Thermophysical Properties of Advanced Ni‐Based Superalloys in the Liquid State Measured on Board the International Space Station

Advanced nickel‐based superalloys combine excellent high‐temperature mechanical strength, creep resistance, and toughness, and therefore find applications in next‐generation aircraft engines and turbines for land‐based power generators. The related fabrication processes are complex, time consuming, and costly, making it necessary to perform supporting computer simulations of the heat and material flow in the melt before and during crystallization. Such models are based on reliable thermophysical property data in the solid and liquid phase. However, measurements of surface‐ and volume‐dependent properties, such as liquid surface tension, viscosity, and specific heat of these complex liquid metal alloys, are very challenging, due to the melts high chemical reactivity at temperatures of interest. The method of choice is electromagnetic levitation, a containerless method. The measurements of surface tension, viscosity, mass density, and specific heat capacity, performed in the electromagnetic levitator (EML) on board the Space Laboratory Columbus in the International Space Station (ISS), are presented and discussed.


Introduction
Advanced nickel-based superalloys are a class of complex metallic materials with an exceptional combination of high-temperature strength, creep resistance, and toughness, as well as excellent corrosion and oxidation resistance. The materials' superior combination of mechanical properties result from the precipitation of an ordered cubic γ 0 -phase within the disordered cubic γ-phase [1][2][3] of nearly the same lattice constant.
Nickel-based superalloys are commonly used in turbine components for aircrafts and in land-based turbines for power generation. [4,5] Turbine blades are typically manufactured by complex casting procedures, such as directional solidification casting. [6] This way, polycrystalline morphologies with equiaxial grains or grains aligned along the length of the blade can be achieved. [7] Furthermore, for more enhanced performance, single crystalline turbine blades can be manufactured. [7,8] That way, a higher lifetime compared with blades with usual polycrystalline morphology can be attained. [6] Also, further heat treatments after casting are typically involved to achieve the final microstructure. [1,3,9] In addition to advances in alloy chemistry and manufacturing process, component geometries are becoming increasingly complex with multiple thin walls and very controlled interwall spacing to support advanced cooling methods. These complex geometries are extremely challenging to manufacture, so modeling and simulation of liquid metal filling of mold cavities and subsequent solidification is absolutely required. To enable such material and process simulation, accurate material properties and process boundary conditions are required. The results from this careful study of the thermophysical properties of nickelbased superalloys can be directly utilized for such simulation efforts.
Especially for the investment casting technique, a purely empirical methodology for the optimization of process conditions, such as process temperatures, mold withdrawal rate, and so on, is tedious and costly. Therefore, numerical computer simulations of turbine blade casting procedures are of great importance for the development of new casting designs. [8,[10][11][12] Good predictions of the heat flow and the mass flow patterns in the liquid melt are necessary for a successful casting simulation. [8,11] Therefore, precise knowledge of the thermophysical properties of the alloys melt is a prerequisite for meaningful numerical models of the casting procedure.
Measuring surface tension and viscosity of liquid metals is challenging, what is reflected in the large scatter of results. [13] Different methods for the measurement of viscosity of liquid metals exist. In most of them, such as the capillary viscosimeter or the oscillating vessel viscosimeter, the liquid is in direct contact to a part of the measurement device. A consequent shortcoming of such methods is the obvious difficulties while applying them to very reactive liquid metals.
The same is true for the measurement of specific heat capacity of reactive liquid metals in container-based methods. Reactions of the liquid with the container wall will lead to erroneous results.
For high-quality casting products that need high-dimensional precision, it is necessary to know the density/volume change of the liquid alloys with temperature. Moreover, porosity or cavities due to shrinkage are common casting defects.
Therefore, containerless, dynamic methods are most suitable for the measurement of surface tension, viscosity, and specific heat capacity and mass density of liquid metals. Surface contaminations, which can be inherently present in other methods, such as in the sessile drop method, can be eliminated in containerless methods. A well-suited containerless method for measurements of surface tension and viscosity is the oscillating drop method in an electromagnetic levitator (EML). [14][15][16][17] The containerless measurement of specific heat capacity and total hemispheric emissivity can be done by modulation calorimetry, which was demonstrated by Fecht and Wunderlich. [18][19][20][21][22][23] Also, the density can be obtained from the optical imaging of the sample diameter as a function of temperature.
