Surface tension and viscosity of liquid Pd43Cu27Ni10P20 measured in a levitation device under microgravity

ARTICLE

OPEN

Surface tension and viscosity of liquid Pd

measured in a levitation device under microgravity

Markus Mohr

, Rainer K. Wunderlich

, Kai Zweiacker

, Silke Prades-Rödel

, Romuald Sauget

, Andreas Blatter

, Roland Logé

Alex Dommann

, Antonia Neels

, William L. Johnson

and Hans-Jörg Fecht

Here we present measurements of surface tension and viscosity of the bulk glass-forming alloy Pd

performed during

containerless processing under reduced gravity. We applied the oscillating drop method in an electromagnetic levitation facility on

board of parabolic

fl

ights. The measured viscosity exhibits a pronounced temperature dependence following an Arrhenius law over

a temperature range from 1100 K to 1450 K. Together with literature values of viscosity at lower temperatures, the viscosity of

can be well described by a free volume model. X-ray diffraction analysis on the material retrieved after the

parabolic

fl

ights con

fi

rm the glassy nature after vitri

fi

cation of the bulk samples and thus the absence of crystallization during

processing over a wide temperature range.

npj Microgravity

(2019) 5:4 ; https://doi.org/10.1038/s41526-019-0065-4

INTRODUCTION

Bulk metallic glasses (BMGs) represent a new development in

materials science with the major advantage to possess superior

mechanical properties (and others) compared with materials in

their conventional crystalline state. Metallic glasses, solid metallic

materials with a disordered liquid-like atomic-scale structure, are

formed when they get cooled faster from the liquid state than a

critical cooling rate. The higher the glass-forming ability (GFA) of a

metallic glass, the lower is the critical cooling rate and the better it

is suitable for industrial applications. Some of the most robust

BMG alloys in terms of critical cooling rates and oxidation/

corrosion resistance are based on precious metal

–

metalloid alloy

systems, such as Pd-, Pt-, and Au-based alloys, combined with

typically 20 at% of phosphorous. Within this material class, the

metallic glasses composed of Pd-Cu-Ni-P and Pd-Ni-P have an

outstanding GFA, re

fl

ected by their very high reduced glass

temperature.

–

During cooling of a liquid, the increasing

thermodynamic driving force for crystallization and the reducing

atomic kinetics are competing.

The formation of a glass during

cooling of a liquid demands the bypass of the nose of the so-

called temperature-time transformation (TTT) diagram, describing

the time in isothermal conditions, after which considerable crystal

nucleation occurs.

The strong decrease in atomic kinetics during

cooling is one important factor in order to obtain BMGs during

industrial processing, such as casting or injection molding. The

strong increase of viscosity during cool down, and the demand to

achieve vitri

fi

cation of the liquid also establishes boundary

conditions for the right choice of process parameters for industrial

production procedures. Also, for superplastic forming technolo-

gies, precise knowledge of the temperature-dependent viscosity

of the alloy is of importance.

Thus, it is important to provide basic thermophysical property

data over a wide temperature range to design production

processes and models for supporting process simulations.

The precise measurement of thermophysical properties such as

surface tension and viscosity of metallic alloys in their liquid phase

(at high temperatures) demand clean conditions, especially the

absence of foreign materials that could contaminate the surface or

bulk of the measured liquid sample. This makes processing under

ultra-high vacuum or inert gas mandatory. Additionally, the high

reactivity of typical metallic melts makes containerless methods

necessary for many metallic alloys. A very versatile containerless

processing method that offers wide applicability to electrically

conductive samples is electromagnetic levitation (EML).

–

This

method enables the determination of surface tension and

viscosity by the oscillating drop method,

where surface

oscillations of the liquid sample are excited, observed, and

analyzed.

