Deciphering atomistic mechanisms of the gas-solid interfacial reaction during alloy oxidation

See allHide authors and affiliations

Science Advances  24 Apr 2020:
Vol. 6, no. 17, eaay8491
DOI: 10.1126/sciadv.aay8491


Gas-solid interfacial reaction is critical to many technological applications from heterogeneous catalysis to stress corrosion cracking. A prominent question that remains unclear is how gas and solid interact beyond chemisorption to form a stable interphase for bridging subsequent gas-solid reactions. Here, we report real-time atomic-scale observations of Ni-Al alloy oxidation reaction from initial surface adsorption to interfacial reaction into the bulk. We found distinct atomistic mechanisms for oxide growth in O2 and H2O vapor, featuring a “step-edge” mechanism with severe interfacial strain in O2, and a “subsurface” one in H2O. Ab initio density functional theory simulations rationalize the H2O dissociation to favor the formation of a disordered oxide, which promotes ion diffusion to the oxide-metal interface and leads to an eased interfacial strain, therefore enhancing inward oxidation. Our findings depict a complete pathway for the Ni-Al surface oxidation reaction and delineate the delicate coupling of chemomechanical effect on gas-solid interactions.


Gas-solid reactions are central to many technological processes, including stress corrosion cracking (14), hydrogen embrittlement (5, 6), and heterogeneous catalysis (711). These reactions generally initiate via gas adsorption, followed by nucleation and growth of a new phase, whereas canonical models are traditionally based on gravimetric measurements that oftentimes exclusively consider the growth process only. Moreover, new phase formation generates a new heterointerface between the base solid and newly formed phase that can play an important role in the subsequent reaction processes. For instance, epitaxial oxide formation on an alloy surface produces an interfacial strain that can result in cracking and/or spalling of the oxide layer, and cracks or strained intergranular paths will affect the ion diffusivity and lead consequently to an altered, often accelerated, oxidation rate. These coupled mechanical-chemical effects can markedly affect the oxidation- or corrosion-resistant properties of alloys and also lead to a sudden mechanical failure of alloys (12, 13). These mechanically coupled chemical processes are also found in other gas-metal interactions, i.e., the compressive strain in supported metal nanoparticle catalysts can largely enhance the catalytic activity of the metal (e.g., Pt) due to a shift of the electronic band structure and weaken the chemisorption of oxygenated species (14, 15).

While substantial efforts have been exerted to decouple the mechanical and chemical factors in macroscale corrosion phenomena (16), little is known about their initial entanglement at the nanoscale to atomic scale. For the technologically important Ni-Al alloys, attempts have been made to determine the atomic structure of ultrathin surface layer formed on β-NiAl under ideal conditions (17, 18). Recent studies have shown that strain-mediated ionic transport can enhance the oxidation of metal nanoparticles (19), and in situ studies have revealed the critical role of surface defects during the initial oxidation of alloys (2022). However, a full picture of the atomic-scale processes during gas-metal reactions including possible strain-coupled mechanisms has yet to be reported.

Here, we use in situ atomic-scale environmental transmission electron microscopy (ETEM) to visualize the structural and phase evolution on a single-crystalline Ni–5 atomic % (at %) Al (A1 phase) surface at 350°C, showing a typical gas-metal reaction pathway from initial surface dynamics to bulk phase change accompanied by chemomechanical coupling. Comparative studies demonstrate that different oxidizing agents (O2 and H2O vapor) lead to distinctive atomic mechanisms of oxide growth and kinetics, where the evolving interface plays a substantial role. The experimental observation is corroborated by density functional theory (DFT) calculations and simulations, providing unprecedented atomic insight into the unique role of water vapor in modifying the structures of growing oxides and interfacial structure during oxidation.