Containerless processing is not the only prerequisite for precise measurements; also, the fluid flow inside the liquid sample has effects on the measurement results and solidification behavior. [24] The measurement of surface tension by the oscillating drop method under normal gravity conditions is complicated by the frequency split of the surface oscillations, even though the latter can be corrected by the Blackburn-Cummings correction. [25] Viscosity measurements by the oscillating drop method are not possible in ground-based electromagnetic levitation as the high positioning field that needs to be applied on ground leads to significant additional heating and turbulences in the liquid droplet. This makes it necessary to perform surface tension and viscosity measurements using the oscillating drop method under microgravity conditions. But most importantly, the absence of significant sample heating by the positioning field in the microgravity environment enables independent heating and positioning and, therefore, allows measurements in the undercooled liquid phase. Also, the precision of the modulation calorimetry and optical density measurements benefits from the very stable positioned sample.
Containerless processing using electromagnetic levitation under microgravity can be performed on board of parabolic flights. [26][27][28] But the limited time of microgravity (typically 20 s) per parabola limits the measurement precision, as well as the type of measurements that can be performed. The International Space Station (ISS), in contrast, offers a long duration of microgravity which enables the measurement cycles to be as long as 20 min (limited only by the duration of the signal link to the ISS). This enables the performance of modulation calorimetry cycles with a typical duration of 15-20 min.
In this study, thermophysical properties (surface tension, viscosity, electrical resistivity, mass density, specific heat capacity and, total hemispheric emissivity) of three Ni-based superalloys (LEK94, MC2, and CMSX-10) were investigated using the EML on board the European Science Laboratory Columbus in the ISS.
Ni-based superalloys are based on the precipitation strengthening of the Ni 3 Al-based γ 0 -phase particles in an face centered cubic Ni-based γ-matrix phase. [5] Additional alloying elements are introduced to improve the mechanical properties further, such as Ti, Ta, and Nb, which are understood to strengthen the γ 0 -phase. [4] Additions of C and Cr, W, Mo, Nb, Ta, and Ti can lead to the formation carbides, whereas B, together with Cr or Mo, promotes boride formation. [7] Further additions can be divided into elements that add to the γ 0 -phase and to those that add to the γ-phase. The elements Co, Cr, Mo, Ru, Rh, and W, having a similar atomic radius than Ni, prefer to be within the γ-matrix phase, and are used to stabilize the phase. [7] Elements such as Al, Ti, Nb, and Ta are added to promote the formation of the γ 0 -phase. [4,7] In this study, three alloy compositions were studied: LEK94, MC2, and CMSX-10. The specific alloy compositions are shown in Table 1. The composition was verified using energy dispersive X-ray spectroscopy (EDX) measurements. These measurements revealed also trace amounts of Y (<0.2 wt%) in MC2. For the LEK94 and CMSX-10 samples, the trace amounts of other elements were below 200 wt ppm. The oxygen content of the three alloy materials was about 10 wt ppm for LEK94 and CMSX-10 and about 50 wt ppm for MC2.
LEK94, a rhenium-containing single crystal superalloy with a low density between 8.1 and 8.3 g cm À3 , was developed by MTU Aero Engines, München, Germany. [29,30] The lightweight alloy LEK94 was first used in the series production of the low-pressure part of the GP7000 turbine for the Airbus A380. [31] The second-generation nickel-based single crystal superalloy MC2 was developed by ONERA in the late 1980s. The rhenium-free alloy is used for high-pressure turbine blades in turbo-engines of Safran Helicopter Engines for operating temperatures in the range from 750 to 1000 C. [32] CMSX-10 is a third-generation single-crystal alloy, developed by Cannon-Muskegon for a temperature range of 850-950 C, [33,34] and is, e.g., used in the Rolls-Royce engine TRENT 800. [34] The specific composition of the investigated CMSX-10 sample was chosen from Erickson. [34]  The LEK94 and MC2 master alloys were delivered by industrial partners. In both cases, spherical samples of 6.5 mm diameter were produced by arc melting and suction casting. The nominal composition was verified by EDX in a scanning electron microscope.