However, under earth

’

s gravitational conditions, a liquid sample

in its natural geometry or levitated by an electromagnetic

positioning

fi

eld will be considerably deformed. For levitated

drops, this leads to a split of oscillating frequencies.

addition, simultaneous temperature and levitation control is

limited under normal 1

gravitational conditions, since the

positioning

fi

eld required for lifting the samples may already heat

the sample to signi

fi

cant temperatures, even beyond the melting

point (especially true for low melting BMGs). Furthermore, the

fl

uid

fl

ow in the constantly heated, deformed droplet under

terrestrial conditions is not well controlled (laminar to turbulent

transition), which makes it necessary to perform the experiments

in reduced gravity conditions (microgravity, μ

Received: 6 December 2018 Accepted: 5 February 2019

Institute of Functional Nanosystems FNS, Ulm University, Albert-Einstein-Allee 47, 89081 Ulm, Germany;

Center for X-ray Analytics, Empa Swiss Federal Laboratories for

Materials Science and Technology, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland;

PX Services SA, Boulevard des Eplatures 42, 2304 La Chaux-De-Fonds, Switzerland;

Thermomechanical Metallurgy Laboratory

–

PX Group Chair, Ecole Polytechnique Fédérale de Lausanne (EPFL), Neuchâtel, Switzerland and

California Institute of Technology,

1200 East California Boulevard, Pasadena, CA, USA

Correspondence: Markus Mohr (markus.mohr@uni-ulm.de)

www.nature.com/npjmgrav

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

One possibility to reach μ

for a short period of time (10

–

20 s)

are parabolic

fl

ights, such as those performed by Novespace using

an Airbus A310.

The presented results were collected during two parabolic

fl

ight

campaigns in year 2016 and 2017, employing the EML facility

TEMPUS (

’

Tiegelfreies elektromagnetisches Prozessieren unter

Schwerelosigkeit

’

, engl.

‘

Containerless electromagnetic processing

under weightlessness

’

We present the surface tension and

viscosity of Pd

in the liquid phase and show that the

temperature dependence of the viscosity of Pd

can

be well described by a free volume model. Furthermore, we

con

fi

rm the absence of long-range order, characteristic for BMGs,

using X-ray diffraction (XRD). The samples morphology is analyzed

by X-ray computed tomography (CT) before and after the

fl

ights.

RESULTS

The photograph in Fig.

a gives an impression of the TEMPUS

facility on board the parabolic

fl

ight airplane. The samples are

contained in a sample chamber during

fl

ight, and the desired

sample is brought to the experiment chamber for processing in

several subsequent parabolas. A coil system connected to two rf-

generators is used to position and heat the sample independently.

The sample temperature is measured using a pyrometer, while

two high-speed cameras observe the sample during processing.

Typical sample diameters are between 6 and 7 mm. Figure

shows photographs of the PdCuNiP sample before and after the

parabolic

fl

ight in 2017.

In Fig.

c, a representative temperature-time pro

fi

le in the μ

phase during processing of Pd

is shown. The sample

temperature is shown in red in the

fi

rst diagram, while the control

voltage of the rf-heater with pulses for the excitation of surface

oscillations is shown in the second diagram in blue. Additionally,

the level of vertical acceleration is shown in green below. In every

parabola, after positioning the sample (see I in Fig.

c), the heater

is turned on to melt the sample (see II in Fig.

c) and overheat the

liquid melt by about 450 K (see III in Fig.

c). Afterwards, the heater

is turned off (IV in Fig.

c), giving the sample the opportunity to

cool down by heat radiation and heat conduction in the inert gas

atmosphere until the end of the μ

phase (see IV and V in Fig.

c).

A series of short heat pulses is applied in order to initiate

surface oscillations of the liquid droplet. The excitation of surface

oscillations is detected by two different methods.

The surface oscillations, effectively being a modulation of the

sample diameter, also modulate the apparent impedance of the

heater coil circuitry. This way, the sample oscillations are detected

inductively by an electronic measurement equipment, the so-

called sample coupling electronics (SCEs). Figure

d shows the

high-pass

fi

ltered apparent electrical sample resistivity determined

from the modulated impedance of the heater circuit. As seen in

Fig.

d, after turning off the heater after the initial melting step,

the sample radius oscillates with an exponentially decaying

amplitude. The time constant varies as a function of temperature

and is related to the viscosity, as described below in more detail.

The same happens also after every heater pulse.

