts of energy that can be related to plasmons or to the ionization energy of specific species. Analyzing the spectrum of the kinetic energy of the electrons coming out of the specimen allows for determining the specimen’s composition and bonding state of the different species present. As the electron beam in STEM is tightly focused and scanned across the specimen’s surface, STEM–EELS enables mapping of the species potentially within each individual atomic column along the pathway of the electron beam,49,50 as schematically shown in Figure 1d. There are also ongoing efforts to enable in situ or quasi in operando observations by STEM–EELS51 to help better understand materials processes in conditions mimicking service.

Complementarities and challenges

First, there is a complementarity in the scale of the features that can be analyzed by the different techniques and a compatibility in the spatial resolutions. The spatial resolution of each of these techniques is anisotropic and depends on an array of parameters. What matters more is not the absolute value of the spatial resolution but the size of the smaller resolvable feature both laterally and in-depth. For both XPS and STEM–EELS, the spatial resolution not only depends on intrinsic instrumental parameters—the spot size of the illuminating beam, but also depends on extrinsic factors—for instance, the nature of the specimen itself. Indeed, in XPS, the penetration depth is dependent on the composition itself, typically below 10 nm, whereas the lateral extent depends on the size of the focused x-ray beam. In STEM–EELS, the signal collected is integrated through the thickness of the thin specimen (20–100 nm) and depends on the lateral extent of the focused electron beam, and its potential spread from scattering going through the specimen.52 The actual spatial extent of the probed volume of the material can hence be difficult to ascertain. This is where the inherent strength of APT can be best used. In APT, the spatial resolution or size of the smaller resolvable microstructural feature of interest depends on its composition,53,54 and it can be assumed to be below 1 nm in all three dimensions55 but is typically an order of magnitude better in depth,53,56 as summarized in Figure 2a.

Figure 2
figure 2

(a) Summary view of the resolvable feature size for atom probe tomography (APT), scanning transmission electron microscopy–electron energy-loss spectroscopy (STEM–EELS), and x-ray photoelectron spectroscopy (XPS). (b) Schematic view of the main microstructural features and their typical size to provide a perspective on what can be addressed by the various techniques discussed herein (inspired and modified from Reference 10).

Full size image

Now, let us consider common microstructural features, for instance, stacking faults, dislocations, grain boundaries, and secondary phases, along with their respective dimensions, as shown in Figure 2b. The signal in the recorded spectra from XPS or STEM–EELS can be extremely complex as it is a combination of the signals originating from the features within the probed volume. Fitting approaches are hence used to decompose the signals originating from the different bulk or surface features averaged over the volume probed by the electron or x-ray beam.33,50,57 This can lead to components and phases with a low volume fraction to be missed or overlooked, yet their influence on the materials properties may remain substantial.

This is where the correlation with APT can become a game changer: APT provides compositional mapping in three dimensions within the reconstructed volume, with a compositional sensitivity that can be in the range of parts-per-million,58 and individual features can be interrogated separately after data reconstruction and segmentation. This information can then be used to better understand the origins of the signal in the XPS or EELS spectrum. In turn, the bonding state remains inaccessible to APT. This is a first critical aspect of the complementarity with XPS and STEM–EELS, as the bonding state is required to understand the chemical activity or reactivity of a compound.

APT has a higher chemical sensitivity58 than these spectroscopic techniques, especially for light elements (e.g., H,59,60 C, and N).61 Yet the compositional accuracy strongly depends on the local intensity of the electric field62,63,64 during the analysis. These issues can make the analysis of certain materials or features challenging, and the complementary insights into the composition of some interfaces, for instance, that can be gleaned from STEM–EELS, can help support65,66 the validity of some measurements.

Finally, the analysis of a sample’s outermost surface requires dedicated strategies to avoid damage and contamination during specimen preparation and transport. For APT and STEM–EELS, specimens are typically prepared using a focused ion beam (FIB),67,68 in which the energetic incoming ions get implanted, can cause amorphization, and can push surface atoms to penetrate inside the subsurface region.69,70,71 The use of cryogenic temperatures during the preparation can help alleviate some issues,72,73 but not all. A thin metallic coating can be used to protect the very surface, yet it often involves transporting the sample through ambient air, which can have an influence on the surface species themselves. There are efforts to develop approaches to transport samples under protective environments or (ultra-) high vacuum conditions and possibly under cryogenic conditions.74,75,76,77 The use of transport suitcases has been more common in surface sciences, and most XPS instruments are equipped with an ion-gun to clean the specimen’s surface by sputtering and remove the first few atomic layers.

