Introduction

Alloys applied in nuclear reactors1,2, gas turbines3,4, heterogeneous catalysis5,6, automotive internal combustion engines7, and electronic devices8,9 are routinely exposed to high-temperature, corrosive, and oxidising environments. Oxidation of alloys in service can degrade material properties and performance, raising significant safety and economic concerns. Hence, a comprehensive understanding of oxidation mechanisms is necessary to evaluate applicability of a material in a desired environment. Characterising one or more oxides formed during the initial stages of oxidation is a significant challenge because oxidation of alloying elements is simultaneous with the nucleation and growth of metastable phases10,11. Further, the composition across buried oxide/metal interfaces can also change, often heterogeneously, depending on the crystallographic orientation of the base metal grains below6, requiring a spatially resolved analysis of such changes.

Zirconium (Zr) alloys are widely used in nuclear applications, most commonly as nuclear reactor fuel cladding in pressurised water reactors12,13, and as a hydrogen isotope absorbing material in tritium producing burnable absorber rods14,15. Zr alloys have been selected for these applications because they offer a favourable combination of mechanical durability, corrosion resistance, and a low thermal neutron cross-section16,17. However, oxidation and hydrogen pickup can significantly affect the integrity of these alloys in various environments, such as normal reactor operating conditions and short- and long-term storage (in spent fuel storage pools or canisters)2. Since oxidation of these alloys has serious implications for environmental and economic viability of nuclear energy, a deeper understanding of oxide evolution and oxidation mechanisms is needed.

Most of our understanding of Zr alloy oxidation comes from long-term autoclave experiments, ranging from tens to hundreds of days, involving weight gain measurements and post-test, ex situ microscopy2. These measurements are fit to a power–law relationship, with oxide thickness proportional to tn, where t is time and n is a constant between 0.25 and 0.5018. The ‘initial’ oxide layer (referred to as ‘pre-transition’) is adherent and protective. Once the oxide thickness reaches a critical value (~2–3 µm), the protective characteristic is lost, and oxidation proceeds rapidly, essentially resetting the power–law growth rate19,20,21. This sudden increase in oxidation rate is referred to as a kinetic transition. Some Zr-based alloys experience cyclic kinetics (with multiple transitions), while others have an initially protective oxide film, followed by breakaway corrosion and oxide spallation2. It has been hypothesised that stress accumulation in the oxide layer leads to the development of cracks and pores that act as short-circuit paths for oxygen diffusion to the oxide/metal interface, leading to accelerated corrosion2,22,23. Although models and computational studies have explored this topic24,25, reported experimental observations in Zr alloys cannot be described by such models or fundamental theories (e.g., Wagner’s theory of parabolic oxide growth rate26) alone. The topic of Zr alloy oxidation is further complicated by the numerous oxide compositions and structures reported in literature, with several allotropic transformations possible in the oxide19,27. Thus the mechanisms governing the evolution of composition and structure across the oxide/metal interface and how the protective character of the oxide film is lost remain elusive2.

The application of high-resolution analytical techniques, such as atom probe tomography (APT) and scanning transmission electron microscopy (STEM)/electron energy loss spectroscopy, has helped improve the understanding of structure, composition, and chemical states across oxide/metal interfaces after long-term oxidation18,22,27,28,29,30,31,32,33. Critical to improved mechanistic understanding of Zr alloy oxidation is tracking the early stages of oxide film formation. However, such analyses have received relatively less attention given the difficulties in preserving oxide/metal interfaces. High-resolution in situ/operando transmission electron microscopy (TEM) is emerging as a way to analyse atomic-scale structural changes near alloy surfaces during high-temperature oxidation11,34,35,36. Nevertheless, such TEM studies have inherent challenges when quantitatively analysing composition of light elements and low concentrations of alloying or impurity elements, particularly across complex interfaces. The recently developed reaction chambers attached to the vacuum systems of APT instruments have been used to study composition changes of oxide layers. These chambers have demonstrated their usefulness in analysing near-surface compositional changes during oxidation of catalyst materials such as Pt–Rh, Pt–Ru, Pt–Ru–Rh, Pd–Rh, and Rh5,6,37 and Mg alloys38.

