Lixin Cheng Dieter Mewes Advances in Multiphase Flow and Heat Transfer Vol. 1, 2009

Advances in Multiphase Flow and Heat Transfer
Vol. 1, 2009
Lixin Cheng and Dieter Mewes (Editors)
Bentham Science Publishers Ltd.

Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009)
Contents
Preface i
Contributors v
Research and Review Studies vi
Chapter 1 Advances in Void Fraction, Flow Pattern Maps and Non-Boiling

Heat Transfer Two-Phase Flow in Pipes with Various Inclinations 1
Afshin J. Ghajar and Clement C. Tang
Chapter 2 Boiling Heat Transfer in Reduced Gravity Environments 53

Paolo DiMarco, Jungho Kim and Haruhiko Ohta
Chapter 3 Two-Phase Flow Regime Identification Methodologies in

Thermal-Hydraulic Applications 93
J. Enrique Julia, Takashi Hibiki and Mamoru Ishii
Chapter 4 Thermal Radiation Modeling in Multiphase Flows Typical

of Melt-Coolant Interaction 114
Leonid A. Dombrovsky
Chapter 5 Marangoni Motion of Multiphase Compound Drops 158

O.M. Lavrenteva and A. Nir
Chapter 6 Annular Flow Effects on Pore Formation in High-Intensity

Beam Welding or Drilling 213
P. S. Wei and S. C. Kuo
Chapter 7 Ground Heat Transfer from Civil Infrastructure with Seasonal

Freezing and Thawing 233
X. Duan and G. F. Naterer
Chapter 8 Evaluation of Four Nucleate Flow Boiling Models 267

J.P. Kroes, C. W. M. van der Geld and E. van Velthooven
Index 284
Lixin Cheng and Dieter Mewes (Ed)

All right reserved Bentham Science Publishers Ltd.
Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) i
Preface
Multiphase flow and heat transfer have been found a wide range of applications in
nearly all aspects of engineering and science fields such as mechanical engineering,
chemical and petrochemical engineering, nuclear engineering, energy engineering,
material engineering, ocean engineering, mineral engineering, electronics and micro-
electronics engineering, information technology, space technology, micro- and nano-
technologies, bio-medical and life science etc. With the rapid development of various
relevant technologies, the research of multiphase flow and heat transfer is growing very
fast nowadays than ever before. It is highly the time to provide a vehicle to present the
state-of-the-art knowledge and research in this very active field.
To facilitate the exchange and dissemination of original research results and state-
of-the-art reviews pertaining to multiphase flow and heat transfer efficiently, we have
proposed the e-book series entitled Advances in Multiphase Flow and Heat Transfer
to present state-of-the-art reviews/technical research work in all aspects of multiphase
flow and heat transfer fields by inviting renowned scientists and researchers to
contribute chapters in their respective research interests. The e-book series have now
been launched and two volumes have been planned to be published per year since 2009.
The e-books provide a forum specially for publishing these important topics and the
relevant interdisciplinary research topics in fundamental and applied research of
multiphase flow and heat transfer. The topics include multiphase transport phenomena
including gas-liquid, liquid-solid, gas-solid and gas-liquid-solid flows, phase change
processes such as flow boiling, pool boiling, and condensation etc, nuclear thermal
hydraulics, fluidization, mass transfer, bubble and drop dynamics, particle flow
interactions, cavitation phenomena, numerical methods, experimental techniques,
multiphase flow equipment such as multiphase pumps, mixers and separators etc,
combustion processes, environmental protection and pollution control, phase change
materials and their applications, macro-scale and micro-scale transport phenomena,
micro- and nano-fluidics, micro-gravity multiphase flow and heat transfer, energy
engineering, renewable energy, electronic chips cooling, data-centre cooling, fuel cell,
multiphase flow and heat transfer in biological and life engineering and science etc. The
e-book series do not only present advances in conventional research topics but also in
new and interdisciplinary research fields. Thus, frontiers of the interesting research
topics in a wide range of engineering and science areas are timely presented to readers.
In volume 1, there are eight chapters on various topics. Chapter 1 deals with the
advances in void fraction, flow pattern maps and non-boiling heat transfer two-phase
flow in pipes with various inclinations. Void fraction correlations for various pipe
inclinations, both theoretical and empirical, that are widely available in the literature are
compared with experimental data from various sources with different experimental
facilities. The study produced a recommendation of six void fraction correlations and a
proposed improved general void fraction correlation for horizontal to vertical pipe
orientations. Further investigation has also been done specifically on the performances
of available void fraction correlations with available experimental data for upward
vertical two-phase flow. The work demonstrated that more accurate predictions can be

ii Advances in Multiphase Flow and Heat Transfer 1 (2009) Cheng and Mewes
obtained by giving attention to specific pipe inclination and ranges of void fraction. The
result is a recommendation of a void fraction correlation that performs satisfactorily on
the entire range of void fraction for upward vertical flow. Flow pattern maps of different
pipe inclinations and gas-liquid combinations available in the literature are compared
and discussed. The validity and limitations of the numerous two-phase non-boiling heat
transfer correlations that have been published in the literature over the past six decades
are discussed. Practical heat transfer correlations for a variety of gas-liquid flow
patterns and pipe inclination angles are recommended.
Chapter 2 presents a comprehensive boiling heat transfer in reduced gravity
environments. The effects of low gravity on boiling heat transfer are presented. A short
overview of boiling is provided to introduce non-specialists to the field of boiling heat
transfer to make them aware of the fundamental physical mechanisms. The facilities
used to obtain variable gravity environments are then described, and recent results are
reviewed. Topics covered include bubble dynamics, nucleate pool boiling, critical heat
flux, transition boiling, flow boiling, boiling of mixtures, electric field effects, acoustic
effects, quenching. Planned research activities in Japan, the United States, and Europe
are presented.
Chapter 3 deals with the two-phase flow regime identification methodologies in
thermal-hydraulic applications. Two-phase flow regimes have a profound influence on
all the two-phase transport processes. Consequently, their correct identification is a task
of major importance. Two main components are needed in the identification process:
flow regime indicator and classifier. In the first pioneering works, visual flow regime
maps were obtained. In this case, the visual information was the flow regime indicator
and the researcher judgement was used as flow regime classifier. This approach presents
a high level of subjectivity. In the last decades, important work in obtaining more
objective flow regime indicators and classifiers has been done. In this review the current
knowledge about flow regime indicators and classifiers in thermal-hydraulic
applications is summarized. Flow regime indicators comprise different statistical
parameters of void fraction and bubble chord length distributions. Flow regime
classifiers cover different artificial neural network architectures such as self-organized
and probabilistic neural networks. Finally, the main flow regime identification works
performed in different flow channel geometries are reported.
Chapter 4 deals with thermal radiation modeling in multiphase flows of typical of
melt-coolant interaction. The chapter is concerned with radiation heat transfer modeling
in multiphase disperse systems, which are formed in high-temperature melt-coolant
interactions. This problem is important for complex interaction of the core melt with
water in the case of a hypothetical severe accident in light-water nuclear reactors. A
considerable part of thermal radiation emitted by the melt droplets lies in the range of
water semi-transparency. As a result, the radiation is not completely absorbed in water
and one needs to account for radiation heat transfer between the particles which have
different temperatures. The scattering of radiation by steam bubbles and melt droplets
separated from ambient water by a thin steam layer is also important. The problem is
further complicated by semi-transparency of small oxide droplets and temperature
differences between the center and surface of the melt droplets during their
solidification. Nevertheless, the specific radiative properties of the multiphase flow
Preface Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) iii
components allow for a simplified approach, which is implemented in a problem-

oriented CFD code. A more sophisticated approach for visible radiation of the
multiphase media is also presented. The latter is expected to be important for optical
diagnostics of the flow in small-scale experiments including those using various
stimulant melts.
Chapter 5 presents a topic on Marangoni motion of multi-phase compound drops.
Compound drops are comprised of two of more immiscible phases. They occur in many
natural and technological processes and environments, e.g. the atmosphere, liquid
membranes and liquid bi-layers, direct contact heat exchange and phase separation
processes. Compound liquid-liquid drop can model a living eukaryotic cell. Two-phase
compound drops may consist of two touching, partially or completely engulfed parts, or
have more complex geometry. Most processes involving multiphase drops are
accompanied by intensive heat and mass transfer and, thus, a significant influence of
Marangoni flow is anticipated, i.e. fluid motion induced variation of surface tension due
to surfactant or temperature gradient along the interface. In the case of completely
engulfed drop, Marangoni effect induces a relative motion of the inner particle towards
the center of the outer one or in the opposite direction and, thus, stabilizes or destabilizes
the liquid shell. In this chapter we overview the theoretical advances concerning the
motion of multiphase compound drops of various types induced by externally imposed
temperature and concentrations gradient and spontaneous Marangoni motion of such
systems due to heat and mass transfer between the phases including the results of our
ongoing research. Most of the works on the subject are devoted to quasi-stationary
Stokes flow and non-deformable spherical interfaces corresponding to the case of small
Reynolds and capillary numbers. Solutions of the Stokes equations are obtained making
use of special coordinate systems (bi-spherical, toroidal, or tangent spheres), depending
on the compound drop type.
Chapter 6 deals with the annular flow effects on pore formation in high-intensity
beam welding or drilling. This work is to review and study the collapse of the molten
layer surrounding the keyhole filled with vapor and droplets during high intensity beam
welding or drilling. Investigating collapse of the liquid layer is essentially required for
understanding of pore formation in keyhole welding. The keyhole collapse is similar to
transition between the slug and annular two-phase flows. This study thus discusses
mechanisms of pore formation from a viewpoint of vertical annular flows. Porosity is
formed from solidification of the liquid enclosing bubbles or voids, which can be
considered as either gas tapped into liquid or liquid trapped into gas. Penetrative
understanding annular two-phase flows released from this work helps us to know how to
reduce or avoid pore formation in the products in high intensity laser or electron beam
manufacturing and materials processing.
Chapter 7 is concerned with the ground heat transfer from civil infrastructure with
seasonal freezing and thawing. This chapter discusses ground heat transfer from
underground structures, particularly with seasonal freezing and thawing in permafrost
regions. Ground freezing and thawing cycles and permafrost degradation can cause
severe foundation problems for infrastructure in cold regions. Better understanding of
ground heat transfer and the associated frost actions is essential for successful
foundation design and development of protection techniques to overcome these
iv Advances in Multiphase Flow and Heat Transfer 1 (2009) Cheng and Mewes
foundation problems. This chapter briefly reviews mechanisms of water freezing in soil
systems, related frost actions, as well as advances in the modeling of ground heat
transfer and active layer thickness. Foundation design methods for infrastructure in
permafrost are also discussed, with a focus on traditional and latest techniques for
reducing the effects of ground freezing and thawing cycles and for maintaining the
frozen ground as a stable foundation. In a case study, recent research results are
presented of ground heat transfer from a power transmission line foundation.
Chapter 8 presents the evaluation of four flow boiling models. The merits of various
existing 1D models for nucleate boiling heat transfer are assessed with four data sets
from the literature. Several implementation problems and physical inconsistencies of
these models have been identified. Surprisingly, the Chen correlation predicts the
measurements at least as good as more sophisticated, recent models. For heat fluxes in
the range of a diesel engine the Chen correlation even predicts best. Prospects are
discussed.
As the founding editors of the e-book series, we are very happy to see that the e-
books are now available to our readers. We are very much grateful to the authors who
have contributed to the chapters. It is our great wishes if the e-book series are able to
provide useful knowledge for our community and to facilitate the progress of the
research in the field of multiphase flow and heat transfer.
We would like to express our gratitude to our families for their great support to our
work.
Editor-in-Chief: Dr. Lixin Cheng

School of Engineering, University of
Aberdeen, Kings College, Aberdeen, AB24 3UE,
Scotland, the UK,
Email: lixincheng@hotmail.com
Co-editor: Prof. Dieter Mewes

Institute of Multiphase Process, Leibniz
University of Hanover, Callinstrae 36, D-30167
Hannover, Germany,
E-mail: mewes@ifv.uni-hannover.de
20 10 2009
Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) V
Contributors
Paolo DiMarco, Dipartimento di Energetica, Universit di Pisa, Italy
Leonid A. Dombrovsky, Joint Institute for High Temperatures, Moscow, Russia
X. Duan, University of Manitoba, Canada
Afshin J. Ghajar, School of Mechanical and Aerospace Engineering, Oklahoma State

University, USA
Takashi Hibiki, School of Nuclear Engineering, Purdue University, USA
Mamoru Ishii, School of Nuclear Engineering, Purdue University, USA
J. Enrique Julia, Departamento de Ingenieria Mecanica y Construccion. Universitat Jaume I,

Spain
Jungho Kim, Department of Mechanical Engineering, University of Maryland, USA
J.P. Kroes, Department of Mechanical Engineering, Technische Universiteit Eindhoven,, The

Netherlands
S. C. Kuo, Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-

Sen University, TAIWAN, China
O.M. Lavrenteva, Department of Chemical Engineering, TECHNION Israel Institute of

Technology, Israel
G. F. Naterer, University of Ontario Institute of Technology, Canada
A. Nir, Department of Chemical Engineering, TECHNION Israel Institute of Technology,

Israel
Haruhiko Ohta, Department of Aeronautics and Astronautics, Kyushu University, Japan
Clement C. Tang, School of Mechanical and Aerospace Engineering, Oklahoma State

University, USA
C. W. M. van der Geld, Department of Mechanical Engineering, Technische Universiteit

Eindhoven, The Netherlands
E. van Velthooven, DAF Trucks, CAE-Engines, The Netherlands
P. S. Wei, Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-

Sen University, TAIWAN, China

Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) VI
Research and Review Studies

Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 1-52 1
Chapter 1
Advances in Void Fraction, Flow Pattern Maps and Non-

Boiling Heat Transfer Two-Phase Flow in Pipes with Various
Inclinations
Afshin J. Ghajar 1 and Clement C. Tang
School of Mechanical and Aerospace Engineering, Oklahoma State University,
Stillwater, Oklahoma, 74078, USA
Abstract
Void fraction correlations for various pipe inclinations, both theoretical and empirical, that are
widely available in the literature are compared with experimental data from various sources with different
experimental facilities. The study produced a recommendation of six void fraction correlations and a
proposed improved general void fraction correlation for horizontal to vertical pipe orientations. Further
investigation has also been done specifically on the performances of available void fraction correlations
with available experimental data for upward vertical two-phase flow. The work demonstrated that more
accurate predictions can be obtained by giving attention to specific pipe inclination and ranges of void
fraction. The result is a recommendation of a void fraction correlation that performs satisfactorily on the
entire range of void fraction for upward vertical flow. Flow pattern maps of different pipe inclinations and
gas-liquid combinations available in the literature are compared and discussed. The validity and
limitations of the numerous two-phase non-boiling heat transfer correlations that have been published in
the literature over the past six decades are discussed. Practical heat transfer correlations for a variety of
gas-liquid flow patterns and pipe inclination angles are recommended.
Introduction
In many industrial applications, such as the flow of oil and natural gas in flow lines
and wellbores, the knowledge of non-boiling two-phase, two-component (liquid and
permanent gas) heat transfer is required. During the production of two-phase
hydrocarbon fluids from an oil reservoir to the surface, temperature of the hydrocarbon
fluids changes due to the difference in temperatures of the oil reservoir and the surface.
The change in temperature results in heat transfer between the hydrocarbon fluids and
the earth surrounding the oil well, and the ability to estimate the flowing temperature
profile is necessary to address several design problems in petroleum production
engineering [1].
In subsea oil and natural gas production, hydrocarbon fluids may leave the reservoir
with a temperature of 75C and flow in subsea surrounding of 4C [2]. As a result of the
temperature gradient between the reservoir and the surrounding, the knowledge of heat
transfer is critical to prevent gas hydrate and wax deposition blockages [3]. Wax
deposition can result in problems including reduction of inner pipe diameter causing
1
E-mail address: afshin.ghajar@okstate.edu, tel: 405-744-5900, fax: 405-744-7873.

2 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Ghajar and Tang
blockage, increase surface roughness of pipe leading to restricted flow line pressure,
decrease in production, and various mechanical problems [4]. Some examples of the
economical losses caused by the wax deposition blockages include direct cost of
removing the blockage from a subsea pipeline was $5 million, production downtime
loss in 40 days was $25 million [5], and the cost of oil platform abandonment by Lasmo
Company (U.K.) was $100 million [6].
In situations where low velocity flow is necessary while high heat transfer rates are
desirable, heat transfer enhancement schemes such as coil-spring wire insert, twisted
tape insert, and helical ribs are used to promote turbulence thus enhancing heat transfer.
Although these heat transfer enhancement schemes have been proven to be effective,
however they do come with drawbacks, such as fouling, increase in pressure drop, and
sometimes even blockage. Celata et al. [7] presented an alternative approach to enhance
heat transfer in pipe flow, by injecting gas into liquid to promote turbulence. In the
experimental study performed by Celata et al. [7], a uniformly heated vertical pipe was
internally cooled by water, while heat transfer coefficients with and without air injection
were measured. The introduction of small air flow rate into the water flow resulted in
increase of the heat transfer coefficient up to 2040% for forced-convection, and even
larger heat transfer enhancement for mixed-convection [7].
Two-phase flow can also occur in various situations related to ongoing and planned
space operations, and the understanding of heat transfer characteristics is important for
designing piping systems for space operations limited by size constraints [8]. To
investigate heat transfer in two-phase slug and annular flows under reduced gravity
conditions, Fore et al. [8, 9] conducted heat transfer measurements for air-water and air-
50% aqueous glycerin abroad NASAs Zero-G KC-135 aircraft.
In the assessment of non-boiling heat transfer, void fraction and flow pattern are two
critical parameters that need to be taken into consideration. In any gas-liquid system,
void fraction, the volume of space occupied by gas in the system, has great effects on
heat transfer, since the liquid phase generally has significantly larger thermal
conductivity than the gas phase. The knowledge of flow patterns in gas-liquid system is
also critical in the assessment of non-boiling heat transfer, for example heat transfer in
slug flow is likely to be different than heat transfer in annular flow. Void fraction, flow
pattern and heat transfer are also affected by the pipe inclination. In inclined pipes, the
hydrodynamics of two-phase flow are influenced by gravitational force. The influence
of gravitational force affects the slippage between the gas and liquid phases, which in
turn affects the void fraction. Pipe inclination also affects the flow patterns, which is the
reason behind the difference in flow pattern maps between horizontal and vertical flows.
The influence of pipe inclination on the hydrodynamics of two-phase flow also leads to
its effect on heat transfer. Due to the complexity and the lack of fundamental
understanding of void fraction, flow patterns, and non-boiling heat transfer in two-phase
flow, majority of the documented works, whether they are flow pattern maps or
correlations for void fraction and heat transfer, are only applicable to selected flow
conditions. Thus, the content within this chapter addresses the recent advances in void
fraction, flow patterns, and non-boiling heat transfer with the effect of pipe inclination
from a systematic and fundamental point of view.
Advances in Void Fraction, Flow Pattern Maps and Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 3
Void fraction correlations for various pipe inclinations, both theoretical and
empirical, that are widely available in the literature are presented. In addition, measured
void fraction data from various sources and experimental facilities is also collected. The
collected measured void faction data is used to validate the correlations, and a selected
few correlations are recommended based on their predictive performance. In the interest
of a fundamental point of view, the comparison of the measured void fraction data with
the available correlations encompasses a wide range of gas and liquid flow parameters,
flow patterns, and pipe inclinations for different gas-liquid combinations. Flow pattern
maps for gas-liquid flow are also discussed. Flow pattern maps of different pipe
inclinations and gas-liquid combinations available in the literature are compared. Since
the assessment of flow patterns in a given gas-liquid system is somewhat subjective, the
comparison of flow maps from various sources reveals the similarities and
dissimilarities thereof. Lastly, the validity and limitations of the numerous two-phase
non-boiling heat transfer correlations that have been published in the literature over the
past six decades are discussed. The extensive results of the recent developments in the
non-boiling two-phase heat transfer in air-water flow in horizontal and inclined pipes
conducted at Oklahoma State Universitys two-phase flow heat transfer laboratory are
presented. Practical heat transfer correlations for a variety of gas-liquid flow patterns
and pipe inclination angles are recommended. The application of these correlations in
engineering practice, and how they can influence the equipment design and
consequently the process design are discussed.
Void Fraction in Two-Phase Pipe Flow
Comparison of void fraction correlations for different flow patterns and pipe
inclinations
Due to the importance of void fraction in influencing the characteristics of two-
phase flow in pipes, Woldesemayat and Ghajar [10] conducted a very extensive
comparison of 68 void fraction correlations available in the open literature against 2845
experimental data points. The experimental data points were compiled from various
sources with different experimental facilities [11-18]. Out of the 2845 experimental data
points, 900 were for horizontal, 1542 for inclined, and 403 for vertical pipe orientations
(see Table 1). Based on the comparison with experimental data, six void fraction
correlations [19-24] were recommended for acceptably predicting void fraction for
horizontal, upward inclined, and vertical pipe orientations regardless of flow patterns.
The percentage of data points that correctly predicted the 2845 experimental data points
within three error bands for each correlation is summarized in Table 2.
The three more accurate correlations out of the six correlations recommended by
Woldesemayat and Ghajar [10] are developed based on drift flux model. A recent
discussion on the concept of drift flux is given in [25]. Void fraction correlations based
of drift flux model can be expressed generically as
VSG C0 VSG VSL (uGM ) (1)

The gas drift velocity (uGM) represents the local relative velocity between gas and liquid
phase. Both the two-phase distribution coefficient (C0) and the gas drift velocity (uGM)
are determined empirically. In the three more accurate correlations recommended by
Woldesemayat and Ghajar [10], the appropriate expressions for the two-phase
distribution coefficient (C0) and the gas drift velocity (uGM) are listed in Table 3.
Table 1. Summary of experimental database sources, Woldesemayat and Ghajar [10].

Source Pipe diameter [mm] Measurement
Fluids
(no. of data points) (orientation) technique
Eaton [11] Natural

52.5 & 102 (H) Quick-closing valves
(237) gas-water
Beggs [12] 25.4 & 38.1

Air-water Quick-closing valves
(291) (H, U, V)
Spedding & Nguyen [13]

45.5 (H, U, V) Air-water Quick-closing valves
(1383)
Mukherjee [14]
38.1 (H, U, V) Air-kerosene Capacitance probes
(558)
Minami & Brill [15] Air-water & Air-

77.9 (H) Quick-closing valves
(54 & 57) kerosene
Franca & Lahey [16]

19.0 (H) Air-water Quick-closing valves
(81)
Abdul-Majeed [17]
50.8 (H) Air-kerosene Quick-closing valves
(83)
Sujumnong [18]
12.7 (V) Air-water Quick-closing valves
(101)
The pipe orientations are designated with H, U, and V for horizontal, uphill inclined, and vertical, respectively.
Among the six void fraction correlations recommended by Woldesemayat and

Ghajar [10], Dix [20] showed better performance in regards to general overall
comparison with the experimental data points summarized in Table 1. The performance
of the void fraction correlation by Dix [20] is shown in Fig. 1. Woldesemayat and
Ghajar [10] proposed an improved void fraction correlation that gives better predictions
when compared with available experimental data. The correlation proposed by
Woldesemayat and Ghajar [10] was developed based on the drift flux model and takes
on the following expression:
VSG

C0 VSG VSL uGM
(2)
where the two-phase distribution coefficient (C0) and the gas drift velocity (uGM) are
given as
C0
VSG
VSG VSL

1 (VSL / VSG ) ( G / L )
0.1
(2a)
and
0.25
gD (1 cos )( L G )
uGM 2.9(1.22 1.22 sin )
Patm / Psys
(2b)
L2
Note that the leading constant value of 2.9 in Eq. (2b) has a unit such that the drift flux
velocity (uGM) carries the units of meter per second, and Eq. (2) should be used with
parameters conformed to the International System of Units (SI).
Table 2. Number and percentage of data points correctly predicted by the six
recommended void fraction correlations and Eq. (2) for the entire experimental
database summarized in Table 1, Woldesemayat and Ghajar [10].
No. of data points within
Correlation
5% 10% 15%
1065 2137 2427

Morooka et al. [19]
(37.4%) (75.1%) (85.3%)
1597 2139 2363

Dix [20]
(56.1%) (75.2%) (83.1%)
1082 2059 2395

Rouhani & Axelsson [21]
(38.0%) (72.4%) (84.2%)
1244 2003 2322

Hughmark [22]
(43.7%) (70.4%) (81.6%)
1643 2084 2304

Premoli et al. [23]
(57.8%) (73.3%) (81.0%)
1369 1953 2294

Filimonov et al. [24]
(48.1%) (68.6%) (80.6%)
1718 2234 2436

Woldesemayat & Ghajar [10], Eq. (2)
(60.4%) (78.5%) (85.6%)
A total of 2845 experimental data points (see Table 1) were used in this comparison.
The number in ( ) represents the percentage of the data points within the error band.
Table 3. Expressions for two-phase distribution coefficient (C0) and gas drift velocity
(uGM) of different void fraction correlations.
Two-phase distribution coefficient (C0) and gas
Source
drift velocity (uGM)
C0 1.08
Morooka et al. [19]
u GM 0.45
C0
VSG
VSG VSL
1 (VSL / VSG ) ( G / L )
0.1

Dix [20] 0.25
g ( L G )
uGM 2.9
L2
C 0 1 0.2(1 x)
0.25
Rouhani & Axelsson [21] g ( L G )
uGM 1.18(1 x)
L2
1.0 V VH IH
V IH IIIIIIIII
H
IIH
H
VVV V
IIHV IV
H VVH
IIIHVIVIH
H IIIH
IIH
IV
H IH
IIIH
H
H
IIHIH
HII
+15% I I II HIVV IVIV IHIVH V
HIH H
IIV
IV
IHIH
IIHIHIV
H IHH
IVIHIH HIHIHIHIIHI
HI I V IIIII IH I I
IH V HI
I
IIIH
H I
H
I V
I
H
VII
V
H
I
V
H
IH
IIIV
H
I H
IV
IV
H
I
I
H
HIV
IH
HHHIH
V HIH
H
HI I
H HIV
I
I
H H H H I
IIV I VI HIV IV II II VVH II IH
I
I
VH
VV
HIIIV
IH IVVVH
H
IVIH
H
VH
IVIHH
H
I
H IH
I H
V IIVVV H IIIH VI
IVH
H III
VIV
HI
H VHV
VHI IH H
I H
I HH H H
H Horizontaldata I VIIV IIIV IIH IIIIH HIIHIIIH
IIHIIIH
IVIH IV
H
V V
H IIIH
V HIIH
H HH H H H
V IH I H HH
IIHHIHHHH
I I V II I IVIV V IIHIH
IIV VIIH IV H V V
H
IIIH IV
HHIH VIVH
IH
H
H
IHHIVH
IH VIIIIV IH IIH
HIIIV IIHIVH
H H IIV
H
V VVIV HV
H
V H IH H
H
IH H HV
HIH
IHH HHIHHVIIH I I HI
I H H V IH IIIIIIH
IIH IIV
H IH
H V
H
IVIIVH H HHHIIH HH HHH
IV IIH V IH HV HIIH H
I Inclineddata I
V HHIV I I
IIIH
I IIH
I VH
V
H IH
I
IH
IV I
IV
I
V V
H H V
VI IHIH V H H H I H
I H I
IH I
HI VHI HH H
I VH
H I H
I I IH I IV
H V IIIIH
IIVVH
IIVII HIIVIIIIIV
H V HH HI H
H HIH HHH
IIHIHIH
H
IV H
HV I HIHHII H HH
H
I HI I H H
IIIH
H VIH
VV HIIV
VHH HH HHH IIH H I HH
V Verticaldata I
IIIIHH
I I
V VV
I I HV H IHVIIV
I HI I IH H
H
H
V IH
IIH H I H
0.8 I H
I H HI HI I IH
H
IH IV
I VI IIIII VH
HV I
IHII IIH
IIVH
I H
VH
H
IV
H
IIHIH
IIIIIIH VH
V
HHIIH
H V
I I HI IHI H
HH
H
I H
I H
H H HH
H H
I
IIIIIH
H
H I IV IIH
I IIH
VH V II I IH I HI HHH H H
H I II IHH VIHIVIVIIV
IH IV
H
IIH
H VI V HIH I
H IH IH I HVHV
I II II IIIHH
H
IIH
V
HV
IH
V HV
IH HIH H
IHH IH
IIHI V IH H II HH VHI H
I I
H IH II I V I IIVHIV
IIIH
H
VHI
H V
H
H HHV
VIIHHI IHIIHHV
HI H
H H
I I I II H H 15%
IV H I IVIVIVIH
IIVHVI H HIH H
VH H I H HH H
I H I I V IH VH
II H II HIV
I H
V H HH H
IIHH HI IIIH H
I I I H V
I H
H I H HII IVHIV IHHHHV
Calculatedvoidfraction
I I H H I
VHI V HIVV II I HH I I H V H
I HH H
V HI IHII I I V H H
H H I V IH I H
I I II H IIVI IVHH
VI HIIV IH HHIVIIIIIH I H HH
I
I I H HI H H H H
H I I VI I H H
IV HIH
II I
H H H H
H I I I IIHHI HHV I I VV II I IHI H IV H V H
0.6 H I I I II H I IIIIH I I
IIV
V I VII V
I VI
VIIH
VI
HH I II VI I HH
H II I I V VH III H IH II H H HI H H
H I HI IIVH IH
I IV IVHVII V
HHH I VVV H IIHI I I H
IV I HI II H H H H H H
I HI H II I H IIIVIIIH
HH
IIVH H V
III I H
II H VHI IH H HH
II V I H H
I I I V
HI V H H
HV I IIHIIHH I II H I V VV H I I
H V V I IH IVIHIIV I I H I IH H
I I I H
I I H I I
I IV V I H
V VI HH
H
V
III I H IHHIVIIHI IH IHH H I VI I HI
HI IIH
H IIIH HIII II V IHVHI IH I H VI
II IIHI IHII IHIV
V H
HHI H
I H I HIII I VHI H H II H H
I HI H VIHH I IH V
H HIIVH I I I I V I H
I IH H VV I
H H V I I V I H H I
I HV V II H
H I IIIVIIH
IV
H I HI V I HI I H I I H H H
0.4 I
VI HI I
I I I I VII I IIIHHH
III IHII IIH
I H
H HII I HIV
V
I
V H HV H
IH
V I I I I V
VI V HV I I I I I I I
IIVV IIII H IIV I I V IV
I V HI IIV I IIIHIH IIHI VIH H HVH IHI I I I V
I I II VHH III I IIH
V
I I H
IH V H I IIIII IHIIHI V I I H
H H
I V H HHI
H H
I
V
VI
I I I I I VIHHV I IH
I V
V I I I I I IH V H I I I
I V V I
I I IVII I V II I H I I V H
V V H I I
V IIHI I IH I
II HI I H
V II
V I I I I I IIVHIHI II III IIIHVI I I I
V I I II I V II I I VI I I I I I I V H
I VVHI I IVI III IIII I IV III VI IV I HI H V II VV I I
IHI VH I I H
0.2 VV I II IV I I H
II IIHII III IIII I VV I IVI I I HH
H VI I VH H IIHI V V I VI
V
I VI HIIH I I IV II II VV HI H HI H H H
I V I II II IVVIIVIII HH H
I IV H
H
V H H H
V I I
V IV III I V I II IIII IH IHIIVI II I I I I VI I I I I V
H I HV H I H I I I I I
IIIIIIIH
H II I I V V
I IVII IIV I II
V I V V
VI VIIII
II I
I
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Measuredvoidfraction
Fig. 1. Comparison of void fraction correlation by Dix [20] (see Table 3), with 2845 experimental
data points summarized in Table 1, Woldesemayat and Ghajar [10].
The performance of Eq. (2) on the 2845 experimental data points in comparison
with the recommended six void fraction correlations is also summarized in Table 2. As
shown in Table 2, the void fraction correlation, Eq. (2), introduced by Woldesemayat
and Ghajar [10] gives noticeable improvements over the other six correlations. The
results of the comparison for Eq. (2) with the 2845 experimental data points are also
illustrated in Fig. 2. Both Table 2 and Fig. 2 show the capability and robustness of Eq.
(2) to successfully predict void fraction for various pipe sizes, inclinations, and two-
phase fluid mixtures from various sources with different experimental facilities. The
benefit of comparing with experimental data from different facilities is the minimization
of sample bias.
1.0 H H
IIV H
IIIH
H III
IIII
V V
H V
H
IIIIIH
V IV
H IH
IH
V
HVIV
IIIH
H
H IIH
H IH II
I II II IH IVV IV
IH
VHIV
HH V IH
IIIIHHIHV
HIIH
HIV
IV H
IH
IH IIH
IH
V IH
IV
IIH
HH
IV H
IH
H
IH
IH IHII
HI
I
+15%
VI I IH
II VV IIII V
I I
II
I
V HIV
HIV
IH
H
II
IH
H
I
V
II
H
V
H
V
H
IIVH
H
IIIIIH
IH
V
H H H
V
I
I
IIIV
V H
H
V
HH
I
IV
I
V
I
HI
IV
HH
IIV
II
H
HH
IVH
VH
II
H
IIH
V
IHH
H
IV
IIVI
H
V HIH
H
H
V
H
H
V
V
H
I
IIH H
H
H H
I
I
HH
H
H
I
H Horizontaldata I I I
VH V
II I
HIVIH I I
HHH I
V
VV
H
VHH
IIIH H I
V
H IH H H
V I VV I IIHV IVIVII VIIVH
IV
IH IIH
H
IIH V
IH
IV IIH
IH
VH
HH
IV
V
H IIV
V IV
HV
VHH H
V H HHII H
I H IH IHIIIIH
IIIIH
IH
IVH IH IHIIV
H
IH IIVHH
IIV
H
VH VIH IHII
H V
H V
IIHV
V
HIHHH
HH IH
H
IIH
H
H VIH
H
H
I
I IH IH IH
IV
IIIIIHIIIIV H IV IH
H HIIH
H HV
HHVH H H H
H H
H I IH HIH V H IIHHI H
I
H
I Inclineddata H I VHV VI H IIHH
IH HIIH
IH IV IH V H
VIIVIH
IIHIVVIIH
VH
IV
H IVH HHIIIH
H V
HIIIVH II HIIH
H HIHVIHI H
HH HII H IVHI IVH IIH V
H
V V
VI VIV IIHIH
V V
H VH H I H
HH HI V I H
HH I HIH H I V H
I II IH II
H IH VHH IIIIIIIH
IIV
H
I I
VH
VI
IHIVH
H I
IH
I
VH
V
V
HIH H
I IIH H
HH IHHH
I H
H
HH II H H HV I I II H
V Verticaldata H
HI HH
I IHIIV
HIIVHIH IH VIV IIIIH
IIIIV I
IHIVVH
I I IH
VIIIHIIIV
IIHVIH H
I HH IIHHIH HI I
HI HII HII VIH
IH HH
IIIH H
V
IIH I IH HHIHH H
0.8 HI I H HIIIH I IIV
IVI HHII IV V
IIHI IH IH I HIH IH HH
I H IIH IV
I IHV IIHHV IIIH HIHVI HI H
V II H HI H H
IHI IHHII H I H
HV HIH H
V H
V
IV
HHHHV
I I H V
I
H H IH V
H H
I IH I I H III II IIHV VIVH
IH
I HIH
VIHIH
IH
H
IVI V I IH I
V H
V
H HI
I
V H H
H II HIIIH H IIIIIIIIIV IIH
IHV
H H
IV
IH IHIIHI H
VIH HH I H 15%
H HHH II H VVIIVHHIH
VHH
HH VHH V H II HH H
I H H
VH IIVHHH IHI II H
H HH H IHI VHIIV IIIIHH
VIIHV V II I V H HH H
H
I HH H HI H I I H
HVVH IH I I H I HII I V
HI HI HI IIVH
Calculatedvoidfraction
I IHH I H
I H VH
I II I I HH HI
V
V H H III H II H I I
I I I VHIV H
H I IVI I VIV
IHI H
I IV
VIIHIIVIII IV IVH
IIHI IVHH V
HHH H H IH H
IHI HI VH HIH IH
IIV
H
VV I IIIHI I I I H
I H IH H H
0.6 H H IH II II I IIII III III V IH
I I
V VI HI HIH H I V
H I H H IV V
I I V
V
H
I H
I
H
V HHI H
H I V
I HVI VI I IVHHI I I V H H I HI H H H
H I H I IIIH H
IVIHH IH I I IIIH IH I H IVH V I II
H I IIH II IIVIHIIIHVH I I IIHHIIH H H I
I I V VI H
H H I H I I HIV H IVIHIV I IH IHI IIIHVI V I H
I II I I VV I
H I V HHH IVI II V IHIVII V
HH HII HH
H I II IIII I I IV V V
H H I
HI I HVV H I H IH
II H II HII IH
V I I V I I
H V H H
I I H H HV II I H VHH I I I
V I H II I II
I H IHH HIHI IH HI I II HHIHI IVI VII HI
HIH H
H I HH I I
V
HH VIII I
I I I I V III I I
I HV IVIIHIIIH
HHHI HI I I IVHHHH V
H V I IH V
V H I II VI VI I II IV V V
0.4 II II HI II I VIII II V II H
I H
VH V IHIH I II I HH I H I HI VV I H V HI
I
I H VV I IH II H I I
IH I V
V H I I II II I H I I HIHI II HI H I II I V V
V II I
I I V
I I I I
IH HI V I I VII I V I V V
I I VIII II HI IVH II I IIHH I I I HI I
H
I V II II H I II II II III I V I I I V I
I I VV I IIH IV I H V I III VIIVV II I I I
I II I IH I IIV I IV II I V H
V I I HH II I
H
II II II I I
I
H I V I I I
V
V
V I V VI H I I V H
II II I I V V I H
I VVVHI HII H
I IIIH I I I I IIII H I H H
H I I III III IV IV I IV H V VV V V H H
I VI II IHHI V I V I H
H VV I I HV HH IIII I VI IV H VI I I H H
0.2 V V I II HIH II IIVIII IIII VH
III IH I
II I I
H
I I I I VIII VI I V I V I
II II I II VI I I VV I V
V I VH I III I V IVI I I V
H
VII IVHII I I IIIIIHI I I VI
HIVH I
II I I V I V
I IIIIII IVI V
I I I IV
VIIVI
V VIII
I
I
I II
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Measuredvoidfraction
Fig. 2. Comparison of void fraction correlation by Woldesemayat and Ghajar [10], Eq. (2), with
2845 experimental data points summarized in Table 1, Woldesemayat and Ghajar [10].
Comparison of void fraction correlations for vertical pipes

Further scrutiny has also been done specifically on the performances of available
void fraction correlations with available experimental data for upward vertical two-
phase flow. The work demonstrated that more accurate predictions can be obtained by
giving attention to specific pipe inclination and ranges of void fraction. The effort
resulted in the categorization of void fraction correlations recommended for specific
void fraction ranges in upward vertical two-phase flow. Results of these categorical
comparisons would allow the access to correlations with higher accuracies for specific
void fraction range of interest. A database with a total of 1208 experimental data points
was used for the comparison with void fraction correlations available in the literature.
The database encompasses experimental data points for different gas-liquid

combinations and pipe diameters. Table 4 presents a summary of the experimental
database, compiled from ten independent sources, for the comparison with void fraction
correlations.
In total, 52 flow pattern independent void fraction correlations are considered and
compared with data points in the experimental database. Out of the 52 correlations, 11
correlations were considered to be in generally good agreement with the entire
experimental database of 1208 data points. The 11 correlations were selected on the
basis that their predictions, when compared with the experimental data, have more than
75% and 85% of the predicted data points within 15% and 20% error bands,
respectively. The sources of the 11 selected correlations are listed in Table 5, along with
the results of the comparison.
The 11 correlations listed in Table 5 were selected on the basis of overall
performance, which overlooks the strengths and weaknesses in specific ranges of void
fraction. Hence, the subsequent logical approach is to analyze the selected correlations
in ranges, by dividing the entire void fraction range into four ranges: 0 to 0.25, 0.25 to
0.5, 0.5 to 0.75, and 0.75 to 1.0. The qualitative outcome of the 11 correlations and their
performances in each of the four ranges are summarized in Table 6. By comparing the
void fraction correlations with experimental data in each of the four specific ranges, the
correlation by Rouhani & Axelsson [21] was identified as the best correlation for
upward vertical two-phase flow. Figures 3 and 4 show the comparison of the Rouhani &
Axelsson [21] and Nicklin et al. [26] correlations with the entire experimental database
listed in Table 4. Although Table 5 indicates that the correlation by Nicklin et al. [26]
has predicted more data points within the error bands of 15% and 20% than the
correlation by Rouhani & Axelsson [21], it was noted that the Nicklin et al. [26]
correlation performed unsatisfactorily in the 0.75 to 1.0 void fraction range. The
Rouhani & Axelsson [21] correlation, on the other hand, is the only correlation that was
found to perform satisfactorily on each of the four void faction ranges (see Table 6).
Table 4. Summary of experimental database for upward vertical two-phase flow.

Source Pipe diameter [mm] Fluids No. of data points
Present study 12.7 Air-water 153
Schmidt et al. [27] 54.5 Nitrogen-water 20
Sujumnong [18] 12.7 Air-water 104
Sujumnong [18] 12.7 Air-glycerin 77
Chokshi [28] 76.0 Air-water 103
Fernandes [29] 50.7 Air-water 88
Mukherjee [14] 38.1 Air-kerosene 65
Spedding & Nguyen [13] 45.5 Air-water 224
Beggs [12] 25.4 & 38.1 Air-water 27
Oshinowo [30] 25.4 Air-water 153
Oshinowo [30] 25.4 Air-glycerin 172
Isbin et al. [31] 22.2 Steam-water 22

The experimental results of present study are discussed in a subsequent section.
Table 5. Results of the selected 11 correlations that compared satisfactorily with all
1208 experimental data points listed in Table 4.
Percentage of data points predicted within the
Correlation error band of
5% 10% 15% 20%
Nicklin et al. [26] 33.1 62.1 84.4 91.7
Hughmark [22] 33.9 58.2 76.7 86.1
Nishino & Yamazaki [32] 43.2 66.6 78.6 84.7
Guzhov et al. [33] 28.1 54.6 77.6 88.7
Rouhani & Axelsson [21] 39.9 68.5 83.5 89.3
Bonnecaze et al. [34] 33.1 62.1 84.4 91.7
Ishii [35] 37.9 66.6 80.5 87.3
Sun et al. [36] 31.3 58.1 78.1 91.1
Kokal & Stanislav [37] 33.0 61.9 84.4 91.6
Morooka et al. [19] 32.5 62.1 79.1 87.9
Takeuchi et al. [38] 27.6 52.7 78.1 88.6
Table 6. Qualitative performance of 11 selected correlations in four void fraction ranges.

Void fraction range
Correlation
0 to 0.25 0.25 to 0.50 0.50 to 0.75 0.75 to 1.0
Nicklin et al. [26] S S S NS
Hughmark [22] NS NS S S
Nishino & Yamazaki [32] NS NS S S
Guzhov et al. [33] S NS S S
Rouhani & Axelsson [21] S S S S
Bonnecaze et al. [34] S S S NS
Ishii [35] S NS S S
Sun et al. [36] S S S NS
Kokal & Stanislav [37] S S S NS
Morooka et al. [19] S NS NS S
Takeuchi et al. [38] S S NS NS
Note: S = satisfactory & NS = not satisfactory

1.0
+10%
+15%
0.8 +20%
Calculated void fraction
-10%
0.6
-15%
-20%
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Measured void fraction
Fig. 3. Comparison of void fraction correlation by Rouhani & Axelsson [21], with 1208
experimental data points summarized in Table 4.
1.0
+10%
+15%
0.8 +20%
Calculated void fraction
-10%
0.6
-15%
-20%
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Measured void fraction
Fig. 4. Comparison of void fraction correlation by Nicklin et al. [26], with 1208 experimental data
points summarized in Table 4.
Table 7. The experimental data used in Kim et al. [39].

Source Orientation Fluids No. of data points
Vijay [40] Vertical Air-water 139
Vijay [40] Vertical Air-glycerin 57
Rezkallah [41] Vertical Air-silicone 162
Aggour [42] Vertical Helium-water 53
Aggour [42] Vertical Freon 12-water 44
Pletcher [43] Horizontal Air-water 48
King [44] Horizontal Air-water 21
Table 8. List of recommended correlations from the general comparisons with regard to fluid
combinations and major flow patterns, Kim et al. [39].
Source Heat transfer correlations
hTP / hL (1 ) 1/ 3 Laminar
where Nu L 1.615(Re SL PrL D / L)1 / 3 ( B / w ) 0.14
Aggour [42]
hTP / hL (1 ) 0.83 Turbulent
where Nu L 0.0155Re0SL.83 PrL0.5 ( B / w )0.33
0.17
Patm
0.14

Chu & Jones [45] Nu TP 0.43 ReTP
0.55
PrL1 / 3 B
P
w sys
1/ 3
hTP VSG
1
Knott et al. [48] hL VSL
where hL is from Sieder & Tate [49]
1/ 8 0.6 0.14
V G
Kudirka et al. [50] NuTP 125 SG Re1SL/ 4 PrL1 / 3 B
VSL L w
1/ 2
hTP V
1 0.64 SG
Martin & Sims [51] hL VSL
0.3 0.2 0.14
V G
Ravipudi & Godbold [46] NuTP 0.56 SG Re0SL.6 PrL1 / 3 B
VSL L w
hTP
(1 ) 0.9
Rezkallah & Sims [52] hL
1/ 4
hTP VSG
1
hL VSL
Shah [53] where
Nu L 1.86(Re SL PrL D / L)1 / 3 ( B / w )0.14 Laminar
Nu L 0.023Re Pr ( B / w )
0.8
SL
0.4
L
0.14
Turbulent
Note: is taken from the original experimental data. ReSL < 2000 implies laminar flow, otherwise turbulent;
and for Shah [53], replace 2000 by 170. With regard to the eqs. given for Shah [53] above, the laminar two-
phase correlation was used along with the appropriate single phase correlation, since Shah [53]
recommended a graphical turbulent two-phase correlation.
Table 9. Results of recommended correlations with regard to fluid combinations and major flow
patterns, Kim et al. [39].
Vertical pipes
Correlations Air-water Air-glycerin Air-silicone
B S F A B S F A B S C A F
Aggour [42]
Chu & Jones [45]
Knott et al. [48]
Kudirka et al. [50]
Martin & Sims [51]
Ravipudi & Godbold [46]
Rezkallah & Sims [52]
Shah [53]
Vertical pipes Horizontal pipes
Correlations Helium-water Freon 12-water Air-water
B S F A B S F A A S
Aggour [42]
Chu & Jones [45]
Knott et al. [48]
Kudirka et al. [50]
Martin & Sims [51]
Ravipudi & Godbold [46]
Rezkallah & Sims [52]
Shah [53]
Note: = Recommended correlation with and without restrictions. Shaded cells indicate the
correlations that best satisfied the 30% two-phase heat transfer coefficient criterion.
A = annular, B = bubbly, C = churn, F = froth, S = slug.
Comparison of Twenty Two-Phase Heat Transfer Correlations with

Seven Sets of Experimental Data
Numerous heat transfer correlations and experimental data for forced convective
heat transfer during gas-liquid two-phase flow in vertical and horizontal pipes have been
published over roughly the past six decades. In a study published by Kim et al. [39], a
comprehensive literature search was carried out and a total of 38 two-phase flow heat
transfer correlations were identified. The validity of these correlations and their ranges
of applicability have been documented by the original authors. In most cases, the
identified heat transfer correlations were based on a small set of experimental data with
a limited range of variables and gas-liquid combinations. In order to assess the validity
of those correlations, they were compared against seven extensive sets of two-phase
flow heat transfer experimental data available from the literature, for vertical and
horizontal tubes and different flow patterns and fluid combinations. For consistency, the
validity of the identified heat transfer correlations were based on the comparison
between the predicted and experimental two-phase heat transfer coefficients meeting the
30% criterion. Out of the 38 two-phase flow heat transfer correlations, Kim et al. [39]
identified 20 correlations for comparison with experimental data from the various
sources. Eighteen of the 38 two-phase flow heat transfer correlations were not tested,
since the required information for those correlations was not available through the
identified experimental studies.
A total of 524 data points from the five available experimental studies [40-44] were
used for these comparisons (see Table 7). The experimental data included five different
gas-liquid combinations (air-water, air-glycerin, air-silicone, helium-water, Freon 12-
water), and covered a wide range of variables, including liquid and gas flow rates and
properties, flow patterns, pipe sizes, and pipe inclination. Five of these experimental
data sets are concerned with a wide variety of flow patterns in vertical pipes and the
other two data sets are for limited flow patterns (slug and annular) within horizontal
pipes. In assessing the ability of the 20 identified heat transfer correlations, their
predictions were compared with the experimental data from the sources listed in Table
7, both with and without considering the restrictions on superficial liquid Reynolds
number (ReSL) and superficial gas velocity to superficial liquid velocity ratio (VSG/VSL)
accompanying the correlations. The two-phase flow heat transfer correlations
recommended by Kim et al. [39] from the general comparisons with regard to fluid
combinations and major flow patterns are listed in Table 8. The results from comparing
the 20 two-phase flow heat transfer correlations and the experimental data are several
correlations for major flow patterns recommended by Kim et al. [39], which are
summarized in Table 9.
There were no remarkable differences for the recommendations of the heat transfer
correlations based on the results with and without the restrictions on ReSL and VSG/VSL,
except for the correlations of Chu and Jones [45] and Ravipudi and Godbold [46], as
applied to the air-water experimental data of Vijay [40]. Details of this discussion can
be found in Kim et al. [39]. Based on the results without the authors restrictions on
ReSL and VSG/VSL, the correlation of Chu and Jones [45] was recommended for only
annular, bubbly-froth, slug-annular, and froth-annular flow patterns of air-water in
vertical pipes. While the correlation of Ravipudi and Godbold [46] was recommended
for only annular, slug-annular, and froth-annular flow patterns of air-water in vertical
pipes. However, when considering the ReSL and VSG/VSL restrictions by the authors, the
correlation of Chu and Jones [45] was recommended for all vertical pipe air-water flow
patterns including transitional flow patterns except the annular-mist flow pattern. While
the correlation of Ravipudi and Godbold [46] was recommended for slug, froth, and
annular flow patterns and for all of the transitional flow patterns of the vertical air-water
experimental data.
All of the above recommended correlations have the following important parameters
in common: ReSL, PrL, B/w and either void fraction () or superficial velocity ratio
(VSG/VSL). It appears that void fraction and superficial velocity ratio, although not
directly related, may serve the same function in two-phase flow heat transfer
correlations. From the comprehensive literature search, Kim et al. [39] found that there
is no single correlation capable of predicting the flow for all fluid combinations in
vertical pipes. In the following section, the effort of Kim et al. [47] in developing a heat
transfer correlation that is robust enough to span all or most of the fluid combinations
and flow patterns for vertical pipes is highlighted. Kim et al. [47] developed a
correlation that is capable of predicting heat transfer coefficient in two-phase flow
regardless of fluid combinations and flow patterns. The correlation uses a carefully
derived heat transfer model which takes into account the appropriate contributions of
both the liquid and gas phases using the respective cross-sectional areas occupied by the
two phases.
Development of a Heat Transfer Correlation for Vertical Pipes
In this section, the effort of Kim et al. [47] in developing a heat transfer correlation
that is robust enough to span all or most of the fluid combinations and flow patterns for
vertical pipes is highlighted. The void fraction () is defined as the ratio of the gas-flow
cross-sectional area (AG) to the total cross-sectional area, A (= AG + AL),
AG
(3)
AG AL
The actual gas velocity (VG) can be calculated from
QG m G m x
VG (4)
AG G AG GA
Similarly, the actual liquid velocity (VL) is defined as
QL m L m (1 x)
VL (5)
AL L AL L (1 ) A
The total gas-liquid two-phase heat transfer coefficient is assumed to be the sum of
the individual single-phase heat transfer coefficients of the gas and liquid phases,
weighted by the volume of each phase
hG
hTP (1 ) hL hG (1 )hL 1 (6)
1 hL
Based upon this correlation, the single-phase heat transfer coefficients in Eq. (6), hL and
hG, can be modeled as functions of Reynolds number, Prandtl number and the ratio of
bulk to wall viscosities. Thus, Eq. (6) can be expressed as
f (Re, Pr, B w ) G
hTP (1 )hL 1 (7)
1 f (Re, Pr, B w ) L
or
Re Pr ( / )
hTP (1 )hL 1 f G , G , B w G (8)
1 Re L PrL ( B / w ) L
Substituting the definition of Reynolds number (Re = VD/B) for the gas (ReG) and
liquid (ReL) yields
hTP ( VD / B ) G PrG ( B / w ) G
1 f , , (9)
(1 )hL 1 ( VD / B ) L PrL ( B / w ) L
Rearranging yields
hTP V D Pr ( )
1 f G G G , G, w L (10)
(1 )hL 1 LVL DL PrL ( w ) G
where the assumption has been made that the viscosity ratio in the Reynolds number
term of Eq. (9) is exactly cancelled by the last term in Eq. (9), which includes the same
viscosity ratio. Substituting Eq. (3) for the ratio of gas-to-liquid diameters (DG/DL) in
Eq. (10) and based upon practical considerations assuming that the ratio of liquid-to-gas
viscosities evaluated at the wall temperature is comparable to the ratio of those
viscosities evaluated at the bulk temperature (L/G)B, Eq. (10) reduces to
hTP V PrG L
1 f G G , , (11)
(1 )hL 1 LVL 1 PrL G
Further simplifying Eq. (11), combine Eqs. (4) and (5) for gas velocity (VG) and
liquid velocity (VL) to get the ratio of VG/VL and substitute into Eq. (11) to get
x PrG L
hTP (1 ) hL 1 f , , , (12)
1 x 1 PrL G
Assuming that two-phase heat transfer coefficient can be expressed using a power-law
relationship on the individual parameters that appear in Eq. (12), then it can be
expressed as
x m n Pr
p
G
q

hTP (1 )hL 1 C G (13)
1 x 1 PrL L
where C, m , n, p, and q are adjustable constants, and hL comes from the Sieder and Tate
[49] correlation for turbulent flow,
0.14
k
hL 0.027 Re 45
Pr L B
13
(14)
D W
L L
L
Although there are several well-known single-phase heat transfer correlations in the
literature. In this study the Sieder and Tate [49] equation was chosen as the fundamental
single-phase heat transfer correlation because of its practical simplicity and proven
applicability. Any other well-known single-phase turbulent heat transfer correlation
could have been used in place of the Sieder and Tate [49] correlation. The difference
resulting from the use of a different single-phase heat transfer correlation will be
absorbed during the determination of the values of the leading coefficient and exponents
on the different parameters in Eq. (13). For the Reynolds number (ReL) in Eq. (14), the
following relationship is used to evaluate the in situ Reynolds number (liquid phase)
rather than the superficial Reynolds number (ReSL) as commonly used in the
correlations available in the literature [39]:
VD 4m L
Re L (15)
L 1 L D
The values of the void fraction () used in Eq. (13) were either taken directly from the
original experimental data sets (if available) or were calculated based on the equation
provided by Chisholm [54], which can be expressed as
1

12
1 x G
1 1 x x L (16)
G x L
In the next section the proposed heat transfer correlation, Eq. (13), will be tested with
four extensive sets of vertical two-phase flow heat transfer data available from the
literature (see Table 7).
Heat Transfer Correlation for Gas-Liquid Flow in Vertical Pipes

Four sets of experimental data (see the first column in Table 10) for vertical pipe
flow were used to validate the two-phase flow heat transfer correlation, Eq. (13),
developed in previous section. The ranges of these four sets of experimental data can be
found in Kim et al. [39]. The experimental data (a total of 255 data points) included
four different gas-liquid combinations (air-water, air-silicone, helium-water, Freon 12-
water), and covered a wide range of variables, including liquid and gas flow rates,
properties, and flow patterns. The selected experimental data were only for turbulent
two-phase heat transfer data in which the superficial Reynolds numbers of the liquid
(ReSL) were all greater than 4000. Table 10 and Fig. 5 provide the details of the
correlation and how well the proposed correlation predicted the experimental data.The

two-phase heat transfer correlation, Eq. (13), predicted the heat transfer coefficients of
255 experimental data points for vertical flow with an overall mean deviation of about
2.5% and a root-mean-square deviation of about 12.8%. About 83% of the data (212
data points) were predicted with less than 15% deviation, and about 96% of the data
(245 data points) were predicted with less than 30% deviation. The results clearly
show that the proposed heat transfer correlation is robust and can be applied to turbulent
gas-liquid flow in vertical pipes with different flow patterns and fluid combinations.
Table 10. Results of the predictions for available two-phase heat transfer experimental data in
vertical pipes using Eq. (13), Kim et al. [47].
Values of constant and No. of
RMS Mean Range of parameters
Fluids exponents data
dev. dev.
(ReSL > 4000) within
C m n p q (%) (%) ReSL ReSG PrG / PrL G / L
30%
All
12.78 2.54 245
255 data points
3.64 103 to 23.7 103

9.99 103 to 137 103
Air-water [40]
12.98 3.53 98
105 data points
4000 to 127000
14 to 209000
Air-silicone
[41] 7.77 5.25 56
0.27 0.04 1.21 0.66 0.72
56 data points
Helium-water
[42] 15.68 1.66 48
50 data points
Freon 12-water
[42] 13.74 1.51 43
44 data points
Only data points for ReSL > 4000 are considered. Data of Vijay [40], see Table 7, for air-glycerin did not
satisfy the ReSL > 4000 condition.
A General Two-Phase Heat Transfer Correlation for Various Flow

Patterns and Pipe Inclinations
The heat transfer correlation developed by Kim et al. [47], Eq. (13), was meant for
predicting heat transfer rate in two-phase flow in vertical pipes. In order to handle the
effects of various flow patterns and inclination angles on the two-phase heat transfer
data with only one correlation, Ghajar and Kim [55] and Kim and Ghajar [56]
introduced the flow pattern factor (FP) and the inclination factor (I). The void fraction
(), which is the volume fraction of the gas-phase in the tube cross-sectional area, does
not reflect the actual wetted-perimeter (SL) in the tube with respect to the corresponding
flow pattern. For instance, the void fraction and the non-dimensionalized wetted-
perimeter of annular flow both approach unity, but in the case of plug flow the void
fraction is near zero and the wetted-perimeter is near unity. However, the estimation of
the actual wetted-perimeter is very difficult due to the continuous interaction of the two
phases in the tube. Therefore, instead of estimating the actual wetted-perimeter,
modeling the effective wetted-perimeter is a more practical approach. In their model,
Ghajar and his co-workers have ignored the influence of the surface tension and the
contact angle of each phase on the effective wetted-perimeter. The wetted-perimeter at

the equilibrium state, which can be calculated from the void fraction, is
2
~ S L ,eq
S L2,eq 1 (17)
D
50000
Air-water [40]
Air-silicone [41]
Helium-water [42]
20000 Freon 12-water [42]
(W/m K)
10000
2
+30 %
CAL
5000
hTP
2000
-30 %
1000
1000 2000 5000 10000 20000 50000
2
hTP (W/m K)
EXP
Fig. 5. Comparison of the predictions by Eq. (13) with the experimental data for vertical flow (255
data points), Kim et al. [47].
However, as shown in Fig. 6, the shape of the gas-liquid interface at the equilibrium
state based on the void fraction () is far different from the one for the realistic case.
The two-phase heat transfer correlation, Eq. (13), weighted by the void fraction (1),
is not capable of distinguishing the differences between different flow patterns.
Therefore, in order to capture the realistic shape of the gas-liquid interface, the flow
pattern factor (FP), an effective wetted-perimeter relation, which is a modified version
of the equilibrium wetted-perimeter, Eq. (17), is proposed,
2
~ S L ,eff
FP S L2,eff (1 ) FS2 (18)

D
For simplicity, the equation for the effective wetted-perimeter relation, Eq. (18), is
referred to as the flow pattern factor (FP). The term (FS) appearing in Eq. (18) is
referred to as shape factor which in essence is a modified and normalized Froude
number. The shape factor (FS) is defined as
V V 2
FS
2
tan 1 G G L (19)
g D L G

The shape factor (FS) is applicable for slip ratio greater than or equal to one (K 1),
which is common in gas-liquid flow, and represents the shape changes of the gas-liquid
interface by the force acting on the interface due to the relative momentum and
gravitational forces.
Fig. 6. Gas-liquid interfaces and wetted-perimeters.
Due to the density difference between gas and liquid, the liquid phase is much more
affected by the orientation of pipe (inclination). The detailed discussion of the
inclination effect on the two-phase heat transfer is available in Ghajar and Tang [57]. In
order to account for the effect of inclination, Ghajar and Kim [55] proposed the
inclination factor
g D L G sin
I 1 (20)
L VSL2
The expression for the inclination factor (I), Eq. (20), includes representation of the
relative force acting on the liquid phase in the flow direction due to the momentum and
the buoyancy forces.
The two proposed factors for the flow pattern (FP) and pipe inclination (I) effects
are introduced into the two-phase heat transfer correlation for vertical pipe, Eq. (13). By
substituting (FP) for (1) which is the leading coefficient of (hL) and introducing (I) as
an additional power-law term in Eq. (13), the two-phase heat transfer correlation
becomes
x m 1 F
n
PrG
p
L
q

hTP FP hL 1 C P
I r (21)

1 x P
F PrL G
where (hL) comes from the Sieder and Tate [49] correlation for turbulent flow, see Eq.
(14). For the Reynolds number needed in the (hL) calculation, Eq. (15) presented and
discussed earlier was used. The values of the void fraction () used in Equations (15),
(18), and (21) were calculated based on the correlation provided by Woldesemayat and
Ghajar [10], which was given by Eq. (2).
Other void fraction correlations could also be used in place of the Woldesemayat
and Ghajar [10] correlation. Tang and Ghajar [58] showed that Eq. (21) has the
robustness that it can be applied with different void fraction correlations. The difference
resulting from the use of different correlations will be absorbed during the
determination of the values of the constant and exponents of Eq. (21). The two-phase
heat transfer correlation, Eq. (21) was validated with a total of 763 experimental data
points for different flow patterns and inclination angles [56, 58, 59]. Overall, the
correlation, Eq. (21), has successfully predicted over 85% of the experimental data
points to within 30% for 0, 2, 5, and 7 pipe orientations.
However, upon revisiting the two-phase heat transfer correlation, Eq. (21), along
with the equations for flow pattern factor (FP), Eq. (18), and inclination factor (I), Eq.
(20), it was realized that the correlation has not accounted for the surface tension force.
Since surface tension is a variable that can affect the hydrodynamics of gas-liquid two-
phase flow, it is sensible to include the surface tension into the correlation. In order to
do that, the equation for the inclination factor (I), Eq. (20), is modified. The modified
inclination factor takes on the following form,
I 1 Eo sin (22)
where the Etvs number (Eo) is defined as
Eo
L G gD 2
(23)

The Etvs number (Eo), also known as Bond number (Bo), represents the
hydrodynamic interaction of buoyancy and surface tension forces that occur in two-
phase flow. With the modification of the equation for the inclination factor, two-phase
heat transfer coefficients can be estimated using the general two-phase heat transfer
correlation, Eq. (21), along with the flow pattern factor (FP), Eq. (18), and modified
inclination factor (I *), Eq. (23),
n
x 1 FP PrG
m

p
L r
q
hTP FP hL 1 C ( I ) (24)
1 x FP PrL G
The proper values of the constant and exponents are discussed in a later section.
Experimental Setup
Experimental setup for horizontal and slightly upward inclined flow
A schematic diagram of the overall experimental setup for heat transfer
measurements is shown in Fig. 7. The test section is a straight standard stainless steel
schedule 10S pipe with an inner diameter of 27.9 mm and a length to diameter ratio of
95. The setup rests atop a 9-m long aluminum I-beam that is supported by a pivoting
foot and a stationary foot that incorporates a small electric screw jack. In order to apply
uniform wall heat flux boundary condition to the test section, copper plates were silver
soldered to the inlet and exit of the test section. The uniform wall heat flux boundary
condition was maintained by a Lincoln SA-750 welder for superficial liquid Reynolds
number greater than 2000 and a Miller Maxtron 450 DC welder for superficial liquid
Reynolds number less than 2000. The Lincoln SA-750 welder has the capability of
supplying 300 to 750 A of current, while the Miller Maxtron 450 DC welder is capable
of supplying 5 to 450 A of current. The entire length of the test section was wrapped
using fiberglass pipe wrap insulation, followed by a thin polymer vapor seal to prevent
moisture penetration. The calming section (clear polycarbonate pipe with inner diameter
of 25.4 mm and length to diameter ratio of 88) served as a flow developing and
turbulence reduction device, and flow pattern observation section. One end of the
calming section is connected to the test section with an acrylic flange and the other end
of the calming section is connected to the gas-liquid mixer. For the horizontal flow
measurements, the test section, and the observation section (refer to Fig. 7) were
carefully leveled to eliminate the effect of inclination on these measurements.
T-type thermocouple wires were cemented with Omegabond 101, an epoxy adhesive
with high thermal conductivity and electrical resistivity, to the outside wall of the
stainless steel test section as shown in Fig. 8. Thermocouples were placed on the outer
surface of the pipe wall at uniform intervals of 254 mm from the entrance to the exit of
the test section. There were 10 thermocouple stations in the test section (refer to Fig. 8).
All the thermocouples were monitored with a National Instruments data acquisition
system. The average system stabilization time period was from 30 to 60 minutes after
the system attained steady state. The inlet liquid and gas temperatures and the exit bulk
temperature were measured by Omega TMQSS-125U-6 thermocouple probes.
Calibration of thermocouples and thermocouple probes showed that they were accurate
within 0.5C. The operating pressures inside the experimental setup were monitored
with a pressure transducer. To ensure a uniform fluid bulk temperature at the inlet and
exit of the test section, a mixing well of alternating polypropylene baffle type static
mixer for both gas and liquid phases was utilized.
The fluids used in the test loop are air and water. The water is distilled and stored in
a 55-gallon cylindrical polyethylene tank. A Bell & Gosset series 1535 coupled
centrifugal pump was used to pump the water through an Aqua-Pure AP12T water
filter. An ITT Standard model BCF 4063 one shell and two-tube pass heat exchanger
removes the pump heat and the heat added during the test to maintain a constant inlet
water temperature. From the heat exchanger, the water passes through a Micro Motion
Coriolis flow meter (model CMF100) connected to a digital Field-Mount Transmitter
(model RFT9739) that conditions the flow information for the data acquisition system.
From the Coriolis flow meter it then flows into the test section. Air is supplied via an
Ingersoll-Rand T30 (model 2545) industrial air compressor. The air passes through a
copper coil submerged in a vessel of water to lower the temperature of the air to room
temperature. The air is then filtered and condensation removed in a coalescing filter.
The air flow is measured by a Micro Motion Coriolis flow meter (model CMF025)
connected to a digital Field-Mount Transmitter (model RFT9739) and regulated by a
needle valve. Air is delivered to the test section by flexible tubing. The water and air
mixture is returned to the reservoir where it is separated and the water recycled.
Fig. 7. Schematic of experimental setup.

Fig. 8. Schematic of heated test section illustrating the placement of thermocouples.
The heat transfer measurements at uniform wall heat flux boundary condition were
carried out by measuring the local outside wall temperatures at 10 stations along the
axis of the pipe and the inlet and outlet bulk temperatures in addition to other
measurements such as the flow rates of gas and liquid, room temperature, voltage drop
across the test section, and current carried by the test section. A National Instruments
data acquisition system was used to record and store the data measured during these
experiments. The computer interface used to record the data is a LabVIEW Virtual
Instrument (VI) program written for this specific application. The peripheral heat
transfer coefficient (local average) was calculated based on the knowledge of the pipe
inside wall surface temperature and inside wall heat flux obtained from a data reduction
program developed exclusively for this type of experiments [60]. The local average
peripheral values for inside wall temperature, inside wall heat flux, and heat transfer
coefficient were then obtained by averaging all the appropriate individual local
peripheral values at each axial location. The variation in the circumferential wall
temperature distribution, which is typical for two-phase gas-liquid flow in horizontal
pipes, leads to different heat transfer coefficients depending on which circumferential
wall temperature was selected for the calculations. In two-phase heat transfer
experiments, in order to overcome the unbalanced circumferential heat transfer
coefficients and to get a representative heat transfer coefficient for a test run, the
following equation was used to calculate an overall two-phase heat transfer coefficient
( hTPEXP ) for each test run:
q
N ST N ST

1 1 1
hTPEXP h dz hk z k z k (25)
L L k 1
L k 1 Tw TB k
where L is the length of the test section, h , q , Tw , TB are the local mean heat transfer
coefficient, the local mean heat flux, the local mean wall temperature, and the bulk
temperature at a thermocouple station, respectively; k is the index of the thermocouple
stations, NST is the number of the thermocouple stations, z is the axial coordinate, and z
is the element length of each thermocouple station. The data reduction program used a
finite-difference formulation to determine the inside wall temperature and the inside
wall heat flux from measurements of the outside wall temperature, the heat generation
within the pipe wall, and the thermophysical properties of the pipe material (electrical
resistivity and thermal conductivity).
The reliability of the flow circulation system and of the experimental procedures
was checked by making several single-phase calibration runs with distilled water. The
single-phase heat transfer experimental data were checked against the well established
single-phase heat transfer correlations [61] in the Reynolds number range from 3000 to
30,000. In most instances, the majority of the experimental results were well within
10% of the predicted results [61, 62]. The uncertainty analysis of the overall
experimental procedures using the method of Kline and McClintock [63] showed that
there is a maximum of 11.5% uncertainty for heat transfer coefficient calculations.
Experiments under the same conditions were conducted periodically to ensure the
repeatability of the results. The maximum difference between the duplicated
experimental runs was within 10%.
Experimental setup for horizontal to upward and downward vertical flow

The recently constructed experimental setup is equipped for measuring heat transfer,
pressure drop, void fraction, and also conducting flow visualization in air-water flow for
all major flow patterns and inclination angles from 0 (horizontal) to 90 (vertical).
The capabilities of the new experimental setup allow an undertaking that combines the
study of heat transfer, flow patterns, pressure drop, void fraction, and inclination effects.
The flow loop using air and water for this experimental setup is similar to that
illustrated in Fig. 7. Detail discussions on the design, construction and functionality of
this experimental setup are documented by Cook [64].
In this writing, only the experimental results of flow patterns and void fraction in
upward vertical flow are discussed. The test section for flow visualization and void
fraction is illustrated in Fig. 9. The flow visualization section is the central portion of
the void fraction section. The flow visualization section is constructed from a
polycarbonate tube with an inner diameter of 12.7 mm. Pressure taps are included in the
flow visualization section for measuring pressure drop across the section.
The void fraction section is constructed to trap mixture of two-phase flow in order
to measure the volume of the liquid portion. With the known volume of the void
fraction section and the measured volume of the liquid portion, the value of the void
fraction can be determined. To trap the two-phase mixture in the void fraction section,
three quick closing valves are used. Two normally open valves are used for controlling
fluid movement at the inlet and exit of the void fraction section, while a normally closed
valve is for controlling the entry of fluid into a bypass line. The quick closing valves are
W. E. Anderson Model ABV1DA101 Pneumatic Ball Valves and they exhibit a positive
seal when closed and have a closing time of 0.03 seconds. When the valves are
triggered, the two normally open valves close and the normally closed valve opens
simultaneously. In this manner, a two-phase sample is trapped in the void fraction
section while the air-water mixture is allowed to continue flowing through the bypass
line. Backflow from the mainline into the exit of the bypass line is prevented through
the use of a check valve. The experimental procedure of measuring the void fraction
with this experimental setup is also discussed in [64]. The uncertainties associated to the
measured void fraction results are estimated to be between 1.25% and 4.16%.
Fig. 9. Test section for flow visualization and void fraction.
Flow Patterns
Flow patterns in horizontal and upward inclined pipes
The various interpretations accorded to the multitude of flow patterns by different
investigators are subjective; and no uniform procedure exists at present for describing
and classifying them. In this study, the flow pattern identification for the experimental
data was based on the procedures suggested by Taitel and Dukler [65], and Kim and
Ghajar [61]; and visual observations as deemed appropriate. All observations for the
flow pattern judgments were made at the clear polycarbonate observation sections
before and after the stainless steel test section (see Fig. 7). By fixing the water flow rate,
flow patterns were observed by varying air flow rates. Flow pattern data were obtained
at isothermal condition with the pipe in horizontal position and at 2, 5, and 7 inclined
positions. These experimental data were plotted and compared using their
corresponding values of ReSG and ReSL, and the flow patterns. Representative digital
images of each flow pattern were taken using a Nikon D50 digital camera with Nikkor
50mm f/1.8D lens. Flow patterns that can be observed in horizontal two-phase flow are
illustrated in Fig. 10. The flow map for horizontal flow with the representative
photographs of the various flow patterns is shown in Fig. 11. The various flow patterns
for horizontal flow depicted in Fig. 11 show the capability of our experimental setup to
cover multitude of flow patterns. The shaded regions represent the transition boundaries
of the observed flow patterns.
The influence of small inclination angles of 2, 5, and 7 on the observed flow
patterns is shown in Fig. 12. As shown in this figure, the flow pattern transition
boundaries for horizontal flow were found to be quite different from the flow pattern
transition boundaries for inclined flow when slight inclinations of 2, 5, and 7 were
introduced. The changes in the flow pattern transition boundaries from horizontal to
slightly inclined flow are the transition boundaries for stratified flow and slug/wavy
flow. When the pipe was inclined from horizontal to slight inclination angles of 2, 5,
and 7, the stratified flow region was replaced by slug flow and slug/wavy flow for ReSG
< 4000 and 4000 < ReSG < 10000, respectively. Other shifts in the flow pattern
transition boundaries were observed in the plug-to-slug boundary and the slug-to-
slug/bubbly boundary. In these two cases, the flow pattern transition boundaries were
observed to be shifted slightly to the upper left direction as inclination angles were
slightly increased from horizontal to 7. For slightly inclined flow of 2, 5, and 7,
there were no drastic changes in the flow pattern transition boundaries.
Table 11. Number of two-phase heat transfer data points measured for different flow
patterns and pipe orientations.
Test section orientation
Flow patterns
Horizontal 2 inclined 5 inclined 7 inclined
Stratified 20
Slug 39 44 43 40
Plug 13 14 11 12
Slug/Wavy 7 15 15 15
Wavy 10 8 10 10
Wavy/Annular 22 11 9 9
Slug/Bubbly/Annular 40 47 50 52
Annular 57 45 46 49
For verification of the flow pattern map, flow patterns data from Barnea et al. [66]
was used and compared with the flow pattern maps for horizontal and 2 inclined pipe.
Using flow pattern data from Barnea et al. [66] for air-water flow in 25.5 mm diameter
horizontal pipe, the data points plotted on the flow map for horizontal flow (see Fig. 11)
are illustrated in Fig. 13. The comparison between the data points from Barnea et al.
[66] and the flow pattern map for horizontal flow showed very satisfactory agreement,
especially among the distinctive major flow patterns such as annular, slug and stratified.
It should be noted that Barnea et al. [66] had successfully compared their horizontal
flow pattern data with the flow map proposed by Mandhane et al. [67]. In a similar
manner, using flow pattern data from Barnea et al. [66] for air-water flow in 25.5 mm
diameter 2 inclined pipe, the data points plotted on the flow map for 2 inclined flow
(see Fig. 12) are illustrated in Fig. 14. The comparison between the data points from
Barnea et al. [66] and the flow pattern map for 2 inclined flow also showed very
satisfactory agreement.
Although the flow patterns may have similar names for both horizontal and inclined
flow; that does not mean that the flow patterns in the inclined positions have identical
characteristics of the comparable flow patterns in the horizontal position. For example,
it was observed that the slug flow patterns in the inclined positions of 5 and 7 have
reverse flow between slugs due to the gravitational force, which can have a significant
effect on the heat transfer. To understand the influence of flow patterns on heat transfer,
systematic measurement of heat transfer data were conducted. Table 11 and Fig. 15
illustrate the number of two-phase heat transfer data points systematically measured for
different flow patterns and test section orientations. Heat transfer data at low air and
water flow rates (ReSG < 500 and ReSL < 700) were not collected. At such low air and
water flow rates, there exists the possibility of local boiling or dry-out which could
potentially damage the heated test section.
Stratified Slug Slug/wavy Wavy/annular
Plug Wavy Slug/bubbly Annular
Fig. 10. Flow patterns in horizontal two-phase flow.
20000 Plug Slug/Bubbly

or
Slug/Bubbly/Annular
10000
Slug
5000
ReSL
Annular
2000
1000
Slug/Wavy
Stratified
500
Wavy/Annular
500 1000 2000 5000 10000 20000 50000
ReSG
Wavy
Fig. 11. Flow map for horizontal flow with representative photographs of flow patterns.
25000
Plug
Slug/Bubbly
10000 or
Slug/Bubbly/Annular
Slug
5000 Annular
Wavy/Annular
ReSL
Slug/Wavy
2500
1000 Stratified
(horizontal flow) Wavy
500
500 1000 2500 5000 10000 25000 50000
ReSG
Flow pattern transition boundaries

Horizontal flow
2 inclined flow
5 inclined flow
7 inclined flow
Fig. 12. Change of flow pattern transition boundaries as pipe inclined from horizontal position.
Fig. 13. Flow patterns data points from Barnea et al. [66] plotted on the flow map for horizontal
flow (see Fig. 11).
Flow patterns in vertical pipes

Flow patterns in upward vertical two-phase air-water flow were conducted at
isothermal condition using the robust experimental setup capable for pipe orientation
from downward vertical to upward vertical. The experimental setup is equipped for
measuring heat transfer, pressure drop, void fraction, and also conducting flow
visualization in air-water flow for all major flow patterns and inclination angles, from
0 (horizontal) to 90 (vertical). All observations for the flow pattern judgments were
made at the flow visualization section (see Fig. 9). By fixing the water flow rate, flow
patterns were observed by varying air flow rates. Using visual observation and digital
photography, distinctive flow patterns were recognized and transition boundaries
between flow patterns were determined. The five distinctive major flow patterns
observed in the upward vertical two-phase flow are dispersed bubble, slug, churn, froth,
and annular. Based on the experimentally documented flow patterns and flow pattern
transition boundaries, the two-phase flow pattern map for the upward vertical pipe was
delineated. The flow map for vertical flow with the representative photographs of the
various flow patterns are shown in Fig. 16. Flow patterns that can be observed in
upward vertical two-phase flow are illustrated in Fig. 17. The technique for obtaining
the digital images was similar to that employed for horizontal pipe using a Nikon D50
digital camera with Nikkor 50mm f/1.8D lens.
25000
Plug
10000 Annular/Bubbly/Slug
5000
ReSL
Annular/Wavy
Slug/Wavy
2500
Slug Annular
1000
Wavy
Annular
Elongated bubble
500
Slug
Stratified wavy
500 1000 2500 5000 10000 25000 50000
ReSG
Fig. 14. Flow patterns data points from Barnea et al. [66] plotted on the flow map for 2 inclined
flow (see Fig. 12).

25000 25000
10000 10000
5000 5000
ReSL
ReSL
2500 2500
1000 1000
500 500
500 1000 2500 5000 10000 25000 50000 500 1000 2500 5000 10000 25000 50000
ReSG ReSG
(a) Horizontal flow (b) 2 inclined flow
25000 25000
10000 10000
5000 5000
ReSL
ReSL
2500 2500
1000 1000
500 500
500 1000 2500 5000 10000 25000 50000 500 1000 2500 5000 10000 25000 50000
ReSG ReSG
(c) 5 inclined flow (d) 7 inclined flow
Fig. 15. Flow maps for horizontal, 2, 5, and 7 inclined flows with distribution of heat transfer
data collected.
The slug-churn and churn-annular transition boundaries in this experimental study

were compared with correlations available in the literature. The correlation suggested by
Wallis [68] is widely used and can be written as
VSL1 / 2 mVSG
1 / 2
C (26)
where the dimensionless superficial velocities for gas and liquid are expressed as

VSG VSG G1 / 2 [ gD( L G )]1 / 2 (26a)
and
VSL VSL 1L/ 2 [ gD( L G )]1 / 2 (26b)
In Wallis original expression of Eq. (26), the values of the dimensionless parameters
are C = m = 1 [68]. Equation (26) may be treated as an empirical correlation where the
parameters C and m depend on the flow conditions at the inlet and outlet as well as
geometry. The parameters, C and m, vary approximately within the 0.7 C 1.0 and
0.8 m 1.0 ranges [25]. McQuillan and Whalley [69] applied Wallis correlation, Eq.
(26), and showed generally good agreement with experimental flow pattern data. The
comparison of the experimentally documented slug-churn transition with Eq. (26) is
listed in Table 12 and illustrated in Fig. 16. The agreement between the experimental
data and Eq. (26), with C = 0.94 and m = 1.0, is satisfactory and the percentage error is
within 6%, see Table 12. At the churn-annular transition, the experimental data was
compared with the results from McQuillan and Whalley [69], and agreement is also
generally good, see Fig. 16.
Table 12. Comparison of experimental data with Eq. (26), for

the slug-churn transition.
VSG [m s-1]
VSL [m s-1] Error [%]
Experimental Eq. (26)
0.080 1.02 0.975 4.37
0.165 0.90 0.873 3.06
0.310 0.70 0.717 2.44
0.460 0.53 0.560 5.73
0.600 0.42 0.417 0.69

C = 0.94 and m = 1
Experimental Results of Void Fraction in Upward Vertical Flow

The experimental results of void fraction in upward vertical flow were measured
from the test section for flow visualization and void fraction illustrated in Fig. 9. The
variation of void fraction with gas mass flow rate for vertical pipe flow is shown in Fig.
18. As liquid mass flow rate increases, the increase in liquid holdup cause the void
fraction versus gas mass flow rate curves shift lower. On Fig. 18, the groupings of
various flow patterns on the variation of void fraction with gas mass flow rate curves
are shown. Slug flow is confined to low-range gas mass flow rate with 0.25 < < 0.72,
while churn and froth flows are found in mid-range gas mass flow rate. At any given
gas mass flow rate, churn flow has higher void fraction than froth flow. Annular flow is
in the high gas mass flow rate region with 0.72 < < 0.90. The range of void fraction
values observed in the present study for different flow patterns in upward vertical flow
is listed in Table 13.
.
Annular
Dispersed
bubble
1
0.8 Froth
Froth 0.6
0.5
0.4
V SL [m/s]
0.2
Slug Churn
Churn Annular
0.1
0.08
0.2 0.4 0.6 0.8 1 2 4 6 8 10 20
VSG [m/s]
Flow pattern transition boundaries
Slug Experimental
McQuillan and Whalley [69]
Eq. (26) with C =0.94 & m = 1
Fig. 16 Flow map for vertical flow with representative photographs of

flow patterns.
Dispersed
bubble
Fig. 17. Flow patterns in upward vertical two-phase flow.

Table 13. Range of void fraction values for different flow patterns.
Flow pattern Range of void fraction
Dispersed bubble 0.16 to 0.48
Slug 0.25 to 0.69
Churn 0.35 to 0.77
Froth 0.32 to 0.78
Annular 0.72 to 0.90
A A A
A A
A A A
A A A
C A A A A
A
A A A
0.8 C
F
A A
A
A A
F F F AA A A
C F A A
C F A A
C C F F F F A
F
F F F A
C
SS C F
F F
C C F
F
Void fraction
S C F
S F
0.6 S F
F
F
S C
S
S
C
S SC
Flow pattern Liquid mass flow rate
S
0.4 S
S F
A Annular 0.6 kg/min
S
S C 1.2 kg/min
S
S
C Churn 2.3 kg/min
S
S F Froth 3.4 kg/min
4.5 kg/min
0.2 S Slug 5.6 kg/min
0.00 0.01 0.02 0.03 0.04 0.05 0.10 0.15 0.20 0.25
Gas mass flow rate (kg/min)
Fig. 18. Variation of void fraction with gas mass flow rate for vertical pipe flow.
Systematic Investigation on Two-Phase Gas-Liquid Flow Heat

Transfer in Horizontal and Slightly Upward Inclined Pipes
In this section, an overview of the different trends that have been observed in the
heat transfer behavior of the two-phase air-water flow in horizontal and inclined pipes
for various flow patterns is presented. The non-boiling two-phase heat transfer data
were obtained by systematically varying the air and water flow rates and the pipe
inclination angle. The summary of the experimental conditions and measured heat
transfer coefficients are tabulated in Table 14. Detailed discussions on the complete
experimental results are documented by Ghajar and Tang [57].
Figures 19 and 20 provide an overview of the pronounced influence of the flow
pattern, superficial liquid Reynolds number (water flow rate) and superficial gas
Reynolds number (air flow rate) on the two-phase heat transfer coefficient in horizontal
flow. The results presented in Fig. 19 clearly show that two-phase heat transfer
coefficient is strongly influenced by the superficial liquid Reynolds number (ReSL). As
shown in Fig. 19, the heat transfer coefficient increases proportionally as superficial
liquid Reynolds number (ReSL) increases. In addition, for a fixed ReSL, the two-phase
heat transfer coefficient is also influenced by the superficial gas Reynolds number
(ReSG) and each flow pattern shows its own distinguished heat transfer trend as shown
in Fig. 20. Typically, heat transfer increases at low ReSG (the regime of plug flow), and
then slightly decreases at the mid range of ReSG (the regime of slug and slug-type
transitional flows), and increases again at the high ReSG (the regime of annular flow).
Table 14. Summary of experimental conditions and measured two-phase heat transfer data.
Test section orientation
Horizontal 2 inclined 5 inclined 7 inclined
No. of data points 208 184 184 187
ReSL range 74026100 75025900 78025900 77026000
ReSG range 70047600 70047500 59047500 56047200
Heat flux range [W m-2] 186010800 282010800 290010800 260010900
hTPEXP range [W m-2 K-1] 1015457 2425140 2865507 3645701
Flow Pattern
6000
Annular
Slug/Bubbly/Annular
5000 Wavy/Annular
Slug & Slug/Bubbly
Plug & Plug/Slug
Slug/Wavy & Wavy
4000 Stratified
[W/m K]
2
3000
EXP
h TP
2000
1000
0
0 5000 10000 15000 20000 25000
ReSL
Fig. 19. Variation of two-phase heat transfer coefficient with superficial liquid Reynolds number in
horizontal flow.
Flow Pattern
Annular
5000 Slug/Bubbly/Annular
Wavy/Annular
Slug & Slug/Bubbly
4000 Plug & Plug/Slug
Slug/Wavy & Wavy
[W/m K]
Stratified
2
3000
Re SL
EXP
700
h TP
2000
1400
2200
3400
1000
4500
6200
7800
0 10500
13000
15100
10000
20000
30000
50000
1000
2000
3000
5000
7000
700
19000
22000
ReSG
Fig. 20. Variation of two-phase heat transfer coefficient with superficial gas Reynolds number in
horizontal flow.
Comparison of General Heat Transfer Correlation with Experimental

Results for Various Flow Patterns and Pipe Inclinations
The two-phase heat transfer correlation, Eq. (21) was validated with a total of 763
experimental data points for different flow patterns in horizontal and slightly inclined
air-water two-phase pipe flows [56, 58, 59]. Equation (21), performed relatively well by
predicting over 85% of the experimental data points to within 30% for 0, 2, 5, and
7 pipe orientations. Recently, Franca et al. [70] compared their mechanistic model
developed for convective heat transfer in gas-liquid intermittent (slug) flows with the
general heat transfer correlation proposed in this study. For void fraction, Franca et al.
[70] used their own experimental data which was obtained for air-water flow in a 15 m
long, 25.4 mm inside diameter copper pipe. When comparing their mechanistic model
with Eq. (21), the agreement is within 15%, which is considered to be excellent.
However, when comparing the heat transfer correlation, Eq. (21), with data from
vertical pipes and different gas-liquid combinations, Eq. (21) has shown some
inadequacy in its performance. Equation (21) was validated with 986 experimental data
points for different flow patterns, inclination angles, and gas-liquid combinations. The
986 experimental data points were compiled from various sources with different
experimental facilities (see Table 15) with a wide range of superficial gas and liquid
Reynolds numbers (750 ReSL 127000 and 14 ReSG 209000) and inclination
angles (0 90). Figure 21 shows the comparison of Eq. (21), with all 986
experimental data points for different inclination angles and gas-liquid combinations.
Figure 21 shows that Eq. (21) performed well for two-phase flow with heat transfer
coefficient between 1000 W m-2 K-1 and 5000 W m-2 K-1. However, Eq. (21) has shown
some inadequacy in predicting two-phase flow with heat transfer coefficients below
1000 W m-2 K-1 and above 5000 W m-2 K-1. Overall, Eq. (21) successfully predicted
83% of the 986 experimental data points within 30% agreement (see Table 15). The
results shown in Table 15 and Fig. 21 prompted further investigation and improvements
were made on Eq. (21).
Table 15. Results of the predictions for 986 experimental heat transfer data points with different
gas-liquid combinations and inclination angles by using Eq. (21).
No. of No. of No. of
Avg. Range of parameters
RMS data data data
dev.
Data set dev. points points points
range
(%) within within within ReSL ReSG PrG / PrL G / L
(%)
20% 25% 30%
All 16.9 750 14
649 746 817
986 data points 33.1 to to to
(66%) (76%) (83%)
0 90 30.8 127000 209000
Air-water ( = 0)
12.6 2100 700
160 data points 111 140 154
20.5 to to to
[57] (63%) (80%) (88%)
18.6 67000 48000
16 data points [70]
Air-water ( = 2) 12.7 750 700
143 154 168
(78%) (84%) (91%)
[57] 23.0 26000 48000
3.64 103 to 26.3 103

9.99 103 to 148 103
Air-water ( = 5) 15.9 780 600
124 137 150
(67%) (74%) (82%)
[57] 64.5 26000 48000
Air-water ( = 7) 16.3 770 560
110 132 149
(59%) (71%) (80%)
[57] 74.7 26000 47000
Air-water ( =
22.3 4000 43
90) 67 79 85
25.0 to to to
105 data points (64%) (75%) (81%)
2.4 127000 154000
[40]
Air-silicone ( = 4.6 8400 52
56 56 56
90) 5.9 to to to
(100%) (100%) (100%)
56 data points [41] 6.1 21000 42000
Helium-water ( = 25.9 4000 14
22 31 37
90) 25.4 to to to
(44%) (62%) (74%)
50 data points [42] 6.9 126000 13000
Freon 12-water ( 33.3 4200 860
16 17 18
= 90) 39.1 to to to
(36%) (39%) (41%)
44 data points [42] 0 55000 209000
Values of constant and exponents: C = 0.82, m = 0.08, n = 0.39, p = 0.03, q = 0.01, and r = 0.40.
With the proposed constant and exponents, C = 0.55, m = 0.1, n = 0.4, and p = q = r
= 0.25, Eq. (24) was successfully validated with a total of 986 experimental data points
for different flow patterns, inclination angles, and gas-liquid combinations. The 986
experimental data points were compiled from various sources with different
experimental facilities (see Table 16) with a wide range of superficial gas and liquid
Reynolds numbers (750 ReSL 127000 and 14 ReSG 209000) and inclination
angles (0 90). As summarized in Table 16, the comparison of the predictions by
the general two-phase heat transfer correlation, Eq. (24), confirmed that the correlation
is adequately robust. Of all the 986 experimental data points, Eq. (24) has successfully
predicted 90% of the data points within 25% agreement with the experimental results.
Overall, the prediction by Eq. (24) has a root-mean-square deviation of 18.4% from the
experimental data.
Figure 22 shows the comparison of the calculated hTP values from the general heat
transfer correlation, Eq. (24), with all 986 experimental data points for different
inclination angles and gas-liquid combinations. The comparison of the predictions by
Eq. (24) with experimental data for air-water horizontal flow is shown in Fig. 23. The
results illustrated in Fig. 23 show that the introduction of the flow pattern factor, Eq.
(18), into the general heat transfer correlation, Eq. (24), provides the needed capability
to handle different flow patterns. Figure 24 shows the comparison of the predictions by
Eq. (24) with experimental data for air-water in slightly inclined pipes (2, 5 and 7).
Finally, as illustrated in Figure 25, the comparison of the predictions by Eq. (24) with
experimental data for various gas-liquid combinations in vertical pipes shows that the
modified inclination factor (I *), see Eq. (22), has adequately accounted for the
inclination effects.
100000
60000
40000
20000
10000
(W/m2 K)
6000
4000
2000
CAL
1000
hTP
600 +25 %
400
Horizontal
200 2 inclined
5 inclined
100
7 inclined
60
-25 % Vertical
40
1000
2000
4000
6000
10000
20000
40000
60000
100
200
400
600
100000
40
60
hTP
EXP
(W/m2 K)
Fig. 21. Comparison of the predictions by Eq. (21) with all 986 experimental data points for
different inclination angles and gas-liquid combinations (see Table 15).
100000
60000
40000
20000
10000
[W/m2 K]
6000
4000
2000
CAL
1000
hTP
600
400
+25 % Horizontal
200 2 inclined
100 5 inclined
7 inclined
60 -25 % Vertical
40 1000
2000
4000
6000
10000
20000
40000
60000
100
200
400
600
100000
40
60
hTP
EXP
[W/m2 K]
Fig. 22. Comparison of the predictions by Eq. (24) with all 986 experimental data points for
different inclination angles and gas-liquid combinations (see Table 16).
Fig. 23. Comparison of the predictions by Eq. (24) with experimental data for air-water horizontal
pipe flow (see Table 16).
Fig. 24. Comparison of the predictions by Eq. (24) with experimental data for air-water in slightly
inclined pipes (see (Table 16).
Fig. 25. Comparison of the predictions by Eq. (24) with experimental data for various gas-liquid
combinations in vertical pipes (see Table 16).

Table 16. Results of the predictions for 986 experimental heat transfer data points with different
gas-liquid combinations and inclination angles by using Eq. (24).
No. of No. of No. of
Avg. Range of parameters
RMS data data data
dev.
Data set dev. points points points
range
(%) within within within ReSL ReSG PrG / PrL G / L
(%)
20% 25% 30%
All 15.3 750 14
793 884 922
(80%) (90%) (94%)
0 90 12.5 127000 209000
Air-water (= 0) 16.2 2100 700
127 152 164
160 data points [57] 22.2 to to to
(72%) (86%) (93%)
16 data points [70] 20.4 67000 48000
9.2 750 700
Air-water ( = 2) 161 178 184
13.0 to to to
184 data points [57] (88%) (97%) (100%)
12.9 26000 48000
3.64 103 to 26.3 103

9.99 103 to 148 103
7.7 780 600
Air-water ( = 5) 154 169 174
12.1 to to to
184 data points [57] (84%) (92%) (95%)
11.8 26000 48000
10.3 770 560
Air-water ( = 7) 164 174 176
12.3 to to to
187 data points [57] (88%) (93%) (94%)
9.5 26000 47000
24.5 4000 43
Air-water ( = 90) 79 92 95
23.8 to to to
105 data points [40] (75%) (88%) (90%)
11.4 127000 154000
Air-silicone ( = 1.7 8400 52
37 42 47
90) 10.3 to to to
(66%) (75%) (84%)
56 data points [41] 9.4 21000 42000
Helium-water ( = 25.9 4000 14
41 42 46
90) 28.3 to to to
(82%) (84%) (92%)
50 data points [42] 17.6 126000 13000
Freon 12-water ( = 24.9 4200 860
30 35 36
90) 29.8 to to to
(68%) (80%) (82%)
44 data points [42] 4.0 55000 209000
Values of constant and exponents: C = 0.55, m = 0.1, n = 0.4, and p = q = r = 0.25.
Practical Illustrations of Using the General Two-Phase Heat Transfer

Correlation
The general two-phase heat transfer correlation, Eq. (24), is applicable for
estimating heat transfer coefficients for non-boiling two-phase, two-component (liquid
and permanent gas) flow in pipes. In this section, three illustrations of using the general
two-phase heat transfer correlation, Eq. (24), are discussed. The first illustration is about
the application of the correlation on air and gas-oil flow in vertical pipes with gas-to-
liquid volume ratio of approximately two. The second illustration is on air and silicone
(Dow Corning 200 Fluid, 5 cs) in a vertical pipe with liquid-to-gas volume ratio of
approximately ninety. Finally, the third illustration is an application of the correlation
on air and water pipe flow in microgravity condition.
Application in air and gas-oil flow

Dorresteijn [71] conducted an experimental study of heat transfer in non-boiling
two-phase flow of air and gas-oil through 70-mm diameter vertical tubes. The liquid
phase consists of domestic grade gas-oil with kinematic viscosity (L) of 4.710-6 m2 s-1
and Prandtl number (PrL) of approximately 60 [71]. In the conditions at which VSG = 8
m s-1, VSL = 3.16 m s-1, G = 2.5 kg m-3, L = 835 kg m-3, and = 0.67, Dorresteijn [71]
measured a value of 1.65 for hTP/hL. The following example calculation illustrates the
use of the general two-phase heat transfer correlation, Eq. (24), to predict the hTP/hL
value measured by Dorresteijn [71].
From the measured superficial gas and liquid velocities, and void fraction, the gas
and liquid velocities are found to be
VSG VSL
VG 11.9 m s -1 and VL 9.58 m s -1
1
The gas and liquid mass flow rates are calculated as
m G GVSG A 0.0771 kg s -1 and m L LVSL A 10.2 kg s -1
Using the gas and liquid mass flow rates, the quality is determined to be
m G
x 0.0075
m G m L
Equations (19) and (18) are then used for calculating the flow pattern factor (FP),
2 (V V ) 2
FS tan 1 G G L 0.0969 and FP (1 ) FS2 0.336
gD ( L G )

Using Eqs. (23) and (22), the inclination factor (I *) for vertical tube ( = 90) is
calculated to be
( L G ) gD 2
Eo 1600 and I 1 Eo sin 1601

The surface tension () of gas-oil is assumed to be 2510-3 N m-1, since the surface
tension for live gas-oil at 1380 kPa ranges from 2010-3 to 3010-3 N m-1 [72]. Using
the general two-phase heat transfer correlation, Eq. (24), the value for hTP/hL is
estimated to be,
0.4
x 1 FP PrG
0.1

0.25
L
0.25

hTP
FP 1 0.55 I
* 0.25
1.53
hL 1 x FP PrL G
The Prandtl number (PrG) and dynamic viscosity (G) for air are 0.71 and 18.210-6 kg
m-1 s-1, respectively. Comparing with the measured value of hTP/hL = 1.65 by Dorresteijn
[71], the general two-phase heat transfer correlation, Eq. (24), under-predicted the
measured value by 7.3%.
Application in air and silicone flow

Liquid silicone such as the Dow Corning 200 Fluid, 5 cs, is used primarily as an
ingredient in cosmetic and personal care products due to its high spreadability, low
surface tension ( = 19.710-3 N m-1), non-greasy, soft feel and subtle skin lubricity
characteristics. A two-phase flow of air and silicone (Dow Corning 200 Fluid, 5 cs)
with m L = 0.907 kg s-1, x = 2.08105, G = 1.19 kg m-3, L = 913 kg m-3, G = 18.410-6
kg m-1 s-1, L = 45.710-4 kg m-1 s-1, w = 39.810-4 kg m-1 s-1, PrG = 0.71, PrL = 64, kL =
0.117 W m-1 K-1, and = 0.011 flows inside an 11.7-mm diameter vertical ( = 90)
tube. Using the general two-phase heat transfer correlation, Eq. (24), the two-phase heat
transfer coefficient for this flow can be estimated.
With known liquid mass flow rate ( m L ) and quality (x), the gas mass flow rate ( m G )
is determined using
x
m G m L 1.89 10 5 kg s -1
1 x
From the gas and liquid mass flow rates, the superficial gas and liquid velocities can be
calculated:
m G m L
VSG 0.149 m s -1 and VSL 9.24 m s -1
G A L A
Using the superficial velocities and void fraction, the gas and liquid velocities are found
to be
VSG VSL
VG 13.5 m s -1 and VL 9.34 m s -1
1
2 (V V ) 2
g D ( L G )

Using Eqs. (23) and (22), the inclination factor (I *) for vertical tube ( = 90) is
calculated to be
( L G ) gD 2
Eo 62.1 and I 1 Eo sin 63.1

Finally, with the general two-phase heat transfer correlation, Eq. (24), the value for hTP
is estimated to be,
x 1 FP
0.1

0.4
PrG
0.25
L
0.25

hTP hL FP 1 0.55 I
* 0.25
3550 W m -2 K -1
1 x FP PrL G
When compared with the measured two-phase heat transfer coefficient of 3480 W m-2
K-1 by Rezkallah [41] in similar flow conditions, the general two-phase heat transfer
correlation, Eq. (24), over-predicted the measured value by 2%.
Application in microgravity condition

An air-water slug flow heat transfer coefficient in microgravity condition (less than
1% of earths normal gravity) was measured by Witte et al. [73] in a 25.4-mm diameter
horizontal tube. In the conditions at which VSG = 0.3 m s-1, VSL = 0.544 m s-1, G = 1.16
kg m-3, L = 997 kg m-3, G = 18.510-6 kg m-1 s-1, L = 85.5105 kg m-1 s-1, w =
73.9105 kg m-1 s-1, PrG = 0.71, PrL = 5.0, kL = 0.613 W m-1 K-1 and = 0.27, Witte et
al. [73] measured a value of 3169 W m-2 K-1 for the two-phase heat transfer coefficient
(hTP). The following example calculation illustrates the use of the general two-phase
heat transfer correlation, Eq. (24), to predict the hTP value measured by Witte et al. [73].
From the measured superficial gas and liquid velocities, and void fraction, the gas
and liquid velocities are found to be
VSG VSL
VG 1.11 m s -1 and VL 0.745 m s -1
1
The gas and liquid mass flow rates are calculated as
m G GVSG A 1.76 10 4 kg s -1 and m L LVSL A 0.275 kg s -1
Using the gas and liquid mass flow rates, the quality is determined to be
m G
x 6.40 10 4
m G m L
2 (V V ) 2
g D ( L G )

The inclination factor (I *) has a value of one in horizontal tube ( = 0). Thus, using the
general two-phase heat transfer correlation, Eq. (24), the value for hTP is estimated to be,
x 1 FP
0.1

0.4
PrG
0.25
L
0.25

hTP hL FP 1 0.55 I
* 0.25
2810 W m -2 K -1
1 x FP PrL G
Comparing with the measured two-phase heat transfer coefficient of 3169 W m-2 K-1
by Witte et al.[73], the general two-phase heat transfer correlation, Eq. (24), under-
predicted the measured value by 11%. Although the above example showed that Eq.
(24) can satisfactorily estimate heat transfer coefficient for one case of two-phase flow
under reduced gravity condition, it should be noted that Eq. (24) was not developed to
handle reduced gravity conditions. Validation with experimental results needs to be
done before the use of Eq. (24) in reduced gravity conditions can be recommended.
Summary
The work documented in this manuscript initiated with the motivation to
understand, in both fundamental and industrial applications, the importance of non-
boiling two-phase flow heat transfer in pipes. Through the survey of literature and
tracing the validity and limitations of the numerous two-phase non-boiling heat transfer
correlations that have been published over the past six decades, it was established that
there is no single correlation capable of predicting the two-phase flow heat transfer for
all fluid combinations in vertical pipes [39].
The results from the literature survey prompted the development of a two-phase
non-boiling heat transfer correlation that is robust and applicable to turbulent gas-liquid
flow in vertical pipes with different flow patterns and fluid combinations [47]. Since the
development of the two-phase non-boiling heat transfer correlation for vertical pipes by
Kim et al. [47], extensive efforts have been invested in the development of the general
two-phase heat transfer correlation, Eq. (24). When compared with experimental data
from horizontal, slightly inclined, and vertical pipes with various fluid combinations
and flow patterns, the general two-phase heat transfer correlation successfully predicted
90% of the data points within 25% agreement with the experimental data and has a
root-mean-square deviation of 18.4% from the experimental data. In addition, practical
illustrations of using the general two-phase heat transfer correlation were also discussed.
In the efforts of investigating non-boiling two-phase flow heat transfer in pipes,
significant amount of work has also been done on understanding void fraction. A very
extensive comparison of 68 void fraction correlations available in the literature against
2845 experimental data points was conducted by Woldesemayat and Ghajar [10]. From
this work an improved void fraction correlation, Eq. (2), was proposed. The improved
void fraction correlation gives noticeable improvements over other correlations when
compared with 2845 experimental data points of various pipe sizes, inclinations, and
two-phase fluid mixtures from various sources with different experimental facilities.
Further scrutiny has also been done specifically for upward vertical two-phase flow on
the performances of available void fraction correlations. A database of 1208
experimental data points, for different gas-liquid combinations and pipe diameters, was
used for the comparison with void fraction correlations available in the literature. The
effort resulted in the categorization of void fraction correlations recommended for
specific void fraction ranges in upward vertical two-phase flow. Results of these
categorical comparisons would allow the access to correlations with higher accuracies
for specific void fraction range of interest.
The overall objective of this on-going research has been to establish a fundamental
understanding of two-phase flow in pipes. Although the initial effort has been on the
development of heat transfer in non-boiling two-phase flow, the focus now has been
expanded to encompass void fraction, flow patterns and even pressure drop in two-
phase flow. The direction is certainly toward gaining understanding of the fundamentals
in the hydrodynamic and thermal aspects of gas-liquid two-phase flow in pipes of all
possible orientations.
Acknowledgements
Sincere thanks are offered to Emerson Micro Motion for generously donating one of
the Coriolis flow meters and providing a substantial discount on the other one. Thanks
are also due to Mr. Martin Mabry for his assistance in procuring these meters. The
authors also thank Mr. Pranav Godbole for his contributions in the experimental part for
void fraction and flow patterns as well as the analysis of void fraction correlations in
upward vertical two-phase flow [74].
Nomenclature
A cross sectional area, m2
C constant value of the leading coefficient in Eqs. (13), (21) & (24); and parameter
of Wallis correlation, Eq (26), dimensionless
C0 two-phase distribution coefficient, dimensionless
c specific heat at constant pressure, J kg-1 K-1
D pipe inside diameter, m
Eo Etvs number, dimensionless
FP flow pattern factor, Eq. (18), dimensionless
FS shape factor, Eq. (19), dimensionless
g gravitational acceleration, m s-2
h heat transfer coefficient, W m-2 K-1
I inclination factor, Eq. (20), dimensionless
I* modified inclination factor, Eq. (22), dimensionless
K slip ratio, VG/VL, dimensionless

k thermal conductivity, W m-1 K-1
L length, m
m constant in Wallis correlation, Eq. (26), dimensionless
m mass flow rate, kg s-1
NST number of thermocouple stations, dimensionless
Nu Nusselt number, hD/k, dimensionless
P pressure, Pa
Pr Prandtl number, c/k, dimensionless
Q volumetric flow rate, m3 s-1
q heat flux, W m-2
Re Reynolds number, VD/, dimensionless
ReL in situ liquid Reynolds number, Eq. (15), dimensionless
SL wetted perimeter, m
T temperature, K
uGM drift velocity for gas, m s-1
V velocity, m s-1
V* superficial velocity in Wallis correlation, Eq. (26), dimensionless
x flow quality, m G /( m G + m L ), dimensionless
z axial coordinate, m
z element length of each thermocouple station, m
Greek Symbols
void fraction, dimensionless
inclination angle, rad.
dynamic viscosity, kg m-1 s-1
kinematic viscosity, m2 s-1
density, kg m-3
surface tension, N m-1
Subscripts
atm atmosphere
B bulk
CAL calculated
eff effective
eq equilibrium state
EXP experimental
G gas
k index of thermocouple station
L liquid
SG superficial gas
SL superficial liquid
sys system
TP two-phase
w wall
Superscripts
m exponent on the quality ratio term in Eqs. (13), (21) & (24), dimensionless
n exponent in Eqs. (13), (21) & (24), dimensionless
p exponent on the Prandtl number ratio term in Eqs. (13), (21) & (24),
dimensionless
q exponent on the viscosity ratio term in Eqs. (13), (21) & (24), dimensionless
r exponent on the inclination factor in Eqs. (21) & (24), dimensionless
Diacritical Marks
local mean
~ non-dimensionalized
References
[1] K. C. Shiu and H. D. Beggs, "Predicting temperatures in flowing oil wells,"
Journal of Energy Resources Technology, vol. 102, pp. 2-11, 1980.
[2] O. V. Trevisan, F. A. Franca, and A. C. Lisboa, "Oil production in offshore
fields: An overview of the brazilian technology development program," in
Proceedings of the 1st World Heavy Oil Conference, Beijing, China, 2006.
[3] E. M. Furuholt, "Multiphase technology: Is it of interest for future field
developments?," in Society of Petroleum Engineers European Petroleum
Conference, London, England, 1988.
[4] G. G. McClaflin and D. L. Whitfill, "Control of paraffin deposition in
production operations," Journal of Petroleum Technology, vol. 36, pp. 1965-
1970, 1984.
[5] H. S. Fogler, "Paraffin research," Porous Media Research Group, Website:
<sitemaker.umich.edu/sfogler/paraffin_deposition>, Retrieved Sep. 29, 2009.
[6] P. Singh, R. Venkatesan, H. S. Fogler, and N. Nagarajan, "Formation and aging
of incipient thin film wax-oil gels," AIChE Journal, vol. 46, pp. 1059-1074,
2000.
[7] G. P. Celata, A. Chiaradia, M. Cumo, and F. D'Annibale, "Heat transfer
enhancement by air injection in upward heated mixed-convection flow of
water," International Journal of Multiphase Flow, vol. 25, pp. 1033-1052, 1999.
[8] L. B. Fore, L. C. Witte, and J. B. McQuillen, "Heat transfer to annular gas-liquid
mixtures at reduced gravity," Journal of Thermophysics and Heat Transfer, vol.
10, pp. 633-639, 1996.
[9] L. B. Fore, L. C. Witte, and J. B. McQuillen, "Heat transfer to two-phase slug
flows under reduced-gravity conditions," International Journal of Multiphase
Flow, vol. 23, pp. 301-311, 1997.
[10] M. A. Woldesemayat and A. J. Ghajar, "Comparison of void fraction
correlations for different flow patterns in horizontal and upward inclined pipes,"
International Journal of Multiphase Flow, vol. 33, pp. 347-370, 2007.
[11] B. A. Eaton, "The prediction of flow patterns, liquid holdup and pressure losses
occurring during continuous two-phase flow in horizontal pipelines," Ph.D.
thesis, University of Texas, Austin, Texas, 1966.
[12] H. D. Beggs, "An experimental study of two phase flow in inclined pipes,"
Ph.D. thesis, University of Tulsa, Tulsa, Oklahoma, 1972.
[13] P. L. Spedding and V. T. Nguyen, "Data on holdup, pressure loss and flow
patterns for two-phase air-water flow in an inclined pipe," Eng. Report 122,
University of Auckland, Auckland, New Zealand, 1976.
[14] H. Mukherjee, "An experimental study of inclined two-phase flow," Ph.D.
thesis, University of Tulsa, Tulsa, Oklahoma, 1979.
[15] K. Minami and J. P. Brill, "Liquid holdup in wet-gas pipelines," SPE Production
Engineering, vol. 2, pp. 36-44, 1987.
[16] F. Franca and R. T. Lahey Jr., "The use of drift-flux techniques for the analysis
of horizontal two-phase flows," International Journal of Multiphase Flow, vol.
18, pp. 787-801, 1992.
[17] G. H. Abdul-Majeed, "Liquid holdup in horizontal two-phase gas-liquid flow,"
Journal of Petroleum Science and Engineering, vol. 15, pp. 271-280, 1996.
[18] M. Sujumnong, "Heat transfer, pressure drop and void fraction in two-phase,
two component flow in a vertical tube," Ph.D. thesis, University of Manitoba,
Winnipeg, Manitoba, Canada, 1998.
[19] S. Morooka, T. Ishizuka, M. IIzuka, and K. Yoshimura, "Experimental study on
void fraction in a simulated bwr fuel assembly (evaluation of cross-sectional
averaged void fraction)," Nuclear Engineering and Design, vol. 114, pp. 91-98,
1989.
[20] G. E. Dix, "Vapor void fractions for forced convection with subcooled boiling at
low flow rates," Ph.D. thesis, University of California, Berkeley, California,
1971.
[21] S. Z. Rouhani and E. Axelsson, "Calculation of void volume fraction in the
subcooled and quality boiling regions," International Journal of Heat and Mass
Transfer, vol. 13, pp. 383-393, 1970.
[22] G. A. Hughmark, "Holdup in gas-liquid flow," Chemical Engineering Progress,
vol. 58, pp. 62-65, 1962.
[23] A. Premoli, D. Francesco, and A. Prima, "An empirical correlation for
evaluating two-phase mixture density under adiabatic conditions," in European
Two-Phase Flow Group Meeting, Milan, Italy, 1970.
[24] A. I. Filimonov, M. M. Przhizhalovski, E. P. Dik, and J. N. Petrova, "The
driving head in pipes with a free interface in the pressure range from 17 to 180
atm," Teploenergetika, vol. 4, pp. 22-26, 1957.
[25] S. M. Ghiaasiaan, Two-phase flow, boiling, and condensation. New York:
Cambridge University Press, 2008.
[26] D. J. Nicklin, J. O. Wilkes, and J. F. Davidson, "Two-phase flow in vertical
tubes," Chemical Engineering Research and Design, vol. 40, pp. 61-68, 1962.
[27] J. Schmidt, H. Giesbrecht, and C. W. M. van der Geld, "Phase and velocity
distributions in vertically upward high-viscosity two-phase flow," International
Journal of Multiphase Flow, vol. 34, pp. 363-374, 2008.
[28] R. Chokshi, "Prediction of pressure drop and liquid hold up in vertical two-
phase flow through large diameter tubing," Ph.D. thesis, University of Tulsa,
Tulsa, Oklahoma, 1994.
[29] R. C. Fernandes, "Experimental and theoretical studies of isothermal upward
gas-liquid flows in vertical tubes," Ph.D. thesis, University of Houston, Houston,
Texas, 1981.
[30] O. Oshinowo, "Two-phase flow in a vertical tube coil," Ph.D. thesis, University
of Toronto, Toronto, Canada, 1971.
[31] H. S. Isbin, N. C. Shear, and K. C. Eddy, "Void fractions in steam water two-
phase flow," AIChE Journal, vol. 3, pp. 136-142, 1957.
[32] H. Nishino and Y. Yamazaki, "A new method of evaluating steam volume
fractions in boiling systems," Journal of Society of Atomic Energy Japan, vol. 5,
pp. 39-59, 1963.
[33] A. L. Guzhov, V. A. Mamayev, and G. E. Odishariya, "A study of transportation
in gas liquid systems," in 10th International Gas Union Conference, Hamburg,
Germany, 1967.
[34] R. H. Bonnecaze, W. Erskine, and E. J. Greskovich, "Holdup and pressure drop
for two-phase slug flow in inclined pipelines," AIChE Journal, vol. 17, pp.
1109-1113, 1971.
[35] M. Ishii, "One-dimensional drift-flux model and constitutive equations for
relative motion between phases in various two-phase flow regimes," Argonne
National Lab. Report, ANL-77-47, 1977.
[36] K. H. Sun, R. B. Duffey, and C. M. Peng, "A thermal-hydraulic analysis of core
uncovery," in Proceedings of the 19th National Heat Transfer Conference,
Experimental and Analytical Modeling of LWR Safety Experiments, Orlando,
Florida, 1980, pp. 1-10
[37] S. L. Kokal and J. F. Stanislav, "An experimental study of two-phase flow in
slightly inclined pipes--ii. Liquid holdup and pressure drop," Chemical
Engineering Science, vol. 44, pp. 681-693, 1989.
[38] K. Takeuchi, M. Y. Young, and L. E. Hochreiter, "Generalized drift flux
correlation for vertical flow," Nuclear Science and Engineering, vol. 112, pp.
170-180, 1992.
[39] D. Kim, A. J. Ghajar, R. L. Dougherty, and V. K. Ryali, "Comparison of 20 two-
phase heat transfer correlations with seven sets of experimental data, including
flow pattern and tube inclination effects," Heat Transfer Engineering, vol. 20,
pp. 15-40, 1999.
[40] M. M. Vijay, "A study of heat transfer in two-phase two-component flow in a
vertical tube," Ph.D. thesis, University of Manitoba, Winnipeg, Manitoba,
Canada, 1978.
[41] K. S. Rezkallah, "Heat transfer and hydrodynamics in two-phase two-component
flow in a vertical tube," Ph.D. thesis, University of Manitoba, Winnipeg,
Manitoba, Canada, 1987.
[42] M. A. Aggour, "Hydrodynamics and heat transfer in two-phase two-component
flow," Ph.D. thesis, University of Manitoba, Winnipeg, Manitoba, Canada,
1978.
[43] R. H. Pletcher, "An experimental and analytical study of heat transfer and
pressure drop in horizontal annular two-phase, two-component flow," Ph.D.
thesis, Cornell University, Ithaca, New York, 1966.
[44] C. D. G. King, "Heat transfer and pressure drop for an air-water mixture flowing
in a 0.737inch i.D. Horizontal tube," M.S. thesis, University of California,
Berkeley, California, 1952.
[45] Y.-C. Chu and B. G. Jones, "Convective heat transfer coefficient studies in
upward and downward, vertical, two-phase, non-boiling flows," AIChE
Symposium Series, vol. 76, pp. 79-90, 1980.
[46] S. R. Ravipudi and T. M. Godbold, "The effect of mass transfer on heat transfer
on heat transfer rates for two-phase flow in a vertical pipe," in Proceedings of
the 6th International Heat Transfer Conference, Toronto, Canada, 1978, pp.
505-510.
[47] D. Kim, A. J. Ghajar, and R. L. Dougherty, "Robust heat transfer correlation for
turbulent gas-liquid flow in vertical pipes," Journal of Thermophysics and Heat
Transfer, vol. 14, pp. 574-578, 2000.
[48] R. F. Knott, R. N. Anderson, A. Acrivos, and E. E. Petersen, "An experimental
study of heat transfer to nitrogen-oil mixtures," Industrial & Engineering
Chemistry, vol. 51, pp. 1369-1372, 1959.
[49] E. N. Sieder and G. E. Tate, "Heat transfer and pressure drop of liquids in
tubes," Industrial & Engineering Chemistry, vol. 28, pp. 1429-1435, 1936.
[50] A. A. Kudirka, R. J. Grosh, and P. W. McFadden, "Heat transfer in two-phase
flow of gas-liquid mixtures," Industrial & Engineering Chemistry
Fundamentals, vol. 4, pp. 339-344, 1965.
[51] B. W. Martin and G. E. Sims, "Forced convection heat transfer to water with air
injection in a rectangular duct," International Journal of Heat and Mass
Transfer, vol. 14, pp. 1115-1134, 1971.
[52] K. S. Rezkallah and G. E. Sims, "Examination of correlations of mean heat
transfer coefficients in two-phase two-component flow in vertical tubes," AIChE
Symposium Series, vol. 83, pp. 109-114, 1987.
[53] M. M. Shah, "Generalized prediction of heat transfer during two component gas-
liquid flow in tubes and other channels," AIChE Symposium Series, vol. 77, pp.
140-151, 1981.
[54] D. Chisholm, "Research note: Void fraction during two-phase flow," Journal of
Mechanical Engineering Science, vol. 15, pp. 235-236, 1973.
[55] A. J. Ghajar and J. Kim, "A non-boiling two-phase flow heat transfer correlation
for different flow patterns and pipe inclination angles," in Proceedings of the
2005 ASME Summer Heat Transfer Conference, San Francisco, CA, 2005.
[56] J. Kim and A. J. Ghajar, "A general heat transfer correlation for non-boiling gas-
liquid flow with different flow patterns in horizontal pipes," International
Journal of Multiphase Flow, vol. 32, pp. 447-465, 2006.
[57] A. J. Ghajar and C. C. Tang, "Heat transfer measurements, flow pattern maps
and flow visualization for non-boiling two-phase flow in horizontal and slightly
inclined pipe," Heat Transfer Engineering, vol. 28, pp. 525-540, 2007.
[58] C. C. Tang and A. J. Ghajar, "Validation of a general heat transfer correlation

for non-boiling two-phase flow with different flow patterns and pipe inclination
angles," in Proceedings of the 2007 ASME-JSME Thermal Engineering Summer
Heat Transfer Conference, Vancouver, BC, Canada, 2007.
[59] A. J. Ghajar, J. Kim, and C. Tang, "Two-phase flow heat transfer measurements
and correlation for the entire flow map in horizontal pipes," in Proceedings of
the 13th International Heat Transfer Conference, Sydney, Australia, 2006.
[60] A. J. Ghajar and J. Kim, "Calculation of local inside-wall convective heat
transfer parameters from measurements of local outside-wall temperatures along
an electrically heated circular tube," in Heat transfer calculations, M. Kutz, Ed.
New York: McGraw-Hill, 2006, pp. 23.3-23.27.
[61] D. Kim and A. J. Ghajar, "Heat transfer measurements and correlations for air-
water flow of different flow patterns in a horizontal pipe," Experimental
Thermal and Fluid Science, vol. 25, pp. 659-676, 2002.
[62] W. B. Durant, "Heat transfer measurement of annular two-phase flow in
horizontal and a slightly upward inclined tube," M.S. thesis, Oklahoma State
University, Stillwater, Oklahoma, 2003.
[63] S. J. Kline and F. A. McClintock, "Describing uncertainties in single-sample
experiments," Mechanical Engineering, vol. 75, pp. 3-8, 1953.
[64] W. L. Cook, "An experimental apparatus for measurement of pressure drop, void
fraction and non-boiling two-phase heat transfer and flow visualization in pipes
for all inclinations," M.S. thesis, Oklahoma State University, Stillwater,
Oklahoma, 2008.
[65] Y. Taitel and A. E. Dukler, "A model for predicting flow regime transitions in
horizontal and near horizontal gas-liquid flow," AIChE Journal, vol. 22, pp. 47-
55, 1976.
[66] D. Barnea, O. Shoham, Y. Taitel, and A. E. Dukler, "Flow pattern transition for
gas-liquid flow in horizontal and inclined pipes. Comparison of experimental
data with theory," International Journal of Multiphase Flow, vol. 6, pp. 217-
225, 1980.
[67] J. M. Mandhane, G. A. Gregory, and K. Aziz, "A flow pattern map for gas-
liquid flow in horizontal pipes," International Journal of Multiphase Flow, vol.
1, pp. 537-553, 1974.
[68] G. B. Wallis, "Flooding velocities for air and water in vertical tubes," Report
AEEW-R123, UKAEA, Harwell, UK, 1961.
[69] K. W. McQuillan and P. B. Whalley, "Flow patterns in vertical two-phase flow,"
International Journal of Multiphase Flow, vol. 11, pp. 161-175, 1985.
[70] F. A. Franca, A. C. Bannwart, R. M. T. Camargo, and M. A. L. Gonalves,
"Mechanistic modeling of the convective heat transfer coefficient in gas-liquid
intermittent flows," Heat Transfer Engineering, vol. 29, pp. 984-998, 2008.
[71] W. R. Dorresteijn, "Experimental study of heat transfer in upward and
downward two-phase flow of air and oil through 70-mm tubes," in Proceedings
of the 4th International Heat Transfer Conference, Paris & Versailles, France,
1970.
[72] G. H. Abdul-Majeed and N. B. Abu Al-Soof, "Estimation of gasoil surface

tension," Journal of Petroleum Science and Engineering, vol. 27, pp. 197-200,
2000.
[73] L. C. Witte, W. S. Bousman, and L. B. Fore, "Studies of two-phase flow
dynamics and heat transfer at reduced gravity conditions," NASA Contractor
Report 198459, NASA Lewis Research Center, Cleveland, Ohio, USA, 1996.
[74] P. V. Godbole, "Study of flow patterns and void fraction in vertical upward two-
phase flow," M.S. thesis, Oklahoma State University, Stillwater, Oklahoma,
2009.
Chapter 2
Boiling Heat Transfer in Reduced Gravity Environments

Paolo DiMarco1, , Jungho Kim2 and Haruhiko Ohta3
1
Dipartimento di Energetica, Universit di Pisa, via Diotisalvi 2, 56122 PISA Italy,
2
Department of Mechanical Engineering, University of Maryland, College Park, MD
20742, USA
3
Department of Aeronautics and Astronautics, Graduate School of Engineering, Kyushu
University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
Abstract
The effects of low gravity on boiling heat transfer are presented. A short overview of boiling is
provided to introduce non-specialists to the field of boiling heat transfer to make them aware of the
fundamental physical mechanisms. The facilities used to obtain variable gravity environments are then
described, and recent results are reviewed. Topics covered include bubble dynamics, nucleate pool
boiling, critical heat flux, transition boiling, flow boiling, boiling of mixtures, electric field effects,
acoustic effects, quenching. Planned research activities in Japan, the United States, and Europe are
presented.
Introduction
The increase in the size and power requirements of space platforms results in
increased waste heat generation and transport of heat to the radiator and ultimately to
the cold of space. To date, spacecraft active thermal control has been accomplished
using pumped single-phase liquid loops. Systems of this kind (termed Active Thermal
Control Systems, ATCSs) have been used on Mercury, Gemini, Apollo, MIR space
station and are currently used on the Russian Soyuz spacecraft, US Space Shuttle and
the International Space Station. Heat removal in a spacecraft is generally performed via
a thermal bus, i.e., a loop in which a fluid transports the rejected heat from the utilities
to the external radiators where power is radiated to space. The thermal bus must provide
a stable thermal regime for a given number of attached modules and for variations in the
thermal load. At present, thermal buses are generally mechanically pumped, single
phase loops so their heat removal capability is based on the sensible heat of the fluid,
i.e., its capacity to absorb energy through an increase in temperature. However, fluids
can also exchange energy through latent heat at constant temperature as well through a
change of phase from liquid to vapor. Boiling is recognized as a very effective
Email address: p.dimarco@ing.unipi.it, kimjh@umd.edu, ohta@aero.kyushu-u.ac.jp.

54 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) DiMarco et al.
technique to remove large amounts of heat at relatively low temperatures and is widely
applied in earth gravity based technologies. Space-based cold plate surface areas can be
significantly reduced if boiling can be exploited, resulting in significantly lower heat
exchanger mass, fluid mass flow rates, and pumping power. Furthermore, lower wall
temperature variation with heat flux can be attained in a boiling fluid, allowing better
temperature control.
The relatively poor understanding of gravity effects on multiphase flow and phase-
change heat transfer has been identified as one of the primary obstacles to reliable
design of space based hardware and processes such as heat exchange, cryogenic fuel
storage and transportation, propulsion, and electronic cooling. Although much research
in this area has been performed over the last 50 years, the mechanisms by which heat is
removed from surfaces in low-g environments are still unclear. In this chapter, the
existing boiling data in low-gravity environments are reviewed and future planned
research is described. Because of space limitations, research performed in low gravity
environments using vapor/liquid systems is emphasized. The large body of data
obtained in earth gravity using surfaces inclined to the gravity vector and/or gas/liquid
systems are generally not included.
Boiling Overview
In any heat transfer process between a solid surface and a fluid, the heat rate q is
commonly expressed using Newtons law of cooling:

q hA Tw Tref (1)
where his the heat transfer coefficient, A is the heater area, and (Tw-Tref) is the difference
between the temperature of the surface and a convenient reference one taken in the
fluid; in boiling phenomena, the saturation temperature Tsat is generally adopted for Tref.
Both the heat transfer area and the temperature difference should be kept as small as
possible, the former to minimize weight and investment costs, and the latter to minimize
entropy generation and avoid surface overheating, which in turn may lead to equipment
failure. Consequently, the heat transfer coefficient is desired to be as high as possible to
accommodate large heat fluxes.
Pool boiling
Following the approach originally developed by Nukiyama [1] in his early
experiment, heat transfer performance in pool boiling is commonly reported as a plot of
heat flux vs. wall superheat Tsat = Tw Tsat (boiling curve). The curve generally
exhibits the trend shown in Figure 1. Several heat transfer regimes can be identified.
Between A-B no boiling exists and heat transfer is by natural convection even though
the wall is above the saturation temperature. In microgravity, if buoyancy and other
driving forces are excluded, natural convection cannot take place and is replaced by
transient conduction into the liquid. When the temperature at the surface exceeds the
Boiling Heat Transfer in Reduced Gravity Environments Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 55
saturation value by a certain amount, bubbles are generated in surface cavities by

heterogeneous nucleation and boiling starts. This implies a strong increase in heat
transfer performance, and the superheat is suddenly decreased, path B-C. The
temperature overshoot
Critical heat flux E

D
Tr
ng
a
ili
ns
Heat flux, q
bo
iti
lm
on
Fi
bo
boiling
ili
G ng F
Nucleate
Minimum film boiling

Temperature overshoot at ONB
C B n
ve ctio
. con
nat
ha se
A gl e-p
Sin
Wall superheat, Twall - Tsat
(point B) may be so high as to compromise the operation of temperature sensitive

equipment such as electronic devices. Along the path CD (nucleate boiling), the
nucleation site density increases and the heat flux q increases steeply with superheat;
this heat transfer mode is termed nucleate boiling and is the most important regime for
industrial applications due to its high efficiency. However, it cannot be sustained
indefinitely: beyond a maximum value referred to as the critical heat flux (CHF, point
D), it is suppressed. Two paths can be followed depending on the controlling variable. If
the variable is heat flux, as in electric or nuclear equipment, a small rise in q causes a
transition from point D to E with a very large increase in wall temperature which often
leads to the destruction of the heater (the so-called burnout phenomenon). Beyond point
F the curve has a much smaller slope than CD and the heat transfer regime is termed
film boiling. In film boiling, the surface is completely blanketed with vapor and at most
sporadic liquid contacts may occur. An unstable vapor film covers the heater, with
bubbles detaching periodically from the film surface; radiation contributes significantly
to the total heat transfer in this regime, especially at high superheat. If heat flux is now
progressively reduced (see the arrows), this curve is followed down to point F, the
minimum film boiling heat flux (MFB), where a further decrease takes the system back
to G (hysteresis loop). On the other hand, if the wall temperature is the controlling
variable, as in heat exchangers, the unstable path DF (termed transition boiling) can be
covered, by increasing or decreasing the wall superheat. To summarize, although
nucleate boiling is a very suitable heat transfer regime, care must be taken in
establishing it from cold conditions without damaging the equipment due to temperature
overshoot, and the heat flux must always be maintained well below the critical heat flux
value.
Dispersed-flow
film boiling
transition
regime
Flow
QF
DO film boiling
Inverted-
boiling flow
annular
convective
Forced
SAT
Transition
Slug flow boiling
Nucleate boiling
NVG
Single-phase
Fig.2. Flow boiling regimes. Case A (left): low flow rate, Case B (right): high flow rate.
Flow boiling
In forced convective boiling the phenomenology is complicated by additional
system effects. Using a vertical heated pipe as reference, two kinds of regimes are
generally possible depending on the inlet subcooling and flow rate (Figure 2). In
situation A (low flow rate), nucleation is initiated at the wall when the bulk of the fluid
is still subcooled (subcooled nucleate boiling). After the point is attained in which the
bulk of the fluid becomes saturated (quality x=0), saturated nucleate boiling takes place.
The flow pattern in this region is gradually modified from bubbly to slug and annular,
with a progressive increase in heat transfer coefficient. When the liquid film at the wall
is destroyed due to thinning and instability, the liquid deficient region is entered, with a
sudden decrease of heat transfer coefficient and a consequent increase in wall
temperature. The flow regime is now dispersed drop. The droplet evaporation causes an
increase in vapor velocity and heat transfer coefficient, up to the point where heat
transfer to a single-phase vapor is eventually reached. In situation B, at higher flow rate,
a vapor layer can be formed at the wall (reversed annular flow regime) directly from
nucleate boiling conditions. Both conditions have been extensively studied, mainly in
the field of nuclear reactor safety and design, and the reader is addressed to specialized
literature (e.g Delhaye et al, [2] and Collier and Thome [3] in which appropriate
correlations for each regime are given. In both the outlined situations, the nucleate
boiling regime cannot be sustained indefinitely. A boiling transition is eventually
reached, which leads to severe heat transfer degradation. Usually, this transition is
generally termed dryout in A and departure from nucleate boiling (DNB) in B. In
contrast to the pool boiling situation, the crisis in this case depends not only on local
conditions, but on the whole evolution of the fluid before the point of the crisis. In
forced convective boiling, inertial effects are expected to dominate over buoyancy in
driving phases, except at low flow conditions: this threshold, however, has yet to be
determined in low-g conditions.
Pool boiling correlations and extension to low-g conditions

Boiling is significantly affected by buoyancy. In low gravity environments, the
magnitude of effects related to natural convection and buoyancy are small and physical
mechanisms normally masked by natural convection in earth gravity such as Marangoni
convection can substantially influence boiling and vapor bubble dynamics. Experiments
to date have shown that commonly used correlations do not properly account for the
effect of gravity on boiling processes. Generally, the dependence on gravity in
correlations has been expressed by a power law [4]
n
h g

h0 g0
(2)
where is the so-called heat transfer efficiency and the subscript 0 represents conditions
in earth gravity. For example, the correlation by Rosenhow [5] for nucleate pool boiling
q" c 3p g l v
Csf3 (3)
Tw Tsat
3 2 3s
h Pr
fg
gives n=0.5 and predicts that the heat flux goes to zero in the absence of gravity, but
experiments to date have shown this to be erroneous. Zhang & Chao [6] proposed to
retain the Rohsenows model also in microgravity, supplementing it with the actual
bubble departure diameter in place of the Laplace length. Another commonly used
correlation for nucleate pool boiling is given by Cooper [7]
hnb 55 q " M 0.5 Prm log10 Pr

0.67 0.55
(4)
where
m 0.12 0.2 log 10 Rp
and Pr = reduced pressure, M = molecular weight, and Rp = rms surface roughness in

microns. This correlation does not have any dependence on gravity (n=0), which is also
inconsistent with experiments. Similarly, Stephan and Abdelsalam [8] used regression
analysis to obtain correlations for various classes of fluids based on a large body of
data, but the data were limited to fully developed nucleate boiling on horizontal surfaces
under the influence of gravity. The role of gravity was accounted for in their correlation
through the bubble departure diameter given by
1/2
2
d 0.146 (5)
g l v
The dependence on g varies depending on which of their correlations is used.

Another well established correlation proposed in the VDI Heat Atlas [9] is due to
Stephan and Preusser
0.371
h fg d 2
0.674 0.156 0.35
q"d q"d v al2 l
Nu 0.1 2 Prl 0.16
Tsat kl Tsat kl l al Dd (6)
where Dd is the bubble detachment diameter, given by

d =C = C lL (7)
g (l -v )
The gravity acceleration appears through the capillarity length inside d, resulting in a
small value of n (n=-0.033 for a given heat flux). The VDI Heat Atlas also proposes an
alternate methodology. Once a reference value href is determined by experiments, by the
tables given in the VDI Heat Atlas, or by Eq. (6), the value of h at other heat fluxes can
be determined by:
m
q"
0.3
p
h href where m 0.9 0.3 (8)
q"
ref pcrit
DiMarco and Grassi [10] found good agreement with this correlation for low and
intermediate heat fluxes on a wire, both in normal and in reduced gravity, indicating that
nucleate pool boiling performance on a wire is not affected by gravity, except for a
reduction in CHF. For pool boiling on plates in microgravity, however, a different value
of the exponent m seems to be necessary [11].
CHF is also substantially affected by microgravity. In 1-g environments, Bo has been

used as a correlating parameter for CHF. Zubers [12] CHF model for an infinite
horizontal surface assumes that vapor columns formed by the merger of bubbles become
unstable due to a Helmholtz instability, blocking the supply of liquid to the surface. The
jets are spaced D apart, where
1/2

D 2 3 2 3L Bo 1/2 3c (9)
g l v
and is the wavelength that amplifies most rapidly. The critical wavelength, c, is the
wavelength below which a vapor layer underneath a liquid layer is stable. For heaters
with Bo smaller than about 3 (heaters smaller than D), the above model is not
applicable, and surface tension dominates. Bubble coalescence is thought to be the
mechanism for CHF under these conditions. Small Bo can result by decreasing the size
of a heater in earth gravity, or by operating a large heater in a low gravity environment.
In microgravity environments, even large heaters can have low Bo, and models based
on Taylor and Helmholtz instabilities should not be applicable. If true, this has a strong
physical implication, since heaters that are considered large in normal gravity may
become small as gravity decreases. Zubers correlation
"
qmax 0.131v1/2 h fg 4 g l v (10)
predicts a vanishing heat transfer as the acceleration approaches zero, as does the
similar correlation of Lienhard and Dhir [13]. The macrolayer model of Haramura and
Katto [14] is of similar form, and also predicts a vanishing heat transfer. Recent
experimental data suggests that Bo is not the only quantity that determines whether
buoyancy or instability controls the boiling process. For example, Di Marco and Grassi
[15] studied boiling on a 0.2 mm diameter wire using R113 and FC-72 in low gravity.
They observed that for a given R, defined as
R
R Bo , lL

(11)
lL g l v
earth gravity data showed a higher CHF relative to that for a flat plate than the
microgravity data. Such findings suggest different non-dimensional groups containing
acceleration and wire diameter are needed.
Low-Gravity Facilities
Any free-fall trajectory produces a microgravity environment, since in the non-
inertial reference system the weight is balanced by inertia forces. Two aspects to be
considered are the residual level of gravity, which may range from 10-2 to 10-5 g: 750s
of very good quality microgravity is obtained (down to 10-5 g, with very low g-jitter),
and the so called g-jitter, i.e. the noise oscillation in gravity level at different
frequencies. Both parameters vary according to the carrier. Low gravity environments
have been produced using drop towers, aircraft, sounding rockets, and through orbiting
platforms (satellites, Space Shuttle, and the International Space Station). Drop towers
are the lowest cost method of producing good quality low-g environments of limited
duration. Numerous drop towers are available for use throughout the world. The Zero
Gravity Research Facility at NASA Glenn Research Center currently provides the
longest freefall distance (132 m, 5.18 s) after the closure of the 10 s JAMIC dropshaft in
Hokkaido, Japan. The MGLAB 4.5 s drop tower directed by the Science and
Technology Agency in Japan has been available since the mid 1990s. The Bremen Drop
Tower in Europe provides 4.76 s of microgravity (10-5 g) time up to three times per day,
but this time can be doubled to 9.3 s through use of a catapult at the bottom of the
towerthis facility currently provides the longest drop tower test duration.
Aircraft flying approximately parabolic trajectories can provide low-g
environments for 20-25 seconds 30-40 times per flight, but with a relatively high g-jitter
of 10-2 g unless the test package is free-floated. Low-g aircraft platforms are available
for use through Diamond Air Service in Japan (Mitsubishi MU-300), ESA in Europe
(Airbus A300), and NASA in the US (McDonnell Douglas C-9B).
Sounding rockets are single or two stage launcher fired on a nearly-vertical sub-
orbital trajectory, to an apogee up to several hundreds of kilometers. They are mainly
dedicated to measurements in earth science, heliophysics and astrophysics; however,
during the free-fall part of their trajectory up to 750 s of very good quality microgravity
is obtained (down to 10-5 g, with very low g-jitter). Active programs in the European
frame are the MAXUS launcher, where the microgravity time can attain 15 min, with a
payload of 800 kg, and the TEXUS and MASER carriers, both providing about 6 min of
microgravity for a payload of about 400 kg; a similar program (TR-1A) was operated by
JAXA seven times, with an apogee of 260 km. The payloads can be as large as 750 kg
and 4 to 5 experiments can be supported simultaneously. The NASA Sounding rocket
program provides many launches per year, with some flights dedicated to microgravity
science (http://sites.wff.nasa.gov/code810/index.html). In most of these facilities,
interaction and control of the experiment is allowed via telecommands and real-time
video images transmitted to/from ground.
Spacelab, the Space Shuttle, and other orbiting satellites have been used to
conduct boiling experiments. The Space Station is expected to become the facility of
choice for long-duration, high-quality boiling experiments. Boiling experiments are
currently planned for the Microgravity Glovebox in the US module, Japans KIBO
module, and the Fluid Science Laboratory (FSL) in the European Columbus laboratory.
Pool Boiling in Low Gravity

Many of the early experimental studies regarding boiling heat transfer in
microgravity environments were first performed under NASA sponsorship in drop
towers during 1960s. The results of these early experiments were somewhat con-
tradictory, with some experiments showing no effect of gravity on heat transfer and
others showing a strong dependence. Much of the discrepancy can be attributed to the
relatively short test times that were available since natural convection from before drop
initiation could not be eliminated during the short drop time. Visual observations of the
boiling process, however, revealed that a large increase in bubble size (up to a few
millimeters) occurred under microgravity conditions, with small bubbles coalescing into
larger bubbles a small distance from the heater. Siegel and Keshock [16], for example,
-1/3
found the bubble departure radius varied approximately as a for 0.1<a/g<1, and
-1/2
according to a for lower gravities. Thin wires have been used for their low thermal
inertia allowing for fast transients like in parabolic flights, and for simple data
conditioning (they can be used as resistance thermometers). Plates (up to 50 mm
diameter) are more difficult to operate, but have more practical significance for
applications; finally, small heaters were adopted to investigate individual bubble
behavior. The reader is referred to Siegel and Keshock [16], Clark [17] and Straub [18]
for a review of the early literature. A review of the work through 2000 is provided by Di
Marco and Grassi [10].
Bubble dynamics
The effects of low gravity on the dynamics of single bubbles using water were
studied by Qui et al. [9]. A single artificial cavity that served as nucleation sites was
micromachined into a silicon wafer that was heated from the backside by numerous
strain gauge heating elements kept at a set temperature. Measurements of the bubble
shape vs. time were obtained from inception to departure under saturated and slightly
subcooled conditions during the low gravity environment provided by an aircraft. The
bubbles grew to a much larger size in low gravity compared to earth gravity (~2.5 mm
in earth gravity up to 22 mm in low gravity). The bubble growth time also increased
significantly. Because the bubbles grew slowly, bubble departure could be predicted
using a force balance between surface tension and buoyancy forces just prior to bubble
departure when gx and gy were small compared to gz . The departure diameter was
found to scale with g0.5
z , consistent with the results of Siegel and Keshock [16] and the
Fritz [20] correlation. Similar scaling was found from numerical simulations of the
bubble growth by Son et al. [21]. The bubble growth time was found to scale with g1.05
z
from the experiments, which agreed well with the numerically obtained values of g0.93z
between 0.01 gz / ge 1.8. These results are for cases where the bubble size is much
larger than the superheated liquid layer thickness for the majority of the bubble growth
period. Subcooling had a negligible influence on the bubble departure diameter, but
strongly influenced the bubble growth rate. Sliding bubbles occurred when
gxy gx2 gy2 was similar to 0.01 ge . The asymmetric forces on the bubble caused the
bubble to distort, and the shear-lift force that developed resulted in a smaller departure
diameter.
Local heat transfer measurements in low-g during single bubble dynamics and
merger of two bubbles were studied by Qui et al. [22]. Five cavities micromachined into
a silicon wafer provided nucleation sites (only two were active in this study). The power
supplied to the heating elements were monitored during the growth, merger, and
departure process. The power supplied to the heating elements directly under the cavity
for the single bubble study were observed to increase during bubble nucleation,
decrease to a low value indicating dryout, then spike at bubble departure as liquid
rewetted the surface just after liftoff. When two cavities were activated, the heat transfer
under one of the cavities increased during bubble growth, decreased as dryout occurred,
then increased at liftoff. Similar behavior was observed for the other cavity during
growth and dryout, but the heat transfer remained low after merger, indicating continued
dryout at that cavity. Abarajith et al. [23] used the level set method to numerically
simulate the single bubbles. Both constant gravity level and the gravity level measured
in the aircraft were used in the simulations. Good agreement was observed with the
observed bubble shapes, especially when the measured gravity was used. The bubble
was observed to depart when it reached the lift-off diameter predicted by the Fritz
correlation using the gravity at that instant.
The group of IUSTI Marseille, in collaboration with MRC laboratory in Brussels,
designed a 3-D experimental set up and a 2-D Hele-Shaw cell to study pool boiling on a
single artificial site. The bubble was created on the upper downward facing element
with an artificial nucleation site. In the 2-D apparatus, the bubble was confined between
the lateral walls. This simple configuration permitted the quantitative evaluation of the
volume, contact line diameter and wetting angle. In the 3D configuration, the bubble
growth radius varied as time power n (n varied with sub cooling level). This was also
found in 2-D configuration. The exponent n tended to in the two cases where the
subcooling was small. The associated wall heat transfer components were attributed to
several phenomena. The heat transfer linked to the phase change was found to be small
compared to the total wall heat flux, whereas the largest heat transfer was due to the
convection induced by the bubble dynamics (growth and detachment from the wall).
The wetting angle tended to a constant value as soon as the depinning effect occurred,
leading to contact line moving at a constant velocity [24].
Stephan and coworkers [25] set up an apparatus for the study of nucleation of a
single bubble in variable gravity. The heater consisted of a thin stainless steel foil with
an artificial nucleation cavity (20 m diameter, 20 m depth) in contact with FC-72.
Infrared, high-resolution, high-speed thermography was used to obtain the temperature
distribution on the back side of the foil, and high-speed video of the bubble growth was
taken from the side. The local heat flux could be derived by means of an ad-hoc
numerical inverse technique. Bubble detachment diameter and frequency, wall
temperature and heat flux were measured in the variable gravity phase of parabolic
flight, from hypergravity (1.8 g0) to 0.02 g0, see Fig. 3. In microgravity, vapor bubbles
did not detach from the heated surface during the experiment. The expression of
gravitational influence in the correlations of Fritz for departure diameter and of
Malenkov in conjunction with Fritz for bubble frequency was compared to the
experimental data: good agreement was found. The evaluation of the heat flux
distribution indicated high heat transfer rates at the three phase contact line and an
almost adiabatic zone of the heating wall inside the bubble. The heat flow at the contact
line was compared to the overall heat flow into the bubble: at the early state of the
bubble growth, the heat flow ratio was independent of subcooling and gravity.
t=0 ms t=4 ms t=8 ms t=12 ms t=16 ms
1 mm
1 mm
1 mm
Figure 3. Bubble shape, temperature field and calculated local heat flux vs. time [25]
Nucleate pool boiling

The heat transfer mechanisms for steady state pool boiling in reduced gravity have
been described by several authors (e.g. Kim et al. [26], Lee et al. [27], Oka et al. [28]),
and there is substantial agreement. A large bubble is attached to or resides at a short
distance from the heater and acts as a reservoir, engulfing bubbles forming on the
surface, see Figure 4. This large bubble maintains its size due to the balance between
condensation at its cap and coalescence of new, small bubbles at the base. Lateral
coalescence of bubbles along the surface was observed, with consequently induced
motion in the fluid, causing small oscillations in the heater temperature. The dimensions
of the surface may affect liquid renewal under the large bubble and thus the possibility
of maintaining steady state conditions. The probability of having steady state conditions
increases with subcooling. Both enhancement and degradation of performance with
respect to terrestrial conditions were encountered, as detailed in the following.
liquid liquid
inflow SLUG inflow
heated surface
Figure 4. Saturated or quasi-saturated boiling pattern encountered in microgravity over a flat

plate.
Lee et al. [29] described the results of boiling on a flat plate heater obtained in the
long duration microgravity provided by the Space Shuttle. Data was collected at three
heat fluxes (2, 4, and 8 W/cm2) and three subcoolings (0, 2.7, and 11 C) during three
space experiments (STS-47 in 1992, STS-57 in 1993, and STS-60 in 1994) for a total of
27 data runs. A large bubble formed during the initial nucleation, and could hover just
over the heater or stay attached depending on how rapidly it grew initially. This bubble
served as a vapor sink for the smaller bubbles generated on the surface in either case
and allowed liquid to rewet the wall, resulting in higher than expected heat transfer. The
size of this bubble remained roughly constant, indicating a balance between
condensation at the bubble cap and vapor addition by coalescence with small bubbles at
its base. The heat transfer coefficient for the lower wall heat flux was about 30% higher
in microgravity than in earth gravity, and was speculated to be due to the increase in
size of the bubbles on the surface. Transient dryout occurred at 8 W/cm2 for all the
subcoolings tested, and was usually associated with an increase in the dry area on the
heater. Quasi-steady nucleate boiling sometimes occurred until the size of the large
bubble increased until contact with the wall opposite the heater was made, inducing
dryout on the heater. Based on the transient data obtained, CHF was estimated to occur
between 4-6 W/cm2, about 1/3 of the earth gravity values. For a given heat flux, lower
wall temperatures were observed for higher subcooling.
On the basis of his experiments using sounding rocket (R113 on wire and flat
plates) [30], Get Away Special payload on the Space Shuttle (R134a on wires) [31, 32],
and the BDPU facility (R123a on hemispherical heaters) owned by ESA [32, 33], Straub
identified three basic boiling configurations: saturated or slightly subcooled boiling
characterized by bubble detachment and coalescence, subcooled boiling with bubbles
adherent to the surface acting as heat pipes and strong thermocapillary flow along their
surface, and highly subcooled boiling with very small bubbles growing and collapsing
very quickly with no thermocapillary flow. The main outcomes of this research have
been summarized in Straub [4]: a) the enhancement or degradation are limited ( 30%
max); b) the heat transfer efficiency (see Eq. 2) always decreases with increasing heat
flux but can be greater than unity at low heat flux and lower than unity at high heat flux;
c) drops off at high reduced pressure. The role of bulk subcooling is complex. For
wires, subcooling seems not to appreciably influence the value of , while in small
heaters decreases with increasing subcooling. Finally, in a flat plate the degradation
reduces with increasing subcooling.
Abe and Iwasaki [34] conducted pool boiling experiments with n-pentane in a
sounding rocket (TR-1A No.1). A transparent heater made of ITO coated glass was used
to observe the bubble behavior from the underside. The surface had artificial cavities to
promote nucleation. A Fizeau interferometer was used for detailed observation of the
film under attached bubbles. They confirmed the existence of a stationary microlayer at
low heat flux. Bubble coalescence occurred also in reduced gravity consistent with other
investigators, and they observed a large single bubble surrounded by many small
bubbles. Oka et al. [35] confirmed that small bubbles newly formed under a large
coalesced bubble were successively swallowed into the coalesced bubble resulting its
further growth. Ohta et al. [36] made another attempt to clarify the bubble structure. A
transparent surface with a large heating diameter of 50 mm was developed and housed
in a cylindrical boiling vessel of 120 mm inner diameter. They observed a large
coalesced bubble that eventually grew to become the same diameter as that of the vessel
for saturated water at 0.1MPa and a heat flux q " =30 W/cm2 under reduced gravity
conditions created by the MU-300 aircraft. A macrolayer of 10-4 m order thickness was
formed and small bubbles were observed to generate and collapse on the surface of the
macrolayer. No "vapor stems" (see Gaertner [37]) were observed in the macrolayer.
Tanaka-Nishio [38] used a narrow vertical gap to confine the boiling to two-dimensions,
and investigated the structure of a coalesced bubble. Ohta et al. [39] reported on the
bubble structure for saturated ethanol at 0.01MPa in the environment produced by the
TR-1A No.5 sounding rocket. The coalesced bubble grew to a diameter of about 100
mm and was surrounded by smaller nucleating bubbles with a maximum diameter of 10
to 30 mm. Microlayers and dry patches were observed at the base of the nucleating
bubbles, and the compound structure of a coalesced bubble was confirmed. The
microlayer thickness was measured at Tsub =0 K, P=0.01MPa. Based on variation in
film thickness with time, the change in local heat flux was predicted by transient
conduction across the microlayer. The values agreed well with the local heat flux
evaluated from the conduction across the heating surface substrate (Ohta et al. [40]).
Ohta [36] also measured the macrolayer thickness using sensors directly coated on
the transparent substrate of a heating surface with water as the test liquid onboard the
MU-300 aircraft. The macrolayer thickness decreased with an increase of the heat flux
in reduced gravity, similar to what was observed in earth gravity (Bhat et al. [41],
Gaertner [37]). The heat transfer due to conduction across the macrolayer was limited to
30% even at the heat flux just below burnout ( q " =27 W/cm2, Tsub =0 K, at
P=0.1MPa). The remaining heat flux was transferred by through nucleate boiling in the
macrolayer, indicating the dominant role of small bubbles to the heat transfer at high
heat flux.
Merte et al. [42] and Lee [43] describe results from two additional space flights
(STS-72 and STS-76 in 1996). The heat fluxes on STS-72 were similar to those on the
earlier three flights, but the subcooling was increased to a maximum of 22C. The
subcooling levels on STS-77 were similar to the earlier three flights, but the heat flux
was decreased to a minimum of one-fourth that previously used. They found that the
large vapor bubble above the heater tended to remain closer to the heater surface as the
subcooling increased, and this was attributed to condensation at the bubble cap and the
formation of Marangoni convection at higher subcoolings. Marangoni convection
caused flow of liquid from the base of the bubble to the bubble cap causing the bubble
to be impelled towards the surface. The bubble served as a sink for the small bubbles
generated on the surface promoting heat transfer, but it also caused partial dryout on the
heater, increasing the surface temperature. Large spikes in heat transfer were
occasionally observed when dual large bubble formed on the heater, then coalesced to
form a single large bubble that then became somewhat removed from the surface.
Marangoni convection was also thought to cause the observed migration of the
nucleating bubbles from their location of origin towards the coalesced bubble at high
subcooling (16.7-22C) and a heat flux of 2 W/cm2. Bubble velocities of 2 cm/s were
common. The heat transfer coefficient when bubble migration was present was higher
than that for 4 W/cm2 at the same subcooling. A heat transfer enhancement of up to
40% was attributed to bubble migration. Nucleate pool boiling curves that summarize
the results of their space flights are shown on Figure 5. Higher heat fluxes occur in
microgravity compared to earth gravity at lower superheats, but the CHF is about half
that in earth gravity except for the high subcooling case in microgravity.
10
a/g=+1, 0g subcool (22.2 C)
Tsub=22.2 C, 16.7 C, 11.1 C, 0.3 C
9 0g subcool (16.7)
-4
0g subcool (11.1)
a/g=10 ,
8 0g subcool (2.7)
Tsub=22.2 C,
0g subcool (0.3)
7 +1g subcool (22.2 C)
+1g subcool (16.7)
6 a/g=10-4, +1g subcool (11.1)
q(W/cm 2 )
Tsub=11.1 C +1g subcool (2.7)

5 +1g subcool (0.3)
a/g=10-4, Tsub=0.3 C
4
1
Natural Convection * Partial Dryout
0
0 10 20 30 40 50 60 70 80
TW =TW-Tsat ( C)
Figure 5: Nucleate boiling curve from Lee [43].
Kim et al. [26] used a microheater array consisting of ninety-six individual heaters
each 0.27 x 0.27 mm2 in size to measure time and space resolved heat transfer during
subcooled boiling of FC-72 using a parabolic aircraft to provide low gravity and high
gravity environments. Their use of electronic feedback circuits to keep each of the
heaters in the array at a constant temperature enabled them to obtain heat transfer in the
nucleate boiling, CHF, and transition boiling regimes. The heater design also allowed
images of the boiling process to be obtained from below. Data was obtained at wall
superheats as high as 45 K and subcoolings between 7-34 K. The formation of a
coalesced bubble surrounded by many smaller nucleating bubbles was observed, similar
to the observations of other researchers, despite the large difference in heater size.
Dryout occurred under the coalesced bubble, causing CHF in low gravity to be
significantly lower than in earth gravity. The restriction in the detachment of the
coalesced bubble and its larger size prevent the liquid supply to the bottom of the
coalesced bubble under reduced gravity conditions. Boiling curves showing the effect of
subcooling and gravity are shown on Figure 6. Subcooling is observed to have a strong
effect on the size of the coalesced bubble and CHF increases with higher subcooling
and gravity level. By sampling the heat transfer from each individual heater in the array
only when nucleating bubbles were present on that heater (nucleating bubbles in low
gravity, the bubbles that form everywhere on the heater during nucleate boiling in earth
gravity, or when liquid contacted the surface during transition boiling in earth and high
gravity) it was possible to measure the nucleate boiling heat flux. It was shown that
the nucleate boiling heat flux collapsed onto a single curve, indicating that the small
scale boiling was independent of subcooling and gravity level. This suggests that if one
is able to predict the extent of the dry area in microgravity (or Lunar and Martian
gravity), then one can predict the microgravity boiling curve from earth gravity boiling
data. It also suggests that CHF in low gravity simply results from the competition
between increasing heat transfer from the satellite bubbles and the increase in the dry
area under the coalesced bubble.
Figure 6: Nucleate boiling curve from Kim et al. [26].

Raj, Kim, and McQuillen [44] obtained boiling heat transfer data over a continuous
range of gravity levels 0 g1.8 g under subcooled liquid conditions (n-perfluorohexane,
Tsub=26 K, and 1 atm), two gas concentrations (220 ppm and 1216 ppm), and three
heater sizes (full heater-7 x7 mm2, half heater-7 x 3.5 mm2, and quarter heater-3.5 x 3.5
mm2) using the ESA Airbus A-300 aircraft. The transition between high-g and low-g
was slow enough that the heat transfer was quasi-steady in the nucleate boiling regime.
As the gravity level changed, a sharp transition in the heat transfer mechanism was
observed at a threshold gravity level (Figure 7). Below this threshold (low-g regime), a
nondeparting coalesced bubble governed the heat transfer and the effect of residual
gravity was small. Above this threshold (high-g regime), bubble growth and departure
dominated the heat transfer and gravity effects became more important. An increase in
noncondensable dissolved gas concentration shifted the threshold gravity level to lower
accelerations. Heat flux was found to be heater size dependent only in the low-g regime.
The slope of the heat flux versus acceleration curve was different in the two regimes,
ruling out the possibility of a unified power law dependence across all gravity levels.
Figure 7: Example of heat flux vs. acceleration graph illustrating the sharp transition in heat flux
below a certain gravity level.
Recently, Kawanami et al. [45] conducted pool boiling experiments on the ESA A-
300 aircraft. They employed a transparent heating surface where 88 pairs of temperature
sensors and small heaters were directly coated on a 40mm diameter circular area of a
sapphire substrate, and investigated the relation between the liquid-vapor interface and
local distribution of heat transfer coefficients. A preliminary analysis indicated a
deterioration in heat transfer in reduced gravity based on the changes in temperature
distribution.
Suzuki et al. [46] conducted experiments onboard a Gulfstream II aircraft to clarify

the effect of gravity on CHF for water at P=0.1MPa under subcooled liquid conditions
(Tsub =10- 40C) using a thin stainless steel plate of 205 mm2. The values of CHF in
reduced gravity were much smaller than those in normal gravity, but larger by 200 to
400% than the values predicted by Zuber [12]). CHF increased with increasing liquid
subccooling, similar to what is observed in normal gravity (e.g. Ivey-Morris [47]).
However, the residual g-jitter of 0.01-0.03 g during the flight may have resulted in
higher CHF values. Suzuki et al. [48] introduced a transparent heater to observe the
bubble behaviors from underneath in the JAMIC drop tower. Heating was started 10 s
before the drop. Under high subcooling and high heat flux conditions for water, the
bubble contact area instantaneously increased when the drop was started, but the
bubbles were detached from the surface to form coalesced bubbles just above the
heating surface resulting in almost constant bubble contact area during the drop of 10s.
The reduction of CHF values in reduced gravity was evident from the boiling curves
(Oka et al. [1]). It is noteworthy that MEB (Micro-bubble Emission Boiling) was also
observed under reduced gravity conditions with high liquid subcooling.
Di Marco and Grassi carried out experiments of pool boiling with imposed heat flux
of FC-72 on a flat, semi-transparent heater 2020 mm2 in orbital flight (ARIEL
experiment) [35]. Side and bottom views of the boiling phenomena were collected; the
usual formation of a coalesced bubble of a size comparable with heater size was
observed. The nucleate boiling heat transfer was degraded, and a boiling transition was
identified (Figure 8). This transition, however, was not associated with a temperature
jump, but rather to a change in slope of the curve. A flow pattern analysis revealed that
a change in behavior of the coalesced bubble occurs: after the transition, the bubble
resides steadily in the proximity of the surface; before the transition, the bubble
periodically hovered from the surface and was replaced by fresh liquid, and the
oscillations in wall temperature are well correlated with the patch extension (Figure 9).
200
150
q",(kW/m2)
100
50
0
0 40 80 120
T sat , (K)
Figure 8: Nucleate boiling curve in normal gravity (squares) and microgravity (circles),
from [11]
49
48
47
46
45
Tsat [K]
44
43
42
41
40
39
1003298 1003299 1003300 1003301 1003302 1003303 1003304 1003305
t [s]
Figure 9: Interrelationship between flow pattern and wall temperature, from [49].
Di Marco and Grassi [50] conducted experiments of pool boiling of R113 and FC-72
on a 0.2-mm platinum wire, in slightly subcooled conditions, both in parabolic flight
and in sounding rockets. High values of heat flux up to CHF were tested. Data were also
recorded in the enhanced gravity phase of trajectory and during special trajectories
resulting in a constant gravity value of 1.5 g for 40 s. Pool boiling data at Martian
gravity level (0.4 g) were also collected. Despite a very evident change in bubble size
and velocity, no appreciable effect of gravity on the heat transfer coefficient in nucleate
boiling on a wire was found. Critical heat flux (CHF) was clearly detected and it was
found to be reduced of about 50% in low gravity; analysis of high speed images
indicated that CHF was triggered by bubble coalescence along the wire (Di Marco and
Grassi [51]). For R > 0.08, the CHF data obtained in reduced or enhanced gravity was
in the range of most of the other experimental data. The Bond number (or alternatively
its square root, R, see Eq. 9) seems to be a suitable parameter to scale the effect of
gravity for R > 0.08. The situation appears to be drastically different for R < 0.04: the
data obtained in micro-gravity are very well separated from those of thinner wires,
corresponding to the same value of R, obtained in normal gravity. At low values of the
Bond number R is no more suitable to scale critical heat flux, as it could be inferred by
extending the model of Lienhard and Dhir [13] to low gravity conditions, and the effects
of gravity and size have to be accounted for separately [50]. Straub [4] found that his
own CHF data for wires in parabolic flight and orbital flight were consistently lower
than the corresponding terrestrial values, and could be successfully correlated if the
constant 0.131 in Zubers correlation is replaced by 0.94 R 0.25 for values of R as low
as 5x10-4. However, no fitting of CHF values at different gravity levels was made.
Asano et al. [52] compared nucleate boiling characteristics of enhanced surfaces
with those of a smooth one under different gravity conditions. The surface roughness of
copper rods with a diameter of 20mm was increased by the coating of copper grains
through the vacuum plasma spraying method. By using HCFC123 at saturation
temperature of 30C at near atmospheric pressure, heat transfer coefficients in nucleate
boiling were measured in a wide range of heat flux up to 160 W/cm2 along a parabolic
trajectory of aircraft. The rods were horizontally located in a pool of test liquid under
normal gravity. The experimental results show no heat transfer deterioration for the
coated surface even at high heat flux as was usually observed for enhanced surfaces
having porous micro structures. The heat transfer coefficients in nucleate boiling for the
coated surface is enhanced by three to four times from those of smooth surface in both
normal gravity and microgravity.
The research carried out so far worldwide allowed a suitable qualitative picture of
pool boiling in microgravity to be obtained. The small-scale boiling surrounding the
coalesced bubble provides the dominant contribution to the overall heat transfer in
boiling. Remarkably, the governing phenomena in this region, namely intermolecular
forces of adsorption, capillary forces, molecular interfacial phase change resistance and
change of phase equilibrium, are independent of gravity. This was experimentally
confirmed in parabolic flight experiments by Kim et al. [26], who found that the small
scale boiling, i.e. heat transfer in the zones where only small, uncoalesced, nucleating
bubbles exist, was independent of gravity level and subcooling. Straub [53, 4] also
divided boiling mechanisms into primary mechanisms which are independent of gravity
and determined by evaporation and capillary forces in what he calls the microwedge
underneath the bubble, and secondary mechanisms which are responsible for mass and
energy transport away from the surface and include buoyancy, coalescence processes,
momentum transfer due to bubble growth and formation, and thermocapillary flow for
subcooled states. Gravity thus plays a role in the macroscale, for convective removal of
energy away of the layer, and can largely be replaced by the other secondary
mechanisms. Secondary mechanisms influence the number and size of the bubbles
residing close to the surface. When the vapor fraction is larger, vapor may adhere to the
heated surface leading to substantial heat transfer degradation and eventually to CHF.
Gravity may affect the two preceding factors in ways which depend on subcooling and
heater geometry, and this may explain why both enhancement and degradation of heat
transfer in microgravity were reported. As a consequence, a correct prediction of the
nucleation site density and of the bubble detachment diameter associated with a gravity-
independent bubble growth model (e.g. Dhir [54], Bai and Fujita [55], Stephan and
Hammer [56]) may lead to a substantially correct prediction of boiling heat transfer in
any gravitational field, at least for low to intermediate heat fluxes.
Marangoni convection
The role of thermocapillary convection in boiling has still to be completely
assessed, and it is expected that this mechanism will becomes more important in the
absence of buoyancy. Straub points out that thermocapillary flow around the bubbles
was observed in reduced gravity in subcooled boiling, but never in saturated conditions.
The origin of this flow has yet to be understood. Marek and Straub [57] suggest that it is
due to a gradient of dissolved gas concentration along the bubble interface. Straub [58]
also stresses that quantitative measurements indicate that the contribution of
thermocapillary convection to overall boiling heat transfer is small, however he gives
large evidence that thermocapillary flow is an important mechanism for transporting
energy away from the liquid-vapor interface into the bulk fluid (and not from the heated
surface to the fluid). In the absence of buoyancy and subcooling, like in reduced gravity
saturated boiling, the energy transport into the bulk of the fluid uniquely relies on
bubble motion.
Betz and Straub [59] numerically modeled the Marangoni convection around a gas
bubble in microgravity and 1g, and found that strong enhancement of single-phase
convection around the bubble can be obtained in microgravity if the cell size is
comparable with the bubble diameter. The model compared well with experimental
results and the relevant dimensionless parameters were identified. The authors
appropriately point out that Marangoni convection is strongly influenced by surface
contamination, so the purity of the fluid or its degradation in time may substantially
affect the experimental behavior of the system.
Tadrist and coworkers studied the thermocapillary convection around single
bubbles in different liquids with or without liquid-vapor phase change by means of two
analogue experiments in the cells mentioned before [60, 61]. The occurrence threshold
of the 3D oscillatory thermocapillary convection was characterized. In the case of the
silicone oil (Pr=228), it was found that only the stationary state existed for the operating
conditions tested. For the two other liquids (silicone oil Pr=16.7 and FC-72), an
unstationary state also existed and the critical threshold was defined using the critical
radius and a critical temperature gradient. The system was tested also in reduced gravity
in parabolic flight [62].
Henry et al. [63] and Henry, Kim, and McQuillen [64], obtained data during boiling
of >99.3% pure n-perfluorohexane with and without noncondensible gas in a low-g
environment using a 7.0x7.0 mm2 microheater array. In their results, however,
thermocapillary convection around the coalesced bubble that formed in the gassy fluid
was found to be much weaker than in the degassed fluid, and the coalesced bubble
diameter was much larger in the gassy fluid due to the accumulation of noncondensible
gas within the bubble. The results suggest that the accumulation of noncondensible gas
in the bubble can result in temperature variations along the interface but due to the
increased vapor/gas bubble size, the driving thermocapillary temperature gradient along
the interface is significantly reduced. This results in much weaker thermocapillary flow.
The highest CHF values in a reduced gravity environment (19 W/cm2) occurred when
the fluid was highly subcooled and degassed.
Transition boiling
Xu and Kawaji [65] were perhaps the first to study transition boiling in
microgravity. The surface-averaged heat flux and liquid-solid contact frequency were
measured with a fast response, 25.4 mm diameter heat flux gauge built onto a stainless
steel plate during a quench with PF-5060. The low gravity environment was obtained
using a DC-9 aircraft. They observed that the frequency of liquid re-wetting in low
gravity was significantly lower than that in earth gravity. Vapor film collapse and
spreading of the liquid film were thought to be the dominant modes of liquid-solid
contact. In earth gravity, both modes are equally significant, but vapor film collapse
became less significant in low gravity resulting in lower contact frequencies. The heat
flux was observed to fluctuate by 101-102 W/cm2 during the quench.
Henry and Kim [66] obtained post-CHF boiling data for FC-72 using the same
microheater array used by Kim et al. [26] on the KC-135. At low subcooling, the heat
transfer after CHF monotonically decreased with superheat due to vapor covering an
increasingly larger fraction of the heated area. At higher subcoolings, however, they
observed a decrease in heat flux with superheat immediately after CHF, followed by a
sharp increase due to the formation of a strong Marangoni convection around the
coalesced bubble. In fact, the heat flux levels reached in the post-CHF region were
higher than at CHF. Visual observations indicated that Marangoni convection caused
colder fluid to come into contact with the coalesced bubble cap decreasing its size,
resulting in higher wall heat transfer.
Boiling of mixtures
Concentration gradients can arise at the heated surface during boiling of mixtures
due to preferential evaporation of the more volatile component. This can lead to
Marangoni convection and pumping of liquid to the wall, delaying the onset of CHF.
Abe et al. [67] tested ethanol-water mixtures in the JAMIC drop tower. Non-
azeotropic concentrations of 11.3 wt% and 27.3 wt% ethanol were selected. The
mixture was "positive", i.e., the surface tension decreased with increasing concentration
of the more volatile component. Enhanced bubble detachment was observed for the
mixtures. The structure of the coalesced bubble was reported to be the same as that of
the pure liquid. The heat transfer was enhanced for the mixture throughout the entire
heat flux range in low gravity compared to pure fluids. The reduction in CHF values for
the mixture was only 20-40% of the terrestrial values, which is small compared to that
for pure liquids. Abe and Iwasaki [68] observed bubble growth of a CFC12-CFC112
mixture by using a two-wavelength interferometer to separate the effects of
concentration and temperature on the optical index. They concluded that the Marangoni
effect due to temperature distribution along the interface was dominant for pure
CFC113, while the effect due to concentration gradient was dominant for this mixture.
Thick temperature and concentration boundary layers nearly one order of magnitude
larger than those due to thermal and mass diffusion indicated the existence of induced
flows around a bubble due to the Marangoni effect.
Ahmed and Carey [69] studied the effects of gravity on the boiling of a non-
azeotropic binary mixture of water and 2-propanol with low subcooling using the KC-
135. Boiling curves were obtained at various gravity levels (0.01 g, 1 g, and 1.8 g) and
mixture concentrations (2-propanol concentrations of 0.015, 0.025, and 0.1). For a given
gravity level, the nucleate boiling heat transfer was similar for all mixtures. CHF varied
significantly, however, with the highest CHF observed for a 2-propanol concentration of
0.015. CHF was found to correlate directly with the surface tension gradient at the base
of the bubble. All mixtures showed much higher CHF values than for pure water.
Although visual observations were not given, the high CHF was attributed to the
development of strong Marangoni convection around the bubble that helped maintain a
liquid layer on the surface, similar to the results obtained by Abe et al. For a given
concentration, similar nucleate boiling curves were observed in low gravity and earth
gravity, and a slight enhancement was observed for high gravity. CHF values for low
gravity were only 10% smaller than those for earth gravity at a concentration of 0.015.
Correlations commonly used to predict boiling of mixtures and CHF predicted the earth
and high gravity data well, but significantly underpredicted the low gravity data.
Sun and Carey [70] studied gravity effects on boiling of 2-propanol/water mixtures
within the gap formed by a copper heating element 12.7 mm in diameter and a cold
plate. The spacing varied between 6.412.7 mm, and the cold plate temperature was
subcooled by 10-30 K to provide a temperature gradient within the gap. Low gravity
and high gravity data was obtained in the KC-135. In low gravity, nucleation was rarely
found for superheats less than 20 K for pure water and a 6.4 mm gap. When boiling did
occur, a large bubble formed between the gap and transition to a pseudo film boiling
regime was observed where nucleation only occurred at the perimeter of the bubble base
and dry patches formed on the heater. Similar behavior was observed when the molar
concentration of 2-propanol was increased to 0.015. CHF was observed to vary
according to g 0.17 .
Electric field effects

Electric fields have been shown to mitigate the effects of reduced gravity on boiling
heat transfer by providing a body force on the bubbles. Snyder et al. [71] studied the
effects of electric fields during saturated and subcooled (15 K) boiling of FC-72 on a
0.25 mm diameter platinum wire. The wire served as both temperature sensor and
heating element, and was placed between two 25.425.4 mm2 electrodes oriented 13 to
each other to provide a non-uniform electric field. Two electrode spacings were used
such that dielectrophoretic forces (DEP) equal to 1 g and 2 g could be produced at a
voltage of 23 kV. By varying the orientation of the test rig with respect to gravity, the
DEP could assist or oppose gravity. Microgravity data was determined in a 2.1 s drop
tower. They defined an effective gravity ratio g'b,e as the ratio between the sum of
gravity and DEP forces to that of gravity alone. In the nucleate boiling regime, they
found the bubble behavior close to the wire and heat transfer for g'b1,e 0 1 and
'
gb1,e2 1 to be very similar. They also found similar behavior for g'b1,e 0 1 and
g'b 0,e1 1, indicating that the boiling behavior can be determined using an earth
gravity or variable gravity heat transfer correlation if the effective gravity is known.
CHF followed a g0.25 dependence for 1 g'b,e 3, but significantly higher CHF values
were observed as g'b,e became small.
Snyder et al. [71] studied the effects of electrode geometry during boiling of FC-72
on a flat plate heater (25.4 mm 25.4 mm) constructed by sputtering a thin gold film
onto a Pyrex substrate. Flat plate, pin, and diverging-plate electrodes were used to
provide the DEP. The DEP distribution was observed to have a very strong influence on
the bubble behavior. Strong bubble motion was observed directly underneath the pin
electrode where the DEP force was strongest, but the bubble motion quickly decayed as
the radial distance from the electrode increased. The flat plate electrode produced strong
bubble motion at the edges where the electric field gradient was strong, but a substantial
amount of stationary vapor bubbles were observed directly under the electrode where
the DEP force is uniform. The diverging-plate electrode produced the highest heat
transfer, and pushed the bubbles across the heater surface. The magnitude of the bubble
motion increased with increasing voltage for all electrode geometries. Significant
enhancement in wall heat transfer was observed when the DEP forces were present.
A numerical and experimental study was carried out by Snyder et al. [72-74] to
clarify the effect of electrode geometry on boiling. Numerical simulations were carried
out for infinite and finite diverging-plate electrodes. The numerical results indicated that
the finite electrode was found to produce very large DEP forces at the edges which
caused a large acceleration on the bubble, but this force decreased rapidly. The infinite-
plate electrode produced a smaller but more continuous driving force and resulted in the
furthest bubble transport. The bubble detachment diameters were found to scale with
g'b,e . Potentially more effective electrode geometries were proposed.
0.5
Di Marco and Grassi [50] imposed a cylindrical electrostatic field around a 0.2 mm
platinum wire during pool boiling of R-113 and FC-72 (up to 10 MV/m at the heater
surface) during part of a parabolic flight path. No significant effect was detected on the
heat transfer coefficient, but the imposition of an electric field was found to be effective
in drastically reducing bubble size and increasing CHF in microgravity. At high values
of applied voltage, the same value of CHF as in terrestrial conditions was measured,
thus demonstrating the dominance of the electric force on buoyancy under these
conditions. The value of CHF exhibited a three-fold increase even in terrestrial gravity,
after the application of the electric field. However, existing models of CHF in the
presence of electric field revealed their limitation when applied in reduced gravity [51].
The detachment of bubbles from the heated surface took place in both the presence and
absence of an electric field, the difference being that in the former case the bubbles
slowed down and stopped at a small distance from the surface and started to coalesce.
The experiments were repeated on a sounding rocket flight (MASER-8) and similar
results were obtained in low g-jitter conditions (about 10-5 g) [75].
Di Marco et al. [76] performed experiments of pool film boiling of FC72 on a wire in
parabolic flight, in the presence of an electric field. Variation in the oscillation
wavelength of the film, due to gravity and electric field, was observed. In particular, a
reduction of gravity caused an increase in oscillation wavelength and a decrease in heat
transfer, while the reverse occurred with electric field. Above a certain electric field
threshold, the heat transfer coefficient became insensitive to gravity changes,
demonstrating the dominance of electric forces.
Di Marco and Grassi also performed experiments on a flat, semi-transparent heater
20x20 mm2 in orbital flight (ARIEL experiment), imposing an electric field up to 2
MV/m, while obtaining side and bottom video images of the boiling flow patterns [49].
The electric field was effective in delaying CHF in low gravity and in reducing the
detachment diameter and coalescence of bubbles, thus easing the subsequent
condensation of the produced vapor. For the highest value of the applied electric field,
the coalesced bubble disappeared, the void fraction inside the cell was substantially
reduced [77], and the boiling performance at low and intermediate heat flux was
restored to the same, or even greater, value as in terrestrial gravity [71].
Acoustic field effects

Acoustic fields can also be used to provide a body force in place of gravity to
remove bubbles from a heater wall. Hao, et al. [78] theoretically studied removal of
growing vapor bubbles from a flat wall using acoustic pressure (Bjerknes) forces in the
absence of gravity. Because potential flow was assumed, the wall could be accounted
for by inclusion of an image bubble. The motion of the bubble was influenced by the
imposed sound field and by the attractive force of the image bubble. With the wave
fronts parallel to the wall, it was found that the imposed sound field was not able to
overcome the attractive force imposed by the image bubble, making bubble removal
difficult. They suggested using a sound front parallel to the surface to move bubbles
along the wall where they could be removed or condensed.
Sitter et al. [79, 80] studied the effect of acoustic fields during boiling on a 0.25 mm
diameter platinum wire in FC-72 using a 2.1 s drop tower to provide the microgravity
environment. The wire was placed in a chamber with a high-intensity acoustic sounding
wave (2.6 atm) at 10.18 kHz. The heat transfer for a given wall superheat was always
higher with the acoustic field than without, and the highest heat transfer occurred with
the heater placed at the antinode for both terrestrial and microgravity boiling. When the
wire was placed at the antinode in microgravity, the vapor bubbles were driven off the
wire and accumulated were the pressure node was located. Cavitation caused large
bubbles to break up into smaller bubbles, which then moved about the chamber due to
acoustic streaming. These small bubbles eventually coalesced at the pressure node.
Acoustic streaming was observed to bring colder ambient fluid to the wire, enhancing
heat transfer.
Flow Boiling in Low Gravity

Bubbles can be swept off a heated surface during flow boiling, thereby providing
high heat transfer and CHF levels. At high flow velocities, the heat transfer becomes
independent of gravity, but the velocities above which this occurs is not yet known.
Terrestrial experiments performed with the test section oriented in various directions
relative to the gravity vector have indicated that bubble departure size decreases as the
flow velocity is increased due to additional lift and drag forces on the bubble. Relatively
little work has been performed on flow boiling in low gravity due to the large power
requirements and the long transients required for steady state to be reached.
Flow pattern
Misawa-Anghaie [81] introduced two test sections for boiling experiments, a
transparent square channel of pyrex glass with a thin, transparent heating film for flow
pattern observation and a copper tube wrapped with a nichrome coil for the pressure
drop measurements. Drop tower experiments in JAMIC were conducted for CFC113
flowing in vertical test sections. It was found that the slip ratio under reduced gravity
was less than unity and the pressure drop was larger than predicted by the homogeneous
model due to the enhanced contribution of acceleration resulted from the increase in
void fraction. Kawaji et al. [82] observed flow patterns during flow boiling of subcooled
CFC113 and saturated LN2 in both earth and low gravity, and reported marked
difference in the shape of the liquid droplets in the dispersed flow region. Saito et al.
[83] performed flow boiling experiments using subcooled and saturated water in a
horizontal transparent duct with a concentric heater rod on an aircraft. Under reduced
gravity conditions, the bubbles moved along the heating rod without detachment and
grew and coalesced to become large bubbles in the annulus. The local heat transfer
coefficients along the periphery of the heater rod, however, were quite insensitive to
gravity. Yamada et al. [84] employed a vertical annuli for flow boiling of HCHF123
heated by hot liquid, and observed the process of flow pattern transition from bubbly to
annular flow at low mass flow rate and high heat flux. They reported a decrease in the
pressure drop in the annular flow regime by the reduction of gravity.
Ohta [85] developed a transparent heated tube for flow boiling experiments. The
tube was made from a Pyrex tube with 8mm ID and wall thickness of 1mm to minimize
the heat capacity so measurements could be obtained on a MU-300 aircraft. The heated
length was varied from 20 mm to 260 mm. A thin, electrically heated gold film about
0.01m order thick was deposited on the inside of the tube. Visual observation of the
liquid-vapor behavior could be made through the tube wall, and the film was operated
as a resistance thermometer to evaluate the inner wall temperature averaged over the
entire heated length. From MU-300 aircraft experiments using vertical tubes, followings
were clarified (Ohta [86]). At low mass velocity (G=150 kg/m2s) and low quality, the
bubbles generated from the tube wall grew much larger in reduced gravity, and slid
downstream as they detached from the surface. When the mass velocity was increased
to G=600 kg/m2s, the bubbles detached at a much smaller diameter due the increased
shear force exerted by the flow. At moderate quality, annular flow with a smooth
interface was realized in reduced gravity. A reduction in the frequency and the length of
disturbance waves was observed. At even higher mass velocity or higher quality, the
effect of gravity disappeared and the behavior of annular liquid film was dominated by
the inertia force exerted by the vapor core flow.
Ohta et al. [86-88] used a transparent, parallel plate heating surface with
temperature sensors to evaluate local heat transfer in narrow channels. The experiments
were conducted at very low inlet liquid velocity onboard the MU-300 aircraft. The gap
size was varied from 0.7 mm to 10 mm for the boiling of water (P=0.1 MPa, inlet liquid
subcooling Tsub= 0 to 10 K with a constant inlet liquid velocity of 0.06 m/s). When the
gap size was larger than the bubble diameter in pool boiling in normal gravity, the
coalescence of the bubbles occurred in low gravity and the bubble size increased. For
smaller gaps, flattened bubbles covering almost entire heating area of 3050 mm2 were
observed in reduced gravity. For extremely small gap size, the increased growth rate of
flattened bubbles eliminated the influence of gravity, and a vigorous exchange of
generated bubbles and the bulk flow of liquid was observed on the entire part of local
positions on the heating surface.
Westheimer and Peterson [89] performed a visualization study of R113 using a
glass annular heat exchanger on the KC-135 aircraft. They found that earth gravity
based flow regime maps did not correlate well with the low-g maps, but that all of the
low-g maps produced similar results for calculation of quality, heat transfer coefficient,
and heat exchanger temperature, indicating. The maximum heat transfer was associated
with the transition from bubble to slug flow.
Heat transfer
Little data is available regarding gravity effects on flow boiling due to limited
power and the need to establish steady state during the short low-g duration in drop
towers and aircraft. Ma and Chung [90, 91] studied flow boiling of FC-72 over a
25.425.4 mm2 flat plate heater using a 2.1 s drop tower. The liquid velocity across the
heater varied from 0 to 30 cm/s. Increasing velocity resulted in increased wall heat
transfer with a corresponding decrease in the average bubble size in microgravity. For a
heat flux of 5.1 W/cm2, the average surface superheat in microgravity was almost the
same as in earth gravity when the flow velocity reached 20 cm/s. Similarly, the bubble
generation frequency and bubble shape became similar to those in earth gravity at high
flow rates.
CHF in flow boiling using a 0.25 mm diameter wire in cross-flow with FC-72
subcooled by 26 K was studied by Ma and Chung [92] using a 2.1 s drop tower.
Feedback control was used to set the wire temperature and the power required to do this
was measured, enabling measurements to be made into the transition and film boiling
regimes. The maximum flow Reynolds number was 188 (30 cm/s). CHF for earth
gravity was always larger than for microgravity, but the difference decreased with
increasing velocity. For example, the ratio between earth gravity and microgravity CHF
was 1.9 for Re = 49, but it decreased to 1.16 for Re = 188. Visualization of the flow also
indicated increasing similarity between earth gravity and microgravity bubble behavior
with increasing velocity.
Lui et al. [93] reported the results from a KC-135 aircraft experiment on subcooled
flow boiling of R113 in a horizontal tube with12 mm ID and mass velocity 230-
840kg/m2s. They reported an increase in heat transfer coefficients due to nucleate
boiling in reduced gravity up to 20% from those in normal gravity if subcooling is low.
They attributed the increase to the enhanced local turbulence caused by the movement
of generated vapor bubbles.
Ohta [94] made heat transfer measurements in a vertical tube in the moderate
quality, annular flow region (CFC113, P =0.1MPa, G=150kg/m2s, x0.3) onboard MU-
300 aircraft in Japan. In low gravity, the heat transfer coefficient due to two-phase
forced convection deteriorated by 7% from the value in normal gravity, while it was
enhanced by 25% during the 2g hypergravity phase. The gravity effect disappeared
when the mass velocity was increased to 600kg/m2s. The gravity effect did not affect
the heat transfer at low quality near zero, where nucleate boiling dominated, despite the
drastic change in the size of bubbles as mentioned above. Gravity also did not affect the
heat transfer for x>0.5. Acquisition of critical heat flux data was attempted in
microgravity (Ohta et al. [95]) using MU-300. No appreciable difference in CHF values
were observed for R113 (P=0.1MPa) over the quality range x0.5-0.9, although
inverted annular flow was observed at low quality. These results had considerable
uncertainty due to difficulties in the setting the experimental parameters given the
restrictions of the aircraft.
Ohta et al. [96] performed an analytical investigation to clarify the mechanisms
relating to the gravity-dependent behavior of annular liquid films. A simple model was
first developed for the upward flow in a vertical tube to obtain the velocity and
temperature profiles in the annular liquid film. Three trends were taken into
consideration; a decrease in liquid film thickness due to an increase in the upward liquid
velocity in the absence of gravity, an increase in film thickness due to a reduction of
interfacial shear force exerted by the vapor core flow, and an increase in film thickness
due to a decrease in the liquid flow rate due to disturbance waves. The analysis
indicated an increase in the annular liquid film in reduced gravity for given conditions
(R113, mass velocity 150-600 kg/m2-s, 0.2<x<1.0, vertical upward flow). The heat
transfer was influenced, in general, by the liquid film thickness and the turbulence in the
liquid film. The analysis indicated that the change of the film thickness had little
influence on the heat transfer, while the reduction in the turbulence in the film reduced
the heat transfer due to two-phase forced convention in microgravity. The analysis
qualitatively reproduced the gravity effect on the heat transfer at various mass velocity
and qualities. A more detailed model concentrating on the periodic heat transfer due to
the passage of disturbance waves is required to improve the knowledge on the
phenomena.
The dryout mechanisms at moderate quality were investigated by comparing the
temperature fluctuations at heat fluxes just above and below the critical value (Ohta et
al. [96]). The temperature fluctuation was found to be synchronized with the passage of
disturbance waves that supply liquid to the dry patches extending between the
disturbance waves. The statistical nature of a longer interval of disturbance waves could
become a trigger of temperature excursion. A summary of Ohtas work on flow boiling
under reduced gravity conditions is given in Ohta [97].
Bower and Klausner [98] measured heat transfer using FC-87 in a 5.625.4 mm2
flow channel with one side heated. The channel orientation relative to the gravity
vector could be changed. Data at a given orientation was obtained as a function of the
flow parameter
Ul l We l

l v Re l v
l C pl Tsat
and the Jakob number ( Ja ). A gravity-dependent/gravity independent
v h fg
regime map was developed for 16<Ja<40 and 0.02<<0.06.
Zhang, Mudawar, and Hasan [99] developed an Interfacial Lift-off CHF model to
predict CHF during flow boiling of FC-72 in a 2.5x5.0x10.2 mm3 channel heated on one
side. The model was used to explain the lower CHF values observed in variable gravity
environments (low-g, lunar, Martian, and 1.8g) environments at flow velocities up to
1.6 m/s. Convergence between the low-g and 1-g data for bulk fluid velocities above 1.5
m/s showed that it is possible to design heat exchangers that operated in the gravity
independent regime. Zhang, Mudawar, and Hasan [100] assessed seventeen subcooled
flow boiling correlations to identify which most accurately predicted 1-g and low-g
data. The correlations of Hall and Mudawar [101, 102agreed best with the data, with
Mean Absolute Errors of about 15% when based on inlet conditions and using the
hydraulic heated diameter.
Flow boiling experiments of flow boiling of FC-72 in circular Pyrex tubes of
different diameters (2, 4 and 6 mm inner diameter) were performed by Celata et al.
[103, 104] in parabolic flight campaigns. The tubes were heated externally by an ITO
layer, and the local heat transfer could be estimated from the measurements given by
thermocouples attached externally at different locations, after appropriate corrections

for heat losses and conduction in the pipe wall. The results showed that at low mass flux
in microgravity, the bubbles become larger and the local heat transfer coefficient is
increased at the tube inlet and decreased at the tube outlet, while it is almost unchanged
at tube center. The influence of gravity disappeared for the highest mass flux tested,
consistently with the fact that inertia forces dominate over buoyancy forces in this
regime; the threshold, however, is still to be determined.
Tadrist and coworkers [105] performed experiments on a square minichannel
heated from one side; the local heat transfer coefficient was derived from temperature
measurements inside the heater, by an inverse technique; high speed images of the
boiling pattern were taken from the transparent (unheated) side of the channel.
Experiments were carried out in parabolic flight, and data were collected also in the
enhanced-g phase of the parabola. The results showed that the average heat transfer
coefficient is increased in microgravity but, consistently with the findings of Celata
reported above, at local level the increase is located at tube inlet and the heat transfer
coefficient is unaffected afterwards. In the enhanced-g phase, heat transfer coefficients
and flow patterns were found unchanged with respect to normal gravity. The
experiments of Tadrist and Celata, although providing glimpse of the phenomena,
confirmed the need of a more extensive experimentation to clarify all aspects of flow
boiling in microgravity.
Quenching
Kawaji et al. [82] investigated onboard KC-135 aircraft the behavior of two-phase
flow and heat transfer during the quenching of a preheated quartz tube. The tube, heated
externally by a spiral nichrome tape, was initially empty and CFC113 was pumped into
the tube. In reduced gravity, thicker vapor film is formed on the tube wall making the
rewetting of the wall more difficult and resulting the reduction of heat transfer rate.
Kawanami et al. [106] investigated the heat transfer during quenching of tubes by
cryogenic fluids for the development of a system of fuel re-supply in orbit. Experiments
were conducted in JAMIC using LN2 as the test fluid. A transparent tube developed by
Ohta and Fujiyama [107] was used for the observation and the measurement of
temperature transition. The test fluid was introduced from underneath to the tube to
minimize the shear force in the counter direction exerted by the flow of generated
vapor. They reported the quenching time in reduced gravity was reduced by about 1.2
times, and the maximum and the minimum heat fluxes were larger by 1.4 and 1.2 times
respectively. The reduction of quenching time appeared to be due to the increased
velocity of the quenching front under reduced gravity conditions.
Yuan et al. [108] studied the chilldown of a horizontal Pyrex glass tube using liquid
nitrogen in a 1.7 s drop tower. The wall temperature was measured at various locations
along the tube, and the entering mass flux varied between 3.6-10.8 kg/m2s. The gravity
force was about two order of magnitude larger than the inertia force in earth gravity.
The chilldown process started out as film boiling, transitioned to nucleate boiling, and
was followed by single phase convection. The stratified flow that was observed during
nucleate boiling in earth gravity did not occur in low-g, resulting in longer chilldown
times.
Celata and Zummo [109] studied the quenching of a heated tube in normal and
reduced gravity, in parabolic flight experiments. Vertical Pyrex tubes with internal
diameters of 2,4 and 6 mm were flooded (both upwards and downwards in normal
gravity) with FC-72 at different mass flux velocities (50 to 450 kg/m2-s), and the
quench velocity was estimated by the temperature history given by the thermocouples
attached at various axial positions on the external side of the pipe. The results showed a
reduction of the rewetting velocity in microgravity.
Spray Cooling
Spray cooling is one of promising methods applied to the heat management system
in space. In general, the heat transfer, however, seems to be insensitive to the gravity.
On the other hand, there is an evidence from the experiments that the phenomena is
dependent on the surface orientation especially at high heat flux around the conditions
of maximum and minimum heat fluxes. Kato et al. [110] investigated the effect of
gravity on the spray cooling onboard MU-300. Decrease in CHF is observed for
CFC113, while the trend is reversed for water. Heat transfer is enhanced for both liquids
in the low heat flux region. In the transition boiling region, the boiling curve for
CFC113 is shifted toward smaller surface superheat in reduced gravity. The effect of
gravity disappears at high spray volume flux. Yoshida et al. [111] carried out the
experiments by using water and FC72 onboard MU-300. Two types of heaters were
employed, i.e., a transparent glass heater for the observation from underneath and a
copper block heater for the measurement of heat transfer data. At low spray volume
flux, effect of gravity or of surface orientation in the terrestrial experiments was
observed, while the significant influence of gravity or of the orientation was observed in
the values of CHF and heat transfer in the transition region if spray volume flux is high
and liquid film covers the heating surface. The heat transfer in film boiling region is
deteriorated when gravity is reduced or the heater is facing downward at low Weber
number less than 80.
Yerkes et al. [112] obtained data on spray cooling of ITO heaters deposited on glass
posts using FC-72 was the working fluid. The preliminary results indicated an increase
in heat transfer as the gravity decreased. Recently, Elston et al. [113] studied the cooling
performance of a 16-nozzle spray array using FC-72 on NASAs C-9 aircraft. A
25.4x25.4 mm2 thick film resistor was used as the heated surface. The spray cooling
performance in low-g was generally enhanced compared to 1-g and 1.8 g. At all gravity
levels, higher subcooling resulted in higher heat transfer levels. The gas content had
little effect, however, and this was confirmed recently for a single nozzle by
Puterbaugh, Yerkes, and Thomas [114]. The mass flow rate effect was mixed. An
increase in mass flow rate from 14.0 to 17.5 g/s resulted in higher heat transfer in low-g,
but a further increase to 21 g/s resulted in heat transfer degradation that may have been
due to liquid buildup on the nozzle. The heat transfer improved monotonically with
mass flow rate in 1.8g.
Future Work
Two pool boiling experiments are expected to be conducted on the ISS by NASA
within the next few years. The Nucleate Pool Boiling Experiment (NPBX) from UCLA
will study the dynamics of single and multiple bubbles growing on a silicon wafer with
micromachined cavities. The Microheater Array Boiling Experiment (MABE) will use
two microheater arrays (2.7 mm x 2.7 mm and 7.0 mm x 7.0 mm) to obtain boiling
curves under various combinations of heater size, bulk subcooling, and bulk pressure.
Both NPBX and MABE will be conducted within the Boiling Experiment Facility
(BXF) that will be flown on the Microgravity Science Glovebox (MSG).
A Japanese research team directed by JAXA is planning to perform flow boiling
experiments onboard the ISS. The experiment employs the MSPR (Multi-purpose Small
payload Rack) under development for Japanese Experimental Module "Kibo". The rack
volume of 900x600x660 mm3 makes possible the integration of an entire setup for flow
boiling. Two heated sections of a transparent glass heated tube and a stainless steel tube
are connected in parallel to a test loop. The objectives of the experiment are as follows:
i) Utilization of a long term microgravity period, systematic data on flow structures,
pressure drop, heat transfer are measured in a wide range of parameters, ii) Critical heat
flux data will be measured for both of low and high vapor mass quality regions, and
differences in the mechanisms for CHF conditions from the terrestrial ones will be
clarified, iii) Parameter ranges with no gravity effect is clarified on the dominant force
maps by selecting appropriate definition of dimensionless parameters such as Bond,
Weber, and Froude numbers etc.
The experimental module RUBI, to be implemented in the European Columbus
laboratory on ISS, is currently being developed by a European team of scientists, under
the sponsorship of ESA [115]. RUBI is dedicated to the study, with advanced
diagnostics, of the growth on a single vapor bubble on a heating foil with a single
nucleation site etched on it, and is aimed at clarifying the fundamental physical aspects
of boiling and their variation in variable gravity. Different stimuli, like electric field and
shear flow can also be applied, and the use of a fluid mixture is foreseen by the addition
of a second fluid during the latter part of the experiment. Diagnostic include
microthermocouples in the fluid, high speed video images, and infrared high speed
imaging of the bottom of the heated foil. The infrared images will allow the heat flux
magnitude to be derived.
Conclusions
The fundamental characteristics of boiling heat transfer and its advantages for high
demanding heat removal applications in low gravity applications have been outlined in
this book chapter. Experiments to date have shown that boiling can be used to provide
substantial heat transfer in microgravity, especially in subcooled conditions, and that
vapor removal from the proximity of the heated surface is the key problem to be solved
to avoid early degradation of heat transfer performance. However, we are still far from
the elaboration of quantitative models, and extensive experimentation is still required.
Future pool boiling experiments should provide information on the basic mechanisms
by which heat is transferred. Critical heat flux is reduced in low gravity, but it is higher
than predicted by the extrapolation of correlations obtained in earth gravity. The scaling
of CHF in microgravity based on Bond number has been found to be unsatisfactory, and
a separate dependence on gravity and heater size seems to exist.
Electric and acoustic fields along with boiling of mixtures have shown potential
for increasing wall heat transfer during pool boiling. Small amounts of dissolved gases
can significantly affect subcooled pool boiling heat transfer through the formation of
Marangoni convection around the primary bubble, resulting in a smaller primary bubble
and higher wall heat transfer. Models for Marangoni convection need to be developed
and verified against quantitative data. The effects of other parameters on pool boiling
remain largely unknown. Enhanced surfaces offer the possibility of greatly increasing
wall heat transfer, but no work is currently available. Surface energy gradients on a
heated surface might be utilized to remove bubbles without using flow boiling. Even
basic information on the effect of surface geometry on reduced gravity boiling heat
transfer is not understood. Continued development of numerical techniques is needed so
that two-phase flows systems can be efficiently designed. The experiments in flow
boiling are still insufficient to elaborate flow maps or to identify the minimum flow
above which the role of gravity is still significant. The boundaries should be clarified by
both analysis and experiments in short reduced gravity duration environments in order
to determine the experimental conditions for future long-term experiments.
The opportunities available for space-based experiments are extremely limited.
Significant international collaboration will be required to avoid the duplication of
experimental capabilities in the proposed experiments and to ensure facilities are used to
the maximum extend possible at the lowest cost.
Nomenclature
A Area [m2]
a Acceleration [m/s2]
Bo Bond number
cp Specific heat [J/kg-K]
Csf Constant
d Bubble departure diameter [m]
G Mass velocity [kg/m2-s]
g Earth gravity [9.81 m/s2]
h Heat transfer coefficient [W/m2-K]
hfg Heat of fusion [J/kg]
Ja Jakob number
M Molecular weight
m Constant or exponent
n Exponent
Nu Nusselt number
P pressure [N/m2]
Pr Reduced pressure
Pr Prandtl number
q Heat transfer [W]

q" Heat flux [W/m2]
Rp rms surface roughness [m]
R Non-dimensional radius
T Temperature
L length scale [m]
R Radius [m]
We Weber number
x Vapor quality
Subscripts
e Equivalent
l Liquid
max Maximum
nb Nucleate boiling
ref Reference condition
sat Saturation conditions
v Vapor
w Wall
x Coordinate direction
y Coordinate direction
z Coordinate direction perpendicular to heater
Greek
Heat transfer efficiency
Flow parameter
Contact angle
c Critical wavelength [m]
d Most dangerous wavelength [m]
Viscosity [Pa-s]
Density [kg/m3]
Surface tension [N/m]
References
[1] S. Nukiyama, The Maximum Values of the Heat Q Transmitted from Metal to
Boiling Water Under Atmospheric Pressure, Trans: JSME, 37 (1934) p.357
(reprinted in Int J. Heat Mass Transfer, v.9, pp. 1419-1433.1
[2] J.M. Delhaye, M. Giot and M.J. Rietmuller (eds.), Thermohydraulics of Two-
Phase Systems for Industrial Design and Nuclear Engineering, Hemisphere Pub.
Corp., NY, 1980.
[3] J.G. Collier and J.R. Thome, Convective Boiling and Condensation, 3rd ed.,
Oxford Univ. Press, 1996.
[4] J. Straub, Boiling Heat Transfer and Bubble Dynamics in Microgravity, Adv. Heat
Transfer, 35, (2001), pp.57-172.
[5] W.M. Rohsenhow, A Method of Correlating Heat Transfer Data for Surface
Boiling of Liquids, Transactions of the ASME, Vol. 84, (1962), p p. 969.
[6] N., Zhang and D.F. Chao, Models for enhanced boiling heat transfer by unusual
Marangoni effects under microgravity conditions, Int. Comm. Heat Mass
Transfer, 26, 8, (1999) pp. 1081-1090.
[7] M.G. Cooper, Saturation Nucleate, Pool Boilinga Simple Correlation, Institution
of Chemical Engineers Symposium Series, Vol. 86, (1984) pp. 785-792.
[8] K. Stephan and M. Abdelsalam, Heat-transfer correlations for natural convection
boiling, Int. J. of Heat and Mass Transfer, Vol. 23, (1980) pp. 73-87.
[9] VDI Heat Atlas, ed. Verein Deutscher Ingenieure, (English version), VDI-Verlag,
Dusseldorf, D, 1993.
[10] P. Di Marco, W. Grassi, EHD Effects on Pool Boiling in Reduced Gravity, Proc.
of the 5th ASME/JSME Joint Thermal Engineering Conference, San Diego, CA,
USA, March 1999, paper AJTE99/6275.
[11] P. Di Marco and W. Grassi, Effects of external electric field on pool boiling:
comparison of terrestrial and microgravity data in the ARIEL experiment, 7th ECI
International Conference on Boiling Heat Transfer, Florianopolis, Brazil, May 3-
7, 2009.
[12] N. Zuber, Hydrodynamics of Boiling Heat Transfer, AEC Report AECU-4439,
(1959).
[13] J.H. Lienhard and V.K. Dhir, Hydrodynamic prediction of peak pool-boiling heat
fluxes from finite bodies, J. Heat Transfer, ASME Serial C, Vol. 95, (1973) pp.
152-158.
[14] Y. Haramura, and Y. Katto, (1983) A New Hydrodynamic Model of the Critical
Heat Flux, Applicable Widely to Both Pool and Forced Convective Boiling on
Submerged Bodies in Saturated Liquids, Int. J. of Heat and Mass Transfer, Vol.
26, pp. 389-399.
[15] P. Di Marco and W. Grassi, About the scaling of critical heat flux with gravity
acceleration in pool boiling, Proc. XVII UIT National Heat Transfer Conference,
Ferrara, June 30-July 2, (1999).
[16] R. Siegel and E.G. Keshock, Effects of reduced gravity on nucleate boiling
bubble dynamics in saturated water, AICHE Journal, Vol. 10, No. 4, (1964), pp.
509-517.
[17] J.A. Clark, Gravic and agravic effects in cryogenic heat transfer, Advances in
Cryogenic Heat Transfer 87, Vol. 64, (1968) pp. 93-102.
[18] J. Straub, M. Zell, B. Vogel, Pool boiling in a reduced gravity field, Proc. 9th Int.
Heat Transfer Conference, Jerusalem, Israel, Aug. 1990, pp. 91-112.
[19] D.M Qui, V.K. Vhir, D. Chao, M.M. Hasan, E. Neumann, G. Yee, and A.
Birchenough, Single-bubble dynamics during pool boiling under low gravity
conditions, J. of Thermophysics and Heat Transfer, Vol. 16, No. 3, (2002), pp.
336-345.
[20] W. Fritz, Berechnungen des Maximalvolumens von Dampfblasen, Phys. Z., Vol.
36, (1935) pp. 379-384.
[21] G. Son, V.K. Dhir and N. Ramanujapu, Dynamics and heat transfer associated
with a single bubble during nucleate boiling on a horizontal surface, J. of Heat
Transfer, Vol. 121, No. 3, (1999), pp. 623-632.
[22] D. M. Qui, G. Son, V.K. Vhir, D. Chao and K. Logsdon, Dynamics of single and
multiple bubbles and associated heat transfer in nucleate boiling under low gravity
conditions, Annals of the New York Academy of Sciences, Vol. 974, (2002), pp.
378-397.
[23] H.S. Abarajith, V.K. Dhir, G. Warrier, and G. Son, Numerical Simulation and
Experimental Validation of the Dynamics of Multiple Bubble Merger during Pool
Boiling under Microgravity conditions, Microgravity Transport Processes
Conference, Davos, Switzerland, Sep. 14-19, (2003).
[24] M. Barthes, C. Reynard, R. Santini, L. Tadrist Liquid vapour phase change on a
single nucleation site : heat and mass transfer, Proceedings of Second
International Topical Team Workshop on Two-Phase Systems For Ground And
Space Applications, Kyoto, Japan, 26-27 October 2007.
[25] N. Schweizer, P. Stephan, Experimental study of bubble behavior and local heat
flux in pool boiling under variable gravitational conditions, to appear in
Multiphase Science and Technology, 2009.
[26] J. Kim, J.F. Benton, D. Wisniewski, Pool Boiling Heat Transfer on Small Heaters:
Effects of Gravity and Subcooling, Int. J. Heat Mass Transfer, 45 (2002) pp.3921-
3934.
[27] H.S. Lee, H. Merte, Pool boiling phenomena in microgravity, Heat Transfer 1998,
Proc. of 11th Int. Heat Transfer Conference, pp.395-406.
[28] T. Oka, Y. Abe, K. Tanaka, Y.H. Mori, H. Yasuhiko, A. Nagashima,
Observational study of pool boiling under microgravity, JSME International
Journal, Series 2, 35, 2, (1992) pp. 280-286.
[29] H.S. Lee, H. Merte, and F. Chiaramonte, Pool boiling curve in microgravity, J. of
Thermophysics and Heat Transfer, Vol. 11, No. 2, (1997) pp. 216-222.
[30] M. Zell, J. Straub, A. Weinzierl, Nucleate pool boiling in subcooled liquid under
microgravity results of Texus experimental investigations, 5th European Symp. on
Material Science under Microgravity, Schloss Elmau, 1984, pp.327-333.
[31] J. Straub, S. Micko, Boiling on a Wire under Microgravity Conditions: First
Results from a Space Experiment Performed in 1996, Proc. EUROTHERM
Seminar n.48, Pool Boiling 2, Gorenflo D., Kenning D.B.R., Marvillet C. Eds.,
Paderborn, D, Sept. 1996, pp.275-282.
[32] M. Steinbichler, S. Micko, J. Straub, Nucleate boiling heat transfer on a small
hemispherical heater and a wire under microgravity, Heat Transfer 1998, Proc. of
11th Int. Heat Transfer Conference, 2, Seoul, Korea, Aug. 1998, pp539-544.
[33] J. Straub, G. Picker, M. Steinbichler, J. Winter, M. Zell, Heat transfer and various
modes of bubble dynamics on a small hemispherical heater under microgravity
and 1g conditions, Proc. EUROTHERM Seminar n.48, Pool Boiling 2, Gorenflo
D., Kenning D.B.R., Marvillet C. Eds., Paderborn, D, Sept. 1996, pp.265-275.
[34] Y. Abe and A. Iwasaki, Pool Boling Under Microgravity, Advances In Space
Research, Vol.13, No.7, pp.165-(7)168, 1993.
[35] T. Oka, Y. Abe, Y.H. Mori and A. Nagashima, Pool boiling of n-pentane, CFC-
113, and water under reduced gravity: Parabolic flight experiments with a
transparent heater, Trans. ASME, J. Heat Transfer, Vol. 114, pp. 408-417, 1995.
[36] H. Ohta, K. Kawasaki, S. Okada, K. Inoue, S. Yosida. and T.S. Morita, Heat
Transfer in Microgravity Nucleate Boiling on a Transparent Heating Surface,
Fluids in Space, ESA, pp. 531-536, 1996.
[37] R,F, Gaertner, Photographic Study of Nucleate Pool Boiling on a Horizontal
Surface, Trans. ASME J. Heat transfer, Vol.87, pp.17-29, 1965.
[38] H. Tanaka and S. Nishio, Bubble Structure in Subcooled Boiling at High Heat
Fluxes, Proc. 35th national Heat Transfer symposium of Japan, Vol.III, pp.923-
924, 1998 (in Japanese).
[39] H. Ohta, M. Kawaji, H. Azuma, K. Kawasaki, H. Tamaoki, K. Ohta, T. Takada, S.
Okada, S. Yoda and T. Nakamura, TR-1A Rocket Experiment on Nucleate Pool
Boiling Heat Transfer Under Microgravity, Heat Transfer in Microgravity
Systems, ASME HDT-Vol.354, pp. 249-256, 1997.
[40] H. Ohta, K. Kawasaki, S. Okada, H. Azuma, S. Yoda and T. Nakamura, On the
Heat Transfer Mechanisms in Microgravity Nucleate Boiling, Advances In Space
Research, Vol.24No.10, pp. 1325-1330, 1999.
[41] A.M. Bhat, R. Prakash and J.S. Saini, On the Mechanism of Macrolayer
Formation in Nucleate Pool Boiling at High Heat Flux, Int. J. Heat Mass
Transfer, Vol.26, No.5, pp.735-740, 1983.
[42] H. Merte, H.S. Lee, and R.B. Keller, Dryout and rewetting in the pool boiling
experiment flown on STS-72 (PBE-IIB) and STS-77 (PBE-IIA), NASA/CR
1998-207410, (1998).
[43] H. Lee, Mechanisms of steady-state nucleate pool boiling in microgravity, Annals
of the New York Academy of Sciences, Vol. 974, (2002), pp. 447-462.
[44] R. Raj, J. Kim and J. McQuillen, Subcooled Pool Boiling in Variable Gravity
Environments, ASME Journal of Heat Transfer, Vol. 131, 091502-1:091502-10,
September, 2009.
[45] O. Kawanami, H. Ohta, O. Kabov, Y. Sakata, Y. Kotani, Y. Asada, T. Nagayasu,
Y. Shinmoto, S. Chikov, P. Queeckers, J. Straub, Heat Transfer and Bubble
Behaviors in Microgravity Pool Boling in ESA Parabolic Flight Experiment,
Microgravity Science and Technology, Vol.21, Suppl.1, pp.S3-S8, 2009.
[46] K. Suzuki, H. Kawamura, et al., Burnout in Subcooled Pool Boiling of Water
under Microgravity, Proc. of Joint Xth European and VIth Russian Symposium on
Physical Science in Microgravity, pp. 366-369, 1997.
[47] H.J. Ivey and D.J. Moris, On the Relevance of the Vapor liquid Exchange
Mechanism for Subcooled Boiling Heat Transfer at Higher Pressure, UK Atomic
Energy Authority, Winfrith, England, AEEW-R, Vol.137, 1962.
[48] K. Suzuki, M. Suzuki and H. Kawamura, On Subcooled Pool Boiling of Water in
the Drop-shaft Experiment, Proc. Drop Tower Days 2002 in Japan, pp.131-135,
2002.
[49] P. Di Marco and W. Grassi, Effect of force fields on pool boiling flow patterns in
normal and reduced gravity, Heat Mass Transfer, V. 45, 7, 2009, pp. 959-966.
[50] P. Di Marco W., Grassi, Motivation and Results of a Long-Term Research on
Pool Boiling Heat Transfer in Low Gravity, Int. J. Th. Sciences, 41, 7 (2002) pp.
567-585.
[51] P. Di Marco, W. Grassi, An Experimental Study on the Influence of Force Fields
on CHF in Pool Boiling, Proc. 12th International Heat Transfer Conference,
Grenoble, F, August 18-23, 2002, vol. 3, pp. 617-622.
[52] H. Asano, K. Akita and M. Inoue, Effect of Gravity on Pool Boiling Heat Transfer
on Thermal Spray Coating, Microgravity Science and Technology, Vol.19,
No.3/4, pp.90-92, 2007.
[53] J. Straub, Transport phenomena in micro- and zero- gravitational fields, Transport
Phenomena and Technology 1992, Proc. 3rd Int. Symp. on Heat Transfer and
Transport Phenomena, Beijing, 1992, pp.16-28.
[54] V.K. Dhir, Numerical Simulations of Pool Boiling Heat Transfer, AIChe J., 47
(2001), pp.813-834.
[55] Q. Bai, Y. Fujita, Numerical Simulation of the Growth of a Single Bubble in
Nucelate Boiling, Thermal Science and Engineering, 7 (1999), pp.45-53.
[56] P. Stephan and J. Hammer, A New Model for Nucleate Boiling Heat Transfer,
Heat and Mass Transfer, 30 (1994), pp.119-125.
[57] R. Marek and J. Straub, The Origin of Thermocapillary Convection in Subcooled
Nucleate Pool Boiling, Int. J. Heat Mass Transfer 44 (2001) pp. 619-632.
[58] J. Straub, Origin and Effect of Thermocapillary Convection in Subcooled Boiling,
Ann. N.Y. Acad. Sci., 974 (2002) pp.348-363.
[59] J. Betz and J. Straub, Thermocapillary Convection around Gas Bubbles, Ann. N.Y.
Acad. Sci., 974 (2002) pp.220-245.
[60] M. Barthes, C. Reynard, R. Santini, L. Tadrist Non-condensable gas influence on
the Marangoni convection during a single vapour bubble growth in a subcooled
liquid, Europhysics letters, Vol. 77, N1, pp. 14001-14007, 2007.
[61] C. Reynard, M. Barthes, , R. Santini, L. Tadrist Experimental study of the onset of
the 3d oscillatory thermocapillary convection around a single air or vapor bubble.
Influence on heat transfer, Experimental Thermal and Fluid Science, Vol. 29, pp.
783793, 2005.
[62] C. Reynard, R. Santini and L. Tadrist, Experimental study of the gravity influence
on the periodic thermocapillary convection around a bubble, Experiments in
Fluids , Vol. 31, (2001), pp. 440-446.
[63] C.D. Henry, J. Kim, B. Chamberlain and T.G. Hartmann, Heater aspect ratio
effects on pool boiling heat transfer under varying gravity conditions,
Experimental Thermal and Fluid Science, Vol. 29, No. 7, (2005) pp. 773-782.
[64] C.D. Henry, J. Kim and J. McQuillen, Dissolved Gas Effects on Thermocapillary
Convection During Boiling in Reduced Gravity Environments, Heat and Mass
Transfer, Vol. 42, (2006). pp. 919-928.
[65] J.J. Xu and M. Kawaji, Study of liquid-solid contact in transition boiling regime
under reduced and normal gravity conditions, Proceedings of the 11th
International Heat Transfer Conference, Kyongju, Korea, Vol. 2, (1998) pp. 419-
424.
[66] C.D. Henry and J. Kim, Heater size, subcooling and gravity effects on pool
boiling heat transfer, Int. J. Heat Fluid Flow, Vol. 25, No. 2, (2004) pp. 262-273.
[67] Y. Abe, T. Oka, Y.H. Mori and A. Nagashima, Pool Boiling of a Non-azeotropic
Binaryu Mixture Under Microgravity, Int. J. Heat Mass Transfer, 37 (16), (1994)
pp. 2405-2413.
[68] Y. Abe and A. Iwasaki, Single and Dual Vapor Bubble Experiments in
Microgravity, Proc. Int. Conf. on Microgravity Fluid Physics and Heat Transfer,
pp. 55-61, 1999.
[69] S. Ahmed and V.P. Carey, Effects of gravity on the boiling of binary fluid
mixtures, Int. J. of Heat and Mass Transfer, Vol. 41, No. 16, (1998), pp. 2469-
2483.
[70] C.I. Sun and V.P. Carey, Effects of gap geometry and gravity on boiling around a
constrained bubble in 2-propanol/water mixtures, Proceedings of the 2003 ASME
Summer Heat Transfer Conference, Las Vegas, (2003).
[71] T.J. Snyder, J.N. Chung and J.B. Schneider, Competing effects of
dielectrophoresis and buoyancy on nucleate boiling and an analogy with variable
gravity boiling results, J. of Heat Transfer, Vol. 120, ,(1998), pp. 371-379.
[72] T.J. Snyder and J.N. Chung, Terrestrial and microgravity boiling heat transfer in a
dielectrophoretic force field, Int. J. of Heat and Mass Transfer, Vol. 43, (2000),
pp. 1547-1562.
[73] T.J. Snyder, J.B. Schneider and J.N. Chung, Dielectrophoresis with application to
boiling heat transfer in microgravity. I. Numerical Analysis, J. of Applied
Physics, Vol. 89, No. 7, (2001), pp. 4076-4083.
[74] T.J. Snyder, J.N. Chung and J.B. Schneider, Dielectrophoresis with application to
boiling heat transfer in microgravity. II. Experimental investigation, J. of Applied
Physics, Vol. 89, No. 7, (2001), pp. 4084-4090.
[75] P. Di Marco, W. Grassi, Influence of Electric and Gravitational Force Fields on
Nucleate Boiling of FC72 on a Wire: Results of a Sounding Rocket Campaign,
AIP Conference Proceedings, 608 (2002) pp. 172-180.
[76] P. Di Marco, W. Grassi, F. Trentavizi, Pool Film Boiling Experiments on a Wire
in Low Gravity: Preliminary Results, Ann. N.Y. Acad. Sci., 974, (2002) pp. 428-
446.
[77] Di Marco P., Grassi W., Pool boiling in microgravity and in the presence of
electric field: evaluation of the void fraction in ARIEL experiment, to appear in
Multiphase Science and Technology, 2009.
[78] Y. Hao, H.N. Oguz and A. Prosperetti, The action of pressure-radiation forces on
pulsating vapor bubbles, Physics of Fluids, Vol. 13, No. 5, (2001), pp. 1167-1177.
[79] J.S. Sitter, T.J. Snyder and J.N. Chung, Acoustic field interaction with a boiling
system under terrestrial gravity and microgravity, J. of Acoustical Society of
America, Vol. 104, no. 5, (1998), pp. 2561-2569.
[80] J.S. Sitter, T.J. Snyder and J.N. Chung, Terrestrial and microgravity pool boiling
heat transfer from a wire in an acoustic field, Int. J. of Heat and Mass Transfer,
Vol. 41, No. 14, (1998), pp. 2143-2155.
[81] M. Misawa and S. Anghaie, Zero-gravity Void Fraction and Pressure Drop in a
Boiling Channel, AIChE Symp. Ser., Vol.87, No.283, (1991), pp.226-235.
[82] M. Kawaji, C.J. Westbye and B.N. Antar, Microgravity Experiments on Two-
phase Flow and Heat Transfer during Quenching of a Tube and Filling of a
Vessel, AIChE Symp. Ser.,Vol.87, No.283, (1991), pp.236-243.
[83] M. Saito, K. Miyazaki, M. Kinoshita and Y. Abe, Experiments of Boiling Two-
Phase Flow under Microgravity, Proc. 5th Int. Topical Meeting on Reactor
Thermal Hydraulics, pp.1738-1748, 1992.
[84] H. Yamada, T. Nakamura, T. Matsumoto, T. Fujii, H. Asano and T. Nakazawa,
Two-phase Flow Experiment under Microgravity Using MU-300, Proc. JSME
Conf. Kansai No.253, No. 944-4, pp. 66-68, (1994) (in Japanese).
[85] H. Ohta, K. Inoue, Y. Yamada, S. Yoshida, H. Fujiyama and S. Ishikura,
Microgravity Flow Boiling in a Transparent Tube, Proc. 4th ASME-JSME
Thermal Engineering Joint Conf., Vol.4, pp. 547-554, 1995.
[86] H. Ohta, Experiments on Microgravity Boiling Heat Transfer by Using
Transparent Heaters, Nuclear Engineering and Design, Vol. 175, pp. 167-180,
1997.
[87] H. Ohta, Watanabe, T. Sabato, S. Okada, S. Takasu and H. Kawasaki, Gravity
Effect on Flow Boiling Heat Transfer in Narrow Gaps, Proc. 5th ASME-JSME
Thermal Engineering Joint Conference, AJTE 99-6421, CD Rom, 1999.
[88] H. Ohta, Y. Shinmoto and K. Matsunaga, Effect of Gravity on Flow Boiling in
Narrow Ducts and Enhancement of CHF Values, Heat Transfer 2002, Proc. 12th
Int. Heat Transfer Conference, Vol.3, pp.725-730, 2002.
[89] D.T. Westheimer and G.P. Peterson, Visualization of flow boiling in an annular
heat exchanger under microgravity conditions, J. of Thermophysics and Heat
Transfer, Vol. 15, No. 3, (2001), pp. 333-339.
[90] Y. Ma and J.N. Chung, An experimental study of forced convection boiling in
microgravity, Int. J. of Heat and Mass Transfer, Vol. 41, No. 15, (1998), pp.
2371-2382.
[91] Y. Ma and J.N. Chung, A study of bubble dynamics in reduced gravity forced
convection boiling, Int. J. of Heat and Mass Transfer, Vol. 44, (2001), pp. 399-
415.
[92] Y. Ma and J.N. Chung, An experimental study of critical heat flux (CHF) in
microgravity forced-convection boiling, Int. J. of Multiphase Flow, Vol. 27,
(2001), 1753-1767.
[93] R.K. Lui, M. Kawaji and T. Ogushi, An Experimental Investigation of Subcooled
Flow Boiling Heat Transfer under Microgravity Conditions, Heat Transfer 1994,
Proc. 10th Int. Heat Transfer Conf., Vol.7, 18-FB-16, pp.497-502, 1994.
[94] H. Ohta, K. Inou and S. Yosida, S., Experiments on Dryout Phenomena in
Microgravity Flow Boiling, Proc. 10th International Symposium on Transport
Phenomena in Thermal Science and Process Engineering, Vol.2, pp. 373-378,
1997.
[95] H. Ohta, K. Inoue and H. Fujiyama, Analysis of Gravity Effect on Two-phase
Forced Convective Heat Transfer in Annular Flow Regime, Proc. 3rd Int.
Conference on Multiphase Flow, CD Rom, No.389, 1998.
[96] H. Ohta, Y. Shinmoto, S. Morioka, H. Tanouchi and H. Fujiyama, Liquid-vapor
Behavior and Temperature Fluctuation near Dryout Point in Flow Boiling under
Different Gravity Conditions, Proc. 3rd European Thermal Science Conference
2000, Vol.II, pp. 845-850, 2000.
[97] H. Ohta, Microgravity heat Transfer in Flow Boiling, Advances in Heat Transfer,
Academic Press, Vol.37, pp.1-76, 2003.
[98] J.S. Bower, and J.F. Klausner, Gravity independent subcooled flow boiling heat
transfer regime, Exp. Therm. Fluid Sci., Vol. 31, (2006) pp. 141-149.
[99] Zhang, H., Mudawar, I., Hasan, M. (2005) Flow boiling CHF in microgravity
International Journal of Heat and Mass Transfer 48 (2005) 31073118.
[100] H. Zhang, I. Mudawar and M. Hasan Assessment of dimensionless CHF
correlations for subcooled flow boiling in microgravity and Earth gravity, Int. J.
Heat Mass Transfer, Vol. 50, (2007), 4568-4580.
[101] D.D. Hall and I. Mudawar, Ultra-high critical heat flux (CHF) for subcooled water
flow boiling II. High-CHF database and design equations, Int. J. Heat Mass
Transfer 42, (1999), 14291456.
[102] D.D. Hall and I. Mudawar, Critical heat flux (CHF) for water flow in tubes II.
Subcooled CHF correlations, Int. J. Heat Mass Transfer, Vol. 43, (2000), 2605
2640.
[103] G.P. Celata, M. Cumo, M. Gervasi and G. Zummo, Flow pattern analysis of flow
boiling in microgravity, Multiphase Science and Technology, Vol. 19, No. 2,
(2007), pp. 183210.
[104] G.P. Celata, M. Cumo, F. D'Annibale, L. Saraceno and G. Zummo, Rewetting
Velocity in Quenching at Reduced Gravity, Proc. of 27th UIT National
Conference, Reggio Emilia, I, June 2009, submitted to Int. J. Therm. Sci., (2009),
[105] S. Luciani, D. Brutin, O. Rahli, C. Le Niliot and L. Tadrist, Flow Boiling in

Minichannels under Normal, Hyper and Microgravity: Local Heat Transfer
Analysis Using Inverse Methods, Journal of Heat Transfer, Vol. 130, Issue 10,
2008.
[106] O. Kawanami, T. Hiejima, H. Azuma and H. Ohta, Experimental Study on
Dryogenic Flow Boiling under Low Gravity, Proc. 6th ASME-JSME Thermal
Engineering Joint Conference, TED-AJ03-361, CD Rom, 2003.
[107] H. Ohta and H. Fujiyama, Local Heat Transfer Measurement and Visualization of
Liquid-vapor Behavior for Microgravity Flow Boiling Experiments, Two-Phase
Flow Modeling and Experimentation 1999, Vol.1, pp. 255-262, 1999.
[108] K. Yuan, Y. Ji, J.N. Chung and W. Shyy, Cryogenic boiling and two-phase flow
during pipe chilldown in earth and reduced gravity, J. Low Temperature Physics,
Vol. 150, (2008), pp. 101-122.
[109] G.P. Celata and G. Zummo, Flow boiling heat transfer in microgravity, ECI
International Conference on Boiling Heat Transfer Florianpolis-SC-Brazil, 3-7
May 2009, pp.1-13.
[110] M. Kato, Y. Abe, Y.H. Mori and A. Nagashima, Spray cooling characteristics
under reduced gravity, AIAA J. Thermophysics and Heat Transfer, Vol. 9, No.2,
(1994), pp. 378-381..
[111] K. Yoshida, Y. Abe, T. Oka, Y.H. Mori and A. Nagashima, Spray cooling under
reduced gravity condition, Trans ASME, J. Heat Transfer, Vol. 123, (2001), pp.
309-318.
[112] K.L. Yerkes, T.E. Michalak, K.M. Baysinger and R. Puterbaugh, Variable-
Gravity Effects on a Single-Phase Partially Confined Spray Cooling System, J.
Thermophysics and Heat Transfer, Vol. 20, No. 3, (2006), pp. 361-370.
[113] L.J. Elston, K.L. Yerkes, S.K. Thomas and J. McQuillen, Cooling Performance of
a 16-Nozzle Array in Variable Gravity, J. Thermophysics and Heat Transfer, Vol.
23, No. 3, (2009), p. 571-581.
[114] R.L. Puterbaugh, K.L. Yerkes and S.K. Thomas, Partially confined FC-72 Spray-
cooling performance: Effect of Dissolved Air, J. Thermophysics and Heat
Transfer, Vol. 23, No. 3, (2009), p. 582-591.
[115] A. Pacros, O. Minster and T. Dewandre, Future ESA experiments in heat and
mass transfer research, 3rd Int. Symposium on Physical Sciences in Space, Kyoto,
Japan, October 2007.
Chapter 3
Two-Phase Flow Regime Identification Methodologies in

Thermal-Hydraulic Applications
J. Enrique Julia1 , Takashi Hibiki2 and Mamoru Ishii2

1
Departamento de Ingenieria Mecanica y Construccion. Universitat Jaume I
Campus de Riu Sec. 12071-Castellon de la Plana. Spain
2
School of Nuclear Engineering. Purdue University
400 Central Drive, West Lafayette, IN 47907-2017, USA
Abstract
Two-phase flow regimes have a profound influence on all the two-phase transport processes.
Consequently, their correct identification is a task of major importance. Two main components are
needed in the identification process: flow regime indicator and classifier. In the first pioneering works,
visual flow regime maps were obtained. In this case, the visual information was the flow regime indicator
and the researcher judgement was used as flow regime classifier. This approach presents a high level of
subjectivity. In the last decades, important work in obtaining more objective flow regime indicators and
classifiers has been done. In this review the current knowledge about flow regime indicators and
classifiers in thermal-hydraulic applications is summarized. Flow regime indicators comprise different
statistical parameters of void fraction and bubble chord length distributions. Flow regime classifiers
cover different artificial neural network architectures such as self-organized and probabilistic neural
networks. Finally, the main flow regime identification works performed in different flow channel
geometries are reported.
Introduction
Many industrial applications handle systems involving two or more phases, as for
example, nuclear power plant steam generators, boilers, paper-making industry,
processing of chemicals for the production of food, pharmaceuticals, petroleum
transportation etc. The two phases can flow according to several topological
configurations called flow patterns or flow regimes, which are determined by the
interfacial structure between both phases. The existence of a particular flow regime
depends on a variety of parameters. These parameters include the fluids properties, the
flow channel size, geometry and orientation, body force field and flow rates.
Despite the wide usage of multiphase systems, their hydrodynamic and kinematic
mechanisms are not completely understood. The correct identification of the flow
regimes is necessary because they have a profound influence on all the two-phase
Email address: bolivar@emc.uji.es, hibiki@purdue.edu, ishii@purdue.edu.

94 Advances in Multiphase Flow and Heat Transfer 1 (2009) Julia et al.
transport processes. Various models have been developed to predict the transition
criteria between the flow regimes. The first studies in this field have been confined to
circular flow geometry [1-2]. Additionally, models for other flow channel geometries
such as annuli [3-5], rod bundles [6] and rectangular channels [7] have been developed.
In all the cases, consistent experimental flow regime maps are needed to understand the
physical phenomena involved in the flow regime transitions as well as to validate the
models.
This chapter is focused on flow regime identification methodologies in thermal-
hydraulic applications. Consequently, vertical two-phase flow regimes will be
considered. However, the methodologies shown in the chapter are fully applicable to
other two-phase flow configurations, i.e. inclined and horizontal two-phase flows.
Vertical two-phase flows are usually classified into four or five basic flow regimes
[1-2]. Considering an upward liquid flow in a vertical pipe with constant liquid flow
rate, a gradual increment in the gas flow rate produces, consecutively, the following
flow patterns (Fig (1)):
1. Bubbly flow (designated as B in the following sections). The liquid phase is

continuous and small dispersed bubbles flow within the liquid (Fig. (1a)).
2. Cap-Bubbly flow (CB). The number density of small bubbles increases and
bigger bubbles (cap bubbles) are formed due to coalescence (Fig. (1b)).
3. Slug flow (S). The cap bubbles coalescence is increased due to wake entrainment
and their diameter approaches that of the pipe. When this occurs, large, characteristic
bullet-shaped bubbles, known as Taylor bubbles, are formed. Taylor bubbles are
separated by liquid slugs containing small bubbles. Liquid flows downward between the
Taylor bubbles and the pipe wall in the form of a thin falling film, although both gas
and liquid net flows can be upward (Fig. (1c)). This flow regime does not occur in large
diameter pipes.
4. Churn-Turbulent flow (CT). By increasing the gas flow rate, a breakdown in the
Taylor bubbles leads to an unstable flow regime, and the continuity of the liquid slug is
repeatedly destroyed. This liquid accumulates, forms a bridge and is again lifted by the
gas. This oscillatory or alternating direction of the liquid motion is typical of the churn-
turbulent flow (Fig. (1d)).
5. Annular flow (A). The gas phase flows in the pipe core and the liquid phase
flows on the pipe wall as a film. Generally, part of the liquid phase is entrained as small
droplets in the gas core, but it is quite uncommon for bubbles to be entrained in a liquid
film (Fig (1e)).
Two-Phase Flow Regimes Identification Methodologies Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 95
a) b) c) d) e)
Fig. (1). Two-phase flow regimes in upward flow. (a) bubbly, (b) cap-bubbly,(c) slug, (d) churn-
turbulent and (e) annular.
Many researchers have been working on developing objective flow regime

identification methodologies. Every identification process comprises at least two main
performers: indicator and classifier. First of all, a feature that is intrinsic to the subject
to be identified and that makes it different from the other subjects in the group is needed
(flow regime indicator). Second, a procedure to classify the subjects according to this
feature should be developed (flow regime classifier).
Consequently, most flow regime identification approaches have two steps in
common: the first step consists of developing an experimental methodology for
measuring certain parameters that are intrinsic to the flow and are also suitable flow
regime indicators. In the second step, a non-linear mapping is performed to obtain an
objective identification of the flow regimes in accordance with these indicators.
Flow regime identification results are plotted in flow regime maps (FRM). These
maps are graphs displaying the flow condition information in which the flow regime is
identified in the x-y coordinates. Different options for the x-y axis variables in the flow
maps are available. If the superficial velocity values are known this is the preferred
option. The superficial velocity, jg or jf, is defined as the volumetric flow rate divided
by the flow channel area. However, in steam-water flows the local superficial gas
velocity is not usually known accurately. In this case, other options for the x-y axis are
considered, for instance mass velocity vs. flow quality have been used in different
works [8]. Fig. (2) shows an example flow regime map for adiabatic flow.
1
10
B
CB
Superficial liquid velocity, <jf> [m/s]
S
CT
0 A
10
-1
10
-2
10 -3 -2 -1 0 1 2
10 10 10 10 10 10
Superficial gas velocity, <jg> [m/s]
Fig. (2). Two-phase flow regime map. Example of flow regime map for vertical upward adiabatic
flow in a 25.4 inner diameter pipe.
Flow Regime Indicators

In the firsts flow regime identification works, the visual information obtained by the
researcher was used as the flow regime indicator [9-10]. Certainly, this is the easiest
flow regime indicator to obtain, since only the researchers sight is necessary.
However, the information that can be extracted from it is frequently partial. This fact is
easily understandable for high velocity or large diameter pipe flows. In addition, this
indicator is highly subjective given that it is not recordable. Further, only the researcher
judgement can be used as flow regime classifier. Despite its weakness this flow regime
indicator is still used in research works [11-12] although some improvements have been
made, such as the use of high speed cameras [13-14].
In the last decades, important work in obtaining more objective flow regime
indicators has been done. There are two important parameters to characterize the two-
phase flow geometry, namely void fraction and interfacial area concentration ( and ai).
Consequently, the flow regime indicators have to be related to such parameters. In
order to obtain the flow regime indicator two different procedures are needed. First, a
signal measured with an experimental device related to void fraction or interfacial area
concentration is needed. Then, some statistical parameters of the signal are calculated
and used as indicators. The latter step is not mandatory since the direct signal can be
used as indicator though it is not recommended for the majority of flow regime
classifiers.
Experimental techniques used to provide flow regime indicator

Several experimental techniques have been used to provide flow regime indicators.
The first experimental techniques that were used provided void fraction information.
Thus, the flow regime is defined as a time-averaged, volume-averaged or area-averaged
pattern. The flow regimes identified in this way are known as Global Flow Regimes
(GFR). In this regard, Jones and Zuber [15] employed a linearized X-ray system to
obtain very accurate area-averaged void fraction measurements, <>. However, the use
of radioactivity complicates this approach and it is only indicated for applications in
which other experimental techniques can not be applied.
In addition, local pressure transducers have been also used for volume-averaged
void fraction measurements, [16-19]. The volumetric void fraction or pressure
fluctuations are measured with this technique, which vary with the distance of the test
locations. In addition, some difficulties are not easily overcome owing to the presence
of gas which might be trapped in pressure lines [20]. Nevertheless, it represents a
simple experimental approach, so it is still used in some experimental work.
Finally, area-averaged void fraction measurements obtained from impedance meters
have been used to obtain flow regime indicators [21-22]. An impedance void meter is a
non-intrusive conductance type probe that utilizes the difference in electrical
conductivity between the air and water. An alternating current is supplied to the
electrodes of the meter and the electrodes are connected to the electronic circuit, which
is specially designed so that the output voltage of the circuit is proportional to the
measured admittance between the electrodes. At the end of the process, the area-
averaged void fraction can be obtained from the measured admittance by a calibration
process. The void fraction is almost a linear function of the non-dimensional
admittance. Thus, the void fraction measured by the void meter is an instantaneous,
area-averaged void fraction.
A completely different approach can be followed in order to select the flow regime
indicators. In this case the indicator is composed of a set of variables that can be easily
measured in the two-phase flow. In this regard the superficial velocity values as well as
the pipe diameter and inclination angle have been used [23].
The second parameter to characterize the two-phase flow geometry is the interfacial
area concentration. Local conductivity probes are used to measure it. The conductivity
probe, proposed by Neal and Bankoff [24], is based on the difference of conductivity
between water and air. The first probe designs were composed of two sensors;
however, these probes were only utilizable with spherical bubbles (bubbly flows). This
limitation was overcome with the use of four-sensor conductivity probes [25-27]. In a
four-sensor conductivity probe, three pairs of double sensors can be formed with one
front common sensor in the upstream and three rear sensors in the downstream.
Therefore, three components of interfacial velocities can be obtained at a local point by
measuring the time delay between the signals from three pairs of double sensors. It is
important to reduce the probe cross-section as much as possible in order to minimize
both the number of missing bubbles and the deformation of passing bubble interfaces.
The ai calculation is a time consuming process, so bubble chord lengths, LB, are
used as flow regime indicators. LB is directly connected with ai and it can be obtained
by multiplying the local residence time and the bubble interface velocity. The local
residence time of the bubbles can be obtained by dividing the sum of the time fraction
occupied by gas-phase by the total measurement time. The bubble interface velocity
can be obtained by using two sensors of the conductivity probe. This measurement is
performed by utilizing the ratio of the distance between the sensors and the time delay
between the bubble reaching these sensors.
Thus, if the local conductivity probe is used to get a flow regime indicator, the flow
regime is defined by time-averaged bubble chord length patterns. The flow regimes
identified in this way are known as Local Flow Regimes (LFR). The main advantages
of this approach over the conventional GFR identifiers are the following [28]:
- The LFRs provide local information which can be very useful for a multi-
dimensional computational fluids dynamics (CFD) two-phase codes where different
flow regimes in the flow geometry can be used.
- The simultaneous identification of LFR in different radial locations provides 2D
local flow regime maps that can be used to identify new GFR configurations, not
available with the conventional identification methods.
- The experimental apparatus needed for the identification (conductivity probe) is
simple, affordable and commonly used in two-phase flow characterization.
Statistical parameters used as flow regime identifiers

Generally, the flow regime indicator signals can not be directly used as flow regime
indicator. Instead, some statistical distributions obtained from these signals are used to
obtain the indicators. The probability distribution function (PDF) and its cumulative
version (CPDF) are frequently used. The CPDF is an integral parameter and is
therefore more stable than the PDF [29]. In addition, it has a smaller input data
requirement that makes it useful for fast identification purposes.
The complete distributions are not directly used as flow regime indicators. In order
to facilitate the classification, some statistical parameters of the distributions are
employed:
- Mean value, standard deviation and skewness of the PDF (Mean+std+skw): this
set of 3 statistical parameters represents the indicator most used in flow regime
identification work. The PDF is calculated by differentiation of the CPDF and the
mean, standard deviation and skewness are calculated in the usual method [21-22].
- Principal Component Analysis (PCA): PCA is a useful statistical technique that
compresses the data by reducing the number of dimensions, without much loss of
information. This is done by transforming the data into orthogonal components.
Those components which contribute less than a specified fraction (min_frac) of the
total variation in the input data are eliminated [30-32]. Following this
methodology, the input CPDFs are reduced from 100-200 components to fewer
than 10 (depending on the temporal length of the signal and flow regime), retaining
only those components that contribute more than 0.1 percent to the variance in the
data set [33].
- Four indexes (4IND): In order to characterize the CPDF following a simple and
fast procedure, the four indexes of the components at which 0.25, 0.5, 0.75 and 1
values of the CPDF when reached, were selected as characteristic of the CPDF
shape. Afterwards, these four values were normalized so that they fall in the range
[-1,1] [33].
- Entropy (Si): The entropy computation is based on the CPDF distribution of the
area-averaged void fraction or local bubble chord length. The area-averaged void
fraction entropy (S) and the bubble chord length entropy (SB) are defined as a
conventional Shannon entropy. Thus, the entropy provides quantitative information
about the disorder in the area-averaged void fraction or bubble chord length
distributions. Entropy is able to reduce the relevant information of the flow
regimes to only one significant variable. In addition, the computation time for
entropy is much lower than for PCA. Additionally, when using neural networks
using one variable as input results in less training time and faster predictions [34].
The signal duration needed for obtaining good classification results depends on the
type of signal and statistical parameters used. If area-averaged void fraction signals
obtained from an impedance meter are used a duration of 0.1 seconds is enough for
obtaining accurate classifications. If conductivity probe signals are used a temporal
length of 1 second is necessary for accurate classifications to be obtained (96%
agreement with visual information). This difference in the minimum temporal length
requirement can be explained since conductivity probes provide local measurements
that are more dependent on the input temporal length.
More sophisticated flow regime indicators are also available. Ruzicka et al. [35]
used the Kolmogorov entropy from pressure signals in order to discriminate between
flow regimes. Furthermore, Zhang and Shi [36] developed a methodology that could be
used to identify flow regimes in a two-phase flow loop using the Shannon entropy of the
power spectral density, PSD, of pressure oscillation signals in a two-phase flow loop. In
addition, Elperin and Klochko [37] used the wavelet transform in order to identify two-
phase flow regimes. In this case, the basis of the wavelet decomposition was the entropy
and the sparsity of differential pressure transducer signals. Finally, Lee et al. [38] used
the chaotic characteristics of time sequential impedance probe signals in order to
identify flow regimes in upward vertical two-phase flows. However, these approaches
require time consuming calculations and the indicators obtained are difficult to classify
by conventional methods. These facts make these approaches not applicable for
practical purposes at the present time.
Table 1 summarizes the main experimental techniques and flow regime indicators
used in flow regime identification methodologies in the last decades. If a more detailed
explanation of the methodologies is needed, please refer to the published references
given in the table.
Table 1. Flow regime indicators. Most important flow regime indicators and experimental
techniques used to obtain them.
Experimental Technique Flow Regime Indicator Flow Regime Ref.
Researcher sight 9-12
Visual information
CCD cameras 13-14
X-ray PDF of <> 15
Pressure transducers PDF of <> 16-19
PDF of <>:
21-22
mean, std and skw GFR
CPDF of <> 29
Impedance meters
CPDF of <>:
4IND, PCA, mean, std and 34
skw, entropy
Flow meters jf, jg values 23
CPDF of <> 28
Conductivity probes CPDF of <>: LFR
4IND, PCA, mean, std and 33-34
skw, entropy
Flow Regime Classifiers

In the first flow regime identification works the flow regime indicator was the
visual information obtained by the researcher. Then, the only flow regime classifier is
the researcher him- or herself [9-14]. Obviously, this mapping methodology is strongly
subjective, and the results are highly dependent on the researcher.
Progress in flow regime classifiers was introduced when some statistical parameters
of the void fraction signals were introduced as flow regime indicators [16-19, 38-39].
Thus some rules regarding the mean and standard deviation of the PDF distributions
were used to perform the classification. However, these rules were decided based on
the researchers knowledge, so the objectivity of the classification is not guaranteed. In
addition, some classification problems arise in the transition boundaries between flow
regimes since the rules are static by nature. Nevertheless, this type of classifier is still
used.
A significant advance in objective flow regime mapping was achieved by the use of
artificial neural networks (ANN) [19, 21-22]. Artificial neural networks (ANNs) are an
emerging tool, which has been shown to be effective in solving a wide range of
problems including many applications in engineering, finance and industry [40]. ANNs
are also able to self-identify these complicated relationships, generating non-parametric
methods which are capable of non-linear mapping. In the case of flow regime
identification, where the traditional classifier systems present some difficulties in
classification, this ANN advantage implies an enormous simplification. Another
important advantage of the ANN approach is its fast response, which enables it to be
included in more complex procedures, such as decision trees and optimization
applications. More information about the general features and working principles of
ANNs can be found in [41-43].
The specific use of a certain ANN determines its structure and therefore its learning
rule. In addition, the number of layers, nodes and epochs (one sweep through the entire
training data) should be also specified. In the case of two-phase flow regime
identification, two ANN types have been used as classifiers: Self-Organized Neural
Networks (SONN) and Probabilistic Neural Networks (PNN):
- Self-Organized Neural Network (SONN): This neural network architecture is
trained without supervision and was developed by Kohonen [44]. It consists of two
layers that during the training procedure adjust themselves based on regularities and
correlations in the training patterns input data. Specifically, Kohonen maps take
training patterns (input vectors) from a certain input space and project them into a
lower-ordered data space (determined by the output neurons) in such a way that
similar feature vectors project onto points in close proximity to one another. The
number of categories in which the data are to be classified should be specified. Fig.
(3) shows an schematic of a SONN.
Weighting matrix
(m, n)
Input values
Input layer Output layer

(n dimensions) (m categories)
Fig. (3). Self-organized neural network (SONN). SONN composed of one input layer (flow regime
indicator vector with n components), and an output layer with m flow regime categories.
- Probabilistic Neural Network (PNN): These ANNs belong to the family of radial
basis networks and are trained by supervised learning. PNNs are based on the
combination of Bayes decision strategy and Panzens method of PDF approximation
[45-47]. The classification decision is made by means of the PDF of each class
calculated using the training patterns. During training, the goal is to approximate the
PDF of the underlying training patterns distribution. Finding the PDF of the
training patterns is based on its approximation by superposition of a certain function,
in this study the Gaussian function. When applying a PNN is it important to properly
select the width parameter in the curve, which controls how spread the Gaussian
curve is. The spread parameter ranges between 0 and 1, the smaller the spread
parameter is, the more selective the corresponding neuron is, and in turn, the more
the ANN acts as a nearer neighbour classifier. PNNs train quickly, as it uses only
one pass at the same training set, not many, like in other ANNs. Fig. (4) shows an
schematic of a PNN.
Input Hidden Class Decision

nodes nodes nodes node
Input values
Fig. (4). Probabilistic neural network (PNN). PNN composed of one input layer (flow regime
indicator vector with 3 components), a hidden layer, a pattern layer and and an output layer.
In the first flow regime identification works SONNs were used [19, 21]. The main
advantage of this ANN methodology is related to its self-organization. That means that
only the number of flow regimes that are present in the test matrix should be given by
the researcher. Since the flow regime transition is a gradual process, if a higher number
of flow regimes are introduced, a more detailed FRM would be obtained. However, this
depends on the quality of the flow regime indicator used. Flow regime transition models
[1-7] consider four or five flow regimes, so this number is used in the SONN training.
Although SONNs are considered the most objective ANN classifiers, some extra
subjective information is needed for its correct performance. During the SONN training
and classification process the number epochs should be introduced. In order to select
the correct values of these variables the identification results given by the SONN should
be compared with a pattern. Usually, the FRM obtained by the visual observation is
selected as pattern and this map is completely subjective. Consequently, the results
given by the SONN are influenced; in some extent, by a subjective flow regime map as
well as the number of flow regimes expected by the researcher. In addition, the training
and identification process are very time-consuming, so this ANN architecture is not
suitable for fast identification purposes.
In addition, PNNs have been used for two-phase flow regime identification [23, 32-
33]. This type of ANN is supervised, so an external flow regime map is needed as input
pattern. Visual FRMs are used as input patterns, so the flow regime map obtained is
influenced by subjective results. However, the training and identification processes are
very fast. So its use has been extended in the last few years for on-line monitoring
applications.
The steps followed in the flow regime identification methodology using ANNs as
classifier are depicted in Fig. (5). This figure is interpreted from left to right. The first
column provides examples of images of the typical five flow regimes considered for
upward vertical flow in a pipe (B, CB, S, CT and A). The first step in the identification
procedure consists of obtaining the flow regime indicator. It should be noted here that
selected flow conditions to be utilized for neural network training must contain all the
considered flow regimes with sufficient amount from each flow regime. If this
condition is not satisfied, biased identification results are obtained [33]. The second
column of Fig. (5) shows examples of typical conductivity probe signal time series for
the flow regimes considered. The next step in the identification procedure consists of
calculating the bubble chord length CPDF. In this regard, the signal is pre-processed,
where the signals are normalized in relation to the maximum and minimum voltage
values in the raw signal and are converted into step-signals using a threshold technique.
Once the raw signals are pre-processed, the bubble residence time information is
obtained from the front sensor signal. The last step consists of calculating, sorting and
normalizing the bubble chord length. In order to transform the residence time
information into the bubble chord length distribution it is necessary to calculate the
interfacial velocity of each interface.
The next step in the identification procedure consists of selecting the set of
statistical parameters of the LB distributions that will be used as flow regime indicators.
A set composed of vectors for each flow condition is used as a flow regime indicator.
Once the set of flow regime indicator vectors is obtained, a blind training process is
applied to the ANN. The set of flow regime indicator vectors are divided into two
separated groups used for training the neural network (training group) and for obtaining
the identification results (identification group). The blind training process assures
unbiased identification results and increased objectivity of the final result. The vectors
are selected randomly, but a minimum amount of flow conditions for every flow regime
is needed in the training group in order to avoid biased results [33]. The amount of flow
conditions needed in the training group is constrained by the flow regime with the
smallest number of conditions and the minimum amount of flow conditions needed to
avoid biased results. Usually, the training group contains the 90% of the vectors and the
identification group has the other 10% of the vectors (right hand side of Fig. (5)). The
training and identification processes are applied 10 times until all the flow regime
indicator vectors, i.e. flow conditions, have been identified. It is imposed that any flow
condition can be identified twice. The outputs of the SONN are ordinal numeric values
between 1 and 5 having a natural ordering based on the observed typical sequence of
regime transitions.
In addition, in order to minimize the effect of the fuzzy flow regime boundaries on
the flow regime identification results, a committee of neural networks is assembled. If a
single neural network is used, some of the flow regime identification results are
unstable, mainly in the flow regime transition zones. If a committee of 50 neural
networks is used, flow regime map repeatability higher than 95% can be achieved [33].
The number of ANN needed in the committee depends on the experimental matrix, flow
regime indicators and flow channel geometry. In most cases 50 ANNs in the committee
are enough to obtain stable identification results [33-34]. If an ANN committee is used
the PNN architecture is preferred due to it fast training process. The same training and
identification groups are used for all the neural networks that integrate the committee.
Finally, the identification result is obtained by averaging the results provided by all the
neural networks that compose the committee. More details about the neural network
methodology can be found in [33].
Bubbly Flow
4.5
Sensor 0
4
Sensor 1
3.5
Output Voltage [V] 3
2.5
2 1
1.5
1
0.5
0
0 1000 2000 3000 4000 5000 6000
Samples [-]
Cap-Bubbly Flow
4.5
Sensor 0
4 Sensor 1
3.5
Output Voltage [V]
3
2.5 90%
2 Train. ANN 2
1.5
1
0.5
0
0 1000 2000 3000 4000 5000 6000
Samples [-]
Slug Flow Control Probe

4.5 1
Sensor 0
4 Sensor 1
3.5 0.8
Output Voltage [V]
3
2.5
0.6
10% 3
CPDF
2
1.5
1
0.4 Bubbly
Cap-Bubbly
Indent.
Slug
0.5 Chun-Tunteser
0.2
Annular
0
0 5000 10000 15000
Samples [-]
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.5 1.8 2
L (m)
Churn-Turbulent Flow
4.5
Sensor 0
4 Sensor 1
3.5
Output Voltage [V]
3
2.5
2
4
1.5
1
0.5
0
0 0.5 1 1.5 2 2.5
Samples [-] 4
x 10
Annular Flow
4.5
4
3.5
Output Voltage [V]
3
Sensor 0
2.5
2
Sensor 1
5
1.5
1
0.5
0
0 2 4 6 8
Samples [-] 4
x 10
Fig. (5). Flow regime identification methodology using ANN as classifier. Steps followed in the flow
regime identification using CPDF of LB as flow regime indicator and ANN as classifier.
Table 2 summarizes the main flow regime classifiers used in flow regime
identification methodologies in the last decades. If a more detailed explanation of the
methodologies is needed, please refer to the published references given in the table.
Table 2. Flow regime classifiers. Most important flow regime classifiers and indicators used with
them.
Flow Regime Classifier Flow Regime Indicator Flow Regime Ref.
Researcher Visual information 9-14
Researcher based on rules on
PDF of <> 15-18, 38-39
PDF
PDF of <> GFR
19, 21-22
Mean+std+skw
ANN (SONN) CPDF of <>
29
4IND, PCA
CPDF of LB
LFR 28
4IND, PCA
PDF of <>:
34
mean, std and skewness
CPDF of <> GFR 34
ANN (PNN) 4IND, PCA
jg, jf 23
CPDF of LB
LFR 33-34
4IND, PCA
Applications of Flow Regime Identification Methodologies in Thermal-

Hydraulics
Flow regime methodologies have been utilized in the last decades in applications
related to thermal-hydraulics. In this research field two-phase flow is found in vertical
upward conditions. The majority of the applications are related to the validation of flow
regime transition models [1-7]. Also, some applications have been developed in order
to obtain on-line flow regime identification systems. Both types of applications are
intended for improving the performance and safety of industrial systems that use
multiphase flows such as nuclear reactors, boilers, etc. These applications can be
classified in different ways depending on the type of flow (boiling or adiabatic), flow
channel geometry and fast classification purposes. The main applications are
summarized in the next paragraphs:
Adiabatic and boiling flows

Existing thermal-hydraulic applications involve boiling flow, however accurate
experiments in boiling flow conditions present important technical difficulties.
Consequently, adiabatic air-water flow works are widespread in the literature. Air-
water flows can be only considered to be representative of the real flow for high void
fraction values. Most of the published works on flow regime identification in two-
phase flows have been performed in air-water flows [3-5, 7, 11-19, 21-23, 28-29, 33-39,
48-49, 51-53, 54-58].
Only a few works can be found for boiling flow regime identification. In most of
these works only visual flow regime maps are available [8, 59-62]. Only the work of
Meftha and Ruggles [63] provides a flow regime identification using conductivity
probes measurements as flow regime indicator.
Flow channel geometry

Four flow channel geometry types have been used in flow regime identification
studies: round pipes, annuli, rod bundles and rectangular channels. Most of the studies
on flow regime identification have concentrated on gas-liquid two-phase flows in tubes
due to the simple geometry and many practical applications. However, in many
chemical and nuclear systems more complex geometries exist such as rod bundles. The
annulus channel is often utilized to simulate some phenomena encountered in complex
geometries such as a sub-channel of a rod bundle in a nuclear reactor core, yet it is
simple enough to perform fundamental studies. Consequently, the annulus flow channel
can be considered as an intermediate case between the pipe and rod bundle flow
channels. Rectangular channels are of interest for some special applications.
As commented, most of the works on flow regime identification have been
performed in pipes [13-19, 21-23, 28-29, 33-39]. This fact is easily explained since it
represents the simplest test section from the experimental point of view. Consequently,
the major part of the flow regime identification methodologies have been developed and
benchmarked for this flow channel geometry. Yet only some of them have been applied
to more complex flow channel geometries. In addition, the first flow regime transition
models were developed for this flow channel geometry [1-2].
Smaller amounts of published work can be found for the case of the annulus flow
channel geometry. Sadatomi and Sato [48] and Furukawa and Sekoguchi [49] studied
the flow regimes of gasliquid two-phase flows in non-circular flow ducts, including the
concentric annulus. Kelessidis and Dukler [3] and Das et al. [50] investigated the flow
patterns in vertical upward flow for concentric and eccentric annulus channels. In these
works only visual FRMs are available. In the work of Das et al. conductivity probe
signals were used to generate the flow regime indicators, though the classification was
done directly by the researcher. Sun et al. [5] investigated the cap-bubbly to slug flow
regime transition criteria in an annulus using impedance meter signals. This work
represents the first use of ANN classifiers in this flow channel geometry. Finally, the
work of Julia et al. [51] studies the axial development of gas-liquid two-phase flow
regimes in adiabatic upward vertical flow in an annular channel. The flow regime
indicator was obtained from area-averaged void fraction signals, which are measured by
impedance meters at three axial locations and SONNs have been used as the mapping
system. The obtained information has been used to analyze the axial development of
flow regime as well as to compare the predictions given by the existing flow regime
transition models.
Only a few published works dealing with flow regime identification in rod bundle
geometry can be found. The experimental complexity of this type of flow channels
explains this fact. In addition, most of the rod bundle work does not consider complete
flow channel geometries present in industrial applications. The flow regime map
dependence on the number of rods in the flow channel should be investigated since
noticeable changes can be observed between the FRMs in the different sub-channels
that compose the rod bundle. Bergles et al [59], Peterson and Williams [60] and
Osakabe et al [61] provide visual FRMs in diabatic conditions and several pressure
conditions. In these works, simplified of 2x2, 1x4 and 5x5 rod bundle geometries were
used, respectively. Meftha and Ruggles [63] provided a FRM in a 2x2 rod bundle and
diabatic conditions. They used conductivity probe signals as flow regime indicators;
although the classification was performed by the researcher. Venkateswararao et al [7]
and Mizutani et al. [11] obtained a visual FRM in adiabatic flow in 4x4 rod bundle
assemblies. Harvel et al. [52] also obtained a FRM in adiabatic flow in a hexagonal rod
bundle assembly with 36 rods. In this case, differential pressure signals were used as
the flow regime indicator; although the classification was made by the researcher. Only
the work of Paranjape et al. [53] provides flow regime maps in an 8x8 rod bundle
assembly. This work is performed in adiabatic conditions and statistical parameters of
void fraction signals are used as the flow regime identifier. The classification is
performed by a SONN.
Finally, the flow regime identification work performed in rectangular or narrow
channels is related to high performance microelectronics and special types of nuclear
reactors. Consequently, its study has been increased in the last decades. There are
some flow regime identification works in rectangular channels [48, 54-58, 62].
However, all of them are based on visual information.
Table 3. Flow regime identification works. Most important flow regime identification works
classified by the flow channel geometry.
Classification Case Flow Comments Ref.
All indicators and 13-19, 21-23, 28-29, 33-

Pipe Adiabatic
classifiers 39
Flow channel
geometry All indicators and
Annulus Adiabatic 3, 5, 48-51
classifiers
Boiling Only visual FRM 59-61, 63

Rod bundle
Adiabatic Mainly visual FRM 7, 13, 52-53
Rectangular Boiling Only visual FRM 62
channel Adiabatic Only visual FRM 48, 54-58, 62
Steady state flow or fast identification applications

Most of the flow regime identification work was performed to check the prediction
capabilities of existing models or to develop new ones [1-7]. Only a few works have
been performed in order to develop fast flow regime identification methodologies [33,
34] that can be applicable to on-line monitoring systems. These works attempt to
minimize the flow regime identification process by selecting the correct flow regime
indicators and classifiers. The most time-consuming step during the flow regime
identification process is due to the experimental signal acquisition. The best results are
obtained using the entropy or PCA components of CPDF, LB or <> as flow regime
indicators and a PNN as classifier [33]. Thus, the identification process can be done in
0.1 seconds for almost all the flow conditions and flow channel geometries.
Table 3 summarizes the main flow regime identification works classified by the
type of flow channel geometry used in the experiments.
Conclusions
Two-phase flow regime identification methodologies are important in order to
understand the physical phenomena involved in the flow regime transitions as well as to
validate the flow regime transition models. Two main components are involved in the
identification process: flow regime indicator and classifier. In the first flow regime
identification works, visual information and the researcher him- or herself were used as
the flow regime indicator and classifier. However, this identification methodology
provides subjective results. Since then, a great effort has been made in order to improve
performers and obtain more objective results, however visual flow regime maps can still
be found in recent literature.
Concerning the flow regime indicators, statistical parameters of the PDF and CPDF
distributions related to void fraction and bubble chord lengths are used. Void fraction
measurements can be obtained by differential pressure transducers or impedance meters.
Bubble chord lengths are obtained by conductivity probes. Statistical parameters
obtained from experimental signals are used as final flow regime indicators. The mean
value, standard deviation and skewness formed the first set of statistical indicators;
however more advanced statistical parameters such as those provided by PCA or Si
provide better identification results.
ANNs are widely used as flow regime classifiers. ANNs are accessible and they
provide objective identification results, however some considerations should be made in
order to obtain unbiased results. A minimum amount of flow conditions representing
every flow regime are needed. A blind training process is mandatory. In addition, a
committee of ANNs should be used in order to obtain reliable identification results in
the flow regime transition zones. Two ANN architectures have been employed. The
SONN provides more objective results since its training process is self-organized.
Though some subjective input parameters provided by the researcher are still needed.
However, the training and identification processes by SONNs are time consuming. The
PNN provides more subjective results since an input pattern is needed for the training
procedure; however the training and identification processes are very fast.
Consequently, PNNs are preferred when an ANN committee is used or for fast
identification applications.
Most of the flow regime identification work has been performed in round pipes in
adiabatic conditions. These experimental conditions have been used to benchmark the
methodologies since they represent the more affordable approach to the two-phase flow
problem. Only some of the flow regime identification methodologies have been applied
to other experimental conditions. Smaller flow regime identification work can be found
in boiling flow or more complex flow channel geometries such as annuli, rectangular
channels or rod bundles.
Nomenclature
Roman
ai Interfacial area concentration
j Superficial velocity
LB Bubble chord length
Si Entropy
Greek symbols
void fraction
Subscripts
B chord length
f liquid phase
g gas phase
Mathematical symbols
<> Area average
Abbreviations
4IND 4 index method
A Annular flow regime
ANN Artificial Neural Network
B Bubbly flow regime
CB Cap-bubbly flow regime
CFD Computer Fluid Dynamics
CPDF Cumulative Probability Distribution Function
CT Churn-turbulent flow regime
FRM Flow Regime Map
GFR Global Flow Regime
LFR Local Flow Regime
PCA Principal Component Analysis
PDF Probability Distribution Function
PNN Probabilistic Neural Network
PSD Power Spectral Density
S Slug flow regime
SONN Self-Organized Neural Network
References
[1] Y. Taitel, D. Barnea, and A. E. Dukler, Modeling flow pattern transitions for
steady upward gasliquid flow in vertical tubes, AIChE J., vol. 26, pp. 345
354, 1980.
[2] K. Mishima, and M. Ishii, Flow regime transition criteria for upward two-phase
flow in vertical tubes, Int. J. Heat Mass Transf., vol. 27, pp. 723-737, 1984.
[3] V. C. Kelessidis, and A. E. Dukler, Modeling flow pattern transitions for
upward gasliquid flow in vertical concentric and eccentric annuli, Int. J.
Multiph. Flow, vol. 15, pp. 173191, 1989.
[4] G. Das, P. K. Das, N. K. Purohit, and A. K. Mitra, Flow pattern transition
during gas liquid upflow through vertical concentric annulipart II: mechanistic
models, J. Fluids Eng.-Trans. ASME, vol. 121, pp. 902-907, 1999.
[5] X. Sun, S. Kuran, and M. Ishii, Cap bubbly-to-slug flow regime transition in a
vertical annulus, Exp. Fluids, vol. 37, pp. 458-464, 2004.
[6] T. Hibiki and K. Mishima, Flow regime transition criteria for upward two-
phase flow in vertical narrow rectangular channels, Nucl. Eng. Des., vol. 203,
pp. 117131, 2001.
[7] P. Venkateswararao, R. Semiat, and A. E. Dukler, Flow pattern transition for
gasliquid flow in a vertical rod bundle, Int. J. Multiph. Flow, vol. 8, pp. 509
524, 1982.
[8] T. Crawford and J. Weisman, Two-phase (vapour-liquid) flow pattern
transitions in ducts of non-circular cross-section and under diabatic conditions,
Int. J. Multiph. Flow, vol. 10, pp. 385-391, 1984.
[9] J. Weisman, D. Duncan, J. Gibson, and T. Crawford, Effects of fluid properties
and pipe diameter on two-phase flow patterns in horizontal lines, Int. J.
Multiph. Flow, vol. 5, pp. 437-462, 1979.
[10] P. L. Spedding and V. T. Nguyen, Regime maps for air water two phase flow,
Chem. Eng. Sci., vol. 3, pp. 779793, 1980.
[11] M. Sadatomi, A. Kawahara, K. Kano, and S. Tanoue, Flow characteristics in
hydraulically equilibrium two-phase flows in a vertical 2x3 rod bundle channel,
Int. J. Multiph. Flow, vol. 30, pp. 1093-1119, 2004.
[12] J. J. Jeong, B. Ozar, A. Dixit, J. E. Julia, T. Hibiki, and M. Ishii, Interfacial area
transport of vertical upward air-water two-phase flow in an annulus channel,
Int. J. Heat Fluid Flow, vol. 29, pp. 178-193, 2008.
[13] Y. Mizutani, A. Tomiyama, S. Hosokawa, A. Sou, Y. Kudo and K. Mishima,
Two-phase Flow Patterns in a Four by Four Rod Bundle, J. Nucl. Sci.
Technol., vol. 44, pp. 894901, 2007.
[14] K. Nilpueng, and S. Wongwises, Flow pattern and pressure drop of vertical
upward gas-liquid flow in sinusoidal wavy channels, Exp. Therm. Fluid Sci.,
vol. 30, pp. 523-534, 2006.
[15] O. C. Jones, and N. Zuber, The interrelation between void fraction fluctuations
and flow patterns in two-phase flow. Int. J. Multiph. Flow, vol. 3, pp. 273306,
1975.
[16] [N. G. Hubbard, and A. E. Dukler, The characterization of flow regimes for
horizontal two-phase flow, in Proceedings of the 1996 Heat Transfer and Fluid
Mechanics Institute. Stanford University Press, CA, pp. 100121, 1966.
[17] [N. K. Tutu, Pressure fluctuations and flow pattern recognition in vertical two-
phase gasliquid flows, Int. J. Multiph. Flow, vol. 8, pp. 443447, 1982.
[18] G. Matsui, Identification of flow regimes in vertical gas-liquid two-phase flow
using differential pressure fluctuations, Int. J. Multiph. Flow, vol. 10, pp. 711-
720, 1984.
[19] S. Cai, H. Toral, J. Qiu, and J. S. Archer, Neural network-based objective flow
regime identification in air-water two-phase flow, Can. J. Chem. Eng., vol. 72,
pp. 440445, 1994.
[20] O. C. Jones, and J. M. Delhaye, Transient and statistical measurement
techniques for two-phase flows: a critical review, Int. J. Multiph. Flow, vol. 3,
pp. 89-116, 1976.
[21] Y. Mi, M. Ishii, and L. H. Tsoukalas, Vertical two-phase flow identification
using advanced instrumentation and neural networks, Nucl. Eng. Des., vol. 184,
pp. 409420, 1998.
[22] Y. Mi, M. Ishii, and L. H. Tsoukalas, Flow regime identification methodology
with neural networks and two-phase flow models, Nucl. Eng. Des., vol. 204,
pp. 87100, 2001.
[23] H. Sharma, G. Das, and A. N. Samanta, ANN-based prediction of two-phase
gas-liquid flow patterns in a circular conduit, AICHE J., vol. 52, pp. 3018-
3028, 2006.
[24] L. G. Neal and S. G. Bankoff, A high resolution resistivity probe for
determination of local void properties in gas-liquid flow, AICHE J., vol. 9, pp.
490-494, 1963.
[25] I. Kataoka, M. Ishii, and A. Serizawa, Local formulation and measurements of
interfacial area concentration in two-phase flow, Int. J. Multiph. Flow, vol. 12,
pp. 505-529, 1986.
[26] S. T. Revankar, and M. Ishii, M., Theory and measurement of local interfacial
area using a four sensor probe in two-phase flow, Int. J. Heat Mass Transf.,
vol. 36, pp. 2997-3007, 1993.
[27] S. Kim, X. Y. Fu, X. Wang and M. Ishii, Development of the miniaturized
four-sensor conductivity probe and the signal processing scheme, Int. J. Heat
Mass Transf., vol. 43, pp. 4101-4118, 2000.
[28] J. E. Julia, Y. Liu, S. Paranjape, and M. Ishii, Local flow regimes analysis in
vertical upward two-phase flow, Nucl. Eng. Des., vol. 238, pp. 156-169, 2008.
[29] J. Y. Lee, M. Ishii, and N. S. Kim, Instantaneous and objective flow regime
identification method for the vertical upward and downward co-current two-
phase flow, Int. J. Heat Mass Transf., vol. 51, pp. 3442-3459, 2008.
[30] J. Mohamad-Saleh, and B. S. Hoyle, Determination of multi-component flow
process parameters based on electrical capacitance tomography data using
artificial neural networks, Meas. Sci. Technol., vol. 13, pp. 1815-1821, 2002.
[31] I. T. Jolliffe, Principal Components Analysis. New York: Springer, 1986.
[32] H. Demuth, and M. Beale, Neural Network Toolbox for use with Matlab: users
guide version 3. Massachusetts: The Mathworks Inc., 1998.
[33] L. Hernandez, J. E. Julia, S. Chiva, S. Paranjape, M. Ishii, Fast classification of
two-phase flow regimes based on conductivity signals and artificial neural
networks, Meas. Sci. Technol., vol. 17, pp. 15111521, 2006.
[34] J. E. Julia, L. Hernandez, S. Paranjape, and M. Ishii, On the use of area-
averaged void fraction and local bubble chord length entropies as two-phase
flow regime indicators, in 6th International Conference on Multiphase Flow.
Germany, 2007.
[35] M. C. Ruzicka, J. Drahos, J. Zahradnik, and N. H. Thomas, Intermittent
transition from bubbling to jetting regime in gas-liquid two-phase flow, Int. J.
Multiph. Flow, vol. 23, pp. 671-682, 1997.
[36] Z. Zhang, and L. Shi, Shannon entropy characteristics in two-phase flow
systems, J. Appl. Physics, vol. 85, pp. 7544-7551, 1999.
[37] T. Elperin, and M. Klochko, Flow regime identification in a two-phase flow
using wavelet transform, Exp. Fluids, vol. 32, pp. 674-682, 2002.
[38] J. Y. Lee, N. S. Kim, and M. Ishii, Flow regime identification using chaotic
characteristics of two-phase flow, Nucl. Eng. Des., vol. 238, pp. 945-957, 2008.
[39] [D. Barnea, O. Shoham, and Y. Taitel, Flow pattern characterization in 2-phase
flow by electrical conductance probe, Int. J. Multiph. Flow, vol. 6, pp. 387-397,
1980.
[40] C. M. Bishop. Neural Networks for Pattern Recognition, Oxford: Oxford
University Press., 1995.
[41] S. Haykin. Neural Networks: a Comprehensive Foundation, London:
Macmillan, 1999.
[42] J. M. Zurada. Introduction to Artificial Neural Systems, Boston: PWS Publishing
Company, 1992.
[43] A. J. Maren, C. T. Herston and R. M. Pap. Handbook of Neural Computing
Applications, Orlando: Academic Press, 1990.
[44] T. Kohonen. Self-Organization and Associative Memory, Berlin: Springer-
Verlag. 1987.
[45] D. F. Specht, Probabilistic neural networks for classification, mapping or
associative memory, in Proc. of IEEE Int. Conf. on Neural Networks, 1988,
vol. 1, pp. 527-30.
[46] D. F. Specht, Probabilistic neural networks, Neural Netw., vol. 3, pp. 109-118,
1990.
[47] E. Parzen, On estimation of a probability density function and mode, Ann.
Math. Stats., vol. 3, pp. 1065-1076, 1962.
[48] M. Sadatomi, and Y. Sato, Two-phase flow in vertical noncircular channels,
Int. J. Multiph. Flow, vol. 8, pp. 641655, 1982.
[49] T. Furukawa, and K. Sekoguchi, Phase distribution for air-water two-phase
flow in annuli, Bull. JSME, vol. 29, pp. 3007-3014, 1986.
[50] G. Das, P. K. Das, N. K. Purohit, and A. K. Mitra, Flow pattern transition
during gas liquid upflow through vertical concentric annuli. I. Experimental
investigations, J. Fluids Eng.-Trans. ASME, vol. 121, pp. 895-901, 1999.
[51] J. E. Julia, B. Ozar, A. Dixit, J. J. Jeong, T. Hibiki, and M. Ishii, Axial

development of flow regime in adiabatic upward two-phase flow in a vertical
annulus, J. Fluids Eng.-Trans. ASME, vol. 131, paper 021302, 2009.
[52] G. D. Harvel, J. S. Chang, and V. S. Krishnan, Two-phase flow regime map for
a maple-type nuclear research reactor fuel channel: Effect of hexagonal finned
bundle, Nucl Technol., vol. 118, pp. 151-161, 1997.
[53] S. Paranjape, D. Stefanczyk, Y. Liang, T. Hibiki, and M. Ishii, Two-phase flow
structure in a rod bundle geometry, J. Eng. Gas. Turbines Power-Trans. ASME,
in press, 2009.
[54] B. Lowry, and M. Kawaji, Adiabatic vertical two-phase flow in narrow flow
channels, in AICHE Symp., Series 84, 1988, pp. 133139.
[55] M. Ali, M. Kawaji, and M. Sadatomi, Adiabatic two-phase flow in narrow
channels between two flat plates, Can. J. Chem. Eng., vol. 71, pp. 657666,
1993.
[56] K. Mishima, T. Hibiki, and H. Nishihara, Some characteristics of gasliquid
flow in narrow rectangular ducts, Int. J. Multiph. Flow, vol. 19, pp. 115124,
1993.
[57] T. Wilmarth, and M. Ishii, Two-phase flow regimes in narrow rectangular
vertical and horizontal channels, Int. J. Heat Mass Transf., vol. 37, pp. 1749
1758, 1994.
[58] J. L. Xu, P. Cheng, and T. S. Zhao, Gasliquid two-phase flow regimes in
rectangular channels with mini:micro gaps, Int. J. Multiph. Flow, vol. 25, pp.
411432, 1999.
[59] A. E. Bergles, J. P. Roos and J. G. Bourne, Investigation of Boiling Flow
Regimes and Critical Heat Flux, NYO-3304-13, Dynatech Corporation,
Cambridge Massachusetts, 1968.
[60] C. L. Williams, and A. C. Peterson, Jr., Two phase flow patterns with high-
pressure water in a heated four-rod bundle, Nucl. Sci. Eng., vol. 68, pp. 115-
168, 1978.
[61] M. Osakabe, Y. Koizumi, T. Yonomoto, H. Kumamaru, and K. Tasaka, Slug to
annular flow transition during boiloff in a rod bundle under high-pressure
conditions, Nucl. Eng. Des., vol. 98, pp. 69-76, 1986.
[62] E. R. Hosler, Flow patterns in high pressure two-phase (steamwater) flow with
heat addition, in Chem. Eng. Prog. Symp., vol. 64, 1968, pp. 5466.
Chapter 4
Thermal Radiation Modeling in Multiphase Flows

Typical of Melt-Coolant Interaction
Leonid A. Dombrovsky
Joint Institute for High Temperatures, Moscow, Russia
Abstract
The chapter is concerned with radiation heat transfer modeling in multiphase disperse systems, which
are formed in high-temperature melt-coolant interactions. This problem is important for complex
interaction of the core melt with water in the case of a hypothetical severe accident in light-water nuclear
reactors. A considerable part of thermal radiation emitted by the melt droplets lies in the range of water
semi-transparency. As a result, the radiation is not completely absorbed in water and one needs to account
for radiation heat transfer between the particles which have different temperatures. The scattering of
radiation by steam bubbles and melt droplets separated from ambient water by a thin steam layer is also
important. The problem is further complicated by semi-transparency of small oxide droplets and
temperature differences between the center and surface of the melt droplets during their solidification.
Nevertheless, the specific radiative properties of the multiphase flow components allow for a simplified
approach, which is implemented in a problem-oriented CFD code. A more sophisticated approach for
visible radiation of the multiphase media is also presented. The latter is expected to be important for
optical diagnostics of the flow in small-scale experiments including those using various stimulant melts.
Introduction
The present state of the art in modeling of thermal radiation transfer in disperse
systems is characterized by numerous methods ranging from simple approaches based
on diffusion-type models to detailed mathematical procedures for a discrete or
continuous description of the process. One can find a comprehensive analysis of these
models in the book by Modest [1]. The detailed mathematics for radiative transfer is
usually too complicated to be employed in engineering solutions. Fortunately, some
specific simplifications appear to be very effective in solving the radiation heat transfer
problems in a host medium containing numerous particles, bubbles, or fibers [2]. As an
example, one can also remember the problem of thermal radiation in combustion
considered in the recent book by Viskanta [3].
One can expect that a set of engineering models elaborated for various applied
problems is complete and it is sufficient to choose simply one of the known approaches
to solve one or another heat transfer problem. But there are some examples which
demonstrate that engineers should be ready to develop a modified or new approach
adopted to specific physical conditions of the real problem.
Email address: ldombr@yandex.ru.

Thermal Radiation Modeling in Multiphase Flows Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 115
In the present paper, we consider a model for thermal radiation in the so-called fuel-
coolant interaction (FCI) which takes place in hypothetical severe accident of light-
water nuclear reactors. This process is usually treated as a penetration of the core melt
of temperature about 3000 K into the water pool. The melt jets are first fragmented to
numerous droplets by a hydrodynamic interaction with ambient water. After this
premixing stage, the thermal interaction of melt droplets with water is considered as a
major process. The physical picture of this interaction is very complex and its analysis
is beyond the scopes of the paper. Various aspects of combined hydrodynamic and
thermal FCI have been widely investigated during last two-three decades with an aim to
solve a problem of energetic steam explosion [46]. The efforts of many researchers
have been focused on hydrodynamic simulation of melt jet breakup [79]. At the same
time, the radiation heat transfer in the multiphase medium containing high-temperature
polydisperse core melt (corium) particles was not a subject of detailed analysis.
There is no doubt that the role of thermal radiation coming from droplets of high
temperature melt (usually close to the eutectic composition: 70%UO2+30%ZrO2) is of
great importance. It was first discussed about ten years ago by Dinh et al. [10], Fletcher
[11], and Dombrovsky [12, 13]. After more recent analysis, there is no doubt in the
predominant role of thermal radiation at least at the initial stage of the melt-coolant
interaction [14, 15].
The general problem of radiation heat transfer between corium particles and ambient
water can be divided into the following problems of different scale: the thermal
radiation from a single particle through a steam blanket to ambient water and the
radiation heat transfer in a large-scale volume containing numerous corium particles,
steam bubbles, and water droplets. The single-particle problem has been analyzed in
some details by Dombrovsky [12, 13]. The main attention was paid to significant
contribution of electromagnetic wave effects in the case of a very thin steam layer. The
effect of semi-transparency of nonisothermal oxide particles on their thermal radiation
has also been studied [1618]. The resulting physical pictures of particle solidification
have been reported recently by Dombrovsky [19] and Dombrovsky and Dinh [14]. The
large-scale problem of radiation heat transfer in FCI can be solved on the basis of
detailed modeling of both the spectral radiative properties and radiative transfer in
absorbing, scattering and emitting multi-temperature medium. But implementation of
advanced computational models for thermal radiation into the problem-oriented
multiphase flow codes is unpractical at the moment. For this reason, a simplified
approach based on spectral radiation balance in the computational cells (the so-called
large-cell radiation model LCRM) has been recently developed by Dombrovsky [20].
This model is sufficiently simple to be easily implemented into CFD-codes for
multiphase flow calculations.
The present-day codes used in FCI calculations are based on the known methods of
multiphase medium dynamics [21]. The state-of-the-art in computational modeling of
mixing and fragmentation of the melt in FCI has been discussed in some details by
Nourgaliev et al. [22]. We do not consider the theoretical model, numerical methods,
and computational techniques in this field. A reader can be addressed to a general study
by Dinh et al. [7] as well as to recent papers by Abe et al. [23] and Pohlner et al. [9]
which are concerned with particular models of the melt fragmentation. In the present
116 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Dombrovsky
paper, we use the regular integrated code VAPEX-P developed at Electrogorsk

Research & Engineering Center on NPP Safety (Russia). One should distinguish the
code VAPEX-D for vapor explosion modeling [24] and the code VAPEX-P for the so-
called premixture stage or non-explosive FCI modeling [2527]. In the VAPEX-P code,
three phases are considered: the liquid water, the vapor (it may be a mixture of steam
and hydrogen), and the melt. The Eulerian approach is used for water and vapor
dynamics and heat transfer, while the Lagrangian approach for the melt dynamics and
cooling. In previous versions of VAPEX-P, as well as in other theoretical models of FCI
and debris bed formation [9, 23], thermal radiation of core melt particles was considered
on the basis of too crude models, without account for real optical properties of water,
melt particles, steam bubbles, and water droplets in the visible and near-infrared spectral
ranges.
In this paper, we give a systematic presentation of the theoretical and computational
models for radiation heat transfer in FCI multiphase systems including some results of
the implementation of these models into the code VAPEX-P. It would be great to
validate the complete computational model by use of some experimental data. But it is
problematic at the moment because of two reasons. The first one is very limited
measurements of the local parameters for representative comparison with theoretical
predictions. The second reason is not so detailed and reliable hydrodynamic and thermal
FCI models realized in the code VAPEX-P. To improve the thermal model, we consider
not only the specific radiation effects but also the effect of temperature difference in
large solidifying particles. The analysis is based on the approximate model suggested by
Dombrovsky and Dinh [14]. This model was modified by Dombrovsky et al. [15] to
account for the temperature profile after solidification and also by Dombrovsky [28] to
account for partial transparency of a thin layer on the surface of opaque corium
particles.
Having in mind an importance of comparison between the computational predictions
and the measurements, we discuss an advanced radiation model based on LCRM for the
source function and subsequent integration of the radiative transfer equation along a set
of rays. This combined model can give us the spectral radiation flux from the zone of
melt-coolant interaction.
Computational Models for Radiation Transfer

To suggest an adequate model of radiation heat transfer in water containing hot
corium particles and steam bubbles, one should take into account the specific optical
properties of water in the visible and near infrared spectral ranges. The optical constants
of water in the most important wavelength range from 0.5 to 5m are shown in Fig. 1.
The values of nw and w are obtained by linear interpolation of tabulated data from
papers [29] (in the range of 0.5 2 m) and [30] (in the range of 2 5 m).
1.5
nw
1.4
1.3
1.2
1.1
1 2 3 4
, m 5
0.1 w
0.01
1E-3
1E-4
1E-5
1E-6
1E-7
1E-8
1E-9
1 2 3 4 , m 5
Fig. 1. Optical constants of water in the visible and near infrared.
One can see in Fig. 1 that water is transparent in a short-wave range and there is
a strong absorption band at the wavelength 3 m. The spectral absorption
coefficient of water can be calculated as follows:
0,w 4 w (1)
One can also introduce the characteristic penetration depth of the collimated radiation in
water: l 1 0,w . The spectral dependence of l in the most interesting intermediate
range 0.8 1.4 m is illustrated in Fig. 2.
3
10
l, mm
2
10
1
10
0
10
0.8 0.9 1.0 1.1 1.2 1.3 1.4
, m
Fig. 2. The characteristic propagation depth of collimated near-infrared radiation in water.
One can see that l decreases from about 0.5m at the visible range boundary
0.8 m to l 1 mm at the wavelength 1.38 m. It makes reasonable to consider
separately the following conventional spectral regions:
- The short-wave semi-transparency range 1.2 m, where l 10 mm. There is
a considerable radiation heat transfer between corium particles in this spectral range
because the distance between neighboring particles is usually less than 10mm. One can
use the traditional radiation transfer theory to calculate the volume distribution of
radiation power. Both absorption and scattering of radiation by particles should be taken
into account [2].
- The opacity range , where l 10 mm. In this range, one can neglect the
radiation heat transfer between the particles. The radiative transfer problem degenerates
because of strong absorption at distances comparable with both particle sizes and
distances between the particles. One can assume that radiation emitted by the particle in
this spectral range is totally absorbed in ambient water.
It was shown in paper [20] that the above two-range scheme with 1.2 m can be
used in engineering calculations.
The radiative transfer equation (RTE) for emitting, absorbing, and scattering medium
containing N components of different temperatures can be written as follows [1, 2, and
31]:

N
I r , I r , I r , d n2 p ,i (2)
4 4 i 1
The physical meaning

of Eq. (2) is evident: variation of the spectral radiation intensity
I in direction takes place due to extinction by absorption and by scattering in other
directions, as well as due to scattering from other directions (integral term) and thermal
radiation of the medium. The thermal radiation emitted by hot particles is described by
the last term of Eq. (2). The extinction coefficient is defined as follows:
N
,i (3)
i 1
where ,i is the absorption coefficient of the i-th medium component, is the

scattering coefficient of the composite medium. In the general case, the spectral

coefficients ,i , , , the scattering (phase) function , and the spectral

radiation powers p ,i depend on the coordinate r .
One can use obvious equation
p ,i ,i B Ti (4)
when every component of the medium is characterized by a definite temperature Ti .

There are two simple limiting cases of Eq. (4) applicability: isothermal particles or large
totally opaque particles. In the last case, it is sufficient to treat the value of Ti as a
surface temperature of the particles of the i-th fraction. An essentially more complex
problem should be considered for semi-transparent particles when thermal radiation
comes from the particle volume. It is a realistic situation for particles of aluminum
oxide or other light oxides used as simulant substances in experimental studies of the
core melt-coolant interaction [32, 33]. The problem of thermal radiation from semi-
transparent non-isothermal particles has been studied in some details by Dombrovsky
[1619], and a differential approximation was developed for total radiation power and
radiation field in the particle. This approach has been recently applied to the problem of
cooling and solidification of the metal oxide melt droplet in water [14]. The solution
obtained can be combined with the large-scale problem under consideration. Note than
semi-transparency of the particle surface layer should be taken into account even in the
case of optically thick corium particles. We will consider this particular problem in the
corresponding section of the paper.
It is very difficult to use the complete description of the radiation heat transfer based
on RTE (2) in the range of water semi-transparency. Therefore, the simplified radiation
models should be considered for engineering calculations. The integration of Eq. (2)
over all values of solid angle yields the following important equation of spectral energy
balance:

q p I 0 r (5a)

I 0 r c0 E r
N
p 4n2 p ,i (5b)
i 1

where q is the spectral radiation flux, p is the spectral radiation power emitted in a

unit volume of the medium, E r is the spectral radiation energy density, c0 is the
velocity of light. For brevity, the name spectral radiation energy density is often used
for the quantity I 0 . The spectral radiation flux and radiation energy density are
expressed as follows:

I 0 r , d
I r

(6a)
4

q r

, d
I r (6b)
4
The spectral balance equation (5a) is considered as a starting point for simplified
models for radiation heat transfer in multiphase disperse systems.
Opaque medium model (OMM)

In the case of not too hot particles, the main part of the thermal radiation is emitted in
the range of water opacity. Therefore, it is reasonable to ignore the specific feature of
the process in the short-wave range and assume water to be totally opaque over the
whole spectrum. This approach can be called opaque medium model (OMM).
According to OMM, thermal radiation emitted by single hot particles is absorbed in
water at very small distances from the particle. In this case, the total power absorbed by
water in a unit volume is equal to the power emitted by particles in this volume:
2
Pw Pc p d (7)
1
where 1 and 2 are the boundaries of the spectral range of significant thermal
radiation. Obviously, this model overestimates the heat absorbed in water. Besides, it
cannot be employed to distinguish the radiation power absorbed at the steam/water
interface near the particle and the power absorbed in the volume. The latter may be
important for detailed analysis of heat transfer from corium particles to ambient water in
calculations of water heating and evaporation.
Simple estimates showed that contribution of short-wave radiation increases rapidly
with the particle temperature and one cannot ignore the spectral range of water semi-
transparency when corium particles have temperature greater than 2500 K. In other
words, one can expect the OMM error to be considerable in this case.
At the same time, the present-day computer codes for multiphase flows use
computational cells of size about 510 cm or greater. All parameters of the multiphase
flow are assumed to be constant in every cell. It means that OMM is really applied to
the large (almost opaque) cell and may give fairly good results, especially at low
volume fraction of corium particles when the average distance between the neighboring
particles is greater than the mean free path of near-infrared radiation in water. It is a
favorable situation for OMM which ignores the radiation heat transfer between the
particles of different temperature.
Note that the formal use of OMM for the cells with high void fraction (volume
fraction of steam) seems to be problematic even for big cells because steam in the cell at
not too high pressure is practically transparent for thermal radiation and water droplets
cannot totally absorb the radiation emitted by corium particles.
Large-cell radiation model (LCRM)

The large-cell radiation model suggested recently by Dombrovsky [20] is based on
the assumption of negligible radiation heat transfer between the cells. In the range of
water semi-transparency, the local radiative balance in every cell yields the following
relation for radiation energy density instead of Eq. (5a):
I 0 p (8)
It means that the divergence of the spectral radiation flux is assumed to be zero in the
cell. As a result, the expressions for the integral radiation power absorbed in water can
be written as:
Pw Pc Pw1 Pw2 (9a)
2
1 2
Pw ,w p d Pw p d (9b)
1
where , w is the spectral absorption coefficient of water containing steam bubbles. The
components Pw1 and Pw2 of the absorbed power correspond to the ranges of water
semi-transparency and opacity. One can assume that Pw1 causes volume heating of
water whereas Pw2 only causes surface heating and evaporation of water near the hot
particles. Obviously, the predicted contribution of the semi-transparency range to the
total absorbed power appears to be less than the corresponding value estimated by use
of OMM. Note that LCRM does not include any characteristics of radiation scattering in
the medium.
Radiation model based on P1 approximation

The radiation balance equation (5a) can also be employed without ignoring the
radiation flux divergence. To realize such a possibility, one should find a relation
between the spectral radiation flux and radiation energy density. In the diffusion
approximation, the following representation of the radiation flux is assumed to make the
problem statement complete:

q D I 0 (10)
where D is the radiation diffusion coefficient. Sometimes, as in the monograph by

Zeldovich and Raizer [34], the term diffusion approximation is related only to the
case of Eddington approximation when
D 1 3 tr (11a)
tr tr tr 1 (11b)
where tr and tr are the transport coefficients of scattering and extinction, and is
the asymmetry factor of scattering defined as

1
4
d
(12)
4
It is known that Eqs. (10) and (11) can be derived by assuming the linear angular
dependence of the radiation intensity:

I r ,
1 0
4

I r 3q r (13)
Therefore, this approximate model is also known as the first approximation of spherical
harmonics method or P1 approximation [1, 2]. Substituting Eq. (10) into Eq. (5a), we
obtain nonhomogeneous modified Helmholtz equation for the spectral radiation energy
density:

D I 0 p r I 0 r (14)
The Marshak boundary condition can be used to complete the boundary-value problem:

D I 0 en I 0 2 (15)
The condition (15) is written here for the case of zero external radiation and no
reflection from the boundary surface. The assumption of linear angular dependence of
radiation intensity seems to be rather good for the problem under consideration because
of large optical thickness of the medium and smooth variation of the physical
parameters. Therefore, we consider P1 approximation instead of the RTE to analyze the
quality of LCRM in this paper.
Note that boundary-value problem (14)(15) is formulated for the complete
computational region (not for single cells). After solving this problem for several
wavelengths in the range of 1 , one can find the radiation power absorbed in
water:

Pw r Pw1 Pw2 (16a)
2
Pw1 r ,w I 0 r d Pw2 r p r d (16b)
1
The total radiative heat loss from the corium particles is

Pc r Pc1 Pw2 (17a)

1

Pc r p r ,c r I 0 r d (17b)
1
where ,c is the absorption coefficient of polydisperse corium particles. It is important

that Pc1 Pw1 due to heat transfer by radiation in semi-transparent medium:

1
Pc Pw 1
pr

I

0
r

d (18a)
1
, w ,c (18b)
The P1 approximation takes into account the radiative transfer between all the
computational cells. It is an important advantage of this model, especially in the case of
semi-transparent steam cells. A long-time experience in the use of P1 for solving
various engineering problems has showed that the predicted field of radiation energy
density is usually very close to the exact RTE solution. One can see that P1 gives also
the radiation flux at the boundary of the computational region. In contrast to the
radiation energy density, the radiation flux error may be significant [35]. Therefore,
more sophisticated approach should be employed to determine the radiation coming
from the FCI region. This problem is discussed in the last section of this paper.
The complete solution to the two-dimensional radiation heat transfer problem in a
multiphase flow typical of fuel-coolant interaction is too complicated even in the case
when the P1 approximation is employed. The main computational difficulty is related to
the wide range of optical thicknesses of the medium at different wavelengths. One
should consider not only the visible radiation when optical thickness of the medium is
determined by numerous particles but also a part of the near-infrared range
characterized by large absorption coefficient of water. As a result, the numerical
solution of the boundary-value problem (14)(15), generally speaking, cannot be
obtained by using the same computational mesh at all wavelengths. There is no such
difficulty in LCRM, which is simply an algebraic model and can be easily implemented
into any multiphase CFD code.
Approximate Relations for Radiative Characteristics of Multiphase

Disperse Medium
In the computational cells with not too high volume fraction of steam f vs 0.7
(hereafter water cells), it is assumed that water contains separate bubbles of steam and
polydisperse corium particles separated from ambient water by thin steam layers. In the
opposite case of steam cells, when f vs 0.7 , the multiphase medium is treated as
steam containing water droplets and corium particles.
To simplify the estimates, we assume that all the particles and bubbles are ideal
spheres. It is also assumed that absorption and scattering characteristics of a small
element of the medium can be determined on the basis of the far-field single-scattering
approximation [36]. There is no doubt that the latter assumption, which is known also as
the independent scattering approximation (ISA), is correct for the engineering problem
under consideration because particle positions are random and uncorrelated and the
distances between particles are usually greater than the particle size. In our problem, the
majority of particles and bubbles are much greater in size than the wavelength. The
latter simplifies an estimate of ISA applicability as it was done by Coquard and Baillis
[37] by use of direct Monte Carlo simulation for randomly located large opaque
spheres. It was shown that ISA underestimates extinction coefficient of the disperse
medium, but this effect is less than about 11% at volume fraction of particles f v 0.1 .
The correction factor S is well described by the mean-beam-length approach suggested
by Brewster [38]. This approach gives the following simple formula for packed beds of
large opaque spheres:
S 1 1 f v (19)
that appears to be fairly good approximation at f v 0.3 .

Since the simplified models for radiation heat transfer are used in this paper, we do
not consider the detailed angular characteristics of radiation scattering by particles and
bubbles. According to the above formulation, only two characteristics of the composite
medium are considered: absorption coefficient and transport scattering coefficient. The
general expressions for these quantities are as follows [2, 19, and 39] (for brevity,
hereafter we omit subscript ):
w c (20a)

f vs
w w0 0.75 Q b a 2 Fb a da
b a
(20b)
a30 0

f vc
c 0.75 Q c a 2 Fc a da
c a
(20c)
a30 0
tr trs trc (21a)
f vs b ,tr 2
0.75 b Qsc a Fb a da
s
tr (21b)
a30 0

f vc
0.75 c Qscc ,tr a 2 Fc a da
c
tr (21c)
a30 0
w c (22a)

f vw
w 0.75 Q w a 2 Fw a da
w a
(22b)
a30 0

f vc
c 0.75 Q c 0 a 2 Fc a da
c a
(22c)
a30 0
tr trw trc (23a)

f vw
0.75 w Qscw,tr a 2 Fw a da
s
tr (23b)
a30 0

f vc
0.75 c Qscc 0,tr a 2 Fc a da
c
tr (23c)
a30 0
Equations (20) and (21) refer to the water cells and Eqs. (22) and (23) to the
steam cells. The value w0 is the absorption coefficient of pure water (without any
particles and bubbles), w is the absorption coefficient of water with bubbles (20) or
water droplets (22), f vs , f vc , f vw are the local volume fractions of steam, corium, and
water ( f vs f vc f vw 1 ), Fb , Fc , Fw are the normalized size distributions of steam

bubbles, corium particles, and water droplets, Qa and Qstr are the efficiency factor of
absorption and the transport efficiency factor of scattering for single particles in the host
medium, while the parameter a30 can be computed from the following definition of aij :

aij a i F a da a F a da
j
(24)
0 0
Steam bubbles
The radiative properties of gas bubbles in a refracting and weakly absorbing medium
have been studied by Dombrovsky [39] on the basis of the rigorous Mie theory [40, 41].
It was shown that the efficiency factors of large steam bubbles in water can be
approximated as follows:
Qab 4 w0 3 Qscb ,tr 0.9 nw 1 (25)
Equations (25) overestimate the absolute values of Qab and Qscb ,tr by less than 5% when
a 10 and w0 a 1 . Obviously, these conditions are satisfied for the majority of
steam bubbles in the range of water semi-transparency. The resulting approximate
expressions for the absorption coefficient and transport scattering coefficient of water
containing polydisperse steam bubbles are as follows:
w 1 f vs w0 (26a)
f vs
trs 0.675nw 1 b
(26b)
a32
It is important that absorption does not depend on the bubble size distribution and
scattering does not depend on absorption coefficient of water. The only parameter
related to the bubbles which affects the transport scattering coefficient of the medium is
the ratio of steam volume fraction to the average radius of bubbles: f vs a32 b
. Note that
these equations have been verified by Dombrovsky et al. [42] in the experimental study
of radiative properties of fused quartz containing gas bubbles.
Corium particles
In the case of water cells, we assume that every particle of corium is surrounded by a
concentric steam layer that separates the particle from ambient water. It was shown by
Dombrovsky [20] that one can use the following simple formula when the steam layer
thickness is much greater than the wavelength:
Qac Qac 0 nw2 (27)
where Qac 0 is the absorption efficiency factor of the oxide particle in a vacuum. For
large opaque particles, Qac 0 , where is the spectral hemispherical emissivity of
bulk corium. In contrast to absorption, the transport efficiency factor of scattering
referred to the steam bubble cross-section is equal to the value of Qscb ,tr for steam bubble
without oxide particle:
Qsc ,tr 0.9nw 1 (28)
In this paper, the thickness of the steam layer is assumed to be much less than the
particle radius. Therefore, equation (25) can be used for the transport efficiency factor
of scattering referred to the particle radius. Note that the radiation scattering by opaque
corium particles in steam cells are determined by reflection of the radiation from the
particle surface. Assuming that the particle surface is diffusely reflecting, one can use
the theoretical value 4 9 0.44 for asymmetry factor of scattering (the negative
value of should be treated as a predominant scattering to the back hemisphere). The
resulting approximate relation for the efficiency factor of scattering of corium particle is
as follows:
Qsc ,tr 1 c 1 1.441 c (29)
It is assumed here that emissivity of corium in the range of water semi-transparency is

approximately the same as that in the near-infrared range.
Water droplets
The efficiency factor of absorption of large (as compared with the wavelength) water
droplets can be approximately calculated by use of the following expression suggested
by Dombrovsky [43] for large particles of a semi-transparent substance:
Qaw
4n w

1 exp 2 w0 a (30)
nw 12
It can be shown that the monotonic dependence of Qaw on droplet radius given by Eq.
(29) is appropriate for the use of monodisperse approximation with the Sauter radius
a32 . The resulting expression for the absorption coefficient of polydisperse water
droplets is as follows:
w
f vw 3nw
a32 1 nw
w 2

1 exp w0 a32
w
(31)
The diffraction of radiation by large water droplets is characterized by a strong peak in

the vicinity of the forward direction and can be neglected. Therefore, the transport
efficiency factor is determined by geometrical optics (reflection and refraction of the
radiation) and is independent of droplet size in the case of transparent droplets. The
radiation absorption in the droplets decreases the contribution of refraction but this
effect is insignificant in the problem under consideration and one can use an upper
estimate of scattering based on exact calculations for perfectly transparent droplets.
These calculations give the following value of transport scattering efficiency factor for
large water droplets in the visible range:
Qscw,tr 0.23 (32)
Final expressions
According to the above consideration of absorption and scattering characteristics of
the multiphase medium components, one can write the following final expressions for
absorption coefficient and transport scattering coefficient of the medium in various
computational cells:
f vc
1 f vs w0 0.75 c
(33a)
a32 nw2
f vs f vc
tr 0.675nw 1 b
c (33b)
a32 a32
w (34a)
w
1 f vc f vs 3nw

1 exp w0 a32w 0.75 c
f vc
(34b)
a32w
1 nw 2 c
a32
f vc f vw
tr 1.08 c 1 w 0.17 w (34c)
a32 a32
Equations (33) are written for water cells, and Eq. (34) for steam cells. Note that index
of refraction n nw in water cells and n 1 in steam cells.
Thermal Radiation from Single Corium Particles

Simple estimates show that thermal radiation affects considerably the temperature of
a hot corium particle. As a result, thermal radiation from single corium particles cannot
be calculated without a computational analysis of the transient thermal state of the
particle.
The transient thermal state of a solidifying semi-transparent particle can be

determined by solving a combined radiative-conductive heat transfer problem in the
particle. Strictly speaking, the general problem statement should take into account a
complex behavior of the solidifying melt. Even in the case of a pure substance or
eutectic chemical composition of the melt, the internal radiative cooling of a semi-
transparent particle may lead to existence of a two-phase zone or mushy zone in the
particle. This effect was originally discussed by Chan et al. [44]. The other physical
cause of the mushy zone formation is a non-eutectic composition of the melt when the
temperatures of liquidus and solidus corresponding to the beginning and completion of
equilibrium solidification do not coincide with each other. One can expect that the
radiative transfer in solidifying particle is complicated by the radiation scattering by
small solid particles in the mushy zone [45, 46]. An analysis of possible formation of
the mushy zone and its effect on the process is not a specific subject of this paper.
Therefore, we consider the case of optically thick particles and small temperature
difference between the liquidus and solidus [47]. It is assumed that these conditions
enable us to neglect specific radiation effects in a narrow mushy zone.
The radiative cooling and solidification of semi-transparent particles has been
recently analyzed by Dombrovsky [19] and Dombrovsky and Dinh [14]. This thermal
regime is characterized by possible internal solidification of melt particles. It may take
place for particles of weakly absorbing light oxides used as corium simulants in
laboratory experiments. The effect of predominant internal cooling of such particles is
explained by the effect of total internal reflection of the radiation emitted from a
peripheral zone of the particle [16, 17].
The one-dimensional mathematical formulation of the problem for a single semi-
transparent spherical particle is based on the following transient energy equation
accompanied by the boundary and initial conditions [14]:
T 1 2 T
Lf T 2 r k W t , r (35a)
t r r r
t 0 T Tliq (35b)
T T
r0 0 ra k hT Te (35c)
r r
Where
T Tsol T Tliq
f T (35d)
Tliq Tsol
and is the Heaviside unit step function. The last term in Eq. (35a) is a volumetric
heat loss rate due to the thermal radiation.
As was mentioned above, the problem of thermal radiation emitted by a semi-

transparent spherical particle can be formulated on the basis of the MDP0
approximation. The corresponding boundary-value problem is as follows [17, 48]:
1 d 2 dI p ,
0
r D p , 1 I p0 , 4 p , 1 n2B T (36a)
r 2 dr

dr
dI p0 ,
r 0 0 (36b)
dr
dI p0 , I p0 ,
ra D
n n2 1
(36c)
dr
where the variable radiation diffusion coefficient is
1
D
4 p ,

1 2 (36d)
r2
1 r r (36e)
r2
The spectral radiation flux from the particle and the spectral radiation power inside the
particle are calculated as follows:
p I p0 , a
q
n n2 1
(37)

w r p , 1 4n2 B T I p0 , r (38)
The subscript p is used in Eqs. (36)(38) to remember that the marked quantities are
referred to a single particle and its material. The independent variable t is omitted in
these equations because it can be considered as a parameter in the steady-state radiative
transfer formulation. The integral radiation flux from the particle and the volumetric
heat loss rate due to thermal radiation are expressed as:
2 2
Q p p
q d W w r d (39)
1 1
The data for optical properties of metal oxides near the melting temperature are very
limited. Molten alumina is used in some experiments as a simulant material instead of
corium [32]. The analysis of experimental data for infrared radiative properties of
alumina showed that this material can be considered as a gray substance near the
melting point [14]. Note that alumina is one of the most transparent materials used
instead of the real corium in model experiments. There is no data for near-infrared
properties of the oxide composition Bi2O3-WO3 and some others used in laboratory
experiments by Kudinov et al. [49]. It is natural to employ the gray model for these
compositions too. To the best of the authors knowledge, there is no data for infrared
properties of corium, i.e. binary composition of uranium dioxide and zirconium dioxide.
Experimental data by Bober et al. [50, 51] for pure UO2 were obtained in the visible
range 0.45 0.75 m, where this substance is opaque. The value of index of
absorption 0.8 corresponds to the absorption coefficient p , 4 ~ 10 7 m-1
which is in four orders of magnitude greater than the absorption coefficient of alumina.
It is known that the absorption coefficient of uranium dioxide strongly decreases by
transfer from the visible to the near-infrared spectral range [52, 53]. But this coefficient
remains much greater than that for alumina. Anderson [52] has recommended using the
Rosseland average value p 5103 m-1 for UO2 melt. It may be the minimal estimate
for the absorption coefficient of UO2. The recent study of the spectral absorption of UO2
in the temperature range from room temperature up to 1173 K by Ruello et al. [53]
confirms the high absorption in the visible and near-infrared spectral ranges. It was also
shown that the boundary of the strong optical absorption band moves to the infrared
range when the sample is heated.
Having in mind the above information on optical properties of uranium dioxide, we
employ the gray model and consider only the case of highly absorbing material of
solidifying melt droplets with the absorption coefficient p 10 4 m-1. The values of
other physical parameters of corium used in calculations are given in Table 1 (see
papers [5456] for more details).
Table 1. Physical parameters used in the calculations

, kg/m3 8000
c , kJ/(kg K) 0.6
k , W/(mK) 3
L , kJ/kg 400
Tliq , K 2840
Tsol , K 2820
c 0.85
n 2.3
The values of Tsol and Tliq from Table 1 correspond to 20% of ZrO2 in the corium [47].
It the gray approximation, equations (36)(39) can be written as follows:
1 d 2 dI p
0
2 r D p 1 I p0 4 p 1 n 2T 4 (40a)
r dr dr

1
D
4 p

1 2 (40b)
dI p0
r 0 0 (40c)
dr
dI 0p I p0
ra D
dr
n n2 1 (40d)
I p0 a
Qp

n n2 1 (40e)

W r p 1 4n 2T 4 I p0 r (40f)
Note that solution for problem (40) is insufficient even in the case of a gray particle
substance because (according to the LCRM) we should determine the spectral radiation
flux q(1) in the range of water semi-transparency.
The combined heat transfer problem formulated by equations (35) and (40) is a
typical problem of the radiative-conductive interaction. The problems of this kind have
been investigated in some details starting from the classic study by Viskanta and Grosh
[57]. The comprehensive review of combined conduction-radiation heat transfer
investigations in the period before 1998 has been given by Siegel [58, 59]. Information
on more recent studies can be found in the book by Modest [1]. In our case, the
interaction of the two modes of heat transfer is determined by three additional
dimensionless parameters: the index of refraction n , the optical thickness of the particle
p p a , and the conduction-to-radiation parameter
k
N (41)
4T3 p
which characterizes the relative part of conductive and radiative heat transfer in an
optically dense medium when N 1 . In further calculations, we use the value of Tliq
as a characteristic temperature T in Eq. (41). It is known that the problem degenerates
in the limit of highly absorbing medium when the effect of radiation on temperature
field can be estimated by introducing the effective total conductivity
16 T 3
k t k k rad k rad (42)
3 p
where T is the current local temperature. Equation (42) can be written in the form:
4 T
3

k t k 1 (43)
3 N Tliq

In our case, the parameter N is equal to 5.77 even at p 10 4 m-1. It means that the
effect of thermal radiation on the temperature profile is expected to be like the effect of
a small increase in thermal conductivity. As for effect of the semi-transparency on
radiation flux from optically thick particle, it is characterized by the difference between
the surface temperature Ts and the temperature T Ts at distance from the surface
a p , where ~ 1 . In the case of predominant radiative cooling, the relative
increase in radiation flux is estimated as follows:
Q p 1
p (44)
Q 2N
Let us consider some numerical results obtained at parameters from Table 1 and
convective heat transfer coefficient h 300 Wm-2 K-1. This average value of heat
transfer coefficient for the conditions of film boiling on the particle surface has been
used in calculations of [14, 60, and 61]. Note that it was recently confirmed by
numerical simulation of film boiling on a sphere by Yuan et al. [62]. The temperature
profiles shown in Fig. 3 indicate a considerable effect of the particle semi-transparency
especially in the central region of the particle.
This effect should be taken into account even in the case of absorption coefficient
p 4 10 4 m-1. It is explained by specific temperature profiles characterized by
significant temperature difference in a solid crust on the particle surface. This surface
layer is not optically thick at the initial stage of solidification. As a result, the time of
particle solidification depends on absorption coefficient of the particle material even in
the case of optically thick particle: the variation of absorption coefficient from
p 810 4 m-1 to 10 4 m-1 leads to about 20% decrease in the solidification time (see
Fig. 4).
T, K
2800 t =0.2s
2600
0.4s
1
2400 2
3
(A) 0.6s
4
2200
0.0 0.2 0.4 0.6 0.8 r, mm 1.0
3000
T, K
t =1s
2500
2s
3s
2000 1
2 4s
3 (B)
4
1500
0 1 2 r, mm 3
a 1 mm (A) and 3 mm (B)

Fig. 3. Transient temperature profiles in solidifying particles of radius
at various absorption coefficient of the particle substance: 1 10 m-1, 2 2 10 m-1, 3
4 4
4 10 4 m-1, 4 8 10 4 m-1.
_1.0
rf
0.8
(A)
1
0.6 2
3
4
0.4
0.2
0.0
0.0 0.2 0.4 0.6 t, s 0.8
_1.0
rf
0.8 (B)
1
0.6 2
3
0.4 4
0.2
0.0
0 1 2 3 4 t, s
a 1 mm (A) and 3 mm (B)

Fig. 4. The relative position of solidification front in particles of radius
at various absorption coefficient of the particle substance: 1 10 m-1, 2 2 10 m-1, 3
4 4
4 10 4 m-1, 4 8 10 4 m-1.
Note that the conventional position of solidification front r f r f a shown in Fig. 4

is treated as a current position of the isotherm T f Tsol Tliq 2 . The calculated values
of the time of solidification, t sol , are in a good agreement with the theoretical
predictions based on equation (43) for the effective thermal conductivity:
0
t sol
t sol (45)
1 4 3 N
0
where t sol is the solidification time for the totally opaque particle.
The integral radiation flux from the particle is not so sensitive to the variation of
absorption coefficient (see Fig. 5), and one can neglect the effect of semi-transparency
at p 2 10 4 m-1. This result is also predicted by approximate equation (44).
2
Q , MW/m
3.0
(A)
2.5
(p)
1
2.0 2
3
1.5 4
1.0
0.5
0.0 0.2 0.4 0.6 0.8 t, s 1.0
2
Q , MW/m
3.0
(B)
2.5
(p)
2.0
1
2
3
1.5 4
1.0
0.5
0 1 2 3 t, s 4
a 1 mm (A)
Fig. 5. Time variation of integral radiation flux from solidifying particles of radius
and 3 mm (B) at various absorption coefficient of the particle substance: 1 10 m-1, 2
4
2 10 4 m-1, 3 4 10 4 m-1, 4 8 10 4 m-1.

Approximate Solution for Cooling and Solidification of Single Particles

The computational model for the transient thermal state of solidifying corium
particles and time variation of the integral radiation flux from single particles should be
radically simplified to be implemented into the multiphase CFD codes for FCIs
calculations. For the limiting case of totally opaque particles it was done by
Dombrovsky et al. [14, 15] on the basis of the linear approximation of temperature
profile in solid crust layer at t t cs :

T r , t Tliq Tliq Ts t ar rr r r
f
f (46)
f
and the parabolic temperature profile in the particle just after solidification (at t t cs )
[63-65]:
T r , t Tc t Tc t Ts t r a
2
(47)
The same idea can be used to derive ordinary differential equations for functions Ts t
and r f t in the case of optically thick particles at a finite value of the absorption
coefficient p 10 4 m-1. Two additional assumptions are made to derive the modified
equations: (1) The effect of radiation on temperature profile in the particle can be
described by the constant effective thermal conductivity:
4 k p
k t k 1 N (48)
3N 4Tliq3
and (2) The integral radiation flux from the particle can be estimated as:
~

Q ( p ) c Ts4 Te4 c 1 0.5 N
~
(49)
where c is the hemispherical emissivity of bulk corium. After the transformations

similar to those of [14], we obtain the Cauchy problem for coupled ordinary differential
equations:

Tliq Ts 1 2rf 3rf 3 Lrf2 c drf 3 rf rf rf dTs 3qt
2 2 3
(50a)
4 dt 4 dt ca
dTs qt dr f

k t a ht 1 r f dt
(50b)
dt
qt hTs Te Q ( p )
~
ht h 4 Ts3 (50c)
r f 0 1 Ts 0 Tliq (50d)
Note that we use here a simplified term for thermal radiation on the particle surface. The
complete formulation taking into account a contribution of other particles from the same
computational cell (according to the LCRM) can be found in paper [15].
It is instructive to note that Moriyama et al. [66] and Pohlner et al. [9] employed the
parabolic model to obtain approximate relations for the temperature difference between
the center and the surface of a corium particle. The so-derived approximate relations
were recommended for implementation into FCI simulation codes. It is well-known that
the parabolic model gives good results for quasi-steady heating or cooling of the particle
with heat absorption or generation inside the particle without phase changes [6365].
However, the parabolic model is not applicable to a solidifying melt droplet.
The approximate differential model (50) has two advantages over the simplest model
of isothermal particles. The first obvious advantage is that calculations of heat transfer
between the corium and water are more accurate. But it is important also that we obtain
quantitative information on the dynamics of solidification of the corium particles. The
crust layer on the surface of solidifying particles should be taken into account in the
estimates of possible further fragmentation of the melt particles. The latter is especially
important for predicting the melt explosivity at specific conditions of FCI. Obviously,
the break-up models cannot be based only on the value of the volume-averaged
temperature of the particle.
Equations (50) should be used before complete solidification of the particle, i.e. from
t 0 to t t cs which is defined by equation r f t cs 0. After solidification, one can
consider the simple problem:
dTs 3 qt
(51)
dt ca 1 0.2a k t ht
At initial time moment, one can assume Ts to be equal to Ts t cs taken from solution of
the problem (50). Note that there is no rigorous matching condition at t t cs because it
is impossible to satisfy all the physical conditions in this approximation: the
conservation of energy and the absence of jumps in heat flux and surface temperature.
Fortunately, it is not so important because the temperature difference in solid particle is
much less than that during the particle solidification.
Equations (50) and (51) are similar to those for opaque particle. The only difference
~
is in coefficients kt and instead of k and . It makes the extension of the
previously developed model very simple and there is no need in additional
computational resources to use the modified approximation in the multiphase code

VAPEX-P [15] or in other CFD codes of the same type. As was discussed above, the
corrected approximation is applicable to optically thick particles. But it is not a limiting
condition for the millimeter-size corium particles and, most likely, for the particles of
stimulant oxide compositions.
Some results of the calculations using the approximate model (50) are presented in
Fig. 6. A comparison with the exact numerical solution indicates that the effect of the
particle semi-transparency is well described by this approximate model.
1.0
_
rf
0.8 I
II
III
0.6
0.4
0.2 1 2
0.0
0 1 2 3 t, s 4
3.5
2
Q , MW/m
3.0
(p)
2.5
I
II
2.0 III
1.5
1
2
1.0
0.5
0.0 0.5 1.0 1.5 2.0 2.5 t, s 3.0
Fig. 6. The relative position of solidification front and the integral radiation flux for particles of
radius a 1 mm (1) and 3 mm (2) at various absorption coefficient of the particle substance: I
(opaque substance), II, III 10 m-1 (II calculations by using the approximate
4
model (50), III exact numerical solution).

The approximate differential model (50) can be used to predict the dynamics of
solidification of corium particles. As was noted above, the crust layer on the surface of
solidifying particles should be taken into account in the estimates of possible further
fragmentation of the melt particles. A role of the crust layer in stability of solidifying
corium droplets by action of pressure drop in steam blanket and then due to the tensile
thermal stresses in the growing crust layer has been analyzed recently by Dombrovsky
[60, 67]. These papers were concerned with the totally opaque corium particles when
the simplified version of the problem (50) for the limiting case of N can be
employed. It goes without saying that the results reported in [60, 67] can be easily
generalized for the case of optically thick but not totally opaque particles.
A Model Problem of Melt-Coolant Interaction
In this section, we consider one of the model problems of different scales analyzed in
recent study [15]. The initial diameter of the melt jet falling into water is assumed to be
equal to d j 0.1 m, the discharge of the melt through the orifice in the pressure vessel is
assumed to be at a height H j 1 m above the water pool, the depth and diameter of the
water pool are H w 8 m and d w 1 m, the total mass and initial temperature of corium
(80%UO2, 20%ZrO2) are mc 300 kg, Tc 0 3000 K, the temperature of water is
Tw 373 K, the initial pressure p 0.1 MPa. In the calculations, we assumed that heat
losses through the walls of the water pool are negligibly small.
The data for thermophysical properties of corium have been reported by Harding et
al. [68], Petukhov et al. [69], Fink [55], and Journeau et al. [56]. In the calculations of
[15] , we have used the data by Annunziato et al. [70] and Okkonen and Sehgal [54].
The integral hemispherical emissivity of bulk corium was assumed to be equal to
0.7 as in earlier versions of the code VAPEX. As was reported by Harding et al.
[68] and Fink [55], this value is about 0.85 . We do not give the results for both
values of the corium emissivity and our calculations at 0.7 can be treated as a
lower estimate of the contribution of thermal radiation. A time variation of some
integral parameters of the model problem are illustrated in Fig. 7.
One can see that practically all the corium particles reach the pool bottom during 10
seconds from the process beginning. The detailed calculations showed that about 2% of
the total mass is found in corium particles of size about 0.1 mm. The sedimentation of
these fine particles is very slow and they remain suspended in water during a long time.
300
m, kg
250
200
150 1
2
3
100
4
50
0
0 2 4 6 8 t, s 10
Fig. 7. Time variation of corium mass (the calculation by use of LCRM): 1 melt jet, 2 suspended
particles, 3 debris bed on the pool bottom, 4 total mass of corium.
Numerical Results Based on Simplified Radiation Models

Consider first the numerical data for the integral mass of corium particles observed at
a certain depth in the pool during all the process. The calculations were performed by
use of a uniform dividing of the computational region with 5 cells in the radial direction
and 45 along the axis, so that every circular cell has the radial thickness 10 cm and
height 20 cm. The typical distributions of the mass between the particles of different
size fractions at the depth hw H w 8 m (the bottom level) are presented in Fig. 8. The
calculated pictures for different values of the depth from 2m to 8m appeared to be
almost similar to each other. Therefore, having in mind the limited length of the paper,
we consider only the data for one depth value. For convenience of presentation, only
twelve fractions of particles with diameters in the ranges of d 0.5 mm, 0.51, 12,,
910, and d 10 mm are considered. The relative mass of particles of every fraction,
m , was calculated as a ratio of the mass of these particles to the total mass of corium.
One can see that thermal radiation affects significantly even the integrated over time
mass characteristics of corium particles suspended in water. The difference between
calculations by use of OMM and LCRM is also considerable. The relative mass of the
most important fraction 4 d 5 mm is about 25% less in OMM as compared with
LCRM. But this local difference in numerical data should not be considered as a
definite effect of the radiation model. One can speak only about small displacement of
the size distribution maximum. Note that size distributions of corium particles obtained
in these calculations are in good qualitative agreement with the known experimental
data [71].
_
m
0.3
1
2
3
0.2
0.1
0.0
1 2 3 4 5 6 7 8 9 10
d, mm
Fig. 8. The calculated relative mass of particles of different diameters near the pool bottom: 1 by
ignoring thermal radiation, 2 radiation according to OMM, 3 LCRM.
2000
T, K
1
1500 2
3
1000
500
0
1 2 3 4 5 6 7 8 9 10
d, mm
Fig. 9. Effect of thermal radiation on temperature of corium particles near the pool bottom: 1
calculations without thermal radiation, 2 OMM, 3 LCRM.
Obviously, thermal radiation decreases significantly the temperature of large corium

particles. This effect of radiation on heat transfer from relatively cold small particles is
not so strong. It is illustrated in Fig. 9 for corium particles near the pool bottom.
The difference between OMM and LCRM is observed only for large particles of
diameter only about d 6 mm and it is not greater than 80 K. Note that all the particles
are cold near the bottom: the maximum temperature is about 1300 K. It means that
greater differences can be found far from the bottom, in a zone of high-temperature
particles. We have an example of such difference between OMM and LCRM
calculations for large particles. For particles of diameter in the range of 9 d 10 mm
at the depth hw 4 m, the OMM gives T 2200 K instead of 2900 K in LCRM. But the
mass fraction of these large particles is very small. For particle of average size, the
difference between temperatures calculated by use of OMM and LCRM is not large. It
is illustrated in Fig. 10, where the dependences T hw for particles of several sizes are
shown. The presented mass-averaged temperature of corium particles was averaged also
over the process time. One can see that OMM gives fairly good prediction of this value
and corrections by using the LCRM are negligible.
T, K
2500
OMM
2000 II LCRM
1500
I
1000
500
2 3 4 5 6 7 hw8, m
Fig. 10. Variation of averaged temperatures of corium particles with the depth: I d 3 4 mm,
II d 6 7 mm.
The contribution of radiation to the heat transfer from corium is illustrated in

Fig. 11. The values averaged over the pool volume including the melt jet and all the
particles are presented in this figure.
W, MW
100
1
50
2
3
0
0 1 2 3 4 5 6 7 8 9 t, s 10
Fig. 11. Integrated components of heat transfer between corium and water calculation by use of
LCRM): 1 radiation emitted by corium, 2 radiation absorbed by corium, 3 convective heat
transfer.
As one can expect, thermal radiation is more important during the initial period of
FCI. Note that radiation absorbed by corium (this effect is taken into account in LCRM)
is relatively small.
On the basis of the presented numerical results, one can draw the following
conclusions: (1) The contribution of thermal radiation to heat transfer from corium is
really significant, especially at the initial stage of the process; the radiation affects not
only the temperature of corium particles but also their size distribution; (2) The integral
effect of radiation is well described by the opaque medium model, but LCRM gives
some corrections of the particle size distribution and the temperature of large melt
droplets at the initial stage of FCI. Fortunately, the algebraic relations of LCRM are
simple and the use of LCRM instead of OMM leads to only about a 20% increase in the
computational time of the code VAPEX-P.
In this section of the paper, we considered only the integral parameters of corium-
water interaction in the particular case of water at saturation temperature. It the case of a
lower temperature of water, one should take into account that considerable part of
radiation power is absorbed in water far from the steam/water interface and does not
contribute to steam production. In contrast to OMM, the LCRM includes a correct
estimate of volumetric radiation absorption in water. It is an important advantage of
LCRM which can be used to improve the current model of water heating and steam
production in the general case of arbitrary temperature of water.
Both approximations, OMM and LCRM, are based on the assumption of negligible
radiative transfer between the computational cells. It means that even a good agreement
between the results obtained by use of OMM and LCRM cannot be considered as a
confirmation of applicability of these approaches.
Verification of Large-Cell Radiation Model

To verify the main assumption of negligible radiative transfer between the
computational cells, we compare LCRM with the P1 approximation. One can solve a
two-dimensional problem boundary-value problem (14)(15) by use of the finite-
element method (FEM) and the earlier developed computer code [2, 72, and 73]. But a
good estimate can be also obtained on the basis of the following one-dimensional
formulation for spectral radiative transfer in the radial direction:
1 d dI 0
rD p r I 0 r (52a)
r dr dr
dI 0
r0 0 (52b)
dr
dI 0 I 0
r rw D (52c)
dr 2
where rw is the radius of water pool. In our particular case of constant medium
properties in every computational cell, it is convenient to use an analytical solution to
the problem (52). This solution can be easily obtained by use of matching of the general
solutions for single cells:
I 0,k r C1k I 0 ,k r C2 k K 0 ,k r p ,k ,k 2,k ,k D ,k k 1,2,, N

(53)
where k is the current computational cell number, N is the number of cells in the
radial cross-section of the computational region, I 0 x and K 0 x are the modified
Bessel functions of the first and second kind, respectively [74]. The matching
conditions at the boundaries between the neighboring cells are the continuity of the
function I 0 r and the spectral radiation flux D dI 0 dr . The resulting system of
linear algebraic equations for the coefficients of the analytical solutions is solved by use
of a standard procedure. We do not give here all the relations because of the limited
space of the paper. Note, that the analytical solution is preferable in our particular case
because the usual finite-difference solution and factorization procedure degenerates in
the limit of an optically thin computational region, i.e. in the visible spectral range.
It is interesting to consider a model problem with the following variable coefficients:
r12 1 I 01
D p (54)
r r r
which yields the radical simplification of the problem (52):
r12 1 (55a)
0 0 rw rw
rw
(55b)
2r12
with the obvious analytical solution for dimensionless function I 0 I 01 :
r 1
cosh r
(56a)
cosh rw 2 sinh rw rw
r r r1 rw rw r1 (56b)
In the limiting case of rw 1 , the analytical solution (56) degenerates to the constant
value of 2 3 . It is a good test problem for the limit of optically thin medium. We
have used this solution to check our algorithm developed for piecewise constant
medium properties.
The calculations using the P1 approximation were performed for several cross
sections of the computational region at the moments corresponding to intensive heat
transfer by thermal radiation. Some numerical data are presented in Fig. 12.
80
3
Pw , MW/m
60 t=1.5s
hw=2m
40
1
2
20
0
0.0 0.1 0.2 0.3 0.4 0.5
r, m
Fig. 12. Radial variation of radiation power absorbed by water: 1 P1 approximation, 2 LCRM.
Vertical lines the boundaries of computational cells.
To understand the computational results on radiation power absorbed by water, one

should analyze the fields of void fraction (volume fraction of steam) shown in Fig. 13.
One can see that considerable error of LCRM (about 20%) in the computational cells
near the axis is explained by high volume fraction of steam in these cells. Obviously,
the assumption of radiative balance is not correct for these semi-transparent cells. In the
water cells, i.e. in the majority of the computational cells during the process, the error of
LCRM is very small. It means that LCRM is a good approach which can be
recommended for practical calculations at realistic scale of the problem.
t = 1.5 s
8
0.90
7 0.80
0.70
0.60
0.50
6 0.40
0.30
0.20
5 0.10
hw ,m
0
0 0.25 0.5
r, m
Fig. 13. Volume fraction of steam. Calculation by use of LCRM.
Advanced Radiative Transfer Models Based on LCRM Solution

The problem discussed in this section does not concern the heat transfer calculation
in multiphase flows. But it may be important to predict thermal radiation exiting the
zone of intensive melt-coolant interaction. This short-wave radiation (in the visible and
near-infrared) is a source of additional information on the process and can be used both
in laboratory studies and in industrial alarm systems.

It is natural to use the field of spectral radiation energy density I 0 r from LCRM
calculations as a first step of the problem solution. After that, one can employ the
transport approximation of scattering (phase) function and the corresponding simplified
RTE [1, 2]:

I r , tr I r , S r (57a)

S r
4

1 tr 0

I r p r (57b)
where tr 1 and tr tr are the transport coefficients of scattering

and extinction, and is the asymmetry factor of scattering [2]. One can see that the
radiation source function S is known from LCRM solution. The subsequent
integration of Eq. (57a) by ray-tracing method is not a too difficult task. It is
additionally simplified due to constant properties of the medium and even the constant
source function in every computational cell. As a result, one can use obvious analytical
expressions instead of numerical integration. It was shown by Dombrovsky [2] that the
combined two-step solution gives fairly good results even in the case of complex
angular dependences of the radiation intensity. This result is in a qualitative agreement
with the analysis by Modest [1] on possible improving the solution obtained using the
P1 approximation.
Of course, one can use another computational model such as Monte Carlo simulation
which allows the exact numerical solution of the problem for an arbitrary scattering
function to be obtained [1]. This general procedure is successfully used for thermal
radiation from the exhaust plumes of solid-propellant rocket engines [75]. Note that the
exhaust plume radiation in the visible and near infrared is also determined by molten
and solid metal oxide particles [76]. But these micron-size aluminum oxide particles are
much smaller than corium particles in FCI. As a result, the optical thickness of the
rocket plumes is usually very large and the continuous radiative transfer model is
appropriate for this problem.
One can remember also thermal radiation of metal oxide particles (usually,
aluminum oxide or zirconium oxide) in plasma spraying [77]. This problem is an
example of the opposite limiting case of optically thin medium because of the
comparably small cross size of the two-phase jet. In the optically thin limit, the radiative
transfer problem degenerates and it is sufficient to consider thermal radiation from
single particles [78].
It was shown by Dombrovsky et al. [15] that either limiting case of optically thin or
optically thick multiphase medium can be realized in the visible spectral range in FCI
problems and the case of intermediate optical thicknesses allows employing alternative
approaches (continuous or discrete) for the problem analysis. The author prefers to use
the continuous radiation model and the two-step approximate solution described above,
which is recommended for engineering calculations.
Some computational results for radiation observed at large distance from the
cylindrical laboratory vessel with transparent walls are presented below. It is assumed
that the image of FCI zone is formed by parallel rays in the normal direction to the axis
of the vessel and the melt jet. All the parameters are assumed to be the same as in the
model problem considered above. Some coordinate dependences of the integral (over
the spectrum) radiation flux at various time moments are shown in Fig. 14. At time
moment t 1 s, there is a strongly emitting zone determined by a penetration depth of

the melt jet ( hw 3.5 m from the pool surface). The radiation flux from this upper zone
of the pool decreases with time but one can see an increasing thermal radiation from the
region of hw 3.5 m. It corresponds to fragmentation of the jet and penetration of
numerous (but not so hot) corium particles in the lower part of the pool.
200
2
q, kW/m
150 t, s
1.0
1.5
2.0
100
3.0
50
0
0 1 2 3 4 5 6 7 8
hw, m
Fig. 14. Profiles of integral radiation flux in the normal direction from a local zone near the axis.
It is important that there is a qualitative correspondence between the radiation flux

profiles shown in Fig. 14 and the time variation of corium mass in the melt jet,
suspended particles, and debris bed on the pool bottom (see Fig. 7). For instance, one
can see in Fig. 14 that radiation of the region close to the pool bottom appears at t 3 s,
i.e. practically simultaneously with the beginning of the debris bed formation shown in
Fig. 7. In addition, a sharp decrease of the radiation flux with the distance hw from the
pool surface corresponds approximately to the penetration depth of the melt jet. One can
see this characteristic kink on the curves in Fig. 14 at t 1 s and 1.5 s, but it is not
observed on the profile at t 2 s, when the corium mass remaining in the jet is very
small (see Fig. 7). The above comparison of Fig. 14 and Fig. 7 indicates that some
important parameters of the FCI can be estimated or even identified by using the
measurements of thermal radiation from the zone near the pool axis.
Some more detailed pictures of thermal radiation exiting the zone of melt-coolant
interaction are illustrated in Fig. 15 where the typical fields of the calculated integral
radiation flux in the plane of experimental observation are presented. The fast
displacement of the high-temperature corium particles towards the pool bottom is quite
clear from comparison of the pictures for t 2 s and 4 s. One can also observe the
deformation of the region occupied by suspended corium particles. Most likely, some
estimates of the particle average temperature can be also done on the basis of the
thermal radiation measurements. But this problem is beyond the scope of the present
paper. At least, one can compare the measured and calculated radiation fields to validate
the computational models of multiphase flow and heat transfer. The detailed spectral
calculations, which are not presented in this paper, showed that optical measurements at
the wavelength about 0.6 m are also representative. In other words, the fields of
spectral radiation flux in the visible range are similar to the field of the integral radiation
flux and the spectral measurements can be used in the experimental study of thermal
radiation exciting the FCI zone.
t = 2.0 s t = 4.0 s
8 8
q, kW/m 2 q, kW/m 2
7 93.0 7 4.0
87.8 3.8
82.7 3.6
6 77.5 6 3.4
72.3 3.1
67.2 2.9
62.0 2.7
5 5
56.8 2.5
hw ,m
hw ,m
51.7 2.2
46.5 2.0
4 41.3 4 1.8
36.2 1.6
31.0 1.3
3 25.8 3 1.1
20.7 0.9
15.5 0.7
2 10.3 2 0.4
5.2 0.2
1 1
0 0
0 0.25 0.5 0 0.25 0.5
r, m r, m
Fig. 15. Integral radiation flux in the registration plane.
Conclusions
The problem of thermal radiation modeling in multiphase flows typical for melt
coolant interaction is considered in some details including the contribution of the
radiation to the thermal effects and the use of radiation measurements to validate
computational models of multiphase flow and heat transfer.
To describe the thermal effects of the radiation, the recently developed large-cell
radiation model (LCRM) based on the spectral radiation balance for single
computational cells was implemented into the integral code VAPEX-P for multiphase
flow simulation of fuel-coolant interaction (FCI) in hypothetical severe accident of a
light-water nuclear reactor. The analysis of thermal radiation effects and verification of
LCRM were performed for a representative model problem.
The calculations for an example model problem showed that the role of thermal
radiation in hydrodynamic and thermal interaction of the melt jet with ambient water is
significant. The integral effect of radiation is well described by the opaque medium
model, but LCRM gives noticeable corrections of the particle size distribution and
temperature of large melt droplets at the initial stage of FCI. The verification of LCRM
by comparison with the diffusion approximation (P1) confirmed that LCRM is very
good approach for the majority of computational cells, which are characterized by not
too high volume fraction of steam. Moreover, the typical error of LCRM is moderate
even for steam cells. Therefore, this model can be recommended for engineering
calculations.
All the relations of LCRM, including those for radiative characteristics of
polydisperse multiphase medium and for emitted and absorbed radiation power, are
presented in the paper. It makes possible the use of this model in other multiphase CFD
codes. The implementation of algebraic LCRM into VAPEX-P was not a source of any
computational difficulties and the increase in the computational time of about 20% is
considered as a reasonable cost for additional possibilities of this improved radiation
model. The use of LCRM enables us to take into account the radiation heat transfer
between corium particles of different temperatures and to determine the field of spectral
radiation energy density in the computational region. It is important that LCRM
includes a correct estimate of the radiation power absorbed in water far from the
steam/water interface. This radiation power does not directly contribute to steam
production. The latter circumstance can be used to improve the model of water heating
and steam production realized in the present-day multiphase CFD codes.
The recently developed approximate model for solidification of a single opaque
corium particle is modified to include the period of the particle cooling after complete
solidification. The modified model was implemented in the code VAPEX-P to take into
account nonisothermicity of corium particles. The calculations showed that the
difference between surface and average temperatures of solidifying particle of diameter
about 5mm may reach several hundred degrees and the overall effect of particle
nonisothermicity of FCI parameters is considerable. It is especially important for the
predicted bulk temperature of corium particles during formation of a particulate debris
bed at the pool bottom. The typical increase in computational time to account for
particles nonisothermicity in comparison with the isothermal model is less than 5%, that
is acceptable for practical calculations. In addition, the LCRM is generalized to account
for transient temperature profile in large semi-transparent particles of solidifying melt.
The proposed modification of LCRM is based on approximate solution for coupled
radiative and conductive heat transfer in optically thick spherical particles of a refractive
material during the solidification of these particles in water pool. The LCRM extension
makes possible the use of this approach not only for core melt in nuclear FCIs but also
for other melt substances which are widely used in the laboratory experiments. The
numerical data demonstrate an effect of the absorption coefficient of the particle
substance on the rate of particle cooling and solidification. The new model includes also
determination of the transient position of the solidification front in the particle. The
predicted thickness of solid crust on the particle surface is used to improve the current
models for corium particles fragmentation at the final stage of the FCI premixing
process.
A combined two-step radiation model based on the field of radiation energy density
obtained by LCRM is developed to calculate the visible and near-infrared thermal
radiation at arbitrary direction from the zone of intensive melt-coolant interaction. The
calculations of thermal radiation exiting the FCI zone in the normal direction to the
water pool axis for the conditions of a model problem were performed and some
computational results for the integral radiation flux in the registration plane are
presented. It is shown that important characteristic times and some other parameters of
the FCI can be estimated or even identified by using the measurements of thermal
radiation from the zone near the pool axis. In all cases, one can compare the measured
and calculated radiation fields to validate the computational models of multiphase flow
and heat transfer.
Acknowledgemens
The author is grateful to the KTH (Royal Institute of Technology, Stockholm) and
the Russian Foundation for Basic Research (grants 07-08-00015 and 10-08-00218) for
partial financial support of this work. The author is also grateful to Prof. Truc-Nam
Dinh, Dr. Pavel Kudinov and Mikhail Davydov for useful cooperation and discussions.
References
[1] M. F. Modest, Radiative Heat Transfer, Second Edition. New York: Acad. Press,
2003.
[2] L. A. Dombrovsky, Radiation Heat Transfer in Disperse Systems. New York: Begell
House, 1996.
[3] R. Viskanta, Radiative Transfer in Combustion Systems: Fundamentals and
Applications. New York: Begell House, 2005.
[4] T. G. Theofanous, The study of steam explosions in nuclear systems, Nucl. Eng.
Des., vol. 155, no. 1-2, pp. 1-26, 1995.
[5] D. F. Fletcher, Steam explosion triggering: A review of theoretical and
experimental investigations, Nucl. Eng. Des., vol. 155, no. 1-2, pp. 27-36, 1995.
[6] G. Berthoud, Vapor explosions, Ann. Rev. Fluid Mech., vol. 32, pp. 573-611,
2000.
[7] T. N. Dinh, V. A. Bui, R. R. Nourgaliev, J. A. Green, and B. R. Sehgal,
Experimental and analytical studies of melt jet coolant interactions: A
synthesis, Nucl. Eng. Des., vol. 189, no. 1-3, pp. 299-327, 1999.
[8] M. Brger, Particulate debris formation by breakup of melt jets, Nucl. Eng. Des.,
vol. 236, no. 19-21, pp. 1991-1997, 2006.
[9] G. Pohlner, Z. Vujic, M. Brger, and G. Lohnert, Simulation of melt jet breakup
and debris bed formation in water pools with IKEJET/IKEMIX, Nucl. Eng. Des.,
vol. 236, no. 19-21, pp. 2026-2048, 2006.
[10] T. N. Dinh, A. T. Dinh, R. R. Nourgaliev, and B. R. Sehgal, Investigation of film
boiling thermal hydraulics under FCI conditions: Results of analyses and numerical
study, Nucl. Eng. Des., vol. 189, no. 1-3, pp. 251-272, 1999.
[11] D. F. Fletcher, Radiation absorption during premixing, Nucl. Eng. Des., vol. 189,
no. 1-3, pp. 435-440, 1999.
[12] L. A. Dombrovsky, Radiation heat transfer from a spherical particle via vapor
shell to the surrounding liquid, High Temperature, vol. 37, no. 6, pp. 912-919,
1999.
[13] L. A. Dombrovsky, Radiation heat transfer from a hot particle to ambient water
through the vapor layer, Int. J. Heat Mass Transfer, vol. 43, no. 13, pp. 2405-2414,
2000.
[14] L. A. Dombrovsky and T. N. Dinh, The effect of thermal radiation on the
solidification dynamics of metal oxide melt droplets, Nucl. Eng. Des., vol. 238,
no. 6, pp. 1421-1429, 2008.
[15] L. A. Dombrovsky, M. V., Davydov, and P. Kudinov, Thermal Radiation
Modeling in Numerical Simulation of Melt-Coolant Interaction, Comp. Therm.
Sci., vol. 1, no. 1, pp. 1-35, 2009.
[16] L. A. Dombrovsky, Thermal radiation of a spherical particle of semitransparent
material, High Temp., vol. 37, no. 2, pp. 260-269, 1999.
[17] L. A. Dombrovsky, Thermal radiation from nonisothermal spherical particles of a
semitransparent material, Int. J. Heat Mass Transfer, vol. 43, no. 9, pp. 1661-1672,
2000.
[18] L. A. Dombrovsky, A modified differential approximation for thermal radiation of
semitransparent nonisothermal particles: Application to optical diagnostics of plasma
spraying, J. Quant. Spectr. Radiat. Transfer, vol. 73, no. 2-5, pp. 433-441, 2002.
[19] L. A. Dombrovsky, Thermal radiation of nonisothermal particles in combined
heat transfer problems, Proc. 5th Intl Symp. Radiat. Transfer, 1722 June 2007,
Bodrum, Turkey, (dedication lecture).
[20] L. A. Dombrovsky, Large-cell model of radiation heat transfer in multiphase
flows typical for fuel-coolant interaction, Int. J. Heat Mass Transfer, vol. 50, no.
17-18, pp. 3401-3410, 2007.
[21] R. I. Nigmatulin, Dynamics of Multiphase Media, vol. 1. New York: Hemisphere,
1991.
[22] R. R. Nourgaliev, T. N. Dinh, A. T. Dinh, H. O. Haraldsson, and B. R. Sehgal,
The multiphase Eulerian-Lagrangian transport (MELT-3D) approach for modeling
of multiphase mixing in Fragmentation Processes, Prog. Nucl. Energy, vol. 42, no.
2, pp. 123-157, 2003.
[23] Y. Abe, E. Matsuo, T. Arai, H. Nariai, K. Chitose, K. Koyama, and K. Itoh,
Fragmentation behavior during molten material and coolant interactions, Nucl.
Eng. Des., vol. 236, no. 19-21, pp. 1668-1681, 2006.
[24] V. I. Melikhov, O. I. Melikhov, and A. V. Sokolin, Melt/water explosive

interaction: VAPEX-D code simulation, High Temperature, vol. 40, no. 3, pp.
428-436, 2002.
[25] V. I. Melikhov, O. I. Melikhov, Yu. A. Parfenov, and A. V. Sokolin, VAPEX-
code analysis of a melt-coolant interaction experiment on the FARO setup, Atomic
Energy, vol. 92, no. 2, pp. 95-99, 2002.
[26] V. I. Melikhov, O. I. Melikhov, and S. E. Yakush, VAPEX code-aided analysis of
large-scale experiments in corium/water interaction, High Temperature, vol. 45,
no. 4, pp. 509-517, 2007.
[27] M. V. Davydov, V. I. Melikhov, and O. I. Melikhov, Numerical analysis of
multiphase premixing of steam explosions, Proc. 3rd Intl Conf. Multiphase Flow
(ICMF-98), 812 June 1998, Lyon, France.
[28] L. A. Dombrovsky, An extension of the large-cell radiation model for the case of
semi-transparent nonisothermal particles, ASME J. Heat Transfer, 2010, in press.
[29] G. M. Hale and M. P. Querry, Optical constants of water in the 200 nm to 200 m
wavelength region, Appl. Optics, vol. 12, no. 3, pp. 555-563, 1973.
[30] V. M. Zolotarev and A. V. Demin, Optical properties of water in the wide
wavelength range 0.11m, Opt. Spectr., vol. 43, no. 2, pp. 271-279, 1977.
[31] R. Siegel and J.R. Howell, Thermal Radiation Heat Transfer, Fourth Ed., New York:
Taylor & Francis, 2002.
[32] I. Huhtiniemi, D. Magallon, and H. Hohmann, Results of recent KROTOS FCI
tests: Alumina versus corium melts, Nucl. Eng. Des., vol. 189, no. 1-3, pp. 379-
389, 1999.
[33] T. N. Dinh, Material property effect in steam explosion energetics: revisited,
Proc. 12th Intl Topical Meeting Nucl. Reactor Therm. Hydraulics (NURETH-12),
30 Sept. 4 Oct. 2007, Pittsburg, Pennsylvania, paper 150.
[34] Ya. B. Zeldovich and Yu. P. Raizer, Physics of Shock Waves and High-
Temperature Hydrodynamic Phenomena. New York: Acad. Press, 1967 (also Dover
Publ., 2002; orig. Russian ed.: Moscow: Nauka, 1966).
[35] L. A. Dombrovsky, Evaluation of the error of the P1 approximation in calculations
of thermal radiation transfer in optically inhomogeneous media, High
Temperature, vol. 35, no. 4, pp. 676-679, 1997.
[36] M. I. Mishchenko, J. W. Hovenier, and D. W. Mackowski, Single scattering by a
small volume element, J. Opt. Soc. Am. A, vol. 21, no. 1, pp. 71-87, 2004.
[37] R. Coquard and D. Baillis, Radiative characteristics of opaque spherical particles
beds: A new method of prediction, J. Thermophys. Heat Transfer, vol. 18, no. 2,
pp. 178-186, 2004.
[38] Q. Brewster, Volume scattering of radiation in packed beds of large, opaque
spheres, ASME J. Heat Transfer, vol. 126, no. 4, pp. 1048-1050, 2004.
[39] L. A. Dombrovsky, The propagation of infrared radiation in a semitransparent
liquid containing gas bubbles, High Temp., vol. 42, no. 1, pp. 133-139, 2004.
[40] H. C. van de Hulst, Light Scattering by Small Particles. New York: Wiley, 1957
(also Dover Publ., 1981).
[41] C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small
Particles. New York: Wiley, 1983.
[42] L. Dombrovsky, J. Randrianalisoa, D. Baillis, and L. Pilon, Use of Mie theory to

analyze experimental data to identify infrared properties of fused quartz containing
bubbles, Appl. Optics, vol. 44, no. 33, pp. 7021-7031, 2005.
[43] L. A. Dombrovsky, A spectral model of absorption and scattering of thermal
radiation by droplets of diesel fuel, High Temperature, vol. 40, no. 2, pp. 242-248,
2002.
[44] S. H. Chan, D. H. Cho, and G. Kocamustafaogullari, Melting and solidification
with internal radiative transfer A generalized phase change model, Int. J. Heat
Mass Transfer, vol. 26, no. 4, pp. 621-633, 1983.
[45] F. O. Oruma, M. N. ziik, and M. A. Boles, Effects of anisotropic scattering on
melting and solidification of a semi-infinite, semi-transparent medium, Int. J. Heat
Mass Transfer, vol. 28, no. 2, pp. 441-449, 1985.
[46] N. A. Rubtsov and N. A. Savvinova, Effect of radiation scattering on the
dynamics of the transitional zone during solidification of semitransparent material,
J. Appl. Mech. Tech. Phys., vol. 44, no. 5, pp. 676-680, 2003.
[47] B. T. Min, J. H. Song, Y. S. Park, and J. G. Kim, A physical and chemical
analysis of fast quenched particles of UO2 and ZrO2 mixture, J. Nucl. Mater., vol.
358, no. 2-3, pp. 243-254, 2006.
[48] L. A. Dombrovsky, Absorption of thermal radiation in large semi-transparent
particles at arbitrary illumination of the polydisperse system, Int. J. Heat Mass
Transfer, vol. 47, no. 25, pp. 5511-5522, 2004.
[49] P. Kudinov, A. Karbojian, W. Ma, and T.-N. Dinh, An experimental study on
debris formation with corium simulant materials, Proc. Int. Cong. Adv. Nucl.
Power Plants (ICAPP08), Anaheim, CA, USA, June 8-12, 2008, paper 8390.
[50] M. Bober, J. Singer, and K. Wagner, Determination of the optical constants of
liquid UO2 from reflectivity measurements, Proc. Eighth Symp. Thermophys.
Properties, Gaithersburg, MD, vol. II, pp. 234-244, 1981.
[51] M. Bober, J. Singer, and K. Wagner, Bestimmung der optischen Konstanten von
geschmolzenen Kernbrennstoffen, J. Nucl. Mater., vol. 124, pp. 120-128, 1984.
[52] E. E. Anderson, Radiative heat transfer in molten UO2 based on the Rosseland
diffusion method, Nucl. Tech., vol. 30, pp. 65-70, 1976.
[53] P. Ruello, K. D. Becker, K. Ulrich, L. Desgranges, C. Petot, and G. Petot-Ervas,
Thermal variation of the optical absorption of UO2: determination of the small
polaron self-energy, J. Nucl. Mater., vol. 328, pp. 46-54, 2004.
[54] T. Okkonen and B. R. Sehgal, Influence of melt freezing characteristics on steam
explosion energetic, Proc. ASME/JSME 4th Intl Conf. Nucl. Eng. (ICONE-4),
vol. 1, part B, pp. 923-934, 1996.
[55] J. K. Fink, Thermophysical properties of uranium dioxide (Review), J. Nucl.
Mater., vol. 279, pp. 1-18, 2000.
[56] C. Journeau, P. Piluso, and K.N. Frolov, Corium physical properties for severe
accident R&D, Proc. ICAPP04, Pittsburg, PA USA, paper 4140, pp. 1062-1071,
2004.
[57] R. Viskanta and R. J. Grosh, Heat transfer by simultaneous conduction and
radiation in an absorbing medium, ASME J. Heat Transfer, vol. 84, no. 1, pp. 63-
72, 1962.
[58] R. Siegel, Transient thermal effects of radiant energy in translucent materials,

ASME J. Heat Transfer, vol. 120, no. 1, pp. 4-23, 1998.
[59] R. Siegel, Transient effects of radiative transfer in semitransparent materials, Int.
J. Eng. Sci., vol. 36, no. 12-14, pp. 1701-1739, 1998.
[60] L. A. Dombrovsky, Approximate model for break-up of solidifying melt particles
due to thermal stresses in surface crust layer, Int. J. Heat Mass Transfer, vol. 52,
no. 3-4, pp. 582-587, 2009.
[61] L. A. Dombrovsky, A model for solid bubbles formation in melt-coolant
interaction, Int. J. Heat Mass Transfer, vol. 52, no. 5-6, pp. 1085-1093, 2009.
[62] M. H. Yuan, Y. H. Yang, T. S. Li, and Z. H. Hu, Numerical simulation of film
boiling on a sphere with a volume of fluid interface tracking method, Int. J. Heat
Mass Transfer, vol. 51, no. 7-8, pp. 1646-1657, 2008.
[63] L. A. Dombrovsky and S. S. Sazhin, A parabolic temperature profile model for
heating of droplets, ASME J. Heat Transfer, vol. 125, no. 3, pp. 535-537, 2003.
[64] L. A. Dombrovsky and S. S. Sazhin, A simplified nonisothermal model for
droplet heating and evaporation, Int. Comm. Heat Mass Transfer, vol. 30, no. 6,
pp. 787-796, 2003.
[65] L. A. Dombrovsky and W. Lipiski, Transient temperature and thermal stress
profiles in semi-transparent particles under high-flux irradiation, Int. J. Heat Mass
Transfer, vol. 50, no. 11-12, pp. 2117-2123, 2007.
[66] K. Moriyama, H. Nakamura, and Y. Maruyama, Analytical tool development for
coarse break-up of a molten jet in a deep water pool, Nucl. Eng. Des., vol. 236, no.
19-21, pp. 2010-2025, 2006.
[67] L. A. Dombrovsky, An estimate of stability of large solidifying droplets in fuel
coolant interaction, Int. J. Heat Mass Transfer, vol. 50, no. 19-20, pp. 3832-3836,
2007.
[68] J. H. Harding, D. G. Martin, and P. E. Potter, Thermophysical and
Thermochemical Properties of Fast Reactor Materials, Commiss. Eur. Commun.
Report, EUR 12402 EN, 1989.
[69] S. M. Petukhov, M. S. Trakhtengertz, R. Kh. Khasanov, and B. I. Nigmatulin,
Prediction of thermal properties of the nuclear reactor core melt, Preprint Res.
Eng. Center Nucl. Power Plant Safety (ENITs VNIIAES), Electrogorsk, No. L11
20, 1995 (in Russian).
[70] A. Annunziato, C. Addabbo, and G. Leva, OECD-CSNI international standard
problem 39 on FARO test L-14 Reference specification, Joint Res. Centre
Europ. Comiss. Tech. Note, I.96.64, 1996.
[71] D. Magallon, Characteristics of corium debris bed generated in large-scale fuel-
coolant interaction experiments, Nucl. Eng. Des., vol. 236, no. 19-21, pp. 1998-
2009, 2006.
[72] L. A. Dombrovsky and L. G. Barkova, Solving the two-dimensional problem of
thermal-radiation transfer in an anisotropically scattering medium using the finite
element method, High Temperature, vol. 24, no. 4, pp. 585-592, 1986.
[73] L. A. Dombrovsky, Approximate methods for calculating radiation heat transfer
in dispersed systems, Therm. Eng., vol. 43, no. 3, pp. 235243, 1996.
[74] M. Abramowitz and I. A. Stegun, Eds. Handbook of Mathematical Functions. New

York: Dover Publ., 1965.
[75] S. T. Surzhikov, Three-dimensional model of the spectral emissivity of light-
scattering exhaust plumes, High Temperature, vol. 42, no. 5, pp. 763-775, 2004.
[76] D. Laredo and D. W. Netzer, The dominant effect of alumina on nearfield plume
radiation, J. Quant. Spectr. Radiat. Transfer, vol. 50, no. 5, pp. 511-530, 1993.
[77] P. Fauchais, Understanding plasma spraying, J. Phys. D: Appl. Phys., vol. 37, no.
9, R86-R108, 2004.
[78] L. A. Dombrovsky and M. B. Ignatiev, An estimate of the temperature of
semitransparent oxide particles in thermal spraying, Heat Transfer Eng., vol. 24,
no. 2, pp. 60-68, 2003.
Chapter 5
Marangoni Motion of Multiphase Compound Drops
O.M. Lavrenteva and A. Nir

Department of Chemical Engineering, TECHNION Israel Institute of Technology,
Israel
Abstract
Compound drops are comprised of two of more immiscible phases. They occur in many natural and
technological processes and environments, e.g. the atmosphere, liquid membranes and liquid bi-layers,
direct contact heat exchange and phase separation processes. Compound liquid-liquid drop can model a
living eukaryotic cell. Two-phase compound drops may consist of two touching, partially or completely
engulfed parts, or have more complex geometry. Most processes involving multiphase drops are
accompanied by intensive heat and mass transfer and, thus, a significant influence of Marangoni flow is
anticipated, i.e. fluid motion induced variation of surface tension due to surfactant or temperature
gradient along the interface. In the case of completely engulfed drop, Marangoni effect induces a relative
motion of the inner particle towards the center of the outer one or in the opposite direction and, thus,
stabilizes or destabilizes the liquid shell. In this chapter we overview the theoretical advances concerning
the motion of multiphase compound drops of various types induced by externally imposed temperature
and concentrations gradient and spontaneous Marangoni motion of such systems due to heat and mass
transfer between the phases including the results of our ongoing research. Most of the works on the
subject are devoted to quasi-stationary Stokes flow and non-deformable spherical interfaces
corresponding to the case of small Reynolds and capillary numbers. Solutions of the Stokes equations are
obtained making use of special coordinate systems (bi-spherical, toroidal, or tangent spheres), depending
on the compound drop type.
Introduction
Multiphase compound drops occur in processes such as melting of ice particles in the
atmosphere, liquid membrane technology, direct-contact heat exchange between two-
immiscible liquids, separation technology, where drops and bubbles are used to capture
particulate impurities (see e.g. Sadhal et al. 1997 [1]), as well as in other industrial
operations. Such drops are also found to exist in lipid bilayer and polymer grafted
membranes in concentrated solutions (Evans Macromolecules 1988) [2], (Evans
Langmuir 1996) [3]. The configurations with partially and entirely encapsulated drops
are found also during rapid evaporation of drops in a super heated liquid (Aston COB
2003) [4] and disruptive combustion of free droplets in multi-component fuels
(Shepherd JFM 1982) [5], (Lasheras CST 1980) [6]. Hybrid drops are typical for
concentrated multi-component emulsions and might have a considerable effect on the
rheology of such media.
Several modern technologies employing compound drops are based on the idea of
protecting of an active ingredient core by enclosing it within a shell of second material.
Email address: ceolga@techunix.technion.ac.il, avinir@tx.technion.ac.il

Marangoni Motion of Multiphase Compound Drops Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 159
The active substance is released at a determined rate, time, and location to achieve
desired localized effects. These include controlled release of drugs (Tai FASEB 1993)
[7], flavors (e.g. Leclercq FFJ 2009) [8], enzymes (Hassan 1996) [9] or latent heat (Goel
IJHMT 1998) [10], (Mulligan JCR 1996) [11].
Leukocytes and other eukaryotic cells are often modeled by compound drops with
complete engulfment (Kan POF 1998) [12] (Liu JBM 2004]) [13], while aggregates of
cells and properly tailored colloidal droplets constructed in laboratory allow to
determine drops energy of adhesion of living cells (Ravaine JCIS 2002) [14]. Various
technological and natural processes involving hybrid drops are described in (Johnson
ARFM 1985) [15] and (Sadhal et al. 1997) [1].
Most processes involving multiphase drops are accompanied by intensive heat and
mass transfer and, thus, a significant influence of Marangoni flow is anticipated, i.e.
fluid motion is induced by variation of surface tension due to surfactant or temperature
gradient along the interface. Non-homogeneous temperature or concentration field in a
suspension of drops may be imposed by distant boundary conditions or it may be
caused by heat transfer between the phases that eventually results in a spontaneous
thermocapillary migration of a compound drop.
The Marangoni effect is especially important in the case when the natural buoyancy
driven motion is suppressed, e.g. in the absence of gravity force on board a spacecraft or
when the densities of the phases are nearly equal. Most of the numerous studies devoted
to the thermocapillary migration, interaction and deformation of drops were aimed to
simulate these special applications and were so far restricted to the case when the
ambient fluid is at rest far from the inclusion that is typical for the micro gravity
conditions. A comprehensive review is given in the monograph by Subramanian and
Balasubramaniam 2001 [16]. See also recent papers (Jin JDST 2008) [17] and (Kang
MST 2008) [18] and the literature cited therein.
The thermocapillary effect is of great importance also in various ground-based
processes with micro and nano characteristic dimensions, when the surface forces are
relatively large compared to the body forces. These include such modern advanced
technologies as actuation of droplet based microfluids (Darhuber ARFM 2005) [19],
(Verneuil Langmiuir 2009) [20].
Marangoni induced motion of a multiphase drop exhibit several new feautures:
Geometry change in the course of flow due to relative motion of drops parts and
various kinds of spontaneous motion induced but heat/ mass transfer between the
phases.
We begin with formulation of the equations and interfacial conditions governing
Marangoni flow and thermocapillaryinduced migration of drops and bubbles. Further,
we describe various types of compound drops and overview the theoretical advances
concerning the motion of multiphase compound drops of each type. In these we induce
motion induced by externally imposed temperature and concentration gradients, and
spontaneous Marangoni motion of such systems due to heat and mass transfer between
the phases. The discussion includes results of our ongoing research.
160 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Lavrenteva and Nir
The Marangoni Effect on the Motion of Fluid Particles in Viscous Flow

When a bubble or drop is placed in a fluid in which the temperature changes from
one place to another, temperature variations can be expected to arise at the interface.
The consequence is a variation of the local interfacial tension along the interface. This
variation causes elements of fluid in the interfacial region to experience a pull from the
lower interfacial tension side toward the side with higher interfacial tension. This pull,
expressed as tangential force per unit area, is often termed the thermocapillary stress
because its origin is from temperature gradients and capillary effects, the latter being a
term used to describe effects associated with the interface. The consequence of this
stress is to cause the fluid adjacent to the interface on either side to move.
The stress transmitted to fluid within the drop cannot cause any net force to be
exerted on the drop. The stress transmitted to the exterior fluid will cause a reaction of
the same magnitude but opposite in direction on the drop. If the drop is held fixed it
would experience this reaction force. If it is free to move in the continuous phase, it will
begin to translate in the direction of this reaction force. Since the interfacial tension in
many systems decreases with the increase of temperature, bubbles and drops are found
to swim toward warmer regions in the fluid. The schematics of the thermocapillary
induced migration of a drop is given in Fig. 1.
Fig. (1). Thermocapillary induced migration of a drop. Schematics
For axisymmetric stationary migration of non-deformable particles that is of

primary interest in this paper, the velocity of the drop is found from the zero net force
condition. The velocity of the steady Marangoni migration of a drop of radius R,
viscosity 1 and heat conductivity and diffusivity 1 , 1 in a liquid matrix with a
constant temperature gradient T was given in (Young JFM 1959) [21]:
2 R T
V T . (1)
(3 2 2 )(2 1 2 )
1
Here is the interfacial tension, while 2 and 2 denote viscosity and heat conductivity
of the liquid matrix, respectively.
The complex
V * L T T L (2)
is conventionally used as a velocity scale for problems involving Marangoni flow. In

(2), L, and denote characteristic length, variation of the surface tension and
viscosity, respectively. In this chapter, L is an equivalent radius of the compound drop,
is the viscosity of the ambient fluid and is a characteristic variation of the
surface tension along one of the interfaces, specified for each case separately.
Dimensionless parameters, governing the processes under consideration are the
Reynolds number,
Re V * L / , (3)
Peclet number,
Pe V * L / , (4)
capillary numbers at each interface involved,
Cai V * i* (5)
the ratios of the material parameters of the phases and the ratio of the volumes of the
two phases, compising the compound drop.
For most application involving Marangoni motion of drops and bubbles, negligible
Reynolds number is typical. Peclet and capillary numbers vary from very small to very
large numbers. Here we restrict our considerations to the case of negligible Pe (quasi-
stationary approximation) and small Ca (non-deformable and slightly deformable
interfaces). Thus, the dynamics of the multiphase flow is governed by the Stokes
equations,
i 2 u i p i , u i 0, xi, (6)
where i = 1,..,n, with n being the number of phases involved. i denotes the domain
occupied by phase i, while ui , pi and i are the velocity, pressure and dynamic
viscosity in the respective domain. At the interfaces separating the immiscible phases i
and j, ij , the velocity field is continuous,
ui u j 0, x ij , (7)
and the kinematic condition,
u i n ij Vnij , x ij , (8)
is satisfied as well as the stress balance (dynamic conditions),
nij H ij ij nij s ij , x ij . (9)

Here is the stress tensor, s nij nij is the surface gradient, while
H ij and ij denote the curvature and variable surface tension at the interface ij ,
respectively. Vnij is the velocity of the interface ij in normal direction that, in the
general case, depends on time and on the position at the interface. If the interface is
described by the equation f (x, t ) 0, Vnij f t f . In the case of non-deformable
interface moving with the velocity V ij , Vnij V ij nij . Note, that the reader should not
be deceived by the linear appearance of the above problem since, once the interfacial
deformation or the dependence of material properties such as surface tension on field
variables (as described below) are taken into account, the system becomes highly non-
linear.
The last term in the right-hand-side of (9) is a so-called Marangoni, or thermocapillary
traction. Obviously, when the process is accompanied with heat or mass transfer and the
temperature or concentration varies along the interface, the interfacial tension that
depends on temperature is not constant, and tangential stress jump occurs at the
interface. In order to determine this jump, one must solve a simultaneous transport
problem,
T i
ui T i i 2T i , xi, (10)
t
with appropriate boundary conditions (to be specified for every particular case). Here
T i and i denote temperature and thermal diffusivity, respectively.
The limit of vanishing Peclet number yields the quasi-steady approximation'',
whereupon the temperature field at every instant is computed by solving a Laplace
equation. An implication of this assumption is that the velocity and temperature fields
are established on a time scale that is much shorter than the time scale of geometrical
change of the boundaries. The simplified description significantly facilitates the solution
by allowing us to employ powerful methods for linear differential equations, such as
series expansions and surface potentials. As soon as the velocity field is available, the
Eulerian description of the interface is advanced in time based on a kinematic condition
involving the normal component of the interfacial velocity.
The presence of impurities is known to have a noticeable effect on the motion of
drops and bubbles and on deformation of the interfaces. Not only surface tension but
also the interface mobility depends strongly on the surfactants concentration that, in
turn, is determined by additional processes on the interface, such as adsorption-
desorption, surface diffusion and convection, etc. A discussion of the influence of
various interfacial phenomena on the Marangoni flow is given in (Edwards 1991) [22].
However, in the case when the surfactants concentration on the surface is predominantly
determined by the mass transfer in the bulk and when it does not affect any material
properties except for the surface tension coefficient (weak surfactant), the governing
equations are the same as in the case of thermocapillary motion.
In what follows we consider two main types of transport processes: The first
imposed by an external temperature field and the second, termed spontaneous,
induced by lack of equilibrium across respective interface resulting in the interfacial
heat or mass transfer. In all that follows a reference to a heat transfer and temperature
field applies also to mass transfer and concentration field when no special activities
occurs on fluid-fluid interface (diffusion controlled transfer of weak surfactant's
transfer) and vice versa. The motion that results from both a temperature field and a
residual contamination applied at the surface of a liquid system was analyzed in (Lyell
ASR 1993) [23].
Multiphase Compound Drops: Various Types

The first thorough analysis of static two-liquid drops configuration was performed by
Torsa and Mason (Torsa JCIS 1970) [24]. In a dynamic situation when the interfacial
forces dominate over the viscous forces in each phase, and the capillary number
associated with each of the phases is vanishingly small, the deformation of these
interfaces due to the flow can be neglected and, hence, the drop preserves its static
configuration at all time. The results of [24] are briefly reviewed in this section.
Various typed of compound drops are presented in Fig. 2. The occurrence of a
particular type of configuration (complete engulfment, partial engulfment or non-
engulfment) is determined solely by surface tension between the 3 pairs of media
involved. When two immiscible liquid drops, designated as phase-1 (yellow) and 3
(red), suspended in a third immiscible liquid, phase-2 (blue), are brought into contact,
the resulting equilibrium configuration is readily predicted from the interfacial tensions
between phases i and k, ik , and the - spreading coefficients
Si jk ij ik , (i j k 1, 2,3) . In terms of the convention 12 23 ( S1 0) , phase-
1 is completely engulfed by phase-3 when S3 0 (see Fig. 2 (b), (d) and (f)); no
engulfing occurs when S 2 0 (Fig. 2 (a) and (c)); while S1 , S 2 , S3 0 leads to partial
engulfing and formation of 2-phase drops (Fig. 2 (e)), with three interfaces being
segments of spheres intersecting with angles i , which are determined by the force
balance at the triple junction,
cos i 2jk ij2 ik2 2 ij ik (11)
Fig. (2). (a) separated drops, (b) complete engulfment, separated interfaces, (c) touching drops, (d)
complete engulfment, touching interfaces, (e) partial engulfment, (f) complex compound drop.
Phases 1, 2 and 3 are marked by yellow, blue and red, respectively.
Touching drops configurations are limiting cases of partial engulfment, 2 0 or

3 0 and also serve as approximation of the cases with complete, Fig 2 (d), and no
engulfment, Fig 2 (c), when the interfaces approach each other as a result of the motion
and are separated by a thin fluid film. The resulting configuration of a compound drop
depends on the volumes of the phases as well as on contact angles. Several examples of
partially engulfed compound drop are shown in Fig 3. Note that one of the phases
comprising a compound drop may be solid or gaseous for all types of geometrical types
of compound drop.
Fig. (3). Various types of partially engulfed compound drops. Phases 1, 2 and 3 are marked by
yellow, blue and red, respectively. First column: 1 2 3 2 3. Second column:
1 2 2, 3 . Third column: 1 17 18, 2 , 3 18. Fourth column:
1 18, 2 17 18, 3 . Volume ratio between phases 1 and 3 equals 8, 1 and 1/8 for the
first, second and third rows, respectively
All the configurations of compound drops except for the complex one in Fig, 2 (f)
are axisymmetric. In what follows various orthogonal coordinate systems are employed
to study the axisymmetric dynamics of such drops. Some details concerning these
coordinate systems are presented in the following section.
Curvilinear Coordinates
For the axi-symmetric flow, it is convenient to apply orthogonal coordinates,
, , conjugate to the cylindrical system z , , by a transformation
z i f ( i ), , (12)
having the interfaces as coordinate surfaces. Introduce an axi-symmetric Stokes stream

function such that the velocity components are
h h
u , u (13)

where h is a metric coefficient,

d ( z i )
2
1 1 1 1
2 2 2 d ( i ) , h3 .
(14)
h1 h2 h
The stream function satisfies the following equation (Happel & Brenner 1965) [25]
E 2 E 2 i 0, x i , (15)
with
1 1
E 2 h2 . (16)

Laplace operator in such coordinate system is
h2

2
.

(17)
The choice of a specific coordinate system, where the interfaces are described by
coordinate level lines, facilitates solving problem (6)-(10). 3 systems that are used for
various types of compound drops are presented below.
Bi-spherical coordinate system

The bi-spherical coordinates , , are related to the cylindrical coordinates z , by
the analytic function z i ia cot i 2 , R, (0, ) with the components
given by
a sin a sinh
, z . (18)
cosh cos cosh cos
The metric coefficient is h cosh cos a , and the coordinate surface

const is the sphere ( z a coth ) 2 a 2 sinh 2 (see Fig. 4).
2
Fig 4: Bi-spherical coordinate system
In this curvilinear system, Laplace equation for the function c ( , ) is
sin c sin c
0, (19)
cosh cos cosh cos
and the operator E 2 takes the form
cosh cos cosh cos cosh cos

E2 sin
sin sin
2
a
(20)

h h 1 2 h , cos

The general solution of the Laplace equation (14) is expressed in the form of Fourier
series (see e.g., Lebedev 1965 [26] and Subrananian & Balasubramaniam 2001 [19])

U n ( ) Pn (cos ),
1
c (cosh cos ) 2
i 1, 2, (21)
n 0
where
U n ( ) En cosh n 1 2 Gn sinh n 1 2 , (22)
while Pn ( ) are Legendre polynomials.

The general solution of (15) in bi-spherical coordinates was given by Stimson and
Jeffery (Stimson PRSLS 1926) [27] as

3 1
(cosh cos ) 2
W
n ( )Cn 12 (cos ), i 1, 2,3, (23)
n 1
1
where Cn 12 ( ) are the Gegenbauer polynomials and with the general form of the
stream function coefficients Wni in phase i being
Wni Ani cosh n 1 2 Bni sinh n 1 2 Cni cosh n 3 2 Dni sinh n 3 2 . (24)
The tangential component of the viscous stress tensor at the interface
u u 1
h
sinh u sin u . (25)
a
Tangent spheres coordinates

Tangent-spheres coordinates ( , ) are related to cylindrical coordinates ( z , ) by
the mapping z i i / i , R, (0, ). The components and the metric
coefficient are given by

z , , h 2 2 , (26)

2 2
2 2
where h is the metric coefficient. Coordinate surface const is a sphere
2 z 1/ 2 1 4 2 (see Fig. 5).

2
Fig. (5): Tangent sphere coordinate system
c c
2 2 0, (27)

2 2
2 2 2 2

E 2 2 2 . (28)

The general solution of the Laplace equation (27) is expressed in the form (see e.g. Nir
JFM 1973) [28]
1
c ( 2 2 ) 2 L s sinh s K s cosh s J 0 ( s )ds. (29)
0
The general solution of (15) in tangent spheres coordinates was given by Cooley and
ONeill (Cooley PCPS 1969) [29] as

A s C s e B s D s e s J1 ( s )ds,
s
(30)
3
( )
2 2 2 0
where J 0 and J1 are the Bessel functions of the first kind.
Toroidal coordinate system

Toroidal coordinate system ( , ) is related to the cylindrical system by the analytic
i
function (Lebedev 1965) [26] z i ic coth with the components given by
2
a sin a sinh
z , , 0 , (31)
cosh cos cosh cos
Fig. (6). Toroidal coordinate system.
The metric coefficient h cosh cos a . A coordinate surface const is

a sphere 2 ( z a cot ) 2 a 2 sin 2 (see Fig. 6).
sinh c sinh c
0, (32)
cosh cos cosh cos
cosh cos cosh cos cosh cos

E2 sinh . (33)
sinh sinh
2
c
The general solution of (15) for the Stokes stream function expressed in toroidal
coordinates is of the form (see e.g. (Payne JFM 1960) [30] and (Ouz, 1987) [31]):

sinh 2
(cosh cos )3/ 2 0
( , ) ( , )P '1/ 2i (cosh )d , (34)
where P '1/ 2 i is the derivative of the Legendre function with respect to its argument
and ( , ) consists of a combination of four linearly independent expressions,
( , ) cos A( ) cosh B ( ) sinh

(35)
sin C ( ) cosh D ( ) sinh
with the coefficients A( ), B( ), C ( ), D( ) to be determined using the boundary

conditions in each case.
Compound Drops with Complete Engulfment

Consider a compound drop of radius R2 containing another smaller droplet of radius
R1 that moves steadily in an unbounded viscous fluid. The three different immiscible
phases are denoted by the indices 1, 2 and 3 as shown in Fig. 7.
Fig (7): Compound drop schematics
Introduce dimensionless variables choosing the large drop radius as a length scale and a
bi-spherical coordinate system. The geometry of the system in dimensionless
coordinates is shown in Fig. 7.
The interfaces of the large drop and the small droplet are described by the coordinate
surfaces 0 and , respectively. If the ratio of the radii R = R1/ R2 and
the scaled separation between the centers of the spheres d are given, then and can be
defined by
1 R 2 1 R d 1 R d 1 R2
2 2
cosh , cosh (36)

2 R 1 R d 2 1 R d
The forces exerted on the small, F1, and large, F2, drops by the flow can be computed
according to (Happel & Brenner 1965 [25])
2 i 2
Fi Ani Bni Cni Dni ,
sinh n 1
i 1, 2. (37)
Note that the domains 2 (continuous fluid) and 3 (small droplet) contain singular
points, . The requirement of stream function to be finite in these domains
results in
An2 Bn2 , Cn2 Dn2 , An3 Bn3 , Cn3 Dn3 . (38)
Substituting presentation (23), (24) into boundary conditions (7) - (9) and free stream
condition at infinity results in a finite system of 12 linear equations for the coefficients
of the stream function for any choice of n. The right-hand side of this system depends
linearly on the velocities of the drops and on the temperature distribution along each
interface that will be specified below). The force exerted on each particle by the
creeping flow due to the linearity of the problem it can be expressed as a sum (see e.g.
Tsemakh IJMF 2004 [32],
Fi U1 Fi1 U 2 Fi 2 Fi M , i 1, 2. (39)
Here U1 V 31 i z and U 2 V12 i z denote velocities of small and large drops,

respectively. Fi i is the drag on the drop i translating with a unit velocity in the direction
of z axis if the other drop is in rest. Fi j (i j ) denotes the force exerted on the quiescent
drop i by the flow induced by the motion of the drop j, while Fi M is the force exerted on
the quiescent drop i by the Marangoni flow when the drop j does not move as well.
The total force Fi exerted on each drop by a creeping flow vanishes or is balanced
by gravitational force. The latter condition allows determination of the velocities of the
drops.
Thermocapillary motion induced by an externally imposed temperature gradient

Thermocapillary-induced motion of a compound drop with complete engulfment
under externally imposed temperature gradient was studied in (Morton PoF 1990) [33]
assuming quasi-stationary temperature field. For the concentric configuration, the
authors analytically solved the problem for temperature and stream function in a
spherical coordinate system.
For the general eccentric position of the inner drop, the natural choice is a bi-
spherical coordinate system. The temperature field can be found making use of a
general solution of the Laplace equations (21) and boundary conditions, that results in
an infinite system for the coefficients in (22). Morton, Subramanian and
Balasubramaniam (Morton PoF 1990) [33] employed an alternative approach
introducing axisymmetric heat potential to obtain directly the heat flux q as

q i , (40)

or, in bi-spherical coordinate system,
cosh cos cosh cos .

2 2

q , q (41)
a 2 sin a 2 sin
It follows that satisfies
E 2 0 . (42)
The general solution of (42) is of the form

E cosh( n 1/ 2) Gn sinh( n 1/ 2) Cn 12 (cos ), (43)
1 1
(cosh cos ) 2
n
n 1
and conditions at infinity and continuity of temperature and heat fluxes at the interfaces
yield
2 2 2 as , 0
(44)
0, i i j j , , .
i
Substituting (43) into (44) results in an infinite system for the coefficients of the
presentation (43), that is to be solved by truncating (43) at some N. However, for a
special case when one of the phases is non conducting 0 , this approach
considerably facilitates the solution (see also Sadhal JCIS 1983 [34]; Feuillebois JCIS
1989 [35]; Loewenberg JCIS 1993a and JFMb [37, 37]). The obtained tangential heat
fluxes at the interfaces are substituted into the tangential stress balance (8) that allows to
determine the stream function in the form (23). The analysis of [33] revealed an unusual
far flow field pattern in the continuous fluid with a dividing
streamline 0 downstream the drop and a reverse flow region separated by the rest of
the flow by this streamline.
The general conclusion of [33] is that in the case of a normal Marangoni effect
(interfacial tension decreases with temperature), the inner droplet always moves faster
that the large one. However, stationary locations of a droplet inside the large drop are
possible if the surface tension gradients at the interfaces are of different signs.
Locomotion
In this section, we follow the development of (Tsemakh IJMF 2004) [32]. Assume
now that the ambient fluid far from the drop is quiescent and free of surfactants, and
that the concentration of the surfactant on the inner interface, C3 , is uniform. In what
follows C3 is used as the concentration scale.
For any given configuration, the scaled concentration fields ci Ci CS C3 CS
satisfy Laplace equation
2 ci 0, x i , i 1, 2. (45)
The concentration vanishes in the ambient fluid far from the drops
c2 0 | x | , (46)
and it is constant at the surface of the small drop,
c1 1, x 13 . (47)
The concentration and the mass flux are continuous through the surface of the large
drop, i.e.,
c2 c
c1 c2 , 1 , x 12 . (48)
n n
The velocity, u, and pressure (modified by adding the gravity force potential), p, fields
in each phase satisfy the quasi-stationary Stokes equations (6).
Substituting these series into boundary conditions (7)(9) results in an infinite system of
linear algebraic equations for the coefficients Eni and Gni , which are then solved for a
specified level of accuracy (see Golovin IECR 1995 [38] for more details). As soon as
the concentration field is available, the stream function can be determined using the
algorithm developed for computing the interaction of two drop under Marangoni
convection (Morton PoF 1990) [33] described in the previous section.
Self-induced motion
We discuss first the case where the motion is driven solely by the Marangoni effect.
It is obvious that if the droplet is located exactly in the center of the large drop, the
concentration depends solely on the distance to the compound drop center, thus being
uniform on the large drop interface. No motion occurs in this case. In contrast to this if
the drops are not concentric, our calculations revealed that in the common case when
C 0, the inner droplet acquires a velocity directed from the center to the interface
of the large drop. The latter drop also begins to move in the same direction. The
velocity of the droplet always exceeds that of the drop and the distance between the
centers increases. Thus, our results show that the stationary concentric configuration is
unstable.
Fig. (8). Iso-concentration contours for R = 0.4, d = 0.08 (a) = 2, (b) =0.2.
Iso-concentration contours for diffusivity ratio = 2 and = 0.2 (R = 0.4, d = 0.08)

are shown in Figure 8a and 8b, respectively. For = 1, iso-concentration lines are
circles concentric with the inner droplet. For = 0, the concentration in the outer phase
and on the surface of the large drop is uniform and, hence, no Marangoni flow is
induced.
Fig. (9). Streamline patterns for R = 0.4, d = 0.12, = 1, 3 1 2 .

(a) Laboratory reference frame. (b): Frame moving with the large drop.
Streamline patterns for R = 0.4, d = 0.12, = 0.2, 1 3 2 are shown in the

laboratory reference frame in Figure 9a and in the reference frame linked to the large
drop in Fig. 9b. The motion of the two drops in the same direction and the drift of the
inner droplet in the direction of the surface of the large one are evident in Fig 9a and Fig
9b, respectively. An interesting feature of the motion is the onset of a region of reverse
flow (Fig 9a) similar to that observed in (Morton PoF 1990) [32] for a compound drop
subjected to an externally imposed temperature gradient. The streamline pattern of the
induced flow, shown in Figure 3b in a frame of reference attached to the large drop,
reflects the integral results that the migration velocity of the inner drop is higher than
that of the outer drop, and an eventual contact is expected.
The dependence of the drops velocities on the diffusivity ratio, viscosity ratio, radii
ratio, and interface separation was studied. The results are illustrated in Figs 10 12.
The relative velocity of the drops is depicted in Fig. 10 versus separation distance for
various values of the other governing parameters. This velocity vanishes when the drops
are concentric and when they touch, and it reaches a maximum value at a certain
separation distance that is on the order of the droplets radius. A lubrication analysis of
the flow at the thin gap between the surfaces, similar to that conducted in (Loewenberg
JFM 1993) [36], shows that at small gap thickness d the lubrication resistance is of
order Fr ~ O (d 1/2 ), which yields that non-deformable drops touch at a finite time.
The dependence of the relative velocity on the radii ratio, R, and diffusivity ratio, ,
are illustrated in Figs 10 a, b respectively. For a fixed separation, the relative velocity
decreases with R and vanishes in the limit R 0, corresponding to vanishing radius of
the inner drop. With the growth of R, the velocity first increases, reaches maximum
value at R~0.5 and for larger R it decreases (see Fig. 10 a). The distance, at which the
maximum relative velocity is realized, increases with the diffusivity ratio. For a fixed
separation, the dependence of the relative velocity on the diffusivity ratio is not
monotonic, and it exhibits a maximum at some critical value of that decreases with the
growth of separation distance. At the limit 0, corresponding to the case of non-

conducting large drop, the migration velocity vanishes, while at the limit ,
corresponding to the case of non-conducting ambient liquid, the velocity tends to a
finite non-zero value (see Fig. 10 b).
Fig. (10). The inner-sphere relative velocity as a function of the separation between the interfaces.
(a): 1 3 1, 1. Dashed, dashed-dotted, dotted and solid curves correspond to R = 0.05, 0.2,
0.5 and 0.75, respectively. (b): 1 3 1, R 0.2. Solid, dashed, dotted and dashed-dotted
curves correspond to = 1, 2, 10 and 0.2, respectively. (c): 1, R 0.2. Solid, dashed and
dashed-dotted curves correspond to 1 3 1; 3 10, 1 0.1 and 3 0.2, 1 5 ,
respectively. (d): 1. R 0.2 . Solid, dashed and dashed-dotted curves correspond to
1 3 1, 0.2 and 10, respectively
The dependence of the relative velocity on the viscosity of the inner droplet and on
the ambient fluid is illustrated in Figs. 10 c and d. One can see that the influence of the
viscosity of the droplet on the migration velocity is significant for small separation
distances (Fig 10 c), while the growth of the viscosity of the large drop results in a
substantial retardation of the motion for any configuration (Fig 10 b).
The velocity of the large drop versus separation distance is shown in Fig. 11. The
dependence of the velocity on the radii ratio, diffusivity ratio and the viscosity ratios are
illustrated in plots 11a, b, c and d, respectively. In Fig. 11 a, it is evident that the
migration velocity grows with the radii ratio and it tends to zero with R 0. At a fixed
separation, velocity first increases with R, reaches maximum value at R ~ 0.5 and for
larger R it decreases.
In Fig. 11 b one can see that, for large , the velocity of the outer drop decays
monotonically with the growth of the separation distance, while for smaller it exhibits
a maximum at some d. For a fixed separation, the dependence on the diffusivity ratio is
not monotonic, and it achieves its maximum value at ~2.
Fig. (11). The outer-sphere velocity as a function of the separation between the interfaces. (a):
1 3 1, 1. Solid, dotted, dashed-dotted and dashed curves correspond to R = 0.75, 0.5, 0.2
and 0.05, respectively. (b): 1 3 1, R 0.2. Solid, dashed, dotted and dashed-dotted curves
correspond to = 1, 2, 10 and 0.2, respectively. (c): 1 1, 1, R 0.2. Solid, dashed and dashed-
dotted curves correspond to 1 2 1; 1 10, 2 0.1 and 1 0.2, 2 5 , respectively.
(d): 1, R 0.2. Solid, dashed, dotted and dashed-dotted curves correspond to 1 3 1, 0.2
and 10, respectively.
Figure 11 c demonstrates that the velocity of the large drop decays with the growth
of the viscosity of the inclusion. The influence of viscosity of the droplet on the
migration velocity is pronounced for moderate separations and becomes insignificant

when the surfaces come closer. Fig. 11 d illustrates the dependence of the aggregate
migration velocity on the ratio of its viscosity to that of the ambient medium. The
growth of the viscosity of the large drop results in a significant retardation of the motion
for any configuration.
From Fig. 11 it is evident that, similar to the relative velocity, the velocity of the
larger drop vanishes for the concentric configuration, while when the separation
distance tends to zero, the velocity of the large drop tends to a finite non-zero value.
Thus, starting from some non-concentric configuration, both drops move in the same
direction with the inner droplet having a larger velocity. The eccentricity of the system
increases. When the droplet comes closer to the interface of the large drop, its relative
motion is retarded by the strong viscous resistance and both particles continue to move
with the same speed. As was mentioned above, the drops interfaces come in contact in a
finite time. After that, depending on the interfacial tensions, the drop either separate, or
a partially engulfed configuration is established, or the drop proceeds to move as a
touching drop aggregate. Marangoni induced motion of the latter two types of
compound drops are studied in the following sections.
Combined action of gravity and Marangoni effect

In all the cases discussed thus far in this paper, the flow is driven solely by the
Marangoni traction. In many applications, however, gravitational effects also play an
important role. In this section we study the combined effect of gravity and
thermocapillarity on the motion of a compound drop. We concentrate on the
axisymmetric configuration where the centerline of the drops is parallel to the gravity
acceleration vector, a case that can be studied using the technique developed in the
previous sections. It is anticipated that such axisymmetric configurations are realized in
most cases following an initial period during which 3-dimensional flow is present.
Compared to the case of self-induced motion, here the problem has two additional
governing parameters, that characterize the relative influence of gravity on the motion
of each particle,
4 ( 2 -1 ) R23 ( 1 -3 ) R13 4 -
Ar1 g and Ar1 gR13 *1 3 . (49)
3 V * 2 R2 3 V 2 R2
Note that these parameters may be negative as well as positive. If the averaged density
of the aggregate exceeds that of the ambient liquid, the number Ar1 is negative while in
the opposite case it is positive. The other parameter, Ar2 , characterizes the relative
influence of buoyancy on the motion of the inner drop.
We do not conduct a full parametric analysis of the problem but rather consider some
interesting special cases, where we describe typical flow patterns and give qualitative
description of drops behavior. Some typical dependence of the relative velocity of the
two drops (velocity of the droplet in the reference frame attached to the drop) rescaled
by the characteristic buoyancy driven velocity of the large drop, Vr / Ar2 , on the center-
to-center distance, Z Z 2 Z1 , are shown in Fig 12 for a variety of governing

parameters. Positive values of Z correspond to the cases when the inner drop is close to
the upper surface of the large one. Zero relative velocity corresponds to an equilibrium
configuration of the system. The velocity is negative in the case of a downward motion
(see Fig. 7).
Sadhal and Oguz (Sadhal JFM1985) [39] studied the buoyancy-driven motion of a
compound drop and found that, depending on the governing parameters, there are cases
where zero, two or four equilibrium positions of the inner drop inside the outer one
exist, in addition to the touching aggregate configuration. Figure 12 illustrates the
combined action of buoyancy and Marangoni effect on the relative motion of a light
inner droplet ( Ar1 0 ) within the heavy large drop ( Ar2 0 ) for R=0.4,
3 0, Ar1 / Ar2 0.0206 and the values of the other parameters corresponding to
each of the cases mentioned above.
In Fig. 12 a, 1 10.52, 1. In the absence of Marangoni effect the internal
circulating flow within the large drop is not strong enough to reverse the motion of the
droplet. The latter always moves upwards and eventually touches the upper interface.
The dashed-dotted curve with Ar2 12100 illustrates the case of a relatively weak
Marangoni effect when the interaction pattern is qualitatively similar to that driven
solely by buoyancy. The solid curve is calculated for Ar2 2450 . Here the Marangoni
effect is stronger and results in the appearance of an unstable equilibrium position in the
vicinity of the lower interface. The inner droplet released at the position above this
equilibrium would move to the upper interface and that released below migrates to the
lower interface. With the further increase of Ar2 this position is shifted upwards (see
dashed, Ar2 1210 , and dotted, Ar2 80, lines) and tends to the center of the large
drop where it should be located for the motion driven solely by the Marangoni effect.
In Fig. 12 b 1 9.52, 1, The viscosity of the large drop is lower and the internal
circulation is more intensive than in the previous case. In the absence of Marangoni
effect the droplet motion within the large drop is always reversed. At any configuration,
it moves downwards and eventually touches the lower interface. The dashed line
illustrates the case of a weak Marangoni effect and the interaction pattern qualitatively
similar to that driven solely by buoyancy. The dashed-dotted curve is calculated for
Ar2 8067 . Here the Marangoni effect is more pronounced and results in the
appearance of a pair of equilibrium positions in the upper part of the large drop, with the
upper of them being stable. With the further decrease of | Ar2 | , the stable position is
shifted upwards and eventually merges with the interface of the large drop, as illustrated
by the dashed line calculated for Ar2 1152 . The unstable position is shifted
downwards and tends to the center of the large drop where it should be for the motion
driven solely by the Marangoni effect.
In Fig. 12 c 1 9.7, 1, In the absence of Marangoni effect the buoyancy of the
droplet is almost compensated by the action of the circulating flow. The direction of its
relative motion depends on its position within the large drop. There exist two
equilibrium positions of the inner drop within the large one, the lower of them being
stable. After being released at a position lower than the unstable equilibrium position
the drop would move towards the stable equilibrium, while after release above this
critical position it would move to the upper interface. The dashed-dotted curve is
calculated for Ar2 12 090 , when the Marangoni effect is relatively very weak, and
the curve is similar to that observed for the motion driven solely by buoyancy. The solid
line corresponds to Ar2 2418 . Here the influence of thermocapillarity is evident. The
stable equilibrium position is shifted downwards and another unstable equilibrium
appears near the lower interface. With the further decrease of | Ar2 | , the lower unstable
position moves upwards, while the stable one downwards. For Ar2 1680 (dotted
line) they merge into one unstable position and for smaller | Ar2 | the interaction pattern
is similar to that observed for the motion induced solely by the Marangoni effect when
there exists a single equilibrium position of the droplet inside the drop. It is illustrated
by dotted line ( Ar2 1209 ).
Fig. (12). The inner droplet relative velocity as function of the separation between the centers of the
spheres, R=0.4, 3 0 , Ar1 / Ar2 0.0206 . (a): 1 10.52, 1, dashed, solid, dashed-dotted
and dotted lines correspond to Ar2 12100, 2450, 1210 and 8 , respectively. (b):
1 9.52, 1, solid, dashed-dotted and dashed lines correspond
to Ar2 8067 , 12100 and 1152, , respectively. (c): 1 9.7, 1, dashed-dotted, solid,
dotted and dashed lines correspond to Ar2 12 090, 2418, 1680 and 1209, , respectively. (d):
1 9.7, 2128, dashed, solid, dashed-dotted and dotted lines correspond
to Ar2 10 , 16.4, 12.1 and 8.6 , respectively.
5
In Fig. 12 d, 2 9.7, 2128. In the absence of Marangoni effect such a pair of

drops has four equilibrium positions of the inner drop within the large one, two of them
being stable. After being released below the first lower unstable equilibrium position the
drop would move downwards and eventually touch the lower interface. An inner drop
initially placed between two unstable equilibrium positions migrates toward the lower
stable configuration and the drop with the initial position higher than the upper unstable
equilibrium migrates to the upper stable equilibrium position. The dashed-dotted curve
calculated for Ar2 105 demonstrates a similar behavior. The dotted curve
( Ar2 16.4 ) is already deformed by the Marangoni effect. The two lower equilibrium
points merge and a single stable equilibrium in the vicinity of the upper interface is
possible. With the further decrease of | Ar2 | this equilibrium point is shifted upwards as
it is demonstrated by the dotted curve ( Ar2 12.1 ) and eventually merges with the
upper interface. For smaller | Ar2 | the interaction pattern is qualitatively similar to that
observed for the motion induced solely by the Marangoni effect as it is shown by the
dashed-dotted curve calculated for Ar2 8.6 .
Fig. (13). The outer drop velocity as function of the separation between the centers of the spheres.
1 2 1, 1, Ar1 0.03625 Ar2 0.12. Dashed, solid and dashed-dotted curves
correspond to = 500, 250 and 160, respectively.
An interesting sample of the dependence of the velocity of the large drop on the
position of the droplet is depicted in Fig. 13 for R=0.4,
1 2 1, Ar1 0.03625, Ar2 0.12 . Dashed, solid and dashed-dotted curves are
calculated for = 500, 250 and 160, respectively. The average density of the compound
drop exceeds that of the ambient fluid and, when driven solely by buoyancy, it would
move downwards for any configuration. The Marangoni effect results in acceleration of
this motion, when the inner drop is located below the center of the large one, and in
retardation in the opposite case. When the thermocapillary force is relatively strong, this
retardation may lead to the reversal of the motion of the large drop as it is demonstrated
by the solid and dashed-dotted curves. The relative velocity of the inner drop in these
cases is directed upwards. If initially the droplet was placed below the large drop center,
then in all the cases the compound drop first moves downwards and, in the case of a
weak Marangoni effect, the motion maintains its direction (see the dashed curve). In
contrast to this, for smaller values of the motion of the large drop is reversed when the
inner droplet reaches some critical position inside it. The aggregate migrates upwards
and the inner drop continues to approach the upper interface. For strong enough
thermocapillarity, the upward motion is kept afterwards (see the dashed-dotted curve),
while for intermediate values of , the motion is reversed once more when the inclusion
reaches the second critical position shown by the solid curve.
Fig. (14). Streamline patterns with combined action of gravity and Marangoni effect. R = 0.2,
1 2 1. (a): d = 0.46, Ar1 0.587, Ar2 1.038. The large drop moves upwards while
the small one downwards. (b): d = 0.5, Ar1 5.19, Ar2 7 .785. The larger drop moves
downwards while the small one upwards. (c): d = 0.46, Ar1 5, Ar2 67.5. Both particles move
upwards. (d): d = 0.46, Ar1 0.5, Ar2 63. Both particles move downwards.
As it was shown above, the competition between buoyancy and the Marangony flow
induced on the surface of the large drop results in complex combinations of the
direction of motion of two particles. In Figs 14 - 16 we present a sample collection of
flow patterns and streamlines for a few of these possible realizations that depend on the
combination of physical parameters and the position of the small drop within the large
one. The streamline patterns are depicted in the laboratory reference frame. In all the
cases 1 2 1, R=0.2.
In Fig 14 a d=0.46, Ar1 0.587, Ar2 1.038. The large drop moves upwards
while the small one downwards. The opposite case, where the larger drop moves
downwards while the small one upwards, is illustrated in Fig. 14 b, where d=0.5,
Ar1 5.19, Ar2 7.785. Note the separation streamline within the large drop. The
streamline patterns in the cases when the drops move in the same direction, upwards or
downwards, are illustrated in Figs 14 c and d.
Fig. (15). Streamline patterns with combined action of gravity and Marangoni effect.
1 2 1, 1, R = 0.2. (a): d = 0.166, Ar1 Ar2 1. (b): d = 0.26, Ar1 Ar2 1. (c): d = 0.6,
Ar1 2.158, Ar2 0.917 . (d): d = 0.6, Ar1 1.268 , Ar2 4.659.
We notice that there exist dynamic cases for which one of the drops is immobile in
the laboratory frame although all fluids in the bulks and at the surfaces are circulating.
Some of such cases are illustrated in Fig. 15. In Fig 15 a, d=0.166, Ar1 Ar2 1. Both
drops are buoyant and would move upwards in the absence of Marangoni effect.
Thermocapillarity results in a downward migration of the large drop and immobilization
of the small one. A closed separation streamline is evident in outer fluid around the
large drop.
In Fig. 15 b, d = 0.26, Ar1 Ar2 1. Here the small drop is closer to the center of the
large one and the influence of the Marangoni effect is weaker than in the previous case.
The large drop is quiescent while the small one drifts upwards. The separation line is
evident inside the large drop around the small one. In Fig 15 c, d = 0.46,
Ar1 2.158, Ar2 0.917. Density of the large drop is higher than that of the ambient
medium, while the density of the droplet is lower. The Marangoni force that is directed
downwards makes the small particle move against buoyancy, while the large one
remains immobile. The separation streamline coincides with the interface of the drop.
In Fig. 15 d, d = 0.46, Ar1 1.268, Ar2 4.659. A separation streamline is evident in
the outer fluid near the drops interface.
Fig. (16). Streamline pattern for the case where the two bodies remain motionless. R = 0.2,
1 2 1, d = 0.46, Ar1 0.286 , Ar2 0.661.
For a certain values of parameters the Marangoni effect may compensate the
buoyancy forces acting on both drops and result in stationary configuration of neutrally
buoyant particles. An example for the streamline pattern in such a case is illustrated in
Fig. 16 (d = 0.46, Ar1 0.286 , Ar2 0.661 ), where it is evident that all fluids are
circulating and the surfaces of both drops are stream surfaces.
Deformable interfaces
If the capillary numbers are small, Ca Ca2 , Ca1 O( ), it is natural to assume
that the deformations of the interfaces are small and to construct solutions of the
problem making use of regular perturbation techniques. Following the derivation of
(Berejnov PoF 2002) [40] we describe the interfaces using two separate polar
coordinates, (ri ,i , ) , with origins at the centers of the un-deformed drops,
ri Ri (i ). An alternative approach based on the bi-spherical coordinates was developed
in Chervenivanova & Zapryanov 1989 [41]. We construct a solution in the form of
regular asymptotic expansions in terms of powers of ,
2 2 2 Ca1
p1 p10 p11 , p2 p20 p12 , p3 p30 p31 ,
Ca2
ci ci ci , Vi Vi Vi , Ri (i ) Ri fi (i ) ,
0 1 0 1
i 1, 2, (50)
ui ui ui ,
0 1
i 1, 2,3.
Substituting these expansions into equations and boundary conditions gives, at the
leading order, the problem that coincides with the problem for non-deformable drops
solved in the previous section. The normal stress balance, to within terms of O ( ) takes
the form
1 d df Ca
sin i i 2 fi i Qi (i ) bi , i 1, 2, (51)
sin i di d i Ca1
where
Qi (i ) 2c 0 (i ) 0 : nn (52)
and : nn denotes the jump of the normal stresses across the interfaces evaluated at
the unperturbed boundaries. The appearance of the dimensionless constants bi reflects
the fact that the pressure inside the drops can be found up to an additive constant in
excess of the static values. These constants can be found from the requirement of the
incompressibility of the drops,

sin fi ( )d 0.
0
(53)
The homogeneous solutions of (51) are s1 ( ) cos and s2 ( ) F ( 1 2 ,1,1 2 , cos 2 ) ,

with F being the hypergeometric function. Note, that s2 ( ) has logarithmic singularities
at 0, 2 . The complete bounded solution of (51) can be written as

fi ( ) cos Qi ( ) s2 ( ) sin d s2 ( ) Qi ( ) cos sin d bi ci1 cos . (54)
0 0
Fig. (17): Small deformations of the drops. (a) Spontaneous motion. (b) Equilibrium under
combined action of gravity and Marangoni effect. The dashed curves denote the unperturbed drops
surfaces.
The terms proportional to cos correspond to the displacement of the drops while
retaining their spherical form. The results of numerical solutions for the perturbations of
the shape of the drops are presented schematically in Fig. 17. It should be noted that, as
can be expected, the deformations are negligibly small. Hence, we used O(1) values of
Ca to make the small deformations visible. The case of spontaneous motion is
illustrated in Fig. 17a. The inner and the outer drops are, respectively, of the form of
prolate and oblate spheroids, with higher deformations in the near-contact region. In the
case of combined action of gravity and thermocapillarity we illustrate a special case of
equilibrium drops (see Fig. 17b). Here, the inner drop has an oblate form, while the
outer is prolate. Magnitude of deformations increases with the approach of the
interfaces.
The above analysis is valid when the separation distance is much larger than the
magnitude of deformations. When the interfaces come closer, the analysis should be
modified. The problem can be solved by matched asymptotic expansions for small Ca
following e.g. (Yiantsios IJSC and 1991) [42]. This is beyond the scope of this
presentation, but by the analogy with the case of two separated drops, it can be
anticipated that the deformability will prevent interface touching in finite time. Note
that zero order approximation in the above mentioned expansion is not separate but
touching drops configuration that is considered in the next section.
Touching Drops
Consider an aggregate of two externally or internally touching drops and chose larger
drops radius as a length scale. Denote the ratio of drops radii by R. Let
1 , 2 and 3 be domains occupied by larger drop, ambient medium and smaller
droplet, respectively (externally touching drops can be of equal size). The configuration
is schematically shown in Fig. 18.
Fig (18). Internally (a) and externally (b) touching drops
In the tangent spheres coordinate system the interfaces of the large drop and the
small droplet are described by the coordinate surfaces 1 2 and
1 2R , respectively. Positive corresponds to an internal touching, and
negative one to external.
External touching
For externally touching drops, a solution of (15), bounded in the outer region 2 and
satisfying the free stream condition far from the drop, is

A s C s e
s
2 3 2 2
( )
2 2 2 0
(55)
1
B s D s
2 2 e s e s| | | | U J1 ( s )ds,
1
2 s
while bounded solution for the interior fluid within the upper drop, 1 , and lower drop,
3 , are

A s C s e J1 ( s )ds,
s
1 3
( )
2 2 2 0

(56)

B s D s e J1 ( s )ds,
s
3 3
( )
2 2 2 0
respectively. The forces exerted on the each drop by the flow are

F1 4 2 s ( A2 ( s ) B2 ( s )) ds and F3 4 2 s[ A2 ( s ) B2 ( s )]ds. (57)
0 0
The requirements of vanishing normal velocities and the tangential stress jump yield the
following
i ( ) i ( ) 0, i 1, 2,3
0, Ma q .
(58)
,
,
Here
2
2 3/ 2
2 2
2
2 .
3/ 2
(59)

Similarly to the case of separated interfaces, these forces can be separated into
thermocapillary, Fi M , and hydrodynamic resistance, UFi R , parts. Here U denotes the
migration velocity of the aggregate. However, these two parts are not balanced: a
contact force, F, acting on each drop with equal magnitude in opposite direction at the
contact point completes the force balances:
F1M UF1R F 0, F3M UF3R F 0, (60)
which yield the pair migration velocity and contact force
F1M F3M F1M F3R F3M F1R

U , F . (61)
F1R F3R F1R F2R
Spontaneous motion
Leshansky, Golovin and Nir (Leshansky PoF 1997) [43] were the first to consider a
spontaneous thermocapillaryinduced motion of an aggregate of a hot solid particle and
a non-conducting bubble in contact submerged into an ambient liquid having a constant
temperature for from the inclusion. The temperature of the solid particle was assumed to
be constant during the motion and different from that of the continuous phase. The
authors did not introduce heat potential and found the temperature field in the form of
(29). The stream function was found in the form of (55), (56).
Fig. (19): Thermocapillary induced migration velocity U of a bubble and a particle in a point
contact (the particle surface temperature is constant) as a function of the bubble-to-particle radii
ratio, R.
The dependence of the solid-gas thermocapillary migration velocity on the bubble

to particle radii ratio is depicted in Fig 19. As anticipated, the velocity vanishes at R = 0
and decays at R , exhibiting maximum at R = Rm~3. The motion of such a
compound drop under the combined action of gravity, buoyancy and Marangoni effect
was studied in [43] as well. It was demonstrated that spontaneous thermocapillaly effect
can make the aggregate to move against buoyancy forces.
Another type of a spontaneous Marangoni migration of two touching drops
aggregate caused by mass transfer between the phases was studied in (Berejnov PoF
2002) [40]. The drops were assumed to consist of the same material and mass transfer
between the phases was modeled by a condition of the third kind for the concentration
at the interfaces,
c
Sh(c 1), x 1 3 . (62)
n
Here c is the difference between the concentration and its undisturbed value, scaled buy
the equilibrium concentration in the dispersed phase, Sh is the Sherwood number.
Authors looked for the concentration in the form (29) and derived a 4-th order system of
ordinary differential equations for the functions L(s) and K(s) that was solved
numerically. The two-drop aggregate migration velocity is plotted versus the radii ratio
in Fig. 20.
Fig.(20): The migration velocity of two touching drops aggregate induced by mass transfer between
the phases versus the radii ratio.
Fig. (21): Streamlines of spontaneous Marangoni flow in and around two equal spherical drops in
contact.
One can see that the migration velocity is zero at the limit of one drop being
vanishingly small and in the case of equal drops. However, in the latter case the fluid is
not quiescent. Non-homogeneous distribution of the concentration along the drops
surfaces results in a Marangoni flow. The streamline pattern of this flow is illustrated in
Fig. 21.
Motion in external temperature gradient

Thermocapillary motion of two non-conduction drops induced by an externally
imposed temperature gradient was studied in (Loewenberg JFM 1993) [35], making use
of heat potential techniques described above. Thus, components of the heat flux are
2 2 2
2 2 2
q , q , E 2 ( ) 0. (63)

General non-singular solution for the heat potential satisfying condition of a given
constant temperature gradient, T , at infinity,
2
1
2
2 T , 2 2 0, (64)
2
is of the form

s
+H ( s )e s e s| | T J1 ( s )ds.
s
1 G s e 2
(65)
( )
2 2 2 0
Condition of non-conducting drops yields
, 0, , 0, (66)
and, hence
se s sinh( s ) se s sinh( s )
G ( s) , H ( s) . (67)
sinh(( ) s ) sinh(( ) s)
Substituting presentations (55) and (56) into boundary conditions (58) and employing
(63), (65) for the tangential heat flux results in an explicit solution for the stream
functions.
Computations of [35] revealed that the thermocapillary-induced velocity of two non-
conducting drops in contact is always lower than the velocity of the larger drop when
isolated, while for the buoyancy-driven motion the effect is reversed.
Rother and Davis (Rother JCIS 1999) [44] employed the same approach for conducting
drops in a special case of equal conductivities of the dispersed and continuous phases.
In this case the temperature gradient is uniformly constant in the entire flow domain and
the heat potential is given by (65) with G(s) = H(s) = 0.
Note that in all the publications on the Marangoni motion of externally touching drops,
this geometry was considered as an intermediate stage of coalescence and, thus solely
drops of the identical material were studied. However the methods described above are
equally valid for drops comprised of different matter. If two dispersed phases are non-
wetting in a continuous one, the touching configuration is stable, while in other cases
one of the other types of compound drop will be formed with the passage of time.
Internal touching
In the case of internally touching drops, a solution of (15), bounded in the outer
region 2 and satisfying the free stream condition far from the drop, is

1
A s C s e
1
2 2 2
s
e s V1 J1 ( s ) ds, (68)
s
3
2
( )
2 2 2 0
while a bounded solution for the interior fluid within the inner drop, 3 , is

A s C s e J ( s )ds
s
3 3 3 3 1 (69)
( )
2 2 2 0
and for the outer drop stream function has the general form (30). The boundary
conditions are the same as in the case of the external touching.
The forces exerted by the flow on large and small drops are calculated according to

4 2
F1 41 sA2 ( s ) ds and F3 s[ A( s) B( s)]ds, (70)
0
1 0
respectively. The force balances are modified according to
F1M UF1R 0, F3M UF3R F 0, (71)
since, in the internal touching case, contact force acts on the ambient fluid and has no
direct contribution to the balance of forces exerted on the large drop.
Motion in external temperature gradient

Thermocapillary motion of two non-conduction drops induced by an externally
imposed temperature gradient can be solved similar to the case of externally touching
drops. The introduction of the heat potential facilitates the use of tangential stress
balances. For special cases, it is useful to solve the problem for the heat transfer. Thus,
in the case of equal heat conductivities of the phases, the heat problem is trivial, and the
heat flux is readily given by (63), (65) with H = G = 0, while in the case when thermal
conductivity of the large drop equals that of the continuous phase and the small droplet
is non-conducting, H 0, G s exp( 2 ) / 2. When the temperature at the interfaces
is known the problem for the stream function can be solved similarly to the case of
externally touching drops. To the best of our knowledge, no results on the
thermocapillary migration of such drops in an externally imposed temperature gradient
are available in the literature.
Spontaneous motion
Spontaneous motion of two internally touching drops induced by the secretion of a
surface active substance at the interface of the small droplet was investigated in
(Tsemakh IJMF 2002) [37]. The authors assumed the mass transfer to be governed by
(44) (47) and searched for the concentration field in the form (28). The concentration
was assumed to be constant at the interface of the inner droplet and tend to another
constant at infinity. Substituting the general presentation for the concentration into the
boundary conditions (45) (47) results, after certain manipulations, in a system of
linear ordinary differential equations and boundary conditions for the functions Li (s )
and Ki (s ) (i=1,2,3) that was solved numerically. Note that in a special case of equal
diffusivities of the phases, the concentration field is described by the fundamental
expression c=1/, were is distance from the center of the small sphere and, hence, in
this case
1 2
(c) .
2
(72)
,
As soon as the concentration field at the interface of the large drop is known, the
problem for the stream function and migration velocity can be solved similarly to the
case of external touching described above. The results of calculations are illustrated in
Fig 22. Figure 22 shows the dependence of the migration velocity of the drops pair on
the radii ratio R for the particular case of equal diffusivities and for a range of viscosity
ratios. In Fig. 22a, 2 1 , solid, dashed and dashed-dotted curves correspond to
1 1, 0.1 and 5, respectively. In Fig. 22b, 1 1 , solid, dashed and dashed-dotted
curves correspond to 2 1, 0.1 and 5, respectively. It is evident that the velocity
decreases monotonically with the growth of each of the viscosities, with much stronger
sensitivity to the viscosity of the large drop. Every curve in Fig. 22 exhibits a maximum
at R ~ 0.5, and vanishes in the limits R 0, corresponding to vanishing radius of one of
the drops, and R 1, corresponding to equal-sized drops.
Fig. (22). Velocity of a compound drop with surfaces in contact as function of size ratio. (a):
1 3 , solid, dashed and dashed-dotted curves correspond to 1 1, 0.1 and 5, respectively. (b):
1 1 , solid, dashed and dashed-dotted curves correspond to 2 1, 0.1 and 5, respectively.
Partially Engulfed Drops

Partially engulfed drops are composed of more than one phase with a finite size
boundary, separating each couple of phases that are in direct contact. When the phases
are immiscible these boundaries are sharp interfaces while when some miscibility is
present the boundaries can be blurred due to processes of diffusion across them. In this
presentation it is assumed that the former case prevails. The simplest partially engulfed
drops are comprised of only two phases. In this case, when such a drop is embedded in
an immiscible viscous domain with infinite extent the system has three interfaces
separating the three phases as is depicted in Fig. 23.
Fig (23): Configuration of a partially engulfed fluid-fluid compound drop

In this case it is useful to describe the partially engulfed drop by the toroidal
coordinate system with each interface separating the phases i and j, being a spherical
segment, denoted by the coordinateij .
The isothermal case

In the absence of a Marangoni effect, when the drop of volume V with an equivalent
dimension R 3V 4 , is translating with a uniform velocity U in an isothermal
1/ 3
fluid (phase 2) with viscosity 2 , the force on the drop is of the form (Ouz, 1987) [31]
1 1 (2) 1
F

82 RU 2 0, N 0, 2 d . (73)
20 4
Here P '1/ 2 i (1) 2 1 4 2 and N ( , ) , which emerges from the uniform

undisturbed flow field, is constructed to enhance a rapid convergence of the integrals in
the eigenvalue domain (see also Rosenfeld Pof 2008 [45]).
Limiting cases for compound drops with one of the phases being inviscid were
reported by Vuong and Sadhal (Vuong JFM 1989a,b) [46,47], who studied compound
drops comprised of liquid and vapor phases. The general case, covering any choice of
contact angles and of viscosity and volume ratios of the phases comprising the drop and
the ambient fluid, was recently calculated in (Rosenfeld Pof 2008) [45]. Fig. 24 depicts
the drag force for particular choices of the contact angles.
Another limiting case that was studied is when the advancing compound drop
consists mostly of one of the phases while the trailing phase has the form of a stagnant
spherical cap. This case was derived asymptotically by Sadhal and Johnson (Sadhal
JFM 1983)[48]. The reported drag force, scaled by 8 2 RU , has the form
sin 3( 12 )
2( 12 ) sin( 12 ) sin 2( 12 ) 2 3 3 2
F 3 . (74)
8 (1 3 2 ) 4 4 3 2
This result is recovered by the calculation (Rosenfeld PoF 2008) [45] when one of
the internal contact angle diminishes while the other two approach, each, the limit
value , as is depicted in Fig. 25.
In Fig. 24 the viscosity 3 equals that of the ambient fluid while the ratios 1 3
and V1 V3 vary. It is evident that, when the volume ratio V1 V3 is large with the hybrid
drop consisting mostly of phase 1, and when 1 3 becomes exceedingly small or
large, the drag force approaches the limits of those akin to a bubble or a solid sphere,
respectively. On the other end of the volume ratio scale, when V1 V3 diminishes and the
drop is comprised mostly of phase 3, all curves converge to the force on a drop
characterized by the Hadamard-Rybczynski formula for 1 3 1 .
(a)
(b)
Fig. (24). The drag force versus volume ratio of the two dispersed phases for various values of the
viscosity ratio; 2 3 1 , the contact angles are. (a) 1 0.8722 ,2 0.95 ,3 0.1778 . (b)
1 0.5 ,2 0.95 ,3 0.55 .
Fig. (25): Drag force versus cap angle. A comparison between the model of a drop with a stagnant
cap for V1 V3 (solid curve) and the model of a hybrid drop in which (V1/V3>>1); The contact
angles are: 2 3 0.95 ,1 0.01 (dotted curve); 2 3 0.995 ,1 0.01 (dashed curve),
for both cases, 2 3 1; 1 10000 .
Thermocapillary induced dynamics

In systems that are subject to a temperature field, the temperature variations result in
surface tension gradients, which prevail on each interface. These gradients induce
tangential motion on the interfaces and in the bulk of all the phases. The coupling
between the temperature field and the dynamics of the drop is embedded in the
interfacial stress condition on each boundary ij separating phases i and j which, in the
toroidal coordinate formulation has the form (Rosenfeld Pof 2008, Rosenfeld JFM
2009) [45,49]

2i j 2 j a2 Tij
2
0

i 2

sinh P '1/ 2 i (cosh ) d h
(cosh cos ij ) 3/ 2

. (75)
Here, the (+) sign applies at 12 and 13 and the (-) sign at 23 .
In the studies reported here it was assumed that heat conduction dominates over
advection and that the corresponding Peclet number is vanishingly small. Under this
assumption the temperature is a potential function and can be readily expressed using
the toroidal coordinates. Nevertheless, the use of such expression in (75) requires the
application of the inverse Mehler-Fock transform (Sneddon 1972) [50] (Zabarankin
PRSA 2007) [51] that, for the general case expressed in (70), is yet to be explicitly
obtained. An alternative approach, used e. g. in (Sadhal JCIS 1983) [33], (Feuillebois
JCIS 1989) [43], (Loewenberg JCIS 1993) [35] and (Loewenberg JFM 1993) [36],
employs axisymmetric potentials to obtain directly the heat flux q (see (34)) that, in
toroidal coordinate system, is related to the heat potential by
(cosh cos ) 2 (cosh cos ) 2

q , q , E 2 0, (76)
a 2 sinh a 2 sinh
from which application in (75) can be made for various particular cases.
Themocapillary dynamics induced by an externally applied temperature field

Whenever the auxiliary function is employed it is solved subject conditions of
continuity of the heat flux at the interfaces. At infinity, for an imposed field with a
constant temperature gradient defined in (76), the condition for has the form
2
2 2 2 sinh
2 , 0 . (77)
2 2 cosh cos
We next examine several case studies.
Non conducting drop

In this case, it is assumed that both phases of the compound drop have low
conductivity and hence the stress variations occur at the outer interfaces 12 , 23 . Since
no heat flux penetrate these interfaces 2 0 and the constant 2 is set to zero
without loss of generality. The general solution for 2 satisfying (77) may be written in
the form

2 2 cosh
2 cosh cos
1/ 2
2 ( , ) sinh 2 P '1/ 2i (cosh )d , (78)
0
cosh
with the function 2 ( , ) given explicitly in (Rosenfeld JFM 2009) [49]. Since the
temperature gradient on each surface can be written as
T2 q (cosh cos ) 2 2
h 2 , (79)
, 2 12 ,23 a 2
sinh ,
12 23 12 23
the interfacial tangential stress conditions driving the flow become
2 (2) 1 2 (1) 2 2 2 sinh 12

, (80)
2 2 cosh
2
12
12
22 3 23 2 2 2 sinh 23
2 . (81)
2 2
23
cosh
23
Fig. (26): Velocity versus volume ratio for various viscosity ratios of the dispersed phases for the
case of non-conducting compound drop. T 1 on all interfaces; 1 3 0, 2 1; 2 3 1 . The
contact angles are (a) 1 0.8722 , 2 0.95 ,3 0.1778 ; (b) 1 1.02 , 2 0.5 ,3 0.48 .
Results for the dependence of the migration velocity on the viscosity ratio, 1 3 ,
and on the volume ratio, V1 V3 , of the phases in the drop, while keeping the total drop
volume constant at 4 3 , are depicted in Fig. 26 a and b for two sets of contact angles
at the three-phase contact line. The first set was used earlier by Vuong and Sadhal
(Vuong JFM 1989a,b)[46,47].
When the drop is relatively non-conductive the only thermally active interfaces are
those bordering the ambient phase, 12 , 23 . The figure also depicts the expected
configuration of the compound drop for these sets of contact angles at small, unity and
large volume ratios. When V1 V3 , the drop consists mostly of phase 1. As the
viscosity ratio varies from 1 3 0 to 1 3 , the velocity varies from that of a
non-conductive vapor bubble (Young JFM 1959)[23] to the diminishing velocity of a
solid sphere, respectively. On the other hand, as V1 V3 0 , the drop consists mostly of
phase 3 with a viscosity equals that of the ambient fluid and, thus, at all viscosity ratios
the curves approach the velocity of a non-conductive single drop.
Partially conducting drop, 1 0, 3 2

In this case, it is assumed that the phase 1 is non-conductive and the conductivity of
phase 3 equals that of the ambient fluid. The method of solution is similar to that of the
case of non-conducting compound drop described above, except that the boundary
conditions in this case are
2
2 3 0, 2 3 and 3 . (82)
12 13 23 23
23 23
The contributions to the dynamics arise, in this case, from all three interfaces. The
mathematical approach to assess the contributions from the interfaces 12 , 13 follows
the procedure described above for the previous case. However, the calculation of the
contribution from 23 requires the use of 3 , which is of the form

sinh 2
3
(cosh cos ) 1/ 2 0
3 ( , )P '1/ 2i (cosh )d (83)
A substitution in the tangential stress condition on 23 and application of the Mehler-

Fock transform of order 1, given in (Sneddon, 1972) [50] and (Zabarankin PRSA 2007),
[51] yields

2 (2) 3 2 (3) tanh( )
2 1 4 0 1
H ( ) sinh P '1/ 2 i (cosh )d (84)
2
2
2
23
(see, Rosenfeld JFM 2009 [49] for the derivation of H1 ( ) ).

A typical result for the case of a partially conducting drop, in which only phase 1 is
relatively non-conducting, is depicted in Fig. 27. Here all three interface are thermally
active and contribute to the drop dynamics. The dependence on the volume ratio and the
viscosity ratio has the same pattern as for the case where the entire drop is non-
conductive. Since phase 3 has the same conductivity as phase 2 the convergence to the
migration velocity of a single drop as, V1 V3 0 , for all viscosity ratios is faster than in
the previous non-conductive case.
Fig (27): Velocity versus volume ratio for various viscosity ratios of the dispersed phases for the
case of partially conducting compound drop with all the interfaces thermally active. The contact
angles are: 1 0.8722 , 2 0.95 ,3 0.1778 ; T 1 on all interfaces;
1 0, 3 2 1; 2 3 1 .
In Fig. 28 a, where the case of thermally inert lower interface, 23 , is depicted, one
can see that when V1 V3 0 the velocity changes its sign, which indicates that the drop
starts to move in the direction opposite to the temperature gradient. In contrast to this,
when all the interfaces are equally active (the case depicted in Fig. 27), the compound
drop migrates to warmer fluid regions as a single-phase drop. Figure 28 b presents a
similar calculation with a different set of contact angles. When examining the
dependence of the migration velocity on the choice of contact angles, which reflect
various choices of the interfacial tensions involved, one realizes that in some cases
negative values appear. A negative velocity means that the migration takes place against
the temperature gradient, i.e., the hybrid drop is translating toward the colder region.
This occurs primarily at small volume ratios and the drop cosists mainly of phase 3. It
suggests that it may appear mostly when the contribution along the surface 12 is
relatively small and the anomalous migration results from the competition of the
thermocapillary activities along the surfaces 13 and 23 .
Indeed, this observation was examined in (Rosenfeld PoF 2008) [44] and (Rosenfeld
JFM 2009) [45] by rendering the interface 23 relatively inactive assuming that the
dependence of surface tension on temperature there can be neglected. The migration
velocity was calculated against the volume ratio and varying the importance of the
thermocapillary activity at the interface 12 relative to that separating the drop
phases 13 . If the ratio of these activities is denoted by GD T12 T13 , for the case of a
partially conducting drop and at small GD , the region of negative migration velocity
(i.e., migration against temperature gradient) was extended to volume ratios of O(1) and
beyond. This is depicted in Fig. 29. For the choice of set of contact angles and of
viscosity and conductivity for the phases in Fig. 29a, the magnitude of the negative
velocity is rather small and it completely disappears at GD 0.1 or larger. However, for
a different choice of contact angles, and even when the interface 23 is allowed to be
active at the same GD intensity as 12 , it was observed that in this case the magnitudes
of the negative and positive migration velocity are similar (see Figure 29b).
Fig. (28): Velocity versus volume ratio for various configurations of the compound drop for the case
of partially conducting drop with the lower interface ( 23 ) thermally inert. T 1 at all other
interfaces; 1 0, 3 2 1; 2 3 1.
(a)
(b)
Fig. (29): Velocity versus volume ratio for various ratios of tension temperature dependence for the
case of partially conductive compound drops. (a) the lower interface is thermally inert;
1 1.02 , 2 0.5 , 3 0.48 ; 1 0, 2 3 1; 1 2 3 1; GD T12 T13 , T13 1. (b) the
lower interface is thermally active; 1 0.8722 , 2 0.95 , 3 0.1778 ; 1 0, 2 3 1;
1 2 3 1; GD T12 T13 T23 T13 , T13 1.
Fig. (30): Velocity versus volume ratio for various viscosity ratios
GD 0.1; 1 0.8722 , 2 0.95 , 3 0.1778 , 1 2 3 1.
This phenomenon was also calculated when all phases are conductive with equal
conductivity (Rosenfeld et al., 2008) and negative velocity was observed in a similar
interval of V1 V3 up to O(1) for values of GD not exceeding 0.1, approximately. There
is however a stronger dependence on the viscosity ratio. Figure 30 shows that a
significant negative migration exists when the viscosity of phase 1 is negligible
compared to that of phase 2 and 3. However, when all three phase have a similar
viscosity the magnitude of the negative velocity decreases considerably, and at higher
viscosity of phase 1 it disappears completely.
Cases of special geometry

There exist cases of special geometry, which may facilitate a simpler and
straightforward derivation. Such an example was already described above for the
isothermal migration of a compound drop, in which one phase is a stagnant spherical
cup (Sadhal JFM 1983) [48]. Following that work, a thermocapillary-induced migration
of a compound drop with this geometry was solved by Kim and Subramanian (Kim
JCIS1989) [52]. The reported scaled migration velocity is of the form
2(2 )
U , where 12 ( 12 ),
3[ 2 (2 3 3 2 )]
(85)
1
212 sin 12 sin 212 sin 312 .
3
Their result was recently recovered in (Rosenfeld PoF 2008) [45] by letting the contact
angle in the cap layer approach zero and each of the other two angles approach , (see
configurations depicted in Fig. 3, the third and forth columns) and by rendering the
interfaces bounding this phase thermally inactive. These results are depicted in Fig. 31.
Fig. (31) Velocity of a compound drop versus cap angle. A comparison between the model of a drop
with a stagnant cap for V1 V3 (lower curve, dotted) and the model of a hybrid drop in which
V1 V3 1 with the contact angles 2 1 0.995 ,3 0.01 (upper curve, dashed),
2 3 1 1 .
Another thermocapillary induced dynamics case with a special geometry exists when
one of the drop phases (say, phase 1. See Fig 3, second column) or the entire compound
drop (see Fig. 3 third column) appears to be a complete sphere. In this case the
temperature in the three phases T1 , T2 and T3 can readily and explicitly be obtained.
Rosenfeld et al. (Rosenfeld JFM 2009) [48] derived explicit expressions for the
temperature in the first case, with a further simplification assuming that 3 2 , and
obtained the following expressions for the three tangential stress balances
2 2 1 2 1 4 2 sinh( ( 12 ))
,
2
2
2
1 2 2 cosh
12
2 3 1 2 1 4 2 sinh( ( 13 ))
, (86)
2
3
2
1 2 2 cosh
13

2

2 2 3
tanh( ) c sinh P '1/ 2i (cosh ) T2
2 1 0 (cosh cos 23 )1/ 2
3 d .
2
2 2
23
4
In the case of a spherical compound drop, the presentations are similar.

A case of special conductivity

A unique case of special conductivity exists when the conductivity in all phase is
very similar and can be assume to be equal. Hence, when heat conduction is the
dominant mode of heat transfer, to a first approximation the imposed linear temperature
field is unperturbed by the presence of the drop, and the tangential gradients on the
interfaces are readily obtained. This case is discussed in details in (Rosenfeld PoF 2008)
[45]. The dependence of the migration velocity on V1 V3 and on 1 3 0 , reported
there, resembles closely those of the cases with different conductivities for the drop
phases, which were discussed above.
Themocapillary dynamics induced spontaneously

In cases that we term 'spontaneous dynamics' no external temperature field is
imposed on the system. Temperature variation on the surfaces may result in situations
when the drop phases as well as the ambient fluid are not in thermal equilibrium, and
heat is transferred between them across the interfaces. Indeed, the temperature field,
being a harmonic function, can be formally written in terms of the toroidal coordinates.
Rosenfeld et al. (Rosenfeld JFM 2009) [49] demonstrated such spontaneous
dynamics for the case in which phase one is the source of heat that is transported across
its bounding surfaces 12 , 13 , with the temperature on these surfaces being uniform
and equals TS . This situation can be facilitated by assuming a large heat capacity in
phase 1 with 1 . A tangential temperature gradient is induced on 23 by the heat
that is conducting from 12 and 13 toward that surface and, assuming further the
simplifying equality 3 2 1 , the non-uniform dimensionless temperature
distribution in phases 2 and 3 is

T T0 TS T0 (cosh cos )1/ 2 G ( , )P1/ 2i (cosh )d (87)
0
and the tangential stress balance at that interface has the form

2 (2) 3 2 (3) tanh
2

2 2

2 1 g ( ) sinh P ' 1/ 2 i
(cosh )d (88)
23 0
4
The explicit expressions for G ( , ) and g ( ) are given in (Rosenfeld JFM 2009) [48].
Fig. (32) Velocity versus volume ratio for various values of the viscosity
ratio ; 1 2 1; 2 3 1 . The contact angles are (a) 1 0.8722 , 2 0.95 , 3 0.1778 , (b)
1 1.02 , 2 0.5 , 3 0.48 .
Figure 32 presents the velocity of the compound drop versus the volume ratio with
viscosity ratio as a parameter for two choices of contact angles. It can be seen that for
the limit V1 V3 , when the drop is comprised mostly of phase 1 which is of uniform
temperature and hence no Marangoni effect is expected, the velocity approached zero
with the diminishing of the driving force for its motion. Moreover, the velocity decays
also at the limit V1 V3 0 , with phase 1 shrinking to a point source of diminished
intensity. Hence, the velocity obtains a maximum value at a certain volume ratio in
which the heat transfer between the two dispersed phases is the most efficient to induce
dynamics. A comparison between the two figures reveals that the location of this
maximum depends on the choice of contact angles and, more weakly, on the viscosity
ratio between the phases of the compound drop.
Fig. (33): Temperature distribution at the lower interface, 23 , versus the coordinate for various
values of the volume ratio between the dispersed phases. 1 0.8722 , 2 0.95 , 3 0.1778 ;
1 2 3 1 ; 2 3 1 .
The distribution of temperature on the interface 23 for various volume ratios,

depicted in Figure 33, helps to further illuminate the profiles of migration velocity.
When V1 V3 , the temperature distribution at the interface is almost constant and
hence, as was demonstrated in Fig 32, the velocity of the compound drop diminishes.
The middle curve (V1/V31) represents the case in which the heat transfer between the
phases is the most efficient. The efficiency of the heat transfer decays as the volume
ratio decreases, as phase 1 become less pronounced.
References
[1] Sadhal SS, Ayyaswamy PS, Chung JN. Transport Phenomena with Drops and
Bubbles. Springer, New York, 1997.
[2] Evans E, Needham D. Attraction between lipid bilayer membranes in concentrated
solutions of non-adsorbing polymers comparison of mean-field theory with
measurements of adhesion energy. Macromolecules 1988 21(6): 18221831.
[3] Evans E, Klingenberg DJ, Rawicz W, Szoka F. Interactions between polymer-
grafted membranes in concentrated solutions of free polymer. Langmuir 1996,
12(12): 30313037.
[4] Aston R, Padala C, Kane RS. Microfluid separation of DNA. Current Option in
Biotechnology 2003, 14(5): 497-504.
[5] Shepherd JE, Sturtevant B. Rapid evaporation at the super heat limit. J. Fluid
Mech. 1982, 121: 379-402.
[6] Lasheras JC, Fernandezpello AC, Dryer FL. Experimental observations on the
disruptive combustion of free droplets of multi-component fuels. Combust. Sci.
Technol. 1980, 22 (5-6): 195-209.
[7] Tai IT, Sun AM. Microencapsulation of recombinant cells: a new delivery system
for gene therapy. FASEB J. 1993 7(11): 1061 1069.
[8] Leclercq S, Harlander KR, Reineccius GA Formation and characterization of
microcapsules by complex coacervation with liquid or solid aroma cores. Flavour
and Fragnance J. 2009, 24(1) 17-24.
[9] Hassan MM, Atiqullah M, Beg SA, Chowdhury MHM. Effects of enzyme
microcapsule shape on the performance of a nonisothermal packed-bed tubular
reactor J. Chem. Tech. Biotech. 1996, 66(1), 41-55.
[10] M. Goel M, Roy SK, Sengupta S. Laminar forced convection heat transfer in
microcapsulated phase change material suspensions. Int. J.Heat Mass Transf. 1994,
37(4) 593-604.
[11] Mulligan JC, Colvin DP Bryant YG. Microencapsulated phase-change material
suspensions for heat transfer in spacecraft thermal systems. J. Spacecr. Rockets.
1996, 33(2): 278 284.
[12] Kan HC, Udaykumar HS, Shyy W, Tran-Son-Tay R. Hydrodynamics of a
compound drop with application to leukocyte modeling. Phys. Fluids 1998, 10(4):
760 774.
[13] Liu XH, Wang X. The deformation of an adherent leukocyte under steady shear
flow: a numerical study. J. Biomech. 2004, 37(7): 1079-1085.
[14] Ravaine V, Bibette J, Henry N. Wetting of liquid droplets on living cells. J Collid
Intef. Sci. 2002,255(2):270-273.
[15] Johnson RE, Sadhal SS. Fluid mechanics of compound multiphase drops and
bubbles, Ann. Rev. Fluid Mech. 1985, 17: 289320.
[16] Subramanian RS, Balasubramaniam, R. The Motion of Bubbles and Drops in
Reduced Gravity, Cambridge University Press, 2001.
[17] Jin J, Niu JY, Wang HP, Luo GX, Wang QD. The thermocapillary migration of
drop and fluid in microgravity environment. J. Dispersion Sci. Techn. 2008, 29(6):
855-860.
[18] Kang Q, Cui HL, Hu L, Duan, L. On boad experimental stuffy of bubble
thermocapillary migration in a recoverable satellite. MIicrogravity Sci. Techn.2008,
20(2): 67-71.
[19] Darhuber AD, Troian SM. Microfluid actuation by surface stresses. Ann. Rev.
Fluid Mech. 2005, 37:425-455.
[20] Verneuil E, Cordero ML, Gallaire F, Baroud CN. Laser-induced force on a
microfluidic drop: Origin and magnitude. Langmuir 2009, 25(9): 51275134.
[23] Young NO, Goldstein JS, Block MJ. The motion of gas bubbles in a vertical
temperature gradient. J. Fluid Mech. 1959, 6: 350-364.
[22] Edwards D, Brenner H, Wasan DT, Interfacial Transport Processes and Rheology,
Boston, Butterworth-Heinemann, 1991.
[23] Lyell MJ, Carpenter, MJ, The effect of residual contamination on Marangoni
convection in a spherical liquid system. Applied Scientific Research 1993, 14: 639-
662.
[24] Torza S, Mason SF. Three phase interaction in shear and electrical field. J Colloid
Interface Sci. 1970, 33(1): 67-83.
[25] Happel J and Brenner H. Low Reynolds Number Hydrodynamics. 1965. Prentice-
Hall. N.J.
[26] Lebedev N N. Special functions and their applications. New Jersey: Prentice-Hall,
1965.
[27] Stimson M, Jeffrey GB. Proc. R. Soc. London Ser. 1912, A 87: 109-113.
[28] Nir A, Acrivos A. On the creeping flow of two arbitrary-sized touching spheres in
a linear shear field. J. Fluid Mech. 1973, 59: 209-223.
[29] Cooley MDA, ONeill ME. On the slow motion of two spheres in contact along
their line of centers through a viscous fluid. Proc. Cambridge Philos. Soc. 1969, 66:
407-415.
[30] Payne LE, Pell WH. The Stokes flow problem for a class of axially symmetric
bodies. J. Fluid Mech. 1960, 7: 529-549.
[31] Oguz, H.N. 1987 Fluid dynamics of compound multiphase drops and bubbles.
Ph.D. thesis University of Southern California.
[37] Tsemakh D, Lavrenteva, OM, Nir A. On the locomotion of a drop induced by the
internal secretion of surfactant, Int. J. Multiphase Flow 2004, 30(11): 1337-1367.
[33] Morton DS, Subramanian RS, Balasubramaniam R. The migration of a compound
drop due to thermocapillarity. Phys. Fluids 1990, A 2(12): 2119-2133.
[34] Sadhal SS. A note on the thermocapillary migration of a bubble normal to a plane
surface. J Collid Intef. Sci. 1983, 95(1): 283-285.
[35] Feuillebois F. The thermacapillary migration of two equal bubbles parallel to their
line of centers. J Colloid Interf. Sci. 1989, 131(1): 267-274.
[36] Loewenberg M, Davis RH. Near contact thermocapillary migration of a non-
conducting viscous drop normal to a plane interface. J Collid Interf. Sci. 1993a,
160(2): 265-274.
[37] Loewenberg M, Davis RH. Near contact thermocapillary motion of two
nonconducting drops. J. Fluid Mech. 1993b, 256: 107-131.
[38] Golovin AA, Nir A, Pismen LM. Spontaneous motion of two droplets caused by
mass transfer, Ind. Eng. Chem. Res. 1995, 34(10): 3278-3288.
[39] Sadhal SS, Oguz HN. Stokes flow past compound multiphase drops: the case of
completely engulfed drops/bubbles. J. Fluid Mech. 1985, 160: 511-529.
[40] Berejnov V, Leshansky AM, Lavrenteva OM, and Nir A. Spontaneous
thermocapillary interaction of drops: effect of surface deformation at nonzero
capillary number, Phys. Fluids 2002, 14(4): 1326-1339.
[41] Chervenianova E., Zapryanov Z. On the deformation of compound multiphase
drops at low Reynolds numbers. Physico-Chemical Hydrodynamics 1989, 11: 243-
259.
[42] Yiantsios SG, Davis RH. Close approach and deformation of 2 viscous drops due
to gravity and Van der Waals forces. J Collid Intef. Sci. 1991, 144(2): 412-433.
[43] Leshansky AM, Golovin AA, Nir A. Thermocapillary interaction between a solid
particle and a liquid-gas interface: Phys. Fluids 1997, 9(10): 2818-2827
[44] Rother MA, Davis RH. The effect of slight deformation on thermocapillarydriven
droplet coalescence and growth. J Collid Intef. Sci. 1999, 214(2): 297-318.
[45] Rosenfeld L, Lavrenteva OM, Nir A. Thermocapillary motion of hybrid drops.

Phys Fluids 2008, 20(7) 072102.
[46] Vuong ST, Sadhal SS. Growth and translation of a liquid-vapor compound drop in
a second liquid. Part 1. Fluid mechanics . J. Fluid Mech. 1989, 209: 617-637.
[47] Vuong ST, Sadhal SS. Growth and translation of a liquid-vapor compound drop in
a second liquid. Part 2. Heat transfer. J. Fluid Mech. 1989, 209: 639-660.
[48] Sadhal SS, Johnson RE. Stokes flow past bubbles and drops partially coated with
thin films. Part 1. Stagnant cup of surfactant film exact solution. J. Fluid Mech.
1983. 126: 237-250.
[49] Rosenfeld L, Lavrenteva OM, Nir A. On the thermocapillary motion of partially
engulfed compound drops. J. Fluid. Mech. 2009, 626: 263289.
[50] Sneddon IN. 1972 The use of integral transforms. McGraw-Hill.
[51] Zabarankin M. Asymmetric three-dimensional Stokes flows about two fused equal
spheres. Proc. R. Soc. A. 2007, 463: 2329-2349.
[52] Kim HS, Subramanian RS. Thermocapillary migration of a droplet with insoluble
surfactant I: Surfactant cap. J. Colloid Interf. Sci. 1989, 127(2): 417-428.
Chapter 6
Annular Flow Effects on Pore Formation in High-Intensity

Beam Welding or Drilling
P. S. Wei and S. C. Kuo
Department of Mechanical and Electro-Mechanical Engineering

National Sun Yat-Sen University, Kaohsiung 80424, TAIWAN, China
Abstract
This work is to review and study the collapse of the molten layer surrounding the keyhole filled with
vapor and droplets during high intensity beam welding or drilling. Investigating collapse of the liquid
layer is essentially required for understanding of pore formation in keyhole welding. The keyhole collapse
is similar to transition between the slug and annular two-phase flows. This study thus discusses
mechanisms of pore formation from a viewpoint of vertical annular flows. Porosity is formed from
solidification of the liquid enclosing bubbles or voids, which can be considered as either gas tapped into
liquid or liquid trapped into gas. Penetrative understanding annular two-phase flows released from this
work helps us to know how to reduce or avoid pore formation in the products in high intensity laser or
electron beam manufacturing and materials processing.
Introduction
High power density beam laser or electron beam welding have been contributed to a
significant increase of productivity in the welding industries. The merits are increased
service speed, decrease in thermal distortion, good weld quality, high flexibility of the
tool, and then deep penetration in comparison with conventional welding processes. The
deep penetration of the joint in the high power density beam welding resulted from a
keyhole formation, however, is usually accompanied with porosities, whose
distributions, sizes and volume fractions determine the extent of any degradation in
mechanical properties, and deteriorate the strength of the weld [1-5]. It is therefore
necessary to understand the porosity formation mechanism and to find methods to
reduce or eliminate porosity defects.
The mechanisms for the formation of porosity are still under debate [6, 7]. Aside
from solidification shrinkage, pore formation has been linked to solute gas rejection
from the solid phase during solidification [7-11] and collapse of the induced keyhole [2,
12-15], and gas entrapped by the flow of the molten layer around the keyhole [16]. In
fact, we can consider the last two mechanisms as the gas trapped into liquid and liquid
trapped into gas, respectively, as revealed from Figs 1 and 2 [17, 18], respectively.
Email: address: pswei@mail.nsysu.edu.tw, Fax: 886-7-5254214; Tel: 886-7-5254050

214 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Wei and Kuo
Keyhole collapse therefore is related to the liquid trapped into the gas phase in the
keyhole.
Hydrogen in Al, for example, is mainly formed by the reaction of molten aluminum
and water vapor. Hydrogen may originate from the base material itself, oxide or
hydroxide, grease, oil, dirt, paint, or other contaminated surface layers on the workpiece
and filler metal, shielding gas, and welding atmosphere. During the cooling and
solidification of molten Al, dissolved hydrogen in excess of the extremely low solid
solubility is rejected at the solidliquid interface, resulting in the formation of bubbles
[7, 19]. Unless these bubbles can escape to the melt surface, gas porosity will be present
in the solidified metal. The high cooling rate is unfavorable to the growth and flotation
of hydrogen bubbles. Thus the high cooling rates or welding speed experienced in laser
welding lead to the formation of small bubbles (micropores) [6] and lower pore number
densities and volume fractions [11, 20]. The tolerable porosity in welds depends on
welding process, alloy composition, local solidification time, thermal gradients, the
structure of welds, inclusion concentration, etc.
Relative time Keyhole geometry

(ms)
5 mm
0.00
0.44
0.89
1.33
1.78
Fig. 1 An in situ microfocused X-ray transmission images and schematic illustration of keyhole
collapse and pore formation [17].
Annular Flow Effects on Pore Formation in Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 215
Time t =0.2 s t + 5ms t + 10ms
X-Ray
transmission Porosity 3 mm
images
Keyhole Bubble Bubble Bubble
3 mm
Schematic
illustration Bubble
Bubble Bubble
Keyhole Porosity
(a) 1.5 m/min welding speed
Time t =0.3 s t + 5ms t + 10 ms
X-Ray
transmission 3 mm
images
Keyhole 3 mm
Schematic
illustration
Keyhole
(b) 6 m/min welding speed

Fig. 2 X-ray transmission in situ images and schematic illustration of pore formation for welding
speeds of (a) 1.5 m/min, and (b) 6 m/min. [18].
Even with proper preparation of material composition and surface condition to

minimize contaminations, incident energy parameters, and shielding gases, aluminum
alloys are still susceptible to porosities during welding [6]. The gas analysis and SEM
analysis of porosity showed that the bubble was composed of evaporated metal vapor
and entrained shielding gases [13]. Porosities therefore were caused by mechanisms
such as collapse or instability of the keyhole [2, 6, 12, 13] and entrainment by liquid
flow of the liquid layer around the keyhole [16]. Matsunawa et al. [13] observed that
when the pulsed laser power was terminated, the keyhole opening collapsed and
enclosed with a shorter time than that of solidification. A large pore took place near the
keyhole base. Porosities also formed during laser welding using shielding gas He, for
example, were chiefly due to the occurrence of metal vapor bubbles formed from near
the keyhole base [21]. The formed bubbles entrained metal vapor, and to some extent
shielding gas and airborne gas components. They were trapped by the solidification wall
as porosities during transport by the intense molten metal flow. Understanding and
controlling keyhole shape variations therefore is a crucial problem of much practical
importance for their affecting the quality and consistency of the finished weld.
This study reviews and discusses mechanisms of pore formation in keyhole welding.
The phenomena involve understanding of annular two-phase flow and collapse of the
keyhole. Pore formation encountered in workpieces in high intensity laser or electron
beam welding therefore can be suppressed.
Keyhole Physics
Keyhole formation is a complicated process involving rapid interactions between

the incident high energy, the flows of the evaporating atoms and the liquid layer along
the wall of the keyhole. At the beginning of the laser pulse, the absorbed laser intensity
heats up the substrate and increases temperature beyond the melting point. In view of
surface vaporization, an opposite recoil pressure is generated, and it exerts a force on
the melt surface. Together with thermocapillary force induced by a negative surface
tension coefficient for all liquid metals or alloys in the absence of surface active solute,
the melt is expelled out of the molten pool. A deep and narrow depression in the molten
material or a keyhole thus takes place. In view of strong vapor flow, the keyhole is thus
comprised of atmospheric gas, vapor and liquid droplets from the molten layer around
the keyhole wall.
Chun and Rose [22] measured and found that for laser intensity less than 109 W/m2
and long pulse durations of several millseconds or incident flux greater than 1011 W/m2
and duration longer than 10-4 s, liquid expulsion dominates mass transfer during
drilling. A further increase in incident flux and decrease in pulse duration, which is
valid for a laser in a Q-switched mode, the evaporation dominates the metal removal.
Relationships between the incident flux, the depth of the cavity and liquid expulsion in
copper irradiated by a Nd-YAG-laser pulse of a rectangular shape were also
experimentally and analytically determined by von Allmen [23]. The cavity grew at a
constant rate after a delay of 10-7 s. Strong oscillations of the drilling velocity resulted
for an incident flux between 109 and 5 1010 W/m2. This was proposed to be an
insufficient energy for full liquid expulsion, even though vapor pressure overcame
surface tension. Keyhole formation due to liquid expulsion can also refer to
experimental work [24-28].
The flows within and around the keyhole are very complicated and unstable, as first
experimentally observed by Tong and Giedt [29] using a high energy pulsed x-ray
source. The radiographs not only verified the existence of the keyhole but also showed
its oscillations in size and shape. Arata et al. [30] further used transmission X-ray
method to observe the molten pool flow during electron beam welding of stainless steel,
finding that without the influence of keyhole fluctuation four simple patterns were a
simple circulating pattern at upper part of the molten pool, a fast downward flow at the
rear of the molten pool, a very fast upward flow near the rear surface of the keyhole,
and a violent flow near the base of the keyhole. In view of to observe materials with low
density, narrow keyhole and small bubbles etc., Katayama and Matsunawa [31] further
applied a microfocused X-ray transmission imaging system with high speed real-time to
observe the flow and bubble and pore formation and keyhole behavior during
continuous and pulsed laser welding.
Although experimental observations can provide some useful information, the
underlying physics, such as evolutions of molten pool shapes, temperature, velocity et
al. during the keyhole formation and collapse cannot be revealed. Modeling keyhole
dynamics thus is required. A simple and reasonable one-dimensional steady-state model
for drilling was proposed by Chan and Mazumder [32] to predict energy, momentum
and mass transport near the keyhole base in drilling aluminum, superalloy and titanium.
In order to more reasonably calculate the evaporation rate [33, 34], the non-equilibrium
Knudsen layer at the vapor-liquid interface was accounted for. Liquid expulsion induced
by the vapor pressure gradient and satisfied by mass balance between the incoming
solid and evaporation rate was found to overwhelm evaporation rate in drilling for a
beam power less than 5 kW in contrast to a higher beam power. The predicted surface
temperature of aluminum was extrapolated to be around 3,200 K. This value is higher
than an experimental finding of 2,200 K [35], and theoretical result of 2350 K [36, 37]
for a beam power of 4 kW. The reason appears to be due to the assumption of one-
dimensional model as discussed by Wei and Ho [38]. Wei and Chiou [36] proposed a
steady-state, axisymmetric model to study momentum, energy, and mass transport near
the cavity base during drilling. Both the vapor pressure and thermocapillary force [39,
40] were included. The velocity profile, temperature and thickness of the liquid layer
were determined. The ratio of evaporation rate to liquid expulsion was found to be
around 1/200. Drilling resulted from the induced outward and upward flow of the
molten pool by effective surface pressure and thermocapillary force. Kar and Mazumder
[41] developed a transient 2D model to predict molten pool shape satisfied by energy
balances along the solidliquid and liquidvapor interfaces during drilling process. The
working parameters accounted for were laser power, beam diameter, pulse-on time, and
the number of pulses per second. Wei and Ho [38] also developed a steady-state,
axisymmetric heat conduction model to determine the shapes of the solidliquid and
liquid vapor interfaces as a function of laser power and beam distribution parameter
for a Gaussian beam. It presented magnitudes of heat conduction and latent heats for
evaporation and melting, and demonstrated irrelevance of pure evaporation model and
one-dimensional heat transfer. Kar et al. [42] extended previous model [41] by taking
into account multiple reflections of the laser beam inside the cavity, and shear stress-
induced liquid metal flow. Multiple reflections led the formation of deeper cavities,
thinner recast layer, and more cylindrical in the presence of multiple reflections than in
the absence of any reflections. The effects of the liquid metal flow, which was caused
by the shear stress at the liquid-vapor interface, on the cavity parameters such as the
cavity depth and the recast layer thickness were found to be insignificant in this study.
Batteh et al. [43] presented a theoretical Fourier analysis to predict drilling process by
accounting for conduction, convection, melting and vaporization subject to laser beam
of arbitrary pulse shape, duty cycle and intensity. Drilling can be enhanced by
augmented recoil pressure gradient for pulsed mode laser in comparison with
continuous wave mode laser. Low et al. [26] also extended the one-dimensional steady
state model including vaporization and melt ejection provided by Semak and
Matsunawa [44] to study the effects of exothermic reaction, assist gas pressure and
convection cooling of assist oxygen gas on temperature, drilling velocity, ejection
velocity, threshold time, etc. in drilling EN3 low carbon steel.
Elaborate and comprehensive numerical computations for the fluid flow and heat
transfer in the keyhole welding have been provided by Ki et al. [45, 46], Lee et al. [47],
Rai et al. [48], and Zhou et al. [49]. Ki et al. [45] proposed a comprehensive model
together with a level set method to predict the evolution of keyhole considering recoil
pressure, multiple reflections and fluid flow by tracking free surface movement at the
gas/liquid interface. The gas flow was incompressible and mass transfer was absent on
the keyhole surface. Ignoring the effects of vapor flow in the keyhole, Lee et al. [47]
found that the keyhole was unstable even in the stationary condition. A protrusion was
initiated by collision of the upward and downward flows due to recoil pressure and
surface tension, respectively. Marangoni force had only minor effects. Since the
evolution of keyhole geometry considering convection in liquid and the two phase solid
and liquid regions requires significant computational effort, Rai et al. [48] provided a
computationally efficient model for predicting geometry of the keyhole by considering
thermal equilibrium at the gas and liquid interface and point by point heat balance at the
keyhole wall. The heat transfer outside the keyhole was calculated by solving equations
of continuity, momentum and energy. A vorticity based turbulence model was used to
estimate the effective viscosity and thermal conductivity of liquid metal in the weld
pool. More recently, Zhou et al. [49] calculated the heat transfer, fluid flow of the
plasma and molten pool in the keyhole hole subject to plasma absorption, multiple
Fresnel reflections during a stationary pulsed keyhole laser welding. Deformation of the
free surface of the keyhole was determined by using the volume-of-fluid (VOF) method.
It showed that pores were formed during the keyhole collapse process. The pore
formation has a strong relationship with the solidification rate and time for backfill of
the molten pool.
Pore Formation
Porosities classified by Arata et al. [50] or Arata [2] were root porosity, active zone
porosity, and arcing porosity. The root porosity occurred in the vicinity of the keyhole
base, whereas the active zone porosity took place near the travelling path of the beam
active zone due to violent action of the active zone of the beam. The arcing porosity was
a long porosity, which was composed of a series of porosity to the bead surface. This
porosity occurred due to strong eject of vapor. Seto et al. [21] also classified similarly
three types of porosity, large spherical porosity in the upper part, elongated porosity in
the middle part, and spike-like root porosity in the base part of the molten pool. The
formed bubbles or voids were trapped by the solidification wall as porosities during
transport by the intense molten metal flow. Compositions of porosity were metal vapor
and some entrained shielding gas and air. This has been confirmed by gas analyses of
porosity using the Q-mass spectrograph by drilling the portion of pores in high vacuum
[13]. Since the metallic vapor was the major gas inside the keyhole, and some shielding
gas and air possibly entrained into the keyhole, it proves that instability or collapse of
keyholes is responsible for the formation of porosity [6].
Collapse or instability of the keyhole was studied by Klemens [51], who provided
simple analytical models to interpret energy and material flow during keyhole welding
and determine the factors, governing the shape of the vapor-supported cavity and of the
molten zone. The pressure inside the vapor cavity and its variation with depth was
governed by surface tension, hydrostatic pressure in the liquid, and viscous forces
acting on the vapor stream. The keyhole taper implying collapse of the keyhole took
place if gas pressure was greater than hydrostatic pressure of the liquid layer. Collapse
of the keyhole later was experimentally confirmed by Schauer and Giedt [52]. On the
basis of temperature distributions measured with an infrared radiation pyrometer, it
showed that the vapor pressure dominated in the lower region and the capillary pressure
in the upper region. This led to a projection and collapse of the keyhole and a spiking
formation of the molten pool.
Girard et al. [7] proposed that the solidification stage of pore formation, as
illustrated in Fig. 3, can be the molten metal was removed from the crown and went
back into the keyhole, the material cooled and solidified and the bubble rose up towards
the pool surface.
Keyhole refilling Cooling and solidification Bubble rising
Molten
Molten
metal
metal
base metal base metal
Fig. 3 Three phenomena occurring during solidification stage Girard et al. [7].
Kaplan et al. [53] proposed that pore formation can be divided into (1) drilling, (2)
expansion, detachment and flow inversion, (3) pulse termination, post-vaporization and
vapor relaxation, (4) cooling, re-condensation and shielding gas reverse flow, (5)
keyhole collapse and bubble formation, (6) bubble contraction, and (7) final sphere and
buoyancy.
Zhou et al. [54] developed mathematical models and calculated the keyhole
dynamics, heat transfer and fluid flow, and the porosity formation process in pulsed
laser welding. Their studies confirmed that porosity formation was affected by two
competing factors. One was the solidification rate of the molten metal and the other was
the back filling speed of the molten metal during the keyhole collapse process.
Prevention of Porosity
There are many methods to suppress pore formation [1]. The ways to reduce this
type of porosity are to keep the keyhole stable. The methods used to maintain stable
keyhole can be (1) modulating incident flux [54-62], (2) choosing an appropriate
shielding gas [2, 14, 20, 63-65], (3) selecting appropriate welding speed [6, 20, 66], (4)
using electromagnetic force [67,68], (5) reducing vapor pressure [69], (6) conducting
full penetration [20, 21, 64, 70, 71], (7) cleaning surface [20, 70], and (8) other methods
[72, 73], etc. Modulating incident flux involves choosing power [20, 54], frequency [13,
14], wave forms [14, 54-56], oscillation [57-59], defocus [6, 14, 20, 60], inclination
[21], and hybrid or tandem energy beam [61, 62].
Vertical Annular Flow Model
The fluid flow and heat transfer in the molten pool in thin layer surrounding the core
region filled with vapor and droplets in the keyhole are actually an annular two-phase
flow in a vertical pipe [74-78], which occurs in many practical systems such as
evaporators, condensers, water-cooled nuclear reactors, and film cooling in rocket
motors. The collapse of the keyhole is the same phenomenon as the bridging of the gas
core by liquid from the liquid film and a consequent transition to slug flow. Bridging an
gas entrapment of the liquid layer are also related to instability affected by inviscid
pressure, viscous shear, inviscid Reynolds stress, turbulent fluctuations of pressure,
gravitational force and capillary pressure for the two-phase flow [79-81].
If the gas is blown upward through the center of a vertical tube in which there is a
falling film, a shear stress which retards the film is set up at the interface. As long as the
film remains fairly smooth and stable, this shear stress is usually small and the film
thickness, and consequently the void fraction, is also virtually unchanged from the value
which is obtained with no gas flow. However, for a given liquid rate there is a certain
gas flow at which very large waves appear on the interface, the whole flow becomes
chaotic, the gas pressure drop increases markedly. This condition is known as flooding,
at which liquid flow begins to travel upward. Sometimes, the flooding point is defined
as the conditions where the waves grow sufficiently to bridge the core region and lead
to a slug flow. Unlike the phenomenon which occurs in dispersions of drops or bubbles,
the flooding point is not approached as the limit of a continuous process, but is the
result of a sudden and dramatic instability which increases the pressure gradient by an
order of magnitude. As the gas and liquid flow rates further increase, the concentration
of drops in the gas core increases; ultimately, droplet coalescence in the core leads to
large lumps or wisps of liquid in the gas core. This regime is characteristic of flows with
high mass flux. Transitions between the slug and annular regimes have been
investigated, for example, by Taitel et al. [81], and reviewed by Wallis [74], Hewitt
[75], Whalley [76], and Collier and Thome [77]. The dimensionless speeds of gas and
liquid can be obtained by balancing the shear stresses at the liquid-vapor interface and
liquid-solid interface against the body force. The dimensionless speeds of gas and liquid
were used to correlate the flooding of countercurrent falling film flow in a vertical tube.
Two simple methods have been widely used to study complicated problems of two-
phase flow. The integral analysis is conducted by assuming a velocity profile, which is
integrated over each phase in the pipe to give the flow rates. The wall and interfacial
shear stresses are related to the profile by means of single-phase flow correlations which
are already well established. Because of the complexity, the integral method is replaced
by a quasi-steady averaged one-dimensional model. The latter has been widely,
extensively and flexibly used in treating different flow patterns and their transitions,
instabilities and wave phenomena of two-phase flows [74, 77].
The unsteady, averaged one-dimensional model can be used to simulate collapse and
gas entrapment in the keyhole, which are identical to the annular two-phase vertical
flow. This simple, general and flexible method has been extensively and efficiently used
to investigate the complicated annular two-phase flows and their transitions to the slug,
churn or mist flows. Even though elaborate numerical methods have been provided,
penetrative analysis of different working parameters on flow patterns in the core region
and thin liquid layer and pore formation are still lacking. The collapse and instabilities
of the keyhole can be determined by different parameters including compressibility of
vapor, droplet entrainment from the liquid layer, mass transfer, and surface tension.
The co-ordinate system used and the geometry of the keyhole are illustrated in Fig.
4. For convenience, the solid can be considered as moving upward at a constant speed U
relative to the liquid. A cylindrical coordinate system rz thus is chosen with the origin
at the solid-liquid interface on the axisymmetric axis pointed in the upward direction. In
view of the incident flux, a molten pool in a thin layer is formed around the keyhole in
~
depth h and radius rc ( z ). The core region is comprised of vapor and droplets entrained
from the liquid layer. The mixture can also be deposited onto the surface of the liquid
layer. Without loss of generality, the major assumptions made are (1) axisymmetric
model due to because the ratio between welding speed and velocity in the liquid layer is
usually around 0.01, (2) neutral, ideal and compressible, (3) the core region filled with a
mixture of vapor and liquid particles, and (4) supersonic flow at the keyhole base. A
Knudsen layer takes place on the free surface. Fluid is at most at a choked state with
Mach number equal to unity at the edge of Knudsen layer. A high pressure gradient near
base can induce a supersonic rarefaction flow for strong evaporation, especially in
vacuum surroundings of electron beam welding.
Shape of free surface

The free surface of the liquid layer in a deep and narrow keyhole can be determined
from Young-Laplace equation
j2 1 1
p c + E
= p + ( + ) (1)
c
R1 R 2
where the terms on the left-hand side are gas and recoil pressures from the core region,
respectively. The terms on the right-hand side are liquid and capillary pressures. The
first and second principal curvatures in the last term are, respectively, defined as
1 1 dr 1 d 2 r dr
/ 1+( ) 2 , - 2 /[1+( ) 2 ]3/2 (2)
R 1 r dz R 2 dz dz
Boundary conditions of Eqs. (1) and (2) are r 0 , dr / dz at z = 0.
High Energy Beam
r oh
Vapor
Solid rc
ro h
Liquid
u
s
r o0 r
U*
Fig. 4 Physical model and coordinate system.
Vapor flow in keyhole

Unsteady conservation equations of mass, momentum and energy are, respectively,

) Wc
) ( u A
( c A (3)
c c c c
t z z
) ( u 2 A
( c u c A p c A
) gA P (4)
c c c c c c c im c
t z z
) (H
(H c c A
u A p c
c c c c c ) qA c A c (5)
t z t
where the term on the right-hand side of Eq. (3) represents entrainment of mass from the
liquid layer to the keyhole. Liquid pressure in Eq. (1) is determined from Bernoulli
equation
1 1
p p (0) U 2 u 2 gz K u 2 (6)
2 2
where the terms on the right-hand side represent liquid pressure at the same level of the
base, dynamic pressures due to drilling speed, dynamic pressures due to drilling speed
and liquid velocity, hydrostatic and loss pressures at any depth location, respectively.
Averaged liquid velocity at a given depth in Eq. (6) can be determined from mass
conservation in the liquid layer

s
1 1
u (Uro2 2rc jE ds ') (7)
ro2 - rc
2 0
where the terms on the right-hand side are referred to the incoming mass of the solid to
the liquid layer and total entrainment from the base to the depth location considered,
respectively. The shape of the interface between liquid and solid can be specified as a
paraboloid of revolution. Given independent parameters governing the shape of the
solid-liquid interface, density, Mach number, and temperature can be determined from
Eqs. (3)-(5) together with Eqs. (6) and (7) and Young-Laplace equation (1),
respectively.
Keyhole shape evolution

Regardless of zero or nonzero incident flux, energy absorbed in the keyhole is
ignored for simplicity. Values of the independent dimensionless parameters governing
the process are Mach number at the base M c (0) = 2, and parameters governing drilling
speed U*=10-4, energy absorption Q = 0, friction fim = 0, total enthalpy HE = 2, and
surface tension = 10-3. Dimensional variables appearing in these parameters are
velocity, density, and temperature of the vapor at the keyhole base u c (0) , c (0) and
T (0) , and and R the specific heat ratio and gas constant. Spatial vertical variations
c c
of dimensionless radius, pressure, temperature and Mach number of the vapor in the
keyhole subject to entrainment are shown in Figure 5(a)-(d), respectively. The keyhole
shape changes with time from a paraboloid of revolution to sphere. The dimensionless
pressure in the keyhole decreases and then increases from the base to the upper region
of the keyhole. The keyhole thus is enclosed or collapsed as a spherical pore in the late
stage. Provided that the keyhole depth is 0.01 m, the radius of the pore can be found
around 1.5 m near the keyhole tip. This size is smaller than that observed for a pore
[74]. The interpretation is unclear and it may be due to strong oscillations and
complicated processes near the keyhole tip. The result also shows that Mach number
has the similar trend as dimensionless keyhole radius.
Final pore shape

The final pore shape can be determined by introducing equation of state at two
different times [73], which are the instant when the keyhole enclosure occurs and the
time when the pore shape is at the melting point.
t* = 0.05 M c(0)=2 Q=0 U* =10-4

G =10-3
H E=2 6
4
z x 10
2
3
0 0
3
2 1
rc x 1 t*
10 4 0 2
(a)
t* = 0.05 M c (0)=2 Q=0

U*=10-4 G =10-3
H E=2
0.8
pc
0
0.6
0.5
6
1
4 *
zx 1.5
t
2
10 3
02
(b)
t* = 0.05 M c (0)=2 Q=0

U*=10-4 G =10-3
H E=2
1
0.5
Tc
0
0
6 0.5
4 1
rx
10 3 2 1.5 t*
0 2
(c)
t* = 0.05 M c(0)=2 Q=0 U* =10-4

G =10-3
H E=2
4
Mc
0
0
6
0.5
4 1
zx t*
10 3 2 1.5
0 2
(d)
Fig. 5 Dimensionless axial variations of dimensionless (a) radius, (b) pressure, (c) temperature and
(d) Mach number as functions of dimensionless time.
Conclusions
This work reviews pore formation affected by collapse or instability of the keyhole
during high power density beam welding or drilling. The collapse of keyhole is actually
an unsteady vertical annular two-phase flow process. It is quite complicated due to
involvement of unsteady and 3-dimensional interactions between absorption,
transmission, specular reflections of the incident flux, flows in the keyhole and molten
pool around the keyhole, as well as transitions between liquid, gas and solid phase. This
work demonstrates that penetrative understanding of mechanisms of pore formation in
weldment needs to study vertical annular flows. It is of great interest in that pore
formation is closely related to fluid flow and heat transfer in the keyhole in the course of
processing.
Nomenclature
A cross section
f im friction factor
g gravitational acceleration
H total energy, H H / c T(0)

p
h keyhole depth
jE local entrainment flux across keyhole wall, jE jE /[ c (0) R cT c (0)]
K loss coefficient
Mc Mach number, M c u c / R c T c
M c (0) Mach number at keyhole base
~
p vapor pressure p ~ p c (0) ~
p/~ p /[~
c (0)R c Tc (0)]
p(0) pressure at keyhole base
Pc perimeter
Q absorbed energy parameter
~
r radius, r ~r / h
Rc specific gas constant
R1 , R 2 principal curvature
s arc length
t time, t* t R T (0) / h
c c
T keyhole gas temperature, T T / T c (0)
Tc (0) vapor temperature at keyhole base
U drilling speed, U* U / R c T c (0)
u c (0) c (0)
velocity at keyhole base, u c (0) = M c (0) R cT
~ ~
Wc axial flow rate across keyhole, Wc Wc /(~
c (0)~
u c (0) h 2 )
z axial coordinate, z z / h
Greek letters
surface tension
surface tension parameter, /[h c (0)R c T c (0)]
im wall friction, im fim c u c2 / 2

vapor density, ~ /~ c (0)
c (0) vapor density at keyhole base

specific heat ratio
Superscripts
~ dimensional quantity
Subscripts
c core region or keyhole

E entrainment
liquid
o solid-liquid
References
[1] V. Ya. Agarkov, "Porosity in electron-beam welds," Avt. Svarka, no. 2, pp. 63-68,
1982.
[2] Y. Arata, Plasma, electron and laser beam technology, Ohio: American Society
for Metals, 1986.
[3] S. Kou, Welding metallurgy. New York: Wiley, 1987.
[4] H. Zhao, D. R. White, and T. DebRoy, Current issues and problems in laser
welding of automotive aluminum alloys, Int. Mater. Rev., vol. 44, pp. 238-266,
1999.
[5] W. W. Duley, Laser welding, New York: Wiley, 1999.
[6] M. Pastor, H. Zhao, R. P. Martukanitz, and T. DebRoy, Porosity, underfill and
magnesium loss during continuous wave Nd:YAG laser welding of thin plates of
aluminum alloys 5182 and 5754, Weld. J., vol. 78, pp. 207-s-216-s, 1999.
[7] K. Girard, J. M. Jouvard, and Ph. Naudy, Study of voluminal defects observed
in laser spot welding of tantalum, J. Phys. D: Appl. Phys., vol. 33, pp. 2815-
2824, 2000.
[8] B. Chalmers, "How water freezes," Scientific American, vol.200, pp.114-122,
1959.
[9] P. S. Wei, "Pore formation in metals processing- research trends," Trends in
Heat, Mass & Momentum Transfer, Research Trends, Poojapura, Trivandrum,
India, vol. 5, pp. 101-125, 1999.
[10] P. S. Wei, and C. Y. Ho, "An analytical self-consistent determination of a bubble
with a deformed cap trapped in solid during solidification," Metall. Mater. Trans.
B, vol. 33B, pp. 91-100, 2002.
[11] P. S. Wei, C. C. Huang, and K. W. Lee, "Nucleation of bubbles on a solidification
front-experiment and analysis, Metall. Mater. Trans. B, vol. 34B, pp. 321-332,
2003.
[12] H. Tong, and W. H. Giedt, "A dynamic interpretation of electron beam welding,"
Weld. J., vol.49, pp.259-s-266-s, 1970.
[13] A. Matsunawa, J. D. Kim, N. Seto, M. Mizutani, and S. Katayama, Dynamics
of keyhole and molten pool in laser welding, J. Laser Appl., vol. 10, pp. 247-254,
1998.
[14] S. Tsukamoto, I. Kawaguchi, G.. Arakane, T. Kamata, and K. Maekawa,
Suppression of welding defects in deep penetration CO2 laser welding,
ICALEO, Section C, pp. 7-15, 2000.
[15] M. Pastor, H. Zhao, and T. DebRoy, Pore formation during continuous wave
Nd:YAG laser welding of aluminum for automotive applications. Weld. Int.,
vol.15, pp. 275-281, 2001.
[16] D. J. Rod r iguez , and T. A. Shedd, Entrainment of gas in the liquid film of
horizontal, annular, two-phase flow, Int. J. Multiphase Flow, vol. 30, pp. 565-
583, 2004.
[17] S. Fujinaga, H. Takenaka, T. Narikiyo, S. Katayama, and A. Matsunawa, Direct
observation of keyhole behavior during pulse modulated high-power Nd:YAG
laser irradiation, J. Phys. D: Appl. Phys., vol. 33, pp. 492-497, 2000.
[18] Y. Kawahito, M. Mizutani, and S. Katayama, Elucidation of high-power fibre
laser welding phenomena of stainless steel and effect of factors on weld
geometry, J. Phys. D: Appl. Phys., vol. 40, pp. 5854-5859, 2007.
[19] P. S. Wei, C. C. Huang, Z. P. Wang, K. Y. Chen, and C. H. Lin, "Growths of
bubble/pore sizes in solid during solidification -an in situ measurement and
analysis," J. Crystal Growth, vol. 270, pp. 662-673, 2004.
[20] S. Katayama, A. Matsunawa, and K. Kojima, CO2 laser weldability of
aluminum alloys (2nd report): defect formation conditions and causes, Weld. Int.,
vol. 12, pp. 44-59, 1998.
[21] N. Seto, S. Katayama, and A. Matsunawa, Porosity formation mechanism and
suppression procedure in laser welding of aluminum alloys, Weld. Int., vol. 15,
pp. 191-202, 2001.
[22] M. K. Chun, and K. Rose, "Interaction of high-intensity laser beams with metals,"
J. Appl. Phys., vol. 41, pp.614-620, 1970.
[23] M. von Allmen, "Laser drilling velocity in metals," J. Appl. Phys., vol. 47,
pp.5460-5463, 1976.
[24] T. Miyazaki, S. Yoshioka, and T. Kimura, "Ejection of molten material produced
by pulsed electron and laser beams," Precision Eng., vol. 10, pp.141-146, 1988.
[25] B. S. Yilbas, "Study of liquid and vapor ejection processes during laser drilling of
metals,"J. Laser Appl., vol.7, pp.147-152, 1995.
[26] D. K. Y. Low, L. Li, and P. J. Byrd, Hydrodynamic physical modeling of laser
drilling, J. Manufacturing Sci. Eng., vol. 124, pp. 852-862, 2002.
[27] K. T. Voisey, S. S. Kudesia, W. S. O. Rodden, D. P. Hand, J. D. C. Jones, and
T. W. Clyne, Melt ejection during laser drilling of metals, Mater. Sci. Eng. A.,
vol. 356, pp. 414-424, 2003.
[28] X. He, J. T. Norris, P. W. Fuerschbach, and T. DebRoy, Liquid metal expulsion
during laser spot welding of 304 stainless steel, J. Phys. D: Appl. Phys., vol. 39,
pp. 525-534, 2006.
[29] H. Tong, and W. H. Giedt, "Radiographs of the electron beam welding cavity,"
The Review of Scientific Instruments, vol.40, pp.1283-1285, 1969.
[30] Y. Arata, M. Tomie, N. Abe, and X.-Y. Yao,"Observation of molten metal flow
during EB welding," Trans. JWRI, vol.16, pp.13-16, 1987.
[31] S. Katayama, and A. Matsunawa, Microfocused x-ray transmission real-time
observation of laser welding phenomena, Weld. Int., vol. 16, pp. 425-431, 2002.
[32] C. L. Chan, and J. Mazumder, "One-dimensional steady-state model for damage
by vaporization and liquid expulsion due to laser-material interaction," J. Appl.
Phys., vol. 62, pp. 4579-4586, 1987.
[33] S. I. Anisimov, "Vaporization of metal absorbing laser radiation," Sov. Phys.
JETP, vol.27, pp.182-183, 1968.
[34] C. J. Knight, "Theoretical modeling of rapid surface vaporization with back
pressure," AIAA J., vol. 17, pp. 519-523, 1979.
[35] D. A. Schauer, W. H. Giedt, and S. M. Shintaku, "Electron beam welding cavity
temperature distributions in pure metals and alloys," Weld. J., vol. 57, pp. 127-s-
133-s, 1978.
[36] P. S. Wei, and L. R. Chiou, "Molten metal flow around the base of a cavity during
a high-energy beam penetrating process," J. Heat Transfer, vol. 110, pp. 918-923,
1988.
[37] P. S. Wei, T. H. Wu, andY. T. Chow, "Investigation of high-intensity beam
characteristics on welding cavity shape and temperature distribution," J. Heat
Transfer, vol.112, pp.163-169, 1990.
[38] P. S. Wei, and J. Y. Ho, "Energy considerations in high-energy beam drilling,"
Int. J. Heat Mass Transfer, vol.33, pp.2207-2217, 1990.
[39] N. A. Ol'shanskii, "Movement of molten metal during electron-beam welding,"
Svar. Proiz., vol.21, pp.12-14, 1974.
[40] P. S. Wei, and W. H. Giedt, "Surface tension gradient-driven flow around an
electron beam welding cavity," Weld. J., vol.64, pp.251-s to 259-s, 1985.
[41] A. Kar, and J. Mazumder, "Two-Dimensional model for material damage due to
melting and vaporization during laser irradiation," J. Appl. Phys., vol. 68, pp.
3884-3891, 1990.
[42] A. Kar, T. Rockstroh, and J. Mazumder, Two-dimensional model for laser-
induced materials damage: effects of assist gas and multiple reflections inside the
cavity, J. Appl. Phys., vol. 71, pp. 2560-2569, 1992.
[43] J. J. Batteh, M. M. Chen, and J. Mazumder, An analysis of drilling enhancement
in pulsed-mode laser operation due to recoil pressure effects, Proc. NHTC00,

34th National Heat Transfer Conf., Pittsburgh, Pennsylvania, Aug. 20-22, 2000,
pp. 305-312, 2000.
[44] V. Semak, and A. Matsunawa, The role of recoil pressure in energy balance
during laser materials processing, J. Phys. D: Appl. Phys., vol. 30, pp. 2541-
2552, 1997.
[45] H. Ki, P. S. Mohanty, and J. Mazumder, Modeling of laser keyhole welding: part
I. Mathematical modeling, numerical methodology, role of recoil pressure,
multiple reflections, and free surface evolution, Metall. Mater. Trans. A, vol.
33A, pp. 1817-1830, 2002.
[46] H. Ki, P. S. Mohanty, and J. Mazumder, Modeling of laser keyhole welding: part
II. Simulation of keyhole evolution, velocity, temperature profile, and
experimental verification, Metall. Mater. Trans. A, vol. 33A, pp. 1831-1842,
2002.
[47] J. Y. Lee, S. H. Ko, D. F. Farson, and C. D. Yoo, Mechanism of keyhole
formation and stability in stationary laser welding, J. Phys. D: Appl. Phys., vol.
35, pp. 1570-1576, 2002.
[48] R. Rai, G. G. Roy, and T. DebRoy, A computationally efficient model of
convective heat transfer and solidification characteristics during keyhole mode
laser welding, J. Appl. Phys., vol. 101, pp. 054909-1 to11, 2007.
[49] J. Zhou, H. L. Tsai, and P. C. Wang, Transport phenomena and keyhole
dynamics during pulsed laser welding, J. Heat Transfer, vol. 128, pp. 680-690,
2006.
[50] Y. Arata, K. Terai, and S. Matsuda, "Study on characteristics of weld defect and
its prevention in electron beam welding (report I)- characteristics of weld
porosities," Trans. JWRI, vol. 2, pp. 103-112, 1973.
[51] P. G. Klemens, Heat balance and flow conditions for electron beam and laser
welding, J. Appl. Phys., vol. 47, pp. 2165-2174, 1976.
[52] D. A. Schauer, and W. H. Giedt, "Prediction of electron beam welding spiking
tendency," Weld. J., vol. 57, pp.189-s-195-s, 1978.
[53] A. F. H. Kaplan, M. Mizutani, S. Katayama, and A. Matsunawa, Unbounded
keyhole collapse and bubble formation during pulsed laser interaction with liquid
zinc, J. Phys. D: Appl. Phys., vol. 35, pp. 1218-1228, 2002.
[54] J. Zhou, and H. L. Tsai, Porosity formation and prevention in pulsed laser
welding, J. Heat Transfer, vol. 129, pp. 1014-1024, 2007.
[55] H. N. Bransch, D. C. Weckman, and H. W. Kerr, "Effects of pulse shaping on Nd:
YAG spot welds in austenitic stainless steel," Weld. J., vol.73, pp.141-s-151-s,
1994.
[56] S. Katayama, S. Kohsaka, M. Mizutani, K. Nishizawa, and A. Matsunawa,
Pulse shape optimization for defect prevention in pulsed laser welding of
stainless steels, ICALEO, pp. 487-497, 1993.
[57] X. Zhang, W. Chen, G.. Bao, and L. Zhao, Suppresion of porosity in beam
weaving laser welding, Sci. Technol. Weld. Joining, vol. 9, pp. 374-376, 2004.
[58] G. A. Shilov, K. A. Akopyants, and O. G. Kasatkin, Effects of the frequency
and diameter of circular travel by an electron beam on the penetration of metal
during electron-beam welding, Avt. Svarka, no. 8, pp. 25-28, 1983.

[59] K. S. Akopyants, Preventing weld root defects from forming during electron-
beam welding, Avt. Svarka, no. 6, pp. 59-61, 1984.
[60] J. S. Kim, T. Watanabe, and Y. Yoshida, Effect of the beam-defocusing
characteristics on porosity formation in laser welding, J. Mater. Sci. Letters, vol.
14, pp. 1624-1626, 1995.
[61] T. Hayashi, K. Matsubayashi, S. Katayama, N. Abe, A. Matsunawa, and A.
Ohmori, Reduction mechanism of porosity in tandem twin-spot laser welding of
stainless steel, Quarterly J. Japan Weld. Soc., vol. 20, pp. 228-236, 2002.
[62] S. Katayama, S. Uchiumi, M. Mizutani, J. Wang, and K. Fujii, Penetration
and porosity prevention mechanism in YAG laser-MIG hybrid welding, Weld.
Int., vol. 21, pp. 25-31, 2007.
[63] D. H. Abbott, and C. E. Albright, CO 2 shielding gas effects in laser welding
mild steel, J. Laser Appl., vol. 6, pp. 69-80, 1994.
[64] Y. Miyazaki, S. Furusako, and M. Ohara, Porosity formation in CO2 laser
welding of steel sheets, Weld. Int., vol. 16, pp. 26-37, 2002.
[65] N. Seto, S. Katayama, and A. Matsunawa, Porosity formation mechanism and
reduction method in CO 2 laser welding of stainless steel, Weld. Int., vol. 16, pp.
451-460, 2002.
[66] I. Masumoto, M. Kutsuna, and K. Yasuda, Fundamental study of high power
electron beam welding (report 2)-effect of beam oscillation on porosity
elimination, J. Japan Weld. Soc., vol. 51, pp. 27-32, 1982.
[67] M. Kern, P. Berger, and H. Hugel , Magneto-fluid dynamic control of seam
quality in CO 2 laser beam welding, Weld. J., vol.79, pp. 72-s-78-s, 2000.
[68] J. Zhou, and H. L. Tsai, Effects of electromagnetic force on melt flow and
porosity prevention in pulsed laser keyhole welding, Int. J. Heat Mass Transfer,
vol. 50, pp. 2217-2235, 2007.
[69] S. Katayama, Y. Kobayashi, N. Seto, M. Mizutani, and A. Matsunawa, Effect
of vacuum on penetration and defects in laser welding, Section C, ICALEO, pp.
182-191, 2000.
[70] K. Takahashi, and S. Sato, Porosity reduction in CO 2 laser welding of aluminum
alloys influence of penetration, joint, oxygen gas and oxide films, Weld. Int.,
vol. 14, pp. 439-446, 2000.
[71] S. Tsukamoto, G.. Arakane, H. Honda, and S. Kuroda, Formation mechanism
and prevention of weld defects in full penetration laser welding of thick steel
plates, Proc. the 23th Int. Congress on Applications of Lasers and Electro-Optics
2004, pp. 11-17, 2004.
[72] K. Kamimuki, T. Inoue, K. Yasuda, M. Muro, T. Nakabayashi, and A.
Matsunawa, Prevention of welding defect by side gas flow and its monitoring
method in continuous wave Nd:YAG laser welding, J. Laser Appl., vol. 14, pp.
136-145, 2002.
[73] H. Zhao, and T. DebRoy, Pore formation during laser beam welding of die-cast
magnesium alloy AM60B- mechanism and remedy. Weld. J., 80, pp. 204-s-210-
s, 2001.
[74] G. B. Wallis, One-dimensional two-phase flow, New York: McGraw-Hill, 1969.

[75] G. F. Hewitt, Chapter 2, Liquid-gas systems. in Hetsroni, G. (editor), Handbook
of Multiphase Systems, New York: Hemisphere Pub., 1982.
[76] P. B. Whalley, Boiling, condensation, and gas-liquid flow, Oxford: Clarendon
Press, 1987.
[77] J. G. Collier, and J. R. Thome, Convective boiling and condensation, 3rd ed.,
Oxford: Clarendon Press, 1994.
[78] L. Cheng, Modeling of heat transfer of upward annular flow in vertical tubes,
Chem. Eng. Comm., vol. 194, pp. 975-993, 2007.
[79] M. Ishii, Chapter 2.4, Wave phenomena and two-phase flow instabilities. in G.
Hetsroni (editor), Handbook of multiphase systems, New York: Hemisphere Pub.,
1982.
[80] P. A. M. Boomkamp, and R. H. M. Miesen, Classification of instabilities in
parallel two-phase flow. Int. J. Multiphase Flow, vol. 22, pp. 67-88, 1996.
[81] Y. Taitel, D. Bornea, and A. E. Dukler, Modelling flow pattern transitions for
steady upward gas-liquid flow in vertical tubes. AIChE J., vol. 26, pp. 345-354,
1980.
Chapter 7
Ground Heat Transfer from Civil Infrastructure with

Seasonal Freezing and Thawing
X. Duan1 and G. F. Naterer2
1
University of Manitoba, Winnipeg, Manitoba, Canada
2
University of Ontario Institute of Technology, Oshawa, Ontario, Canada
Abstract
This chapter discusses ground heat transfer from underground structures, particularly with seasonal
freezing and thawing in permafrost regions. Ground freezing and thawing cycles and permafrost
degradation can cause severe foundation problems for infrastructure in cold regions. Better understanding
of ground heat transfer and the associated frost actions is essential for successful foundation design and
development of protection techniques to overcome these foundation problems. This chapter briefly
reviews mechanisms of water freezing in soil systems, related frost actions, as well as advances in the
modeling of ground heat transfer and active layer thickness. Foundation design methods for infrastructure
in permafrost are also discussed, with a focus on traditional and latest techniques for reducing the effects
of ground freezing and thawing cycles and for maintaining the frozen ground as a stable foundation. In a
case study, recent research results are presented of ground heat transfer from a power transmission line
foundation.
Introduction
Detailed knowledge of ground heat transfer is crucial for the design and
maintenance of ground structures, especially for those in cold regions where seasonal
freezing and thawing in the ground is involved. There are generally two types of frozen
ground: seasonally frozen ground and perennially frozen ground, or permafrost. The
former indicates ground materials that freeze and thaw on an annual basis; the latter
refers to any subsurface material that remains at or below 0 C continuously for a
significantly long period of time (from 2 years to thousands of years) [1]. Even in the
coldest permafrost regions, there is usually an active layer that undergoes seasonal
thaw and freeze effects. Economic development in the northern regions, particularly the
extraction and transportation of the abundant natural resources (oil and gas) to the
south, has brought significant expansion of human infrastructure there. Although the
fundamental phase change mechanism is the same for both types of frozen ground, this
chapter will focus on thawing and freezing in permafrost regions. It is in these regions
that ground heat transfer has more significant effects on human infrastructure, especially
with permafrost thawing caused by the global climatic warming.
Email address: greg.naterer@uoit.ca.

234 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Duan and Naterer
Seasonal freezing and thawing in the ground can cause severe damage to ground
structures, particularly to those built over permafrost. Different frost actions, including
frost heave and thaw settlement, can cause differential movements of a structure and
may eventually cause failure of the foundation and severe damage to the structure. The
problem becomes worse with permafrost degradation caused by global climate change.
The U.S. Arctic Research Commission in 2002 organized a series of studies on the
impact of global climate change on permafrost in the Arctic and their linkages to natural
and human systems [2]. The Geological Survey of Canada initiated research in the
Machenzie Valley [3] to examine the impact of permafrost degradation on the
surrounding infrastructure. Foundation failure and structural deformation caused by
active layer growth and freezing/thawing cycles were found with different types of
infrastructure, including old buildings, pipelines, and gas/water distribution systems.
Nelson et al. [4, 5] mapped the risks for infrastructure damage in northern Canada,
Alaska, and Russia as a result of thawing of permafrost. Different risks are reported
with transportation facilities (railroads, airfields), major electrical transmission lines and
pipelines. Permafrost thawing problems have also been reported for road embankments
in northwest China [6].
This book chapter discusses ground heat transfer from civil infrastructure,
particularly the freezing and thawing process in active layer near underground
structures. Section 2 gives an introduction of ground thermal regime in permafrost
regions. Section 3 discusses mechanisms and characteristics of ground freezing and
thawing, frost actions and their effects on engineering structures. Section 4 presents a
brief review on advances in the modeling of ground heat transfer, focusing on ground
thawing/freezing and active layer thickness. Section 5 discusses foundation design
approaches for infrastructure in permafrost regions as to address the freezing and
thawing problems. Various techniques will be reviewed for maintaining the frozen
ground in permafrost regions, with some typical application examples. The last section
of this book chapter presents recent studies on ground heat transfer from a power
transmission tower foundation. This case study addresses thermal effects of an
underground metal structure on the seasonal freezing and thawing in the local active
layer. The analytical model, measurement data, and protection techniques could be
useful for other ground structures in cold regions.
Ground Thermal Regime in Permafrost Regions

Permafrost, or permanently frozen ground, refers to a thermal condition of earth
materials with a temperature that remains at or below 0 oC continuously for a
significantly long period of time (from 2 years to thousands of years) [1]. Permafrost is
defined on the basis of temperature alone; it can be any material such as sand, gravel,
silt, peat, or bedrock. Permafrost also includes soil that is below 0 oC and still not frozen
hard due to salinity and other dissolved substances in the pore water. Permafrost
underlies about 20% of the land surface of the world. Canada and Russia have most of
the land areas of permafrost, each having about 50% of their total land underlain by
permafrost [1, 7]. United States also have significant area of permafrost, mainly in the
Alaska. China, Mongolia, Finland, and other countries also have permafrost.
Ground Heat Transfer from Civil Infrastructure Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 235
Two major types of permafrost exist in the cold regions, i.e. continuous
permafrost and discontinuous (including sporadic) permafrost. In the continuous
regions, permafrost is found everywhere under the ground surface and the ground is
never completely thawed even in the warmest days. Continuous permafrost is usually
found in the northern part of the arctic region, where the thickness of permafrost table
can be several hundreds of meters. In the southern areas, seasonal thawing and freezing
penetrate deeper to the ground and the permafrost becomes patchy with a thickness as
low as only a few meters.
The uppermost layer of the ground usually undergoes seasonal thaw and freeze
effects, even in the most severe climatic regions of continuous permafrost. This layer is
called the active layer. It is defined for a given location as the maximum depth,
determined over a number of years, to which seasonal frost effects will penetrate. In
some cases, winter freezing does not extend down to the permafrost table, an unfrozen
layer remains between the permafrost and the frozen active layer known as talik [7].
The occurrence and distribution of permafrost depends on many factors, such as air
temperature, soil type and properties, relief of the land, vegetation, drainage, snow
cover, etc.
Figure 1 shows a typical ground temperature profile in a permafrost region,
without the presence of any ground structures (based on [8, 9]. Fluctuations in air
temperature during the year produce corresponding fluctuations in ground temperatures.
With increasing depth into the ground, the seasonal temperature range decreases. The
point where there is no discernable change in temperature is called the "depth of zero
annual amplitude". Below this depth, temperature changes are very small throughout the
year. Seasonal temperatures in the upper layer of the ground determine the thickness of
the active layer, which is dependent on the same climatic and terrain factors as affect the
permafrost. Generally this thickness increase from a few centimetres in the north
continuous permafrost zone to several meters in the southern sporadic regions.
Heat transfer in the active layer is very complex [10, 11], due to not only the
phase change processes in ground freezing and thawing, but other processes such as
water flow, moisture migration and vaporization/condensation etc. Local terrain factors
and any thermal disturbance by human activities (construction of engineering structures)
can alter the heat transfer processes and consequently a change in the active layer
thickness. Below the permafrost table - the upper limit of material that experiences a
maximum annual temperature of 0 C - heat transfer occurs largely by conduction.
Global climate change has significant effects on permafrost and the active layer.
Generally climatic warming will lead to permafrost degradation, with increases of
permafrost temperature and growth of the active layer thickness. These changes can
severely disrupt ecosystems and human infrastructure throughout northern regions of
Alaska, Russia, Canada and other countries. It was found that permafrost plays three
important roles in the context of climatic changes [2]: (1) as a record keeper by
functioning as a temperature archive, (2) as a translator of climate change through
subsidence and related impacts, and (3) as a facilitator of further change through its
impact on the global carbon cycle.
Temperature ( C)
-20 -10 0 10 20
0
Active layer
5 Warmest day
Annual average
Zero amplitude level
Coldest day
10
Depth (m)
15
Permafrost
20
Geothermal
25
gradient
Permafrost base
30
Figure 1: Annual temperature variations in the ground.
Ground Freezing/Thawing and Frost Heaving
This section will briefly discuss the physical processes involved in ground freezing,
thawing and the resulted frost heaving, as well its effects on ground structures. Frost
damage to building foundations, roads, underground pipelines, power transmission lines
and similar structures is common in Canada and other northern regions, and it has
incurred significant cost.
Mechanism of freezing in soil system

A soil system is usually a combination of solid materials, liquids (usually water),
and gases (usually air and water vapour). The solid materials (minerals and possible
organic materials) form the skeleton of the soil, with voids between the solid particles.
These voids may contain liquids, gases, or both. Soil is therefore often considered as
porous media in engineering analysis and modeling. Phase change of water in a soil
system plays a key role in soil freezing and thawing processes. It should be noted that
there are two types of liquid water in a soil system: free water and bound water. The
free water is the bulk pore water. It generally acts like normal liquid water and tends to
freeze at 0 oC (at 1 atm). The bound water is a layer of liquid water bond between the
mineral surface and the surrounding medium by the energy of chemical and electrical
bounds. The physical properties, particularly equilibrium thermodynamic characteristics
of bound water are different from the free water. Therefore it is not surprising that water
in soil does not freeze at the same temperature.
Figure 2: Freezing curve of a soil system.
Figure 2 shows a typical freezing process of a wet soil, generally consisting of 4

stages: (1) cooling and supercooling of water, (2) freezing of free water, (3) freezing of
bound water, and (4) cooling of soil system. In Fig. 2, T * is the freezing point of free
water (0 oC at 1 atm). It should be noted that when the temperature of the system drops
to T * , the pore water does not start to freeze immediately, but will supercool down to a
critical point of Tsc . At this point, abrupt transformation of free water to ice is triggered
and the system temperature rises to T f due to the release of latent heat. Then the bulk
free water freezes at this essentially constant temperature. Then the bound water start to
freeze at lower temperatures until a temperature Te is reached, at which point the whole
soil system freezes and there is almost no unfrozen water. More thermophysical details
about the bound water and freezing process in a soil system can be found in [1].
Ground freezing starts at the ground surface when the air temperature drops in the
fall of the year. With the air temperature continuing to drop in the winter, the freezing
plane slowly penetrates down the soil. The rate at which soil freezes is dependent upon
its thermal properties, moisture content, and the ambient air temperature. Snow cover
also plays an important role and often makes the heat transfer process more complex,
especially during thawing. Accurate prediction of freezing and thawing in the ground
requires careful modeling work, which will be briefly reviewed in Section 4.
Variations of soil mechanical and thermal properties during freezing and thawing
The mechanical properties of porous media like soil are dependent on properties of
both the solid matrix and of water in the voids. When wet soil freezes, water changes
from liquid to solid and ice fills the voids between soil grains. The soil system becomes
rigid and its strength will be significantly increased as the soil particles are now
enclosed in a solid block. The frozen ground then becomes a hard mass with high
mechanical strength like concrete. This explains why frozen ground can provide a good
foundation for engineering structures and facilities in permafrost regions.
The strength of frozen ground increases with decrease of its temperature and, in
general, with increase in ice content. It is also dependent on the composition and texture
of the soil. For example, frozen sands that are well cemented by ice usually have
considerably greater strength than fine-grained materials, particularly at temperatures
near thawing [12]. On the other hand, mechanical strength of the frozen ground
decreases with increases of its temperature. When frozen ground with high ice content
thawed, the ground becomes a muddy mass and may totally lose its strength as to
support the engineering structure. In permafrost regions, the thawed active layer loses
its mechanical strength and reduces the bond between the ground and buried part of the
structure, particularly piles or footings. Freezing and thawing cycles may cause
substantial heave, settlement, tilting or other differential movements of a structure and
may eventually cause failure of the foundation and damage to the structure.
10 1
k
8 0.8
(m /s)
ksoil , (W/m K)
2
6 0.6
-7
soil , x10
4 0.4
2 0.2
0 0
-20 -15 -10 -5 0 5 10 15 20 25
o
Soil Temperature Tsoil ( C)
Figure 3: Thermal conductivity and diffusivity of a soil sample at different temperatures (moisture
contents: soil = 16.5%).
When soil freezes, its thermal properties will also change. Figure 3 shows thermal
properties of a soil sample measured by at different temperatures [13]. It can be
observed that the frozen soil has a higher thermal conductivity than the unfrozen soil.
This occurs because of the higher thermal conductivity of ice (2.03 W/mK at -20 C)
compared to liquid water (0.56 W/mK at 0.01 C). When the liquid water in the soil
pores becomes ice, the apparent thermal conductivity of the soil increases. But the
thawed soil retains a relatively constant thermal conductivity. Also, the frozen soil has a
much higher thermal diffusivity than the thawed soil. Similarly, this arises because the
thermal diffusivity of ice (approximately 11.6 10 7 m 2 s ) is much higher than liquid
water (approximately 1.42 10 7 m 2 s ).
Frost actions
Frost actions here refer to the movement of soil mass caused by different forces as
a result of soil freezing and thawing. These forces are in general created during the
freeze and melt process of water, and possible temperature-induced changes in other
materials. Repeated cycles of freezing and thawing can provide powerful forces for
differential movement of soil materials.
There are several types of frost actions: frost heave or frost heaving indicates the
upward movement of the ground surface; frost thrusting is the lateral movement of
ground materials. Other movement mechanisms caused by freezing and thawing cycles
include frost pull and frost-push. Sometimes thaw weakening loss or weakened
strength of ground as ice within the soil melts is also regarded as a type of frost action.
In some cases, the strength of frozen ground changes with temperature change, without
visible change at the ground surface. But for many other cases, soil actions can be so
severe as to damage or even destroy engineering structures.
In these frost actions, frost heave has the most significant impact on ground
structures. For the cause of frost heave, one may think of the expansion of water when it
changes to ice. But the freezing of stationery water in the soil the in situ freezing
usually only contribute a 9% increase in the soil volume [1]. The more important cause
of frost heave is the migration of water to the freezing interface and the formation of ice
lenses in soil. Continuous ice lenses can be formed at the freezing interface and a
pressure is developed so that the ice and soil above it are lifted. It is not fully
understood yet how water is drawn from the surrounding soil to the freezing interface.
One explanation is that ice formation at the freezing plane creates a drying action and
water in the unfrozen soil beneath moves toward the freezing plane in the same way that
water will move from moist soil to dry soil. Other theories proposed to explain the
forces for this water movement include the adsorption-film theory, the capillary theory,
ice crystallization potential theory and others [1]. Change of water supply or disturbance
of heat balance can lead to advancement of the freezing interface. As a result, a series of
ice lenses are created and separated by layers of frozen soil. It is the most common
situation in nature.
The occurrence and magnitude of frost heave depend on the temperature, moisture
content, particle size, and pore size of the soil. Fine-grained soil with high water content
tends to create the highest heaving potential. In general, dry soil cant freeze and has no
frost heave potential, coarse sands and clean gravels will not develop noticeable heave.
It is usually helpful to prevent frost heave by replacing the fine grain soil with a coarse
granular material, or by improving drainage around a ground structure.
Effects of ground freezing/thawing on engineering structures

Frost actions can damage different types of infrastructure in cold regions. The
problems can become disastrous with repeated freezing and thawing cycles. Adverse
effects of frost actions on infrastructure are common in most northern regions,
particularly in those areas with ice rich permafrost.
Roads in many northern areas of Canada are susceptible to frost problems. Under
cold winter weather, ice lenses are developed in wet soil under the road surface and
cause it to heave. The frost heave can lift the pavement up to 20 cm [14]. Uneven frost
heave leads to bumps on the roads. In addition, excessive contraction of the pavement in
cold weather results in longitudinal cracks of the pavement. Frost heave also causes the
pavement to bend, causing more longitudinal frost cracks. In the following spring and
summer, ground thawing proceeds downward from the surface. The ice lenses melt, so
does possible snow near the road. Water can then accumulate within the soil, speeding
up the weakening of the ground. In this situation, even a light load from traffic on the
road can cause the paved road surface to break due to loss of sufficient support.
Figure 4: Heave of a buried structure in frost heaving soils.
Frost actions cause similar problems to foundation of buildings and other

engineering structures with a buried portion in the ground. Movements in the soil can be
transmitted to the structure if soil freezes to the foundation walls or supporting columns,
as illustrated in Fig. 4 [15]. During ground freezing, upward frost thrust creates stresses
on these buried units. This can cause heave of the whole structure, along with heave of
the ground surface, unless the dead load of the building or other resisting forces exceed
the uplift force. In most cases the displacement amount is not uniform over the whole
structure and distortion occurs. When thawed, the ground loses its strength and the bond
strength between the wet soil and the buried units greatly reduced. Thaw settlement or
subsidence will usually occur. The freezing and thawing cycles can consequently cause
the buried structure move up and down repeatedly for many years, leading to damage of
the foundation and the whole structure. In other cases, soil material drops into the
cavities produced by the heaving process so that the structure cannot settle back into its
original position. This effect can also accumulate from year to year and eventually cause
serious damage to the structure.
Frost actions in the active layer of permafrost regions can and has causes serious
damage to a wide range of foundations and facilities. In addition, permafrost is
particularly sensitive to thermal disturbance. Any natural or man-made disturbance can
affect the natural thermal equilibrium and cause local permafrost degradation (active
layer increase). With permafrost thawing, the problems caused by frost action in the
active layer will become worse. Special considerations must be made in construction
and operation of civil infrastructure in permafrost areas. For example, underground oil
or gas pipelines will cause thawing of surrounding permafrost. Special design has been
adopted to prevent or reduce this problem, as will be discussed in Section 5. In the last
section of this chapter, it will also be shown that a metal structure in the ground can
cause additional thawing and freezing in the ground. Permafrost protection techniques
must be employed for these structures.
Modeling Ground Heat Transfer and the Active Layer

Ground heat transfer modeling is a powerful tool to predict ground temperatures and
provide guidance for infrastructure design and construction. Various models have been
developed. They include analytical models and numerical models. Some models neglect
phase change and other models consider freezing and thawing of the soil.
Analytical models
It was shown in Fig.1 that ground temperatures without disturbance from any buried
structure undergo periodic changes with heat flow in the vertical direction. This
temperature profile is often called the far-field temperature [16] or study of ground heat
transfer near a structure. The one-dimensional transient heat conduction equation forms
the basis of all models of ground heat transfer without a buried structure. Seasonal
ground temperatures in this case can be easily estimated if the ground is assumed to be
homogenous with constant thermal properties [17, 18]. One of the basic formulations
considers steady periodic heat transfer in a solid semi-infinite medium with a sinusoidal
expression,
Tgs (t ) Tgsm Ags sin(wgs t gs ) (1)

where Tgsm is the average ground surface temperature, Ags is the amplitude of the
temperature variation, wgs is the frequency of variation, and gs is the phase lag. The
periodic ground temperature can be solved as [19],
z
wgs
z wgs
Tg = Tgsm erfc A e sin wgs t z gs
2 g
(2)
2 gt gs
2 g

where g is the thermal diffusivity of the ground. Many other periodic problems can be
solved analytically by the methods of separation of variables, Lapclace transform and
Duhamels integral.
Heat transfer from some buried structures, such as buried tubes of ground heat
exchangers, can be modeled by an analytical heat source method. Solutions of heat
conduction problems with various heat sources have been formulated by Carslaw and
Jaeger [20]. A one-dimensional infinite line source model was documented by Ingersoll
et al. [21] and applied to earth coupled heat transfer problems by Hart and Couvillion
[16]. Recently, Zeng et al. [22] presented a 2-D finite line-source model to consider
axial heat flow in the ground for long-term performance of GSHPs. Duan and Naterer
[19] extended the finite line heat source model to allow seasonal variations of the
strength of the heat source.
The solution of phase change in a homogeneous, semi-infinite medium is a basis
for most of the analytical and empirical models for ground freezing/thawing and the
active layer thickness (ALT). It is documented in several books, i.e. Carslaw and Jaeger
[20], Lunardini [1] and Naterer [23]. The quasi-steady approximations to the Neumann
problem [1, 20, 24] and Stefan solution [1, 23] are well-known and widely used to
predict the depth of thaw in soils when little site-specific information is available. The
general form and different variations of the Stefan equation have been used frequently
for examination of the maximum active layer thaw depth and spatial variability of the
active layer [25, 26].
An approximate analytical model developed by Kudryavtsev and co-workers [27]
has also been used widely in predicting the ALT, particularly the spatial distribution of
the ALT over a large area [24, 28]. It accounts for the effects of snow cover, vegetation,
moisture and thermal properties of soil, and regional climate. Romanovsky and
Osterkamp [24] reviewed several analytical solutions for the calculation of active layer
thickness. They developed a modified Kudryavtsev equation, which was found to be
advantageous over other analytical models (by overcoming the limitation of zero latent
heat). In the Kudryavtsev model, the active layer thaw depth, Za, is expressed by the
following equation [24, 27]:
2 AC
Z 2 c Q ph Z a Q ph
k th P
C th
th
k th C th P
Z a 2 AC th Q ph 2 Ags T ps

2 AC th Z 2 c Q ph Z a 2 AC th Q ph k th P
C th
(3)
where
Ags T ps Q ph
2 Ags T ps k th Cth P

A and Z 2 c
Q ph 2C th 2 ACth Q ph
Ags
2C th
ln
Q ph
T ps
2C th
and T ps is the mean annual permafrost surface temperature, oC; P is the period of
temperature variation (one year), s; Cth is the volumetric heat capacity of thawed soil,
J/m3K; k th is the thermal conductivity of the thawed soil, W/mK; and Q ph is the
volumetric latent heat of phase change of water, J/m3.
The equation can be simplified by introducing the following dimensionless
2Cth Ags 2Cth T ps
variables, and , which are analogous to the Stefan number
Q ph Q ph
(the ratio of sensible and latent heats) in the Neumann equation. Using these two new
variables, the active layer thawing depth can then be expressed as follows [24],
k th P
Za Z * (4)
Cth
where Z* is a solution to the following quadratic equation,
1
Z * ln
1
Z* (5)
1 1
ln Z * ln
1 1 ln 1
1
The solution of the above quadratic equation can be expressed as

Z * B B2 D (6)
2 1
where B , D 2 , and , ln .
2 2 2 1
There are several other analytical formulations, such as [29, 30], used to calculate
active layer thickness and mean annual permafrost surface temperatures. For work
involving point locations where subsurface data are available, the importance and utility
of analytical solutions is diminished.
In permafrost regions, thermal disturbance from heated ground structures can leads
to the growth of active layer in the local area and subsequent foundation problems. For
example, Sivanbaev [31] examined thawing of permafrost subjected to a steady heat
source. Structures with temperatures above 100 C inside rooms of furnaces or electrical
power plants were studied. Taking the thermal effects of the heat source into account,
the depth of ground thawing was calculated by an analytical solution and found to be
about one-half of the Neumann solution.
Pipelines for oil and gas transportation are important part of infrastructure in the
north areas where abundant fossil fuels must be transported to the south. The
introduction of hot pipeline into permafrost will lead to considerable thawing around the
pipe, which is highly possible to cause pipe failure due to excess settlement or flow of
the thawed soil. No exact solutions are available for general cylindrical phase change
problems. But various approximate formulations have been obtained using the quasi-
steady approximation. For a pipe buried at an infinite depth (Fig. 5a), Carslaw and
Jaeger [20] presented a solution when the surrounding soil is at the phase change
temperature,
R k th T p T f
2 R 2 ln R 2 r02 4 t (7)
r0 L
where r0 is the radius of the pipe, R r0 is the thaw depth beneath the pipe, k t is
thermal conductivity of the thawed soil, T p is temperature of pipe, T f is the phase
change temperature of the soil system. This solution will overestimate the thaw depth. A
correction can be made by using an effective latent heat.
For finite superheat problems, a heat balance integral method can be used, although
the solution is a little more complex. It was evaluated for thawing around buried pipes
and detailed results were plotted for the thawing depth and pipe surface heat transfer
[1]. There are no analytical solutions for the complete problem of thawing around a pipe
at finite burial depth (see Fig. 5b). The quasi-static approximation has been used when
the original ground temperature is assumed at a constant value of the given ground
surface temperature. A solution presented by Porkhavyev [32] was discussed in [1] and
the results were plotted in several diagrams. It was found that this solution yields an
underestimate of the thaw depth.
Figure 5: Permafrost thawing around buried pipes at (a) infinite and (b) finite burial depth.
Numerical models
More detailed modeling of ground heat transfer, particularly when freezing and
thawing are involved, requires numerical solutions. One of the numerical models is the
simultaneous heat and water (SHAW) model [33], which has been used mainly in
hydrological applications. It estimates the evaporation rate, snow depth, and soil water
profiles, in addition to the soil freezing depth. A finite element thermal analysis
program ADINAT has been used to determine two-dimensional heat transfer between
an earth-sheltered building and the ground [34]. Effects of phase change of water in the
soil were included in the model. The study showed that the thermal conductivity of the
soil significantly influences the simulation results and must be carefully determined.
Tarnawski [35] investigated the effect of snow cover and soil freezing on ground heat
transfer by computational simulations. An experimental study was also reported [36].
These studies confirm that winter snow cover acts as natural insulation of the ground.
Numerical models are widely available for permafrost problems, and they can be
used with some confidence when adequate information about climatic, boundary layer,
and subsurface parameters is available. Waelbroeck [37] developed a numerical model
of the soil thermal regime in a permafrost terrain and validated it against soil
temperature measurements in Barrow, Alaska. The model calculates daily soil
temperatures as a function of depth and average moisture contents of the organic and
mineral layers, using a set of 5 climatic variables (i.e., air temperature, precipitation,
cloudiness, wind speed and relative humidity). It was used to estimate the effects of
climate change on the thermal regime and moisture of permafrost, as well as the
consequences of carbon exchange between the terrestrial biosphere and atmosphere in
northern ecosystems. Degaetano et al. [38] reported a one-dimensional heat flow model
to predict soil freezing and thawing using a daily time step. The model assumes near-
saturated soil moisture conditions and uses meteorological data including daily
temperature, precipitation, and snow depth.
The numerical models are able to account for most processes involved in climate
permafrost systems, such as soil moisture migration, variations of soil thermal
properties, snow melting and others. They can describe gradual changes of seasonal
thawing and freezing in soils within small temporal intervals. However, the numerical
models are usually very complex and require significant computational resources. They
also require detailed meteorological and soil data as input. These factors limit the
applicability of the numerical models, especially to engineering applications on
permafrost structures. In addition, problems arise when complex numerical models are
employed in a spatial context, particularly when little is known about the spatial
variability of parameters. In such cases, stochastic modeling [39] or the use of analytic
procedures in a GIS environment [40, 41] may yield better results. Recently,
Riseborough et al. [42] conducted a review of permafrost modeling. They reported
various recent advances, including application of transient thermal models within spatial
models.
Protection Techniques against Frost Actions and Permafrost Thawing
In areas of deep seasonal frost and permafrost, special considerations are needed
in the design, construction and operation of ground structures. A proper foundation is
the most crucial element for engineering structures in permafrost regions. This section
discusses methods for foundation design and structural maintenance for protection
against frost actions and permafrost thawing.
There are several reports and manuals that provide engineering guidance for the
design and construction of foundations in areas of deep seasonal frost and permafrost. A
report prepared by Linell and Lobacz [43] considered the soil mechanics, seasonal
variations, and effects of permafrost to several types of underground structures. Another
technical manual [44] provides criteria and guidance for design of foundations for
structures for military facilities in arctic and subarctic regions. It considers the various
special environmental conditions in cold regions that affect the design, construction and
maintenance of foundations of these structures.
Foundation design approaches in cold regions

Adequate site investigations are essential prior to engineering design and
construction in deep seasonal frost or permafrost areas. The results will indicate the
approach to be taken in the design of foundations and the construction techniques to be
used. Information must be obtained on the distribution of permafrost, the thermal
regime of the ground, the active layer thickness, as well as the physical and mechanical
properties of the soils.
Although many general concepts are commonly applicable for all ground structures
in cold regions, specific methods of foundation design and techniques against
permafrost thawing should be considered for different types of infrastructure in different
ground conditions. Foundation design in permafrost is generally based on the following
approaches [1, 12]: (1) Neglecting the seasonal frost or the permafrost underlying the
structure and using conventional foundation design methods, (2) Allowing for thawing
and freezing during construction and operation of structures, (3) Thawing of permafrost
prior to construction, and (4) Maintaining the permafrost in a frozen state. These
approaches are detailed in the following.
Conventional foundation design can usually be used in non-frost-susceptible soils,
such as dry soils, well-drained granular soils, or solid rock. Variations in ground
temperature regime usually have little influence on these materials. Therefore there is
usually no significant frost action in the ground, nor heave or settlement of the structure.
This approach can also be used for very small and short-lived structures.
The second approach designs flexible foundations that allow ground thawing but
in the meanwhile can be adjusted to eliminate structural deformation when differential
settlement occurs. The foundation materials must be thaw-stable. It also requires
reliable prediction of the thaw and settlement rate caused by the structure. Special
settlement joints that permit individual sections of the building to move without
producing deformations in adjacent sections can also be employed. A report of
Manitoba Hydro discussed several methods to deal with seasonal frost problems with
power transmission line foundations [45]. Polyurethane foam can be used to replace the
soil between tower legs, so that the adfreeze force exerted by the soil on the legs is
reduced. Guyed towers can also be used to allow ground movement due to frost action.
With the third approach, foundation soils is thawed and compacted or removed and
replaced with compacted, well drained, non-frost-susceptible materials. High-velocity
water jets, steam jets, solar energy and other approaches have been used for thawing
purposes. The depth to which pre-construction thawing is carried out will depend on the
estimated rate of subsidence that will result from further thawing when the structure has
been completed. This method is most suitable for discontinuous permafrost regions and
where the ground constrains excessive amounts of ice and it is not possible to preserve
the frozen condition. In areas where ground temperatures are close to 0 oC, the ground
can easily be thawed and excavated. This method has been used extensively in Alaska
and Russia in early constructions of the mining industry.
Foundation design that preserves the underlying permafrost is by far the most
widely used approach. If the frozen state of the ground can be preserved, it can usually
be utilized as an adequate foundation material to support a structure. As discussed
earlier, the mechanical strength of frozen ground decreases rapidly when its temperature
increases. Effective ground temperature control techniques must be employed to prevent
thawing of the permafrost (these techniques will be discussed in details in the
following). This is particularly important in continuous permafrost, particularly for fin-
grained soils with high ice content.
Ground temperature control techniques for permafrost preservation

Ground temperature control is the key to preserve the underlying permafrost for
foundations of infrastructure in the cold regions. Either active or passive techniques can
be used for ground temperature control. Active techniques usually need power and
additional construction to operate. Examples of active techniques include forced
circulation of mechanically heated or cooled fluids through buried piping, steam and
water injection, and electrical heat tracing. Passive techniques are more preferable than
active methods for many applications, since they do not require power to operate, and
they need much less construction and maintenance work. Euler et al. [46] provided a
review of various passive techniques for ground temperature control. They include: (1)
Ground surface modification - ranging from vegetative disturbance to painting that
modifies solar radiation absorptivity, (2) Elevated construction - the structure is
removed from the ground, usually through pile foundations, (3) Thermal barrier (or
insulation layer) layers of material with a low thermal conductivity to reduce heat
transfer to the underlying ground, (4)Thermal sinks - a volume of material, usually soil,
which can be used to temporarily store heat and prevent freezing or thawing outside of
the sink. (5) Natural convection devices including closed thermosyphons (heat pipes)
and open air convection pipes.
Cheng [47] summarized measures for cooling down roadbeds for the construction
of the railway in permafrost regions. These measures include: (1) Lowering solar
radiation by using light or white-colored material for the roadbed surface, using
shading sheds, or placing shading boards along embankment side slopes, (2) Lowing
heat convection by using ventilation foundations with air ducts embedded into the
roadbed, embankment of coarse rocks, or use of thermosyphons, and (3) Lowering heat
conduction by using peat, sodding, or dry bridges over the railway. It is believed that
under the natural cyclic variations of air temperature, the coarse rocks act as thermal
semi-conductors and cool down the underlying soils. It has been demonstrated through
experiments and numerical simulations that coarse rocks can be used as embankment
fill and placed on side slopes to lower the embankment temperature [48].
Pile foundations are the most common design for thaw-unstable permafrost. They
are widely used in northern regions of Canada, Alaska and other northern areas. In this
design, the buildings or other heated structures are raised above the ground surface by
piles, allowing air circulation in the space between the ground surface and the bottom of
the heated structure. This can minimize or prevent heat flow from the structure to the
ground. Piles should be well buried in the permafrost to a depth of at least twice the
active layer, with a minimum depth of 10 feet [1]. The clearance between the ground
surface and the bottom of the building should be at least 2 feet to allow adequate
ventilation and access for cleaning of snow, debris, etc. In the design it must be ensured
that the adfreezing force plus the load is greater than the frost uplift force. Creosoted
wood and steel pipe or H-section steel piles are usually used, though precast or cast-in-
place concrete piles are occasionally found. Sometimes pile placing could be difficult as
in very stony soils. Anchors for footings, guyed towers, or pipelines present special
difficulties because they are subject to uplift, thrust, and creep [7].
Ground thermal barrier (insulation layer) is widely used as a passive temperature
control technique. It increases the thermal resistance of the foundation, so as to reduce
heat exchange between the atmosphere and ground, thereby mitigating the
freezing/thawing cycle and possibly delaying degradation of permafrost. Thickness of
the thermal barrier should be calculated according to the structure and the heat load.
Thermal insulation was used in the 1980s in Alaska for roadway and airfield
engineering [49], and widely used in the arctic regions of Russia to prevent permafrost
degradation and settlements [50]. Insulation sheets were also used to protect power
transmission line foundations in northern regions of Canada [45]. In China, extensive

studies have been conducted in recent years for thermal insulation protection of railway
embankment in permafrost regions [51-53].
Different insulation materials have been used for foundation protection, including
styrofoam [7, 45], expanded polystyrene and polyurethane foams [51, 52]. In recent
years, alternative materials have been proposed or applied. One of these materials is the
aerogel insulation [54], which has very low thermal conductivity and resistance to
moisture. The aerogel has a porosity as high as 90% and pore sizes that are much
smaller than the mean free path of air, which makes its thermal conductivity even lower
than air. Another alternative method is using a layer of phase change materials (PCMs)
as a thermal barrier. The PCM layer absorbs and releases heat in summer and winter
seasons, respectively, but remains at the phase change temperature for a significant
period of time, thereby reducing the temperature amplitude at the ground surface.
Naterer [23] discussed how PCMs can prevent icing of bridges and reduce cracking of
materials caused by repeated freezing and thawing. Duan and Naterer [55] presented a
PCM thermal barrier for protection of foundations in permafrost. Ground temperatures
beneath a PCM thermal barrier were measured and its effectiveness to alleviate seasonal
freezing/thawing problems was demonstrated. Figure 6 shows the modified soil surface
temperatures below a PCM layer. It can be observed that the soil surface temperature
range was greatly reduced. The soil surface remains nearly constant at a certain
temperature for two periods in each cycle, between the highest and lowest air
temperatures (marked by the portion between two horizontal dashed lines in the figure).
It can be observed that this constant temperature is about 18 oC, which corresponds to
the phase change temperature of the PCM.
The thermal insulation method has many advantages, such as construction
simplicity, low capital investment, and wide adaptability to different structures. But
previous studies have shown significant disagreement in regards to overall thermal
effectiveness of this method. Field observations of railway embankments with EPS
insulation [51, 52] revealed that insulation can greatly reduce the range of annual
ground temperatures and heat exchange. However, reducing the ground temperature
range does not necessarily prevent freezing and thawing processes in the foundation.
Daigle and Zhao [56] showed that rigid polystyrene insulation promoted a fast frost
penetration, due to the higher thermal diffusivity and lack of latent heat of the insulation
board, as compared to ground material. For the problem of permafrost degradation,
some studies predicted and observed that thermal insulation can keep or raise the local
permafrost table [52, 53]. But studies also found an increased temperature of shallow
ground and deep permafrost under the insulation layers [51]. Cheng et al. [57] showed
that the effectiveness of insulation depends on its adjustment to the mean annual ground
temperature, and the temperature amplitude in the permafrost. With global warming, it
may delay permafrost degradation, but cannot change the eventual tendency of
degradation.
Air convection devices are also widely used for protection of structure against
permafrost thawing. The devices are buried near the foundation of ground structures and
cold air in winter is circulated through underground pipes to remove heat from the
foundation. When the air temperature exceeds the ground temperature during the
summer, the underground air circulation stops, and the motionless air becomes a
thermal insulator. In this way, the ground temperature is lowered and foundation
thawing is prevented or reduced. Different air ventilation devices can be applied for the
ground cooling purpose, such as an air duct [46], or an air convection pile [58], as
depicted in Fig.7. The device should be made of materials with low thermal
conductivity to prevent heat conduction through the device. Various methods have been
used previously for controlling the wind flow through the air inlets and outlets. A
ventilator, evertor, venturi effect or turbo-fan arrangement can provide this effect, either
manually or automatically [59].
Figure 6: Variations of upper surface temperature of the PCM layer (TuPCM) and the soil surface
temperature under the PCM layer (TbPCM).
There are many other ground cooling devices, including two-phase devices,
particularly thermosyphons (heat pipes), and forced convection devices that use air or
liquids. Adding a pump or fan to the passive air convection devices makes them active
systems with increased heat transfer capacity due to forced convection of air. Buchko et
al. [60] discussed several seasonally functioning cooling devices for controlling ground
temperature of dam-foundations, river banks and open mine walls. These include forced
air convection devices, liquid devices with natural convection of cooling agent
(kerosene), liquid devices with forced convection of the cooling agent, and vapour-
liquid devices.
Thermosyphon is a very effective ground cooling device due to its increased heat
removal capacity through the phase change process between liquid and gas. Most
thermosyphons used for ground temperature control are essentially heat pipes, with
boiling and condensation of a pressurized medium by natural convection in a closed
chamber. These heat pipes allow heat transfer from the ground to the atmosphere in the
winter, and prevent heat transfer in the reverse direction in the summer. In this way, the
mean annual ground temperature is lowered to preserve the permafrost below heated
structures. Unlike most air convection devices, thermosyphons are closed system. In
addition, air convection devices usually operate near atmospheric pressure, but
thermosyphons must be designed, installed, and operated as pressure vessels. Extended
surfaces are required on the upper portions of the devices to improve heat transfer to the
atmosphere (see Fig. 8).
Figure 7: Schematic of an air duct (left) and an air pile (right).
Figure 8: Vertical support members with heat pipes for oil pipeline in warm permafrost regions.
Thermosyphons have been widely used for infrastructure in permafrost regions.

