For decades, Ni-based single-crystalline superalloys have been extensively used in important hot end components (such as blades and vanes) of advanced aerospace engines and gas turbine engines because of their remarkable high-temperature mechanical properties [1]. Oxidation performance is one of the most important evaluation indicators for the high-temperature properties of superalloys, because oxidation can lead to dealloying corrosion, surface strength loss, crack initiation and ultimate failure [[2], [3], [4], [5]]. Due to the superalloy matrix with relative poor oxidation resistance, coatings are used to extend their service lifetime at high temperatures [6]. However, the occurrence of severe thermal shock, mechanical fatigue and partial stress overload on turbine blades during service can compromise the reliability of protective coatings [7]. Moreover, coatings cannot be applied to all regions of a blade. Therefore, understanding the intrinsic behavior of superalloys under an oxidizing atmosphere is critical.

The effect of external factors on the oxidation properties of alloys has been extensively investigated by comparison. Currently, the major factors known to affect the oxidation behavior include the concentration of solute atoms [[8], [9], [10], [11], [12], [13], [14], [15]], oxidation temperature [11,[16], [17], [18]], the oxidizing atmosphere [11,[19], [20], [21]] and surface strengthening technologies [22]. However, the effect of substrate microstructure on alloys′ oxidation performance has received much less attention. The mechanical properties of Ni-based superalloys are acknowledged to depend strongly upon the γ′+γ two phase microstructure [[23], [24], [25]], because it gives rise to precipitation hardening, which strengthens the superalloys. The γ′ precipitates with an L1₂-type ordered structure embedded into the γ matrix act as the main strengthening phase in Ni-based superalloys. The morphological evolution and corresponding coarsening mechanism of the γ′+γ microstructure during aging treatment have been systematically studied [[26], [27], [28], [29]]. The results of numerous experiments have shown that the changes of γ′ volume fraction, morphology and size distribution play a vital role in determining the mechanical properties of superalloys in service [25,[30], [31], [32], [33]]. Therefore, clarifying the effect of the γ′+γ two-phase microstructures on the oxidation mechanism of Ni-based single-crystal superalloys is important.

So in the work, the γ′ phase strengthened Ni-based model superalloy (Ni-Al binary single-crystal alloy) was adopted, and the effect of γ′+γ two-phase structure on the oxidation behavior under atmospheric-pressure (at 650 °C isothermal oxidation) was systematically investigated. Two heat treatments were utilized to obtain obvious different γ′+γ microstructures. The relative low oxidation temperature was chosen to ensure that the microstructure do not change during the isothermal oxidation process. The results are helpful for understanding and even predicting the oxidation behaviors of multicomponent Ni-based superalloys with the coarsening of γ′ phase in service at high temperatures. Further, it may be of reference value for researching the oxidation process of other strengthened superalloys.

The specimens used in the oxidation experiments with dimensions of 22 mm × 10 mm × 3 mm were cut from the Ni-Al single-crystal alloy by wire electro-discharge machining. The nominal chemical composition of the alloy is given in Table 1. Before oxidation, the alloy was subjected to two different heat-treatment regimens to obtain different microstructures of γ′+γ: (1) 1330 °C/8 h + 1350 °C/4 h with air cooling (A. C.) for the first group of alloys, denoted hereafter as SC_S alloy, and (2) 1330 °C/8 h + 1350 °C/4 h with furnace cooling + 900 °C/10 h+700°C/20 h with A. C. for the second group of alloys, denoted hereafter as SC_L alloy.

The samples of SC_L and SC_S alloys were first ground with a series of polishing papers and then polished with progressively finer diamond pastes, with the final polishing performed with 1 μm paste. After polishing, the samples were ultrasonically cleaned in acetone and ethanol.

The as-polished specimens were first placed in a pre-annealed Al₂O₃ crucible to protect them from pollutants. The crucibles with specimens were then placed in a muffle furnace and heated under air atmosphere at 650 °C (heating rate 10 °C /min) for oxidation. Oxidation time was set as 0.5, 1, 3, 5, 8, 14, 30, 54, 78, or 100 h. Finally, the oxidized specimens were removed from the furnace and naturally cooled to room temperature. Three groups of specimens were used to measure the mass change for each oxidation condition during the isothermal oxidation test.

