Materials, heat-treatment processes, and characterization

A 20 $\times$ 20 $\times$ 2 mm³ rectangular Zr-4 alloy was selected as the substrate material. Before the coating deposition, the substrate surfaces were polished using #1500 SiC papers and then ultrasonically cleaned. Subsequently, Cr coatings were deposited on the Zr-4 substrates using the magnetron sputtering technique. The detailed deposition process is reported by Jiang et al. [21].

To study the interdiffusion-induced microstructure evolution, several coated samples were subjected to vacuum annealing. To prevent high-temperature oxidation, the coated samples were sealed in vacuum quartz packages before heating. Subsequently, the samples in the quartz packages were placed in a maffle furnace to undergo isothermal exposure after the furnace temperature reached 1160 °C, and then naturally air cooled after removal from the furnace. Notably, owing to the limited vacuum in the quartz packages, a small amount of oxygen may remain. The samples were annealed at different times ranging from 15 min to 4 h.

The residual stress of the original Cr coatings and vacuum-annealed Cr coatings were estimated using the 2 $\text{θ}$-sin² $\text{ψ}$ X-ray diffraction (XRD) technique. Detailed information on the testing process was reported by Yang et al. [16]. The microstructures of the coatings were studied by scanning electron microscopy (SEM). The elemental distributions of Zr and Cr in the coated samples were determined by energy-dispersive X-ray spectroscopy (EDS). The phase compositions of the vacuum-annealed samples were evaluated using XRD analyses. Moreover, the grain size and orientation of the original and vacuum-annealed Cr coatings were identified by electron backscatter diffraction (EBSD, EDAX) measurements.

In situ three-point bending tests

To investigate the interdiffusion effect on the crack evolution in Cr coatings, the original and vacuum-annealed coatings underwent three-point bending tests in a mechanical testing system with SEM. Various mechanical tests at room/high temperatures were conducted using this system [19,20,22]. The three-point bending tests provide more detailed information on the formation and evolution of both channel and interfacial cracks in real time than the in situ tensile tests. The three-point bending samples were cut from a Zr-4 alloy plate (which was also used for preparing rectangular samples for microstructure characterization, as mentioned in Sect. 2.1) using a spark-discharging machine. These samples underwent the same pre-treatment and coating processes as the rectangular samples. After deposition, the coated three-point bending samples were vacuum-annealed. Subsequently, the observing surface was ground to reveal the cross section. Prior to testing, the coated specimen was fastened to the testing system. Then, a bending test was performed, and the displacement rate remained constant at 0.005 mm/s. To acquire the load (P)-deflection (w) relation, the force and displacement of the indenter were logged by a computer. Furthermore, when the deflection reached the required magnitude, the bending test was paused to obtain SEM images of the cracking behavior in the coating by setting the scanning electron microscope above the sample. Finally, the bending test was stopped when the after-target deflection was reached.

Interdiffusion behaviors

Figure 1 presents the cross-sectional morphology and EDS line scan results of the original Cr-coated sample. As shown in Fig. 1(a), the Cr coating is 13 µm thick and possesses highly dense microstructures without microvoids at the interface. Notably, there is a thin element transition area at the interface in Fig. 1(b), which was caused by the smoothing effect due to the limited spatial resolution of EDS, related to different factors such as the stray electrons, beam broadening, and step size. During line scanning, the stray electrons could induce X-rays from both sides of the interfaces when the spot neared the interface, causing a gradual variation in the intensities of Cr and Zr.

Fig. 1Full-screen View

(Color online) (a) SEM image of the cross-sectional morphology and (b) the corresponding EDS line scan of the original Cr-coated sample.

