An Introduction to

Transport Phenomena in
Materials Engineering
This book elucidates the important role of conduction, convection, and radiation heat transfer,
mass transport in solids and fluids, and internal and external fluid flow in the behavior of
materials processes. These phenomena are critical in materials engineering because of the
connection of transport to the evolution and distribution of microstructural properties during
processing. From making choices in the derivation of fundamental conservation equations, to
using scaling (order-of-magnitude) analysis showing relationships among different phenomena,
to giving examples of how to represent real systems by simple models, the book takes the
reader through the fundamentals of transport phenomena applied to materials processing. Fully
updated, this third edition of a classic textbook offers a significant shift from the previous
editions in the approach to this subject, representing an evolution incorporating the original
ideas and extending them to a more comprehensive approach to the topic.
FEATURES
• Introduces order-of-magnitude (scaling) analysis and uses it to quickly obtain
approximate solutions for complicated problems throughout the book
• Focuses on building models to solve practical problems
• Adds new sections on non-Newtonian flows, turbulence, and measurement of heat
transfer coefficients
• Offers expanded sections on thermal resistance networks, transient heat transfer,
two-phase diffusion mass transfer, and flow in porous media
• Features more homework problems, mostly on the analysis of practical problems, and
new examples from a much broader range of materials classes and processes, including
metals, ceramics, polymers, and electronic materials
• Includes homework problems for the review of the mathematics required for a
course based on this book and connects the theory represented by mathematics with
real-world problems
This book is aimed at advanced engineering undergraduates and students early in their
graduate studies, as well as practicing engineers interested in understanding the behavior
of heat and mass transfer and fluid flow during materials processing. While it is designed
primarily for materials engineering education, it is a good reference for practicing materials
engineers looking for insight into phenomena controlling their processes.
A solutions manual, lecture slides, and figure slides are available for qualifying adopting
professors.
 An Introduction to
Transport Phenomena in
Materials Engineering
Third Edition

David R. Gaskell
Matthew John M. Krane
Designed cover image: Alicia Krane
Third edition published 2024
by CRC Press
2385 NW Executive Center Drive, Suite 320, Boca Raton FL 33431
and by CRC Press
4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
CRC Press is an imprint of Taylor & Francis Group, LLC
© 2024 Taylor & Francis Group, LLC
First edition published by Macmillan Publishing Company 1991
Second edition published by Momentum Press 2012
Reasonable efforts have been made to publish reliable data and information,
but the author and publisher cannot assume responsibility for the validity of
all materials or the consequences of their use. The authors and publishers
have attempted to trace the copyright holders of all material reproduced in this
publication and apologize to copyright holders if permission to publish in this form
has not been obtained. If any copyright material has not been acknowledged please
write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be
reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including
photocopying, microfilming, and recording, or in any information storage or
retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access
www.copyright.com or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are not
available on CCC please contact [email protected]
Trademark notice: Product or corporate names may be trademarks or registered
trademarks and are used only for identification and explanation without intent to
infringe.
Library of Congress Cataloging-in-Publication Data
Names: Gaskell, David R., 1940– author. | Krane, Matthew J. M., author.
Title: An introduction to transport phenomena in materials engineering / David R.
Gaskell and Matthew John M. Krane.
Description: Third edition. | Boca Raton, FL : CRC Press, 2024. | Includes
bibliographical references and index.
Identifiers: LCCN 2023025079 (print) | LCCN 2023025080 (ebook) | ISBN
9780367821074 (hardback) | ISBN 9780367611316 (paperback) | ISBN
9781003104278 (ebook)
Subjects: LCSH: Materials—Fluid dynamics. | Mass transfer. | Heat—
Transmission.
Classification: LCC TA418.5 .G37 2024 (print) | LCC TA418.5 (ebook) | DDC
620.1/1296—dc23/eng/20230729
LC record available at https://lccn.loc.gov/2023025079
LC ebook record available at https://lccn.loc.gov/2023025080
ISBN: 978-0-367-82107-4 (hbk)
ISBN: 978-0-367-61131-6 (pbk)
ISBN: 978-1-003-10427-8 (ebk)
DOI: 10.1201/9781003104278
Typeset in Times
by Apex CoVantage, LLC
Access the Support Material: www.routledge.com/9780367821074
 This work is dedicated to my father,
Robert Joseph Krane, PhD
Eternal rest grant unto him, O Lord,
and may perpetual light shine upon him.
Contents
Preface to the Third Edition .................................................................................. xiii
Authors ...................................................................................................................xvii
Acknowledgments ...................................................................................................xix

Chapter 1 Introduction to Transport Phenomena in Materials

Processing ............................................................................................1
1.1 Transport Phenomena ................................................................ 1
1.2 Examples of Transport Phenomena in Materials
Processing ..................................................................................3
1.2.1 Fluid Flow .................................................................... 3
1.2.2 Heat Transfer ................................................................ 4
1.2.3 Mass Transfer ............................................................... 7
1.2.4 Multimode Transport Phenomena ................................ 8
1.2.5 Nonuniform Distribution of Microstructure ................ 9
1.3 Constitutive Relations for Transport Phenomena .................... 10
1.3.1 Shear in Fluids............................................................ 11
1.3.2 Modes of Heat Transfer .............................................. 22
1.3.3 Mass Transfer ............................................................. 33
1.3.4 Summary of Constitutive Equations .......................... 39
1.4 Finding Solutions to Models of Transport Phenomena ...........40
1.4.1 Mathematical Solutions .............................................. 41
1.4.2 Numerical Solutions ................................................... 43
1.4.3 Scaling Analysis ......................................................... 45
1.4.4 A Note on Selection of Solution Approach ................ 48
1.5 Engineering Units .................................................................... 50
1.6 Summary ................................................................................. 52
References ..........................................................................................53

Chapter 2 Steady State Conduction Heat Transfer.............................................. 55

2.1 Introduction ............................................................................. 55
2.2 Thermal Resistances................................................................ 55
2.2.1 Conduction Resistance ............................................... 55
2.2.2 Convection Resistance................................................ 59
2.2.3 Radiation Resistance ..................................................60
2.2.4 Interface Resistance ...................................................60
2.3 Resistance Networks................................................................ 63
2.4 General Heat Conduction Equation ......................................... 67
2.5 Heat Transfer Boundary Conditions ........................................ 71
2.6 One-Dimensional Heat Conduction ........................................ 73
2.7 Conduction with Heat Generation ........................................... 85

vii
viii Contents

2.8 Multidimensional Conduction ................................................. 91

2.8.1 Scaling Analysis ......................................................... 91
2.8.2 Two-Dimensional Paths in Resistance Networks:
Branches and Shape Factors .......................................97
2.8.3 Exact Solutions ......................................................... 105
2.9 Summary ............................................................................... 109
2.10 Homework Problems ............................................................. 109
References ........................................................................................118

Chapter 3 Transient Conduction Heat Transfer................................................. 120

3.1 Introduction ........................................................................... 120
3.2 Scaling Analysis of General Transient Conduction .............. 120
3.3 Lumped Capacitance Analysis: Convection Resistance
Dominated (Bi << 1) .............................................................. 122
3.3.1 Convection Heat Loss............................................... 122
3.3.2 Radiation Heat Loss ................................................. 126
3.4 Spatial Dependence: Conduction Resistance Dominated
(Bi >> 1) ................................................................................. 129
3.4.1 Cooling in a Slab: Early Times for Bi >> 1 .............. 130
3.4.2 Cooling in a Slab: Late Times for Bi >> 1 ............... 137
3.4.3 Heating in a Radial System ...................................... 144
3.5 Spatial Dependence: The General Solution with a Balance
of Conduction and Convection Resistances (Bi ~ 1) ................ 150
3.5.1 Heating in a Slab: Early Times for Bi ~ 1 ................ 150
3.5.2 Heating in a Slab: Late Times for Bi ~ 1 .................. 153
3.6 Solidification.......................................................................... 160
3.6.1 Energy Balances with Phase Change ....................... 161
3.6.2 Solidification of a Pure Substance............................ 165
3.7 Summary ............................................................................... 174
3.8 Homework Problems ............................................................. 174
References ........................................................................................185

Chapter 4 Mass Diffusion in the Solid State .................................................... 187

4.1 Introduction ........................................................................... 187
4.2 Steady State Mass Diffusion ................................................. 187
4.3 Fick’s Second Law of Diffusion: Transient Diffusion ........... 188
4.4 Infinite Diffusion Couple....................................................... 195
4.5 Diffusion Involving Solid-Solid Phase Change..................... 197
4.6 Diffusion in Substitutional Solid Solutions ...........................208
4.7 Darken’s Analysis ..................................................................209
4.8 Self-Diffusion Coefficient ..................................................... 213
4.9 Measurement of the Interdiffusion Coefficient:
Boltzmann–Matano Analysis ................................................ 216
4.10 Influence of Temperature on the Diffusion Coefficient......... 221
Contents ix

4.11 Summary ............................................................................... 225

4.12 Homework Problems ............................................................. 227
References ........................................................................................ 231

Chapter 5 Fluid Statics ...................................................................................... 232

5.1 Introduction ........................................................................... 232
5.2 Concept of Pressure ............................................................... 232
5.2.1 Pressure at a Point and in a Column ........................ 232
5.2.2 Atmospheric Pressure............................................... 235
5.3 Measurement of Pressure ...................................................... 237
5.4 Pressure in Incompressible Fluids ......................................... 241
5.5 Buoyancy ...............................................................................244
5.6 Summary ...............................................................................246
5.7 Homework Problems ............................................................. 247
Reference .......................................................................................... 250

Chapter 6 Mechanical Energy Balance in Fluid Flow ...................................... 251

6.1 Introduction ........................................................................... 251
6.2 Laminar and Turbulent Flows ............................................... 251
6.3 Bernoulli’s Equation .............................................................. 253
6.4 Friction Losses....................................................................... 255
6.5 Influence of Bends, Fittings, and Changes in
Pipe Radius ............................................................................ 261
6.6 Steady-State Applications of the Modified
Bernoulli Equation ................................................................ 263
6.7 Concept of Hydrostatic Head................................................. 269
6.8 Fluid Flow in an Open Channel ............................................ 270
6.9 Transient Applications of the Modified
Bernoulli Equation ................................................................ 273
6.10 Summary ............................................................................... 277
6.11 Homework Problems ............................................................. 278
References ........................................................................................ 283

Chapter 7 Equations of Fluid Motion ............................................................... 285

7.1 Introduction ........................................................................... 285
7.2 Conservation of Mass ............................................................ 285
7.3 Momentum Balance: The Navier-Stokes Equations.............. 291
7.4 Boundary Conditions for Fluid Flow .................................... 296
7.5 Characteristics of Pressure-Driven Flow Behavior in
a Channel ............................................................................... 299
7.6 Summary ............................................................................... 303
7.7 Homework Problems ............................................................. 303
References ........................................................................................306
x Contents

Chapter 8 Internal Flows...................................................................................307

8.1 Introduction ...........................................................................307
8.2 Simplifications of Equations of Motion for Internal Flows ...307
8.3 Shear-Driven Flow between Flat Parallel Plates ...................309
8.4 Pressure-Driven Flow between Flat Parallel Plates .............. 311
8.5 Fluid Flow in a Vertical Cylindrical Tube............................. 319
8.6 Capillary Flowmeter.............................................................. 325
8.7 Non-Newtonian Internal Flows ............................................. 328
8.7.1 Shear-Driven Flow of a Power-Law Fluid ................ 328
8.7.2 Pressure-Driven Flow of a Power-Law Fluid ........... 329
8.7.3 Pressure-Driven Flow of a Bingham Plastic ............ 334
8.8 Flow through Porous Media .................................................. 336
8.8.1 Resistance to Flow.................................................... 336
8.8.2 Effect of Porous Media Structure on Flow .............. 342
8.9 Fluidized Beds ....................................................................... 349
8.10 Summary ............................................................................... 354
8.11 Homework Problems ............................................................. 354
References ........................................................................................ 359

Chapter 9 External Flows..................................................................................360

9.1 Introduction ...........................................................................360
9.2 Fully Developed Flow Down an Inclined Plane....................360
9.3 Flow over a Horizontal Flat Plane ......................................... 363
9.4 Momentum Integral Solution for Boundary Layer on a
Horizontal Flat Plate.............................................................. 367
9.4.1 Entry Length at Entrance to a Pipe .......................... 373
9.5 Turbulent Flow....................................................................... 375
9.5.1 Characteristics of Turbulent Flows........................... 376
9.5.2 Transition and Turbulent Flow over a Flat Plate ...... 383
9.6 Flow Past Submerged Bluff Objects ...................................... 390
9.7 Summary ............................................................................... 398
9.8 Homework Problems ............................................................. 399
References ........................................................................................ 403

Chapter 10 Convection Heat Transfer .................................................................405

10.1 Introduction ...........................................................................405
10.2 General Energy Equation with Advection and Diffusion .....408
10.3 Advection in Rigid, Moving Media ....................................... 413
10.4 External Forced Convection .................................................. 420
10.4.1 Forced Convection from a Horizontal Flat Plate ..... 420
10.4.2 Forced Convection Correlations in Other
Geometries ............................................................... 439
10.5 Internal Forced Convection ................................................... 451
Contents xi

10.6 Natural Convection Heat Transfer .........................................460

10.6.1 Natural Convection from an Isothermal Vertical
Flat Plate................................................................... 462
10.6.2 Natural Convection from Other Geometries ............ 467
10.7 Boiling Heat Transfer ............................................................469
10.8 Summary ............................................................................... 474
10.9 Homework Problems ............................................................. 475
References ........................................................................................481

Chapter 11 Mass Transfer in Fluids ....................................................................484

11.1 Introduction ...........................................................................484
11.2 Mass and Molar Fluxes in a Fluid .........................................484
11.3 Equations of Diffusion with Advection in a Binary
Mixture A-B ........................................................................... 486
11.4 Equimolar Counterdiffusion.................................................. 489
11.5 One-Dimensional Steady-State Diffusion of Gas A
through Stationary Gas B ...................................................... 490
11.6 Sublimation of a Sphere into a Stationary Gas...................... 496
11.7 Film Model ............................................................................ 498
11.8 Catalytic Surface Reactions ..................................................500
11.9 Diffusion and Chemical Reaction in Stagnant Film ............. 502
11.10 Mass Transfer at Large Fluxes and Large Concentrations ....506
11.11 Influence of Mass Transport on Heat Transfer in
Stagnant Film ........................................................................509
11.12 Mass Transfer Coefficient for Concentration Boundary
Layer on a Flat Plate .............................................................. 512
11.13 Simultaneous Heat and Mass Transfer: Evaporative
Cooling.................................................................................. 517
11.14 Summary ............................................................................... 520
11.15 Homework Problems ............................................................. 520

Chapter 12 Radiation Heat Transfer ................................................................... 523

12.1 Introduction ........................................................................... 523
12.2 Intensity and Emissive Power ................................................ 524
12.2.1 Emissive Flux ........................................................... 526
12.2.2 Irradiation ................................................................. 527
12.2.3 Radiosity................................................................... 528
12.3 Blackbody Radiation ............................................................. 528
12.4 Surface Properties ................................................................. 531
12.5 Kirchhoff’s Law and the Hohlraum ...................................... 537
12.6 Radiation Exchange in an Enclosure: View Factors ..............540
12.7 Radiation Exchange among Blackbodies .............................. 551
12.8 Radiation Exchange among Diffuse-Gray Surfaces ............. 554
12.9 Notes on the Electrical Analogy ........................................... 559
xii Contents

12.10 Radiation Shields ................................................................... 562

12.11 Reradiating Surfaces .............................................................564
12.12 Summary ............................................................................... 568
12.13 Homework Problems ............................................................. 569
References ........................................................................................ 574

Appendix I Math Practice for Transport Phenomena Course ........................ 575

Appendix II Equations of Motion and Thermal Energy Balance ................... 579
Appendix III Unit Conversions........................................................................... 581
Appendix IV Selected Thermophysical Properties .......................................... 583
Index ...................................................................................................................... 591
Preface to the Third Edition
APPROACH
This new edition represents a significant shift from the previous version in the
approach to this subject, which should not be unexpected with the addition of a new
author. The shift is not a discontinuous one but represents an evolution over the last
15 years, incorporating the original ideas and extending them to a more comprehen-
sive approach to the topic.
In his original preface, Prof. Gaskell, who passed away in 2013, posited the need
for materials engineers to study transport phenomena and, given the constraints of
the materials science and engineering (MSE) curriculum, to do so in one semester.
Heat and mass transfer and fluid flow in materials processing control microstructural
development and the associated distribution of properties in engineered products and
so understanding transport phenomena is absolutely critical to the understanding and
design of those processes and products. To support this activity, Prof. Gaskell wrote
this text and developed his third-year course at Purdue University. Regarding the
focus of his text, he wrote in his first-edition preface that

a careful balance must be made between an explanation of the fundamentals

that govern the dynamics of fluid flow and the transport of heat and mass, on
the one hand, and, on the other, illustration of the application of the fundamen-
tals to specific systems of interest in materials engineering.

His response to this challenge for an introductory class was to focus the first two
editions of the text on the development of the governing equations and the small set
of exact solutions. It is my view that he correctly eschewed a more handbook-style
approach that tends to center on a catalog of processes, with ad hoc descriptions of
associated transport phenomena included as needed, in favor of a more strictly fun-
damental approach, inspired by Bird, Stewart, and Lightfoot’s Transport Phenomena
(1960). The understanding of the origins and solutions of the governing equations is
vital to their application to materials processing, either in closed form or by numer-
ical analysis.
However, I believe that being restricted to this approach gives necessary, but not
sufficient, coverage of the subject in response to the needs of MSE students and prac-
ticing materials engineers, and that it can be improved in two ways. First, the previ-
ous approach gives little explicit motivation or demonstration for why this subject is
in the MSE curriculum. Students who have used this text at Purdue and elsewhere
have repeatedly made this point, so a case must be made for the allocation of scarce
time in the curriculum to this subject instead of others. New examples of processing
in the text and end-of-chapter problems are included to show the applicability of this
topic and help to justify its place in the plan of study. As part of this effort, exam-
ples are given from across the broad range of materials, including the processing of
metals, ceramics, polymers, and electronic materials. The second aspect addressed
in this edition is the need for more substantial discussion of the physical basis and

xiii
xiv Preface to the Third Edition

meaning of the derived balance equations, which are the bedrock of the previous
editions, and how to apply them to practical problems. These equations should be
viewed as balances of physical phenomena, and my teaching approach in many cases
has been to begin with scaling analysis to determine the order of magnitude of these
phenomena and their interrelationships before attempting mathematical or numer-
ical solutions. Further exploration is by approximating complex reality to one or
more simple problems, which can be solved and from which solutions we can draw
conclusions about the system’s behavior. This simple approach focuses the student
on fundamental physics and provides the rationale for model simplification during
application to practical problems. It helps the student connect the theory represented
by the mathematics to the real world. In fact, in industrial settings, the scaling of
a practical problem, perhaps followed by the approximate solution of a simplified
model, frequently provides sufficient information to make an engineering decision.
When not sufficient, the preliminary scaling and approximate analysis are still useful
to understand the problem, guide development of a numerical model, and to interpret
its results. I strongly believe that including practical examples and the practice of
problem-solving methods adds depth to the framework provided by Prof. Gaskell in
the previous editions and brings it more in line with a more materials processing–
focused teaching approach. Many former students who took the course on which this
text is based have reported the utility of this approach in exploring real situations in
an industrial context.
USE OF THIS TEXT
This text is aimed at junior/senior-level engineering undergraduates and students
early in their graduate studies, as well as practicing engineers interested in under-
standing the behavior of heat and mass transfer and fluid flow during materials
processing. It assumes familiarity with calculus, basic differential equations, and
introductory thermodynamics and mechanics. An introductory course in materials
science covering phase diagrams, time-temperature-transformation diagrams, and
basic microstructural forms is very useful for understanding many applications.
The course is a useful prerequisite to any materials processing lecture or lab course.
The text is designed primarily for materials engineering education but will also be a
good reference for anyone desiring a one-semester treatment of introductory fluids
and heat transfer or for the practicing materials engineer looking for insight into the
transport phenomena controlling her process. In all cases, the main purpose is to
understand the basic physics of transport phenomena and how to model and apply
them to materials processing.
Looking through many syllabi for courses at many universities that could use this
text, I discovered many different orders and choices of topics. Looking through a
variety of options, certain general restrictions became apparent:

• fluid mechanics should be before convection,

• conduction should be before convection heat transfer,
• mass diffusion should be before convection mass transfer, and
• convection heat transfer should be before boiling.
Preface to the Third Edition xv

One difficulty with any text is convincing students it is worth their time to read it.
To incentivize reading the book, I recently added daily reading quizzes to the course.
These short evaluations, usually just a few multiple choice or true/false questions, are
taken online and graded by the course management software used at Purdue. They
have been an effective tool to encourage reading before lecture and have improved
class participation and quiz scores.
As a last note on teaching, I will address the evaluation of the students in this
course, for which I have tried two different methods for testing. The first is to require
weekly homework assignments, three hour-long midterms, and a two-hour final
exam. This practice is common and does have the advantage of giving students direct
incentives to work problems, which is the best way to learn the material. However,
there are some difficulties. Many students tend to rely on others for their homework
and so do not pay serious attention to the class for the five weeks between exams, but
the material is too difficult to cram for exams three or four times a semester. In recent
years, we have still assigned homework problems, but we did not collect the work.
While it seems counterintuitive, the students have proven to be more likely to do their
own work because of another structural change: we replaced the few midterms with
biweekly, 25-minute quizzes. Having to perform as individuals on a much more fre-
quent basis seems to keep the students’ attention, and their overall performance has
improved (with no significant change in the types or total number of test questions).
Authors
David R. Gaskell (1940–2013) was Professor of Materials Engineering at Purdue
University from 1982 to 2013. Dr. Gaskell was born in Glasgow, Scotland, and at-
tended the Royal College of Science and Technology, receiving First Class Honors
in metallurgy and technical chemistry for a BSc in 1962. He moved to Hamilton,
Canada, to pursue graduate studies at McMaster University and then immigrated
to the United States, teaching first at the University of Pennsylvania and then at
Purdue. During his career, he served as a visiting professor at the NRC Atlantic Re-
gional Laboratory, Canada, and at the G. C. Williams Cooperative Research Centre
for Extraction Metallurgy in the Department of Chemical Engineering, University
of Melbourne, Australia. Professor Gaskell was dedicated to teaching and was the
recipient of the Reinhardt Schumann Jr. Best Undergraduate Teacher Award in Mate-
rials Engineering several times over.
Matthew John M. Krane (1964– ) is Professor of Materials Engineering at Purdue
University and a member of the Purdue Center for Metal Casting Research and the
Purdue Heat Treatment Consortium. Using modeling and experiments, his research
focuses on the connection between macroscopic transport phenomena and defect
formation during materials processes, particularly the study of the solidification of
metal alloys. Professor Krane has been with Purdue’s School of Materials Engineer-
ing since 1996, but his education is in mechanical engineering (Cornell, BS, 1986;
Pennsylvania, MS, 1989; Purdue, PhD, 1996), with a concentration in heat transfer
and fluid flow. He has been a visiting researcher at the University of Birmingham
(UK), the University of Greenwich (UK), and the Université de Lorraine (Nancy,
France). In addition to consulting, research programs, and undergraduate projects
with the metals processing industry during his time at Purdue, he worked in indus-
try for three years on thermal issues in the design and manufacturing of electronic
packaging.

xvii
Acknowledgments
This edition was produced with the help of many people. First and foremost, my wife
Kathy has been a great friend and support throughout my adult life. Ever since we
met as freshmen at Cornell, she has been interested in my work and has always given
good counsel on matters personal and professional. My life would have been much
less interesting and happy without her in it. The Four Horsemen of the Apocalypse
(our sons: Stash, Pat, Mick, and Bob) and our daughter, Mary Kate, have been a
real joy in our lives and have been very supportive in this endeavor. My father and
mother, Robert and Antoinette Krane, sacrificed much to provide all their children
with educational opportunities.
My father was my most influential teacher in technical and professional matters
through countless discussions over a few decades. Two science teachers in high
school, Gil Luttrell and George Wilson, took a special interest in developing my
scientific curiosity and helped turn my professional aspirations toward the technical;
I probably would not be an engineer now without them. At Cornell, Sidney Leibo-
vich, C. Thomas Avedesian, and Bart Conta gave me my first formal introductions
to fluid and thermal sciences, inspiring a real love for these subjects and providing
models on how to teach them. My thesis advisor at Penn, Benjamin Gebhart, one of
the pioneers in buoyancy-induced flows, introduced me to that very interesting field.
The very different teaching, research, and advising styles of Frank Incropera, Ray-
mond Viskanta, David DeWitt, and Satish Ramadhyani, all professors in the Purdue
Heat Transfer Laboratory during my doctoral work, were instrumental in forming
my thinking on how to study, apply, and teach the thermal sciences. Many other col-
leagues over the years helped form my ideas on practical application and the teach-
ing of transport phenomena, including Greg Gallagher (formerly Digital Equipment
Corporation), Chris Vreeman, Jay Rozzi, Frank Pfefferkorn, and Marcus Bianchi
(grad students with me in Purdue’s School of Mechanical Engineering) and Adam C.
Powell (now of Worcester Polytechnic Institute). Jon Dantzig, professor emeritus at
UIUC, has provided much encouragement and sound advice on how to write about,
teach, and model transport phenomena.
I owe a great debt to my colleagues in the School of Materials Engineering at
Purdue, who took in a young mechanical engineer ignorant of materials and turned
him into a passable metallurgist. Prominent in this effort were David Johnson, Keith
Bowman, and Mysore Dayananda. Some of my polymers colleagues (Kendra Erk,
Chelsea Davis, and John Howarter) helped with the inclusion of more soft materials
topics in the book. Special mention goes to Kevin Trumble, who is a passionate
evangelist for the field of materials engineering and has always been a generous
mentor and guide through my second field of study. His enthusiasm for materials
engineering is infectious, and much of my success over the years is due directly to
his knowledge, advice, and encouragement.
I thank Allison Shatkin and Hannah Warfel, my editor and editorial assistant at
CRC Press/Taylor & Francis, for help producing this book and the Gaskell family
for giving me the opportunity to work on it. I hope this new edition lives up to their

xix
xx Acknowledgments

expectations. My wife provided great practical assistance, converting old, unedita-

ble files into readable Word documents and finding obscure pieces of freeware. My
daughter retyped equations from previous editions, and my son Stash drew most of
the figures. The cover art was designed and drawn by my daughter-in-law, Alicia
Krane. The head of my school, David Bahr, has encouraged this project from the
first and supported it along the way. Several colleagues, including Kyle Fezi, Chelsea
Davis, Michael J. Krane, Michael H. Krane, Patrick Krane, Michael Titus, Kevin
Trumble, and Igor Vušanovi , read different pieces of this text and made helpful
suggestions. I greatly appreciate all their comments and enjoyed the (occasionally
heated) discussions they sparked. Any remaining mistakes, awkward passages, and
ineffective pedagogy are my responsibility.
Finally, I would like to acknowledge the help the late David R. Gaskell gave me
in my career. We first met when I asked him to be a member of my doctoral advisory
committee in 1991. The topic was transport phenomena in alloy solidification, and
he agreed to serve as long as I would “do it properly and use a metal” (instead of
a salt-water analogue). I originally signed on to that project because it was rich in
interesting heat transfer and fluid mechanics, but Professor Gaskell from the begin-
ning showed me that it was much more and that transport phenomena are necessary
to understand the property-processing-structure relationships at the heart of materials
engineering. That conversation was the beginning of my journey into materials engi-
neering, a journey he facilitated in many ways over the next two decades.

Matthew John M. Krane

Professor of Materials Engineering
Purdue University
West Lafayette, Indiana, USA
on The Feast of St. Athanasius I of Alexandria
May 2, 2023
 1 Introduction to Transport
Phenomena in Materials
Processing

1.1 TRANSPORT PHENOMENA

In any effort, it is wise to start by clearly stating the purpose of the endeavor. Our goal
in writing this textbook is to help future and practicing engineers better understand
and predict the behavior of mass, momentum, heat, and species movement, espe-
cially during materials processing. The study of the movement of these quantities is
the field of transport phenomena.
Before progressing with this topic, we should address an important question
regarding transport phenomena: why should materials engineers find time to study
this topic, either during a university degree program or on their own? There are many
interesting and useful subjects to learn and a very finite amount of time available, so
a case must be made, in a cramped curriculum or in a busy professional life, for the
allocation of scarce time to this subject instead of many other interesting alternatives.
While many engineers are concerned with transport phenomena for their own
sake, the materials engineer heats, cools, pours, pumps, and diffuses materials in
the pursuit of beneficial materials properties, to make the material economical to
produce in the first place. (An example is age-hardened aluminum alloys, which
make the aerospace industry possible and would not have the required mechani-
cal properties without heating- and cooling-induced and mass diffusion–controlled
solid-solid phase transformations [1]. Other examples are described in the next sec-
tion.) The transport phenomena (along with the phase equilibria in the material) are
the mechanisms that produce many material properties throughout processing by
the control of microstructure development. This interrelationship is illustrated by the
oft-cited “MSE triangle” shown in Figure 1.1. The three vertices of this triangle are
interdependent, and understanding the physical mechanisms of transport phenomena
is necessary to understanding many of these relationships.
The primary approach we will use here to improve our understanding and our
predictions is the making of models of the transport phenomena. These models are
mathematical representations of reality, not reality itself. In an excellent descrip-
tion of model-making methodology in materials science and engineering, Ashby
[2] wrote:

A model is an idealization . . . (it) unashamedly distort(s) the inessentials in order to

capture the features that really matter. . . . At best, it captures the essential physics of the
problem, it illuminates the principles that underline the key observations, and it predicts
behavior under conditions which have not yet been studied.

DOI: 10.1201/9781003104278-1 1
2 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.1 The “MSE triangle,” a representation of the primary, interconnected aspects
of materials science and engineering.

FIGURE 1.2 Balance of generic quantity transport in control volume.

In this text, we will follow Ashby’s lead with the goal of understanding and esti-
mating the behavior of transport phenomena in materials processing.
The simplest way to arrange our thinking about many models of transport phe-
nomena is in terms of a control volume, a fixed portion of space through and in which
we observe the motion of some quantity. Figure 1.2 shows a schematic of transport in
a control volume, which leads to a very general balance:

quantity in quantity out + quantity generated = quantity stored (1.1)

Mass, momentum, heat, and species each have their own possible mechanisms
of transport across a control volume boundary and generation (or destruction) in the
volume, and these mechanisms and how to model them are the subject of much of
this text. The sum of these three phenomena is not necessarily zero, but is the amount
accumulated in (or depleted from) the control volume. For example, momentum can
be added to a control volume through pressure gradients, buoyancy, or friction, all of
which can drive a fluid’s acceleration (storing momentum) or transport through the
volume. Heat might be generated by a chemical reaction and transported across the
boundary by thermal radiation. Most of the models of transport here will be derived
starting from the simple balance in Eq. (1.1).
Introduction to Transport Phenomena in Materials Processing 3

1.2 EXAMPLES OF TRANSPORT PHENOMENA

IN MATERIALS PROCESSING
The study of transport phenomena is sometimes divided into three areas: fluid flow,
heat transfer, and mass transfer. To drive home the importance of transport phenom-
ena in materials processing, we show in this section some examples of industrial
processes in which these phenomena have a significant effect on their behavior.

1.2.1 Fluid Flow

The shot peening process work hardens a metal surface by high-speed bombardment
of that surface by small steel shot. One method of the production of that shot is
sketched in Figure 1.3. Steel is melted in one furnace and then poured into the hold-
ing furnace in the figure, which in turn is drained in a controlled manner. The jet of
liquid steel pouring from the holding furnace is intercepted by a high-speed water
spray. The interaction of these two jets fractures the steel stream, atomizing it into a
cloud of quickly frozen metal droplets. The size distribution of steel shot is a strong
function of the water and steel stream velocities and the angle at which they interact
[3]. Controlling the water spray velocity is done through the design of a piping sys-
tem, taking into account frictional losses in the circuit and the spray nozzle and the
behavior of the pump. That flow is typically steady, but the exit jet velocity of
the steel and the angle at which it enters the water spray change substantially during
the draining, as they are strong functions of the height of the fluid above the nozzle.
In Chapter 6, methods for characterizing the steel jet behavior are explored. The
physics of the liquid metal’s atomization are beyond the scope of this text; interested
readers are referred to monographs on that topic [3, 4].

FIGURE 1.3 Atomizing a steel stream draining from a holding furnace. The water jet frac-
tures the steel into a range of small droplet sizes.
4 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.4 Filling a cavity with polymer during injection molding.

One of the cheapest and fastest methods for making plastic parts is to form a
polymer into the desired shape by injection molding. This process uses pressure
to force a liquid polymer into the mold cavity, where it is quickly solidified and
ejected. In Figure 1.4, we see the melt filling the mold. The process finishes when
the mold is filled and perhaps some polymer has pushed into an air vent (which pre-
vents air from building up a back pressure in the last-filled region of the mold). The
front is propelled through the mold by the upstream inlet pressure and is resisted
by the friction at the mold wall, the latter increasing as the filled region lengthens.
For a constant fill rate, the pressure must therefore increase with time; alterna-
tively, the flow rate can be allowed to slow while maintaining a constant pressure,
although this option will slow the process and increase cycle time. Because the
mold is colder than the injected melt, the polymer loses heat, increasing the melt
viscosity and resistance to flow, and must fill the mold before that resistance grows
enough to prevent complete filling. Models of this process in simple geometries,
including the dependence of resistance to flow on the fluid deformation rate, are
found in Chapter 8.

1.2.2 Heat transFer

One example of using thermal transport phenomena to induce desirable materials
behavior is the glass tempering process, Figure 1.5 [5]. Here, air is blown on the
top and bottom of the moving glass sheet starting at a temperature around 640 oC
[6]. This convection heat transfer (see Section 1.3.2 and Chapter 10) extracts heat
from the glass in a way similar to how one cools hot coffee by blowing on it. The
glass cools quickly at the top and bottom surfaces and, inside the solid sheet, heat
moves from the hotter interior toward the edges much more slowly by conduction
(see Section 1.3.2 and Chapters 2 and 3). Initially, the surface contracts and becomes
stronger due to the cooling, while the centerline is still relatively hot and formable.
Later, the centerline region finally cools, but that volume’s shrinkage is constrained
in the in-plane directions by the stronger, colder outer layers. This constraint causes
the centerline to be in tension and the surfaces in compression. With a careful con-
trol of the heating temperature and the convection process, the nonuniform residual
stress field and surface compressive stresses can be tailored to improve the maximum
fracture strength of the glass.
Introduction to Transport Phenomena in Materials Processing 5

The continuous casting of steel involves the introduction of liquid metal through
a water-chilled copper mold (the primary cooling region), in which a strand is solid-
ified to a small depth from its surface by the time it exits the bottom of the mold [7].
There the steel moves down from the mold and the strand’s direction is altered from
vertical to horizontal by guide rolls (Figure 1.6). In this secondary cooling section,
solidification is completed by heat extraction by forced convection to the water jets
impinging on the steel surface and by conduction losses to the rolls. The convection

FIGURE 1.5 Side view of tempering process, with glass moved by rollers and cooled by
air jets.

FIGURE 1.6 Schematic of steel continuous casting after the mold as the solidification of the
strand is completed (not to scale).
6 An Introduction to Transport Phenomena in Materials Engineering

is similar to that in glass tempering, but is much more effective because of boil-
ing of the water on the very hot surface (see Chapter 10). The liquid-vapor phase
change significantly enhances the heat transfer rate from the steel, required due to
the high speed of the strand (~1 m/s); gas or single-phase liquid convection would
be insufficient. The heat transfer from the steel to boil the water drives the other
phase change in the system, the steel solidification (see Chapter 3). The thermal
energy removed in only a few seconds is supplied by the drop in steel temperature
and by the liquid-solid phase change. The secondary cooling finishes solidification
but leaves a steep temperature gradient through the strand’s thickness; the leveling of
that gradient and/or the further cooling of the steel is due to many other downstream
convection and radiation processes.
A third heat transfer example is found in Figure 1.7, which shows a green, or
unfired, ceramic part in a furnace. A ceramic green body is made of small particles,
formed while wet into a desired shape and then dried, producing a friable (easily
crumbled) solid. Firing the body sinters the structure, growing the connections among
the powder particles, increasing part density, and perhaps initiating solid-solid phase
changes. One way to provide the heat for firing is to expose the body to high temper-
ature–resistant heating elements or to a gas flame. In either case, heat is transferred
from the source to the part by radiation (see Section 1.3.2 and Chapter 12), which is
an electromagnetic phenomenon with wavelengths between ~0.1 μm and ~100 μm
(depending on the source temperature). At these high temperatures, radiation is the
dominant mechanism of heat addition to the part, but conduction controls how fast

FIGURE 1.7 A ceramic part being fired in a furnace with high-temperature interior walls.
Introduction to Transport Phenomena in Materials Processing 7

FIGURE 1.8 Dopant diffusing into unprotected Si, with dashed lines indicating constant
concentration profiles. Most diffusion is normal to surface, but some dopant travels laterally,
undercutting the SiO2 layers.

the body distributes the heat internally. Ceramics tend not to conduct heat quickly, so
large temperature gradients may develop. These gradients cause a spatial distribution
of thermal expansion, leading to high internal stresses and possible cracking.

1.2.3 Mass transFer

One of the many steps in the production of integrated circuits is the doping of
exposed silicon to form n-type junctions (e.g., with phosphorous) or p-type junctions
(e.g., with boron) [8]. The dopant is supplied to the surface either as a solid or a gas
and moves into the silicon layer by means of mass diffusion (see Section 1.3.3 and
Chapter 4), with the maximum concentration on the surface. The dopant concen-
tration falls exponentially deeper into the substrate, which initially has no dopant.
Figure 1.8 shows such a Si layer, partially masked by silicon oxide. The dopant is
applied to the unmasked regions and most diffuses normal to the surface. At the
edges of the masking, the dopant diffusion becomes two dimensional, as it diffuses
laterally underneath the SiO2. The junction formed by the dopant addition is some-
what wider than the exposed Si region. The spatial distribution of the dopant in the
silicon is a function of the solid-state diffusion process, and control of these process
outcomes is required for the reliable manufacture of integrated circuits.
Mass may also be transferred by convection mass transfer, where it is carried
through a vessel by a moving fluid. The chemical vapor deposition (CVD) process
deposits thin solid films by encouraging a chemical reaction on a substrate in contact
with various reactants mixed in a carrier gas. The flowing gas carries the reactants
through a reactor vessel to the substrate and transports the products (and unused
reactants) away from it. Ideally, the reaction will take place uniformly over the sub-
strate and produce a uniform coating thickness [5]. In the case shown in Figure 1.9,
in which the reaction occurs very rapidly compared to the rate of mass transfer to
and from the surface, the difficulty is to configure the flow field and substrate so that
the desired layer uniformity is achieved. One example of CVD is the deposition of a
silicon nitride layer on silicon wafers by the reaction of dichlorosilane and ammonia
at the heated substrate surface,
8 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.9 Chemical vapor deposition vessel, where a solid reaction product coats the
wafers.

with a reaction temperature around 750 oC [8]. The wafers are kept at the reaction
temperature while the vessel walls are heated to levels outside the range in which the
reaction occurs (so the reaction is only on the wafers), but always high enough so
that there is no gas condensation. The solid coating, Si3N4, can be applied to mask
areas of a wafer, preventing oxidation of the underlying silicon layer during process-
ing and acting as a resistance to diffusion from the atmosphere to the silicon. The
effectiveness of the coating is partially a function of its uniformity over the wafers,
which is a strong function of how the carrier gas and reactants are moved from the
gas stream to the surface.

1.2.4 MultiMode transport pHenoMena

Almost every example of a process (including the ones mentioned earlier) has
more than one significant phenomenon, and usually they interact as they influ-
ence the development of microstructure. A classic example of the role of multiple
transport phenomena in processing, structure, and properties is the quenching
of plain carbon steel, a topic usually discussed in an introductory materials
engineering course [e.g., 9]. Beginning with a carbon-bearing austenite phase
(a face-centered cubic crystal structure) at an elevated temperature, the rate of
cooling to room temperature controls the developing microstructure. If heat is
extracted slowly, two new, lower-temperature, equilibrium phases are formed as
the carbon diffuses in the austenite from regions where ferrite (BCC) forms to
regions where cementite (Fe3C) forms. Chemical equilibrium predicts the for-
mation of these two phases, given enough time for the carbon to move from
one phase to the other. Rapid heat extraction will produce a cooling rate fast
enough that the carbon cannot diffuse quickly enough to avoid being trapped in
a new body-centered tetragonal phase (similar to a BCC structure, but elongated
slightly by the trapped carbon atoms). In this case without long-range diffusion
of carbon, only the BCT phase (distorted BCC) forms from FCC. This nonequi-
librium phase, martensite, is much stronger than the ferrite-carbide structure and
much less tough. This example illustrates the mechanism by which the relative
rates of two transport phenomena (diffusion of heat and carbon) contribute to
form different microstructures and properties.
Introduction to Transport Phenomena in Materials Processing 9

1.2.5 nonuniForM distribution oF Microstructure

In almost all real processes, temperatures, velocities, and/or concentrations are non-
uniform and so affect microstructural development differently at different locations
in a part. A simple example is during the cooling of a polyethylene oxide melt, in
which this biocompatible polymer begins above its melt temperature (TM = 70 oC)
and is quickly quenched to below TM. In Figure 1.10, the temperature field in such
a quenched sample was not uniform. The upper left corner of the image cooled
quickly enough that there was no time for crystallization to begin, while the lower
right corner, farther from the chill, was hotter for long enough to nucleate and grow
spherulites. This disparity in cooling rates over the sample caused the nonuniform
microstructure. The notion of a distribution of microstructures in a material is cen-
tral to most real processes. As will be seen throughout this text, the behavior of
transport phenomena is rarely uniform and so neither are the consequent micro-
structures and properties.
Another example of how transport phenomena affect distribution of microstruc-
ture is in Figure 1.11, which shows the solidification structure of a 2 m tall stainless

FIGURE 1.10 PeO microstructure resulting from a nonuniform quench from a melt. The
upper left of the microstructure cooled quickly enough that no crystallization occurred,
while the slow cooling of the right side allowed time for the nucleation and growth of
spherulites.
Source: (Image courtesy of John Howarter.)
10 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.11 Microstructure distribution in a section of a 2 m tall 310 steel ingot, cast
in a permanent mold, with a nonuniform microstructure. Specimen on display in School of
Materials Engineering, Armstrong Hall, Purdue University.
Source: (Image by Mathew John M. Krane.)

steel ingot solidified in a cast iron mold. Here the effects of fluid flow are seen
in the columnar grain orientation in the outer regions being slightly up from the
horizontal. A transition from a columnar to an equiaxed grain structure nearer the
ingot center is caused by a decreasing temperature gradient at the solidification
front. Large cavities near the top are due to metal motion caused by solidification
shrinkage. Many details of the transport behavior that influence these structure dis-
tributions over a wide range of length scales are beyond the scope of this text, but
are ably described in [10].

1.3 CONSTITUTIVE RELATIONS FOR TRANSPORT PHENOMENA

The behavior of transport phenomena is governed by balance equations of the form
of Eq. (1.1), and the terms in those balances include certain constitutive relations dis-
covered through experimentation. A constitutive relation is a model of the response
of a material to some stimulus. A familiar example from classical mechanics is
Hooke’s law,
Introduction to Transport Phenomena in Materials Processing 11

where stress ( ) and strain ( ) are related by the material stiffness (E), a resistance to
elastic deformation. The definition of that material property is Hooke’s law, and it is
derived from measurements of stress and strain (actually, force and displacement).
This model of one-dimensional elastic deformation is only an approximation of the
behavior of the material under load, and the elastic constant of the bulk sample, E,
frequently varies as a function of load, crystallographic texture, and microstructure.
Like most material properties, the stiffness in this model is defined by the behavior
of a material during a specific, standardized test. In general, constitutive relations
are not derived from first principles but are models of measured material responses
integrating many smaller-scale phenomena glossed over by the model.
The approximation of material behavior as the integration of influences of many
atoms/molecules over some small volume eliminates the need for detailed under-
standing of how these atoms interact; the only concern is their aggregate behavior.
Treating materials in this way is known as the continuum approximation, which mod-
els materials as if they exhibit the same behavior independent of their size (even to a
point in space). While this approximation breaks down in a real material if the vol-
ume considered is not much larger than the mean free path of the constituent atoms
or molecules, it is reasonable to be used down to the scale of tens of nanometers
in most condensed phases. (Many texts refer to averaged, “macroscopic” behavior
when discussing this approximation, but they miss the fact that many microstructural
phenomena may be treated as continuous materials.)
In this section, several constitutive equations and their measured properties are
introduced for fluid flow and heat and mass transfer. They are observed, but simpli-
fied, representations of reality and provide the foundations for the many mathemati-
cal models of the phenomena studied in this text.

1.3.1 sHear in Fluids

Isaac Newton [11] described a fluid as “a body, parts of which are started by and
give way to any force and, yielding easily, is moved among itself.”1 So, a fluid will
deform and “move among itself” continuously upon application of a shear force.
This response is different from a solid material, which under a constant load typically
will deform to a certain point and stop. The fundamental constitutive equations for
fluid motion relate the applied shear to the rate of material deformation and so help
to describe the behavior of fluids during processing. The first constitutive equation is
a simple linear one (defining Newtonian fluids), while other more complex (non-
Newtonian) fluids may be modeled by one of many nonlinear relations. A more
complete discussion of these matters can be found in many sources, e.g., [12].

