Microstructural Investigation of Oxidized Complex Refractory High Entropy Alloys
2018 Volume 59 Issue 4 Pages 556-562
Details
2018 Volume 59 Issue 4 Pages 556-562
The microstructures of two newly developed refractory high entropy alloys were examined after isothermal oxidation at 1200°C for 10 hours. Scanning electron microscope analysis showed the formation of aluminosilicate layer on the sample surface, and the structure of oxide layers appears to be greatly affected by the content of Al and V. With increased Al content and decreased V content, the size of pores within the internal oxidation zone was decreased. Future directions to improve the oxidation resistance of complex refractory alloys were proposed.
In recent years, high entropy alloy (HEA) emerged as a new strategy for alloy design. By utilizing the high mixing entropy of multi-element combination, the Gibbs free energy of disordered solid solution can be more negative at high temperature, promoting the formation of single phase solid solution microstructure.1) With this novel alloying strategy, the refractory HEA was created by using refractory elements, such as Mo, Nb, Ta, W, V… as constituents. Several refractory HEAs have shown attractive properties, such as good high temperature strength,2) room temperature ductility,3) and low density.4,5) However, the high temperature surface stability of refractory HEA is generally poor.6–8) Also, due to the complexity of the oxidation behavior of refractory HEA, very few studies have been reported. CrMo0.5NbTa0.5TiZr, firstly reported by Senkov et al. in 2012,6) exhibited near-parabolic dependence in weight-gain test at 1000°C for 100 hours, and the thickness of oxide layer was approximately 1.75 mm. In 2013, the oxidation behavior of several newly synthesized refractory alloys containing Al, Cr, and Si at 1300°C was reported by Liu et al.7) The formation of visible Al2O3 layer at oxidized sample surface was observed, but there was no report on the extent of internal oxidation beneath. By the time the authors completed this report, the latest report on the oxidation behavior of a refractory high entropy alloy AlCrMoTiW was published in 2014 by Gorr et al.8) The oxidation mass gain of AlCrMoTiW at 1000°C exhibited parabolic dependence within 40 hours, but the oxidation rate was still high when comparing with those of Ni-based superalloys. These previous reports indicated that generally oxidation kinetics of HEAs tends to be complex, because of the co-existence of elements whose oxides are both protective (Al, Si) and volatile (W, Nb, etc.) with high alloying additions. Therefore, once HEAs are exposed under oxidative atmosphere, several phenomena, such as formation of protective oxide layers, surface volatilization, and internal oxidation could take place in one alloy system. In other words, understanding the oxidation kinetics of HEAs is of utmost importance when they are newly designed. In the authors’ research group, to explore the possibility of designing an intrinsically oxidation-resistant refractory HEA, two Al, Cr, Si-bearing refractory high entropy alloys, ML-SP2 and ML-SP2A alloy were developed. The aim of present study is to characterize the oxide structure of the newly developed Al, Cr, Si-bearing refractory high entropy alloys. By comparing the microstructure of the studied alloys before and after the oxidation at 1200°C, the correlation between oxidized microstructure, as-cast microstructure, and the change in Al and V content is presented. Future directions to improve the oxidation resistance of complex refractory alloys are also proposed in this article.
Two refractory alloys, ML-SP2 and ML-SP2A, are alloys of interest. Table 1 shows the nominal compositions of ML-SP2 and ML-SP2A alloys. The major difference between these two alloys is the Al and V content. V is a low density refractory element, and is ideal for decreasing the density of refractory HEA, but its oxide is volatile at high temperature9) and may have detrimental effect on oxidation resistance of refractory HEA.
The ML-SP2 and ML-SP2A alloys were prepared by vacuum arc melting nominal mixture of pure elements (purity > 99.9 wt%) in high purity argon atmosphere. The ingots were flipped and re-melted for 4 times to ensure chemical homogeneity, and Ti getter was melted before each re-melting process to minimize the O2 and N2 contamination in the ingot. The as-cast button-shaped ingot was 32 mm in diameter, and 12 mm in thickness at the center. The as-cast ingot was cut into several bar-shaped samples with dimension of 10 × 7 × 4 mm3 by electrical discharge machining (EDM). Surfaces of these samples were grinded with #80, #120, #240, #400, #600, #800, and #1200 SiC sandpapers in sequence, and subsequently oxidized at 1200°C for 10 hours in atmosphere, followed by furnace cooling.
