Professional Documents
Culture Documents
3 Amorphous
and Nanocrystalline Materials 9 Nano- and Micromaterials
Preparation, Properties, Editors: K. Ohno, M. Tanaka, J. Takeda,
and Applications and Y. Kawazoe
Editors: A. Inoue and K. Hashimoto
11 High-Temperature Measurements
5 Structure and Properties of Materials
of Aperiodic Materials Editors: H. Fukuyama, Y. Waseda
Editors: Y. Kawazoe and Y. Waseda
High-Temperature
Measurements
of Materials
ABC
Professor Hiroyuki Fukuyama
Professor Dr. Yoshio Waseda
Tohoku University, Institute of Multidisciplinary Research for Advanced Materials
2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan
E-mail: fukuyama@tagen.tohoku.ac.jp, waseda@tagen.tohoku.ac.jp
Series Editor-in-Chief:
Professor Yoshiyuki Kawazoe
Institute for Materials Research, Tohoku University
2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan
Series Editors:
Professor Masayuki Hasegawa
Professor Akihisa Inoue
Professor Norio Kobayashi
Professor Toshio Sakurai
Institute for Materials Research, Tohoku University
2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
List of Contributors
Ivan Egry
Institute for Materials Physics Taketoshi Hibiya
in Space, German Aerospace Center Graduate School of System Design
(DLR) and Management, Keio University
51170 Cologne German, 4-1-1 Hiyoshi, Kohoku-ku, Yokohama
Ivan.Egry@dlr.de 223-8528, Japan,
t.hibiya@sdm.keio.ac.jp
1.1 Introduction
The levitation technique can also be a very elegant way to handle liquid
samples on diffraction experiments. For these experiments, they offer many
advantages. As the liquid sample is held containerless, there is no need to
subtract the diffraction contribution of a crucible from the total scattering
intensity. Hence, this reduces by half the measurement time because it is
not necessary to measure the diffraction of the empty cell. In addition, the
symmetrical shape (nearly spherical) of the sample enables us to evaluate
easily the correction of absorption and multiple scattering. Moreover, the most
important advantage is the fact that the liquid samples under containerless
conditions can easily reach deep undercooled states because the heterogeneous
nucleation sites, which are usually the inner wall of the crucible, are absent.
The observation of the structure of undercooled liquids can be realized by
these techniques.
Up-to-date, several levitation techniques have been applied to the diffrac-
tion experiments for the structural analysis of liquid matters, as shown in
Table 1.1, in which the typical works of respective levitation techniques are
summarized. The aerodynamic levitation is simple, yet useful for such exper-
iments [8–25]. In this method, a sample is levitated in a conical nozzle at the
location of the minimum potential well of the gas flow. Since the sample is
small (1–3 mm diameter), this levitator is well suited for with the application
to a high-energy synchrotron radiation X-ray beam. Moreover, this process can
be applied to several types of materials under various atmospheres, though
this cannot be used under vacuum conditions. So far, X-ray diffraction exper-
iments of several high-temperature melts (e.g., liquid boron, alumina) have
been performed by this technique and provided the static structure factors of
these liquids [8–25]. Recently, using this method, liquid structures of glass-
forming alloys and ceramics have been studied together with the discussion
of total and partial structure factors with the aid of the computer simula-
tions [11, 12]. The electromagnetic levitation is another technique applicable
to diffraction experiments. In this method, a sample of conductive material
is levitated in a RF coil. The high frequency current of the coil induces an
eddy current in the metallic sample and the electromagnetic force is induced
for the levitation. The levitated sample is positioned at a stabilized point,
which depends on the shape of the coil and on the electromagnetic properties
of the sample. Since the sample size is large (6–8 mm diameter), this method
is especially well suited to neutron scattering experiments. Schenk et al. [29]
applied this levitation method to neutron and X-ray scattering experiments of
equilibrium and non-equilibrium liquid metals. Recently, Watanabe et al. [26]
studied the structure of liquid silicon by using this technique.
It is also possible to levitate matter by applying electrostatic forces, under
an active feedback system, to charged samples by electronic emission [39–43].
Electrostatic levitation is extremely attractive for X-ray diffraction experi-
ments for several reasons. The size of the levitated sample (1–2 mm diameter)
is suitable for the diffraction of high-energy X-rays from synchrotron radiation
source, taking into account the X-ray absorption coefficient and the atomic
1 Electrostatic Levitation for Liquid Structural Analysis 3
Table 1.1. Liquid structural analysis by using diffraction methods coupled with
levitation techniques
j
A
k i
l
e
b B
f
c a
d
Fig. 1.1. Side view of the chamber for electrostatic levitation. a, levitated sample; b,
upper electrode; c, lower electrode; d, side electrodes; e, ceramic support; f, mirrors;
g, positioning rod; h, solenoid; i, beam expander; j, He–Ne laser; k, position detector;
l, ZnSe window; A, beam path of heating CO2 laser; B, beam path of positioning
He–Ne laser
e
A f
d
d
B
b
D
e
Fig. 1.2. Top view of the chamber for electrostatic levitation. a, sapphire window for
the incident X-ray beam; b, glass window for the pyrometer; c, beryllium window for
diffracted X-rays; d, mirrors for the He–Ne laser for positioning in X-Z directions; e,
mirrors for the He–Ne laser for positioning in Y direction; f, mirror for CCD camera;
A, path of incident X-rays; B and C path of He–Ne lasers; D, path of pyrometer
1 Electrostatic Levitation for Liquid Structural Analysis 7
1.3 Experimental
Electrostatic levitation is, in principle, applicable to a wide variety of materials
because all charged materials can be levitated by the action of electrostatic
forces. For the first experiment, zirconium was selected and the structural
analysis of its liquid phase was carried out by high-energy X-ray diffraction
measurement at SPring-8, which is the third generation synchrotron radiation
facility in Japan. Similar experiments were performed for molten silicon and
alumina samples by using a laboratory X-ray source.
8 T. Masaki et al.
For the present experiments, the typical sample size was about 2 mm in
diameter. Spherical zirconium samples were prepared in the following manner:
99.5 mass% pure zirconium wire was cut into 30–32 mg pieces; these pieces
were melted with a diode laser (wavelength, 808 nm; max. power, 200 W) in a
glove box filled with purified argon; the melted pieces took a spherical shape
spontaneously because of the surface tension. Spherical samples of silicon and
alumina were made similarly.
Heating is the most delicate task on the electrostatic levitation of sample
since the sample charge has a tendency to decrease due to the evaporation of
absorbed gas or metallic oxides from its surface. In particular, heating of the
sample from room temperature must be done carefully because the discharge
of sample starts at a temperature of about 800 K. However, the charge can
be increased through electronic emission if the sample reaches a temperature
at which thermionic emission dominates (∼1,500 K in the case of metals). To
overcome these difficulties, the “hot launch” method [48] was used. To initiate
levitation, the sample was heated for the removal of surface oxides. When
it reached ∼1,500 K, at which thermionic emission dominates, high voltage
was applied to the electrodes and the feedback control system was activated.
Once levitated, the sample could then be brought to temperatures beyond the
melting point or be maintained under undercooled conditions for hours. The
aforementioned method was adopted for the levitation of liquid zirconium and
alumina. However, low melting point materials (e.g., silicon) have a tendency
to stick to the positioning rod and, therefore, the heated sample must be
tossed while heating. In the present apparatus, a small solenoid, which created
vibrations, was fixed at the lower part of the positioning rod. The solenoid
was activated remotely during the monitoring of the sample temperature.
The temperature of the sample was measured with the use of a single-color
pyrometer. The emissivity of the sample is necessary to obtain the exact value
of the temperature. However, the emissivity strongly depends on the sample
condition, the sample size, the focus of collimation lens of the pyrometer,
and the transparency of the window of the chamber. In the present research,
the emissivity calibrated referred to the melting points of zirconium and sil-
icon. The undercooled liquid state was established simply by decreasing the
laser power. After a certain depth of undercooling was reached, the sudden
recalescence of the sample, due to the release of the latent heat of fusion,
was observed on solidification. Therefore, the undercooled liquid state was
confirmed by monitoring the signal of the pyrometer. The laser power was
controlled to keep the temperature of the sample constant.
The high-accuracy measurement of the liquid structure is one of the ma-
jor purposes of this research. The two-axis diffractometer is the most typical
instrument used nowadays for the diffraction experiments. The X-ray source
was selected by considering the absorption coefficient of the material. For zir-
conium, which has a high absorption coefficient, high-energy X-rays (113 keV)
from the BL04B2 of SPring-8 were used. For silicon and alumina, the labora-
tory X-rays from a Mo rotation target were sufficient to carry out preliminary
1 Electrostatic Levitation for Liquid Structural Analysis 9
diffraction experiments. The size of the incident beam was 0.7 mm in width
and 2.0 mm in height for the synchrotron radiation X-rays, and 3 × 3 mm2 for
the laboratory X-rays. The incident beam was collimated by the slit and deliv-
ered to the vacuum chamber through a sapphire window. The angular depen-
dence of the intensity of the X-rays diffracted from the sample was measured
in transmission geometry by a germanium detector or a proportional counter
with a graphite monochrometor. Slit collimation eliminated the scattering
from the windows on the chamber. This configuration is very helpful in per-
forming the precise measurements of the diffraction only from the sample. The
intensity of diffracted X-rays was acquired in each diffraction angle by the step
scan method. The diffraction data was collected over a 2θ range of 0.3◦ –25◦
for measurements by synchrotron radiation and 0.5◦ –80◦ for measurements by
−1
laboratory X-ray one. The obtained Q range of S(Q) was 0.3–24.7 Å in the
−1
former case and 0.08–11.4 Å for the latter case. The duration of acquisition
of data for each diffraction angle was greater than 5 s, which is sufficiently long
with high statistics. To obtain the static structure factor of liquids, data cor-
rection of the absorption, background, polarization, and multiple scattering
must be performed [1–3]. For the laboratory experiments, the cross-sectional
area of the sample is smaller than that of the X-ray beam, and the correc-
tion of absorption and polarization can be performed by normal methods [2].
Though the width of incident X-rays (0.7 mm) is narrower than the sample
diameter (2 mm) in the synchrotron radiation experiments, the influence of to-
tal angular dependence of the absorption and geometric factor was negligibly
small, because diffraction experiments could be performed with rather small
scattering angles for high-energy X-ray experiments. In addition, it is worth
mentioning that the absorption coefficient itself is very small for high-energy
X-rays (mass absorption coefficient is 0.673 cm2 g−1 for zirconium [49]). This
implies that the contribution of absorption correction is extremely small. The
contribution of multiple scattering has been mentioned for the structural anal-
ysis of disordered matter not only for the neutron scattering experiments but
also for X-ray diffraction measurements. In the present study, the sample was
spherical and quite small. We evaluated the contribution of double scatter-
ing compared with that of the single scattering by the method reported by
Warren [1]. The ratio of the double scattering to the single scattering in the
present case was less than 1%, and therefore, the contribution was neglected.
After the correction of the absorption [49], background, and multiple scat-
tering, the contribution of Compton scattering [50] was subtracted and then
the X-ray static structure factor [51], S(Q), was derived from the corrected
coherent intensity, I(Q), based on the equation
2 2
I(Q) = f (Q) [S(Q) − 1] + f (Q) , (1.2)
where the angular brackets represent averages over all atoms and f (Q) is the
atomic form factor [52].
10 T. Masaki et al.
4
Intensity(arb.)
3
2125K(mp)
2
2035K(mp-90)
1
back ground
0
0 5 10 15 20
2θ(deg)
Fig. 1.3. High-energy synchrotron X-ray diffraction pattern of liquid and under-
cooled zirconium and background
3000
Intensity(count)
2000
1000
Si
back ground
0
0 20 40 60 80
2θ(deg)
Fig. 1.4. X-ray diffraction pattern of liquid silicon at the melting point and back-
ground
1 Electrostatic Levitation for Liquid Structural Analysis 11
4
1.2x10
1.0
0.8
I(count)
0.6
0.4
0.2
0.0
0 20 40 60 80
2θ(deg)
Fig. 1.5. X-ray diffraction pattern of liquid alumina at the melting point
MC
3
S(Q)
2125K(m.p.)
2
2035K(m.p.-90K)
1
0
0 5 10 15 20
Q(A-1)
Fig. 1.6. Static structure factor of normal and undercooled liquid zirconium ob-
tained from high-energy X-ray diffraction experiment and Monte Carlo simulation
(2,125 K); normal, melting point; undercooled, the temperature of 90K below than
the melting temperature
This small error from the background correction in the present experiments
is remarkably different from the case of conical nozzle levitation. Thus the
combination of containerless conditions and high vacuum enabled us to obtain
the very reliable liquid structure.
The static structure factor S(Q) of liquid samples can be obtained from
the diffraction intensity. The static structure factors shown in Figs. 1.6–1.8
demonstrate that we have succeeded in performing precise observations of the
liquid structure with the present electrostatic levitation apparatus, not only
for synchrotron radiation X-rays but also for laboratory X-ray source. The liq-
uid structures of the materials investigated in this study have been measured
12 T. Masaki et al.
1.6
1.4
1.2
1.0
S(Q)
0.8
0.6
0.4
0.2
0.0
0 2 4 6 8 10
-1
Q(A )
Fig. 1.7. Static structure factor of liquid silicon at the melting point
2.0
1.5
S(Q)
1.0
0.5
0.0
0 2 4 6 8 10 12
-1
Q(A )
Fig. 1.8. Static structure factor of liquid alumina at the melting point
10
6
u(r)/kBT
4
0 2 4 6 8 10 12
r(A)
Fig. 1.9. Effective pair potential of liquid zirconium (2,125 K) based on modified
hypernetted chain approximation
In addition, high-Q data are necessary to reduce the truncation error in the
Fourier transform of S(Q) on the evaluation of g(r) and c(r), which is nec-
essary for the precise determination of the repulsive part of ueff (r), as shown
in (1.3). The BHS (r, η) is the bridge function of the hard sphere fluid and
η is the packing fraction. For the conventional estimation of BHS , the η of
liquid zirconium was taken as 0.46, which is generally adopted for the pack-
ing fraction at the melting point of liquid metals [54]. The ueff (r) obtained
is shown in Fig. 1.9. The validity of obtained ueff (r) was investigated by a
Monte Carlo (MC) simulation [55, 56], in which the obtained ueff (r) was em-
ployed. The temperature of MC was 2,125 K. The S(Q) derived from MC is in
good agreement with the experimental results, as can be seen in Fig. 1.6. The
obtained ueff (r) is widely applicable to the evaluation of not only the static
properties but also the dynamic properties. For example, transport proper-
ties, such as self-diffusion and viscosity coefficients, can be estimated from the
combination of ueff (r) and molecular dynamics (MD) simulation. The viscos-
ity coefficient of liquid zirconium has been already measured by the oscillation
drop method coupled with the electrostatic levitator [57]. The detail analysis
for the viscosity coefficients by the MD simulation by using this ueff (r) is in
progress.
The evaluation S(Q) of liquid silicon was performed for the laboratory
X-ray source. The Q range of S(Q) was rather small for the calculation of
g(r) by the Fourier transform shown in (1.1). Nevertheless, the present S(Q)
for liquid silicon was sufficiently good as the structure data. For example, the
g(r) of liquid silicon was calculated by introducing the present S(Q) into the
Reverse Monte Carlo (RMC) simulation [58]. The obtained coordination num-
ber of nearest neighbors was 5.9, which agrees well with previous research [35].
14 T. Masaki et al.
The present S(Q) of liquid alumina shows a better agreement with the results
of molecular dynamics simulation [59] compared with the experimental S(Q)
reported by Krishnan et al. [59]. This can be explained by the fact that the
reliability of our S(Q) in the low Q region is much better than the experiment
reported in [59]. It was previously reported that the structure of liquid alumina
depends on the ambient atmospheric oxygen concentration [59]. No such be-
havior was found in our data. Furthermore, our preliminary high-energy X-ray
diffraction experiments using a conical nozzle did not show such a behavior.
This may suggest that the structural difference of liquid alumina in oxidizing
and reducing condition is still open to question. Diffraction experiments on
liquid silicon and alumina by synchrotron radiation X-ray source are being
planned in a near future.
It is desired that different types of materials can be successfully studied by
using a single apparatus. For the diffraction experiments, this feature is very
advantageous because a common optical set-up and background calibrations
can be used. We developed such an electrostatic levitator which is applicable
to a wide variety of materials and X-ray sources. Furthermore, we confirmed
that the quality of structure data for high-temperature liquids by the presnt
apparatus is much better than that in previously published literatures. We
believe that we present a reliable apparatus for the structure study of high-
temperature liquids and undercooled ones. The present apparatus will be able
to perform the experimental analysis of high-temperature melts with high
precision and will contribute to the fundamental understanding of the essential
feature of liquids in normal and undercooled states.
References
1. B.E. Warren, X-Ray Diffraction (Addison-Wesley, Reading, MA, USA, 1969)
2. Y. Waseda, The Structure Non-Crystalline Materials (McGraw-Hill, NY, USA,
1980)
3. H.E. Fisher, A.C. Barnes, P.S. Salmon, Pep. Prog. Phys. 69, 233 (2006)
4. J.K.R. Weber, S. Krishnan, S. Ansell, S.D. Hixson, T.C. Nordine, Phys. Rev.
Lett. 84, 3622 (2000)
5. T.G. Wang, J. Fluid Mech. 308, 1 (1996)
6. I. Egry, A. Diefenbach, W. Dreier, J. Piller, Int. J. Thermophys. 22, 569 (2001)
7. W.-K. Rhim, S.K. Chang, D. Barber, K.F. Man, G. Gutt, A. Rulison, R.E.
Spjut, Rev. Sci. Instrum. 64, 2961 (1993)
8. S. Ansell, S. Krishnan, J.J. Felten, D.L. Price, J. Phys. Condens. Matter 10,
L73 (1998)
9. N. Jaske, L. Hennet, D.L. Price, S. Krishnan, T. Key, E. Artacho, B. Glorieux,
A. Pasturel, M.-L. Saboungi, Appl. Phys. Lett. 83, 4734 (2003)
10. A. Alatas, A.H. Said, H. Sinn, E.E. Alp, C.N. Kodituwakku, B. Reinhart, M.-L.
Saboungi, D.L. Price, J. Phys. Chem. Solids 66, 2230 (2005)
11. A. Mizuno, T. Kaneko, S. Matsumura, M. Watanabe, S. Kohara, M. Takata,
Mater. Sci. Forum 561–565, 1349 (2007)
1 Electrostatic Levitation for Liquid Structural Analysis 15
37. T. Volkmann, J. Strohmenger, J. Gao, D.M. Herlach, Appl. Phys. Lett. 85, 2232
(2004)
38. O. Heinen, D. Holland-Moritz, D.M. Herlach, K.F. Kelton, J. Cryst. Growth
286, 146 (2006)
39. T.H. Kim, G.W. Lee, B. Sieve, A.K. Gangopadhyay, R.W. Hyers, T.J. Rathz,
J.R. Rogers, D.S. Robinson, K.F. Kelton, A.I. Goldman, Phys. Rev. Lett. 95,
085501 (2005)
40. G.W. Lee, A.K. Gangopadhyay, K.F. Kelton, R.W. Hyers, T.J. Rathz, J.R.
Rogers, D.S. Robinson, Phys. Rev. Lett. 93, 037802 (2004)
41. K.F. Kelton, G.E. Lee, A.K. Gangopadhyay, R.W. Hyers, T.J. Rathz, J.R.
Rogers, M.B. Robinson, D.S. Robinson, Phys. Rev. Lett. 90, 195504 (2003)
42. G.W. Lee, A.K. Gangopadhyay, T.K. Croat, T.J. Rathz, R.W. Hyers, J.R.
Rogers, K.F. Kelton, Phys. Rev. B 72, 174107 (2005)
43. A. Ishikura, A. Mizuno, M. Watanabe, T. Masaki, T. Ishikawa, S. Kohara, J.
Am. Ceram. Soc. 90, 738 (2007)
44. A.K. Gangopadhyay, G.W. Lee, K.F. Kelton, J.R. Rogers, A.I. Goldman, D.S.
Robinson, T.J. Rathz, R.W. Hyers, Rev. Sci. Instrum. 76, 073901 (2005)
45. H. Aoki, P.-F. Paradis, T. Ishikawa, T. Aoyama, T. Masaki, S. Yoda, Y. Ishii,
T. Itami, Rev. Sci. Instrum. 74, 1147 (2003)
46. T. Masaki, T. Ishikawa, P.-F. Paradis, S. Yoda, J.T. Okada, Y. Watanabe,
S. Nanao, A. Ishikukra, K. Higuchi, A. Mizuno, M. Watanabe, S. Kohara, Rev.
Sci. Instrum. 78, 026102 (2007)
47. S. Kohara, K. Suzuya, Y. Kashihara, N. Matsumoto, N. Umesaki, I. Sakai, I.
Nucl. Instrum. Methods Phys. Res. Sect. A 467, 1030 (2001)
48. T. Ishikawa, P.-F. Paradis, S. Yoda, Rev. Sci. Instrum. 72, 2490 (2001)
49. S. Sasaki, X-ray Absorption Coefficients of the Elements (Li to Bi, U). KEK
Report 90-16 (National Laboratory for High Energy Physics, Japan, 1991)
50. J.H. Hubbell et al., J. Phys. Chem. Ref. Data, 4, 471 (1975)
51. D. Waasmaier, A. Kirfel, Acta Crystallogr. A 51, 416 (1995)
52. T.E. Faber, J.M. Ziman, Philos. Mag. 11, 153 (1965)
53. M.W.C. Dharma-wardana, G.C. Ares, Phys. Rev. B 28, 1701 (1983)
54. M. Shimoji, Liquid Metals (Academic Press, London, 1977)
55. N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, E. Teller, J.
Chem. Phys. 21, 1087 (1953)
56. W.W. Wood, F.R. Parker, J. Chem. Phys. 40, 517 (1981)
57. P.-F. Paradis, T. Ishikawa, S. Yoda, Int. J. Thermophys. 23, 825 (2002)
58. R.L. McGreevy, L. Pusztai, Mol. Simul. 1, 359 (1988)
59. S. Krishnan, L. Hennet, S. John, T.A. Key, P.A Madden, M.-L. Saboungi, D.L.
Price, Chem. Mater. 17, 2262 (2005)
2
Viscosity and Density Measurements
of High Temperature Melts
Yuzuru Sato
2.1 Introduction
Since the viscosity and density are most fundamental properties for any fluids,
many efforts to obtain reliable values have been made. However, the measure-
ments are not so easy, especially at high temperature in molten state. The
high temperature melts are typically classified into molten metals, molten
salts, and molten oxides. They appear in many industrial processes, for exam-
ple, steelmaking, nonferrous metallurgy, aluminum smelting, foundry, glass
making, etc. The adaptable methods for the measurements should be chosen
carefully by considering some physical and chemical properties of the melt.
Iida published the review on the properties including viscosity and density of
molten metals [1], and the comparison among the viscosities of molten iron
reported by many researchers showed considerable difference of several dozen
percent. The viscosity value is in considerably wide range depending on the
groups of the melts, for example, in general low for molten metals and high for
molten silicates, including slag and glass, and the difference reaches more than
ten orders by reflecting the difference in the melt structure. On the other hand,
density is mainly depending on atomic mass and not so different to each other
because of not so big difference in molar volumes of the components. Various
methods for viscosity and density measurement were also introduced [2] and
also the viscometries were summarized [3].
The typical methods used for viscosity measurement are a capillary method,
an oscillating method, a rotating method, etc. Although it is difficult to adapt
them to some liquids with non-Newtonian behavior, there are almost no prob-
lems because most high temperature melts show Newtonian behavior. The
capillary method, which is used as the prescriptive method at room temper-
ature, is simple and very precise in relatively wide viscosity range, although
the application at high temperature is limited due to the problem in the use
of refractory materials for apparatus and the difficulty for detecting meniscus.
However, the successful application was reported on high temperature molten
salts [4]. The oscillating method, especially an oscillating cylinder method [5],
is also precise and sensitive although the measurable viscosity is limited in
lower range. The big advantage of oscillating method is in wide selection of
the container material. Therefore, it is used for high temperature melts with
low viscosity such as most of molten metals and many molten salts. The rotat-
ing method [6] is suitable generally for high viscosity range because it contains
power train as the driven device with friction loss. The method is suitable for
measuring the viscosity of molten glass and slag.
Filtration chamber
Quartz frit
Connection tube
to be sealed
Funnel
Upper fiducial mark
Timing bulb
Capillary
Melt
than 100 to maintain the stable layer flow. The inverse funnel at the end
of capillary is a device to reduce the surface tension effect, which gives the
error on the effective height. Typical capacity of a timing bulb is about 3 cm3 .
The time for flowing out is measured as a time interval between the meniscus
passes through two fiducial marks. The viscometer was made available by the
combination with a transparent “Gold Furnace” in which the temperature
uniformity within 0.5 K was kept around the viscometer and the meniscus
detection was possible visually as shown in Fig. 2.2. The measurement can
be repeated by taking the liquid back to the timing bulb from the bottom by
rotating the whole apparatus together with the furnace. The reproducibility of
the measurement was within 0.2 s, which was corresponding to 0.05%–0.2% of
the flow time. The total error accompanied in the measurement was reported
not exceeding 1% and the viscosity determination was highly reliable. The
viscosities measured of alkaline halides are shown in Fig. 2.3. The results of
fluorides were measured with oscillating viscometer [8]. The feature of the
viscosities of molten alkaline halides are similar to each other, especially for
common anion. As mentioned earlier, the capillary viscometer can be improved
to be adapted to high temperature molten salts. However, it has limitation as
being made of quartz and cannot be adapted to erosive melt such as molten
fluorides, carbonates, or slag. Furthermore, it is hardly adaptable to molten
metals because of the difficulties to remove trace of oxide film on the metals
or the reactivity with active metals such as aluminum or magnesium.