The second approach to detect the surface oscillations utilizes

high-speed videos (at typical frame rates between 150 and 200 Hz)

that are taken from the sample during the cooling period. Figure

e shows six frames recorded after the time

, when the heater

pulse was turned off. The initial deformation is shown in the frame

taken at

, and evidently the sample starts to oscillate thereafter.

An edge detection algorithm is used to obtain several deforma-

tion measures, such as the

and

radius of the sample, shown in

Fig.

e, from which the oscillation amplitudes can be derived.

In all parabolas, at the end of the μ

phase, the sample touched

the sample holder pedestal while still in the liquid phase. This also

gives rise to the non-spherical shape of the samples observed

after processing in 2017 (see Fig.

b) and in 2016 (see Fig.

). No

sign of solidi

fi

cation such as recalescence was observed during

levitation in the μ

phase before. During all processing cycles, no

sign of surface precipitates could be observed on either sample.

No visible signs of surface contamination could be observed when

the sample was taken out of the TEMPUS facility after processing.

Surface tension and viscosity

The surface tension

can be deduced from the samples surface

oscillation frequency

. In order to obtain the surface oscillation

frequency from the optically and inductively measured amplitudes

(see Figs.

a, c), Fourier spectra are calculated through a dedicated

discrete Fourier transformation (DFT) algorithm. These spectra are

used to obtain the surface oscillation frequency at different

temperatures, by

fi

tting a Lorentzian function (see Figs.

b, d for

Fourier spectra of the optically and inductively obtained signals).

The surface tension

, the sample mass

, and the oscillation

frequency

are generally related by

ν

(1)

and hence, the oscillation frequency at different temperatures

during the sample cooling period can be determined (see IV in Fig.

c).

Due to internal friction in the liquid state, the surface

oscillations are considerably damped, exhibiting an exponentially

decreasing oscillation amplitude, according to

(

)

=

exp(

-t/

While Fig.

d shows a general overview of six electromagnetic

pulses in a time window of ca. 10 s., Figs.

a, c show details of

damped oscillations after one single heater pulse. The damping

time constant

of the surface oscillations can be obtained by

proper

fi

tting of the signal envelope. The viscosity of the liquid

can be obtained by

(2)

where

is the averaged radius of the sample. See Methods section

for further details.

The inductive method is less prone, but not immune to effects

of sample translation, rotation, and precession. As a consequence,

most of the sample oscillation analysis was performed with the

inductive method.

In Fig.

a, the surface tension as a function of temperature is

shown as determined in the parabolic

fl

ight in 2016 and 2017. The

fi

rst campaign covered a temperature range of 1050 K

–

1400 K,

whereas in the second one, a temperature range of 1350 K

–

1850 K

was investigated.

The surface tension values in the lower temperature range show

a positive temperature coef

fi

cient. Reduction of surface tension is

often observed by surface-active species, while, less of them

would be adsorbed on the surface at higher temperatures. The

same effect could be due to a temperature-dependent surface

segregation of surface-active elements of the alloy, such as

phosphorus. The values of the higher temperature range can be

well represented by a linear temperature dependence having a

negative temperature coef

fi

cient according to:

ðÞ¼

53 ± 0

ðÞ

28 ± 0

ðÞ

́

́

827 K

ðÞ

(3)

The formula represents an average of the data obtained from

four parabolas. The extrapolation of the obtained surface tension

data to the liquidus temperature gives

(

liq

)

=

(1.53 ± 0.10) N/m.

In Fig.

b, the viscosity of Pd

is shown in an

Arrhenius plot. The values comprise data obtained from the

inductive and optical method in the parabolic

fl

ight campaign

2016. The scatter of the viscosity values shown in Fig.

b can partly

be understood by the quality of the measured data. Sample

movement and the related distortion of the sample edges

M. Mohr et al.

2

npj Microgravity (2019) 4

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

1234567890():,;

contribute to disturbances of the optical data. Also, slow

modulations of the apparent electrical conductivity or the sample

radius can be present due to the precession of the sample

or possibly due to mode jumps between the degenerated

2,m

modes. These phenomena are limiting the precision of viscosity

determination.