The mass gains of specimens oxidized for different times were discontinuously measured with an analytical balance (Mettler Toledo XP205) with 10^-4 g accuracy.

The morphology and microstructure of the oxidized alloys were investigated by scanning electron microscopy (SEM, JEOL JSM-7800F) in secondary-electron (SE) and backscattered-electron (BSE) modes. The distributions of O, Al, and Ni elements in the SC_L and SC_S alloys were examined by energy-dispersive X-ray spectrometry (EDS, EDAX Octane Plus), using an apparatus installed on the scanning electron microscope. The oxidation products on the surface of the SC_L and SC_S alloys were investigated by X-ray diffraction (XRD) on a Bruker D8 Advanced diffractometer equipped with a Cu X-ray anode (40 kV/40 mA, λ = 1.5418 Å). A focused ion beam (FIB) system (FEI 3D Quanta Nanolab FIB/SEM) was used to prepare cross-sectional transmission electron microscopy (TEM) lamellae of the specimens. The composition and microstructure of the oxidation products formed on the SC_L and SC_S alloys were examined by cross-sectional TEM (FEI Tecnai G2 F30). The three-dimensional morphology of the surface oxide layer was measured and reconstructed by atomic force microscopy (AFM, Agilent 5500 AFM). Auger electron spectroscopy (AES, ULVAC-PHI 700) was used to obtain the qualitative and quantitative description of the chemical constitution of the oxidized alloys. The approximate depth of the AES sputter profiles was estimated relative to the instrument calibrated rate for SiO₂ (11 nm/min).

To better understand the effect of the γ′+γ microstructure on the oxidation behaviors of the alloys, different heat-treatment regimens were utilized to obtain clearly different γ′+γ microstructures in the SC_L alloy and the SC_S alloy. The microstructure images of two alloys are shown in Fig. 1. The regularly packed γ′ precipitates in both alloys are coherently embedded in the γ matrix. The cuboidal γ′ precipitates in the SC_L alloy are distributed uniformly, with a mean size of 570 nm. Moreover, adjacent cuboidal γ′ precipitates are separated by γ-phase channels that range from 150 nm to 600 nm in width (Fig. 1(a)). The nanoscale γ′ particles are dispersedly distributed within the γ channels, which are precipitated between cuboidal γ′ phase. Herein, the nanoscale γ′ particles are defined as γ′_Ⅱ. The effect of γ′_II phase on the oxidation behavior of SC_L alloy can be negligible, because most of the γ′_II phase in the γ channel will dissolve at 650 °C. The SC_S alloy has slender γ channels (about 20 nm), and the average size of the γ′ precipitates is about 80 nm, as shown in Fig. 1(b).

The compositions of the γ and γ′ phases in the SC_L and SC_S alloys are investigated by TEM, as shown in Table 2. The composition of γ′ phase in the SC_L and SC_S alloys is basically the same, and the Al content of γ′ phase is much higher than that of γ channel. Owing to the existence of γ′_Ⅱ phase, the SC_L alloy has relative high Al content in γ channel, compared to the SC_S alloy.