The composition of CMSX-10 was patented in very wide ranges. [33] Therefore, a specific composition from the literature [34] was used for the CMSX-10 samples. First, an ingot of CMSX-10 was produced from the elemental materials by arc melting. Subsequently, rods were produced by suction casting. Pieces were cut from the rod for EDX investigation and for suction casting of spheres of 6.5 mm diameter. The composition and its homogeneity were confirmed by the EDX results.

Electromagnetic Levitation Facility On Board the ISS
We limited the description of the EML on ISS to what was essential for this investigation. For further information, we refer to the literature. [35][36][37] The facility, developed and built by Airbus Defence and Space, is centered around a high-vacuum experiment chamber that can be operated under high vacuum or in inert gas atmospheres (argon or helium). Each sample is stored in its individual sample holder, of which up to 18 are contained in an exchangeable sample chamber with a dedicated sample transport system. For operation, the desired sample can be moved into the experiment chamber. The core of the levitator consisted of a coil system (SUPOS coil system [35] ) on which two radiofrequency (RF) generators are connected. One generator is used to produce a quadrupole field, imposing the positioning forces onto the sample. The second RF generator is used to establish a dipole field for heating the sample. The sample, being loosely confined in a wire cage, was placed within the coils during the experiment, leading to the free levitation of the sample. Using the special SUPOS coil system, the sample positioning was independent from sample heating. [35] In addition, this was achieved by preserving an undisturbed view on the sample. [35] As the positioning forces in a microgravity environment are much smaller compared with ground-based EML, sample heating by the positioning field can often be neglected and the sample heating was mainly controlled by the heating field. [36] In addition, the small positioning field resulted in a nearly perfect spherical sample in the liquid phase, when the RF heater was turned off for free cooling.
The oscillation of the liquid sample could be recorded by two high-speed cameras: one mounted axially and one radially, relative to the coil axis. The axial camera usually operates at a frame rate of 150 Hz, whereas the radial camera is typically operated at 200 or 400 Hz. Therefore, the used frame rates were at least a factor 3.5 higher than the surface oscillation frequency.
A pyrometer allowing monitoring the sample temperature in a range between 300 and 2100 C (resolution of <0.1 K above 600 C) was mounted axially (along the coils axis). Within an equivalent facility on ground, samples with the same dimensions as the flight sample were used to derive properties (such as the sample's electrical coupling coefficients to the RF fields and the pyrometer's emissivity setting) important for proper experiment planning. Readjustment of heater and positioner power settings as well as recalibration of the pyrometer's emissivity setting at the solidus temperature could be performed per telecommanding during flight operation if needed. [26] 2.3. Experimental Methods

Oscillating Drop Method
To measure the surface tension and viscosity of the samples under investigation, the oscillating drop technique [14,15] was used. A typical measurement cycle is shown in Figure 1a. It consisted of heating and melting of the solid sample, heating of the liquid, and a cooling phase (heater control voltage was off ). In the cooling phase, one or more short heating pulses were applied, to excite surface oscillations of the sample.
The oscillating drop technique is based on the observation of the surface oscillations of a levitating liquid droplet in vacuum or an inert gas atmosphere. [14,15] The frequencies f R of the fundamental oscillation modes of small amplitude oscillation of free, inviscid incompressible droplets in vacuum or a gas with negligible density and viscosity were derived by Rayleigh. [38] In the case of a microgravity environment, the liquid droplets equilibrium shape is a perfect sphere and the Rayleigh frequency f R that is related to the surface tension γ as [38] γ with M being the mass of the liquid sample. For experiments under normal 1 g conditions, the equilibrium shape of the droplet is not spherical and therefore exhibits a split in surface oscillation frequencies. [15] The surface oscillations were measured by edge detection in the high-speed camera recordings. Subsequently, several measures of the surface oscillations amplitude were determined (X-radius, Y-radius, area, etc.). [39] Figure 1b shows a typical measurement of the time-dependent sample Y-radius, together with the discrete Fourier transformation of the marked time frame (Figure 1c). The time signal was high-pass filtered with a cutoff frequency of 20 Hz. In Figure 1c, a single frequency peak without detectable peak splitting was observed in the frequency spectrum. This is expected due to the stable microgravity environment. The center frequency could be determined with a high accuracy using a Lorentzian fitting curve.