1.3.1.1 Newtonian Fluids

To begin the examination of a constitutive equation for stress in a fluid, we observe
the situation in Figure 1.12, in which a fluid is contained between two horizontal
flat plates separated by the vertical distance H. (This configuration is known as
Couette flow.) If the lower plate is held stationary and a force of constant magni-
tude F is applied to the upper plate in the x-direction, the latter begins to accelerate
in the same direction. Movement of the upper plate sets up a shear stress between
12 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.12 Configuration of experiment defining viscosity in Newton’s law of viscosity,

showing variation of local flow velocity with position in the fluid between the two parallel
plates moving at different velocities.

the plate and the fluid, which opposes plate motion, and a steady state is reached
when the applied force F is balanced by the shear force, in which state the upper
plate has a constant velocity V. That shear is transmitted to the lower plate held in
place by a resisting force equal and opposite to the force of the upper plate. Newton
posited2 that a velocity V is directly proportional to the applied force F and to the
spacing H, and is inversely proportional to the upper plate surface area in contact
with the fluid, A [13]:

FH F V
V~ or ~ . (1.2)
A A H

The shear is transmitted between the plates and the fluid due to a no-slip con-
dition, where the fluid and plate have the same velocity where they are in contact.
Under this condition, the fluid at the lower plate (y = 0) is stationary and at the upper
(y = H) it has the velocity, V, causing the development of a velocity gradient, V/H,
in the fluid in the y-direction. So each layer is dragged by friction from the layer
above it and restrained by friction with the layer below. In Eq. (1.2), the ratio V/H
can be expressed in differential form as du/dy, where u is the velocity of a layer of
liquid in the x-direction. Also, in Eq. (1.2), F/A is the shear stress at the interface
between the upper plate and the fluid. Designating the shear stress as , Eq. (1.2)
can be written as

du
~ , (1.3)
dy

that is, the shear stress is proportional to the velocity gradient in the fluid. The pro-
portionality constant, μ, defined as

(1.4)
Introduction to Transport Phenomena in Materials Processing 13
14 An Introduction to Transport Phenomena in Materials Engineering

Table 1.1 shows gas viscosities two orders of magnitude lower than some liquids
(e.g., water, aluminum). A gas is a population of atoms or molecules moving ran-
domly in all directions within the vessel containing it. The interactions of molecules
in a gas, by which they exchange energy and momentum, are almost entirely due to
their occasional collisions while other intermolecular forces are relatively unimpor-
tant. An estimation of viscosity can be made by application of kinetic theory using
a rigid-sphere molecular model (e.g., in [14]), assuming that the distribution of the
atom velocities is determined only by the temperature of the gas and by the masses of
the atoms. This derivation gives the functional dependence of the viscosity,

~ MT d 2 , (1.7)

where M is the gas atomic or molecular weight, T the temperature, and d the molec-
ular diameter. Eq. (1.7) shows that gas viscosity is independent of the pressure of the
gas and is a linear function of T1/2.
The theoretical prediction that the viscosity of a gas is independent of pressure
might, at first sight, seem surprising. It might seem that with increasing pressure,
there would be more jostling of the atoms, such that, for a given velocity gradi-
ent, there would be a greater shear stress and hence greater viscosity in the gas.
The independence of viscosity on density, and hence on pressure, is because μ is
proportional to both density and , the gas mean free path [14]. However, as is
also inversely proportional to density, these influences on viscosity are canceled.
This canceling effect can be explained as follows: With decreasing density, there
are fewer atoms, which travel a proportionately greater distance between collisions
(higher ). The greater distance traveled exactly compensates for the smaller num-
ber of atoms moving, and hence the viscosity is independent of density. The theoret-
ical calculation of the viscosity of a gas was first made by Maxwell in 1860, and the
subsequent experimental observation that gas viscosity is independent of pressure
(at least at reasonably low pressures) was a distinct triumph for the kinetic theory.
While kinetic theory predicts that ~T 1/ 2 , experiments tell a slightly different
story. As examples, measurements of viscosities of He and Ne as a function of tem-
perature reveal that He ~T 0.65 and Ne ~T 0.67 . The dependence on temperature of the
viscosities of all real gases is ~T n , with n in the range 0.6–1.0. This discrepancy
occurs because the atoms of a gas are not nonattracting hard spheres, as assumed
in the derivation of Eq. (1.7); they are “soft” spheres surrounded by force fields
that attract at large and repel at short interatomic distances. Atoms colliding in a
high-temperature gas penetrate more deeply into each other’s force fields than in
a low-temperature gas. The diameters of the atoms thus appear to decrease with
increasing temperature and, according to Eq. (1.7), the viscosity increases.
Introduction to Transport Phenomena in Materials Processing 15

In contrast with gases, there is no reasonable kinetic theory for liquids, and most
theories of liquids are based on models which employ parameters that lack funda-
mental significance and which require a priori assumptions as to the structures, inter-
actions, or mechanisms of transport in the fluid. For Newtonian liquids in shear, the
molecular orientations and positions in the very loose (if closely packed) structure
may change, the aggregate effect of the breaking and reforming of molecular interac-
tions is that viscosity is relatively unchanged [15]. Figure 1.13 shows the measured
temperature dependence of the viscosities of liquid metals in the form

(1.8)

in which expression H* is the activation enthalpy for the thermally activated pro-
cess of viscous flow. Comparison of Eq. (1.8) to Eq. (1.7) shows that although the
constitutive relation for the flow of Newtonian liquids and gases is the same, the
kinetic-molecular mechanisms are very different, leading to different responses to
temperature. The viscosities of gases increase and liquids decrease with increasing
temperature. The elementary flow process in a liquid involves either the squeezing
of material units (atoms or molecules) between pairs of other material units or the
creation of small holes in the liquid into which a material unit can move, and H* is
the energy barrier that a flow unit must overcome to squeeze successfully between its
neighbors or is the energy required to form a hole. The fraction of flow units that are

FIGURE 1.13 Inverse temperature dependence of the dynamic viscosities of several liquid
metals.
16 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.14 Variation with temperature of viscosities of several gases and liquids at atmo-
spheric pressure.

sufficiently energetic to overcome the energy barrier to flow increases exponentially

with increasing temperature and hence the viscosity of the liquid decreases exponen-
tially with increasing temperature.
The variations of the viscosities of several common fluids with temperature
at atmospheric pressure are shown in Figure 1.14. The difference in temperature
dependence between liquids and gases is obvious, as is the fact that the heavier, more
complicated molecules tend to have higher viscosities.
Of course, the constitutive relation, Eq. (1.4), defines dynamic viscosity using
only one-dimensional shear and deformation, which simplifies measurement of μ.
However, in modeling multidimensional flows, the constitutive equations are more
complicated. For a three-dimensional flow field, with the velocity vector written in
Introduction to Transport Phenomena in Materials Processing 17

FIGURE 1.15 Viscous shear stress vectors on the visible sides of a control volume.

and

where the components ui are u1 = u, u2 = v, and u3 = w in Cartesian coordinates

(x1 = x, x2 = y, x3 = z). These constitutive equations will be used to derive momentum
balances for fluid flow in Chapter 7. They have the same proportionality constant,
μ, as Newton’s law of viscosity, Eq. (1.4), which can be measured in the simpler,
one-dimensional flow configuration.

1.3.1.2 Non-Newtonian Fluids

Newtonian fluids are typically simple enough that their structure is little altered when
sheared and so the resistance to shear stress is proportional to shear strain rate and
the viscosity is constant. However, fluids with more complicated structures exhibit
viscosity changes with level of shear. These non-Newtonian fluids include polymer
18 An Introduction to Transport Phenomena in Materials Engineering

melts and solutions, ceramic slurries, many foods, house paint, and magma [17]. The
definition of viscosity from Eq. (1.6),

(1.13)

still holds, but the ratio is no longer constant. Because the viscosity varies with , μapp
is termed the apparent viscosity.
Rheology is the study of the behavior of different complex, non-Newtonian flu-
ids,4 some examples of which can be seen in Figure 1.16, where they are compared
to the response of a Newtonian fluid. The top curve is an example of a fluid which, as
the strain rate increases, the shear stress required to maintain deformation rises faster.
An example of such shear thickening behavior is a concentrated suspension, a liquid
with a high solids loading [17]. With no strain, there is typically liquid between the
solid particles. At low , that liquid lubricates the particles so they slide past each
other easily and μapp of the suspension is low. With higher strain rates, the lubrication
layers break down and there are increasing incidents of solid-solid contact, rapidly
increasing the frictional resistance to fluid movement. In Eq. (1.13), it is seen that,
because increases faster than , the apparent viscosity increases with higher strain
rate. Porcelain slurries are examples of such suspensions; other shear thickening flu-
ids include magma and wet cement aggregates.
Behavior of shear thinning fluids is also sketched in Figure 1.16. These fluids are
incrementally less difficult to shear as the strain rate increases. With higher , the

FIGURE 1.16 Shear stress and shear strain rate behavior of different types of fluids: shear
thickening, Newtonian, and shear thinning.
Introduction to Transport Phenomena in Materials Processing 19

slope of shear curve decreases and the apparent viscosity is lower. As an example,
one class of shear thinning fluids is thermoplastic polymer melts, which are generally
long, flexible, covalently bonded chains the backbones of which are (mostly) car-
bon atoms [18]. (While important materials such as polymer solutions and uncured
epoxy resins exhibit shear thinning behavior, as an example of such behavior we
will focus in this discussion on polymer melts.) Simple examples of the side groups
from the main chain include H atoms (polyethylene), H and Cl (polyvinyl chloride),
H and CH3 (polypropylene), while polymethylmethacrylate and polycarbonate have
much more complicated side structures. The polymer backbone in polycarbonate
and polyethylene terephthalate includes other structures that are complex enough to
affect the backbone flexibility and consequent properties. At rest, the chains are not
straight but have kinks, bends, and loops leading to extensive chain entanglements.
These chains are long, from thousands to tens of thousands of repeating units (mers),
so the molecular weight (MW) is much larger than hydrocarbons, such as octane.

than a few mers). If these side branches are linked covalently to more than one chain,
the polymer is said to be “cross-linked” and, with a sufficiently dense cross-linking,
will not melt or flow.
With no large-scale motion in the melt, polymer chains are in a low-energy state
and so are in a random orientation, not straight, and entangled. At low and , these
entanglements cause resistance to deformation and so higher viscosity. The shear
tends to straighten the chains, but the strain rate is low enough that there is time for
them to relax back to their original state. With no stable alteration of structure, the
viscosity is not a function of shear; this constant value is the zero shear viscosity,
μapp = μo. At higher strain rates, there is less time to relax, so structural changes in the
melt persist as the polymer chains begin to straighten in the direction of flow. With
fewer entanglements, the fluid becomes progressively easier to move even though
there is more sliding friction along the chains. Recalling again Eq. (1.13), we see
that μapp decreases at higher strain rates because the strain rate increases faster than
the shear stress. When finally the chains are mostly aligned with the flow and the
only friction is from sliding past each other in parallel, no further structural changes
take place. At this very high strain rate, the fluid returns to Newtonian behavior and
μapp = μ∞, a constant and the lowest value of viscosity. (This state is rarely achieved

uncured epoxies.) For practical processing of thermoplastic polymers,

is a typical range of strain rates. These three regimes are sketched in Figure 1.17, in

polymer. This curve shows the two extreme, Newtonian regimes with constant slopes
(log μ) and the thinning regime in the middle.
Given the nature of the structural changes described earlier, obviously the viscos-
ity will also be a function of the structure of the individual chains. More branched
polymers will have more difficulty untangling and sliding past each other and so will
20 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.17 The dependence of apparent viscosity and shear stress on strain rate for a
shear thinning polymer.

have higher viscosities. The curve in Figure 1.18 shows the dependence of the zero
shear viscosity (μo) on the molecular weight of a given polymer. Experiments con-
firm that below a critical molecular weight (MWc), the small chain length polymer
has fewer chain interactions and μo ~ MW. Above the critical value, there are enough
interactions so the resistance to flow increases rapidly as μo ~ MW3.4. This measured
exponent is remarkably constant across a wide range of thermoplastic melts. The crit-
ical molecular weight is a function of temperature and polymer structure, but most
useful thermoplastic polymers are found to have MW >> MWc. The relation plotted
in Figure 1.18 suggests that a polymer will become less viscous the more times it is
recycled. The shearing during that processing tends to break polymer chains, leading
to shorter average chain lengths and broader size distributions in the plastic. This
structural degradation deteriorates the plastic’s mechanical properties and necessi-
tates the mixing of virgin material (usually above 50% by volume) with “post-con-
sumer” plastic. It is also the reason for the “recycling cascade” in which high-quality
polymer is recycled into products with less stringent mechanical requirements. This
progresses down the cascade until such weak items as plastic shopping bags are pro-
duced, which must be reused, buried, or burned.
How to model the behavior of such fluids? What sort of constitutive equation can
we propose to relate shear stress and velocity gradient (or shear strain rate)? A simple
relation that is a reasonable model for many fluids of this type is (in one dimension,
as in Figure 1.12):

(1.14)
Introduction to Transport Phenomena in Materials Processing 21

FIGURE 1.18 Dependence of zero shear viscosity of polymer melts on average molecular
weight.

This relation, defining a power-law fluid, models shear as a simple nonlinear func-
tion of the velocity gradient. The coefficient, k, is the consistency index and has the
odd units [kg/(m s2-n)], where n is the measured, unitless flow behavior index. (Note
that if n = 1, then k = μ and Eq. (1.14) reduces to Newton’s law of viscosity, Eq.
(1.4).) The power-law relation can also be written as

(1.15)

-
ening (n > 1) fluids to reproduce the behavior seen in Figure 1.16. However, the
consistency and flow behavior indices generally are only a good fit to measured data
over a few orders of magnitude of , and Eq. (1.14) is only applicable in the middle
range shown in Figure 1.17. See table 1.2 for examples.
A model that is applicable over a broader range of strain rate is due to Cross [19]:

(1.16)

-
ters: μo, μ∞, k, and n μo and μ∞) are
22 An Introduction to Transport Phenomena in Materials Engineering

The reference temperature, T*, is usually taken to be the glass transition temper-
ature (Tg) of the polymer. For a wide range of thermoplastic melts, Eq. (1.20) is a
surprisingly good fit, if C1 = 17.44, C2 = 51.6, the temperatures are in Kelvin, and if
the temperature is between Tg and Tg + 100 K.

1.3.2 Modes oF Heat transFer

The first law of thermodynamics states that energy is conserved in any process, but
the second law insists that if heat energy flows spontaneously, it moves from regions
of higher to lower temperature [22, 23]. The study of heat transfer is the effort to
understand the rate at which that thermal energy flows and how that rate depends on
the temperatures of the heat sources and sinks.
The discussion of constitutive relations for heat transfer begins with the classifi-
cation of that phenomenon into its three primary modes, conduction, convection, and
radiation, each of which has its own mechanism for moving thermal energy.
Introduction to Transport Phenomena in Materials Processing 23

TABLE 1.2
Sample Viscous Properties of Non-Newtonian, Power-Law Fluids (Compiled
in [17])
Material T (oC) n (-) k (Pa sn)
High-density polyethylene (HDPE) 180–220 0.6 3.8–6.2 × 103
Polystyrene (PS) 190–225 0.2 3.5–7.5 × 104
Polypropylene (PP) 180–200 0.4 4.5–7.0 × 103
Low-density polyethylene (LDPE) 160–230 0.45 4.3–9.4 × 103
Nylon 220–235 0.65 1.8–2.6 × 103
Polymethyl methacrylate (PMMA) 220–260 0.25 2.5–9.0 × 104
Polycarbonate (PC) 280–320 0.65–0.8 1.0–8.5 × 103
Toothpaste 27 0.28 1.2 × 102
Chocolate 30 0.5 7.0 × 10–1
Blood 25 0.9 4.0 × 10–3

1.3.2.1 Conduction Heat Transfer

Heat transfer in a rigid material with no bulk motion is effected by the diffusion
(or conduction) of thermal energy from a hot source to a cold sink. This mode of
heat transfer requires a medium through which the energy can transit. Macroscopic
experiments suggest that the heat flux, q (in W/m2), the rate at which thermal energy
travels per unit cross-sectional area A, is proportional to the temperature gradient
down which it moves:

dT
q ~ . (1.21)
dx

Inserting a proportionality constant, k, we have Fourier’s law for conduction,

(1.22)

which is the constitutive equation for conduction heat transfer. The negative sign is

T(x); heat always flows down a temperature

The proportionality constant, k, in Eq. (1.22) is the definition of a material’s ther-

mal conductivity, which has units of [W/mK]. These units are those of the ratio of
heat flux and temperature gradient, (W/m2)/(K/m). Thermal conductivities of mate-
rials are found over a range of five orders of magnitude, from 3400 W/mK for pure
diamond, to 400 W/mK for pure copper, to 0.04 W/mK for fiberglass insulation and
0.015 W/mK for silica aerogel (all evaluated at room temperature). Table 1.3 has
more examples of k for common materials. Thermal conductivity in solids can be
measured in many ways, including the apparatus shown in Figure 1.19(a). A power
source on the left (e.g., an electric resistance heater) provides a heat flow, q q A in
24 An Introduction to Transport Phenomena in Materials Engineering

TABLE 1.3
Typical Thermal Conductivity Values for Selected Materials
Material T (oC) k (W/mK) Ref.
Aluminum (99.5%) 20 218 [24]
Aluminum (99.0%) 20 209 [24]
Stainless steel (316) 20 16 [9]
Copper 20 394 [24]
Copper 1037 (solid) 244 [24]
Copper 1083 (liquid) 166 [24]
Low-density polyethylene 20 0.33 [9]
High-density polyethylene 20 0.48 [9]
ZrO2 20 2.0 [9]
Al2O3 20 30 [9]
SiC 20 90 [9]
Yttria-stabilized zirconia thermal barrier coating 700–1200 1.2–1.5 [25]

[W], through the material with cross-sectional area A (normal to heat flow direction)
and length L. The temperature gradient is obtained from the temperatures measured
by the thermocouples and their positions. The steady state data may be plotted as in
Figure 1.19(b) and the material thermal conductivity may be calculated as

(1.23)

where T1 and T4 are the temperatures measured near the hot and cold ends of the sample.
Fourier’s law, like all experimentally observed constitutive equations, represents
the integration of many physical mechanisms occurring at smaller scales than the
experiment in Figure 1.19. The thermal conductivity, a property of the material, is
dependent on the microstructure and atomic bonding. In condensed phases, when
linking the value of k to the material structure, it can be broken down into two parts:

(1.24)

where klat is the contribution of lattice vibrations and kfe of the motion of free electrons.
The first component of thermal conductivity in Eq. (1.24), klat, is present to some
degree in all condensed phases, as it has its origin in the vibrational energy of the
atoms. This energy increases monotonically with temperature, and its transfer is
modeled as elastic waves (phonons) traveling through a material. In the hot regions,
the phonons pick up more energy from the vigorously vibrating lattice and lose it
heating colder, less energetic areas. In perfect crystals, the phonons would travel
easily through the material and it would exhibit very high thermal conductivity.
However, in real materials, imperfections such as dislocations and grain and phase
boundaries scatter and misdirect these elastic waves, reducing the rate of transfer
from hot to cold regions. When more imperfections exist in a volume, more scatter-
ing occurs, lowering k.
Introduction to Transport Phenomena in Materials Processing 25

FIGURE 1.19 A method for measuring thermal conductivity. (a) Schematic of test appara-
tus, showing heat input and positions of thermocouples (TC). (b) Temperature readings along
length of a sample in which k is only a weak function of temperature.

The second part of the bulk thermal conductivity in Eq. (1.24) is due to the transfer
of thermal energy by the motion of free electrons. These electrons move through-
out a body in a way to maintain charge neutrality, but the ones from hotter regions
have more kinetic energy, which they impart on the lattice and other free elec-
trons as they move into the colder areas, thereby moving heat down a temperature
gradient.
As noted, the total thermal conductivity is a sum of these two physical effects. In
metals, the influence of phonon motion is less than that of the free electrons charac-
teristic of metallic bonding. In pure metals, k is higher than in alloys, as impurities
strain the lattice and provide possible phonon scattering sites. Evidence for the dom-
inance of conduction by free electron transport is found by the direct proportionality
of thermal (k) and electrical ( e) conductivities in metals, as predicted by the Wiede-
mann–Franz law,

(1.25)

where L is a constant (which theory predicts as 2.44 × 10–8 ΩW/K2). Ceramics have
much lower thermal conductivities than metals, as free electrons are not present and
heat transport is entirely by phonons. Large differences in atomic size among different
elements in a ceramic, as well as large, complicated units cells, may further lower k.
At high temperatures, some ceramics are semitransparent to infrared electromagnetic
26 An Introduction to Transport Phenomena in Materials Engineering

radiation, causing an increase in heat transfer that appears (in tests such as in Fig-
ure 1.19) to be an increase in thermal conductivity. Actually, the increase is due to
internal thermal radiation, not conduction. The porosity often found in bulk ceramics
also affects the bulk value of k, lowering it due to the reduction of cross-sectional
area and the complexity of the conduction path. Polymers, like ceramics, have almost
no free electrons and also much less crystallinity than metals or ceramics. The less
ordered structure of polymers gives rise to elastic waves (having their origin in the
vibration and rotation of the molecular chains) more disorganized than in more crys-
talline materials, with many more opportunities for scattering. The less crystalline
the polymer, the lower its thermal conductivity.
Fourier’s law, Eq. (1.22), is written as a one-dimensional relationship between heat
flux and temperature gradient, defined by experiments such as Figure 1.19. However,

The experiment shown in Figure 1.19 measures only the conductivity in a princi-
pal direction, kii.

Example 1.2
Heat Loss to a Buried Pipe
Consider a fluid at temperature Tf = 85 oC, running through a polyvinyl chloride (PVC)
pipe buried in soil at Ts = 75 oC. Assume the inside pipe surface is at Tf and the outside
is at Ts. The inner radius of the pipe is Rf = 9 cm, the outer radius is Rs = 10 cm, and
the length of pipe is L = 1 m. The thermal conductivity of the PVC is k = 1 W/mK.

Find: The heat loss, q, from the fluid to the soil through the pipe wall.
Solution: We start with the conduction rate equation in cylindrical coordinates:
Introduction to Transport Phenomena in Materials Processing 27

The area normal to the radial heat flow, unlike in Cartesian coordinates, is a function

1.3.2.2 Convection Heat Transfer

Another mode of heat transfer that also requires a medium through which thermal
energy is transported is convection. This mechanism is based on the transport of heat
by conduction and bulk motion of the medium. If a body is hotter than the fluid sur-
rounding it, heat conducts from the surface into the nearby fluid. If the fluid moves
and the hot fluid near the surface is replaced quickly with cold fluid from elsewhere,
heat is moved away from the body by advection, the motion of heat caused by bulk
movement of the medium. The rapid replacement of hot fluid with cold maintains a
high temperature gradient at the surface, keeping the heat transfer rate high. Convec-
tion heat transfer is a combination of the advection and conduction.
Experimentally, such conditions can be observed in order to discover the relation-
ship between the heat flux at the solid-fluid interface and the fluid temperature field.
Such experiments show a dependence of the heat flux on the temperature difference
between the surface (Ts) and the ambient fluid (T∞):

(1.28)

The exponent taken from these data depends on the mechanism driving the flow.
In forced convection, in which the flow is propelled by some exterior force, such as a
fan, a pump, or electromagnetic effects, n ≈ 1. In other cases, density gradients due to
spatial variations in temperature or composition give rise to buoyancy-induced flows,
also known as natural, or free, convection. For natural convection, n is usually in the
range of 1 < n < 1.3.
The experimental result represented by Eq. (1.28) is usually written as the rate
equation for convection,

(1.29)

known as Newton’s law of cooling. The heat transfer coefficient, h [W/m2K], is

the proportionality constant between the heat flux and temperature difference,
28 An Introduction to Transport Phenomena in Materials Engineering

TABLE 1.4
Approximate Ranges of Heat Transfer Coefficients (W/m2K)
Convection type Gas Single phase liquid Boiling liquid
Forced 10–1000 100–30,000 30,000–150,000
Natural 1–100 10–10,000 1000–50,000

determined by experiment. Unlike thermal conductivity or viscosity, the heat trans-

fer coefficient is not a material property, as it is a function of fluid properties,
velocity, and surface shape. For forced convection (n = 1), h is generally a very
weak function of temperature (mostly through temperature-dependent fluid prop-
erties), while

(1.30)

for natural convection.

The methodology for calculating and measuring heat transfer coefficients for forced
or natural convection conditions will be discussed in Chapter 10. Typical ranges of
heat transfer coefficient values for different conditions are found in Table 1.4.

Example 1.3
Cooling with a Heat Sink
The surface of a heat sink of area, A = 0.1 m2, must not exceed a maximum tempera-
ture of Ts,max = 100 oC while dissipating heat at 10 W. The surrounding air tempera-
ture is T∞ = 30 oC.

Find: (a) Plot the heat transfer coefficient required to dissipate the power as
a function of the heat sink surface temperature and (b) find the minimum
value of the heat transfer coefficient allowable.
Solution: (a) We can rearrange the rate equation for convection, Eq. (1.29),
and have plotted the heat transfer coefficient as a function of the surface
temperature in Figure 1.20:

The value of the minimum heat transfer coefficient can be found from the previous
equation or the plot:

The heat transfer coefficient must be at least that value to maintain the surface tem-
perature below 100 oC. How to control the value of h is the topic of Chapter 10.
Introduction to Transport Phenomena in Materials Processing 29

FIGURE 1.20 Required heat transfer coefficient as a function of heat sink surface.

1.3.2.3 Radiation Heat Transfer

A third mode of heat transfer, radiation, is fundamentally different from conduc-
tion and convection in several ways. First is that the heat flow is not driven by a
temperature difference (or gradient), but emission of heat by thermal radiation. This
emission, a direct function of the absolute surface temperature of the body, is an elec-
tromagnetic phenomenon; thermal radiation is part of the electromagnetic spectrum
and as such does not require a medium for transfer (although it can travel through
some matter). This phenomenon is most obvious when one considers the heat our
planet receives from the sun, having traveled through 93 million miles of vacuum
before interacting with the earth. Electromagnetic radiation travels at the speed of
light and may be scattered or absorbed by an intervening material.
To thoroughly and quantitatively understand the origins of the emission of ther-
mal radiation from matter, one must delve deeply into quantum mechanics, electro-
magnetic wave theory, and thermodynamics, but that study is outside the purview
of this book. Qualitatively, if heat is added to matter, the temperature rises, thereby
increasing the vibrational energy of the atoms and molecules, causing the emission
of electromagnetic radiation over a range of wavelengths. That distribution of emit-
ted energy flux over a wavelength spectrum (known at Planck’s distribution) depends
on the types of changing energy states. The distribution is a strong function of tem-
perature, as seen in Figure 1.21.
The thermal radiation emitted by matter at room temperature (≈300 K) is entirely
at wavelengths in the infrared region, much higher than the spectrum of visible light
(0.4 µm < < 0.7 µm). As temperature increases, more radiation is emitted (higher
Eb, ) and the range shifts to lower wavelengths. The distribution in Figure 1.21 for
T = 300 K moves left and up as T increases.
30 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.21 The distribution of emissive flux from a perfect surface as a function of
temperature and wavelength.

The emitted light first seen by the human eye as the temperature increases is red,
at the higher end of the visible. As T continues to rise, the distribution continues
to move to lower and the visible light changes from red to yellow to white (the
whole visible spectrum). This shift can be observed by watching a high temperature
furnace as it heats from room temperature. Eventually we see in Figure 1.21 that the
peak of the emitted spectrum at about 5800 K is in the center of the visible range;
this temperature is approximately what is found on the surface of the Sun. (From the
viewpoint of the evolution of the eye of a daytime creature, this fact is probably not
coincidental.)
If the Planck’s distribution shown in Figure 1.21 is integrated over at a given T,
the total rate of emission per unit area (qe ) turns out to be proportional to the abso-
lute temperature of the surface (Ts) to the fourth power:

(1.31)

where the proportionality constant is = 5.67 × 10–8 W/m2K4, the Stefan–Boltzmann

constant. This quantity is a universal constant and not a material property. (The law
was deduced from experimental data by Josef Stefan in 1879 and then derived theo-
retically by Ludwig Boltzmann in 1884.)
The Stefan–Boltzmann law, Eq. (1.31), and Planck’s spectral distribution of radi-
ative heat flux, Figure 1.21, assume a perfect emitter, known as a blackbody. No such
surface exists (although approximations will be discussed in Chapter 12), and all real
surfaces emit at some level less than Eq. (1.31). The first approximation of a less-
than-perfect emitter is a gray body, emitting

(1.32)

which introduces the surface property of emissivity, . This property tends to be a very
strong function of temperature, wavelength, angle from the surface ( ), roughness,
Introduction to Transport Phenomena in Materials Processing 31

TABLE 1.5
Averaged Emissivity Values for Selected Materials
(Excerpted from a Compilation in [26])
Material T (oC) (-)
Aluminum (highly polished) 225–575 0.039–0.057
Aluminum (oxidized) 95–500 0.20–0.31
Alumina 540–1000 0.65–0.45
Brass (oxidized) 200–600 0.6
Brick, fireclay 1000 0.75
Brick, red 20 0.93
Cast iron (machined) 22 0.44
Cast iron (oxidized) 40–250 0.95
Concrete, rough 38 0.94
Copper (polished) 115 0.023
Copper (oxidized) 25 0.78
Gold (highly polished) 225–625 0.018–0.035
Graphite (pressed) 250–510 0.98
Ice 0 0.985

and the atomic structure of the surface material. The value (0 < < 1) used in Eq.
(1.32) is found for a specific surface, integrated over T, , and . Some typical values
of this averaged for common surfaces are found in Table 1.5.
The previous discussion focused on the emission of thermal radiation from mat-
ter, but if all surfaces emit, where does that radiation go? The answer shows another
unique feature of radiation as a heat transfer mode: all bodies in a line of sight with
each other exchange thermal radiation, both emitting and absorbing heat. The inci-
dent thermal radiation increases the vibrational energy of the atoms in the receiving
material, raising its temperature. A black body (a perfect emitter) will absorb all
incident radiation, but real surfaces absorb something less. The surface absorptivity
( ), like emissivity, is presented for now as an average value, so we can write an

(1.33)

∞ , then

(1.34)

The range of ) not absorbed is

approximation of

(1.35)
32 An Introduction to Transport Phenomena in Materials Engineering

This equation can be linearized so that it can be handled like the convection rate equation:

(1.36)

(1.37)

This rate equation, Eq. (1.35), while useful as a starting point to understanding radi-
ation heat transfer, is much less general than the other constitutive relations for con-
duction and convection heat transfer. It is strictly applicable to one small “gray”
surface at Ts in a much larger environment at T∞. The equations for a more general
form of radiation exchange between two bodies will be derived in Chapter 12.

Example 1.4
Heat Transfer from a Surface by Parallel Radiation and Convection
A surface at temperature TS loses heat to the environment (at T∞) by both forced
convection and radiation. Assume the surface emissivity is = 0.5 and T∞ = 300 K.

Find: (a) Find an expression for the fraction (R) of heat loss due to radiation.
(b) Plot R = R(TS) for h = 2, 20, and 200 W/m2K, over the range 300 K < TS
< 1000 K.
Solution: (a) From the rate equations, Eqs. (1.29) and (1.35), the total heat flux

Rewriting

Because h rad s
can see that as T 0, R 0 (only convection). If the surface
temperature is very large, rad and R 1 (only radiation). These trends are
seen in Figure 1.22.
For this specific example, the results show that, at the lowest h value, radiation is
always important and it dominates heat transfer above 500 K, while at h = 200 W/m2K
Introduction to Transport Phenomena in Materials Processing 33

FIGURE 1.22 Fraction of heat loss due to radiation as a function of surface temperature and
heat transfer coefficient.

radiation is only a small contribution over the entire range of surface temperatures.
The trend at all convection heat transfer coefficients is that radiation, however unim-
portant at low temperatures, becomes increasingly important at higher temperatures.

1.3.3 Mass transFer

The purpose of materials processing is not only making a shape; it is also a quest for
desirable properties, and that is accomplished by forming specific microstructures.
The evolution of these microstructures in many processes (e.g., solidification, heat
treatment, sintering) is governed by the movement of solute atoms in a solid solution.
Equilibrium phase diagrams indicate a material’s thermodynamically favored state(s)
at given process temperatures, but moving atoms to effect changes from one state to
another generally takes a finite amount of time, so the study of mass transfer is nec-
essary to understand microstructural development.
Experimental observations in binary and many multicomponent metal alloys
show that there is a net flow of solute atoms down concentration gradients. (In this
text we will focus on binary alloys.) This diffusion is a manifestation of the random
motion of atoms. In the solid state at any finite temperature, the atoms vibrate about
their mean positions in a crystal lattice, and every now and then, when conditions are
favorable, an atom can jump from one site to a neighboring vacant site. The probabil-
ity that an atom may move in a given direction is affected by many factors, such as:
crystal structure and orientation; the location and distribution of vacancies, including
in and near dislocations; grain boundaries; surfaces; and the presence of other types
of atoms (as in multicomponent alloys). In the end, the question is: why is there a net
flow of atoms down a concentration gradient?
34 An Introduction to Transport Phenomena in Materials Engineering

The explanation for the net flow of atoms from high to low concentrations begins
with a thought experiment known as the “drunkard’s walk” (or “random walk” in
more polite company). Imagine a tipsy sailor leaves a bar and attempts to return to his
ship. Although the sailor is capable of walking, he does not have control of the direc-
tion of his steps and thus his motion is random. To simplify the analysis, consider that
the random motion of the sailor is in one dimension, in which steps are made either
forward to the ship or backward to the bar. The sailor begins at the origin, x = 0, and
takes n steps; he travels a distance l1 on the first step, l2 on the second step, and so
on. For now, assume that li is a random distance and there is equal probability that
its direction is positive or negative. After n steps, the net distance that the sailor has
traveled from the origin is

(1.38)

li to give the mean

square distance

(1.39)

-
tions of li

distances at steps n

After the first step

Introduction to Transport Phenomena in Materials Processing 35

If the n steps are made at the rate of r steps per second, the time taken to make the n
steps is t n / r (i.e., n rt ); hence,

xn2 rtl 2 ,

so we see that
12
xn2 ~ t1 2 . (1.40)

In this example, the mean square distance from the bar, Eq. (1.40), is the average
distance from the bar traveled by randomly walking sailors, if we have averaged
over enough sailors, each with their own random series of steps. The analogy to
atoms moving in a perfect, uniform lattice leads us to expect that for a very large
number of atoms, each making a very large number of jumps, in the aggregate the
local concentration of solute atoms will move a distance proportional to the square
root of time.
Figure
x-direction. The space on the left has a concentration of C
volume, and that on the right contains particles at the concentration of CR parti-
cles per unit volume. Over time t, the average distance traveled by all12
randomly
walking particles is the square root of the mean square distance, xn2 . For each
atom, either from the right or the left, there is an equal probability of jumping,
so the number of particles moving across
12
the interface plane between the left and
right volumes over time t is ½C x 2 . Thus the flux of particles from left to right,
jL R , is
12
1C
L
x2
jL R .
2 t

FIGURE 1.23 Two rectangular volumes containing randomly walking particles at concen-
tration CL in the left volume and CR in the right volume.
36 An Introduction to Transport Phenomena in Materials Engineering

Similarly, the flux of particles from the volume on the right to the volume on the
left is

The net

(1.41)

(1.42)

(1.43)

(1.44)

which is Fick’s Law of diffusion of component i and is a rate equation for mass
diffusion in a binary solution. Examples are found in Table 1.6.
Like other constitutive relations in this section, Fick’s first law can also be
observed experimentally, and the diffusion coefficient, Di, is the ratio of the mass
flux and the concentration gradient. It is similar to the thermal conductivity, k, in
Fourier’s law, Eq. (1.22), in that it is a proportionality constant between a net flux
and the gradient of a transport quantity. Unlike thermal conductivity, which is usu-
ally a weak function of temperature, Di tends to be a strong function of composition.
This dependence is due to the effect on jump probabilities of strain fields in the bulk
crystal lattice caused by the presence of solute atoms. The form of Eqs. (1.22) and
Introduction to Transport Phenomena in Materials Processing 37

TABLE 1.6
Mass Diffusion Coefficients (Excerpted from a
Compilation in [9])
Material system T (K) Di (m2/s)
Cu in Al 700 5.9 × 10–14
Zn in Cu 1200 1.6 × 10–13
C in -Fe 900 3.5 × 10–9
C in -Fe 1250 2.8 × 10–9
Al+3 in Al2O3 1000 3.4 × 10–28
O−2 in Al2O3 1000 1.1 × 10–34
CO2 in polyvinyl chloride 350 1.2 × 10–11
O2 in polyethylene 350 4.7 × 10–10

(1.44) make heat conduction and binary mass diffusion look similar from a macro-
scopic point of view, and similar techniques can be used to derive and solve heat and
mass conservation equations. However, even though the phenomenological formula-
tion is similar (at least in binary systems), it must be remembered that the physics of
transport of these two phenomena are quite different and the similarities only apply
in the simplest cases. We have also treated the jump mechanisms and probabilities
as homogeneous and isotropic; if they are not, for example, due to the presence of
a distribution of atoms of other elements in a multicomponent solid solution, then
the net effect may be an atomic flux up concentration gradients. These issues will be
addressed in Chapter 4.

Example 1.5
Diffusion of Hydrogen through a Planar Iron Wall
A storage tank for hydrogen has an iron wall of thickness L = 0.001 m, as shown in
Figure 1.24. The stored H2 gas is on the left side of the wall where the pressure is
P1 = 1 atm, while there is practically no H2 in the outside atmosphere (P2 ≈ 0 atm).
At 400 oC,

Find:
Solution:

where CH(P) is the hydrogen concentration in iron at equilibrium with hydrogen at

pressure P.
38 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.24 Diffusion of hydrogen through a planar iron wall.

To calculate the concentration of hydrogen in iron in contact with H2 gas, we

remember from basic thermochemistry that hydrogen dissolved in iron obeys
Sievert’s law, which states that, for the equilibrium reaction between hydrogen in the
gas and in solution in the iron,

the equilibrium constant is

where [H]Fe is the hydrogen concentration in iron (in kg H/kg Fe), in equilibrium
with hydrogen gas at pressure PH2 . At 400 oC and a hydrogen pressure of 1.013 × 105
Pa (1 atm), the solubility of hydrogen in iron is 3 ppm by weight, or [H]Fe = 3 × 10–6
kg H/kg Fe. Thus at 400 oC,
Introduction to Transport Phenomena in Materials Processing 39

or

This result gives us the rate of H2 loss per unit area of tank surface, which is about
31 g H2/day.

1.3.4 suMMary oF constitutive equations

TABLE 1.7
List of Constitutive Equations Described in Section 1.3

Shear (Newtonian)

Shear (power-law fluid)

Conduction heat transfer

Convection heat transfer

Radiation heat transfer

Mass diffusion (binary alloy)

40 An Introduction to Transport Phenomena in Materials Engineering

1.4 FINDING SOLUTIONS TO MODELS OF

TRANSPORT PHENOMENA
In the following chapters, we will derive several transient, multidimensional par-
tial differential equations, which are the bases of our models of the conservation of
various transport quantities (e.g., thermal energy, momentum, species). These der-
ivations typically will begin with the constitutive relations in Section 1.3 and the
general balance in Eq. (1.1). As we discuss different approaches to representing these
physical phenomena, we should keep in mind that the purpose served by the models
is to provide a useful theoretical framework for understanding reality and for esti-
mating behavior. The reader should remember that the models are useful only as far
as they do represent the real behavior of physical phenomena and must be supported
and informed by carefully designed and executed experiments. A researcher who
helped lay both the theoretical and experimental foundations of the field of natural
convection heat transfer put it best: “When nature talks, you have to listen” [27, 28].
At first, especially for those whose exposure to partial differential equations is thin, the
complexity of these models can be intimidating. However, there are a variety of methods
for extracting useful information from them that are within the scope of an introductory
text such as this one. Because we have covered little of transport phenomena so far, but
all engineers have been exposed to basic mechanics early in their university experience,
a problem from that field is used to illustrate the different solution techniques.

Example 1.6.a
Definition of the Falling Body Problem
Consider the system shown in Figure 1.25 in which a body of mass mb is falling in
a gravity field, restrained only by drag proportional to the fluid density, f, and the

(1.45)

or

(1.46)

-
tion, gravity, and drag, respectively. The mass begins at some point, y = 0, with zero
velocity (V = 0), and it is released at t = 0. We know from experience that the mass
accelerates until the drag balances the gravity force, at which point V reaches a con-
stant terminal velocity
CD, is
-
tion and all forces are in the y direction and
Introduction to Transport Phenomena in Materials Processing 41

FIGURE 1.25 Free body diagram of falling mass, subject to gravity and aerodynamic drag.

Find: The velocity of the body as a function of time, V(t), and the terminal veloc-
ity, Vterm, if there is one. (The various solutions will be produced and com-
pared in extensions of this example throughout this section.)

1.4.1 MatHeMatical solutions

The first approach to solving a mathematical model is to find an exact, closed-form
solution. This type of solution gives a general equation describing the behavior of
the system, valid for all cases within the assumptions used to make the model. This
continuous result is more easily applicable than discrete results of individual experi-
ments or numerical simulations, which rely on some form of statistical correlation of
data to generate a mathematical expression of behavior.
Unfortunately, the opportunity for exact solutions of the balance equations found
in the study of transport phenomena is limited. These solutions are possible only if
significantly limiting assumptions are made about the geometry, the properties, and
the boundary conditions. In many cases such limitations still produce a reasonable
representation of reality, which can be used for a design calculation at best and, at
least, to better understand the physics or to verify a numerical solution. Because a
model is only a simplification of reality, it must always be remembered that even an
exact mathematical solution is only an approximation.
In that light, it is apparent that, in addition to exact solutions, a wide range
of mathematically approximate methods can be used to solve models. Gener-
ally, these methods involve assumptions about the shape of the function sought
(e.g., temperature or velocity fields). This approach also may provide useful
solutions if care is taken in the choice of these shapes. The approximations do
42 An Introduction to Transport Phenomena in Materials Engineering

not necessarily make for less desirable solutions, as long as their limitations are
recognized and respected.

Example 1.6.b
Exact Solution to Falling Body Problem

(1.47)

(1.48)

t on the right,

(1.49)

(1.50)

= Vterm (the “late time

regime”):

(1.51)

(1.52)

at t

(1.53)
Introduction to Transport Phenomena in Materials Processing 43

(1.54)

We find

(1.55)

and

(1.56)

(1.57)

(1.58)

or

(1.59)

this approximation will work best at early times.

1.4.2 nuMerical solutions

If the model is complex enough not to admit an exact solution or even a useful
approximation, it may be necessary to turn to numerical simulation. The fundamen-
tal idea with almost all numerical methods is the conversion of sets of continuous
differential equations to systems of discrete algebraic equations by dividing both
space and time into discrete pieces. Each simulation for specific boundary and initial
conditions is similar to an individual experiment, in that the results are valid for only
that set of conditions and neither method gives the continuous functions of the exact
or approximate solutions.
The numerical solution has many distinct advantages, beginning with allowing
the relaxation of many simplifying assumptions and the prediction of much more
complicated behavior. Numerical simulation can be an excellent diagnostic tool for
investigating details of physical phenomena that are quite frequently unavailable
from experiments due to the difficulty of observation. For opaque systems, or ones
in which probes might alter the process or not survive high temperature or stress,
44 An Introduction to Transport Phenomena in Materials Engineering

many experiments in materials processing are forced to rely on post mortem anal-
ysis of samples from the final product. Even when in situ observation is possible, a
well-validated simulation can probe a process with much finer spatial and temporal
resolution than is possible in a real system. Trends in behavior with changes in pro-
cessing parameters or materials can be suggested by the results of numerical studies,
allowing the engineer to better understand the underlying mechanisms that control a
process. Simulations may also be used as a stand-in for some experiments, to cheaply
and easily estimate trends in process behavior; they are a helpful tool to reduce, but
not eliminate, the need for expensive and sometimes dangerous experiments.
Numerical simulation of processes does have these many advantages, but its use
also comes with several caveats. The main difficulty from the point of view of a
practicing engineer is also a major advantage: the proliferation of commercial soft-
ware and freeware to perform these calculations. It is quite straightforward to obtain
a code, be trained in the mechanics of its use, and begin calculations. The major
pitfall may appear when the engineer does not have a good grasp of both the phys-
ics included in the model and the numerical techniques employed. Without under-
standing the first, the code user cannot judge the results of the simulation and may
believe incorrect results arising either from his error in formulating the problem or
from misapplication of the code. Without the second, unknown assumptions in the
numerical treatment of the governing equations and the behavior of solution algo-
rithms can cause non-physical artifacts in the predictions. For both these reasons, the
number of engineers who have reported physically unreasonable results because of
a lack of knowledge of physics and/or numerical methods is legion. There is also a
surprising, if human, tendency for some engineers and managers to believe detailed
and well-presented numerical results, even in the face of contrary reason and experi-
mental data. In sum, giving a powerful code to an engineer unfamiliar with numerics
and/or the problem physics is like giving a teenager whiskey and car keys: you are
never sure of the outcome, but it is a safe bet that it will not be good. We need to
understand numerical methods (to get started, see [29–31]) as well as the dominant
physical mechanisms, aided by experiments and simpler solutions (as in this text), in
order to produce correct and useful results and to interpret them properly.
So, numerical solutions can be tricky and seductive, but that should not discour-
age us from using this powerful tool with discretion and caution. A famous meteor-
ologist and pioneer numerical analyst, L. F. Richardson, has been quoted as writing
that there are “in the mathematical country, . . . precipices and pit-shafts down which
it would be possible to fall, but that need not deter us from walking about” [32]. We
certainly encourage the reader to delve more deeply into this field of study.

Example 1.6.c
Numerical Solution to Falling Body Problem
Again, we begin with the first order, nonlinear differential equation for the velocity

(1.47)
Introduction to Transport Phenomena in Materials Processing 45

The approach here is to discretize time, so we only solve the equation at distinct
instants. These instants are separated by a period of time (the time step, t). Denoting
the velocity and the time of the nth step as Vn and tn and at the previous step as Vn-1
and tn-1 = tn – t, we can rewrite the velocity derivative as a ratio of differences and

(1.60)

or

(1.61)

Eq. (1.61) is the formulation of a time marching problem. We start at n = 1, where Vo

is known from the initial condition (here V(0) = Vo = 0). Once V1 = V( t) is found,
the index is incremented to n = 2 and V2 = V(2 t) is calculated as a function of V1, a
process repeated at each successive step n (tn = n t). When Vn ≈ Vn-1, the simulation
has reached terminal velocity and can be ended. This technique is known as Euler’s
Method, which is found in any basic text on numerical methods [e.g., 30, 32].