Distinct atomic processes for initial oxidation of Ni-Al alloy in O2 and H2O vapor

Figure 1 shows the in situ atomic-scale observations of the onset of oxidation process from a clean Ni-Al (110) surface in O2 and H2O (1 × 10−6 mbar) at 350°C, respectively, via high-resolution TEM (HRTEM). In O2, the oxidation proceeds through a “step-edge” growth mechanism, i.e., step edges on the alloy surface are the primary source of reacting metal atoms that enable nucleation and growth of oxide islands as indicated by ledge migration (white arrows in Fig. 1, A and B). An energy barrier exists at the oxide-alloy interface, inhibiting O subsurface diffusion to form a “subsurface” oxide (20). As a consequence, the alloy surface atoms are peeled off “layer by layer” as the oxide islands only grow on top of an intact surface. In direct contrast, the oxidation in H2O vapor under the same conditions exhibits a direct subsurface growth mode (Fig. 1, C and D), i.e., oxide islands grow into the alloy surface (yellow dashed lines), absorbing adjacent metal atoms to sustain inward oxide growth. This growth model indicates that the oxide-alloy interface does not prohibit subsurface diffusion, enabling a direct “attack” to the pristine alloy material by advancing oxidizing species. In the regime of initial oxidation, an increased oxidation rate is also found for the oxidation in H2O by comparing the consumed atomic layers of alloy substrates in Fig. 1 (A and C). These observations demonstrate different atomistic processes for initial alloy oxidation under O2 and H2O gas environment.

Fig. 1 Step-edge oxide growth in O2 versus subsurface oxide growth in H2O.

(A and C) Time-lapse HRTEM images demonstrate the atomic-scale process of Ni–5 at % Al (100) alloy surface oxidation in O2 and H2O, respectively, at 350°C. (A) Surface-only oxide growth on the Ni-Al surface in O2 through a step-edge mechanism, where only Ni atoms from step edges participate in the oxide growth. This mechanism is evidenced by the migration of step edges (white arrows), leaving intact alloy surfaces (white dashed lines). (C) Subsurface oxide growth into the Ni-Al in H2O vapor, where alloy atoms adjacent to the oxide island are “absorbed” into the growing oxide. The initial Ni-Al surface is indicated by the white dashed line, and the oxide islands are indicated by yellow dashed lines. (B and D) Illustration of different atomistic processes in O2 and H2O, respectively. Step edges are indicated by white arrows in (A). Scale bars, 1 nm (A and C).

Changing oxidizing species from O2 to H2O introduces adsorbed hydroxyl through water dissociation, and OH can further dissociate into atomic H and atomic O, where atomic O serves as oxidizing agents (23). It may be argued that, associated with the charge transfer between Ni and Al (24), alloying of Ni with Al could possibly change the surface adsorption behavior. However, the oxidation process is thermodynamically dominated by the oxide formation energy, therefore making the alloying effect less important, especially for the case of a dilute solid solution of ~5 at % Al. The initial oxide formed on dilute Ni alloys has been reported to be NiO (2529) before possible alloying elements (e.g., Cr and Al) incorporated into the oxides, which is in accordance with the results of our phase identification based on lattice imaging and the subsequent analysis as detailed in fig. S2. To understand the oxidation reaction mechanisms in H2O, we use DFT and ab initio molecular dynamics (AIMD) calculations to elucidate the possible effect of dissociated H2O molecules on oxide formation at a preexisted NiO/Ni interface. The major difference between the oxidation by O2 and H2O lies in the existence of H in the case of H2O. Hence, our simulation focuses on the response of the preformed oxide to H. As shown in Fig. 2, a pyramid-shaped NiO is built on Ni surface according to experimental observations in Fig. 1. Two views of the relaxed NiO/Ni interface are illustrated in Fig. 2 (A and B), where seven Ni (002) atomic layers match with six NiO (002) atomic layers according to their lattice misfit. To simulate the oxide growth process in a water vapor environment, we introduce preadsorbed H atoms and OH groups to the surface of the relaxed structures in Fig. 2 (A and B). As shown in Fig. 2 (C and D), after 10 ps, the NiO became defected, and Ni and O atoms in the oxide became disordered, which shows a clear perturbation on the growing oxide brought by H atoms. For initial oxide embryo nucleation and growth, it usually has a metastable atomic structure due to its small size (<2 nm); hence, the H can easily affect the stoichiometric growth of the oxide as evidenced by HRTEM imaging (Fig. 1). At the same time, H is found to diffuse from the adsorption surface to the oxide-alloy interface (Fig. 2D). It is well known that H can easily diffuse through oxides via interstitial sites with a small diffusion barrier (29, 30); hence, we usually do not expect H to aggregate at specific sites. As shown by the simulation results, the interfacial enrichment of H indicates that the oxide-alloy interface acts as an energetically favorable “sink” for H ions. This is critical for understanding distinct growth mechanisms between dry oxidation and oxidation by water vapor observed in Fig. 1. Specifically, the H generated from water dissociation diffuses to the oxide-alloy interface, which could lead to the destabilization of the oxide-alloy interface, as shown by the direct attack to the alloy substrate in Fig. 1 (C and D). It should be noted that upon adsorption and dissociation process of gas molecules on metal surface, it takes a time scale of several seconds to initiate the oxidation reaction. Therefore, there is a gap in the time scale between the molecular dynamics simulation of several picoseconds and the experimentally observed surface reaction of several seconds. Apparently, the picosecond simulation reveals a snapshot of the dynamics of the whole system. These combined experimental and theoretical results reveal a possible effect of water vapor on the initial stages of oxide formation. It is, however, intriguing to also consider how water vapor affects later stages of the oxidation process.