At the high temperatures (1100 K

–

1450 K) far above the glass

transition temperature, the data can be satisfactorily represented

by an Arrhenius dependence:

(

)

=

× exp(

) with

=

(18.2 ± 9) μPa s and

=

(0.77 ± 0.07) eV. This is evident from the

good linear

fi

t shown in Fig.

In the higher temperature range, investigated in parabolic

fl

ight

campaign 2017, only one data point at 1420 K could be analyzed,

for the higher temperatures, the sample oscillation was not taking

place in a single mode or the sample was performing movements,

which obscured the exponential damping of the surface

oscillations.

Structural and chemical analysis

Samples were weighted before and after they were processed in

the TEMPUS sample chamber. The mass was determined as

1.3387 g (precision ±0.0004 g) and afterwards as 1.3385 g. There-

fore, the mass loss can be estimated to be lower than the

precision of the balance (~ 0.4 mg). The sample was heated six

times to 1100

–

1200 °C. The time for which the temperature was

above 900 °C was about 7 s in each parabola. The evaporation rate

is therefore below 10 μg/s in the temperature range between

900 °C and 1200 °C.

The energy dispersive X-ray spectroscopy (EDX) analysis of as-

cast and processed samples shows homogeneous mixture of all

Fig. 1

Photograph of the TEMPUS facility on board the parabolic

fl

ight airplane.

Photographs of the samples before and after the

processing in the parabolic

fl

ight 2017 are shown. The deformation of the sample after the

fl

ight is due to the contact of the sample with the

sample holder at the end of the parabola.

Temperature-time pro

fi

le of processing in the electromagnetic levitator on board a parabolic

fl

ight

(red). The control voltage of the rf-heater (blue) shows pulses for the excitation of surface oscillations. The level of vertical acceleration (gree

shows the ~20-s time window of μ

Variation of the high-pass

fi

ltered

“

apparent

”

electrical resistivity as a function of time

—

the

exponentially decaying surface oscillations can be detected after heater turn-off and after every heater pulse.

Series of frames recorded by

the high-speed camera, showing the surface oscillations of the droplet

M. Mohr et al.

3

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

npj Microgravity (2019) 4

constituents of the alloy. Within the measurement uncertainty of

the EDX analysis, no change of sample composition was

measurable, which is expected, considering the very low mass

loss during processing.

Furthermore, the samples returned from the parabolic

fl

ights

were analyzed using XRD and CT methods in order to elaborate

the homogeneity of the solidi

fi

ed microstructure and to deter-

mine the presence or absence of voids.

During the processing cycles in the TEMPUS, the sample was

repeatedly molten and vitri

fi

ed. By the forced gas cooling with

–

Ar gas mixture and the

fi

nal contact of the sample with the

sample holder at the end of the μ

cycle, the cooling rate was

higher than the critical cooling rate of Pd

, thus

avoiding crystallization. Typically, the phase change from liquid to

solid is concurrent to a reduction in volume. This volume

reduction can cause small pores and voids within the sample.

Figure

shows an absorption X-ray CT image and the volumetric

reconstruction of the melted and subsequent solidi

fi

ed Pd-alloy

sample. Absorption-based X-ray CT is a powerful nondestructive

technique to visualize the interior features within solid objects and

obtaining digital information on three-dimensional (3D) geome-

tries (Fig.

a). Here, CT is utilized to probe the solidi

fi

ed Pd-alloy

specimen to determine larger

fl

aws. An absorption X-ray CT

image, typically referred to as slice corresponds to a certain

thickness of the sample, each pixel of a slice is corresponding to a

volume element (voxel). Figure

a shows the result of complete

volumetric reconstruction after six melt-solidi

fi

cation cycles. The

reconstruction is composed of 1392 slices and each slice can be

analyzed separately. The gray levels in the displayed representa-

tive slice, i.e., Fig.

b, correspond to the X-ray attenuation, which

fl

ects the proportion of X-rays scattered or absorbed as they

pass through each voxel. X-ray attenuation is mainly a function of

the X-ray energy, the density and composition of the material

being imaged. The slice shown in Fig.

b shows the absence of

any kind of internal features. However, a strong beam-hardening

effect on the outer periphery of the sample slice (bright contrast)

is visible due to nonlinearity of the absorption coef

fi

cient of the

non-monochromatic X-ray beam; this effect can safely be

neglected. The sample, being spherical during the microgravity

phase, is vitri

fi

ed in the shape shown in Figs.

a, b due to the

deformation happening after the contact with the sample

pedestal at the end of the microgravity phase.