Fig. 2 shows the values of mass gain and the square of mass gain of the SC_L and SC_S alloys during isothermal oxidation under atmospheric pressure at 650 °C. The corresponding values of mass gain are determined from ΔmA, where Δm is the mass change, and A represents the total surface area of the tested alloy. Fig. 2(a) shows that the SC_L and SC_S alloys exhibit similar trends during the isothermal oxidation at 650 °C. The final mass gain of the SC_S alloy after isothermal oxidation at 650 °C for 100 h is 0.42 mg/cm², which is lower than that of the SC_L alloy (0.51 mg/cm²). The square of the mass gain of the SC_L and SC_S alloys with fitted lines is shown in Fig. 2(b). Note that the two parts of the oxidation mass gain curve were fitted using the parabolic law. According to the classic parabolic rate law [34], parabolic rate constants K_p can be determined from the relation between the square of the mass gain ΔmA2 and the oxidation time t, i.e. the slope of the fitted lines. Thus, the values of K_p for the SC_L alloy are 6.06 × 10^-3 mg²/(cm⁴ h) and 2.17 × 10^-3 mg²/(cm⁴ h) during the periods of 0-14 and 14-100 h, respectively. The values of K_p for the SC_S alloy are determined to be 4.65 × 10^-3 mg²/(cm⁴ h) and 1.46 × 10^-3 mg²/(cm⁴ h) during the periods of 0-14 and 14-100 h, respectively. These fitting results show that the two regions of oxidation mass gain curve can be controlled by the different oxide growth process.

The value of K_p for the SC_S alloy is consistently lower than that of SC_L alloy, indicating that the SC_S alloy exhibits better oxidation resistance. That is to say, the γ′+γ microstructure with small γ′ precipitates and slender γ channels is beneficial to the decrease of alloy′s oxidation rate.

The surface morphologies of the SC_L and SC_S alloys during isothermal oxidation tests at 650 °C for 14, 54, and 100 h are shown in Fig. 3. The formation of the neck structure among the oxide particles in both alloys led to the formation of oxides with a submicron porous sintered-like microstructure, which is a sign of rapid growth of oxides [35]. Fig. 3(a) shows that the oxides formed on the SC_S alloy are not compact. They exhibit a large cellular morphology with oxide ridges after 14 h of exposure. In contrast, the oxides formed on the SC_L alloy are flat and dense, and the cellular oxides and holes are relatively small (Fig. 3(d)). After 100 h of oxidation, the oxide layers on both alloys become flat, and comprise fine closely packed particles (Fig. 3(c) and (f)).

Fig. 4 shows cross-sectional BSE images of the SC_L and SC_S alloys oxidized at 650 °C for different times. The SC_L and SC_S alloys exhibit similar oxidation-affected zones (OAZs). In OAZs, two separate regions can be clearly differentiated: a regular outer oxide zone (OOZ) and an extensive uniform internal oxidation zone (IOZ). Comparing the oxide thickness of the two alloys oxidized at the same oxidation time (Fig. 4), the results reveal that the OAZ in the SC_L alloy is always thicker than that in the SC_S alloy, consistent with the results of the mass gain curve (Fig. 2(a)). After oxidation at 650 °C for 100 h, the thickness of OAZ reaches up to about 12 μm (in SC_L alloy) and 7 μm (in SC_S alloy), respectively (Fig. 4(c) and (f)). The bright contrast nodule precipitates in the OOZ of SC_L and SC_S alloys undergo coarsening as oxidation proceeds (Fig. 4). Throughout the whole oxidation process, the γ′+γ microstructure in the SC_L alloy remains basically unchanged (Fig. 4(d-f)), indicating that the γ′ phase is stable at 650 °C for at least 100 h. The dark contrast regions (Fig. 4) at the interface of the OOZ/IOZ as well as in the IOZ are low-density areas or voids.

The XRD profiles of the OAZs in SC_L and SC_S alloys (oxidized at 650 °C for 14 h and 100 h) are shown in Fig. 5. NiO, Al₂O₃, and NiAl₂O₄ phases (oxidation products in Ni-Al alloys) were detected in all samples. Ni peaks were obviously detected in the other three samples, except in the SC_S alloy oxidized for 14 h, which is consistent with the appearance of bright contrast nodules in the cross-sectional SEM images (Fig. 4). These results and the cross-sectional EDS line-scans (Fig. 6) show that the bright contrast nodule precipitates in the OAZ of the SC_L and SC_S alloys are almost entirely metallic Ni (depleted in O and Al).