1.4.3 scaling analysis

In addition to using exact, approximate, or numerical solutions, one method of
extracting useful information from our models very quickly is known as scaling
analysis, which is described and applied extensively to transport phenomena in [33,
34]. The model equations for transport phenomena derived in later chapters (based
on Eq. (1.1) and constitutive equations) represent balances of mass, momentum,
energy, or species, and each term in these equations represents a physical phenome-
non in the balance. One purpose of a scaling analysis is to estimate the order of mag-
nitude5 of each term in the governing equations. The relative magnitude of each term
guides us to making rational, defensible simplifications to the problem by telling us
which terms are small enough to be neglected in the solution process. Another result
is the functional relationships among the remaining terms. These relationships are
only known to an order of magnitude, but are in the proper functional form.
The procedure for this analysis is roughly outlined as follows:

1. Write the full balance equations and boundary and initial conditions for
whatever quantities are transported.
2. In many problems, it is possible to break up the problem into different
regions in space and/or time and examine each separately using the follow-
ing steps:
3. Pick reference values for each of the dependent and independent variables,
the latter based on careful definitions of the spatial and temporal extent
of the problem. They should be selected so that the ratios of variables and
46 An Introduction to Transport Phenomena in Materials Engineering

their reference values vary between 0 and 1 in the region of interest. Often
reference values are not known at the beginning, but are found during the
analysis.
4. Approximate the order of magnitude of each term in the equations by
replacing each variable with its reference value. At this point the balance
equation is no longer exact, but approximate because each term is only an
order-of-magnitude estimate.
5. Select a physical phenomenon suspected to be important throughout the
problem. Divide the entire equation by the term representing that effect,
making all terms dimensionless and giving the dominant term a value of
(1).
6. Compare the orders of magnitude of each term in the equation to that of
the dominant physical mechanism, which is now (1). If the order of mag-
nitude of a term is much less than 1, then the term has only a small effect
and it can be neglected in further analysis. If the term is (1), then it should
be retained, as it is as important as the dominant effect. Finally, if a term
turns out to be much larger than (1), then that term should be considered
dominant and the analysis should return to step (5).
7. Compare the remaining terms to find the functional forms of the unknown
reference values.

As will be seen in the many examples in this book, to perform this analysis we need
some knowledge of system behavior in order to make many of these decisions, espe-
cially in steps (2), (3), and (5). Using this procedure, we continue with the example
of the falling body.

Example 1.6.d
Scaling Analysis of Falling Body Problem
Starting once again with the force balance for the falling body with aerodynamic
drag,

(1.47)

, the velocity at the

reference time,

(1.62)

three terms by dividing by g, leaving the gravity term of magnitude (1):

Introduction to Transport Phenomena in Materials Processing 47

(1.63)

(1.64)

(1.65)

(1.66)

(1.67)

So, what have we done here? Without solving exactly the force balance, Eq. (1.46), the
scaling analysis has extracted useful information (estimates of early behavior, terminal
velocity, time to terminal velocity) from the equation quickly and easily. The estimates
are only good to an order of magnitude, but the functional forms are known. Considering
both the uncertainty of the answers and the short time it took to obtain them, the process
gives a good return on investment [33]. The problem here is simple, almost trivial, but the
method is also useful for much more complicated problems in transport phenomena with
no exact solution, and this example serves as a preview of things to come.
48 An Introduction to Transport Phenomena in Materials Engineering

1.4.4 a note on selection oF solution approacH

In this section, we have introduced different approaches to solving the models of
transport phenomena described later in this text and illustrated those methods with a
simple mechanics problem. The question that immediately arises is: which approach
is the best? The answer can only be found by first answering another question: what
is the purpose of the model? If only a quick “back of the envelope” check is needed,
then a scaling analysis or a simple mathematical model amenable to an exact or
approximate solution should be sufficient. These should always be the first steps in
analyzing transport phenomena, as they will improve our understanding of the dom-
inant physics in a problem, estimate the gross behavior of the system, and show the
important groupings of variables. Once this step has been taken, an engineer has a
better idea if further work is needed, usually in the form of detailed numerical anal-
ysis and/or careful experiments. In either case, we should always start with simple
models, as the initial analyses will guide the more complicated and expensive steps
and reduce the effort needed for them.
This approach is of great practical use to a practicing engineer who, in addi-
tion to being conversant with the technical disciplines involved with the projects on
which she works, must have some area in which she is the local expert. With such
expertise, the engineer frequently has a steady stream of customers from inside and
outside her company asking for technical advice. In such a situation, an engineer
may respond, “Do you want the one month solution, the one week solution, or the
five o’clock solution?” indicating the deadline for the answer. A longer project may
lead to lower uncertainty, but at a higher cost. Very often, the answer is that only
a quick response is requested because of limited time and money to obtain a more
complete answer; there is only a need or the time to perform “back-of-the-envelope”
calculations, that is, to make reasonable approximations and obtain quick-and-dirty
solutions. If the customer needs or can afford only an answer to an order of mag-
nitude or within some factor of safety, or if they only need to know the sensitivity
of a dependent variable to an independent one, then the five o’clock solution is the
best path to take. In those cases, a full, detailed 3D numerical analysis will not give
them any additional useful information, only beautiful color animations devoid of
any real added value. At other times, there is enough time, money, and need for a
more detailed, less uncertain, answer. Then the initial approximate answer is still
important, to give the engineer the order of magnitude of the effects involved and
the relationships among important variables. This knowledge is an important check
on the behavior of the more complicated and detailed solutions provided by the
numerical codes, as well as a guide for understanding and organizing experimental
results. In that spirit, and recognizing that most readers are at the very beginning of
the long path to becoming an expert in this field, the goal of this text is to prepare
materials engineers to generate these “five o’clock solutions” by teaching them the
basic physics in transport phenomena and some of the simpler solution techniques.
Mastering this approach will elucidate the physics and prepare readers for more
advanced study in this field and to make decisions regarding if and how to proceed
to more detailed simulations and experimental design.
Introduction to Transport Phenomena in Materials Processing 49

Example 1.6.e
Comparison of Solutions of Falling Body Problem
In order to demonstrate the kinds of results generated by the different solution meth-
ods introduced earlier, we finish the example running through Section 1.5. To easily
compare the various solutions to this problem, we nondimensionalize the governing

(1.68)

where

or

(1.69)

(1.70)

(1.71)

(1.72)

Scaling:

(1.73)
50 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 1.26 Comparison of the solutions to falling body problem.

The solutions all match at low , which is not surprising, as the nonlinear drag
term is less important there. The exact, numerical, and scaling analyses all report the
same shape of the velocity behavior over time, although the scaling analysis is off
by a factor of 2 on the terminal velocity. This difference is not surprising, as the
scaling is only good to an order of magnitude (in that light, a factor of 1.4 is excel-
lent). The scaling does predict the proper dependence of the early time and terminal
velocities on the problem parameters. The approximate solution approach is more
problematic. It correctly estimates the terminal velocity ( >> 1) and the behavior at
early times ( < 1), but the truncated power series solution deviates from the exact
result at intermediate times ( ~ 1). The series used was centered on = 0 and so
works best for early times. Up to 30 terms are needed to nurse a good agreement into
later times ( ~ 3 or 4), illustrating that one must be aware of the applicability and
limitations of any solution method used.

1.5 ENGINEERING UNITS

Standards and units for weights and measures were originally developed for the pur-
poses of commerce, necessitated by the desire to get what one pays for while not
relying on trust in strangers to ensure fair trade. Historian Ken Adler described this
necessity when he wrote that “measures are a consequence of man’s fall, a human
invention for a world outside Eden, where scarcity and mistrust rule, and labor and
exchange are our lot” [35]. Over the course of history, different units have been
developed, and their usefulness, first to commerce and then to engineering and sci-
ence, has been a function of how well and widely people accepted the standards. The
importance to commerce of standard and verifiable units is the main reason that the
establishment or improvement of such standards is one of the first tasks of any new
Introduction to Transport Phenomena in Materials Processing 51

government. As an example, one of the few powers the U.S. Constitution explicitly
grants to the Congress is “To . . . fix the Standard of Weights and Measures” [36].
The earliest standardized units, linear measure (length, area, volume), weight, and
time, were developed for commercial and religious reasons. Across space and time,
there have been a vast number of these types of units, but most were based on arbitrary
but agreed-upon measures and were local in their use, usually within the jurisdiction
of a single government that could regulate their implementation. Units for mass, force,
and pressure were first defined at the time of Newton’s revolution (1600s), which
provided the theory to discuss these mechanical concepts. Later work in classical ther-
modynamics (in the 1700–1800s) defined quantitative concepts such as temperature,
energy, and power, necessitating the invention of standard units for these phenomena.
Up to this point in history, standard units typically were based on what was at
hand or could be easily replicated. As trade, technological, and religious practices
accumulated and changed over the years, the number of units grew beyond counting.
The European Age of Reason and the related (but not equivalent) Scientific Revo-
lution in the 1600s and 1700s gave rise to a tendency to uniformity in practice and
regularity in organization. One consequence of this tendency was the push to reduce
the confusion of measurement systems, with their bewildering array of local options.
This trend was also impelled by the simultaneous centralization of governmental
power in the new, large nation-states.
In France, during the ancien régime, a group of scientists in the French Royal
Academy of Sciences proposed a unified, decimal-based system, with units based on
unchanging natural phenomena. The first two standards proposed were the length of
the meter, to be defined as 1/10,000,000th of the distance between the North Pole and
the equator [35], and the gram, the mass of one cubic centimeter of water at its melt-
ing point. During the French Revolution, the Academy (no longer “Royal”) refined
the metric system but worked for several years to precisely measure these quantities.
(This process was slowed somewhat by the imprisonment or execution of several
Academy members.) The system was approved in 1795 and formally adopted in
1799 when the measurements for the meter were complete. But the people of France
were very reluctant to embrace the change in their everyday lives and widely ignored
it. In a recognition of this reality, Napoleon I in 1812 allowed the old systems to con-
tinue in business, while the metric system still was required for government-related
work. This arrangement survived the emperor and the metric system was fully not
reintroduced until 1840.
Throughout the nineteenth and twentieth centuries, these standard metric units
spread across most of the world. The level of precision in these standards was suffi-
cient for most business purposes during that time, but as technology and science have
advanced, more and more precision has been required. In 1960, the 11th General
Conference on Weights and Measures defined the Système International d’Unités
(SI), based on the metric system. The conference began a process, ending in 2019,
of using various universal constants to define the units. Five of the new base units of
interest here are defined as [37]:

• second (time): the unperturbed ground-state hyperfine transition frequency of

the caesium 133 atom is defined as 9,192,631,770 s-1;
52 An Introduction to Transport Phenomena in Materials Engineering

• meter (length): if the speed of light in a vacuum is defined as 299,792,458 m/s,

the meter is defined by that speed and the definition of the second;
• kilogram (mass): using Planck’s constant (defined as 6.62607015 × 10−34 kg
m2/s), the meter and second define the kilogram;
• ampere (electric current): taking the value of the elementary charge to be
1.602176634 × 10−19 A s, and using definition of a second, gives the ampere;
• kelvin (thermodynamic temperature): beginning with the value of the
Boltzmann constant as 1.380649 × 10−23 kg m2/s2K, the kilogram, meter,
and second define Kelvin.

The International System has been adopted almost universally, with the United States
being the only major exception, although it does use SI in some manufacturing and
more widely in technical education. Some local exceptions are occasionally made in
many countries, such as the United Kingdom still using miles along roads and pints
in pubs.
The SI units for many quantities of interest in this text, and their equivalents in the
United States Customary System of units are listed in Appendix III. In this text, the
unit systems of “mks” (meter, kilogram, second) or “cgs” (centimeter, gram, second)
will be used primarily, with temperatures in oC or K. Other, derived units of interest
in this text include the newton (N = kg m/s2) for force, the joule (J = N m = kg m2/s2)
for energy, and the watt (W = J/s = kg m2/s3) for power.

1.6 SUMMARY
Introduced in this chapter is the study of transport phenomena, the purpose of
which is to understand and predict the behavior of mass, momentum, heat, and
species movement, especially during materials processing. We show the influence
of different transport phenomena, including different modes of heat transfer (con-
duction, convection, radiation), fluid mechanics, and mass transfer, on examples
of processing of metals, ceramics, polymers, and electronic materials. Constitu-
tive equations for dependence of transport rates on different stimuli are introduced,
and these functions are discussed in terms of relationships between material prop-
erties and structure. Anticipating the use of these constitutive equations in heat,
mass, and momentum balance as models of transport mechanisms, the use of exact,
approximate, numerical and scaling methods to solve these models are compared
and contrasted. As an example, a simple mechanics problem is solved with all these
methods and the results are compared.

NOTES
1. “Fluidum est corpus omne cuius partes cedunt vi cuicunque illatæ, & cedendo facile move-
tur inter se.” [11] Author’s translation.
2. Newton actually worked out the ideas behind the viscosity law in flow around a cylinder,
but the result is the same.
3. Other common SI units for dynamic viscosity: kg ms= Ns m 2 = Pa s=10 g cm s=10 poise.
Many other units are in limited use, each based on a different technique for measuring
viscosity.
Introduction to Transport Phenomena in Materials Processing 53

4. Rheology is sometimes described as the “study of all fluid flows,” but this definition is
fairly broad and not universally accepted. It is certainly used for the study of behavior of
fluids with complex structures, but fluid mechanicians who study Newtonian flows (e.g.,
turbulent water flow past boats or liquid metal motion in casting processes) generally do
not consider themselves to be rheologists. Such is the imprecision and, one might say, flu-
idity of the boundaries of technical fields.
5. The order of magnitude of a quantity, x, is written as (x). As a rough guide, one can say
that y is “order x,” or (y) ~ (x), if nx < y < 10nx, where n may be somewhere approxi-
mately between 0.3 and 0.5.

REFERENCES
1. Polmear, I., D. St John, J.-F. Nie, and M. Qian, Light Alloys: Metallurgy of the Light
Metals, 5th ed., Butterworth-Heinemann, 2017.
2. Ashby, M., “Physical modelling of materials problems,” Materials Science and Technol-
ogy, pp. 102–111, 1992.
3. Yule, A. J., and J. J. Dunkley, Atomization of Melts for Powder Production and Spray
Deposition, Clarendon Press, 1994.
4. Lavernia, E. J., and Y. Wu, Spray Atomization and Deposition, John Wiley & Sons, 1996.
5. Francis, L. F., Materials Processing: A Unified Approach to Processing of Metals,
Ceramics and Polymers, Academic Press, 2016.
6. McMaster, R. A., D. M. Shetterly, and A. G. Bueno, “Annealed and tempered glass,” in
Engineered Materials Handbook. Volume 4: Ceramics and Glasses, S. J. Schneider, Jr.
(ed.), ASM International, 1991.
7. Thomas, B. G., “Continuous casting,” in Modeling for Casting and Solidification Pro-
cessing, K.-O. Yu (ed.), CRC Press, 2002.
8. May, G. S., and S. M. Sze, Fundamentals of Semiconductor Fabrication, John Wiley &
Sons, 2004.
9. Callister, W. D., Materials Science and Engineering: An Introduction, 3rd ed., John
Wiley & Sons, 1994.
10. Dantzig, J., and M. Rappaz, Solidification, 2nd ed., CRC Press, 2017.
11. Newton, I., Philosophiæ Naturalis Principia Mathematica, Book II, Section V, 1687.
12. Barnes, H. A., J. F. Hutton, and K. Walters, An Introduction to Rheology, Elsevier, 1989.
13. Newton, I., Philosophiæ Naturalis Principia Mathematica, Book II, Section IX, 1687.
14. Panton, R. L., Incompressible Flow, 2nd ed., John Wiley & Sons, 1996.
15. Rosen, S. L., Fundamental Principles of Polymeric Materials, 2nd ed., John Wiley &
Sons, 1993.
16. Schlichting, H., Boundary-Layer Theory, 7th ed. (trans. J. Kestin), McGraw-Hill, 1979.
17. Chhabra, R. P., and J. F. Richardson, Non-Newtonian Flow and Applied Rheology: Engi-
neering Applications, 2nd ed., Elsevier, 2008.
18. Callister, W. D., and D. G. Rethwisch, Callister’s Materials Science and Engineering,
10th ed., John Wiley & Sons, 2020.
19. Cross, M. M., “Rheology of non-Newtonian fluids: A new flow equation for pseudoplas-
tic systems,” Journal of Colloid Science, v. 20, pp. 417–437, 1965.
20. Carreau, P. J., “Rheological equations from molecular network theories,” Transactions
of the Society of Rheology, v. 16, p. 99, 1972.
21. Williams, M. L., R. F. Landel, and J. D. Ferry, “The temperature dependence of relax-
ation mechanisms in amorphous polymers and other glass-forming liquids,” Journal of
the American Chemical Society, v. 77, pp. 3701–3707, 1955.
22. Gaskell, D. R., Introduction to the Thermodynamics of Materials, 5th ed., Taylor and
Francis, 2008.
54 An Introduction to Transport Phenomena in Materials Engineering
 2 Steady State Conduction
Heat Transfer

2.1 INTRODUCTION
The transfer of heat by conduction was introduced in Chapter 1, both in its physi-
cal mechanisms and the experimentally observed relationship between heat flux and
temperature gradient that makes up Fourier’s law:

(1.28)
In this chapter, we will derive a transient, three-dimensional partial differential equa-
tion for temperature with thermal storage, heat diffusion, and volumetric heat gener-
ation, based on the first law of thermodynamics, Fourier’s law, and several idealized
boundary conditions. Several solution approaches for this equation in steady state are
presented to approximate the thermal behavior of real systems. To begin, however,
we introduce the idea of thermal resistance based on the heat transfer rate equations
in Chapter 1, and construct heat flow models using thermal resistance networks.

2.2 THERMAL RESISTANCES

We begin the study of heat transfer by modeling systems with a simple approach
using resistance networks, based on an analogy with Ohm’s law for electrical circuits:

(2.1)

where I is the current flowing through the resistors (with overall resistance R), driven
by the voltage drop, V. The thermal analogue has heat flow (q) and a potential drop,
the temperature difference, T. Expressions for the thermal resistances are found
here from the heat flow rate equations in Chapter 1.
In an electrical resistance network, the potential (voltage) is only known at the
nodes of the network, and the overall resistance represents the effect of all the indi-
vidual resistances between the nodes. Similarly, we only model the temperatures at
the nodes in the thermal network and treat the resistances between them. In addition,
all heat inputs and outputs occur at nodes (not in the middle of a resistor) and all
models will be steady state.

2.2.1 conduction resistance

For one-dimensional conduction, the rate equation can be written in terms of the heat flow,

(2.2)

DOI: 10.1201/9781003104278-2 55
56 An Introduction to Transport Phenomena in Materials Engineering

in Cartesian coordinates. If we assume a uniform k and q, we can separate variables

in Eq. (2.2),

(2.3)

(2.4)

x, so we can com-
plete the integration.

(2.5)

written as

(2.6)

FIGURE 2.1 Schematic of steady conduction heat transfer through a plane wall.
Steady State Conduction Heat Transfer 57

We see that we can rewrite Eq. (2.5) in a way similar to Ohm’s law,

(2.7)

Here the potential is -

(2.8)

The area normal to the heat flow is Ax and L = x2 – x1. We see in Eq. (2.8) that a thicker
slab, a lower thermal conductivity, and/or a smaller cross-sectional area will increase
the thermal resistance.

Example 2.1
Conduction through a Glass Window
The temperatures of the inner and outer surfaces of a glass window in a room are
Ti = 25 oC and To = 0 oC. The glass is L = 5 mm thick, has an area normal to heat flow
of 1 m2, and has a mean thermal conductivity of k = 0.84 W/mK.

Find: (a) the thermal resistance of the window; (b) the rate of heat loss from
the room through the window; and (c) the glass thickness (L) required to
decrease the heat flow to 2000 W.

Solution

We can perform a similar analysis on the annulus (the space between two concentric
cylinders) in Figure 2.2. The conduction resistance in cylindrical coordinates is a bit
more complicated than for the planar wall, because the cross-sectional area normal
to the heat flow increases with radius. The heat flow is entirely radial (neglecting end
effects by assuming L>>r2
58 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.2 Schematic of steady conduction heat transfer through an annulus.

The expression of Fourier’s law of conduction (the rate equation for conduction) in
radial coordinates is

(2.9)

(2.10)

thickness of the pipe wall, as either increasing r2 or decreasing r1 will increase the
r2/r1 ratio. We could also force the heat flow through a smaller area (increasing heat
flux, q”) by shortening the pipe length (L). Finally, we could decrease the thermal
conductivity by changing materials.
Steady State Conduction Heat Transfer 59

Example 2.2
Radial Conduction through a Glass Tube
Hot water flows through a glass tube of length L = 50 cm, inner radius r1 = 3 cm, and
outer radius r2 = 5 cm. The temperatures of the inner and outer surfaces are T1 = 90
°C and T2 = 85 °C, and the glass thermal conductivity is k = 0.84 W/mK.

Find: (a) the thermal resistance of the tube in the radial direction; (b) the radial
heat loss rate; and (c) the heat loss rate if the thickness of the annulus is
doubled, while keeping r1 constant.

Solution

In contrast to the plane wall in Example 2.1, doubling the tube thickness here does
not double the resistance or halve the heat flow. The resistance increases by only a
factor of 1.66. The area through which the heat flows increases with the radius and
so the resistance of the added material is lower than the original. A slab with uniform
cross-sectional area has no analogous behavior.

2.2.2 convection resistance

An expression for the thermal resistance due to convection on a surface is found from
rewriting the convection rate equation

(1.31)
60 An Introduction to Transport Phenomena in Materials Engineering

as

(2.11)

where As and Ts are the surface area and temperature, respectively. The potential for

(2.12)

The resistance can be lowered by the addition of surface area or by increasing the
convective heat transfer coefficient. Unlike conduction, the convection resistance
always takes the same form regardless of geometry. Changes in geometry influence
Rconv through the determination of the heat transfer coefficient, and methods for con-
trolling h are presented in Chapter 10.

2.2.3 radiation resistance

Radiation heat transfer between a surface (at Ts) and a much larger environment (at
T∞) may be characterized by the rate equation presented in Chapter 1, combining
Eqs. (1.37)–(1.39).

Eq. (2.14) shows that radiation resistance is a very strong function of the surface and
environmental temperatures and raising either or both reduces the surface resistance.
Increasing either the total surface area, As, or the emissivity, , also will increase heat
flow from the surface to the environment.
We should recognize that the rate equation for radiation is not based on a linear
temperature difference as a potential between the two surfaces, but on a net exchange
of radiation. Such a net flow of heat is governed by a different potential (radiosity,
the total radiation leaving a surface), which will be explained at length in Chapter 11.
To fit radiation effects into a pattern using T as the potential driving heat flow, Eq.
(2.14) is a reasonable approximation.

2.2.4 interFace resistance

When two materials are put in contact, the simplest assumption is that there is no
temperature drop at the interface as heat flows through it, i.e., there is perfect ther-
mal contact at the material interface. However, there are several reasons why real
Steady State Conduction Heat Transfer 61

surfaces do not have perfect thermal contact, including (i) different shapes that do
not fit together on the macroscopic scale (nonconforming interfaces); (ii) roughness,
or microscopically uneven surfaces; or (iii) both.
Two macroscopically flat, conforming surfaces with microscopic roughness are
shown in Figure 2.3(a). When pressed together, as in Figure 2.3(b), there are areas
of contact between the two surfaces, but gaps remain. In [1], Yovanovich reviews a
plethora of models of thermal resistance at such joints, for a wide variety of combi-
nations of conditions: conforming and nonconforming interfaces, smooth and rough
surfaces, and with and without filler material in the gaps. The details of these models
are beyond the scope of this text, but they do elucidate significant dependencies of

FIGURE 2.3 Surface roughness effects at a conforming interface. (a) Two microscopically
rough surfaces before contact. (b) Rough surface in contact and under pressure. (c) Temperature
profile influenced by interface resistance, showing the T at the interface.
62 An Introduction to Transport Phenomena in Materials Engineering

the important resistances on the characteristics of the surfaces and the materials.
Three general types of mechanisms affecting the interface thermal resistance are
included in these models.
The first type of effect is the interface morphology. If the two surfaces are
not the same size and shape (nonconformity), there is an ill-fitting interface with
macroscopic gaps between the surfaces, reducing the heat transfer. The resist-
ance across the gaps is governed by conduction and radiation heat transfers which
are represented by the one dimensional resistances, Eqs. (2.8) and (2.14). At the
microscale, e.g. Figure 2.3(b), a one-dimensional heat flow perpendicular to
the interface becomes two dimensional as the heat approaches the joint. Because
the heat travels farther and the cross-sectional area decreases, Eq. (2.8) suggests a
higher interface resistance, known as a constriction resistance, due to the complex
local geometry.
The mechanical behavior of the materials also influences the joint thermal resist-
ance; Figure 2.3(b) has been sketched assuming some plastic deformation in both
surfaces. They could both be stronger and more brittle, which would limit the sol-
id-solid contact area almost to point contacts at the asperities, or, at the other extreme,
they both could be ductile enough to flow and fill all the gaps. The deformation of the
asperities and the final solid contact area are a strong function of the clamping pres-
sure, as well as the elastic and plastic mechanical properties of both materials. More
deformation forms more contact area, making the heat path more one-dimensional
and lowering the constriction resistance.
The third class of mechanism contributing to the overall interface resistance is
heat transfer across the gaps. These gaps may be empty (if the joint was prepared
under vacuum), in which case the only mode of heat flow is radiation. The resist-
ance is then a function of the local temperatures on either side of the void. The gaps
may be filled with a gas (e.g., air, or the more conductive argon or helium), which,
at lower temperatures, would have conduction as the dominant heat transfer mode.
Intermediate cases with transparent gases and higher temperatures are also common,
with the conduction and radiation paths in parallel. To reduce the gap conduction
resistance significantly, addition of thermally conductive compounds (often known
as “thermal grease”) is a common practice, especially in microelectronic applica-
tions. The removal of the parallel radiation path is more than compensated by the
increase of the gap material conductivity by several orders of magnitude. The use of
a malleable metal foil (e.g., indium or silver) to flow under pressure into the small
voids is a logical conclusion of this approach. Such techniques and associated mod-
els are reviewed in [2].
All of these physical effects may contribute to the overall interface thermal resist-
ance. The effect of having an imperfect contact can be seen in Figure 2.3(c), which
shows temperature profiles in two solids with a steep temperature drop at the inter-
face. This decrease is confined to the small region of the joint in which the afore-
mentioned physical mechanisms dominate local heat transfer. The overall interface

(2.15)
Steady State Conduction Heat Transfer 63

TABLE 2.1
Measured Thermal Interface Resistances ( Ri ) for Static Joints
Material 1 Material 2 P (MPa) T (oC) Filler
Ri (Km2/W) (×10–5) Ref
Fabric laminated phenolic resin 0.003 20 air 33 [3]
Fabric laminated phenolic resin 0.005 20 air 25 [3]
Copper 6.7 45 vacuum 8.4 [4]
Tungsten Graphite 6.7 130 air 8.4 [4]
Aluminum 0.1 air 27.5 [5]
Aluminum 0.1 helium 10.5 [5]
Aluminum 0.1 silicone oil 5.3 [5]
Brass Mild steel 0.1 20 air 20–100 [3]
SS304 .01 30 air 500 [3]
SS304 .5 30 air 100 [3]
Al 6061-T4 .01 70 thermal grease 4 [3]
Al 6061-T4 .05 70 thermal grease 2 [3]

Here TA and TB are the limits of the temperatures at the interface in materi-
als A and B as the interface is approached from either side, hint is the interface
heat transfer coefficient or interface conductance, and A is the total macroscopic
area of the joint, including both gaps and contact regions. The contact resistance
of an interface is defined as the overall interface resistance multiplied by the
surface area,

(2.16)

Representative values of interface contact resistance are found in Table 2.1.

The configuration assumed up to now is static contact, where the surfaces are not
moving relative to each other. We can also consider dynamic contact, during which
the surfaces touch only for a short period of time and may be moving relative to each
other while in contact; examples include forging and rolling processes. Selections
of measured thermal interface resistances are found Table 2.2, for dynamic contact
during materials processing.

2.3 RESISTANCE NETWORKS

The resistances to heat transport discussed in the previous section are useful models
to understand the relationships between a temperature difference and a heat flow
in various configurations and by different mechanisms. They are even more useful
when combined into thermal resistance networks that model systems, including sev-
eral resistances. These equivalent thermal circuits follow the same rules as simple
electrical circuits, namely Ohm’s and Kirchhoff’s laws. Ohm’s law has been covered
in Eq. (2.1). Kirchhoff’s law applied to a thermal network is conservation of heat
flow at a node. It states that all heat flowing into a node must flow out again, given
64 An Introduction to Transport Phenomena in Materials Engineering

TABLE 2.2
Measured Thermal Interface Resistances ( Ri ) for Dynamic Contacts in
Materials Processing
Material 1 Material 2 P (MPa) T (oC) Filler
Ri (Km2/W) Ref
(×10–5)
Forging
Al6061-O H12 tool steel 0 400 MoS2 lubricant 50 [6]
Al6061-O H12 tool steel 14 400 MoS2 lubricant 8.3 [6]
Al6061-O H12 tool steel 28 400 MoS2 lubricant 4.8 [6]
Injection molding
Polypropylene metal mold 2 air 40 [7]
Polypropylene metal mold 16 air 6 [7]
Plasma sprayed droplet impacting surface
(T is substrate temperature)
Mo (drop) In625 (substrate) 27 air 1.9 ± 0.15 [8]
Mo (drop) In625 (substrate) 400 air 0.12 ± 0.02 [8]
Zirconia (drop) glass (substrate) 27 air 2.2 ± 0.3 [8]
Zirconia (drop) glass (substrate) 400 air 0.1 ± 0.03 [8]
hot stamping
Usibor 1500PTM unspecified tool steel 10 200–400 Al-Si alloy 55 ± 5 [9]
(steel coating)
low pressure die casting
A356 H13 tool steel 7 300 ceramic die coating 52.6 [10]
A356 H13 tool steel 14 300 ceramic die coating 45.5 [10]
A356 H13 tool steel 21 300 ceramic die coating 37.7 [10]

that the system is assumed to be at steady state (i.e., it has no capacitance). So, at a
node with N paths connected to it,

(2.17)

Generally, nodes are placed at surfaces, with resistors for conduction across materi-
als and resistors for convection, radiation, and contact resistance placed at material
interfaces.

Example 2.3
Heat Loss through a Furnace Window
A glass window is placed in a furnace wall as a port to view the load during process-
ing (Figure 2.4). Heat is transferred into and out of the glass by convection on both
sides. Conduction is the mode of heat transfer in the glass, and we assume radiation
heat transfer is negligible. For now we assume the two heat transfer coefficients, hf
and hp, the furnace and plant temperatures, Tf and Tp, and the window surface area,
Aw, are known.
Steady State Conduction Heat Transfer 65

FIGURE 2.4 Schematic of heat loss through a furnace window.

FIGURE 2.5 Thermal resistance network for furnace window in Figure 2.4.

When designing such a window, we should estimate the penalty for having such a
window, in terms of the heat loss (q) through it. We may also wish to know, for safety
reasons, the outside surface temperature of the window (To).

Find: (a) an expression for the total heat loss, q; and (b) the outside window
temperature, To.
Solution: To begin the analysis, we approximate this system by three resis-
tances: (i) convection from furnace gas to window, (ii) conduction through
the window, and (iii) convection from window to air in the plant. The cor-
responding thermal resistance network is shown in Figure 2.5.

Note that a complete drawing for a thermal network includes labels for all resis-
tances and for the temperature at each node. It will also show where heat flows enter
and exit the network. The direction of the flow is not known for certain until values
for a specific furnace are used, so a reasonable guess is made as a sign convention.
A negative value of q will be calculated where the actual heat flow moves opposite
to the assumed direction.
66 An Introduction to Transport Phenomena in Materials Engineering

Kirchhoff’s law applied at any node is trivial (qin = q = qout) because there is only
one path and the heat flow is the same everywhere. We can also see that, if RTOT is
independent of the window temperatures (the usual first approximation), then the

(2.18)

using our definitions of conduction and convection resistances, Eqs. (2.8) and (2.12).
How could we decrease the heat flow? Many possibilities occur to us while exam-
ining the result in Eq. (2.18). Increasing resistance and decreasing potential are the
two choices. Having a smaller window (Aw ) is an obvious solution, as is choosing
a window that is thicker (L ) or a material with a lower thermal conductivity (k ).
Decreasing the heat transfer coefficients also may be effective, and ways to do so are
found in Chapter 10. One might also suggest reducing the overall temperature differ-
ence, but this value is usually determined by the requirements of furnace operations
and worker comfort and safety in the plant.

(b) Once the heat loss is known, the window temperatures can be found by
applying Ohm’s law to pieces of the circuit. To write an expression for To,
we examine only the convection outside of the window and equate the heat

Given that Tf and Tp are fixed, and changing hf might alter furnace performance, how
do we lower the outside window temperature? Changing the window area has no
effect, but a thicker or lower k of the window helps. Another alteration that might be
suggested is to cool actively the outside window by blowing air on it and so increas-
ing hp. However, while To falls as hp increases, our solution for q, Eq. (2.18), shows
that the heat loss will also increase.
The purpose of this example is to illustrate the use of a thermal resistance network
to quickly describe gross system behavior. While crudely approximate, it does show
the general dependencies of quantities of interest (e.g., To, q) on parameters we can
Steady State Conduction Heat Transfer 67

control (k, L, hp, Aw). Surely that is what we typically require from a first-order engi-
neering model.

Example 2.3 shows the usefulness of making simple resistance networks to predict
general thermal behavior, limited to steady conditions with uniform properties and
a coarse spatial resolution. These crude models based on such networks can provide
estimates for the thermal resistances governing that behavior as well as the func-
tional dependence of resistances on the material properties, geometry, and boundary
conditions.

2.4 GENERAL HEAT CONDUCTION EQUATION

The previous sections develop a simple approach that estimates heat flows and tem-
peratures in terms of thermal resistances, but only predicts the temperatures at spe-
cific locations. During the processing of many materials, more detailed information
is needed. Phase transformations and microstructural development as a function of
position in a material are determined by the temperature history and spatial distribu-
tion, T(x, y, z, t). These data are also useful in predicting the development and relaxa-
tion of thermal strains and stresses. The temperature field also can be used to predict
the pattern of heat flow and so can inform process and product design.
To find the temperature field, we apply the first law of thermodynamics to an arbi-
trary control volume in space and track energy flows in that volume, as suggested in
Section 1.1. This general control volume is shown in Figure 2.6, and the balance of
thermal energy rates is

U in U out U gen Ustor . (2.19)

The first two terms, Uin U out , are the net rate of thermal energy transfer across the
surface area of the volume. Another contribution to this balance, U gen , represents heat
generation inside the volume due, for example, to electrical current running through
a resistor (Joule heating) or chemical energy changing to thermal during a reaction.

FIGURE 2.6 Thermal energy flows in an arbitrary control volume.

68 An Introduction to Transport Phenomena in Materials Engineering

If the sum of all the heat flows across the surface and the heat generation is not zero,
then thermal energy is either stored or depleted and U stor 0 .
The thermal energy balance is repeated for a two-dimensional flow in and through
a differential control volume in Cartesian coordinates in Figure 2.7. Given the depth
into the page is defined as dz, the differential volume is d = dx dy dz. By a sign

(2.20)

Assuming that conduction is the only mode of heat transfer across the boundaries,
then the conduction rate equation gives

(2.21)

The conduction out in the x direction, qx+dx, can be rewritten in terms of the enter-
ing heat flow, qx, by the use of a Taylor series in the vicinity of x:

(2.22)

(2.23)

FIGURE 2.7 Differential Cartesian control volume for thermal energy balance.
Steady State Conduction Heat Transfer 69

(2.24)

(2.25)

(2.26)

(2.27)

(2.28)

(2.29)

(2.30)
70 An Introduction to Transport Phenomena in Materials Engineering

Dividing by the differential volume, dx dy dz, gives the general form of a partial dif-
ferential equation for temperature, assuming that conduction is the only heat transfer
mode at the boundaries:

(2.31)

(2.32)

(2.33)

(2.34)

(2.35)

(2.36)
Steady State Conduction Heat Transfer 71

(2.37)

Both Eqs. (2.34) and (2.37) are equivalent to a more general form of the energy
equation:

(2.38)

2.5 HEAT TRANSFER BOUNDARY CONDITIONS

In whatever form we find the energy conservation equation, its solution requires
an initial temperature field (if transient) and several boundary conditions. A model
of a system may be one-, two-, or three-dimensional and needs two constraints on
temperature in each thermally active direction. In this section, the forms of three
common boundary conditions are presented.
The reader must remember that the conditions described here, while mathe-
matically convenient, are idealizations. We will describe where each form is more
applicable, but boundaries are seldom if ever as uniform and unvarying as these
approximations. While the mathematics may be exact, reality is messier. (This prob-
lem can be lessened by the use of numerical solutions, but it is not eliminated.)
However, these simple models of reality applied to the energy conservation equation
are useful for finding reasonable and useful estimates of temperature histories and
distributions.
The boundary condition of the first kind is that of a specified temperature, To, at a
specified location, xo:

(2.39)

xo to be uniform,
T(xo) to

A specified heat flux, qo

kind
qo at the boundary. As

(2.40)
72 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.8 Heat transfer boundary conditions: (a) specified temperature; (b) specified
flux; (c) zero flux; and (d) specified relationship between temperature and heat flux. (The gray
areas are outside the domain of interest and m is the slope of the T(x) at the boundary.)

Such a condition is a good approximation if heat is generated at the surface (as in

,
Figure 2.8(c).

(2.41)

Symmetry in a temperature field at xo is an exact example of this condition, as

seen by the dashed line in Figure 2.8(c). An insulated (adiabatic) boundary can be
approximated as zero flux, if the thermal conductivity of the domain is much higher
than the conductivity of the surroundings (kd >> ks). The heat flux is continuous
across the boundary, so
Steady State Conduction Heat Transfer 73

where “+” indicates the side inside the domain and “-” the surroundings. Rearrang-
ing, we get

) is relatively very small.

specified

or

(2.42)

This boundary condition of the third kind is shown in Figure 2.8(d). The resistance to
heat flow is inversely proportional to the heat transfer coefficient, h. This boundary
condition is a more general case in that there is a finite, non-zero resistance at xo.
Looking at the extremes in values of h, we reduce Eq. (2.42) to a zero-flux case, Eq.
(2.41), when h 0 (resistance ∞) and to the prescribed temperature (T(xo) T∞)
case, Eq. (2.39), when h ∞ (resistance 0).

2.6 ONE-DIMENSIONAL HEAT CONDUCTION

Armed with a thermal energy conservation equation, Eq. (2.38), and the appropriate
number of boundary conditions from the previous section, we can begin finding tem-
perature fields and heat flows due to conduction. The first problem to be solved will
be what must be the simplest problem in heat transfer, finding the steady temperature
field in a one-dimensional slab with uniform properties and two fixed temperature

(2.43)

to remind us that there is no change in temperature gradient or heat flux, if properties

are uniform.
74 An Introduction to Transport Phenomena in Materials Engineering

As shown in Figure 2.9(a), the boundary conditions are

(2.44)

(2.45)

(2.46)

The heat flux is uniform across the domain.

The solution in Eq. (2.45) shows the temperature as a function of three parameters
(To, TL, L). Throughout this text, the practice of nondimensionalization of equations,
boundary conditions, and solutions will be used often to reduce the solution param-
eter space. These transformations give more general solutions, valid for all combi-
nations of the dimensional parameters in the original problem. In this case, we can
define a nondimensional temperature and length as

(2.47)

Solving for T and x,

which are substituted into the original governing equation and boundary conditions,
Eqs. (2.43) and (2.44).
 -

or

The two solutions are plotted in Figure 2.9. Given the nondimensionalization of the
problem, the curve in Figure 2.9(b) is always the same, whether TL > To or TL < To.
In a slightly more complicated problem, we can examine the steady tempera-
ture field inside a tube wall with a convection condition (boundary condition of the
third kind) at the outside surface and a fixed temperature at the inside surface, as in
Figure 2.10. This configuration is a reasonable model for a tube with a conductive
liquid flowing strongly inside, so the interior convection resistance is very small.
-

or

(2.49)

(2.50)
76 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.9 Temperature fields for simple steady conduction in a slab. (a) Dimensional
solutions, assuming To < TL. (b) Dimensionless solution, regardless of relative values of
To and TL.

The two boundary conditions are the isothermal inside wall,

(2.51)

(2.52)
Steady State Conduction Heat Transfer 77

FIGURE 2.10 Schematic of geometry and boundary condition for heat flow through a
tube wall.

(2.53)

(2.54)

(2.55)

because the area normal to the flux is increasing with

78 An Introduction to Transport Phenomena in Materials Engineering

r (A = 2πrL, where L is the length of the tube). Note that the heat flow is still
uniform in r as

, R1, R2).

R2

h 0) is

-
h, T1,
T∞ -

(2.56)

-
Steady State Conduction Heat Transfer 79

and substituting into Eqs. (2.49), (2.51), and (2.52), we have the nondimensional

(2.57)

(2.58)

(2.59)

(2.60)

This solution can be more easily plotted. Figure 2.11 shows temperature profiles
with 1 = 0.2 and various Bi values. This family of curves has individual lines cor-
responding to specific values of Bi. The plot shows that, under all circumstances in
the stated problem, the temperature is held fixed at the inner radius, ( = 1) = 1 or
T(r = R1) = T1, according to the boundary condition. In low convection resistance
cases (Bi ∞), the outer surface temperature approaches the environmental condi-
tion, ( = 1) 0 or (r = R2) T∞. As the convection resistance increases (Bi 0),
the body becomes isothermal as the entire tube wall goes = 1 or T(r) = T1.
Another type of one-dimensional, steady problem arises if there are two or more
materials along the heat flow path. We could easily estimate thermal behavior with
resistance networks, but the solution to the energy conservation equation will provide
more detail about the temperature field. As an example, here we find the solution to
the behavior of an oven wall with a low thermal conductivity ceramic layer supported
by an outer steel shell, as in Figure 2.12.
If the inside temperature is at T(x = −Lc) = To, the outside is T(x = Ls) = T∞, and
To > T∞, then we know two boundary conditions in x, which we might think enough
to solve the one-dimensional, steady conduction equation without heat generation.
80 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.11 Nondimensional temperature in the tube wall for 1 = 0.2 and various Bi.

FIGURE 2.12 Temperature field in a two material oven wall.

Alas, it is not. The energy equation, Eq. (2.43), assumes uniform properties in the
domain, which is not true here because kc ≠ ks. The mathematically simplest solution
is to write the energy equation for the two materials separately:

(2.61)
Steady State Conduction Heat Transfer 81

where TS is the steel temperature and TC is in the ceramic. With two equations, we
now need four boundary conditions, so this neat division does raise an issue: we
only know the temperatures at the outer boundaries, so where do we find two more
boundary conditions?
While we cannot write what happens at the interface (x = 0) exactly in terms of
the three conditions in Section 2.5, two boundary conditions can be obtained by
assuming perfect contact between the steel and ceramic. The conditions are a contin-
uous temperature (no contact resistance) and a continuous heat flow between the two
materials. These conditions at the interface are written as

s ∞
is approximately isothermal at the outside temperature.
82 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.13 Influence of the temperature dependence of thermal conductivity on the tem-
perature profile in a plane wall during heat transfer by conduction.

Up to this point, the thermal conductivity has been assumed constant and uniform,
whereas it may vary significantly with temperature. Also, sometimes a geometry
is such that, while the cross-sectional area varies with position, it changes gently
enough that the system can be treated as one-dimensional. Imagine that the solid
slab shown in Figure 2.1 is of a material with a thermal conductivity that increases
with increasing temperature; the temperature profile through the slab is shown by
curve (a) in Figure 2.13. If the values of q and A are uniform across the slab, then the
product k (dT/dx) is also uniform, and thus k decreases with decreasing temperature
and dT/dx must increase. Similarly, if k decreases with increasing temperature, the
temperature profile is as shown by curve (b) in Figure 2.13.

Any variation of the cross-sectional area through which heat is conducted, A, with
distance, x, must also be taken into account in the integration of Eq. (2.21). Both
effects are accounted for by writing the conduction rate equation as

(2.65)

(2.66)
Steady State Conduction Heat Transfer 83

(2.67)

(2.68)

or

(2.69)

x gives

(2.70)

(2.71)

a m
extends from x1 = 0.1 m to x2 = 0.5 m.
84 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.14 Unidirectional heat conduction in conically shaped conductor in Example 2.4.

Find: The heat flux and temperature profile for both (a) T1 = 1000 K at x1 and
T 2 = 500 K at x2 and (b) T1 = 500 K at x1 and T 2 = 1000 K at x2.

which is shown as curve (b) in Figure 2.15. This figure also illustrates the influence of
a change in cross-sectional area on the temperature gradient; for heat flow in the +x
direction, A increases with x and hence dT/dx decreases with x, and for heat flow in
the -x direction, A decreases and hence dT/dx increases in the heat flow direction. If
the variation of A with x is significant enough that the heat flow has a radial compo-
nent, the problem can no longer be considered as one of simple unidirectional flow.
Steady State Conduction Heat Transfer 85

FIGURE 2.15 Temperature profiles in the conical conductor shown in Figure 2.14.

2.7 CONDUCTION WITH HEAT GENERATION

For a one-dimensional, steady heat flow in a plane slab with internal heat generation,

(2.72)
86 An Introduction to Transport Phenomena in Materials Engineering

(2.73)

(2.74)

the higher

(2.75)

(2.76)

(2.77)

(2.78)

where the negative sign indicates that the heat flow is in the -x direction. Thus, as
required by energy conservation, the sum of the magnitudes of the heat fluxes given
by Eqs. (2.77) and (2.78) gives Eq. (2.75).
Steady State Conduction Heat Transfer 87

Example 2.5
Steady Conduction in Slab with Heat Generation and Convection Losses
Consider a plane slab of thickness 2L = 0.1 m, which generates heat at a rate of 250
kW/m3 and sheds it convectively to an environment at T∞ =15 °C.

h = 60 W/m2K k = 25 W/mK.

Find: The temperature profile in the slab and Tmax.

Solution: As the heat transfer conditions are identical at both faces of the slab,
the rate at which heat is transferred through unit surface area of the slab, q”,
is equal to the rate of heat generation in the volume, = LA:
q
q q L.
A

This profile is plotted in Figure 2.16 and the maximum temperature Tmax = T(x=0)=
236 oC.
The heat flow balance is seen in the first equation in this section, where the heat
generation rate inside the slab is the same as the rate of conduction out. For a higher
q, the temperature gradient at the surfaces (x = ±L) must be higher to maintain a
higher conduction rate out.

FIGURE 2.16 Temperature profile in a slab with heat generation.

88 An Introduction to Transport Phenomena in Materials Engineering

The passage of electrical current through matter causes heat generation in that body,
the rate of which is related to the electrical conductivity of the material. This resist-
ance to current flow is the source of heat dissipated by a mobile phone or a high ten-
sion transmission line. Here, we will examine the steady state relationship between
that heat generation and a convective heat loss to the environment, as well as the
resulting temperature field in the conductive body (Figure 2.17).
The governing equation for this problem is the energy equation, here written in
cylindrical coordinates:

(2.79)

),

, s

(2.80)

FIGURE 2.17 System geometry for heat generation in a wire.