Samples for X-ray diffraction (XRD) analysis were grinded with #80, #120, #240, #400, #600, #800, and #1200 SiC sandpapers in sequence, while those for scanning electron microscope (SEM) analysis were further grinded with #2500 SiC sandpaper and polished with 0.05 µm SiO2 colloid. The microstructure and composition of as-cast and oxidized samples were analyzed with JEOL-JSM 5410 scanning electron microscope equipped with backscatter electron (BSE) and energy dispersive spectroscopy (EDS). The crystal structure of the samples was identified by using Shimadzu XRD-6000 X-ray diffractometer.
Figure 1 shows the as-cast microstructure of ML-SP2 (Fig. 1(a)) and ML-SP2A (Fig. 1(b)) alloys. Dendritic structure and large amount of secondary phase could be observed in the interdendritic region for both alloys. The size of the secondary phases is generally under 3 µm in ML-SP2 alloy, and 6 µm in ML-SP2A alloy.
Backscattered electron images (BEI) of as-cast (a) ML-SP2 and (b) ML-SP2A alloys. Dendritic structure with finely dispersed precipitates in interdendritic region could be observed.
Table 2 shows the chemical composition of observed phases in as-cast ML-SP2 and ML-SP2A alloys. For ML-SP2 alloy, the dendrite arm (Fig. 1(a), marked by black arrowhead No. 1) contains higher Mo, Nb, and Ta contents, while the interdendritic region (marked by black arrowhead No. 2) is rich in Al, Cr, Ti, and V. Si has been found to distribute uniformly in the bulk. However, since there were lots of fine precipitates in interdendritic region, an accurate EDS analysis was not available on each phase. As a result, the EDS analysis result of interdendritic region could only be considered as the overall composition of the interdendritic region. Also, due to the characteristic X-ray signal overlapping between Si and Ta, the content of Si was significantly higher, and the EDS analysis results for both Ta and Si were merely qualitative. The enrichment of specific elements in different dendritic structure resulted from the casting segregations. Mo, Nb, and Ta partitioned toward dendrite arms, while other elements partitioned toward interdendritic region.
For ML-SP2A alloy, the dendrite arm (Fig. 1(b), marked by black arrowhead No. 1) was rich in Al, Mo, Ta, and V, while the precipitates (marked by black arrowhead No. 2, 3) in the interdendritic region were rich in Cr, Nb, Si, and Ti. The bright precipitates (marked by black arrowhead No. 2) were rich in Cr and Si, indicating that it was chromium silicide (CrSix). The dark region (marked by black arrowhead No. 3) surrounding the white precipitates was rich in Nb, Si, and Ti, indicating that this region was niobium-titanium co-silicide. The chemical segregation behavior of ML-SP2A was different from that of ML-SP2 alloy. For example, Al segregated to the dendrite arms, while Nb partitioned to the interdendritic region. In addition, the dendritic morphology and dendrite arm spacings of as-cast ML-SP2A alloy differed from those of as-cast ML-SP2 alloy. With increased Al and decreased V content, ML-SP2A possessed smaller dendrites and shorter dendrite arm spacings.
3.2 Phase identification before and after oxidationFigure 2(a) shows the X-ray diffraction (XRD) pattern of as-cast ML-SP2 and ML-SP2A. For both alloys, peaks from BCC lattice were the major feature of XRD pattern, indicating the fact that most of the bulk material had BCC crystal structure. The other signal found in XRD pattern was orthorhombic TiSix. Despite most of its signal overlapped with the signal from BCC crystal structure, there was still one of its peaks located at 2θ = 30°∼40°, which could be distinguished from other signals. TiSix signal might be resulted from (Nb, Ti)-rich silicide observed in the interdendritic region of both alloys (Fig. 1). Since the volume fraction of this silicide was small comparing to the bulk material, the XRD signal from this silicide was expected to be weak.
X-ray diffraction (XRD) pattern of (a) as-cast and (b) oxidized ML-SP2 (lower lines) and ML-SP2A (upper lines). The oxidation occurred by heating the samples at 1200°C for 10 hours in laboratory atmosphere. The oxide species presenting in both oxidized samples are the same, but (Cr, Nb, Ti) oxide showed slight difference in structure due to different chemical composition in this oxide. For the (Cr, Nb, Ti) oxide in oxidized ML-SP2, the identified oxide is (Ti0.2Cr0.4Nb0.4)O2, while it is (Ti0.8Cr0.1Nb0.1)O2 in oxidized ML-SP2A.