XRD on the other hand is based on constructive interference of

monochromatic X-rays on periodic arrangements of atoms or

molecules in a crystalline sample. Amorphous materials do not

possess that periodicity at a long range. Here, the scattering of X-

rays by atoms is considered; hence strong and narrow re

fl

ections

that represent long-range ordering are absent and instead broad

diffuse peaks will be present that are indicative of nearest

neighbor distances.

Six locations for XRD measurements were chosen based on the

sample reconstruction from CT. The sample shows a clear

rotational symmetry along the axis drawn between points 1 and

6, points 2

–

4 are located around the largest circumference of the

displayed reconstructed volume. Figure

c shows six individual

measurements from position 1 to 6 around the surface of the Pd-

alloy sample from the

fi

rst parabolic

fl

ight experiment. All of the

resulting diffraction patterns show symmetrical broad

fi

rst peaks

with maxima at 2.88 Å and second peaks maxima at 5.022 Å, the

symmetricity of the

fi

rst peak and the ratio between

fi

rst and

second peak maxima (1.74) are representative of amorphous

metallic samples.

The diffraction patterns for the experiments

performed in the parabolic

fl

ight campaign 2017 are comparable,

hence not shown here.

Fig. 2

Oscillation amplitude, as detected by the inductive method as a function of time.

Discrete Fourier transformation (DFT) spectrum of

the amplitude variation between 2.0 and 2.5 s.

Amplitude of the optically determined surface oscillations.

DFT spectrum of a selected time

slice

M. Mohr et al.

4

npj Microgravity (2019) 4

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

DISCUSSION

During two parabolic

fl

ight campaigns, the surface tension of

was successfully measured in the temperature

range of 1050 K

–

1850 K. While in 2016, at the lower temperature

range (below 1300 K), the surface tension was probably in

fl

uenced

by surface-active species leading to a positive temperature

coef

fi

cient, in the higher temperature range, a negative tempera-

ture coef

fi

cient is obvious. Even though we could not

fi

measurable composition gradients to the sample surfaces by EDX

measurements, surface segregation of a small fraction of surface-

active species, could be responsible for a change of the surface

tension. Further investigations on surface segregation and

fl

uences of adsorbents from the gas atmosphere are necessary

to explain this phenomenon.

In this work, the viscosity of Pd

was measured in

the liquid phase in a temperature range of 1100 K

–

1450 K, above

the liquidus temperature.

Directly measured values for the viscosity of liquid Pd

are scarce in literature, however, measurements of the

viscosity at and above the glass temperature are available. The

only direct measurement of viscosity of liquid PdCuNiP (Pd

) was performed by Haumesser et al.

using a gas-

fi

levitation technique. The investigated temperature range was

880 K

–

1137 K, where the viscosity varied between 69 mPa s and

11 mPa s. In contrast to that, our measurements were performed

in a temperature range from 1110 K to 1420 K and we obtained

viscosities between 63 mPa s and 7.5 mPa s for a PdCuNiP alloy

with a slightly higher Pd/Cu ratio of 43/27. The Pd/Cu ratio was

shown to be a critical parameter determining the thermophysical

properties in the Pd

+

30-x

system.

As was shown by Lu

et al., the glass transition temperature (

+

7 K) and the onset of

crystallization (

+

36 K) is higher, when the Pd/Cu ratio is increased

from 40/30 (

=

0) to 43/27 (

=

3).

This may also explain the

slightly different viscosities obtained for slightly different PdCuNiP

compositions in literature. Kato et al.

investigated the deform-

ability of Pd

42.5

7.5

at and slightly above the glass

transition temperature, obtaining lower viscosities than those by

Fan et al., who measured a slightly different composition

(Pd

Measurements of Lu et al.

were performed

in the solid state between 610 and 680 K, and are in very good

agreement with the measurements of Fan et al.

performed in

the same temperature range.