Figure 2 displays the cross-sectional appearance and EDS results of the sample vacuum-annealed for 15 min. No oxide layer formed on the vacuum-annealed coating surface. In addition, a thin diffusion layer formed at the interface. From Fig. 1(b) and the transmission electron microscopy results reported by Brachet et al. [12] and Jiang et al. [21], the diffusion layer was determined to be an intermetallic ZrCr₂ layer possessing the Laves phase. As shown in the SEM and EDS results, numerous scattered precipitates appeared in the area beneath the ZrCr₂ layer, but its composition was not determined with certainty because the resolution of the SEM-EDS map or line scan was limited. These micron-sized Cr-rich precipitates were considered to be a ZrCr₂ or/and Zr(Fe, Cr)₂ phase, which formed because of Cr precipitation in the substrate during the cooling period [15,16,23]. Furthermore, the scattered Cr-rich precipitates were mainly located in the area with a relatively long distance from the coating/substrate interface, instead of the area near the interface, as was reported by Yang et al. [16]. The authors considered that the non-uniform distribution of the Cr-rich precipitates was attributed to the presence of the oxygen-stabilized α-Zr(O) phase in the area underneath the ZrCr₂ layer. Although a vacuum environment was provided in the sealed quartz packages, a small amount of oxygen may remain. At high temperatures, oxygen might penetrate the Zr-4 substrate, which promotes a β-to-α phase transformation. As the α-Zr(O) phase had a lower solubility of Cr than the β-Zr phase, the Cr-rich spots mainly precipitated in the β-Zr phase [24].

Fig. 2Full-screen View

(Color online) (a) SEM image of the cross-sectional microstructure, (b) results of the EDS line scan and (c-e) EDS maps of the sample vacuum-annealed for 15 min.

When the annealing time was longer, the diffusion layer became thicker, and the more significant the separated Cr precipitates (see Fig. 3). As the high-temperature exposure continued, Cr and Zr constantly reacted to form intermetallic ZrCr₂, which thickened the diffusion layer. However, owing to the local uneven diffusion rate at the interface, the ZrCr₂ layer had an uneven thickness. In Fig. 4, the cross-sectional appearance and elemental distribution of the Cr coating annealed for 2 h followed a similar trend to that annealed for 1 h; namely, the interface became rougher and the ZrCr₂ layer became thicker. In addition, significant consumption of the Cr coating occurred simultaneously owing to the presence of the ZrCr₂ layer and continual Cr diffusion.

Fig. 3Full-screen View

(Color online) (a) SEM image of the cross-sectional microstructure, (b) results of the EDS line scan and (c-e) EDS maps of the sample vacuum-annealed for 1 h.

Fig. 4Full-screen View

(Color online) (a) SEM image of the cross-sectional microstructure, (b) results of the EDS line scan and (c-e) EDS maps of the sample vacuum-annealed for 2 h.

Figure 5 displays the cross-sectional appearance and EDS results of the sample vacuum-annealed for 4 h. After a long diffusion time, the remaining Cr coating became thin due to remarkable Cr diffusion. Therefore, the ZrCr₂ thickness was even larger than that of the coating, and the Cr/ZrCr₂/Zr-4 interfaces became significantly rougher because of the uneven diffusion rate. As shown in Fig. 5, a local maximum Zr concentration appears in the diffusion layer. This phenomenon was believed to be caused by Zr diffusing into the coating, considering that the solubility of the Zr atoms in the Cr coating was much lower than that of the Cr atoms in the $\text{β}$-Zr phase [12,16,17]. In addition, Zr was generally enriched in paths along the grain boundaries of the coating [12,16,17], which is not consistent with the presented results. Despite this, a small amount of Zr might also have dissolved into the Cr coating. Moreover, it was reported that the Cr in the ZrCr₂ phase had a much lower diffusion coefficient than Cr in the Zr substrate [16]. The amount of Cr that precipitated from the ZrCr₂ to the Zr-4 layer was greater than that from the Cr coating to the ZrCr₂ layer after extended annealing time. From this point, the atomic percentage of Cr decreased locally in the interlayer, which increased the atomic percentage of Zr (see Fig. 5(b)). Moreover, some microvoids occurred near the interface, which were formed due to the Kirkendall effect, where the differences in the Cr and Zr diffusion rates caused a back diffusion of vacancies at the interfaces [12,16].