The XRD pattern of the oxidized ML-SP2 and ML-SP2A is shown in Fig. 2(b). The identified oxides were the same for both alloys with difference in the peak intensities that could be a result of different fractions of oxides. The oxides near the sample surface could be identified as aluminosilicate, SiO2, and (Cr, Nb, Ti)-rich oxide. This result is consistent with the following SEM observations. The VOx observed in ML-SP2 was not detected by XRD due to its small volume fraction at the surface of oxide layer. Also, the XRD pattern confirmed the presence of aluminosilicate at the surface of both samples, which is the most significant discovery in this work.
3.3 Microstructures after oxidationFigure 3 shows the cross-section area of as-cast ML-SP2 (Fig. 3(a)) and ML-SP2A (Fig. 3(b)) alloys after oxidizing at 1200°C for 10 hours. The average thickness of oxidized layer was 670 µm and 810 µm for ML-SP2 and ML-SP2A alloy, respectively. The outer surface of oxide was flat for ML-SP2 alloy, and on the contrary, the surface of oxidized ML-SP2A was rough. This could result from non-uniform oxidation of the larger interdendritic regions containing large amount of silicides in ML-SP2A, since the silicides in this region showed lower oxidation rate than other phases in this alloy, which will be shown later.
Cross-sectional BEI of (a) ML-SP2 and (b) ML-SP2A alloys after oxidizing at 1200°C for 10 hours. The thickness of oxide layer is uniform for ML-SP2, but this is not the case for ML-SP2A.
Near the surface of the oxidized sample, a distinct aluminosilicate layer (Fig. 4(a)(b), marked by white arrowhead No. 1) could be observed on both alloys, and there appeared to be a mixture with (Cr, Nb, Ti)-rich oxide particles (marked by white arrowhead No. 2). It is possible that aluminosilicate has similar thermodynamic stability as that of (Cr, Nb, Ti)-rich oxide; aluminosilicate layer would grow competitively against (Cr, Nb, Ti)-rich oxide layer. In this case, the morphology of the surface layer should contain at least one layer of either (Cr, Nb, Ti)-rich oxide or aluminosilicate, followed by a layer with mixture of both oxides. A typical structure of such oxidation mechanism can be found in the oxidation of γ-TiAl alloy.10) In Fig. 4, the surface layers showed sign of partial spallation (marked by thick white arrowheads). This indicates the “surface oxide layer” observed in this study might not be the original oxide surface layer. The original surface layer seemed to have spalled off before the samples were extracted from the furnace, leaving the mixed oxide layers. Nevertheless, this aluminosilicate layer could not provide sufficient oxidation resistance for both alloys due to limited degree of continuity. But it is still noteworthy that with increased Al and decreased V content, the degree of continuity of aluminosilicate layer appeared to be enhanced. Also, the retained VOx observed in ML-SP2 (Fig. 4(a), arrowhead No. 3) was not found in the surface layer of ML-SP2A. The chemical composition of different phases presented in Fig. 4 is listed in Table 3.
Cross-sectional secondary electron images (SEI) of oxidized (a) ML-SP2 and (b) ML-SP2A alloys near the surface. Notice the aluminosilicate layer (marked by white arrowhead No. 1) on the surface of the oxide layer.
For both alloys, the internal oxidation zone under the surface layer was porous (Fig. 4, marked with thick black arrowheads), and it resembled the microstructural feature from the substrate. (Cr, Nb, Ti)-rich oxide region corresponded the dendrite arms and was the main constituent. On the other hand, of the interdendritic region was filled with pores, Al2O3, and SiO2. Figure 5 shows the typical structure of the internal oxidation zone for both alloys, and the chemical composition of different phases shown in the figure is listed in Table 4. According to EDS analysis, traces of the VOx (marked by black arrowhead No. 2) could be found scattered across the whole internal oxidation zone for ML-SP2. They were usually adjacent to the (Cr, Nb, Ti)-rich oxide (marked by black arrowhead No. 1), or at the edges of pores, along with cracks and Al2O3 or SiO2.
Cross-sectional SEI of oxidized (a) ML-SP2 and (b) ML-SP2A alloys at the middle of oxide layer. The grain-like, bright phase is Cr-rich oxide, while other dark region is either pores or Al2O3/SiO2 – filled region. White area near the cracks is caused by electron accumulation induced by low conductivity region near the cracks.