Manitoba Hydro has experience in using thermosyphons for tower foundations of power
transmission lines in northern Manitoba. Thermosyphons were successfully used in the
Trans-Alaska Pipeline System [2] for permafrost preservation. The pipeline carries oil at
temperatures above 60 oC and crosses many ice-rich permafrost areas. In order to
prevent permafrost thawing and damage to the pipe, the pipeline is elevated in these
permafrost areas on vertical support members (VSMs). Most of these VSMs are
equipped with pairs of heat pipes to maintain permafrost stability. Figure 8 shows a
schematic of such a VSM. In recent years, thermosyphons have also been used in
Chinas Qinghai-Tibet Railway project for protection of railway embankments [61] in
permafrost areas.
Figure 9: A real (left) and simplified (right) power transmission tower foundation in permafrost.
Case Study Ground Heat Transfer in a Power Transmission Tower

Foundation
Thermal effects of a metal structure can induce significant additional thawing and
freezing near its foundation. Figure 9 shows an example of such a structure a power
transmission tower. In the summer, solar radiation is strong and the air temperature is
usually higher than the ground temperature. Due to the high thermal conductivity of a
tower footing (relative to soil), heat is conducted through the tower downwards into the
ground. This leads to an additional temperature increase of the surrounding soil and
causes more thawing. In the winter, the tower footing allows for heat flow from the
ground to the surrounding air, leading to more severe freezing in the ground. The
subsequently enhanced heave, settlement, or tilting has caused significant damage to
towers in permafrost regions.
Both analytical and experimental studies have been conducted recently for ground
heat transfer in power transmission line foundations [13, 19, 62]. Particularly the
seasonal thermal effects of the metal tower have been studied. A typical high voltage
power transmission tower is a metal structure that has four footings buried in the
ground. In order to simplify the heat transfer analysis, a single metal rod buried in the
ground was used to approximate a tower footing, as shown in Fig.7. A finite line heat
source model was developed to predict the transient thermal effects of the metal tower
footing [19]. Seasonal freezing and thawing in the active layer were simulated and
measured in an experimental tower foundation test cell [13]. This section will presents
some results of these studies.
Thermal analysis
The extended finite line heat source model with a variable heating strength [19, 62]
provides a good approximation of the seasonal thermal effects of the tower on the
temperature variations near the foundation. In this model, the tower footing is
approximated as a finite line heat source with a strength of q(h, t ) , as shown in Fig.10.
The thermal response at a ground point, P(r, z), to the finite line heat source can be
obtained by a method of images and integration of a series of time-varying point heat
sources as follows,
q ( h, t )
H t 2 2
1 D1 D2
gl ( r , z, t )
8 k g g
1.5 0 0 (t t )1.5 4 g (t t )
exp exp

4 (t t
dt dh (8)
)
g
where D1 r 2 ( z h ) 2 , D2 r 2 ( z h ) 2 , h is the distance from an element of

the heat source to the ground surface, m; gl is the thermal response, oC; k g is the
thermal conductivity of the ground, W/m oC; and g is the thermal diffusivity of the
ground, m2/s. The solution of Eq. (8) can be expressed alternatively with respect to the
D1 (t t ) 0.5 D2 (t t ) 0.5
following variables: 1 , 2 :
2 g 2 g

1 1
H
D12 12 1

D
2
2
gl (r , z, t ) 1.5 q h, t e d1 D q h, t 4 g 22 e 2 d 2 dh (9)
2
2
2 k g 0 D1 D1 4 g 1 D2 2
2 gt 2 gt
The seasonal variation of heat flow through the tower can be approximated by the a
line heat source with the following sinusoidal strength,
q( h, t ) qm Aq sin( wq t q ) (10)
where qm is the average value of the heat source strength, Aq is the amplitude of the
seasonal variation, wq is the frequency of variation, and q is the phase lag. Using the
variable heating strength model, the thermal response to this line heat source can be
evaluated by either equation (8) or (9). In particular, the long-term thermal response can
be calculated by the following approximate solution, which needs much less
computational time:
qm H 1 D 1

D

dh +
gl (r , z, t ) 1 2
erfc erfc
4k g 0 D1
2 gt 2 gt
D2

D1 wq D2 wq
2 g
2 g

dh
Aq eH
wq wq
wq t D1 e wq t D2
4k g 0 D1
sin sin (11)
2 g
q
D2 2 g
q

Figure 10: Schematic of a finite line heat source in the ground.
These results also apply to inclined heat sources (to represent an inclined tower
footing), except that D1 and D2 should be changed to include the inclination angle, ,
of the line heat source, i.e., D1 (z - hcos ) 2 ( r h sin ) 2 and

D2 (z hcos ) 2 ( r h sin ) 2 .
In an experimental study, the simplified tower foundation in Fig.1 was modeled in a
test cell with an examined local area of depth Ze = 200 mm and radius Re=100 mm
centered on an aluminum rod of Db=6.67 mm in diameter, and Hb = 66.67 mm buried
length in soil. Other details of the experimental setup were described in [13]. Seasonal
freezing and thawing cycles were simulated, by applying periodic air temperature
variations around the tower as well as over the ground surface. The Canadian Climate
Normals or Averages data [63] were used to design the simulated air temperature, which
was 5 oC lower than the corresponding temperatures at the Thompson Airport, Manitoba.
Figure 11: Ground temperature variations without a tower.
Figure 11 shows the measured results of the ground temperatures at 5 representative

positions in the test cell for a period of 292 minutes (corresponding to two years in a real
foundation situation). At most positions, the ground temperatures remain frozen,
indicating the permafrost table. At positions located in the active layer (z = 8.5 mm in
Fig.11), the soil temperature undergoes thawing in the summer and freezing in the other
period. A flat section of the temperature curve near 0 oC (dashed line) indicates the
freezing process.
Figure 12 illustrates the summer ground temperatures without presence of a tower,
with the active layer thawing depth shown by the vertical dashed line. It clearly shows
that the ground temperature decreases dramatically with depth in the top soil levels. The
results also confirm the one-dimensionality of these measurements, since temperature
readings from thermocouples of three sides fall onto the same trend line. The active layer
thaw depth is found by examining the depth level where the ground temperatures reach 0
o
C, at around 13 mm in this case.
Figure 13 shows the measured soil temperatures in the foundation, with the tower, at
two typical times, t = 688 minutes (Fig.13a) and t = 625 minutes (Fig. 13b), which
correspond to a time instant in the summer and another time in the winter,
respectively. These results show the thermal effects of the tower footing on soil thawing
and freezing in the foundation. As compared to the summer ground temperatures in Fig.
12, the ground temperatures with the tower reveal a two-dimensional character. Since
thermocouples with R = 39 mm are relatively far away from the metal rod, the
temperatures at these positions are almost the same as those without the tower.
Figure 12: Summer ground temperatures without a tower.
(a) (b)
Figure 13: Thermal effects of a metal tower on the thawing and freezing in the tower foundation.
0 6 6
6 6
4 4 4 4
2 2 2
10 2
0 0
0
20 0
30 -2
-2
z (mm)
-2
40
50
-2
60
70
10 20 30 40 50
R (mm)
Figure 14: Two-dimensional contour of ground temperatures in the summer.
In the summer (Fig. 13a), the tower footing makes the surrounding soil warmer.
This effect increases at locations closer to the tower footing (as expected) to a
temperature difference as high as 1.2 oC in Fig. 13a. The thermal effect causes
additional thawing and makes the active layer increase in a local region around the
tower footing. This can also be observed by the two-dimensional measured temperature
contours shown in Fig. 14. The temperature gradient and thawing depth in the lateral
direction indicate thermal effects of the tower footing. The isotherm of 0 oC
approximately indicates the level of active layer thawing, which is around 13 mm at
locations far from the tower footing, but it increases to about 19 mm at locations very
close to the footing. Similarly at t = 625 minutes (winter), the tower footing makes the
surrounding soil colder (to a maximum temperature difference of 3 oC) and causes an
additional freezing effect. As discussed earlier, this intensified thawing and freezing
could result in more severe heave and settlement of the tower and structural damage to
the transmission line.
Figure 15 shows the relative increase of the temperature amplitude, (AT*-AT)/AT,
around the tower footing, indicating its thermal effect on ground temperatures. It can be
observed that this thermal effect varies in both radial and vertical directions. In the
radial direction, it decreases at further distances from the tower footing (with almost no
change at R = 39 mm). In the vertical direction, the effect is most significant around the
middle and lower portion of the tower footing, but insignificant at the top and deeper
levels below the buried depth of the tower.
Thermal effects of the tower vary not only with time, but also at different positions
around the tower footing. In the vertical direction, the thermal effect starts from zero at
the ground surface and increases to a maximum level at a certain depth (about 60 mm in
the current example). Then it decreases to zero at another depth beyond the buried depth
of the footing. Again, thermal effects of the tower are confined to a limited ground
space surrounding the tower footing. The magnitude of the temperature differences
caused by this thermal effect is relatively small ( 3 oC in this experiment) compared to
amplitudes of seasonal temperature variations in the foundation. This seasonal thermal
effect enhances the possibility of ground thawing in the summer and freezing in the
winter, if the tower foundation lies in permafrost regions.
Figure 15: Difference of amplitude of ground temperature variations with and without the tower.
Thermal protection methods

Many of the ground temperature control techniques discussed in Section 5 can be
used for protection of the tower foundation in permafrost regions. Thermosyphons have
been used by Manitoba Hydro for tower foundations in northern Manitoba. Air
convection devices can also be placed near the foundation for cooling down ground
temperature in the foundation. In this section, results of recent studies on a phase
change material (PCM) thermal barrier will be presented, along with concept design of a
solar/snow shade.
An RPCM pack was used as a PCM thermal barrier. It has a specific heat of
2.1kJ/kgK, a phase change temperature of 18 oC, and a latent heat of fusion of 195
kJ/kg. Another insulation material, Spaceloft, was also used. It is a flexible, nanoporous
aerogel insulation blanket with very low thermal conductivity, as detailed in [54]. Table
1 lists the nominal thickness and other thermal properties of these materials obtained
from the manufacturers.
Modified ground surface temperatures under the PCM layer have been illustrated in
Fig.6 of Section 5. Figure 16 shows the measured results at two particular locations (R =
7 mm, z = 20 mm, and R = 39 mm, z = 20 mm) at the same depth in a test cell, with
three different ground surface conditions, i.e., (1) bare ground without any insulation,
(2) ground surface covered with a 10 mm RPCM thermal barrier, and (3) ground surface
covered with a 3 mm Spaceloft insulation sheet, respectively. Thermal effects of the
metal tower can be observed by temperature differences at the location closer to the
tower footing (R = 7 mm, z = 20 mm) and the location further from the footing (R = 39
mm, z = 20 mm). Under the same ground surface conditions, thermal effects of the
tower cause an increase of the temperature range near the tower footing, i.e., higher
temperature in the warm periods and lower temperature in the cold periods.
Table 1 Thickness and Thermal Properties of Materials
Name (mm) (kg/m3) k (W/mK)

Spaceloft 3 150 0.014
RPCM 10 840 0.55
Figure 16: Thermal effects of a metal tower under different soil surface conditions.
It was also found that none of the ground surface insulations can prevent the
thermal effects induced by the tower. On the contrary, the measured results suggest that
the thermal barriers actually increase thermal effects of the tower. Clearly, thermal
effects of the tower show the largest amplitude under the best ground surface insulator,
i.e. Spaceloft sheet. The ground surface insulation reduces heat transfer through the
ground surface, but it cannot prevent heat conduction through the metal tower. The
insulation creates a larger temperature difference between the atmosphere and deeper
levels of the foundation, thereby causing more heat conduction through the tower to the
foundation.
Figure 17: Schematic of a snow/solar shade.
In order to keep the permafrost foundation stable, a technique of removing or

preventing the winter snow over a tower foundation could be developed. There are two
ways to implement this technique. The first is to remove the snow by line maintenance
crews in the winter. However, this may be infeasible considering the irregular snow
events, particularly for transmission lines in remote areas that are hard to access. The
second way is to design a snow prevention mechanism integrated with the tower
structure, as shown in Fig. 17. The objective is to prevent snowfall on the ground
surface above the tower foundation, but allow the cold winter wind to blow through. In
addition to snow prevention, the device can also prevent solar radiation on the ground
surface over the tower foundation, thereby cooling the foundation and reducing ground
thawing, particularly in the summer. Figure 17 shows this conceptual design.
Development of such a device requires further research in the future.
Conclusion and Remarks on Future Study
Ground heat transfer in cold (particularly permafrost) regions is usually complex due
to the seasonal freezing and thawing of soil and other transport processes. Freezing and
thawing of soil often lead to changes in its mechanical and thermal properties, as well as
various frost actions in the ground and on ground structures. Frost heave, thaw
settlement and other movements can cause severe damage to the ground structures.
Particularly permafrost thawing in the northern areas has significant impact on human
infrastructure. Different foundation problems have been reviewed in this chapter as to
several types of ground structures.
Special care must be taken in the design and construction of ground structures in cold
regions. This chapter discussed several foundation design approaches and ground
temperature control techniques to preserve the frozen ground under the structures. It
should be noted that different types of infrastructure and different ground situations
usually require different foundation design or protection techniques. It has been shown
in a case study with a power transmission tower foundation that metal ground structures
can introduce additional thawing and freezing in the ground and therefore may need
special considerations in foundation protection.
Careful site investigation and accurate prediction of the ground thermal regime is
crucial for effective design and operation of ground structures in cold regions. Literature
review on the ground heat transfer models has shown that analytical and numerical
models both have their advantages and limitations. It seems that an integration of those
two would provide better solutions for practical engineering applications. Future
research may also include improvement of existing foundation design and protection
techniques, as well as integration of several protection techniques for one type of
infrastructure.
Nomenclature
A amplitude
B intermediate variable in Eq. (6)
c specific heat capacity (J/kgC)
C volumetric heat capacity (J/m3 oC)
D distance, diameter (m, or mm), or intermediate variable in Eq. (6)
h distance from an element of the heat source to the ground surface (m)
H height (m, or mm)
k thermal conductivity (W/mC)
L length (m or mm)
P cyclic period (seconds, minutes, or months)
q heating strength (W/m)
r radial coordinate (m or mm)
R radius (m or mm)
T temperature (C)
t time (seconds, minutes, or months)
w frequency
z vertical coordinate (m, or mm)
Z depth (m, or mm)
Greek
thermal diffusivity (m2/s), or intermediate variable in Eq.(5)
intermediate variable in Eq. (5) and Eq. (9)
thickness
phase lag
intermediate variable in Eq. (6)
inclination angle
moisture content (%)
temperature response or difference (C)
density (kg/m3)
intermediate variable in Eq. (6)
Subscripts
a active layer
b buried metal rod, or tower
f freezing, phase change
g ground
gl ground response to line heat source
gs ground surface
m mean or average
p pipe
ps permafrost
q heat source
th thawed soil
References
[1] V. J. Lunardini, Heat Transfer in Cold Climates, Van Nostrand Reinhold Co.,
New York, 1981.
[2] U.S. Arctic Research Commission Permafrost Task Force Climate Change,
Permafrost, and Impacts on Civil Infrastructure, Special Report 01-03, U.S.
Arctic Research Commission, Arlington, Virginia, 2003.
[3] R. Couture, S. D. Robinson, and M. M. Burgess, Climate change, permafrost
degradation, and infrastructure adaption: preliminary results from a pilot
community case study in the Machenzie valley, Geological Survey of Canada,
Current Research 2000-B2, 2000.
[4] F. E. Nelson, O.A. Anisimov, and N.I. Shiklomanov, Subsidence risk from
thawing permafrost, Nature, vol. 410, no. 6831, pp. 889890, 2001.
[5] F. E. Nelson, O.A. Anisimov, and N.I. Shiklomanov, Climate change and
hazard zonation in the circum-Arctic permafrost regions, Natural Hazards, vol.
26, no. 3, pp.203-225, 2002.
[6] Y. Sheng, J. Zhang, Y. Liu, J. Wu, Thermal regime in the embankment of
Qinghai-Tibetan Highway in permafrost regions, Cold Regions Science and
Technology, vo. 35, no. 1, pp. 35- 44, 2002.
[7] Alaska Regional Profiles, http://www.alaskool.org/resources/regional

/nw_reg_pro/ permafrost.html, accessed on Nov. 10, 2008.
[8] T. A. Hammer, W. L. Ryan, W. L. Zirjacks, Ground temperature observations,
Thermal Design Considerations in Frozen Ground Engineering (Edited by
Thomas G. Krzewinski and Rupert G. Tart, Jr.), New York, ASCE, 1985.
[9] Natural Resource of Canada, http://gsc.nrcan.gc.ca/permafrost/wheredoes_e.php,
accessed on Nov.10, 2008.
[10] K. M. Hinkel, S.I. Outcalt, and A.E. Taylor, Seasonal patterns of coupled flow
in the active layer at three sites in northwest North America, Canadian Journal
of Earth Sciences, vol.34, no.5, pp.667-678, 1997.
[11] D. L. Kane, K.M. Hinkel, D.J. Goering, L.D. Hinzman, and S.I. Outcalt, Non-
conductive heat transfer associated with frozen soils, Global and Planetary
Change, vol.29, no.3-4, pp.275-292, 2001.
[12] National Research Council Canada, Canadian Building Digests, CBD-64.
Permafrost and Foundations, originally by G.H. Johnston, 1965, accessed
online at http://www.nrc-cnrc.gc.ca/eng/ibp/irc/cbd/digest-index.html.
[13] X. Duan and G.F. Naterer, Heat conduction with seasonal freezing and thawing
in an active layer near a tower foundation, International Journal of Heat and
Mass Transfer, vol. 52, no.7-8, pp.2068-2078, 2009.
[14] Ministry of Transportation of Quebec, Qubec Pavement Story,
http://www.mtq.gouv.qc.ca/portal/page/portal/entreprises_en/zone_fournisseurs/
reseau_routier/chaussee/chaussees_climat_quebecois.
[15] National Research Council Canada, Canadian Building Digests, CBD-128.
Adfreezing and Frost Heaving of Foundations, originally by E. Penner, and
K.N. Burn, 1970, accessed online at http://www.nrc-
cnrc.gc.ca/eng/ibp/irc/cbd/digest-index.html.
[16] D. P. Hart, R. Couvillion, Earth-coupled Heat Transfer, Dublin, OH, National
Water Well Association, 1986.
[17] G. Mihalakakou, M. Santamouris, J.O. Lewis, D.N. Asimakopoulos, On the
application of the energy balance equation to predict ground temperature
profiles, Solar Energy, vol.60, pp.181-190, 1997.
[18] M. M. S. El-Din, On the heat flow into the ground, Renewable Energy, vol.
18, no.4, pp. 473-490, 1999.
[19] X. Duan and G. F. Naterer, Ground thermal response to heat conduction in a
power transmission tower foundation, Heat and Mass Transfer, vol.44, no.5,
pp.547-558, 2008.
[20] H. S. Carslaw, J. C. Jaeger, Conduction of Heat in Solids, Second Edition,
Oxford University Press, 1993.
[21] R. Ingersoll, O. J. Zobe, A. C. Ingersoll, Heat Conduction with Engineering,
Geological and Other Applications, Revised Edition, The University of
Wisconsin Press, 1954.
[22] H. Y. Zeng, N. R. Diao, Z. H. Fang, A finite line-source model for boreholes in
geothermal heat exchangers, Heat Transfer Asian Research, vol. 31, no. 7,
pp. 558 567, 2002.
[23] G. F. Naterer, Heat Transfer in Single and Multiphase Systems, CRC Press,
Boca Raton, FL, 2003.
[24] V. E. Romanovsky, T. E. Osterkamp, Thawing of the active layer on the coastal
plain of the Alaskan arctic, Permafrost and Periglacial Processes, vol. 8, no. 1,
pp. 1-22, 1997.
[25] K. M. Hinkel, J. R. J. Nicholas, Active layer thaw rate at a boreal forest site in
central Alaska, USA, Arctic and Alpine Research, vol. 27, no.1, pp. 7280,
1995.
[26] J. R. Mackay, Active layer changes (1968 to 1993) following the forest-tundra
fire near Inuvik, N.W.T., Canada, Arctic and Alpine Research, vol. 27, no. 4,
pp. 323-336, 1995.
[27] V. A. Kudryavtsev, L. S. Garagulia, K. A. Kondratyeva, V. G. Melamed,
Fundamentals of Frost Forecasting in Geological Engineering Investigations,
Nauka, Moscow, Draft Translation 606, Cold Regions Research And
Engineering Lab Hanover N H, Hanover, NH, 1977.
[28] A. Anisimov, N. I. Shiklomanov, F. E. Nelson. Global warming and active-
layer thickness: results from transient general circulation models, Global and
Planetary Change, vol.15, no. 3, pp. 61-77, 1997.
[29] J. P. Zarling, Approximate solution to the Neumann problem, Proceedings of
the International Symposium on Cold Region Heat Transfer, American Society
of Mechanical Engineering, New York, pp. 47-57, 1987.
[30] M. W. Smith, and D.W. Riseborough, Permafrost monitoring and detection of
climate change, Permafrost and Periglacial Processes, vol. 7, no.4, pp.301-
309, 1996.
[31] V. Sivanbaev, Thawing of permafrost under a high temperature heat source,
Soil Mechanics and Foundation Engineering, vol. 8, no.1, pp.50-53, 1971.
[32] G.V. Porkhayev, Temperature fields in foundation, Proceedings of the First
International Conference on Permafrost, NRC No. 1287, pp. 285-291.
[33] G. N. Flerchinger and K. E. Saxton, Simultaneous heat and water model of a
snow-residue-soil system I. Theory and development, Transactions of ASAE,
vol. 32, no. 2, pp.565-571, 1989.
[34] T.P. Bligh and T.E. Willard, Modeling the thermal performance of earth-
contact buildings, including the effect of phase change due to soil freezing,
Computers and Structures, vol. 21, no. 12, pp.291-318, 1985.
[35] V. R. Tarnawski, Effect of snow cover on ground heat pump performance and
soil moisture freezing, International Journal of Refrigeration, vol.12, no. 2, pp.
71-76, 1989.
[36] V. R. Tarnawski, P. K. Yuet, Winter performance of residential heat pump,
Heat Recovery Systems and CHP, vol.8, no. 3, pp.271-278, 1988.
[37] C. Waelbroeck, Climate-soil processes in the presence of permafrost: a systems
modelling approach, Ecological Modelling, vol. 69, no. 3, pp. 185-225, 1993.
[38] T. Degaetano, D. S. Wilks, M. Mckay, A physically based model of soil
freezing in humid climates using air temperature and snow cover data, Journal
of Applied Meteorology, vol. 35, no. 6, pp. 1009-1027, 1996.
[39] A. Anisimov, N. I. Shiklomanov, F. E. Nelson, Variability of seasonal thaw

depth in permafrost regions: a stochastic modeling approach, Ecological
Modelling, vol. 153, no. 3, pp. 217227, 2002.
[40] F. E. Nelson, N. I. Shiklomanov, G. Mueller, K. M. Hinkel, D. A. Walker, J. G.
Bockheim, Estimating active layer thickness over a large region: Kuparuk river
basin, Alaska, USA, Arctic and Alpine Research, vol. 29, no. 4, pp.367378,
1997.
[41] N. I. Shiklomanov, F. E. Nelson, Analytic representation of the active layer
thickness field, Kuparuk River Basin, Alaska, Ecological Modelling, vol.123,
no.2, pp.105-125, 1999.
[42] D. Riseborough, N. Shiklomanov, B. Etzelmuller, S. Gruber, and S. Marchenko,
Recent advances in permafrost modeling, Permafrost and Periglacial
Processes, vol. 19, no. 2, pp.137156, 2008.
[43] K. A. Linell and E. F Lobacz, Design and Construction of Foundations in Areas
of Deep Seasonal Frost and Permafrost, Cold Regions Research and
Engineering Lab, Hanover, NH, 1980.
[44] Joint Departments of the Army and the Air Force, USA, Technical Manual TM
5-852-4/AFM 88-19, Chap. 4, Arctic and Subarctic Construction, Foundations
for Structures, 1983.
[45] A. Staudzs, Frost and permafrost: effects and remedial work on transmission
tower foundations, Manitoba Hydro, report, 1982.
[46] E. Heuer, E. L. Long, and J. P. Zarling, Passive techniques for ground
temperature control, in: Thermal Design Considerations in Frozen Ground
Engineering, Edited by T. G. Krzewinski and R. G. Tart, Jr., New York, ASCE,
1985.
[47] G. Cheng, A roadbed cooling approach for the construction of QinghaiTibet
Railway, Cold Regions Science and Technology, vol. 42, no.2, pp.169 176,
2005.
[48] W. Yu, Y. Lai, X. Zhang, S. Zhang, J. Xiao, Laboratory investigation on
cooling effect of coarse rock layer and fine rock layer in permafrost regions,
Cold Regions Science and Technology, vol.38, no.1, pp.31 42, 2004.
[49] D. C. Esch, Insulation performance beneath roads and airfields in Alaska,
Cold Region Engineering. Proceedings of 4th International Specialty
Conference, ASCE, New York, 1986, pp. 713722.
[50] V. N. Feklistov, N. L. Rusakov, Application of foam insulation for remediation
of degraded permafrost. Cold Regions Science and Technology, vol.24, no.2,
pp. 205-212, 1996.
[51] Z. Wen, Y. Sheng, W. Ma, J. Qi, Evaluation of EPS application to embankment
of QinghaiTibetan railway, Cold Regions Science and Technology, vol. 41,
no.3, pp. 235- 247, 2005.
[52] Y. Sheng , Z. Wen, W. Ma, Y. Liu, J. Qi, J. Wu, Long-term evaluations of
insulated road in the Qinghai-Tibetan plateau, Cold Regions Science and
Technology, vol.45, no.1, pp.2330, 2006.
[53] J. Liu, Y. Tian, Numerical studies for the thermal regime of a roadbed with
insulation on permafrost, Cold Regions Science and Technology, vol. 35, no.1,
pp.113, 2002.
[54] E. R. Bardy, J. C. Mollendorf, D. R. Pendergast, Thermal conductivity and
compressive strain of aerogel insulation blankets under applied hydrostatic
pressure, ASME Journal of Heat Transfer, vol. 129, no.2, pp. 232-235, 2007.
[55] X. Duan and G. F. Naterer, Experimental investigation on a phase change
ground thermal barrier, The 20th International Symposium on Transport
Phenomena, Victoria, British Columbia, Canada, July 7-10, 2009.
[56] L. Daigle, J. Q. Zhao, Effectiveness of rigid insulation for thermal protection of
buried water pipes in rock trenches, CSCE 1999 Annual Conference-1st Cold
Regions Specialty Conference, June 1999, pp. 389-398.
[57] G. Cheng, J. Zhang, Y. Sheng and J. Chen, Principle of thermal insulation for
permafrost protection, Cold Regions Science and Technology, vol. 40, pp.71-
79, 2004.
[58] R. Reid, A. Evans, Investigation of the air convection pile as a permafrost
protection device, Proc. 4th Int. Conf. Permafrost, National Academy Press,
pp.1048-1053, 1983.
[59] G. Lock, K. Chong, A. Dyckerhoff, J. Huyer, B. Jones, On the design of wind-
augmented thermosyphons, Cold Regions Science and Technology, vol. 16,
pp.1123, 1989.
[60] N. A. Buchko, I. A. Maksimov, M. P. Pavchich and V. A. Turchina, Analysis
of the operating capacity of seasonally functioning cooling devices using
operational data of waterworks facilities, Journal Power Technology and
Engineering, vol. 38, no. 2, pp. 65-70, 2004.
[61] W. Zhi, S. Yu, W. Ma; J. Qi, J. Wu, Analysis on effect of permafrost protection
by two-phase closed thermosyphon and insulation jointly in permafrost regions,
Cold Regions Science and Technology, vol. 43, no.3, pp.150-163, 2005.
[62] X. Duan and G. F. Naterer, Ground heat transfer from a varying line source
with seasonal temperature fluctuations, ASME Journal of Heat Transfer,
vol.130, no.11, pp.111302, 2008.
[63] Canadian Climate Normals or Averages 1971-2000, Environment Canada,
http://www.climate. weatheroffice.ec.gc.ca/climate_normals/index_e.html,
accessed on Nov. 1, 2008.
Chapter 8
EvaluationofFourNucleateFlowBoilingModels
J.P. Kroes1, C. W. M. van der Geld1 and E. van Velthooven2

1
Department of Mechanical Engineering, Technische Universiteit Eindhoven, P.O. Box
513, 5600 MB Eindhoven, The Netherlands
2
DAF Trucks, CAE-Engines, Hugo van der Goeslaan 1, 5643 TW Eindhoven, The
Netherlands
Abstract
The merits of various existing 1D models for nucleate boiling heat transfer are
assessed with four data sets from the literature. Several implementation problems and
physical inconsistencies of these models have been identified.
Surprisingly, the Chen correlation predicts the measurements at least as good as
more sophisticated, recent models. For heat fluxes in the range of a diesel engine the
Chen correlation even predicts best. Prospects are discussed.
Introduction
Let us examine a less known example of nucleate boiling in a well-known

application. The lifetime of the cylinder head of a truck diesel engine is very dependent
on its thermal loading. Cooling of the head is achieved by pumping a water-glycol
mixture (e.g. 50%50%) through a system of connected cavities that we shall consider
as and name channel here. Nucleate boiling occurs near the cylinder head flame deck
since heat fluxes are locally high. The channel geometry is characterized by a varying
cross-section area, see Fig. 1, a complex connectivity and spatially varying wall tempe-
rature and heat flux. Typical temperature difference between engine cooling in- and
outlet is 6 oC, typical cooling power about one third of the total engine power and
typical total pressure drop is 0.3 MPa. Even in the case of a single component coolant,
which is this case that we shall further consider in this chapter, it is hard to predict heat
transfer in this coolant geometry.
State-of-the art of prediction of heat transfer in the cylinder head is done with 3D
CFD/FE analysis of the engine block in combination with 1D-modeling of nucleate
flow boiling in the cooling channel. 1D means using bulk properties averaged over a
cross-section area normal to the mean flow. At some locations this definition is
ambiguous, but other intrinsic difficulties exist that are more important. Each 1D model
Email address: c.w.m.v.d.geld@tue.nl, tel: +31 40 2472923, fax: +31 40 2475399.

268 Advances in Multiphase Flow and Heat Transfer 1 (2009) Kroes et al.
that can be applied and is known from the literature is essentially a set of correlations
based on experimental data for well-established flow without a mean swirl component
and measured over some length of tube with constant cross-section. In the present
application, these prescriptions have to be applied to swirling, accelerated, undeveloped
flows. On places with void fraction redistribution in a cross-section bubble-bubble
interaction may differ significantly from that in straight channels. To account for this,
an essentially two-dimensional description of nucleate, subcooled boiling would need to
be applied. Like in CFD, heat transfer would be evaluated locally, in elements near the
wall, say, with the aid of prescriptions based on local measurements. Unfortunately,
such two-dimensional modeling does not exist, as far as we know. The more recent
boiling models that are applicable, to be listed below, are 1D despite the fact that they
attempt to incorporate some of the transport phenomena that occur in the cross-sectional
plane. It may of course be attempted to use these 1D models in 3D CFD of heat transfer
in a coolant system, but difficulties will arise because of, for example, the use of a bulk
temperature. In the 1D models analyzed below the bulk temperature of the liquid is
used. This does not facilitate direct application in CFD computer codes.
Exhaust
Intake
Exhaust
Intake
Figure 1: Coolant flow in the jacket right above the combustion chamber. The water side is shown
in a top view. Coolant flows in the bridges between the intake- and exhaust manifolds. Reaching the
injector it flows upwards in an annulus-type channel.

The present study attempts to assess the merits of various existing 1D models for
nucleate boiling. To this end, prediction results will be compared with published results
of experiments in the literature. Inlet and heating parameters will be varied in a wide
range.
The complexity of implementation of the models will be analyzed. To this end, all
models have been implemented in a new code and no use of existing or available codes
Evaluation of Four Nucleate Flow Boiling Models Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 269
has been made. No sensitivity analysis will be carried out and it will not be attempted to
modify the models. The new code is made with Matlab and Simulink, and will be
made available at our website w3.wtb.tue.
Distance
FLOW
Temperature
Bubble Increasing flow
Decreasing temperature at bubble center
Figure 2: Principle of boiling suppression. At increased flow the steeper temperature gradients
reduce the net superheat and henceforth bubble growth.
Main Features of the 1D Models
The models investigated all introduce parameters in and ad hoc way, and are
therefore more like correlations than models. However, the name correlation will
henceforth be used for one-formula recipes and the name model used otherwise. Only
the main features of these models will here be described shortly since often many details
are involved in describing the models completely, details which are all described in the
open literature. The corresponding references will be given below. The focus here is on
elucidating intrinsic difficulties with implementing the models and on physical
inconsistencies of the models. They will be described in detail in section 3, after the
short introduction in the models given below.
The models examined are the following approaches based on mechanistic modeling:
The well-known correlation of Chen (1966) that computes the total heat transfer as
the sum of a forced convection contribution and a nucleate boiling one. So-called
bubble growth suppression is taken into account in the nucleate boiling contribution.
Bubble growth suppression is due to decreasing mean temperatures near the wall
with increasing mean velocity, see Fig. 2. Chen was the first to consider this effect
and he interpreted it in terms of effective superheat available for bubble growth
(Chen, 1966).
The extension of Chens correlation by Steiner et al. (2005) which is called by these
authors the boiling departure lift-off model. It utilizes two suppression parameters.
The first accounts for bubble growth suppression with the aid of estimates of bubble
diameters at detachment and at take-off (lift-off) from the wall. The other
suppression parameter of Steiner et al. accounts for condensation at the bubble top
in subcooled boiling.
The Basu, Warrier and Dhir model (2005a). This model assumes that all energy
from the wall is transferred to a superheated liquid layer immediately adjacent to the
wall. Heat transfer to this layer is split in two contributions, see also Fig. 3. One
contribution is what we call enhanced mixing due to departing or sliding bubbles
disrupting the fluid layer adjacent to the wall, the other contribution is merely
enhanced forced convection. Characteristic times of bubble growth and waiting until
another bubble is initiated are used to quantify the enhanced mixing. Also a
correlation for the nucleation site density is a necessary ingredient. From the
superheated liquid layer heat is transferred to the bulk
o either by enthalpy of evaporation of bubbles migrating away from the wall,
Qev,
o by condensation at the bubble top of a bubble growing at the wall, Qcond, or
o by a mechanism that is called direct heating of the bulk liquid.
The so-called direct heat of the bulk is conveniently defined as the heat flux from the
wall to the superheated layer minus the sum of the other two contributions,
(Qev+Qcond). Some consequences of this definition will be discussed in section 3. In
Fig. 3 the contributions are depicted.
Although the model is fully described in the Basu et al. paper of 2005 (2005a),
some details are explained in previous work of these authors (Basu et al., 2002,
2003).
The model of Kolev (2006). A mechanistic model for single bubble detachment in
the presence of other bubbles is combined with, amongst others, a correlation for the
nucleation site density to predict a characteristic velocity scale, see also Zuber
(1963). Combination with a characteristic length scale yields the time scale for what
Kolev calls bubble induced turbulence. The nucleate boiling heat flux then follows
from enhanced mixing, utilizing this time. This approach is to some extent
comparable to the Basu et al. (2005a) approach.
Direct communication with Kolev revealed that for forced convective boiling a
correlation for convective heat transfer has to be used in conjunction with the above
nucleate boiling model, without suppression factors as in the Chen correlation.
Kolev evaluates physical properties as surface tension coefficient and evaporation
enthalpy at the wall temperature, Twall, and at the corresponding saturation pressure,
psat(Twall), rather than at system pressure and corresponding saturation temperature.
Kolev (1994-2006) proposed three correlations in total for the active nucleation site
density, Nwall, to be used. Direct private communication with Kolev taught us that
different site densities are expected to correspond to different properties of the
surface at which boiling takes place. All three site correlations for site density might
therefore be appropriate.

In this chapter the Dittus-Boelter correlation for forced convection is used for the
convective parts of models; this correlation is a basic constituent in many boiling heat
transfer correlations (Collier and Thome, 1994).
Application Problems and Physical Inconsistencies
During implementation and testing of the models described briefly in section 2 the
following findings were made.
The predictions of Chen and Steiner et al. could readily be reproduced (see e.g. Fig.
4). Boiling suppression as defined by Chen is in the Steiner (2005) model computed
with the aid of a bubble growth model that yields a ratio of two radii. This ratio is used
as suppression factor. Actual boiling suppression is caused by the temperature drop
from wall to bulk, as explained in section 2. And the bubble growth rate in the Steiner
model is not adapted for suppression. Hence the Steiner way of accounting for
suppression is a rather indirect one.
Condensation
FLOW in bulk liquid
Direct heating
of bulk liquid Condensation
at bubble top
Boundary
layer Evaporation
Sliding bubble Bubble lift-off

Transient Enhanced
conduction convection
Subcooled liquid Superheated liquid Wall

Figure 3: Schematic of the Basu, Warrier and Dhir model (2005a).

Steiner et al. (2005)

600
p = 1.5 bar, ReD = 5547
500 p = 1.5 bar, Re = 43264

D
p = 1.5 bar, Re = 129791
D
qwall ( kW / m 2 )
400
300
200
100
0
-20 -10 0 10 20 30
Wall superheat Twall-Tsat(p) (K)
Figure 4. Reproduction of boiling curves predicted by Steiner et al. (2005). Subcooled flow boiling
of water along a smooth surface (= 0) at a pressure of 1.5 bar; subcooling of 16 K; bulk
temperature, Tbulk, is 95 C; Dhydr = 34 cm.
Kolev, active nucleation site density, Nwall

9
10
Gaertner & Westwater (1960)
8 Wang & Dhir (1993)
10
new correlation by Kolev (2006)
7
10
6
10
Nwall (#/m2)
5
10
4 Kolev (2006)
10
3
10
2
10
1
10
3 4 5 6 7 8 9 10 20 30
Figure 5. The three correlations of Kolev (19942006) for the nucleation site density at 0.1 MPa and
a static contact angle, , of 35 o. Gaertner and Westwater (1960) data correlated by Kolev (1994).
Properties in the Wang and Dhir correlation (1993) are evaluated at wall temperature, see Kolev
(1995).
Kolev, bubble departure diameter

(saturated pool boiling)
6 Nwall acc. to Gaertner & Westwater (1960)
Nwall acc. to Wang & Dhir (1993)
5
Nwall acc. to Kolev (2006)
Db,dep (mm)
3
Kolev (2006)
2
0
0 5 10 15 20 25 30 35 40 45 50
Figure 6: Predictions with the Kolev model for 0.1 MPa and a static contact angle of 35 o. acc.
means according. The labeled curve is given by Kolev (2006).
Kolev, wall heat flux

(saturated pool boiling)
6
10
Kolev (2006)
qwall ( W / m 2 )
5
10
4
10
Nwall acc. to Gaertner & W (1960)
3
Nwall acc. to Wang & Dhir (1993)
10
Nwall acc. to Kolev (2006)
0 1 2
10 10 10
Figure 7: Predictions with the Kolev model for 0.1 MPa and a static contact angle of 35 o. The
labeled curve is given by Kolev (2006).
Figure 8: Predictions with the Kolev model (nucleation site density according to 1994, Gaertner and
Westwater data) for 0.1 MPa and a static contact angle, , of 35 o. Note the in the equation. To be
compared with Fig. 9. The thick solid line is the desired solution, where the sum of governing forces
is zero.
Figure 9: Predictions with the Kolev model (nucleation site density according to 1994, Gaertner and
Westwater data) for 0.1 MPa and a static contact angle of 35 o. Note the | in the equation. See
caption Fig. 8. Part of the surface yields the solid line in Fig. 6.
vbulk = 1.9 m/s Dhydr = 15 mm

p = 2 bar (Tsat=120.21 C) Subcooling = 2 K
Wall temperature T (C)

wall
120.21 130.21 140.21 150.21 160.21 170.21
( mv hlv / qtotal ) 100 (%)
Basu et al.
Kolev
100
50
0
0 10 20 30 40 50
Figure 10: The ratio of the evaporation enthalpy flux to the total wall enthalpy flux as computed
with Kolev (2006) and Basu et al. (2005a) models. In Kolevs model the nucleation site density is
computed according to the Wang and Dhir correlation (1993), with properties evaluated at the
temperature of the wall.
MODEL PREDICTIONS FOR SUBCOOLED FLOW BOILING
p = 2 bar (Tsat=120.21 C) Subcooling Tsat-Tbulk = 30.21 K (Tbulk=90 C)

2
vbulk = 1.5 m/s Dhydr = 15 mm = 49 = 260 m g = 9.81 m/s
Wall temperature T (C)

wall
90.21 100.21 110.21 120.21 130.21 140.21 150.21 160.21 170.21
Wall heat flux qwall ( kW / m 2 )
1200
1000
800
600
400 Chen
Steiner
200 Basu et al.
Kolev
0
-30 -20 -10 0 10 20 30 40 50
Figure 11. Comparison of predicted boiling curves. In Kolevs model the nucleation site density is
computed according to the Wang and Dhir correlation (1993), with properties evaluated at the
temperature of the wall.
The model of Basu et al. (2005a) takes sliding of bubbles into account with the aid
of if-then criteria. These criteria inevitably result in discontinuities in predictions of for
example the boiling curve. Some examples will be mentioned below.
The approach of Kolev consists of the computation of a bubble detachment
diameter, Ddep, basically from correlations for the nucleation site density, and
subsequently the calculation of a heat flux from the wall, qwall. Figures 5 7 show that
predictions of Ddep and qwall published in Kolev (2006) correspond to different
correlations for the nucleation site density. When his latest expression (2006) is used,
previous results for the departure diameter, Ddep, and the heat flux, qwall, are not
recovered anymore. This should be considered as a consequence of variation of surface
properties, as discussed in section 2.
The procedure to compute the departure diameter as given by Kolev contains the
following mistake, see Fig.s 8 9. The thick lines in these Figures represent the values
of Ddep and wall superheat where the sum of all force components is zero, as required.
In Fig. 8 the solution is not single-valued but in Fig. 8 the solid line of Fig. 6 is
recovered. In the force balance on a bubble at the wall an angle, occurs that the
resulting force on the bubble makes with the wallOnly after taking the absolute value
of the results published by Kolev could be reproduced, see Fig. 9. The necessity to
take the absolute value of makes the model physically inconsistent. Moreover, the
absolute value is required to obtain a decreasing tendency of Ddep with increasing
superheat at superheats exceeding 20 K, see also Fig. 6. This decreasing tendency
causes the change in the Nukiyama curve, where heat flux is plotted versus superheat.
It is noted that assessment of physical properties, as surface tension , at the wall
temperature (as required in the Kolev model) can be cumbersome, especially if
multicomponent mixtures are considered (which is not done here).
Both the Basu et al. and Kolev models comprise detailed descriptions of detaching
bubble dynamics which enable to compute the latent heat carried away by the bubbles
from the wall to the bulk liquid, mvhlv. The total heat flux away from the wall, qtotal, is
in these models predicted in an independent way, which makes it interesting to compare
mvhlv with qtotal. It turns out that in some cases the latent heat flux mvhlv exceeds the
total heat flux, see Fig. 10. Obviously this is physically impossible. This unexpected
finding is a consequence of the way the models of Basu et al. and Kolev have been
composed. It might be possible to improve the various submodels (e.g. those related to
bubble dynamics) as stated at the end of a paper of Basu et al. (2005b). This might
however turn out to be insufficient to warrant energy conservation and to prevent
mvhlv to exceed qtotal. Model partitioning should be done in a way that is consistent
with energy conservation.
To come back to the discontinuities of the Basu et al. model as mentioned above,
notice Fig. 10.
Predicted boiling curves are compared in Fig. 11 for arbitrary conditions. Significant
prediction differences occur at superheats exceeding 10o. Notice again the
discontinuities in the predictions of Basu et al. These differences found between the
predictions of various models for circumstances occurring in common practice
necessitate a further assessment of the merits of these models. This is done in the next
section, 4, where a comparison is made with experimental boiling results found in the
literature.
a) Steiner et al. (2005) b) Boyd et al. (2003)
c) Gollin et al. (1995) d) McAdams (1954)
10 mm
Water Metal
Figure 12: Schematics of cross-sections and locations where heat is added in the experiments used
for validation.
Comparison to Experiments
Model predictions have been compared with four data sets from literature. In
principle any arbitrary coolant can be chosen, but only water has been selected as
coolant fluid because its physical properties can easily be obtained at all temperatures.
The data reported by Steiner et al. (2005), Boyd et al. (2003), Gollin et al. (1995)
and McAdams (1954) have been used to compare with predictions of the models briefly
described in section 2 and criticized in section 3. Table 1 summarizes typical flow and
heat transfer parameters of the four data sets selected, and those of a real diesel engine
for comparison as well. In Fig. 12 the shape of the cross-sections are depicted.
In Figs. 13 16 predictions of all four models are compared with the experimental
data. Surface properties in the Kolev model are based on the Wang and Dhir correlation
(1993) with properties evaluated at the temperature of the wall.
None of the models provides an overall good prediction. For heat fluxes below about
300 kW/m2 deviations to 50 % occur. Errors are bigger at heat fluxes exceeding 300
kW/m2, indicating that the boiling contribution is not predicted correctly.
The data set of Steiner et al. (2005) is best predicted by their own correlation, as was
to be expected, although surprisingly deviations over 50% can be observed.
Table 1 Measuring conditions.
Table 1a
Author Shape/ Coolant Dhydr vbulk qwall
Material [kW/m2]
[mm] [m/s]
Diesel engine misc/ iron mixture 1320 12 5001200
Steiner 2005 rect/ alu water 34.3 0.051.17 0558
Boyd 2003 circ/ copper water 10 0.6 01309
Gollin 1995 circ/ alu water 15.9 0.22/0.32 0644
McAdams 54 circ/ alu water 12.2 0.3-3.6 3513000
Table 1b
Author Tsub Tsup p Tsat [C] Tbulk [C] Twall [C]
[bar]
[C] [C]
Diesel engine 60 050 2 150 90 150200
Steiner 2005 16/25 -2536 1.5/2 111/ 120 95 94150
Boyd 2003 6498 -9332 2.07 121 2357 28153
Gollin 1995 1336 -2051 2.05 121 85108 102172
McAdams 54 1156 -4639 4.13 145 89134 99183
Compared to the other data sets, the one of Boyd et al. (2003) is most representative
for the conditions that occur in a diesel engine. It covers for example the complete heat
flux range in the engine. This set is best predicted by the Chen (1966) correlation.
The prediction results of attempts to incorporate bubble dynamics in 1D models for
nucleate boiling are not very convincing. Possible ways to proceed are discussed after
some concluding remarks in section 5.
Prediction by Chen (1966)

4
10
predicted qwall ( kW / m 2 )
3
10
2
10
1
10 Steiner et al.
Boyd et al.
0 Gollin et al.
10
McAdams
0 1 2 3 4
10 10 10 10 10
2
measured qwall ( kW / m )
Figure 13: Comparison of measurements and prediction with the correlation of Chen (1966).
Dotted lines indicate the 50 % limits.
Prediction by Steiner et al. (2005)

4
10
3
10
2
10
1
10 Steiner et al.
Boyd et al.
0 Gollin et al.
10
McAdams
0 1 2 3 4
10 10 10 10 10
2
Figure 14: Comparison of measurements and prediction with Steiner et al. (2005). Dotted lines
indicate the 50 % limits.
Prediction by Basu et al. (2005)

4
10
3
10
2
10
1
10 Steiner et al.
Boyd et al.
0 Gollin et al.
10
McAdams
0 1 2 3 4
10 10 10 10 10
2

Figure 15: Comparison of measurements and prediction with the model of Basu et al. (2005a).
Dotted lines indicate the 50 % limits.
Prediction by Kolev (2006)

4
10
3
10
2
10
1
10 Steiner et al.
Boyd et al.
0 Gollin et al.
10
McAdams
0 1 2 3 4
10 10 10 10 10
2
Figure 16: Comparison of measurements and prediction with the model of Kolev (2006). The
nucleation site density is computed according to the Wang and Dhir correlation (1993), with
properties evaluated at the temperature of the wall. Dotted lines indicate the 50 % limits.
Conclusions
Next to the correlation of Chen, three more recent models for the prediction of heat
transfer in subcooled boiling have been used to compare predictions with each other and
with four experimental data sets from the literature. In the three recent models attempts
were made to account for the dynamics of bubbles at or near a heated wall. The
implementation of these models turned out to be fairly complex not merely because of
an extensive set of equations and definitions, but also because of occasionally
ambiguous or even erroneous descriptions in the literature. A library of Matlab and
Simulink procedures has been set up that will be made available via our website. The
original results of the authors 1D model have been reproduced. A number of physical
modeling inconsistencies have been identified. Some stem from the indirect way of
modeling applied.
Surprisingly, the Chen correlation predicts the measurements best, in particular for
the data set of Boyd et al. (2003) which is most representative for the situation in a
diesel engine. It is apparently difficult to incorporate detailed physical descriptions of
bubble behavior in a 1D model and subsequently obtain good agreement with
experiments of heat transfer in various operating conditions. The reasons for this
difficulty may be in the stochastic nature of boiling, as discussed by Dhir (2006) in his
Max Jakob Award paper.
However, the mechanisms of bubble growth and detachment are essential
ingredients of nucleate boiling and need to be accounted for properly if a reliable
prediction method is to be obtained. If the 1D framework fails, 2D modeling, i.e. ac-
counting for variations in radial direction in a cross-sectional plane, might be the proper
way to proceed. As discussed in the introduction this might require some new, dedicated
boiling experiments to circumvent the use of e.g. the bulk temperature.
Acknowledgement
The authors thank Dr. Kolev for valuable information and a lively discussion of the
first version of this chapter.
Nomenclature
Roman
D diameter (m)
F Force (N)
F Adhering force related to surface tension (N)
g gravitational acceleration (9.81) (m/s2)
h heat transfer coefficient (W/m2K)
mv mass flux of vapor (kg/m2)
N nucleation site density (#/m2)
p pressure (Pa)
q heat flux (W/m2)
Q heat flow rate (W)

T temperature (K)
v velocity (m/s)
Greek
contact angle (rad)
hlv latent heat of evaporation (J/kg)
surface roughness (m)
surface tension (N/m)
Subscribts
b bubble
bulk bulk
cond condensing
dep departure
ev evaporative
hydr hydraulic
lv liquidvapor
n normal to the wall
sat saturation conditions
sub subcooled
sup superheated
t tangential or parallel to the wall
tot total
v vapor
wall wall
References
Basu, N., Warrier, G.R., Dhir V.K., 2002, Onset of nucleate boiling and active
nucleation site density during subcooled flow boiling, ASME J. Heat Transfer, Vol 124
no 4 pp 717-728.
Basu, N., Warrier, G.R., Dhir V.K., 2003, Wall heat flux partitioning during
subcooled flow boiling at low pressures, Proceedings of HT2003 ASME Summer Heat
Transfer Conference, July 21-23, 2003, Las Vegas, USA.
Basu, N., Warrier, G.R. and Dhir, V.K., 2005a, Wall heat flux partitioning during
subcooled flow boiling: Part I model development, Journal of Heat Transfer, Vol.127,
pp. 131-140.
Basu, N. Warrier, G.R. and Dhir, V.K., 2005b, Wall heat flux partitioning during
subcooled flow boiling: Part II model validation, Journal of Heat Transfer, Vol.127,
pp. 141-148.
Boyd, R.D., Strahan, M., Cofie, P., Ekhlassi, A., Martin, R., 2003, High heat flux
removal using water subcooled flow boiling in a single-side heated circular channel, Int.
Journal of Heat and Mass Transfer, Vol 46 pp 4105-4117.
Chen, J.C., 1966, Correlation for boiling heat transfer to saturated fluids in
convective flow, Ind. & Eng. Chem. Process Design and Development, Vol 5 no 3 pp
322-329.
Collier, J. and Thome, J., 1994, Convective boiling and condensation, Clarendon
Press, Oxford.
Dhir, V.K., 2006, Mechanistic prediction of nucleate boiling heat transfer-
achievable or a hopeless task?, J. of Heat Transfer, Vol. 128, pp 1-12.
Gaertner, R.F., Westwater, J.R., 1960, Population of active sites in nucleate boiling
heat transfer, Chem. Eng. Prog. Symp. Ser., Vol 30 pp 39-48.
Gollin, M., McAssey, E.V., Stinson, C., 1995, Comparative performance of ethylene
glycol/water and propylene glycol/water coolants in the convective and forced flow
boiling regimes. SAE Technical Paper Series, Paper 950464.
Kolev, N.I., 1994, The influence of mutual bubble interaction on the bubble
departure diameter, Experimental Thermal and Fluid Science, Vol 8, pp 167-174.
Kolev, N.I., 1995, How accurately can we predict nucleate boiling? Experimental
Thermal and Fluid Science, Vol 10 pp 370-378.
Kolev, N.I., 2006, The internal characteristics of boiling at heated surfaces, Proc. of
the ECI International conference on Boiling Heat Transfer, Spoleto.
McAdams, W.H., 1954, Heat transmission, McGraw Hill, New York.
Steiner, H., Kobor, A. and Gebhard, L.A., 2005, A wall heat transfer model for
subcooled boiling flow, Int. J. of Heat and Mass Transfer, Vol. 48, pp. 4161-4173.
Wang, C.H., Dhir, V.K., 1993, Effect of surface wettability on active nucleation site
density during pool boiling of water on a vertical surface, Trans. ASME J. Heat
Transfer, vol 115 pp 659-669.
Zuber, N., 1963, Nucleate boiling, the region of isolated bubbles and the similarity
with natural convection, Int. J. Heat Mass Transfer, Vol 6 pp 53-78.
Index Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) 284-287 284
Index
A Classifier, 100
Cold, 247
Abbreviation, 109 Combined, 179
Abstract, 1, 53, 93, 114, 158, 213, 233, Comparison, 278
268 Complete, 171
Acknowledgement, 45, 152, 226, 262, Compound, 158, 163, 171
282 Computational, 116
Acoustic field, 75 Conclusion, 82, 108, 150, 202, 226, 260
Action, 179, 239, 246 Convection, 71
Active, 241 Condition, 43, 57
Adiabatic, 105 Conducting, 201
Advanced, 147 Conductivity, 207
Air, 41, 42 Control, 248
Analysis, 254 Cooling, 81
Analytical, 241 Coordinate, 165, 166, 168, 170
Annular, 213, 220 Corium, 126, 128
Annular flow, 213, 220 Correlation, 3, 7, 12, 14, 16, 40, 57
Application, 41-43, 93, 105, 107, 272 Curvilinear, 165
Applied, 199
Approach, 247 D
Approximate, 124, 137
Approximation, 122 Deformable, 185
Design, 247
B Diacritical mark, 47
Disperse, 124
Beam, 213 Drilling, 213
Bi-spherical, 166 Drop, 158, 163, 171, 187, 195, 199, 201
Boiling, 53, 54, 57, 61, 63, 72, 73, 76, Droplet, 127
105 Dynamics, 61, 98, 199, 207
Boiling flow
Bubble, 61, 126 E
C Effect, 74, 75, 160, 179, 213, 240

Electric field, 74
Case, 196, 205, 207, 253 Engineering, 240
Cell, 121, 145 Engulfed, 195
Channel, 106 Engulfment, 171
Chapter, 1, 53, 93, 114, 158, 213, 233, Environment, 53
268 Expression, 128
Characteristics, 124 Experiment, 278
Civil, 233 Experimental, 21, 24, 31, 35, 97
Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Index 285
Extension, 57 Horizontal, 21, 24, 25, 33

External, 188, 192, 193 Hydraulic, 93, 105
Externally, 173, 199
Evolution, 223 I
F Identification, 93, 105, 107
Identifier, 98
Facility, 60 Illustration, 40
Fast, 107 Imposed, 173
Feature, 270 Inclined 21, 23, 33
Field, 74, 75, 199 Inconsistency, 272
Final, 128, 224 Indicator, 96, 97
Flow, 21, 24, 31, 33, 41, 42, 105-107, Induced, 173, 198, 199, 207
114, 158, 160, 163, 213, 220, 222 Infrastructure, 233
Flow boiling, 56, 76 Intensity, 213
Flow map, 1 Interaction, 114, 140
Fluid, 160 Interface, 185
Flow pattern 1, 17, 25, 29, 35, 76 Internal, 193
Formation, 213, 218 Introduction, 1, 53, 93, 114, 158, 213,
Flow regime, 93, 96-100, 105 233, 268
Flow visualization, 1 Investigation, 33
Foundation, 247, 253 Isothermal, 196
Free, 221
Freezing, 233, 236, 238, 240 K
Frost, 236, 239, 246
Future, 82, 260 Keyhole, 216, 222, 223
G
L
Gas-liquid, 16, 33
Gas oil, 41 Large, 121, 145
Geometry, 106, 205 Layer, 241
Gradient, 173, 192, 193 LCRM, 121, 147
Gravity, 53, 170 Letter, 227
Greek, 46, 84, 109, 227, 262, 282 Locomotion, 174
Ground, 233, 234, 236, 240, 241, 248, Low gravity (Low-g), 57, 60, 61, 76
253
M
H
Main, 270
Heat transfer, 1, 12, 14, 16, 33, 35, 40, Marangoni, 71, 158, 160, 179
53, 77, 233, 241, 253 Mathematical, 109
Heaving, 236 Mathematical symbol, 109
High, 213 Mechanical, 238
High-intensity, 213
286 Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Index
Mechanism, 236 Prevention, 220

Medium, 116, 124 Problem, 140, 272
Melt-coolant, 114, 140 Property, 238
Method, 259
Methodology, 93, 105 Q
Microgravity, 43
Mixture, 73 Quenching, 80
Model, 116, 120-122, 140, 141, 145,
220, 241, 245, 270
R
Modeling, 241
Motion, 158, 160, 173, 175, 190-194
Radiation, 114, 116, 121, 122, 128, 141
Multiphase, 114, 124, 158, 163
Radiative, 124, 147
Multiphase flow, 114
Reduced, 53
Reference, 47, 84, 100, 152, 209, 227,
N 262, 282
Regime, 234
Nomenclature, 45, 83, 109, 261, 282 Region, 234, 247
Non-conducting, 199 Relation, 124
Non-boiling, 1 Remark, 260
Nucleate, 63 Result, 31, 35, 141
Numerical, 141, 245 Roman, 109, 282
O S
OMM, 116 Seasonal, 233
Opaque, 116 Self-induced, 175
Overview, 54 Setup, 21, 24
1D Shape, 221, 223, 224
Silicone, 42
P Simplified, 141
Single, 128, 137
P1, 122 Soil, 236, 238
Parameter, 98 Solidification, 137,
Partially, 195, 201 Solution, 137
Particle, 126, 128, 137, 160 Special, 205, 207
Permafrost, 234, 246, 248 Spontaneous, 190, 194
Phsical, 272 Spontaneously, 207
Physics, 216 Spray cooling, 81
Pipe inclination 1, 3, 17, 35 Spray, 81
Pool boiling, 54, 61, 63 Sphere, 168
Pore, 213, 218, 224 Statistical, 98
Porosity, 220 State, 107
Practical, 40 Steady, 107
Preservation, 248 Steady state, 107
Protection, 246, 259
Advances in Multiphase Flow and Heat Transfer Vol. 1 (2009) Index 287
Steam, 126 Verification, 145

Structure, 240 Vertical, 220
Study, 253, 260 Viscous, 160
Subscript, 46, 84, 109, 227, 262, 282 Void fraction, 1, 3, 31
Summary, 44
Surface, 221 W
Superscript, 46, 227
Symbol, 46, 109 Water, 127
System, 166, 170, 236 Welding, 213
Systematic, 33 Work, 82
T
Tangent, 168
Technique, 97, 246, 248
Temperature, 173, 192, 193, 199, 248
Temperature field, 199
Temperature gradient, 173, 192, 193
Thawing, 233, 236, 238, 240, 246
Thermal, 93, 105, 114, 128, 234, 238,
254, 259
Thermal-hydraulic 93, 105
Thermocapillary, 173, 198, 199, 207
Toroidal, 170
Touching, 187, 188, 193
Touching drop, 187
Tower, 253
Transfer, 116, 147
Transmission, 253
Transition, 72
Two phase, 12, 17, 40, 93
Two-phase flow 1, 3, 7, 93
Typical, 114
Type, 163
U
Upward 21-25, 31, 33
V
Vapor, 222
Variation, 238
Various, 163
Vertical 24, 29, 31

Lixin Cheng Dieter Mewes Advances in Multiphase Flow and Heat Transfer Vol. 1, 2009

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Lixin Cheng Dieter Mewes Advances in Multiphase Flow and Heat Transfer Vol. 1, 2009

Diunggah oleh

Hak Cipta:

Format Tersedia

Advances in Multiphase Flow and Heat Transfer

Lixin Cheng and Dieter Mewes (Editors)

Bentham Science Publishers Ltd.

Research and Review Studies vi

Chapter 1 Advances in Void Fraction, Flow Pattern Maps and Non-Boiling

Chapter 2 Boiling Heat Transfer in Reduced Gravity Environments 53

Chapter 3 Two-Phase Flow Regime Identification Methodologies in

Chapter 4 Thermal Radiation Modeling in Multiphase Flows Typical

Chapter 5 Marangoni Motion of Multiphase Compound Drops 158

Chapter 6 Annular Flow Effects on Pore Formation in High-Intensity

Chapter 7 Ground Heat Transfer from Civil Infrastructure with Seasonal

Chapter 8 Evaluation of Four Nucleate Flow Boiling Models 267

Lixin Cheng and Dieter Mewes (Ed)

Lixin Cheng and Dieter Mewes (Ed)

components allow for a simplified approach, which is implemented in a problem-

Editor-in-Chief: Dr. Lixin Cheng

Co-editor: Prof. Dieter Mewes

Paolo DiMarco, Dipartimento di Energetica, Universit di Pisa, Italy

Leonid A. Dombrovsky, Joint Institute for High Temperatures, Moscow, Russia

X. Duan, University of Manitoba, Canada

Afshin J. Ghajar, School of Mechanical and Aerospace Engineering, Oklahoma State

Takashi Hibiki, School of Nuclear Engineering, Purdue University, USA

Mamoru Ishii, School of Nuclear Engineering, Purdue University, USA

J. Enrique Julia, Departamento de Ingenieria Mecanica y Construccion. Universitat Jaume I,

Jungho Kim, Department of Mechanical Engineering, University of Maryland, USA

J.P. Kroes, Department of Mechanical Engineering, Technische Universiteit Eindhoven,, The

S. C. Kuo, Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-

O.M. Lavrenteva, Department of Chemical Engineering, TECHNION Israel Institute of

G. F. Naterer, University of Ontario Institute of Technology, Canada

A. Nir, Department of Chemical Engineering, TECHNION Israel Institute of Technology,

Haruhiko Ohta, Department of Aeronautics and Astronautics, Kyushu University, Japan

Clement C. Tang, School of Mechanical and Aerospace Engineering, Oklahoma State

C. W. M. van der Geld, Department of Mechanical Engineering, Technische Universiteit

E. van Velthooven, DAF Trucks, CAE-Engines, The Netherlands

P. S. Wei, Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-

Lixin Cheng and Dieter Mewes (Ed)

Research and Review Studies

Lixin Cheng and Dieter Mewes (Ed)

Advances in Void Fraction, Flow Pattern Maps and Non-

Lixin Cheng and Dieter Mewes (Ed)

Void Fraction in Two-Phase Pipe Flow

VSG C0 VSG VSL (uGM ) (1)

Table 1. Summary of experimental database sources, Woldesemayat and Ghajar [10].

Eaton [11] Natural

Beggs [12] 25.4 & 38.1

Spedding & Nguyen [13]

Minami & Brill [15] Air-water & Air-

Franca & Lahey [16]

Among the six void fraction correlations recommended by Woldesemayat and

1065 2137 2427

1597 2139 2363

1082 2059 2395

1244 2003 2322

1643 2084 2304

1369 1953 2294

1718 2234 2436

Comparison of void fraction correlations for vertical pipes

The database encompasses experimental data points for different gas-liquid

Table 4. Summary of experimental database for upward vertical two-phase flow.

Table 6. Qualitative performance of 11 selected correlations in four void fraction ranges.

Nicklin et al. [26] S S S NS

Nishino & Yamazaki [32] NS NS S S

Guzhov et al. [33] S NS S S