Degradation of Yb–Gd–Ti-Si middle entropy silicides in oxidizing atmosphere

Using arc melting, a novel rare-earth silicide, middle entropy Yb–Gd–Ti-Si, was fabricated for high-temperature material applications. The oxidation behavior of the Yb–Gd–Ti-Si was evaluated through oxidation tests conducted at 1200 °C for 1, 2, 4, and 8 h in air. The oxidation rate at 1200 °C was almost the same, regardless of the addition of Ti. The oxidation rate of Yb–Gd–Ti-Si at lower temperatures (600 °C to 900 °C) was lower than that of Yb–Gd–Si. In addition, detailed microstructural observations indicate that the formation of TiO2 and other oxides in the Yb-Ti–O (Yb–Ti–Si–O) phase suppresses the formation of Yb2O3, causing a drastic oxidation at intermediate temperatures (600 °C to 900 °C). These results indicate that the addition of Ti to Yb–Gd–Si is effective at preventing the preferential oxidation of the grain boundaries at 600 °C to 900 °C, which is commonly observed in metal silicides.


Introduction
Transition metal silicides (e.g., MoSi 2 , CrSi 2 , ZrSi 2 , and NbSi 2 ) have been used as high-temperature materials in oxidizing atmospheres owing to the formation of SiO 2 scales on their surfaces (at temperatures of 1000°C), which prevent further oxidation [1][2][3][4][5][6][7][8][9][10]. At temperatures of <1000°C, drastic oxidation with the formation of metal oxides occurs in the absence of SiO 2 scales. For example, MoSi 2 forms SiO 2 scales even at above 1400°C, and the oxidation rate is relatively low in air (7.06×10 -14 m·s -1 ). [2] At temperatures between 600°C and 900°C , a drastic oxidation of MoSi 2 proceeds owing to the preferential oxidation of the grain boundaries, which prevents MoSi 2 from maintaining its form through complete oxidization. However, the oxidation behavior of MoSi 2 at intermediate temperatures improves with the addition of B, Al, Ti, Ta, Zr, and Y because the value of Gibbs-free energy of the oxide formation at 600°C to 900°C is lower than that of Si, preventing the preferential oxidation of the grain boundaries [4,5,11].
In addition to transition metal silicides, rare-earth silicides have also been studied as coating materials in aero gas turbine engines [12][13][14][15]. Such use is due to the oxidation of a conventional coating material (Si) in a combustion environment (T<1300°C, 10−30 atm [16]), forming cristobalite, which is a polymorphous type of SiO 2 , thereby causing the delamination of the coating. Recent preliminary results demonstrated the hightemperature oxidation of ytterbium silicides (Yb 3 Si 5 and Yb 5 Si 3 ) at up to 1200°C in both air and steam [15]. The preferential oxidation of Yb prevents the formation of SiO 2 and protects it from further oxidation, forming Yb 2 O 3 , Yb 2 SiO 5 , and Yb 2 Si 2 O 7 . However, the Yb 3 Si 5 and Yb 5 Si 3 phases are unstable because the Si-Yb 3 Si 5 eutectic phase (at a melting temperature of 1125°C) evolves during oxidation, which is caused by a decrease in the Yb content in these silicides. Specifically, the rapid weight gain and formation of Yb 2 O 3 (at between 700°C and 900°C), caused by the preferential oxidation of Yb, is still a critical stability issue.
Because compounds in Gd-Si systems also have higher melting points than Si, Morigayama et al studied the phase stability of Yb 25 Gd 20 Si 55 ternary silicides at high temperatures (1500°C) using a calculated phase diagram (CALPHAD) to improve the phase stability [17] The results proved that the addition of Gd prevents oxidation, and the Yb-Si phase stability improves because no liquid phase is formed and the weight gain is suppressed at up to ∼1200°C [17,18]. However, at between 700°C and 900°C, a rapid weight gain has been observed owing to the formation of Yb 2 O 3 , leading to the preferential oxidation of the grain boundaries, which is still a critical issue in Yb-silicides and Yb-Gd-Si systems. The prevention of oxidation at 700∼900°C is one of the most important challenges to design Yb-silicides and Yb-Gd-Si as a candidate of coating materials in aero gas turbine engines. However, the reports on the improvement of oxidation behavior of them is limited to the best of the author's knowledge.