To obtain the surface oscillation frequency f R , as well as the damping time constant τ of the oscillations, an edge detection algorithm was applied to the high-speed videos, revealing the edge of the sample. From this, time-dependent data for different radii and other measures of surfaces oscillations (visible area, etc.) were extracted. [39] These oscillation signals were used to perform a series of discrete Fourier transformations on time windows of 1.5 s, which were moved in steps of 0.75 s along the time axis. For each spectrum, a Lorentzian curve was used to fit the peak. The obtained center frequencies were then stored together with their www.advancedsciencenews.com www.aem-journal.com corresponding sample temperature values obtained by the pyrometer. Finally, the temperature-dependent surface tension values were calculated according to Equation (1). The damping of the surface oscillations was characterized by a damping time constant τ, being the time within which the oscillation amplitude decayed to 1/e of the original amplitude. As such, the time dependence of the oscillation amplitude A(t) could be described as where A 0 denotes the initial oscillation amplitude at the time t 0 . Lamb described the damping of the surface oscillations of free oscillating droplets with small viscosities (1/τ << f R ). [40] If the damping is small due to a small viscosity, the Rayleigh frequency remains unchanged. [14,[41][42][43] Therefore, for small amplitude oscillations, as in our investigations, typically only the exponential decay of the Y 2,m modes was observable. Knowing the damping time constant τ, the viscosity η could be calculated by [14,36] where M is the sample mass and R 0 is the equilibrium radius of the spherical sample. However, this is only true for force-free samples with only laminar flow in the droplet. That allows viscosity measurements with the oscillating drop technique only under microgravity conditions. Due to the relative small investigated temperature range, the Arrhenius model was assumed as temperature dependence. Therefore, the viscosity could be expressed as www.advancedsciencenews.com www.aem-journal.com where η 0 is the viscosity at infinite temperature, k B is the Boltzmann constant, and E A is the activation energy. Using an Arrhenius plot, the activation energy and the viscosity of infinite temperature η 0 were determined. The single data points for surface tension could be obtained with an accuracy of about 0.6% (oscillation frequency obtained with 0.25% accuracy, sample mass accuracy 0.1%). Additional influences that altered the measured values could be effects of rotation [44] or oscillation amplitude. [45] We did not observe any strong sample rotations and the data were only evaluated for relative sample oscillations below 1.5% to avoid nonlinear effects. [45] The accuracy of the measured oscillation damping times was about 1% and the uncertainties of sample radius (0.8%) and sample mass (0.1%) led to the measurement uncertainty of the viscosity of about 1.9%. The uncertainties for surface tension and viscosity shown in Table 2 are, however, the uncertainties obtained from fitting. Here, it should be mentioned that the relatively small temperature interval of fitting led to values of η 0 and E A for the viscosity that described the measured temperature interval but did not have physical meaning.

Modulation Calorimetry
The measurement of specific heat capacity by the modulation calorimetry is based on the modulation of the heating power and the observation of the temperature response. [18][19][20][21][22][23] The preferred way was the sinusoidal modulation of the heating power, which could be achieved by modulation of the heating current I H such that I H ¼ (I H0 2 þ I Hω 2 sin(ωt þ φ 0 )) 1/2 . This resulted in an amplitude of the power modulation which was given by where G H is a coupling coefficient which describes the inductive coupling between the conducting sample and the RF heating generator oscillating circuit. It depends on the coil geometry, the resistivity and radius of the sample, as well as the applied RF frequency ω H . The coupling coefficient was determined applying an analytical solution derived by Fromm and Jehn, [46] which assumed coil geometries with rotational symmetry and coils with infinitesimally thin wires. The amplitude of the temperature modulation could be described as [18,22] ΔT ¼ ΔP where the heat capacity of the sample is C p and the function f(ω,τ 1 ,τ 2 ) is a correction function taking the effects of external and internal heat loss into account. For the case that ω is chosen such that the system is in the adiabatic regime, f(ω,τ 1 ,τ 2 ) ¼ 1. The external relaxation time τ 1 , characterizing the heat loss to the external heat sink (heat radiation, and in case of gas atmosphere also heat conduction and convection) can be determined experimentally from the temperature transient following a step-like change in the heating power. The relaxation time τ 2 describes the internal heat transfer from the directly heated equator of the sample to the sample poles (where the temperature is measured). [20] The external relaxation time τ 1 can be used to determine the total hemispherical emissivity. In vacuum, with only the heat radiation as heat-loss mechanism, it holds [20] where A surface is the surface area of the sample, σ B is the Stefan-Boltzmann constant, and ε is the total hemispherical emissivity. When combined with an appropriate heat-loss model in a gas atmosphere, [47] the total hemispherical emissivity can also be determined for cycles performed in a gas atmosphere. A typical measurement cycle is shown in Figure 2, where the sample gets melted, overheated, and subsequently heated, while the heating power is modulated at four different average power levels and therefore the measurement is performed at four different average sample temperatures. It is worth noting that the  www.advancedsciencenews.com www.aem-journal.com sample shown here undercools more than 200 K, demonstrating the cleanness of the sample and the measurement environment.

