F.H.temperature Measurements of Materials (Springer 2008)

advances in materials research 11
advances in materials research

Series Editor-in-Chief: Y. Kawazoe
Series Editors: M. Hasegawa A. Inoue N. Kobayashi T. Sakurai L. Wille
The series Advances in Materials Research reports in a systematic and comprehensive

way on the latest progress in basic materials sciences. It contains both theoretically and
experimentally oriented texts written by leading experts in the f ield. Advances in Materials
Research is a continuation of the series Research Institute of Tohoku University (RITU).
1 Mesoscopic Dynamics of Fracture 7 Advanced Materials Characterization
Computational Materials Design for Corrosion Products
Editors: H. Kitagawa, T. Aihara, Jr., Formed on the Steel Surface
and Y. Kawazoe Editors: Y. Waseda and S. Suzuki
2 Advances 8 Shaped Crystals

in Scanning Probe Microscopy Growth
Editors: T. Sakurai and Y. Watanabe by Micro-Pulling-Down Technique
Editors: T. Fukuda and V.I. Chani
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
10 Frontiers in Materials Research

4 Materials Science Editors: Y. Fujikawa, K. Nakajima,
in Static High Magnetic Fields and T. Sakurai
Editors: K. Watanabe and M. Motokawa
11 High-Temperature Measurements
5 Structure and Properties of Materials
of Aperiodic Materials Editors: H. Fukuyama, Y. Waseda
Editors: Y. Kawazoe and Y. Waseda
12 Oxide and Nitride Semiconductors

6 Fiber Crystal Growth from the Melt Processing, Properties
Editors: T. Fukuda, P. Rudolph, and Applications
and S. Uda Editors: T. Yao, S.-K. Hong
Hiroyuki Fukuyama
Yoshio Waseda
(Eds.)
High-Temperature
Measurements
of Materials
With 125 Figures
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
Series Editors:
Professor Masayuki Hasegawa
Professor Akihisa Inoue
Professor Norio Kobayashi
Professor Toshio Sakurai
Institute for Materials Research, Tohoku University
Professor Luc Wille

Department of Physics, Florida Atlantic University
777 Glades Road, Boca Raton, FL 33431, USA
Advances in Materials Research ISSN 1435-1889
ISBN 978-3-540-85917-8 e-ISBN 978-3-540-85918-5
Library of Congress Control Number: 200893 61 4 5
© Springer Berlin Heidelberg 2009

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or
parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its
current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable
to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and
regulations and therefore free for general use.
Typesetting: Data prepared by SPi using a Springer LATEX macro package
Cover: eStudio Calmar Steinen
SPIN: 12254330 57/3180/SPI
Printed on acid-free paper
987654321
springer.com
Preface of Series by the Editor-in-Chief
This book titled High-Temperature Measurements of Materials is the 11th

volume in the series “Advances in Materials Research” edited by Professors
Hiroyuki Fukuyama and Yoshio Waseda. The book is composed of nine chap-
ters, the contents of which try to solve the currently important basic prob-
lems in high-temperature melts from scientific basics to real application fields
related to industrial problems. The contents of this present volume are ex-
pected to contribute to solving the most important problem of sustainability
of the global society by applying recent high-level scientific and engineering
researches in materials science. Recent important contributions from large
scale experimental facilities from over the world are collected and explained
in detail in this book, where we can now measure clearly the important behav-
iors of high-temperature melts which have never been observed by traditional
experimental techniques. This book also introduces recent advancements in
computer simulations, which have become very effective not only in explaining
experimentally observed phenomena but also in predicting materials proper-
ties and processes which are difficult to be measured for the materials under
high-temperature environments.
As the series editor, I thank Dr. Claus Ascheron of Springer-Verlag, who
always has interest in and kindly takes care of our research activity to encour-
age publication in this series of books.
Sendai, July 2008 Yoshiyuki Kawazoe

Preface
A variety of industries – information technology, aerospace, automobile, and

basic and new materials manufacturing – need technological innovations,
which bring high-value-added and high-quality products at low cost not only
because of global competition, but also because of the perspective of envi-
ronmental consciousness and regulation. Thermophysical properties of high-
temperature melts are indispensable for numerical simulations of material
processes such as semiconductor and optical crystal growth of the melt, and
casting of super-high-temperature alloys for jet-engine turbine blades, in addi-
tion to welding in automobile manufacturing. Recent developments in process
modeling provide 3D unsteady analysis of melt convection, temperature, and
heat flux distribution, which enables us to predict product quality. In fact,
3D process visualization using computer modeling helps us to understand
complicated phenomena occurring in the melt and to control the process.
Accurate data are necessary to improve the modeling, which cost-
effectively engenders high-quality products. However, crucial obstacles render
measurements of thermophysical properties difficult at elevated temperatures
because of high chemical reactivity and fluidity of melts. Substantial and
persistent challenges have been made to ascertain the precise thermophysical
properties of high-temperature melts. This book describes the new techniques
and latest developments in the measurements of atomic structure, density,
surface tension, viscosity, heat capacity, thermal and mass diffusivity, ther-
mal conductivity, emissivity, and electrical conductivity of high-temperature
melts.
In addition to up-to-date improvements in conventional techniques,
some new attempts are introduced to open a new scientific field, that is,
physics of high-temperature melts. Space-related organizations such as Japan
Aerospace Exploration Agency (JAXA) and German Aerospace Center (DLR)
demonstrate the importance of noncontact measurements and microgravity
environments. Recent progress in levitation techniques enables high-precision
measurements of various properties of stable and deeply undercooled metallic
melts. An electrostatic levitation apparatus is specially designed not only
VIII Preface
to measure thermophysical properties but also to study the short-range

order of metallic melts in combination with a synchrotron radiation facil-
ity: SPring-8 (Japan Synchrotron Radiation Research Institute, JASRI). A
noncontact measurement of thermal conductivity was developed using an elec-
tromagnetic levitator incorporating a superconducting magnet (High Field
Laboratory for Superconducting Materials, Institute for Materials Research,
Tohoku University). An electromagnetically levitated droplet behaves as a
hard sphere in a static magnetic field. The oscillation and convection of the
droplet are suppressed because of the Lorentz force, which enables measure-
ment of its true thermal conductivity using modulation laser calorimetry.
Utilization of microgravity conditions provides an ideal environment without
a fluid flow driven by buoyancy force. Diffusion coefficients of metallic melts
have been determined using the shear cell technique under microgravity using
a sounding rocket.
As described earlier, this book is a unique compilation of information re-
lated to recent advances in high-temperature measurements and thermophys-
ical properties data. The editors earnestly hope that this book is a useful
guide for the scientists and engineers who are working in the field of materials
science and processing, and that it is attractive to students interested in the
physics of high-temperature melts.
The editors thank Yoshimasa Ito and Miwa Sasaki at the Institute of Mul-
tidisciplinary Research for Advanced Materials (IMRAM), Tohoku University
for preparing TeX manuscripts and figures. The editors gratefully acknowledge
the encouragement and patience of Dr. Claus Ascheron of Springer-Verlag.
Sendai Hiroyuki Fukuyama

July 2008 Yoshio Waseda
Contents
1 Measurement of Structure of High Temperature

and Undercooled Melts by using X-Ray Diffraction
Methods Combined with Levitation Techniques
Tadahiko Masaki, Akitoshi Mizuno, and Masahito Watanabe . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Electrostatic Levitator for the Structural Analysis by X-Ray
Diffraction Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Viscosity and Density Measurements

of High Temperature Melts
Yuzuru Sato . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Viscosity Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 Capillary Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2 Oscillating Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.3 Rotating Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Density Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.1 Archimedean Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.2 Pycnometric Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.3 Manometric Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.4 Maximum Bubble Pressure Method . . . . . . . . . . . . . . . . . . . . . 33
2.3.5 Sessile Drop Method and Levitation Method . . . . . . . . . . . . . 34
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
X Contents
3 Marangoni Flow and Surface Tension

Taketoshi Hibiya and Shumpei Ozawa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Marangoni Effect on High-Temperature Melts . . . . . . . . . . . . . . . . . . . 39
3.2.1 Definition of Marangoni Flow . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2 Crystal Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3 Welding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Electron Beam Melting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5 Methods for Measuring Surface Tension: Oscillating Drop Method
Using Electromagnetic Levitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Surface Tension of Molten Silicon: Influence
of Oxygen on Surface Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.7 Surface Tension of Molten Iron and Iron-based Alloy . . . . . . . . . . . . . 54
3.8 Thermodynamic Approach for Adsorption of Oxygen
at Melt Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.9 Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4 Diffusion Coefficients of Metallic Melts Measured

by Shear Cell Technique Under Microgravity
and on the Ground
Shinsuke Suzuki . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Design of Shear Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.1 Principle of Shear Cell Technique . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.2 Minimization of Shear Convection . . . . . . . . . . . . . . . . . . . . . . 64
4.2.3 Minimization of Free Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4 Structure of the Shear Cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Experimental Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.1 Diffusion Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.2 Evaluation of Mean Square Diffusion Depth . . . . . . . . . . . . . . 67
4.4 Quantitative Measurement of Shear Convection
and Correction Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.4.1 Short-Time Diffusion Experiments . . . . . . . . . . . . . . . . . . . . . . 68
4.4.2 Time Dependence of Mean Square Diffusion Depth . . . . . . . . 70
4.4.3 Influence of Shear Convection . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Correction Method for the Determination
of Diffusion Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6 1g-Diffusion Measurements with Stable Density Layering . . . . . . . . . 72
4.6.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.6.2 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.6.3 Effect of Density Layering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Contents XI
4.7 Microgravity Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

4.7.1 Utilization of Microgravity Environment . . . . . . . . . . . . . . . . . 77
4.7.2 Microgravity Diffusion Experiments in Foton-M2 . . . . . . . . . 77
4.8 Temperature Dependence of the Diffusion Coefficients . . . . . . . . . . . . 79
4.9 Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5 Thermal Diffusivity Measurements of Oxide and Metallic

Melts at High Temperature by the Laser Flash Method
Hiroyuki Shibata, Hiromichi Ohta, and Yoshio Waseda . . . . . . . . . . . . . . . . 85
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2 A Brief Background of the Present Requirement for the Thermal
Property Measurements of High Temperature Materials . . . . . . . . . . 86
5.3 Experimental Procedures and Theoretical Basis
for the Laser Flash Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.4 Selected Examples of Thermal Diffusivities of Oxide Melts . . . . . . . . 94
5.5 Selected Examples of Thermal Diffusivities
of Metallic Melts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6 Emissivities of High Temperature Metallic Melts
Masahiro Susa and Rie K Endo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.2 Definition of Emissivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.3 Measurement Techniques for Emissivities . . . . . . . . . . . . . . . . . . . . . . . 112
6.3.1 Method Based on Wien’s Formula . . . . . . . . . . . . . . . . . . . . . . 112
6.3.2 Method Based on Optical Constants . . . . . . . . . . . . . . . . . . . . 113
6.3.3 Method Based on Direct Measurements of Radiation
Intensities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3.4 Other Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.4 Emissivity Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.4.1 Noble Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.4.2 Transition Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.4.3 Semiconducting Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.4.4 Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7 Noncontact Thermophysical Property Measurements

of Metallic Melts under Microgravity
Ivan Egry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.2 Microgravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
XII Contents
7.3 Containerless Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

7.4 Thermophysical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.4.1 Electrical Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.4.2 Density and Thermal Expansion . . . . . . . . . . . . . . . . . . . . . . . . 139
7.4.3 Specific Heat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.4.4 Viscosity and Surface Tension . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.5 Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8 Noncontact Laser Calorimetry of High Temperature Melts

in a Static Magnetic Field
Hiroyuki Fukuyama, Hidekazu Kobatake, Takao Tsukada,
and Satoshi Awaji . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.2 Theory of Modulation Calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.2.1 Heat Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.2.2 Thermal Conductivity and Emissivity . . . . . . . . . . . . . . . . . . . 153
8.2.3 Verification of the Assumptions of Conduction-Dominated
Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.2.4 Verification of the Model and Sensitivity Analysis . . . . . . . . . 159
8.2.5 Emissivity Determination from Cooling Curve . . . . . . . . . . . . 163
8.3 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.4.1 Motion of the Silicon Droplet . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.4.2 Temperature Response and Phase Difference . . . . . . . . . . . . . 164
8.4.3 Isobaric Molar Heat Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.4.4 Hemispherical Total Emissivity . . . . . . . . . . . . . . . . . . . . . . . . . 167
8.4.5 Thermal Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
9 Noncontact Thermophysical Property Measurements

of Refractory Metals Using an Electrostatic Levitator
Takehiko Ishikawa, Paul-François Paradis . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2 Electrostatic Levitation System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.3 Thermophysical Property Measurements . . . . . . . . . . . . . . . . . . . . . . . 177
9.3.1 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
9.3.2 Surface Tension and Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9.3.3 Experimental Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.4 Results of Thermophysical Property Measurements of Refractory
Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.4.1 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Contents XIII
9.4.2 Surface Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

9.4.3 Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
List of Contributors
Satoshi Awaji Hiroyuki Fukuyama

Institute for Materials Research, Institute of Multidisciplinary
Tohoku University Research for Advanced Materials
2-1-1 Katahira, Aoba-ku, Sendai (IMRAM), Tohoku University
980-8577, Japan, 2-1-1 Katahira, Aoba-ku, Sendai
awaji@imr.tohoku.ac.jp 980-8577, Japan,
fukuyama@tagen.tohoku.ac.jp
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
Rie K Endo Hidekazu Kobatake

Department of Metallurgy Institute of Multidisciplinary
and Ceramics Science, Tokyo Research for Advanced Materials
Institute of Technology (IMRAM), Tohoku University
2-12-1 Ookayama, Meguro-ku, Tokyo 2-1-1 Katahira, Aoba-ku, Sendai
152-8552, Japan, 980-8577, Japan,
rie@mtl.titech.ac.jp kobatake@tagen.tohoku.ac.jp
Paul-Francois Paradis Takehiko Ishikawa

Central Reseach Institute Japan, Aerospace Exploration
Mitsubishi Materials Corporation Agency (JAXA)
1002-14 Mukohyama, Naka, Ibaraki 2-1-1 Sengen, Tsukuba, Ibaraki
311-0102, Japan, 305-8505, Japan,
paradis@mmc.co.jp ishikawa.takehiko@jaxa.jp
XVI List of Contributors
Tadahiko Masaki Masahiro Susa

Shibaura Institute of Technology Department of Metallurgy
3-7-5 Toyosu, Koto-ku, Tokyo and Ceramics Science, Tokyo
135-8548, Japan, Institute of Technology
t masaki@sic.shibaura-it.ac.jp 2-12-1 Ookayama, Meguro-ku, Tokyo
152-8552, Japan,
Akitoshi Mizuno
susa.m.aa@m.titech.ac.jp
Department of Physics, Gakushuin
University
1-5-1 Mejiro, Toshima-ku, Tokyo Shinsuke Suzuki
171-8588, Japan, The Institute of Scientific
akitoshi.mizuno@gakushuin.ac.jp and Industrial Research, Osaka
University
Hiromichi Ohta
Faculty of Engineering, Ibaraki 8-1 Mihogaoka, Ibaraki, Osaka
University 567-0047, Japan,
4-12-1 Nakanarusawa-machi, Hitachi, suzuki@sanken.osaka-u.ac.jp
Ibaraki 316-8511, Japan,
ohta@mx.ibaraki.ac.jp Takao Tsukada
Shumpei Ozawa Department of Chemical
Department of Aerospace Engineering, Tohoku University
Engineering, Tokyo Metropolitan 6-6-07 Aoba, Aramaki, Aoba-ku,
University Sendai 980-8579, Japan,
6-6 Asahigaoka, Hino, Tokyo tsukada@pcel.che.tohoku.ac.jp
191-0065, Japan,
shumpei.ozawa@tmu.ac.jp Masahito Watanabe
Yuzuru Sato Department of Physics, Gakushuin
Department of Metallurgy, University
Tohoku University 1-5-1 Mejiro, Toshima-ku, Tokyo
6-6-02 Aramaki, Aoba-ku, Sendai 171-8588, Japan,
980-8579, Japan, masahito.watanabe@gakushuin.ac.jp
satoz@material.tohoku.ac.jp
Hiroyuki Shibata Yoshio Waseda
Institute of Multidisciplinary Institute of Multidisciplinary
Research for Advanced Materials Research for Advanced Materials
(IMRAM), Tohoku University (IMRAM), Tohoku University
2-1-1 Katahira, Aoba-ku, Sendai 2-1-1 Katahira, Aoba-ku, Sendai
980-8577, Japan, 980-8577, Japan,
shibata@tagen.tohoku.ac.jp waseda@tagen.tohoku.ac.jp
1
Measurement of Structure of High
Temperature and Undercooled Melts
by using X-Ray Diffraction Methods
Combined with Levitation Techniques
Tadahiko Masaki, Akitoshi Mizuno, and Masahito Watanabe
1.1 Introduction
The knowledge of the microscopic feature of matter is of paramount impor-

tance in materials science. In particular, the information about the atomic
configuration is essential for the understanding of the characteristic proper-
ties of disordered matter. Therefore, a huge amount of efforts has been devoted
to the development of experimental techniques coupled with X-ray or neutron
diffraction techniques [1–3] to study the structure of liquids. In this decade,
intense and high-energy X-ray beam sources, especially synchrotron radiation
facilities, have emerged, and can be used for diffraction experiments of disor-
dered matter. Compared to former experimental facilities, they enabled us to
perform the highly precise investigation of the structure of liquids in a much
shorter time.
Although the methods and facilities for diffraction experiments have been
improved rapidly, the sample handling techniques of high-temperature liquids
have not been advanced so much because of the difficulty in the selection of
crucible materials. In the case of liquid metals, several ceramics (e.g., fused
silica, sintered alumina, sapphire, graphite, boron nitride) have been used
for crucibles. Despite the adoption of these various materials, the maximum
temperature of the experiments has been limited due to the corrosion of the
crucible.
Levitation techniques use a variety of external forces (e.g., aerodynamic
[4], acoustic [5], electromagnetic [6], and electrostatic [7]) to hold a small
amount of material in space without a crucible. When a levitated sample is
in its liquid phase, it takes a spherical shape because the lack of crucible
minimizes the surface free energy. In particular, a great deal of attention has
been given to the measurement of the thermophysical properties of extremely
high-temperature melts and the study of solidification phenomena from deeply
undercooled liquids.
2 T. Masaki et al.
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
Levitation Material Structural analysis Ref

CNL Si XRD [8]
Si XRD [9]
Si IXS [10]
CuZr, NiZr XRD [11]
CuZr XRD [12]
Co80 Pd20 XRD [13]
Y2 O3 , Al2 O3 , B, Si, CoNi, Ni XRD, NS [14]
Al2 O3 IXS [15]
Al2 O3 XRD [16]
Al2 O3 , ZrO EXAFS [17]
Al2 O3 XRD [18]
MgAl2 O4 , CaAl2 O4 , Al2 O3 XRD, NS, IXS [19]
BaB2 O3 XRD [20]
Mg2 SiO4 , CuZr XRD [21]
Y2 O3 XRD, NS [22]
Y2 O3 AXRD [23]
YAG EXAFS [24]
SiO2 XRD [25]
EML Si XRD [26]
Si XRD [27]
Si ED-XRD [28]
Ni, Fe, Zr NS [29]
Ti NS [30]
Fe, Zr, Ni, Al65 Cu25 Co10 NS [31]
Al–Cu, Al–Ni XRD [32]
AuCu, PdCuSi, Si, CoPd EXAFS [33]
Co80 Pd20 EXAFS [34]
Co–Pd, Au–Cu–Co EXAFS [35]
Ni–V XRD [36]
Nd–Fe–B XRD [37]
Ti–Fe–Si–O ED-XRD [38]
ESL Si XRD [39]
Ni, Ti XRD [40]
Ti39.5 Zr39.5 Ti21 XRD [41]
TiZrNi XRD [42]
Ba–Ge XRD [43]
CNL Conical nozzle levitation, EML Electromagnetic levitation, ESL Electrostatic
levitation, XRD X-ray diffraction, NS Neutron scattering, ED-XRD Energy
dispersive X-ray diffraction, IXS Inelastic X-ray diffraction, AXRD Anomalous
X-ray diffraction, EXAFS Extended X-ray absorption fine structure
4 T. Masaki et al.
scattering factor of typical high-temperature metallic melts. In addition, as

the charged liquid sample is levitated between pairs of electrodes, the sample
is free from any obstacle, such as the nozzle or the coil in other levitators.
Moreover, to avoid electrical breakdown on the application of a high voltage
between electrodes, electrostatic levitators have to be operated either under
pressurized atmospheres (∼0.4 MPa) or under high vacuum. The high vacuum
conditions are particularly advantages for X-ray diffraction because there is
no need to consider the scattering from the ambient gas. Recently, Gangopad-
hyay et al. [44] used such levitators for X-ray diffraction in a synchrotron
radiation facility and observed the static structure factors and solidification
behavior of several metallic melts. Electrostatic levitation has also been ap-
plied by Aoki et al. [45] to neutron diffraction experiments. They successfully
measured the diffraction pattern of sintered alumina at room temperature. Al-
though the validity of electrostatic levitation for diffraction experiments has
been recognized, the previous electrostatic levitators exhibited limitations for
precise measurements. In particular, the limitation was present for the ob-
servable range of the diffraction angle, which affects the resolution of the data
obtained through a Fourier transform. The atomic configuration of liquids
in real space can be investigated from the radial distribution function, g(r),
which is obtained by a Fourier transformation of S(Q) as follows:
∞
1
g(r) = 1 + 2 S(Q) − 1 Q sin Qr dr, (1.1)
2π ρr 0
where ρ is the number density, S(Q) is the static structure factor, and Q is the
momentum transfer. The g(r) obtained from diffraction experiments involves
experimental errors, essentially because of the limited Q range of S(Q) due to
the wavelength of X-rays and the available range of 2θ, and the diverted tail of
the direct beam. In addition, the previously developed electrostatic levitators
could be used only in large beam source facilities (e.g., synchrotron radiation
facilites [44] or nuclear reactors [45]). Therefore, this situation restricts the
opportunities for experiments because of the limited machine time.
Recently, we developed an electrostatic levitator for X-ray diffraction mea-
surements with a high applicability to various beam sources [46]. Since the
system is very compact, it can be utilized coupled not only with the diffrac-
tometer at the high-energy X-ray diffraction beamline, BL04B2 of the syn-
chrotron radiation facility, SPring-8 [47], but also with a laboratory X-ray
system (RIGAKU RINT). The present electrostatic levitator was designed
for the X-ray diffraction measurements by a two-axis diffractometer with slits
collimation coupled with a germanium detector or a proportional counter.
The scattering intensity for each scattering angle, 2θ, was acquired by the
counter with the step-scan method. The high-energy X-ray beam from the
synchrotron source of over 100 keV is a very attractive probe for the liquid
structure analysis compared with the laboratory X-ray sources. The structure
of liquid 3d or 4d metals can be measured by using the high-energy X-rays
due to the high penetration of incident X-ray to samples. In addition, since
the momentum transfer, Q = 4π sin θ/λ (2θ, scattering angle; λ, wavelength

of incident X-rays), is proportional to the X-ray energy, the static structure
factor, S(Q), in sufficiently wide Q range can be obtained from the measure-
ment of diffraction pattern with small scattering angles. On the other hand,
the laboratory X-ray source can be used for diffraction experiments of lighter
materials, such as silicon. Since the laboratory X-ray source is free from the
restriction of user time of the facility, preliminary or challenging experiments
can be performed with trial and error. Therefore, the laboratory X-ray exper-
iments complement synchrotron X-ray ones.
In this report, we describe the electrostatic levitation system for the ob-
servation of high temperature liquid structures and present the results of a
preliminary application to the atomic structure analysis by X-ray diffraction
measurements.
1.2 Electrostatic Levitator for the Structural Analysis

by X-Ray Diffraction Technique
The design of the present apparatus was based on an electrostatic levitator

developed by Rhim et al. [7]. However, the present levitator was optimized
for the liquid structure analysis of high-temperature melts by X-ray diffrac-
tion technique. The apparatus consists of a vacuum chamber, a sample po-
sition control system, and a sample heating source. The sample, charged by
electronic emission, is levitated by applying an electrostatic field (typically
20–30 kV cm−1 for metallic materials) between two electrodes. To prevent the
electrical breakdown, the electrodes are contained in a chamber that is evacu-
ated to a level of vacuum lower than 1 × 10−4 Pa by a turbo molecular pump
attached directly to the side of the chamber. Figures 1.1 and 1.2 illustrate the
side and top views of the chamber, respectively. The chamber has a cylindrical
shape (height, 200 mm; diameter, 200 mm) and comprises several view ports.
A thin sapphire window (thickness, 0.5 mm; diameter, 17 mm) enables the in-
cident X-ray beam to reach the sample. With the use of a rectangular and
curved beryllium window, the intensity of the X-rays diffracted by the levi-
tated sample can be detected over a wide angle. The available range of 2θ is
−5◦ to 80◦ , which is wider than that previously reported [44]. Sufficiently wide
−1
Q range (Q ∼ 11.5 Å ) can be obtained even for laboratory X-ray source (Mo
Kα). Five silica glass windows, located on the top of the chamber, are used,
along with mirrors inside the chamber, for the position control system and
the sample observation by a camera. A ZnSe window (or lens) in the middle
of the top plate was used for the sample heating by a CO2 laser (wavelength,
10.6 μm; max. power, 240 W). A glass window on the side of the chamber was
employed for the temperature measurement by a single-color pyrometer. Two
valves located on the top plate acted as an air lock that is equipped to insert
samples without breaking the vacuum.
6 T. Masaki et al.
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
The design of the electrodes is of utmost importance for electrostatic levita-

tors. In our levitator, there are two main electrodes for vertical and horizontal
control and four side electrodes for additional horizontal control. The main
electrodes consist of two parallel disks. The upper electrode (40 mm diame-
ter), which is connected to a high voltage amplifier, is suspended from the top
plate by using insulating ceramic rods. The upper electrode has a spherical
end, which generates a concave electrostatic field that helps the position of
sample to stabilize laterally. In addition, a hole (3 mm diameter) in its cen-
ter is available for the sample heating by the CO2 laser. The lower electrode
(20 mm diameter) is electrically grounded and possesses a hole, which is avail-
able for sample handling by a positioning rod. This rod can be moved up and
down from the outside of the chamber to set the initial position of sample.
Four small spherical electrodes are distributed around the lower electrode for
additional control of the sample position along the horizontal direction.
To maintain stable sample levitation, the voltage of the electrodes is con-
trolled actively coupled with a position sensing system, a computer, and high-
voltage DC amplifiers. The position of the levitated sample is detected by two
sets of position sensors and associated He–Ne lasers. In each set, the expanded
He–Ne laser beam (10 mm diameter) illuminates the sample and its shadow
is projected on the position detector, which is located on the opposite side of
the He–Ne laser. The computer receives from the position detectors an electric
signal that corresponds to the sample position. It then calculates the control
signal by using a PID control scheme and sends the proper information to the
voltage amplifier that changes the voltage of the electrodes. By continuing this
sequence at a feedback rate of 1,000 Hz, the sample can maintain a fixed posi-
tion. The position control system used in this study is similar to that reported
in this literature [7], though the optical paths for position sensing are modified
because of the constraint of X-ray scattering facility. The He–Ne lasers and
the position detectors are located on the top plate, and therefore, the optical
paths of the lasers are bent twice by mirrors. This optical configuration offers
a wide observation view of the sample as well as the miniaturization of the
chamber. These enable us to set up the experimental configuration easier at
the synchrotron radiation facility.
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
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.
1.4 Results and Discussion

The zirconium, silicon, and alumina samples were representative materials for
metals, semiconductors, and ceramics, respectively. All these samples could be
levitated successfully in their molten states and could be maintained at a fixed
temperature for more than 1 h, which was sufficient for the measurement of
X-ray diffraction. The fluctuation of the sample position was less than 0.1 mm
for all materials during measurements. The diffracted X-rays were counted for
5 s for each diffraction angle. In addition, in the case of the laboratory X-ray
facility, the area of the incident X-ray beam (3 × 3 mm2 ) was much larger than
the cross-sectional area of the sample. Therefore, the sample fluctuation did
not affect the diffraction data. The diffraction patterns for all three materials
are shown in Figs. 1.3–1.5. As can be seen in Figs. 1.3 and 1.4, diffraction
−1
from the empty chamber (background) was almost negligible at Q > 1 Å .
This was attributed to the use of vacuum and proper shields to the detector.
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
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
by means of other types of levitators coupled with neutron or X-ray scatter-

ing. Our experimental results are in good agreement with those previously
published(Zr, [29, 31]; Si, [26–28, 39]; Al2 O3 , [16, 18]). However, the quality of
data, such as the Q range of S(Q) and low background, was significantly im-
proved. For example, the S(Q) of liquid zirconium was observed in the range
of Q = 0.5–20.0 Å, which was wider than that of a previous research [44].
To demonstrate the quality of the obtained S(Q), the effective pair potential,
ueff (r), was deduced for liquid zirconium based on the Modified Hypernetted
Chain approximation [53] as follows:
ueff (r)/kB T = g(r) − 1 − c(r) − ln g(r) + BHS (r, η), (1.3)
where c(r) is the direct correlation function, kB is the Boltzmann constant,
and T is the temperature. The c(r) was calculated from S(Q) as follows:

1 1
c(r) = 1− Q exp(−iQ · r) dQ (1.4)
2πρr S(Q)
The reliability of S(Q) in the small Q region is quite important for this calcu-
lation because the Fourier transform of the term, 1/S(Q), must be calculated.
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)
12. A Mizuno, S. Matsumura, M. Watanabe, S. Kohara, M. Takata, Mater. Trans.