In an effort to uncover atomic-scale changes that contribute to oxidation behaviour in the pre-transition regime and gain insight into characteristics of the protective oxide formed, we investigated the initial oxide formation and its evolution on zirconium alloy (Zircaloy-4) at 300 °C in a controlled oxygen gas (O2) environment. Ultrahigh vacuum (UHV) transfer of the sample between an environmental reaction chamber and the APT analysis chamber enables cyclic oxidation and analysis of the same sample with direct site registry, in contrast with conventional ex situ approaches where different areas are analysed in each experiment39. The APT results were coupled with STEM analysis of oxidised APT needles for spatially resolved structural analysis of the oxide phases. Synchrotron-based X-ray absorption near-edge spectroscopy (XANES) provided insights into electronic structure changes. The collective insights from all three experimental techniques provide a comprehensive understanding of structural, compositional, and chemical state changes during the early stages of Zircaloy-4 oxidation. The approach utilised here provides a rapid, high-throughput methodology for detailing early stages of alloy oxidation, relevant to several systems and application areas.

Results

Initial oxide formation at 300 °C

The native oxide formed on needle samples during atmospheric transfer from the focused ion beam (FIB) microscope to the APT microscope (Fig. 1) had a stoichiometry close to that of Zr3O. The O:Zr ratio was ~1:3.5 (~29 at. % O), which is consistent with the solid solubility limit of O in the α-Zr matrix40 and with previously reported APT analysis of Zr oxides27. The native oxides on all needle samples were removed via field evaporation in the APT microscope, exposing an oxide-free top surface prior to oxidation experiments in an environmental reaction chamber.

Fig. 1: Compositional analysis of the Zircaloy-4 needle before oxidation.
figure 1

a 2-D contour plot for O, where concentration is shown on the colour scale to vary between 0 and 30 at. %. b 3-D element distribution map showing Zr (purple dots) and ZrO (light blue dots) ions and a 10 at. % O isoconcentration surface (purple surface). A 10 at. % O isoconcentration surface is shown in purple in a, b. c Proximity histogram for Zr and O concentrations across the oxide/metal interface.

Next, this same needle was oxidised for 1 min at 300 °C–10.35 mbar O2 and characterised thereafter using APT (Fig. 2). The two-dimensional (2-D) compositional map of O in Fig. 2a highlights the oxidation being most prominent at the tip apex, up to a maximum thickness of 8 nm. The oxide/metal interface is represented by an isoconcentration surface at 30 at. % O (Fig. 2b). A quantitative proximity histogram across a subvolume of this interface (white dashed outline in Fig. 2b) in Fig. 2c shows that the O:Zr ratio throughout the oxide is approaching 1:1, suggesting a nearly stoichiometric ZrO phase. The interfacial width (calculated as the distance between O concentrations of 20 and 80 at. % of the maximum in the proximity histogram given in Fig. 2c) is 4 nm, whereas the physical roughness (defined as root-mean-square roughness) is 1.02 nm. Roughness values were calculated from a cube region of interest (30 × 40 × 80 nm) placed in the centre of the reconstruction, containing the oxide/metal interface, with voxel size 1 × 1 × 1 nm and delocalisation parameters of 3 nm (x) × 3 nm (y) × 1.5 nm (z).

Fig. 2: Compositional analysis of Zircaloy-4 after oxidation at 300 °C–10.35 mbar O2–1 min.
figure 2

a 2-D contour plot for O, where concentration is shown on the colour scale to vary between 0 and 70 at. % O, where white arrows identify regions of non-uniform O penetration in the base alloy, b 3-D element distribution map showing Zr (purple dots) and ZrO ions (light blue dots), and a 30 at. % O isoconcentration surface (light blue). c Proximity histogram for Zr and O concentrations across the oxide/metal interface. Inset in c shows the subvolume analysed, which corresponds to the region outlined by the dashed line in b.

Oxygen penetration into the base alloy below the oxide–metal interface is evident in the 2-D concentration plot of Fig. 2a. White arrows in Fig. 2a indicate local regions of O enrichment within the alloy and highlight the nonuniformity of O penetration.