20 Y. Sato
Adjusting screw
Halogen
lamp Capillary viscometer
Thermocouple
0.3
F
Li
Br
Br
Cs
Li
Rb sF
0.2
Br
l l
RbCsC
F
l
C
C
Cs
Na
Li
Rb
Cl
Rb
KF
log h
Kl
0.1
r
KBr
aB
al
lN
N
KC
Li Fluoride
Cl
Na
Na
Chloride
0 K Bromide
Rb
Cs Iodide
-0.1
T -1, 10-3K-1
Fig. 2.3. Viscosities of alkaline halides
2 Viscosity and Density Measurements of High Temperature Melts 21
L
C
t
-L
t1 t2 t3 t4
Fig. 2.4. Oscillating curve with decay and the time intervals measured in a period
2 Viscosity and Density Measurements of High Temperature Melts 23
τ = t1 + t2 + t3 + t4 . (2.9)
The constants A and C in (2.8) are defined by (2.10) and (2.11), respectively,
for each period.
A = 2L/[sin{π(0.5 − t1 /τ )} − sin{π(1.5 − t3 /τ )}], (2.10)
δ = B τ . (2.13)
1.0
Logarithmic decrement
0.0 Period
-1.0
1st Approx.
-2.0
Log (Error/%)
-3.0
-4.0
2nd Approx.
-5.0
-6.0
-7.0
0.001 0.01 0.1
Logarithmic decrement
Fig. 2.5. Error contained in primary and secondary approximation values through
the approximation method
24 Y. Sato
t⬘
C
t
-L
Fig. 2.6. Relation between true and apparent periods in the wave with decay
C
ta t
tb
1 -1 w p 1 -1 w
w tan ( B ) w
-
w tan ( B )
where the parameter, τ is taken from (2.9) and ω/B = 2π/δ is taken from δ
in (2.13), which are the first approximate values. The latest period, τ , is used
to calculate the second approximate value of logarithmic decrement through
(2.10)−(2.13) again.
The second approximate values of the period of oscillation and the log-
arithmic decrement are sufficiently precise. The error accompanied by the
approximation is about from 10−4.5 % to 10−3.5 % as also shown in Fig. 2.5 in
the same conditions earlier. Therefore, this set of second approximate values
was employed for calculating the viscosity.
An example of the suspension and oscillating system [5] is shown in Fig. 2.8.
A cylindrical vessel is suspended by a wire through a mirror and an inertia
disk made of aluminum. The rotational force is given to the inertia disk elec-
tromagnetically. The oscillation is started by removing the rotational force
and is attenuated by the viscosity resistance without the change of period of
oscillation. The reflected light from a laser passes through two photo detectors
to measure the time intervals. Figure 2.9 shows an example of the viscometer.
The uniform temperature profile along the crucible is very important, espe-
cially at high temperature to avoid the convection flow and inhomogeneity in
the melt. The furnace in the figure consists of three independent heaters, and
many thermal shielding plates made of refractory metal are installed above
and under the crucible suspended in an inner tube to obtain good temper-
ature profile. Additionally, it is also important to keep the temperature of
suspension wire constant, because the modulus of rigidity varies with tem-
perature and affects the measurement. As the vessel with simple geometry
is used in this method, there is a great advantage. It is the wide flexibility
on the choice of materials for the vessel. The vessel made of metal or various
refractory ceramics are available for molten salts and molten metals for high
temperatures. It should be noted that the viscosity range measurable in the
26 Y. Sato
Data processing
computer
wire(Pt-13%Rh)
(PC-9801)
mirror(Al)
Time
Counter
photo
transistor
inertia disc
laser light
oscillating vessel method is limited to low viscosity, typically less than about
30 mPa s due to the principle of the method. Therefore, the oscillating method
cannot be applied to high viscosity melts such as molten slag or glass. How-
ever, most of molten salts and molten metals have low viscosity typically less
than 10 mPa s and the oscillating method is successfully adaptable to them.
The viscosities of molten metals and semiconductors measured with the
oscillating method are shown in Fig. 2.10. The viscosities of molten metals
generally show relatively low values and good Arrhenian behavior, including
semiconductors was obtained. In general, viscosity and activation energy are
high as the melting temperature increases.
As the viscosity is originated from the friction between the hypothetical layer
in the liquid, torque measurement is principle and the method is prevailing.
Therefore, the measurable torque created by the liquid between coaxial in-
ner cylindrical column and outer cylinder with different angular velocities is
2 Viscosity and Density Measurements of High Temperature Melts 27
Fig. 2.9. Whole view of the oscillating viscometer for high temperature
1.00
Fe
Co Co
0.80 Fe Au Ni
Au
Log (Viscosity, h / mPa.s)
Ag Cu
0.60 Ni Ag
Zn Zn
Cd Al
0.40 Cu Tl Cd
Pb
Bi Sb
Pb
Sn Bi
0.20 Sn
InSb Tl
In
Al In
0.00 Sb
Ga
Si
GaSb Ga Ge
−0.20 InSb
GaSb
Si Ge
−0.40
0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
(a)Differential transformer
Differential
transformer Torsion wire To the torsion wire
Thermocouple
Water-cooled Primary coil
Pt-20Rh bob brass cylinder Core
MoSi2 (b)Crucible & Bob
heating element To the rod
10
Pt-20Rh crucible Molten slag 2
Crucible
supporter 10
4
Alumina tube 28 32
Gas inlet
Rotating axis Water-cooled 28
(alumina tube) brass cylinder 32 /mm
To the motor
To balance To balance
Sample
Sinker Crucible
Liquid metal
sample
Liquid metal
Buoyant
sample
liquid
(a) (b)
Fig. 2.12. Principle of direct and indirect Archimedean method
one hand, in the former, the sinker with known volume is immersed into
the immiscible liquid. On the other hand, in the latter, liquid or solid sample
contained in a basket is immersed in the liquid with known density. In general,
the former is more popular and is widely used due to the simplicity in the
experimental setup. The sinker should be made of refractory materials at high
temperature such as platinum or tungsten for molten salts, molten slag, or
molten glass, and the ceramics for molten metals. For using the combination
of molten metal and ceramics sinker, heavy weight should be attached to
immerse the sinker by taking account of high density metal.
In the case of using electric balance, as a force, F acting on the balance is
a product of the gravitational acceleration g and the reading of the balance
w, which is corresponding mass, w = F/g. Therefore, the buoyancy force
on the sinker with volume V in a medium with density ρ is given as −ρgV .
The readings of the balance for the sinker in air and liquid, wair and wliq ,
are W − ρair V and W − ρliq V , respectively. W is real mass of the sinker. In
the case of using suspension wire, additional force caused by surface tension,
Fsurf = 2πrγ cos θ appears for immersing in liquid. r, γ, and θ are radius of
wire, surface tension of the liquid and contact angle, respectively. Then, the
following (2.17) is derived:
Δw = wair − wliq = (ρliq − ρair )V − Fsurf /g. (2.17)
Wetting condition of smaller contact angle decreases apparent buoyancy and
vice versa. The radius of suspension wire should be as fine as possible to
decrease the surface tension effect. However, to use very fine wire is not easy
at high temperature, and the surface tension effect by Fsurf is an essential
error source. Therefore, a popular solution is using two sinkers with different
volumes and same radius of suspension wire, which contacts the surface of
liquid to cancel the surface tension effect. Equation (2.18) is obtained for
using large and small sinkers:
Δwlarge − Δwsmall
ρliq − ρair = . (2.18)
Vlarge − Vsmall
For this purpose, replacing two sinkers alternately was frequently used at low
temperature, or for molten salts and slag, which are not so sensitive against at-
mosphere. However, to replace the sinkers is very difficult at high temperatures
and the control of atmosphere is almost impossible. It is a big disadvantage
for oxidizable molten metals. Another way is to use the moving one sinker
with two humps vertically to make pseudo condition like two sinkers [11]. But
the moving one sinker may introduce position error and need complex devices.
Therefore, another apparatus was developed [12]. It contains two sinkers that
are suspended from independent balances and are immersed into the melt si-
multaneously as shown in Fig. 2.13. In this method, replacing the sinkers is
not necessary, then closed system can be achieved to control the atmosphere.
It is a big advantage for high temperature melts such as atmosphere sensitive
molten metals.
2 Viscosity and Density Measurements of High Temperature Melts 31
1 1. Electric balabce
(for small sinker)
2. Electric balance
2
(for large sinker)
3
3. Moving unit (Balance)
4. Pt-Rh ( or Ta) wire
5. Furnace
6. Tungsten weight
7. small sinker
4 8. Large sinker
9. Crucible
10. Zirconium case (BN)
5 11. Alumina tube
6 12. Moving unit (Tube)
13. Thermocouple
7
8
9
10
11
13 12
This method is based on the measurement of the mass with fixed volume by
using a pycnometer with known capacity. The pycnometer is made of refrac-
tory materials such as ceramics, boron nitride, or graphite and machined to
desired shape and dimension. The solid sample with enough volume contained
in the pycnometer is melted and the excess volume more than the capacity of
pycnometer is let brim over. The remaining melt in the pycnometer is brought
out after solidification to measure the mass for determining the density. This
method is adequate for molten metals with large surface tension to avoid the
leakage through the gap between the body and the cap at high temperature,
although the pycnometric method is used as simple device for the measure-
ment of density of liquid or solid at room temperature. This method has an
advantage for oxidizable molten metal due to easy atmosphere control and
has relatively high precision. However, the experimental procedure is cum-
bersome because only one density at certain temperature can be obtained in
one experiment. Several experiments are required to obtain the temperature
dependence for one composition of the sample. And also the experimental han-
dling at high temperature is not necessarily easy. In this method, the choice of
adequate material and the machining precision are important. The capacity
should be calibrated by means of the liquid with well known density such as
mercury at room temperature, and also determined at high temperature using
32 Y. Sato
φ 0.8mm
through hole
12mm
6mm screw
Ar Vac.
A D
B E
F
C
A: Hg reservoir
B: Gold furnace
C: Molten sample
D: Thermocouple
E: Thermostat
F: Standard sample
used in the apparatus shown in the figure to make the detection of the menis-
cus easy. In this apparatus, the material of U-type tube is made of quartz
because of the transparency and the easiness of glass blowing to observe the
meniscus. Furthermore, the advantage of this method is the closed system
to make atmosphere control easy. However, the method is hardly adaptable
to molten metals and silicate melts because the small amount of oxide film
remaining adheres on the quartz tube and makes the observation difficult, and
the molten oxides or other erosive melts attack the quartz. The molten salts
without erosive property for quartz such as molten chlorides are preferable for
the method.
2r
Bubble
capillary h
A bubble
Fig. 2.16. Bubble formed at the end of capillary immersed in the liquid
where Pmax and Psurf are maximum pressure and the pressure caused by the
surface tension. As the hydrostatic pressure is proportional to the immersion
depth, the gradient of the relation between the depth and the pressure is a
product of gravitational acceleration and the density of the melt as shown.
This method is relatively simple and the feature is in easy atmospheric control.
Furthermore, it is adaptable to any high temperature melts such as molten
metals and molten salts by choosing adequate material for the nozzle. How-
ever, it is noted that the measurement may be often interfered by the oxide
film in the case of molten metals due to the use of thin nozzle.
Both these methods are based on the volume measurement of the liquid with
known mass and are mainly adapted to metallic liquids that have high sur-
face tension and make spherical droplet on the substrate or under levitation,
although these methods are also used for surface tension measurement. The
former’s advantage is the relatively simple apparatus and the easy availability
to stable droplet under controlled atmosphere. The latter’s advantage is es-
sentially contactless with container and is also actually limitless in increasing
temperature as a result of containerless measurement.
In the sessile drop method, the substrate is made of refractory material
such as the ceramic, which is chosen for not to react with the melt droplet.
The profile of the droplet as shown in Fig. 2.17a is taken by using a digital
camera and the volume is calculated based on the sectional integration [1].
The problem of this method is uncertainty in the degree of coaxial symmetry
of the droplet and in the magnification of the image recorded from real dimen-
sion. Therefore, two axial shooting must be necessary to correct the volume
calculation and also precise calibration by using reference sample with known
shape and dimension is required. A modified one is called as constrained drop
method, in which a cylindrical crucible with sharp edge is often used as shown
in Fig. 2.17b [15]. This has some advantages in determining the volume. The
2 Viscosity and Density Measurements of High Temperature Melts 35
Liquid metal
specimen
X
Z
Z
X
(a)
Liquid metal
specimen
Edge
Vessel
(b)
Fig. 2.17. Sessile drop and constrained drop for density measurement
sample melt with excess volume is contained in the crucible. The convex shape
on the edge is expected to be coaxial and the calculation of the volume is rel-
atively easy by summating the volume in the crucible and the volume of the
convex part, although the problem in the magnification still remains.
The levitation methods are classified by the levitation force, for example,
electromagenetic, electrostatic, acoustic, and aerodynamic levitations. Former
two methods are available under vacuum. Figure 2.18 shows an example of
electrostatic levitation [16]. In these methods, small sample, typically 2–3 mm
in diameter, is required to maintain near sphere form to make levitation and
calculation easy. The major problem of the methods is uncertainty in the
temperature measurement because only the optical thermometer is available.
Additionally, the rapid oscillation of the droplet and the positional instability
accompanied by the levitation make the volume determination difficult. To
solve the problems, the temperature is usually calibrated by the recalessence
at melting point for pure metals and average of the volumes obtained from
many photo images are used. The superior advantage of the levitation method
of limitless temperature was demonstrated in the density measurement of
molten tungsten by the electrostatic levitation method [16].
36 Y. Sato
Pyrometer
Top Pyrometer
He-Ne electrode YAG Laser
Laser Beam He-Ne
Laser
Rotation
detector
CO2 Laser
beams(3)
Sample
Side
TelePhoto electrodes(4)
Camera1
Oscillation
Position detector Position
sensor sensor
Beam Bottom TelePhoto
Splitter electrode Camera2
2.4 Summary
Viscosity and density are widely recognized to be the fundamental and im-
portant properties for any fluids. However, measurement is more difficult as
the temperature increases. Therefore, the data available are not necessarily
reliable and much effort to obtain precise values was made by developing new
methods and apparatus. As the principles of the measurements are gener-
ally well known, new techniques are desired for obtaining reliable values. The
methods mentioned in this chapter are not necessarily enough, but it is great
pleasure for the author if this chapter is of some help for researchers to study
the viscosity and density of high temperature melts.
References
1. T. Iida, R.I.L. Guthrie, The Physical Properties of Liquid Metals (Clarendon
Press, Oxford, 1988)
2. R.A. Rapp (ed.), Techniques of Metals Research Vol. IV, Physicochemical Mea-
surements in Metals Research Part 2 (Interscience, New York, 1970)
3. R.F. Brooks, A.T. Dinsdale, P. Quested, Meas. Sci. Tech. 16, 354 (2005)
4. Y. Sato, M. Fukasawa, T. Yamamura, Int. J. Thermophys. 18, 1123 (1997)
5. Y. Sato, K. Sugisawa, D.A Oki, T. Yamamura, Meas. Sci. Tech. 16, 363 (2005)
6. S. Sukenaga, N. Saito, K. Kawakami, K. Nakajima, ISIJ Int. 46, 352 (2006)
7. Y. Sato, Y. Matsuzaki, M. Uda, A. Nagatani, T. Yamamura, Electrochemistry
67, 568 (1999)
8. T. Ejima, Y. Sato, S. Yaegashi, T. Kijima, E. Takeuchi, K. Tamai, J. Jpn. Inst.
Metals, 51, 328 (1987)
2 Viscosity and Density Measurements of High Temperature Melts 37
3.1 Introduction
Marangoni flow plays an important role in the heat and mass transport for
highly value-added high-temperature processes, such as crystal growth, weld-
ing, casting, and electron beam melting. For silicon single crystal growth, the
effect of the oscillatory Marangoni flow on the introduction of growth striation
was discussed by Chen and Wilcox for the first time in 1972 [1]. The existence
of the Marangoni flow within molten silicon was proved through microgravity
experiments in space on board a sounding rocket in 1983 by Eyer et al. [2], who
found formation of growth striation in single crystals even under microgravity,
where buoyancy-driven flow was suppressed. To explain the Marangoni effect
at the melt surface, surface tension is essential. Keene [3] discussed the oxy-
gen contamination in the surface tension measurement and recommended the
use of a levitation technique, which is a containerless process and assures the
contamination-free condition from measurement devices. It is well known that
flow direction in the weld pool is dependent on surface contamination and that
this is related to weldability [4, 5]. Flow direction is controlled by the temper-
ature coefficient of surface tension for molten steels; contaminants are oxygen
and sulfur. In the electron beam button melting system, the Marangoni flow
is dominant because of intense heating at the melt surface [5]. In this chapter,
surface tension of high temperature metallic melts is discussed from the view-
point of the Marangoni effect in the value-added high temperature processes,
particularly from the viewpoint of the effect of oxygen and sulfur. Theoretical
treatment for oxygen adsorption is also discussed.
Gas Gas
σ small Δ TH σ large σ large σ small σ large
Temp Low
Liquid Liquid Δ TV
Temp High
a b
Fig. 3.1. Definition of the Marangoni effect; (a) thermocapillary effect and (b) the
classical Marangoni effect [6]
Almost all melt crystal growth processes are accompanied by a free surface
of melts; the existence of a surface tension gradient causes the Marangoni
flow. A typical example for the Marangoni flow is that of float zone crystal
growth. The existence of the Marangoni flow for the float zone growth of
silicon and its effect on crystal quality, that is, the introduction of growth
striation, were predicted for the first time by Chen and Wilcox in 1972 [1].
For the float zone case, the free surface is comparatively larger than volume
and temperature difference between the hottest part of the molten zone and
solid/liquid interfaces caused by the Marangoni flow. This was proven during
silicon crystal growth experiments conducted by Eyer et al. under microgravity
[2]. Since then, a study of the Marangoni flow of molten silicon has been carried
out extensively by the Freiburg group [7,8]. A Japanese group also investigated
the Marangoni flow on an experimental basis, particularly from the viewpoint
of fluid dynamics under microgravity conditions [9]. Lan and Kou carried out
precise work on the Marangoni flow of molten silicon numerically [10].
Since oscillatory Marangoni flow degrades the quality of grown crystals, as
shown in Fig. 3.2 [11], the critical Marangoni number is one of the main con-
cerns of Marangoni flow study. For the float zone configuration, a flow mode
changes from axisymmetric to three-dimensional steady flow at the fist criti-
cal Marangoni number, M ac1 . Above the second critical Marangoni number,
M ac2 , flow becomes oscillatory. Cröll et al. experimentally determined M ac2
using partially confined float zone geometry [7]. For the Czochralski growth
case, the formation of a network pattern has been observed at the melt surface,
and this has been attributed to the existence of a thermal inverse layer [12–14].
A network pattern was reported for the first time for fluoride and oxide melts,
as shown in Fig. 3.3. This inverse layer causes not only formation of a cell
structure, known as the Benard–Rayleigh cell, but also formation of the cell
pattern caused by the Benard–Marangoni–Rayleigh mechanism [15, 16].
For the Czochralski silicon growth, as shown in Fig. 3.4, oxygen O in the
melt is provided through the dissolution of a crucible wall made of fused quartz
(SiO2 ) and evaporates easily as SiO. The concentration of O in the silicon melt
is highest at the crucible wall and then most diluted near the growing crystal.
The amount of adsorbed oxygen is believed to exhibit a similar gradient to
42 T. Hibiya and S. Ozawa
feed
solidified
last melt
zone
grown
crystal
seed
10mm
Fig. 3.2. Growth striation of silicon single crystal grown under microgravity, where
buoyancy flow is suppressed but the Marangoni flow is dominant [11]
Fig. 3.3. Cell structures on the inverse layer of the Czochralski melt [14]
3 Marangoni Flow and Surface Tension of High Temperature Melts 43
Fig. 3.4. Marangoni and buoyancy effects within the Czochralski melt of silicon [6]
Fig. 3.5. Network pattern observed at the Czochralski silicon melt. Arrows show
surface flow direction [17]
that of concentration. Since at the crucible wall there exist three phases, that
is, SiO2 (solid), Si (melt), and O2 (gas), the system is univariant according to
calculation of Gibbs’ number-of-freedom f as shown in 3.3:
f = c − p + 2 = 1, (3.3)
where c is the number of independent components and p is the number of
phases. Concentration of O in the silicon melt must be saturated at the cru-
cible wall at a given temperature under equilibrium condition. Melt surface
temperature is highest at the crucible wall and lowest at the crystal, where
it exhibits the melting temperature of silicon, such as 1,693 K. Thus, there
is a surface tension gradient due to temperature, oxygen concentration, and
oxygen adsorption [6].
As shown in Fig. 3.5, a cellular pattern appears at the surface of a silicon
melt 700 mm in diameter and 200 mm deep during crystal growth; Fig. 3.5
44 T. Hibiya and S. Ozawa
was observed using an IR(infrared)-camera [17]. The cell shape shows hexag-
onal symmetry and is about 100 mm across. Formation of this cell is due to
the existence of a thermal inverse layer several centimeters thick [14]. Within
the thermal inverse layer, the melt is heated from the bottom side and is
cooled at the surface by radiation. Thus, both Rayleigh and Marangoni mech-
anisms should be taken into account in the formation of the cell, that is, a
Rayleigh–Marangoni–Benard cell. The pattern moves from the crucible wall
to the crystal at about 10 mm s−1 . Cell motion was damped when a vertical
magnetic flux density of 50 mT was applied. This suggests that cell flow is
driven by a strong buoyancy flow beneath the inverse layer, whose direction
is inbound. The application of a magnetic field, however, damped this flow
and thus the motion of the cell pattern [6]. At the same time, a magnetic
field also elongated the hexagonal pattern along the radial direction. This
suggests that the surface tension-driven flow, which is caused by difference
of temperature and that of amount of oxygen adsorption (oxygen concentra-
tion) between the crucible wall and the crystal, contributes to the formation
of the hexagonal pattern and that this radial flow cannot be easily damped,
because the radial flow component due to the aforementioned gradient would
be exceedingly strong compared with that in the azimuthal direction. For the
EMCZ (electromagnetic Czochralski) configuration (magnetic flux density of
0.1 T and electric current of 8 A), a flow pattern was highly twisted toward the
azimuthal direction and moves clockwise slowly [6]. This is due to a strong
forced flow induced by the Lorentz force, which resulted from coupling an
applied magnetic field and an electric current passing between an inserted
electrode and a growing crystal.
Recently, Kalalev simulated crystal growth of a 400 mm diameter silicon
ingot considering the stress at the free surface due to the Marangoni effect,
so as to understand and improve the silicon crystal growth process [18].
3.3 Welding
In a TIG (tungsten inert gas) welding process for steel, the shape of a weld
pool is a key factor in determining weldability [4]. Because of a strong tem-
perature difference between the center and rim of the pool surface, Marangoni
flow takes place at the melt surface. The center of the melt surface is heated
by a torch, whereas at the rim a melt coexists with the solid; the temperature
of the melt is lowest at the rim. The temperature coefficient of surface tension
depends on contamination, particularly that due to oxygen and sulfur. Since
a temperature coefficient is negative in a reducing atmosphere, the direction
of the Marangoni flow is outbound; the flow takes place from the center of
the pool to the rim, as shown in Fig. 3.6a [19]. The pool is shallow. However,
if the surface tension displays a positive temperature coefficient due to con-
tamination, the Marangoni flow takes place from the rim to the center of the
pool; the shape of the pool is deep and good weldability is expected, as shown
3 Marangoni Flow and Surface Tension of High Temperature Melts 45
Fig. 3.6. Relationship between temperature coefficient of molten steel and weld
pool shape [19]
in Fig. 3.6b. This positive temperature coefficient of surface tension has been
reported for systems adsorbed with the group-16 elements, such as oxygen,
sulfur, etc. [20–22]. Sulfur also affects the temperature coefficient of surface
tension for stainless steel [4, 5]. Surface tension indicates the maximal value
at a given temperature; this is due to desorption with increase in temperature
(Fig. 3.6c). In such cases flow is complicated within a weld pool.
Lu et al. experimentally and quantitatively studied the effect of O2 and
CO2 in the Ar-based shielding gas on formation of oxide layer, weld pool shape,
and flow structure [23]. In this case, O in the liquid pool is supplied from O2
or CO2 and the melt surface is coated with an oxide layer. Even though the
surface is coated with an oxide layer, Marangoni flow was observed to exist
and to affect the pool shape. When the O2 or CO2 concentration is below
0.6 vol.% in Ar gas, the oxide layer is thin. This thin oxide layer on the pool
peripheral area is easily destroyed and the weld pool surface is exposed to
the shielding gas; thus, Marangoni flow takes place. When the oxygen in the
weld pool is below 100 ppm (O2 or CO2 content in the shielding gas is below
0.2 vol.%), the direction of the Marangoni convection at the quasi-free surface
is outbound; the pool is wide and shallow, as shown Fig. 3.7a. When the
oxygen content is over 100 ppm (O2 or CO2 is over 0.2 vol.%), the direction of
Marangoni convection changes from outward to inward; the weld pool shape
is deep and narrow, as shown in Fig. 3.7b. With a further increase in oxygen
concentration in the melt (O2 or CO2 is over 0.6 vol.%), a thick, rigid oxide
layer is formed at the peripheral area of the liquid pool surface. However, in
the pool center area, inward Marangoni convection still exists. This inward
convection transports the hot melt from the surface to the bottom. As a result,
the weld shape becomes wide again but convex toward the bottom, as shown
in Fig. 3.7c.
46 T. Hibiya and S. Ozawa
Fig. 3.7. The effect of oxygen on a weld pool shape. (a) Oxygen concentration
of Ar-shielding gas is 0.1 vol% (30 ppm in the melt), (b) 0.4 vol% (150 ppm), (c)
0.9 vol% (200 ppm) [23]
Electron beam button melting (EBBM) is used for processing titanium, re-
fractory metals, and refractory alloys for aerospace applications. The process
is also used for preparation of thin films for microelectronic chips. Button-
shaped material c.a. 50 mm in diameter contained within a cold crucible is
melted by using an electron beam. The configuration of the melt is similar to
that of a weld pool. Lee et al. examined the Marangoni effect for the EBBM
process not only from the viewpoint of heat and mass transport in this pro-
cess but also from that of texture of the solidified material, that is, secondary
dendrite arm length [24]. They paid particular attention to the effect of sur-
factant sulfur on Ni-based IN718 alloy; the temperature coefficient of surface
tension is dependent on the amount of sulfur.