46, 2799 (2005)
13. S. Krishnan, S. Ansell, D.L. Price, J. Non-Crystalline Solids 250–252, 286
(1999)
14. S. Krishnan, D.L. Price, J. Phys. Condens. Matter 12, R145 (2000)
15. H. Sinn, B. Glorieux, L. Hennet, A. Altas, M. Hu, E.E. Alp, F.J. Bermejo, D.L.
Price, M.-L. Saboungi, Science 299, 2047 (2003)
16. S. Ansell, S. Krishnan, J.K.R. Weber, J.J. Felten, P.C. Nordine, M.A. Beno,
D.L. Price, M.-L. Saboungi, Phys. Rev. Lett. 78, 464 (1997)
17. C. Landron, L. Hennet, D. Thaudiere, D.L. Price, G.N. Greaves, Nucl. Instrum.
Methods Phys. Res. B 199, 481 (2003)
18. C. Landron, L. Hennet, T.E. Jenkins, G.N. Greaves, J.P. Coutures, A.K. Soper,
Phys. Rev. Lett. 86, 4839 (2001)
19. L. Hennet, I. Pozdnyakova, V. Cristiglio, S. Krishnan, A. Bytchkov,
F. Albergamo, G.J. Cuello, J.-F. Brun, H.E. Ficher, D. Zanghi, S. Brassamin,
M.-L. Saboungi, D.L. Price, J. Non-Crystalline Solid, 353, 1705 (2007)
20. S. Matsushita, M. Watanabe, A. Mizuno, S. Kohara, J. Am. Ceram. Soc. 90,
742 (2007)
21. A. Mizuno, S. Kohara, S. Matsumura, M. Watanabe, J.K.R. Weber, M. Takata,
Mater. Sci. Forum 539–543, 2012 (2007)
22. L. Cristiglio, L. Hennet, G.J. Cuello, I. Pozdnyakova, A. Bytchkov, P. Palleau,
H.E. Fischer, D. Zanghi, M.-L. Saboungi, D.L. Price, J. Non-Crystalline Solid
353, 993 (2007)
23. L. Hennet, D. Thiaudiere, C. Landron, J.-F. Berar, M.-L. Saboungi, D. Matzen,
D.L. Price, Nucl. Instrum. Methods Phys. Res. B 207, 447 (2003)
24. C. Landron, X. Launay, J.C. Rifflet, P. Echegut, Y. Auger, D. Ruffier, J.P.
Coutures, M. Lemonier, M. Gailhanou, M. Bessiere, D. Bazin, H. Dexpert, Nucl.
Instrum. Methods Phys. Res. B 124, 627 (1997)
25. Q. Mei, C.J. Benmore, J.K.R. Weber , Phys. Rev. Lett. 98, 057802 (2007)
26. M. Watanabe, M. Adachi, T. Morishita, K. Higuchi, H. Kobatake, H. Fukuyama,
Faraday Discuss. 136, 279 (2007)
27. K. Higuchi, K. Kimura, A. Mizuno, M. Watanabe, K. Katayama,
K. Kuribayashi, Meas. Sci. Technol. 16, 381 (2005)
28. H. Kimura, M. Watanabe, K. Izumi, T. Hibiya, D. Hollad-Moritz, T. Schenk,
K.R. Bauchspeiss, S. Schneider, I. Egry, K. Funakoshi, M. Hanfland, Appl. Phys.
Lett. 78, 604 (2001)
29. T. Schenk, D. Holland-Moritz, V. Simonet, R. Bellissent, D.M. Herlach, Phys.
Rev. Lett. 89, 075507 (2002)
30. D. Hollad-Moritz, O. Heinen, R. Bellissent, T. Schenk, Mater. Sci. Eng. A 449–
451, 42 (2007)
31. D. Holland-Moritz, T. Schenk, V. Simonet, R. Bellissent, P. Convert, T. Hansen,
D.M. Herlach, Mater. Sci. Eng. A 375–377, 98 (2004)
32. J. Brillo, A. Bytchkov, I. Egry, L. Hennet, G. Mathiak, I. Pozdnyakova, D.L.
Price, D. Thiaudiere, D. Zanghi, J. Non-Crystalline Solids 352, 4008 (2006)
33. I. Egry, J. Non-Crystalline Solids 250–252, 63 (1999)
34. G. Jacobs, I. Egry, Phys. Rev. B 59, 3961 (1999)
35. G. Jacobs, I. Egry, D. Holland-Moritz. D. Platzek, J. Non-Crystalline Solid
232–234, 396 (1998)
36. C. Notthoff, B. Feuerbacher, H. Franz, D.M. Herlach, D. Hollad-Moritz, Phys.
Rev. Lett. 86, 1038 (2001)
16 T. Masaki et al.
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
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].
2.2 Viscosity Measurement
Viscosity of the liquid decreases with an increase of temperature. The temper-

ature dependence obeys Andrade law, which is similar to Arrhenius behavior
as shown in (2.1).
E
η = A exp . (2.1)
RT
18 Y. Sato
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.
2.2.1 Capillary Method
Principle of the capillary method is simple. The viscosity, η is expressed by

well known Hagen–Poiseuille’s equation (2.2) in the case of vertical type of
the capillary viscometer.
πr4 ρgh mρV 1
η= t− , (2.2)
8(L + nr)V 8π(L + nr) t
where ρ is the density of liquid, r and L are the radius and length of the
capillary, h is the effective height of the liquid column, V is the volume of
the liquid that has flowed, g is the gravitational acceleration, t is the time for
the flow out of liquid, m and n are constants. For an identical viscometer, the
coefficients of right hand terms in (2.2) except t and ρ are constant; therefore
(2.2) is rewritten to (2.3) by introducing kinematic viscosity, ν:
ν = η/ρ = C1 t − C2 /t. (2.3)
As the determination of the constants, C1 and C2 based on the apparatus
dimensions in (2.2) are not realistic. They are usually calibrated by using
a standard liquid of known kinematic viscosity through the measurement of
the time for it to flow out of the liquid. Distilled water is most popular as a
standard liquid for low viscosity due to the well known kinematic viscosity.
Figure 2.1 shows a capillary viscometer made of quartz for high temperature
molten salts [4]. The melt under measurement is sealed completely inside.
Then the melt with relatively high vapor pressure is allowed even if it is
molten AlCl3 with considerably high vapor pressure and strong hygroscopicity
[7]. The inner diameter and the length of the capillary were about 0.4 and
80 mm, respectively. Therefore, Reynolds number in the capillary was less
2 Viscosity and Density Measurements of High Temperature Melts 19
Filtration chamber
Quartz frit
Connection tube
to be sealed
Hole for flow of melt
Funnel
Upper fiducial mark
Timing bulb
Lower fiducial mark
Capillary
Slot for thermocouple
Outlet for melt of

suspended level type
Melt
Fig. 2.1. Capillary viscometer made of quartz
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
Quartz supporting tube

with silica radiation baffles
Halogen
lamp Capillary viscometer
Thermocouple
Axis of Gold furnace

rotation
Steel frame
Fig. 2.2. Whole apparatus of high temperature capillary viscometer

l
Li
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
0.8 0.9 1.0 1.1 1.2 1.3
T -1, 10-3K-1
Fig. 2.3. Viscosities of alkaline halides
2.2.2 Oscillating Method
Principle of the oscillating viscometer is based on the internal friction be-

tween the hypothetical layers assumed in the liquid. The oscillation is created
by giving the alternated rotational motion to axially symmetrical vessel con-
taining liquid or vibrating the plate immersed in the liquid. The former of
suspension type by a wire is popular and found to be precise. The amplitude
of the rotational oscillation caused in the liquid is attenuated by the inter-
nal friction if no external force is applied and the attenuation goes along the
rule of logarithmic decrement. However, to obtain the numerical solution of
the viscosity based on the logarithmic decrement measured is very difficult.
Therefore, the calibration method based on “Knappwost’s relationship” [9]
has been used previously. In this method, some reference calibration liquids
with precisely known viscosity and density are required. However, it is con-
siderably difficult to find the adequate reference liquids at high temperatures,
and also it is easily predicted that the calibration curve is not so reliable if con-
siderably different liquids such as molten metals and organic liquids are used
for the calibration of identical apparatus. Therefore, it is not expected that
precise and reliable determination of the viscosity can be achieved by using
the calibration method. On the other hand, noncalibration methods usually
called as absolute methods are recently used in most cases. Their principle is
to determine the viscosity based on the mathematical calculation using loga-
rithmic decrement and the period of oscillation, together with other physical
parameters such as the mass and density of the liquid, the moment of inertia,
the inner diameter of the cylindrical vessel, etc. The absolute methods are,
in general, very complicated mathematically. However, it is great advantage
that no reference liquids are required. Roscoe’s equation [10] shown in (2.5)
may be most popular absolute algorithm. It was confirmed that the equation
gave good reproducibility to obtain the viscosity even if the parameters are
changed, for example, considerable changing the logarithmic decrement and
the period of oscillation by alternating the dimension of the crucible or the
suspension wire. The equation for cylindrical vessel is described as follows:

2
Iδ 1
η= , (2.4)
πR3 HZ πρτ

1R 3 4R 1 3 9R a2
Z = 1+ a0 − + + + +0
4H 2 πH p 8 4H 2p2

63 45R a4
− − + ···,
128 64H 4p4
√ 12 √ 12
1 + Δ2 + 1 1 + Δ2 − 1
a0 = (1 − Δ) − (1 + Δ) ,
2 2
√ 12 √ 12
1 + Δ2 + 1 1 + Δ2 − 1 a2
a2 = + , a4 = √ ,
2 2 1 + Δ2
22 Y. Sato

12

πρ δ Ai W
p= R, Δ= δ = ln , H= ,
ητ 2π Ai+1 πR2 ρ
where η, δ, τ , A, I, R, W , and ρ are viscosity, logarithmic decrement, period of
oscillation, amplitude of the oscillation, moment of inertia of whole suspension
system, inner radius of crucible, mass, and density of the melt.
To solve the equation analytically is difficult due to the recursion of the
viscosity given in the aforementioned equation. Therefore, the successive ap-
proximation is used to obtain the viscosity numerically. In the actual calcu-
lation, the logarithmic decrement is given as a difference between the values
obtained for the crucible containing melt and the empty crucible, respectively.
As the parameters experimentally determined from the oscillation are the
logarithmic decrement and the period of oscillation, it is important how to
determine them. Previously, the logarithmic decrement had been determined
through an observation in the amplitude of oscillation by using a lamp scale or
a recorded profile on the film, etc. because PC was not available until 1970’s.
Now the time interval measurement [5] is simpler than direct measurement
of the amplitude of the oscillation and is sufficiently precise compared with
others.
An oscillatory curve with decay expressed by (2.5) is shown in Fig. 2.4. L
and −L are the positions of the detectors. C is the deviation of the center of
oscillation from the center of detectors. The period, τ , and the logarithmic
decrement of oscillation, δ, are defined by (2.6) and (2.7).
Y = A exp(−Bt) sin ωt + C, (2.5)
τ = 2π/ω, (2.6)
δ = 2πB/ω = Bτ, (2.7)
L
C
t
-L
t1 t2 t3 t4
Fig. 2.4. Oscillating curve with decay and the time intervals measured in a period
where ω is an angular frequency and B is a constant related to the decay. Four

time intervals, t1 through t4 , measured in a period were used for determining
τ and δ. It is difficult to obtain analytical solutions for τ and δ if C is not
zero. Following approximation procedure is available. At first, oscillatory curve
without decay is assumed as expressed by (2.8). Therefore, the primary period
of the oscillation, τ is given in (2.9).
Y = A sin ωt + C, (2.8)
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)
C = −A[sin{π(0.5 − t1 /τ )} + sin{π(1.5 − t3 /τ )}]/2. (2.11)
A primary constant B is calculated with (2.12) using a series of Ai , which

is the amplitude obtained for sequential individual periods. And a primary
logarithmic decrement, δ , is given by (2.13).
B = (1/τ , ) ln(A/Ai+1 ), (2.12)
δ = B τ . (2.13)
This set of first approximation values, δ and τ includes a considerable er-

ror depending on the logarithmic decrement as shown in Fig. 2.5. The error
can be simulated by using a sample data of the time intervals, which was
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
generated with given parameters of fixed C of 10 mm, 2L of 100 mm, and

various B, although they are relatively worse in condition than that of the
real measurement. The deviations of the first approximate values, δ and τ ,
from true values were about from 10−1.5 % to 10−0.5 % for the practical range
of logarithmic decrement in the measurement. They are only available when
the decay is impractically small.
The effect of decay on the oscillation was taken into account in the next
step. The oscillatory curve with decay and the relationship between the pri-
mary periods given by (2.9) and the true period are shown in Fig. 2.6. A pri-
mary period, τ , is always longer than a true period, τ . The true period can
be determined if a difference (τ − τ ) is estimated precisely. Figure 2.7 shows
the overlapped consecutive half period with decay. The curve is apparently
distorted sine wave and is vertically similar to the consecutive wave. The dif-
ference, (τ − τ ), is apparently equal to ta from Fig. 2.6, although ta is not
measurable. However, the following relationship in (2.14) is helpful.
t1,i − t1,i+1 = ta + tb , (2.14)
ta can be determined if the measurable left hand term, (t1,i − t1,i+1 ), is rea-
sonably divided into ta and tb . The times, at which a differential coefficient
of a wave equal to zero, are the same for any period as shown in Fig. 2.7.
Therefore, the left hand term in (2.14) was divided at the time mentioned
above to distribute the left hand term into ta and tb by assuming that the
ratio of ta to tb is equal to the ratio of divided parts. As a result, a second
approximate value of τ is obtained as (2.15).
τ = τ − ta = τ − (t1,i − t1,i+1 ) tan−1 (ω/B)/π, (2.15)
C
ta t
tb
1 -1 w p 1 -1 w
w tan ( B ) w
-
w tan ( B )
Fig. 2.7. Overlapped consecutive half waves with decay
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
colimator lens oscillation

initiator
W rod
convex lens
laser source Crucible
Fig. 2.8. Suspension and detection system in the oscillating viscometer
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.
2.2.3 Rotating Method
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
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
(Temperature, T)−1 / 10−3 K−1
Fig. 2.10. Viscosities of molten metals and semiconductors

28 Y. Sato
determined to obtain the viscosity of the liquid. In the rotating viscometer,

the viscosity is represented by following (2.16).
T (r12 − r22 )
η= , (2.16)
4πhr12 r22 (ω1 − ω2 )
where T , r, h, and ω are torque, radius of cylinders, length of cylinders, and
angular velocity of cylinders. Suffixs 1 and 2 mean outer and inner cylinders.
Two different types of the apparatus are used for the rotating method. One
consists of a rotating inner cylindrical column and a fixed outer cylinder which
allows infinite radius. In this type, the rotating inner column accepts the roles
of both creating and measuring torque. Therefore, measurable range of the
torque is, in general, in considerably large viscosity due to the mechanical
friction loss along the route from the column to the device of measurement
and the value obtained is generally not so precise. This type of the viscometer
is relatively simple for the construction of apparatus, which makes the mea-
surement easy and most of the apparatus commercially available are of these
types. It is adaptable to measure the viscosity not only of the liquid in small
crucible but also for large scale of the liquid such as a big pool of molten glass.
Another one is more essential and precise. It consists of rotating outer cylin-
der and fixed inner column that detects the torque. The feature of the type
of viscometer is of higher precision than former type due to nonfriction loss
for torque measurement, although the dimensions of inner and outer cylinder
should be fixed.
For the cases of rotating inner or outer cylinder, ω1 = 0 or ω2 = 0 in
(2.16), respectively. However, the equation can be adapted only for semi-
infinite length of cylinders. Therefore, calibration method is usually used for
determining the apparatus constant regarding to the dimensions at various
angular velocities by means of standard viscosity liquids such as silicon oils
with various viscosities. An example for the outer crucible rotating viscometer
is shown in Fig. 2.11 [6].
2.3 Density Measurements

As density is the most fundamental property for all the matter in any state
and temperature, many methods were propounded in the long history. Typical
methods for high temperature melts [2] are an Archimedean method, a pycno-
metric method, a manometric method, a maximum bubble pressure method,
a sessile drop method, a levitation method, a dilatometric method, a float
method, a γ-ray transmission method, etc. Among them, the Archimedean
method is the most popular and precise one for any liquids even at high tem-
peratures although it is considerably traditional. The levitation method is
recently developed for mainly molten metals. It has an advantage of no con-
tamination from the container and limitless in the experimental temperature
(a)Differential transformer
Differential
transformer Torsion wire To the torsion wire
Oil damper Secondary coil Coil holder
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
Fig. 2.11. Rotating viscometer for silicate melts
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
because of the containerless method. Additionally, the maximum bubble pres-

sure method and the levitation method are basically same to those used for
also surface tension measurement.
2.3.1 Archimedean Method
In the Archimedean method, the buoyancy worked on the immersed sinker

in the liquid is measured by using a balance. The buoyancy measurement is
relatively easy and considerably precise by using a recent electric balance.
It consists of direct and indirect methods as shown in Fig. 2.12 [1]. On the
30 Y. Sato
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.
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
Fig. 2.13. Archimedean densitiometer with dual sinkers
2.3.2 Pycnometric Method
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
Fig. 2.14. Pycnometer made of refractory material
precise data of expansion coefficient, because the direct determination of the

capacity at high temperature is difficult and error-prone. Figure 2.14 shows
an example of pycnometer made of boron nitride or high purity alumina for
measuring the density of molten silicon [13].
2.3.3 Manometric Method
As the height of liquid column under gravitational acceleration is proportional

to the liquid density and the pressure applied, density is obtained by measur-
ing the height of liquid column and the pressure. As the direct measurement
of the height is difficult, the difference between the heights of two columns
connecting in the shape of U-type tube is usually measured. The advantage
of using U-type tube is the cancellation of surface tension effect by using uni-
form inner radius of the tube. For this method, in many cases, two U-type
tubes connected in atmospheric path are used. One U-type tube contains high
temperature sample melt and other contains reference liquid with well known
density at certain temperature. The density of the sample melt is expressed
as in (2.19).
Δhref
ρmelt = ρref , (2.19)
Δhmelt
where Δh is the height difference in a U-type tube, and the suffixs “melt”
and “ref” mean the sample melt and reference liquid. Figure 2.15 shows an
example of the apparatus [14]. In this method, significant difficulty is in the
detection of the meniscus. A transparent furnace up to about 1,000◦ C was
Ar Vac.
A D
B E
F
C
A: Hg reservoir
B: Gold furnace
C: Molten sample
D: Thermocouple
E: Thermostat
F: Standard sample
Fig. 2.15. Apparatus of manometric method for molten salts
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.
2.3.4 Maximum Bubble Pressure Method
This method is usually adapted to density measurement although it is essen-

tially used for surface tension measurement. The principle is to measure the
hydrostatic pressure similar to the manometric method. A nozzle of thin tube
is immersed in the melt and inert gas is sent to the nozzle very slowly. Then,
a small bubble is formed at the end of the nozzle made of refractory materials
as shown in Fig. 2.16. The pressure in the bubble is summation of hydrostatic
pressure of the melt and the maximum pressure caused by the surface tension
of the bubble when the radius of the bubble is minimum as indicated in (2.20).
Pmax = Psurf + ρgh, (2.20)
34 Y. Sato
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.
2.3.5 Sessile Drop Method and Levitation Method
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
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
Fig. 2.18. Electrostatic levitation method for thermophysical properties measure-

ment
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)
9. A. Knappwost, Z. Metallk. 39, 314 (1948)

10. R. Roscoe, Proc. Phys. Soc. 72, 576 (1958)
11. H. Sasaki, E. Tokizaki, K. Terashima, S. Kimura, J. Cryst. Growth 139, 225
(1994)
12. Y. Sato, Y. Anbo, K. Yanagase, T. Yamamura, in CD-ROM Reprint Volume
for the 17th European Conference on Thermophysical Properties, 2005
13. Y. Sato, T. Nishizuka, K. Hara, T. Yamamura, Y. Waseda, Int. J. Thermophys.
21, 1463 (2000)
14. Y. Sato, K. Kobayashi, T. Ejima, J. Jpn. Inst. Metals 43, 97 (1979)
15. K. Mukai, Z. Yuan, Mater. Trans. JIM 41, 323 (2000)
16. T. Ishikawa, P.-F. Paradis, S. Yoda, Space Utiliz. Res. 21, 42 (2005)
3
Marangoni Flow and Surface Tension of High
Temperature Melts
Taketoshi Hibiya and Shumpei Ozawa
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.
3.2 Marangoni Effect on High-Temperature Melts

3.2.1 Definition of Marangoni Flow
The Marangoni effect appears on the melt surface where (a) the surface tension
gradient exists parallel to the melt surface (Fig. 3.1a), or (Fig. 3.1b) where the
40 T. Hibiya and S. Ozawa
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]
temperature gradient is normal to the surface. In the latter case, temperature

is distributed homogeneously at the melt surface, but the surface tempera-
ture is lower than that inside, as shown in Fig. 3.1b. The former is called the
thermocapillary or solutocapillary effect, depending on the origin of surface
tension difference at the melt surface. The latter, case (b), is a classic def-
inition of the Marangoni effect. For case (a), because of shear stress at the
melt surface, a melt surface with large surface tension pulls that with small
surface tension. Beneath the surface, flow takes place due to viscosity of the
melt. The surface tension gradient is caused not only by difference in tem-
perature, but also by that in concentration or adsorption of impurities. They
are called thermocapillary and solutocapillary flow, respectively. For case (b),
thermocapillary flow takes place locally, due to fluctuation of temperature at
the melt surface with nominally homogeneous temperature. To compensate
volume lost by this flow, it flows upward from the bottom at a high temper-
ature. This flow supplies high temperature melt from the bottom. Therefore,
once this flow is generated, it maintains the temperature difference at the melt
surface to enable continuous flow.
The balance of forces at the liquid surface is written as follows to describe
a Marangoni flow:
∂u ∂u ∂σ ∂T ∂σ ∂Ci ∂σ ∂Γ
−μl − μg = + + . (3.1)
∂z ∂z ∂T ∂x ∂Ci ∂x ∂Γ ∂x
Here, μl and μg are the dynamic viscosity for liquid and gas phases. u is
the velocity; z and x are the axes of the ordinate; T , σ, Ci , and Γ are the
temperature, surface tension, concentration of solute at the liquid surface,
and number of adsorbed atoms per unit area, respectively. The second term
of the right hand equation is related with solute, whose mass transport is rapid
between the adsorbed layer and the bulk liquid. The third term corresponds
to a condition where atoms or molecules exist only at the interface, or mass
transport between the adsorption layer and the bulk liquid is very slow. For
example, for molten silicon case, the second term can be neglected, since
number of adsorbed oxygen is supposed to be equilibrated with that of oxygen
3 Marangoni Flow and Surface Tension of High Temperature Melts 41
in melt; Ci can be written using Γ through the equilibrium constant. The

Magnitude of Marangoni flow is written as follows, depending on the driving
force of the flow:
∂σ ΔT · l ∂σ ΔC · l ∂σ ΔΓ · l
Ma = + + . (3.2)
∂T μl · κ ∂C μl · κ ∂Γ μl · κ
Here, l and κ are the characteristic length of the system and thermal dif-
fusivity, respectively. For thermocapillary flow, the magnitude of the flow is
expressed with the first term of the right side equation.
3.2.2 Crystal Growth
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
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]
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
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
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.
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]
3.4 Electron Beam Melting
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.
3.5 Methods for Measuring Surface Tension: Oscillating