The objective of this study is to improve the oxidation behavior of a Yb-Gd-Si ternary system at low temperatures. In this research, Ti was added to Yb-Gd-Si to improve the oxidation behavior of silicides and ceramics (e.g., MoSi 2 and Ni-based superalloys) [9,16]. This is mainly because the value of Gibbs-free energy for the oxidation of Ti is lower than that of Si and Ni, which causes the preferential oxidation of Ti and prevents the oxidation of the grain boundaries. In addition, as the entropy of the mixture increases, thermodynamic oxidation is expected to decrease.

Materials and methods
Yb-Gd-Ti-Si was fabricated using an arc-melting furnace. In addition, Yb (99.9% purity, Mitsui Kinzoku Trading Co. Ltd, Japan), Gd (99.9% purity, Mitsui Kinzoku Trading Co. Ltd, Japan), Ti (99.9% purity, Nirako Co. Ltd, Japan), and Si (99.99% purity, Kojundo Chemical Co., Japan) granules were used as the raw materials. Granular mixtures of Yb:Gd:Ti:Si with relative proportions of 30:14.5:0.5:55 (at%) and 30:14:1:55 (at%) were prepared (which are hereafter denoted as Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si, respectively). The compositions were determined from the phase diagram of Yb-Gd-Si using CALPHAD (FactSage 7.3, Centre for Research in Computational Thermochemistry, Canada) with a melting point of 1500°C. Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si were prepared by arc-melting methods. First, granular mixture of them were put onto arcfurnace. To avoid oxidation during process, evacuation and the introduction of Ar is repeated by three times. After that, arc-furnace was filled with Ar with the pressure of ∼0.06 atm. Owing to the rapid evaporation of Yb in an inert atmosphere, arc melting was repeated four times with the addition of Yb. A similar description of the preparation process can be also found in our previous report [17].
The microstructure and composition of the samples were characterized using scanning electron microscopy (SEM; JSM-6510SA, JEOL, Japan) equipped with x-ray energy dispersive spectroscopy (EDX). To study the structural properties, x-ray diffraction (XRD) using a MiniFlex 600 Rigaku, Japan) was carried out within the 2θ range of 20°-80°using CuKα radiation (at a wavelength of 0.154 nm). Furthermore, using the intercept method, the surfaces of the samples were etched using aqua regia to evaluate the mean grain size. Since quantitative define of crystallite was quite difficult in the present study, analyses for grain size were conducted. The grain size can be estimated using the mean linear intercept length (l¯), which is expressed through the following equation: where L and n L represent the length of a line drawn in a micrograph and the number of grains intersecting that line, respectively. The area of the grain boundaries (S v ) is then calculated using l¯as follows:

Oxidation tests
To evaluate the oxidation rates, the samples (6.0 mm×6.0 mm×4.0 mm) were exposed to air at 1200°C for 1, 2, 4, and 8 h in an infrared (IR) furnace (IR-QP1-4S, Yonekura MFG Co., Ltd, Osaka, Japan). To avoid oxidation while reaching the desired temperature, a flow of Ar (500 ml min -1 ) was used to maintain an inert atmosphere. After reaching 1200°C, the Ar flow was closed and air was introduced at a rate of 500 ml min -1 . After the test time at 1200°C, the air flow was stopped, the Ar flow (500 ml min -1 ) was reinitiated to prevent further oxidation, and the samples were cooled to room temperature (25°C) at a cooling rate of ∼50°C min -1 . The detailed conditions of the heat-exposure test using an IR furnace can be found in [17]. Additional oxidation tests were conducted to evaluate the oxidation of the samples at lower temperatures. The samples were heated to 700°C and 900°C at a heating rate of 50°C min -1 , and after reaching 700°C or 900°C, the samples were cooled to room temperature at a cooling rate of ∼50°C min -1 .