Fig. 2 Ab initio simulation of H2O-mediated NiO growth on Ni.

(A and B) DFT-relaxed structure of the island and multilayers NiO (100)/Ni (100) interfaces from different views. (C and D) Structures of (A) and (B) after 10 ps of AIMD of NiO (100)/Ni (100) in the presence of preadsorbed H atoms and OH molecules, where defected and disordered NiO is formed by the presence of H2O breakdown products (simulating oxidation in H2O).

Metal-oxide interface structure and its effect on oxidation of Ni-Al alloy in O2 and H2O

The atomic-scale transitions of the oxidation process from initial oxide nucleation to bulk growth in O2 and H2O at a higher pressure (1 × 10−4 mbar) at 350°C are depicted in Fig. 3 (A and B, respectively). Although the oxide growth on a clean NiAl (110) surface features similar stages for the two different gas environments (i.e., proceeding from the formation of metastable surface oxides to coalescence of these surface oxides and subsequent inward oxide growth), the oxidation by O2 (Fig. 3A) exhibits a highly strained interface (white dashed lines) between oxide and alloy, propagating into the alloy and forming a dislocation zone (between white and yellow dashed lines). On the contrary, the oxidation by H2O vapor exhibits neither lattice distortion nor dislocation formation. Instead, a disordered oxide and incoherent interface between oxide and alloy accommodates the lattice mismatch. While the breakdown of the step-edge growth mechanism in O2 is due to a higher chemical potential for the oxidation reaction at a higher gas pressure, the lattice mismatch–induced strain at the coherent alloy-oxide interface contrasts that in H2O with a noncoherent interface, as shown in Fig. 3 (C and D). Figure 3C shows four structurally distinctive zones, i.e., from surface to the inland of alloy, disordered surface oxide, well-ordered or “bulk” NiO, dislocation zone, and alloy substrate as sequentially labeled as zones 1 to 4. In direct contrast, the interfacial region formed in H2O lacks both a distinct dislocation zone and a zone of well-ordered epitaxial NiO, as shown in Fig. 3D. Detailed analysis of the oxide-alloy interface formed in O2 finds that six layers of oxide lattice correspond to seven layers of alloy lattice, which is in accordance with the lattice mismatch (16.67%) between cubic NiO (α = 4.17 Å) and face-centered cubic Ni–10 at % Al (α = 3.575 Å) (fig. S3A). It also finds that the interfacial strain is reduced from 16.67 to 11.76% through the incoherent interface formed in H2O (fig. S3B). It needs to be mentioned that the Moiré patterns shown on the alloy substrate lattice in Fig. 3 (B and D) were due to the surface oxide formation on the Ni-Al (100) surface (the projection view), as shown in fig. S4.

Fig. 3 Strain-mediated Ni-Al oxidation by O2 and H2O.

Time-lapse HRTEM images demonstrate atomic-scale cross-sectional view of strain-mediated alloy oxidation process (1× 10−4 mbar and 350°C) on the Ni-Al surface. (A) Oxidation in O2 initiates from surface oxide islands (white arrows; 6 s) to inward oxide growth into the alloy substrate (the interface indicated by white dashed lines; 73 s) and forms a highly strained dislocation zone (between yellow and white dashed lines; 100 s). (B) Oxidation in H2O initiates from surface oxide islands (30 s) to inward oxide growth into the alloy substrate (the interface indicated by white dashed lines; 42 and 64 s), but the growing interface is absent of dislocation zone compared with that in O2. (C) HRTEM image shows four zones at the growing oxide-alloy interface in O2, from the surface to bulk: 1. Surface oxides; 2. Bulk NiO; 3. Dislocation zone (the interfacial area between oxide and substrate); 4. Alloy substrate, as illustrated by the atomic structure model. (D) HRTEM image shows only two zones presented at the growing oxide-alloy interface in H2O: 1. Oxide (less ordered) and 4. Alloy substrate, as illustrated by the atomic structure model. Scale bars, 2 nm (A) and 1 nm (B to D).