Drop Method Using Electromagnetic Levitation
Surface tension is measured by several methods, such as (a) capillary rise
method, (b) maximum drop method, (c) maximum bubble pressure method,
(d) sessile drop method, (e) pendant drop method, (f) drop weight method,
and (g) oscillating drop method [25]. For surface tension measurements of high
temperature melts, such as molten metals, alloys, and semiconductors, data
for the ambient atmosphere must be attached, because surface tension is sensi-
tive to surfactants of the group-16 elements, particularly oxygen. Temperature
range for measurement must also be extended, because uncertainty for the
temperature coefficient is reduced by measurement in the wide temperature
range. For these purposes, an oscillating drop method using levitation can sup-
ply a solution, because the containerless technique assures contamination-free
conditions from the crucible wall and the wide temperature range, including
undercooled condition due to the fact that there is neither a contamination
source nor a nucleation site. Oxygen partial pressure can be controlled only
by using electromagnetic levitation, although contamination-free conditions
and a wide rage of temperature measurements can be realized using electro-
static levitation. For electrostatic levitation, it is very difficult to control the
ambient atmosphere, because discharge takes place between electrodes, when
introducing gas for oxygen partial pressure control. For the oscillating drop
method, surface tension can be obtained through the Rayleigh’s equation as
follows [26]:
3 2
σ= πν M. (3.4)
8
Here, σ is the surface tension. ν and M are the oscillation frequency and the
mass of the droplet, respectively. The above ideal relationship is valid only
under microgravity [27], whereas frequencies of m = 0, ±1, and ±2 oscilla-
tions for the l = 2 mode are not degenerated but split into five peaks, when
the oscillation frequency of a rotating droplet is observed from the top of the
rotation axis on earth as shown in Figs. 3.8a and 3.8b. Here the droplet is
no longer spherical but is deformed into an egg shape due to gravitational
acceleration as well as magnetic forces. Thus, surface tension must be calcu-
lated using the Cummings and Blackburn correction, including the frequency
of translational mode νt as follows [28]:
z 2
1 2
2
0
2
ν = ν − νt 1.9 − 1.2
2
, (3.5)
5 m=−2 2,m a

g 3M
z0 = 2 2 , a= 3 .
8π νt 4ρπ
Here, ν2,m is the frequency of m = 0, ±1, and ±2 oscillations for the l = 2
mode. The νt is the oscillation frequency of the translational mode, that is,
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
Table 3.1. Identification rule for the oscillations

m=0 m = ±1 m = ±2 Rotation
Area Yes Yes No No
R+ Yes Yes No No
R− No Yes Yes Yes
movement of gravitational center. a is the equilibrium radius of the droplet. g

and ρ are the gravitational acceleration and the density of the droplet, respec-
tively. The frequency of the m = 0, ±1, and ±2 oscillations can be identified
based on the rule shown in Table 3.1. However, both the frequencies of the
m = ±2 oscillation and of the rotation of elliptically deformed droplets ap-
pear in the R− signals, that is, the difference in radius along the x and y axes,
because a droplet is deformed into a twofold symmetric shape in both cases. If
there is no droplet rotation, m = +2 and m = −2 exhibit the same frequency
for m = |2|; a single peak appears. When the droplet rotates, frequencies for
m = +2 and m = −2 oscillations appear separately apart by two times of
rotation frequency [29]. In another word, the real frequency of the m = |2|
oscillation is located just between the frequencies of m = +2 and m = −2
oscillations. The difference in the m = +2 and m = −2 is featured with the
temporal phase difference between two oscillations, ΔΨ . This is derived from
the hemispherical harmonics to describe the shape of the droplet. It causes
apparent rotation, which exhibits a much higher frequency than that of real
droplet rotation. Thus, analysis of oscillation frequency becomes complicated.
The height of the m = +2 and m = −2 oscillations depends on the phase
difference, ΔΨ ; usually the height of these peaks is uneven. When a droplet
shows real rotation and the phase difference is ΔΨ = 0 or π, both m = +2 and

m = −2 oscillation peaks exhibit the same height. However, when ΔΨ = +π/2
or −π/2, only a single peak appears, even though the droplet really rotates.
The peak appears at the frequency shifted from that for m = |2|. Thus, it
is possible to misidentify the frequency for m = |2| oscillations. To avoid the
error, it is advisable to identify the two peaks for m = +2 and m = −2
oscillations [30, 31].
By superimposing a weak static magnetic field parallel to the axis of the
levitation coil axis, the rotation axis of a droplet is constrained so that it is
parallel to that of the coil, and mutation is also suppressed. Otherwise, the
rotation axis and the rotation rate exhibit time dependence. The application
of a magnetic field would make it possible to clearly identify the m = |2| fre-
quency and improve uncertainty of surface tension. However, the application
of a strong magnetic field damps surface oscillation and makes it almost im-
possible to identify frequencies. Furthermore, the application of the magnetic
field appears to shift m = 0 and ±1 frequencies higher. On the other hand,
the application of a strong magnetic field enables us to measure thermal con-
ductivity and density of metallic melts easily, because a strong magnetic field
can suppress flow and surface oscillation, that is, a droplet behaves as if it
were solid and axisymmetry of the droplet shape is assured [32, 33]
3.6 Surface Tension of Molten Silicon: Influence

of Oxygen on Surface Tension
For molten silicon, Keene collected and evaluated a great deal of published
data [3]. As shown in Fig. 3.9, the surface tension of molten silicon has been
classified into two groups; the larger the surface tension is, the larger the
temperature coefficient is [34–54]. This classification is attributed to oxy-
gen contamination from measurement devices and atmosphere. Keene advised
measuring surface tension using a contamination-free environment, such as the
use of levitation. Huan et al. and Mukai et al. measured surface tension and
its temperature coefficient as a function not only of temperature but also of
oxygen partial pressure of an ambient atmosphere P o2 [55, 56]. Figures 3.10
and 3.11 show the measurements carried out by Mukai et al. They are written
using equations as follows:
σ/mNm−1 = 831 − 29.5 ln(1 + 3.88 × 1010 P o2 2 )

1
at 1, 693 K, (3.6)
1
814 − 30.1 ln(1 + 3.06 × 1010 P o2 )
2 at 1, 723 K,
1
793 − 30.6 ln(1 + 2.47 × 1010 P o2 )2 at 1, 753 K,
1
774 − 31.0 ln(1 + 1.01 × 1010 P o2 )2 at 1, 773 K,
Fig. 3.9. Surface tension of molten silicon reviewed by Keene [3]
∂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)
860
Po2 <Po2sat Po2 >Po2sat

1693K
840
Po2sat = 1.32 ×10-20 MPa
Surface Tension: s (mN/m)
820 1723K
Po2sat = 4.27 ×10-20 MPa
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:
Sessile drop method

BN substrate
−0.2
−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. 3.11. Temperature coefficient of surface tension of molten silicon as a function

of oxygen partial pressure of the ambient atmosphere. Even above the equilibrium
P o2 for SiO2 formation, the temperature coefficient of surface tension is obtained [56]
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].
As mentioned earlier, surfactants such as oxygen or sulfur are recognized

as playing an important role in describing Marangoni flow through the surfac-
tant effect on surface tension and its temperature coefficient. This means that
surface tension must be reported to include data of oxygen partial pressure
or concentration of oxygen or sulfur in the melt. However, it is very diffi-
cult to ascertain the real oxygen partial pressure at the melt surface besides
that in the bulk atmosphere, whose oxygen partial pressure can be measured
using an oxygen sensor. Ricci’s group has tried to determine the relation-
ship between real oxygen partial pressure at the melt surface and effective
oxygen partial pressure in the bulk atmosphere [61]. This has been clarified
through considering not only chemical equilibria for oxidation of melts but
also oxygen transport under stationary conditions. The effective oxygen pres-
sure, that is, oxygen partial pressure in the bulk atmosphere, is specific for
each metal/metal oxide system and is several orders of magnitude larger than
equilibrium oxygen partial pressure determined by chemical thermodynamics.
Ratto et al. derived the equation, which shows the relationship between the
actual oxygen partial pressure at the melt surface P o2 surface and that at in-
let of the systems P o2 inlet for molten silicon under gas flow conditions [62].
Actual oxygen partial pressure at the melt surface was also described as a
function of oxygen partial pressure at the outlet of the system. Equation (3.9)
is useful in practically controlling the oxygen partial pressure:
2
SiO,SiO
1 + Pe αj
POinlet = POsurface saturation
+ PSi αj Kj POsurface . (3.9)
2 2
1 + P e/|Ψj | 2
j=reaction
Here, two chemical equilibria are considered as follows:

Si + αj O2 = SiO2αj α1 = 1/2, α2 = 1,
for j = 1, Si(l) + 1/2O2 (g) → SiO(g),
and for j = 2, Si(l) + O2 (g) → SiO2 (s).
Kj is the dimensionless equilibrium constant for the jth reaction, Ψj is the
dimensionless diffusion constant of oxygen, and P e, Peclet number, is defined
as the ratio of flow to diffusion of oxygen.
Ratto’s model was experimentally validated by Azami and Hibiya [63].
There was a magnitude difference of over tenfold in oxygen partial pressure
between Huan’s measurement [55] and Mukai’s measurement [56]; Huan et al.
reported the higher oxygen partial pressure, such as 10−6 to 10−5 MPa, and
Mukai the lower, such as 10−24 to 10−14 MPa for similar surface tension val-
ues. Azami also reported higher oxygen partial pressure for visualization of
the Marangoni flow, that is, 10−8 to 10−4 MPa. Differences in the oxygen par-
tial pressure reported by various researchers were universally understood by
employing Ratto’s interpretation, because a lower P o2 was obtained at the
outlet than that at the inlet.
Fig. 3.13. Dependence of temperature coefficient of iron on oxygen concentration

[20]
3.7 Surface Tension of Molten Iron and Iron-based Alloy

For molten iron, the temperature coefficient for surface tension was observed
to change from negative to positive along with increases in oxygen concen-
tration as shown in Fig. 3.13 [20]. Ozawa et al. measured the surface tension
of molten SUS304 stainless steel (8%Ni-18%Cr–Fe) by the oscillating droplet
method using electromagnetic levitation in a wide temperature range includ-
ing the undercooled condition of 170 K [64]. Measurement was carried out
on two types of ambient atmosphere, that is, low P o2 of 10−28 MPa mea-
sured at the exit of the gas flow system and high P o2 of 10−21 MPa. The
low P o2 atmosphere was prepared by a flow of 5% H2 –Ar mixture, whereas
the 5% H2 –He mixture was used for lowering temperature. Pure Ar or He gas
flow was employed to prepare high P o2 condition; residual O2 concentration
was <0.1 ppm for both Ar and He. As shown in Fig. 3.14, a negative tempera-
ture coefficient was observed in the low P o2 atmosphere, whereas in the high
P o2 case a positive temperature coefficient was observed at a lower tempera-
ture (1,620–2,070 K), and the temperature coefficient changed into a negative
value above 2,070 K. This would be attributed to desorption of oxygen from
the melt surface. It is noteworthy that temperature dependence showed almost
the same behavior above 2,070 K.
McNallan and Debroy theoretically calculated surface tension for molten
iron and 18% Cr–Fe stainless steel as a function of both temperature and sul-
fur concentration [5], as shown in Fig. 3.15. As oxygen does, sulfur significantly
affects surface tension and temperature coefficients. It should be noted that,
for Fe and Fe-based alloys, surface tension decreased with increases in oxy-
gen and/or sulfur concentration, and that the temperature coefficient changed
2400
Surface tension: σ / 10−3 N.m−1

2200
Ar-5%H2 / He-5%H2(Po2:10−9−10−15Pa)
2000 s = −0.3063(T−1694)+1676
1800
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]
from negative to positive. However, as mentioned in the previous section, sil-

icon does not display a positive temperature coefficient. This difference is
attributed to that of the solubility of oxygen between molten silicon and iron.
For molten silicon, the solubility limit is 0.001 mass%, whereas for iron, solu-
bility is in the order of 0.1 mass%. The solubility of oxygen in molten silicon
is two orders of magnitude smaller than that in molten iron. This is why
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.
3.8 Thermodynamic Approach for Adsorption of Oxygen

at Melt Surface
As discussed in the previous sections, adsorption of oxygen at the melt surface
plays a significant role in high temperature processes such as silicon crystal
growth and welding, because adsorption not only decreases surface tension
but also changes the temperature coefficient of surface tension. However, sur-
face structure and mechanism of adsorption have not yet been clarified suf-
ficiently. This is attributed to difficulties in controlling PO2 during surface
tension measurement. Molten silver is drawing interest as a model liquid ma-
terial for studying the Marangoni flow of low-Pr-number fluids, such as molten
metals, alloys, and semiconductors [65], because silver cannot be oxidized at
elevated temperatures, yet can sustain bare surfaces. Also molten silver shows
inversion of a temperature coefficient of surface tension. Bernard and Lupis
measured surface tension of silver as a function of oxygen partial pressure
and tried to understand the surface structure adsorbed with oxygen [66]. By
applying the monolayer approximation, it has been clarified that the surface
of molten silver is saturated with oxygen when each adsorbed oxygen atom
occupies an area of 3.3 ± 0.5 × 10−1 nm2 . This corresponds to the formation of
the two-dimensional compound of Ag3 O; the ratio of the amount of metal to
that of oxygen is in the same order as that of the silicon case, that is, SiO0.1 .
Sangiorgi et al. also measured surface tension of molten silver as a function
of oxygen partial pressure and found a surface area per oxygen atom to be
8.8 × 10−2 nm [67]. They suggested the presence of a “corrugated” surface
structure of Ag–O with the oxygen ions protruded towards the gas phase.
Yoshikawa et al. introduced the “associate compound” model to understand
the adsorption of oxygen at the melt surface of Fe–Cr, Ag–Au, and Ag–Cu
melts [68]. An observed decrease in surface tension was systematically ana-
lyzed using the Belton equation. Because the surface tension of oxide melt is
smaller than that of metals, it is reasonable to understand that oxygen ad-
sorption decreases surface tension by the formation of associate oxides at the
melt surface.
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,
etc. Numerical modeling with high-performance computers is used to improve

the quality of products and to shorten the turn-around-time for production.
Marangoni flow plays a key role in heat and mass transport for crystal growth
and welding; the driving force of the Marangoni flow is surface tension differ-
ences at the melt surface. Since thermophysical properties of molten materi-
als are required to be measured with small uncertainty for use in large scale
numerical calculation, there is significant demand for improvement or new
developments in measurement technology. An oscillating drop method using
a levitation technique is an elegant means of measurement; particularly elec-
tromagnetic levitation is suitable for molten metals, alloys and semiconduc-
tors. Oxygen partial pressure can be controlled for electromagnetic levitation.
Nowadays, surface tension must be measured under controlled oxygen partial
pressure, because surface tension of molten metals, alloys, and semiconduc-
tors are sensitive to the oxygen partial pressure of an ambient atmosphere.
Electromagnetic levitation techniques have been improved upon recently to a
great extent, because this technique has been used in microgravity conditions
on board the space shuttle and will be used in the International Space Sta-
tion. Using modern measurement technologies, highly reliable thermophysical
properties including surface tension will be prepared. This technique supplies
not only basic data for engineering science but also for the new field of physics
and chemistry of high temperature melt surfaces, which is indeed a cutting
edge science.
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)
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)
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
a strong magnetic field is effective. Recently, the results of diffusion experi-

ments in a strong magnetic field have been reported and showed agreement
with microgravity experiments [6–8].
Another possibility of diffusion experiments on the ground is a method
using a stable density layering. The diffusion axis is set parallel to the gravity
vector and the samples are set such that the density increases monotonously
downwards during the whole diffusion time. In such a case, even if forces
are generated by temperature gradients and try to start convection rolls, the
force generated by the density gradient can cancel them mostly. The results
with a stable density layering on the ground [9–14] showed that the obtained
concentration curves in four parallel experiments are lying over one another
and each of them can be fitted well with the theoretical equation, and the
calculated diffusion coefficients agree well with that obtained in former μg-
experiments. The temperature dependence of the diffusion coefficients in 1g-
experiments follows the power law with n 2. Although the principle of the
arrangement for a stable density layering seems very simple, it is not until
the completion of the shear cell technique that the stable density layering was
realized, as explained in Sect. 4.2.1.
In June 2005, the research group including the author performed diffusion
experiments in the Russian recovery satellite Foton-M2 successfully using the
shear cell technique. At the same time, the group developed the diffusion
experiment technique on the ground, expanded the potential of the diffusion
research, and changed the concept of microgravity experiments. This chapter
introduces the followings, with showing the results of Foton-M2 mission as
examples:
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
4.2 Design of Shear Cell
4.2.1 Principle of Shear Cell Technique
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
heat up + homogenization diffusion cooling

Tdiff
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.
4.2.2 Minimization of Shear Convection
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.
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 .
4.2.3 Minimization of Free Surfaces
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
4.2.4 Structure of the Shear Cell
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).
4.3 Experimental Procedure
4.3.1 Diffusion Experiments
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
(a) (b) (c)
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)
axial temperature gradient ΔT should be kept small. In case of the Foton-M2

v0 = 0.5 mm min−1 , ΔT = 100 K m−1 . At the end of the diffusion time, the
final shear is started at a constant speed, and then the heater is switched off.
After cooling down each sample piece is pressed out of the cell mechanically,
weighed, and dissolved in acid solutions for the analysis byAAS, ICP-MS. The
concentration of the components in each cell is plotted with an error of ±2%
at the center position of each cell and then the concentration profiles can be
obtained.
4.3.2 Evaluation of Mean Square Diffusion Depth
Here the cases of the diffusion of an impurity component in an alloy into a

pure material are described.
The concentration profile for the interdiffusion between two semi-infinite
sources can be described:

c0 x − x0
c(x, t) = efrc √ . (4.1)
2 2 Dt
The concentration profile for the thick layer diffusion is fitted with the thick
layer solution:

c0 h+x h−x
c(x, t) = erf √ + erf √ . (4.2)
2 2 Dt 2 Dt
68 S. Suzuki
Here D is the diffusion coefficient, t is the diffusion time, c0 is the initial

concentration of the impurity in the alloy side, x is the distance from the
edge of the diffusion sample in (4.1) (Fig. 4.2a) and in (4.2) (Fig. 4.2c) and
the distance from the center of the intermediate cell in (4.2) (Fig. 4.2b) and
x0 is the position of the interface. The initial thickness of the planar source is
either 2h in case of Fig. 4.2b or h in case of Fig. 4.2c. Because x2 meas includes
contributions from the shear convection and the averaging effect, as mentioned
later, (4.3) has to be dealt carefully. The details of the calculation of D from
x2 meas is described in Sect. 4.5.
The fitting parameters were h, x0 , and the product Dt, which can be
described by a measured mean square diffusion depth:
x2 meas = 2Dt. (4.3)
By fitting the following functions to the concentration profiles obtained by,
for example, AAS, the measured mean square diffusion depths x2 meas can be
determined.
4.4 Quantitative Measurement of Shear Convection

and Correction Method
If a new type of shear cell is developed, it is recommended to investigate the
shear convection quantitatively before measurements of diffusion coefficients.
It has been reported that the mass transport caused by the shear convec-
tion consists of three parts: convection during the initial shearing, relaxing
convection just after the initial shearing, and during the final shearing [23].
The relaxation time is reported to be a few seconds, and the total time
of the shear convection is about 10 s, which is independent from the diffusion
time. Therefore, several methods to investigate the shear convection were pro-
posed. Here two kinds of experiments under conditions, where the buoyancy
convection and the Marangoni convection are suppressed, are introduced.
One method is short-time diffusion experiments, where the contribution of
the shear convection to the mass transport is larger than that of the diffusion
process. The other is measurements of the time dependence of x2 meas , where
the contribution from the diffusion process is proportional to the diffusion
time and the additional mass transport is independent from the diffusion
time. Thus extrapolation to the initial time (t = 0) yields the contribution of
the additional mass transport.
4.4.1 Short-Time Diffusion Experiments
For short-time diffusion experiments in low gravity, parabolic flight in an

airplane is available, when the shear convection lasts shorter than 20 s.
Using the Foton shear cell, short-time diffusion experiments were per-
formed on ESA’s parabolic flight campaign (PFC31) by Airbus A300 [23].
Two parallel experiments were performed simultaneously with two capillar-

ies in the shear cell. One capillary was prepared for interdiffusion between
Sn47 In53 alloy (sample A) and Sn (sample B), the diffusion type shown in
Fig. 4.2a. The other capillary was prepared for diffusion of Sn47 In53 (sample
A) from a thick layer into Sn (sample B), the diffusion type shown in Fig. 4.2b.
Each shear process took 2.5 s and the completely overlapped position of the
capillaries and the intermediate cell remained for 10 s. The diffusion time tdiff
was defined as the time from the half-overlapped position at the initial shear
to the half-overlapped position at the final shear, which was 12.5 s. A low
gravity parabola lasted about 20 s with a residual gravity of ±0.05g. Within
one of the low gravity parabolas, all experiments were performed from the
start of the initial shearing to the end of the final shearing. The experiments
were performed at 250◦ C. The concentration profiles are shown in Fig. 4.3.
Similar short-time diffusion experiments can be performed also under 1g
conditions with a stable density layering. The diffusion type was thick layer
diffusion from Sn90 In10 (sample A) into pure Sn (sample B), the diffusion type
shown in Fig. 4.2c. The density increases from Sn90 In10 to Sn monotonically
(a)
c0
t=0
after diffusion
x0
(b)
c0 t=0
after diffusion
2h
Fig. 4.3. Obtained In concentration profiles after short-time diffusion experiments

under low-g at 250◦ C for 12.5 s: (a) interdiffusion between In53 Sn47 -Sn, (b) thick
layer diffusion In53 Sn47 into Sn
70 S. Suzuki
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.
4.4.2 Time Dependence of Mean Square Diffusion Depth
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
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).
4.4.3 Influence of Shear Convection
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.
4.5 Correction Method for the Determination

of Diffusion Coefficients
By using these results, a correction method of these influences was sug-
gested for the determination of D-values in long time diffusion experiments
using (4.4):
72 S. Suzuki
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 1g-Diffusion Measurements with Stable Density

Layering
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.
Table 4.1. 1g-diffusion experiments performed during the Foton projects and ob-
tained mean square diffusion depths and diffusion coefficients
Samplea Typeb Gravity Tdiff (◦ C) Dc ± ΔDd ΔD/D References

vector (tdiff [s]) ×109 [m2 · s−1 ] [%]
(A)PbAg5 (b) +x 350 (28,800) 2.05 ± 0.05 2.3 [13]

(B)Pb 500 (21,600) 3.63 ± 0.09 2.3
670 (12,600) 7.69 ± 0.16 2.0
(a) +x 450 (5,104) 2.65 ± 0.06 2.3
(A)PbGa5 (b) +x 350 (12,600) 2.27 ± 0.08 3.6 [14]
(B)Pb 500 (21,600) 3.57 ± 0.19 5.2
630 (12600) 5.45 ± 0.11 2.0
(A)SnBi1.5 (a) −x 300 (28,800) 2.35 ± 0.06 2.5 [10, 11]
(B)Sn Reference 2.4 ± 0.14 5.8 [28]
2.38 ± 0.14 5.8 [6]
(A)SnBi3 (a) −x 300 (28,800) 2.41 ± 0.05 2.1 [11, 12]
(B)Sn 500 (16,200) 4.35 ± 0.10 2.3
800 (9,000) 7.62 ± 0.20 2.7
(b) −x 300 (12,600) 2.31 ± 0.08 3.5 [12]
(A)SnIn10 (a) +x 275 (28,800) 2.44 ± 0.07 2.8 [9, 11]
(B)Sn Reference 2.74 ± 0.27 9.9 [28]
2.6 ± 0.27 10.3 [6]
400 (21,600) 3.73 ± 0.11 3.0 [9, 11]
600 (21,600) 5.82 ± 0.14 2.4
800 (14,400) 8.62 ± 0.62 7.2
(A)SnIn3 Bi5 (b) −x 500 (21,600) (Bi)4.26 −− [29]
(B)Sn (In)4.99
(A)SnSb3 (a) +x 630 (12,600) 6.80 ± 0.20 2.9 [29]
(B)Sn
(A)AlNi (a) +x 730 (18,800) 3.81 ± 0.12 3.1 [10, 11]
(B)Al Reference 3.9 ± 0.29 −− [28]
a
(A), and (B) are the sample parts shown in Fig. 4.2
b
(a), (b), and (c) are the sample types shown in Fig. 4.2
c
Average value of D obtained from the four capillaries A–D
d
Full statistical errors from four capillaries including temperature error
4.6.2 Data Analysis
It is generally known from the previous results that if convection exists in

the liquid sample, the measured concentration curve has wave-like deviations
from the fitting function, the reproducibility of the diffusion measurements is
low and the diffusion coefficient becomes larger than that in the case without
74 S. Suzuki
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]
convection. To check the influence of the buoyancy-convection on the diffusion

measurements, the obtained data were investigated from these three points of
view [11].
(1) Comparison with fitting function
As a typical example, Fig. 4.5 shows Bi concentration profiles in four par-
allel experiments with SnBi3 -Sn at 300◦ C. In the other experiments in
Table 4.1, the concentration profiles are similar to that shown in Fig. 4.5.
All data were well described by the fitting function (coefficient of determina-
tion R2 > 0.995). This means that there was no indication of buoyancy-driven
convection.
This profile can be also approximately described by a thin layer profile
using

c0 h −x2
c(x, t) = exp , (4.9)
πD(t + t0 ) 4D(t + t0 )
where t0 = h2 /6D is the time offset at t = 0, which gives the same mean
square diffusion depth as the real start profile. Because of the linearity of the
relationship between x2 and ln c, (4.9) can be used to show the absence of
nondiffusion-like convection (Fig. 4.6).
(2) Reproducibility of diffusion experiments
As shown in Figs. 4.5 and 4.6, the four concentration curves gave about the
same diffusion coefficient. The reproducibility can be investigated by the
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.
4.6.3 Effect of Density Layering
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
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.
4.7 Microgravity Experiments