Steady State Conduction Heat Transfer 89

Two boundary conditions are required to solve this equation. The first is at the center-
line (r = 0), around which the system is symmetrical, so
dT
0. (2.81)
dr r 0

(2.82)

(2.83)

(2.84)

and

C2.
90 An Introduction to Transport Phenomena in Materials Engineering

(2.85)

-
dinate using

, and

where Bi = h R/k. This solution is plotted in Figure 2.18.

Examining the solution and the plot, we see certain features of the curves. First,
we notice that the slope is always zero at the symmetry boundary, = 0. The slope
is negative throughout, down to the edge of the body, = 1, where all the heat
generated leaves the wire. For lower convection resistance (where h and Bi are large),

FIGURE 2.18 Temperature profiles for a heated wire for a range of Bi values.
Steady State Conduction Heat Transfer 91

the temperature of the body is lower, as the heat generated is lost more easily to the
environment. In the limit of infinite heat transfer coefficient, the surface temperature
becomes the same as the environment (T∞). The opposite is true as the convection
resistance increases and the heat requires a larger potential, T(L) – T∞, to force the
heat out of the body. In the limit of h = 0, there is no escape for the heat. Under this
condition, the temperature in the body is infinite at steady state, which means there
is no steady solution.

2.8 MULTIDIMENSIONAL CONDUCTION

The preceding discussion has been confined to unidirectional conduction heat flow
either in the x-direction through a plane wall or radially through a cylinder or annu-
lus. These cases show that heat flows in a direction normal to the lines of constant
temperature (the isotherms) in the solid. In Figure 2.19(a), the isotherms are straight
lines parallel to the faces of the plane wall (T = T(x) only), and in Figure 2.19(b) the
isotherms are concentric circles (T = T(r) only). Figure 2.19(c) shows the isotherms
and the directions of heat flow in a square plate, the top edge of which is maintained
at the constant nondimensional temperature, = 1, and the other three edges of which
are set to = 0. The direction of conduction heat flow is such that it is always per-
pendicular to the isotherms, and hence the heat flow is two dimensional. Thus, the
positions of the isotherms, which are determined by the temperatures of the edges,
define lanes through which heat flows and there is no transfer of heat between lanes.
In a two- or three-dimensional geometry, the heat flow is a vector in the direction of
the temperature gradient and must be considered in terms of its components in the
multiple directions.
There are many approaches to solving multidimensional conduction problems,
some of which we will describe briefly in this section. First, we will demonstrate
scaling analysis, and then return to examples of thermal networks to which some
two-dimensional geometries may be reduced to the equivalent one-dimensional
resistances. Finally, some exposure to exact solutions of Eq. (2.31) is given. Of
course, all of the methods represent approximations of one sort or another. The
most obvious is the approximate solution of a well-posed, exact mathematical
form, but even those model formulations are only crude representations of reality.
The methods here produce our “five o’clock solutions,” which may be augmented
by numerical solutions if need be. All in all, the purpose of this section is to
demonstrate some simple, approximate solutions for temperatures and heat flows
in multidimensional conduction problems. More precise and detailed answers
may be obtained for more complex geometries and boundary conditions using
numerical methods.

2.8.1 scaling analysis

As a first example of scaling analysis in multidimensional systems, let us consider
a case of a square body (H × H), with convection cooling on the top (h, T∞) and the
other three sides held at To (> T∞), as in Figure 2.20. A general scaling procedure is
outlined in Section 1.4.3.
92 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.19 Isotherms and heat flow lines in (a) a plane wall, (b) a cylinder, and (c) a
simple system undergoing heat flow in two directions.
Steady State Conduction Heat Transfer 93

FIGURE 2.20 Square body cooled on the top by convection.

Starting with the steady, two-dimensional steady energy equation with no heat

(2.86)

(2.87)

(H), so

(2.88)

-
where, but it is not particularly helpful.
More promising is an examination of the top convection boundary condition
94 An Introduction to Transport Phenomena in Materials Engineering

If we represent the surface temperature (Ts) with Tmin, scaling this boundary condi-
tion gives

(2.89)

(2.90)

where Eq. (2.90) is an estimate of the minimum temperature of the body based on
the Biot number (BiH). If the heat transfer coefficient is very large (no convection
resistance, as BiH ∞), then Tmin T∞. When the top is insulated (h and BiH 0),
then the entire body is at To.
The estimate for temperatures in Eq. (2.90) is used to find the heat loss rate sup-

(2.91)

Using

(2.92)

Again we look at two extreme cases, the first of which gives no heat flow (q 0)
when h 0. When there is no convection resistance on the top (h ∞), the heat

(2.93)
Steady State Conduction Heat Transfer 95

In this case, the top temperature is T∞, so this equation shows that all the resistance
is due to conduction in the body.
Another example of the use of scaling analysis is to obtain estimates of heat flow
during multidimensional, steady conduction in a rectangular body with uniform
heat generation and a uniform, fixed temperature, To, around its perimeter. The three
configurations used with different aspect ratios (H/L) are shown in Figure 2.21. In
this example, we are interested in an estimate of the maximum temperature rise in
the body.
The temperature reference is chosen as the maximum temperature difference in
each body,

(2.94)

FIGURE 2.21 Two-dimensional bodies at different aspect ratios with uniform heat genera-
tion and isothermal boundaries. (a) H/L = 1, (b) H/L < 1, and (c) H/L << 1.
96 An Introduction to Transport Phenomena in Materials Engineering

and, due to symmetry, Tmax is located at the center of each body. The distances in the
x and y directions over which Eq. (2.94) applies are

(2.95)

as

(2.96)

(2.97)

or

(2.98)

(2.100)

Notice the subtle difference between this result and the square case, Eq. (2.99). While
most of the surface area in Figure 2.21 (c) is on the long horizontal surfaces, the
length of the body in the x direction (L) is not the important geometric parameter in
determining Tmax. Comparing the conduction terms in Eq. (2.96),
Steady State Conduction Heat Transfer 97

so the conduction out the short ends can be neglected. There is very much less resist-
ance to conduction over (Tmax – To) in the vertical direction than in the horizontal, so
that lesser resistance controls the heat flow and only its length scale (H) influences
the value of the maximum temperature.

2.8.2 two-diMensional patHs in resistance networks:

brancHes and sHape Factors
Another approach to modeling conduction through multidimensional domains is the
use of thermal resistance networks, as introduced in Section 2.3. In a simple example
of a resistance network for two-dimensional heat flow, we consider a composite wall,
shown in Figure 2.22(a), constructed from four different solid materials, in which the
thermal resistances of materials 2 and 3 are in parallel. The heat flows from left to

FIGURE 2.22 Transfer of heat by conduction through a plane composite wall in which the
thermal resistances of materials 2 and 3 are in parallel. (a) System schematic (b) Resistance
network.
98 An Introduction to Transport Phenomena in Materials Engineering

right, and the top and bottom surfaces are insulated. The thermal resistance network
is constructed as in Figure 2.22(b), and the total thermal resistance to heat flow from
the hotter surface at Ti to the cooler surface at To is

This model treats the individual materials in Figure 2.22 as if each had one-
dimensional heat flow and that the right border of material 1 and the left border of
material 4 are isothermal. The assumptions are reasonable only where there are small
differences in the parallel resistances (R2 and R3) or when the outer conduction resist-
ances are much smaller than the inner terms (R1, R4 << R2, R3).
A comparison between examples of one- and two-dimensional heat flow is in
Figure 2.23. A one-dimensional conduction in a rectangular body (H X L), insulated
left and right, is shown in Figure 2.23 (a). The top horizontal surface has a uniform
applied heat flux, and the bottom is isothermal at To. The isotherms (lines of uniform
temperature) are horizontal in this case, and the heat is conducted top to bottom
between the evenly spaced heat flow lines perpendicular to the isotherms. The length
of each heat flow line is the same as the height of the domain (H). The isotherms are
of length L (the horizontal length) and are parallel to the cross-sectional area normal
to the heat flow; A = LW, where W is the depth into the page. In this geometry, the

where ℓ is the length over which heat is conducted. The length and area in this one-di-
mensional case are ℓ1D = H and A1D = A = LW.
Figure 2.23 (b) tells a different story. The uniform heat flux is limited to the left-
most half of the top surface (with the rest insulated), thereby constricting the heat
Steady State Conduction Heat Transfer 99

FIGURE 2.23 Isotherms and heat flow paths through (a) a one-dimensional body
and (b) a two-dimensional body of the same length and height, but with the heat flux
constricted.
100 An Introduction to Transport Phenomena in Materials Engineering

flow. The isotherms shown here are shorter on average than L, the length of those
in the one-dimensional case. Also, the heat flow lines are no longer straight lines,
but are curved and all are longer than the length in the 1D case, H. We can write the
two-dimensional heat conduction resistance as

(2.101)

where ℓ2D
ℓ1D (always) and
A2D < A1D

This increase in thermal resistance for the same source to sink distance is due to the
two-dimensional heat path and is known as constriction (or spreading) resistance.
Eq. (2.101) introduces the definition of a shape factor, S, which represents the
average over the domain of the (A/L) ratio; it has units of length (m). The shape fac-
tor can be derived from exact or numerical solutions to the energy equation or from
measurement. Its primary use is to reduce the two- or three-dimensional conduction
to an equivalent “one-dimensional” path to be used in a thermal resistance network.
Table 2.3 lists examples of published shape factors for several different geometries.

TABLE 2.3
Shape Factors (S) for a Variety of Geometries [11]. For Cases That Are
Uniform into the Page, the Depth of the Geometry Is W.
(a) Thin vertical strip in semi-infinite solid

(b) Infinite rectangular hole in semi-infinite solid

Steady State Conduction Heat Transfer 101

TABLE 2.3 (Continued )

(c) Hemispherical depression in surface of semi-infinite solid

(d) Horizontal cylindrical hole in semi-infinite solid

(e) Infinite cylindrical tube in semi-infinite solid

(f)

(Continued)
102 An Introduction to Transport Phenomena in Materials Engineering

TABLE 2.3 (Continued )

(g) Row of infinite cylindrical tubes buried in a semi-infinite solid

(i) Two parallel infinite cylinders in infinite solid

(j) Infinite cylindrical tube in infinite square solid

(Continued)
Steady State Conduction Heat Transfer 103

TABLE 2.3 (Continued )

(k) Two plane sections and the edge at their junction

(l) Corner of junction of three plane sections

S
104 An Introduction to Transport Phenomena in Materials Engineering

Example 2.6
Use of Shape Factor to Find Heat Loss from a Furnace
A tube furnace has a cylindrical interior held at Tf = 700 oC and a convectively cooled
square exterior. The outside heat transfer coefficient is h = 15 W/m2K, and the sur-
rounding air temperature is T∞ = 25 oC. The interior is D = 8 cm in diameter, the front
exterior face is L = 16 cm on a side, and the device is W = 50 cm into the page. The
furnace insulation has a thermal conductivity of k = 1 W/mK.

Find: Treating the furnace as two-dimensional by ignoring the end effects,

write expressions for and calculate (a) the heat flow per unit length that must
be applied to the interior surface to maintain Tf and (b) a Biot number for
this furnace wall.
Solution: (a) The thermal resistance network is in Figure 2.24.

LW , where the area thermally

where the shape factor is found inTable 2.3 (j). From Ohm’s law applied to the resist-

so both convection and conduction resistances are important.

Steady State Conduction Heat Transfer 105

FIGURE 2.24 Resistance network for conduction shape factor example.

2.8.3 exact solutions

There is a class of multidimensional conduction problems that can be reasonably
represented by simple geometries and the idealized boundary conditions in Sec-
tion 2.5, and many of these configurations may be solved with exact mathematical
solutions, providing detailed predictions of T(x,y) and of heat flows. If such res-
olution is required, these results are a marked improvement over the approximate
methods. However, the math required to obtain these limited numbers of solutions
is generally beyond the level typical of the undergraduate engineering student and
would be too much of a detour (albeit an interesting one) for this text. Instead, the
reader is referred to any or all of the texts on conduction heat transfer by Carslaw and
Jaeger [12], Arpaci [13], Hahn and Özi ik [14], and Poulikakos [15], or on advanced
engineering mathematics by Greenberg [16] and Kreysig [17] for extensive coverage
of these topics. In this section, we will present two such solutions as examples of the
typical form and the uses to which they can be put.
In the first example, we seek a steady-state solution for temperature in a rectan-
gular shape, of length L and height H, with the top surface maintained at T0 and the
other three at T

(2.102)

(2.103)

which is plotted in Figure 2.25 (a). While the top boundary condition is specified
as T = T0 ( = 1), Eq. (2.104) does not give that result even for 104 terms in the
series. Instead, it gives the temperature profile in Figure 2.25 (b), where the temper-
ature is usually within 1% of the expected result, except near the corners of the top
boundary, where becomes increasingly divergent from = 1, in spite of being a
convergent series. This difficulty is well-known near boundaries that exhibit surface
temperature discontinuities, as in the upper corners of Figure 2.25 (a) [e.g., 18–20].
106 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.25 Plot of nondimensional temperature, Eq. (2.104), for an aspect ratio H/L = 1.
(a) Temperature field in entire rectangular body with 200 terms. (b) Temperature profile along top
surface (y/L = 1) with 10,000 terms.

In a real system, such a discontinuity is impossible, as the finite temperature differ-

ence over zero distance produces an infinite heat flux. In the mathematical system,
the temperatures can be maintained but only with an infinite number of terms in the
solution. This discontinuity also manifests itself in the calculation of a rate of heat
crossing the top boundary, which is a quantity of practical interest here. Applying
Fourier’s law at the top surface, we can write an expression for the heat flux distri-
bution there.

(2.104)
Steady State Conduction Heat Transfer 107

(2.105)

(2.106)

= 0, and a

(2.107)

(2.108)

(2.109)

Figure 2.26 (a–c) shows isotherms from small to large Bi, all for an aspect ratio
of H/L = 1. For the Bi << 1 case, the outside convection resistance dominates
and the body has a shallow temperature gradient. More than half the body has a
108 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.26 Temperature fields from an exact solution, with convection on the top and
right side, a uniform temperature bottom surface, and an insulated left side. Aspect ratio is
H/L = 1 and (a) Bi = 0.1, (b) Bi = 1, and (c) Bi = 10. All of these cases are from Eqs. (2.109)–
(2.110) using only the first six terms.

nondimensional temperature between 0.9 and 1.0, while all but the upper right cor-
ner is above 0.8, so it is fairly uniform close to prescribed temperature at the bottom
surface (T = To, = 1). The minimum temperature is in the upper right corner (where
it is cooled on two surfaces), and there are large temperature differences between the
surface and the environment (at = 0). The opposite is true when Bi >>1, Figure 2.26
(c). With conduction resistance dominating, a steep gradient is seen across the solid
between the bottom, held at To, and the two cooled surfaces. The largest temperature
drop is across the solid, with the surface near = 0, except near the bottom surface.
The low between surface and ambient and the large between top and right
surfaces and the hot bottom surface are both due to the overall resistance from To to
T∞ dominated by the conduction resistance.
Steady State Conduction Heat Transfer 109

Careful inspection in the Bi = 10 case shows a wavy = 0.9 isotherm near the
bottom surface, while the other isotherms are smooth. This inaccuracy in the solu-
tion is due to the representation of the uniform temperature boundary condition at
y/L = 0 using the infinite series with only the first six terms. Most texts only tabulate
the first few solutions of Eq. (2.109) for the eigenfunctions ( n). This fluctuation in
the solution to Eq. (2.108) at and near y/L = 0 is propagated to the calculation of the
local and overall heat flux there, even with significantly more terms (as in the first
example in this section).

2.9 SUMMARY
Quantitative treatment of heat flow in a system is facilitated by the electric analogy, in
which the resistance to flow is defined as the ratio of the temperature difference to the
heat flux. Resistances are derived for one-dimensional heat flow from the rate equa-
tions for conduction, convection, and radiation and steady heat flow in systems can
be approximated by resistance networks following thermal versions of Ohm’s and
Kirchhoff’s laws. The electric analogy can be applied to heat flows through multiple
materials in series and in parallel in one-dimensional or multidimensional systems.
Heat flow through complicated geometries can be approximated by shape factors,
reducing a two- or three-dimensional path to a one-dimensional overall resistance
in a network.
Application of a heat balance to a control volume yields the general energy
equation that describes transient conduction heat flow in three-dimensional space.
Steady-state exact solutions to this equation are explored in one dimension with and
without volumetric heat generation. Multidimensional conduction heat transfer is
explored through scaling analysis, resistance networks using shape factors, and a
limited number of exact solutions.

2.10 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

2.1 A window is constructed of two panes of glass, each W = 5 mm thick,

between which is a L = 2.5 mm gap containing stagnant argon. The inside
room is at Ti = 25 °C, the outside temperature is To =15 °C, the heat transfer
coefficients on the inside and outside are, respectively, hi =10 W/m2K and
ho =80 W/m2K, and the thermal conductivities of the glass and Ar are, respec-
tively, kg =2 W/mK and k Ar = 0.02 W/mK. (a) Draw a resistance network
for this double-pane window from the inside to the outside. (b) Calculate
the heat flux through the window. (c) Determine what thickness of the air
gap is required for the heat flux to be less than 100 W/m2. (d) In order to
compare the effectiveness of the double pane window, find the thickness of
a single pane required to do the job of the specified double-pane window.
Is that a practical value?
110 An Introduction to Transport Phenomena in Materials Engineering

2.2 A furnace wall has an inner layer of silica brick (kB = 1 W/mK), a Wm = 5 cm
thickness of an insulating material (km = 0.035 W/mK), and an outer stainless
steel plate (kS = 26 W/mK) WS = 1 cm thick. The furnace gas temperature
is Tf = 1000 °C, and the outer wall of the steel plate is at To = 20 °C. Inside
the furnace, hf = 25 W/m2K. (a) If the maximum service temperature of the
insulating material is 800 °C, what is the minimum thickness of silica brick
that can be used? (b) What is the overall heat transfer coefficient and the heat
flux through the composite furnace wall with this thickness of silica brick?
2.3 A cast-iron steam pipe (ROD = 0.11 m outer and R ID = 0.10 m inner diameter,
(kCI = 20 W/mK) carries steam at TS = 200°C. It is to be insulated with a
Irri = 0.025 m thickness of an insulating material (km = 0.035 W/mK). The
ambient temperature is 0 °C, and the outer and inner heat transfer coef-
ficients are, respectively, hi = 75 and ho = 225 W/m2K. (a) Find the overall
thermal resistance between the steam and the ambient. (b) Find the total
heat loss rate for a L = 2 m pipe.
2.4 A thermal barrier coating made of yttria-stabilized zirconia (YSZ) is fre-
quently plasma sprayed on nickel superalloy (e.g., IN 625) turbine blades
for the hottest stages of gas turbines. The YSZ acts as insulation, which,
with the use of a cooling gas internal to the blade, can allow it to operate
when the gas temperatures are above the thermal limit of the alloy and
sometimes above its melting temperature. Assume the inside (cooling air)
and outside (combustion gas) temperatures and heat transfer coefficients
are: Tin = 400 K, hin = 500 W/m2K, hout = 900 W/m2K. (a) Draw a resis-
tance network for the heat transfer path, modeling it as one-dimensional
heat flow with uniform cross-sectional area. The thickness of the metal is
dm = 4 mm between internal cooling air and the YSZ coating. (b) From the
resistance network, write an expression for the maximum temperature in
the blade, in terms of the properties and thicknesses of the superalloy and
the YSZ, the convection coefficients, and the inside and outside tempera-
tures. (c) If km = 23 W/mK and kYSZ = 2.5 W/mK (thermal conductivities of
the metal and ceramic, respectively), plot the maximum blade temperature
vs. gas temperature (800 K–1400 K), for dYSZ = 100 μm, 300 μm, and 500
μm. Explain the trends in the three curves.
2.5 An electrically conducting copper wire (kCu = 390 W/mK) of radius
R1 = 0.5 mm is covered with an electrically insulating plastic (kP = 0.35 W/mK)
of outer radius of R2 = 1.5 mm. When a current is passed through the wire,
heat is generated at the rate q = IV = I2 eL/Ac, where I is current, V voltage
drop, e (= 1.96 × 10 –8 Ωm) electrical resistivity, L (= 1 m) wire, and Ac wire
cross-sectional area. No current passes through the plastic due to its very
high resistivity. All the heat generated by the current escapes through the
insulator (conduction) and into the environment (convection with T∞ = 30 oC
and h = 8 W/m2K). (a) If the maximum temperature at which the plastic
may be used is 100 oC, write an expression for and calculate the maximum
current that can be passed through the wire. (b) We assume that the wire
is isothermal in the radial direction; is this a reasonable assumption under
these conditions?
Steady State Conduction Heat Transfer 111

2.6 A flowing fluid is carried in and loses heat to a tube (R1 = inner diam-
eter, R2 = outer diameter) with a layer of insulation (R3 = outer diameter)
wrapped around it, all of which is cooled by convection on the outermost
surface. (a) Sketch the temperature as a function of radius from the fluid
(Tf ) to ambient (T∞) temperature. (b) Find the total resistance from fluid to
ambient.
2.7 Consider a steel steam pipe (I.D. = 3 cm, O.D. = 3.5 cm, ksteel = 41 W/
mK) covered with an insulating material (O.D. = 4.75 cm, kins = 0.04 W/
mK). The steam in the pipe is at 120 oC and the ambient air temperature
is 15 oC. The inner and outer convection coefficients are hi = 150 W/m 2K
and h∞ = 30 W/m 2K. (a) Using the resistance network found in Problem
2.6(b), find the heat flux on the outer surface. (b) What fraction of the
total resistance is due to the insulation? (c) What is the maximum tem-
perature of the pipe?
2.8 Consider a slab wall (L thick) with the left side maintained at TL. The right
side has heat leaving to the environment (T∞) in parallel by radiation and
convection. Let L = 0.1 m, TL = 800 K, T∞ = 300 K, k wall = 20 W/mK,
= 0.8, h = 100 W/m2K. (a) Sketch a resistance network from the left wall
to the environment on the right. (b) Write an expression for the temperature
of the right wall (TR), based on the network in part (a). (c) Guess a value for
the temperature on the right (TR) and iterate with the expression in part (b)
to find a solution for that temperature.
2.9 A computer chip in an automobile engine is packaged in polyimide (kP = 0.5
W/mK), as shown in Figure 2.27. While in operation, the chip generates
heat, q = 1.5 W, and has a uniform temperature, Tc, which must be kept at or

FIGURE 2.27 Problem 2.9.

112 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.28 Problem 2.11.

below 95 oC. The left side of the package is attached to an aluminum alloy
heat sink (k Al = 100 W/mK), the left side of which is maintained at To. The
right side of the polyimide is exposed to the air in the engine at T∞ = 50 oC,
with a heat transfer coefficient h = 30 W/m2K. Assume one-dimensional
heat flow from the chip along two paths, both to the environment and the
heat sink. The geometry is L1 = 0.005 m, L2 = 0.002 m, and A = 0.0001 m2.
(a) Draw the thermal resistance network with all of the resistances, temper-
atures, and heat flows labeled. In symbolic form, write the forms of all the
resistances. (b) Find the heat sink temperature (To) necessary to maintain
the chip temperature at or below 95 oC. (c) Is the sink temperature found in
(b) practical for an automobile engine? Explain.
2.10 A cubical picnic chest of length L = 0.5 m, constructed of sheet Styrofoam
of thickness t = 0.025 m, contains ice at Tin = 0 °C. The thermal conductiv-
ity of the Styrofoam is kS = 0.035 W/mK, and the ambient temperature is
T∞ = 25 °C. If the convection resistences are negligible, calculate the rate at
which the ice in the chest melts. The latent heat of melting of ice is 3.34 ×
105 J/kg.
2.11 The average heat transfer coefficient for flow over the top side of a flat
surface can be determined experimentally by passing a current through a
very thin sheet of electrically resistive metal (e.g., IN718, a common nickel
superalloy). For structural stability, the bottom side of the sheet is sup-
ported by foam insulation of thickness L, with the opposite surface of the
foam maintained at Tsink. (a) Sketch and label a complete resistance network
for this system. (b) Write an expression for the heat transfer coefficient
in terms of the other parameters in the resistance network, including the
foam. (c) For q = 2500 W/m2, kfoam = 0.1 W/mK, L = 0.1 m, and T∞ = 25 oC,
the average sheet temperature was measured as TS = 60 oC when air was
blown parallel to the sheet. What is the value of the heat transfer coef-
ficient? (d) Write an expression for and calculate the % error in the heat
transfer coefficient if heat loss through the foam is ignored? (e) What is the
maximum thickness (L) if that error is held to half the value in (d)?
2.12 To understand the origin of the concept of “wind chill,” consider the human
body as a vertical cylinder L = 2 m long and D = 30 cm in diameter. A sub-
skin layer (k = 0.4 W/mK, t = 1 cm) separates the isothermal body core,
at Tcore = 36.5 oC, from the skin exposed to air. This outer boundary is
treated with a convection condition. The ambient temperature is T∞ = 10 oC.
Steady State Conduction Heat Transfer 113

(a) Sketch and label a complete resistance network for this system. (b) Solve
the system to find an expression for the outer skin temperature. (c) For
natural convection (no wind), the convection coefficient is hNC = 1 W/m2K,
while for forced convection (with a wind velocity of around 2 m/s), the con-
vection coefficient is hFC = 30 W/m2K. What is the skin temperature and
heat loss rate in these two cases?
2.13 In direct chill casting of aluminum, the metal alloy is transferred from melt-
ing furnace to molds through a trough, which has a length into the paper of
L = 10 m. The heat is added by radiation exchange with the heaters above,
which gives a net heat flow into the aluminum, qrad. The heat leaves the
isothermal (TAl = 700 oC) liquid metal by two paths: (i) through the insula-
tion by conduction and then by convection from the bottom and sides of the
trough (hins), and (ii) from the exposed Al surface by convection (h Al). For
each of the two geometries shown in Figure 2.29 : (a) Sketch the resistance
network and write the formulae for the resistances. (Assume that there is

FIGURE 2.29 Problem 2.13.

114 An Introduction to Transport Phenomena in Materials Engineering

negligible resistance through the metal, so it can be treated as isothermal at

TAl.) (b) At what rate must heat (qrad ) be added to the alloy if it is maintained
at TAl? What fraction of the heat is lost on each of the two paths?

hAl = 30 W/m2K T∞ = 30 oC hins = 10 W/m2K kins = 0.2 W/mK

= 0.2
For rectangular geometry: t = 0.05 m Scorner = 0.54 L
H = 0.25 m W = 0.3 m
For cylindrical geometry: ro = 0.21 m ri = 0.16 m

2.14 In the continuous casting of steel, a slag layer is used as a lubricant and
an insulator between the moving steel and the copper mold. Inside the
mold, water flows through a bank of tubes (radius R) to extract the heat.
In Figure 2.30, we see tubes in the mold running parallel to the slag-mold
interface and perpendicular to the page. The length of each tube is L and
the height of the region cooled by a single tube is H, so the cross-sectional
area of the slag-steel interface (normal to the heat flow) is As = LH. (You
can see that H is the tube spacing.)
Assume that the steel-slag and slag-mold interfaces and the inside of the
tube are isothermal at temperatures T1, T 2, and T3, respectively. The flowing
water is at Tw (< T3).

T1 = 1550 oC Tw = 50 oC ds = 0.01 m dt = 0.05 m

kslag = 0.5 W/mK
R = 0.02 m H = 0.1 m L=1m kslag = 0.5 W/mK
htube = 500 W/m2K kmold = 250 W/mK
(a) Draw a resistance network from the steel-slag interface to the cooling
water in the tube. (Hint: You need to use a shape factor in this network.) (b)

FIGURE 2.30 Problem 2.14. The dashed lines are symmetry planes.
Steady State Conduction Heat Transfer 115

Find the values of the three resistances and rank them. (c) What is the rate of
heat loss (q) in the area (= LH) due to the cooling water in the one tube? (d)
Plot T2 and q” as a function of dt (between 0.03 m and 0.2 m). What would you
recommend for a value of dt ? Justify your recommendation.

2.15 The curved surface of a solid cylinder (length = L = 0.2 m and diame-
ter = D = 0.02 m) is covered with insulation. The ends are maintained at con-
stant temperatures of 250 °C and 150 °C. Calculate the rate of axial heat flow
through the cylinder if it is constructed of (a) pure copper (k = 401 W/mK),
(b) pure aluminum (k = 236 W/mK), (c) graphite (k = 1800 W/mK),
(d) quartz (k = 2 W/mK), and (e) magnesia (k = 30 W/mK). (f) For the same
end temperatures, what would the length of a copper rod be to have the
same heat flux as in a 0.2 m length of aluminum cylinder?
2.16

Heat flows by conduction through a plane slab of this material of thickness

0.1 m, the left face of which is at 100 °C and the right face of which is at 0
°C. Calculate (a) the mean thermal conductivity of the material in the range
0–100 °C; (b) the heat flux through the slab; (c) the actual variation of T
with distance through the slab; (d) the temperature at x = 0.05 m; (e) the
temperature at x = 0.05 m calculated using the mean thermal conductivity;
(f) the actual temperature gradient in the slab at x = 0.05 m; and (g) the
temperature gradient at x = 0.05 m calculated using the mean thermal
conductivity.
2.17 A horizontal cylindrical tube furnace has heating elements maintained at
Tf = 500 oC on the internal surface (r = Rin = 0.1 m). The furnace is con-
vectively cooled on the outside (r = Rout = 0.25 m) by air at T∞ = 25 oC and
with hair = 10 W/m2K. What is the largest kins (thermal conductivity of insu-
lation) allowable if we keep the temperature of the outside of the furnace
below Ts = 50oC?
2.18 A cast iron piston head in a diesel engine is subjected to the heat gener-
ated by the fuel combustion, which can be treated as a steady-state heat
flux (because the time scale of each cycle is much shorter than the time for
thermal diffusion). A YSZ (yttria-stabilized zirconia) coating is applied
to insulate the piston, which is convectively cooled at x = LP. (a) Write
two steady-state energy conservation equations for temperature, one in
each material (Tp(x) in the piston and Tc(x) in the coating), and appropriate
boundary conditions for each equation. Keep in mind that there are no dis-
continuities in the temperature or heat flux at the coating-piston interface,
x = 0. You should have two differential equations and four boundary condi-
tions. (b) Solve the two equations and write the expressions for the tem-
perature distributions in both layers. (c) What is the maximum temperature
of the cast iron at steady state?
116 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 2.31 Problem 2.18.

Steady State Conduction Heat Transfer 117

(f) What is the steady-state current (I) limit for the fuse (T < Tmelt), using the
material properties and conditions: h = 0.01 W/m2K, L = 0.0005 m, T∞ = 30 oC,
k = 20 W/mK, Ac = 5 × 10–7 m2, Tmelt = 600 oC, and = 107 (Ωm)-1?
118 An Introduction to Transport Phenomena in Materials Engineering

condition on the side to estimate the order of magnitude and functional

form of the maximum and surface temperatures.
2.25 A rectangle (H × L) with uniform internal heat generation is cooled uni-
formly by convection on all four sides by the same heat transfer coefficient
and ambient temperature. Scale (a) the energy equation (2.32) and bound-
ary conditions on the (b) right side and (3) top to estimate the order of
magnitude and functional forms of the maximum temperature and the top
and right side surface temperatures.
2.26 A slab with uniform heat generation is cooled convectively on both sides
(at x = 0 and x = L), but h 0 ≠ hL. (a) Simplify the energy equation (2.32) and
solve, writing expressions for T(x) and Tmax. (b) Write an expression for the
fraction of heat lost to the left side of the slab.

NOTE
1. If there is a contact resistance at the interface, Eq. (2.64) is still valid. However, there

replaces Eq. (2.63) as a boundary condition for Eq. (2.62).

REFERENCES
1. Yovanovich, M. M., “Four decades of research on thermal contact, gap, and joint resis-
tance in microelectronics,” IEEE Transactions on Components and Packaging Technolo-
gies, v. 28, pp. 182–206, 2005.
2. Savija, I., M. M. Yovanovich, J. R. Culham, and E. E. Marotta, “Review of thermal
conductance models for joints incorporating enhancement materials,” Journal of Ther-
mophysics and Heat Transfer, v. 17, pp. 43–52, 2003.
3. Thomas, T. R., and S. D. Probert, “Thermal contact in solids,” Chemical Process Engi-
neering, pp. 1–12, 1966.
4. Schneider, P. J., “Conduction,” in Handbook of Heat Transfer, W. M. Rosenhow and J.
P. Hartnett (eds.), McGraw-Hill, 1973.
5. Fried, E., “Thermal conduction contribution to heat transfer at contacts,” in Thermal
Conductivity, R. P. Tye (ed.), v. 2, Academic Press, 1969.
6. Jain, V. K., “Determination of heat transfer coefficients for forging applications,” Jour-
nal of Materials Shaping Technology, v. 8, pp. 192–202, 1990.
7. Bendada, A., A. Derdouri, M. Lamontagne, and Y. Simard, “Analysis of thermal contact
resistance between polymer and mold in injection molding,” Applied Thermal Engineer-
ing, v. 24, pp. 2029–2040, 2004.
8. McDonald, A., C. Moreau, and S. Chandra, “Thermal contact resistance between
plasma-sprayed particles and flat surfaces,” International Journal of Heat and Mass
Transfer, v. 50, pp. 1737–1749, 2007.
9. Blaise, A., B. Bourouga, B. Abdulhay, and C. Dessain, “Thermal contact resistance esti-
mation and metallurgical transformation identification during hot stamping,” Applied
Thermal Engineering, v. 61, pp. 141–148, 2013.
10. Griffiths, W. D., and K. Kawai, “The effect of increased pressure on interfacial heat
transfer in the aluminum gravity casting process,” Journal of Materials Science, v. 45,
pp. 2330–2339, 2010.
Steady State Conduction Heat Transfer 119
 3 Transient Conduction
Heat Transfer

3.1 INTRODUCTION
If a body initially at some uniform temperature is suddenly immersed in a quenching
liquid or gas that is at a lower temperature, the consequent heat transfer from the solid
to the quenchant causes the local temperature in the body to vary with both position
and time. In principle, these variations are obtained from the integration of the gen-
eral heat conduction equation (2.32) with appropriate boundary conditions. In this
chapter, we will examine such examples of transient heat conduction using the tools
introduced in the previous two chapters. Scaling analysis will be used to determine the
relative importance of various thermal resistances, which will allow us to make certain
approximations to simplify the solutions. Approximate and exact solutions will be
found, applications of these zero-, one-, and two-dimensional solutions will be shown,
and the discussion will be extended to include the effect of solid-liquid phase change.

3.2 SCALING ANALYSIS OF GENERAL TRANSIENT CONDUCTION

Here we examine heat transported from a fluid to a surface by convection and then
transferred from the solid surface to its interior by conduction.1 In this case, there are
thus two thermal resistances to the heating process – resistance to internal conduc-
tion and resistance to external convection – and the relative magnitude of these resist-
ances is of great importance. If the convection resistance is significantly greater than
the conduction resistance, a bottleneck in the heat transfer process occurs at the solid
surface. Heat is easily transported through the body by conduction but experiences
difficulty in being transferred into the surface from the fluid. In such a situation, the
temperature gradients developed in the solid can be small enough to be ignored. At
the other extreme, if the convection resistance is very much smaller than the conduc-
tion, then the surface temperature is approximately the same as the ambient temper-
ature and most of the temperature change is inside the body.
Figure 3.1 shows the general case of the problem at hand. The body starts (t ≤ 0)
with an initial uniform temperature, Ti, and is insulated at x = L. At t = 0, the body is
suddenly heated by convection at the x = 0 boundary,

(3.1)

where the transient surface temperature is To = T(x=0). Eq. (3.1) is the energy bal-
ance at x = 0 between the internal conduction (x > 0) and external convection (x = 0).
(We continue to use the sign convention that heat flow is positive in the positive

120
120 DOI: 10.1201/9781003104278-3
Transient Conduction Heat Transfer 121

FIGURE 3.1 Schematic of temperature profiles in a conducting solid heated at x = 0 by

convection to the environment at T∞ (t1 < t2 < t3 < t4).

coordinate direction, and the sign of the heat flow discovered by the analysis shows
its real direction.) The T(x) curves are plotted at increasing times, as the body heats.
As heat enters the body, the temperature at x = 0, To(t), rises and the interior is heated.
The effect of the higher surface temperature reaches further into the body as time
progresses and eventually impacts the far wall (x = L). With an insulated (zero tem-
perature gradient) condition there, the temperature at the far wall, TL(t), increases.
The heating continues until the entire body is at the ambient temperature, T∞.
To make a first quantitative estimate of the thermal behavior inside this body, we
select

(3.2)

(3.3)

(3.4)
122 An Introduction to Transport Phenomena in Materials Engineering

For the case with the dominant thermal resistance due to convection, BiL 0 and
To TL, so there is only a very small temperature drop across the body. If BiL ∞
because the convection resistance is negligible in the heat flow path, then To T∞. If
BiL ~ (1), the surface temperature is in between TL and T∞. The next three sections
will explore these three Biot number regimes.

3.3 LUMPED CAPACITANCE ANALYSIS: CONVECTION

RESISTANCE DOMINATED (BI << 1)
For Bi << 1, the conduction resistance is much smaller than convection resistance at
the surface and heat spreads inside the body more easily than it leaves. The surface
and inside temperatures are transient, but, as shown by Eq. (3.4), they have effec-
tively the same value in this case. Although this situation is clearly impossible, as
conduction heat transfer requires some temperature gradient, that gradient is so small
when Bi << 1 that we can treat the body as isothermal while the body discharges or
stores heat, hence the description of the system as a “lumped capacitor.”2

3.3.1 convection Heat loss

To model the lumped thermal capacitor, we start with the thermal balance on a vol-
ume in Figure 2.6, and the energy balance with no heat generated or transferred into
the solid:

(2.18)

The heat storage rate is the change in enthalpy with time, in this case with no phase
change,

(3.5)

For a convection heat loss, the total thermal energy balance is

(3.6)

for constant properties and where As is the surface area through which heat flows.
Separation of variables in Eq. (3.6) gives

(3.7)

which, on integration from Ti at t = 0 to T at time t, produces

(3.8)
Transient Conduction Heat Transfer 123

or

T T hAs
exp t . (3.9)
Ti T c

Thus, under the conditions that significant temperature gradients do not develop
within the solid during cooling, the temperature of the solid decreases exponentially
with time. Eq. (3.9) shows that either an increase in the surface area or heat transfer
coefficient or a decrease in thermal capacity c will increase the rate at which the
solid cools. The solution also does not include thermal conductivity, as the derivation
of Eq. (3.9) is based on the determination that the conduction resistance does not play
a role in the system’s thermal response.
We can nondimensionalize Eq. (3.9) using

T T hLc k t t
, Bi , and Fo ,
Ti T k c L2c 2
L
c

which gives a simple result:

exp Bi Fo . (3.10)

The Fourier number, Fo t L2c , is a nondimensional time, using a time for thermal
diffusion across a length Lc as a reference, tref L2c . Figure 3.2 shows the exponen-
tial decay in temperature during cooling.

FIGURE 3.2 Nondimensional temperature ( ) in a lumped capacitor with convection heat

loss, as a function of nondimensional time (Fo) and a range of Bi.
124 An Introduction to Transport Phenomena in Materials Engineering

Example 3.1
Quenching an Aluminum Plate
An aluminum alloy plate of dimensions 0.01 m × 0.05 m × 0.05 m is solution-treated
at 520 °C to form a completely phase microstructure. The plate is removed from the
furnace and quenched by forced convection in 25 °C air, governed by a heat transfer
coefficient of h = 500 W/m2K. In the temperature range between 25 °C and 520 °C,
the following mean property values are used:

= 2700 kg / m 3 c = 900 J / kgK k = 120 W / mK.

The goal of the quench is to cool the plate fast enough so that there is no precipita-
tion of a second ( ) phase, which can be achieved if the cooling curve does not cross
the transformation start line on the alloy’s continuous cooling transformation (CCT)

The temperature below which phase precipitation is thermodynamically favor-

able is Ttrans. The transformation does not begin there, but is delayed by kinetics
effects. The time scale here is t* = t – to; t* is the time beginning when T = Ttrans,
which is reached at to, the time between the start of cooling and when T = Ttrans.
(For more information on quenching aluminum alloys, see [1] for the metallurgy
of the phase change and [2] for data such as Figure 3.3 for a wide range of nonfer-
rous alloys.)

FIGURE 3.3 Schematic of a continuous cooling transformation diagram for ( ) phase pre-
cipitation in an aluminum alloy. The elapsed time, t*, begins when the temperature falls into
the - region of the alloy phase diagram (T < Ttrans), when it becomes thermodynamically
favorable to begin precipitation.
Transient Conduction Heat Transfer 125

Find: (a) the initial rate of cooling, T t t 0 ; (b) the time required for the
plate to cool to Ttrans; and, (c) whether or not the phase begins to precipi-
tate during this quench.
Solution: The characteristic length of the plate is defined as the ratio of the
volume and area, Lc 0.005 , neglecting the thin edges. The Biot
As
number is then

hLc 500 W/m 2 K 0.005 m

Bi 0.021 1,
k 120 W/mK
which indicates that the plate can be treated as a lumped capacitor.

(a) At T = 500 °C, rearranging Eq. (3.6) will give us the cooling rate at t = 0:

dT hAs Ti T h Ti T
= =
dt c cLc
500 W/m K 500 C 25 o C
2 o

= 3
19.5 o C/s.
2700 kg/m 900 J/kgK 0.005 m

(b) Rearranging Eq. (3.8) and using the characteristic length definition, the
time to reach Ttrans = 490 °C (t* = 0) is

cLc T T
to ln
h Ti T
2700 kg/m 3 900 J/kgK 0.005 m 490 o C 25 o C
ln
500 W/m 2 K 520 o C 25 o C

to 1.52 s.

(c) To ensure that the cooling curve does not cross the precipitation line, we
find the elapsed time since T = Ttrans, t*, at which the cooling curve reaches
Tnose.

cLc T T
t* ln nose
h Ttrans T
2700 kg/m 3 900 J/kgK 0.005 m 300 o C 25 o C
ln
500 W/m 2 K 490 o C 25 o C

t* 12.8s
*
This value is slightly larger than the minimum time for precipitation to begin (t nose ).
Considering the uncertainty in these diagrams and in the analysis, this cooling path
126 An Introduction to Transport Phenomena in Materials Engineering

may just nip the nose of the precipitation curve and we might expect only a very
small amount of phase.

Example 3.1 shows what most would consider a small body treated as a lumped
capacitor, and there is a tendency to think that only “small” parts can be treated that
way. This assumption is only true with the proper definition of “small.” An adult
human is small compared to an elephant, but not to a mouse. For determining the
applicability of the lumped capacitance analysis, the characteristic length of the
plate is small or large, in the context of the problem, based on its comparison to the
ratio of the heat transfer coefficient and thermal conductivity. Hence, the physics
of the problem will tell us if the solid is small enough to be considered a lumped
capacitor; for example, even a solid as large as a direct chill cast aluminum ingot
with a diameter of 0.5 m may be treated this way under the right conditions [3].

3.3.2 radiation Heat loss

Another possible mode of heat loss during quenching is radiation, in which case the
energy balance is

(3.11)

using the radiation rate equation, Eq. (1.35). This mode of heat transfer is dominant
only when quenching in a vacuum or from a high temperature in a gas. Again using
separation of variables,

where Ti is the initial temperature. This temperature history can be generalized by

defining the nondimensional variables

The generalized equation is time as a function of the body temperature with one free

(3.13)
Transient Conduction Heat Transfer 127

A simpler solution is available if T 4 T 4. In that case, the energy balance is

dT d
As T 4 c and 4
. (3.14)
dt d

The temperature histories from solving these energy balances are

1/ 3
Ti 3 1/ 3
T Ti 1 3 t and 1 3 . (3.15)
cLc

This approximate solution behaves as if T∞ = 0 K, as seen when plotted with the

exact solution, Eq. (3.13). Figure 3.4 shows the exponential decay in the temperature
response to radiative cooling for several values of ∞ = T∞/Ti. The approximate solu-
tion compares very well to the behavior of the exact solution until the temperature
approaches the ambient temperature. At that point, the approximation T 4 T 4 fails
and the exact solution is needed for more detailed predictions of temperature and
heat flow.

Example 3.2
Quenching a Stainless Steel Turbine Wheel3
A stainless steel turbine wheel is shown in Figure 3.5, with the general shape in part
(a) and the dimensions of the cross-section in part (b). The wheel is austenitized at
1000 oC and must be quenched quickly enough to form a martensitic microstructure
even in the center of the thickest section (the outer rim). The requirement is to reach
below 700 oC in less than 1000 s. Oil quenching is suggested, but may be a problem
if the quenching is too fast and causes unacceptable plastic thermal strains; it is much

FIGURE 3.4 Nondimensional temperature histories for a lumped capacitor cooled by radia-
tion to various environmental temperatures.
128 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.5 (a) Turbine wheel. (b) Schematic of turbine wheel cross-section. The thicker
lines indicate the thermally active surfaces on the rim. (c) Predicted centerline temperature
history of thick rim of wheel.

more costly than air cooling. Such passive cooling in air is desirable, but is it rapid
enough to meet the metallurgical requirements?
Because of the high temperature, we can show that convection heat losses are
much less than those by radiation, therefore we will neglect convection. (This is a
conservative estimate; if the cooling predicted is sufficient, then it would only be
more so if convection was important.)
Transient Conduction Heat Transfer 129

Find: (a) the characteristic length, Lc, of the rim (the thickest section); (b) esti-
mate the Biot number based on hrad at the highest surface temperature; and
(c) assuming the rim can be treated as a lumped capacitor, plot the T(t). Will
air cooling be sufficient to produce martensite at the center of the rim?
Solution: (a) To find the characteristic length of the rim, we identify its sur-
faces that are thermally active; these are shown in

Figure 3.5 (b). The areas of these surfaces and the volume of the rim are:

outer rim surface: AOR 2 R2 H 2 0.188 m 2 inner rim surface: AIR 2 R1 H1 0.047 m 2
top rim surface: ATR R22 R12 0.086 m 2 ri volume: R ATR H 2 0.00864 m 3 .

The characteristic length is

Lc R / ( AOR AIR ATR ) 0.0268 m.