The measurements done by parallel plate rheometry, dilato-

metry, or beam bending around and above the glass transition

temperature can only cover a limited temperature range. Contain-

erless methods can complete the available data with viscosities in

the liquid phase. Figure

shows the viscosity of Pd

obtained by Fan et al.,

together with the viscosity obtained in

this study. The large temperature range (~ 1000 K), covering 16

orders of magnitude in viscosity, is used to compare theoretical

viscosity models.

The free volume model,

expresses viscosity as

log

Þ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ν

(4)

with

=

log(e), where

is the molecular volume and

and

are constant parameters of the model, used in the description

of the local free energy.

An alternative model is the hybrid model

that divides the

temperature dependence of viscosity in two temperature regimes.

Decreasing temperature in the low temperature regime increases

viscosity by increased chemical short-range order, while the

activated annealing of

fl

ow defects at high temperatures

decreases the viscosity for increasing temperature.

This can be

expressed as

Þ¼

exp

(5)

For the characteristic temperature, we used in both models the

Kauzmann temperature

=

507 K.

Figure

shows the

fi

t of both models to the experimental data.

The coef

fi

cients of determination (COD), calculated for the

fi

show that the free volume model

fi

ts the experimental data better

than the hybrid model. The determined parameters from both

model

fi

ts are shown in Table

The possible presence of surface-active species during the

parabolic

fl

ight in 2016, signi

fi

ed by the positive temperature

dependence of the surface tension is important for the evaluation

of the measured viscosity. Since an inhomogeneous coverage of

the surface with adsorbents could lead to convective Marangoni

fl

ows on the surface, the measured viscosity has to be viewed with

caution. Further analysis, such as magneto-hydrodynamic simula-

tions are necessary to investigate, if inhomogeneous coverage of

surface-active species would induce Marangoni convection and to

what extent such Marangoni

fl

ows could affect the apparent

viscosity measured by the oscillating drop method.

That the samples could be vitri

fi

ed during the process cycles in

the parabolic

fl

ight can be rationalized when the TTT diagram of

is compared with the cooling curve, as it is done

in Fig.

. The TTT diagram of Pd

was determined by

Schroers et al. and its nose was found to be at about 680 K and

200 s.

The increased negative slope of the temperature, after around

15 s, is due to the contact of the sample with the sample holder at

the end of the parabola. It is apparent that the sample

temperature dropped fast enough to bypass the nose of the TTT

diagram to prevent crystallization.

Fig. 3

Surface tension data obtained in both parabolic

fl

ight campaigns 2016 and 2017.

Arrhenius plot of the viscosity obtained for liquid

in the temperature range between 1100 K and 1450 K

M. Mohr et al.

5

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

npj Microgravity (2019) 4

Furthermore, re

fl

ection-based XRD analyses have been per-

formed for the both samples retrieved from the parabolic

fl

ight

campaigns 2016 and 2017. The resulting diffraction patterns show

the amorphous state of the specimens after the last melt and

solidi

fi

cation cycle, it shows the symmetry of the external heat

extraction geometry is maintained in the diffraction pattern, i.e.,

direct gas

fl

ow cools the fastest. Further, volumetric absorption-

based X-ray CT showed the uniformity of the specimen, and

con

fi

rmed only minimal weight loss (below 0.03%).

Fig. 4

Three-dimensional (3D) reconstructed volume of a X-ray computed tomography (CT) scan;

an example slice of Pd

Diffraction patterns obtained after the PF experiment (PF 2016) the inset shows a sketch of the vitri

fi

ed sample shape

Fig. 5

Temperature-dependent viscosity of Pd

, mea-

sured on board a parabolic

fl

ight (this work), and measured by

parallel plate rheometry and three-point beam bending by Fan

et al.

Table 1.