4.7.1 Utilization of Microgravity Environment
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.
4.7.2 Microgravity Diffusion Experiments in Foton-M2
In the Foton-M2 mission, μg-diffusion experiments were performed in six shear

cell units with a total of 24 capillaries (four capillaries A–D in each shear cell)
in the AGAT-facility. The mission procedure of the Foton-M2 is shown in
Fig. 4.7. The Soyuz-U rocket was launched on 30 May 2005 from Baikonur
Cosmodrome in Kazakhstan, and the Foton capsule was put into the low-
earth orbit. During 2 weeks of the flight, the experiments were performed in
one shear cell after another cell. During the experiments it was possible to con-
trol the AGAT via the payload control center ESRANGE located in Kiruna,
Sweden, in the Arctic Circle. The downlink was aquired from the satellite,
which contained the temperature in the shear cell, torque of the motor during
both the initial and the final shear, and the vacuum pressure. After the data
analysis it was confirmed that all of the experiments had performed normally.
On 16 June 2005 the capsule landed in Kazakhstan. The sample and the
78 S. Suzuki
μg-experiments in orbit (June 2005)
Soyuz-U was launched

from Baikonur Recovery to earth
Cosmodrome (May 2005) Kasakhstan (June 2005)
Fig. 4.7. Mission procedure of Foton-M2. The pictures were provided by ESA
Fig. 4.8. Sb concentration profiles in four parallel 1g-experiments and in a μg-

experiment with SnSb5 –Sn (at 630◦ C for 12,600 s) with the error of ΔC/C = 2%
dependent on the performance of the AAS equipment
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.
4.8 Temperature Dependence of the Diffusion

Coefficients
The temperature dependence of the diffusion coefficients in liquid metals has
been discussed for a long time. One of the descriptions is the power law.
The results of molecular dynamics simulations [34] showed a temperature de-
pendence of D = AT n , with n = 1.7–2.3. Several results of μg-experiments
(self-diffusion of Sn (FSLP [1], MSL-1 [4, 5]), In, Pb, and Sb (D2 [3]), inter-
diffusion of Sn-In (D1 [2])) showed D = AT 2 , which is considered as a special
case of the power law.
The function D = AT n was fitted to the measurement results of the ex-
periment of In in Sn, Bi in Sn (Fig. 4.9), Ag in Pb (Fig. 4.10), Ga, and Pb207
(Fig. 4.11) using weighted least squares according to ΔD. In every case, the
diffusion coefficient Dx109 [m2/s]
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]
Fig. 4.9. Temperature dependence of diffusion coefficients of Bi in Sn and In in Sn
D=1.4x10 −15 T
2.2
Fig. 4.10. Temperature dependence of diffusion coefficients of Ag in Pb. 1g-data

[13], and μg-data [29]
80 S. Suzuki
10
measured diffusion coefficient

Experiments with shear cell[14]:
9 : Ga in Pb (1g)
8 : Fitting curve for Ga in Pb (1g)
: Ga in Pb (μg, Foton−M2)
7
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
temperature dependence is described better with a fitting parameter n be-

tween 1.5 and 2.4 than with a fixed n of 2. This suggests the possibility of
the power law with a variable exponent according to the measured system.
But the number of the fitting points is so small that this discussion does
not exclude the possibility of the T 2 -law. In any case it was confirmed that
the temperature dependence can be well described by the power law. But the
power law has not yet been explained conclusively with a clear physical model.
This is still under work.
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
Table 4.2. Classification of diffusion types and possible conditions
Possibility of Segregation Possible Possible

Diffusion type stabilization by (in general) gravity techniques
density gradient conditions
Self- In elements No (or low) No μg SCa, LCb

diffusion In alloys No Yes μg SC
Impurity In elements Yes (at If “no” μg, 1g SC, LC
diffusion sufficient high If “yes” μg, 1g SC
concentrations)
No (at low No μg SC, LC
concentration)
In alloys If “yes” Yes μg,1g SC
If “no” Yes μg SC
Interdiffusion In binary If “yes” Yes μg, 1g SC
alloys If “no” Yes μg SC
In multi- No (because of Yes μg SC
component drag effects)
alloys (>2
components)
Combination In binary If “yes” Yes μg, 1g SC
of inter and alloys If “no” Yes μg SC
self-diffusion In multi- No (because of Yes μg SC
component drag effects)
alloys (>2
components)
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
For a more detailed discussion on the diffusion mechanism, more measure-

ments are required. The influence of the atomic weight, the atomic radius, and
the electronic structure of the tracer element on the diffusion coefficients is
significant. From these points of view, the next experiments will be planned.
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.
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
26. G. Müller-Vogt, R. Kößler, J. Cryst. Growth 186, 511 (1998)

27. M. Uchida, Y. Watanabe, S. Matsumoto, M. Kaneko, T. Fukazawa, T. Masaki,
T. Itami, J. Non-Crystalline Solids 312–314, 203 (2003)
28. J.P. Garandet, G. Mathiak, V. Botton, P. Lehmann, A. Griesche, Int. J. Ther-
mophys. 25, 249 (2004)
29. S. Suzuki, K-H. Kraatz, G. Frohberg, R. Rosu-Pflumm, W. Wendl,
G. Müller-Vogt, to be published
30. R.W. Smith, Micrograv. Sci. Technol. 11/2, 78 (1998)
31. T. Itami, A. Mizuno, H. Aoki, M. Kaneko, T. Fukuzawa, A. Tanji, Y. Arai,
K. Goto, Y. Yamaura, N. Tateiwa, M. Koyama, T. Morita, H. Kawasaki,
M. Fujishima, T. Masaki, S. Yoda, Y. Nakamura, N. Koshikawa, T. Nakamura,
A. Ogiso, J. Jpn. Soc. Micrograv. Appl. 3, 198 (1999)
32. T. Itami, Jpn. Soc. Micrograv. Appl. 19, 185 (2002)
33. A. Griesche, M-P. Macht , S. Suzuki, K-H. Kraatz, G. Frohberg, Scripta Mater.
57, 477 (2007)
34. M. Shimoji, T. Itami, Atomic Transport in Liquid Metals, p. 250, Trans. Tech.
Aedermannsdorf (1986)
5
Thermal Diffusivity Measurements of Oxide
and Metallic Melts at High Temperature
by the Laser Flash Method
Hiroyuki Shibata, Hiromichi Ohta, and Yoshio Waseda
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.
5.2 A Brief Background of the Present Requirement

for the Thermal Property Measurements of High
Temperature Materials
For unidirectional heat flow in an isotropic medium, the following Fourier–Biot
equation holds with the thermal conductivity λ,
J = −λ∇T, (5.1)
where J is the quantity of heat flowing in unit time through unit area un-
der influence of a temperature gradient ∇T by conduction. For the transient
condition where the temperature changes with time t, (5.1) becomes
∇(λ∇T ) = Cp ρ(dT /dt), (5.2)
where Cp is the specific heat at constant pressure and ρ is the density. In
cases where the thermal properties of λ, Cp , and ρ are treated as constants,
independent of both position and temperature, (5.2) can be written as follows:
∇2 T = (Cp ρ/λ)(dT /dt) (5.3)
or
∇2 T = α−1 (dT /dt), (5.4)
where α = λ/Cp ρ is the quantity called thermal diffusivity, since all diffusivity
processes can be represented by an equation similar to (5.3), and since it has
the same dimensions as diffusion coefficient. Usually, measurements of thermal
diffusivity or conductivity are made under conditions in which the assumption
that thermal properties of the sample are independent of temperature and
position in the sample is well justified. Of course, particular attention should
5 Thermal Diffusivity of Melts 87
Table 5.1. A comparison for steady state measurement with transient state mea-
surement for thermal conductivity, λ, and thermal diffusivity, α [9]
Steady state measurement Transient state

measurement
Obtained value λ λ or α
Mathematical basis Equation (5.1) Equations (5.2) or

Eqs.(5.1) and (5.2)
Boundary conditions Constant in time Varies with time and

position
Temperature Stationary in time Varies with time

distribution at all points and position
Measured quantities J and ∇T T vs. t or J and T vs. t
be given to measurements at temperatures close to any phase transformation,

where apparent heat capacity Cp shows strong temperature dependence.
The thermal diffusivity values can be determined by any of the experimen-
tal methods based on (5.1)–(5.4). The experimental methods for measuring
thermal diffusivity can be classified into two main categories: steady state mea-
surement and transient state measurement, respectively. Comparisons of these
two methods for measuring thermal diffusivity are summarized in Table 5.1,
along the line suggested by Taylor [9].
The experimental uncertainties due to heat leak, such as radiation and con-
vection, can be generally reduced when experimental time is decreased, and
much shorter times are convenient for transient experiments. Much shorter
times can be used in transient-state experiments than when a steady state
distribution has to be established. Steady-state experiments also have a theo-
retical disadvantage in measurements on a fluid mixture, because the existence
of the temperature gradient in the sample for a long period may cause the
separation of the components due to thermal diffusion.
Improved electronic devices that have allowed temperature change to be
accurately determined over short time intervals have led to an increase of
interest in transient state measurements for determining thermal diffusivity
particularly at high temperatures. Owing to these factors, the transient mea-
surement such as the laser flash method has gained a dominant position at
the present time. Some characteristic features of the laser flash method cur-
rently being used for thermal diffusivity measurements of various substances
including melts at high temperatures will be given in Sects. 5.4 and 5.5.
5.3 Experimental Procedures and Theoretical Basis

for the Laser Flash Method
The laser flash apparatus for measuring thermal diffusivity of various sub-
stances at variety of conditions has already been described [5,10]. One typical
example is given in Fig. 5.1 using the schematic diagram of the experimental
arrangement in the vertical line-up. Figure 5.2 shows a photograph of the
facility at the Institute of Multidisciplinary Research for Advanced Materials
(IMRAM), Tohoku University. The front surface of a sample is irradiated by a
Nd glass (or ruby) laser beam with a pulse duration of the order of 1 ms. The
temperature response signals at the back surface of the sample are detected by
using an infrared detector (e.g., InSb: effective wavelength 2–5.5 μm) without
mechanical contact. The sample is usually heated up to the desired temper-
ature, within ±2 K controlled by a thermocouple installed near the sample
cell and holder, using an additional heating element such as a tungsten mesh
heater under vacuum of less than 2 × 10−3 Pa or a platinum wire heater under
the air atmosphere.
The three layered cell assembly for oxide melts is illustrated in Fig. 5.3 [7].
In this case, the liquid sample is sandwiched with two platinum plates. A
platinum crucible containing a small amount of a liquid sample is placed on
the alumina pipe. The platinum crucible size (corresponding to the third layer)
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
Fig. 5.1. A laser flash apparatus at high temperature

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
Pulse laser Pulse laser
Micrometer
Mullite tube
Upper plate Platinum crucible

Δl (upper plate)
Molten l2 Molten sample
sample α Platinum crucible
Lower plate (lower plate)
Infrared detector
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
is 0.2 mm thick, 10 mm deep, and 19 mm in diameter. Another relatively small

platinum crucible of 0.2 mm thickness, 10 mm depth, and 14 mm diameter is
employed as the first layer and it is centered to the end of the alumina tube
by ceramic cement. The top platinum crucible is located just right above the
bottom one to consist of a three-layered cell. The alumina tube is connected
to a holder on a three axial movable arm for adjusting the sample thickness.
A typical flow of a measuring process is as follows.
The inner and outer platinum crucibles were separated several times dur-
ing the course of the experiments to check that there was no bubble in the
sample melt. The alumina pipe was lowered vertically by a micrometer. Then,
a laser pulse was irradiated on the top surface of the inner crucible (the first
layer). The temperature response signals were recorded by means of an in-
frared detector focused on the back surface of the outer crucible (the third
layer) through a gold mirror. Another temperature response signals were also
measured after varying slightly the thickness of a sample melt. The variation
of the sample thickness can be determined by reading a micrometer scale at-
tached on the holder. In this type apparatus, the relative difference of sample
thickness l in two sets of the measurements was fixed to be 0.2 mm, on the
basis of the previous results on high temperature melts at various conditions
using the three-layered laser flash method [7, 8]. After measurement, the up-
per platinum crucible was raised to again check visually that there were no
bubbles in the sample melt. From the results of measurements by applying
the three-layered cell filled with distilled water, and molten carbonate salts,
it was found that the experimental uncertainties were within ±5%.
As easy and reliable data processing method has also been developed to
estimate the thermal diffusivity value of a sample melt at high temperature
using the three layered laser flash method [6]. This data processing method
is based on a point that the initial time region of the temperature response
curves is unlikely to be affected by heat loss by conduction from the sample
to the holder and by radiation. The essential points are summarized later for
convenience of discussion.
The temperature response of the three-layered cell at the initial time re-
gion of the temperature response curve can be given in the following form,
analogous to the approach for a two-layered cell proposed by James [11]:
√
∂ ln(T t) (η1 + η2 + η3 )2
=− , (5.5)
∂(1/t) 4
√
where ηi is li / αi , T is temperature rise, and t is time after irradiating a
laser pulse. The thickness and thermal diffusivity value √ of the ith layer are
denoted by li and αi , respectively. Thus, a plot of ln(T t) against 1/t for (5.5)
is considered to show a straight line with slope of β = −(η1 + η2 + η3 )2 /4.
The thermal diffusivity value of α2 of the second layer (a sample melt in the
present case) can readily be estimated, when the thermal diffusivity values of
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)
where, l2 is the thickness of a sample melt in the first measurement, l

is the relative change of sample thickness corresponding to the difference
between two measurements obtained by lifting √ up the inner crucible (the
first layer in the present case), β(l2 ) = ∂ ln(T(l2 ) t)/∂(1/t) and β(l2 +l) =
√
∂ ln(T(l2 +l) t/∂(1/t) are the slopes of temperature response Tl2 for the first
measurement and Tl2 +l for the second measurement.
The relation between the sample thickness l2 and the thermal diffusivity
value of α2 for sample is expressed as a solid line in Fig. 5.4 [7]. For a given
l, the similar relation between l2 and α2 is also described as a dashed line
in Fig. 5.4. The intersection of these two lines provides the resultant thermal
diffusivity value of α2 and its thickness l2 . The temperature response curve
is estimated at one thickness l2 as shown in Fig. 5.5a. The same procedure
is applied for a thickness l2 + l. The two response curves are converted as
provided in Fig. 5.5b and the respective gradients β(l2 ) and β(l2 +l) are esti-
mated by the least squares method. The values of η1 , η3 and l are given, the
values of α2 and l2 to satisfy (5.6) and (5.7) are calculated as the intersection
(see Fig. 5.4).
Sample thickness: l 2
Sample thickness: l2 + Δl
a2
0
0 l2
Fig. 5.4. Relationship of α2 as a function of thickness estimated from (5.6) and

(5.7)
(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
In practice, to derive more accurate values from temperature response

curves, numerical iterative procedure combined with a solver of heat transfer
equations is employed using the derived values by this method as an initial
trial values, because (5.5) is exact only for short duration after laser pulse
flashed. Detailed information for the data processing with high accuracy has
been given in [7].
It may also be worth mentioning from recent simulated results by apply-
ing the finite element method that the heat leakage through a side wall of
a platinum crucible is able to reduce less than 5%, when only the values of
sample thickness and its variation are carefully selected [12]. In addition, from
the results of systematic measurements with samples having different diame-
ters larger than that of the laser beam, it was proved that the shape of the
three-layered cell presently used reduces heat leakage to the sample container
and the radial heat flow was well-confirmed to be insignificant [6–8].
It is worthy of note that another three-layered cell has been developed for
keeping the shape of metallic melt uniform for a given thickness. A schematic
diagram of the devised cell is illustrated in Fig. 5.6 [13]. This cell system
Pulse laser
Hole
Carbon fixture Sapphire plate

Molten metals
Mullite or Infrared
alumina tube detector
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].
5.4 Selected Examples of Thermal Diffusivities of Oxide

Melts
Boron oxide is widely used as a liquid capsule material to encase molten
semiconductors for reducing a trouble due to diffusion of high vapor pressure
components. The laser flash method with a three-layered cell has been ap-
plied to measure the thermal diffusivity data of boron oxide melt at intervals
50–100 K over the temperature range of 1,000–1,500 K [8]. With different wa-
ter contents, Figure 5.7 shows the results for boron oxide melts containing
0.08 mol% In2 O3 and molten boron oxide containing a small amount of P2 O5 ,
respectively.
In all cases, the thermal diffusivity values of boron oxide melt increase as
the temperature rises. Changes in thermal diffusivity, which are not promi-
nent, systematically decrease as the contents of In2 O3 and P2 O5 increase. It
can be qualitatively understood that the results are attributed to an increase
of the anharmonicity of boron oxide melt resulting from the addition of other
oxide, which causes the mean free path of phonons to decrease as compared
with the values for pure boron oxide melts [14]. Of course, some further studies
are required to obtain the definite answer about these experimental results. It
may also be noted that unless a forced dehydrating treatment such as bubbling
(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]
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
8
(a) (a)
6
4
TiO2 4.9% T.Fe 0.4%
a / 10-7m2s-1
2 TiO2 0.7% TiO2 7.4% T.Fe 1.2%

TiO2 2.6% TiO2 9.6% T.Fe 2.6%
0
(b) (b)
6
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]
such scattering is considered beyond the experimental uncertainty. This is at-

tributed to the effect of radiative heat transfer component involved in a molten
sample. Although the initial time region of the temperature response curve is
known not to be severely affected by the radiative heat transfer [11], the ra-
diative component should be separated from measured temperature response
signals. This is particularly true for the higher temperature measurements.
The absorption coefficient of samples of interest is needed for quantitative
discussion of the radiative heat transfer in high temperature substances. It
should be mentioned that the absorption coefficient data are essentially re-
quired for the semitransparent media, in comparison with the transparent and
opaque cases. The absorption coefficients of three CC powder samples contain-
ing titanium oxide and iron oxide are given in Fig. 5.10 as a function of wave
length together with that of the hemispherical emissive power of blackbody
at 1,573 K [21]. It is also very interesting that the absorption coefficients of
CC powder samples in the range between 1 × 10−6 and 4 × 10−6 m−1 is found
not to be significantly affected by the change in CaO/SiO2 ratio or the addi-
tion of zirconium oxide and hafnium oxide to the reference composition [21].
These results suggest that the correction by the transparent body approxima-
tion is sufficiently accepted for these CC powder samples with respect to the
radiative heat transfer behavior. In other words, the contribution from the
radiative component should be explicitly considered only for cases containing
titanium oxide and iron oxide.
In samples for which the transparent body approximation is valid, a liquid
layer appears to be transparent to radiative heat transfer where no radiation
3000
Hemispherical emissive power, 1010 Wm-3

1573K
Reference
T.Fe 1.2% 10
2000
TiO2 9.6%
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
Given value of thermal disffusivity for calculation

4
1350 1400 1450 1500 1550 1600
T/K
Fig. 5.11. Apparent thermal diffusivity indicating radiative heat flow theoretically
calculated under the transparent body approximation for a sample of thermal dif-
fusivity of 4 × 10−7 m2 s−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
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)
c = 2.549 × 10−7 T 2 − 4.504 × 10−4 T + 1.244 for l2 = 0.1 mm,

c = 6.986 × 10−7 T 2 − 1.466 × 10−4 T + 1.883 for l2 = 0.2 mm, (5.9)
c = 1.253 × 10−6 T 2 − 2.675 × 10−3 T + 2.606 for l2 = 0.3 mm.
On the other hand, both the absorption and emission of the radiative com-
ponent in a liquid layer should be included when the transparent body ap-
proximation is not well accepted as for CC powder samples. With respect to
this subject, Darby [23] has reported fundamental equations by considering
all modes of radiative heat transfer. Ohta et al. [22] computed the results for
the gray body approximation where the spectral absorption coefficient of the
semitransparent media is reduced to the mean absorption coefficient κm on the
basis of Darby equations using the control volume method [24]. In this com-
putation, the temperature response signals are estimated as a function of κm
from which the thermal diffusivity values can be obtained. Figure 5.12 shows
the results for a case of thermal diffusivity of 4×10−7 m2 s−1 [22]. The validity
and usefulness of these results may be confirmed by suggesting the following
points. Good agreement is found with the value proposed by Rosseland [25],
who suggests for the condition of κl > 1, in the higher absorption region
l = 0.3mm
573K
Rosseland
a/10−7m2s−1
approximation
4.5
Transparent
approximation κ l >1
4 Given value of thermal diffusivity
101 102 103 104 105 106

k /m−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
.
Table 5.2. Mean absorption coefficient κm for samples containing TiO2 and iron
oxide estimated from measured optical properties [21]
Temperature(K) 1,373 1,473 1,573

Type of Powder κm
Fe 0.40% 546 572 596
Fe 1.20% 774 810 844
TiO2 2.60% 21 22 24
TiO2 4.90% 39 41 43
TiO2 9.60% 81 87 94
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.
(1) The contribution from radiative component could be estimated by the

transparent body approximation with the experimental uncertainty less
than 4% for samples recognized as transparent.
(2) The contribution from radiative component should be estimated, at least,
by the gray body approximation, with the mean absorption coefficient for
samples considered semi-transparent if we want to hold the experimental
uncertainty down to less than 4%.
Table 5.3. Apparent thermal diffusivity estimated from theoretical temperature

response curves at 1,575 K by the three-layered laser flash method
Type of Powder Approximation

Transparent Gray Band
Reference 4.90 (0.03) 5.05
Fe 0.40% 4.90 (0.09) 5.29 (0.02) 5.39
Fe 1.20% 4.90 (0.11) 5.44 (0.01) 5.5
TiO2 2.60% 4.90 (0.04) 4.91 (0.04) 5.1
TiO2 4.90% 4.90 (0.04) 4.93 (0.03) 5.11
TiO2 9.60% 4.90 (0.05) 4.97 (0.03) 5.13
Given, thermal diffusivity, 4 × 10−7 m2 s−1 ; sample thickness and its variation, 0.2
and 0.2 mm. The numerical values in the parenthesis correspond to the ratio of
deviation from the case estimated under the band approximation [20]
The thermal diffusivity values at high temperature were again estimated

for CC powder samples for steel by applying the transparent body approxima-
tion or the gray body approximation, depending upon their optical properties
characterized by the absorption coefficient behavior. The resultant thermal
diffusivity values are given in Fig. 5.9b. It is clearly found from a comparison
of the results of Fig. 5.9a that scattering detected in measured values is con-
siderably reduced by correcting the contribution from radiative component,
depending on both temperature and sample thickness. In addition, the ther-
mal diffusivity values of continuous casting powders for steel are of the order
to 4 ± 0.5 × 10−7 m2 s−1 and almost insignificant with the variation of both
temperature and composition presently investigated.
5.5 Selected Examples of Thermal Diffusivities

of Metallic Melts
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
(3) All thermophysical properties of three layers are given

(4) The thermal contact resistance at the interface between layers is uniform
and has same value at both the upper and the lower interfaces
(5) A heat pulse is uniformly absorbed at the front surface
(6) Radiative heat loss is proportional to the temperature difference, Tm −
Te , where Tm is temperature rise in metallic melt and Te is steady-state
temperature, respectively
(7) No absorption of the energy in medium of the cell because the sapphire
presently used is transparent to both the laser pulse and infrared
(8) Radiative heat loss occurs only at the surface of the metallic melt
The most important and significant points are given below for further work
on the thermal diffusivity measurements of metallic melts at high temperature.
Figure 5.13 shows a typical temperature response curve where the dimen-
sionless parameter of thermal contact resistance at interface, R+ = λm R/lm =
102 , and Biot number, Y = 4εσTe3 lm /λm = 0.1, and the normalized temper-
ature, T ∗ = T /Tmax . This result is demonstrated using the molten nickel
case as an example. Here λm is thermal conductivity of metallic melts, R is
the thermal contact resistance, lm is the sample thickness of metallic melts,
ε is the thermal emissivity of the surface of metallic melts, and σ is the
Stefan–Boltzmann constant, respectively. It is interesting to note that the
decay part of the temperature response curve normalized by the maximum
temperature rise, Tmax , is approximately given in the normalized time region
between 8 and 12.