Sample Coupling Electronics
The specific resistivity of the liquid is a quantity of practical importance for industrial processes that, e.g., involve inductive melting. And therefore, the EML got upgraded with a sample coupling electronics (SCE), [48] allowing the precise determination of the sample resistivity and radius in the solid and liquid phase. The method allows the precise measurement of the complex impedance of the oscillating circuit. In performing a measurement without a sample, the generators/oscillators circuit parameters (inductance L, capacity C, series resistance R, and their temperature dependences) could be obtained. By the measurement with a sample, the change in circuit impedance was used to obtain the specific resistivity and radius of the sample. The measurement of a known sample with known resistivity could be used to find the calibration constant to obtain absolute values for specific resistivity and sample radius. Further details can be found in the literature. [48] Furthermore, one could also calibrate the measurement by comparing the sample density (calculated from the sample mass and sample radius obtained by the SCE measurement) with the optically obtained sample density. In this study, the resistivity measurement was only performed for the LEK94 sample. It was shown that the measurement method could obtain a measurement accuracy of about 0.05%. [48]

Optical Measurement of Density
Mass density of the liquid alloys was analyzed by optical means. In equilibrium, due to microgravity conditions, the sample was a perfect sphere. As described earlier, the shape of the liquid droplet was recorded by axial and radial high-speed cameras. Using an edge-detection algorithm, [39] the circumference of the sphere was detected and fitted by an ellipse. The projected area of the sample A sample was then calculated from the fitting. To neglect the small surface oscillations that were present after the heater was turned off, the time signal of the sample area A sample was low-pass filtered with a cutoff frequency of 3 Hz. The sample volume was then calculated (assuming perfect spherical symmetry) by The calibration of the image scale (mm per pixel) was done by the measurement of a pure, solid zirconium sample with precisely known diameter and the comparison with the literature data on thermal expansion and density change of Zr. [49] Knowing the mass of the samples, one obtains the mass density ρ ¼ M/V at different temperatures.
The measurement uncertainty of the optical method used here was influenced by the used edge-finding algorithm, which did not operate on a sub-pixel level, hence introducing an error of AE0.5% on the sample radius. Together with the mass inaccuracy of AE0.1%, this led to an inaccuracy of about 1.6% for the optically determined density. The inductive method, however, was not prone to the limitations of the optical system and had therefore a much higher accuracy. [48]

Results
The liquidus temperatures of the samples LEK94, MC2, and CMSX-10, measured using a high-temperature DSC, are shown in Table 2.

Surface Tension and Viscosity
The linear temperature dependence of the surface tension γ(T ) can be expressed as In Figure 3a, the surface tension as a function of temperature for LEK94 is shown. Three cycles were combined for the shown data, using recordings of the axial and radial high-speed camera. Figure 3b shows the temperature dependence of the surface tension for MC2, and Figure 3c shows the surface tension of CMSX-10.