After each test, the heat-exposed samples were embedded in an epoxy resin to observe their cross-sections. The microstructures were observed and characterized using SEM. The oxidized region was measured by using image processing software (Image J, NIH) from cross-sectional images. To evaluate the crystal structures of the samples, an XRD analysis was conducted under the same conditions as those mentioned above. In addition, the weight gain of the samples during heat exposure was measured through a thermogravimetric analysis (TGA) using a Netzsch 2000 SE-a (Netzsch, Germany) with a uniform weight of ∼10 mg, and the samples were heated to 1200°C at a heating rate of 5°C min -1 .  figure 1(c)) and a darker gray region (P2 in figure 1(c)), which indicates a two-phase region. The EDX map and point analysis of these regions (see figure 1(c) and table 1) indicate two Yb-Gd-Si phases, with the Yb, Gd, and Si contents within the range of 15-30, 10-30, and 50-65 at%, respectively. These results were expected because Yb-Gd-Si has two stable phases within these compositional ranges. In the EDX map analysis, the darkest randomly distributed regions were attributed to the Ti-Si phases because the signals of Ti and Si overlapped ( figure 1(c)). The formation of the Ti-Si phases decreased the Si content in the Yb-Gd-Si phase, which revealed that the melting point decreased, as shown in the Yb-Gd-Si ternary phase diagram ( figure 2). However, the melting point of the silicides is 1500°C owing to the relatively low Ti content (0.5-1.0 at%), which indicates that the stability of the Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si phases is almost the same as that of Yb-Gd-Si. Figure 3 shows the XRD patterns of Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si. Reference peaks for Yb-Gd-Si and Y 2 O 3 is also shown [17,19]. No significant changes were observed in the XRD patterns upon the addition of Ti. This result was expected because the amount of Ti in the samples (0.5%-1.0%) was much lower than that of Yb, Gd, and Si. Furthermore       Yb 30 Gd 14.5 Ti 0.5 Si, whereas S v is lower. These results suggest that the grain size of the samples increased with the addition of Ti, and the area of the grain boundaries decreased.

Oxidation behavior of Yb-Gd-Ti-Si
The TG curves of Yb-Gd-Si, Yb 30 Gd 14.5 Ti 0.5 Si, and Yb 30 Gd 14 Ti 1 Si in air are shown in figure 6. The total weight gain during the TG test increased with the addition of Ti, and the rate of weight gain for both silicides was basically the same at above 900°C ( figure 6(a)). However, it seems that the weight gain of Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si at intermediate temperatures (∼600°C-900°C) is lower than that of Yb-Gd-Si. The largest difference was observed at between 800°C and 900°C. An enlarged view within the range of 850°C and 900°C shows that the oxidation rates of Yb-Gd-Si and Yb 30 Gd 14.5 Ti 0.5 Si were within the range of ∼1.2×10 -2 and 2.3×10 -2°C-1 , whereas that of Yb 30 Gd 14 Ti 1 Si was ∼9.9×10 -3°C-1 ( figure 6(b)). These results clearly show that the oxidation rate decreases with an increase in the Ti content at 600°C to 900°C. This implies that the addition of Ti suppresses the formation of Yb 2 O 3 during oxidation. BSE images along with EDX analyses for Yb-Gd-Si [17] and Yb 30 Gd 14 Ti 1 Si after oxidation at 1200°C for both 1 and 8 h time frames are shown in figure 7. For Yb-Gd-Si, the continuous bright contrast region corresponds to Yb 2 O 3 , and the other contrast regions represent oxides in the Yb-Gd-Si-O system with different contents of Yb, Gd, and Si, as reported in a literature [17].    ( ) D =where S a and S b represent the entropies before and after the reaction, respectively. An increase in S mix D induced an increase in S , b and the absolute value of G r D decreased. Assuming that the difference in S mix D for the silicides affects G , r D for Yb-Gd-Ti-Si (and Yb-Gd-Si), G r D decreases by ∼4-5 kJ mol −1 compared with that of Yb3Si5 because the weight gain of these silicides at 1200°C decreases by 2% in comparison with Yb 3 Si 5 . Unfortunately, the weight gain and oxidation rate of these silicides depend on the oxides formed on their surfaces, and it is difficult to extract the effect of the mixed entropy on the oxidation behavior. As discussed previously [18], the formation of complex oxides containing Yb-Gd-Si-O (and Yb-Gd-Ti-Si-O) acts as a barrier for oxygen diffusion into the unoxidized area, and similar mechanisms have been reported in studies on the oxidation mechanism of mid-and high-entropy materials [20][21][22].