T ∗ = TM exp −kt∗ , (5.10)
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]
where TM corresponding to the temperature when extrapolated the attenua-

tion curve down to time at laser beam irradiation. As shown in the results of
Fig. 5.13, the linear relationship between the logarithmic value of T ∗ and t∗
may be described in the following form:
ln T ∗ = −kt∗ + ln TM . (5.11)
The gradient k and the intercept ln TM provide information with respect to
the effects of radiative and conductive heat loss at interface between melt
sample and cell material. Similar calculations were made for various given
conditions and the simulated results of the relationship between k and ln TM
with conduction for the thermal resistance parameter R+ = 10n , n = 0.5−∞,
and Biot number Y = 0 − 0.25 are illustrated in Fig. 5.14. This figure includes
the experimental results of molten nickel obtained with different conditions
and a point at the origin in Fig. 5.14 corresponds to the adiabatic condi-
tion. It has been intention of the present authors to suggest that most of the
experimental results of k and ln TM for molten nickel are found to be situ-
ated near the curve of R+ = ∞. Small dispersion is likely to be attributed
to the fluctuations in the signal-to-noise ratio detected in the high tempera-
ture measurements. Accordingly, the present authors maintain the view, from
the results of Fig. 5.14 that the effect of the conductive heat loss at inter-
face between metallic melt and cell material is negligibly small within the
present experimental condition. In other words, thermal contact resistance at
the interface between molten sample and cell material can be considered to
Molten nickel : Run1

: Run2
0.20 : Run3
: Run4
: Calculated
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]
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]
be sufficiently large. Consequently, only the radiative heat leakage needs to

be taken into account for the present experimental condition with the devised
three-layered cell. Sufficiently reasonable reproducibility of the experimental
thermal diffusivity data was confirmed by repeating the measurements under
a given condition more than three times [13].
The thermal diffusivity values of molten iron, cobalt, and nickel were mea-
sured for iron in the temperature range from 1,808 to 1,868 K, for cobalt in
the temperature range from 1,768 to 1,838 K, and for nickel in the tempera-
ture range from 1,728 to 1,928 K [13]. The results are summarized in Fig. 5.15.
The thermal diffusivity value of molten iron at its melting point (1,808 K) is
12% smaller than the reported value of solid iron at 1,600 K [4]. The liquid
value for cobalt with its melting point (1,768 K) is 21% smaller than that of
solid cobalt at 1,700 K and for nickel case (with melting point 1,728 K) the
liquid value is 36% smaller than that of solid nickel at 1,500 K [4]. Such dis-
tinct variation in thermal diffusivity of three metals should be attributed to
a change in the structure from the long-range periodic atomic distribution
to that with nonperiodicity on melting. It is also worth mentioning that the
thermal diffusivity values of three metallic melts indicate slightly positive tem-
perature dependence, although the origin of this behavior cannot be certainly
identified at the present time.
Thermal diffusivity values of molten three metals are summarized in the
following equations (m2 s−1 ) [13].
αFe = 4.51 × 10−9 (T − 1808) + 5.97 × 10−6 (1808 ≤ T /K ≤ 1868),

αCo = 6.59 × 10−9 (T − 1768) + 6.14 × 10−6 (1768 ≤ T /K ≤ 1838),
αNi = 6.61 × 10−9 (T − 1728) + 1.02 × 10−5 (1728 ≤ T /K ≤ 1928),
(5.12)
Thermal diffusivity values of molten germanium and silicon have also been
obtained by the laser flash method with a three-layered cell, although silica
glass, known to be transparent to Nd glass laser and infrared ray, was used in
the measurement for silicon [26].
αGe = 1.40 × 10−8 (T − 1218) + 2.28 × 10−5 (1218 ≤ T /K ≤ 1398),
αSi = 4.48 × 10−9 (T − 1685) + 2.23 × 10−5 (1685 ≤ T /K ≤ 1705).
(5.13)
A large number of attempts to produce metallic glasses have been reported
by direct melt processing techniques such as chill block casting of ribbon and
free jet spinning of wire. The quenching rate is of the order of 106 K s−1 . The
glassy samples were usually produced in the form of ribbon with thickness of
30 μm or wire with diameter of 100 μm. Further, the addition of third elements
that exhibit different properties, such as atomic size and crystalline symmetry
to a binary alloy, is known to enhance drastically the ease of glass formation.
Then some exceptional systems such as Pd–Cu–Si, Pd–Ni–P, and Pt–Ni–P
have produced as cylindrical rods of 1–3 mm diameter in the glassy state at
quenching rate of only 102 K s−1 [27, 28]. Inoue and his colleagues [29, 30]
discovered new Zr-based alloy systems such as Zr–X–Y (X= Al, Y = Co, Ni,
Cu), which show a wide super-cooled liquid region of more than 100 K and do
not include any nonmetallic elements such as P and Si. Not only a cylindrical
rods of several millimeters diameter, but also pipes with inner diameter of
10 mm and wall thickness of 2 mm and ingots with diameter or thickness
larger than several centimeters [30, 31] have been recently fabricated with
bulk metallic glass. These results prompt us to investigate thermal properties
of these alloy melts because of their importance to estimate a temperature
profile in a mold. A critical cooling rate to obtain glassy phase may also be a
function of thermal diffusivity.
Figure 5.16 shows the measured temperature response curve of molten
Zr55 Al10 Ni5 Cu30 alloy at 1,218 K, as an example [32]. Measurements using
the laser flash method with a three-layered cell were also made for molten
Zr60 Al15 Ni25 and Zr65 Al7.5 Cu27.5 alloys for comparison. The thermal diffu-
sivity values of molten Pd-based alloys which are known to show good glass
formability were systematically measured by the laser flash method [33]. All
these results for six alloy melts are summarized as a function of temperature
in the following form (m2 s−1 ).
αZr55 Al10 Ni5 Cu30 = 3.73 × 10−9 (T − 1154) + 4.53 × 10−6 (1154 ≤ T /K ≤ 1228),
αZr60 Al15 Ni25 = 7.89 × 10−9 (T − 1170) + 5.01 × 10−6 (1170 ≤ T /K ≤ 1263),
αZr65 Al7.5 Cu27.5 = 4.75 × 10−9 (T − 1180) + 5.39 × 10−6 (1180 ≤ T /K ≤ 1303),
(5.14)
T/a.u.
Zr55Al10Cu30Ni5
in the liquid state at 1218K
0 200 400 600

t/ms
Fig. 5.16. Measured temperature response curve of molten Zr55 Al10 Ni5 Cu30 alloy
at 1,218 K [32]
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)
Positive temperature dependence of the thermal diffusivity values is found

for six alloys in the liquid state and the best glass forming alloy, such as
Zr55 Al10 Ni5 Cu30 and Pd40 Cu30 Ni10 P20 , indicates the lowest value among the
investigated alloys with respect to both Zr-based and Pd-based cases. The
thermal diffusivity values of these six molten alloys are summarized in Fig. 5.17
as a function of normalized temperature, (T − Tl )/Tl , where Tl is the liquidus
temperature. The thermal diffusivity value can be estimated at a liquidus
temperature by extrapolating the obtained equations to that temperature, so
that such values are plotted against the critical cooling rate of each alloy in
Fig. 5.18 [32].
It is interesting to find that the lower the thermal diffusivity values of
Zr-based alloys at liquidus temperature, the lower the critical cooling rate to
obtain metallic glass phase becomes. Such behavior has also been confirmed in
the Pd-based alloy case. Although any definite comment cannot be certainly
obtained yet, the number of free electrons in molten alloys is likely to be close
relationship between transport property and the critical cooling rate. The
thermal diffusivity values of molten alloys may be one of the good indicators
for predicting the glass forming ability of alloys of interest.
2 -1 4
a /10 m s
-6
3 :Zr 65 Al 7.5 Cu 27.5

:Zr 60 Al 15 Ni 25
2 :Zr 55 Al 10 Ni 5 Cu 30
:Pd 40 Cu 40 P 20
1 :Pd 40 Ni 40 P 20
:Pd 40 Cu 30 Ni 10 P 20
0
0.0 0.1 0.2 0.3 0.4 0.5
(T - Tl) / Tl
Fig. 5.17. The thermal diffusivity values of molten Zr-based and Pd-based alloys
as a function of normalized temperature, (T − Tl )/Tl , where Tl is the liquidus tem-
perature [32]
5
2 -1
a /10 m s
4
-6
3 :Zr 65 Al 7.5 Cu 27.5

:Zr 60 Al 15 Ni 25
2 :Zr 55 Al 10 Ni 5 Cu 30
:Pd 40 Cu 40 P 20
1 :Pd 40 Ni 40 P 20
:Pd 40 Cu 30 Ni 10 P 20
0
1 10 100
-l
R c / Ks
Fig. 5.18. Relationship between the estimated thermal diffusivity values at a liq-
uidus temperature and the critical cooling rate for glass phase formation of molten
Zr-based and Pd-based alloys [32]
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
The authors are deeply indebted to Professors Y. Tomota and S. Kitamura

for their sustained encouragement of our projects on the thermal property
measurements of materials. We are also grateful to Professors A. Inoue and
J. Saida, and Drs. T. Nishi and N. Nishiyama for their collaboration on the
metallic melt projects.
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
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
Masahiro Susa and Rie K Endo
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.
6.2 Definition of Emissivity

The emissivity ε is a value to specify how well a real body radiates energy in
comparison with a blackbody at the same temperature as the real body, and is
defined using radiation intensities emitted from a real body (sample) and from
a blackbody. However, the radiation intensity from a blackbody IB depends
upon wavelength of radiation λ and temperature T and, furthermore, the
112 M. Susa and R.K. Endo
radiation intensity from a real body IR depends upon direction of radiation

as well. Thus, the emissivity should be strictly expressed as a function of
wavelength, direction, and temperature.
The most fundamental emissivity is directional spectral emissivity ε(λ, θ,
ϕ, T ) that includes dependencies on wavelength, direction, and temperature,
and is defined as
ε(λ, θ, ϕ, T ) = IR (λ, θ, ϕ, T )/IB (λ, T ), (6.1)
where θ is the polar angle measured from the normal to the surface of
a sample and ϕ is the azimuth angle. Strictly, the radiation intensity as a
function of wavelength is termed the spectral radiation intensity (or radiance)
and the value of IB (λ, T ) is given by Planck’s law of radiation as follows:
IB (λ, T ) = 2C1 /λ5 [exp(C2 /λT ) − 1] , (6.2)
where C1 and C2 are Planck’s first and second constants, respectively,
5.96 × 10−17 W m2 sr−1 and 1.44 × 10−2 m K. In particular, the spectral emis-
sivity in a direction normal to the surface is termed normal spectral emissivity,
which is often measured experimentally and is very important in temperature
measurements using optical pyrometers (radiation thermometers).
Integrations of (6.1) over all wavelengths and over all directions give di-
rectional total emissivity and hemispherical spectral emissivity, respectively,
and integration of (6.1) over all wavelengths and all directions gives hemi-
spherical total emissivity. The hemispherical total emissivity is indispensable
to the calculation of energy loss from the surface and can also be measured.
However, this direct measurement is rather difficult and the hemispherical
total emissivity is usually estimated from the normal spectral emissivity. As a
consequence, the following review focuses measurement techniques and data
for normal spectral emissivity.
6.3 Measurement Techniques for Emissivities

6.3.1 Method Based on Wien’s Formula
For small values of λT , that is, exp(C2 /λT )

1, (6.2) for blackbody radiation
can be approximated to the following:
IB (λ, T ) = 2C1 /λ5 exp(C2 /λT ). (6.3)
This equation is called Wien’s equation, which is accurate to within 1% for
λT less than 3,000 μm K [2]. For real body (non-blackbody) at the same tem-
perature as blackbody, the normal spectral radiation intensity IR could be
expressed by Wien’s formula using an apparent or brightness temperature
TA as
IR (λ, T ) = 2C1 /λ5 exp(C2 /λTA ). (6.4)
6 Emissivities of High Temperature Metallic Melts 113
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]
Thus, the normal spectral emissivity can be derived through (6.1) as

ε(λ, T ) = IR (λ, T )/IB (λ, T ) = exp [(C2 /λ)(1/T − 1/TA )] (6.5)
The normal spectral emissivity is sometimes determined based on (6.5) from
measurements of true and apparent temperatures T and TA [3–6]. For ex-
ample, Ratanapupech and Bautista [5] have determined the normal spectral
emissivities of molten iron–nickel alloys over the complete range of composi-
tion for a wavelength of 645 nm. As shown in Fig. 6.1, they used two crucibles
to hold molten samples: the outside crucible was made of tantalum and the
inside crucible was made of alumina, which kept molten samples from chem-
ical contamination. The tantalum crucible had a tantalum rod with a cavity
(ca 1.6 mm diameter and 16 mm depth), which emitted blackbody radiation
at the same temperature as the sample. The true and apparent temperatures
T and TA were measured using an optical pyrometer on radiations from the
cavity and the sample surface, respectively. The true temperature is some-
times measured using a thermocouple. Tomita et al. [6] have determined the
normal spectral emissivity of titanium arc-melted on a water-cooled copper
plate for a wavelength of 650 nm on the basis of (6.5), where the apparent
temperature was measured using an optical pyrometer and the true temper-
ature was measured using a W·5%Re–W·26%Re thermocouple sheathed in a
silica tube.
6.3.2 Method Based on Optical Constants
When a light beam is incident on a medium, several optical processes take

place. Some of the light is reflected from the front surface, some is absorbed in
the medium, and the rest is transmitted through the medium. The fractions of
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
ambient (n1)
φ1
linearly elliptically
polarised polarised
material (n,k)
Fig. 6.2. Reflection of linearly polarized light from material surface
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.
D1
Rotating
Analyser
D2
Sample
Electromagnetic Levitator
Linear
Polariser
Laser
Mirror
Fig. 6.4. Electromagnetic levitation with laser ellipsometry [14]
6.3.3 Method Based on Direct Measurements of Radiation

Intensities
According (6.1), the normal spectral emissivity can be obtained by measuring

radiation intensities from the sample and the blackbody at the same tempera-
ture, and this method has very often been applied to emissivity measurement
due to the simplicity in principle [16–19]. For example, Takasuka et al [18, 19]
have determined the normal spectral emissivities of solid and liquid silicon
at the melting point in the wavelength range 500–800 nm, using an experi-
mental setup shown in Fig. 6.5 [18]: A silicon sample is held in a graphite
crucible with cavities for blackbody radiation in a furnace. The temperature
of the sample is measured with an optical pyrometer using blackbody radi-
ation from a cavity. The normal radiation from the sample surface is led to
a monochromator using the lens and optical fibre system, and its intensity
is measured with a detector as a function of wavelength. The radiation in-
tensity from the cavity is also measured similarly and both intensities are
compared according to (6.1) to obtain the normal spectral emissivity. On oc-
casion, these measurements are made in different experiments [16,17]. Here it
should be noted that the term “radiation intensity” strictly means the spec-
tral radiation intensity in W m−3 sr−1 derived from Planck’s law of radiation
but not the output count (or voltage) from the detector simply. If the linear
relationship between the intensity and the output count is guaranteed over
a wide range of intensity, the emissivity can simply be derived as (output
count from the sample)/(output count from the cavity) [16]; however, the re-
lationship does not always obtain at any intensity, depending on the detector.
Because of this, the output count is usually calibrated using blackbody radia-
tion at various temperatures – this can be realized using fixed-point blackbody
Optical Filter Monochromator
Quartz
Window
Pyrometer
Filter
Quartz Lens
Ar (1 atm)
Carbon Crucible
Carbon Heater
Cavity
Heat Shield
Sample
Chamber
Fig. 6.5. Direct measurement of radiation intensity [18]
with various fixed-point metals – or combination of blackbody radiation at a

certain temperature and optical filters having various transmissivities.
Recently, Watanabe et al. [20] have established an emissivity measurement
system using a cold crucible furnace based upon the above principle, as shown
in Fig. 6.6, followed by many measurements using this system [21–27]. In the
cold crucible, eddy currents induced in a metallic sample by a high frequency
current heat the sample, while its surroundings are kept at lower temperature
because the cold crucible itself is water-cooled. Accordingly, the cold crucible
enables only a metal sample to be melted, which can eliminate the effect of
stray radiation on emissivity measurements. Furthermore, a repulsive force
between the cold crucible and the molten sample prevents the melt from
being in contact with the inner wall of the crucible, which helps to reduce
chemical contamination of the sample. Using this method, Watanabe and his
co-workers have made measurements on noble metals (Cu, Ag, and Au) [20,24,
26], transition metals (Fe, Co, and Ni) [25], and semi-conducting materials (Si
and Ge) [21] in visible and near-infrared regions; however, these measurements
have been limited only to the melting point of each substance due to the
difficulty in temperature measurement. Afterwards, Hayashi et al. [23] and
Tanaka et al. [27] have installed a two-colour pyrometer to this measurement
system and determined normal spectral emissivities of Ag–Cu and Ni–Cu
alloys, respectively, at temperatures higher than the liquidus temperatures.
More recently, Fukuyama and his co-workers [28] have applied an electro-
magnetic levitation technique to emissivity measurements. This technique can
Spectroscope
Prism
Window
Diaphragm
Camera
Window Window
Sample
Induction coils Ar- 5 %H2 (1 atm)
Cold crucible
Graphite plate Alumina tube
Air-tight chamber Positioning rod

Fig. 6.6. Normal spectral emissivity measurement using cold crucible [20]
realize completely containerless melting and thereby eliminate chemical con-

tamination of samples from containers. Another advantage in this technique
enables the sample to be supercooled to a large extent. They have determined
normal spectral emissivities of silicon in liquid state including supercooled
state (1,553–1,797 K) over a wavelength range 550–1,600 nm, where tempera-
ture measurements were made using a two-colour pyrometer. They have also
developed another technique for determining thermophysical properties in-
cluding hemispherical total emissivity, which is covered in detail in Chap. 8 of
this book.
6.3.4 Other Methods
(1) Microsecond pulse heating

The microsecond pulse heating system is basically a technique for heating
samples; however, thermophysical property measurements using this tech-
nique have been highlighted as “Subsecond Thermophysics” since 1990 [29]
and thus are outlined here.
Pottlacher and his co-workers [30–38] have made a vigorous effort to es-
tablish the microsecond pulse heating method to apply to measurements of
various thermophysical properties, including normal spectral emissivity. In
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:
ελ = ελ,M · (ISR /IBR ). (6.12)

Consequently, they have determined the emissivity for liquid nickel for a wave-
length of 850 nm in the temperature range 1,726–2,250 K.
More recently, Pottlacher and his co-workers have combined the microsec-
ond pulse heating system with ellipsometry and produced data of normal
spectral emissivities for niobium [31, 32], tungsten [32], nickel [32], tanta-
lum [32, 33], molybdenum [34], titanium alloy [35], nickel alloy [36], and pal-
ladium [37]. The readers can find a review about the pulse heating technique
in [38].
(2) Electrostatic levitation
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.
6.4 Emissivity Data

This chapter summarises normal spectral emissivities for some common metals
and alloys in the liquid state, which are important from practical and/or
scientific viewpoints.
6.4.1 Noble Metals
There have been considerable numbers of experimental investigations for the

spectral emissivity of Cu, Ag, and Au in liquid states.
Figure 6.8 shows the temperature dependence of the normal spectral emis-
sivity of liquid copper around 650 nm. The temperature coefficients are pos-
itive except the value by Burgess; however, there are large discrepancies in
the reported data. Figure 6.9 shows the emissivity of Cu as a function of
wavelength at the melting point. Krishnan et al. have used an apparatus
based on an electromagnetic levitation technique combined with rotating an-
alyzer ellipsometry, and the other data included in Fig. 6.8 were measured
based on the direct measurement methods. Even though Watanabe et al. and
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.
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]
6.4.2 Transition Metals
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
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]
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
Fig. 6.16. Normal spectral emissivities of liquid silicon as function of wavelength;

1 Shvarev et al. [55], 2 Takasuka et al. [56], 3 Tanaka et al. [22], 4 Kawamura
et al. [28], 5 Watanabe et al. [21]
temperature for the entire measured wavelengths, whereas those for W show
significant temperature dependence.
6.4.3 Semiconducting Materials
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
1
0.3
6 2
5
3
ε
0.2 4
Si
0.1
1600 1700 1800 1900 2000
T/K
Fig. 6.17. Normal spectral emissivity of liquid silicon at 633 nm as a function of

temperature; 1 Lampert [57], 2 Shvarev et al. (800 nm) [53], 3 Kawamura et al. [28],
4 Krishnan [58], 5 Lange et al. [47], 6 Takasuka [54]
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
Ag-Cu 1300 nm 2
0.06
1
ε
0.05 4
6
5
0.04
1200 1300 1400 1500
T/K
Fig. 6.19. Temperature dependencies of spectral emissivities of Ag–Cu alloys mea-

sured at a wavelength of 1,300 nm; 1 Cu, 2 Ag-80 at%Cu, 3 Ag-60 at%Cu, 4 Ag-
40 at%Cu, 5 Ag-2 at%Cu, 6 Ag [23]
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]
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)
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)
19. E. Takasuka, E. Tokizaki, K. Terashima, S. Kimura, J. Appl. Phys. 81, 6384

(1997)
20. H. Watanabe, M. Susa, K. Nagata, Metall. Mater. Trans. A 28A, 2507 (1997)
21. H. Watanabe, M. Susa, H. Fukuyama, K. Nagata, High Temp. High Press. 31,
587 (1999)
22. R. Tanaka, M. Susa, High Temp. High Press. 34, 681 (2002)
23. M. Hayashi, M. Murata, H. Fukuyama, K. Nagata, Metall. Mater. Trans. B
33B, 47 (2002)
24. H. Watanabe, M. Susa, H. Fukuyama, K. Nagata, Int. J. Thermophys. 24, 223
(2003)
(2003)
(2003)
27. R. Tanaka, M. Susa, in Proc of 7 th ECTP Conf., CD-ROM (2005)
28. H. Kawamura, H. Fukuyama, M. Watanabe, T. Hibiya, Meas. Sci, Technol. 16,
386 (2005)
29. A. Cezairliyan, G.R. Gathers, A.M. Malvezzi, A.P. Miiller, F. Righini, J.W.
Shaner, Int. J. Thermophys. 11, 819 (1990)
30. E. Kaschnitz, G. Pottlacher, H. Jäger, Int. J. Thermophys. 13, 699 (1992)
31. A. Seifter, F. Sachsenhofer, S. Krishnan, G. Pottlacher, Int. J. Thermophys. 22,
1537 (2001)
32. A. Seifter, F. Sachsenhofer, G. Pottlacher, Int. J. Thermophys. 23, 1267 (2002)
33. G. Pottlacher, A. Seifter, Int. J. Thermophys. 23, 1281 (2002)
34. C. Cagran, B. Wilthan, G. Pottlacher, Int. J. Thermophys. 25, 1551 (2004)
35. M. Boivineau, C. Cagran, D. Doytier, V. Eyraud, M.-H. Nadal, B. Wilthan,
G. Pottlacher, Int. J. Thermophys. 27, 507 (2006)
36. B. Wilthan, G. Pottlacher, Rare Metals 25, 592 (2006)
37. C. Cagran, G. Pottlacher, Platinum Metals Rev. 50, 144 (2006)
38. M. Boivineau, G. Pottlacher, Int. J. Mater. Prod. Technol. 26, 217 (2006)
39. W.K. Rhim, K. Ohsaka, J. Cryst. Growth 208, 313 (2000)
40. A.J. Rulison, W.K. Rhim, Metall. Mater. Trans. B 26B, 503 (1995)
41. G.K. Burgess, Bull. Nat. Bur. Stand. 6, 111 (1909)
42. C.M. Stubbs, Proc. Roy. Soc. (London) Ser. A 88, 195 (1913)
43. C.C. Bidwell, Phys. Rev. Ser. 2 3, 439 (1914)
44. G.K. Burgess, R.G. Waltenberg, Bull. Nat. Bur. Stand. 11, 591 (1915)
45. D.B. Smith, J. Chipman, Trans. AIME 194, 643 (1952)
46. von K.W. Lange, H.Schenck, Arch. Eisenhuettenw. 39, 611 (1968)
47. L.A. Stretz, R.G. Bautista, Temperature, Its Measuremant and Control in Sci-
ence and Industry vol. 4, Part 1, ed. by H. Preston-Thomas, T.P. Murray, R.L.
Shepard (Instrument Society of America, Pittsburgh, PA, 1972) p 489
48. D.W. Bonnell, J.A. Treverton, A.J. Valerga, J.L. Margrave, Temperature, Its
Measurement and Control in Science and Industry, vol. 4, Part 1, ed. by
H. Preston-Thomas, T.P. Murray, R.L. Shepard (Instrument Society of America,
Pittsburgh, PA, 1972), p 483
49. S. Krishnan, G.P. Hansen, R.H. Hauge, J.L. Margrave, High Temp. Sci 26, 143
(1990)
50. R. Tanaka, M. Susa, in 17th European Conference on Thermophysical Properties,
Collection of Contributions CD-ROM, Bratislava, Slovak Republic, 2005
51. M.N. Dastur, N.A. Gokcen, J. Metals 1, 665 (1949)

52. J.C. d’Entremont, Trans. AIME 227, 482(1963)
53. S. Krishnan, D. Basak, Int. J. Thermophys. 20, 1811 (1999)
54. C. Cagran, G. Pottlacher, M. Rink, W. Bauer, Int. J. Thermophys. 26, 1001
(2005)
55. K.M. Shvarev, B.A. Baum, P.V. Gel’d, Sov. Phys. Solid State 16, 2111 (1975)
56. E. Takasuka, E. Tokizaki, K. Terashima, S. Kimura, J. Appl. Phys. 81, 6384
(1997)
57. M.O. Lampert, J.M. Koebel, P. Siffert, J. Appl. Phys. 52, 4975 (1981)
58. S. Krishnan, J.K.R. Weber, P.C. Nordine, R.A. Schiffman, R.H. Hauge, J.L.
Margrave, High. Temp. Sci. 30, 137 (1991)
59. K. Schaefers, M. Röner-Kuhn, M.G. Frohberg, Int. J. Thermophys. 16, 997
(1995)
7
Noncontact Thermophysical Property
Measurements of Metallic Melts
under Microgravity
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
Microgravity is the term commonly used to describe weightlessness conditions

onboard a spacecraft. Unfortunately, this term is quite misleading: first of all,
gravity remains essentially unaltered, that is, the strength of the gravity field
300 km above the surface of the earth is still about 90% of its sea-level value.
Second, the apparent weight is typically reduced by a factor of 103 , rather
than 106 .
The physical origin for weightlessness lies in the fact that the experiments
are carried out in a frame of reference which is accelerated with respect to
an inertial frame. Consequently, Newton’s equations of motion have to be
supplemented by a fictitious force:
Fs = −ms̈,
where s̈ is the acceleration of the frame of reference along its trajectory, s,
with respect to an inertial system (e.g., the earth.) For a circular motion with
angular frequency ω and radius R, like an orbiting satellite, this force is the
centrifugal force:
Fc = mω 2 R.
If this force cancels the gravitational force, there is no net force on the body,
it is weightless. In reality, a complete cancellation of the forces cannot be
achieved, due to the spatial dependence of the earth’s gravitational field (tidal
forces), the still existing air drag, and internal disturbances within the space-
craft (g-jitter). Weightlessness conditions can be achieved in drop tubes or
drop towers, in parabolic flights, onboard sounding rockets, and on orbiting
spacecrafts like the International Space Station, ISS.
7 Metallic Melts under Microgravity 133
Table 7.1. Available microgravity platforms
g level(g0 ) μg duration(s) payload volume(m3 )

Drop tower 10−3 − 10−6 1 − 10 ≈ 0.1
Aircraft 10−2 − 10−3 ≈ 20 ≈ 1.0
Sounding rocket 10−4 − 10−6 100 − 1,000 ≈ 0.1
Space Station 10−3 − 10−6 ∞ ≈ 1.0
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.
7.3 Containerless Methods
Thermophysical properties of high temperature, and highly reactive, melts can

be conveniently measured by containerless methods. These methods provide
the purest environment possible. Since the surface of the liquid sample is not in
contact with a wall, Marangoni convection will occur, if there is a temperature
or concentration gradient along the surface.
There are a number of containerless processing techniques, based on dif-
ferent levitation fields [3]. Acoustic and aerodynamic levitation techniques use
standing waves or pressure gradients of a carrier gas, respectively. In electro-
static levitation, an electrostatic field is used to position a charged or polarized
sample. In diamagnetic levitation, the molecular magnetic moments couple
to an external dc magnetic field of typically 1–5 T. Electromagnetic levita-
tion uses high frequency electromagnetic fields and exploits the Lorentz force
to levitate electrically conducting specimen. Except for the latter, all other
methods require an independent heating system in order to melt the sample,
usually an infrared laser. Although there have been some early attempts to
operate an acoustic levitator under microgravity conditions, up to date the
only successful implementation of a microgravity levitator is TEMPUS [4, 5],
based on the electromagnetic levitation principle. Thermophysical properties
of levitated samples have been measured in TEMPUS during two Spacelab
missions, IML-2 in 1994 [6] and MSL-1 in 1997 [7]. Later on, TEMPUS was
flown regularly on parabolic flight campaigns. Its successor, MSL-EML, the