In all three cases, the samples show a negative temperature coefficient. The obtained values for the surface tension at the liquidus temperature γ 0 and the temperature coefficients dγ/dT are shown in Table 2. Figure 3d shows the viscosity obtained for LEK94. The viscosity at the liquidus temperature was found to be 8.4 mPa s. The viscosity of MC2 is shown in Figure 3e; the value at liquidus temperature is 12 mPa s. Figure 3f shows the viscosity values for CMSX-10 as a function of temperature. The viscosity at the liquidus temperature is 8.6 mPa s. An overview over all determined Arrhenius parameters is shown in Table 2.

Specific Heat Capacity and Total Hemispherical Emissivity
The specific heat capacity of LEK94 was obtained in a temperature range from 1550 to 1800 K, where measurements at about 100 K undercooling were possible. The obtained values are shown in Figure 4a. The measured phase shift between power and temperature confirmed that the experiment took place in the adiabatic regime.
The given heat capacities of MC2 and CMSX-10 were obtained the same way; however, for the electrical resistivity and sample radius, the same values as for LEK94 were taken, as no measurements of electrical resistivity of these samples were performed. However, the relative influence of electrical resistivity on the sample coupling to the RF field and therefore the influence on the calculated specific heat capacity is about 10 À3 (μΩ cm) À1 , meaning that a specific resistivity that would be 10 μΩ·cm higher would increase the c p by a factor 10 À2 . As the specific resistivity of Ni-based superalloys will be with high certainty between 100 and 200 μΩ cm (comparing with resistivities of other elemental materials in the liquid phase [50] ), we add this as uncertainty to the determined specific heat value accordingly.
A higher influence on the coupling and therefore the power dissipation in the sample has the sample radius, which is in www.advancedsciencenews.com www.aem-journal.com all three samples known from electronic measurements (LEK94) or from the optical assessment of the sample mass density (LEK94, MC2, and CMSX-10). Other influences on the measurement uncertainty are the finite accuracy of the temperature measurement, which is 0.1 K, which translates to an accuracy of about 1% accuracy of determination of the temperature amplitude.
In all investigated samples, the specific heat capacity does not have a significant temperature dependence, and the average values for specific heat capacities of the three alloys are shown in Table 2, together with the standard deviation. Figure 4b shows the total hemispherical emissivity determined for the three alloy samples in the stable and undercooled liquid phase. Within the measurement accuracy, the total  hemispherical emissivity is about 0.25 for all three samples. The increase in emissivity for CMSX-10 above 1725 K cannot be explained at present. It could be due to changes in local structure in the liquid, which could lead to different charge carrier scattering and therefore emissivity behavior. Also, the influence of surface-active species cannot be ruled out; however, this is not supported by the surface-tension measurements.

Specific Electrical Resistivity
The specific resistivity of LEK94 obtained from the dedicated measurement cycles is found to be 138 μΩ·cm in the stable and undercooled liquid phase, without significant temperature dependence.

Mass Density
The mass density as a function of temperature for the three alloys under investigation is shown in Figure 5. The three lines show the density data obtained by the optical method, together with the uncertainty interval. The data points obtained by the inductive method for LEK94, having a much higher accuracy, [48] are shown for comparison. Five cycles were analyzed for each sample. The temperaturedependent mass density of LEK94 can be described as ρðTÞ ¼ ð7.57 AE 0.15Þ g cm À3 À ð6.31 AE 0.5Þ Â 10 À4 g cm À3 K À1 ðT À 1653 KÞ (10) for MC2 can be expressed as ρðTÞ ¼ ð7.91 AE 0.15Þ g cm À3 À ð6.58 AE 0.5Þ Â 10 À4 g cm À3 K À1 ðT À 1668 KÞ (11) and for CMSX-10, the mass density can be written as ρðTÞ ¼ ð8.21 AE 0.15Þ g cm À3 À ð5.35 AE 0.5Þ Â 10 À4 g cm À3 K À1 ðT À 1681 KÞ (12) For all three liquid alloys, the mass density shows a negative temperature coefficient. The values of the mass density at the liquidus temperature ρ 0 and the temperature coefficient dρ/dT are also shown in Table 2. The mass density for LEK94 derived from the inductive measurement of sample conductivity and projected radius is also shown in Figure 5.