Oxide evolution after 5 min at 300 °C

Further development of the oxide film was studied by analysing the oxide layers after exposing the APT needle at 300 °C for 5 min at 10.35 mbar O2 (Fig. 3). The mass-to-charge spectra before and after oxidation experiments are provided in Supplementary Fig. 1. The 30 and 55 at. % O isoconcentration surfaces (light blue and green, respectively) overlaid on the 2-D concentration map of O in Fig. 3a and the ionic view in Fig. 3b are markers for the oxide/metal interface (30 at. % O) and the compositional transition in the oxide (55 at. % O). The oxide–metal interface has a root-mean-square roughness of 1.48 nm, while the 55 at. % O isoconcentration surface has a root-mean-square roughness of 1.93 nm, which serves as a proxy for the diffuse ZrO/ZrO2 interface. Note that the oxide layer formed on the base alloy needle is asymmetrical (Fig. 3a, b). The compositional partitioning across the two different oxide/metal interface facets, identified using the 30 at. % O isoconcentration surfaces, were calculated using proximity histograms. The specific facet subvolumes selected for proximity histograms are marked by white dashed outlines in Fig. 3b. These subvolumes are shown as insets within the corresponding proximity histograms in Fig. 3c, d.

Fig. 3: Compositional analysis of Zircaloy-4 after oxidation at 300 °C–10.35 mbar O2–5 min.
figure 3

a 2-D contour plot for O at. %, where the colour scale indicates that concentration varies between 0 and 70 at. %. Isoconcentration surfaces of 30 at. % O (light blue surface) and 55 at. % O (green surface) are provided to highlight the layered composition. b 2-D contour plot for O at. % without a 55 at. % O isosurface with same O concentration scale as a. c 3-D element distribution map showing Zr (purple dots) and ZrO (light blue dots) ions, with 30 at. % O and 55 at. % O isoconcentration surfaces. c Proximity histograms for Zr and O concentration across the oxide/metal interface for the two regions outlined in c. Dashed white lines and numbers in c indicate the subvolumes that were used to generate proximity histograms shown in d, e. Insets in d, e correspond to the regions labelled 1 and 2 in c.

The composition profiles show a layered oxide structure that appears to be, from base metal outward, ZrO, ZrO2−x (0 < x < 1), and ZrO2 phases. The inner ZrO layer is conformal with the metal surface and uniformly ~7 nm thick, whereas the outermost ZrO2 layer varies between 7 and 15 nm thick. These two layers are separated by an intermediate layer of ZrO2−x (~5 nm thick). The O:Zr ratio varies monotonically from ZrO to ZrO2 through this region. This monotonic variability in O:Zr through the intermediate oxide seems invariant to direction of analysis in the reconstructed data across all facets, suggesting that it is a real feature of the oxide structure and not a field evaporation artefact.

APT also reveals that the oxygen concentration in the base alloy can vary across the oxide/metal interface. For example, in Fig. 3c the O concentration declines from 12 at. % to 5 at. % over 23 nm. This region of higher O concentration (relative to 1.2 at. % O in the base alloy) is indicated with a white arrow in Fig. 3a. Conversely, in the subvolume shown in the inset of Fig. 3d taken from another facet, the interfacial width of the oxide–metal interface is <1 nm.

Zr-based alloys are known to form hydrides readily, and hydride formation and oxidation are reported to be interrelated phenomena2; hence, hydrogen (H) in the Zircaloy-4 base alloy and oxide layers was also considered in APT analysis. Results, reported in Supplementary Fig. 2, show that H concentrations in the base alloy were similar before and after oxidation, and H in the oxide was minimal (~2–4 at. %).

Structural characterisation of the oxide layers

Correlative STEM analysis was conducted to investigate the structure of the oxide phases formed on the APT needle sample after 5 min of oxidation at 300 °C. The oxide layer is 20–25 nm thick, polycrystalline with equiaxed grains of size ~5–10 nm, and encapsulates the specimen apex and sides, as seen in Fig. 4a–c. The yellow dotted lines in Fig. 4d highlight several discrete oxide grains. The atom probe needle was tilted in TEM to align with a low-index zone axis. Lattice spacings and crystal symmetry were identified for the outer regions of the oxide layer with the needle aligned to the [\(01\overline 1 2\)] zone axis (N/L = 1.52, where N = 2.55 and L = 1.68) in the hexagonal close packed (HCP) structure of Zircaloy-4. A selected area diffraction pattern taken from the base alloy is shown in Fig. 4e. The lattice parameters of the base alloy were estimated to be a = 3.36 Å and c = 5.28 Å. The values for a and c are ~4% higher than those reported in the literature (a = 3.23 Å and c = 5.15 Å)41, which may be attributed to lattice expansion (and thus volume expansion) associated with oxidation.

Fig. 4: Structure and morphology of the initial oxide layer formed on Zircaloy-4.
figure 4

a HAADF-STEM image showing the oxide film formed on the APT needle after 5-min exposure at 300 °C–10.35 mbar O2. b, c EDS maps of Zr and O, respectively. d High-resolution (HR)-TEM image of the oxide formed on the outer edge of the needle. e SADP, with brightest spots from the Zircaloy-4 matrix aligned with the [\(01\overline 1 2\)] HCP zone axis. f, g Indexed FFT images from grains labelled ‘1’ and ‘2’ in d, respectively.

As we examined the nanograins near the periphery of the needle sample, we found that there were several overlapping oxide grains. Grains labelled ‘1’ and ‘2’ in Fig. 4d were analysed by examining their respective fast Fourier transformations (FFTs), provided in Fig. 4f, g. The atom spacing and crystal symmetry in grain 1, shown in Fig. 4f, indicate the cubic crystal structure (N/L = 1.65, where N = 2.8 Å and L = 1.7 Å), which compositionally agrees well with the ZrO suboxide identified via APT. Based on only FFT examination of grain 1, it is difficult to distinguish between the cubic structure aligned in the [211] direction with lattice parameter 4.85 Å and the tetragonal structure aligned in the [011] direction with lattice parameters a = 3.40 Å and c = 4.93 Å. The analysed grain close to the outermost surface of the needle sample (grain 2 in Fig. 4g) matches the tetragonal crystal structure, which agrees well with the ZrO2 composition (Fig. 3). The grain was consistently indexed as tetragonal structure aligned in the [011] direction (N/L = 1.68, where N = 3.10 and L = 1.84) with lattice parameters close to a = 3.68 Å and c = 5.75 Å. Note that the extra reflections observed at (\(\overline 1 00\)) locations are kinematically forbidden but can be expected due to dynamic scattering. A wide range of lattice parameters for Zr oxide phases have been reported in the literature. Lattice parameters commonly reported for the m-ZrO2 (monoclinic) are a = 5.1690 Å, b = 5.2103 Å, and c = 5.3106 Å and for t-ZrO2 (tetragonal) are a = 3.6070 Å and c = 5.1660 Å42. In our results, we do not see a clear indication of m-ZrO2 after 5 min of oxidation. The outermost region corresponds to t-ZrO2.

Redistribution of Sn during oxidation

Tin (Sn) is the major alloying element in Zircaloy-4, and its distributions after 1 and 5 min of oxidation are presented using 2-D concentration maps of Sn and one-dimensional (1-D) concentration profiles in Fig. 5. We observed a heterogeneous distribution of Sn with local enrichments up to 4 at. % after 1 min of oxidation (Fig. 5a). A 1-D concentration profile in Fig. 5b from the same condition indicates enrichment of Sn within oxide (~2.5 at. %), with a Sn-depleted region of ~5 nm wide (Sn <0.5 at. %) in the base alloy adjacent to the oxide/metal interface. Interestingly, after 5 min of oxidation, we observed a redistribution of Sn across different oxide/metal interface facets. Sn appears to either preferentially segregate at the oxide/metal interface (~2 at. %) or show a depletion (e.g. Fig. 5d versus Fig. 5e). This observation suggests that crystallographic features influence local segregation of Sn at the oxide/metal interface.

Fig. 5: Redistribution of Sn after 1 and 5 min of oxidation at 300 °C–10.35 mbar O2.
figure 5

a 2-D Sn concentration map after oxidation for 1 min and b 1-D concentration profile across the oxide/metal interface, highlighting the change in Sn concentration. c 2-D Sn concentration map after oxidation for 5 min and d, e corresponding 1-D concentration profiles across the oxide/metal interface, highlighting the change in Sn concentration. Cylinder placements from which the concentration profile plots were generated are shown as insets in d, e. 30° at. % O isoconcentration surfaces are overlaid in a, c and insets of d, e show the location of the oxide/metal interface.

The Sn distribution inside the oxide layer after 5-min oxidation is shown in Fig. 6. The 2-D Sn concentration maps from several locations, consisting of 1-nm-thick rectangular slices from the oxide layer (Fig. 6b–d), reveal local variation of Sn concentration ranging from 0 to 2.3 at. %. The heterogeneous enrichment of Sn outlines nanocrystalline oxide grain boundaries, as seen in the high-resolution TEM image in Fig. 4d. The grain size was measured to be 5–10 nm in TEM, which is consistent with the size of Sn-free regions enclosed within the Sn-rich regions observed here.

Fig. 6: Distribution of Sn in the oxide layer after oxidation at 300 °C–10.35 mbar–5 min.
figure 6

a APT reconstruction with overlaid locations of 2-D concentration maps from oxide bd shows magnified 2-D Sn concentration maps inside the oxide. A slice thickness of 1 nm was used, from which the Sn 2-D concentration maps were determined. Colour scale for each 2-D map shows the range of Sn concentration variation.

XANES analysis of changes in electronic structure of oxide layer

O K-edge total fluorescence yield XANES was used to study the changes in electronic structure of the surface of Zircaloy-4 in the as-received condition and after oxidation at 300 °C–10.35 mbar for 1 min to 16 h. Figure 7 shows absorption intensity versus energy for Zircaloy-4 oxidised for 1 min and 16 h overlaid with a standard m-ZrO2 sample and unoxidised Zircaloy-4. Spectral features contain two regions describing different electronic transitions: (1) O 2p hybridised with Zr 4d, and (2) O 2p hybridised with Zr 5sp43. In the first transition region, two peaks (eg and t2g) are seen in both the 16-h oxidised sample and the m-ZrO2 standard. These two peaks at the O K-edge are attributed to hybridisation of O 2p states with Zr 4d-eg and 4d-t2g states44. At energies >540 eV, broad spectral features (corresponding to O 2p and Zr 5sp interactions) are observed for each oxidation condition and the m-ZrO2 standard. The O 2p-Zr 5sp spectral features are known to be related to hybridisation45, where similar spectra are observed for all sample conditions. This similarity in the post-edge regime suggests that there are no considerable changes in hybridisation for different Zr oxide structures, consistent with prior work46. Results clearly show a change in electronic structure as a function of oxidation conditions. The spectra for unoxidised Zircaloy-4 and Zircaloy-4 oxidised for 1 min do not match the O K-edge spectra of the m-ZrO2. Only after 16-h oxidisation do the spectra resemble the m-ZrO2 standard. This result suggests that mixed metastable suboxides are present after short oxidation time steps, which is consistent with the APT and STEM results.

Fig. 7: Synchrotron-based O K-edge XANES spectra for Zircaloy-4 samples.
figure 7

A monoclinic ZrO2 standard compared with as-received Zircaloy-4, and Zircaloy-4 oxidised at 300 °C and 10.35 mbar for 1 min and 16 h.

Discussion

In this study, APT, STEM, and XANES were used to investigate the multiple atomic-level changes that occur across oxide/metal interfaces and within the oxide during the early stages of Zircaloy-4 oxidation. The composition and structure (crystal and electronic) changes that occur during the initial stages of oxidation are of particular importance for Zircaloy-4 because this alloy experiences cyclic oxidation kinetics. Hence, these initial stages of oxidation may occur multiple times throughout a material’s lifetime in an oxidising environment. This approach developed and applied to studying Zircaloy-4 oxidation provides insight both into the initial changes accompanying oxidation and demonstrates a methodology for assessing such changes in accelerated time scales than available using traditional long-term experiments in autoclave or dry oxidation environments.

Prior to analysis of oxides formed after 300 °C oxidation, we identified a thin native oxide layer on APT needle samples comprising ~30 at. % O, with stoichiometry close to Zr3O. This native oxide was removed by field evaporating the needle in APT past the oxide layer. Then the needle was transferred in vacuum to the reaction chamber and oxidation was performed at 300 °C to investigate structural and compositional changes that occur during oxidation. After 1 min of oxidation at 300 °C–10.35 mbar O2, a ZrO layer was formed. After 5 min at 300 °C–10.35 mbar O2, a ZrO layer formed adjacent to the metal/oxide interface, followed by suboxides with a stoichiometry of ZrO2−x (0 < x < 1), and a tetragonal structured ZrO2 layer formed as the dominant outer oxide layer. Figure 8a–c shows sections of the APT reconstruction after 5-min oxidation at 300 °C–10.35 mbar, where Fig. 8d is a schematic representation of the multilayered film. The oxide layer contained equiaxed, nanocrystalline grains, which could be a result of oxide grains nucleating on several surface defects47. The Zr suboxides appear to retain an epitaxial relationship with the base alloy, which has lattice parameters slightly larger than ones reported in the literature. Larger lattice parameters may be attributed to oxygen atoms occupying interstitial sites in the lattice of the HCP base alloy. The tetragonal ZrO2 phase is observed in the outermost region of the oxide after 5 min, with no evidence of the stable monoclinic ZrO2. The presence of only the tetragonal phase in the outer oxide may be attributed to insufficient time for the transformation from tetragonal to monoclinic to proceed. It may also be possible for the tetragonal phase to be stabilised in the outer oxide due to a combination of factors detailed in prior work, such as macroscopic compressive stresses, nanoscale oxide grain size, presence of Sn in the oxide, and oxygen vacancies48,49,50.

Fig. 8: Summary of composition and structure across the oxide/metal interface.
figure 8

a Element distribution after 300 °C–10.35 mbar O2–5 min for Zr (purple dots) and ZrO (light blue dots) ions. b 2-D concentration plot illustrating the O concentration gradient across the oxide/metal interface. c 2-D concentration plot for Sn. d Schematic of the nanocrystalline oxide and redistribution of Sn developed after 5 min of oxidation treatment. Here c-ZrO refers to face-centred cubic (FCC) ZrO phase, ZrO2−x indicates the region with graded composition, and t-ZrO2 refers to the tetragonal ZrO2 phase.

Addition of Sn, the major alloying element in Zircaloy-4, is known to improve yield strength and creep resistance via solid solution strengthening but negatively affect oxidation resistance29,31,51. During oxidation, Sn is incorporated into the oxide31, which can influence oxide structure and composition51. The grain size was measured to be 5–10 nm in TEM shown in Fig. 4d, which is consistent with the size of Sn-free regions enclosed within the Sn-rich regions observed in APT. Sn segregation to oxide grain boundaries has previously been suggested and may stabilise tetragonal ZrO2 grains31,51. The Sn segregation observed after 5-min oxidation at 300 °C can therefore indicate that a similar phenomenon is evident in experiments performed as part of this work (Fig. 6). These local regions of Sn enrichment are up to ~3 at % Sn and may be precursors to larger Sn clusters (~20 nm in diameter) with higher Sn content that require more time at elevated temperature to develop31. The lack of Sn precipitates observed here may be attributed to the shorter time scales used in oxidation experiments when compared to the longer-term studies performed after several days in a 400 °C autoclave31.

Our results indicate that Sn is distributed nonuniformly within the base alloy, along the oxide/metal interface, and within the oxide layer. The segregation of Sn to ZrO after oxidation for just 1 min (Fig. 2) and subsequent partitioning/depletion of Sn along the oxide/metal interface after 5 min (Fig. 3) depict the early stages of Sn partitioning into the oxide, which have not been captured in prior ex situ analyses of corroded Zr alloys. The multistage oxide growth and associated Sn redistribution is schematically described in Fig. 8d.

Prior work investigating the relationship between Zircaloy-4 oxidation behaviour and oxygen partial pressure demonstrated that structure changes during the oxidation process in various environments (e.g. oxygen gas, water, steam) did not vary with oxygen partial pressure (via X-ray diffraction analysis) in the pre-transition regime52. Post-transition, however, oxidation rate is weakly dependent on oxygen partial pressure52. Here our analysis is focused in the pre-transition regime, so the oxidation environment is not expected to strongly contribute to any structural or compositional changes of the oxide layer but is expected to affect kinetics. Other important factors contributing to oxidation kinetics in the present study include crystallographic orientation of the base alloy and needle specimen geometry. It is well established that oxidation rate depends on crystallographic orientation of the metal substrate15,]53. Here each needle was extracted from the same grain (where Zircaloy-4 grains were ~15–25 µm in diameter) and thus are expected to have same orientation. However, the non-uniform penetration of oxygen into the base alloy after 1 and 5 min (Figs. 2 and 3) suggests that different facets may be present on the surface of the needle. Additionally, needle specimen shape has previously been shown via in situ TEM to increase oxidation rate and lead to different oxide thicknesses at the apex and along the shank54. The impact of curvature has also been investigated by studying oxide phases developed on flat coupons versus nanoparticles (with similar diameter to an APT needle apex)35,55. Substrate geometry (flat versus curved) was found to accelerate kinetics but not impact oxide phases formed.

Metastable suboxides including hexagonal Zr3O, cubic ZrO, and tetragonal ZrO2 have all been reported to form after exposure to an oxygen-rich environment for a range of times, from <1 day56 to hundreds of days19,29,30 in various environments (aqueous19,27,30, steam50,57,58,59, and dry oxygen gas56). More specifically, pre-transition suboxides have previously been reported with thicknesses of 30–80 nm (studied via APT) developed over ~34–90 days in an aqueous environment19. Here a similar oxide thickness of ~20 nm is developed on the APT-STEM needle specimens in a matter of minutes, with comparable suboxide phase compositions and structures. Owing to the needle geometry, oxidation kinetics is significantly accelerated using the in situ APT approach, allowing us to rapidly assess composition/structure changes occurring during initial stages of Zr alloy oxidation, and effectively reduce the time required for oxidation experiments.

The ZrO2−x phase with a large lattice parameter identified in our work consisted of much higher O concentration than the previously reported composition range (40–50 at. % O) for suboxides with unidentified crystal structure19,60. Even though the needle-shaped geometry in the current analysis results in multiple overlapping nanocrystalline grains—making it difficult to unambiguously ascertain the crystal structure of this suboxide layer using STEM—we observe face-centred cubic ZrO and tetragonal ZrO2 suboxides, consistent with prior work19; we find no evidence for hexagonal Zr3O19,27,61 or monoclinic ZrO2 after oxidation at 300 °C for up to 5 min in the current work. We also find variable and significantly lower local oxygen concentration (relative to the ~30 at. % solid solubility limit of O in α-Zr) in the base alloy, ranging from 2 to 20 at. %, below the oxide–metal interface (see Fig. 3c, d). The differences in the structures and compositions of oxide phases between our experiments and prior work from literature is attributed to our ability to capture very early stages of the oxidation progression that have not been possible to resolve previously.

Quantifying oxide stoichiometry is a significant challenge given the known oxygen deficits in laser-assisted APT62,63,64,65. Here the application of complementary microscopic and spectroscopic methods allows us to assign oxide stoichiometries to the phases formed during the in situ experiments. The oxide stoichiometries reported here, in particular the ZrO and ZrO2m, are confirmed by STEM measurements that identified cubic and tetragonal oxide phases (after oxidation at 300 °C for 5 min), which have known stoichiometries of ZrO and ZrO2, respectively. Hence, we assign the cubic inner oxide to ZrO and outer oxide to tetragonal ZrO2, and thus the region of composition ZrO2−x is a transition region between these two known phases. To further investigate error associated with Zr oxide quantification via APT, standard ZrO2 may be run using similar user-selected parameters.

In this work, structural, compositional, and electronic structure changes during the early stages of Zircaloy-4 oxidation were studied using a multimodal chemical imaging approach. Needle-shaped Zircaloy-4 samples were oxidised at 300 °C in 10.35 mbar, high-purity O2 gas for 1 and 5 min in an environmental reaction chamber attached to an APT system. APT and correlative STEM analysis of the oxidised needles permitted a rapid assessment of metastable suboxides, revealing formation of an initial cubic ZrO layer, followed by a transition layer of ZrO2−x (with composition and structure previously not reported in literature) and tetragonal ZrO2 as the outermost layer. Sn enrichment into the ZrO suboxide phase, followed by its segregation to different regions of the oxide/metal interface, was also observed. XANES analysis of Zircaloy-4 coupons complemented the APT/STEM characterisation and showed that the stable monoclinic oxide did not form until longer oxidation times, on the order of hours. The approach presented here can be valuable for rapidly assessing structure and composition changes during early stages of oxidation in several commercial alloys for oxidation- and corrosion-sensitive applications.

Methods

Materials and needle sample preparation

A Zircaloy-4 tube, with composition 1.02 ± 0.03 Sn, 0.72 ± 0.02 O, 0.38 ± 0.01 Fe, and 0.23 ± 0.01 Cr (all in at. %), with balance Zr, was sectioned and ion milled to produce a polished cross-section. Composition specification and impurities are detailed in Supplementary Tables 13 and Supplementary Fig. 3. Ion milling was used in lieu of standard metallographic polishing procedures to minimise hydrogen uptake of Zircaloy-4 from aqueous metallographic polishing. Needles for APT and TEM analysis were prepared from the polished cross-section via the standard FIB-based lift-out process66. Samples were mounted to pre-sharpened tungsten APT/TEM correlation grids67 and silicon microtip arrays for TEM and APT analysis, respectively.