3 Marangoni Flow and Surface Tension of High Temperature Melts 47
Fig. 3.8. Frequency of surface oscillation observed on earth (a) and in microgravity
obtained by a parabolic flight of a jet aircraft (b). Multiple frequencies are observed
on earth; m = 0, ±1, and ±2 frequencies split. All oscillations for the l = 2 mode
were degenerated and a single peak was observed in microgravity
∂σ
/Nm−1K−1 = −0.90 + 0.370 ln(1 + 6.62 × 1010 P o2 2 )
1
(3.7)
∂T
1
− 0.387 ln(1 + 8.22 × 109 P o2 2 )
(1, 693 K < T < 1, 773 K, P o2 < P o2 sat ).
Comparing Keene’s review with Mukai’s measurements, it is clear that low
surface tension of ca. 700×10−3 N m−1 (See Fig. 3.9) corresponds to that mea-
sured at high oxygen partial pressure; the oxygen partial pressure of ambient
atmosphere exceeds the equilibrium oxygen partial pressure for SiO2 forma-
tion, and the surface was oxidized. Mukai suggests that surface tension-like
values can be obtained even above the equilibrium oxygen partial pressure for
SiO2 formation. In Fig. 3.10a, kinks correspond to the formation of SiO2 at
the melt surface. Oxygen partial pressure was not controlled quantitatively for
any of the measurements in Fig. 3.9. Hardy and Kingery changed the ambient
atmosphere qualitatively, such as the use of hydrogen and helium [34, 43].
According to Mukai’s measurements, the excess amount of adsorbed oxy-
gen atoms Γ0 was calculated to be 2.1 × 10−6 mol m−2 at 1,693 K and at P o2
of 10−21 − 5 × 10−20 MPa using the following equation, that is, the Gibbs
adsorption isotherm:
ΓO = −2(1/RT )(∂σ/∂ ln P o2 ). (3.8)
3 Marangoni Flow and Surface Tension of High Temperature Melts 51
860
800
1753K
Po2sat =1.32 ×10-19 MPa
780
760
Observed
1693K 1723K
740 1753K 1773K Po2sat = 2.75×10-19 MPa
1773K
720
-26 -24 -22 -20 -18 -16 -14
log(Po2/MPa)
Fig. 3.10. Surface tension of molten silicon as a function of oxygen partial pressures
at various temperatures. Kinks near the P o2 = 10−18 –10−19 MPa correspond to the
formation of SiO2 at the melt surface [56]
0
Temperature Coefficient of Surface Tension:
−0.4
−0.6
1773K
Po2sat = 2.75×10−19 MPa
−0.8
1693K
Po2sat = 1.32×10−20 MPa
−1
−26 −24 −22 −20 −18 −16 −14
log(Po2/MPa)
Fig. 3.12. Oxygen partial pressure dependence of the Marangoni flow velocity ob-
served as tracer particle velocity using X-ray radiography under microgravity. Actual
velocity would be almost one order of magnitude larger than that of tracer particle
velocity [59]
Here, R and T are the gas constant and temperature, respectively. The ex-
cess amount of adsorbed oxygen of ΓO = 2.1 × 10−6 mol m−2 corresponds to
the fact that one out of every ten atoms is oxygen on the melt surface; this
suggests that the oxygen concentration is extremely high at the surface [57],
whereas the bulk concentration of oxygen was calculated to be as low as ca.
0.001 mass% (8.5 × 1017 atoms cm−3 or 18 ppm) [58].
As shown in Fig. 3.11, it should be noted that the temperature coefficient
was never converted from negative to positive, although the absolute value of
temperature coefficient decreased with increase in oxygen partial pressure and
the temperature-coefficient-like value was obtained even above the equilibrium
P o2 for SiO2 formation. This is consistent with the result of Marangoni flow
visualization under microgravity as shown in Fig. 3.12 [59]. Marangoni flow ve-
locity decreased with increase in P o2 and flow stopped at P o2 inlet = 10−2 MPa
due to a SiO2 film formation. Oxygen partial pressure at the inlet condition
of P o2 inlet = 10−2 MPa far exceeds that for equilibrium oxygen partial pres-
sure for SiO2 formation defined at the inlet condition of P o2 inlet ≈ 10−3 MPa.
Through crystal growth experiments under the various Po2 condition, it was
made clear that a surface tension temperature coefficient ∂σ/∂T of less than
0.25 × 10−3 N m−1 K−1 was not realistic; the corresponding oxygen partial
pressure was 4.9 × 10−4 MPa defined at the inlet condition. Above this value
the melt surface was coated with a thin SiO2 film [60].
3 Marangoni Flow and Surface Tension of High Temperature Melts 53
2400
1600
1400
Ar / He (Po2:10−2Pa)
1200
Melting point
s = 0.3896(T−1694)+1423
1000
1500 1600 1700 1800 1900 2000 2100 2200 2300
Temperature / K
Fig. 3.14. Surface tension of molten SUS304 stainless steel measured by the oscil-
lating droplet method using electromagnetic levitation [64]
Fe-18%Cr
2.00
SURFACE TENSION. 1.75
N/m
-S
Cr
8%
-1
Fe
1.50
Fe-S
1.25
00
0.
UR
SU 1
0
LF
180
Fe-S
0.
0
T
2
TE 200
EN
PE 02
MP
RC
0.
ER 260
AT 0
UR
E. 300
.03
K 0
Fig. 3.15. The effect of sulfur concentration on surface tension of molten Fe and
8Ni–18Cr–Fe stainless steel (results of calculations) [5]
the amount of adsorbed oxygen is so high for molten silicon, that is, SiO0.1
for estimated chemical composition at the melt surface, even for low bulk
oxygen concentration; the melt surface is fully coated with SiO2 prior to the
occurrence of temperature coefficient inversion from negative to positive.
3.9 Perspective
Development of high technology society requires products fabricated by highly
value-added high-temperature process, such as crystal growth of semiconduc-
tor, fine casting of turbine blades for jet engines, welding of automobiles,
3 Marangoni Flow and Surface Tension of High Temperature Melts 57
References
1. C.E. Chang, W.R. Wilcox, J. Cryst. Growth 28, 8 (1975)
2. A. Eyer, H. Leiste, R. Nitsche, in Proceeding of Vth European Symp. Materials
and Fluid Sciences in Microgravity, ESA SP-222, 1984, pp 173–182
3. B.J. Keene, Surf. Interface Anal 10, 367 (1987)
4. K.C. Mills, B.J. Keene, R.F. Brooks, A. Shirali, Phil. Trans. R. Soc. Lond. A
356, 911 (1998)
5. M.J. McNallan, T. Debroy, Metall. Trans. B 22B, 557, (1991)
6. T. Hibiya, T. Azami, M. Sumiji, S. Nakamura, Surface Tension-Driven Flow
of Molten Silicon: Its Instability and the Effect of Oxygen, Lecture Notes in
Physics – Interfacial Fluid Dynamics in Physicochemical Phenomena ed. by R.
Narayanan, D. Schwabe (Springer Verlag, Berlin Heidelberg New York, 2003)
pp 131–155
7. A. Cröll, W. Muller-Sebert, K.W. Benz, R. Nitsche, Microgravity Sci. Technol.
3, 204 (1991)
8. P. Dold, A. Cröll, K.W. Benz, J. Cryst. Growth 183, 545 (1998)
9. S. Nakamura, T. Hibiya, K. Kakimoto, N. Imaishi, S. Nishizawa, A. Hirata,
K. Mukai, S. Yoda, T.J. Morita, J. Cryst. Growth 186, 402 (1998)
10. C.W. Lan, S. Kou, J. Cryst. Growth 108, 351 (1991)
11. M. Schweizer, A. Cröll, P. Dold, T. Kaiser, M. Lichtensteiger, K.W. Benz,
J. Cryst. Growth 203, 500 (1999)
12. B. Cockayne, M.P. Gates, J. Mater. Sci. 2, 118 (1967)
58 T. Hibiya and S. Ozawa
13. D.C. Miller, T.L. Pernell, J. Cryst. Growth 57, 253 (1982)
14. A.D.W. Jones, J. Cryst. Growth 63, 70 (1983)
15. K.-W. Yi, K. Kakimoto, M. Eguchi, M. Watanabe, T. Shyo, T. Hibiya, J. Cryst.
Growth 144, 20 (1994)
16. C.J. Jing, M. Kobayashi, T. Tsukada, M. Hozawa, T. Fukuda, N. Imaishi,
K. Shimamura, N. Ichinose, J. Cryst. Growth 252, 550 (2003)
17. T. Azami, S. Nakamura, M. Eguchi, T. Hibiya, J. Cryst. Growth 233, 99 (2001)
18. V.V. Kalalev, J. Cryst. Growth 303, 203 (2007)
19. C.A. Heiple, J.R. Roper, Welding J. 61, 975 (1982)
20. N. Takiuchi, T. Taniguchi, Y. Tanaka, N. Shinozaki, K. Mukai, J. Jpn. Inst.
Metals (in Japanese) 55,180 (1991)
21. B.J. Keene, Int. Mater. Rev. 33, 1 (1988)
22. S.I. Popel, B.V. Tsarevskii, V.V. Pavolv, E.L. Furman, Izv. Akad. Nauk SSSR,
Met. 54, (1975)
23. S. Lu, H. Fujii, K. Nogi, Mater. Sci. Eng. A 380, 290 (2004)
24. P.D. Lee, P.N. Quested, M. McLean, Metall. Trans. B 22B, 1027 (1991)
25. T. Iida, R.I.L. Guthrie, The Physical Properties of Liquid Metals (Oxford
University Press, Oxford, 1988)
26. Lord Rayleigh, Proc. Royal Soc. London 29, 71 (1879)
27. Team TEMPUS, in Materials and Fluids Under Low Gravity, ed. by L. Ratke,
H. Walter and B. Feuerbacher (Springer, Berlin, 1996) pp 223–252
28. D.L. Cummings, D.A. Blackburn, J. Fluid Mech. 224, 395 (1991)
29. I. Egry, H. Giffard, S. Schneider, Meas. Sci. Technol. 16, 426 (2005)
30. T. Koda, S. Ozawa, M. Adachi, S. Shiratori, N. Takenaga, T. Hibiya,
M. Watanabe, Netsubussei 21, 207 (2007)
31. T. Koda, T. Tagawa, S. Ozawa, T. Hibiya, 18th European Conference on Ther-
mophysical Properties, Pau, France, 31 August-4 September 2008
32. H. Fukuyama H. Kobatake, T. Tsukada, S. Awaji, Noncontact Laser Calorime-
try of High Temperature Melts in a Static Magnetic Field, pp 149–171
33. K. Higuchi, K. Kimura, A. Mizuno, M. Watanabe, Y. Katayama,
K. Kuribayashi, Meas. Sci. Technol. 16, 381 (2005)
34. S.C. Hardy, J. Cryst. Growth 69, 456 (1984)
35. Y.V. Naidich, V.M. Perevertailo, P. Obushchak, Rus. J. Phys. Chem. 49, 917
(1975)
36. Y.V. Naidich, V.M. Perevertailo, L.P. Obushchak, Sov. Powder Metall. Met.
Ceram. 149, 403 (1975)
37. V.N. Bobkovskii, V.I. Kostikov, V.Y. Levin, A.S. Tarabanov, Konstr. Mater.
Osn. Gafita (in Russian) 5, 138 (1970)
38. N.K. Dzhemilev, S.I. Popel, B.V. Tsarewskiy, Fiz. Metal. Metalloved. 18, 77
(1964)
39. S.I. Popel, L.M. Shergin, B.B. Tsarevskii, Rus. J. Phys. Chem. 44, 144 (1970)
40. E.S. Levin, P.V. Geld, B.A. Baum, Rus. J. Phys. Chem. 40, 1455 (1966)
41. L.M. Shergin, S.I. Popel, B.A. Baum, AN SSSR Ural’sk Mauchn Tsentr Trudy
Inst. Metall. (in Russian) 25, 52 (1971)
42. V.I. Nizhenko, Y.I. Smirnov, Fiz. Khim. Ganits Razdela Lontakt Faz, ed. by
V.N. Eremenko, (Naukova Dumka, Kiev, 1976) (in Russian) pp 154–159
43. W.D. Kingery, M. Humenik Tr., J. Phys. Chem. 57, 359 (1953)
44. S.V. Lukin, V.I. Zhuchkov, N.A. Vatolin, Y.S. Kozlov, J. Less-Common Met.
67, 407 (1979)
3 Marangoni Flow and Surface Tension of High Temperature Melts 59
45. V.P. Elyutin, V.I. Kostikov, V.Y. Levin, Fiz. Khim. Poverkh Yavleni Rasp., ed.
by V.N. Eremenko, (Naukova Dumka, Kiev, 1971) (in Russian) pp 153–156
46. N.K. Dzhemilev, S.I. Popel, B.V. Tsarevskii, Poverkh Yavlen v Rasp. i Voz iz
Nikh Tverd Faz. Nal’chik 1965 (in Russian) pp 306–311
47. Y.M. Shashkov, T.P. Kolesnikova, Rus. J. Phys. Chem. 37, 747 (1963)
48. F.N. Tavadze, M.G. Kekus, D.V. Khantadze, T.G. Tsverdvase, Poverkh
Yavleniya Rasp., ed. by V.N. Eremenko, (Naukova Dumka, Kiev, 1968) pp
159–162
49. G.P. Khilya, Y.I. Vashchenko, Dopov Akad Nauk Ukr RSR Ser B, 35, 69 (1973)
50. P. Keck, W. van Horn, Phys. Rev. 91, 512 (1953)
51. P.V. Geld, M.S. Petrushevski, Izv AN SSSR Otd, Tkhn Nauk Met i Toplivo 3,
160 (1960)
52. V.I. Kositkov, A.S. Tarabanov, Metod Isssl i Svoistva Gran Raz Kontakt Faz,
AN UKR SSR, ed. by V.N. Eremenko, (Naukova Dumka, 1977) pp 79–90
53. S.-I. Chung, K. Izunome, A. Yokotani, S. Kimura, Jpn. J. Appl. Phys. 34, L631
(1995)
54. M. Przyborowski, T. Hibiya, M. Eguchi, I. Egry, J. Cryst. Growth 151, 60
(1995)
55. X. Huang, S. Togawa , S.-I. Chung, K. Terashima, J. Cryst. Growth 156, 52
(1995)
56. K. Mukai, Z. Yuan, K. Nogi, T. Hibiya, ISIJ Int. 40, S148 (2000)
57. T. Hibiya, S. Nakamura, K. Mukai, Z.-G. Niu, N. Imaishi, S. Nishizawa, S. Yoda,
M. Koyama, Phil. Trans. Royal Soc. London A 356, 899 (1998)
58. H. Hirata, K. Hoshikawa, J. Cryst. Growth 106, 657 (1990)
59. T. Azami, S. Nakamura, T. Hibiya, J. Electrochem. Soc., 148, G185 (2001)
60. T. Hibiya, Y. Asai, M. Sumiji, T. Kojima, Cryst. Res. Technol. 38, 619 (2003)
61. E. Ricci, E. Arato, A. Passerone, P. Costa, Adv. Colloid Interface Sci. 117, 15
(2005)
62. M. Ratto, E. Ricci, E. Arato, J. Cryst. Growth 217, 233 (2000)
63. T. Azami, T. Hibiya, J. Cryst. Growth 233, 417 (2001)
64. S. Ozawa, K. Morohoshi, T, Koda, T. Hibiya, ISIJ Spring Meeting, 2008
65. T. Hibiya, K. Nagafuchi, S. Shiratori, N. Yamane, S. Ozawa, Adv. Space Res.
41, 2107 (2008)
66. G. Bernard, C.H.P. Lupis, Metall. Trans. 2, 2991 (1971)
67. R. Sangiorgi, M.L. Muolo, A. Passerone, Acta Metall. 30, 1597 (1982)
68. T. Yoshikawa, K. Yamamoto, T. Tanaka, K. Morita, J. High Temp. Soc. (in
Japanese) 32, 289 (2006)
4
Diffusion Coefficients of Metallic Melts
Measured by Shear Cell Technique Under
Microgravity and on the Ground
Shinsuke Suzuki
4.1 Introduction
Diffusion coefficients in liquid metals are very important parameters for both
material processing and physics of liquid. However, we have not yet had
enough amounts of accurate diffusion data to discuss the diffusion model in
liquid metals, because several kinds of convection bother measurements. The
sources of convections expected in diffusion experiments are classified into the
followings: (1) natural convection caused by gravity, (2) Marangoni convection
caused by difference of the surface tension, (3) convection induced by mechan-
ical force, etc. As a result of existence of convection, the measured diffusion
coefficients show values higher than the real values and the measured values
are scattered.
Until now several attempts were done to suppress natural convection dur-
ing diffusion experiments. One of them is to perform diffusion experiments
under microgravity conditions where no buoyancy force is generated. In 1980s–
1990s Frohberg et al. performed diffusion experiments using the long capillary
technique as explained later in the space shuttle missions [1–3] and demon-
strated: both measured values of diffusion coefficients obtained under micro-
gravity in the space shuttle; their scattering are smaller than that obtained
on the ground. Furthermore, the temperature dependence of the diffusion
coefficients was found to be proportional to the square of the temperature.
Afterward Yoda et al. [4] and Itami et al. [5] performed concerning experi-
ments in another space mission and obtained results that have consistency
with the results of Frohberg’s group. These results showed the effectiveness
of microgravity conditions to suppress natural convection and are helpful to
discuss diffusion mechanism.
Microgravity experiments are very costly and the number of flight oppor-
tunity is limited. Even if we have a flight opportunity, there still remains
a risk of failure either in rocket launches or in hardware. Therefore, an ex-
perimental technique for accurate diffusion measurements on the ground is
strongly desired. To suppress buoyancy-driven convection, it is known that
62 S. Suzuki
1. Design concept of the shear cell for both reliable performance under μg
conditions and minimization of several kinds of convections
2. Quantification and correction of the shear-induced convection
3. Diffusion experiments with a stable density layering under 1g conditions
4. Diffusion experiments under μg conditions
5. Perspective of future in the field of diffusion measurement in liquid metals
Most of the diffusion experiments have been performed using the long capillary
method, where the liquid diffusion sample is held in a one-dimensional fixed
thin capillary through the whole deffusion experiment. Recently, the shear
cell method is replacing the long capillary as a standard measurement tool. A
diffusion experiment with the shear cell technique is shown in Fig. 4.1, which
is the case of the shear cell specially developed for Foton missions (named
Foton shear cell) [10]. A lot of discs with thin holes pile up and the holes
4 Diffusion Coefficients of Metallic Melts 63
temperature [a.u.]
TL
initial shearing
final shearing
time [a.u.]
graphite felts
F
reservoir PU
cell Rc φd
no.
H
1
2 sample A
3
10 inter-
mediate
cell
sample B
20
reservoir PL PL PL
graphite felts
(a) set-up (b) melting and (c) initial shear (d) diffusion (e) final shear (f) cooling and
homogenization (start of diffusion) (end of diffusion) solidification
Fig. 4.1. Path of a diffusion experiment with the Foton shear cell
form capillaries. By rotating the discs we can form and segment the samples.
Because of this mechanism two kinds of samples are connected only during
the diffusion phase. Therefore, the shear cell technique has several advantages
compared with the long capillary method:
1. No necessity to correct diffusion coefficients during the heating and the
cooling phase
2. No necessity to correct diffusion coefficients for the effect of volume ex-
pansion and shrinkage
3. Possibility of homogenization before the diffusion phase starts (e.g. in case
of monotectic alloys)
4. No influence of segregation during solidification
Especially the advantage (4) is the most important to expand the variation
of sample selection into multicomponent systems, which usually show the
segregation problem and are necessary to realize the stable density layering
in ground experiments.
64 S. Suzuki
In spite of these advantages, after the invention of the shear cell technique
by Nachtrieb et al. [15], the amount of the measurement results with using
the shear cell technique is not so large because its complicated structure can
cause experimental failure or additional convection, for example, misalign-
ment, leakage of liquid samples, not smooth rotation of cells, shear-induced
convection, averaging effect inside the cells, etc. From these reasons, several
attempts were done to improve the reliability and to avoid the error sources.
One of them is the shear cell developed for the diffusion experiments onboard
ISS by JAXA [16]. Another is the Foton shear cell with which space experi-
ments have been already successfully done [14]. The following sections explain
the development of the shear cell technique with showing the case of the Foton
shear cell as an example.
Convection takes place during shearing because of the friction between the
liquid sample in a cell and the moving cell wall. This convection is called
shear convection. To minimize the shear convection, the cell size, the shearing
mode, the rotation method, and the rotation speed should be optimized. There
are several reports about the effect of these parameters on the intensity of the
shear convection in a diffusion experiment using the shear cell technique.
Arnold et al. [17] and Oda et al. [18] investigated the effect of the shear
convection by numerical simulations. Arnold et al. concluded that the aspect
ratio of the cell length H (Fig. 4.1f) and the diameter d (Fig. 4.1f) of a capillary
is important; for H/d > 3, the measurement is unaffected by the shearing
action; for 1 < H/d < 3, convective mixing exists but the disturbance is small
enough to be corrected well; for H/d < 1, convective mixing is contributing
strongly. If the cell length H is longer, the averaging effect of AAS(mentioned
later) will be larger, but can be corrected. Considering these results and the
restriction to minimize the whole shear cell length and weight for the flights,
the Foton shear cell employed the cell size of H = 3 mm, d = φ1.5 mm.
The shearing mode is also critical for the shear convection. Mathiak et al.
[19] investigated this problem by numerical simulations. Griesche et al. [20]
observed the shear convection of colored water in a transparent model of shear
cell made of PlexiglasTM in parabolic flight experiments. They showed that
the initial shearing mode with an intermediate cell has a smaller penetration
depth than the mode with two half columns. There was no large difference
between the comb shearing and the helical shearing for the final shearing
process. Considering these results, the Foton shear cell employs the mode
with an intermediate cell as the initial shearing and the comb shearing as
the final shearing mode. If it is needed, the initial shear mode with two half
columns is also possible by changing the set-up of the Foton shear cell.
The smoothness of the rotation is also critical to avoid vibration and con-
vection. From this point of view a stepping motor, which causes vibration,
is not preferable, although it is advantageous for the positioning of the cells.
4 Diffusion Coefficients of Metallic Melts 65
The Foton shear cell employed a DC-motor with a specially designed amplifier
circuit that controls the tachometer voltage (rotation speed). Positioning is
done with software that detects the rapid increase of the motor current at the
end of a shear process and then switches off the motor. The discs are pressed
together by elastically compressed graphite felts. The compression force F in
Fig. 4.1a was set to 2 N, which is large enough to avoid leakage of liquid sam-
ple as mentioned later. With this compression force, the friction force between
the discs is small enough to guarantee a smooth rotation of the discs/cells.
The intensity of the shear convection is proportional to the shear speed
[18]. On the other hand, this kind of motor regulation system has a low torque
at a low speed. Thus the shearing speed of the discs was set between 0.5 and
1 mm s−1 .
In shear cell experiments, misalignment of cells and scratches on the cell wall
cause free surfaces on the liquid diffusion samples. If the free surface has a
large gradient of surface tension, Marangoni convection takes place, which
also disturbs the concentration profile in diffusion experiments [21]. To avoid
this problem and a leakage of liquid sample materials, the graphite parts have
been machined with a tolerance of about 10 μm.
Another technique for the minimization of free surfaces is setting reser-
voirs with sufficient high pressure on both ends of the capillary (Fig. 4.1a).
At the set-up the volume of solid samples in the capillaries must be about
20% smaller than the capillary capacity; otherwise the expanded sample ma-
terial opens the gap between the discs and molten sample material leaks out
through the gap. The material in the reservoir should have identical composi-
tions to that in the capillary. Pressure on the material is given by elastically
compressed graphite felts in the reservoir. At the liquidus temperature of the
sample, the compressed elastic graphite felts press onto the molten metal and
fill the capillary completely. To minimize free surfaces, it is required to keep a
sufficient pressure PL (Fig. 4.1) on the liquid in the capillary during the diffu-
sion phase. Hence, the diameter of the reservoir must be much larger than that
of the capillary so that there is only a small reduction in the spring force of the
felts. At the initial shearing one of the reservoirs is separated from the cap-
illary, but the other remains in the initial position. The curvature radius RC
(Fig. 4.1d) of the free surface at the corner in the capillary can be estimated
by γ/PL , where γ is the surface tension of the molten sample. A too high pres-
sure could cause leakage of the molten sample from the capillary. Considering
also the requirement of the smooth rotation as mentioned earlier, the reservoir
pressures were set at 10–30 kPa in case of the Foton shear cell. RC is estimated
as 20–60 μm, for example, in case of Sn at 300◦ C, γ = 560 mN m−1 [22].
66 S. Suzuki
The materials of the main parts of the shear cell should not react with the melt,
be finely machinable, heat proof, and be lubricant. The shear cell of JAXA is
made of glassy carbon [16] and the Foton shear cell [10] is made of graphite. A
shear cell unit usually has several diffusion capillaries so that several parallel
experiments can be performed simultaneously, for example, three capillaries
in a set of JAXA shear cell and four capillaries in a set of Foton shear cell.
In case of the Foton shear cell, the whole shear cell consists of an outer fixed
tube, an inner rotatable tube, 20 discs, two reservoir units, and an axle. Each
capillary inside the discs has 60 mm length and 1.5 mm diameter. The shear
cell is heated up by a Mo-wire isolated by ceramic tubes and mounted around
the outer fixed tube. Ceramic fiber sheets are wound around the outer fixed
tube for thermal isolation and for damping of mechanical shocks at launching.
Six shear cells can be operated in six separate chambers in the AGAT-furnace
(Foton-M2 mission). The holders on both ends of the shear cell are fixing the
shear cell inside the chamber. The motor rotates the inner tube and thereby
every second disc (comb shearing).
By varying the combination of the samples and the position of the interme-
diate cell, the Foton shear cell can perform several types of diffusion experi-
ments:
1. Interdiffusion (diffusion between semi-infinite sources) (Fig. 4.2a)
2. Thick layer diffusion (diffusion into double semi-infinite liquids from a
planar source) (Fig. 4.2b)
3. Thick layer diffusion (diffusion into a single semi-infinite liquid from a
planar source) (Fig. 4.2c)
Each sample part is weighed before insertion in order to check filling of the
capillary after the experiment. To investigate the reproducibility of measure-
ments, all the four capillaries (named A, B, C, and D) are filled with samples
the same composition from the same charge and the experiments are simul-
taneously performed.