Materials Science Lab – Electro-Magnetic Levitator, was flown in 2006 on a
sounding rocket, and is presently being built as a second generation payload
for ISS, to be flown in 2011.
In the following, the principle of electromagnetic levitation will be briefly
described. Levitation of electrically conducting samples is achieved by placing
the sample into a high frequency alternating inhomogeneous electromagnetic
field, produced by a levitation coil with a conical or cylindrical shape. This
field B induces a current in the sample which, in turn, interacts with the field.
Levitation is caused by the Lorentz force, F . Its magnitude can be expressed,
to lowest order in a multipole expansion [8], as
∇B 2 4π 3
FL = − a Q(q), (7.1)
2μ0 3 0
where μ0 is the magnetic permeability, a0 is the radius of the sample, q = a0 /δ
is a dimensionless quantity, and δ is the skin depth, defined as follows:

ωρμ0
1/δ = . (7.2)
2
Here, ρ is the electric conductivity of the sample and ω is the frequency of
the alternating field. The function Q(q) is given by

3 3 sinh(2q) − sin(2q)
Q(q) = 1− . (7.3)
4 2q cosh(2q) − cos(2q)
For stable levitation, this field has to cancel the gravitational field:
−→ −
→ 4π
FL = −Fg = − a30 ρ− →g. (7.4)
3
Here, ρ is the density of the sample and g is the gravity vector. For a linear
magnetic field, ∂H∂z = Hzz = const, the levitation force is given by [9]
z
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. 7.2. Heating and positioning efficiencies of electromagnetic levitation

7.4 Thermophysical Properties
The measurement of thermophysical properties of liquid metals using elec-

tromagnetic levitation has become a routine task in the past few years, at
least concerning surface tension and density measurements, and comprehen-
sive reviews are available [14–16]. Concerning experiments under microgravity,
the major event was the Spacelab mission MSL-1 in 1997 [7]. Since then, no
long-duration microgravity campaign has taken place. However, a number of
parabolic flights using electromagnetic levitation has been performed, and also
one sounding rocket experiment. Consequently, most of the results discussed
below have been obtained during MSL-1 and are now 10 years old. Where ap-
propriate, they are supplemented by recent results obtained during parabolic
flights.
7.4.1 Electrical Conductivity
It is possible to measure the electrical conductivity of levitated melts using a

noncontact, inductive method. The impedance of a coil surrounding the sam-
ple is influenced by the sample’s electrical conductivity. For spherical samples
and homogeneous magnetic fields, as realized in microgravity, this relation is
rather simple [17], whereas under terrestrial conditions, extensive mathemati-
cal and engineering effort is required to extract the required information from
the measured impedance [18].
The complex impedance Z of the coil is a function of the oscillation fre-
quency ω of the electromagnetic field, properties of the empty coil, and prop-
erties of the sample:
Zcoil = Zcoil (ω, δ, a0 , −
→
r ,−
→
α ) = R0 + iωL0 + ΔRs (δ, a0 , −
→
r ,−
→
α)
+iωΔLs (δ, a0 , −
→
r ,−
→
α ). (7.9)
Here, R0 and L0 are resistance and inductivity of the empty circuit and ΔRs ,
ΔLs are the changes induced by the sample. These latter functions depend
on the skin depth, δ, the radius a0 of the sample, and its position −→r within
the coil. The vector −
→
α is meant to describe the shape of the sample as, for
example, given by the coefficients of a series expansion in spherical harmonics.
In a homogeneous magnetic field, and for a spherical sample, the dependence
on −→
r and −→
α vanish.
The complex impedance Ztot of the oscillatory circuit can be determined
by measuring amplitudes of current I and voltage U , as well as their phase
shift φ. The real and imaginary part of Ztot yield two equations for the two
unknowns δ and a0 . For small ohmic losses in the circuit these are given by
C I0
(R + Rs ) = cos(φ), (7.10)
L U0

C L 2 I0
ω − (LC)−1 + ωLs = sin(φ). (7.11)
L ω U0
138 I. Egry
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.
7.4.2 Density and Thermal Expansion
Density and thermal expansion of levitated drops are determined by record-

ing the visible cross section of the sample. Assuming rotational symmetry,
the volume is calculated. Since the mass of the sample is known and does
not change, this yields the density. Typically, a resolution of δV /V = 10−4
is required. This can be achieved using sub-pixel algorithms for edge detec-
tion, curve fitting of the shape, and statistical averaging [21]. Although such
measurements can be performed on ground [22],the precision of the data is
improved in microgravity due to the spherical shape of the sample.
The volume of an axisymmetric sample can be calculated from
π
2
VP = π <R(ϕ)>3 dϕ, (7.14)
3 0
where <R(ϕ)> is the time-averaged, angle-dependent radius of the sample,
which, in the case of a sphere, reduces to a0 , yielding VP = 4π 3
3 a0 . The volume,
VP , is given in pixel units and must be converted to physical units using a cal-
ibration body. This is the main source of error in this kind of measurements.
Another difficulty arises from the edge detection algorithm. The brightness
of the sample changes with temperature, and the edge detection must be
independent of contrast changes and blooming effects. In terrestrial experi-
ments, this problem is bypassed by using a shadowgraph technique [22], but
the TEMPUS facility used in the microgravity experiments does not have this
feature installed.
During MSL-1, Samwer and coworkers have performed measurements on
glass-forming metallic alloys [23]. For such systems, two interesting questions
need to be addressed: A change in the slope of the density as a function of
temperature, that is, a change in the thermal expansion, in the undercooled
regime would indicate a liquid–liquid phase transition, for example, phase
separation. The other interesting question is related to the glass transition
itself: it has been speculated that, at this temperature, the density of the
undercooled liquid would become equal to that of the crystalline solid. Un-
fortunately, the data of Samwer and coworkers do not extend close enough to
the glass transition temperature to answer this question.
More recently, the same group measured the density of liquid Si–Ge alloys
with TEMPUS during parabolic flights [24]. They found good agreement with
ground based data obtained in an electrostatic levitator, and an interesting,
yet unexplained, anomaly in the thermal expansion around Si75 Ge25 . Their
raw data for the volume of an Si25 Ge75 alloy are shown in Fig. 7.4.
7.4.3 Specific Heat
A noncontact method developed by Fecht and Johnson [25] can be used to

determine the specific heat in levitation experiments. It is a variant of noncon-
tact modulation calorimetry, normally used in low temperature physics. The
140 I. Egry
3.0x107 Sample: Si25Ge75
2.5x107
Volume(pixel3)
2.0x107
1.5x107
1.0x107
5.0x106
1400 1500 1600 1700 1800 1900

Temperature(K)
Fig. 7.4. Volume of liquid Si25 Ge75 as a function of temperature, measured during
a parabolic flight with TEMPUS
heater power is modulated according to Pω (t) = ΔPω cos(ωt), resulting in a

modulated temperature response ΔTω of the sample. A thermal model con-
sidering heat loss to the environment and heat conduction within the sample
has been developed by Wunderlich [26]. It considers heat loss to the exterior
(typically by radiation only if the experiment is carried out under vacuum),
spatially inhomogeneous heating of the sample, and heat conduction within
the sample. For small Biot numbers, Bi = kr /kc 1, where kr is the heat
loss due to radiation and kc the heat loss due to heat conduction, adiabatic
conditions are realized and quantitative modulation calorimetry is possible.
Under such conditions, the following relation for ΔTω is obtained:
"
2 !
2 !#−1/2
ΔPω ω λ2
ΔTω = 1+ 1+ , (7.15)
ωcp λ1 ω
where λ1 and λ2 are functions of kr and kc such that for Bi 1, λ2 λ1 .
This allows to choose the modulation frequency between λ2 and λ1 such that
λ2 ω λ1 . Under these circumstances, a simple relation for the specific
heat, cp , can be derived:
1 ΔPω
cp = . (7.16)
ω ΔTω
The power input into the sample, ΔPω , cannot be measured directly. It is re-
lated to the current I flowing through the heating coil by a coupling coefficient
GH , which has to be determined separately:
ΔPω = GH Iω2 (7.17)
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
It should be noted that a harmonic modulation of the current with ω leads to

components in the power P with ω = 0, ω = ω and ω = 2ω . The static and
frequency-doubled components can also be analyzed along the lines indicated
above.
Modulation calorimetry needs to be carried out under isothermal con-
ditions, that is, the sample must be allowed to thermalize at a prede-
fined temperature before the modulation signal can be applied. Therefore,
such experiments cannot be performed during parabolic flights. Fecht and
coworkers [27] applied this method during the MSL-1 Spacelab mission. The
temperature modulation is shown in Fig. 7.5, where the quaternary alloy
Zr65 Al7.5 Cu17.5 Ni10 was undercooled by 194 K, and modulation calorimetery
could be performed in the equilibrium as well as in the undercooled liquid.
More recently, Wunderlich et al. performed modulation calorimetry during a
sounding rocket experiment on a Ti46 Al8 Nb alloy in the framework of the
IMPRESS project [28]. It was possible to determine both the specific heat,
and the total hemispherical emissivity in the liquid phase [29]. The results are
shown in Table 7.2.
7.4.4 Viscosity and Surface Tension
Viscosity and surface tension of levitated samples are conveniently measured

by the oscillating drop technique [30]. Liquid samples perform oscillations
around their equilibrium shape. In microgravity, this is a sphere and in that
142 I. Egry
Table 7.2. Ti46 Al8 Nb specific heat capacity and total hemispherical emissivity in
the liquid phase
T (◦ C) 1570 1491 1468

−1 −1
cp (J K g ) 1.14 1.06 1.03
εtot 0.44 0.36 0.34
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
Fig. 7.6. Frequency spectrum of an oscillating AuCu drop under 1g (top) and
microgravity (bottom)
be seen in the ground-based spectra, both, a splitting of the single frequency

into 5 peaks, and a shift to higher frequencies occur.
By comparing measurements in microgravity with ground-based results,
the validity of the Cummings correction has been confirmed experimentally.
It is fair to say that this benchmark experiment under microgravity has led to
a calibration of the ground-based measurements applying the oscillating drop
technique, which has consequently become the accepted method for surface
tension measurements of high-temperature liquid metals.
Surface tension measurements using the oscillating drop technique are
short duration experiments. Nowadays, they are routinely carried out during
parabolic flight campaigns. For example, Wunderlich et al. [32, 33] have mea-
sured the surface tension of a number of industrially relevant multicomponent
steels and nickel-based superalloys in the framework of ESA’s ThermoLab
project [32]. A list of the investigated alloys is given in Table 7.3.
Recently, there have been also surface tension measurements using the oscil-
lating drop method under microgravity conditions in Japan, notably by the
group of Nogi [34–36]. This group has either used the JAMIC drop tower [36]
or a short 1.5 m drop tube [34, 35] and measured the oscillations during free
fall of the drop.
The oscillating drop technique also allows to measure the viscosity of liq-
uids by measuring the decay of oscillations and using (7.19). Since no cor-
rection formula, equivalent to the Cummings correction, (7.20), exists, the
application of the oscillating drop technique to viscosity measurements is re-
stricted to microgravity conditions. Egry and coworkers [37] have measured
the viscosity of a PdCuSi alloy during the MSL-1 mission. Their result is
144 I. Egry
Table 7.3. Surface tension measurements of industrial alloys measured during

parabolic flights in the framework of the ThermoLab project
Alloy Tl (◦ C) γ(Tl ) (N m−1 ) Year

CMSX-4 1,384 1.85 2002
IN738LC 1,335 1.85 2003
MM247LC 1,368 1.86 2003
C263 1,368 1.74 2003
Low alloyed Steel 1,480 1.61 2004
Cr-alloyed Steel 1,477 1.76 2004
Cu-alloy 1,067 1.15 2004
Al-75 at% Ni 1,410 1.63 2004
Nickel 1,455 1.71 2004
Al-31.5 at% Ni 1,250 1.16 2005
Al-65 at% Ni 1,550 1.65 2006
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.
7.5 Summary and Outlook
In the preceding sections, we have demonstrated the potential of the combined

use of the microgravity environment and containerless processing, in partic-
ular, electromagnetic levitation, for thermophysical property measurements.
Because of the lack of flight opportunities, this potential has not been fully
exploited yet. The facilities already operational on board the International
Space Station, ISS, are of limited use for thermophysical property measure-
ments. MSL-EML, the Materials Science Lab – Electro-Magnetic Levitator,
has still to be built and will not become available before 2011. Until then,
most thermophysical property measurements have to be carried out on board
parabolic flights within 20 s of microgravity. This restricts the use of TEMPUS
to surface tension measurements, as discussed earlier. For most other prop-
erties, the available microgravity period is not sufficient. Unfortunately, the
Drop Tower at JAMIC has also been closed down. Using sounding rockets as a
microgravity platform is attractive in terms of microgravity duration (about
5 min) and also in terms of microgravity quality (about 10−5 g0 ). However,
there is only one flight every 2 years at best within the national German pro-
gram, which is co-financed by and shared with ESA. Because of restrictions in
payload mass and the available μg-time, only two samples can be processed,
yielding about 160 s experiment time per sample, as shown, for example, in
Fig. 7.1. Therefore, TEXUS is by far not as cost-effective as either parabolic
flights or the ISS.
When the MSL-EML on ISS becomes fully operational, experiment time
will be virtually unlimited. This is the chance for performing systematic
investigations varying the process parameters like, for example, tempera-
ture, cooling and heating rates, and composition of the ambient atmosphere,
that is, the oxygen partial pressure. Generally speaking, the emphasis will
shift from technology development to exploiting the technology for measure-
ments of industrially and scientifically interesting (multicomponent) alloys.
Even in this optimistic scenario, microgravity experiments will not replace a
sound, ground-based measurement program. Microgravity experiments pro-
vide benchmarks for terrestrial measurements owing to their inherent higher
precision. However, the huge amount of data required for numerical mod-
eling of industrial processes by far exceeds the capabilities of spaceborne
experiments.
While most thermophysical properties can be measured in a containerless
environment using, for example, MSL-EML, this is not the case for mea-
surements of diffusion and thermal conductivity. Here, the application of an
additional strong static magnetic field, reducing or even eliminating fluid flows
[39, 40] may open up new possibilities.
146 I. Egry
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)
26. R. Wunderlich, H. Fecht, Meas. Sci. Tech. 16, 402 (2005)

27. R. Wunderlich, R. Sagel, Ch. Ettel, H.-J. Fecht, D. Lee, S. Glade, W. John-
son, Solidification 1999, ed. by W. Hofmeister, J. Rogers, N. Singh, S. Marsh,
P. Vorhees, vol 53 (TMS, Warrendale, 1999)
28. D.J. Jarvis, D. Voss, Mat. Sci. and Engng. A 413–414, 583–591 (2005)
29. R. Wunderlich, private communication
30. S. Sauerland, K. Eckler, I. Egry, J. Mat. Sci. Letters, 11, 330 (1992)
31. I. Egry, G. Lohöfer, G. Jacobs, Phys. Rev. Letts. 75 4043 (1995)
32. R. Aune, L. Battezzati, R. Brooks, I. Egry, H. Fecht, J. Garandet, K. Mills,
A. Passerone, P. Quested, E. Ricci, S. Schneider, S. Seetharaman, R. Wunder-
lich, B. Vinet, Microgravity Sci. Technol. XVI, 11 (2005)
33. K. Higuchi, H.-J. Fecht, R.K. Wunderlich, Adv. Eng. Mat. 9, 349 (2007)
34. H. Fujii, T. Matsumoto, T. Ueda, K. Nogi, J. Mater. Sci. 40, 2161 (2005)
35. T. Matsumoto, H. Fujii, T. Ueda, M. Kamai, K. Nogi, Meas. Sci. Tech. 16, 432
(2005)
36. H. Fujii, T. Matsumoto, S. Izutani, S. Kiguchi, K. Nogi, Acta Mat. 54, 1221
(2006)
37. I. Egry, G. Lohöfer, I. Seyhan, S. Schneider, B. Feuerbacher, Appl. Phys. Lett.
73, 462 (1998)
38. S. Lee, K. Tsang, H. Kai, J. Appl. Phys. 70, 4842 (1991)
39. H. Kobatake, H. Fukuyama, I. Minato, T. Tsukada, S. Awaji, Appl. Phys. Lett.
90, 094102 (2007)
40. H. Kobatake, H. Fukuyama, I. Minato, T. Tsukada, S. Awaji, J. Appl. Phys.
104, 054901 (2008)
8
Noncontact Laser Calorimetry of High
Temperature Melts in a Static Magnetic Field
Hiroyuki Fukuyama, Hidekazu Kobatake, Takao Tsukada, and Satoshi Awaji
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.
An electromagnetic levitator is equipped in a superconducting magnet to hold

a droplet statically. The high-temperature droplet is heated sinusoidally by
laser irradiation on the top instead of using a modulated rf-coil heating; then,
the temperature response is measured at the bottom of the droplet using a
pyrometer. This technique enables us to measure the thermophysical proper-
ties, especially thermal conductivity, over a wide temperature range without
contamination. This chapter presents an explanation of the theoretical back-
ground of this novel method to determine thermophysical properties such as
heat capacity, thermal conductivity, and emissivity of a levitated droplet us-
ing an electromagnetic levitator with a static magnetic field. To confirm the
fact that a static magnetic field actually suppresses the convection in a molten
silicon droplet in the electromagnetic levitator, numerical simulations for con-
vection in the droplet are demonstrated taking into account the buoyancy
force, thermocapillary force, and electromagnetic force in the droplet. Finally,
experimental details are presented for molten silicon as an example.
8.2 Theory of Modulation Calorimetry
8.2.1 Heat Capacity
The principle of the conventional modulation calorimetry is described in de-

tail elsewhere [12, 13]. Figure 8.1a presents a schematic illustration of the
electromagnetic levitation apparatus incorporating a static magnetic field for
laser modulation calorimetry. In the experiment, the upper part of an electro-
magnetically levitated droplet is heated periodically using a modulated light
source, that is, a semiconductor laser (Fig. 8.1b); then the temperature varia-
tion at the lower part of the droplet attributable to heat flow from the upper
part through the droplet is detected using a pyrometer. The simple heat flow
model of this noncontact modulation calorimetry is presented in Fig. 8.2 [9];
it resembles that developed by Wunderlich and Fecht [7]. The sample droplet,
which is levitated in the vacuum chamber, is heated using a modulated laser
at a power of po (1 + cos ωt) per unit area [W m−2 ] from the top of the droplet.
The temperature response is measured at the bottom of the droplet using a
two-color pyrometer. The fraction of the irradiated surface area that is heated
by the laser is defined as Sh ; the corresponding volume fraction is Vh . The
heat balance equations for the system are as follows.
For the laser irradiated part,
dTh
Vh Cp = Qh +αSh Apo (1+cos ωt)−Sh Aσ(Th4 −Ta4 )−Kc (Th −Tl ). (8.1)
dt
For the laser nonirradiated part,
dTl
(1 − Vh )Cp = Ql − (1 − Sh )Aσ(Tl4 − Ta4 ) + Kc (Th − Tl ). (8.2)
dt
8 Noncontact Calorimetry of High Temperature Melts 151
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
Laser irradiated ShKr

part: Sh, Vh
Radiative heat
transfer
Conductive heat Heat
Kc transfer bath
Radiative heat
transfer
Laser non-irradiated
part: (1-Sh), (1-Vh) (1-Sh)Kr
Fig. 8.2. Heat flow model in this noncontact modulated laser calorimetry
In those equations, Cp [J K−1 ]denotes the isobaric heat capacity, T [K]

is the absolute temperature, Ta [K] is the ambient temperature, Q [W] is
the power input from the rf-coil, α is the absorptivity, A [m2 ] is the surface
area of the droplet, is the hemispherical total emissivity, σ [W m−2 K−4 ] is

the Stefan–Boltzmann constant, and Kc [W K−1 ] is the thermal conductance
for conductive heat transfer from the laser-irradiated part to the nonirradi-
ated part. Subscripts h and l, respectively, denote the laser irradiated and
nonirradiated parts. In Fig. 8.2, Kr [W K−1 ] is the thermal conductance for
radiative heat transfer from the sample surface to the heat reservoir in vac-
uum (Kr = 4AσTo3 , To is the initial temperature of the sample). The sample
temperature is expressed as the sum of the initial temperature To , the average
increase in temperature ΔTdc (dc component), and the modulation amplitude
md
ΔTac (ac component) as
md
T = To + ΔTdc + ΔTac . (8.3)
The modulation amplitude is expressed in terms of the phase difference φ
between the laser irradiation and the temperature response as
md
ΔTac = ΔTac cos(ωt − φ). (8.4)
Under the condition of To
ΔTdc , ΔTac md
, the following linearization is rea-
sonably applied for the radiative heat loss term in (8.1) and (8.2).
md

ΔTdc ΔTac
T = (To + ΔTdc + ΔTac ) To 1 + 4
4 md 4 4
+4 . (8.5)
To To
The heat balance between the power input from the rf-coil and radiation heat
loss is given by the following equation when the sample droplet is maintained
at the initial temperature.
Q = Qh + Ql = Aσ(To4 − Ta4 ). (8.6)

Substituting (8.3), (8.5), and (8.6) to (8.1) and (8.2), the ac components are
expressed using the following equations.
For the laser irradiated part:
md
dΔTac,h
Vh Cp = αSh Apo cos ωt − 4Sh AσTo3 ΔTac,h
md
dt
−Kc (ΔTac,h
md
− ΔTac,l
md
). (8.7)
For the laser nonirradiated part:
md
dΔTac,l
(1 − Vh )Cp = −4(1 − Sh )AσTo3 ΔTac,l
md
dt
md
+Kc (ΔTac,h − ΔTac,l
md
). (8.8)
The modulation amplitude ΔTac,l and the phase difference are derived by
solving (8.7) and (8.8). They are simplified under the condition of Kr /Kc ≤
0.01 as
−1/2
αSh Apo 1 2 2
ΔTac,l = 1 + 2 2 + ω τc (8.9)
ωCp ω τr
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.
8.2.2 Thermal Conductivity and Emissivity
Governing Equations and Boundary Conditions Figure 8.3 shows a

schematic diagram of the modulated laser calorimetry in spherical coordi-
nates used in our studies to determine thermal conductivity and emissivity
rlaser
z
q r
y
x
Fig. 8.3. Modulated laser calorimetry in spherical coordinates
of molten materials simultaneously [9–11]. The upper part of an electromag-

netically levitated droplet is heated periodically using a modulated laser; the
temperature variation at the lower part of the droplet is detected using a py-
rometer. By carrying out a similar experiment at various frequencies of the
modulated laser ω, the relation of the phase difference between the modulated
laser and the temperature variation, φ, to ω is obtained, where φ depends on
the thermal conductivity, emissivity, and diameter of the droplet aside from ω.
For analysis of the temperature variations in an electromagnetically lev-
itated droplet whose upper part is irradiated using a modulated laser, the
following are assumed: (1) the system is axially symmetric; (2) the thermo-
physical properties of the droplet are constant; (3) the droplet is opaque to
the light source, that is, the light source is partially absorbed at the surface
and the rest is reflected; (4) the distribution of laser intensity is Gaussian;
(5) the heat loss from the droplet surface is radiation alone; and (6) the heat
transfer in the droplet is governed solely by conduction.
Under the aforementioned assumption, the unsteady-state heat conduction
equation in the spherical coordinate system is expressed as

∂T 1 ∂ 2 ∂T 1 ∂ ∂T
ρcp,mass = κ r + sin θ
∂t r2 ∂r ∂r r2 sin θ ∂θ ∂θ
+Q(r, θ), (8.14)
where cp,mass , κ, r, t, θ, and ρ respectively denote the isobaric mass heat ca-
pacity (specific heat), thermal conductivity, radial distance in the spherical
coordinates, time, polar angle in the spherical coordinates, and density. In
addition, Q(r, θ) in (8.14) is the heat generation rate attributable to electro-
magnetic induction heating.
The boundary conditions are given by the following equations. At the
droplet surface irradiated by the laser,

∂T 2R2 sin2 θ
−κ = σ(T 4 − Ta4 ) − αIo (t) exp − s 2 (−n · elaser ). (8.15)
∂n rlaser
At the droplet surface without the light source,
∂T
−κ = σ(T 4 − Ta4 ). (8.16)
∂n
At the centerline,
∂T
−κ =0 (8.17)
∂θ
and the initial condition is as follows.
T = To (r, θ). (8.18)
Here, elaser and n, respectively, denote unit vector representing the incident
direction of the light source and unit normal vector at the droplet surface.
Rs and rlaser are radial distance from the centerline to the droplet surface
and e−2 radius of laser beam, respectively. Furthermore, Io [W m−2 ] is the
laser beam intensity at the centerline, which is related to the power of the
modulated laser beam, Po (cos ωt + 1) [W], as
2 2
Io (t) = 2 P (t) = 2 Po (cos ωt + 1). (8.19)
πrlaser πrlaser
Therein, To (r, θ) is the initial temperature distribution in the droplet imme-
diately before laser heating, as determined by considering the electromagnet-
ically induced Joule heat.
Simplified Model Basically, the thermal conductivity and emissivity of the
droplet can be determined by fitting the experimental temperature response
with the results obtained by solving (8.14)–(8.19) numerically if the thermo-
physical properties other than these parameters, that is, density and specific
heat, are known. However, because (8.14) gives unsteady-state solutions, and
because the boundary conditions for radiation at the droplet surface, (8.15)
and (8.16), are nonlinear, they are too troublesome to use in a system of equa-
tions for determining thermophysical properties. Furthermore, Ta and To (r, θ)
must be predetermined experimentally, but it is difficult to measure these
values. Consequently, (8.14)–(8.18) are simplified as described later.
When the upper part of the droplet is irradiated by the modulated laser
beam, as presented in Fig. 8.3, the temperature at each point in the droplet,
T (r, θ, t), exhibits increases in average temperature and modulation amplitude
from the initial temperature; then reaches the stationary modulation state
with certain constant average temperature and amplitude, depending on the
laser beam’s power and frequency. Therefore, we consider only the tempera-
ture response at the stationary modulation state and express the temperature
T (r, θ, t) as
md
T (r, θ, t) = To + ΔTdc (r, θ) + ΔTac (r, θ, t), (8.20)
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
out
. (8.25)
∂n
At the droplet surface without the light source, the following are true:
in
∂ΔTac
in
, (8.26)
∂n
out
∂ΔTac
out
. (8.27)
∂n
At the centerline,
in
∂ΔTac
− = 0, (8.28)
∂θ
and
out
∂ΔTac
− = 0. (8.29)
∂θ
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.
8.2.3 Verification of the Assumptions of Conduction-Dominated