A detailed characterization of the Ni-rich nodule precipitates was conducted using cross-sectional TEM lamellae. The HRTEM micrograph of a Ni-rich nodule is shown in Fig. 7(a), and the corresponding selected-area electron diffraction (SAED) pattern is shown in Fig. 7(b). From Fig. 7, it is confirmed that the nodule precipitates are metallic Ni. The mechanism of formation of solvent metal in OOZ has been extensively studied [[36], [37], [38], [39]]. Generally, the noble solvent metal, such as Ni, is expulsed to the alloy surface to relieve the internal stress caused by the formation of internal oxides [36,38,39]. Moreover, the void formed below the oxide/alloy interface can isolate the metal region from the underlying matrix. So the oxidation-prone elements diffuse outwardly and incorporate into the oxide layer, and the matrix element is left to form unoxidized-metal region [37]. The Ni nodules in the OOZ of the SC_L and SC_S alloys are covered with a continuous NiO layer (Fig. 4). The OOZ region is followed by an innermost IOZ pattern, characterized by intermixing of alumina and metallic Ni (Fig. 6).

Fig. 8 shows TEM images of SC_L and SC_S alloys oxidized at 650 °C for 54 h. Distinct intermediate oxide layers (not observed in the previous cross-sectional SEM images) formed between the IOZ and the OOZ of the two alloys (Fig. 8(a) and (b)). The internal oxidation of Ni-based superalloys shows similar regional segregation between oxide types: (1) large, well-developed oxide networks consisting of MAl₂O₄ near the interface of the OOZ/IOZ and (2) small discrete Al₂O₃ locating further down [38,40]. Obviously, the intermediate oxide layer is uniform in the SC_S alloy (Fig. 8(a)). However, a protrusion is formed in the SC_L alloy and extends into the matrix (Fig. 8(b)). In the SC_S alloy, crisscrossed needle-like alumina exists throughout the IOZ (Fig. 8(a)). Moreover, the distribution of alumina in the SC_L alloy is not uniform (Fig. 8(b)). The density of alumina in the γ′ phase is higher than that in the γ channel (Fig. 8(c)). Numerous low-density areas are clearly visible in the IOZ of both alloys, and these areas emerge during the oxidation treatment [37].

Similar oxidation processes and OAZ structures were observed in both SC_L and SC_S alloys during the exposure tests. To better understand the effect of the γ′+γ microstructures on the oxidation behaviors of alloys, further study of the initial stage of oxidation is required. The microstructures of the surface of SC_L and SC_S alloys oxidized for 30 min are shown in Fig. 9. Interestingly, the oxides in the two alloys show substantially different structures. The outermost oxide on the SC_S alloy (Fig. 9(b)) is evenly distributed, with a small difference in contrast (Fig. 9(d)). However, the contrast difference in the SC_L alloy (Fig. 9(a) and (c)) is obvious, indicating that the thickness of the outermost oxide layer on the SC_L alloy is not uniform. To better reveal the height difference of the outermost oxide layer, the details of the oxide morphology profiles were further investigated by AFM. The measurement results are shown in Fig. 10. Island-like oxides are evenly distributed on the SC_S alloy (Fig. 10(b)), whereas substantial rumpling is visible on the surface of the SC_L alloy (Fig. 10(a)). The height difference between recessed region (valley) and convex region (ridge) is about 200 nm.

The AFM data in Fig. 10 shows the results of a local morphology analysis of the oxidized surfaces. To gain insight into the composition profile over a substantially larger surface area (about 0.01 mm²), AES was used in combination with sputter depth profiling (Fig. 11). No substantial difference in element distribution was observed between the SC_L and SC_S alloys except in the thickness. The OAZ structures formed on the two alloys still consist of an outer NiO layer, an intermediate Ni-Al spinel layer and an inner oxide layer mixed with alumina and metallic Ni. The thicknesses of NiO in the SC_S alloy and SC_L alloy are 220 and 380 nm (Fig. 11), respectively. On the basis of the combined results in Fig. 9, Fig. 10, it can be concluded that the NiO thickness difference between the SC_L and SC_S alloys is primarily caused by different γ′+γ two-phase microstructures in the SC_S and SC_L alloys.

According to the thermodynamics of oxidation, the Gibbs energy of formation of Al oxides is more negative than that of the Ni oxides [41], indicating that Al has much larger affinity to oxygen than Ni. However, the formation of alumina severely depletes the Al near the reaction front. Moreover, NiO is a less stoichiometric oxide than Al₂O₃. Thus, a NiO layer develops continuously on the surface of the OAZ.

NiO is controlled by cation outward migration, whereas alumina is mainly controlled by anion inward diffusion [3]. In the initial stage of oxidation, oxygen diffuses inwardly and encounters the outwardly diffused solute component at the reaction front, causing the activity product aBaOv to reach the critical value of oxide nucleation. The reaction responsible for alumina formation is as follows:

where K is the equilibrium constant and a is the activity coefficient. According to the oxygen potential map, after the alumina is formed, the oxygen partial pressure at the oxide/alloy interface is controlled by the decomposition pressure of alumina. So it is insufficient to oxidize the remaining solvent element Ni. As the oxidation process, the Al content (i.e., the Al activity coefficient) at the reaction front begins to decrease. The oxygen activity must increase to maintain the equilibrium condition according to the reaction equilibrium law (Eq. (1)). When it reaches a thermodynamic critical value for reaction with Ni, the oxide is converted into Ni-Al spinel oxide according to the following reaction [42]:

(2)$2\text{Ni}+{{\text{O}}_{2}}+2\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}=2\text{NiA}{{\text{l}}_{2}}{{\text{O}}_{4}}$

The Ni-Al spinel oxide exhibits poor oxidation resistance because of the formation of interfacial voids with volumetric shrinkage caused by solid state reactions among oxides [42]. A continuous interpenetrating network of alumina in IOZ is expected, because the growing Ni-Al spinel reduces the oxygen partial pressure at the oxide-alloy interface and enables the alumina to form thermodynamically [43]. Because lattice diffusion is essentially negligible at relative low temperatures, the extent of internal oxidation and solvent metal transport observed in the present study is likely related to the substantial short circuits provided by the oxide phase and the voids themselves. Therefore, it is necessary to focus on the formation of the intermediate layer, which is the key to controlling the initial diffusion process. As shown in Fig. 11, the intermediate layer thickness of the SC_S alloy and the SC_L alloy are 160 and 240 nm, respectively. The large difference in thickness of intermediate layer in SC_S and the SC_L alloys is also closely related with γ'+γ two-phase microstructures in alloys.

The microstructural evolution of the OAZ in the SC_S and SC_L alloys is illustrated in Fig. 12. The high Al content in the γ′ phase promotes the formation of an alumina layer in the SC_L alloy. Thus, the formation and growth of the oxide layer above the γ′ phase occur uniformly. However, the low Al content in the γ-channels (Table 2) and the competitive growth of NiO can hinder the further growth of the continuous alumina layer on the channels. Both Ni and O are expected to preferentially diffuse through unprotected γ channels, and form an outward convex NiO region and an inward convex IOZ in the γ channels, as shown in Fig. 12(a). Simultaneously, the oxygen activity will increase local vacancies in the γ channels, which facilitates the conversion of alumina to Ni-Al spinel oxides [44] (Fig. 12(a)). Vacancies and oxide/oxide interfaces in Ni-Al spinel oxides provide short circuit paths, and promote the inward diffusion of oxygen and the outward diffusion of Ni, similar to the role of grain boundaries during the oxidation process [45].

During further oxidation, these spinel structures extend into the alloy and form distinct protrusions in the γ channels (Fig. 12(b)). As shown in Fig. 8(b) and (c), the distribution of alumina in the γ′ phase and the γ channel is uneven. The high density of alumina in the γ′ phase results in more metal/oxide and oxide/oxide interfaces. Moreover, these interfaces can act as short-circuit diffusion paths [39,46,47]. And the internal stress generated by the volumetric increase with the formation of alumina provides driving forces for the expulsion of Ni [36,48]. This makes the expulsion of Ni from the γ′ phase faster than that of Ni from the γ channel. Therefore, the height difference of the oxides between γ′ phase and γ phase gradually decreases (Figs. 12(c) and 3 (d)). And the Ni-Al spinel structure becomes serrated (Figs. 12(c) and 8 (b)). The effect of uneven distribution of alumina in γ′ phase and γ channel on the oxidation performance is discussed in Section 4.2.