Parameters, determined by

fi

tting both models to the

experimental viscosity data of Pd

Model

Parameters

Free volume model

=

–

3.93,

=

448.42,

=

8.91 K

Hybrid model

=

2.32 × 10

–

Pa s,

=

1.32 eV,

=

1104 K

–

Fig. 6

The temperature variation during the cooling is shown after

the heater was turned off (red). Here, data of the last parabola

performed on the sample in the parabolic

fl

ight campaign 2016 is

shown. For comparison, the temperature-time transformation (TTT)

diagram measured by Schroers et al.

is shown

M. Mohr et al.

6

npj Microgravity (2019) 4

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

The surface tension and viscosity of Pd

BMG alloy

were measured during levitation in reduced gravity on board

parabolic

fl

ights. The surface tension was measured in the

temperature range from 1050 to 1850 K, while viscosity data

was obtained between 1100 and 1450 K.

The viscosity evaluated in the stable liquid phase appears

generally higher than the values determined by Haumesser et al.

for Pd

. The slightly higher glass transition tempera-

ture and the larger resistance of Pd

against crystal-

lization, compared with the Cu richer Pd

,isin

general agreement with the higher viscosity of liquid Pd

. A free volume model can be

fi

t well to the temperature-

dependent viscosity of Pd

given in literature and the

measurements in the present investigation.

Energy dispersive X-ray spectroscopy measurements have

shown that samples exhibit a homogeneous chemical element

distribution on the sample surface and cross section before and

after processing in μ

Volumetric absorption-based X-ray CT showed the homogene-

ity of the specimen and the symmetry of the cooling conditions in

the solidi

fi

ed samples. Scattering curves resulting from re

fl

ection

based X-ray diffractograms show the amorphous state of the

specimens after the last melt and solidi

fi

cation cycle. Further it

was observed that the excess free volume differs, depending on

the rate of heat extraction, given by the geometry of the specimen

and sample pedestal.

As shown in Figs.

c and

, the μ

time was too short to cool the

liquid melt below the glass transition temperature before it

touched the sample pedestal. As such, it is remarkable that the SiN

sample holder pedestal did not act as a heterogeneous nucleant.

Studies concerned with the solidi

fi

cation or vitri

fi

cation under μ

were therefore not possible in this experiments.

Considering the very small critical cooling rate of the PdCuNiP

BMGs (below 0.1 K/s) and the cooling rates of ca. 20 K/s (see Fig.

c) that are typically achievable in the TEMPUS facility, would

allow the vitri

fi

cation under μ

conditions if the μ

time would last

longer. Vitri

fi

cation on ground (at 1

) is accompanied by effects of

the ubiquitous gravity, which can be avoided by processing under

microgravity. Longer μ

times can be achieved by other μ

platforms such as the International Space Station (ISS), which also

open the possibility to study further material properties, such as

the speci

fi

c heat capacity and thermal conductivity in the

(undercooled) liquid state. Hence, preparations for the processing

of noble metal-based metallic glasses, such as PdCuNiP in the EML

facility (ISS-EML) on board the ISS are ongoing.

The containerless processing in TEMPUS on board parabolic

fl

ights, as presented here, is a good possibility for the measure-

ment of thermophysical properties within μ

times of around 20 s,

like surface tension and viscosity. Besides, it is a testing platform

for sample positioning, heating, and cooling ef

fi

ciency, which give

important knowledge for the design of experiments with similar

samples in the electromagnetic levitator ISS-EML on board the

Columbus module of the ISS.

METHODS

Sample preparation, structural, and chemical analysis

A master alloy with the composition Pd

was prepared at the

precious metal foundry of PX group. A one-kilo batch has been melted in a

vacuum induction furnace using pure components. Purity was 99.95% for

the palladium and nickel, 99.99% for the oxygen-free copper, and 99.999%

for phosphorous. The obtained master alloy is then re-melted and cast in

amorphous rods of 8 mm diameter in a vacuum induction furnace without

any

fl

uxing. Suitable pieces of the rod were used to prepare spheres of

6.5 mm diameter in a water-cooled copper mold using an arc melter. Their

exposure to air was minimized to the time needed for sample integration

in the TEMPUS facility (~20 min). The base pressure in the process chamber

of the levitator was in the range of 2 × 10

-7

mbar.