Workflows and applications

Challenges arise in the development of appropriate workflows enabling analysis of the same sample to correlate the signals from the same microstructural features. Schweinar et al. proposed to use thin films deposited on the commercial flat-top coupons78 used as support for APT specimen preparation,79,80 as shown in Figure 3a. They performed spatially resolved XPS directly onto the film before and after thermal oxidation (Figure 3b–c), revealing the change in the oxidation state of Ir and Ru atoms. The compositional maps appear homogeneous across the 3-µm-wide disc at the top of the microtip. Following this analysis, the sample’s surface was protected by a thin layer of sputter-deposited Cr and subsequently sharpened into an APT specimen by FIB81 and analyzed by APT. The corresponding APT reconstruction is shown in Figure 3d. The analysis starts from the Cr capping layer, goes through the oxide layer, and terminates into the metallic film. Interestingly, by looking at a region of interest across the x–y section of the data set in the oxidized region, Ru is not spread across the material but rather agglomerated along grain boundaries within the thin film, as readily visible in Figure 3e. APT allows for quantification of the Ru segregation, determined as up to nearly 10 at.% (Figure 3f), accompanied by a depletion of Ir. This observation is likely related to the reduction in the free energy of the grain boundary associated with the segregation of Ru prior to oxidation.

Figure 3
figure 3

Composite image of the workflows introduced by Schweinar et al.79,80 (reproduced and modified under a CC-BY license). (a) Schematic of the analysis by spatially resolved x-ray photoelectron spectroscopy of a thin Ir–Ru film deposited on flat-top commercial microtip coupons, normally used as support for lift-out specimen preparation for atom probe tomography (APT). (b) Photoelectron (PE) map and seemingly homogeneous compositional map obtained from the film after thermal oxidation at 600°C for 5 h in air. (c) Comparison of spectra before and after oxidation showing the metallic state and oxidized state, respectively. (d) Reconstructed APT data set following deposition of a Cr capping layer and sharpening. (e) Cross section through the APT reconstruction showing segregation of Ru to grain boundaries within the oxidized film. (f) Composition profile calculated in a region of interest positioned across the grain boundary enabling quantification of the Ru (blue) segregation.

Full size image

This exemplifies the kind of insights that APT can bring—XPS could only detect a single oxide phase, whereas a richer microstructure is resolved by APT. The local changes in composition at grain boundaries, combined with a different local atomic organization, will change the catalytic activity of the grain boundaries, as other studies have also shown with the aid of STEM–EELS.35 If APT can be performed on a thin film suitable for XPS, the opposite was also proposed by Balakrishnan et al.,82 in Figure 4a. They performed XPS first on APT specimens of Ir, which had been used as an electrocatalyst for the oxygen evolution reaction. Following XPS, the specimens were then analyzed by APT to reveal the local compositional evolution of the different sets of atomic planes that form the end surface of the needle.

Figure 4
figure 4

(a) Workflow for the analysis of atom probe tomography (APT) specimens used as electrocatalysts by x-ray photoelectron spectroscopy (XPS) (from Reference 82 reproduced under a CC-BY license). (b) Summary of the oxygen isotope exchange experiment on iridium oxide during the oxygen evolution reaction (OER) (from Reference 83 reproduced under a CC-BY license).

Full size image

Both of these approaches have advantages and drawbacks. For instance, the APT specimen’s shape and size can make direct electrochemical or XPS measurements not necessarily straightforward, making the latter approach not always suitable, particularly when site-specific analysis is necessary. Yet the need to shape the specimen with a FIB after the reaction typically requires transporting the specimen through air, and the energetic ion beam can induce damage to the reacted sample’s surface. Transport under protective atmospheres can be possible by using dedicated infrastructure74 but it is not yet routinely available.

In any case, these studies showcased the spatial sensitivity of APT and the importance of the precise assessment of the species on the surface and in the near-surface regions. The depth probed by XPS is often not known precisely, as the penetration of the x-rays depends on the local composition and structure. In contrast, APT provides direct compositional measurement of the features on the surface, provided that appropriate protection is used prior to specimen preparation and analysis, and the spatial resolution of APT in the depth is extremely high.54,56 There remains some uncertainties estimated to be 10% or more84 in the depth dimension of the APT reconstruction.85 However, calibration of the spatial reconstruction parameters is possible either by using the partial structural information86 from within the data or through the use of TEM.6 This leads to an extreme spatial resolving power of APT that was, for instance, combined with isotope labeling to investigate the exchange of oxygen between the solution and the lattice of iridium oxide during the oxygen evolution reaction,83 as summarized in Figure 4b.

Most showcase studies so far have focused on the analyses of electrocatalysts and electro-photo-catalysts that find application in the hydrogen energy cycle, especially the analysis of Ir metal in the form of thin films,35 metal wire,82 and mixed Ir–Ru alloys,79 but also an array of oxides,17,36,87,88,89 in particular to better understand the microstructural origins of their activity36 or degradation in service.88 In principle, similar approaches could be deployed in the future to a broad range of electro(photo)catalysts.

Conclusions

APT arises from FIM, and both were originally surface science techniques, yet progress in APT made the community progressively turn its attention toward bulk analyses. However, APT with its intrinsic capacity to map the composition or a material with a resolution better than a nanometer in three dimensions has great potential for complementing XPS that can probe the chemistry of a surface, and STEM–EELS, which provides compositional and chemical mapping through the thickness of a thin sample. Information from APT can guide the fitting of the electron energy spectra and facilitate data interpretation, as well as support investigation into processing–microstructure–property relationships that are necessary to design new electrocatalysts and electro-photo-catalysts. The difference in scale can be difficult to reconcile but the development of dedicated workflows has already enabled progress that will continue in the decade to come.