(b) To estimate the Biot number at the initial temperature, hrad, Eq. (1.37) is
required.

hrad T2 T2 T T
2 2
0.8 5.67 10 8 W/m 2 K 4 1273K 573K 1273K 573K

139 W/m 2 K.

The initial Biot number is Bi = hrad Lc/k = 0.155, which is low enough to justify the
use of a lumped capacitance analysis and only decreases as the rim temperature falls.

(c) Given that Bi <<1, the lumped capacitance solutions for radiation in Eqs.
(3.13) and (3.15) are valid and are plotted in Figure 3.5 (c). For both solu-
tions, the results predict that the cooling to 700 oC occurs much sooner than
1000 s, suggesting that the air quench will be sufficient to form martensite
even in the slowest cooling region at the center of the rim.

3.4 SPATIAL DEPENDENCE: CONDUCTION

RESISTANCE DOMINATED (BI >> 1)
When conduction resistance inside the body in Figure 3.1 is much greater than the
external convection resistance, then Bi >> 1, and Eq. (3.4) gives the limit of the sur-
face temperature of
To ~ T .

This result indicates the convection resistance between surface and environment is
so small that the surface boundary condition changes almost immediately from con-
vection cooling to a fixed temperature at the ambient value. (There is a calculable
time delay, modeled in Section 3.5, but it is much faster than the rest of the system
130 An Introduction to Transport Phenomena in Materials Engineering

response.) The initial condition inside is T(x, t=0) = Ti, and the effect of suddenly
changing To to T∞ at t = 0 takes time to penetrate the body. The full range of temper-
ature response is described in the next two subsections.

3.4.1 cooling in a slab: early tiMes For Bi >> 1

Because the heat takes a finite amount of time to move through the slab, there is a
period of time after the boundary condition change before it is felt at the far boundary
of a finite slab ( < L, as in Figure 3.1 at t = t1). At these early times, the heat transfer
behaves as if the slab is semi-infinite,4 as the far wall (at x = L) has no influence on
the evolving temperature field.
In this case, with one-dimensional, transient conduction and no heat generation,
the thermal energy conservation is expressed by

heat sensiblee heat (2.31)

x conduction y conduction z conduction
generation storage

or with constant and uniform properties,

(3.16)

remembering that the thermal diffusivity is = k/ c. The boundary conditions are

(3.17)

and the initial condition is

Scaling the system, Eqs. (3.16)–(3.17), using Tref ~ (Ti – To) and xref ~ , we obtain

or

(3.18)

a noticeable change in temperature. This penetration depth of the boundary condi-

tion is seen in Figure 3.6 (a). This expression for the growing size of such a region,
Transient Conduction Heat Transfer 131

proportional to the square root of a diffusivity multiplied by time, is ubiquitous in

diffusion-controlled problems (as we will see in Chapters 4, 9, and 10). Using the
penetration depth, we can scale the conduction rate equation at x = 0 to get an esti-
mate of the behavior of the heat flux into the body as sketched in Figure 3.6 (b).

dT T To T To
q k ~ k i ~k i (3.19)
dx t

To obtain a solution for the temperature and heat flux throughout the body, we
introduce an approximate technique, integral analysis, which generally agrees with
the exact solution to within 10–20% (and frequently better than that). This level of

FIGURE 3.6 (a) Temperature response and (b) penetration depth and surface heat flux
behavior of semi-infinite body with a sudden change in surface temperature.
132 An Introduction to Transport Phenomena in Materials Engineering

agreement is reasonable given uncertainty in the thermophysical properties and vari-

ous simplifying assumptions about geometry and thermal conditions invariably made
when reducing a real engineering problem to an inexact model.
The integral solution consists of two steps to produce the transient profile of the
dependent variable (here temperature).

Step 1: The first task is to assume a profile for the dependent variable over the
region in which it varies. Frequently, it is best to assume a polynomial form.
(Other possible forms include exponential or trigonometric functions.) The
order of the polynomial depends on the boundary conditions of the problem. If
there is reason to believe from those conditions that a point of inflection exists
in the profile, then a cubic function is best. However, in many cases examined
in this text it is enough to assume a second order (quadratic) profile, by which
three unknown coefficients to be found by application of boundary conditions.
Step 2: The coefficients and/or the independent variable in the assumed profile
will usually contain a quantity or quantities for which the time dependence
is not known. In order to determine these functions and complete the solu-
tion to the problem, one integrates the governing conservation equation
over the entire region of the assumed profile. The procedure will give rise
to a differential equation for the unknown, transient quantity, the solution
to which will produce the time dependence of those functions, completing
the solution to the conservation equation.

The integral method is applied here to the semi-infinite conduction problem intro-
duced in this section. First, we will partially normalize temperature and position in
the energy equation and boundary conditions thus:

(3.20)

where = x/ (t

(t), so the region of interest at any given time is x = 0 to (or = 0 to 1). Note that
the energy equation (3.20) is not entirely dimensionless, as we have no convenient
reference scale for time.
We assume a parabolic profile for the normalized temperature, :

(3.21)

This function has three unknown coefficients (C1, C2, and C3), but we have only two
conditions at the boundaries of the domain. A physical argument can be made for
adding a third condition, as the heat flux approaches zero at the edge of the penetra-
tion depth:
Transient Conduction Heat Transfer 133

Using these three boundary conditions, we can find the three coefficients and the
temperature profile:

( = 0) = C 1(0 ) 2 + C 2(0) + C 3 = 1
( = 1) = C 1(1) 2 + C 2(1) + C 3 = 0

( = 1) = 2 C 1(1) + C 2 = 0

and
2
= 2 +1. (3.22)

It may appear that we have the temperature profile, but we must remember the ther-
mal penetration depth, , and its unknown dependence on time lurking in = f[ (t)].
To obtain that function, we must turn to the energy integral equation,

1 2 1
2 2
d = d , (3.23)
0 0 t

which is the energy conservation equation, Eq. (3.16), integrated over the region that
is affected by the change in the wall temperature (x = 0 to x = or = 0 to = 1). Note
again that the size of this region, (t), has an unknown time dependence. Integrating
the left side of the equation, we get

1
2
= d , (3.24)
1 0
0 t

which is the balance of the heat flux in and out of the domain (on the left) and
the temperature rise (a measure of energy storage) averaged over the domain (on
the right).
We can use the temperature profile ( ) from Eq. (3.22) to evaluate the three terms
in this equation. The chain rule is used to find an expression for the cooling rate in
terms of and the unknown functions.

1 2
= = 2 2 = 2 2
t t t t

= (2 2) 0
= 2 and = 0.
0 1

Plugging these results into Eq. (3.24), and integrating the transient term, we get

2 1 d
2
= .
3 dt
134 An Introduction to Transport Phenomena in Materials Engineering

Solving using separation of variables and the penetration depth initial condition,
(t = 0) = 0,

(3.25)

Now that we know the behavior of the penetration depth, we can write the tempera-
ture as a function of x and t:

(3.26)

We can also find the heat flux at the surface as a function of time:

(3.27)

Finally, an exact solution can be found by reducing the partial differential equation
to an ordinary one and solving it for temperature (for details see, e.g., [4, 5]). Various
procedures for this reduction of order are described in [6]. The solution to Eqs. (3.16)
and (3.17) can be found as

(3.28)

which defines the error function in terms of an independent variable,

(The error function in Eq. (3.28) cannot be solved analytically, so its numerically
calculated and usually presented in tabular form or as a math function in spreadsheets
and other calculation tools.) We can also redefine the nondimensional temperature
as in Eq. (3.20),

(3.29)

which is the definition of the complimentary error function.

Using this exact solution to the energy equation, we can find the transient heat flux
behavior at the surface (x = 0):

(3.30)
Transient Conduction Heat Transfer 135

FIGURE 3.7 Comparison of exact temperature solution with the integral analysis using
second- and third-order polynomial profiles.

which matches the scaling analysis to a factor of 1 and the integral analysis
much closer.
Comparing the exact expression for the heat flux to those of the scaling and inte-
gral scaling solutions, Eqs. (3.19) and (3.27), we see that they all have the same
functional dependence on material properties and thermal conditions. The coeffi-
cients for the integral and exact methods are only 2% different from each other, and
the solutions are the same order of magnitude as the scaling analysis. A comparison
of the nondimensional temperature profiles is shown in Figure 3.7 (remembering
that E = √3). Given the assumptions generally made to reduce a real system to
this simple configuration, as well as uncertainties in the thermophysical properties,
the agreement found is acceptable. The solution from an integral analysis using a
third-order polynomial is also included [5], showing that higher-order profiles are
not always better. In fact, while a third-order polynomial is frequently used, it should
not be expected to be as good as the quadratic because the actual temperature profile
does not have a point of inflection.

Example 3.3
Deposited Metal Droplet Heating a Substrate
A droplet of liquid aluminum is deposited with negligible superheat onto a steel
substrate, initially at Ti = 25 oC, where the droplet freezes at a constant temperature,
TAl = 660 oC. The droplet takes 10 ms to freeze. For the substrate: k = 20 W/mK,
= 8240 kg/m3, c = 470 J/kgK, = 5.16 × 10–6 m2/s. We can assume that the heat
transfer coefficient between substrate and droplet is high enough that Bi >> 1, so
To = T∞ = TAl.
136 An Introduction to Transport Phenomena in Materials Engineering

Find: (a) Calculate the thermal penetration depth at the end of freezing, f. (b)
Plot T(t) at half the final penetration depth: xb = f/2.
Solution: (a) Neglecting the lateral conduction in the substrate, this problem
can be estimated as one-dimensional planar transient conduction with a
constant temperature at x = 0, as shown in Figure 3.8 (a). Therefore, from
Eq. (3.25), the penetration depth at t = 10 ms is

(b) We seek the thermal history at xb = f /2 = 0.38 mm. The thermal effect
reaches that point at

FIGURE 3.8 (a) Schematic of substrate response to a droplet freezing on its surface.
(b) Temperature history during droplet freezing for substrate depth xb = 0.38 mm.
Transient Conduction Heat Transfer 137

For time less than tb, the temperature at xb is still at the initial temperature, Ti = 25 oC.
After that time, the temperature rises according to Eq. (3.22),

Tb t Ti 2 xb2 2xb
t b 2 b +1 +1
TAl Ti 12 t 12 t

The temperature at xb, Tb(t) is plotted in Figure 3.8 (b).

3.4.2 cooling in a slab: late tiMes For Bi >> 1

All bodies can be treated with the previous semi-infinite solution, as long as the pen-
etration depth, (t), has not reached the far end of the body ( < L). At a critical time,
tcrit, the thermal effect reaches the far wall, (tcrit) = L, so the solution in Eq. (3.25) for
the penetration depth gives

tcrit L2 12 . (3.31)

For times greater than tcrit, the problem becomes one of conduction in a finite slab.
The temperature field behavior is now different from the semi-infinite solution in two
ways: (i) the penetration depth is a constant, L, and (ii) the (as yet unknown) temper-
ature at x = L increases in time. The complete system is
2
T 1 T T
T x 0, t 0 To 0. (3.32)
x2 t x x L

To start, we scale the energy equation (3.32) to get estimates for the unknowns in this
situation: the rising temperature at the far end, TL(t), and the falling heat flux at x = 0.
To do so, we pick a length scale, xref ~ L, and the time reference is the elapsed time
since the penetration depth reached the far wall, tref ~ t – tcrit. There are two temper-
ature references used here. Conduction is controlled by the temperature difference
across length L, so Tref,C ~ (TL – To), while the energy storage term is proportional
to the temperature change at the far end during tref, so Tref,S ~ (Ti – TL). Scaling the
thermal energy balance, Eq. (3.32),

Tref ,C Tref , S
2
~
x ref tref
or
TL To Ti TL
2
~ . (3.33)
L t tcrit

Defining a nondimensional time as = (t – tcrit)/L2, we can rearrange Eq. (3.33) for

an estimate of the far wall temperature behavior after tcrit:

To Ti
TL ~ . (3.34)
1
138 An Introduction to Transport Phenomena in Materials Engineering

At the beginning of the finite slab regime, = 0, this relation shows the far wall tem-
perature is at the initial value (TL ~ Ti), and as time goes to infinity, the TL To. The
heat flux at the surface can be estimated by scaling the conduction rate equation at
x T

(3.35)

the nondimensional version of which is

(3.36)

where S indicates the scaling result. An integral analysis can be performed on this
conduction problem, although the unknown quantity is now the far boundary temper-
ature value, TL(t
.

(3.37)

The analysis for the thermal response for t ≥ tcrit starts by assuming a quadratic non-
dimensional temperature profile,

(3.39)

To find the three coefficients, we need one more boundary condition. Here we use
the unknown nondimensional temperature at x = L ( = 1), L( ). Applying the three
boundary conditions, Eq. (3.39) can be written as

(3.40)

So, we have the temperature profile as a function of position, but the time depend-
ence in L is still unknown. We integrate the energy equation, Eq. (3.37), over the
domain of interest (the entire slab, 0 ≤ ≤ 1) and use the assumed profile to find a
differential equation for L( ):
Transient Conduction Heat Transfer 139

1
L
d ,
0
1 0 L

where we have used the chain rule in the storage term. Evaluating the derivatives,

2
1 L 2 2 and 2
L

we get
1
L 2
0 2 1 L 2 d
0

1
L 2 1 3
2 1 L
3 0

L
3 L 3.

This differential equation is solved using separation of variables and a substitution

u = 1 – L.

L u
3
1 L u

u
ln u 3 ln C1 or ln 3
C1

u 1 L C1 exp 3

Applying the initial condition of

TL t tcrit Ti or L 0 0,

we find the unknown constant, C1 = 1. The far wall temperature is then

L 1 exp 3 , (3.41)

and the integral solution (I) for the temperature profile can be written as
2
I exp 3 2 1. (3.42)

We note that at t = tcrit, Eq. (3.42) reduces to the same as the solution to the semi-in-
finite model, Eq. (3.22), because = L at that time. Also, at large times, we see that

lim I 1, exp 12 2 1 1 1,
140 An Introduction to Transport Phenomena in Materials Engineering

which means the entire body is at the surface temperature.

The heat flux at the surface during the finite slab regime can be found using the
temperature profile, Eq. (3.42):

(3.43)

The nondimensional form of the integral solution for heat flux at the surface is

(3.44)

The exact solution (E) to this problem for both the early and late (semi-infinite and
finite) regimes is given by

(3.45)

(3.46)

and

(3.47)

Figure 3.9 is a comparison of temperature profiles from the exact and integral solu-
tions for conduction in a body with a constant and uniform temperature at x = 0 (Bi
>>1). The two solutions are plotted at three different times: (i) in the semi-infinite
regime (t < tcrit, < L); (ii) at the critical time between the two regimes (t = tcrit,
= L); and (iii) in the finite slab regime (t > tcrit, L > 0). The integral solutions are
plotted from Eq. (3.22) for the first two curves and Eq. (3.42) for the finite slab
case; the exact solution (calculated to 10 terms in the series) is plotted from Eq.
Transient Conduction Heat Transfer 141

FIGURE 3.9 Comparison of integral and exact solutions to the temperature in a cooled (or
heated) slab for a constant surface temperature (Bi >> 1). Solutions are shown before, at, and
after penetration of the thermal effect reaches the far wall.

(3.45) for all three curves. At each time, the predictions of temperature profiles are
within a few percent of each other. There is less agreement in Figure 3.10 (a) of the
heat extraction rate and the far wall temperature of the finite slab regime, when Q
has slowed considerably and most of the heat has already been removed from the
solid. We see that the scaling analysis does not match the other solutions quantita-
tively, but correctly estimates the overall behavior. Finally, the estimation of L( )
behavior by the scaling analysis is compared in Figure 3.10 (b) to the more detailed
solutions. The scaling analysis does not have a close agreement to those curves, but
once again does have the correct order of magnitude and the general trend. As the
scaling result is roughly within a factor of two with the other estimates and shows
the correct curve shape, with the proper values at the extremes ( = 0 and ∞),
it fulfills its purpose, especially considering the small investment required to obtain
these results.

Example 3.4
Cooling of Injection Molded Polymer
To form a thin sheet of polystyrene (thickness = 2L = 1 mm), the liquid polymer is
injected at Tinj = 270 oC (above its melting temperature) into a mold maintained by
active cooling at Tmold = 30 oC, see Figure 3.11 (a). The heat transfer is symmetric
around the centerline, so dT/dx = 0 there; our analysis will be mapped onto the top
half of the sheet, where x = 0 is the top surface and x = L is the centerline. The sheet
may be ejected from the mold when the polymer has cooled from its initial tempera-
ture to 10 oC below its glass transition temperature, Tg = 100 oC.

k 0.33W/mK 1000 kg/m 3 c 1100 J/kgK k/ c 3 x 10 8 m 2 /s

142 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.10 (a) Nondimensional heat flow at surface of body cooled by a fixed tem-
perature. The semi-infinite solution is shown for < 0 and the finite slab for > 0. (b)
Nondimensional far wall temperature in the finite slab regime, comparing integral, exact,
and scaling analyses.

Find: (a) Calculate the ejection time (te) required for the entire sheet to reach
Te = Tg – 10 oC = 90 oC. (b) Plot the temperature profiles between the surface
(x = 0) and the centerline (x = L) at t = tcrit and t = te.
Solution: (a) Because the temperature at the centerline must have changed
before te, we know the solution lies in the finite body regime and will use
that solution. The centerline temperature is from Eq. (3.41), which can be
rearranged to find the elapsed time,
Transient Conduction Heat Transfer 143

FIGURE 3.11 (a) Schematic of injected polymer cooling in mold. (b) Temperature profiles
when the thermal effect reaches the centerline and when the polymer is entirely below Tf.

1 TL Tinj
ln 1 L or t tcrit 1 4 ln 1 .
3 Tmold Tinj

The entire sheet is below the desired final temperature when TL = Te, so the final
time is

Te Tinj L2 Te Tinj
te tcrit 1 4 ln 1 1 4 lnn 1
Tmold Tinj 12 Tmold Tinj

2
0.0005m 90 o C 270 o C
1 4 ln 1 4.5 s.
12 3 10 8 m 2 /s 30 o C 270 o C

(b) Using Eq. (3.42) to calculate the temperature profiles in the polymer at
t = tcrit = L2/12 = 0.66 s and t = te = 4.5 s, we plot T(x) at these times in
Figure 3.11 (b).
144 An Introduction to Transport Phenomena in Materials Engineering

3.4.3 Heating in a radial systeM

For another example of one-dimensional transient conduction, let us examine the
axisymmetric (∂/∂ = 0) heating of a cylindrical body from the outside, ignoring the
end effects (∂/∂z = 0). The energy equation is

(3.48)

The initial condition is a uniform temperature,

(3.49)

and at the outer radius the temperature is changed at t = 0,

(3.50)

This system is shown in Figure 3.12. Like the previous slab example, this problem can
be solved in two steps. The first is the semi-infinite regime in which the thermal effect of
the change at the outer boundary is not yet felt at the center of the cylinder, Figure 3.12
(a). The thermal penetration depth from r = R is (t). At the edge of the penetration
depth, the temperature has not yet changed and there is no temperature gradient, so

(3.51)

To find the temperature response in the semi-infinite regime, we start by trans-

forming the governing equation (3.48) and boundary conditions, Eqs. (3.50)–(3.51):

(3.52)

and

(3.53)

where

(3.54)

and

(3.55)
Transient Conduction Heat Transfer 145

FIGURE 3.12 Schematic of heated cylinder with temperature profiles for (a) semi-infinite
and (b) finite body regimes.

Using the boundary conditions in Eq. (3.53) to find the coefficients in a quadratic
polynomial, we have the form of the temperature field:
2
2 1. (3.56)

Rearranging Eq. (3.52), we have

2
R R ,
t
146 An Introduction to Transport Phenomena in Materials Engineering

which is then integrated over the affected area, R – ≤ r ≤ R or 0 ≤ ≤ 1.

1 1
2
R d R d (3.57)
0 0 t

Eq. (3.56) is used in this energy integral equation, and the transient term is
changed by the chain rule; thus,

d
.
t dt

Integrating Eq. (3.57) and using Eq. (3.56) produces a differential equation for the
penetration depth,

2 R 1 2R d
2
,
6 dt

which can be solved by separation of variables:

2
1
t 1 . (3.58)
12 3R

(There is an explicit form, (t), to this equation, but it is unnecessarily complicated;

for our purposes this implicit form, t( ), will suffice.)
With Eqs. (3.56) and (3.58), the temperature behavior is known until the thermal
effect reaches the centerline ( =R), at which time we can write from Eq. (3.58),

tcrit R 2 18 . (3.59)

The time for the penetration depth to reach the center is less than that to travel the
same distance in a slab, see Eq. (3.31). In the radial case, the incremental heated vol-
umes are smaller near the center, so less thermal energy is needed to heat the material
closer to the cylinder axis.
At times after tcrit, the domain affected by the outer boundary condition reaches
the center and we are in a finite body regime,Figure 3.12 (b). Nondimensionalizing
the energy equation, Eq. (3.48), with

R r r t tcrit
1 and , (3.60)
R R R2
we have
2
1 2
1 (3.61)
Transient Conduction Heat Transfer 147

0 1, 0 , and 1 o . (3.62)
1

Eq. (3.62) shows that there is no radial temperature gradient (no heat flux) at r = 0
( = 1) and that the unknown, transient temperature at the centerline, (r=0) = o( ),
is used as a boundary condition there. The assumed temperature profile in the body
is now

2
, 1 o 2 1. (3.63)

Integrating Eq. (3.61) over the radius of the body from = 0 to = 1 (r = R to r = 0)

and applying the chain rule to the time derivative thus

produces a differential equation for o ( ):

5 1 o
1 o
3 4
conduction storage

or
o 20 20
o . (3.64)
3 3
The solution is

0 1 exp 20 3 , (3.65)

where the initial condition is

0 0 0.

This expression for the centerline temperature allows us to rewrite Eq. (3.63) for the
transient temperature field in the finite body regime as

2
, exp 20 3 2 1. (3.66)

This result is similar to Eq. (3.42) for the temperature in a finite slab, except the radial
geometry has less volume per unit surface area, causing a faster centerline tempera-
ture rise than the slab (where R L for comparison). Figure 3.13 shows temperature
fields during the entire process as a function of radius.
148 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.13 Temperature profiles at different times through the heating process in a
cylinder.

Example 3.5
Curing Bakelite in a Metal Sample Mounting Press 5
A common practice to prepare metal samples for microscopic examination is to
mount them in Bakelite (a thermosetting polymer resin) to make for easier polishing
and etching. In a lab with a high rate of sample preparation, the cure time for the
Bakelite under temperature and pressure in the mounting press can be a bottleneck.
It has been suggested that the curing time be shortened by the inclusion of 0.05–0.10
volume fraction of a powder with a thermal diffusivity much higher than Bakelite.
If the thermal diffusivity of both materials is known, we can make a crude first
approximation that an effective value of for the mixture is a volume-weighted aver-
age of the filler ( f) and Bakelite ( B),

mix gf f gB B B gf f B ,

where gf is the volume fraction of filler. The Bakelite diffusivity is B = 2 × 10−7 m2/s
and f = 10−5 m2/s.
The goal is to predict the time for the entire cylindrical mount to rise from the
initial temperature, Ti = 25 oC, to the curing temperature, Tcure = 160 oC. The heat is
supplied only at the outer radius of the cylinder and the top and bottom are treated
as insulated (a dubious but simplifying approximation). Because the mount is cured
under pressure and Bakelite can flow under mounting conditions, the contact resis-
tance between the Bakelite and the press is assumed here to be negligible, and the
boundary condition at R = 2 cm is treated as

T r R, t 0 TR 180 o C.

The thermal influence of the metal sample in the mount is neglected.

Transient Conduction Heat Transfer 149

Find: Compare the cure times (tcure) for the Bakelite to reach Tcure at the center-
line (r = 0) for cases with no filler and g f = 0.05 and 0.10. Plot T(t) for these
three cases.
Solution: The first step is to use the solution for the centerline temperature after
tcrit found in Eq. (3.65); that is, we assume that t > tcrit, so we use the finite
body regime. Rearranging that equation and substituting the definitions of
the nondimensional variables and the critical time, Eqs. (3.59) and (3.60).

3 t tcrit mix 3 T Ti
ln 1 o ln 1
20 R2 20 TR Ti

R2 27 T Ti
t 1 ln 1
18 mix 10 TR Ti

To calculate the cure time, tcure, we substitute the mixture thermal diffusivity and
the cure temperature into this expression. For the case with no filler, we calculate:

2
0.02 m 27 160 o C 25 o C
tcure 1 ln 1 726 s;
18 2 10 7 m 2 /s 10 180 o C 25 C

the results for all cases are found in Table 3.1. The temperature histories for the three
cases are plotted in Figure 3.14. The filler decreases cure time by 71–83%, a signifi-
cant improvement. If the filler does not degrade mechanical properties of the mount,
then this preliminary estimate suggests that further exploration through modeling
and experimentation is warranted. To improve the estimates made by this simple
model, several simplifications may be relaxed, including: the conduction in the axial
direction of the cylinder; using a better model for the mixture thermal diffusivity
based on filler size and shape; checking the contact resistance between mount and
press; and accounting for the heat release upon curing.

A merry band of bandits we: A merrier band you’ll never see.

What is it that makes our spirits light? What else but Bakelite?
When others seek for treasures old, for tarnished silver, hefty gold,
What glimmers in our torches’ light? What else but Bakelite?
From Devil King Kun by H. Albertus Boli [7]

TABLE 3.1
Estimated Bakelite Mount Cure Times for a Range of Volume Fraction of
Filler (gf )
gf mix (m2/s) tcrit (s) tcure (s)
0.0 2.0 × 10 –7 111 726
0.05 6.9 × 10–7 32 210
0.10 1.2 × 10–6 19 123
150 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.14 Centerline (r=0) temperature response as a function of filler volume fraction
in a Bakelite mount with a press temperature of 180 oC.

3.5 SPATIAL DEPENDENCE: THE GENERAL SOLUTION

WITH A BALANCE OF CONDUCTION AND
CONVECTION RESISTANCES (BI ~ 1)
In the previous two sections, we have found solutions for transient conduction in
bodies with very small and very large Biot numbers. These two extremes allowed
simplifications to the transient heat transfer model, but how do we handle the case
with Bi ~ 1, in which the thermal resistances inside and outside a body are the same
order of magnitude?
The schematic in Figure 3.1 shows this case. With the resistance to heat flow
important both inside and outside the body, there are important temperature
differences across the body (To – TL) and from the surface to the environment
(T∞ – To), and both the boundary temperatures TL and To change throughout the
process. As in Section 3.4, there is an initial period in which the effect of surface
convection has not reached the far wall and the conduction appears to be acting
in a semi-infinite body. This regime will be dealt with first, followed by a finite
body case. In this section, we should notice a certain familiarity in the form of
the solutions; at any stage, taking the limit of BiL ∞ will produce the result in
the analogous step in Section 3.4, which begins with the assumption of an infinite
Biot number.

3.5.1 Heating in a Slab: early timeS for Bi ~ 1

The semi-infinite regime is modeled by the energy equation (3.16), which is a bal-
ance of conduction and storage of heat. The initial condition is

T x, t 0 Ti , (3.67)
Transient Conduction Heat Transfer 151

and the boundary conditions at the edge of the penetration depth are

T
T x t ,t Ti and 0. (3.68)
x x

The convection heat loss at the surface is

T
k h T To , (3.69)
x x 0

where To t T x 0, t .
Scaling the energy equation in the heated region 0 x uses xref ~ (t) and
Tref ~ (To – Ti). As in Section 3.4.1, we have

~ t. (3.70)

Using this expression and Tref , we then scale the convection boundary condition,
Eq. (3.69), for the surface temperature

To Ti
k ~h T To (3.71)

or

To Ti Bi 1 1
o ~ ~ ~ , (3.72)
T Ti 1 Bi 1 1 Bi 1 k h t

where the Biot number is defined as Bi h k . We see that o(t = 0) = 0 (To = Ti).
The heat flux at the surface is then found from Fourier’s law applied at x = 0:

T To Ti o T Ti h T Ti k T Ti Bi
qo k ~k ~k ~ ~ (3.73)
x x 0 1 Bi 1 Bi

or

qo L L Bi
QSI ~ ~ . (3.74)
k To Ti 1 Bi

An approximate solution to this system can be found using the integral method intro-
duced in Section 3.4.1. First, the energy equation (3.16) and boundary conditions,
Eqs. (3.68) and (3.69), are transformed to

2 2
(3.75)
t
152 An Introduction to Transport Phenomena in Materials Engineering

and

1 0, 0, and Bi 0 1, (3.76)
1 0

where = (T – Ti)/(T∞ – Ti) and = x/ . Applying these boundary conditions to find

the coefficients in a second-order polynomial for produces

T Ti Bi 2 2
2 1, (3.77)
T Ti 2 Bi Bi

the nondimensional transient temperature field in the semi-infinite regime. The inte-
gral of Eq. (3.75) over the penetration depth,
2
1 1
2
d d (3.78)
0 0 t
is evaluated, as was the energy integral equation (3.23) in Section 3.4.1 to find the
penetration depth,

12 t. (3.79)

Using the solution for temperature in Eq. (3.77), we can find the surface temperature
as it rises during the semi-infinite regime. Defining BiL = hL/k, the nondimensional
surface temperature is

To Ti 2 2
1 1 . (3.80)
T Ti 2 Bi 2 BiL L

Applying the same definition of the part Biot number, BiL, to the temperature profile,
Eq. (3.77),

Bi L L 2 2 L
2 1. (3.81)
2 Bi L L Bi L

Figure 3.15 (a) shows the progression of the temperature field for two different values
of BiL. The shape of the temperature curves is the same as for the BiL >> 1 case seen
in Figure 3.6 (a), except that instead of being instantly pegged to o = 1 (To = T∞),
the surface temperature rises from zero gradually, increasing the surface temperature
gradient and heat flux. At low BiL, the heat added is absorbed at almost the same rate
throughout the body, so the surface temperature rise is much slower. As BiL increases,
the rising conduction resistance slows the heat spreading into the body and keeps it
near the surface, increasing the rate at which o rises.
The surface heat flux can be found using Eqs. (3.77) and (3.79),

T k T Ti k T Ti 2Bi
q0 k ,
x 0 0
2 Bi
Transient Conduction Heat Transfer 153

FIGURE 3.15 Temperature during heating of a semi-infinite body. (a) Progression of pro-
files for two values of BiL. ( /L= 1 occurs at tcrit.) (b) Surface response over time for range
of BiL.

which can be written in nondimensional form as

q0 L L 2Bi L 2Bi
QSI . (3.82)
k T Ti 12 t 2 Bi 2 Bi

3.5.2 Heating in a slab: late tiMes For Bi ~ 1

Once the thermal penetration depth, , has reached the far wall (x=L) at time t = tcrit,
the solution regime shifts to that of a finite body; the critical time is found from Eq.
(3.79):

tcrit L2 . (3.83)
12
154 An Introduction to Transport Phenomena in Materials Engineering

The only difference between this solution and the Bi 1 case in Section 3.4.2 is the
convection condition at the surface, Eq. (3.69). The far wall is still insulated, and the
temperature there is an unknown function we use as a boundary condition.

TL t T x L, t or L 1, (3.84)

We normalize the energy equation and boundary conditions using = x/L and
=(t-tcrit) /L2:
2

2
(3.85)

Bi L 1 0 , 0, and 1 L . (3.86)
0 1

After tcrit, the maximum rate of temperature rise in the body always takes place at the
far wall, so we scale the storage term as

T Tref , S TL Ti
. (3.87)
t t t tcrit

The temperature reference in the conduction term is a spatial difference, not a tem-
poral one. The temperature difference across the body (xref ~ L) is Tref,C ~ (To ~ TL),
so the energy conservation equation is scaled as

TL Ti
~ 2 To TL
t tcrit L
storage ~ conduction
or

TL Ti
~ .
To TL

Rearranging this result,

Ti To
TL ~ and L ~ o . (3.88)
1 1

To find o, we scale the convection boundary condition:

TL To
h T To ~ k
L

hL
T Ti Ti To ~ TL To
k

Bi L 1 o ~ L o . (3.89)
Transient Conduction Heat Transfer 155

Combining Eqs. (3.88) and (3.89) from the energy equation and the boundary con-
dition gives

Bi L Bi L Bi L Bi L
L ~ and o ~ ~ L . (3.90)
1 Bi L Bi L 1 Bi L Bi L 1 Bi L Bi L

These estimates confirm that, at the transition from semi-infinite to finite body solu-
tions (t = tcrit),

Bi L
L 0 0 and 0 0 , (3.91)
1 Bi L

consistent with Section 3.5.1. As ∞, both temperatures go to 1, with L < o

always.
These estimates of the surface and far wall temperatures are shown in Figure 3.16
for two values of BiL. As expected, the two temperatures for the low BiL case are
within 10% of the overall temperature difference throughout the entire heating pro-
cess. The high BiL case has a surface within 10% of the overall difference with a
rate of far wall temperature rise, which is high at first and decreases with time. The
medium value (BiL = 1) has a behavior similar to the high case, except the surface
temperature begins the finite body regime at a lower value.
To find better estimates of the temperature behavior, the integral method is again
employed. A second-order polynomial of ( ) is used and, using the boundary con-
ditions in Eq. (3.86), is found to be

Bi L 2 2
, 1 L 2 1, (3.92)
1 Bi L Bi L

FIGURE 3.16 Surface and far wall temperature histories from scaling in the finite body
regime.
156 An Introduction to Transport Phenomena in Materials Engineering

where L( ) is still unknown. Integrating over the domain (0 ≤ x ≤ L or 0 ≤ ≤ 1) and

using the guessed temperature field, Eq. (3.92),

2
1 1
2
d d
0 0

gives

L 3Bi L 3Bi L 3Bi L

L L 1 exp . (3.93)
3 Bi L 3 Bi L 3 Bi L

The complete expression for the transient temperature field in the finite body
regime is

3Bi L Bi L 2 2
, exp 2 1, (3.94)
3 Bi L 2 Bi L Bi L

and the surface is

3Bi L 2
o 1 exp . (3.95)
3 Bi L 2 Bi L

The temperature profiles at different times and BiL and the histories of o( ) and L( )
are plotted in Figure 3.17. With lower relative convection resistance (BiL = 10 >> 1),
the temperatures in the bodies rise faster and with steeper gradients near the surface
early in the finite body regime; the results are close to the constant surface tempera-
ture case. The low Biot number results (BiL = 0.1 << 1) are in the lumped capacitance
regime; Figure 3.17 (a) shows almost uniform and rising temperature profiles. The
case with BiL = 1 splits the difference, with a noticeable temperature gradient across
the body and surface temperature that begins well below the ambient and rises slowly
through the process. This plot shows that the behavior of the solution approaches
those in Section 3.4 at extreme Bi values.
The heat flux is found:

T k T Ti k T Ti 3Bi L 2Bi L
q0 k exp
x 0 L 0 L 3 Bi L 2 Bi L

and, in nondimensional form,

q0 L 3Bi L 2Bi L
QFB exp (3.96)
k T Ti 3 Bi L 2 Bi L
Transient Conduction Heat Transfer 157

FIGURE 3.17 For finite body regime, thermal behavior for selected Biot numbers (0.1, 1,
10): (a) temperature distributions at three times; (b) surface ( o) and far wall ( L) temperature
histories.

The nondimensional heat fluxes for the semi-infinite solution, Eq. (3.82), and the
finite body solution, Eq. (3.96), are plotted in Figure 3.18. The heat flux is zero at
t = 0 because there is initially no temperature gradient inside the body. The convec-
tion heats the surface, causing a gradient and a heat flux. This flux increases during
the semi-infinite regime, as the gradient at x = 0 grows in time, as shown in Fig-
ure 3.15 (a). Once the heated domain stops growing (t = tcrit), the far wall temperature
rises, decreasing the temperature gradient until it reaches zero.
158 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.18 Nondimensional transient surface heat fluxes at BiL = 1, 10, and 100. Negative
values of are in the semi-infinite regime and positive values are t > tcrit.

Example 3.6
Temperature Gradients during Tempering of a Glass Sheet
Cooling ceramics must be done with care to prevent cracking, as they tend to have
low thermal diffusivities and so can be subject to high temperature gradients during
the process. These gradients cause high gradients in thermally induced strain, which
may lead to brittle fracture. The models developed in this section can be used to
generate the transient temperatures and temperature gradients used in thermal stress
modeling to predict the likelihood of part distortion and cracking. The mechanical
portion of this problem is beyond the scope of this text, but does require the informa-
tion provided by these heat transfer calculations.
A 2-cm thick soda-lime glass sheet is air-cooled on one side from Ti = 600 oC.
External conditions are set so that the heat transfer coefficient can be either h1 = 40
W/m2K or h2 = 250 W/m2K. The ambient temperature is T∞ = 20 oC, and the glass
properties are: = 2500 kg/m3, c = 500 J/kgK, k = 1 W/mK, and = k/ c = 8 ×
10–7 m2/s.

Find: (a) Calculate the Biot numbers (BiL) for the plate. (b) Plot surface tem-
perature, T(x=0, t), the temperature in the sheet, T(x=L/4,t), and the differ-
ence between them, all as functions of time up to t = 300 s.
Solution: (a) The BiL for these two cases are:

h1 L 40 W m 2 K 0.02 m
BiL1 0.8 and
k 1W mK
h2 L 250 W m 2 K 0.05 m
BiL 2 5
k 1W mK

Both cases have BiL ~ (1), so the solutions from Section 3.5 apply.
Transient Conduction Heat Transfer 159

FIGURE 3.19 (a) T(t) for the surface (x = 0) and a point inside the glass sheet (x = L/4 = 0.5 cm)
for two BiL. (b) Transient difference between temperatures at x = 0 and x = 0.5 cm.
160 An Introduction to Transport Phenomena in Materials Engineering

and so there is a large heat loss rate. The difference reaches a maximum when
reaches x = L/4. With lower surface temperatures in the finite body regime, the heat
loss slows and the interior (now of fixed extent) begins to catch up to the temperature
of the surface. The magnitude of this temperature difference plotted in Figure 3.19 (b)
is seen to increase quickly, then with a gradual decline in the finite body regime. This
behavior suggests that the largest differential strains due to thermal contraction will
occur early in the process. The plot also shows that increasing the heat transfer coef-
ficient by about a factor of six affected the maximum T by a lower factor.

3.6 SOLIDIFICATION
Up to this point in our discussion of conduction heat transfer, we have assumed that
any change in enthalpy of a system causes only a change in temperature, a phenom-
enon known as sensible heating or cooling. However, enthalpy changes can cause
changes of phase as well, which require the supply or removal of latent heat, so-called
because it is “hidden” and not apparent until there is a phase change. In this section
we will examine the thermal behavior of materials with solid-liquid phase changes.
If we extract heat at a constant rate from a pure metal at an initial temperature (Ti)
above its melting point (Tm), the material begins cooling at a rate dependent on its
specific heat,

dH d cT or dh cdT, (3.97)

where the latter expression assumes constant properties. This sensible cooling of a
liquid is seen at early times in Figure 3.20, and it continues until the melting temper-
ature is reached. At that point, solid and liquid exist in equilibrium and the thermal
behavior changes. In a pure substance, the solid-liquid equilibrium is at a single tem-
perature, and the metal remains constant at Tm while the enthalpy required to effect
the phase change, the latent heat Lf, is removed. This thermal arrest begins when T
reaches Tm in Figure 3.20 and liquid begins to freeze and ends when the last of the
liquid has solidified. Defining the mass fraction of liquid as

mL
fL ,
mTOT

we say that the enthalpy removed during the isothermal solid-liquid phase change is

dH d mL f fL or dh L f dfL , (3.98)

assuming constant properties. The phase change starts at fL = 1 (all liquid). Once
fL = 0 (all solid), there is no more phase transformation possible with further enthalpy
removal, and sensible cooling resumes according to Eq. (3.97) with solid properties.
The behavior of the freezing binary alloy has some differences compared to
the pure metal, as seen in Figure 3.21. With an alloy composition of Co, the phase
diagram predicts a single phase, sensible cooling until T = TLiq. At that point, the
Transient Conduction Heat Transfer 161

FIGURE 3.20 Temperature history of pure metal with a constant heat extraction rate.

nature of a binary alloy is that the liquid freezes to solid over a temperature range
(TLiq – TEut). In that range, the two phases (L + ) are at equilibrium and the enthalpy
removal causes simultaneous phase change and temperature drop. When the tem-
perature reaches TEut, a second solid phase ( ) begins to freeze along with the first
( ) in a eutectic reaction. With three phases in equilibrium (L + ), there must
be an isothermal change during further enthalpy removal until the last liquid disap-
pears. When that condition occurs, the completely solid metal (in the + phases)
continues to cool sensibly.
The previous descriptions of the thermal behavior of metals during solidifica-
tion are based on equilibrium thermodynamics and do not include non-equilibrium
effects. For a much more thorough description of the physics of solidification and the
relationship of heat transfer to microstructural development, see the excellent text by
Dantzig and Rappaz [8].
In the following sections, we will develop one-dimensional models of the tran-
sient conduction heat transfer with phase change to illustrate the behavior of simple
systems. (Here we find integral solutions to these problems; a thorough presentation
of the exact solutions is found in Dantzig and Rappaz [8].) These solutions are useful
for estimates of the solidification time and the mold and metal temperature histories
during more complex configurations, and can provide results related to microstruc-
tural development.

3.6.1 energy balances witH pHase cHange

Earlier we assumed the mass from which we remove heat behaved in a uniform man-
ner as it cooled and froze, like lumped capacitor. However, most practical systems
exhibit temperature gradients during such a process, so here we derive a thermal
energy balance from a control volume analysis, as in Section 2.4.
Restricting the analysis to one dimension (x), we repeat Figure 2.6 and write a
generic energy balance with no thermal energy generation:

Ustor U in U out . (3.99)

162 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.21 Response of a binary eutectic alloy to a constant heat extraction rate (a) phase
diagram (b) temperature history.

The storage term, Ustor , is merely the time rate of change of enthalpy (∂H/∂t). For
a single phase, a change in enthalpy is from Eq. (3.97), but that expression only
accounts for the sensible heating. If two phases are possible, we can write the spe-
cific enthalpy (h=H/m) as a mixture of solid and liquid enthalpies weighted by mass
fraction,

h fs hs fL hL , (3.100)

where fi mi ms mL and fs fs 1. The solid enthalpy is

hs cT,
Transient Conduction Heat Transfer 163

and the liquid enthalpy is

hL cT Lf ,

assuming that cs cL c, usually a reasonable approximation near the melting point.

The total enthalpy of the control volume is then

H h fs cT fL cT Lf

or, where the volume is that of the infinitesimal control volume in Figure 2.7,
H dxdydz cT fL L f
sensible laten (3.101)
heat heat
With that definition, we can write

H T fL
Ustor dxdydz c Lf . (3.102)
t t t

In Eq. (3.102), an addition of enthalpy to the control volume (∂H/∂t > 0) results in a
temperature rise (∂T/∂t > 0) or melting (∂fL/∂t > 0), or both.
An expression for the net flow of heat in and out of the control volume (in one
dimension) is the same as in Eq. (2.25),
2
T
U in U out dxdydz k , (3.103)
x2

which gives us, with Eqs. (3.99) and (3.102),

2
T T fL
c k Lf .
t x2 t
sensible conduction in latent (3.104)
heat x-direction heat

The latent heat term is only active when and where phase change is occurring, as
determined by the phase diagram, e.g., Figure 3.21 (a). When freezing isothermally,
as with a pure metal (Figure 3.20) or at the eutectic in a binary alloy, ∂T/∂t = 0 and
only freezing occurs. In single-phase regions, only sensible changes in heat are possi-
ble and ∂fL/∂t = 0. In two phase areas of alloy phase diagrams, all terms in Eq. (3.104)
are active.
Focusing on the solid-liquid interface during freezing, we can draw an infinites-
imally thin control volume around that interface. In Figure 3.22, the heat flow from
the solid, qs, enters the interface and qL, the heat flow into the liquid, leaves the inter-
face, and latent heat is released as the interface position, x M t moves to the right.
The heat flows are only by conduction, so

dTs dTL
qs ks A and qL kL A . (3.105)
dx M t
dx M t
164 An Introduction to Transport Phenomena in Materials Engineering

To find the latent heat release rate, we start by stating that the total latent heat
released to form the solid for x = 0 to M is the decrease in total enthalpy to effect the
phase change:

Hf HL Hs mL f s Lf s L f AM .

The latent heat release rate is then

d s L f AM dM
s Lf A s L f AVint , (3.106)
dt dt

when Vint is the interface velocity. The heat balance at the interface is:

dTs dT dM
ks kL L sLf .
dx M dx M dt
conduction conduction latent heat (3.107)
frrom solid to liquid release

This description of thermal behavior at the interface, known as the Stefan condi-
tion, shows that the velocity of the interface (the solidification rate) is controlled
by the net heat flow in and out of the solid-liquid boundary. Eq. (3.107) also shows
why Figure 3.22 is drawn with different temperature slopes in the solid and the liq-
uid. If ks kL , then the temperature gradients will be different even without a phase
change, but the added heat release (or absorption during melting) also contributes to
the change in slope at the interface.

FIGURE 3.22 Heat balance in a control volume at a moving solid-liquid interface.

Transient Conduction Heat Transfer 165

3.6.2 solidiFication oF a pure substance

Figure 3.23 shows a simple one-dimensional, planar configuration that is our basis
for the approximate treatment of a metal freezing in a mold. We have assumed no
superheat (the liquid metal enters the mold at the melting temperature, TM, the metal
is pure, and the semi-infinite mold is initially at Ti. When the liquid is poured, the
interface temperature between metal and mold rapidly goes to To, an as-yet unknown
value. In designing such a process, we seek knowledge of the total solidification time
(it is good to know when your process is finished) and the interface temperature (so
we know if we might damage the mold).
In the general case, the heat transferred to the mold comes from latent heat
released when the liquid freezes and the sensible heat taken to cool the solid. The
heat transfer down the temperature gradient in Figure 3.23 can be considered using
a crude resistance network, Figure 3.24, which has heat flowing through condition
resistances in the solid metal and the mold.
These quantities can be approximated as

M
Rmold and Rsolid . (3.108)
km A ks A

The ratio of these quantities,

Rsolid M km
, (3.109)
Rmold ks

can give us insight into the thermal behavior of the process and sometimes lead to
simplified solutions. From Ohm’s law, Eq. (2.7), and Eq. (3.109), we can estimate the
interface temperature as a function on the ratio of resistances:
Rsolid Ts qm TM To
Rmold qs Tm To Ti

FIGURE 3.23 Temperature field in the metal and mold during the freezing of a pure metal.
166 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.24 Resistance network crudely representing main thermal resistances in

Figure 3.23 .

or

M km M km
To TM Ti 1 . (3.110)
ks ks

These are crude estimates of the resistances and the interface temperature because
the problem is transient, with unaccounted-for sensible heat changes in the solid
metal, and the temperature field is not actually linear. However, the model will suf-
fice as an initial estimate of system behavior.
In the following two subsections we will examine two cases. The first is solidifi-
cation in a thermally resistant mold,

Rsolid Rmold 1, (3.111)

and the second is when both resistances play a role,

Rsolid Rmold 1. (3.112)

The case with all significant resistance in the solid metal, Rsolid >> Rmold, is left as an
exercise for the reader.