After the processing in μ

, the samples were analyzed by EDX, XRD, and

CT. XRD was performed on a Stoe imaging plate diffractometer system

(IPDS II, Stoe & Cie GmBH, Darmstadt, Germany). Two-dimensional (2D)

diffraction data in re

fl

ection mode were collected at room temperature

using MoK

radiation (

=

0.71073 Å); the resulting images were then

azimuthally integrated. An X-ray CT setup was utilized at 280 kV in its

nominal geometry, a total of 1392 z-slices were acquired in 3 h. The CT

setup was composed of a microfocus source from Finetec (model FOMR

300.03Y RT) and a

fl

at-panel detector with 100 μm

pixels from Perkin

Elmer (model XRD 1611-CP3).

Contact-less EML

–

TEMPUS

The experiments were performed using the electromagnetic facility

TEMPUS, which has been run by DLR personnel on board a parabolic

fl

ight airplane

operated by Novespace. TEMPUS consists of a process and

a sample chamber, which are connected to a high vacuum pumping

system and a gas circulation unit. The gas circulation system is equipped

with a gas cleaning cartridge speci

fi

ed to impurity levels <1 ppb for O

and

O. The sample is heated and positioned by two different radio frequency

(rf-) electromagnetic

fi

elds: a dipole

fi

eld for heating and a quadrupole

fi

eld for positioning. The rf-power is supplied by two rf-generators

operating at frequencies of 375 kHz and 150 kHz for heating and

positioning, respectively.

Further details of the experimental setup and

data analysis are described elsewhere.

To determine the electrical resistivity and the sample radius of the

processed liquid sample by contact-less inductive means a measurement

electronics (SCEs) is attached to the facility. It measures the current,

voltage, phase shift, and frequency of the rf generator that establishes the

heating dipole

fi

eld.

Due to its high sensitivity, the SCE also allows the detection and

evaluation of the surface oscillations of a liquid metallic droplet via the

inductive coupling of the rf-heaters oscillating circuit and the induced

current distribution in the sample. The SCE operates with a sampling rate

of 400 Hz. This is very well suited for surface oscillation analysis, as their

typical frequencies are in the range between 20 and 50 Hz.

The process chamber is equipped with several observation windows

allowing the recording of the sample shape in two perpendicular

directions using two high-speed cameras. One camera is mounted axially

(for top view) along the direction of the rf-induction coil axis and the

second one radially, in a direction perpendicular to the former. Both

cameras are typically operating at 200 Hz, compromising between

brightness and temporal resolution.

An optical pyrometer is integrated in the axial camera for temperature

measurement in the range between 300 and 2100 °C. The optical

pyrometer operates at a sampling rate of 100 Hz.

The sample chamber sits below the process chamber. Samples are

contained in a sample holder with either a metallic wire cage structure or a

ceramic cup on top of a SiN pedestal.

Due to the relatively short μ

times, convective cooling with a He

–

Ar gas

mixture is necessary to increase the cooling rate, aiming for solidi

fi

cation

during the μ

phase.

Oscillating drop method

The heater pulses during the nearly force-free cooling phase lead to an

axial elongation of the sample, which leads predominantly to oscillations in

the

2,0

mode. Since in a force-free μ

environment all

2,m

are

degenerated,

only one oscillation frequency

=

(the Rayleigh

frequency) can be observed. Under 1

EML, the sample is not force free

and spherical, but deforms. This leads to a split and shift of the measured

oscillation frequencies

. A correction was developed

and successfully

proven by comparison of μ

with ground-based surface tension

measurements in an EML device.

Periodic sample movements within

the positioning

fi

eld lead to small, periodic forces on the sample. Under

this condition, the application of the so-called Cummings and Blackburn

correction

results in a reduction of the surface tension values in the

range of 2

–

3% when the measured surface oscillation frequency

instead of

is used in the formula for the evaluation of Eq.

. This

correction was not applied to the data presented here. Also, the absence of

excessive sample rotation is necessary for the validity of Eq.

, which was

the case in our experiments.

Viscosity is evaluated from the damping time constant

of the surface

oscillations according to Eq.

. There are, however, some subtleties and

constraints in the application of Eqs.

and

. For small enough viscosities,

M. Mohr et al.

7

Published in cooperation with the Biodesign Institute at Arizona State University, with the support of NASA

npj Microgravity (2019) 4