Fig. 5Full-screen View

(Color online) (a) SEM image of the cross-sectional microstructure, (b) results of the EDS line scan of the red line in (a), and (c-e) the corresponding EDS maps of the Cr-coated Zr-4 substrate annealed for 4 h.

According to the above results, the thickness of all layers, as well as the interfacial roughness, are functions of the annealing time. Figure 6 summarizes their evolution with respect to the annealing time by counting at least three samples for each annealing condition. As shown in Fig. 6(a), the Cr coating thickness decreased at a lower rate with the annealing time, indicating that the consumption rate of Cr was high during the initial period of annealing because of the rapid Cr-Zr chemical reaction and the precipitation of Cr atoms into the substrate; however, it decreased during the late period of annealing because of the lower chemical reaction rate between Cr and Zr. In Fig. 6(b), the ZrCr₂ layer thickened at a lower rate with the annealing time, which roughly followed a parabolic law, as reported in [15,16]. During the initial period of annealing, the fast reaction-diffusion process between Cr and Zr played a dominant role in the formation of the diffusion layer. During the late period, the growth of the ZrCr₂ layer became diffusion-controlled, accompanied by a decrease in the interdiffusion coefficient. Meanwhile, the extent of Cr atoms diffusing from the ZrCr₂ layer into the Zr substrate is greater than that from the coating to the ZrCr₂ layer, slowing down the growth rate of the ZrCr₂ layer during the late period [24]. It is worth noting that the ZrCr₂ layer appeared to be slightly thicker than that in the Cr coating produced by cold spray and magnetron sputtering techniques with other deposition parameters [15,16]. The deposition process and parameters significantly affected the diffusion rate of the ZrCr₂ layer, which will be further studied in future work. Moreover, in Fig. 6(c), the interfacial roughness also increased with the annealing time, indicating that the diffusion process became more irregular at the interface. During the diffusion process, the ZrCr₂ layer growth direction was generally normal with respect to the rough interface, leading to an increase in the interfacial roughness [16]. In addition, the local interfacial stress (corresponding to the thermal mismatch and creep effect) might also result in an irregular diffusion rate and thus an increased interfacial roughness [25]. Furthermore, the intergranular diffusion rate was generally larger than the transgranular diffusion; thus, Zr-Cr interdiffusion was prone to occur at the grain boundaries of the ZrCr₂ layer, further roughening the coating/substrate interface. Based on Figs. 6(a) and 6(b), the remaining ZrCr₂-coating layer was thinner than the as-deposited coating. The thickness loss became more severe with increasing annealing time. This phenomenon is mainly attributed to continual Cr precipitation into the substrate.

Fig. 6Full-screen View

(a) Thickness of Cr coating vs. annealing time, (b) thickness of ZrCr₂ layer vs. annealing time, and (c) interfacial roughness vs. annealing time.

Figure 7 presents the residual stress ${\text{σ}}_{\text{r}}$ in the Cr coating before and after annealing at different times. After deposition, the Cr coatings had a compressive ${\text{σ}}_{\text{r}}$ of -446 MPa. During high-temperature annealing, the residual stress produced during the deposition was strongly released. Nevertheless, because of the thermal mismatch between all layers, a large thermal stress was generated in the Cr coating upon cooling from 1160 °C to room temperature. In Fig. 7, the compressive residual stress of the 15 min vacuum-annealed Cr coating reached -908.2 MPa. After extended high-temperature annealing, the residual stress in the Cr coating decreased significantly and reached a minimum value of -518.1 MPa in the Cr coating annealed for 4 h. The decrease in the residual stress is mainly due to the decrease in the thermal mismatch stress in the Cr coating. After further annealing, the Cr coating became thinner and the ZrCr₂ layer became thicker. Considering that the elastic modulus (173.2 GPa [26]) and thermal coefficient (6.48 $\times$ 10^-6 K^-1 [27]) of the ZrCr₂ layer fall in the range between the magnitudes of the Cr coating (280 GPa [28] and 6.5 $\times$ 10^-6 K^-1 [17]) and the Zr-4 substrate (110 GPa [29] and 5.77~7.62 $\times$ 10^-6 K^-1 [17]), respectively, the ZrCr₂ layer could be regarded as a transition layer to reduce the thermal mismatch. Thus, the thickening of the ZrCr₂ layer upon effective annealing reduced the thermal stress, which decreased the residual stress in the vacuum-annealed Cr coating. Furthermore, during the cooling process, the Zr substrate also undergoes a $\text{β}$ to $\text{α}$ phase transformation, which results in a transformation-induced plasticity (TRIP) effect at temperatures between 900 and 800 °C, greatly releasing the stress caused by thermal mismatch in the coated sample [30,31].