The VOx observed in the sample is likely to be V2O5 because of the fact that V2O5 is thermodynamically stable up to beyond 1400°C in the atmosphere.9) Since VOx is generally very volatile at high temperature,7) very little VOx was found within the pores of oxide layer and V content was greatly reduced in the surface layer and the internal oxidation zone (Table 3 and 4). From the observation mentioned above, it is reasonable to assume that the evaporation of VOx contributed to the formation of porous internal oxidation zone. Furthermore, with lower V and higher Al content in ML-SP2A, the size of the pores in oxide layer, which were roughly the spacing between the dendrite arms in the substrate, appeared to be smaller than that of ML-SP2 (Fig. 5). The change in the pore size of ML-SP2A alloy also appeared to improve the continuity of aluminosilicate layer (Fig. 4(b)).
Another noteworthy phenomenon in this study was that part of the space between (Cr, Nb, Ti)-rich oxide particles was filled with either Al2O3 or SiO2 (Fig. 5). These Al2O3 and SiO2 formation were likely occurred during early stage of oxidation due to their low partial oxygen pressure of formation.11) However, since these Al2O3 and SiO2 were not continuous in form and could not hinder further oxidation, VOx would form, and fit in the spaces within the structurally loose (Cr, Nb, Ti)-rich oxide. But since VOx is very volatile,12) the VOx would disappear, leaving cavities in the oxide layer (Fig. 5(a), marked by black arrowhead No. 3).
Moving downward to the vicinity of the interface between the oxidized region and metallic substrate, ML-SP2 and ML-SP2A exhibited different microstructures (Fig. 6), and the chemical composition of different phases in this region is listed in Table 5. For ML-SP2 alloy, the dendritic structure extending from the substrate transformed into grain-like oxide, which was the major morphology of the oxide layer, in a very narrow region with thickness of 20 µm (Fig. 6(a), marked by red square). The bright grey phase (marked by black arrowhead No. 1), which was the oxidized dendrite arm, contained high amount of V, while the dark grey phase (marked by black arrowhead No. 2), which was the oxidized interdendritic region, contained high amount of Si. Due to the small size of dark grey phase, the result of EDS analysis on this phase would be affected by nearby bright grey phase. A belt-shaped phase rich in V and Mo could also be found in this region (marked by black arrowhead No. 4). The dark phase (marked by white arrowhead) in this region was Al2O3. For ML-SP2A, the thickness of transition region between substrate and oxide layer was 120 µm (Fig. 6(b), marked by red square). The bright grey (V-rich) phase and dark phase (Al2O3) still presented in this region, but the dark grey phase (marked by black arrowhead No. 1) here was identified as (Al, Mo)-rich phase.
Cross-sectional SEI of oxidized (a) ML-SP2 and (b) ML-SP2A alloys at the interface between oxide layer and metallic substrate. The thickness of the interface is very narrow (∼20 µm) for ML-SP2, while the thickness of this interface is relatively wide (∼120 µm) for ML-SP2A.
Since MoOx could not be found elsewhere in the oxidized region for both alloys, and little traces of Mo concentration was found at the internal oxidized region (Table 3, 4, and 5), while the substrate retained the original Mo content, it is possible that the evaporation of MoOx occurred at early stage of the oxidizing progress. The most common form of MoOx observed at high temperature is MoO3, which sublimates at above 600°C.13) However, the retained Mo content in the substrate indicated that Mo did not diffuse outward rapidly during the oxidation, neither was the oxygen partial pressure in the substrate high enough to induce the oxidation and evaporation of Mo. The bright phase (marked by black arrowhead No. 2 in Fig. 6) was identified as CrSix, which was also observed in the interdendritic region of the substrate. This indicates that CrSix can resist oxidation better than other phases in the substrate, but its role in the oxidation behavior still needs to be clarified.
There are two important observations in this study. First, an obvious aluminosilicate layer, which could be an oxygen barrier, can be observed on the surface of both alloys in this work. From the observed correlation between the composition difference and the integrity of aluminosilicate layer, increasing Al content and decreasing V content appear to be a reasonable approach for improving the continuity of aluminosilicate layer. But another possible factor affecting the continuity of aluminosilicate layer is the degree of segregation relating to as-cast dendritic structure. Second, VOx evaporation will add extra contribution to the formation of porous internal oxidation zone, which is not ideal for surface stability. Although V is a low density refractory element, its oxide is volatile and shows no benefit to the oxidation resistance of the alloy. As a result, it is recommended to minimize the content of V in future study.
Two refractory HEAs containing Al, Cr, and Si were prepared. The microstructure of as-cast and oxidized samples for both alloys were investigated. The findings can be summarized as followings:
