Study of solidification pathway of a MoSiBTiC alloy by optical thermal analysis and in-situ observation with electromagnetic levitation

MoSiBTiC alloys are promising candidates for next-generation ultrahigh-temperature materials. However, the phase diagram of these alloys has been unknown. We have developed an ultrahigh-temperature thermal analyser based on blackbody radiation that can be used to analyse the melting and solidification of the alloy 67.5Mo–5Si–10B–8.75Ti–8.75 C (mol%). Furthermore, electromagnetic levitation (EML) was used for in-situ observation of solidification and microstructural study of the alloy. On the basis of the results, the following solidification pathway is proposed: Mo solid solution (Moss) begins to crystallize out as a primary phase at 1955 °C (2228 K) from a liquid state, which is followed by a (Moss+TiC) eutectic reaction starting at 1900 °C (2173 K). Molybdenum boride (Mo2B) phase precipitates from the liquid after the eutectic reaction; however, the Mo2B phase may react with the remaining liquid to form Moss and Mo5SiB2 (T2) as solidification proceeds. In addition, T2 also precipitates as a single phase from the liquid. The remaining liquid reaches the (Moss + T2 + TiC) ternary eutectic point at 1880 °C (2153 K), and the (Moss + T2 + Mo2C) eutectic reaction finally occurs at 1720 °C (1993 K). This completes the solidification of the MoSiBTiC alloy.

primary phase regions in the liquidus projection for the Mo-rich portion using SEM and TEM results. Thus, combining thermal and microstructural analyses is essential to establishing phase diagrams.
However, conventional DTA or differential scanning calorimetry (DSC) is insufficient for studying the phase diagram of a MoSiBTiC alloy because of its more complicated solidification steps and technical problems caused by ultrahigh temperature (>1900 °C). Platinum, usually used as a container and thermocouple material, has a too low melting point (1768 °C) to be used for thermal analysis of MoSiBTiC alloys. In addition, uncertainty in temperature measurement becomes significant at ultrahigh temperatures owing to a lack of well-defined fixed-temperature points.
To overcome these experimental difficulties, we have developed a method for ultrahigh-temperature thermal analysis of MoSiBTiC alloys using blackbody radiation. This involves using a radiation pyrometer to measure blackbody radiation emitted from a sample holder having a blackbody cavity. The radiation pyrometer is calibrated at 1950 °C using the eutectic temperature of the Ru-C system. We use the method to thermally analyse one of the alloys to study the reaction temperatures during its solidification process. Although differential thermal analysers can be operated above 2000 °C using a commercially available W-Re thermocouple, the thermocouple will suffer from drift owing to aging and contamination. The advantage of the new method is in the higher accuracy of the temperature measurement and the larger sample volume, which allows for the detection of smaller heat effects.
The thermal analysis requires a way to relate reaction heat peaks to corresponding microstructure formation to reveal the solidification pathway of the MoSiBTiC alloy. Here, we use EML to in-situ observe its solidification. We developed an EML technique coupled with a static magnetic field in our previous work [24][25][26][27][28] . This technique enables high-precision, non-contact thermophysical measurements of high-temperature melts. In this work, we use the technique to in-situ observe the solidification of the MoSiBTiC melt, and study the microstructure obtained from a quenched sample of the EML experiment. Finally, the solidification of the alloy is discussed in light of the thermal analysis and in-situ observation. Figure 1 is a schematic of our ultrahigh-temperature thermal analysis apparatus. This apparatus consists of a radio-frequency furnace, two pyrometers and a crucible having a blackbody cavity. A sample is filled into the crucible and heated through a graphite susceptor by applying a radio-frequency current to the heating coil. The radiation from the blackbody cavity is observed with a pyrometer at the top of the apparatus through a quartz glass and a mirror, measuring the sample temperature T s . The susceptor temperature T c is monitored with a pyrometer at the bottom of the apparatus, which is used for temperature control. The temperature determination of the sample is based on the Planck equation

Principle of Ultrahigh-Temperature Thermal Analysis
where E b is the emissive blackbody power, λ is the wavelength, T is the absolute temperature, and C 1 and C 2 are the first and second radiation constants equal to 3.7415 × 10 −16 W·m 2 and 1.4388 × 10 −2 m·K, respectively. The emissivity ε c of the blackbody cavity is expressed in terms of the emissivity ε m of the crucible material and aspect ratio α of the cavity 29 : www.nature.com/scientificreports www.nature.com/scientificreports/ The value of ε c was calculated as 0.999 for graphite with the aspect ratio of the blackbody cavity (diameter of 5 mm; depth of 36 mm). The emissive power of the blackbody cavity, E c , is where τ is the absorptivity of the quartz glass and mirror, and ε app is the apparent emissivity. From Eqs (1) and (3), the absolute temperature is expressed using the emissive power of the blackbody cavity, E c : In this study, ε app was obtained using fixed-temperature points. The highest fixed-temperature point defined by the international temperature scale (ITS-90) is the melting point T m of Cu (1084.62 °C) 30 . Yamada et al. 31 have proposed eutectic temperatures (T E ) of metal-carbon binary systems for fixed-temperature points above T m for Cu. Because the liquidus temperature of MoSiBTiC alloys seems to be higher than 1900 °C, T E values for Ru-C (1953 °C 31 ) and Ni-C (1329 °C 31 ) alloys were chosen as the fixed points in addition to T m for Cu. We considered the effect of eutectic reaction kinetics on measuring fixed points on the basis of the Fe-C eutectic transition temperature reported by Sasajima et al. 32

Results
Ultrahigh-temperature thermal analysis. Fixed-temperature points. The ultrahigh-temperature analyser was calibrated with the melting point of pure Cu and the eutectic temperatures of the Ru-C and Ni-C alloys. As an example, Fig. 2 shows the apparent temperature curves of the Ru-C eutectic alloy obtained upon heating and cooling assuming ε app = 1. The sample temperature was initially set lower than T E by 10 °C and kept there for thermal homogenisation. The sample was heated again to a temperature higher than T E by 5, 10 and 20 °C until the eutectic plateau was completed as shown in Fig. 2(a). After the eutectic melting plateau was completed, the sample was cooled to a temperature lower than T E by 5, 10 and 20 °C to obtain the eutectic plateau on cooling ( Fig. 2(b)). In Fig. 2(a), the duration of the eutectic plateau on heating decreases, meaning the eutectic reaction rate increases, with increasing set temperature. However, the eutectic temperature does not change with reaction rate. The cooling curves in Fig. 2(b) show that the eutectic temperature after undercooling slightly depends on set temperature. This difference is understood in terms of eutectic solidification kinetics. Thus, the eutectic plateau obtained on heating was used as the fixed point in this study, and its average apparent T E for the Ru-C alloy was 1928 °C. By substituting the apparent T E and ε app = 1 into Eq. (4), E c was obtained for the Ru-C alloy, and E b was estimated via Eq. (1) for the true T E of the Ru-C alloy (1953 °C). Similar measurements were conducted to obtain T E for the Ni-C alloy and T m for Cu. Figure 3 shows a linear relationship between E b and E c for these fixed-temperature points. The apparent emissivity ε app was determined to be 0.945 from the slope. From these results, the uncertainty in temperature measurements for the thermal analysis was evaluated as ±0.4% within a temperature range from 1084.62 to 1953 °C.
MoSiBTiC alloy. The composition of the MoSiBTiC alloy used in this study was 67.5Mo-5Si-10B-8.75Ti-8.75 C (mol%). The alloy was thermally analysed using the above method with ε app = 0.945. Heating and cooling steps were repeated three times. Figure 4 shows a typical cooling curve of the MoSiBTiC alloy obtained in the second cooling step. Five inflection points appear on the cooling curve. Table 1 shows the temperatures of all inflection points and their averages for three cooling curves. Reactions corresponding to the inflection points will be discussed later. alloy was levitated and melted using EML in a static magnetic field to observe the solidification of the melt. Figure 5 (top) shows the cooling temperature profile of the MoSiBTiC alloy melt after the heating laser was turned off. We recorded multiple light emissions from the alloy melt owing to the release of latent heat (recalescence), and these emissions were assigned to peaks A-E on the temperature profile from the in-situ CCD camera observation. An example is the bottom of Fig. 5, which is a sequential top-view image of the alloy melt around recalescence D, at which the melt surface became brighter. Consequently, solidification of the melt started with recalescence A and finished with recalescence E. Here, the sample temperature was measured using a pyrometer calibrated with the liquidus temperature (1955 °C) of the alloy as determined in the preceding section. However,    www.nature.com/scientificreports www.nature.com/scientificreports/ after recalescence, the sample temperature was affected by solids precipitated on the melt surface because the partially solidified portion had a different emissivity from that of the melt. For this reason, the recalescence points A-E correspond to the inflection points a-e in Fig. 4, but the recalescence temperatures do not agree with the inflection points.
Solidification microstructural analysis. The levitated MoSiBTiC alloy melt was rapidly solidified from recalescence in EML to freeze the microstructure at that temperature. For microstructural analysis, the partially solidified alloys were held at these temperatures for 10 min just after recalescences B and D as shown in Fig. 6(a,b), respectively, to develop the microstructures. This was followed by the rapid solidification under a He gas flow. This figure also shows partial temperature profiles during the rapid cooling with an enlarged time scale, and top-view images of the partially solidified alloys during the holding time. Solid precipitated from the alloy melt is seen in the laser spot at the centre. Liquid and solid phases are distinguished by their emissivity difference on the alloy surface. Figure 7(a,b) show the XRD profiles of the quenched samples obtained via the EML experiments of Fig. 6(a,b), respectively. The diffraction peaks of molybdenum solid solution (Mo ss ) were clearly observed, and Mo 5 SiB 2 (T 2 ), TiC, Mo 2 C and Mo 2 B phases were also identified in the quenched samples. Figure 8(a,b) show the cross-sectional SEM images of the quenched samples obtained via the EML experiments of Fig. 6(a,b), respectively. Phase identification was performed with SEM energy-dispersive X-ray spectrometry (SEM-EDX) considering the XRD results. As shown in Fig. 8(a), the microstructure has large Mo ss grains (white), which should be primarily crystalized in the alloy melt. A TiC (black) phase is seen in the (Mo ss + TiC) binary-eutectic structure, as are some minor phases: Mo 2 B (light grey), T 2 (grey) and Mo 2 C (dark grey). Furthermore, fine (Mo ss + T 2 + TiC) and (Mo ss + T 2 + Mo 2 C) ternary-eutectic structures are observed. In the microstructure in Fig. 8(b), Mo ss and T 2 phases and (Mo ss + TiC) binary-eutectic structure are seen. The (Mo ss + T 2 + TiC) ternary-eutectic structure is more developed than the structure in Fig. 8(a). A fine (Mo ss + T 2 + Mo 2 C) ternary-eutectic structure is also observed.

Discussion
Solidification pathway. We discuss the solidification pathway of the MoSiBTiC alloy (67.5Mo-5Si-10B-8.75Ti-8.75 C in mol%) in light of the thermal analysis and EML experiments. The results are as follows: 1. Five inflection points (a-e) were detected on the cooling curve in Fig. 4. 2. Five recalescences (A-E) were in-situ observed as shown in Fig. 5. 3. Mo ss , Mo 2 B and T 2 single phases, one binary-eutectic structure (Mo ss + TiC) and two ternary-eutectic structures (Mo ss + T 2 + TiC) and (Mo ss + T 2 + Mo 2 C) were observed in the microstructures of Fig. 8(a,b).
From the above results, we propose the solidification pathway for the MoSiBTiC alloy in Table 2. Inflection points a-e in Fig. 4 correspond to recalescences A-E in Fig. 5, respectively. We note again that, except for the liquidus temperature, the recalescence temperatures in Fig. 5 do not agree with the inflection points of Fig. 4  Effect of undercooling on the solidification pathway. Deep undercooling may affect the solidification pathway of alloy melts. For example, Lü and Wang 33,34 reported that the transition of the primary phase crystallization to the non-equilibrium peritectic phase crystallization is dependent on the degree of undercooling of the melt. In  After the first undercooling, recalescence A occurred but the sample temperature did not increase to the initial liquidus temperature. This does not imply a metastable phase formation, but was simply caused by the liquidus composition shift toward the boundary line between Mo ss and TiC primary phases during recalescence A and the subsequent (Mo ss + TiC) binary-eutectic reaction. Therefore, the sample temperature rise was lowered from the initial liquidus temperature and was limited to the (Mo ss + TiC) binary-eutectic temperature (1900 °C). This is understood by the microstructural observation. The microstructures shown in Fig. 8 indicate that the large Mo ss grains were primarily crystalized in the alloy melt, which was followed by the (Mo ss + TiC) binary-eutectic reaction. Thus, the undercooling did not affect the solidification pathway in the present study.
Discussion based on the Mo-Si-B-TiC phase diagram. The chemical composition of the MoSiBTiC alloy (67.5Mo-5Si-10B-8.75Ti-8.75 C in mol%) we used is located on the Mo-Si-B-TiC quaternary phase diagram in Fig. 9(a). The alloy composition is also expressed as 75.6Mo-8.9Mo 5 SiB 2 (T 2 )-15.6TiC in mol% in the  Fig. 6(a,b). Fig. 4 Recalescence appeared in Fig. 5 Reactions occurred at the inflection temperature  www.nature.com/scientificreports www.nature.com/scientificreports/ Mo-T 2 -TiC ternary system, and is plotted in Fig. 9(b). In this ternary system, the alloy composition is initially located in the Mo ss primary phase region. Therefore, after the Mo ss primary phase precipitates at 1955 °C (inflection point a), the liquid composition shifts toward the Mo ss /TiC boundary lines, then the binary eutectic reaction

Inflection point obtained in
Mo TiC (5) ss occurs at 1900 °C (inflection point b) in Fig. 9(b). Eremenko and Velikanova 35 reported a eutectic temperature of 2175 °C for the Mo-TiC binary system. Here, the reaction (5) occurred at 1900 °C in the Mo-Si-B-TiC system. However, the alloy composition is projected onto the basal plane (the Mo-Si-B ternary system) as shown in Fig. 9(c) 23 . The initial alloy composition exists in the Mo ss primary phase region on the connecting line between Mo and Mo 5 SiB 2 (T 2 ). However, Mo ss content in the liquid is decreased by the eutectic reaction (5), and therefore, the liquid composition moves into the Mo 2 B primary phase region as shown in Fig. 9(c). However, this Mo 2 B phase soon reacts with the liquid to form Mo ss and T 2 via the reaction Nunes et al. 36 reported that this reaction occurs in the four-phase equilibrium at about 2100 °C in the Mo-Si-B ternary system. In our study, the reaction (6) occurred at 1894 °C (inflection point c) in the Mo-Si-B-TiC system.
Going back to the Mo-T 2 -TiC ternary phase diagram ( Fig. 9(b)), the liquid reaches the ternary eutectic point after the eutectic reaction (5), then the following reaction occurs at 1880 °C (inflection point d): ss 2 In addition, the elongated shape of the T 2 grains in Fig. 8(b) indicates that the T 2 phase precipitated out as a single phase from the liquid phase between reactions (6) and (7).
Finally, the (Mo ss + T 2 + Mo 2 C) eutectic reaction occurred at 1720 °C (inflection point e), which does not appear on the Mo-Si-B-TiC phase diagram. The order of the two ternary eutectic reactions (6) and (7) is judged from the relatively well-developed (Mo ss + T 2 + TiC) eutectic structure compared with the (Mo ss + T 2 + Mo 2 C) eutectic structure. Thus, the solidification is complete.
Miyamoto et al. 10   www.nature.com/scientificreports www.nature.com/scientificreports/ In summary, a method for ultra-high-temperature thermal analysis was developed. The maximum measurement error was 5 °C with a temperature range from 1084.62 to 1953 °C. By using this technique, the MoSiBTiC alloy (67.5Mo-5Si-10B-8.75Ti-8.75 C in mol%) was thermally analysed. EML experiments were also conducted to observe the in-situ solidification of the MoSiBTiC alloy melt, and to study the microstructure of the rapidly solidified sample. The solidification pathway was finally proposed as given in Table 2.

Methods
Ultra-high-temperature thermal analysis. Fixed-temperature points. The raw materials for the Ru-C alloy were Ru (99.9 mass%; particle diameter < 500 µm; JX Nippon Mining & Metals, Japan) and C (99.99 mass%; particle diameter <10 µm; Kojundo Chemical Laboratory, Japan) powders. Ru-C mixed powder with a composition of Ru-11 mol% C was filled into a graphite crucible having a blackbody cavity with a tape angle of 120°, and then heated and cooled in a radio-frequency furnace under an Ar atmosphere at an Ar flow rate of 10 mL/min. The apparent sample temperature T s was determined assuming ε app = 1 with a top pyrometer measuring the radiation from the blackbody cavity. Initially the Ru-C sample was kept at T s = T E − 10 °C to stabilise the furnace temperature, after which it was heated to T s = T E + (5, 10, 20) °C at a rate of 5 °C/min. After the sample was kept at T s = T E + 10 °C, it was cooled to T s = T E − (5, 10, 20) °C at a rate of 5 °C/min. This procedure was used for other fixed-point measurements with T m for Cu and T E for Ni-C. A Cu rod (99.99 mass%; diameter of 2 mm; length of 2 mm; Kojundo Chemical Laboratory, Japan) and Ni rod (99.99 mass%; diameter of 2 mm; length of 2 mm; Rare Metallic, Japan) were used as raw materials. The Ni-C alloy was prepared with the composition Ni-4 mol% C. The apparent emissivity ε app of the pyrometer was determined in advance using T m for Cu and T E for the Ni-C and Ru-C alloys.
MoSiBTiC alloy. The MoSiBTiC alloy was prepared by arc-melting from Mo (99.99 mass%; diameter of 10 mm; length of 300 mm; A.L.M.T.), Si (99.999 mass%; particle size of 2-5 mm; Kojundo Chemical Laboratory, Japan), B (99.5 mass%; particle size of 3-7 mm; Kojundo Chemical Laboratory, Japan) and TiC (99 mass%; particle size of 2-5 µm; Kojundo Chemical Laboratory, Japan) as raw materials. The chemical composition of the alloy was 67.5Mo-5Si-10B-8.75Ti-8.75 C (mol%). Here, T 2 is a ternary compound (Mo 5 SiB 2 ). A CaO-stabilised ZrO 2 crucible having a blackbody cavity was used to hold the MoSiBTiC alloy (about 87 g) instead of a graphite crucible to prevent carbon contamination. The alloy was heated and cooled in a radio-frequency furnace under an Ar atmosphere at an Ar flow rate of 10 mL/min. The sample was heated at a rate of 10 °C/min and kept around 2000 °C for 30 min to ensure homogeneous melt, and subsequently cooled at a rate of 10 °C/min. The sample temperature was determined during heating and cooling with a top pyrometer measuring the radiation from the blackbody cavity. The inflection points on the cooling curves were treated as the phase transformation temperatures of the MoSiBTiC alloy.
EML experiments on the MoSiBTiC alloy. The schematic of the EML apparatus is shown in Fig. 10.
After the MoSiBTiC sample was placed on the sample holder made of BN, the chamber was evacuated to 10 −2 Pa using a rotary pump and a turbomolecular pump. The chamber was filled with Ar-5 vol% H 2 gas after evacuation. Alternating electric current was applied to the levitation coil to levitate and heat the sample. The sample was melted with a heating laser (wavelength of 807 nm; power of 8-140 W; Jenoptik Laserdiode, Japan; NBT-S140mkII), and the sample temperature was controlled with the laser power. Ar-5 vol% H 2 gas was supplied at a rate of 3 L/min to prevent sample oxidation. A static magnetic field of 10 T was applied to the sample droplet to suppress both its translational motion and convection. For in-situ observation of solidification, the sample droplet was filmed with a CCD camera (MC1310; MIKROTRON GmbH, Germany) with a resolution of 512 × 512 pixels at 60 fps. The sample temperature was measured using a single-colour pyrometer (temperature range of 450-2500 °C; spectral range of 1.45-1.8 μm; IGA140/MB25; IMPAC Pyrometers; LumaSense