For the SC_S alloy, small γ′ particles and slender γ channels (Fig. 12(d)) form a grid-like structure. This structure shortens the lateral growth distance of alumina and promotes the formation of a relatively continuous protective alumina sublayer on the surface. The SC_S alloy cannot form obvious protrusions of the Ni-Al structure because of the slender γ channel. Therefore, the Ni-Al spinel structure in the SC_S alloy becomes regular in the later oxidation stage (Fig. 12(e) and (f)), corresponding to the image shown in Fig. 8a. Moreover, the distribution of alumina in the IOZ is uniform (Fig. 8(a)). Thus, the whole oxidation process of SC_S alloy can be considered as uniform oxidation (Fig. 12(d-f)). During the subsequent oxidation process, the island-shaped oxide (Figs. 9(b) and 10 (b)) uniformly distributed on the surface of the SC_S alloy will grow and become cellular oxide (Fig. 3(a)). The ridges among cellular-structured oxides begin to flatten during the later stage of oxidation (Fig. 3(b) and (c)).

As shown in Fig. 4, the thickness of the IOZ of the SC_S and SC_L alloys changed substantially at the later stage of oxidation, whereas the OOZ of the SC_S and SC_L alloys changed only slightly. These observations and the mass change curves of the SC_S and SC_L alloys (Fig. 2(b)) indicate that the first stage of oxidation is mainly controlled by the Ni-oxide growth process, and the second stage of oxidation is mainly controlled by the Al-oxide growth process. Moreover, the huge internal stress and dense short-circuit paths generated by the formation of internal oxides in the IOZ become the main factor affecting diffusion process in the SC_S and SC_L alloys.

According to the results presented in Section 4.1, the difference in internal oxidation behavior of the SC_L alloy and the SC_S alloy is the main cause of the difference in the later oxidation rate. In the SC_S alloy, the small size of the γ′ phase in the IOZ limits the growth of alumina within it. Thus, the needle-like alumina is formed as discrete particles throughout the IOZ rather than continuous networks (Fig. 8(a)). This disconnected oxide structure does not enhance the diffusion of oxygen [49]. However, the elongated internal oxides are well-developed in the large γ′ phase of the SC_L alloy (Fig. 8(b) and (c)), forming a continuous internal oxide network that contributes to the diffusion of oxygen along the oxide-metal interface [49]. Oxygen diffuses from the alloy surface to the front of the internal oxidation along the internal oxidation network, preferentially reacting with the more reactive solute element Al.

The Ni content of γ channel is much higher than that of γ′ phase (Table 2). Moreover, due to the rapid diffusion of Ni driven by internal stress along the continuous internal oxide in SC_L alloy, the decreasing rate of Ni content in the γ′ phase is faster than that in the γ phase. The higher Ni concentration gradient is created between the γ channel and the γ′ phase in SC_L alloy. Thus, Ni diffuses from the γ channel (high concentration) to the γ′ phase (low concentration), and rapidly diffuses through the internal oxide network in the γ′ phase. That is to say, the formation of continuous internal oxide networks in the γ′ phase is beneficial to the inward diffusion of oxygen and the outward expulsion of Ni. As a result, the oxidation rate of the SC_L alloy is faster than that of the SC_S alloy.