An example of experimental procedures is shown in Fig. 4.1, which was
used in the AGAT-furnace facility (Foton-M2 mission). The 1g and low-g
(parabolic flight) diffusion experiments are performed by using the Foton shear
cell installed in a vacuum chamber.
After evacuating the chamber the furnace is heated up to the diffusion
temperature. After a 1-h homogenization phase, the initial shear process is
performed at a constant shear speed v0 to start the diffusion process. The
4 Diffusion Coefficients of Metallic Melts 67
0h
0
0h
x0
x
x
x
: sample A
: sample B
Fig. 4.2. Possible diffusion types with the Foton shear cell. (a) Interdiffusion (dif-
fusion between semi-infinite sources), (b) thick layer diffusion (diffusion into double
semi-infinite liquids from a planar source), and (c) thick layer diffusion (diffusion
into a single semi-infinite liquid from a planar source)
(a)
c0
t=0
after diffusion
x0
(b)
c0 t=0
after diffusion
2h
Fig. 4.4. Measured mean square diffusion depth for different times and tempera-
tures. The slope gives 2D and the offset represents the mean square diffusion depth
caused by the convection and the averaging effect [25]. (a) shows the magnified short
time range of (b)
[24]. The diffusion temperature was 275◦ C and the diffusion time tdiff was
14 s. The obtained value of x2 meas is shown in Fig. 4.4.
The time dependence of the measured mean square diffusion depth has been
reported [26, 27]. The additional transport except the pure diffusion process
can be determined by extrapolation of the measured mean square diffusion
depth as a function of the diffusion time to t = 0.
The Foton-shear cell was also investigated by this method, using a sta-
ble density layering under 1g-conditions [25]. The configuration was the thick
4 Diffusion Coefficients of Metallic Melts 71
layer diffusion (Fig. 4.2c), where the sample A was SnBi3 or SnIn10 and the
sample B was Sn. The diffusion experiments were performed at several differ-
ent temperatures for several different diffusion times. For each condition four
parallel experiments were performed simultaneously. After the concentration
measurements, x2 meas was determined (Fig. 4.4). The extrapolation from the
measured values to the y-intercept gave x2 add = 1 ∼ 5 × 10−6 m2 . Half of
the inclination of the line represents the diffusion coefficient at each temper-
ature, which corresponds to the result determined using another correction
method (4.6).
The obtained x2 meas is theoretically the sum of the values of the mean square
diffusion depths of different origins under the assumption that the additional
mass transports can be also interpreted as additional diffusion processes:
x2 meas = x2 diff + x2 add = x2 diff + x2 aver + x2 shear . (4.4)
Here x2 diff is the contribution from diffusion (=2Dtdiff ) and x2 aver is the
contribution from the mixing effect inside a single cell. This mixing is induced
by the convection after the final shear process, by the segregation effects during
the solidification and finally by the integrating analysis of AAS or ICP. x2 shear
is the contribution from the initial and the final shear convection and can
be calculated by subtraction of x2 diff and x2 aver from x2 meas . If we consider
x2 shear as a quasi-prolongation of a diffusion time, the extended time tshear is
calculated by
x2 shear
tshear = . (4.5)
2D
For the calculations, diffusion coefficients of D = 2.1 × 10−9 m2 s−1 (250◦ C)
and 2.3 × 10−9 m2 s−1 (275◦ C) were used according to the μg-reference data
in the D1-mission [2].
As a result, the shear convection turned out to be small. The mean square
diffusion depths caused by the shear convection were in the order of 10−7 m2 ,
which is much smaller than 1% of the typical value x2 diff ≈ 10−4 m2 in a
standard diffusion experiment using the Foton shear cell. Nevertheless, this
value can be used for the correction of diffusion experiments.
1 2
D= x meas − x2 shear − x2 aver , (4.6)
2tdiff
1 H2
= x meas − x shear −
2 2 ,
2tdiff 12
where x2 meas is the result of any long time diffusion experiment and x2 shear
is the result of the corresponding short time diffusion experiment, that is,
at the same temperature and in the same diffusion system under the same
gravity conditions. Here x2 aver was approximated as H 2 /12 and depends only
on the cell geometry under the assumption that tdiff is sufficiently long [12].
For the Foton shear cell, H (the thickness of a disc) is 3 mm and x2 aver is
7.5 × 10−7 m2 .
In the present investigation of a short time experiment, x2 shear could be
determined because x2 diff could be calculated from 2Dtdiff and D was known
from μg-experiments. Nevertheless, we can use (4.6) in the following way by
denoting the short time data with a suffix 0. Then from (4.4) we have
H2
x2 shear/0 = x2 meas/0 − 2Dtdiff/0 − . (4.7)
12
Inserting the results in (4.6) in the generally applicable form
x2 meas − x2 meas/0
D= . (4.8)
2 tdiff − tdiff/0
This equation allows to eliminate the disturbing shear convection and the aver-
aging effect of shear cell diffusion experiments by the additional performance
of a short time diffusion experiment (zero-experiment). This is advisable if
an accuracy of <2–3% is envisaged. If a higher deviation is accepted, it will
be normally not necessary to use the correction with (4.8). Then (4.6) with
x2 shear ≈ 0 could be sufficient.
4.6.1 Experimental
To realize the stable density layering, the sample materials should be selected
so that the density increases monotonously parallel to the gravity vector dur-
ing the whole diffusion time. It must be carefully checked before the diffusion
that the density gradient correspond to the concentration gradient during the
diffusion time and also increases monotonously parallel to the gravity vector.
Until now, experiments using the samples of PbAg–Pb, PbGa–Pb, SnBi–
Sn, SnIn–Sn, InSn–In, SnBiIn–Sn, SnSb–Sn, and AlNi–Al were performed
successfully. The experimental conditions are listed in Table 4.1.
4 Diffusion Coefficients of Metallic Melts 73
Table 4.1. 1g-diffusion experiments performed during the Foton projects and ob-
tained mean square diffusion depths and diffusion coefficients
Fig. 4.5. Bi concentration profiles in four parallel experiments with SnBi3 –Sn (at
300◦ C for 8 h) with the error of ΔC/C = 2% dependent on the performance of the
AAS equipment [12]
Fig. 4.6. “Linear plot” of Bi concentration profiles in four parallel experiments with
SnBi3 –Sn (at 300◦ C for 8 h) [12]
standard deviation ΔD among the four parallel diffusion experiments. Since
the ΔD -values only represent the reproducibility under nearly equal external
conditions, all statistical errors should be included there except those from
the temperature measurement. Hence, for the total statistical error ΔD in
Table 4.1, a standard temperature deviation was additionally included on the
basis of the relation D = AT 2 by the equation
N
2
ΔD 1 Dn − D
=
2
+ (2σtemp ) , (4.10)
D N − 1 n=1 D
where the number of the capillaries N is four in the Foton shear cell, Dn is the
diffusion coefficient individually measured for the nth capillary, and D is the
average value among the Dn -values and the standard temperature deviation
σtemp is estimated as 1%. Thus, we arrive at relative total statistical errors of
ΔD/D = 2–5%, which is indeed lower than that in the Foton-12 mission, as
described in Table 4.1.
Garandet et al. [28] analyzed all expected error sources in the diffusion
measurements in Foton-12. They assumed a maximum error ΔD/Dμg of 6%
for the SnBi1.5 –Sn interdiffusion experiment at 300◦ C, 10% for the SnIn1 –Sn
interdiffusion experiment at 300◦ C, and 7.5% for the AlNi1.5 –Al interdiffusion
experiment at 700◦ C. It seems remarkable that the total statistic errors in the
1g-experiments are less than the error values estimated in [28], also for the
concentration measurements, as can be seen from the reproducibility. It was
considered that the reduction of the total statistic errors on the ground is the
76 S. Suzuki
result of the upgrade of the shear cell from the Foton-12 model. From these
new investigations, no evidence of disturbance of convective flow was detected
in the 1g-measurements.
Only the experiments of SnIn10 –Sn at 800◦ C have a relative large total
statistical error of 7.2%. This has not yet been explained clearly. There might
have been a malfunction of the AAS procedure.
(3) Comparison with reference data
Even if the criteria mentioned earlier are satisfied, there still remains possibil-
ity that the diffusion measurement was bothered by convection. The experi-
ments of PbAg–Pb demonstrated that the density gradient antiparallel to the
gravity vector also delivered high reproducibility and concentration curves,
which agree well with the fitting function [14]. Therefore, it is necessary to
compare the obtained diffusion coefficient with reference data, such as that
obtained in a magnetic field or under microgravity.
Although it seems a paradox that the 1g-experiments requires reference
data from μg-experiments, if we have a few μg-data, it is very helpful to check
much more amount of 1g-data.
For the diffusion of Bi in Sn, and In in Sn, the results of 1g-measurement
in a magnetic field are available. The reference values are shown in Table 4.1.
This comparison shows that the 1g values using stable density layering agree
well with both the μg-reference data and the results of 1g-measurements in a
magnetic field.
The results obtained using stable density layering indicate that buoyancy con-
vection was largely suppressed in these 1g-experiments by using the condition
dρ −
e→ →
−
x · g > 0, (4.11)
dx
where − e→
x is the unit vector in the x-direction. Strong buoyancy convection is
normally caused by a density gradient antiparallel to the g-vector, that is, the
left side in (4.11) is negative. Since
dρ − → ∂ρ ∂c − → ∂ρ ∂T −
e→ −
x· g = e→ −
x· g + e→ →
−
x· g (4.12)
dx ∂c ∂x ∂T ∂x
and ∂ρ/∂T is normally negative, the condition for dρ/dx can be met by an
upwards temperature gradient ∂T /∂x and/or a positive contribution from the
solutal term in (4.12). The 1g-experiments shown in this section are designed
to meet both conditions.
Since it is difficult to arrange the temperature gradient exactly antiparallel
to the g-vector, we have to consider horizontal temperature gradients, say
∂T /∂y. Such gradients produce forces that try to start convection rolls, too.
Then it can be shown that there are always convection rolls, but a sufficient
4 Diffusion Coefficients of Metallic Melts 77
negative solutal density gradient can strongly reduce the velocity of the rolls so
that the influence on the determination of the real atomic diffusion coefficient
is below a certain required level. It depends on the material parameters in
(4.12) whether the condition can be fulfilled at least in the main part of the
diffusion zone. It can normally not be met in the capillary parts away from
the diffusion zone, since there c is constant and dT /dx should be small.
Nowadays various kinds of facilities are available for utilization of low grav-
ity or microgravity environments, such as drop towers/shafts, airplanes for
parabolic flight, sounding rockets, recovery satellites, space shuttles, and the
space station ISS. Duration of reduced gravity, cost, preparation time, success
rate, and quality of gravity are different from each other.
The duration of usual diffusion experiments is from several dozens of min-
utes to several hours. Therefore, until now diffusion experiments under micro-
gravity were carried out in the space shuttles or in recovery satellites that fly
about two weeks in the orbit. Some experiments in the space station MIR [30]
and in sounding rockets have been also reported [31]. The diffusion experi-
ments under microgravity are listed up in [16, 32]. Now the diffusion experi-
ments are prepared for JEM (Japanese Experiment Module) on ISS [16]. It is
also pointed out that the residual gravity and the g-jitter also bother diffusion
measurements. We have to pay attention to the quality of gravity [30].
Recently μg-diffusion experiments were carried out in the mission Foton-
M2 [14, 25, 33]. The following sections introduce several parts of the diffusion
experiments in Foton-M2 and concerning experiments.
Fig. 4.7. Mission procedure of Foton-M2. The pictures were provided by ESA
data were collected from the capsule and analyzed, as explained later. Ex-
perimental procedure is same as described in Sect. 4.3. Figure 4.8 shows Sb
concentration profiles in four parallel 1g-experiments and in a μg-experiment
with SnSb5 –Sn (at 630◦ C for 12,600 s), which demonstrate agreement between
1g and μg-experiments.
4 Diffusion Coefficients of Metallic Melts 79
10
Sn-SnIn
In in Sn(1g) [9]
8 Sn-SnBi
In in Sn Bi in Sn(1g) [11]
Bi in Sn(µ g, Foton-M2) [25]
6
Previous reference data
Sn-SnInBi
4 In in Sn(1g, magnetic field) [6]
In in Sn(µ g, Foton-12) [28]
Bi in Sn(1g, magnetic field) [6]
2 Bi in Sn(µ g, Foton-12) [28]
Bi in Sn
0
0 200 400 600 800 1000 1200
temperature T[K]
D=1.4x10 −15 T
2.2
10
Dx10 9 [m2/s]
: Pb in Pb (μg, Foton−M2)
6 Reference with long capillary[3]:
5 : Fitting curve for Pb in Pb (μg, D2)
4
3
2
1
0
0 200 400 600 800 1000
T[K]
Fig. 4.11. Temperature dependence of diffusion coefficients of Ga and Pb207 in Pb
4.9 Perspectives
Since Frohberg’s successful experiments in the space shuttle, it has been be-
lieved that the space environment is the best condition for diffusion measure-
ments. However, the measurement technique for 1g-experiments with a stable
density layering was developed and demonstrated high reliability. Now we have
to change the concept about the reason why we need space environments.
Nowadays the aims of diffusion experiments in the space can be classified
into two categories [11]; one is the experiments that can be performed also on
the ground and the data will be used as reference values to check the reliability
of the ground experiments, and the other is that the experiments that cannot
be performed on the ground, for example, a diffusion pair that does not have
a stable density layering.
If exact diffusion measurements are required, the combination of μg-
conditions and the shear cell is always helpful. But both the shear cell and
μg-conditions are not always available and necessary. Here the diffusion types
were classified as shown in Table 4.2 that shows in which cases the shear cell
4 Diffusion Coefficients of Metallic Melts 81
a
SC shear cell
b
LC long capillary
can be substituted by the classical long capillary method or in which cases dif-
fusion experiments can be performed also under 1g-conditions. If a sufficient
stabilization is possible by density gradients, there is the possibility to per-
form impurity diffusion, interdiffusion in binary alloys, and the combination
of inter and self-diffusion under 1g conditions. Only in case of no segregation
we can use the long capillary, but such a case is rare and the shear cell is
usually needed.
However, if no sufficient stabilization is possible, we need μg-conditions, for
example, in the case of self-diffusion in elements or impurity diffusion at low
concentration. In the case of interdiffusion with more than three components,
because of drag effects it is not possible to obtain a stable density layering,
and hence the shear cell and μg-conditions are necessary. Usual alloys for
commercial use have more than three components. Also if reliable reference
data are needed, this is up to now only by μg-diffusion data.
A more exact quantitative discussion about the necessary density layering
arrangement to avoid the buoyancy convection is a future subject. More exact
thermodynamic data of the samples are required for the discussion.
82 S. Suzuki
4.10 Summary
For accurate diffusion experiments both under μg and 1g, the highly com-
pleted shear cell should be developed and optimized for considering the min-
imization of shear convection, minimization of free surfaces, and a reliable
operation. The short time diffusion experiments under low-g and 1g condi-
tions and the measurement of the time dependence of x2 meas are helpful to
evaluate the mass transport caused by the shear convection quantitatively. To
determine the diffusion coefficient, the correction method proposed here can
be applied.
The 1g-experiments of In in Sn, Bi in Sn, Sn in In, Ga in Pb, Ag in Pb,
and Ni in Al with the Foton shear cell using a stable density layering showed
a good agreement between the concentration profile and the fitting function,
a high reproducibility with a small deviation among parallel experiments, and
a good agreement with μg-reference data. Thus the high reliability of the 1g
D-values is demonstrated.
The μg-experiments in Foton-M2 were performed successfully. It is also
a fact that the apparatuses for μg-experiments were graded up and highly
completed through a lot of the 1g-experiments.
On the basis of these results, the diffusion types were classified according
to the reliability of 1g-measurement. It is expected that the experimental
method in this study can deliver a large amount of exact diffusion data.
Acknowledgement
The author thank the Germany Federal Ministry of Education and Re-
search, BMBF/ German Aerospace Center (DLR) for the financial support
for the Foton-missions under national registration numbers 50WM0048 and
50WM0348, and ESA for giving the opportunity to perform the space and the
parabolic flight experiments. Collaboration with Prof. G. Frohberg and
Dr. K.-H. Kraatz (TU Berlin) for the development of the Foton-shear cell,
1g-experiments with a stable density layering and the Foton experiments are
gratefully acknowledged. Collaboration and discussion with Dr. A. Griesche
(DLR), Dr. G. Müller-Vogt, Dr. R. Rosu (Univ. Karlsruhe) and Dr. T. Itami
(Hokkaido Univ.), and Dr. T. Masaki (JAXA, present Shibaura Inst. Tech.)
are gratefully acknowledged.
4 Diffusion Coefficients of Metallic Melts 83
References
1. G. Frohberg, in Fluid Sciences and Materials Sciences in Space, ed. by H.U.
Walter. Thermophysical Properties (Springer, Berlin, 1987), p. 159
2. G. Frohberg, K-H. Kraatz, H. Wever, A. Lodding, H. Odelius, Defect Diffus.
Forum 66, 295 (1989)
3. G. Frohberg, in Scientific Results of the German Spacelab Mission D-2,
(WPF/DLR, Cologne, 1995), pp. 275–287
4. S. Yoda, H. Oda, T. Oida, T. Masaki, M. Kaneko, K. Higashino, J. Jpn. Soc.
Micrograv. Appl. 16, 111 (1999)
5. T. Itami, H. Aoki, M. Kaneko, M. Uchida, A. Shisa, S. Amano, O. Odawara,
T. Masaki, H. Oda, T. Ooida, S. Yoda, J. Jpn. Soc. Micrograv. Appl. 15, 225
(1998)
6. V. Botton, P. Lehmann, R. Bolcato, R. Moreau, Energy Conversion Manage.
43, 409 (2002)
7. G. Mathiak, A. Griesche, K-H. Kraatz, G. Frohberg, J. Non-Crystalline Solids
205–207, 412 (1996)
8. Y. Inatomi, T. Miyake, K. Kuribayashi, J. Jpn. Soc. Micrograv. Appl. 20, 160
(2003)
9. S. Suzuki, K-H. Kraatz, G. Frohberg, Ann. N. Y. Acad. Sci. 1027, 169 (2004)
10. S. Suzuki, K-H. Kraatz, G. Frohberg, Micrograv. Sci. Technol. 16-1, 120 (2005)
11. S. Suzuki, K-H. Kraatz, G. Frohberg, J. Jpn. Soc. Micrograv. Appl. 22, 165
(2005)
12. S. Suzuki, K-H. Kraatz, G. Frohberg, R. Roşu, G. Müller-Vogt, J. Non-
Crystalline Solids 357, 3300 (2007)
13. S. Suzuki, K-H. Kraatz, G. Frohberg, R. Roşu, W. Wendl, G. Müller-Vogt, Ann.
N. Y. Acad. Sci. 1077, 380 (2006)
14. S. Suzuki, K-H. Kraatz, G. Frohberg, Micrograv. Sci. Technol. 18(3/4), 82
(2006)
15. N.H. Nachtrieb, J. Petit, J. Chem. Phys. 24, 746 (1956)
16. T. Masaki, T. Fukazawa, S. Matsumoto, T. Itami, S. Yoda, Meas. Sci. Technol.
16, 327 (2005)
17. W. Arnold, D. Matthiesen, J. Electrochem. Soc. 142, 433 (1995)
18. H. Oda, S. Yoda, T. Nakamura, T. Masaki, N. Koshikawa, S. Matsumoto,
A. Tanji, M. Kaneko, Y. Arai, K. Goto, N. Tateiwa, J. Jpn. Soc. Micrograv.
Appl. 16, 33 (1999)
19. G. Mathiak, G. Frohberg, W.A. Arnold, Proceedings of 2nd European Sympo-
sium on Fluids in Space, Naples, Italy, April, p. 369 (1996)
20. A. Griesche, K-H. Kraatz, G. Frohberg, G. Mathiak, R. Willnecker, Proceedings
of 1st International Symposium on Microgravity Research and Applications in
Physical Science and Biotechnology, ESA SP-454, Sorrento, Italy, September,
pp. 497–503 (2000)
21. H. Müller, G. Müller-Vogt, Cryst. Res. Technol. 38, 707 (2003)
22. C.J. Smithells, in Metals Reference Book, vol. 2, 3rd edn. (Elsevier, Amsterdam,
2004), p. 699
23. S. Suzuki, K-H. Kraatz, A. Griesche, G. Frohberg, Micrograv. Sci. Technol.
15(1), 127 (2005)
24. P-E.Berthou, R. Tougas, Metal. Trans. 1, 2978 (1970)
25. S. Suzuki, K-H. Kraatz, G. Frohberg, Micrograv. Sci. Technol. 18(1), 155 (2006)
84 S. Suzuki
5.1 Introduction
The importance of heat transfer properties such as thermal conductivity or
thermal diffusivity of various materials at high temperature is strongly em-
phasized, in parallel with recent progress in surface technology for several
electronic devices. Such importance has been well recognized in many pyro-
metallurgical processes related to plant design and accurate control of contin-
uous casting in steelmaking. For example, heat transfer properties of molten
salts are essential to design applications to heat transfer fluids for fusion re-
actors, breeder reactors, and thermal energy storage systems. Then, thermal
property data of molten salts with sufficient reliability are strongly required
to select an optimum composition of salt mixture for the desired condition [1].
We also need thermal property data of molten iron at elevated temperature
and continuous casting powder melts consisting of various oxide components;
SiO2 , CaO, MgO, Al2 O3 , etc. for further improving the present continuous
casting process for steel [2].
In producing single crystals supplied for devices of semiconductor com-
pounds such as GaAs and GaP, using Czochralski method, the components
of high vapor pressure of P and As are likely to diffuse from the master melt,
causing the original compositions to vary. To reduce such trouble, boron oxide
melts have widely been employed as liquid capsules to encase the semiconduc-
tor master melt [3]. It is necessary to minimize the temperature gradient in
the melt by accurate temperature control for producing high quality single-
crystals with a low dislocation density. Although the thermal diffusivity of a
liquid capsule material is one of the important properties, no report is avail-
able on the value of thermal diffusivity of molten boron oxide within the best
knowledge of the present authors.
On the other hand, thermal diffusivity measurements of high temperature
melts are still far from complete in various cases, arising mainly from onset
convective heat flow, heat leakage to the container, and mixed contributions
of radiative and conductive heat transfer components. For this reason, most
86 H. Shibata et al.
values for melts compiled in the TPRC handbook [4] are estimated from mea-
sured values of respective substances in the solid state. Therefore, some differ-
ences are frequently observed between the simulated results and experimental
data.
The laser flash method (frequently referred to as laser pulse method), first
proposed by Parker and his colleagues [5], has been widely employed as one of
the most versatile techniques for measuring thermal diffusivity of various sub-
stances at high temperature. Particularly, the laser flash method with a three
layered cell was successfully developed for determining the thermal diffusivity
values of high temperature melts of both oxides and metals [6–8].
The main purpose of this chapter is to describe the recent developments of
the laser flash method and its successful application to the thermal diffusiv-
ity measurements of oxide and metallic melts at temperatures above 1,000 K,
together with some selected examples of the thermal diffusivity values by
precisely separating the radiative component from measured temperature re-
sponse signals.
Table 5.1. A comparison for steady state measurement with transient state mea-
surement for thermal conductivity, λ, and thermal diffusivity, α [9]
Pulse laser
Al2O3 pipe
Micrometer
Al2O3 pipe
Fe2O3 powder
Pt heater
Pt crucible
Sample liquid
Pt crucible
Al2O3 pipe
I.R. Infrared detector
Mirror
Si lens
Neodymium Concave
glass laser lens
Vaccum
chamber
InSb infrared
detector
Vacuum
pump
Fig. 5.2. Photograph of the laser flash apparatus set at Tohoku University
Micrometer
Mullite tube
T(l2) T(l2+Δl)
t t
(a) (b)
Fig. 5.3. The cell system for oxide melts in the three-layered laser flash method
90 H. Shibata et al.
the first and third layers (platinum in the present case) and the thickness of
all three layers are provided.
Let us consider an infinite slab as shown in Fig. 5.3 consisting of three lay-
ers. At the initial time region, the temperature response of the back surface of
the third layer can be described as in (5.5). Then, we can obtain the following
useful relations:
√
l2 / α2 = 2 −β(l2 ) − η1 − η3 , (5.6)
√
(l2 + l)/ α2 = 2 −β(l2 +l) − η1 − η3 , (5.7)
Sample thickness: l 2
Sample thickness: l2 + Δl
a2
0
0 l2
(a)
Tmax
0
0 ta tb t
(b)
ln(T t1/2)
1/ta 1/tb
1/t
Fig. 5.5. Principle of data processing from experimental output of T vs. t. (a)
Determination of ta and tb in the range 0.2–0.4Tmax . (b) Plot of l/t vs. ln(T t1/2 ) to
determine slopes β1 and β2
Pulse laser
Hole
Fig. 5.6. The cell system for metallic melts in the three-layered laser flash
method [13]
consists of two sapphire plates and ceramic (alumina or mullite) tube. In other
words, alumina or mullite tube is sandwiched between two sapphire plates.
Sapphire is known to be transparent for energies produced from Nd glass or
ruby laser source as well as energies of infrared ray. The advantage of this
sample cell is to vary the sample thickness easily by changing the thickness
of ceramics tube, depending upon thermophysical properties of the desired
sample. For reducing the volume expansion when the metal sample is melted
and residual gas inside the sample, four small holes with size of 0.5 mm in
diameter are made in the upper sapphire plate. As shown in Fig. 5.6, this three-
layered cell is fixed by the carbon corn. Any leakage of molten metal from this
cell system was not detected, probably because of relatively low wettability
between cell materials (alumina or mullite) and molten metal sample. The
following comments may be worthy of note for further application of this
devised cell.
Two kinds of sample cell systems were tested. One consists of a mullite
tube (thickness, 0.8–1.5 mm; inner diameter, 9 mm; outer diameter, 13 mm)
sandwiched by two sapphire plates (thickness, 1.0 mm; diameter, 13 mm). An-
other is a cell with an alumina tube (thickness, 0.8–1.5 mm; inner diameter,
9 mm; outer diameter, 13 mm) and two sapphire plates (thickness, 1.0 mm;
diameter, 13 mm). The mullite tube cell was successively used in the thermal
diffusivity measurements of iron, cobalt, and nickel. However, it should be
kept in mind that mullite is not a suitable material to contain metallic melt
at temperature above 1,700 K for expected time span (about 3 h). In the case
of iron melt, we could not obtain sufficiently reliable temperature response
signals when using an alumina tube. This is considered mainly arising from a
problem for alumina in degassing the residual gas inside the iron sample. In
the cases of cobalt and nickel, there were no experimental problems associated
with neither the mullite nor alumina tubes [13].
94 H. Shibata et al.
(a) H 2 O(120ppm)
9 1.B 2 O 3
2.B 2 O 3 +P 2 O 5 (0.03mol%)
3.B 2 O 3 +P 2 O 5 (0.10mol%)
α / 10 −7 m 2 s −1
7 4.B 2 O 3 +P 2 O 5 (1.00mol%)
1
2
5 3
4
1
(b) B 2 O 3 :H 2 O(120ppm)
9
B 2 O 3 +In 2 O 3 (0.08mol%):H 2 O(120ppm)
B 2 O 3 :H 2 O(500ppm)
α / 10 −7 m 2 s −1
1
1000 1100 1200 1300 1400 1500
T/K
Fig. 5.7. Thermal diffusivity of molten boron oxide (a) containing H2 O and In2 O3
as a function of temperature. (b) Effect of the P2 O5 content on thermal diffusivity
of molten boron oxide (containing 120 ppm water) [8]
5 Thermal Diffusivity of Melts 95
a/10-7m2s-1
3
2
pure LN
1 LN+2mol% MgO
LN+5mol% MgO
0
1500 1520 1540 1560 1580 1600 1620
T/K
Fig. 5.8. Thermal diffusivity values of molten LiNbO3 with doping MgO as a func-
tion of temperature [16]
is employed, the water content of the boron oxide melts does not change even
if samples are heated up to the temperature range between 1,000 and 1,500 K.
Lithium niobate (LiNbO3 ) is one of the most promising nonlinear optical
materials, because of its excellent optical properties [15]. Thermal diffusivity
values of LiNbO3 melt are important with respect to the solid–liquid interface
for making a large size and subgrain free single crystal. Figure 5.8 shows the
results of molten LiNbO3 at 1,523–1,604 K obtained by the laser flash method
[16]. This includes the results of molten LiNbO3 with 0, 2, and 5 mol% MgO.
All data are found in the relatively narrow range around 3×10−7 m2 s−1 , which
is approximately half the value reported for LiNbO3 crystal at 773 K [17]. A
positive temperature dependence of thermal diffusivity value of LiNbO3 melt
is found and its increase is in the order of 5% as the temperature raises from
1,523 to 1,604 K. Their compositional variation of thermal diffusivity is not
detected by doping with MgO, although some variation has been reported
in density, surface tension, and shear viscosity [18, 19]. This implies that the
fundamental mechanism of thermal diffusion in LiNbO3 melt is unchanged
both by temperature change or by doping with MgO up to 5 mol%.
The thermal diffusivity measurements for continuous casting powder (here-
after referred to as CC powder) for steel were systematically made by the
three-layered laser flash method. The results of thermal diffusivity values are
given in Fig. 5.9a, using the results of CC powder melts containing titanium
oxide and iron oxide as an example. The reference composition of CC powder
is 35.6% SiO2 , 19.9%, 17.1% CaF2 , 10.1% Na2 O, 9.3% MgO, and 7.7% Al2 O3
in mass%. The measured values similar to the Fig. 5.9 case were obtained for
19 cases, although there are differences in detail [12]. The results of Fig. 5.9a
clearly indicate a slightly positive temperature coefficient of thermal diffu-
sivity of molten CC powders in the temperature range presently investigated.
However, the present results are somewhat spread in certain temperatures and
96 H. Shibata et al.
8
(a) (a)
6
4
TiO2 4.9% T.Fe 0.4%
a / 10-7m2s-1
0
1350 1400 1450 1500 1550 1600 1400 1450 1500 1550 1600
T/K
Fig. 5.9. Thermal diffusivity values of continuous casting powders for steel con-
taining titanium oxide and iron oxide. (a) Radiative component is included. (b)
Radiative component is excluded [12]
3000
k / m -1
1000 5
0 0
0 2 4 6 8 10
l/10 -6 m
Fig. 5.10. Absorption coefficient of continuous casting powders for steel containing
titanium oxide and iron oxide together with the hemispherical spectral emissive
power of blackbody at 1,573 K [20]
Δl= 0.2 mm
6 l2(mm)
0.3
a/10-7m2s-1
0.2
5
0.1
is absorbed or emitted from a liquid layer. In this condition the radiation from
a platinum plate is considered to be dominant. With respect to this subject,
Ohta et al. [22] have systematically estimated the effect of radiative component
in the transparent body approximation by computing the apparent thermal
diffusivity values for a case of α2 = 4 × 10−7 m2 s−1 using the finite difference
method and the results are illustrated in Fig. 5.11. The essential point of
Fig. 5.11 is to suggest that the apparent thermal diffusivity values increase
with increasing temperature and sample thickness by the contribution due
to radiative component in cases where the transparent body approximation
is valid. These results also suggest that the radiative component induces an
98 H. Shibata et al.
increase of about 20% in thermal diffusivity for the measurement where the
sample thickness of 0.1 mm (for the first measurement) with its variation of
0.2 mm.
As the coefficients for correction are summarized in the following form as
a function of the absolute temperature T , the measured thermal diffusivity
values of α0 should be corrected with the value of l2 .
α = cα0 , (5.8)
l = 0.3mm
573K
Rosseland
a/10−7m2s−1
approximation
4.5
Transparent
approximation κ l >1
Fig. 5.12. Apparent thermal diffusivity including radiative heat flow theoretically
calculated under the gray body approximation for a sample of thermal diffusivity of
4 ×10−7 m2 s−1
.
5 Thermal Diffusivity of Melts 99
Table 5.2. Mean absorption coefficient κm for samples containing TiO2 and iron
oxide estimated from measured optical properties [21]
larger than 105 m−1 . When the absorption coefficient shows the value beyond
106 m−1 , it may safely be said that samples are considered opaque. In such
case, the radiative contribution is insignificant. On the other hand, in the
lower absorption region less than 102 m−1 , the results are found to be con-
sistent with those of the transparent body approximation. The contribution
from radiative component should be separated from measured thermal diffu-
sivity values, at least, the gray body approximation or higher order (band)
approximation in cases where the absorption coefficient is lying in the region
between 102 and 104 m−1 . Samples containing titanium oxide and iron oxide
are included in this category. The mean absorption coefficient of κm for five
CC powder samples containing titanium oxide and iron oxide was determined
from the measurements of spectral transmissivity and reflectivity in the wave
length region between 4×10−7 and 1×10−5 m and the results are summarized
in Table 5.2 [21].
Analysis for estimating the contribution of radiative component at high
temperature measurements for CC powder samples has been made by consid-
ering the variation of the optical properties and provided information about
the apparent thermal diffusivity values theoretically estimated for three cases
of the transparent body approximation [20]. The wavelength dependence of the
absorption coefficient for samples, as exemplified by the results of Fig. 5.10, is
explicitly included in the band approximation, although this requires lengthy
numerical computation. The theoretical details are given in [22]. The most
significant aspects are to provide the variation of the apparent thermal dif-
fusivity values when using three different approximations for estimating the
radiative contribution as summarized in Table 5.3. The following points are
worth of note.
The laser flash method with a three-layered cell shown in Fig. 5.5 has been
applied to obtain sufficiently reliable values of thermal diffusivity of molten
iron, cobalt, and nickel at temperature above 1,700 K [13]. In this experimen-
tal condition with a devised cell, holding the sample of metallic melts, it is
necessary to consider the effect of not only radiative heat loss but also conduc-
tive heat loss at interface between metallic melt and cell material. Regarding
this subject, Nishi et al. [13] reported the computer simulation results us-
ing numerical analysis under the following eight assumptions to construct the
theoretical heat transport equation with the appropriate boundary conditions
and the initial conditions.
(1) One dimensional heat flow and each layer is homogeneous
(2) The whole cell is under adiabatic condition for the conductive heat flow
5 Thermal Diffusivity of Melts 101
0.2
ln TM exp(-kt)
*
0.1
0.0
−0.1
*
ln T
−0.2
−0.3
Calculated : molten nickel
−0.4 (R+ = 10 2, Y= 0.1)
−0.5
0 5 10 15
t*
Fig. 5.13. Temperature response curve for molten nickel obtained by numerical
simulation [13]
102 H. Shibata et al.
0.15
lnTM
0.10
R + = 10 n
n = 0.5
⬁
0.05 1.0
2.0
0.00
0.00 0.01 0.02 0.03 0.04 0.05
k
Fig. 5.14. Relationships between κ and ln TM of molten nickel. Symbols of
open square, open circle, open triangle, and open inverted triangle denote the
results obtained from the different runs [13]
5 Thermal Diffusivity of Melts 103
20
: Molten iron
: Molten cobalt
15 : Molten nickel
a /10-6 m2s-1
10
0
1700 1800 1900
T/K
Fig. 5.15. Thermal diffusivity of molten iron, cobalt, and nickel as a function of
temperature [13]
T/a.u.
Zr55Al10Cu30Ni5
in the liquid state at 1218K
αPd40 Cu30 Ni10 P20 = 5.73 × 10−9 (T − 920) + 3.28 × 10−6 (920 ≤ T /K ≤ 1120),
−9 −6
αPd40 Ni40 P20 = 3.95 × 10 (T − 1000) + 4.12 × 10 (1000 ≤ T /K ≤ 1120),
αPd40 Cu40 P20 = 6.67 × 10−9 (T − 1060) + 5.35 × 10−6 (1060 ≤ T /K ≤ 1120).
(5.15)
2 -1 4
a /10 m s
-6
5
2 -1
a /10 m s
4
-6
5.6 Summary
The current state of the art concerning the laser flash method has been re-
viewed for determining the thermal diffusivity values of high temperature ox-
ide and metallic melts in the temperature range between 1,000 and 2,000 K.
To overcome some experimental difficulties at high temperature, a differen-
tial scheme without information of the absolute value of sample thickness was
used in the oxide melt case with the help of a numerical technique to estimate
the effect of radiative heat transfer in the cell system, where the absorption
and emission behavior of the radiative component in the sample layer were
fully considered. For metallic melt, a new simple cell system was developed
to keep a precise value of sample thickness and the effect of conductive or
radiative heat leak from sample was taken into account.
The values of thermal diffusivity of molten boron oxide, lithium niobate,
and various continuous casting powders were successfully determined by the
differential three-layered laser flash method. The effect of radiative compo-
nent in the cell system on the measured thermal diffusivity was systemat-
ically discussed by considering absorption and emission of radiation in the
sample layer. The thermal diffusivity values of some metallic melts of Fe,
Co, Ni, Ge, and Si were determined at temperature up to 1,700 K, with suf-
ficient reliability by the laser flash method. This method was also system-
atically applied to molten Zr-based and Pd-based alloys, which show good
glass formability. Although the number of results is limited, it would be in-
teresting to find that the thermal diffusivity values of molten alloys provide
a possible good indicator for predicting the glass forming ability of alloys of
interest.
These results clearly suggest that it would be very promising to extend
the laser flash method with a three-layered cell to the thermal diffusivity
measurements of other melts at high temperature and then its validity and
usefulness can be confirmed on a wider base.
Acknowledgment
References
1. R.J. Gale, D.G. Lovering (eds.), Molten Salt Technology (Plenum, New York,
1982; 1991)
2. K.C. Mills, in Proc. 4th Inter. Conf. on Molten Slag and Fluxes, Iron Steel
Institute of Japan, 1992, p 405
3. Editorial Committee of Crystal Growth Society of Japan, Handbook for Crystal
Growth, Crystal Growth Society of Japan, 1995
4. Y.S. Touloukian, C.Y. Ho, P.E. Liley (eds.), Thermophysical Properties of Mat-
ter, (Plenum, New York, 1971)
5. W.J. Parker, R.J. Jenkins, C.P. Butler, G.L. Abbott, J. Appl. Phys. 32, 1979
(1961)
6. H. Ohta, G. Ogura, Y. Waseda, M. Suzuki, Y. Yamashita, Rev. Sci. Insrum, 61,
2645 (1990)
7. Y. Waseda, M. Masuda, H. Ohta, in Proc. 4th Inter. Symp. on Advanced Nu-
clear Energy Research, Japan. Atomic Energy Research Institute, Tokai, Ibaraki,
1992, p 298
8. G. Ogura, I.K. Suh, H. Ohta, and Y. Waseda, J. Ceram. Soc. Jpn. 98, 320
(1990)
9. R.E. Taylor, High Temp. High Pressure 11, 43 (1979)
10. Y. Waseda, H. Ohta, Solid State Ionics 22, 263 (1987)
11. H.M. James, J. Appl. Phys, 51, 4666 (1980)
12. Y. Waseda, M. Masuda, K. Watanabe, H. Shibata, H. Ohta, and K. Nakajima,
High Temp. Mater. Process 13, 267 (1994)
13. T. Nishi, H. Shibata, H. Ohta, Y. Waseda, Mater. Metallur. Trans. 34A, 2801
(2003)
14. T. Nakamura, Ceramics and Heat (Gihodo, Tokyo, 1985), p 81
15. A.A. Ballman, J. Am. Ceram. Soc. 48, 112 (1965)
16. H. Ogawa, H. Ohta, Y. Waseda, J. Cryst. Growth 133, 255 (1993)
17. R.A. Morgan, K.I. Kang, C.C. Hsu, C.L. Kiliopoulos, N. Peyghambarian, Appl.
Optics 26, 5266 (1987)
18. Y. Zhou, J. Wang, P. Wangm, L. Tang, O. Zhu, Y. Wu, H. Tan, J. Cryst.
Growth, 114, 87 (1991)
19. E. Tokizaki, K. Terashima, S. Kimura, J. Cryst. Growth 123, 121 (1992)
20. H. Ohta, M. Masuda, K. Watanabe, K. Nakajima, H. Shibata, Y. Waseda, Tetsu-
to-Hagane 80, 463 (1994)
21. H. Ohta, K. Watanabe, K. Nakajima, Y. Waseda, High Temp. Mater. Process,
12, 139 (1993)
22. H. Ohta, K. Nakajima, M. Masuda, Y. Waseda, in Proc. 4th Inter. Symp. On
Slags and Fluxes, Iron and Steel Inst. Japan, Tokyo, 1992, p 421
23. M.I. Darby, High Temp. High Pressure, 15, 629 (1983)
24. E.R. Eckert, R.M. Drake Jr., Analysis of Heat Transfer (McGraw-Hill,
Kougakusha, Tokyo, 1972), p 254
25. S. Rosseland, Theoretical Astrophysics, Claredon Press, Oxford, (1936); cited in
a book of R. Siegel and J.R. Howel, Themal Radiation Heat Transfer (McGraw-
Hill, Kougakusha, Tokyo, 1972), p 470
26. T. Nishi, H. Shibata, H. Ohta, Mater. Trans. 44, 2369 (2003)
27. H.S. Chen, Acta Metall. 22, 1504 (1974)
28. H.S. Chen, K.A. Jackson, Chapter 3 in Metallic Glasses, American Society for
Metals, Metals park, Ohio, (1978), p 74
5 Thermal Diffusivity of Melts 109
29. A. Inoue, T. Zhang, T. Masumoto, Mater. Trans. JIM 31, 177 (1990)
30. A. Inoue, Bulk Amorphous Alloys, Preparation and Fundamental Characteris-
tics, Trans. Tech. Pub., Zurich, (1998)
31. N. Nishiyama, A. Inoue, Mater. Trans. JIM 37, 1531 (1996)
32. H. Shibata, S. Nishihata, H. Ohta, S. Suzuki, Y. Waseda, M. Imafuku, J. Saida,
A. Inoue, Mater. Trans. 48, 886 (2007)
33. T. Nishi, H. Shibata, H. Ohta, N. Nishiyama, A. Inoue, Y. Waseda, Phys. Rev.
B 70, 174204 (2004)
6
Emissivities of High Temperature
Metallic Melts
6.1 Introduction
The mathematical modeling of heat flow in high temperature processes has
been a useful means of obtaining more efficient process design and stricter
process control. At high temperatures heat can be transferred by three mecha-
nisms, that is, conduction, convection, and radiation. Analysis of heat transfer
requires physical property data of a medium through which heat is transferred,
relevant to the respective mechanisms: the thermal conductivity is indispens-
able to heat flux calculation by conduction from Fourier’s law, and the viscos-
ity, density, and heat capacity are indispensable to heat flux calculation by
convection. On the other hand, the emissivity plays a key role in heat transfer
analysis by radiation because it quantifies how well a substance radiates en-
ergy in the form of light. The emissivity is more important in the analysis for
materials processing involving metallic melts since the radiation contribution
becomes more predominant as temperature rises. Thus, emissivity measure-
ments have been conventionally attempted on various metallic melts at high
temperatures, data of which have been published in the handbook [1], for ex-
ample. However, the data have not been abundant enough for practical use
in mathematical modeling, and measurements are still now being made con-
tinuously. This chapter focuses on emissivities of metallic melts and reviews
recent measurement techniques and data for the emissivity, mainly on and
after the publication of the handbook.
23 mm
Al2O3 crucible
Cavity
18 mm
(1.6mm diameter,
16 mm depth)
19 mm
26 mm
Ta crucible
Fig. 6.1. Al2 O3 and Ta crucibles used by Ratanapupech and Bautista [5]
incident light that are reflected, absorbed, and transmitted by a material are
termed the reflectivity r, absorptivity a, and transmissivity t, respectively. The
sum of these values equal unity since all the incident light is either reflected,
absorbed, or transmitted:
r + a + t = 1. (6.6)
For opaque materials such as metals, t = 0 and (6.6), reduces to
r + a = 1. (6.7)
When the system is in thermal equilibrium, Kirchhoff’s law applies, that is,
a = ε. Thus,
ε = 1 − r. (6.8)
On the other hand, the normal spectral reflectivity r(λ, T ) can be expressed
from Fresnel’s equation as
r(λ, T ) = (n − 1)2 + k 2 / (n + 1)2 + k 2 , (6.9)
where n and k are the optical constants (refractive index and extinction co-
efficient, respectively) of material at wavelength λ and temperature T . Com-
bination of (6.8) and (6.9) leads to
ε(λ, T ) = 4n/ (n + 1)2 + k 2 . (6.10)
Accordingly, measurement of refractive index and extinction coefficient pro-
vides the normal spectral emissivity.
The refractive index and extinction coefficient are usually determined si-
multaneously using ellipsometry. Ellipsometry is based on the measurement
of the change in polarization state of a beam of light upon reflection from the
sample, and can be conducted as a function of wavelength. The instrument
basically consists of a light source, polarizer, sample, analyzer, and detector,
although there are a number of ways that an ellipsometric measurement can
be realized. As shown in Fig. 6.2, in the measurement the linearly polarized
light is incident onto a sample at an angle of incidence φ1 ; as a result, the
reflected light is elliptically polarized, and the polarization state is determined
by the analyzer, providing the ellipsometry parameters Δ and Ψ , which char-
acterizes the polarization state. The values of Δ and Ψ are then converted to
the optical constants (n and k) of the sample using the angle of incidence φ1
and the refractive index of the ambient n1 (usually air, n1 ≈ 1) via (6.11-1)
and (6.11-2) [7].
n2 − k 2 = n21 sin2 φ1 1 + tan2 φ1 (cos2 2Ψ − sin2 Δ sin2 2Ψ )/
(1 + sin 2Ψ cos Δ)2 (6.11-1)
2nk = n21 sin2 φ1 tan2 φ1 sin 4Ψ sin Δ/(1 + sin 2Ψ cos Δ)2 (6.11-2)
Usually, this calculation is conducted on a computer using sophisticated soft-
ware that may correspond to more complicated situations of samples. Using
6 Emissivities of High Temperature Metallic Melts 115
ambient (n1)
φ1
linearly elliptically
polarised polarised
material (n,k)
Air-tight chamber
He-Ne laser Gas in
Polariser Analyser
Furnace
Sample Filter
Diaphragm
Thermocouple
Gas out
Fig. 6.3. Normal spectral emissivity measurement using ellipsometer at high tem-
perature [8]
this method, Tanaka et al. [8] have determined normal spectral emissivities of
solid Cu–Ni alloys at high temperatures (Fig. 6.3), and Shvarev et al. [9] have
determined normal spectral emissivities of Ni–Cr alloys in solid and liquid
states. Measurements on liquid metals and alloys have been vigorously made
by Krishnan and his co-workers [10–15]: They have applied ellipsometry to
a metal droplet that is melted and suspended by electromagnetic levitation
(Fig. 6.4), and produced data for Al [10–12], Fe [15], Ni [14, 15] etc for ex-
ample. Electromagnetic levitation is very useful for preventing samples from
chemical reactions with containers.
116 M. Susa and R.K. Endo
D1
Rotating
Analyser
D2
Sample
Electromagnetic Levitator
Linear
Polariser
Laser
Mirror
Quartz
Window
Pyrometer
Filter
Quartz Lens
Ar (1 atm)
Carbon Crucible
Carbon Heater
Cavity
Heat Shield
Sample
Chamber
Spectroscope
Prism
Window
Diaphragm
Camera
Window Window
Sample
Induction coils Ar- 5 %H2 (1 atm)
Cold crucible
Graphite plate Alumina tube
1
4
3
Fig. 6.7. Spectral emissivity measurement for liquid metals [30]. 1 : tube-shaped
sample, 2 : aperture, 3 : lens, 4 : optical fibre, 5 : optical fibre
this method, a metallic sample in the shape of tube (or wire) is resistively
self-heated with a RCL discharge circuit at a heating rate 107 –109 K s−1 from
room temperature to a temperature higher than the melting point, and during
this heating cycle measurement of normal spectral emissivity is carried out. In
their earlier work [30], this measurement was based on direct measurements of
radiation intensities mentioned in Sect. 6.3.3: the radiation intensities of real
body and blackbody were measured using radiations from the surface and the
hole of a nickel tube (sample), as shown in Fig. 6.7, where note that the tube
retains its shape into the liquid state owing to the very high heating rate.
At the melting plateau, the ratio of the radiation intensity of sample surface
ISR to that of blackbody IBR was normalized by the emissivity at the melting
point ελ,M taken form the literature, and the emissivity in the liquid state ελ
was represented by the following equation:
This system is basically a technique for levitating samples, and Rhim et al.
[39,40] have applied this technique to measurements of various thermophysical
properties including emissivity. In one of their studies [39], a silicon droplet
(ca. 3 mm diameter) was electrostatically levitated in a vacuum of about
10−8 Torr and heated by a focused 1 kW xenon arc lamp until the sample
was melted. After the lamp was blocked, the temperature of sample was con-
tinuously measured using a one-colour pyrometer to obtain the cooling curve
of temperature vs. time. This cooling curve can also be described by the ra-
diative heat transfer equation as follows.
(m/M )CP (dT /dt) = −εT Aσ(T 4 − Ta4 ), (6.13)
where m is the mass of sample, M is the molar mass of sample, CP is the
molar heat capacity at constant pressure of sample, εT is the hemispherical
total emissivity of sample, A is the surface area of sample, σ is the Stefan–
Boltzmann constant 5.67 × 10−8 W m−2 K−4 , T is the temperature of sample,
Ta is the ambient temperature, and t is the time. Values of m, M, A, and Ta
are known from experimental conditions, and values of T and dT /dt can be
derived from the cooling curve. Thus, the ratio of CP /εT can be determined,
giving the value of εT with the help of the published value of CP . Other than
liquid silicon, Rhim and his co-workers have determined hemispherical total
emissivities of liquid nickel and zirconium at the melting points [40]. The
electrostatic levitation technique is covered in detail in Chap. 9.
0.3
6
0.2 11 5
4 2 12
ε
8 1 3
7 9
0.1 10
M.P. Cu 650 nm
0
1200 1400 1600 1800 2000 2200
T/K
Fig. 6.8. Normal spectral emissivity of copper around 650 nm as a function of
temperature; (650 nm): 1 Burgess [41], 2 Stubbs [42]; (660 nm): 3 Bidwell [43], 4
Burgess et al. [44], 5 Smith et al. [45], 6 Lange et al. [46]; (645 nm) 7 Stretz et al. [47],
8 Bonnell et al. [48], 9 Dokko et al.; (633 nm) 10 Krishinan et al. [49]
0.5
Cu
0.4
0.3
ε
0.2
0.1 3 1
2
0
400 500 600 700 800 900
l/nm
Fig. 6.9. Normal spectral emissivity of liquid copper at the melting point as a
function of wavelength; 1 Watanabe et al. [24], 2 Tanaka et al. [50], 3 Krishnan
et al. [49]
Tanaka et al. have used same cold crucible technique to eliminate the reac-
tion between sample and a container, there are large differences between these
data in Fig. 6.9, and the values reported by Tanaka show good agreement with
those by Krishnan et al. These suggest the difficulty in emissivity measure-
ments of liquid metals; the uncertainty remaining would be caused by the
uneven surface of the sample etc. Figure 6.9 also suggests that the emissivity
of Cu decrease drastically around 600 nm. This is because of the interband
transition as well as free electron transition in the lower wavelength region.
Figures 6.10 and 6.11 show normal spectral emissivities of liquid silver and
gold, respectively, at the melting points. Both noble metals show negative
wavelength dependencies.
122 M. Susa and R.K. Endo
0.15
Ag
0.1
2
ε
1
0.05
0
400 500 600 700 800 900
l/nm
Fig. 6.10. Normal spectral emissivity of liquid silver at the melting point as a
function of wavelength; 1 Watanabe et al. [24], 2 Krishnan et al. [49]
0.6
0.5 Au
0.4
2
0.3
ε
0.2
1
0.1
0
400 500 600 700 800 900
l/nm
Fig. 6.11. Normal spectral emissivity of liquid gold at the melting point as a func-
tion of wavelength; 1 Watanabe et al. [24], 2 Krishnan et al. [49]
There have been many literature data of spectral emissivities and optical
constants for transition metals at and near melting points, such as Fe [5, 43,
46, 51, 52], Ni [5, 44, 47, 49, 50], Co [53], Mo [54], and W [54]. In addition,
Krishnan and his co-workers have actively measured emissivities for other
metals, for example, Zr and Ti. In this section, spectral emissivities of Fe, Ni,
and W are reviewed for representative examples.
Figures 6.12 and 6.13 show normal spectral emissivities of Fe and Ni at
650 nm as functions of temperature. It is reported that Fe and Ni were easily
oxidized at the surface which results in the higher emissivity values.
Cagran et al. measured normal spectral emissivities of Mo and W
by the microsecond pulse heating system with ellipsometer; heating tech-
nique which allowed the measurement on the materials having high melting
points (Figs. 6.14 and 6.15). The emissivities of Mo increase with increase in
6 Emissivities of High Temperature Metallic Melts 123
0.6
2
1
0.4
3
ε
4
5
0.2
Fe 650 nm
0
1800 1900 2000 2100
T/K
Fig. 6.12. Normal spectral emissivity of liquid iron at melting point at 650 nm as
a function of temperature; 1 Dastur et al. [51], 2 Bidwell [43], 3 d’Entremont [52],
4 Ratanapupech et al. [5], 5 Lange et al. [46]
0.6
0.4 1
ε
0.2 3
Ni 650 nm
0
1700 1800 1900 2000 2100
T/K
Fig. 6.13. Normal spectral emissivity of liquid nickel at melting point around
650 nm as a function of temperature ; 1 Lange et al. [47], 2 Ratanapupech et al. [5],
3 Bidwell [44]
0.45
684.5 nm
Mo 902 nm
1570 nm
0.4
ε
0.35
0.3
3000 3500 4000 4500
T/K
Fig. 6.14. Normal spectral emissivities of liquid molybdenum as functions of tem-
perature reported by Cagran et al. [54]
124 M. Susa and R.K. Endo
0.5
684.5 nm
902 nm
1570 nm
0.4
ε
0.3
W
0.2
3500 4000 4500 5000
T/K
Fig. 6.15. Normal spectral emissivities of liquid tungsten as functions of tempera-
ture reported by Cagran et al. [54]
0.3 1
2
3
4
ε
2
0.2
5
Si average over 1553 - 1797 K
0.1
1000 2000 3000
l/nm
temperature for the entire measured wavelengths, whereas those for W show
significant temperature dependence.
There are many reports for emissivities of liquid Si (Figs. 6.16 and 6.17),
because it is one of the most important material supporting the information
technology in the modern society. The emissivity decreases with increasing
wavelength (Fig. 6.16) and has almost no temperature dependence (Fig. 6.17).
6.4.4 Alloys
There are few emissivity data reported for alloys comparing with pure metals.
Schaefers et al. [59] reviewed emissivities for binary system such as Ni–Fe,
Ce–Cu, Ni–Cr, and Ti–Al, and measured on Ti–V, Fe–V, and Fe–Nb systems
6 Emissivities of High Temperature Metallic Melts 125
1
0.3
6 2
5
3
ε
0.2 4
Si
0.1
1600 1700 1800 1900 2000
T/K
0.5
0.4
0.3
ε
0.2
0.1
Ce-Cu 645 nm
0
0 20 40 60 80 100
at%Cu
Fig. 6.18. Composition dependence of normal spectral emissivity of liquid Ce–Cu
alloys at 645 nm [3]
at 547 and 650 nm. Recently, Hayashi et al. [23] and Tanaka et al. [50] mea-
sured emissivities for Ag–Cu and Cu–Ni alloys, respectively.
Figure 6.18 shows emissivities of liquid Ce–Cu alloys at 645 nm measured
using the integral blackbody comparison method. It is reported that the tem-
perature dependence of emissivities can be neglected [3]. The value has peak at
about 75%Cu, which corresponds to the congruent intermetallic phase Cu2 Ce.
Figures 6.19 and 6.20 show spectral emissivities of Ag–Cu alloys. The
reported values show negative temperature coefficient except for Ag-80 at%Cu
and has local minimum in concentration dependence for 650 nm data.
Figures 6.21 and 6.22 show normal spectral emissivities for Zr, Ni, and
nickel based alloys. The alloys have larger emissivities at 633 nm than that for
zirconium and nickel, and do not show significant temperature dependence.
As reviewed earlier, there have been not enough emissivity data reported
for liquid metals, epecially for alloys, because of the difficulty in measurements.
In case of Cu and Si, relatively many data had been reported; however, there
126 M. Susa and R.K. Endo
Ag-Cu 1300 nm 2
0.06
1
ε
0.05 4
6
5
0.04
1200 1300 1400 1500
T/K
0.15
Ag-Cu
1 650 nm
2 1000 nm
0.1 3 1500 nm
ε
1
2
3
0.05
0 20 40 60 80 100
at%Cu
Fig. 6.20. Spectral emissivities of Ag–Cu alloys for the wavelength of 650, 1,000,
1,500 nm at 1,373 K as a function of copper concentration [23]
0.5 1 Zr
2
2 Ni-75%Zr
3
3 Ni
0.4
ε
0.3
0.2
400 600 800 1000
l/nm
Fig. 6.21. Wavelength dependencies of spectral emissivities for liquid Ni (1,800 K),
Ni-75%Zr (1,300 K), and Zr (2,350 K) [58]
6 Emissivities of High Temperature Metallic Melts 127
0.45
1 Zr
2 Ni
3 Ni-25%Sn
0.4 4 4 Ni-32.5%Sn
5 Ni-75% Zr
3 6 Ni-40%Nb
ε
5
6 2
0.35
1
0.3
1500 2000 2500
T/K
Fig. 6.22. Plot of normal spectral emissivity at 633 nm as functions of temperature
for zirconium, nickel, and nickel-based alloys [57]
are large discrepancies even in resent data. One thing can be said is that
the temperature dependence of emissivities for liquid metals have no or ex-
tremely small positive temperature dependences. Further measurements and
new approaches are expected in this field.
References
1. Y. Kawai, Y. Shiraishi, Handbook of Physico-chemical Properties at High Tem-
peratures (The Iron and Steel Institute of Japan, Tokyo, 1988)
2. R. Siegel, J.R. Howell, Thermal Radiation Heat Transfer, 4th edn (Taylor &
Francis, New York, 2002) p 17
3. W. Dokko, R.G. Bautista, Metall. Trans. B 10B, 450 (1979)
4. W. Dokko, R.G. Bautista, Metall. Trans. B 11B, 309(1980)
5. P. Ratanapupech, R.G. Bautista, High Temp. Sci. 14, 269(1981)
6. A. Tomita, M. Susa, K. Nagata, Testu-to-Hagané 79, 1329 (1993) (in Japanese)
7. H.G. Tompkins, W.A. McGahan, Spectroscopic Ellipsometry and Reflectometry
(Wiley, New York, 1999) p 63
8. R. Tanaka, T. Sato, M. Susa, Metall. Mater. Trans. A 36A, 1507 (2005)
9. K.M. Shvarev, N.I. Vnukovskii, B.A. Baum, P.V. Gelfd, High Temp. 20
540(1982)
10. S. Krishnan, C.D. Anderson, J.K.R. Weber, P.C. Nordine, W.H. Hofmeister,
R.J. Bayuzick, Metall. Trans. A 24A, 67 (1993)
11. S. Krishnan, P.C. Nordine, Phys. Rev. B 47, 11780 (1993)
12. S. Krishnan, P.C. Nordine, Phys. Rev. B 48, 4130 (1993)
13. S. Krishnan, P.C. Nordine, Phys. Rev. B 49, 3161 (1994)
14. S. Krishnan, P.C. Nordine, J. Appl. Phys. 80, 1735 (1996)
15. S. Krishnan, K. Yugawa, P.C. Nordine, Phys. Rev. B 55, 8201 (1997)
16. T. Makino, H. Hasegawa, Y. Narumiya, S. Matsuda, T. Kunitomo, Trans. Jpn.
Soc. Mech. Eng. 50, 2655 (1984)
17. K. Nagata, T. Nagane, M. Susa, ISIJ Int. 37, 399 (1997)
18. E. Takasuka, E. Tokizaki, K. Terashima, S. Kimura, Appl. Phys. Lett. 67, 152
(1995)
128 M. Susa and R.K. Endo
Ivan Egry
7.1 Introduction
Materials sciences provide the basis for most modern technologies, for ex-
ample, IT and photovoltaics, aerospace and automotive applications, nano-
and biotechnology, just to name a few. In recent years, the trend in devel-
oping new products moved from the conventional try-and-error approach to
computer-based modeling. This has become possible due to the increase in
computer power, but it is still hampered by an incomplete understanding of
all mechanisms involved and by a lack of available thermophysical property
data. To further advance the field, a multiscale and multidisciplinary approach
is necessary, combining experimental and theoretical work with computer sim-
ulations on all length scales, starting from microscopic ab-initio calculations,
covering the mesoscopic scale by, for example, phase field methods, and finally
simulating the finishing steps, such as near net shape casting.
Material scientists originally devoted most of their efforts to studying the
solid state of materials, its microstructure, and its mechanical and thermal
properties. However, in the last 10–20 years, a change in paradigm has taken
place, and the importance of the liquid phase has been recognized. In fact,
almost all industrially used materials have been molten in some processing
step, specifically in casting. Solidification from the melt leaves its fingerprints
in the final material, and hence it is of utmost importance to understand the
properties of the molten state and its solidification behavior. The prominent
feature of fluids, namely their ability to flow and to form free surfaces, poses
the main difficulty in theoretically describing them. The physics of fluids is
governed by the Navier–Stokes equation and by the ubiquitous presence of
convection. In addition, when dealing with metallic materials, the high tem-
peratures involved lead to experimental difficulties, the most trivial, but also
most fundamental, being the availability of suitable containers. Consequently,
the measurement of thermophysical properties of the liquid phase, in partic-
ular at high temperatures, is a very difficult task under terrestrial conditions.
132 I. Egry
With the advent of microgravity platforms, a new tool has become avail-
able to study the properties of fluids. In the absence of gravity, density differ-
ences play no role; consequently, sedimentation effects and buoyancy driven
convection can be minimized, which allows to study fluids and their solidi-
fication in a quiescent environment. A number of fundamental and seminal
microgravity experiments have been performed since then, sheding new light
onto old problems and, in some cases, rendering accepted textbook knowledge
erroneous. An important breakthrough was the application of containerless
processing techniques in the microgravity environment, which gave access to
the metastable region of an undercooled melt and allowed measurements of
growth velocities as function of undercooling, leading to a revision of dendrite
growth and grain refining theories. In addition, containerless processing using
electromagnetic levitation solved the problem of finding a suitable container
for high-temperature, highly reactive metallic melts and allowed to measure
thermophysical properties of these melts with a high precision.
7.2 Microgravity
Fig. 7.1. Temperature (left), heating control voltage (right) as function of time for
containerless processing of a Cu–Co sample during a TEXUS sounding rocket flight
Table 7.1 shows the characteristics of the main facilities for micrograv-
ity studies available today, including the time interval available for a given
measurement.
The available free-fall time in a drop tube/drop tower scales with the
square root of its height. It can be doubled if the payload is shot verti-
cally upwards from the bottom of the tower, instead of dropping it from
its top. This technically challenging catapult mode of operation has recently
been introduced at the drop tower of ZARM in Bremen. An aircraft on a
parabolic trajectory can provide a reduced gravity environment for about
20 s per parabola. Usually, up to 40 consecutive parabolas are performed per
flight. Sounding rockets can provide free-fall periods between 5 and 20 min.
As an example the experiment protocol for a CuCo sample to be processed
in an electromagnetic levitation facility onboard a TEXUS sounding rocket is
shown in Fig. 7.1. For many experiments, these times are not sufficient and it
is necessary to perform the experiments on board a spacecraft.
134 I. Egry
Liquids are the most sensitive of the phases to a gravitational force. This
force causes an acceleration of all masses towards the center of gravitation,
and it makes bodies fall and fluids flow downwards. In an inhomogeneous fluid
system at rest, sedimentation will result. In addition, gravity can also induce
flows in combination with a temperature or concentration gradient. This is
called buoyancy driven convection and can lead to fluid flow instabilities, like
the Rayleigh–Bénard instability [1]. Consequently, in the absence of gravity,
none of these effects will occur. Specifically, this means the following
• No container is needed for positioning a fluid
• There is no sedimentation
• There is no buoyancy driven convection
In the absence of gravity, surface tension becomes the dominant force for a liq-
uid system. It holds a liquid drop together and controls its wetting behavior.
Therefore, capillarity phenomena become important in microgravity condi-
tions. Variations in temperature or concentration along a free surface induce
variations in the local surface tension, leading to surface tension-driven con-
vection. This effect was discovered by Marangoni [2] and is named after him.
In other words, Marangoni convection replaces the buoyancy driven Rayleigh–
Bénard convection in microgravity. The absence of gravity therefore does not
imply the absence of convection.
FL = 2πa30 z0 μ0 Hzz
2
, (7.5)
where z0 is the equilibrium position of the droplet, measured downwards from
the origin of the magnetic field, that is, from the equatorial plane of the coil.
In contrast to other levitation techniques, electromagnetic levitation is in-
trinsically stable, that is, there is a restoring force for deviations from the
equilibrium position in any direction. Consequently, a solid sample performs
oscillations about its equilibrium position with a frequency which is deter-
mined by the “spring constant” of the field, and its mass. For the magnetic
field introduced above, this frequency can be calculated and is given by [9]
3μ0 2 1 2
Ω2 = H , Ω2⊥ = Ω , (7.6)
2ρ zz 4
where Ω2 is the frequency for oscillations along the symmetry axis, and Ω2⊥
is the frequency for oscillations in a plane perpendicular to it.
136 I. Egry
In addition, the sample may also rotate along any axis which, for a liquid
sample, leads to a flattening and eventual fission of the drop, due to the
centrifugal forces. This kind of instability is often encountered in practice,
and has been studied theoretically by Gerbeth and coworkers [10, 11].
Under terrestrial conditions, the joint action of the gravitational and lev-
itation force leads to a deformation of a liquid sample: the drop is elongated
along the z-axis, that is, the direction of the gravity vector. As will be seen
later, such a deformation is detrimental to thermophysical property measure-
ments. On the other hand, the liquid sample remains essentially spherical in
microgravity.
In addition, electromagnetic fields induce fluid flows inside a liquid,
conducting body. The calculation of the corresoponding flow pattern is a
formidable task for magnetohydrodynamic calculations [12]. In particular, if
these fields become turbulent, they render the measurement of viscosity impos-
sible. This can only be avoided in microgravity, where only small positioning
fields are required.
Electromagnetic levitation not only provides positioning, it also enables in-
ductive heating of the levitated sample. The power P absorbed by the sample
due to ohmic losses of the induced currents is given by [9]
B 2 ω 4π 3
P = a H(q), (7.7)
2μ0 3 0
with H(q) defined as
9 sinh(2q) + sin(2q)
H(q) = 2 q −1 . (7.8)
4q cosh(2q) − cos(2q)
The functions Q(q) and H(q) are shown in Fig. 7.2.
Fig. 7.3. Apparent electrical resistivity of solid (lower curve) and undercooled liquid
(upper curve) Co80 Pd20 . The increase below 1,300 K is due to magnetic effects, both
in the solid and in the liquid state
For small skin depth δ a0 , Rs has a relatively simple form, and (7.10) can
be inverted to yield
!
a0 A I0
δ= 1− 1− 3 cos φ − B . (7.12)
2 a0 U0
Here, A and B are two instrument constants, to be determined by calibration.
Using (7.2) yields finally the electrical conductivity σ.
As is evident from (7.2), the skin depth also depends on the magnetic per-
meability and is therefore sensitive to magnetic ordering effects. This was
demonstrated experimentally on deeply undercooled Co80 Pd20 during the
MSL-1 Spacelab mission and is shown in Fig. 7.3. The Curie temperature
of the solid phase is around 1,250 K, and the steep rise in the apparent elec-
trical resistivity reflects the onset of magnetic ordering. The same behavior
is found for the liquid phase, at approximately the same temperature. This
supports the assertion that undercooled liquid Co80 Pd20 becomes a magnetic
liquid [19].
The electrical conductivity σ is of interest in its own, but it also allows
to obtain the thermal conductivity λ through the Wiedemann–Franz relation,
which is known to hold well for liquid metals [20]:
λ = LσT. (7.13)
Here, L is a universal constant, the so called Lorenz number, L =
2.44 W Ω K−2 . Thus, electrical conductivity measurements provide an al-
ternative way, independent of convective effects, to determine the thermal
conductivity.
7 Metallic Melts under Microgravity 139
2.5x107
Volume(pixel3)
2.0x107
1.5x107
1.0x107
5.0x106
Fig. 7.5. Modulation calorimetry on ZrAlCuNi alloy during MSL-1 Spacelab mis-
sion. The melting temperature of this alloy is 1,146 K. Modulation was carried out
in the equilibrium liquid (1), as well as in the undercooled liquid (2,3). At point 4,
solidification occurs from an undercooling level of 194 K
Table 7.2. Ti46 Al8 Nb specific heat capacity and total hemispherical emissivity in
the liquid phase
case, simple formulae can be used to relate frequency ω and damping Γ of the
oscillations to surface tension γ and viscosity η, respectively. They read
32π γ
ω2 = (7.18)
3 M
and
20π a0 η
Γ = , (7.19)
3 M
where M is the mass of the droplet and a0 its radius. Depending on the sample
mass, typical frequencies ν = ω/2π lie in the range of 30–50 Hz.
Under terrestrial conditions, the liquid drop is distorted by gravity and
the compensating levitation field. As a consequence, the single frequency is
split into up to five peaks. In addition, all peaks are shifted with respect to
(7.18). This is due to the fact that the electromagnetic levitation field acts as
an additional pressure term in the Navier–Stokes equation. It, therefore, leads
to an apparent increase in surface tension. A correction formula to account
for these effects on the frequency spectrum was developed by Cummings and
Blackburn [9] who derived a sum rule that contains only measurable frequen-
cies. It reads
32π γ 1 2
= ω − 1.9Ω2tr − 0.3(Ω2tr )−1 (g/a0 )2 , (7.20)
3 M 5 m 2,m
1 2
Ω2tr = Ω + 2Ω2⊥ .
3
Here Ω2tr is the mean translational frequency of the drop, and g is the gravita-
tional acceleration. The sample oscillations are recorded with video cameras
from the top, that is, along the symmetry axis. The frame rate must satisfy
the Nyquist theorem to avoid aliasing and is typically 100–400 fps. The area of
the cross section of the sample, its center of mass, and two perpendicular radii
are calculated for each frame. A number of frames (256–4,096) is taken and
the corresponding time series is Fourier transformed to yield the frequencies.
From the time dependence of the center of mass the translational frequen-
cies are derived, whereas the surface oscillation frequencies are contained in
the temporal behavior of the two perpendicular radii. Because of their dif-
ferent symmetries, the oscillations corresponding to different m-values can be
identified using selection rules, as derived by Egry et al. [30].
In Fig. 7.6, oscillation spectra of a gold–copper alloy are shown, recorded
on ground and in microgravity during the IML-2 Spacelab mission [31]. As can
7 Metallic Melts under Microgravity 143
Fig. 7.6. Frequency spectrum of an oscillating AuCu drop under 1g (top) and
microgravity (bottom)
Fig. 7.7. Viscosity of Pd76 Cu6 Si18 measured during two different Spacelab missions.
The data are fitted with an Arrhenius relation
shown in Fig. 7.7 and compared to terrestrial data by Lee et al. [38]. Two dif-
ferent datasets are displayed: data during the STS-83 mission were taken with
no crew on board Spacelab, while data during STS-94 were recorded during
nominal crew operations. Both data yield the same fit, the higher scatter in
the STS-94 data is due to g-jitter. Although the data cover a temperature
range of 400 K, they are not close enough to the glass transition temperature
to distinguish between Arrhenius and Vogel–Fulcher behavior.
7 Metallic Melts under Microgravity 145
Acknowledgement
This review is the result of continuous discussions with many colleagues in-
volved in microgravity experiments, during the Spacelab missions, parabolic
flights, or sounding rocket campaigns. Their fruitful comments are grate-
fully acknowledged. In particular, the author thanks B. Damaschke and
R. Wunderlich for providing their data prior to publication.
References
1. Lord Rayleigh, Phil. Mag. 32, 529 (1916)
2. C. Marangoni, Nuovo Cim. 3, 97 (1878)
3. D. Herlach, R. Cochrane, I. Egry, H. Fecht, L. Greer, Int. Mater. Rev. 38, 273
(1993)
4. G. Lohöfer, P. Neuhaus, I. Egry, High Temp. High Pressure 23, 333 (1991)
5. J. Piller, A. Seidel, M. Stauber, W. Dreier, Solidification 1999, ed. by
W. Hofmeister, J. Rogers, N. Singh, S. Marsh, P. Vorhees (TMS, Warrendale,
1999) p 3
6. Team TEMPUS, Materials and Fluids under low Gravity, ed. by L. Ratke,
H. Walter, B. Feuerbacher, (Springer, Berlin, 1996)
7. W. Hofmeister, J. Rogers, N. Singh, S. Marsh, P. Vorhees (ed.), Solidification
1999 (TMS, Warrendale, 1999)
8. G. Lohöfer, Quarterly Appl. Math. 11, 495 (1993)
9. D. Cummings, D. Blackburn, J. Fluid Mech. 224, 395 (1991)
10. J. Priede, G. Gerbeth, IEEE Trans. Magnetics, 36, 349 (2000)
11. J. Priede, G. Gerbeth, IEEE Trans. Magnetics, 36, 354 (2000)
12. R.W. Hyers, Meas. Sci. Tech. 16, 394 (2005)
13. G. Lohöfer, SIAM, J. Appl. Math., 49, 567 (1989)
14. Ivan Egry, C. Nieto de Castro, Chemical Thermodynamics, ed. by T. Letcher
(Blackwell, Oxford, 1999) p 171
15. H. Suga, G. Pottlacher, I. Egry, Measurement of the Thermophysical Properties
of Single Phases, Experimental Thermodynamics, vol VI, ed. by A. Goodwin,
K. Marsh, W. Wakeham (Elsevier, Amsterdam, 2003) p 475
16. J. Brillo, G. Lohöfer, F. Schmidt-Hohagen, S. Schneider, I. Egry, Int. J. Mater.
Product Technol. 26, 247 (2006)
17. G. Lohöfer, I. Egry, Solidification 1999, ed. by W. Hofmeister, J. Rogers,
N. Singh, S. Marsh, P. Vorhees, (TMS, Warrendale, 1999)
18. T. Richardsen, G. Lohöfer, Int. J. Thermophys. 20, 1029 (1999)
19. D. Platzek, C. Notthoff, D.M. Herlach, G. Jacobs, D. Herlach, K. Maier, Appl.
Phys. Letts. 65, 1723 (1994)
20. K. Mills, B. Monaghan, B. Keene, Int. Mater. Rev. 41, 209 (1996)
21. E. Gorges, L. Racz, A. Schillings, I. Egry, Int. J. Thermophys. 17, 1163 (1996)
22. J. Brillo, I. Egry, Z. Metallkd. 95, 691 (2004)
23. B. Damaschke, K. Samwer, I. Egry, Solidification 1999, ed. by W. Hofmeister,
J. Rogers, N. Singh, S. Marsh, P. Vorhees, (TMS, Warrendale, 1999)
24. B. Damaschke, private communication
25. H. Fecht, W. Johnson, Rev. Sci. Instrum. 62, 1299 (1991)
7 Metallic Melts under Microgravity 147
8.1 Introduction
Numerical simulations are widely used for high value-added materials process-
ing such as semiconductor crystal growth, casting of super high-temperature
alloys for a jet-engine turbine blade, and for welding in automobile manufac-
turing [1, 2]. Process modeling involving a liquid-to-solid transition requires
precise thermophysical properties of materials in the solid and liquid state at
temperatures near their melting points. However, high-temperature materi-
als such as liquid silicon are chemically reactive and are easily contaminated
by their containers and contact materials. Therefore, it remains extremely
difficult to measure the thermophysical properties of high-temperature liq-
uids. Especially, the thermal conductivity of a high-temperature liquid is a
difficult property to measure because of the existence of the buoyancy and
Marangoni convections in the liquid. Not only from process modeling but also
from a scientific perspective, thermal conductivity data of high-temperature
metallic or semiconductor liquids are important to investigate whether the
Wiedemann–Franz law [3] is applicable to them.
Fecht et al. [4–7] developed modulation calorimetry for electromagnetically
levitated metallic melts. The radio frequency (rf) coil’s power was modulated
to provide sinusoidal heating to the sample melt. The heat capacities and
hemispherical total emissivities of the melts were determined at higher tem-
peratures. However, convections existing in the droplets make it difficult to
measure the true thermal conductivity of the melts. Yasuda et al. [8] reported
that motion of the center of gravity, surface oscillation, and convection of an
electromagnetically levitated liquid metal were suppressed in a static mag-
netic field because of the Lorentz force resulting from interaction between the
fluid flows and the static magnetic field.
Based on the techniques described above [4–8], we devised noncontact
modulation calorimetry, which incorporates a static magnetic field to realize
measurement of the true thermal conductivity of a metallic melt in addition to
measurements of the heat capacity and hemispherical total emissivity [9–11].
150 H. Fukuyama et al.
dc magnetic field
semiconductor laser
rf coil
0.0 s
silicon droplet
laser spot
1.0 s
2.0 s
aperture
3.0 s
pyrometer
(a) mirror (b)
Fig. 8.1. (a) The electromagnetic levitation apparatus incorporating a static mag-
netic field for noncontact laser modulation calorimetry; (b) a top view of the levi-
tated liquid silicon during modulation heating in a static magnetic field
Fig. 8.2. Heat flow model in this noncontact modulated laser calorimetry
and
−1/2
τc 1 1
cos φl = − ω2 1+ + ω 2 τc2 , (8.10)
ω τc τr ω τr2
2
where τr [s] is the relaxation time for radiation from the droplet to heat
reservoirs and τc [s] is the internal relaxation time for the thermal conductance
of the droplet. These are defined as
Cp Cp
τr = = (8.11)
Kr 4AσTo3
and
Cp
τc = Vh (1 − Vh ). (8.12)
Kc
The correction function, f , is defined as
−1/2
1 2 2
f = 1 + 2 2 + ω τc . (8.13)
ω τr
The condition, ω 2 τr2
1
ω 2 τc2 , which satisfies f 1, is achieved using
a proper choice of the modulation frequency. The heat capacity is therefore
determined by the temperature amplitude from (8.9). Under this condition,
the function ωΔTac,l has the maximum value as a function of the modula-
tion frequency. At that modulation frequency, the phase difference is equal to
−π/2, which is derived from the requirement of ∂f /∂ω = 0. Therefore, the
heat capacity is determined experimentally from the temperature amplitude,
with satisfaction of the requirement described earlier.