3.6.2.1 Thermally Resistant Mold

In the case in which the mold is significantly more thermally resistant to conduc-
tion than the metal, Eq. (3.111), we can use Eq. (3.110) to estimate the interface
temperature.

TM M km ks Ti TM Rsolid Rmold Ti
To ~ TM
M km ks 1 Rsolid Rmold 1

Because To ~ TM, almost all the temperature difference between the liquid and the
unheated mold is found in the mold and we can redrawFigure 3.23 with no temper-
ature change in the solid.
The first step in analyzing this special case is to look at it from the point of view
of the mold, which is initially at Ti and, when the liquid enters at t = 0, has its surface
at x ≥ 0 suddenly raised to TM. This case is found earlier in the chapter, where the
penetration depth, temperature profile, and heat flux are

12 t , (3.25)
Transient Conduction Heat Transfer 167

FIGURE 3.25 Temperature field during pure metal solidification with a thermally resistant
mold.

T Ti 2
2 1, (3.22)
Tm Ti

where = −x/ , and

dTm k TM Ti k TM Ti
q0 qo x 0 k . (3.27)
dx 0 0 3 t

Thus the mold is heated during this process, but what is the source of that heat?
ExaminingFigure 3.26 (a), we see the heat flux out of the solid, qo , is balanced by
the latent heat release at the solid-liquid interface.

k c m dM
q0 TM Ti s Lf qLHR (3.113)
3t dt

This equality shows that the rate of growth of solid (dM/dt) is entirely controlled
by how fast the mold can absorb heat (qo ).
Solving this differential equation for the position of the interface, with the initial
condition M(t=0) = 0, gives

2 TM Ti
M t k c m
t 3. (3.114)
sLf

Eq. (3.113) can also be rearranged to find the interface velocity,

dM TM Ti
Vint k c m
3t . (3.115)
dt sLf

Sketching M(t) and Vint(t) in Figure 3.26 (b), we see the solidification front forms
instantly and grows rapidly, although it slows with time. The deceleration is caused
168 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.26 Freezing of a pure metal in a thermally resistant mold. (a) Heat flow in and
out of solid metal. (b) Sketch of solid-liquid interface position and velocity.

by the heating of the mold as the penetration depth, , lengthens and the temperature
gradient at the wall decreases.
If a slab has a thickness of 2Mf and is cooled by the same type of mold on both
sides, then the solid fronts meet at the centerline at the final solidification time, tf,
when M(t=tf)=Mf. From Eq. (3.114), the solidification time is
2
3 s Lf M f 1
tf . (3.116)
4 TM Ti k c m

So we have an estimate for the solidification time, and how we might alter it if need
be. The properties of the metal ( s, Lf, TM) and its geometry (Mf) are fixed by material
selection and part design, but to slow the process, we could change the mold material
to decrease its heat diffusivity, (k c)m, or preheat the mold before pouring to decrease
(TM – Ti).
Transient Conduction Heat Transfer 169

At the beginning of the analysis, we assumed that all significant thermal resist-
ance was in the mold, Rsolid/Rmold << 1, and here we look for a way to evaluate that
assumption. Using the definition for the resistance ratio, Eq. (3.109), and our results
for (t) and M(t), Eqs. (3.25) and (3.114),

Rsolid M km k c m
TM To
1
Rmold ks 3ks s Lf

is required for this approximation to be reasonable.

Example 3.7
Sand Casting of Steel
A steel slab (Mf = 2 cm) is cast in a sand mold. As a first approximation, the steel
will be treated as pure iron, with properties ks = 20 W/mK, s = 8200 kg/m3, and
cs = 800 J/kgK. The mold properties are km = 0.52 W/mK, m = 1600 kg/m3, and
cm = 1170 J/kgK. The initial mold temperature is Ti = 30 oC.

Find: (a) The final solidification time, tf; (b) the thermal penetration depth into
the mold at tf; (c) show that Rsolid/Rmold << 1.
Solution: (a) From Eq. (3.116),

2
3 8200 kg/m 3 270 kJ/kg 0.02 m
tf
4 1450 o C 30 o C
1
749s
0.552 W/mK 1600 kg/m 3 1170 J/kgK

(b) The penetration depth is found by

12 m tf 12 2.78 10 7 m 2 /s 749 s 0.11m.

(c) The ratio of mold and solid resistances is

Rsolid M km k c m
TM To
Rmold ks 3ks s Lf

0.52 W/mK 1600 kg/m 3 1170 J/kgK 1450 o C 30 o C

0.0104 1.
3 20 W/mK 8200 kg/m 3 270 kJ/kg

So the use of the thermally resistant mold model is reasonable in this example.
170 An Introduction to Transport Phenomena in Materials Engineering

3.6.2.2 General Case

A more general case of planar, one-dimensional solidification is one in which the
solid metal and the mold have thermal resistances of the same order of magnitude,

O Rsolid ~ O Rmold .

Practical examples include systems such as permanent mold or high-pressure die

casting of metals in iron or steel molds, or injection molding of polymers into alu-
minum or steel molds. As shown in Figure 3.23, there are significant temperature
gradients on both sides of the mold-metal boundary.
One desired result of the model of this system, the final solidification time, tf, is
the same as in the previous section. Here, we also will derive an expression for the
interface temperature, To, for two reasons. First, a high temperature may damage
the mold, especially if it is subjected to thermal cycling through repeated use, and
the metal, if the thermal gradient is high enough to generate damage through ther-
mal stresses. The interface temperature is treated as constant throughout the process,
assuming a transient much shorter than tf.
To find tf and To for a slab, we must solve a one-dimensional, transient model
of conduction and phase change for the temperature field and solid-liquid interface
position. We start with the mold temperature, Tm, which is governed by the energy
equation
2
Tm Tm
m , (3.117)
x2 t
subject to the conditions

Tm
Tm ( x,t 0) Ti Tm ( x 0, t 0) To Tm ( x ,t ) Ti 0. (3.118)
x x

The reader should recognize this model of conduction into the mold as the same as
the semi-infinite problem in Section 3.4.1. Transforming the system with

Tm Ti x
m and , (3.119)
TM Ti

the system can be solved as before:

2
m o 2 1 and t 12 m t. (3.120)

The nondimensional interface temperature is o = (To – Ti)/(TM – Ti), a constant value

in this problem. This solution gives the heat flux into the mold from the solidifying
and cooling metal:

dTm TM Ti d m km TM Ti
qo km km o . (3.121)
dx x 0
d 0
3 m t
Transient Conduction Heat Transfer 171

If we treat the temperature field in the solid as a linear profile, then we can find it
from the boundary conditions:

Ts x 0,t To and Ts x M (t ),t TM

or

s 0 o and s 1 1,

where s = (Ts – Ti)/(TM – Ti) and = x/M(t). The temperature profile is then

s ( ) 1 o o . (3.122)

The expressions for the two temperature profiles leave two unknown parameters, the
position of the solid-liquid front (M) and the mold-solid interface temperature (To).
To find the first, we apply the Stefan condition, Eq. (3.107), to the solidification front
at x = M(t), where all the latent heat release is happening.

Ts TLiquid dM
ks kL s Lf (3.123)
x M
x dt
M

With ∂ s/∂ = 1 – o , Eq. (3.123) is

ks TM To s dM
s Lf . (3.124)
M 1
dt

Separating variables M and t and applying the initial condition M(t=0) = 0, we solve
Eq. (3.124) for the solidification front position,
1
M t 2ks TM Ti t s Lf 1 o
2 . (3.125)

The final solidification time (tf) for a symmetrically cooled slab 2Mf thick can be
found when M = Mf and t = tf:

M 2f s Lf 1
tf . (3.126)
2ks TM Ti 1 o

How easily that heat is conducted through the solid is represented in the denominator
of Eq. (3.126), where a high thermal conductivity (ks), a high overall temperature
difference (TM – Ti), and a low interface temperature ( o) contribute to a faster solid-
ification time.
A continuous flux occurs across the mold-solid boundary at x = 0, and

Tm Ts
qmold km ks qsolid (3.127)
x x 0
x x 0
172 An Introduction to Transport Phenomena in Materials Engineering

is used here to find the interface temperature. The derivatives of Eqs. (3.120) and
(3.122) give the temperature gradients on either side of the interface.

TM Ti m
TM Ti s
km o ks
0
M 0

1/ 2
2km o s Lf 1 o
ks 1/ 2
12 mt
2ks TM To t 1 o

1/ 2
k c m o ks s Lf 1/ 2
1 o (3.128)
3 2 TM To

Rearranging and defining a constant

1/ 2
2 k c m
TM To
Co , (3.129)
3ks s Lf

Eq. (3.128) can be written as a quadratic polynomial in o ,

1/ 2
Co o 1 o or Co2 2
o o 1 0,

which is solved as

To Ti 1
o 1 4Co2 1 . (3.130)
TM Ti 2Co2

Therefore, the interface temperature can be found by calculating first Co, Eq. (3.129),
then o, Eq. (3.130).

Example 3.8
Casting a NiAl Slab in Different Molds
When a mold is not actively cooled and the thermal resistances of the mold and
the solid metal are the same order of magnitude, heat from the solidifying metal
will raise the mold temperature and reduce the solid temperature. This mold heating
can be a problem because of the induced mold thermal stresses and wear, but also
because, under extreme conditions, the mold may begin to melt. In this example,
we pour NiAl (at TM = 1682 oC, its melting temperature), an intermetallic that melts
congruently, first into a copper slab mold, then into a sand mold; both are initially
at 30 oC.
Properties:
NiAl: k = 76 W/mK = 5900 kg/m3 c = 640 J/kgK Lf = 300,000 J/kg
copper: k = 300 W/mK = 8500 kg/m3 c = 480 J/kgK TM,Cu = 1084 oC
sand: k = 0.52 W/mK = 1600 kg/m3 c = 1170 J/kgK Mf = 0.05 m
Transient Conduction Heat Transfer 173

Find: (a) the interface temperature, To, between the metal and the copper mold.
Will the copper mold melt? (b) The temperature of the interface of liquid
NiAl in a sand mold. (c) How the interface temperatures for the two mold
compare. (d) The solidification times (tf ) for both molds.
Solution: (a) We use Eqs. (3.129) and (3.130) to find the interface temperature
between NiAl liquid and a copper mold.
12
o 3 o
2 300 W m C 8500 kg m 480 J kg C
Co 1682 o C 30 o C 3.17
3 76 W m o C 5900 kg m 3 300,0000 J kg

To Ti 1 2
o 2
1 4 3.17 1 0.270
TM Ti 2 3.17

To TM Ti o Ti 1682 o C 30 o C 0.270 30 o C 476 o C

The interface temperature is much lower than the melting temperature of copper.

(b) Find the temperature of the interface of liquid NiAl in a sand mold. How
does the interface temperature compare to that in part (a) and why?
12
o 3 o
2 0.52 W m C 1600 kg m 1170 J kg C
Co 1682 o C 30 o C
3 76 W m o C 5900 kg m 3 300, 000 J kg

0.0893 o C

To Ti 1 2
o 2
1 4 0.0893o C 1 0.992
TM Ti 2 0.0893 C o

To TM Ti o Ti 1682 o C 30 o C 0.992 30 o C 1669 o C

(c) The large increase in the thermal resistance of the mold causes the inter-
face temperature to rise to almost the melting temperature of the metal.

(d) Calculate the solidification times (tf ) for the cases in part (a) and (b) and
compare.
2
M 2f S Lf 0.05m 5900 kg m 3 300, 000 J kg
Copper mold: t f 24.1s
2 kS TM To 2 76 W m o C 1682 o C 476 o C

2
M 2f S Lf 0.05 m 5900 kg m 3 300, 000 J kg
Sand mold: t f 2246 s
2 kS TM To 2 76 W m o C 1682 o C 1669o C
174 An Introduction to Transport Phenomena in Materials Engineering

The much higher resistance in the sand (compared to the copper) slows heat
extraction so much that the solidification time is two orders of magnitude longer.
Although the thermal capacity ( c) for the two mold materials is different by a fac-
tor of 4, the ~600X difference in thermal conductivity is the dominant factor in the
slower heat transfer through the mold.

3.7 SUMMARY
When heat flows in and out of a body and is not entirely balanced by internal heat
generation, thermal energy is either accumulated or depleted in that body. This
change in thermal energy causes transient temperature fields and, in the case of solid-
ification, phase changes.
The Biot number (the ratio of conduction and conduction resistances) is now
applied to these transient conduction problems, and its value is used to make physical
arguments for simplifications of the thermal energy conservation equation. For cases
with Bi << 1, the internal temperature of a body is uniform while the body is heated or
cooled; all of the thermal resistance is at the boundary with its environment. This sit-
uation is the “lumped capacitance” problem. For the other extreme, Bi >> 1, the inter-
nal resistance dominates, and the convection resistance is so small that the boundary
takes on the environmental temperature. All of the temperature gradient is internal
to the body. In the middle case, Bi ~ (1), the internal conduction resistance and the
external convection resistance are the same order of magnitude, so there are tempera-
ture differences in each region and the surface temperature changes slowly compared
to the overall rate of cooling or heating. In the two latter cases, there is a finite time
before the effect of changed boundary condition spreads throughout the entire body.
Solutions have been developed in one-dimensional slabs and cylinders, using
scaling, exact solutions, and integral analysis. Many examples of practical applica-
tions are studied using these solutions. Transient conduction is also shown to control
the one-dimensional solidification behavior for pure substances, with derivations of
models to predict solidification times and transient temperature fields.

3.8 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

3.1 We want to measure the heat transfer coefficient for forced convection over
a warm sphere immersed in a stream of cooler air. The rate of decrease in
the surface temperature is measured by means of a thermocouple attached
to its surface. The initial uniform temperature of the 0.02 m diameter
sphere is 75 oC, and after immersion in the air stream (T∞ = 25 oC) for 60 s
the surface temperature is 68 oC.
3.2 Steel rods of 4 cm diameter are heat treated by being passed through a
furnace in which the atmosphere is at 700 oC. Assume the heating is all
convection with a heat transfer coefficient of 135 W/m2K and that the rods
Transient Conduction Heat Transfer 175

are much shorter than the furnace. If the furnace is 5 m in length and the
rods enter at 50 oC, at what speed must they pass through the furnace to be
at 600 oC when they leave it? krod = 20 W/m2K.
3.3 A copper-constantan (Ni-55wt%Cu) thermocouple with a bead (diame-
ter = 2 mm) initially at 25 oC is inserted into a molten salt bath (Tbath = 275 oC)
in order to monitor this quenchant used for steel heat treating. The proper-
ties of the thermocouple bead and the heat transfer coefficient are:

= 8900 kg/m3 c = 400 J/kgK k = 20 W/mK h = 100 W/m2K

What is the response time for this bead to reach 99% of the temperature
difference between the initial bead and the salt bath?
3.4 After being annealed at 700 oC, steel spheres of diameter 0.01 m are allowed
to cool slowly in still air at 50 oC under conditions in which the heat transfer
coefficient is 20 W/mK. Calculate the time required for the sphere center
temperature to reach 100 oC. Show that the model you use is valid.
3.5 A plate (1 cm thick) made of eutectoid plain carbon steel is quenched in two
different media (Cases 1 and 2) from an initial temperature, To = 760 oC. In
both cases, check to see if the lumped capacitance method can be used to
estimate the temperature history of the plate. As a first approximation, use
the lumped capacitance model in both cases. Plot the centerline tempera-
tures as a function of time on the transformation diagram in Figure 3.27.
(a) For both cases, how long before the centerline of the plate reaches
200 oC? (b) Do those T(t) curves cross over the transformation start line?

Properties: c = 500 J/kgK = 7850 kg/m3 k = 50 W/mK.

Case 1 (quench in oil): h = 1000 W/m2K T∞ = 100 oC

Case 2 (quench in gaseous nitrogen): h = 80 W/m2K T∞ = −180 oC

FIGURE 3.27 Time-transformation diagram for eutectoid plain carbon steel used in
Problem 3.5.
176 An Introduction to Transport Phenomena in Materials Engineering

3.6 Spherical alumina particles (40 µm diameter) are introduced into an acety-
lene flame to melt them and to spray them onto a surface on which they are
deposited as a coating. Ideally, the particles should be melted before they
strike the surface. (a) Can you use a lumped capacitance approach to model-
ing the particle temperature in this problem? Why or why not? (b) If the jet
velocity is 100 m/s, what is the minimum distance the nozzle should be from
the surface (L) so that the droplets impact the surface after melting has begun?

flame temperature (Tf ) = 3000 K initial particle temperature

(Ti) = 300 K
melting temperature (Tm) = 2300 K heat transfer coefficient =
105 W/m2K
alumina properties: c = 1700 J/kgK = 3800 kg/m3
k = 40 W/mK
air properties: c = 2700 J/kgK = 0.11 kg/m3 k = 0.49 W/mK

FIGURE 3.28 Thermal spray system in Problem 3.6.

3.7 Consider the heat-up of a one-dimensional slab (L thick) initially at tem-

perature Ti and subjected to a constant temperature bath at TB. Assume BiL
>> 1. (a) Scale time, temperature, and distance in this domain by select-
ing appropriate reference quantities. Are all of the reference values known
quantities? (b) Using the results of part (a), estimate the time required for
~ L/2, where is the distance into the part that is affected by the bound-
ary conditions (i.e., T > Ti). (c) If the slab is replaced by one with twice the
thermal conductivity, what happens to the time found in part (b)? What
happens if the part is halved in thickness?
3.8 The copper wall of an electroslag remelting furnace is initially at T W, and
the outer surface of the wall (x = L) is maintained at T W by the swiftly
moving water in a copper cooling jacket. The temperature on the inside
Transient Conduction Heat Transfer 177

surface of the copper wall is increased at a finite rate by the relatively sud-
den contact with a very hot slag. Approximate the thermal contact with the
slag as a constant and uniform heat flux into the wall from the furnace, q f .
(a) Sketch and label the system. Using scaling analysis, find an expression
for the steady state value of (Tf − T W ), where Tf is the furnace side tempera-
ture of the copper (at x = 0). (b) Scale the transient energy equation and
the boundary condition on the inside wall (x = 0) to estimate (Tf (t) – T W )
before heat penetrates to far wall (t = 0). (c) Using results from parts (a) and
(b), estimate the time the temperature field will take to reach steady state;
comment on the dependence of that time on q f .
3.9 A square part (length = 2L) is heated uniformly through its volume by
passing a current through it, while the sides of the part are maintained at
a constant and uniform temperature, Ts. We would like to quickly estimate
the time dependence of the maximum temperature difference, (Tmax – Ts), at
short and long times, assuming the initial condition is that the part is Ts at
t = 0. A quick way to get that estimate is by a scaling analysis. (a) Sketch
and label the system. Write the appropriate form of energy equation for this
problem. Clearly state your assumptions and identify the sources and sinks
of thermal energy. (b) Pick reference values for the temperature difference
( Tref ) and lengths (xref, yref ). (Use t, the time elapsed since heating began,
for time). Which reference values are unknown? (c) Scale the 2D energy
equation (it should have four terms at this point) and identify the physi-
cal phenomena represented by each of the terms. Combine any terms that
are always the same order of magnitude. (d) Which of the three remaining
physical mechanisms is important throughout the entire process? Divide all
terms by that term representing that phenomenon, so it is (1). (e) Which
term can be neglected when it reaches steady state at very large times
(t ∞)? Using the remaining two terms, find the behavior of (Tmax – Ts) at large
times, i.e., find its dependence on t. (f) How large must time be to neglect the
term in part (e)? (g) Which term can be neglected at very small times? How
does (Tmax – Ts) behave at small times? How small must time be to neglect the
term? (h) Sketch (Tmax – Ts) vs. t showing behavior at small and large times.
3.10 This problem examines the question of whether, when you step on hot sand
on a beach, does the sand cool down faster or slower than your foot heats
up? (Think of your own experience with this problem.) The foot is initially
at Tfi and the sand is at Tsi. The foot makes contact with the sand at t = 0,
at which time the interface between the two (almost) immediately takes
on the constant temperature To. As heat is transformed from sand to foot,
we can define thermal penetration depths, s and f. (a) Write the transient
energy equation and boundary conditions at x ∞ and x −∞. (b) Scale
the one-dimensional transient energy equations separately in both the sand
and foot. Find an estimate for the ratio of the penetration depths into the
sand and the foot ( s/ f ). (c) Write the expression for heat flux continuity at
the sand–foot interface. Scale this relationship using these reference val-
ues: Ts = Tsi – To and Tf = To – Tfi and the result from part (b). Find an
estimate for the interface temperature, To, in terms of properties and the
178 An Introduction to Transport Phenomena in Materials Engineering

initial temperatures. (d) Sketch the temperature distribution at three times,

t > 0, on the Figure 3.28.

(k c)s = 3.2 × 105 WJ/m4K2 (k c)f = 2.5 × 106 WJ/m4K2

s = 2.3 × 10 m /s f = 1.4 × 10
−7 2 −7 m 2/s

3.11 You are designing a sintering furnace including a small glass viewing port
in the wall. You want to know how long (tcrit) after you turn on the furnace
the heat propagates through the glass window and begins heating the out-
side of the glass. A one-dimensional semi-infinite analysis seems appropri-
ate, and you assume that there is a constant heat flux at x = 0. Boundary
conditions for the transient energy equation with no heat generation are

T T
T ( x,t 0) Ti qw k T (x , t ) Ti 0.
x x 0 x x

(a) Sketch and label the system. (b) Normalize the energy equation using the
following nondimensional values:

T Ti x
.
(qw / k )

Keep in mind = (t), so the temperature reference is also not constant.

Because of this time dependence, when normalized, the transient term looks like:
[ ]
.
t t t

(c) Using a second-order polynomial C1 2 C2 C3 , apply the normal-

ized boundary conditions to obtain an expression for = f( ). (d) Write the
integral form of the normalized energy equation over = 0 to 1. With integra-
tion, show that the energy equation reduces to:

3 d
.
dt
(e) Solve the differential equation for (t), and write an expression for tcrit.
3.12 An infinite area, initially at Ti is suddenly heated at t > 0 from a hole of
radius R at TR > Ti (Figure 3.29). There is no heat generation, and heat flow
is only radial. (a) Write the governing energy equation and boundary and
initial conditions for this problem. (b) Nondimensionalize T and r with

T r,t Ti r R
and ,
TR Ti

where (t) is the thermal penetration depth. (c) Find the constants in an
assumed temperature profile:
2
C1 C2 C3 .
Transient Conduction Heat Transfer 179

FIGURE 3.29 Schematic of geometry for radial conduction in Problem 3.12.

(d) By integrating the energy equation over the penetration depth, show that
2
t 1 ,
12 R

and plot R vs. 12 t R 2 . How and why does this (t) behave differently
from the slab case?

3.13 The surface temperature of an effectively semi-infinite slab of steel, initially

at uniform temperature (Ti = 25 oC), is raised instantaneously to 50 oC. Find:
(a) the distance from the surface at which the local temperature is 30 oC
after 5 minutes; (b) heat flux into the slab at 5 minutes; and (c) total thermal
energy transferred to a unit area of the slab surface from t = 0 to 5 minutes.
3.14 Liquid steel at Ts = 1650 oC is rapidly poured to a depth of d = 2.5 m in a
ladle of diameter D = 3 m, which is preheated to Ti = 700 oC. The ladle lin-
ing is L = 15 cm thick. Assume that the radius of curvature in the ladle is
large enough that it can be treated as a slab and that the Bi = hL//klining >> 1.
(a) Calculate how much heat is transferred from the steel to the ladle by
convection during the first minute. (b) Is the semi-infinite approximation
still valid at the end of that minute?
Ladle lining: k = 1 W/mK = 3 × 10 –7 m2/s
Liquid steel: = 7050 kg/m 3 c = 750 J/kg K
3.15 To transform austenite near a cooled surface to martensite, this layer must
be quenched from the soaking temperature of Ti = 900 oC to 400 oC in less
than 2 s. (a) If we assume that the surface temperature is suddenly changed
from 900 oC to Ts = 100 oC, what is the thickness of the martensite layer
( m)? (b) Find the minimum heat transfer coefficient required so that the
assumption in part (a) is valid, using the layer thickness as a length scale.
The steel thermal diffusivity is = 1.2 × 10 –5 m2/s.
180 An Introduction to Transport Phenomena in Materials Engineering

3.16 The inside surface (x = 0) of an alumina furnace lining (thickness = L = 10 cm)

is suddenly raised from Ti = 30 oC to To = 900 oC ( alumina = 2 × 10 −5 m2/s).
(a) Sketch the geometry of interest and identify all relevant parameters. (b)
What is the value of tcrit? (c) Plot T(x = L, t) from t = 0 until T(x = L, t) = To +
0.95(Ts – To), that is, L = 0.95. At what (dimensional) time does that occur?
(d) For the same time period and on the same plot as in part (c), draw the
curves for T(x = 0, t), T(x = L/3, t), and T(x = 2L/3, t). (e) How would the
times found in parts (b) and (c) change if L was doubled?
3.17 CMSX-4, a nickel superalloy used to cast single crystal turbine blades, is
poured at temperature TM into a ceramic mold of thickness L and initially at
temperature Ti. Treat the mold as a slab. (a) How long before the outer sur-
face of the mold (x = L) begins to change temperature? Assume the inside
mold temperature is the melting temperature of the metal (TM ) throughout
the process. (b) Once the outer surface begins to heat up, strictly speaking
the finite body solutions found in this chapter do not apply because they
assume an insulated boundary. Actually, heat is lost to the cold furnace by
radiation (this alloy is cast in a vacuum). However, we might be able to use
it for a time when the heat lost by radiation is small compared to the heat
being added to the mold. Plot the dimensional radiation flux from the outer
wall to the environment (x = L) and the dimensional conduction flux into
the inner wall (x = 0) as functions of time on one plot and their ratio (R).
When does the radiative heat loss become a significant fraction of the heat
input into the ceramic (R > 0.1)? (The point here is to understand when the
insulated boundary approximation is no longer reasonable.)

kmold = 0.38 W/mK cmold = 750 J/kgK mold = 1500 kg/m3

L = 0.02 m TM = 1700 K T∞ = 350 K = 0.5

3.18 A PETE beverage bottle is made by extrusion blow molding through a ring-
shaped die. In Figure 3.30, a parison of PETE is extruded as a hollow tube
into the center of the open mold. The mold is then closed and air is blown
into the parison, coating the inside of the close mold with PETE. The bottle
wall has a final thickness of 1 mm. The initial temperature of the polymer
is Ti = 200 oC. The mold surface (x = 0) is maintained at To = 35 oC, and
there is almost no resistance between the hot polymer and the cold mold (Bi
>> 1). (a) Sketch and label the temperatures and distances during the devel-
opment of T(x,t), beginning with the polymer–mold contact through the
time the inside of the layer reaches 70 oC. Include at least one curve each
for the semi-infinite and finite body regimes, and one for tcrit. (b) At what
time (tcrit) does the cooling effect of the mold reach the inside of the PETE
layer? (c) At what time (tfinal) is all of the polymer below 70 oC? (At this
temperature, the polymer is strong enough to be ejected from the mold.)

= 950 kg/m3 k = 0.2 W/mK c = 1000 J/kgK

3.19 A thick (semi-infinite) oak wall initially at the uniform temperature 25 oC

is suddenly exposed to gaseous reaction products at 800 oC. Assume that
Transient Conduction Heat Transfer 181

FIGURE 3.30 Schematic of the blow-molding process in Problem 3.18.

Bi ~ (1) and that the heat transfer coefficient is 20 W/m2K. Find: (a) the
time to raise the surface temperature to its ignition temperature of 400 oC;
(b) the thermal penetration depth at the time in part (a).
3.20 A thick (semi-infinite) slab of alumina at Ti = 600 oC is exposed to an envi-
ronment at 200 oC, and the heat transfer coefficient is 100 W/m2K. Find,
after 10 minutes: (a) surface temperature of the slab; (b) the temperature
1 cm below the surface.
3.21 Rework Example 3.3 with a heat transfer coefficient h = 5,000 W/mK. (a)
How long to get s = 0.95? 0.99? (b) Plot the time the droplet takes to
change the surface temperature to s = 0.95 and 0.99 as a function of heat
transfer coefficient (100 W/m2K < h < 10,000 W/m2K). (c) Is the approxi-
mation of assuming that the surface temperature changes to s = 1 “instan-
taneously” reasonable?
3.22 Rework Example 3.3, assuming there is a drop-substrate interface resis-
tance (1–10 Km2/W). Assume Bi ~ (1). (a) Plot the surface temperature as
a function of time for Ri = 1 and 10 Km2/W. (b) Is the approximation (used
in Example 3.3) of assuming that the surface temperature changes to s = 1
“instantaneously” reasonable?
3.23 Is it feasible to cast iron with no superheat (at its melting temperature,
TM,Fe = 1535 oC) in a thick-walled aluminum mold?

Liquid Fe: = 7200 kg/m3 c = 790 J/kg K Lf = 2.6 × 105 J/kg

k = 63 W/mK
Solid Al: = 2700 kg/m3 c = 900 J/kg K TM,Al = 660 oC
k = 236 W/mK

3.24 Liquid steel at its freezing temperature of 1480 oC is cast as a slab of thick-
ness 0.1 m. Find the times required for completion of the freezing process
with (a) a thick sand mold at 25 oC and (b) a water-cooled copper mold at
25 oC. (For part (b), use the solution found in Problem 3.25.) The thermo-
physical properties are:

Liquid steel: = 7050 kg/m3 c = 750 J/kg K Lf = 2.6 × 105 J/kg

s = 1.1 × 10 m /s
−5 2

Sand mold: k = 0.3 W/m K m = 1.1 × 10 −5 m2/s

182 An Introduction to Transport Phenomena in Materials Engineering

3.25 At t = 0, a pure liquid metal at its melting temperature (TM ) is poured

into a mold that is maintained at some lower temperature, To, as shown in
Figure 3.31. Here, the mold resistance is negligible, and conduction out of the
metal is controlled by the thermal resistance in the solidified metal. The mov-
ing solid-liquid interface is x = M(t). (a) Assuming a linear function of tem-
perature in the solid, find the coefficients (C1, C2) in the temperature profile:

T To
C1 C2 x M (t ).
Tm To
(b) Sketch and label the normalized temperature profile in the solid and liq-
uid metal. (c) Find the unknown function for the interface position, M(t),
using the Stefan condition. (d) Sketch M vs. t and the latent heat release rate
vs. t. (d) Use your result from part (b) to estimate the solidification time for
the aluminum and copper slabs.

Al: TM,Al = 660 oC Lf,Al = 400 kJ/kg Al = 9.2 × 10 –5 m2/s

cAl = 1050 J/kgK
Cu: TM,Cu = 1083 oC Lf,Cu = 205 kJ/kg Cu = 1.4 × 10 –4 m2/s
cCu = 440 J/kgK

3.26 Consider solidification in a flat ceramic shell mold with a mold thickness
D = 10 mm. There is a heat loss to the outside surface to the surroundings at
T∞ = 30 oC with a constant heat transfer coefficient, h = 100 W/m2K. Except
at very early times, the temperature field in the mold is at steady state.
Assume that the total thermal resistance of the outside convection and the
mold wall conduction can be written as
1 D 1
RTOT ,
UA km A hA

FIGURE 3.31 Schematic of geometry of solidifying pure metal with dominant metal resis-
tance from Problem 3.25.
Transient Conduction Heat Transfer 183

184 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 3.32 Schematic of the solidification fronts in a square part in Problem 3.27.

TABLE 3.2
Property Values for Problem 3.28
k (W/mK) (kg/m3) c (J/kgK) Lf (kJ/kg)
Al 210 2500 1200 400
Cu 170 8000 490 210
Fe 40 7700 790 280
Mullite 3.3 2800 200 -
Sand 1.0 1600 1200 -
Transient Conduction Heat Transfer 185

FIGURE 3.33 Schematic of casting of pure metal in thermally resistive mold from Problem
3.28.

Comment on the applicability of the assumption at t = t f. Is the approxi-

mation better or worse at earlier times? (e) We want to cast a Cu cylinder
3 cm in diameter in a 20 cm diameter mullite mold. Assuming no liquid
superheat and an initial mold temperature of 300 oC, what is tf ? What two
practical changes can be made to speed the solidification?

NOTES
1. It is very important to note that while we derive the solutions and work the examples in this
chapter, they are described as either “heating” or “cooling,” but the analyses are indepen-
dent of the direction of heat flow and may be used in either case.
2. Quite frequently, Bi makes an appearance in the discussion of conduction only when exam-
ining the validity of this lumped capacitance model. The present and the last chapter show
that this parameter is more generally useful.
3. Contributed by Bailey McConnell and Jessica Scharrer, from their industry-sponsored
undergraduate capstone project at Purdue University.
4. A semi-infinite body has a beginning (x = 0) but no end as x increases.
5. Contributed by Brandon Wells and Andrea Martin Tovar, from their industry-sponsored
undergraduate capstone project at Purdue University.

REFERENCES
1. Hatch, J. E. (ed.), Aluminum: Properties and Physical Metallurgy, American Society for
Metals, 1984.
2. Vander Voort, G. F. (ed.), Atlas of Time-Temperature Diagrams for Nonferrous Alloys,
ASM International, 1991.
3. Fezi, K., and M. J. M. Krane, “Influence of a wiper on transport phenomena in direct
chill casting of aluminium alloy 7050,” International Journal of Cast Metal Research, v.
30, pp. 191–200, 2017.
186 An Introduction to Transport Phenomena in Materials Engineering
 4 Mass Diffusion in the
Solid State

4.1 INTRODUCTION
The physical and mechanical properties of all solid materials depend on their micro-
structure, and development of desirable engineering properties requires the ability
to manipulate, for example, crystal structure, phase morphology, and grain size.
Although the relative stabilities of phases are determined by thermodynamic consid-
erations, the rates at which phase changes can occur are determined by the rates at
which atoms, molecules, or ions can be transported in the solid state. An understand-
ing of the kinetics of phase change in the solid state thus requires an appreciation of
solid state diffusion.
In this chapter, we will consider steady and transient diffusion behavior. The con-
stitutive relation for binary diffusion, Fick’s first law (Section 1.3.3), is used to derive
a solute conservation equation. Several transient, one-dimensional solutions are
developed and applied to practical problems. These solutions at first assume uniform
diffusion coefficients, but the later sections of the chapter discuss the limitations of
that assumption. The chapter concludes with the measurement of composition-de-
pendent diffusion coefficients and their dependence on temperature.

4.2 STEADY STATE MASS DIFFUSION

In Section 1.3.3, we introduced Fick’s first law, a constitutive equation for binary
mass diffusion in one dimension, relating the concentration gradient and flux of sol-
ute A, thus defining the mass diffusivity.

CA
jA DA (1.46)
x

Here, DA is the diffusivity of solute A, and jA is the net flux of A in the binary
solution. The concentration of component A, CA, can be written with a variety of
units, usually in terms of a quantity of A (e.g., kg, mole) per unit mass, moles, or
volume of the solution. The flux, jA, is the rate at which that quantity of A passes
through an area; its units are [(quantity of A)/m2s] and the units of mass diffusivity
are then [m2/s].

Example 4.1
One-Dimensional Radial Carbon Diffusion through Wall of an Iron Tube
The inner and outer surfaces of an iron tube of length L are in contact with carbu-
rizing gases (mixtures of CO and CO2) with different carburizing potentials. Local
DOI: 10.1201/9781003104278-4 187
188 An Introduction to Transport Phenomena in Materials Engineering

thermodynamic equilibrium is established on each side of the wall between gas and
surface, and thus the carbon contents of the two surfaces are fixed by the activity of
carbon in the gases. Because the concentrations of carbon at the two surfaces are
different and are maintained at constant values by the gases, there is a steady flux of
carbon by diffusion through the wall of the cylinder given by

dCC
jC DC , (4.1)
dr

where the units of CC are [kg C/m3] and of the flux, jC, are [kg C/m2s].
Because the diffusion is steady state, the mass flow rate of carbon, JC = jC A,
is constant throughout the annulus. Writing Eq. (4.1) in terms of carbon flow rate
instead of flux,

dCC
JC 2 rL DC , (4.2)
dr

we can write

dr 2 LDC 2 LDC
dCC and ln r2 r1 CC1 CC 2 .
r JC JC

Rearranging this result gives us:

2 LDC
JC CC1 CC 2 ,
ln r2 r1

which confirms that JC is not a function of radius in the tube. A uniform JC in Eq.
(4.2) should not be confused with a uniform concentration gradient. If JC is uniform
and the area is proportional to radius, A = 2πrL, then the concentration gradient
must go like 1/r, higher closer to r = 0 and shallowing as the area increases with r.

Another cause of a nonuniform concentration gradient in steady-state mass diffu-

sion occurs independently of geometry. If the mass diffusivity is a strong function
of composition, D(C), for instance, if D is higher at lower C, then we might have
a composition profile, as in Figure 4.1, where the gradient changes while the mass
flow is uniform. This effect of a nonuniform D on the C(x) profile when j is uniform
is shown in Figure 4.1.

4.3 FICK’S SECOND LAW OF DIFFUSION: TRANSIENT

DIFFUSION
In many processes, the change of microstructure is controlled by the transient dif-
fusion of a solute. A model for this diffusion can be developed by considering a
mass balance on the one-dimensional control volume of unit cross-sectional area and
length dx shown in Figure 4.2. (Although one-dimensional, this analysis is done in
Mass Diffusion in the Solid State 189

FIGURE 4.1 Sketch of composition profiles when the mass flux is uniform, one with a con-
stant diffusion coefficient (D) and another that depends on composition (C).

FIGURE 4.2 Mass balance in a control volume of unit cross-sectional area and length x.

the same spirit as Figure 2.7.) For diffusion of a solute (i) in the x-direction, the mass
balance is

Ci
d dA jx jx dx .
t
solute solute flux (4.3)
storage in - out

Here the differential volume is d = dx dy dz and the cross-sectional area normal

to the direction of the mass flux is dA = dy dz. Representing jx+ x by a Taylor series
centered at x,
jx 1 2 jx 2
jx dx jx dx dx ,
x 2 x2
190 An Introduction to Transport Phenomena in Materials Engineering

truncating the series at two terms, and substituting into Eq. (4.3) gives

Ci jx
dx jx jx dx .
t x

Ci jx
t x
solute diffusion (4.4)
storage of solute

For a steady case (∂C/∂t = 0), Eq. (4.4) shows that the mass flux is uniform in x.
(This result is different from Example 4.1, in which the area changes in the direction
of mass flow.) Substitution of Fick’s first law, Eq. (4.1), into Eq. (4.4) gives

Ci Ci
Di . (4.5)
t x x

In Eqs. (4.4) and (4.5), we see different versions of Fick’s second law of diffusion, in
which the rate of local accumulation or depletion of species i is equal to the rate of
change of solute flux due to diffusion at that location.
As noted in Section 1.3.3, the diffusion coefficient is certainly a function of many
microstructural features. We begin with some simple solutions by approximating D
as constant and uniform, perhaps as an average of the diffusion coefficient over the
composition range in the problem. With this simplification, we can readily obtain
estimates for diffusion behavior in a variety of processes, including carburization,
decarburizing, and nitriding in metal processing, and dopant drive-in in integrated
circuit manufacturing. These and other processes can be modeled in semi-infinite or
finite domains.
Here we will show solutions to one-dimensional problems that begin with a uni-
form initial composition, Ci, and have a surface changed suddenly to a different fixed
value, Co. For a semi-infinite domain, the initial and boundary conditions are

C x 0, t 0 Co C x ,t Ci and C x,t 0 Ci (4.6)

and are

C
C x 0, t 0 Co 0 and C x, t 0 Ci (4.7)
x x L

for a finite body of thickness L.

With these boundary and initial conditions and the solute conservation equation,
Eq. (4.5), scaling, integral, and exact solutions can be found. A plot of composition
fields in both the semi-infinite and finite regimes is shown in Figure 4.3.
As we can readily see from the governing equation, the boundary and initial con-
ditions, and Figure 4.3, the mathematics of binary diffusion with a uniform mass
diffusivity, D, are the same as those of transient conduction in Section 3.4 and so the
Mass Diffusion in the Solid State 191

FIGURE 4.3 Composition fields in the semi-infinite (t < tcrit) and finite body (t > tcrit) regimes.

solutions have the same form. For the entire process before and after tcrit, the exact
solution for nondimensional composition ( ) can be written as

C Ci
C o Ci
n 2
2 2 1 1 2 1 2n 1
1 exp n cos 1 , (4.8)
n 0 2n 1 2 12 2

and the composition history at x = L ( = 1) as

n 2
C x L Ci 2 1 1 2 1
L 1 exp n , (4.9)
C o Ci 1 2 12
n 0
n
2

where = x/L, = (t – tcrit)D/L2, and tcrit = L2/12D. The nondimensional solute flux
(F) at the surface (x = 0) is

jo L L C
F
D C o Ci C o Ci x x 0 0
2
1 2 1
2 exp . (4.10)
n 0 2 12
192 An Introduction to Transport Phenomena in Materials Engineering

The approximate solutions from the integral method and scaling analysis are broken
into the two regimes. The first case is the semi-infinite case (t < tcrit), in which the
change in the composition at the surface (x = 0) has yet reached the far wall. The
integral solution for this regime is

C Ci 2
2 1, (4.11)
C o Ci

where

x t and t 12 D t . (4.12)

The scaling analysis estimates a similar penetration depth as

t D t. (4.13)

The solute flux at the surface estimated by the integral method is

jo L L C L 2L L
F (4.14)
D C o Ci C o Ci x x 0 0 3Dt

and from scaling analysis

jo L L C L C o Ci L
F ~ . (4.15)
D C o Ci C o Ci x x 0 C o Ci Dt

When tcrit L , the composition at the impermeable far wall (x = L) begins to feel
the effect of the change in the boundary condition at x = 0 and begins to rise.
For t > tcrit, the composition field is in the finite body regime with the boundary
conditions from Eq. (4.7). There the integral method gives the composition field as

C Ci 2
exp 3 2 1. (4.16)
C o Ci

The far wall composition from the integral and scaling analyses are, respectively,

L 1 exp 3 and L 1 . (4.17)

During this regime, the solute flux at the surface is

jo L L C
F 2 1 L 2 exp 3 (4.18)
D C o Ci C o Ci x x 0 0

from the integral method and, from the scaling analysis,

F 1 1 . (4.19)
Mass Diffusion in the Solid State 193

Figure 4.4 contains plots of these nondimensional solutions. In all cases, the exact
and integral solutions agree very well (within a few percent), while the scaling results
have similar behavior and are within a factor of 2–3 of the other solutions.

Example 4.2
Carburization of Steel
During the carburization process, carbon is diffused from a surface exposed to a
carbon-rich atmosphere into the subsurface region of a steel part. The carbon pro-
file gradually decreases from its maximum at the surface to the original level at the
edge of a penetration depth. The increased carbon content lowers the martensite start

FIGURE 4.4 Plots in the finite body regime (t > tcrit) of nondimensional (a) composition
profiles, and (b) far wall composition histories.
194 An Introduction to Transport Phenomena in Materials Engineering

temperature so that, when quenched, the phase transformation from austenite begins
first just below the surface. Because of the higher carbon content, the surface trans-
forms later and at a lower temperature, and a residual compressive stress is formed
there, making the surface tougher and more wear resistant. The increased carbon
content also makes the martensite stronger.
A good introduction to all aspects of carburization is found in [1], but here we
focus on the development of the carbon profile by diffusion. Consider the unidirec-
tional carburization of a plain carbon steel at 950 °C, in which the surface of the steel
is initially low (CCi = 0.2 wt%) before being brought into contact with a carburiz-
ing gas at time t = 0. This atmosphere instantaneously supplies carbon to raise its
concentration in the iron at the surface to CCo = 1 wt%. The diffusion coefficient is
approximated as uniform at the value of DC = 1.09 × 10–11 m2/s.

Find: (a) The value of CC CCi CCo CCi at x = 0.5 mm after 1 h of

carburization. (b) The time at which = 0.5 at x = 0.8 mm. (c) Penetration
depth, (t), and surface mass flux, jo(t), histories over 24 hours.
Solution: The concentration profiles obtained from the integral solution, Eqs.
(4.11), at t = 1, 6, 12, and 24 hours are shown in Figure 4.5. We see from the
penetration depths in that plot that (24 hrs) ≈ 3.1 mm, which is much less
than L, so the composition field is always in the semi-infinite regime.

(a) With x = 5 × 10 –4 m and t = 3600 s,

x 5 10 4 m
0.729
2 3Dt 2 3 1.09 10 11
m 2 s 3600 s

FIGURE 4.5 Carbon composition profiles diffusing into a semi-infinite iron slab in Example
4.2, based on the integral solution, Eq. (4.11).
Mass Diffusion in the Solid State 195

FIGURE 4.6 Carburization penetration depth and surface mass flux as a function of time
from integral solutions in Example 4.2.

and

CC CCi 2 2
2 1 0.729 2 0.729 1 0.074,
CCo CCi

which is point A in Figure 4.5.

(b) From Eq. (4.11),

2 4 4 1
1 1 0.5 0.293
2

for CC CCi CCo CCi = 0.5. The time at which the composition at
x = 0.8 mm is one half the value at the boundary is
2 2
x 1 0.0008 m 1
t 57,0000 s 15.8 hr.
12D 0.293 12 1.09 10 11
m 2 /s

(c) The penetration depth and surface mass flux are found over 24 hours
(86,400 s) in Figure 4.6, using Eqs. (4.12) and (4.14). This plot shows very
diminished incremental returns after four or five hours; the duration of the
process required to make significant changes increases dramatically after
that time.