Internal oxidation is a common phenomenon observed in binary alloys that contains a noble solvent element and a lower concentration of reactive solute element [47,[50], [51], [52], [53], [54]]. The growth of the internal oxide is driven by capillary action (Ostwald ripening), and the reaction kinetics can be expressed by the parabolic rate law [55]:

where X is the thickness of the IOZ measured from the OOZ-IOZ interface to the internal oxide precipitation front, t is the exposure time, and kp(o) is the parabolic rate constant of IOZ. According to Wagner’s theoretical analysis [55]:

where No(s)i is the solubility of oxygen in Ni (at.%) and Do(e)i is the oxygen effective diffusion coefficient in Ni. The relationship among the effective diffusion coefficient of oxygen Do(e)i, the oxygen short circuit diffusion coefficient Do(sc)i and the oxygen bulk diffusion coefficient Do(b)i is as follows:

where η is the fraction of surface sites associated with short-circuit paths, NAl(o) is the initial concentration of Al in the bulk alloy, and ν is the stoichiometric factor for Al₂O₃. The relation that exists between the thickness X of the IOZ and the oxidation time t can be expressed by the following formula [56]:

(6) $X={{\left[ \frac{2N_{\text{o}(\text{s})}^{\text{i}}D_{\text{o}(\text{e})}^{\text{i}}}{\nu N_{\text{Al}}^{\text{i}}}t \right]}^{\frac{1}{2}}}$

Wagner’s theory has been successfully applied to describe the internal oxidation of binary alloy systems. However, it failed to explain the rapid internal oxidation of binary Ni-Al alloys [46]. The high internal oxidation rate is attributed to the accelerated oxygen diffusion along the oxide-metal phase boundary [57,58]. Because the values of No(s)i and Do(e)i reported elsewhere do not consider the influence of the presence of internal oxide precipitates, the oxygen permeability, i.e. No(s)iDo(e)i is used to describe the internal oxidation of the SC_L and SC_S alloys. Substituting the measured IOZ data from Fig. 13 into Eq. (6), the oxygen permeability in the SC_L alloy is 2.963 × 10^-13 cm²/s, which is approximately three times greater than that of the SCs alloy (1.016 × 10^-13 cm²/s). The higher oxygen permeability indicates that the internal oxide structure formed in the SC_L alloy promotes the inward diffusion of oxygen, compared with that in the SC_S alloy. The reason is that the oxygen diffusion through the internal oxide itself is the key factor in controlling the internal oxidation process [46]. That is to say, the γ′+γ microstructure not only affects the oxide structure in the γ channels but also the oxides′ structure in the IOZ, which strongly validates the discussion in Section 4.1.

The effect of the γ′+γ two phase microstructure on the oxidation behavior of Ni-Al single-crystal alloys was systematically investigated. Prior to oxidation, two heat-treatment regimens were used to obtain various γ′+γ two-phase microstructures in the SC_L and SC_S alloys. The width of the γ channels in the SC_L alloy ranges from 150 to 600 nm, and the average size of the γ′ precipitates is about 570 nm. The SC_S alloy has slender γ channels (about 20 nm) and small γ′ precipitates (about 80 nm). Similar OAZ structures were observed in both SC_L and SC_S alloys oxidized at 650 °C. However, the SC_L alloy with wide γ channels and large γ′ precipitates exhibited poor oxidation performance.

(1)In the initial stage of oxidation, the oxidation process was primarily controlled by the Ni-oxide growth process (growth pattern of oxides in the γ channels). The outward convex NiO region was formed by the preferential diffusion of Ni and O in unprotected wide γ channels (150-600 nm in width) of the SC_L alloy. The Ni-Al spinel provides a large amount of short-circuit paths, promoting the inward and outward diffusion of atoms. Due to the existence of slender γ channels (about 20 nm), the SC_S alloy formed a continuous protective alumina layer on the surface, which leads to the formation of relative uniform NiO layer and regular Ni-Al spinel structure.

(2)In the late stage of oxidation, the oxidation process is primarily controlled by the Al-oxide growth process, and the effect of Ni-Al spinel structure (in γ channels) on diffusion is weakened. The oxidation behavior is primarily dependent on the oxide structure in the IOZ. The small size of the γ′ phase (about 80 nm) in the SC_S alloy limits the growth of alumina within it. However, the high density and elongated alumina are well-developed in the large γ′ phase (about 570 nm) of the SC_L alloy. The continuous internal oxide network greatly contributes to the diffusion of oxygen along the oxide/metal interface.

This work was supported financially by the National Natural Science Foundation of China (No. 51774212).