Discussion
Surface tension data for LEK94, MC2, and CMSX-10 were measured by Amore et al. [51] in the pinned drop method, and the surface tension of LEK94 and CMSX-10 was measured by Wunderlich et al. [26] using the oscillating drop method on board a parabolic flight. The surface tension of LEK94, MC2, and CMSX-10 measured in the EML on ISS is lower than the values obtained by earlier studies. [26,51] Furthermore, the temperature dependency of the surface tension dγ/dT of LEK94 and CMSX-10 is smaller than that from the measurements aboard the parabolic flight [26] and on ground. [51] The compared measurement methods have different possible systematic error sources. Although ground-based measurements could be influenced by segregation of heavy elements and contamination effects, the lower dγ/dT measured in our experiment, compared with the results from parabolic flights, could also point toward an influence by oxygen adsorption. The returned samples do not show any obvious signs of oxide formation on the surface; however, an in-depth analysis is still ongoing. Overall, it can be recognized that the surface tension increases for an increasing ratio of Ni at%/(Al at% þ Cr at%).
The viscosity of LEK94, determined in the present investigation, compares well with the values determined by electromagnetic levitation on board a parabolic flight. [26] Even though the viscosity values of CMSX-10 obtained in the undercooled liquid phase by Wunderlich et al. [26] are similar to those obtained in this study, the viscosity in the stable liquid phase is somewhat higher in this study. However, the technologically relevant viscosities at the liquidus temperature, obtained by Wunderlich et al. [26] and in this study, are within error bars identical for CMSX-10 and LEK94.
A first-order approach to estimate theoretically the viscosity of Ni-based superalloys is the reduction of complexity by considering only the most abundant elements, Ni and Al. The results also indicate that the viscosity is generally increasing for an increasing Ni at%/Al at% ratio, which is also in agreement with measurements of Sato et al. [52] Mills et al. [53] developed an empirical expression to predict temperature-dependent viscosity of Ni-based superalloys only using the mass% of Fe, Cr, and the heavy elements (W, Re, Nb, Ta, Mo, Hf ). The predictions following Mills et al. [53] for the viscosity of LEK94, MC2, and CMSX-10 are shown in Figure 3d-f. It is apparent that for LEK94 and MC2, the compositions with relatively small amounts of heavy elements, the temperature dependence of viscosity is not predicted correctly. However, the prediction of viscosity of CMSX-10 is only about 4%, which is too low, using Mills equation. Therefore, the model might underestimate the influence of the Al concentration or of the refractory elements on the viscosity.
The densities of LEK94, MC2, and CMSX-10 in the liquid phase were also measured by Amore et al. [51] in the containerbased pinned drop method on ground. Although the measured www.advancedsciencenews.com www.aem-journal.com densities of LEK94 and MC2 are close to the values obtained in previous study [51] (þ5%), the density of CMSX-10 is 12% higher than the value obtained by Amore et al. However, the density of CMSX-10 obtained by Quested et al., [54] who investigated a similar composition than ours, is in much closer agreement with the values obtained in our study. The density of the Ni-Al system was investigated by Plevachuk et al., [55] and the temperature coefficients of the density of our measured alloys are in agreement with the values obtained in previous study. [55] The Ni/Al ratio is important for the description of the density because their preferential bonding and reduction of molar volume compared with the ideal mixing will reduce the density. [51] In contrast, the weight percentage of heavy elements has a strong influence on the density of the alloy. Here, it is apparent that the overall density is mainly dominated by the amount of heavy elements, which is highest for CMSX-10 and lowest for LEK94. The specific heat capacity of liquid LEK94 (c p ¼ 0.749 AE 0.006 J (g K) À1 , between 1550 and 1850 K) corresponds well to the value obtained by Amore et al. [51] in a ground-based differential scanning calorimeter of 0.75 J(g K) À1 , with an estimated error of AE10% in the temperature range of 1650-1750 K. The value that is obtained using the Kopp-Neumann rule (as suggested by Mills et al. [53] ) is 0.657 J (g K) À1 , which is 12% lower than the measured values. The specific heat capacities of MC2 and CMSX-10 obtained in this work are comparable to the 0.7 J (g K) À1 obtained by Amore et al. [51] for both compositions. The specific heat capacity of MC2 (0.67 AE 0.05 J(g K) À1 ) is only 5% higher than the prediction by Kopp-Neumann rule (0.635 J(g K) À1 ). The specific heat capacity of CMSX-10 is given as 0.63 J (g K) À1 without measurement uncertainty by Quested et al. [54] , which is only 11% below the value of (0.71 AE 0.05) J (g K) À1 determined in our study. The Kopp-Neumann rule predicts a value of 0.609 J (g K) À1 , which is 3% lower than our measured value.