The term αSh Apo in (8.9) represents the laser input power absorbed by
an object. Here, the term is evaluated quantitatively by the products of the
laser power and the normal spectral emissivity at a laser wavelength of the
object. Assuming Kirchhoff’s law, the normal spectral emissivity is used as the
absorptivity. In this study, the value of the normal spectral emissivity of liquid
silicon is 0.223 at the laser wavelength (807 nm) measured at temperatures of
1,553 and 1,797 K with uncertainty of 5% [14]. The distribution of the laser
intensity is Gaussian; the e−2 radius of the laser beam is 2 mm for a silicon
droplet of 4 mm radius. Effects of the sample curvature on the absorptivity
were ignored.
The simple heat flow model described earlier is insufficient to determine
the thermal conductivity; it is necessary to analyze the heat flow in the sphere
more accurately. The theory of the measurements of the thermal conductivity
and hemispherical total emissivity are explained in Sect. 8.2.2.
rlaser
z
q r
y
x
md
where ΔTdc (r, θ) is the increase in average temperature and ΔTac (r, θ, t) is
the modulation amplitude. In addition, the initial temperature To is assumed
to be uniform throughout the droplet. At the stationary modulation state
md
where the temperature varies sinusoidally with time, ΔTac (r, θ, t) is expressed
in the following form:
md in out
ΔTac (r, θ, t) = ΔTac (r, θ) cos ωt + ΔTac (r, θ) sin ωt, (8.21)
in out
where ΔTac (r, θ)and ΔTac (r, θ), respectively,
represent the in-phase and out-
md
of-phase components of ΔTac (r, θ, t). Then,
substituting (8.20) and (8.21)
in
into (8.14), the following steady-state linear equation systems for ΔTac (r, θ)
out
and ΔTac (r, θ) are obtained.
in
in
1 ∂ 2 ∂ΔTac 1 ∂ ∂ΔTac
κ r + sin θ
r2 ∂r ∂r r2 sin θ ∂θ ∂θ
−ρcp,mass ωΔTac
out
= 0, (8.22)
out
out
1 ∂ ∂ΔTac 1 ∂ ∂ΔTac
κ r2 + sin θ
r2 ∂r ∂r r2 sin θ ∂θ ∂θ
in
+ρcp,mass ωΔTac = 0. (8.23)
Additionally, assuming that the increases in both the average temperature and
md
modulation amplitude, ΔTdc (r, θ) and ΔTac (r, θ, t), are much less than initial
in out
temperature To , and the boundary conditions for ΔTac (r, θ) and ΔTac (r, θ)
are given as the following linear equations. At the droplet surface irradiated
by the laser, the following pertain:
∂ΔTacin
2αPo 2R2 sin2 θ
−κ = 4σTo3 ΔTac
in
− 2 exp − s 2 (−n·elaser ), (8.24)
∂n πrlaser rlaser
out
∂ΔTac
−κ = 4σTo3 ΔTac
out
. (8.25)
∂n
At the droplet surface without the light source, the following are true:
in
∂ΔTac
−κ = 4σTo3 ΔTac
in
, (8.26)
∂n
out
∂ΔTac
−κ = 4σTo3 ΔTac
out
. (8.27)
∂n
At the centerline,
in
∂ΔTac
− = 0, (8.28)
∂θ
and
out
∂ΔTac
− = 0. (8.29)
∂θ
8 Noncontact Calorimetry of High Temperature Melts 157
Equations (8.22) and (8.23) are solved with the boundary conditions
in out
(8.24)−(8.29) to determine the distributions of ΔTac (r, θ) and ΔTac (r, θ)
in out
in the droplet. Then, using ΔTac (r, θ) and ΔTac (r, θ), the distributions of
the phase difference, φ(r, θ), in the droplet are obtainable using the following
equation:
out
ΔTac
φ(r, θ) = tan−1 in
. (8.30)
ΔTac
Determination Procedure The experimental results for φ at various fre-
quencies of the modulated laser ω are fitted with the numerical results ob-
tained using the simplified model described earlier to determine the thermal
conductivity and emissivity of the droplet simultaneously. Here, finite element
method is used to solve (8.22)–(8.29) numerically for a droplet with arbitrary
shape. To fit the experimental values of phase difference φ, the integral mean
value of φ in (8.30) over the region corresponding to the spot area of the
pyrometer used in the experiment is evaluated using the following equation:
out
−1 average(ΔTac )
φ = tan in )
. (8.31)
average(ΔTac
Therein, the following are true:
in 1 in
average(ΔTac )= ΔTac (r, θ)r sin θ (dr)2 + (r dθ)2 , (8.32)
Spyrometer S
out 1 out
average(ΔTac )= ΔTac (r, θ)r sin θ (dr)2 + (r dθ)2 (8.33)
Spyrometer S
In addition, Spyrometer is the spot area of the pyrometer. The numerical simu-
lations shown above are incorporated into the Levenberg–Marquardt method,
which enables us to compute the nonlinear least squares solutions, as a model
function, to determine the thermal conductivity and emissivity of the molten
droplet simultaneously from the experimental relation between φ and ω.
The heat capacity and density of molten silicon are necessary to obtain
the thermal conductivity and emissivity, where the heat capacity measured
in this study (see Sect. 8.4.3) was used, and where the density cited from [15]
was used. Here, both the heat capacity and the density are assumed to be
constant during the modulation experiments.
The mathematical model described earlier is based on the assumption that the
heat transfer in the droplet is governed solely by conduction. In our studies, to
realize such an assumption in the experiment, a static magnetic field was ap-
plied to an electromagnetically levitated molten droplet, and the suppression
of convection in the droplet by the Lorentz force because of the interaction
158 H. Fukuyama et al.
between the conductive fluid flow and the static magnetic field was investi-
gated. Here, it is verified by numerical simulation whether convection in the
molten silicon droplet is truly suppressed by applying a static magnetic field
of 4 T used in the experiment.
In the electromagnetic levitation (EML) system used to measure the ther-
mophysical properties of molten materials, the alternating electric current in
the rf coils induces an eddy current in the conductive material sample, which
is melted by Joule heating from the current. In addition, the electromag-
netic force caused by the interaction between the alternating magnetic field
and the induced current lifts and deforms the molten droplet; moreover, it
induces the magnetohydrodynamic (MHD) convection in the droplet. Aside
from the MHD convection, the buoyancy convection, and the thermocapil-
lary, Marangoni convection attributable to the temperature dependence of
the surface tension on the melt surface exist in the droplet. Here, the MHD,
buoyancy and Marangoni convections in the spherical molten silicon droplet
with axial application of a static magnetic field are simulated numerically. In
the numerical simulation for the EML system, the electromagnetic field in the
system is first computed to obtain the electromagnetic force and the distri-
bution of heat generation rate in the droplet; then the flow and temperature
fields in the droplet are calculated.
Figure 8.4 shows (a) the velocity profiles, (b) the stream lines, and (c)
the isotherms in the molten silicon droplet, where a static magnetic field of
T = 1.0500
T = 1.0535
v max = 547.4
T = 1.052
(a) velocity vectors (b) contours of (c) isotherms
stream functions (Tmin = 1.0462, Tmax = 1.0539,
(φmin = −13.80, φmax = 5.22, ΔT = 0.0005)
Δφ = 2.0)
4 T is applied axially. The physical properties of the molten silicon and pro-
cessing parameters used in the calculations, except that of the laser power,
are presented in Table 8.1. Here, the droplet surface is not irradiated using
a laser beam as modulated laser calorimetry; therefore, the laser power is
set as 0 W. The static magnetic field can suppress the flow perpendicular to
the direction of the magnetic field, that is, the radial flow in the droplet,
but cannot suppress the axial component of the flow. Therefore, convection
cannot be suppressed completely and three longitudinal vortices appear in
the droplet. Particularly, the intense flows, the Marangoni and MHD convec-
tions, remain near the surface of the equatorial part of the droplet. However,
considering that the velocity of the MHD convection induced by the electro-
magnetic force is strong, for example, more than 10 cm s−1 , without a static
magnetic field, and that fluid flow in the droplet is within the mild turbulent
flow regime [16–19], it is readily apparent that the convection in the droplet
is markedly suppressed by the applied static magnetic field. In Fig. 8.4, the
velocity of the fluid flow near the centerline is less than 1 cm s−1 . In addition,
the temperature distributions inside the droplet in the figure are the same as
those without convection, that is, those of the conduction dominated model.
From these results, it can be concluded that applying the static magnetic field
of 4 T, which is used in the actual measurement, enables us to measure the
thermal conductivity of molten silicon accurately.
(a) 50 1810
40 1800
Laser power P / W
Temperature / K
30 T (ii) 1790
P (i)
20 1780
10 1770
(iii)
0 1760
0 50 100 150 200 (iv)
Time / s
(b)
1811 K
1789 K
Fig. 8.5. (Top) calculated temperature response at the lower part of the droplet,
and (bottom) temperature distribution in the droplet for (i)–(iv)
for plots in Fig. 8.6, as shown by a fitting line in the figure. Consequently,
the values 63.93 W m−1 K−1 and 0.321 were obtained respectively as thermal
conductivity and emissivity. Comparing them to the input data in Table 8.1,
the estimated value of thermal conductivity, 63.93 W m−1 K−1 , shows good
agreement with the input data, 64.00 W m−1 K−1 ; it can be concluded that
the present simplified model is appropriate for estimating the thermal con-
ductivity of the electromagnetically levitated droplet. On the other hand, the
8 Noncontact Calorimetry of High Temperature Melts 161
140
120
100
−φ/ο
80
60
40 numerical results
fitting line
20
0
0 0.1 0.2 0.3 0.4
ω/
2 π /Hz
Fig. 8.6. Calculated relation between the phase difference and frequency
estimated value of emissivity, 0.321, is larger than that in Table 8.1: 0.3. The
integral mean value of temperature over the region corresponding to the spot
area of the pyrometer before being irradiated using a laser beam is 1,768 K.
Consequently, the linearization expressed in (8.24) to (8.27) is approximately
satisfied. Therefore, it is inferred that the discrepancy between the estimated
value and the emissivity in Table 8.1 results from the fact that, actually, the
initial temperature To is not constant throughout the droplet.
162 H. Fukuyama et al.
140
120
100
-f/o
80
60
40 57.6 W.m−1.K−1
64.0 W.m−1.K−1
20 70.4 W.m−1.K−1
0
0 0.1 0.2 0.3 0.4
w / 2p /Hz
Fig. 8.7. Effect of variation in thermal conductivity on the relation between the
phase difference and frequency
140
120
100
−φ/ο
80
60
40 0.24
0.30
20 0.36
0
0 0.1 0.2 0.3 0.4
ω / 2 π /Hz
Fig. 8.8. Effect of variation in emissivity on the relation between the phase differ-
ence and frequency
The emissivity of the sample is also evaluated from the cooling curve because
of the radiation in vacuum. The change in sample temperature is expressed
as the following equation after the laser irradiation ceases.
dT Q − Aσ(T 4 − Ta4 )
= . (8.34)
dt Cp
The heat balance between the power input from the rf-coil and radiation heat
loss is given by the following equation because T → To , dT /dt → 0 for t → ∞.
Q = Aσ(To4 − Ta4 ). (8.35)
Under the condition of To
ΔT , the following linearization is applied for the
radiative heat loss term in (8.34):
T 4 = (To + ΔT )4 To4 + 4To3 ΔT. (8.36)
Solving (8.34) with (8.35) and (8.36) yields
ΔT = ΔTdc exp(−t/τr ). (8.37)
By fitting the cooling curve with the exponential function shown above, the
value of τr is determined. Consequently, the emissivity is determined using
(8.11).
8.3 Experimental
A single crystalline silicon cube (7×7×7 mm) was placed at the center of the
rf-coil (15 kW, 200 kHz) in a vacuum chamber. Prior to the experiment, the
chamber was filled with high-purity argon gas; it was subsequently evacuated
to 10−3 Pa using a rotary pump in combination with a turbo-molecular pump.
The silicon was preheated by irradiation of a semiconductor laser to a tem-
perature at which the electric resistivity of the silicon was reduced sufficiently
to levitate silicon using the applied electromagnetic force. A superconducting
magnet with 220 mm bore was used to impose a static magnetic field. The
strength of the static magnetic field was varied from 0 to 4 T to suppress
convection in the liquid silicon. The initial temperature of the liquid silicon
was controlled by the electric power supply to the rf coil. The top surface of
the levitated silicon droplet was heated sinusoidally by a semiconductor laser
through a function generator. The modulation frequency varied from 0.02 to
0.3 Hz, which is sufficiently wide to determine both thermal conductivity and
emissivity experimentally, precisely, as suggested by the sensitivity analysis
described in Sect. 8.2.4.
164 H. Fukuyama et al.
Figure 8.9 shows a top view of the motion of the nucleated silicon grains on
the levitated liquid silicon at 1 T. This picture was obtained by superimposing
five consecutive images taken with a 0.024 s interval. The grains are arranged
concentrically, suggesting that the silicon droplet rotated around the z-axis,
similar to a solid sphere. Vertical convection was suppressed in the static
magnetic field. The numerical simulation presented in Fig. 8.4 also reveals
that MHD convection was reduced to less than 1 cm s−1 near the centerline
of the silicon droplet at 4 T.
4 mm
Fig. 8.9. Top view of the motion of silicon grains floating on the surface of the
levitated liquid silicon under a static magnetic field of 1 T. This picture was obtained
by superimposing five consecutive images taken at a 0.024 s interval
60 1930
0.1 Hz 0.2 Hz
50
T ΔTac
Temperature / K
1910
Laser power / W
40
ΔTdc
30 P 1890
20
1870
10
To
0 1850
0 100 200 300
Time /s
Fig. 8.10. Typical time dependence of the laser power (left-hand ordinate) and
temperature response (right-hand ordinate) during modulated laser calorimetry for
molten silicon
laser power was turned off, the sample temperature decreased according to
an exponential cooling curve and reverted to To . From this cooling curve, the
hemispherical total emissivity was also determined.
Figure 8.11 shows that the relation between ωΔTac and ω was obtained
from a series of modulation heating. The value of ωΔTac has a maximum
around the ω, which gives φ = 90◦ . The isobaric heat capacity was obtained
from this maximum value using (8.9). A curve fit of the numerically obtained
φ − ω relation to the experimentally obtained data over the entire frequency
166 H. Fukuyama et al.
7.0 130
−φ
110
ωΔTac / K.rad.s−1
6.0
90
−φ/ ο
ωΔTac
70
5.0
50
4.0 30
0.00 0.10 0.20 0.30
ω/2 π /Hz
Fig. 8.11. ωΔTac as a function of the modulation frequency is shown on the left-
hand ordinate as open squares. The phase difference −φ is shown on the right-
hand ordinate, experimental data as solid circles; the solid line presents fitting
to the experimental data using the numerical analysis explained in Sect. 8.2.2 for
determination of the emissivity and thermal conductivity of liquid silicon
range is shown by a solid line in Fig. 8.11. The hemispherical total emissivity
and thermal conductivity of liquid silicon were determined using this curve-
fitting process.
Figure 8.12 shows the isobaric molar heat capacity of liquid silicon with lit-
erature data [20–23]. The clear temperature of the molar heat capacity was
not observed. The average isobaric molar heat capacity of liquid silicon is
30 ± 5 J mol−1 K−1 at temperatures of 1,750–2,050 K. The experimental un-
certainty presented in the above value is double the standard deviation for
all the present data. The present result shows good agreement with the data
reported by Kantor et al. [20], Yamaguchi and Itagaki [21], and Olette [22],
which were measured using drop calorimetry method.
The condition, Kr /Kc ≤ 0.01, is expected to be satisfied for justification
of the heat capacity measurement, as explained in Sect. 8.2.1. However, it is
difficult to estimate the volume heated directly by laser irradiation (Vh ) for
the quantitative evaluation of Kc using (8.12). Here, the Biot number relevant
to the value of Kr /Kc , which is defined as
4σTo3
Bi = , (8.38)
κ/R
is used for evaluation. The estimated Bi numbers are 0.019–0.031 under the
present condition, depending on temperature. This value indicates that the
experimental postulate that heat transfer to the external heat bath is much
8 Noncontact Calorimetry of High Temperature Melts 167
50
40
Ref.20
cp / J.mol−1.K−1 Ref.21
30
Ref.23
Average of present study
20 Ref.22
10
Melting point of Si
0
1600 1700 1800 1900 2000 2100
Temperature / K
Fig. 8.12. The isobaric molar heat capacity of liquid silicon measured in a static
magnetic field at 0.5 T (diamonds), 1.0 T (crosses), 2.0 T (squares), 3.0 T (triangles),
and 4.0 T (circles) with data from the literature: Kantor et al., [20], Yamaguchi and
Itagaki [21], Olette [22], NIST-JANAF [23]. The solid line represents the average of
this study
smaller than heat transfer to the inner part, that is, the condition of Kr /Kc ≤
0.01 is fundamentally satisfied. The correction function, f , which is expected
to be nearly equal to unity, is also justified using (8.13) as follows. The external
relaxation time, τr , is calculated from (8.11). Using τr and (8.10), the internal
relaxation time, τc , is evaluated by fitting the frequency dependence of the
phase difference. The values of τr and τc are 12 and 0.16 s, respectively, for the
representative experimental condition as presented in Fig. 8.9. Consequently,
the value of f has a maximum value of 0.99 at the frequency between 0.1 and
0.12 Hz, which indicates that the heat capacity can be measured to within 1%
uncertainty.
The hemispherical total emissivity of liquid silicon determined from the phase
difference is presented in Fig. 8.13. Clear temperature dependence of the emis-
sivity was not observed. The average emissivity is 0.27 ± 0.03 at temperatures
of 1,750–1,910 K.
The hemispherical total emissivity of liquid silicon was also obtained from
the radiative cooling curve. Figure 8.14a shows the radiative cooling curve
of the liquid silicon. By curve fitting of (8.37), the external thermal relax-
ation time attributable to the radiative cooling was determined to be 12.1 s.
The emissivity determined from the radiative cooling curve is presented in
Fig. 8.14b; the average value is 0.25 ± 0.04. This result agrees with data ob-
tained from the phase difference (Fig. 8.13) within experimental uncertainty.
168 H. Fukuyama et al.
0.40
0.30
0.20
ε
0.10
Melting point of Si
0.00
1650 1750 1850 1950 2050
Temperature / K
Fig. 8.13. The hemispherical total emissivity of liquid silicon obtained from the
phase difference measured in a static magnetic field at 0.5 T (diamonds), 1.0 T
(crosses), 2.0 T (squares), 3.0 T (triangles), and 4.0 T (circles). The solid line rep-
resents the average of this study
0.30
Temperature / K
1800
ε
0.20
ΔTdc
1795
0.10
To Melting point of Si
1790 0.00
0 20 40 60 1650 1750 1850 1950 2050
Time / s Temperature / K
Figure 8.15 shows the thermal conductivity of liquid silicon against the tem-
perature for various static magnetic fields with data obtained from the rel-
evant literature [24–32]. The apparent thermal conductivity decreases with
increasing strength of the static magnetic field. The value converges on the
value of 66 ± 11 W m−1 K−1 at 2 T or larger. The experimental uncertainty of
8 Noncontact Calorimetry of High Temperature Melts 169
100
Present study
90 0.5T
0.5T
1T
Average (2-4T)
80 1T
( 2T, 3T, 4T)
κ / W.m−1.K−1
70 29
2-4T
60 31
26
24 32
50 30
27
40
(24) Yamamoto, (25 x) Takasuka, (26) Nishi,
30 Melting point of Si (27) Nagai, (28 ) Yamasue, (29) Cusack,
(30) Glazof, (31) Sasaki, (32) Schnyders
20
1700 1800 1900 2000 2100
Temperature / K
Fig. 8.15. Thermal conductivity of liquid silicon as a function of the temperature
in a static magnetic field in 0.5 T (solid diamonds), 1.0 T (open diamonds), 2.0 T
(squares), 3.0 T (triangles), and 4.0 T (circles) together with values measured using
several methods: laser-flash method [24–26]; hot disk method [27], and transient hot
wire method [28]. Calculated κ assuming the Wiedemann–Franz law from electrical
conductivity is also presented [29–32]. The thermal conductivity measured using
laser-flash technique is recalculated from the thermal diffusivity with cp and ρ used
in this study
the thermal conductivity is double that of the standard deviation for all data
greater than 2 T. No obvious temperature dependence of thermal conductiv-
ity was observed in the experimental temperature region. Our data measured
greater than 2 T are in the high-temperature extended region of the litera-
ture thermal conductivities, which were measured [24–28] or calculated using
the Wiedemann–Franz law [29–32]. The agreement of the present data with
the Wiedemann–Franz law indicates that the electron contribution is dom-
inant for thermal transport at the experimental temperature range. On the
other hand, thermal conductivity measured using hot disk method [27] shows
that the lowest value might be attributable to the influence of the interfacial
resistance between the AlN insulator and the hot disk sensor.
8.5 Summary
droplet. The modulated laser calorimetry was modeled to estimate the thermal
conductivity and emissivity of the electromagnetically levitated droplet using
a measured parameter, that is, the phase difference between the modulated
laser and the temperature variations at various frequencies of the modulated
laser. The experimental relation between the phase difference and frequency
was fitted by the mathematical model to estimate the thermal conductivity
and emissivity of molten silicon simultaneously. In addition, the numerical
simulations for unsteady thermal field in the electromagnetically levitated
droplet were carried out to demonstrate the validity of the proposed simpli-
fied model, then to investigate the sensitivity of the thermophysical properties
to the relation between the phase difference and frequency.
Numerical simulations for convection in the droplet were carried out to
confirm the fact that a static magnetic field suppresses the convection in a
liquid silicon droplet in the electromagnetic levitator. Here, the convection
driven by the buoyancy force, thermocapillary force attributable to the tem-
perature dependence of the surface tension on the melt surface, and electro-
magnetic force in the droplet were considered. Results show that applying
a static magnetic field of 4 T can suppress the convection inside the droplet
sufficiently to measure the thermal conductivity of the liquid silicon.
The experimental results for liquid silicon are summarized as follows. The
silicon droplet rotated around the z-axis similarly to a solid sphere: vertical
convection was suppressed in the static magnetic field. The isobaric molar
heat capacity of 30 ± 5 J mol−1 K−1 (1,750–2,050 K), the hemispherical to-
tal emissivity of 0.27 ± 0.03 (1,750–1,910 K), and the thermal conductivity of
66 ± 11 W m−1 K−1 (1,750–2,050 K). The convection in an electromagnetically
levitated liquid silicon droplet is suppressed as sufficiently small for measure-
ment of the thermal conductivity in a static magnetic field greater than 2 T.
Acknowledgements
References
1. T. Hibiya, I. Egry, Meas. Sci. Technol. 16, 317 (2005)
2. M. Mito, T. Tsukada, M. Hozawa, C. Yokoyama, Y.R. Li, N. Imaishi, Means.
Sci. Technol. 16, 457 (2005)
3. C. Kittel, Introduction to Solid State Physics, 7th edn. (Wiley, New York, 1996)
p. 144
4. H.-J. Fecht, W.L. Johnson, Rev. Sci. Instrum. 62, 1299 (1991)
5. R.K. Wunderlich, H.-J. Fecht, Appl. Phys. Lett. 62, 3111 (1993)
6. R.K. Wunderlich, D.S. Lee, W.K. Johnson, H.-J. Fecht, Phys. Rev. B 55, 26
(1997)
7. R.K. Wunderlich, H.-J. Fecht, Meas. Sci. Technol. 16, 402 (2005)
8. H. Yasuda, I. Ohnaka, Y. Ninomiya, R. Ishii, S. Fujita, K. Kishio, J. Cryst.
Growth 260, 475 (2004)
9. H. Fukuyama, H. Kobatake, K. Takahashi, I. Minato, T. Tsukada, S. Awaji,
Meas. Sci. Technol. 18, 2059 (2007)
10. T. Tsukada, H. Fukuyama, H. Kobatake, Int. J. Heat Mass Trans. 50, 3054
(2007)
11. H. Kobatake, H. Fukuyama, I. Minato, T. Tsukada, S. Awaji, Appl. Phys. Lett.
90, 94102 (2007)
12. Y. Kraftmakher, Modulation calorimetry, Theory and Applications (Springer,
Berlin, 2003)
13. P.F. Sullivan, G. Seidel, Phys. Rev. 173, 679 (1968)
14. H. Kawamura, H. Fukuyama, M. Watanabe, T. Hibiya, Meas. Sci. Technol. 16,
386 (2005)
15. K. Higuchi, K. Kimura, A. Mizuno., M. Watanabe, Y. Katayama, K.
Kuribayashi, Meas. Sci. Technol. 16, 381 (2005)
16. J.H. Zong, B. Li, J. Szekely, Acta Astronautica 26, 435 (1992)
17. B.Q. Li, S.P. Song, Microgravity Sci. Technol. XI, 134 (1998)
18. V. Bojarevics, K. Pericleous, ISIJ Int. 43, 890 (2003)
19. R.W. Hyers, Meas. Sci. Technol. 16, 394 (2005)
20. P.B. Kantor, A.M. Kisel, E.N. Fomichev, Ukr. Fiz. Zh. 5, 358 (1960)
21. K. Yamaguchi, K. Itagaki, J. Therm. Anal. Cal. 69, 1059 (2002)
22. M. Olette, Compt. Rend. 244, 1033 (1957)
23. M.W. Chase Jr. (ed.), NIST-JANAF Thermochemical tables, 4th edn.
(American Chemical Society and American Institute of Physics for the National
Institute of Standards and Technology, Washington DC, 1998)
24. K. Yamamoto, T. Abe, S. Takasu, Jpn. J. Appl. Phys. 30, 2423 (1991)
25. E. Takasuka, E. Tokizaki, K. Terashima, S. Kimura, in Proc. the 4th Asian
Thermo5hys. Properties Conf. B1d3, 89 (1995)
26. T. Nishi, H. Shibata, H. Ohta, Mater. Trans. 44, 2369 (2003)
27. H. Nagai, Y. Nakata, T. Tsurue, H. Minagawa, K. Kamada, E. Gustafsson, T.
Okutani, Jpn. J. Appl. Phys. 39, 1405 (2000)
28. E. Yamasue, M. Susa, H. Fukuyama, K. Nagata, J. Cryst. Growth 234, 121
(2002)
29. N.E. Cusack, Rep. Prog. Phys. 26, 361 (1963)
30. V.M. Glazov, V.B. Kolftsov, V.A. Kurbatov, Sov. Phys. Semicond. 20, 1351
(1986)
31. H. Sasaki, A. Ikari, K. Terashima, S. Kimura, Jpn. J Appl. Phys. 34, 3426 (1995)
32. H.S. Schnyders, J.B. Van Zytveld, J. Phys. Condens. Matter 8, 10875 (1996)
9
Noncontact Thermophysical Property
Measurements of Refractory Metals Using
an Electrostatic Levitator
9.1 Introduction
The use of a containerless technique for materials processing has many techno-
logical and scientific advantages. The absence of a crucible allows the handling
of chemically reactive materials such as molten refractory metals, alloys, or
semiconductors, and eliminates the risk of sample contamination in overheated
and in undercooled states (liquid phase below melting temperature). This of-
fers excellent opportunities to characterize the structure of materials and to
determine accurately their thermophysical properties in those states. The lack
of a crucible also suppresses nucleation induced by the walls of a container
(heterogeneous nucleation), thus increasing the possibility of producing new
materials such as metallic glasses.