To clearly understand the effect of Ti on the oxidation behavior of Yb-Gd-Si, an evaluation at temperatures of lower than 1200°C was necessary. BSE images of Yb 30 Gd 14 Ti 1 Si with EDX mapping (oxygen) were obtained after oxidation at 700°C and 900°C and are presented in figure 10(a). After oxidation at 700°C, oxides with a thickness of ∼10-20 m were formed on the surface, and oxidation proceeded along the cracks toward the inside of the sample by ∼200 μm in the depth direction. The size of the oxidized region increased by 300-400 μm with an increase in the oxidation temperature, and a continuous oxide region with a thickness of ∼100 μm was formed on the surface. However, no oxidation along the cracks was observed in the sample after oxidation at 900°C . An EDX analysis of the samples after oxidation at 700°C ( figure 8(b)) indicates the formation of oxides composed of Yb, Ti, and O near the surface.
Although the Gibbs-free energy for the formation of TiO 2 at 700°C (∼-800 kJ mol -1 [23]) is higher than that of Yb 2 O 3 (∼-1000 kJ mol -1 ), the oxides at a depth of 100 μm were composed of Ti and O. This shows that the formation of TiO 2 prevented the formation of Yb 2 O 3 as well as further oxidation at 600°C to 900°C, which was observed through the oxidation of Ti-added MoSi 2 [11]. The TiO 2 also prevented the formation of cracks in the samples and oxidation through intergranular corrosion because Ti was present in the intergranular corrosion regions (see figure 3). Moreover, the authors hypothesize that the formation of oxides in the Yb-Ti-O system prevented the formation of Yb 2 O 3 because a similar behavior was observed in the Yb-Si-O of the Yb-Si system [15,24]. In this case, Yb 2 SiO 5 and Yb 2 Si 2 O 7 were formed despite the Gibbs-free energy during the formation of SiO 2 being much higher than that of Yb 2 O 3 . The formation of silicates is also reported in Yb 2 O 3 -SiO 2 and Gd 2 O 3 -SiO 2 system [25]. The oxides formed in the Yb-Ti-O system also prevented the formation of a continuous Yb 2 O 3 phase. In addition, the effect of S mix D on corrosion at 600°C to 900°C was limited because of the contribution of entropy to the Gibbs-free energy, considering that the reaction increases with an increase in temperature, as shown in equation (4). A detailed observation of the cross-sections in figures 10(c) and (d) reveal that complex oxides composed of Yb, Gd, Ti, and Si are formed after oxidation at 900°C. Detailed thermodynamic and microstructural analyses of complex oxides are required to evaluate the changes in microstructure caused by the addition of Ti. However, the results presented herein suggest that the severe oxidation of Yb-Gd-Si silicides at 600°C to 900°C can be avoided by adding 1 at% Ti. Therefore, the application of rare-earth silicide as a high-temperature material is expanded owing to the suppression of oxidation and Yb 2 O 3 formation by the TiO 2 and oxides present in the Yb-Ti-O system.
Although the mechanism of oxidation behavior is quite complex, the evaluation conducted in the present study clearly reveals that the addition of Ti by 1at% (Yb-Gd 14 Ti 1 Si) in Yb-Gd-Si is effective to improve oxidation of Yb-Gd-Si because the decreases of the oxidation rate at 600∼900°C by ∼60% is realized.

Conclusion
The oxidation behaviors of mid-entropy silicides Yb 30 Gd 14.5 Ti 0.5 Si and Yb 30 Gd 14 Ti 1 Si were evaluated. The conclusions drawn from this study are as follows.
The oxidation behaviors of Yb-Gd-Si, Yb 30 Gd 14.5 Ti 0.5 Si, and Yb 30 Gd 14 Ti 1 Si at 1200°C were almost the same. In addition, the weight gain decreased by 2% compared with that of Yb 3 Si 5 , suggesting that the addition of Ti to Yb-Gd-Si did not affect the oxidation behavior at 1200°C.
The oxidation of Yb-Gd-Si at intermediate temperatures (600°C to 900°C) was improved by the addition of 1 at% Ti. This occurred mainly because the oxidation rate of Yb 30 Gd 14 Ti 1 Si within the temperature range of 850°C to 900°C was ∼9.9×10 -3°C-1 and that of Yb-Gd-Si and Yb 30 Gd 14.5 Ti 0.5 Si was ∼1.2-2.3×10 -2°C-1 . The addition of Ti increased the grain size of the samples, causing the formation of TiO 2 , leading to the formation of oxides in the Yb-Ti-O system and more complex oxides composed of Yb, Gd, Ti, and Si. As a result, the formation of Yb 2 O 3 was suppressed, and the weight gain decreased at intermediate temperatures.