4.4 INFINITE DIFFUSION COUPLE

When two metal slabs with different compositions are placed in intimate contact with
each other, with no resistance to solute flux, they form an infinite diffusion couple.
(In practical terms, diffusion couple is apparently infinite in that, within the period of
time when diffusion occurs, no changes in concentration are seen near the two actual
196 An Introduction to Transport Phenomena in Materials Engineering

ends of the couple. This configuration can be used to measure diffusion coefficients
and also to model processes such as brazing and diffusion bonding.
In this case, the initial conditions are

C 1Ai for x 0
C t 0, (4.20)
C Ai2 for x 0

where C 1Ai and C Ai2 are the initial values of solute A concentration in regions 1 and 2.
The boundary conditions are

C 1Ai for x (t ) C 0 for x (t )

C and , (4.21)
C Ai2 for x (t ) x 0 for x (t )

where (t) is the penetration depth. Nondimensionalizing the distance from the inter-
face (x = 0) as = x/ , these boundary conditions can be rewritten as

C C 1Ai 0 for 1 0 for 1

and . (4.22)
C Ai2 C 1Ai 1 for 1 0 for 1

An integral solution can be found in the range in which the composition has changed
(x = − to x = or = −1 to = 1), assuming that D is uniform over that region. At the
edges of both species penetration depths (x = ± or = ±1), the composition gradients
are zero, so we know that there is a point of inflection in the profile and a third-order pol-
ynomial may be a good approximation. With the previous boundary conditions, we have
1 3
2 3 . (4.23)
4
Integrating the solute conservation equation produces a differential equation for the
penetration depth, the solution of which is

12Dt. (4.24)

The symmetry of the problem gives = 0 = ½ at the interface ( = 0), or

C A x 0 C Ao (C 1Ai C A2i ) / 2 . The flux of solute across the plane of the inter-
face at x = 0 is obtained from Fick’s first law:

CA C Ai2 C 1Ai d C Ai2 C 1Ai 3

jo DA DA DA
x x 0 d 0 12DA t 4
3 2 DA (4.25)
C Ai C 1Ai .
8 3t
The exact solution for this case is

1 x C Ai2 C 1Ai DA
1 erf and jo . (4.26)
2 2 DA t 4 t
Mass Diffusion in the Solid State 197

FIGURE 4.7 An infinite diffusion couple with uniform diffusion coefficient. (a) Normalized
composition profiles (exact and integral solutions). (b) Dimensional composition profiles for
carbon diffusing in an infinite couple of steel at 1000 oC.

The composition profiles in Eqs. (4.23) and (4.26) are plotted in Figure 4.7.

4.5 DIFFUSION INVOLVING SOLID-SOLID PHASE CHANGE

Often we effect a change in microstructure in a solid solution by heating the material
to a temperature at which there is appreciable diffusion. If the microstructure is not
then at equilibrium (a likely situation), phases may grow or dissolve, which requires
mass to diffuse toward, through, and away from phase boundaries. Diffusion-con-
trolled phase change does not occur instantaneously, and the analysis presented here
198 An Introduction to Transport Phenomena in Materials Engineering

is a first step in understanding how the movement of a two-phase interface might

develop. Before we look at a simple case, we need to understand the solute balance
at a moving interface between two different phases.
We can start with a phase diagram with a two-phase ( - ) region and draw an
isothermal tie line between the two single-phase regions on its boundaries, Figure 4.8
(a). Note that no single phase has a composition anywhere on the tie line, except at the
end points (C A * and C A * ); effectively, a two-phase region is a “hole in the diagram”
where no single phase exists. The relative amounts of the phases are determined
then by the mixture composition of the two-phase region. An interface between - ,
shown in Figure 4.8 (b), will have a discontinuity in composition where it jumps
between the two compositions C A * , C A * at the ends of the tie line. Throughout
this treatment we will assume the interface is at equilibrium, so C A xi C A * and
C A xi C A * ; here we see that the two solid phases are also at equilibrium internally
with no composition gradients.

FIGURE 4.8 (a) Phase diagram around a two-phase region. The tie line across the -
region shows equilibrium compositions for both phases. (b) The two phases at equilibrium
internally and with each other for a mixture composition in the - region.
Mass Diffusion in the Solid State 199

FIGURE 4.9 Schematic of interface between two phases showing movement over dx during
time increment dt. The shaded area represents the solute that must be added to the interface
to maintain chemical equilibrium at the interface as it moves.

In Figure 4.9, we have relaxed the internal equilibrium assumption and now have
composition gradients in each phase. We know from Fick’s first law that these gradi-
ents indicate that there is a flux of solute in each phase. Let us look at what happens
in Figure 4.9 after a small increment in time, dt. During dt, the interface is observed
to move to the right by dx and to sweep through a volume Adx. This differential
volume at time t is the phase equilibrium composition (C A * ). At t + dt, after the
interface moves, there has been a phase transformation ( ) and the swept-out
region has changed its composition to the phase equilibrium composition (C A * ).
This change in composition represents a storage of solute A at the interface, indicat-
ing that the net solute transfer into and out of the - boundary must be non-zero for
it to move. How is this “stored” mass provided to the interface?
The net solute flow in and out of the interface is Jnet = Jin – Jout, where the rate
equation for mass diffusion (Fick’s first law) gives

CA CA
J in JA DA A and J out JA DA A .
x x xi
x x xi

The net solute flow at the interface is not zero if there is phase change (the interface
moves), and so it will be equal to the change in composition needed to supply that
phase change. The mass balance over that small time, dt, is

CA CA
J in J out d t DA A DA A dt C A * C A * Adx (4.27)
x xi
x xi
200 An Introduction to Transport Phenomena in Materials Engineering

or

CA CA dx
DA DA CA * CA * ,
x x x x dt
i i (4.28)
flux of A flux of A compposition change
into interface from interface
from phase innto phase interface velocity

where dx/dt is the interface velocity, Vint. So, for the interface to move dx to the
right in time dt, the flux into the interface, j , must exceed the flux out, j , by the
amount required for phase change, C A * C A * dx dt, swept out on the C-x plot,
Figure 4.9. Thus the - interface velocity (proportional to the rate of phase change)
is controlled by the net diffusive flux at that interface. Eq. (4.28) is used to relate the
interface motion to the solute gradients in the two phases. At chemical equilibrium,
as in Figure 4.8 (b), there are no concentration gradients at the interface, and so, by
Eq. (4.28), no phase boundary movement.
We begin with a solution of an isothermal diffusion-controlled phase change in a
semi-infinite domain initially at C Ai2 and the crystal structure in what we will call phase
2. At t = 0, the surface condition is changed to a fixed composition, C A x 0 C 1Ao.
Under any circumstances such a change at the boundary will alter the composition
fields in the domain, but here we also will examine the effect of a phase change
between C 1Ao and C Ai2 , as shown in Figure 4.10. The change in the boundary condition
to CA(x=0) = C 1Ao causes the nucleation and growth of phase 1 there. This phase will
grow at the expense of phase 2, fed by the C Ai2 C 1Ao composition difference and
with a composition discontinuity at the phase boundary.
To this last point, we examine the phase diagram in Figure 4.10 at temperature
T, and we see that the composition range along the tie line between C A2* andC 1* A is a
two-phase region. Therefore, no point in Figure 4.11, a plot of composition profiles
in both phases (t > 0), should have a value in that range. The minimum composi-
tion in the single-phase region for phase 1 is C 1* A , and the maximum in phase 2 is
C A2* . Assuming thermodynamic equilibrium holds at the interface during the phase
change, the compositions at the interface are C 1A x L C 1A* and C A2 x L C A2*.
These values are the end points of the tie line across the 1–2 region of the phase
diagram.

FIGURE 4.10 Binary phase diagram for isothermal, diffusion-controlled phase change in
Figure 4.11.
Mass Diffusion in the Solid State 201

FIGURE 4.11 Schematic of diffusion-controlled growth of phase 1 into phase 2, initially

2
at C Ai , for the case with a surface maintained at C 1Ao for t > 0. The compositions here refer to
the phase diagram in Figure 4.10.

To find the transient composition fields, C 1A ( x,t) and C A2 ( x,t), and the position and
velocity of the interface, L(t) and Vint(t), we begin with Fick’s second law applied to
phase 2:
2
C A2 C A2
DA2 , (4.29)
x2 t

with the initial and boundary conditions

C A2 x, t 0 C A2i C A2 x L, t C A2* C A2 x ,t C Ai2 . (4.30)

Here we begin to solve for this system’s behavior using the integral method. The
first two steps are as we have done before, first assuming a composition profile
shape and then finding a differential equation for an unknown function using that
profile in the integrated conservation equation. However, in this case we also will
need another equation, as we will have two unknown transient functions ( and L),
so we introduce a third step in which we apply the two-phase interface condition in
Eq. (4.28).
We assume a nondimensional quadratic profile for solute in phase 2,

2 C C A2* 2
A B1 B2 B3 . (4.31)
C Ai2 C A2*
202 An Introduction to Transport Phenomena in Materials Engineering

Here the spatial coordinate, , is defined in terms of y = x – L(t), which fixes the
frame of reference to the interface, thus

y x L t
. (4.32)
(t ) t

With this transformation, = 0 (y = 0 or x = L) is always at the phase boundary, and

= 1 (y = or x = L+ ) is at the far edge of the penetration depth into phase 2. Using
these variables, the two boundary conditions in Eq. (4.30) can be rewritten as
2 2
A 0 0 and A 1 1. (4.33)

Because there is no solute flux outside the penetration depth,

C A2 2
A
0. (4.34)
y y 1

Applying those three conditions to Eq. (4.31), we get the approximate composition
profile and its derivative:
2
2 2 d A
A 2 and 2 1 . (4.35)
d

Substituting the previous nondimensional variables into the solute conservation Eq.
(4.29), we get

DA2 2 2
A
2
A
2 2
(4.36)
t

Integrating Eq. (4.36) over the penetration depth (y = 0 to y = or = 0 to =1),

we have
1 1
DA2 2 2
A
2
A
2 2
d d . (4.37)
0 0
t

The left-hand side of the integrated Eq. (4.37), representing the diffusion in the pen-
etration region in phase 2, is
1
DA2 2 2
A DA2 2
A
2
A DA2 2 DA2
LHS 2 2
d 2 2
0 2 2
.
0 1 0

After integration, we have the difference of the first derivatives of 2A at the bounda-
ries of the penetration region. The derivatives are proportional to the solute fluxes in
and out of the region, so the difference is the net solute flux into the region, which is
stored at x = L during the phase change.
Mass Diffusion in the Solid State 203

The right-hand side of the integrated Eq. (4.37) is the storage (or depletion) of
solute in phase 2. It cannot be integrated directly, so d 2A dt must be changed to a
function of what we can integrate ( ) and what we are looking for (L, ). To do so,
we must remember that 2A is a function of time not only through (t), but L(t) also,

2
A f f L t , t ,

and both L and are hidden in the spatial variable . Writing the total derivative
of 2A ,

2 2
2 A A
d A dL d ,
L
2
an expression can be found for the time derivative of A :

2 2 2 2
A A L A A L
. (4.38)
t L t t L t t

This expression shows the separate effects of the dependencies of L and on time on
the rate of change of the composition, 2A . While the time derivatives of L and are
unknown as yet, the other factors in Eq. (4.38) can be found from the definition of
and the assumed 2A profile:
2
A 1 x
2 1 2
.
L

Using these functions, we get

2
A 1 dL d
2 1 ,
t dt t

and so the integrated storage term on the right-hand side of Eq. (4.37) becomes
2
1
A
1 1 dL 2 1d
RHS d 2 1 2 d
0 t 0 dt dt

1 dL 1 d
.
dt 3 dt

Equating the left- and right-hand sides of Eq. (4.37),

2 DA 1 d dL
.
3 dt dt
net flux growth rate of (4.39)
at interface of in 2 phase change
204 An Introduction to Transport Phenomena in Materials Engineering

So we see that the net flux in and out of the phase boundary feeds both the composition
changes inside the penetration depth in phase 2 as well as the solute required to effect a
phase change from 2 to 1. To simplify the solution to this problem, we will assume that
1 d dL
.
3 dt dt
Thus, we may neglect the last term in Eq. (4.39). (We may evaluate applicability of
this convenient assumption after we find L(t).) This assumption allows us to solve
Eq. (4.39) without the last term, with the initial condition (t=0) = 0.

(t ) 12 D2 t (4.40)

Turning to the growing phase 1, we write the nondimensional composition profile

there as

1 C 1A C 1Ao x
A B1 B2 and . (4.41)
C 1A* C 1Ao L (t )

Although we expect 1A may have some curvature, it is assumed linear in Eq.

(4.41) because there are only two known boundary conditions for it,

C 1A x 0 C 1Ao and C 1A x L C 1A*

or
1 1
A 0 0 and A 1 1.

With these conditions, we can find the two constants in Eq. (4.41):
1
A . (4.42)

While we now have the shapes of the composition profiles in both phases and an
expression for the penetration depth in phase 2, the solution is not complete. The goal
of this exercise is to find how phase 1 grows, so we still seek the layer thickness, L(t).
The two-phase interface condition, Eq. (4.28), derived from the balance of solute
flowing down solute gradients on either side of the interface and the solute “stored”
there when the composition changes from the equilibrium value of one phase to that
in the other, is used here as another equation in terms of L(t).

C 1A C A2 dL
D1A DA2 C 1A* C A2* (4.43)
x x L
y y 0
dt

The gradients of composition in the two phases at the interface are

C 1A C 1A* C 1Ao 1
A C 1A* C 1Ao
x x L
L 1
L
Mass Diffusion in the Solid State 205

and

C A2 C Ai2 C A2* 2
A C Ai2 C A2*
2 .
y y 0 0

So,

C 1A* C 1Ao C Ai2 C A2* dL

D1A 2DA2 C 1A* C A2* . (4.44)
L dt

Rearranging Eq. (4.44) and using Eq. (4.40) for (t), the interface condition can be
rewritten as a differential equation for L(t):

dL D1A C 1A* C 1Ao C 1Ai C A2* 2

A
. (4.45)
dt L C 1A* C A2* C 1A* C A2* 3t

If we assume a solution of the form

dL DA2
L (t ) 2 DA2 t and . (4.46)
dt t

substitute it into Eq. (4.45), and rearrange the result into a quadratic in terms of the
unknown parameter, , then:

2 1 C Ai2 C A2* 1 D1A C 1A* C 1Ao

2* 1
0.
3 CA CA * 2 DA2 C A2* C 1A*

In Eq. (4.46), we have assumed that L(t = 0) = 0. Solving for gives

1 C Ai2 C A2* D1A C A2* C 1A* C 1A* C 1Ao

2* 1*
1 1 6 . (4.47)
2 3 CA CA DA2 C Ai2 i C A2* C Ai2 C A2*

Eqs. (4.46) and (4.47) describe the diffusion-controlled behavior of the two-phase
interface.
An exact solution is available for this problem in the form of the composition
profiles in the two phases:

C 1A C 1Ao erf x 2 D1A t

1
A (4.48)
C 1A* C 1Ao erf DA2 D1A

and

C A2 C Ai2 erfc x 2 DA2 t

2
A . (4.49)
C A2* C A2 i erfc
206 An Introduction to Transport Phenomena in Materials Engineering

Using these exact profiles and the interface condition, Eq. (4.43), gives the exact
expression for :
C A2* C Ai2 1
C A2* C 1A* exp 2
erfc
C 1A* C 1Ao 1
. (4.50)
C A2* C 1A* D2
D 2
2 DA2
A
1
exp A
1
erf
DA D A D1A
The root of Eq. (4.50) provides the value of for the composition profiles, Eqs.
(4.48) and (4.49), and the interface position, Eq. (4.46).

Example 4.3
Diffusion-Controlled Phase Change of Austenite to Ferrite during Decarburization
Consider the decarburization of a plain carbon steel, with an initial carbon content
of CCi = 3 at% at 750 °C by a gas that maintains a carbon content of CCo = 0.05 wt%
at the surface (x = 0). From Figure 4.12, the carbon contents of the ferrite ( ) and
austenite ( ) phases at the phase boundary are CC * = 0.09 and CC * =2.7 wt%. In dilute
solutions, as concentration in atom percent is proportional to concentration in
(g mol/m3) and the solutions contain only expressions for differences in concentra-
tion, the concentrations can be written as

CCi 3.0 at% CC * 2.7 at% CC * 0.09 at% CCo 0.05 at%.

At 750 °C, DC = 1.1 × 10–12 m2/s and DC = 5 × 10–11 m2/s.

FIGURE 4.12 Iron-carbon phase diagram near the eutectoid reaction (not to scale).
Mass Diffusion in the Solid State 207

Find: (a) Plot the behavior of the interface over time, L(t), and (b) the time
required for the ferrite layer to reach x = 0.1 mm and the composition pro-
files at that time.
Solution: (a) To plot the position of the - interface over time, we use
Eq. (4.46) and calculate from Eq. (4.47) for the approximate solution and
from the root of Eq. (4.50) for the exact solution. The calculated values are
for approx = 0.558 and exact = 0.537, a 3.9% difference. A plot of the inter-
face position, L, in Figure 4.13, shows that the approximate solution grows
slightly faster due to the neglect of the effect of the moving phase boundary
in Eq. (4.39).

To find the time required to grow a ferrite layer 0.1 mm thick, Eq. (4.46) is used:

2 2
L 1 10 4 m 1
t 0 s.
7300
2 DC 2 0.558 1.1 10 12
m 2 /s

The approximate profiles in the and phases when L = 0.1 mm are found from Eqs.
(4.35) and (4.42):

x x
CC CCo CC * CCo 0.05 at% 0.09 at% 0.05at%
L 0.0001m
0.05at% 4.0 10 6 at% m x

and

2
* * x L x L
CC CC CCi CC 2
12 DC t 12 D t

2.7 at% 3.0 at% 2.7 at%

2
x 0.0001m x 0.0001m
2 12
12 1.1 10 12
m 2 /s 7300 s 12 1.1 10 m 2 /s 7300 s

at% at% 2
2.7 at% 2030 x 0.0001m 3.42 106 x 0.0001m .
m m2

These compositions are plotted in Figure 4.14, where they are compared to the exact
solutions. It is interesting to note that, in this example, the carbon profile in ferrite is
(almost) linear in the exact solution, as we assumed for convenience in the approxi-
mate. This linear shape is because DC 50 DC and so the response of the composition
field to the changing interface position is much faster in than in .
208 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 4.13 Growth of -ferrite into austenite ( ) when exposed to a decarburizing

atmosphere.

FIGURE 4.14 Composition profiles in ferrite ( ) and austenite ( ) phases at the time
(t = 7300 s) at which the phase boundary reaches L = 0.1 mm.

4.6 DIFFUSION IN SUBSTITUTIONAL SOLID SOLUTIONS

In the preceding discussion, the diffusion of interstitial solutes, such as carbon and
hydrogen in iron, was considered, in which the diffusing atoms jump from one inter-
stitial site to the next in the crystal lattice of the solvent. The solute atoms are like
small animals wandering through a dense forest, in which the trees are unaware of the
motion of the animals, and thus in considering the diffusion of carbon in an infinite
diffusion couple comprised of steels of differing carbon contents, no consideration
was given to the possible movements of the individual iron atoms. Also, because the
primary locations for the small solute atoms are on the interstitial sites in the iron
lattice and those sites are almost always available, the probability is roughly the same
for a solute atom to jump to any of its neighboring sites. Also, the availability of sites
Mass Diffusion in the Solid State 209

into which to jump is significantly higher than in substitutional alloys, so the speed
of diffusion is much higher in interstitial solutions.
Consider now a diffusion couple comprising metals of the same crystal structure
and similar atomic radii, such as silver and gold. At what rates do the silver and gold
diffuse into each other, and how is the diffusion process described? The answer to
the first question was provided by the experiments of Smigelskas and Kirkendall [2].
(The second question will be addressed in the next section.) Their experiments begin
with a rectangular bar of brass (Cu-30wt%Zn), shown in Figure 4.15, wound with
fine molybdenum wire and then electroplated with a layer of pure copper. Molyb-
denum is insoluble in both copper and brass, and the wire served as a marker of
the position of the interface. During annealing it was observed that the distance, d,
decreased monotonically with time, which indicated that the flux of zinc atoms from
the brass outward past the markers is greater than the flux of copper atoms inward
past the markers. The phenomenon is known as the Kirkendall effect.

4.7 DARKEN’S ANALYSIS

In the previous section, an experiment is described in which two metals (pure copper
and Cu-30% Zn brass) form a diffusion couple, and the flux of Zn into the Cu is faster
than the flux of copper into the brass, Figure 4.15. We can measure rates of mass
flow in the couple from this experiment, and the theory suggested by the results was
provided by later Darken [3].
To understand the motion of the interface in the Kirkendall effect, consider a train
traveling due north at 60 mph in which passenger A is walking toward the front of the
train at 4 mph and passenger B is walking toward the rear of the train at 5 mph. Are
the passengers traveling in northerly or southerly directions? The answer depends on
the frame of reference in which the passengers are observed. A stationary observer
standing beside the track (the world frame of reference) sees the train traveling in
a northerly direction at 60 mph, passenger A traveling in a northerly direction at
64 mph, and passenger B traveling in a northerly direction at 55 mph. However, a

FIGURE 4.15 Diffusion couple used by Smigelskas and Kirkendall [2].

210 An Introduction to Transport Phenomena in Materials Engineering

passenger seated on the train (the train frame of reference) sees passenger A traveling
in a northerly direction at 4 mph and passenger B traveling in a southerly direction
at 5 mph. In this analogy, the seated passengers are being transported by the bulk
motion of the train and the walking passengers are being transported by both bulk
flow and “diffusion” relative to it.
In Kirkendall’s experiment involving an (effectively) infinite diffusion couple of
metals A and B, in which the original interface is identified by inert markers, an
observer seated on a marker (the marker frame of reference, like the train) sees only
the movement of A and B across the marker plane, caused by diffusion down a con-
centration gradient given by

CA CB
jA DA and jB DB .
x x

However, an observer near an end of the diffusion couple (in the world frame of
reference), in a region where no concentration gradients occur, sees the movement
of A and B at the position of the marker as being the sum of that due to diffusion
across the plane of the marker and the bulk motion of the marker. Thus, if the marker
is moving with a velocity, Vm, relative to the observer near one end of the diffusion
couple and the concentration of A at the position of the marker is CA, this observer in
the lab frame sees the flux of A as

CA
NA j A C AVm DA C AVm . (4.51)
x

The first term on the right-hand side of Eq. (4.51) is the rate of transport of
A across the plane of the marker by diffusion, and the second term is the rate
of movement of the marker plane relative to the observer in the lab frame. For a
system in which both diffusion and bulk movement are occurring, Fick’s second
law, Eq. (4.4), becomes

CA NA CB NB
= and . (4.52)
t x t x

As the total number of moles per unit volume is C C A C B , Eqs. (4.51) and (4.52),
give

C CA CB CA CB
DA DB CVm (4.53)
t t t x x x

If the molar volume of alloys in the system A–B is independent of composition (i.e.,
is constant and uniform), then ∂C/∂t = 0, in which case integration of Eq. (4.53)
becomes

CA CB
DA DB CVm I, (4.54)
x x
Mass Diffusion in the Solid State 211

in which I is an integration constant. Eq. (4.54) is valid along the entire length of
the diffusion couple, and Vm is the velocity of a marker placed at any position in
the couple relative to the observer in the lab frame. By the definition of an infinite
diffusion couple, no concentration gradients occur near the ends, in which case Eq.
(4.54) gives I = -CVm. But with no concentration gradients, and hence no diffusion
occurring near the ends, Vm = 0 and hence I = 0. Thus, Eq. (4.54) becomes

1 CA CB
Vm DA DB . (4.55)
C x m x m

Eqs. (4.55) and (4.51) are then substituted back into Eq. (4.53) for A to obtain

CA NA CA
DA C A Vm
t x x x
(4.56)
CA CA CA CA CB
DA DA DB .
x x C x C x

Because C is uniform, ∂CA/∂x = – ∂CB/∂x, and hence Eq. (4.54) becomes

CA CA CB CA CA CA CA CA
DA DA DB
t x C x C x C x
C B DA C A DB C A
x C x

or, as CA/C = XA and CB/C = XB, the mole fractions of A and B, respectively,

CA CA
X B DA X A DB . (4.57)
t x x

Eq. (4.57) is identical with Fick’s second law of diffusion, Eq. (4.5), if

D X B DA X A DB . (4.58)

Here D is the interdiffusion coefficient for the system A–B, and DA and DB are the
chemical diffusion coefficients for A and B. Also, rewriting Eq. (4.55) as

1 CA CA XA
Vm DA DB DA DB (4.59)
C x x x

gives the velocity of the marker plane at which the gradient of the mole fraction
of A is ∂XA/∂x. The marker planes in the diffusion couple are stationary only when
DA = DB, which corresponds to the case in which the flux of A and B are equal but in
opposite directions.
Eq. (4.58) shows that the interdiffusion coefficient, D, in Fick’s law for the system
A–B is not uniform because it is a function of local mole fraction. The question now
212 An Introduction to Transport Phenomena in Materials Engineering

arises: Are the chemical diffusion coefficients DA and DB uniform? Atoms move as a
result of the application of a gradient in a potential field. In a system A–B with con-
centration gradients, the potentials of interest are the chemical potentials of A and
B. The chemical potential of a gram mole of the species i in an isothermal, isobaric
system is defined as

i i RT ln ai , (4.60)

in which ai is the thermodynamic activity of i relative to some standard state. This

state is usually pure i, in which state the chemical potential is io and the activity of i
is, by definition, unity. Alternatively, the chemical potential per atom of i is written as
o
i i
i = kT ln ai ,
N0 N0

in which N0 is Avogadro’s number. The existence of concentration gradients gives

rise to gradients in the chemical potentials of the species in the system, which causes
atomic motion in the direction of the negative gradient. This atomic movement down
the gradient decreases the chemical potential of A. In this field, the atoms acquire a
mean velocity, V , and the mobility, B, of the atom is defined as:

BA V F o V i x .

The flux of A atoms (atoms/m2s) with this mobility is thus

d i
jA N 0 C AV C A BA . (4.61)
dx

The thermodynamic activity of A, defined in Eq. (4.60), is the product of an activity

coefficient A and the mole fraction of A, XA = CA/C. Thus, Eq. (4.60) gives

d A d ln X A d ln A
RT , (4.62)
dx dx dx

substitution of which into Eq. (4.61) gives the flux of A in (g mol/m2s), jA:

jA d ln X A d ln A 1 dC A d ln A
jA C A BA kT C A BA kT
N0 dx dx C A dx dx
(4.63)
dC A d ln A d ln A dC A
BA kT 1 BA kT 1 ,
dx d ln C A d ln X A dx

which, on comparison with Fick’s first law, gives the chemical diffusion coefficient
for A as

d ln A
DA BA kT 1 . (4.64)
d ln X A
Mass Diffusion in the Solid State 213

If the system A–B obeys Raoult’s law ( A = 1) or Henry’s law ( A = constant) over
some range of composition, Eq. (4.64) becomes

DA BA kT, (4.65)

the Nernst–Einstein equation. The thermodynamic solution behavior of a solute in a

binary system approaches Henry’s law, and the behavior of the solvent approaches
Raoult’s law as the concentration of the solute approaches zero. However, in many
solutions it can be considered that Henry’s law is obeyed by the solute over some
finite range of dilute solution, in which case Raoult’s law is obeyed by the solvent
in the same range of composition. Within these ranges of composition, the chemi-
cal diffusion coefficients of the species are uniform, but in concentrated solutions
they vary with composition in accordance with the nonideal thermodynamic solution
properties of the system.

4.8 SELF-DIFFUSION COEFFICIENT

If a thin film of the radioisotope A* of the metal A, containing moles of A* per unit
cross-sectional area, is plated onto one end of a semi-infinite rod of the stable isotope
A, and the rod is annealed, the radioisotope will diffuse into the rod. The concentra-
tion profile that develops is given by

x2
C A* exp , (4.66)
D*A t 4 D*A t

which defines the self-diffusion coefficient of A, D*A , assuming the two isotopes dif-
fuse at the same rate. Eq. (4.66) is shown for several values of (D*A t) in Figure 4.16.
With increasing time, the concentration profiles flatten as the A* spreads through
the rod. From Eq. (4.66) the concentration of C A* at x = 0 is proportional to 1 t ,
and because

2
exp d 2
0

then

C A* dx 1;
0

hence the area under each of the curves in Figure 4.16 is unity. A plot of the logarithm
of the experimentally measured C A* against x2 gives a straight line, with a slope of −1/
(4D*A t), which allows determination of the self-diffusion coefficient, D*A .
In the rod, because A and A* are chemically identical, there are no chemi-
cal concentration gradients and hence no gradients in chemical potential. Thus,
in the absence of a potential for diffusion, the concentration profiles shown in
Figure 4.16 are developed by a random walking process of A atoms and the
self-diffusion coefficient is a quantitative measure of this process. The probability
214 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 4.16 Concentration profiles of a radioisotope A* diffusing from an initial plane

source into a semi-infinite rod of A.

that an A* atom has randomly walked to a region between x and x + dx is p(x) dx

is the number of A* atoms between x and x + dx divided by the total number of
A* atoms in that region,

C A* 1 x2
p x dx dx exp dx .
DA* t 4 DA* t

The mean square distance, x 2 , over which an atom of A* has diffused in time t
is thus

1 x2
x2 x 2 p x dx x 2 exp dx
0
DA* t 0 4D*A t

1 3/ 2 x x2
2 DA* t erf 2DA* tx exp ,
DA* t 2 DA* t 4DA* t
0
*
2D t A

which is the same as the results in Eqs. (1.42) and (1.44).

Is the measured self-diffusion coefficient of the radioactive atoms A*, D*A , equal to
the chemical diffusion coefficient of A in the A–B alloy? Consider an alloy diffusion
couple in which CB has the same value in both halves of the couple and the initial
value of CA in one half equals the initial value of the sum of CA + C A* in the other half.
From Eq. (4.63), the self-diffusion coefficient is

d ln A*
D*A B*A kT 1 . (4.67)
d ln X *A
Mass Diffusion in the Solid State 215

However, as the stable and radioactive isotopes are chemically identical, A depends
only on the chemical composition of the alloy (i.e., on the sum of CA and C A* )
and is independent of the ratio C A* /CA, and thus, for a constant CA + C A* , the term
d ln A* d ln X *A . in Eq. (4.67) is zero. Furthermore, the mobilities of the two
isotopes of A are the same, and thus

D*A BA* kT BA kT,

which from Eq. (4.63) leads to

d ln A
DA DA* 1 . (4.68)
d ln X A

Thus, the chemical diffusion coefficient of A, DA, in a system containing concentra-

tion gradients is only equal to the self-diffusion coefficient of A, D*A, if the system
obeys Raoultian (and hence Henrian) ideal behavior.
As the variations of DA and DB with composition are caused by nonideal thermo-
dynamic behavior in the system, the interdiffusion coefficient of the system A–B can
be related to the self-diffusion coefficients by means of the Gibbs–Duhem equation
for the binary system:

XAd A XBd B 0
or

X A RT d ln X A X A RT d ln A X B RT d ln X B X B RT d ln B 0. (4.69)

However, as

XA XB 1 dX A dX B 0

XA XB
dX A dX B 0 X A d ln X A X B d ln X B 0,
XA XB

the sum of the first and third terms in Eq. (4.69) is zero, and hence with some
rearrangement,

d ln A d ln B
. (4.70)
d ln X A d ln X A

Thus, combination of Eqs. (4.58), (4.68), and (4.69) gives

d ln A d ln B
D X B DA X A DB X B DA* 1 X A DB* 1
d ln X A d ln X B
(4.71)
* *
d ln A
XB D A XAD B 1 .
d ln X A
216 An Introduction to Transport Phenomena in Materials Engineering

Example 4.4
Doping Silicon with Phosphorus by Diffusion
To produce a semiconductor by doping silicon, a thin layer of phosphorus is placed
on the surface of the silicon by vapor deposition and allowed to diffuse into the sili-
con. A film of thickness L = 1 μm is deposited and diffused into the silicon at 1000
°C for 5 h. At that temperature,

DPSi = 1.2 10 –17 m 2 / s P = 2000 kg P / m 3 Si = 2300 kg Si / m 3

MWP 0.03097 kg P / g mol P MWSi = 0.02809 kg Si / g mol Si

Find: The depth at which the mole fraction of phosphorus has a value of 10 –3.
Solution: The concentration profile developed from a plane source is given by

x2
CP exp (4.72)
DP t 4DP t

For a 1 μm vapor-deposited layer of phosphorus, the quantity per unit area of surface, is

L 2000 kg P/m 3 10 6 m
P
6.46 10 2 g moll P m 2 .
MWP 0.03097 kg P/g mol P

A mole fraction of 10–3 of P in Si corresponds to a concentration of

3 gmol of P 0.02809 kg Si gmol Si gmol of P

CP 10 81.9
gmol of Si 2300 kg Si m 3 m3

Therefore, in Eq. (4.72), with t = 5 h = 18,000 s,

gmol of P 6.46 10 2 g mol P m 2 x2

81.9 exp ,
m3 1.2 x 10 17
m 2 /s 18, 000 s 4 1.2 x 10 17
m 2 /s 18, 000 s

which gives the penetration depth as x = 2.44 × 10–6 m = 2.44 μm.

4.9 MEASUREMENT OF THE INTERDIFFUSION COEFFICIENT:

BOLTZMANN–MATANO ANALYSIS
To measure the interdiffusion coefficient of a solute in an alloy, we might construct
an infinite diffusion couple (as in Section 4.4) and measure composition profiles at
several times after the start of diffusion at particular temperatures. By curve fitting
these profiles to the mathematical solutions found in Eq. (4.23) or (4.26), we can
obtain values for D, if it is not a strong function of composition. When the inter-
diffusion coefficient does vary with composition (a common case for substitutional
alloys), the Boltzmann–Matano is used to obtain D from an experimentally measured
Mass Diffusion in the Solid State 217

concentration profile in an infinite diffusion couple. We begin the analysis by defin-

ing the function as

x t. (4.73)

The use of the chain rule produces these expressions:

CA dC A 1 x dC A
t d t 2 t 3/ 2 d

and

CA dC A 1 dC A
1/ 2
.
x d x t d

Substitution of the latter into Fick’s second law, Eq. (4.5), gives

1 dC A d dC A
D ,
2 d d d

which, being an ordinary differential equation, can be written as

1 dC A
dC A d D . (4.74)
2 d

The infinite binary diffusion couple is made from pure A and pure B and has the
initial conditions

C Ao for x 0 at t 0
CA .
0 for x 0 at t 0

The integration of Eq. (4.74) from CA = CA to CA = 0 gives

CA
1 CA dC A
dC A D
2 0 d 0

or, as the concentration profile is always measured after diffusion has occurred for
some time, t,
CA
1 CA dC A
x dC A Dt . (4.75)
2 0 dx 0

At a position in the couple where CA = 0, and there is no flux (dCA/dx = 0),

1 CA CA CA
x dC A Dt . (4.76)
2 0 dx CA dx 0
218 An Introduction to Transport Phenomena in Materials Engineering

Similarly, at a position in the couple where CA = CAo, dCA/dx is also zero, in which
case Eq. (4.75) gives
C Ao
x dC A 0,
0

defining the plane in the diffusion couple at which x = 0. This plane is known as the
Matano interface. The plane at which x = 0 is shown in Figure 4.17 as that which
makes the light gray regions of equal area. Eq. (4.76) then gives

1 dx CA
D= x dC A (4.77)
2t dC A CA
0

as the value of the interdiffusion coefficient for the composition CA. In Figure 4.17
the value of
0.2 C Ao
x dC A
0

is the dark gray area under the concentration profile between CA = 0.2CAo and CA = 0
and is the slope of the tangent to the concentration profile at CA = 0.2CAo.
The measured self-diffusion coefficients of nickel and gold in the system Ni-Au
at 900 °C are shown in Figure 4.18, and the measured interdiffusion coefficients are

FIGURE 4.17 Concentration profiles in an infinite diffusion couple used in the Boltzmann–
Matano analysis for determination of the interdiffusion coefficient.
Mass Diffusion in the Solid State 219

FIGURE 4.18 Variation with mole fraction Ni of self-diffusion coefficients of Ni and Au in

the Ni-Au system at 900 oC.

shown in comparison with values calculated using Eq. (10.34) in Figure 4.19 [4].
The thermodynamic activities of nickel and gold and the thermodynamic factor in
Eq. (10.34),

d ln Ni
1 ,
d ln X Ni

are shown in Figure 4.20. The system Au-Ni has a miscibility gap in the solid state that
begins at XNi = 0.8 and T = 820°C, and the minimum in the thermodynamic factor at
this composition at 900 °C reflects the tendency of the system toward immiscibility.
The minimum in the interdiffusion coefficient in Figure 4.20 is caused by the thermo-
dynamic factor and not by any peculiarity in the composition dependence of the term,
* *
X Ni DAu X Au DNi ,

in Eq. (4.71). The very low interdiffusion coefficients in the composition range over
the peak of the immiscibility gap are thus caused by a low thermodynamic driving
force in this range and not by any anomaly in the atomic mobilities. The variations
of D with composition in four other binary metal systems are shown in Figure 4.21.
220 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 4.19 Comparison of the measured and calculated interdiffusion coefficients in the
Ni-Au system at 900 oC.

FIGURE 4.20 Variation of the thermodynamic activities and the thermodynamic factor in
the Ni-Au at 900 oC.
Mass Diffusion in the Solid State 221

FIGURE 4.21 Composition dependencies of the interdiffusion coefficients in four binary

systems.

4.10 INFLUENCE OF TEMPERATURE ON THE

DIFFUSION COEFFICIENT
The examination of temperature dependence of D begins with interstitial diffu-
sion through a face-centered cubic (FCC) iron unit cell. Figure 4.23 (a) shows a
different perspective of the unit cell of FCC iron shown in Figure 4.22 and iden-
tifies three interstitial sites (1, 2, and 3) along the x-axis available for occupancy
by a small impurity atom such as carbon. At 950 °C the diameter of the Fe atom
in the structure is 2.58 Å, and thus the diameter of the interstitial site is 1.07 Å.
The equilibrium distance between atoms a and b in Figure 4.23 (a) is 1.04 Å. The
carbon atom has a diameter of 1.54 Å, so the presence of a carbon atom on an
interstitial site causes local distortion of the iron lattice, which is why the solubility
222 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 4.22 Face-centered cubic crystal structure of Fe showing locations of the octahe-
dral interstitial sites.

of carbon in iron is relatively small. An interstitial atom on an interstitial site such

as site 2 vibrates in all directions about the center of the site with a frequency of
vibration, v, which, typically, is on the order of 1013Hz. For most of the time the
vibrating atom is contained in an “energy well” that arises from the presence of
the six neighboring Fe atoms at locations a to f (i.e., if the interstitial moves in any
direction from the center of the well it experiences a repulsive force exerted by
the neighboring Fe atoms). The variation of the energy of an interstitial atom with
position along the x-axis, Figure 4.23 (b), shows that if an interstitial atom is to
squeeze successfully between the atoms a and b and jump to the interstitial site 1
as a result of its vibration, it must have at least the energy Ea, the activation energy
required for the jump. With that qualitative description of how a jump might occur,
the question now arises: at what rate do interstitial atoms jump successfully from
one interstitial site to another?
In a large population of n atoms, the number of atoms that have energies greater or
equal to Ei is ni n exp( Ei kT ). Thus, the fraction of interstitial atoms in a solution
in a host lattice that have an energy greater than or equal to the required activation
energy for a successful jump from one site to another is exp( Ea kT ), which also
represents the probability that any one atom is sufficiently energetic for a successful
jump. In Figure 4.22, the interstitial atom at the center of the unit cell can jump to
any one of 12 neighboring interstitial sites, and thus the frequency, , with which an
interstitial atom at a site such as 2 in Figure 4.23 (a) jumps to site 1 is:

(frequency of vibration)
x (probability the atom is sufficiently energetic to make a successful jump)
x (probability that jump is in direction 1 2),
Mass Diffusion in the Solid State 223

FIGURE 4.23 (a) Interstitial sites in the face-centered cubic Fe lattice. Atoms on
cell faces denoted by letters a–f; some interstitial sites denoted by numbers 1–3.
(b) Variation of energy with position between interstitial lattice sites.

that is,

v exp Ea / kT 12. (4.78)

The activation energy for diffusion of carbon in FCC iron is 2.223 × 10–19 J/atom or
133,900 J/g mol and thus at 950°C,

2.223 10 19 J Catoms 1
1013 s exp 1.59 106 s 1 , (4.79)
1.381 10 23 J Catoms K 1223K 12

(i.e., a carbon atom makes 1.59 × 106 jumps/second).

224 An Introduction to Transport Phenomena in Materials Engineering

If the number of interstitials per unit area on the plane R in Figure 4.23 (b) is NR,
the flux of interstitials from the plane R to the plane L in the -x-direction is

jR L NR ,

and, if the number of interstitials per unit area on the plane L in Figure 4.23 (b) is NL,
the flux of interstitials from the plane L to the plane R in the x-direction is

jL R NL ,

The net flux of interstitials in the x-direction is thus

j jL R jR L NL NR , (4.80)

with the planes L and R being separated by the jump distance , the concentration of
interstitials on a plane, CI, is

CI N , (4.81)

substitution of which into Eq. (4.80) gives

2 CR CL
j CL CR . (4.82)

In Eq. (4.82), the term in parentheses is the concentration gradient of the interstitials
in the x direction, and hence Eq. (4.82) can be written as

2 C
j . (4.83)
x

Comparing Eq. (4.83) to Fick’s first law, j D C x , shows that

2
D

or, from Eq. (4.78),

2
D v exp Ea / kT 12. (4.84)

The preexponential terms in Eq. (4.84) can be lumped together as D0 to give the
exponential dependence of the diffusion coefficient on temperature as

D D0 exp Ea / kT . (4.85)

In Eq. (4.85), Ea is the activation energy for diffusion per atom, and multiplying
this quantity by Avogadro’s number, N0, gives the activation energy for diffusion per
gram mole of diffusing species, in which case the diffusion coefficient is given by

D = D0 exp Ea / RT , (4.86)
Mass Diffusion in the Solid State 225

Fe
FIGURE 4.24 Variation with temperature of the diffusion coefficient DC for carbon
in body-centered cubic iron ( ferrite).

in which R N 0 k is the universal gas constant. Whether or not the activation energy
is in units of energy per atom or energy per gram mole can be determined by observ-
ing whether the expression contains Boltzmann’s constant, k , or the gas constant, R.
The variation of the logarithm of the diffusion coefficient of carbon in body-
centered cubic iron with temperature over the range −38 °C to 800 °C is shown in
Figure 4.24. The variation is linear with the equation

Ea 1 4400 K
log10 D + log10 Do 5.70.
R T T

From the slope of the line the activation energy for diffusion is

4400 K 2.303 8.3144 J gmol K 84, 300 J / g mol.

4.11 SUMMARY
Diffusion in the solid state occurs by the random motions of atoms, and the mean
square distance through which an atom moves by random motion is proportional to
the time for which the diffusion has occurred. In binary alloys diffusion occurs down
concentration gradients, and Fick’s first law states that the diffusion flux is propor-
tional to the concentration gradient, with the proportionality constant being the diffu-
sion coefficient of the diffusing species. The equation for solute conservation (Fick’s
second law) states that the rate of change of concentration of the diffusing species at
226 An Introduction to Transport Phenomena in Materials Engineering

any location is equal to the gradient of the diffusion flux (which itself is proportional
to the second derivative of concentration with respect to distance). Thus, although
the phenomenon of mass transport by diffusion is intrinsically different from the
phenomenon of heat transport by conduction, the similarity of the laws governing
the phenomena makes the mathematical treatment of the diffusion of interstitials in a
solid very much like the mathematical treatment of heat conduction.
Scaling and other approximate analyses are applied to single-phase, diffusion-con-
trolled processes, producing solutions to problems in carburization, doping of silicon,
nitriding, and diffusion bonding. A control volume analysis applied to the interface
of two solid phases gives the solute conservation at that boundary, where the net
flux across the phase boundary may cause growth and dissolution of the phases and
the motion of that interface. Cases involving alloy homogenization and precipitate
growth are studied using the interface solute balance and Fick’s second law.
Observation of diffusion in substitutional solid solution shows that the diffusion
coefficient, defined by Fick’s first law, can vary significantly with composition. This
diffusion, or interdiffusion, coefficient, is thus only a measure of the rate at which
concentration gradients in the alloy are eliminated by diffusion processes, and it is
not an intrinsic property. It is instead a function of the chemical diffusion coefficients
of the individual component species and of local composition. Fick’s laws hold only
for thermodynamically ideal solutions in which the thermodynamic activity of the
solute is proportional to its concentration (i.e., the solute obeys Henry’s law and the
solvent obeys Raoult’s law). In general, mass transport by diffusion occurs down
gradients of chemical potential (thermodynamic activity gradients), with the conse-
quence that the chemical diffusion coefficients of the diffusing species are only inde-
pendent of composition if the system is thermodynamically ideal. In such systems,
the chemical diffusion coefficient of a species is proportional to its mobility, where
the mobility is defined as the velocity of the diffusing atom per unit force acting on
it, and the force arises from the gradient in chemical potential. Darken’s analysis
models these phenomena.
The chemical diffusion coefficient of a diffusing species in an alloy is also only
equal to its self-diffusion coefficient (determined by measuring isotope diffusion in
a one-component system) if the system is thermodynamically ideal. Kirkendall’s
experiments show that, in systems in which the chemical diffusivities of the compo-
nents have different values, the diffusion process causes movement of the positions
of lattice planes relative to a fixed reference and hence mass transport in the system is
the sum of motion by diffusion down a concentration gradient and motion of the lat-
tice planes. Composition-dependent interdiffusion coefficients can be found directly
by measuring composition profiles some time after transport begins in an infinite
diffusion couple and applying the Boltzmann–Mantano analysis.
Diffusion is a thermally activated process, and hence the diffusion coefficient
increases exponentially with increasing temperature, with the rate of increase being
determined by the activation energy required for an atom to jump from one site to
another. This phenomenon shows again that there is not a physical analogy between
mass diffusion and heat conduction. Conduction is not a thermally activated process,
and the temperature dependence of the thermal diffusivity is determined by com-
pletely different phenomena.
Mass Diffusion in the Solid State 227

4.12 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

4.1 A carbon steel containing 0.1 wt% C is carburized at 950 °C by a gas in

which the carbon activity is equivalent to 1 wt% C in iron. (a) Calculate the
time required to obtain a carbon content of 0.5 wt% at a depth of penetra-
tion of 0.05 cm. (b) What carburizing temperature is required to get a con-
centration of 0.5 wt% at a depth of penetration of 0.1 cm in the same time
that 0.5 wt% was attained at 0.05 cm at 950 °C? Assume that the carbon
activity in the carburizing gas at the higher temperature still corresponds
to 1 wt% C in iron. The diffusion coefficient for carbon in austenitic iron is
DC 7 10 6 m 2 s exp 133, 900 J/gmol RT .
4.2 Hydrogen is stored at 400 °C in a long cylindrical steel vessel of inner
diameter 0.1 m. Calculate the wall thickness of the vessel at which the
hydrogen leaks from the vessel by diffusion through the wall at the rate less
than 0.1 cm3(STP)/s per meter length of the vessel. The pressure of hydro-
gen inside the vessel is maintained at 101.3 kPa, and the hydrogen pressure
outside the vessel is negligibly small. At 400 °C, DCFe 10 8 m 2 s , and,
from Sievert’s law, CHFe kg m 3 7.29 10 5 P kg s2 m 5 .
4.3 Hydrogen in solution in a thick slab of solid nickel is removed by sub-
jecting one face of the slab to a vacuum at 600 °C. (a) If the initial con-
centration of H in the nickel is 8 gmol/m3, after what time is the local
concentration of H at a depth of 1 mm below the surface equal to C HNi =
4 gmol/m3? (b) At this diffusion time, what is the flux of hydrogen
from the surface? (c) How much hydrogen has been removed from the
slab per unit area by the time found in part (a)? For hydrogen in nickel,
D 7.77 10 7 m 2 s exp 41, 200 J/gmol RT .
4.4 In March 1944, an organization of mostly British prisoners of war (POWs)
held in a camp in Silesia were planning a mass escape. One difficulty was
fabricating steel tools that could hold an edge, and one POW with metal-
lurgy knowledge came up with a solution:

Travis made the wire-clippers . . . out of tie bars which he ripped

off the huts. He riveted them together like scissors and filed cutting
notches in them. It was very soft steel, so he hardened the metal him-
self. In a home-made forge he heated the clippers till they were red
hot, poured a few grains of sugar on the metal around the cutting
notches, and heated it up again so the carbon in the sugar was baked
into the metal. Then he plunged it into cold water, and the steel came
out hard enough to cut wire [5].