From the obtained surface tension data in this and other studies, such as in refs. [26,51] the compound formation model by Novakovic and co-workers [51,56] predicts the surface tension of Ni-based superalloys satisfyingly well. The viscosity predictions by the equation of Mills et al. [53] only explain the viscosity of CMSX-10 (however, not precisely), but cannot predict the viscosity of LEK94 and MC2.
The specific heat capacities obtained in the containerless processing method applied here can be obtained with higher accuracy than in container-based ground experiments, but more importantly, the absence of container reactions reduces systematic measurement uncertainties. The simple Kopp-Neumann rule, suggested by Mills et al., [53] does not predict the measured values for the investigated alloys. This proofs that a better model must be found to predict the specific heat capacity of liquid Ni-based superalloys.

Conclusions
The EML on ISS was successfully used to obtain thermophysical properties of three Ni-based superalloys in the stable and undercooled liquid phase. This data can assist modeling of casting and other process simulations. Measured thermophysical properties of superalloys and potentially other alloy systems in the future can not only lead to both enhanced material and process modeling but also guide approaches for new materials and process development. In addition, detailed experiments to measure complex material properties will lead to validation of firstprinciples-type modeling to predict such properties, which is significantly challenging today.
The presented precise measurement of thermophysical properties of Ni-based superalloys is an example of deficiencies of existing empirical models based on the alloy composition. However, such models will be an important cornerstone for fast and easy development of casting processes for new Ni-based superalloy compositions.
Altogether this analysis illustrates the importance of precise measurements to improve predictive models for thermophysical properties of liquid alloy.
The measured thermophysical properties are of great use in direct application to melting and solidification modeling processes. Commercial computational codes rely on various critical thermophysical properties to emulate specific processes and related physical mechanisms.
The results from the space and microgravity experiments have been of great use for the application of casting modeling and would be of similar great use in additive manufacturing modeling and simulation. Material property data of particular interest to the materials processing and aerospace industries would include surface (i.e., liquid/atmosphere interfacial) energies and viscosities as a function of temperature (at casting relevant temperatures).
These are critical values for mold filling and other liquid-phase casting simulations, which are used to diagnose and remediate process defects such as mold erosion and nonfill. These properties may also support ongoing efforts related to mold material design and casting core defect analysis for advanced castings. Highquality data for these properties are scarce, and microgravity measurements and analysis are uniquely placed to provide such data.
Surface energies and viscosity measurements as a function of temperature would be directly applicable to additive manufacturing activities. Liquid melt pool (bead) simulation and prediction is critical for design and analysis of additive manufacturing processes. Similarly, surface energy as a function of temperature is the driving force for Marangoni flow in additive manufacturing (and welding) moving melt pools. Having accurate data of this type would be applied to additive manufacture design and process optimization.
The results from this program will be further used to assess first-principle models to directly predict these thermophysical properties. The ability to generate high-quality, highly pedigreed physical data to validate modeling methods is critical for the goal to further establish model-based material definitions and descriptions. Future experiments that can further assess thermophysical properties as a function of chemistry would greatly support the development and implementation of alloy design tools for next-generation materials.
the ESA MAP project ThermoProp (AO-099-022). The authors further acknowledge funding from the DLR Space Administration with funds provided by the Federal Ministry for Economic Affairs and Energy (BMWi) under grant no. 50WM1759. Furthermore, the authors gratefully acknowledge the support by the Microgravity User Support Center (MUSC) at the Institute of Materials Physics in Space of DLR, Cologne, during preparation and conduction of the experiments. The continuous technical support by Airbus Defence and Space concerning the work at and around the EML is acknowledged. The authors also acknowledge MTU Aero Engines for support and providing material samples. Also, the continuous support by Zollern GmbH & Co. KG is acknowledged.