As shown above, the formation of the strained interface during inward/bulk oxidation requires extra energy for the proceeding of oxidation reaction, which became the dominating factor in the oxide growth in O2. In the oxidation by H2O, this strained interface is relaxed by forming an incoherent interface between the oxide and alloy substrate in H2O. Strain mapping based on the in situ HRTEM observation confirms that considerable strain accumulated at the interface between oxide and alloy substrate for the oxidation by O2 (Fig. 4, A and B), while a relatively mild and uniform strain distribution is shown for the oxidation by H2O (Fig. 4, C and D). These different strain levels have an immediate impact on the following growth kinetics. In Fig. 4E, we tentatively quantify the oxide growth rate in two environments by comparatively counting the oxide layers formed in a period of time. By real-time HRTEM imaging, we could also differentiate the “outward oxidation,” which means the addition of surface oxide layer to the initial alloy surface, from the “inward oxidation,” which means that the oxide forms by incorporating oxygen atoms into or below the initial alloy substrate. The former case is dominated by surface diffusion of O and metal atoms, while the latter case is dominated by diffusion of O to the alloy-oxide interface. Assuming a uniform growth of oxide, a faster inward oxidation is seen for the oxidation by H2O comparing with that by O2, while no noticeable difference is observed for outward oxidation between two cases as seen in Fig. 4E, where a clear enhanced inward oxidation in H2O is seen.

Fig. 4 Strain analysis of alloy-oxide interface and kinetics of initial Ni-Al oxidation by O2 and H2O.

(A) HRTEM image of the oxide-alloy interface formed in O2. (B) Strain mapping of the oxide-alloy interface formed in O2. (C) HRTEM image of the oxide-alloy interface formed in H2O. (D) Strain mapping of the oxide-alloy interface formed in H2O. (E) Initial oxidation kinetics measured by the number of atomic layers oxide layers formed during the same time span. On the basis of the initial alloy surface, the oxide layers are grown inward and outward as illustrated in the scheme on the right, labeled as + and − on the y axis. Scale bars, 1 nm (A and C).


In summary, early stages of atomic-scale oxidation process are captured on Ni-Al surface by in situ ETEM. As the oxidation proceeds from the surface to the bulk alloy, the nature of the interface between oxide and alloy governs the atomistic processes and oxidation mechanisms. In O2, a nearly coherent highly strained interface inhibits subsurface diffusion of O, which leads to a step-edge mechanism at lower pressures and keeps imposing an energy barrier to retard the oxidation reaction at higher pressures. On the contrary, oxidation in H2O shows a direct subsurface mechanism with an enhanced oxidation rate at both lower and higher pressures compared with that in O2. DFT calculation and simulation reveal the role of dissociated H2O in this strain-coupled oxidation reaction as a structural interrupter and ion diffusion enhancer. These results clarified the atomistic mechanisms of initial oxidation and provide insights into other strain-mediated chemical reactions.


Ni-Al alloy thin-film deposition

Single-crystalline Ni-Al alloy of 50-nm-thick thin film was deposited on a NaCl substrate at 400°C using a dual-beam electron-beam evaporation system. The concentration of each element was controlled through individual evaporation rate (9.5 Å/s for Ni versus 0.5 Å/s for Al), resulting in an atomic ratio of ~5 at % of Al in the Ni-Al alloy films confirmed by energy-dispersive x-ray spectroscopy analysis. Following deposition, the NaCl substrate was removed by water dissolution, and the as-deposited alloy thin films were confirmed to be single crystal via electron diffraction.

ETEM oxidation experiments of Ni-Al alloy

An FEI Titan ETEM equipped with an objective lens aberration corrector was used for Ni-Al oxidation experiments. The single-crystalline (100) Ni–5 at % Al alloy thin films (~50 nm in thickness) were prepared by washing in acetone and methanol several times and mounted on a holey silicon TEM grid for observation. A partial pressure up to a few millibars can be reached in the ETEM and at a temperature up to 850°C can be reached by a Gatan 652 double-tilt heating holder. Before oxidation, the alloy thin films were annealed at 700°C in a high vacuum of <1 × 10−8 mbar (fig. S1), resulting in a clean surface. Pure oxygen (~99.999%) was introduced into the TEM column through a leak valve to oxidize the thin films at a given temperature and pressure. For oxidation in water vapor, the water vapor was generated and delivered to the ETEM column by a water vapor system (29) developed in-house. The gas pressure can be adjusted from 1 × 10−8 to 15 mbar. The electron beam dose rate was chosen as ~5 × 104 e·nm−2·s−1 on the basis of calibrations of the electron beam effect according to the established methods (detailed in the Supplementary Materials).

Strain analysis of HRTEM images

The strain analysis of the time-resolved HRTEM image is conducted by DigitalMicrograph software with a geometric phase analysis in FRWRtools plugins.

Computational methods

The Vienna Ab initio Simulation Package (VASP) (31, 32) was used to perform DFT and AIMD calculations to model the NiO/Ni interface geometry and structural evolution in the presence of dissociated H2O molecules. Projector augmented wave (PAW) (33) atomic pseudopotentials were used to describe the potentials of nuclei and core electrons. The generalized gradient approximation with the parametrization of Perdew-Burke-Ernzerhof (PBE) was used for the exchange-correlation functional (34). The on-site coulomb interaction of the 3d electrons of Ni in NiO was accounted for by adding a Hubbard-U term (35), with a U value of 5.3 eV (36). The lattice parameters of Ni and NiO were calculated to be 3.51 and 4.20 Å, respectively. The NiO island was modeled as pyramid shape to mimic experimental observation. Previous theoretical work showed that the NiO (100) surface is the most stable among all low-index NiO surfaces (37), so the termination of the NiO island parallel to the substrate was chosen as (100). To construct a periodic simulation cell and minimize the strain of NiO oxide, the NiO/Ni interface was modeled with six repetitions of Ni unit cell and five repetitions of NiO unit cell, both along the [100] direction (Fig. 4). This results in a very large simulation cell with 408 atoms in total. The Ni substrate consists of four Ni (100) layers, and the NiO island consists of three NiO (100) layers. The DFT calculation for the NiO (100)/Ni (100) structural relaxation was spin-polarized. Electron smearing was carried out using Methfessel-Paxton smearing technique (38) with a smearing parameter of 0.2 eV. A vacuum spacing of 20 Å separates each NiO/Ni slab from its periodic images to prevent unphysical coupling. The positions of all atoms, except the bottom Ni layer, were allowed to relax in all three directions until the force components acting on each atom were less than 0.02 eV/Å. Because of the large size of the simulation cell, the Brillouin zone integration was performed using only the Г K-point in the reciprocal space, and the soft version of the oxygen PAW potential supplied by VASP was used, in conjunction with a kinetic energy cutoff of 353 eV. Previous studies demonstrated that H2O molecules are likely to dissociate on the Ni (100) surface after adsorption (39). Therefore, AIMD simulations were performed using the canonical ensemble at 623 K to monitor the interface structural evolution with hydrogen atoms (H) and hydroxy groups (OH) on the surface. The van der Waals–inclusive functional optPBE-vdW (40) was used to account for the nonlocal dispersion interactions. A small time step of 0.5 fs was used in the AIMD simulations due to the presence of O─H bonds, and a total simulation time of 10 ps was performed.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: This work was supported by the U.S. Department of Energy (DOE), Office of Basic Energy Sciences (BES). The work was conducted in the William R. Wiley Environmental Molecular Sciences Laboratory (EMSL), a Department of Energy User Facility operated by the Battelle for the DOE, Office of Biological and Environmental Research. Pacific Northwest National Laboratory is operated for the DOE under contract DE-AC05-76RL01830. Li. L thanks for the resources of the National Energy Research Scientific Computing Center (NERSC), the DOE, Office of Science User Facility supported by the Office of Science of the DOE under contract no. DE-AC02-05CH11231. Author contributions: C.W. and L. Lu. conceived the idea and designed the in situ ETEM experiments. L. Lu. conducted the in situ ETEM analysis. L. Li conducted the DFT simulation. L. Lu. and C.W. wrote the manuscript with input from all authors, and all the authors have given approval to the final version of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials, as well as at the Environmental Molecular Sciences Laboratory of Pacific Northwest National Laboratory. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article