Heat Transfer
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
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)
Fig. 8.4. Velocity and temperature fields in an electromagnetically levitated liquid

silicon droplet while applying a magnetic flux density of 4 T
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.
8.2.4 Verification of the Model and Sensitivity Analysis
Virtual measurement of the properties was carried out by numerical simula-

tion to ascertain the reliability and accuracy of the model described above to
determine thermal conductivity and emissivity of molten materials simultane-
ously. Here, the electromagnetic field in the electromagnetic levitator actually
used in the experiment is analyzed numerically; then the temperature fields
in the electromagnetically levitated droplet during modulated laser calorime-
try are computed by solving (8.14) under the boundary and initial conditions,
(8.15)–(8.18). In this case, it is assumed that the heat transfer in the droplet is
governed by conduction alone because convection is suppressed by applying a
static magnetic field. Figure 8.5a shows the calculated temperature response
at the lower part of the droplet, that is, the integral mean value of tem-
perature over the region corresponding to the spot area of the pyrometer, for
ω/2π = 0.1 Hz, and additionally Fig. 8.5b shows the temperature distributions
in the droplet for (i)–(iv) in Fig. 8.5a. The physical properties of the molten
silicon and processing parameters used in the calculations are presented in
Table 8.1. From those figures, it is apparent that the electromagnetically in-
duced heterogeneous heat affects the isotherms in the droplet.
By carrying out calculations similar to those in Fig. 8.5a at various fre-
quencies of the modulated laser source ω, the relation between the phase
difference φ and ω is obtained numerically as presented in Fig. 8.6. Then, the
simplified mathematical model described in the previous section was adopted
(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)
(i) t = 190.0 s (ii) t = 192.5 s

Tmin = 1790 K, Tmax = 1798 K Tmin = 1798 K, Tmax = 1810 K
ΔT = 0.84 K ΔT = 0.84 K
1811 K
1789 K
(iii) t = 195.1 s (iv) t = 197.6 s

Tmin = 1801 K, Tmax = 1809 K Tmin = 1790 K, Tmax = 1803 K
ΔT = 0.84 K ΔT = 0.84 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
Table 8.1. Physical properties and processing conditions used in calculations
Physical properties of molten silicon

Temperature coefficient of surface tension [N m−1 K−1 ] −4.3 × 10−4
Viscosity [kg m−1 s−1 ] 7.0 × 10−4
Density [kg m−3 ] 2,530
Thermal expansion coefficient [K−1 ] 1.5 × 10−4
Thermal conductivity [W m−1 K−1 ] 64
Emissivity 0.3
Specific heat [J kg−1 K−1 ] 1,000
Electric conductivity [S m−1 ] 1.2 × 106
Melting temperature [K] 1,683
Operating conditions
Droplet diameter [m] 8 × 10−3
Electric current in rf coil [A] 375
Frequency of electric current in rf coil [kHz] 200
Static magnetic flux density [T] 4
Ambient temperature [K] 323
Laser power [W] 9.56
e−2 radius of semiconductor laser beam [m] 2 × 10−3
Spot radius of pyrometer [m] 2 × 10−3
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.
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
Next, the sensitivity analyses of thermal conductivity and emissivity of

molten silicon are carried out, where the effect of variations in these properties
on the relation between φ and ω is investigated numerically. Figures 8.7 and
8.8 show the calculated relations between φ and ω for three different values of
thermal conductivity and emissivity, respectively, where the values in Table 8.1
are varied by ±10% for thermal conductivity and by ±20% for emissivity. In
the calculations for emissivity, the initial temperatures in the droplet before
laser-heating are equalized in all three cases by adjusting the value of the
rf current. The figure shows that it is found that the sensitivity of φ for
thermal conductivity is significant over a wide range of frequencies, although
the sensitivity for emissivity is significant only at a lower frequency region,
that is, ω/2π < 0.05 Hz. Consequently, it is suggested from Figs. 8.7 and 8.8
that the simultaneous determination of thermal conductivity and emissivity
of molten silicon by the present model must be performed from the relation
between φ and ω measured over a wide range of ω; particularly, φ at low ω is

necessary for precise determination of emissivity.
8.2.5 Emissivity Determination from Cooling Curve
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.
The temperature response was measured at the bottom of the sample

through a diaphragm with aperture diameters of 6 mm using a two-color py-
rometer. The signals from the function generator and the pyrometer were
recorded with a sampling interval of 20 ms. A laser apparatus (NBT-S140-mk
II; JENOPTIK Laserdiode Japan Co. Ltd., Tokyo, Japan) equipped with a
fiber-coupling type CW laser diode was used. The maximum output power
was 140 W at a wavelength of 807 ± 3 nm. A calibrated laser power meter
(FieldMate; Coherent Inc., Portland, OR, USA) evaluated the net laser power
through the optical system, which consisted of an optical fiber, a corrective
lens, and a glass window within uncertainty of ±2.4%. The two-color pyrom-
eter (IR-CAQ2CS; Chino Corp., Tokyo, Japan), using the emissivity ratio of
two wavelengths (1,350 and 900 nm) was used. The relative uncertainty of
the pyrometer was ±0.5% of the temperature; the temperature resolution was
1 K. The pyrometer was calibrated using the melting point of the liquid silicon
during solidification of the levitated liquid silicon. The laser power was turned
off after the succession of the modulated laser calorimetry. Then the silicon
liquid was cooled by radiation. The crystalline silicon grains appearing on
the silicon droplet during solidification reflect the flow in the liquid silicon; a
high-speed CCD camera was used to evaluate the effect of the static magnetic
field on the convective flow.
8.4 Experimental Results

8.4.1 Motion of the Silicon Droplet
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.
8.4.2 Temperature Response and Phase Difference
Figure 8.10 depicts an example of the experimental temperature response of

the noncontact modulated laser calorimetry for electromagnetically levitated
liquid silicon. Initially, the silicon droplet temperature was kept at To , with
balancing between the radiative heat loss and rf induction heating. Then,
the laser sinusoidally heated the droplet. The average temperature of the
droplet increased gradually by ΔTdc from To until the radiative heat loss
was balanced with the sum of the rf induction heating and modulated laser
heating; subsequently, the temperature reached an ac steady state. After the
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
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.
8.4.3 Isobaric Molar Heat Capacity
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
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.
8.4.4 Hemispherical Total Emissivity
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.
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
(a) 1805 (b) 0.40
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
Fig. 8.14. (a) An example of change in temperature of silicon droplet following

turning off of the laser power. (b) The hemispherical total emissivity determined
from the radiative cooling curve measured in a static magnetic field at 0.5 T (dia-
monds), 1.0 T (crosses), 2.0 T (squares), 3.0 T (triangles), and 4.0 T (circles). The
solid line in (b) represents the average of this study
Both experimental uncertainties of the emissivity are double the standard

deviation for all present data.
8.4.5 Thermal Conductivity
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
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
We developed a method of measuring the thermophysical properties, heat

capacity, thermal conductivity, and emissivity of high-temperature molten
droplets using the electromagnetic levitation technique. The method was
based on modulated laser calorimetry where a static magnetic field was super-
imposed to suppress the melt convection in an electromagnetically levitated
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
The authors thank Professors T. Hibiya (Keio University), M. Watanabe

(Gakushuin University), H. Yasuda (Osaka University), H. Fecht (University
of Ulm), I. Egry (DLR), R. K. Wunderlich (University of Ulm), and S. Ozawa
(Tokyo Metropolitan University) for their helpful discussions and critical com-
ments. The author (HF) appreciates financial support from the Japan Society
for the Promotion of Science (Grants-in-Aid for Scientific Research), JFE 21st
Century Foundation, and the Iron and Steel Institute of Japan. This study
was subsidized by the Japan Keirin Association through its Promotion funds
from KEIRIN RACE. It was also supported by the Mechanical Social Systems
Foundation and the Ministry of Economy, Trade, and Industry of Japan. This
work was performed at the High Field Laboratory for Superconducting Ma-
terials, Institute for Materials Research, Tohoku University.
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)
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
Takehiko Ishikawa and Paul-François Paradis
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.
9.2 Electrostatic Levitation System

The electrostatic levitation method utilizes the Coulomb force between the
sample and the surrounding electrodes to cancel the gravity force. A posi-
tively charged sample is levitated between a pair of parallel disk electrodes
(top and bottom electrode), typically 10 mm apart, which are utilized to con-
trol the vertical position (z) of the specimen. The typical sample size is 2 mm
in diameter with an electrical field of around 8–15 kV m−1 is necessary to
levitate it against gravity. In addition, four spherical electrodes distributed
around the bottom electrode are used for horizontal control (x and y). Since
the electrostatic scheme can not produce a potential minimum, a feedback
position control system is necessary. Figure 9.1 illustrates the hardware ar-
rangement for the position control. Position sensing is achieved with a set of
orthogonally disposed He–Ne laser (632.8 nm) that projects a sample image
on a position sensor. The beam of a He–Ne laser is expanded and impinges
Top electrode
Band pass filter
Lens
He-Ne laser
Position
Side electrodes Sample detector
Bottom electrode
PC
A/D
High voltage PID calc. Signal

amplifiers conditioner
D/A
Fig. 9.1. Schematic diagram of the sample position control system

9 Thermophysical Property of Metals Using Electrostatic Levitator 175
on a levitated sample. The size of the resulting sample shadow is optimized

with a lens to cover the area of the sensor such that a good dynamic range
is obtained. In addition, a polarization filter is used to optimize the laser in-
tensity reaching the sensor. The sensor is equipped with a band-pass filter at
632.8 nm to eliminate the photon noise coming from sources other than the
laser. The sample position information read by the sensor is then fed into a
computer and analyzed by a program. The program uses PID servo algorithms
for implementing the feedback system, thus allowing a sample to maintain a
fixed position in time. The computer then inputs new values of voltages for
each electrode. The feedback rate is 720 Hz for the z direction and 30 Hz for
the horizontal directions.
Figure 9.2 schematically depicts the electrostatic levitation furnace (ELF)
currently being used at JAXA. It consists of a stainless steel chamber that is
evacuated to a pressure around 10−5 Pa. The chamber houses the above men-
tioned electrode system (a pair of parallel disk electrodes and four spherical
electrodes) in the center.
Sample heating is achieved using two 100 W CO2 lasers emitting at 10.6 μm
and a 500 W Nd:YAG laser emitting at 1.064 μm. The high power Nd:YAG
laser is needed to melt materials with melting temperatures higher than
2,800 K (Mo, Ta, W, Re, and Os).
(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
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
Proper arrangement for heating laser is important to prevent sample move-

ment due to the photon pressures by heating lasers. Currently, the CO2 laser
beams are divided into three beams and separated by 120◦ in a horizontal
plane, while the Nd:YAG laser beam hits directly from the top, through a
hole in the top electrode. Although this quasi-tetrahedral multiple beam con-
figuration improves the sample position stability, it is not satisfactory for the
thermophysical property measurements when the power of the Nd:YAG laser
is very high. This problem is solved by making the top electrode smaller than
the bottom electrode, to strengthen the horizontal restoring forces. Figure 9.3
shows the numerical analysis results of the electric field between the electrodes
when the diameter of the top electrode is diminished. For the large electrode
(left), the surfaces of the iso-electric potential are almost flat whereas those
for the small electrode (right) exhibits a slope that generated a field gradient.
The conical electrical field distribution resulting from this arrangement pro-
vides a horizontal field component and a natural restoring force towards the
center [15].
Sample temperature data are measured using single-color pyrometers (0.90
and 0.96 μm, 120 Hz), equipped with a band stop filter (Rugate notch filter) at
1.064 μm to remove noise coming from the Nd:YAG laser. The levitated sample
is observed by three charged-coupled-device cameras. One camera offers a view
of both the electrodes and the sample. In addition, two black and white high-
resolution cameras, located at right angle from each other and equipped with
telephoto objectives in conjunction with background lamps, provide magnified
views of the sample. This also helps to monitor the sample position in the
horizontal plane and to align the heating laser beams to minimize any photon
induced effects on the sample.
Four coils located below the bottom electrode generate a rotating magnetic
field to control sample rotation.
9.3 Thermophysical Property Measurements
By combining noncontact diagnostics apparatus such as pyrometer or tele-

photo camera, several thermophysical properties can be measured with
the electrostatic levitation furnace. Property measurements by containerless
methods have several advantages compared with conventional methods. First,
samples are free from the risk of contamination from the container and ma-
terials with melting points higher than that of crucibles (e.g., platinum or
alumina) can be processed. Second, since nucleation from the container wall
can be suppressed, molten samples can be maintained in deeply undercooled
condition.
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
Fig. 9.4. Effect of the background on the appearance of a sample at several tem-
peratures
visible on a black background (Fig. 9.4(d)). As it is heated, it becomes more

visible (Figs. 9.4(e), 9.4(f)) but the blooming effect on the camera becomes
more significant as the sample becomes brighter. With the proposed UV tech-
nique, the images keep to exhibit nearly constant contrast from the room
temperature to well over the melting temperature (Figs. 9.4(g)–9.4(i)).
9.3.2 Surface Tension and Viscosity
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
(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
oscillation frequency, ωc , of this signal after correcting for nonuniform surface

charge distribution, the surface tension γ can be found from the following
equation

8γ Q2
2
ωc = 1− {1 − f (γ, q, Y )} , (9.2)
ρr03 64π 2 r03 γε0
where

243.31γ 2 − 63.14q 2 γ + 1.54q 4 Y 2
f (γ, q, e) = (9.3)
176γ 3 − 120q 2 γ 2 + 27γq 4 − 2q 6
and r0 is the radius of the sample when a spherical shape is assumed, ρ is the
liquid density, Q is the drop charge, ε0 is the permittivity of vacuum. The
symbols q and Y are defined by
Q2
q2 = (9.4)
16π 2 r02 ε0
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
levitated sample is always monitored and the sample rotation is controlled

so that the aspect ratio does not exceed 1.02. Moreover, if the aspect ratios
obtained by the image analysis are greater than 1.02, the oscillation data are
discarded.
9.3.3 Experimental Uncertainties
The experimental uncertainty for density measurements is derived from the

respective uncertainty measurements for the mass and volume of samples.
Because the uncertainty in mass is 0.1 mg, while a typical sample mass is
20 mg, the uncertainty can be estimated to be around 0.5%. The uncertainty
of volume (ΔV /V ) can be calculated by
ΔV 3Δr0
= , (9.11)
V r0
where Δr0 is the uncertainty in radius measurement by the image analysis. In
our experiment, the average value of Δr0 is around 1 pixel, while r0 is around
160 pixels. Therefore, ΔV /V can be estimated to be around 1.9%, and the
overall uncertainty of density measurement (Δρ/ρ) is estimated to be around
2%.
According to (9.2), the uncertainty in surface tension measurement is
mainly determined by those of ρ, r0 , and ωc . As described earlier, the un-
certainties of ρ and r0 are 2 and 0.65%, respectively. The uncertainty of ωc
induced by the FFT analysis is negligibly small (0.4%) and evaluated by
considering the transformation error (less than 1 Hz) and the typical charac-
teristic oscillation frequency (around 200 Hz). As a result, the uncertainty of
surface tension measurements (Δγ/γ) can be estimated to be around 3% by
the following equation:

2
2
2
Δγ Δρ 3Δr0 Δωc
≈ + + . (9.12)
γ ρ r0 ωc
The uncertainty in viscosity measurements can be estimated by the uncer-
tainties of ρ, r0 , and τ . The uncertainty of the decay time Δτ is estimated
to be about 15%, which is mainly due to the sample motion with respect to
the detector during drop oscillation. This determines the overall uncertainty
of viscosity.
9.4 Results of Thermophysical Property Measurements

of Refractory Metals
9.4.1 Density
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
Table 9.1. Density of refractory metals
Metal ρ(Tm ) dρ/dT Temperature Reference

Tm (K) (103 kg m−3 ) (kg m−3 K−1 ) (K)
Ti 4.17 −0.22 1,680–2,060 Present work
1943 4.11 1,943 Eljutin [20]
4.15 1,943 Allen [21]
4.11 1,943 Maurakh [22]
4.1 1,943 Eljutin [23]
4.13 1,943 Ivaschenko [24]
4.155 1,943 Seydel [25]
4.13 −0.226 1,993–2,373 Saito [26]
4.208 −0.508 1,650–2,000 Paradis [27]
3.8 1,943 Peterson [28]
V 5,460 −0.49 1,840–2,240 Present work [17]
2183 5,550 2,183 Allen [21]
5734 2,208 Maurakh [22]
6060 −0.32 2,200–2,470 Saito [26]
5300 2,183 Eljutin [29]
5565 2,175–6,600 Seydel [25]
Ni 7.89 −0.65 1,420–1,850 Present work [30]
1728 7.89 −0.67 1,403–1,838 Chung [10]
7.81 −1.08 2,163–2,423 Saito [26]
7.81 −0.87 Saito and Sakuma [31]
7.78 −1.0 Saito and Sakuma [31]
7.77 −1.42 Allen [32]
7.92 −1.01 Lohöfer [2]
Y 4.15 −0.21 1,560–2,100 Present work
1796 4.15 Fogel [33]
Zr 6.21 −0.27 1,850–2,750 Present work [14]
2128 5.80 2,128 Allen [21]
5.60 2,128 Eljutin [29]
6.06 2,108 Maurakh [22]
6.24 −0.29 1,700–2,300 Paradis [6]
5.50 2,125 Peterson [28]
Nb 7.73 −0.39 2,300–3,000 Present work [14]
2742 7.83 2,742 Allen [21]
7.57 2,742 Ivaschenko [34]
7.6 2,742 Eljutin [29]
7.68 −0.54 2,742 Shaner [35]
Mo 9.11 −0.60 2,450–3,000 Present work [36]
2896 9.35 2,896 Allen [21]
9.10 2,896 Eljutin [29]
9.33 2,896 Pekarev [37]
9.10 −0.80 2,896- Seydel [25]
(continued)
Table 9.1. (continued)

Tm (K) (103 kg m−3 ) (kg m−3 K−1 ) (K)
Ru 10.75 −0.56 2,225–2,775 Present work [38]
2607 10.9 2,607 Allen [21]
Rh 10.82 −0.76 1,820–2,250 Present work [39]
2236 11.1 2,236 Allen [21]
10.65 2,236 Eremenko [40]
10.7 −0.90 2,236–2,473 Mitko [37]
10.7 2,236 Popel [41]
12.2 −0.50 2,236–2,473 Dubinin [37]
Pd 10.66 −0.77 1,640–1,875 Present work [42]
1828 10.49 −1.226 1,828–2,073 Lucas [43]
10.7 Eremenko [40]
10.7 Allen [21]
10.379 −1.169 1,828–1,973 Vatolin [44]
10.52 Martsenyuk [45]
Hf 11.82 −0.55 2,300–2,700 Present work [14]
2504 12.0 2,236 Allen [21]
11.1 2,504 Peterson [28]
11.97 2,504 Ivaschenko [24]
11.5 2,504 Arkhikin [46]
Ta 14.75 −0.85 2,650–3,420 Present work [47]
3,290 15.0 3,290 Allen [21]
14.43 −1.3 Shaner [35]
14.6 3,290 Berhault [48]
W 16.43 −1.08 3,125–3,707 Present work
3695 17.5 3,693 Allen [21]
16.37 −0.97 3,693–8,000 Seydel [25]
16.26 3,693 Shaner [35]
16.2 3,693–5,340 Berhault [48]
17.6 3,693 Calverley [50]
Re 18.65 −0.79 2,683–3,710 Present work
3459 18.7 3,459 Allen [21]
18.0 3,459 Thevenin [51]
18.9 3,459 Pekarev [37]
Os 19.1 −1.16 2,670–3380 Present work [52]
3306 20.1 3,306 Allen [21]
19.2 3,306 Vinet [53]
Ir 19.5 −0.85 2,300–3,000 Present work [54]
2,719 20.0 2,719 Allen [21]
19.39 2,719 Martsenyuk [55]
19.23 2,723 Apollova [56]
20.0 2,719 Gathers [57]
(continued)

Tm (K) (103 kg m−3 ) (kg m−3 K−1 ) (K)
Pt 19.22 −0.96 1,691–2,216 Present work [58]
2041 19.7 2,041 Eremenko [37]
18.82 2,073 Kozakevitch [37]
18.91 −2.882 2,041–2,148 Been [37]
18.81 Martsenyuk [37]
19.77 −2.4 2,041–2,473 Dubinin [37]
19.3 −1.7 2,041–5,100 Hixson [59]
19.1 −1.3 2,095–4,500 Gathers [60]
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
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.
9.4.2 Surface Tension
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
Table 9.2. Surface tension of refractory metals
Metal γ(Tm ) dγ/dT Temperature Reference

Tm (K) (10−3 N m−1 ) (10−3 N m−1 K1 ) (K)
Ti 1,557 −0.16 1,750–2,050 Present work [62]
1943 1,510 1,943 Eljutin [20]
1,650 1,943 Alen [32]
1,427 −1.075 1,973–2,323 Maurakh [22]
1,390 1,943 Peterson [28]
1460 1943 Namba and Isobe [63]
1,576 1,953 Tille and Kelly [64]
1,588 1,953 Tille and Kelly [64]
Ni 1,739 −0.22 1,553–1,963 Present work [30]
1728 1,809 −0.39 Popel [37]
1,758 −0.333 1,728–2,473 Mitko [37]
1,778 −0.38 Allen [32]
Y 804 −0.05 1,830–2,070 Present work
1796 872 −0.09 1,796–2,023 Sukhman [65]
610 Fogel [33]
Zr 1,500 −0.11 1,800–2,400 Present work [62]
2128 1,459 −0.24 1,850–2,200 Paradis [6]
1,512 −0.37 Egry [66]
1,480 2,128 Allen [32]
1,400 2,128 Peterson [28]
1,411 2,128 Shunk [67]
1,430 2,128 Kostikov [68]
Nb 1,937 −0.20 2,320–2,915 Present work [62]
2742 1,900 2,742 Allen [32]
1,827 2,742 Flint [69]
1,839 2,742 Ivaschenko [70]
2,040 2,742 Arkhipkin [71]
1,853 2,472 Eremenko [72]
Mo 2,290 −0.26 2,650–3,000 Present work [73]
2896 2,250 2,896 Allen [32]
2,049 2,896 Flint [69]
1,915 2,896 Pekarev [74]
2,080 2,896 Namba [37]
2,130 2,896 Eljutin [37]
2,068 2,896 Man [75]
Ru 2,256 −0.24 2,450–2,725 Present work [38]
2607 2,250 2,607 Allen [32]
2,180 2,607 Martensyuk [76]
Rh 1,940 −0.30 1,860–2,380 Present work [39]
2236 2,000 2,236 Allen [32]
(continued)
Metal γ(Tm ) dγ/dT Temperature Reference

Tm (K) (10−3 N m−1 ) (10−3 N m−1 K1 ) (K)
1,940 2,236 Eremonko [40]
1,915 −0.664 2,236–2,473 Gushchin [77]
Hf 1,614 −0.10 2,220–2,670 Present work [14]
2504 1,630 2,504 Allen [32]
1,460 2,504 Peterson [28]
1,490 2,504 Kostikov [68]
Ta 2,154 −0.21 3,143–3,393 Present work [78]
3290 2,150 3,290 Allen [32]
1,910 3,290 Namba [63]
2,016 3,290 Eremenko [72]
2,360 3,290 Kelly [79]
2,030 3,290 Kelly [79]
W 2,477 −0.31 3,398–3,693 Present work [49]
3695 2,300 3,693 Calverley [50]
2,500 3,693 Allen [32]
2,200 3,693 Pekarev [74]
2,316 3,693 Martsenyuk [80]
2,300 3,693 Agaev [81]
Re 2,710 −0.23 2,903–3,583 Present work [82]
3459 2,700 3,459 Allen [32]
2,610 3,459 Pekarev [74]
Os 2,480 −0.34 3,230–3605 Present work [52]
3306 2,500 3,306 Allen [32]
2,400 3,306 Vinet [53]
Ir 2,230 −0.17 2,373–2,833 Present work [54]
2,719 2250 2,720 Allen [32]
2,264 −0.247 2,720 Apollova [83]
2,140 2720 Martensyuk [84]
Pt 1,800 −0.14 1,743–2,313 Present work [58]
2041 1,869 2,041 Quincke [37]
1,673 2,053 Quincke [37]
1,740 2,041 Eremenko [37]
1,699 2,073 Kozakevitch [37]
1,800 2,041 Allen [32]
1,746 −0.307 2,041–2,148 Dubinin [37]
1,865 2,041 Kingery [65]
1,707 2,043 Martsenyuk [84]
1650
Surface tension (10−3 Nm−1) 1600
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
γmβm (10−7N/mK)
0 500 1000 1500 2000 2500 3000

0
Rb Sb Ga
Y Nd
Sn Present work
Cs Ca
Tl Allen
Ba Bi Hf Lucas
−0.1 Li
Itami
K Sr Zr
Pt
Na
In Pb Ag Zn
Ti
Ir
−0.2 Re
dγ/dT (10−3 N/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
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.
Table 9.3. Viscosity of refractory metals
Metal η(Tm ) η(T ) = η0 exp(E/RT ) Temperature (K) Reference

Tm (10−3 Pa s) η0 E
(K) (10−3 Pa s) (103 J mol−1 )
Ti 4.4 0.033 76.6 1,750–2,050 Present work
1943 2.2 (0.0034) 68 Agaev [88]
5.2 Eljutin [89]
Ni 7.35 0.071 66.6 1,553–1,963 Present work [30]
1728 4.61 0.490 32.2 1,728–2,023 Lucas [90]
4.90 0.1663 50.2 Brandes [88]
4.7 0.265 41.3 Sato [91]
Y 3.6 0.00287 106.5 1,830–2,070 Present work
1796
Zr 4.7 0.76 31.8 1,800–2,300 Present work [14]
2128 4.83 1,850–2,200 Paradis [6]
3.5 2,133 Agaev [92]
5.45 2,138 Elyutin [93]
Nb 4.5 0.55 48.9 2,320–2,915 Present work [14]
2742
Mo 5.6 0.27 73 2,650–3,000 Present work [73]
2896
Ru 6.1 0.60 49.8 2,450–2,725 Present work [38]
2607
Rh 2.9 0.09 64.3 1,860–2,380 Present work [39]
2236 5 2,236 Demidovich [94]
Hf 5.2 0.50 48.7 2,220–2,670 Present work [14]
2504 5.0 2,504 Agaev [88]
Ta 8.6 0.004 213 3,143–3,393 Present work [78]
3290
(continued)
Metal η(Tm ) η(T ) = η0 exp(E/RT ) Temperature (K) Reference

Tm (10−3 Pa s) η0 E
(K) (10−3 Pa s) (103 J mol−1 )
W 6.9 0.11 128 3,398–3,693 Present work [95]
3695
Re 7.9 0.08 133 2,903–3,583 Present work [82]
3459
Os 4.2 0.00167 220 3,230–3,605 Present work [52]
3306
Ir 7.0 1.85 30.0 2,373–2,773 Present work [54]
2719
Pt 4.82 0.25 49.9 1,743–2,313 Present work [58]
2041 6.74 1.53 25.263 2,041–2,273 Zhuchenko [86]
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.
The effect of the feedback control is systematically being checked and

improvement of the viscosity measurement technique is under way.
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)
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)
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)
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)
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)
53. B. Vinet, L. Magnusson, H. Fredriksson, J.P. Desre, J. Colloid Interface Sci.