Fig. 7Full-screen View

Compressive residual stress in the Cr coating vs. annealing time.

Recrystallization in the Cr coating

High-temperature annealing not only generated interdiffusion, but also caused variations in the microstructure of the Cr coating system. Figure 8 shows the EBSD results of the surfaces of the original and vacuum-annealed Cr coatings. In Figs. 8(a) and 8(b), an intensive (001) texture occurred in the original Cr coating in the direction perpendicular to the interface. Based on the IQ image in Fig. 8(b), the grain size of the Cr coating reached an average of 1.49 µm. After annealing for 15 min and 1 h, the grain sizes reached 4.34 and 5.11 µm, respectively. Recrystallization was observed in the Cr coating upon annealing; however, this did not obviously change the intensities of the textures. Note that the EBSD results of the Cr coatings annealed for 2 and 4 h are not presented because the severe Cr consumption decreased the thickness of the Cr coating, resulting in very low calibration rates during EBSD testing. Despite this, the Cr coating annealed for a longer time presumably possessed larger grains with strong textures.

Fig. 8Full-screen View

(Color online) EBSD results of the coating surfaces: (a) orientation map and (b) IQ image of the original coating, (c) orientation map and (d) IQ image of the coating vacuum-annealed for 15 min, and (c) orientation map and (d) IQ image of the coating vacuum-annealed for 1 h.

Figure 9 shows the EBSD maps of the cross-sections of the original and vacuum-annealed Cr coatings. Clearly, the original Cr coating had columnar grains, which is a common feature in the coatings prepared by magnetron sputtering and other physical vapor deposition techniques [12,21,32]. However, as shown in Fig. 9(b), it was surprising that after annealing for 15 min, equiaxial grains occurred in the Cr coating which were transformed from the original columnar grains that resulted from recrystallization at high temperatures. As shown in Fig. 9(c), after annealing for 1 h, the coating thickness declined and the equiaxial grain grew to some extent. The equiaxial grains were presumably constantly coarsened during long-time annealing before the coating was consumed.

Fig. 9Full-screen View

(Color online) EBSD orientation maps of the cross-sections of the (a) original Cr coating, and the Cr coating (b) vacuum-annealed for 15 min, and (c) vacuum-annealed for 1 h.

In situ three-point bending test

The above results suggest that high-temperature vacuum annealing evidently changed the microstructure of the coating, leading to the formation of an intermetallic layer and recrystallization in the Cr coating. The microstructural evolution may significantly change the mechanical properties and crack resistance of the Cr coating. Figure 10 presents the P–w curve of the coating vacuum-annealed for 1 h during the bending test. As shown in Fig. 10, P increased linearly when w < 0.2 mm, and increased slowly when 0.2 ≤w ≤ 0.85 mm. When w > 0.85 mm, the loading dropped significantly because of the macrocracks which formed and propagated rapidly in the Zr-4 substrate. Considering that the test was paused to acquire SEM images, some small drops in the load were observed, but they did not affect the overall mechanical behavior.

Fig. 10Full-screen View

(a) P–w curve of the sample vacuum-annealed for 1 h during the bending test.

Figure 11 presents the results of the cracking behavior in the 1 h vacuum-annealed Cr coating during the bending test. As shown in Fig. 11(b), most of the microcracks appear at the interface, and few cracks penetrate across the coating when w reaches 0.405 mm. In addition, as shown in the magnified view in Fig. 11(b), few microcracks were observed on the coating surface. Under continuous loading, more microcracks were formed at the coating/substrate interface. Remarkable slip lines were generated at the crack tips owing to large deformations. Despite this, the cracks that formed from the interface and the surface barely coalesced, which may be blocked by the grain boundaries in the Cr coatings. In Figs. 11(e) and 11(f), at w = 1.210 mm where long vertical cracks grew rapidly in other areas, interfacial cracks were formed and grew along the interface and coalesced with the vertical cracks at the interface. The formation and evolution of interfacial cracks were caused by the large local interfacial stress in the regions surrounding the vertical crack tips [33,34]. However, no visible interfacial delamination occurred in the coating until the final failure, which indicated the excellent interfacial adhesion of the vacuum-annealed coating.

Fig. 11Full-screen View

SEM results showing crack evolution in the Cr coating vacuum-annealed for 1 h during bending at w of (a) 0.000 mm, (b) 0.405 mm, (c) 0.734 mm, (d) 0.926 mm, and (e,f) 1.210 mm. Insets show magnified views of the sections in red boxes.

Figure 12 compares the cross-sectional microstructures of the original and vacuum-annealed coatings after bending. As shown in Fig. 12(a), numerous vertical microcracks are formed in the original coating upon bending. These cracks primarily formed at the interface, as shown in Fig. 12(a). The large number of vertical cracks reflect the high brittleness of the as-deposited coating. In addition, it was found that some interfacial microcracks formed and grew along the interface. As shown in Fig. 12(a), some interfacial cracks coalesced with the vertical cracks. The formation and growth of interfacial cracks were due to the large local interfacial stress. As shown in Figs. 12(b)-12(d), for the annealed Cr coating, microcracks were mainly formed in the ZrCr₂ layer. In addition, based on the EDS map in Fig. 12(d), these cracks also penetrated somewhat into the substrate, which may be because of the thin α-Zr(O) layer that embrittled the substrate beneath the ZrCr₂ layer. Moreover, several horizontal interfacial cracks lay at the ZrCr₂/Zr interface, illustrating that the ZrCr₂/Zr interfacial adhesion was lower than that for Cr/ZrCr₂. Furthermore, comparing Figs. 12(a) and 12(b), despite the numerous cracks at the interface, the annealed coating had fewer vertical cracks than the as-deposited coating, which indicated that high-temperature annealing significantly improved the crack resistance of the Cr coating. Upon annealing, recrystallization in the coating not only released a substantial amount of internal stress, but also altered the grain morphology of the coating. In Fig. 9, more grain boundaries were found in the equiaxial grains in the direction vertical to the interface, effectively impeding slip deformations in the Cr coating and clogging the propagation of vertical cracks, which could explain why the crack density was substantially reduced in the vacuum-annealed coating.

Fig. 12Full-screen View

(Color online) Cross-sectional microstructures of the (a) original and (b) vacuum-annealed coatings after three-point bending tests, (c) the magnified view, and (d) the corresponding EDS map of (c).

According to the above results, during high-temperature annealing, the formation of the ZrCr₂ layer caused by interdiffusion embrittled the coating/substrate interface and resulted in numerous microcracks under an external load. Meanwhile, the stress release and grain morphology evolution due to recrystallization significantly enhanced the plasticity of the coating and improved the deformation compatibility, which remarkably improved its crack resistance. For the Cr coating annealed for an extended time, microcracking at the interface could presumably be more remarkable in a thicker diffusion layer, and the vertical crack density in the coating will increase under external loading, despite a decrease in residual stress (see Fig. 6). This is because the crack resistance of the coating decreased with fewer grain boundaries, while the large local interfacial stress increased (see Fig. 6(c)). Once the Cr coating is consumed after a long period of diffusing into the substrate, the vertical cracks are mainly located in the brittle ZrCr₂ layer under external loading, which could be regarded as numerous micro-notches for the underlying substrate, causing earlier failure than those uncoated.