Several levitation methods, including acoustic, electromagnetic, aerody-
namic, and electrostatic have been applied for thermophysical property mea-
surements. The electromagnetic levitation method has been most popularly
used for metal samples because instrumentation is rather simple and compat-
ible with high vacuum. Density, surface tension, electrical resistivity, spectral
emissivity, and thermal conductivity of conductive materials are currently
measured using this method [1–4].
An alternative method for metal samples processing and study is the elec-
trostatic levitation, which uses Coulomb force between a charged sample and
electrodes. This method is applicable to conductive as well as nonconductive
samples but requires a high speed feedback control system to stabilize the
sample position. Due to this technical difficulty, the development of electro-
static levitation method was slower than other methods.
Rhim et al. from the Jet Propulsion Laboratory (JPL) developed a ground
based electrostatic levitation system [5] and successfully levitated and melted
refractory metals (zirconium [6] and titanium [7]) and semiconductors (sil-
icon [8] and germanium [9]). This group also developed several techniques
to measure thermophysical properties such as density [10], constant pressure
heat capacity [11], surface tension, viscosity [12], and electrical resistivity [13]
174 T. Ishikawa and P.-F. Paradis
with the electrostatic levitation system. More recently, the Japan Aerospace
Exploration Agency (JAXA) has improved the JPL technology to measure
thermophysical properties of refractory metals whose melting temperatures
are above 2,000 K [14]. This chapter briefly describes the electrostatic levita-
tion system and the thermophysical property measurement techniques. Typi-
cal data of superheated liquid and undercooled materials are also reported.
Top electrode
Band pass filter
Lens
He-Ne laser
Position
Side electrodes Sample detector
Bottom electrode
PC
A/D
D/A
(9) (9)
(2) (8)
(5)
(5)
(7) (7)
(11) (4)
(1) (12)
(13)
(7)
(6) (6)
(14)
(10)
(11)
(3)
Fig. 9.2. Schematic view of the electrostatic levitation furnace and its diagnostic
apparatus: (1) sample, (2) top electrode, (3) bottom electrode, (4) side electrodes,
(5) He–Ne laser, (6) position detector, (7) CO2 laser beam, (8) Nd:YAG laser beam,
(9) pyrometer, (10) CCD camera, (11) CCD camera with telephoto objective lens,
(12) UV lamp, (13) beam Splitter, (14) power meter for oscillation detection
176 T. Ishikawa and P.-F. Paradis
Top electrode
φ 25 mm φ 10 mm
Bottom electrode
Fig. 9.3. Electric field between the top and bottom electrodes obtained by a nu-
merical analysis
9.3.1 Density
The density can be measured using an UV imaging technique. Once the sam-
ple is molten, it will be in spherical shape due to surface tension and the
distribution of surface charge. Also, the molten sample is axi-symmetric since
the sample is rotated in vertical axis by the rotating magnetic field. Because
the sample is axi-symmetric and the mass is known, density(ρ) can be calcu-
lated as:
3m
ρ= , (9.1)
4πr3
where m and r are the mass and the radius of the sample.
The radiance temperature measured by the pyrometer is calibrated by
using the recalescence of the sample (sudden temperature rise from the un-
dercooled temperature to the melting temperature (Tm ) due to the release
of latent heat of fusion). After cooling the sample, both the images and the
cooling curve are recorded to determine the density. The recorded video im-
ages are digitized and matched to the cooling curve. Then, a JAXA developed
program extracts the area from each image and calculates the density at each
temperature.
A good contrast between sample and background is important to get a
precise result by the image analysis. Refractory metals become luminous at
elevated temperatures and it is hard to get an excellent contrast with the
background. Our solution to this problem is to use a high-pass filter at 450 nm
in conjunction with an UV lamp (Hoya–Schott EX200W) and to observe the
sample only in the UV range.
Figure 9.4 clearly depicts the differences between a room temperature
and an overheated liquid zirconium sample under different background con-
ditions [16]. When a white light illumination source is used, a solid zirco-
nium sample, originally appearing black on a white background (Fig. 9.4(a)),
changes to a white and glowing sample while heated (Fig. 9.4(b)). If the sam-
ple is melted and overheated, it becomes so bright as if it makes the camera
blind (Fig. 9.4(c)). If no background light is used, the solid sample is not so
178 T. Ishikawa and P.-F. Paradis
Fig. 9.4. Effect of the background on the appearance of a sample at several tem-
peratures
The surface tension and viscosity are determined by the drop oscillation
method, for which the frequency of the surface oscillation of the levitated
sample is measured around its equilibrium shape. In this method, a sample
is molten and brought to a selected temperature. Then, a P2 (cos θ)-mode of
drop oscillation is induced to the sample by superimposing a small sinusoidal
electric field on the levitation field. Here, P2 (cos θ) is a Legendre polynomial of
second-order. An oscillation detection system, illustrated in Fig. 9.5(a), mea-
sures the fluctuation of the vertical diameter of the molten sample with a sam-
pling frequency of 4096 Hz. The transient signal that followed the termination
of the excitation field is shown in Fig. 9.5(b). This signal is analyzed using an
in-house written LabVIEWT M program. This is done for a few times at given
temperature and repeated in several temperatures. Using the characteristic
9 Thermophysical Property of Metals Using Electrostatic Levitator 179
(a)
Photo detector
with vertical slit Top electrode
He-Ne laser
Sample
shadow Sample
Bottom electrode
To Beam Splitter
position sensor
(b)
40
Amplitude (a.u.)
−40
0 0.1 0.2 0.3 0.4 0.5 0.6
Time (s)
Fig. 9.5. Sample oscillation detection for surface tension and viscosity measure-
ments
and
Y 2 = E 2 r0 ε0 (9.5)
respectively, and E is the applied electric field. The characteristic oscillation
frequency ωc of molten refractory metal droplet (ca. 2 mm in diameter) ranges
from around 180 to 240 Hz.
Similarly, using the decay time τ given by the same signal, the viscosity η
is found by [17]
ρr02
η= . (9.6)
5τ
The density data ρ are known by conducting measurements described in 9.3.1
and the radius of the sample r0 is determined by the image analysis of the
recorded sample image during oscillation experiment. The drop charge Q can
be calculated by
mg = QE, (9.7)
where g is the gravitational acceleration.
While measuring surface tension and viscosity, liquid samples are inten-
tionally rotated to suppress the excitation of oscillation modes other than the
P2 (cos θ)-mode or nonaxisymmetric oscillations. The characteristic oscillation
frequency ωc and the decay time τ are dependent on the rotation rate of the
liquid sample Ω, by the following equations:
2
Δωc 19 Ω
= , (9.8)
ωc 21 ωc
2
1 1 2 Ω
= 1− , (9.9)
τ + Δτ τ 3 ωc
where Δωc and Δτ are the respective deviations of the characteristic oscilla-
tion frequency and decay time due to rotation. Since the rotation rate of the
liquid sample cannot be measured accurately, it is estimated by monitoring the
aspect ratio of the sample (ratio of the horizontal radius and vertical radius)
and by using the empirical equation calculated by Brown and Scriven [18] and
experimentally confirmed by Rhim and Ishikawa [19]:
2
Rhorizontal −2 Ω Ω
≈ 1 + 1.476 × 10 + 1.2532
Rvertical ωc ωc
3
4
Ω Ω
−1.7877 + 3.7385 . (9.10)
ωc ωc
Using (9.8)–(9.10), relations between the aspect ratio and Δωc or Δτ can
be estimated. If the aspect ratio of a rotating sample is less than 1.02, the
reduced sample rotation rate (Ω/ωc ) is less than 0.13 and effects of the sample
rotation to both the characteristic oscillation frequency and the decay time
should not be greater than 1.5%. In our experiment, the aspect ratio of the
9 Thermophysical Property of Metals Using Electrostatic Levitator 181
The density data for each metal are listed in Table 9.1 with literature val-
ues [2,6,10,14,17,20–60]. During these experiments, the density was measured
182 T. Ishikawa and P.-F. Paradis
15500
Density (kgm−3)
15000
Present work
Allen
14500 Shaner
Berhault
Tm
14000
2600 2800 3000 3200 3400 3600
Temperature (K)
Fig. 9.6. Density of tantalum as a function of temperature
over large temperature ranges including regions above and below the melting
temperature. The density, like that of alkaline metals, exhibited a linear be-
havior as a function of temperature. As an example, Fig. 9.6 shows the density
of tantalum as a function of temperature, including literature values. Since
our measurement covers a wide temperature, the temperature coefficient of
density can be determined with higher accuracy.
A simple relationship between the temperature dependence of the den-
sity of liquid metals and their boiling temperatures (Tb ) was proposed by
Steinberg [61]. He collected liquid density data at the melting point and the
temperature dependence of liquid density for 44 elements and found the fol-
lowing empirical relations
9 Thermophysical Property of Metals Using Electrostatic Levitator 185
15
−dρ/dT(10−4 kgm−3 K−1)
Pu
Pb Tl
Pd
Ni Se Cd
Os Bi
Co W Au
U
10 Mn Zn
Ta Fe Re Ag Pt
Cu Ir
Rh
Sn Cr In Pd
Te Presentwork
Mo Ni Steinberg
Cs
Nd Ru
5 Rb V
Ge Ga Hf
Al Si
Zr Nb
Na K Ba
Sr
Ca Pr
Y Mg La Ce
Li Be Ti
0
0 2 4 6 8 10
ρ00 /Tb (kgm−3K−1)
Fig. 9.7. Correlation of dρ/dT with ρ00 /Tb for elements. Pt points by Ref. [61]
falls outside of the range of figure (its ordinate is 28.8 and its abscissas is 6.05). The
solid line is the best fit to data and the dashed lines represent the 20% error cone
from Ref. [61]
dρ ρ00
− ∝ , (9.13)
dT Tb
where ρ00 was the virtual density of the liquid at 0 K as determined by ex-
trapolation from ρm and Tm with:
dρ
ρ00 = ρm − Tm . (9.14)
dT
Figure 9.7 illustrates the correlation of −dρ/dT with ρ00 /Tb . In his study,
most of the metal elements followed the correlation, except for mercury and
platinum. In Fig. 9.7, our measured data of refractory metals including that of
platinum were also plotted. Our platinum data (−0.96 kg m−3 K−1 ), as well as
those for other refractory metals, shows a good agreement with the Steinberg’s
relation.
The measured values and literature data for surface tension and viscosity
are listed in Table 9.2 [6, 20, 22, 28, 30, 32, 33, 37, 62–84]. Figure 9.8 shows
186 T. Ishikawa and P.-F. Paradis
1650
Present work
1550 Allen
Peterson
Kostikov
Tm
1500
1450
2200 2300 2400 2500 2600 2700
Temperature (K)
Fig. 9.8. Surface tension of hafnium as a function of temperature
the surface tension of hafnium. The surface tension could be measured over
large temperature ranges including the undercooled phase in these experi-
ments, whereas measurements by other methods could be done only around
the melting temperatures (Tm ). These tendency become clearer for higher Tm
samples.
Kasama et al. proposed an empirical equation on a physical model [85].
Based on this model, surface tension and its temperature dependence can be
expressed as
2
1 π 2 C 2 δ 2 Tm ρ 3$ 1 1
%2
γ= 2 (α + 1)ρ 3 − ρm3
, (9.15)
2 NA M 3 ρm
dγ 1 π 2 C 2 Tm Λδ 2 $ 2
−1
%
+ ρ− 3 − 3(1 + α)ρm3 ,
1 1
=− 2 2(α + 1)2 ρ 3 ρm
3
dT 3 NA M 3
where NA is the Avogadro constant, M is the atomic number, and Λ is the
temperature dependence of density (-dρ/dT ). In addition, C is a constant
derived from Lindemann’s theory of melting (ranged from 2.8 × 1012 to 3.1 ×
1012 ), δ is a ratio between the characteristic vibration frequency in the liquid
phase and solid phase (estimated to be around 0.5), and α is a constant related
to the distance, where an atomic attractive force is effective (ranged from 0.45
to 0.65).
At the melting temperature, the temperature coefficient of the surface
tension can be calculated by
9 Thermophysical Property of Metals Using Electrostatic Levitator 189
γmβm (10−7N/mK)
La Hg Nb Ni Pd
Ta Pt
Cu
Ru
Cd Mn
Be
−0.3
Mg Al Rh
W
Ce Os
Au
Ni
−0.4 Cr
Fe
Co
−0.5
−0.6
Fig. 9.9. Correlation between λm βm and dγ/dT . Open circles, open squares, and
open diamonds represent data from [32], [86], [87], respectively. The solid line is the
best fit to data
dγ 2 Λγm 2α + 1 Λγm
=− =K = Kγm βm , (9.16)
dT 3 ρm α ρm
−2(2α + 1)
K≡ ,
3α
where γm is the surface tension at the melting temperature. This equation
suggests that the temperature dependence of the surface tension is propor-
tional to the product of the surface tension at the melting temperature by the
thermal expansion coefficient.
Validity of this formula for liquid metals is checked by using literature
data [32, 86, 87] and our measurements. Results are shown in Fig. 9.9. Litera-
ture data of alkaline metals showed good agreements with (9.16), while those
of transition metals exhibited scattering, some of them being far from the
relation. However, our measured data show the same tendency as the alkaline
metals. In particular, platinum and nickel data follow the relation, while those
appearing in the literature do not.
Temperature dependence of the surface tension is an important parame-
ter to evaluate the magnitude of the thermo-capillary flow when the floating
zone method or Czochralski method are used for single crystal growth of
190 T. Ishikawa and P.-F. Paradis
semiconductors. Based on our results and alkaline data, the temperature de-
pendence of the surface tension of metals can be estimated if γm and the
thermal expansion coefficients (βm ) are known.
9.4.3 Viscosity
Table 9.3 shows the viscosity data for refractory metals [88–95], and Fig. 9.10
illustrates the viscosity of tungsten versus temperature. To our knowledge, the
viscosity data of several elements (Y, Nb, Mo, Ru, Ta, W, Re, Os, and Ir) were
the first to be reported. The viscosity measurement is a very unique capability
of the electrostatic levitation among a variety of levitation methods.
15
Viscosity (10−3 Pa.s)
10
Tm
0
3300 3400 3500 3600 3700 3800
Temperature (K)
Fig. 9.10. Viscosity of tungsten as a function of temperature
The relatively large uncertainty is mainly derived from the large uncer-
tainty in τ . Moreover, the drop oscillation method assume that no external
force is applied to the sample. In the case of our experiment, a strong electric
field is applied and a feedback position control is used, which might affect the
measurements.
192 T. Ishikawa and P.-F. Paradis
9.5 Summary
Thermophysical properties of several refractory metals over wide tempera-
ture ranges in the undercooled as well as in the superheated states could be
measured using the unique capabilities of the electrostatic levitation furnace.
This method can measure the temperature dependence of surface tension and
viscosity for refractory metals, which can not be measured by conventional
methods. Another advantage offered by this technique is its ability to deter-
mine the thermophysical properties at the melting temperature without any
risk of having a half-molten or solidified sample.
Acknowledgment
Authors would like to express their gratitude to Dr. J. Yu, Dr. T. Aoyama,
Ms. A. Ishikura, Mr. R. Ishikawa, Mr. R. Fujii, and Mr. N. Koike for help
in some experiments. This work is partially supported by a Grant-in-Aid for
Science Research (B) from the Japan Society for the Promotion of Science.
References
1. J. Brillo, I. Egry, Int. J. Thermophys. 24, 1155 (2003)
2. G. Lohöfer, J. Brillo, I. Egry, Int. J. Thermophys. 25, 1535 (2004)
3. H. Kawamura, H. Fukuyama, M. Watanabe, T. Hibiya, Meas. Sci. Technol. 16,
386 (2005)
4. H. Kobatake, H. Fukuyama, I. Minato, T. Tsukada, S. Awaji, Appl. Phys. Lett.
90, 094102 (2007)
5. W.-K. Rhim, S.K. Chang, D. Barber, K.F. Man, G. Gutt, A. Rulison, R.E.
Spjut, Rev. Sci. Instrum. 64, 2961 (1993)
6. P.-F. Paradis, W.K. Rhim, J. Mater. Res. 14, 3713 (1999)
7. P.-F. Paradis, W.K. Rhim, J. Chem. Thermodyn. 32, 123 (2000)
8. W.K. Rhim, S.K. Chung, A.J. Rulison, R.E. Spjut, Int. J. Thermophys. 18, 459
(1997)
9. W.-K. Rhim, T. Ishikawa, Int. J. Thermophys. 21, 429 (2000)
10. S.K. Chung, D.B. Thiessen, W.K. Rhim, Rev. Sci. Instrum. 67, 3175 (1996)
11. A.J. Ruliuson, W.K. Rhim, Rev. Sci. Instrum. 65, 695 (1994)
12. W.-K. Rhim, K. Ohsaka, P.-F. Paradis, Rev. Sci. Instrum. 70, 2796 (1999)
13. W.-K. Rhim, T. Ishikawa, Rev. Sci. Instrum. 69, 3628 (1998)
14. T. Ishikawa, P.-F. Paradis, T. Itami, S. Yoda, Meas. Sci. Technol. 16, 443 (2005)
15. P.-F. Paradis, T. Ishikawa, R. Fujii, S. Yoda, Heat Transf. Asian Res. 35, 152
(2006)
9 Thermophysical Property of Metals Using Electrostatic Levitator 193
16. T. Ishikawa, P.-F. Paradis, S. Yoda, Rev. Sci. Instrum. 72, 2490 (2001)
17. P.-F. Paradis, T. Ishikawa, T. Aoyama, S. Yoda, J. Chem. Thermodyn. 34, 1929
(2002)
18. R.A. Brown, L.E. Scriven, Proc. R. Soc. London A 371, 331 (1980)
19. W.-K. Rhim, T. Ishikawa, Rev. Sci. Instrum. 72, 3572 (2001)
20. V.P. Eljutin, M.A. Maurakh, Izv. A. N. OTN 4, 129 (1956)
21. B.C. Allen, Trans. AIME 227, 1175 (1963)
22. M.A. Maurakh, Trans. Indian Inst. Met. 14, 209 (1964)
23. V.P. Eljutin, V.I. Kostikov, I.A. Penkov, Poroshk. Met. 9, 46 (1970)
24. Yu.N.I. Ivashchenko, P.S. Martsenyuk, Zavod. Lab. 39, 42 (1973)
25. U. Seydel, W. Kitzel, J. Phys. F. Met. Phys. 9, L153 (1979)
26. T. Saito, Y. Shiraishi, Y. Sakuma, Trans. ISIJ 9, 118 (1969)
27. P.-F. Paradis, W.-K. Rhim, J. Chem. Thermodyn. 32, 123 (2000)
28. A.W. Peterson, H. Kedesdy, P.H. Keck, E. Schwarz, J. Appl. Phys. 29, 213
(1958)
29. V.P. Eljutin, V.I. Kostikov, I.A. Penkov, Poroshk. Met. 9, 46 (1970)
30. T. Ishikawa, P.-F. Paradis, Y. Saita, J. Jpn Inst. Met. 68(9), 781 (2004)
31. T. Saito, Y. Sakuma, J. Jpn. Inst. Met. 31, 1140 (1967)
32. B.C. Allen, in Liqud Metals Chemistry and Physics, ed. by S.Z. Beer (Marcel
Dekker, New York, 1972), p. 161
33. A.A. Fogel, T.A. Sidorova, G.E. Chuprikov, M.M. Mezdrogina, Izv. Akad. Nauk
SSSR Met. 1, 50 (1975)
34. Yu.N. Ivashchenko, P.S. Martsenyuk, High Temp. 11, 1146 (1973)
35. J.W. Shaner, G.R. Gathers, C. Minichino, High Temp. High Press. 8, 425 (1976)
36. P.-F. Paradis, T. Ishikawa, S. Yoda, Int. J. Thermophys. 23, 555 (2002)
37. G. Lang, in CRC Handbook of Chemistry and Physics, 72nd edn. ed. by D.R.
Lide (CRC Press, Boca Raton, FL, 1991–1992), pp. 4–124
38. P.-F. Paradis, T. Ishikawa, S. Yoda, J. Mater. Res. 19, 590 (2004)
39. P.-F. Paradis, T. Ishikawa, S. Yoda, Int. J. Thermophys. 24, 1121 (2003)
40. V.N. Eremenko, Yu.V. Naidich, Izv. Akad. Nauk. SSR O. T. N. Met. Topliva,
6, 100 (1961)
41. S.I. Popel, B.V. Tsareveskii, N.K. Dzhemilev, Fiz. Met. Metall. SSSR 18, 468
(1964)
42. P.-F. Paradis, T. Ishikawa, Y. Saita, S. Yoda, Int. J. Thermophys. 25, 1905-1912
(2004)
43. L.-D. Lucas, C. R. Acad. Sci. 253, 2526 (1961)
44. B.A. Vatolin, P.V. Esin, A.K. Ukhov, E.L. Dubinin, Trudy Inst. Met. Serdlovsk
18, 73 (1969)
45. P.S. Martsenyuk, Yu.V. Ivaschenko, Ukr. Khim. Zhur. (SU) 40, 431 (1974)
46. V.I. Arkhipkin, G.A. Grigoriev, V.I. Kostikov, in Fiz. Khim. Granits Razdeia
kontaktiryuschikh Faz, ed. by V.N. Eremenko (Naukova Dumka-Kiev, USSR4,
1976), pp. 74–77
47. T. Ishikawa, P.-F. Paradis, J. Electron. Mater. 34, 1526 (2005)
48. A. Berhault, L. Arles, J. Matricon, Int. J. Thermophys. 7, 167 (1986)
49. P.-F. Paradis, T. Ishikawa, R. Fujii, S. Yoda, Appl. Phys. Lett. 86, 41901 (2005)
50. A. Calverley, Proc. Phys. Soc. 70, 1040 (1957)
51. T. Thevenin, L. Alres, M. Boivineau, J.M. Vermeulen, Int. J. Thermophys. 14,
441 (1993)
52. P.-F. Paradis, T. Ishikawa, N. Koike, J. Appl.Phys. 100, 103523 (2006)
194 T. Ishikawa and P.-F. Paradis
Ta, 119, 175, 183, 187, 190, 191 uncertainty, 47, 57, 166–168, 181, 191
tantalum, 184 undercooled liquid, 1, 2, 8
telephoto camera, 177 undercooled melt, 132
temperature coefficient, 45 UV imaging technique, 177
temperature coefficient of density, 184 UV lamp, 177
temperature coefficient of surface
tension, 39, 161 V, 182
TEMPUS, 134, 139, 145 vapor pressure, 18
viscometer, 18, 19, 25
TEXUS sounding rocket, 133
viscosity, 13, 17, 19, 22, 36, 40, 111,
thermal conductance, 153
136, 141–143, 161, 173, 178, 180,
thermal conductivity, 85, 86, 101, 111,
181, 185, 190–192
149, 150, 154, 155, 157, 159, 161,
Vogel–Fulcher behavior, 144
162, 168, 170, 173
thermal contact resistance, 101 W, 119, 122, 124, 175, 183, 187, 190,
thermal diffusivity, 85–87, 91, 93, 94, 191
97–100, 103, 104, 107 weld pool, 39, 44, 45
thermal expansion, 139 weldability, 39, 44
thermal expansion coefficient, 161, 189, welding, 39, 44, 56, 149
190 wettability, 93
thermionic emission, 8 wetting, 30, 134
thermocapillary, 40, 158 Wiedemann–Franz law, 138, 149, 169
thermocapillary force, 150, 170 Wien’s equation, 112
thermocouple, 113
ThermoLab project, 143 X-ray absorption coefficient, 2
thermophysical property, 131, 136, 145, X-ray diffraction, 1–5, 9, 10
173, 176 X-ray scattering, 2, 7
Ti, 3, 46, 122, 173, 182, 186, 191 xenon arc lamp, 120
Ti39.5 Zr39.5 Ti21 , 3
Ti46 Al8 Nb, 141 Y, 182, 186, 190, 191
Y2 O3 , 3
Ti–Al, 124
YAG, 3
Ti–Fe–Si–O, 3
Ti–V, 124 ZARM, 133
TIG (tungsten inert gas), 44 ZnSe window, 5
titanium alloy, 119 Zr, 3, 7–10, 12, 120, 122, 125–127, 173,
TiZrNi, 3 177, 182, 186, 191
transition metals, 117, 122, Zr55 Al10 Ni5 Cu30 , 104
189 Zr60 Al15 Ni25 , 104
transmissivity, 99, 114, 117 Zr65 Al7.5 Cu17.5 Ni10 , 141
turbine blade, 56 Zr65 Al7.5 Cu27.5 , 104
two-axis diffractometer, 4 Zr-based alloy, 104, 105
two-colour pyrometer, 117 ZrO, 3