255, 363 (2002)
54. T. Ishikawa, P.-F. Paradis, S. Yoda, Int. J. Thermophys. 26, 893 (2005)
55. P.S. Martensyuk, Yu.N.I. Ivaschtschenko, Ukr. Chim. Sh. 40, 431 (1974)
56. T.A. Apollova, E.L. Dubinin, M.M. Mitko, A.I. Chegodajev, L.L. Besukland-
nikova, Izv. A.N. SSSR Met. 6, 55 (1982)
57. G.R. Gathers, J.W. Shaner, R.S. Hixson, D.A. Jung, High Temp. High Press.
11, 653 (1979)
58. T. Ishikawa, P.-F. Paradis, N. Koike, Jpn. J. Appl. Phys. 45, 1719 (2006)
59. R.S. Hixson, M.A. Winkler, Int. J. Thermophys. 14, 409 (1993)
60. G.R. Gathers, J.W. Shaner, W.M. Hodgson, High Temp. High Press. 11, 529
(1979)
61. D.J. Steinberg, Metall. Trans. 5, 1341 (1974)
63. S. Namba, T. Isobe, Sci. Papers I. P. C. R. 57, 51 (1963)
64. J. Tille, J.C. Kelly, J. Appl. Phys. 14, 717 (1963)
65. B.J. Keene, Int. Mater. Rev. 38, 157 (1993)
66. I. Egry, A. Diefenbach, W. Dreier, J. Piller, Int. J. Thermophys. 22, 569 (2001)
67. R. Shunk, M. Burr, Trans. AIME 55, 786 (1962)
68. V.I. Kostikov, B.D. Grigorjev, P.G. Arkhipkin, A.D. Agaev, Izv. Vyss. Uch. Sav.
Chern. Met. 3, 25 (1972)
69. O. Flint, J. Nucl. Mater. 16, 260 (1965)
70. Yu.N. Ivaschenko, P.C. Marchenuk, Teplov. Vys. Temp. 11, 1285 (1973)
71. V.I. Arkhipkin, A.D. Agaev, G.A. Grigorev, V.I. Kostikov, Ind. Lab. (USSR)
39, 1340 (1973)
72. V.N. Eremenko, High Temp. 22, 705 (1984)
73. P.-F. Paradis, T. Ishikawa, N. Koike, Int. J. Refract. Met. Hard Mater. 25, 95
(2007)
74. V. Pekarev, Izv. Vyss. Uch. Sav. Tsvetn. Met. 6, 111 (1963)
75. K.F. Man, Int. J. Thermophys. 21, 793 (2000)
76. P.S. Martsenyuk, Yu.N. Ivashchenko, Adgez. Rasp. Paika Mater. 20, 15 (1988)
77. S.G. Gushchin, N.A. Vatolin, E.L. Dubinin, A.I. Timofeev, Ogneupory 12, 49
(1977)
78. P.-F. Paradis, T. Ishikawa, S. Yoda, J. Appl. Phys. 97, 053506 (2005)
79. J.C. Kelley, A. Calverley, SERL-Rep. 80, 53 (1959)
80. P.S. Martsenyuk, Yu.N. Ivashchenko, V.N. Eremenko, Tep. Vys. Temp. 12, 1310
(1974)
81. A.D. Agaev, Dissertation, Moscow Steel and Alloys Institute, 1973
82. T. Ishikawa, P.-F. Paradis, S. Yoda, Appl. Phys. Lett. 85, 5866 (2004)
83. T.A. Apollova, E.L. Dubinin, M.M. Mitko, A.I. Tshegodajev, L.L. Besuklad-
nikova, Izv. A. N. Met. 6, 55 (1982)
84. P.S. Martsenyuk, Inst. Tekh. Probl. Mater. Akad. Nauk SSSR, 51 (1980)
85. A. Kasama, T. Iida, Z. Morita, J. Jpn. Inst. Met. 40, 1030 (1976)
86. T. Itami, in Condensed Matter-Liquid Transition Metals and Alloys Condensed
Matter Disordered Solids, ed. by S.K. Srivastava, N.H. March (World Scientific,
Singapore, 1995), pp. 123–250
87. L.D. Lucas, Tech l’Ing. 7, Form. M67 (1984)
88. L. Battezzatti, L. Greer, Acta. Metall. 37, 1791 (1989)
89. E.A. Brandes, G.B. Brook, in Smithells Metals Reference Book, 7th edn.
(Butterworth-Heinemann, Burlington, 1992), pp. 14–17
90. L.-D. Lucas, C.R. Acad. Sci. 260, 4499 (1969)

91. Y. Sato, K. Sugisawa, D. Aoki, F. Nozoe, T. Yamamura, in Proceedings of the
23rd Japan Symposium on Thermophysical Properties, Tokyo, pp. 27–29 (2002)
92. A.D. Agaev, V.I. Kostikov, V.N. Bobkovski, Izv. Akad. Nauk. SSSR Metall. 43
(1980)
93. V.P. Elyutin, M.A. Maurakh, V.D. Turov, Izv. Vyssh. Ucheb. Zaved. Chern.
Met. 8, 110 (1965)
94. O.V. Demidovich, A.A. Zhuchenko, E.L. Dubinin, N.A. Vatolin, A.I. Timofeev,
Izv. Akad. Nauk. SSSR Metall. 1, 73 (1979)
95. P.-F. Paradis, T. Ishikawa, S. Yoda, J. Appl. Phys. 97, 106101 (2005)
Index
γ-ray transmission method, 28 Arrhenius, 17, 144

As, 85
absorption, 2, 9, 98, 107 atomic attractive force, 188
absorption coefficient, 8, 9, 96, 98–100 atomic configuration, 1, 4
absorptivity, 114, 153 atomic form factor, 9
acoustic levitation, 134 atomic scattering factor, 4
activation energy, 26 Au, 117, 120–122
adiabatic condition, 100 Au–Cu, 142
adsorption, 39, 40, 43, 56 Au-Cu-Co, 3
aerodynamic levitation, 2, 134 AuCu, 3
Ag, 56, 79, 117, 120–122 Avogadro constant, 188
Ag–Au, 56
B, 3
Ag–Cu, 56, 117, 125, 126
Ba–Ge, 3
aircraft, 133
BaB2 O3 , 3
Al, 73, 115
Belton equation, 56
Al2 O3 , 3, 85, 95
Benard–Marangoni–Rayleigh mecha-
Al65 Cu25 Co10 , 3
nism, 41
Al–Cu, 3 Benard–Rayleigh cell, 41
Al–Ni, 3 Bi, 75, 79
AlCl3 , 18 Biot number, 101, 140, 166
alkaline halides, 19 blackbody, 96, 111, 112, 116, 119
alkaline metals, 184, 189 blackbody radiation, 116
alloys, 173 boiling temperature, 184
AlNi, 73 Boltzmann constant, 12
AlNi1.5 –Al, 75 boron nitride, 1, 31
AlNi–Al, 72 boron oxide, 94, 107
alumina, 1, 2, 8, 10, 14, 93, 177 brightness temperature, 112
aluminum smelting, 17 buoyancy, 29, 39, 44, 132, 134, 149
Andrade law, 17 buoyancy convection, 68, 74, 76, 81, 158
anomalous X-ray diffraction, 3 buoyancy force, 61, 150, 170
Ar, 45
Archimedean method, 28, 29 CaAl2 O4 , 3
Arrhenian behavior, 26 CaF2 , 95
198 Index
CaO, 85 decay time, 180

CaO/SiO2 ratio, 96 dendrite, 46
capillarity, 134 density, 17, 28, 36, 86, 111, 135, 137,
capillary method, 18 139, 154, 155, 157, 161, 173, 177,
capillary rise method, 47 180–182
capillary viscometer, 19 density layer, 70
casting, 39, 56, 149 density layering, 62, 63, 72, 80, 81
Ce–Cu, 124, 125 desorption, 54
centrifugal force, 132 diffraction, 1
ceramics, 1, 2, 10 diffusion coefficient, 61, 63, 68, 71, 73,
characteristic oscillation frequency, 179, 75–77, 79, 82, 86
180 dilatometric method, 28
chemical thermodynamics, 53 direct correlation function, 12
Co, 93, 100, 103, 107, 117, 122 directional spectral emissivity, 112
CO2 laser, 5, 175 dislocation, 85
Co80 Pd20 , 3, 138 disordered matter, 1
Co–Pd, 3 drop calorimetry method, 166
coherent intensity, 9 drop oscillation method, 178, 191
cold crucible, 46, 117, 121 drop tower, 77, 133
Compton scattering, 9 drop tube, 133
computer simulations, 2 drop weight method, 47
conduction, 111, 157 dynamic viscosity, 40
CoNi, 3
conical nozzle levitation, 3, 11 eddy current, 2, 117, 158
contact angle, 30 electric conductivity, 135, 137, 138, 161
containerless, 2, 11, 34, 39, 47, 132, 134, electric resistivity, 138, 163
145, 173, 177 electrical resistivity, 173
continuous casting, 85 electromagnetic field, 137, 158, 159
continuous casting powder, 95, 100, 107 electromagnetic force, 2, 150, 158, 163,
convection, 87, 111, 149, 150, 158, 163, 170
169, 170 electromagnetic induction heating, 154
convective flow, 164 electromagnetic levitation, 2, 3, 47, 54,
cooling curve, 120, 163, 165, 167, 177 57, 115, 117, 120, 132, 134–136,
coordination number, 13 142, 145, 158, 169, 173
CoPd, 3 electromagnetic levitator, 150, 159, 170
correction function, 153, 167 electron beam button melting (EBBM),
Coulomb force, 173, 174 46
critical cooling rate, 105 electron beam melting, 39
critical Marangoni number, 41 electronic emission, 8
crystal growth, 39, 41, 56, 149 electrostatic force, 2
Cu, 117, 120, 121, 125 electrostatic levitation, 2, 3, 7, 8, 11, 35,
Cu2 Ce, 125 47, 120, 134, 173, 174, 177, 190,
Cu–Ni, 115, 125 192
CuCo, 133 electrostatic levitator, 4, 5, 14, 139
Cummings and Blackburn, 47, 142 ellipsometer, 122
Curie temperature, 138 ellipsometry, 114, 115, 119, 120
CuZr, 3 EMCZ (electromagnetic Czochralski),
Czochralski, 41 44
Czochralski method, 85, 189 emission, 98, 107
Index 199
emissivity, 8, 101, 111, 150, 154, 155, graphite monochrometor, 9

157, 159, 161–163, 167, 170 gravitational acceleration, 47
EML, 3 gravity, 132, 134
energy dispersive X-ray diffraction, 3 gray body, 98–100
equilibrium P o2 , 52 growth striation, 39, 41
ESA, 68
ESL, 3 Hagen–Poiseuille’s equation, 18
EXAFS, 3 He–Ne laser, 7, 174
extended X-ray absorption fine heat and mass transport, 39, 57
structure, 3 heat balance, 152, 163
extinction coefficient, 114 heat capacity, 111, 120, 149, 150, 153,
157, 167, 173
Fe, 3, 54, 85, 93, 100, 103, 107, 115, 117, heat transfer, 85, 92, 111, 157
122, 123 hemispherical emissive power, 96
Fe–Cr, 56 hemispherical spectral emissivity, 112
Fe–Nb, 124 hemispherical total emissivity, 112, 118,
Fe–Ni, 113 120, 141, 149, 152, 153, 165, 167,
Fe–V, 124 170
fixed-point blackbody, 116 heterogeneous nucleation, 2, 173
float method, 28 Hf, 183, 187, 188, 191
floating zone method, 189 Hg, 185
fluid dynamics, 41 high-energy synchrotron radiation
fluorides, 19 X-ray, 2
Foton, 63, 66 high-energy X-ray, 1
Foton-M2, 62, 77, 82 high-energy X-ray diffraction, 14
foundry, 17 high-temperature liquid, 14, 149
Fourier transform, 13 high-temperature melts, 2, 5, 17
Fourier transformation, 4 hot launch, 8
Fourier’s law, 111 hydrostatic pressure, 33
Fourier–Biot equation, 86
free electron, 105 IML-2, 134, 142
free electron transition, 121 IMPRESS project, 141
free surface, 41 impurity diffusion, 81
Fresnel’s equation, 114 In, 79
fused silica, 1 In2 O3 , 94
IN718, 46
g-jitter, 77, 132, 144 inductive heating, 136
Ga, 79 inelastic X-ray diffraction, 3
GaAs, 85 InSb, 88
GaP, 85 InSn–In, 72
Ge, 104, 107, 117, 173 inter band transition, 121
germanium detector, 4, 9 interdiffusion, 66, 67, 69, 75, 79, 81
Gibbs adsorption isotherm, 50 International Space Station, ISS, 57, 64,
glass, 17 77, 132, 135, 145
glass forming alloy, 2, 105 Ir, 183, 187, 190, 191
glass making, 17 isobaric heat capacity, 151, 165
glass transition, 139, 144 isobaric mass heat capacity, 154
glass-forming metallic alloys, 139 isobaric molar heat capacity, 166,
graphite, 1, 31 170
200 Index
JAMIC drop tower, 143, 145 mathematical model, 157, 170

Japan Aerospace Exploration Agency maximum bubble pressure method, 28,
(JAXA), 64, 174 33, 47
Japanese Experiment Module, JEM, 77 maximum drop method, 47
Jet Propulsion Laboratory (JPL), 173 melting temperature, 161
jet-engine turbine blade, 149 metallic glass, 104, 105, 173
Joule heat, 155 metallic melt, 4, 100, 111, 149
metals, 10, 17
kinematic viscosity, 18 metastable, 132
Kirchhoff’s law, 114, 153 Mg2 SiO4 , 3
Knappwost’s relationship, 21 MgAl2 O4 , 3
MgO, 85, 95
laser flash method, 86–88, 100, 104, 107 microgravity, 39, 41, 52, 61, 62, 76, 77,
laser modulation calorimetry, 150 132–134, 136, 141, 143, 145
laser-pulse method, 86 microsecond pulse heating, 118, 119,
latent heat of fusion, 177 122
Legendre polynomial, 178 Mo, 119, 122, 123, 175, 182, 186, 190,
Levenberg–Marquardt method, 157 191
levitation, 2, 39, 47, 57, 134 modulated laser calorimetry, 159, 164,
levitation force, 135 169
levitation method, 28, 34 modulation amplitude, 152
Lindemann’s theory of melting, 188 modulation calorimetry, 141, 149, 150
liquid metals, 1, 61, 79, 115, 137, 184, modulation frequency, 153
189 molar heat capacity, 166
liquid structural analysis, 3 molecular dynamics (MD) simulation,
liquid structure, 2, 5, 8, 11 13, 14, 79
liquidus temperature, 105 momentum transfer, 4
lithium niobate, 107 monochromator, 116
long capillary, 61, 62, 81 Monte Carlo (MC) simulation, 13
long-range periodic atomic distribution,
MSL-1, 134, 137–139, 141
103
MSL-EML, Materials Science Lab –
Lorentz force, 44, 134, 135, 149, 157
Electro-Magnetic Levitator, 135,
Lorenz number, 138
145
low-Pr-number fluids, 56
mullite, 93
multiple scattering, 2, 9
magnetic field, 44, 62, 76, 135
magnetic flux density, 44
magnetic liquid, 138 Na2 O, 95
magnetic ordering, 138 natural convection, 61
magnetic permeability, 135, 138 Navier–Stokes equation, 131, 142
magnetohydrodynamic (MHD) Nb, 119, 182, 186, 190, 191
convection, 158, 159, 164 Nd glass laser, 88, 93
manometric method, 28, 32 Nd–Fe–B, 3
Marangoni convection, 61, 65, 68, 134, Nd:YAG laser, 175, 176
149, 158 near net shape casting, 131
Marangoni effect, 40 neutron diffraction, 1, 4
Marangoni flow, 39, 41, 44, 52, 53, 57 neutron scattering, 2, 3, 9
mass absorption coefficient, 9 Newton’s equations of motion, 132
mass transport, 40, 68, 82 Newtonian, 18
Index 201
Ni, 3, 93, 100, 103, 107, 115, 117, 119, P, 85

120, 122, 123, 125, 126, 182, 186, P2 O5 , 94
189, 191 packing fraction, 13
Ni–Cr, 115, 124 parabolic flight, 64, 66, 68, 77, 135, 137,
Ni–Cu, 117 139, 143, 145
Ni–Fe, 124 Pb, 73, 79
Ni–V, 3 PbAg5 , 73
Ni-75%Zr, 126 PbAg–Pb, 72, 76
nickel alloy, 119, 127 PbGa5 , 73
nickel based alloy, 125 PbGa–Pb, 72
NiZr, 3 Pd, 119, 183
noble metals, 117, 120 Pd–Cu–Si, 104
non-Newtonian, 18 Pd–Ni–P, 104
noncontact, 139 Pd-based alloy, 104, 105
noncontact modulated laser calorimetry, PdCuSi, 3, 143
164 pendant drop method, 47
noncontact modulation calorimetry, permittivity of vacuum, 179
139, 149, 150 phase difference, 48, 152–154, 157, 159,
nonferrous metallurgy, 17 166, 167, 170
nonlinear optical materials, 95 physical properties, 161
normal spectral emissivity, 112, Planck’s law of radiation, 112, 116
114–116, 118–120, 153 polarization, 9, 114
nuclear reactors, 4 proportional counter, 4
nucleation, 47, 173, 177 Pt, 88, 177, 184, 185, 187, 189, 191
numerical analysis, 176 Pt–Ni–P, 104
numerical modeling, 57, 145 pulse heating, 119
numerical simulation, 64, 149, 150, 158, pycnometer, 31
170 pycnometric method, 28, 31
Nyquist theorem, 142
radial distribution function, 4
optical constants, 114 radiance, 112
optical processes, 113 radiation, 87, 97, 111, 116, 153, 155,
optical pyrometers, 112 163, 164
Os, 175, 183, 187, 190, 191 radiation intensity, 112
oscillating cylinder method, 18 radiative heat transfer, 96, 98
oscillating drop method, 47, 57, 141, Ratto’s model, 53
143 Rayleigh, 47
oscillating method, 18, 26 Rayleigh–Bénard convection, 134
oscillating vessel method, 26 Rayleigh–Bénard instability, 134
oscillating viscometer, 19, 21 Rayleigh–Marangoni–Benard cell, 44
oscillation, 22 Re, 175, 183, 187, 190, 191
oscillation frequency, 47, 48, 137 real body, 111, 119
oxidation, 53 recalescence, 8, 177
oxide melts, 94 reflection, 114
oxides, 17 reflectivity, 99, 114
oxygen, 39, 40, 44, 45, 47, 53 refractive index, 114
oxygen partial pressure, 49 refractory metal, 25, 46, 173, 177, 180,
oxygen sensor, 53 182, 185, 186, 190–192
oxygen transport, 53 relaxation time, 153
202 Index
Reverse Monte Carlo (RMC) simulation, SnBi1.5 , 73

13 SnBi1.5 –Sn, 75
Reynolds number, 18 SnBi–Sn, 72
Rh, 183, 187, 191 SnBiIn–Sn, 72
Roscoe’s equation, 21 SnIn1 –Sn, 75
rotating magnetic field, 176, 177 SnIn3 Bi5 , 73
rotating method, 18, 26 SnIn10 , 73
rotating viscometer, 28 SnIn10 –Sn, 76
rotation, 48 SnIn–Sn, 72
Ru, 182, 186, 190, 191 SnSb3 , 73
ruby laser, 93 SnSb5 –Sn, 78
SnSb–Sn, 72
salts, 17 solidification, 1, 4, 8, 63, 131, 164
sapphire, 1, 93, 101 solubility, 55
sapphire window, 5 solutocapillary, 40
satellite, 62, 77 sounding rocket, 39, 77, 133, 135, 137,
Sb, 79 141, 145
scattering angle, 4 Soyuz-U rocket, 77
scattering intensity, 4 space shuttle, 57, 61, 77, 80
segregation, 63 Spacelab mission, 134, 137, 138, 141,
self-diffusion, 13, 79, 81 142
semiconductor, 10, 56, 85, 149, 173 specific heat, 86, 139, 141, 154, 155, 161
semiconductor laser, 150, 163 spectral emissivity, 173
sensitivity, 170 SPring-8, 4, 7, 8
sensitivity analysis, 159, 163 stainless steel, 45, 54
sessile drop method, 28, 34, 47 standard deviation, 166, 168, 169
shear cell, 62, 64–68, 72, 80, 82 static magnetic field, 49, 149, 150, 158,
shear convection, 64, 65, 68, 71, 72, 82 159, 163, 168–170
shear stress, 40 static structure factor, 4, 9, 11
Si, 2, 3, 8, 10, 13, 14, 39–41, 49, 56, steel, 39, 95
104, 107, 116–118, 120, 124, 125, steelmaking, 17, 85
166–168, 170, 173 Stefan–Boltzmann constant, 101, 120,
Si75 Ge25 , 139 152
Si–Ge, 139 Steinberg’s relation, 185
silicon, 149, 150, 153, 157–159, 161, 163 structural analysis, 2
single crystal, 39, 85, 95 STS-83, 144
single crystalline silicon, 163 STS-94, 144
single-color pyrometer, 5, 176 sulfur, 39, 44–46, 53
SiO, 41 superconducting magnet, 150, 163
SiO2 , 3, 41, 50, 85, 95 supercooled, 118
skin depth, 135, 137, 138 surface free energy, 1
slag, 17 surface oscillation, 142, 149, 178
slits collimation, 4 surface tension, 30, 33, 34, 39–41,
Sn, 65, 69, 73, 79 43, 46, 49, 54, 56, 65, 134, 137,
Sn47 In53 , 69 141–143, 145, 170, 173, 177–181,
Sn90 In10 , 69 185, 186, 188, 189, 192
Sn-In, 79 surfactant, 47, 53
SnBi3 , 71, 73 SUS304, 54
SnBi3 –Sn, 74 synchrotron radiation, 1, 4, 7, 9, 11, 14
Index 203
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

F.H.temperature Measurements of Materials (Springer 2008)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

F.H.temperature Measurements of Materials (Springer 2008)

Uploaded by

Copyright:

Available Formats

advances in materials research 11

advances in materials research

Series Editors: M. Hasegawa A. Inoue N. Kobayashi T. Sakurai L. Wille

The series Advances in Materials Research reports in a systematic and comprehensive

2 Advances 8 Shaped Crystals

10 Frontiers in Materials Research

12 Oxide and Nitride Semiconductors

With 125 Figures

Professor Luc Wille

Advances in Materials Research ISSN 1435-1889

ISBN 978-3-540-85917-8 e-ISBN 978-3-540-85918-5

Library of Congress Control Number: 200893 61 4 5

© Springer Berlin Heidelberg 2009

This book titled High-Temperature Measurements of Materials is the 11th

Sendai, July 2008 Yoshiyuki Kawazoe

A variety of industries – information technology, aerospace, automobile, and

to measure thermophysical properties but also to study the short-range

Sendai Hiroyuki Fukuyama

1 Measurement of Structure of High Temperature

2 Viscosity and Density Measurements

3 Marangoni Flow and Surface Tension

4 Diﬀusion Coeﬃcients of Metallic Melts Measured

4.7 Microgravity Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5 Thermal Diﬀusivity Measurements of Oxide and Metallic

7 Noncontact Thermophysical Property Measurements

7.3 Containerless Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

8 Noncontact Laser Calorimetry of High Temperature Melts

9 Noncontact Thermophysical Property Measurements

9.4.2 Surface Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

Satoshi Awaji Hiroyuki Fukuyama

Rie K Endo Hidekazu Kobatake

Paul-Francois Paradis Takehiko Ishikawa

Tadahiko Masaki Masahiro Susa

Tadahiko Masaki, Akitoshi Mizuno, and Masahito Watanabe

The knowledge of the microscopic feature of matter is of paramount impor-

Levitation Material Structural analysis Ref

scattering factor of typical high-temperature metallic melts. In addition, as

the momentum transfer, Q = 4π sin θ/λ (2θ, scattering angle; λ, wavelength

1.2 Electrostatic Levitator for the Structural Analysis

The design of the present apparatus was based on an electrostatic levitator

The design of the electrodes is of utmost importance for electrostatic levita-

1.4 Results and Discussion

by means of other types of levitators coupled with neutron or X-ray scatter-

12. A Mizuno, S. Matsumura, M. Watanabe, S. Kohara, M. Takata, Mater. Trans.

2.2 Viscosity Measurement

Viscosity of the liquid decreases with an increase of temperature. The temper-

2.2.1 Capillary Method

Principle of the capillary method is simple. The viscosity, η is expressed by

Hole for flow of melt

Lower fiducial mark

Slot for thermocouple

Outlet for melt of

Fig. 2.1. Capillary viscometer made of quartz

Quartz supporting tube

Axis of Gold furnace

Fig. 2.2. Whole apparatus of high temperature capillary viscometer

0.8 0.9 1.0 1.1 1.2 1.3

2.2.2 Oscillating Method

Principle of the oscillating viscometer is based on the internal friction be-

where ω is an angular frequency and B is a constant related to the decay. Four

C = −A[sin{π(0.5 − t1 /τ )} + sin{π(1.5 − t3 /τ )}]/2. (2.11)

A primary constant B is calculated with (2.12) using a series of Ai , which

This set of ﬁrst approximation values, δ and τ includes a considerable er-

generated with given parameters of ﬁxed C of 10 mm, 2L of 100 mm, and