Characterisation of composition and structure via APT and STEM

The experimental approach for complementary STEM and APT analysis of oxide and Zircaloy-4 base metal (schematically shown in Supplementary Fig. 4) involves the following steps: (1) needle sample preparation via FIB lift-out method66, (2) analysis and removal of the native oxide via APT field evaporation, (3) oxidation of the needle samples in the reaction chamber, and (4) APT or STEM analysis of the oxide layer formed during the oxidation process.

Oxidation of Zircaloy-4 needles was performed in O2 gas (99.993% purity) at 10.35 mbar and 300 °C for 1 or 5 min. Experimental conditions were selected to perform Zircaloy-4 oxidation at temperatures relevant to nuclear reactor applications. Time and pressure parameters were selected to control Zircaloy-4 oxidation so that the oxide/metal interface was captured in the top ~50 nm of the needle specimen and could then be analysed via APT. After the initial APT analysis and removal of the native oxide, the needle array was transferred under UHV (10−7–10−8 mbar) to the reaction chamber, where the stage was preheated to 300 °C68. The UHV transfer time from APT to reaction chamber was <5 min and prevents any atmospheric exposure during the transfer. Pressure in the reactor prior to oxygen gas exposure was ~10−8 mbar. Next, O2 gas (which was not preheated) was introduced into the reaction chamber to the desired pressure of 10.35 mbar at a rate of 100 cm3/min, and the APT needles were exposed to O2 gas for the desired time interval (here 1–5 min). The reactor was then immediately evacuated to remove all gas and return to UHV conditions. The oxidised sample was then moved directly to the APT analysis chamber, which takes ~2 min. In the analysis chamber, the sample was rapidly cooled by placing it on the cryogenically cooled sample stage set to a temperature of 45 K.

A Cameca LEAP 4000X HR system equipped with a 355 nm wavelength ultraviolet laser was used for all APT data collection with the following user-selected parameters: 125 kHz pulse frequency, 200 pJ/pulse laser energy, 45 K sample base temperature, and 0.005 detected ions/pulse. All APT data sets were reconstructed and analysed using the Integrated Visualisation and Analysis Software (IVAS), version 3.8.2. Manual ranging was used for analysis of the APT mass spectra. Peak overlaps between Sn isotopes and Zr oxide molecular species (e.g. Zr2O3 and ZrO2) were seen in mass spectra; these are detailed in Supplementary Fig. 5. Regions of gallium (Ga) implantation from the FIB processing were excluded from the analysed volumes discussed here. Further details on APT data handling, including the Ga distribution and composition quantifications, are provided in Supplementary Figs. 68.

Concentration profiles across different facets of oxide/metal interfaces were quantified using the proximity histogram method via isoconcentration surfaces generated at 30 at. % O. 2-D concentration plots were also used to quantitatively investigate O and Sn composition in the oxide and the base alloy. 2-D concentration plots were generated using a rectangular cross-section 1 nm thick. Crystallographic and compositional examination of Zircaloy needle samples was done with a probe-corrected FEI Titan 80–300 STEM instrument operated at 300 kV. The observations were performed using STEM with a high-angle annular dark-field (HAADF) detector. The probe convergence angle was 18 mrad, and the inner detection angle on the HAADF detector was three times larger than the probe convergence angle. AZtec energy dispersive spectroscopy (EDS) software was used to analyse the composition captured by EDS in the STEM instrument.

Electronic structure change characterised via XANES

The electronic structure changes of an unoxidised Zircaloy-4 coupon, ones oxidised at 300 °C for 1 min, 5 min, and 16 h, and a m-ZrO2 standard were analysed using synchrotron-based XANES at Advanced Light Source synchrotron beamline 7.3.1 at Lawrence Berkeley National Laboratory. XANES spectra were collected with an overall resolution of 0.1 eV at the oxygen K-edge. Experiments were conducted in a UHV chamber with a background pressure of 2 × 10−10 mbar. Photon energies were varied from 525 to 560 eV. Data were acquired in total fluorescence yield and total electron yield to obtain oxidation information at the surface (<10 nm) and deeper (<100 nm). The O K-edge total fluorescence yield signal was utilised in this work to characterise the top 100 nm of all samples. Acquired spectra were normalised to background incident intensity.