This process was a crude approximation of pack carburization, during which

solid carbon is packed on a surface and, when heated, it diffuses into the steel.
228 An Introduction to Transport Phenomena in Materials Engineering

Mass Diffusion in the Solid State 229

FIGURE 4.25 Infinite diffusion couple for Problems 4.6 and 4.7.

edge of the appropriate penetration depth ( A or B) and one at the interface

and these variables:

2
B1 B2 B3

2
F1 F2 F3 .

Assume that the composition at the interface, CInt, is a constant throughout the
entire process.
(c) Show that integrating the conservation equations separately in each alloy
over the two penetration depths ( A and B), using the assumed profiles in
part (a), gives:

A 12DA t and B 12DB t .

Use the flux continuity condition at x = 0 to obtain an expression for the non-
dimensional interface composition, Int, in terms of only the diffusion coeffi-
cients. (d) Based on your results, sketch C(x) at a time t > 0 for these three
conditions:

(i) DA /DB > 1 (ii) DA /DB < 1 (iii) DA /DB = 1.

4.8 An iron film (with no nitrogen in it) is vapor-deposited on a silicon wafer

with a thickness of d = 10 μm. The film is then exposed at time t = 0 to a
nitriding atmosphere, which fixes the nitrogen concentration in the iron on
the exposed surface (x = 0) at C NFe = Cs. You would like to know the chang-
ing nitrogen field in the iron film as a function of position and time. You
may assume that: (i) Cs is much higher than the nitrogen solubility in the
silicon, so C NSi ≈ 0, and there is no nitrogen flux of nitrogen into the silicon;
and (ii) Cs is sufficiently low that there are no phase transformations in
230 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 4.26 Schematic for Problem 4.8.

the iron and no significant volume change. The diffusion coefficient for
nitrogen in iron is DNFe = 10 −12 m2/s. (a) Sketch the concentration profile of
nitrogen at the start of the process (t = 0), and for several times leading to
steady state. (b) Write the solution of the diffusion equation that is valid
for a short time after the nitriding atmosphere is introduced. What is the
thickness of the sublayer in which nitrogen concentration is at least half of
the surface concentration at 1 s and at 4 s? (c) At what time, tcrit, does your
solution in part (b) cease to be valid? (d) Write the solution of the diffusion
equation that is valid after tcrit. At what time, t, does the nitrogen concentra-
tion on the iron side of the iron-silicon interface reach 0.9Cs? (contributed
by Adam Powell, Worcester Polytechnic Institute)
4.9 During the solidification of a binary alloy, the two components freeze at
different rates, leaving a composition profile that is solute-poor in the cen-
ter of a dendrite and solute-rich at the position where a dendrite impinges
on the one growing next to it (the last region to freeze). This effect, known
as coring, may set up a repetitive composition field, as shown in the figure.
Because coring causes nonuniform properties, the alloy is homogenized
by heat treating at a high enough temperature to allow diffusion to occur.
(a) In Figure 4.27, sketch the composition profile, C(x), between x = 0 and
x = L after the alloy has been held at an elevated temperature for a period
of time such that the composition at x = 0 has fallen 50% of the difference
between Cave and . (Note that Cmax is C(x=0) and = Cmax(t = 0).)
What is the average composition, Cave, at that time? (b) Using these
normalizations:
* *
C Cave x tD Cmax Cmin
*
Cave ,
Cmax Cave L L2 2
find the composition profile, ( , ) = B1 cos(π ), as a function of the nor-
malized composition at x = 0, o ( ):
C x 0, t Cave
o 0, *
.
C max Cave
Mass Diffusion in the Solid State 231

FIGURE 4.27 Schematic of composition variation after solidification at the dendrite scale
for homogenization in Problem 4.9.

(c) Using the normalized mass conservation equation, find an expression for
the normalized composition at x = 0, o ( ). (Hint: integrate between = 0
and = 1/2.) (c) For D = 10–13 m2/s and L = 100 μm, how long will it take
to get o = 0.5? o = 0.01?

4.10 In a Mg-9wt%Al alloy, blocky structures of phase (nominally Mg4Al3)

grow on grain boundaries near the end of solidification. One route to hard-
ening this alloy is to dissolve this phase and spread aluminum through the
matrix (homogenization) and then precipitate many, smaller particles
throughout the metal (aging). To model the aging process, we begin with
Al,i = 9 wt% Al and no present. At t = 0, begins to grow at Al,i = 42
wt% Al because the homogenized metal has been quenched to 150 oC, at
which temperature the metal is in the + region on the phase diagram, with
*
Al = 5 wt% Al. (a) Sketch the system at t = 0 and two later times, showing
the position of the - interface and CAl(x) in both phases. (b) Simplify the
solution given in Section 4.5, given that C Al x ≈0. (c) How thick will the
precipitate be after 48 h (which time gives the maximum hardness for this
m2 155, 00 J mole
heat treat)? DAl 3.9 10 5 exp [6].
s RT
4.11 Show that the expression in (a) Eq. (4.23) and (b) Eq. (4.24) are correct.

REFERENCES
1. Dossett, J., and G. E. Totten (eds.), ASM Handbook: Volume 4A: Steel Heat Treating
Fundamentals and Processes, ASM International, 2013.
2. Smigelskas, D., and E. O. Kirkendall, “Zinc diffusion in alpha brass,” Transactions of
the AIME, v. 171, pp. 130–142, 1947.
3. Darken, L. S., “Diffusion, mobility, and their interrelation through free energy in binary
metallic systems,” Transactions of the AIME, v. 175, pp. 184, 1948.
4. Reynolds, J. E., B. L. Averbach, and M. Cohen, “Self-diffusion and interdiffusion in
gold-nickel alloys,” Acta Metallurgica, v. 5, pp. 29–40, 1957.
5. Brickhill, P., The Great Escape, Faber & Faber, 1951.
6. Brennan, S., A. P. Warren, K. R. Coffey, N. Kulkarni, P. Todd, M. Kilmov, and Y. Sohn,
“Aluminum impurity diffusion in magnesium,” Journal of Phase Equilibria and Diffu-
sion, v. 33, pp. 121–125, 2012.
 5 Fluid Statics

5.1 INTRODUCTION
The study of fluid mechanics begins with the examination of fluids that are not mov-
ing, or at least in which no part of the fluid is moving relative to any other part.
Without such relative motion, there are no velocity gradients and so no shear stresses.
Without shear, the constitutive equations for a Newtonian fluid, Eqs. (1.10) and (1.11)
reduce to only the normal force per unit area, or pressure, throughout the fluid. In this
chapter, we look at the pressure distribution in a static fluid and some applications.

5.2 CONCEPT OF PRESSURE

5.2.1 pressure at a point and in a coluMn
The force exerted by a gas on the walls of a vessel is caused by collisions of the
individual atoms or molecules of the gas with the wall. An individual atom of mass
m approaches the wall from some direction with the velocity V , and, after an elastic
collision with the wall, rebounds in another direction at the same velocity magni-
tude. As momentum, mV , is a vector quantity that depends on direction as well as
magnitude, the change in the direction of motion of the atom causes a change in
its momentum. For example, if the atom of mass m is traveling in the +x direction
normal to the wall, its momentum before the collision is mu, and after the collision
is −mu. The change in the momentum of the atom as a result of the collision is thus
(mu) − (−mu) = 2mu. The population of atoms in the gas is continuously colliding
with the walls of the containing vessel, and hence there is a continuous rate of change
of momentum, which has the units

momentum mass velocity m

kg N .
time time s2

These are also the units of force, and thus the force exerted by the gas on the walls of the
container arises from, and is equal to, the rate of change of momentum of the gas atoms
or molecules caused by their collisions with the walls. The force exerted by a liquid on
the walls of its container arises from a similar phenomenon, and the fluid pressure is
this force per unit area. Any imaginary surface within the body of the fluid experiences
the same force in the direction normal to the plane of the imaginary surface.
The pressure at any point in a static fluid is the same in all directions, as can be seen
by considering the triangular element of fluid with sides of lengths a, b, and c and
unit depth, located within a body of fluid, as shown in Figure 5.1. Mechanical equi-
librium requires that the algebraic sum of the forces acting in the x-direction be zero,

Pa sin a Pb sin b, (5.1)

232 DOI: 10.1201/9781003104278-5
Fluid Statics 233

FIGURE 5.1 Pressures acting on a volume element of a fluid.

and the algebraic sum of the forces acting in the y-direction be zero,

Pa cos a Pb cos b w Pc c. (5.2)

In Eq. (5.2), w is the body force, or the weight of the element of fluid, which arises
from the influence of the gravitational field on the mass of fluid in the element. For a
mass of fluid m of density in an element of volume ,

1
w = mg = g = g a b . (5.3)
2

From the geometry of the triangular cross-section of the element:

y c2
sin cos
b b

y c1
sin cos .
a a

Thus, from Eq. (5.1),

Pa Pb (5.4)

and from Eq. (5.2),

1
Pa c1 Pb c2 g a b Pc c.
2
234 An Introduction to Transport Phenomena in Materials Engineering

As

Pa Pb and c1 c2 c,

a b2
Pa g Pc . (5.5)
c

As the element is made infinitesimally small, the body force approaches zero, and
hence in the limit of a = b = c 0,

Pa Pc ,

and thus

Pa Pb Pc , (5.6)

that is, the pressure at any point in the static fluid is the same in all directions.
The only forces acting in a static fluid are normal forces (such as those giving
rise to the pressures Pa, Pb, and Pc in Figure 5.1) and the body forces that arise
from the influence of gravity on the mass of the fluid. (Note that the normal forces
act on and perpendicular to the infinitesimal control volume surfaces, while body
forces act throughout the entire volume.) This influence of gravity is such that the
pressure on any plane in a static fluid normal to the gravitational field arises from
the weight of fluid above the plane, and thus the pressure at a point in a static fluid
in a gravitational field varies with position along the direction of the field. Consider
the column of fluid of cross-sectional area A in a gravitational field above the level
z shown in Figure 5.2 (a). The body force exerted by the column of fluid above the
level z on the fluid below the level z equals P |z A, where P |z is the normal pressure
in the fluid at the level z. In Figure 5.2 (b) the body force exerted by the column
of fluid above the level z + z on the fluid below the level z + z equals P |z+ z A,
where P |z+ z is the normal pressure in the fluid at the level z + z. Thus, in Fig-
ure 5.2(c), the difference between the normal force at level z and the normal force
at level z + z equals the body force arising from the weight of the fluid between
the two levels:

Pz z
A gA z Pz A

or

Pz z
Pz
g,
z

Which, in the limit of z 0, becomes

dP
g. (5.7)
dz
Fluid Statics 235

FIGURE 5.2 Derivation of the barometric formula from consideration of the body force of
a fluid in a gravitational field.

5.2.2 atMospHeric pressure

Consider atmospheric air above the surface of the earth. The standard atmospheric
pressure at sea level is 1.01325 × 105 Pa (1 standard atmosphere) at a temperature of
15 °C (288 K) and the standard acceleration due to gravity is g = 9.8067 m/s2 (which
is defined in terms of the pull of the earth’s gravitational field on a mass of 1 lb at sea
level and 45 °N latitude). Atmospheric air obeys the ideal gas law,

P RT M , (5.8)

where P = pressure of the gas (Pa), = volume of the gas per unit mass (= 1/ in m3/
kg), R = universal gas constant (= 8.3144 J/(g mol K)), T = absolute temperature of
the gas (K), M = molecular weight of the gas (kg/g mol).
Rearrangement of Eq. (5.8) as

1 PM
.
RT

And substituting it into Eq. (5.7) gives

dP PMg
dz RT
or
dP Mg
dz. (5.9)
P RT
236 An Introduction to Transport Phenomena in Materials Engineering

Assuming that T and g are independent of z, integration of Eq. (5.9) between the lim-
its P = Po (standard atmospheric pressure) at z = 0 (sea level) and P = P at z = z gives
P M gz
ln
P0 RT
or
M g z/RT
P P0 e , (5.10)

which is known as the barometric formula.

Air is a mixture of 21 vol% O2 (molecular weight 0.032 kg/g mol) and 79 vol% N2
(molecular weight 0.028 kg/g mol), and thus the molecular weight of air is

M 0.21 0.032 0.79 0.028 kg/g mol = 0.02884 kg/g mol.

At T = 288 K,

Mg 0.02884 kg/gmol 9.81m/s2

1.181 10 4 / m,
RT 8.3144 J/gmol K 288 K

and hence Eq. (5.10) becomes (where z = height in meters)

4
1.181 10 /m z
P P0 e , (5.11)

which is shown as line A in Figure 5.3.

FIGURE 5.3 Variation of atmospheric pressure with altitude. Line A, which is given by Eq.
(5.11), is drawn for a constant air temperature of 288 K. Line B, which is given by Eq. (5.14),
is drawn for an air temperature that decreases at the rate of 6.5 °C per 1000 m to an altitude
of 11,000 m.
Fluid Statics 237

By convention, the international standard atmosphere is defined on the basis of a

uniform temperature gradient of −6.5 °C per 10,00 m of altitude from sea level to a
height of 11,000 m and a constant temperature of −56.5 °C for heights greater than
11,000 m. Thus, in the range z = 0 to z = 11,000 m,

T 288 K 0.0065 K/m z, (5.12)

which, when substituted into Eq. (5.9) gives

dP Mg
dz, (5.13)
P R 288 K 0.0065 K m z
integration of which produces
5.245
288 K 0.0065 K m z
P P0 . (5.14)
288 K

Equation (5.14) is shown as line B in Figure 0.3.

As an example, from Eq. (5.14) it is calculated that the air pressure at the summit of
Mount Phillips in the Sangre de Christo Range in New Mexico (elevation 3,579 m, or
11,742 ft) is 0.64 atm, which illustrates why hikers have some difficulty breathing near
the summit. At 17,000 ft (5,182 m), which is the maximum altitude at which passengers
without oxygen masks can fly in an unpressurized airplane, the air pressure is 0.52 atm.
For completeness, it must be pointed out that the gravitational constant itself var-
ies with distance above sea level as
2
m 6388 km
g 9.8067 2 ,
s 6388 km z

where z is in kilometers. The gravitation acceleration at the summit of Mount Phillips

is thus 9.7957 m/s2. This effect on pressure is only a few percent at altitude of 50 km,
and so has little effect on problems in this text.

5.3 MEASUREMENT OF PRESSURE

In 1643, the Italian natural philosopher Evangelista Torricelli filled a 4-foot-long
glass tube with mercury and immersed the open end of the tube in a bath of mercury
as shown in Figure 5.4. He observed that the mercury did not run out of the tube and
concluded that a vacuum had been created above the mercury in the tube and that
the column of mercury was supported by the atmospheric pressure acting on the free
surface of the mercury in the bath.
At equilibrium, the normal force per unit area on the surface of the mercury (point
A) exerted by the column of air above the surface is equal to the body force per unit
area at point B exerted by the column of mercury above B. The body force F exerted
by the column of mercury of mass m, height h, and cross-sectional area A is

F mg A h g,
238 An Introduction to Transport Phenomena in Materials Engineering

which gives rise to a static pressure at point B of

F
P gh;
A
hence, measurement of the height h allows determination of the atmospheric pres-
sure at point B. Liquid mercury, like most condensed phases, is considered to be
incompressible, in which case its density is independent of position within the col-
umn. Thus, the static pressure in the mercury column decreases linearly with z from
the value pgh at point B to virtually zero at the free surface at the height h. Stand-
ard atmospheric pressure at 0 °C corresponds to h = 760 mm and is referred to as
“760 mm Hg” or “29.92 in Hg.” The arrangement shown in Figure 5.4 gave rise to
the development of the barometer, which provides a means of measuring the atmos-
pheric pressure.1
Many devices that measure pressure actually measure the difference between
the pressures of two fluids, giving rise to the distinction between absolute pressure
and gauge pressure. Inflating a tire to a measured gauge pressure (Pg) of 240 kPa
(35 psi) means that the difference between the pressure inside the tire and atmos-
pheric pressure is 240 kPa. If the atmospheric pressure outside the tire is 100 kPa
(14.7 psi), the actual pressure inside the tire is 340 kPa. This value is the absolute
pressure (Pabs) and is determined by the rate of change of momentum of the atoms or
molecules of the gas caused by their collisions with the walls of the container. Thus,

Pabs Pg Patm .

FIGURE 5.4 Torricelli vacuum created in an inverted flask containing a liquid.

Fluid Statics 239

FIGURE 5.5 Simple U-tube manometer for measuring differences in pressure.

Figure 5.5 shows a simple U-tube manometer, which allows measurement of the
difference between the pressure of a fluid in a tank, Pt, and the atmospheric pressure,
Po. The U-tube is partially filled with an incompressible fluid of density (which is
immiscible with the fluid in the tank), and, at equilibrium, the absolute pressure at
level A is the same in both legs of the manometer. At level A, the pressure is Pt in
the left leg while, in the right leg, it is that due to the body force of the column of
manometric fluid (the black region in Figure 5.5) of height h plus the atmospheric
pressure, Po, exerted on the surface of the manometric fluid. Thus,

Pt gh Po .

It has been assumed here that the fluid density at the bottom of the manometer is
much greater than the fluid in the tank or the atmosphere, thus allowing us to ignore
the hydrostatic pressure in those fluids. This assumption is valid, for example, if the
fluids in the atmosphere and tank are gases and in the manometer it is a liquid.
The Torricelli barometer shown in Figure 5.4 is also a manometer, in that it meas-
ures the difference between the atmospheric pressure at A and the pressure in the vol-
ume above the column of liquid in the closed end of the tube. While we have a zero
absolute pressure at the top of the column, the pressure of that “vacuum” is actually
the saturated vapor pressure of the liquid in the barometer, which for liquid mercury
at 298 K is 0.76 Pa. This value is negligible in comparison to an atmospheric pressure
of approximately 100 kPa.
Figure 5.6 shows a U-tube manometer containing two immiscible liquids, 1 and 2,
of densities 1 and 2, (where 1 > 2). The cross-sectional area of the tubing connect-
ing the two reservoirs is At, and the cross-sectional areas of each reservoir is Ar. With
atmospheric pressure, Po, exerted on the free surfaces of both liquids, the equilibrium
240 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 5.6 U-tube manometer containing two immiscible liquids.

state is as shown in Figure 5.6(a). The pressure at level z is the same in both legs of
the manometer,

Po 1 g h1 Po 2 g h2

or

1 1 h h.
2 2 (5.15)

In Figure 5.6 (b) the pressure exerted on the surface of liquid 2 has been
increased to Po+ P, which causes the level of the surface of liquid 2 to fall by the
distance Z, the position of contact of the two liquids in the right leg of the U-tube
to change by the distance z, and the level of the surface of liquid 1 to rise the dis-
tance Z. At the new state of equilibrium, the pressures in the two legs are equal
at level 2 and hence

P0 1 g x1 Po P 2 g x2

or

P g 1 x1 2 x2 . (5.16)

From geometric considerations

x1 h1 Z z
Fluid Statics 241

and

x2 h2 Z z.

Substitution into Eq. (5.16) gives

P g 1 h1 Z z 2 h2 Z z
g 1 1 h 2 2 h Z 1 2 z 2 1 .

which in view of the equality in Eq. (5.15), becomes

P g Z 1 2 z 2 1 . (5.17)

Conservation of total volume also give

z Ar Z At ,

substitution of which into Eq. (5.17) gives

ZAt
P g Z 1 2 2 1
Ar
At
g Z 1 2 2 1 . (5.18)
Ar

As Z is the measured quantity, the sensitivity of the manometer is increased by

decreasing the difference between 1 and 2 or the ratio At/Ar.

5.4 PRESSURE IN INCOMPRESSIBLE FLUIDS

In the case of incompressible fluids, it is often necessary to know the pressure at
some depth below the free surface of the liquid, in which case it is convenient to
consider that h increases in the -z direction (i.e., in the direction of the gravitational
field). Thus, taking Po to be pressure at the free surface, the pressure at a depth h
below the surface is obtained from integration of Eq. (5.7) as

P P0 g h, (5.19)

that is, in a fluid of constant density, the pressure increases linearly with increasing
depth below the free surface.
Figure 5.7 shows a plate of length L and width W immersed vertically in a
liquid of density . From Eq. (5.19), the pressure exerted by the liquid on each
face of the plate varies linearly with depth below the free surface from a gauge
pressure of zero at the free surface (z = 0) to a gauge pressure of gL at the
242 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 5.7 Vertical plate immersed in an incompressible fluid and the variation, with
depth, of the pressures exerted by the fluid on both sides of the plate.

bottom of the plate. The total force, exerted on each side of the plate is obtained
from the definition

Ftotal P dA,
A

In which A is area. For a plate of constant width W, dA = W dh, and hence

L
Ftotal ghW dh gL2W 2.
0

The average pressure exerted on each side of the plate is thus

Pave Ftotal WL gL 2,

which is the local gauge pressure exerted halfway down the plate.

Example 5.1
Force on a Submerged Wall
The rectangular metal water tank shown in Figure 5.8 has a glass window of dimen-
sions 2 m × 2 m located at a distance of 1 m below the waterline in the tank.

Find: What force does the water, = 1000 kg/m3, exert on the window?
Solution: From Eq. (5.19), the gauge pressure at the window centroid is

P P0 gzc 1000 kg/m 3 9.81m/s2 zc 9810 Pa/m zc .

Thus, the local pressure exerted at the centroid of the window (at zc = 2 m) is 19,620
Pa and the total force exerted on the window by the water is

Ftotal 19, 620 Pa 4 m 2 78, 480 N.

Fluid Statics 243

FIGURE 5.8 Variation with depth of the pressure exerted by an incompressible fluid on the
walls of its container.

Example 5.2
Force inside Oil Drum
A cylindrical tank of radius 2 m is placed horizontally and is half-filled with oil of
density 888 kg/m3.

Find: Calculate the force exerted by the oil on one end of the tank.
Solution: The force on the submerged wall of the oil drum is

F P dA,
area

where in Figure 5.9

P gy

and

dA 2 x dy.

From the equation of a circle,

x2 y2 R2

and
1/ 2
x R2 y2 .
244 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 5.9 Consideration of the pressure exerted by a fluid on the end of a cylindrical
drum.

Thus,
1/ 2
dA 2 R2 y2 dy

and
R
1/ 2 1 2 2 3/ 2 R3 2
F 2 g y R2 y2 dy 2 g R y 2 g 0 gR 3 .
3 0 3 3

2 2 3
F gR 3 888 kg/m 3 9.81 m/s2 2m 46, 450 N
3 3
In this example,
F 2 g R3 3 4 gR
Pave ,
A R2 2 3
which is the local pressure at the centroid of the semicircle,
4R
y x 0.
3

5.5 BUOYANCY
The phenomenon of buoyancy was first considered in the third century B.C. by
Archimedes of Syracuse, who formulated what is known as Archimedes’ principle: a
body immersed or floating in a fluid is acted upon by an upward buoyant force which
is equal to the weight of the fluid displaced and which acts through the center of grav-
ity of the displaced volume. Figure 5.10 shows a solid object of density s immersed
in a fluid of density f. Consider the cylindrical volume element of cross-sectional
Fluid Statics 245

FIGURE 5.10 Illustration of Archimedes’ principle.

area dA located in the object. The net differential upward force dFup exerted on the
element by the fluid is given by

dFup P2 dA P1 dA.

But this is also the body force of an identical volume element of fluid given by

f g z2 z1 dA.

Thus, integrating over the entire surface of a solid object gives the total upward force
exerted on the object by the fluid as

Fup f g ,

where is the volume of the object.

Example 5.3
Calculation of Density from Buoyancy Force Measurement
In air an object weighs wa = 3 N and, when immersed in water, it weighs ww = 1.5 N.
The density of water is f = 1000 kg/m3. The force balance is illustrated in Figure 5.11.

Find: What is the density of the object?

Solution: The decrease in the weight of the object when immersed in water
equals the buoyancy force, which is the weight of displaced water,

W Wa – Ww buoyancy = weight of displaced water Wdw f g objj .

246 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 5.11 Difference between the weights obtained when an object is (a) weighed in air
and (b) weighed in water.

Therefore,

buoyancy
obj ,
fg

but

s buoyancy
wa s g obj ,
f

so,

f wa 1000 kg m 3 3 N
s 2000 kg / m 3 .
buoyancy 3N 1.5 N

In this example the influence of the buoyancy force of air on the measurement of the
weight of the object in air is ignored.

5.6 SUMMARY
The pressure exerted by a fluid on the walls of its container arises from the collisions
of the individual atoms or molecules with the walls, and the force exerted on the wall
is equal to the rate of change of momentum of the atoms or molecules due to the col-
lisions. The pressure at any point in a fluid is the same in all directions. The influence
of gravity is such that the pressure at any plane in a static fluid, normal to the gravita-
tional field, arises from the weight of the fluid above the plane, and thus the pressure
at a point in a static fluid varies with position in the direction of the gravitational field.
In incompressible fluids, the pressure is a linear function of position in the direc-
tion of the gravitational field, and thus in liquids, the pressure increases linearly with
Fluid Statics 247

depth below the free surface of the liquid. The average pressure exerted by a liquid
on a plate that is submerged vertically in the liquid is equal to the local value exerted
at the centroid of the plate.
The phenomenon of buoyancy is explained by Archimedes’ principle, which
states that a body immersed in or floating in a fluid is acted upon by an upward force
equal to the weight of the fluid displaced by the body, which acts through the center
of gravity of the displaced volume.

5.7 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

5.1 A 2 m × 3 m rectangular floodgate is placed vertically in water with the

2-m side at the free surface of the water. Calculate the force exerted by
the water on one side of the floodgate. The density of the water is 997
kg/m 3.
5.2 The center of a circular floodgate of radius 0.5 m is located in a dam at a
depth of 2 m beneath the free surface of the water contained by the dam.
Calculate the force exerted by the water on the floodgate. The density of
water is 997 kg/m3.
5.3 A hemispherical bowl of radius 0.5 m is filled with water of density 997 kg/m3.
Calculate the force exerted on the bowl.
5.4 A submariner is escaping from a damaged submarine at a depth of 50 m
beneath the surface of the ocean. To prevent experiencing the “bends,” the
rate of decrease in pressure exerted on his body must not exceed 3,400
Pa/s. What is the maximum rate at which he can ascend without experienc-
ing the bends? The density of salt water is 1,030 kg/m3.
5.5 When immersed in water of density 997 kg/m3 a solid sphere is balanced
with a mass of 10 g, and when immersed in an oil of density 800 kg/m3 it
balances with a mass of 11 g. Calculate (a) the volume and (b) the density
of the sphere.
5.6 When placed in water of density 997 kg/m3, a hollow sphere of inner radius
10 cm and outer radius 11 cm floats with half of its volume below the level
of the free surface of the water. (a) Calculate the density of the mate-
rial from which the sphere is made. (b) When placed in an oil the sphere
floats with 60% of its volume below the level of the free surface of the oil.
Calculate the density of the oil.
5.7 A U-tube manometer contains mercury of density 13,600 kg/m3, and oil.
(a) When both ends of the tube are open to the atmosphere, h1 = 17 cm and
h = 1 cm. Calculate the density of the oil. (b) The open end of the left
leg is sealed and is pressurized to a gauge pressure, P, which causes the
level of the free surface of the oil in the left leg to fall 0.5 cm. Calculate
the value of P.
5.8 The altimeter in an airplane records a pressure of 70 kPa when the absolute
pressure at sea level is 101 kPa and the air temperature is 288 K. Calculate
248 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 5.12 Balance for measurements of sphere density in Problem 5.5.

FIGURE 5.13 The U-tube manometer in Problem 5.7.

FIGURE 5.14 Hydraulic press in Problem 5.9.

the altitude of the airplane, assuming that the air temperature is not a func-
tion of altitude.
5.9 In the hydraulic press shown, what mass can be raised by the ram when a
mass of 1 kg is placed on the plunger? (See Figure 5.14.)
Fluid Statics 249

5.10 Calculate the gauge pressure in the oil at point D in Figure 5.15. The mano-
metric fluids are mercury ( Hg = 13,600 kg/m3) and water ( w = 997 kg/m3).
The density of the oil is 800 kg/m3.

FIGURE 5.15 Two-fluid manometer in Problem 5.10.

FIGURE 5.16 Electroslag remelting with atomization of the liquid metal in Problem 5.11.
250 An Introduction to Transport Phenomena in Materials Engineering

5.11 A metal electrode is melted in a hot slag and liquid metal droplets fall
through the slag and form a reservoir. The metal then exits this reservoir
through a small nozzle at the bottom, and this metal stream is atomized
into a spray of droplets. (The chamber below is in vacuum, and the spray
builds up an ingot with a very refined, but somewhat porous, microstruc-
ture.) Here we will find only the pressures in the system; the metal flow
rate and its dependencies will be considered inChapter 6. (a) Calculate the
pressure at the free surface on the top of the slag. (b) Calculate the pressure
at the interface of the slag and metal. (c) Calculate the pressure at the bot-
tom of the liquid metal pool. Use metal = 8,900 kg/m3, slag = 2000 kg/m3,
L M = 0.3 m, LS = 0.2 m, and Patm = 101 kPa. (This problem is based on
information from Carter and Jones [1]).

NOTE
1. Weather reports give barometric pressure in a number, usually without units. In the USA,
the number is around 29 and elsewhere around 760. These reports assume pressure is in
units of the height of a column of mercury (inches in the USA and millimeters in most
other countries).

REFERENCE
1. Carter, W. T., Jr., and R. M. F. Jones, “Nucleated casting for the production of large
superalloy ingots,” JOM, v. 55 (4), pp. 52–57, 2005.
 6 Mechanical Energy
Balance in Fluid Flow

6.1 INTRODUCTION
A fluid moving in a conduit has, by virtue of its mass and motion, a kinetic energy
that may change along the flow direction. If the fluid flow is not horizontal, the
potential energy of the fluid also varies in the flow direction, which change may
work to impel or retard motion. In all cases of real fluids, the friction arising from the
shear stress exerted on the fluid by the containing conduit causes a transformation of
mechanical to thermal energy. All of the mechanical energy added to a flowing fluid
to maintain motion is obtained from the work done on the fluid by some external
agency, and the rate at which this work is provided is the power required to sustain
the flow. The most common power supplies are fans and pumps, which effect an
increase in the pressure of the fluid at some point in the pipeline; the flows can also
be driven by potential energy. Whatever drives or restrains the flow, we are concerned
with how a given fluid responds in a given geometrical configuration. Understanding
that behavior begins with a brief discussion of some flow characteristics, followed by
examining an energy balance over the entire flow system. We derive that mechanical
energy balance and apply it to many steady and transient flows.

6.2 LAMINAR AND TURBULENT FLOWS

Up to this point in the text, we have examined stationary fluid systems, but now we
will discuss moving flows. The nature of these fluid flows is dependent on the magni-
tude of the forces driving and restraining them, which determine the velocity fields.
For the flow of a fluid of given physical properties in a given geometry, a critical
velocity range exists below which the flow is laminar and above which the flow is
turbulent. The two regimes are illustrated in Figure 6.1, which shows a thin stream
of dye being introduced into a liquid flowing in a duct. At lower flow rates, shown in
Figure 6.1 (a), the dye moves with the fluid in a stable line parallel to the axis of the
tube. This stable behavior indicates that, at this low velocity, flow can be regarded
as being the unidirectional movement of lamellae of fluid sliding over one another,
with no macroscopic mixing or intermingling of the fluid in the vertical direction.
This type of flow is known as laminar flow. If the flow velocity is increased beyond
the critical range, then, as in Figure 6.1 (b), the flow becomes turbulent and the
emerging dye stream is rapidly broken up and mixed with the fluid in the direction
perpendicular to the dominant velocity. In the critical range, a transition from laminar
to turbulent flow occurs.
Experiments similar to those illustrated in Figure 6.1 were published in 1883 by
Osborne Reynolds [1], who established a criterion for the transition from laminar

DOI: 10.1201/9781003104278-6 251

252 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.1 Schematic of dye injected into (a) laminar flow and (b) turbulent flow in a duct,
as seen in a series of experiments by Reynolds [1].

to turbulent flow in terms of a dimensionless quantity, known now as the Reynolds

number:

U Lc
Re . (6.1)

The characteristic length (Lc) in this flow is the height of the duct. The magnitude
of the average velocity of the fluid, U, is measured as the volume flow rate divided
by the cross-sectional area of the duct, and is the fluid density and μ its dynamic
viscosity defined by Newton’s law of viscosity in Section 1.3.1. For fluid flow in a
smooth circular pipe of diameter D, the transition from laminar to turbulent flow
begins at a Reynolds number of approximately 2100, using the pipe diameter as the
characteristic length (Lc = D). From Eq. (6.1), we see that (i) increasing the diame-
ter of the tube decreases the average velocity at which the transition occurs, or (ii)
increasing the flow velocity decreases the maximum pipe diameter in which laminar
flow persists. For flow of a fluid at a given velocity in a pipe of given diameter, the
velocity at which the transition occurs is proportional to the viscosity of the fluid and
is inversely proportional to its density.
Mechanical Energy Balance in Fluid Flow 253

For now, we leave our description of turbulent flow as being more chaotic and
disordered than laminar, with significantly more mixing, and we emphasize that for
any specific geometry there is a Reynolds number range in which the flow regime
changes from laminar (at low Re) to turbulent (at high Re). In Chapter 9, characteris-
tics of turbulent flow and mechanisms of transition will be further discussed.

6.3 BERNOULLI’S EQUATION

Figure 6.2 shows the flow of an incompressible fluid through a section of pipe
between locations 1 and 2. We assume initially that the flow is inviscid, i.e., that
viscous effects are not important to the fluid behavior. This assumption allows the
simplification of characterizing the velocity in the tube as the average value across
the tube cross-section, , so we can neglect local variation of fluid velocity in that
area. The fluid enters at 1 with the average flow velocity 1 and leaves at 2 with 2,
and the static pressures in the fluid at planes 1 and 2 are P1 and P2.
The fluid entering the section at plane 1 is being pushed by the fluid behind it, and
the work done on a unit mass of the fluid entering is P1/ . Similarly, the fluid leaving
the section at plane 2 is pushing on the fluid in front of it (downstream of section 1–2)
and the work done by a unit mass of the fluid leaving is P2/ ). The difference between
the work done on the fluid at section 1 and the work done by the fluid at section 2 is
the net flow work done on the fluid.
The kinetic energy of the fluid entering at plane 1 is
1
EK m 12 ,
2
and the kinetic energy per unit mass of fluid entering is 12 2. Similarly, the kinetic
energy per unit mass of fluid leaving at plane 2 is 22 2. The fluid potential energy
at plane 1 is

Ep m g z1 ,

FIGURE 6.2 Fluid flowing through a section of pipe from location 1 to 2.

254 An Introduction to Transport Phenomena in Materials Engineering

and the potential energy per unit mass of fluid entering is gz1. Similarly, the potential
energy per unit mass of fluid leaving at plane 2 is gz2.
Conservation of mechanical energy between locations 1 and 2 requires that

net flow work done on the fluid = change in kinetic energgy

+ change in potential energy ,

that is,

P P 1 2 2
+ 2 1 g z2 z1 0.
2 1
2
net work change in change in (6.2)
on fluid kinetic enerrgy potential energy

Equation (6.2) is Bernoulli’s equation, a mechanical energy balance in frictionless

flow through a conduit with changes in pressure, kinetic energy, and potential energy.
To apply this equation to a flow in a conduit, first one must identify the starting point
upstream (i) and the end point downstream (ii) and then be able to trace a single,
continuous path through the system between these two points. Rearranging Eq. (6.2),
we can see that the sum of the work done on the fluid and the changes in kinetic and
potential energies do not vary along that continuous path:

P 1 2 P 1 2
2 + g z2 = + 1 + g z1 CB , (6.3)
2
2 1
2

where CB is a constant.
In this development of Bernoulli’s equation, we have neglected several possible
physical phenomena. The density is assumed to be not a function of pressure, which
is valid for liquids or if the change in the pressure in a gas is small enough. (This
approximation is reasonable in most applications in materials processing; supersonic
flow in an oxygen lance during oxygen steelmaking is one exception.) A steady flow
has been assumed, so no work must be done (or recovered) from an acceleration (or
deceleration) of the fluid. No work or heat has been directly added to the system,
except due to the imposed pressure difference between points 1 and 2, and we have
neglected any friction in the conduit.
Using the average velocity in the pipe, , we may obtain the volume flow rate (Q)
in the pipe,

Q A, (6.4)

where A is the cross-sectional area perpendicular to the flow direction and Q has
units of m3/s. The mass flow rate in the pipe, m, is the product of the volume flow
rate and the fluid density,

m Q A 1
A 2, (6.5)
Mechanical Energy Balance in Fluid Flow 255

where conservation of mass requires m in and out of the pipe be the same, assuming
no accumulation of fluid.
In any real system, the second law of thermodynamics requires that some of the
mechanical energy be degraded to thermal energy due to the friction in the moving
fluid. The existence of this effect means that the Bernoulli equation, Eqs. (6.2) and
(6.3), is not strictly valid, i.e., CB in Eq. (6.3) is not a constant. Eq. (6.2) must be
modified to account for the dissipation of energy caused by the viscous drag exerted
on the flowing fluid by the tube wall. If, between planes 1 and 2, this quantity, which
is termed the friction loss, is ef, the energy balance becomes

net flow work increase in increase in frictio

i n
+ + 0.
done on the fluid kinetiic energy potential energy loss

Furthermore, if between locations 1 and 2, work w is done on unit mass of the fluid,
the mechanical energy balance becomes

flow work w ke pe ef

or
2 2
P P 2 1
= g z2 z1 w ef . (6.6)
1 2
2 2

Eq. (6.6) is called the modified Bernoulli equation.

6.4 FRICTION LOSSES

To model the contribution of wall friction in a straight length of pipe to the fric-
tion loss term, ef, first consider a length of horizontal pipe, z z , with a uniform
1 2

cross-sectional area, A 1
A , in which the flow velocity is uniform 1
2 2 , and
there is no added work w 0 . For this situation, Eq. (6.6) is reduced to

P1 P2
ef , (6.7)

Illustrating that a pressure drop in the flowing fluid is required to supply the mechan-
ical energy necessary to overcome the viscous drag exerted on the fluid by the pipe
wall. In Eq. (6.7), P1 > P2, and hence ef > 0. The decrease in the normal force exerted
on the fluid over the length of pipe, L, is equal to the shear force exerted on the fluid
at the pipe wall:

normal force = P1 P2 R2 0 2 R L = shear force,

and Eq. (6.7) becomes

2 o L
ef . (6.8)
R
256 An Introduction to Transport Phenomena in Materials Engineering

If we define the nondimensional shear stress, or Darcy friction factor, f, as

8 0 D P
f , (6.9)
2
L 1 2
2

and substitute it into Eq. (6.8), ef can be written as

1 L 2
ef f , (6.10)
2 D

which is a general expression for calculating the friction losses. The friction factor
for laminar flow in a tube can be found from the laminar solutions for the average
velocity, , and wall shear, o, to be later developed in Section 8.4 (if the flow is
horizontal). In that case,

64 64
f . (6.11)
D Re D

Substitution of this friction factor into Eq. (6.10) gives

32 L
ef ,
D2

which shows that the friction loss per unit mass of fluid per unit length of flow is
proportional to the viscosity and the velocity, and is inversely proportional to the
cross-sectional area of flow and the density. This solution is valid in the laminar regime,
which holds for flows with ReD < 2100. For turbulent flow between ReD ≈ 5 × 103
and 105, the friction factor in a smooth-walled pipe is given by a correlation of exper-
imental data

f 0.316 Re D1/ 4 . (6.12)

The Darcy friction factor, f, is defined in Eq. (6.9) and is frequently used in the liter-
ature. However, the Fanning friction factor, fF, is also used and is defined as

0 1 D P
fF . (6.13)
1 2 2 L 2

The relationship between the two definitions of a friction factor is a factor of

four: f = 4fF. In this text, we use by default the Darcy friction factor, f, defined
in Eq. (6.9).
Mechanical Energy Balance in Fluid Flow 257

Example 6.1
Calculation of Pressure Drop in Pipe
Water at 300 K is forced through 300 m of smooth pipe with an inside diameter of 0.05 m
at a volume flow rate of 0.0015 m3/s. Use = 997 kg/m3 and μ = 8.57 × 10–4 kg/ms.

Find: Pressure drop needed to maintain flow through pipe.

Solution: First, find the average velocity from the volume flow rate, Q, in Eq. (6.4):

Q 0.0015m 3 /s
2 2
0.764 m/s,
D/2 0.05 m / 2

so we can calculate the Reynolds number.

D 997 kg/m 3 0.764 m/s 0.05m

Re D 4.44 10
8.57 10 4 kg/ms

At this value of the Reynolds number, the flow is turbulent (ReD > 2100) and the
friction factor can be calculated from the empirical correlation in Eq. (6.12):
1/ 4
f 0.316 44, 400 0.0218.

Rearrangement of Eq. (6.9) gives an expression for the pressure drop:

1 2 L 1 2 300 m
P f 0.0218 997 kg/m 3 0.764 m/s 3.80 10 4 Pa
2 D 2 0.05m

which is roughly a third of a standard atmosphere (Patm = 1.01 × 105 Pa).

Example 6.1 was straightforward because enough information was available to cal-
culate the Reynolds number directly. If a friction factor cannot be found directly from
the information given in a problem statement, the choice of correct expression for it
must be found by trial and error.

Example 6.2
Calculation of Flow Rate in Pipe for Given Pressure Drop
Water at 300 K is forced through a smooth pipe with an inside diameter of 0.07 m by a
pressure gradient of P/L = 125 Pa/m. Use = 997 kg/m3 and μ = 8.57 × 10–4 kg/ms.

Find: (a) Average velocity, (b) volume flow rate, and (c) mass flow rate of
stream.
Solution: (a) Because we do not know the flow rate a priori, we cannot immedi-
ately calculate the average velocity or Reynolds number. Instead, we assume
a flow regime, which in turn gives us a relationship between the friction
258 An Introduction to Transport Phenomena in Materials Engineering

factor (based on pressure gradient) and Reynolds number (based on the aver-
age velocity). If our assumption is correct, then we have the flow rates; if not,
we try another flow regime.

Assumption 1: Laminar Flow (ReD < 2100)

For laminar flow, the combination of Eqs. (6.9) and (6.11) gives

D P D P D2 1 64
f 2
L 1 2 L 2
Re 2D Re D
2
or

3
1 P D3 1 997 kg/m 3 0.07 m
Re D 2
125 Pa/m 2
1.82 106.
32 L 32 8.57 100 4 kg/ms

Because ReD > 2100 when we assumed the opposite, the flow is definitely turbulent
and we can not use laminar flow friction factor.
Assumption 2: Turbulent Flow (ReD > 5 × 103)
For this flow regime, to find an expression for the Reynolds number, we combine
Eqs. (6.9) and (6.12).

D P D P D2 1 0.316
f 2
L 1 2 L 2
Re 2D Re D1/ 4
2
Manipulation of this relationship by isolating the Reynolds number gives
4/7
4/7 3
2 P D3 2 997 kg/m 3 0.07 m
Re D 2
125 Pa/m 2
0.316 L 0.316 8.57 10 4 kg/ms
7.85 10 4.

This Reynolds number is within the range over which the friction factor correlation
is applicable (5 × 103 > ReD > 105), so the assumption is self-consistent. Using this
value of ReD, we can find the average velocity in the pipe,

Re D 7.85 10 4 8.57 10 4 kg/ms

v 0.964 m/s
m ,
D 997 kg/m 3 0.07 m

as well as the volume and mass flow rates:

Q A ( D / 2 )2 0.964 m/s (0.07 m / 2)2 3.71 10 3 m 3 /s

m Q 997 kg/m 3 3.71 10 3 m 3 /s 3.70 kg/s.

Mechanical Energy Balance in Fluid Flow 259

The foregoing discussion has been limited to consideration of fluid flow in smooth-
walled tubes. Any roughness of the tube wall increases the friction factor and hence
decreases the flow rate obtained for a given value of P/L. The influence of surface
roughness on the flow is correlated by the relative roughness, /D, the quotient of a
characteristic height of protuberances on the surface, , and the diameter of the tube,
D. Typically reported values of are listed in Table 6.1. Measurements of rough-
ness’ effect on flow were carried out originally by Nikurasde [2] and Colebrook [3],
controlling roughness by attaching sand grains of a known diameter and a tight size
distribution to the inside of smooth pipes and using that diameter for . How these
results translate to pipes in service, with much more complicated inner surfaces, is
less clear. In the absence of clear guidance from the literature, most practitioners
assume represents some height of protuberances on the pipe wall, but that height
does not seem to be found by standard roughness measurement techniques (e.g., a
profilometer). Also, the absolute roughness values in Table 6.1 are only approximate
values for new pipes. As the pipes are used, they inevitably begin to foul due to debris
gradually accumulating on the walls. Depending on the fluid and the pipe materials,
significant corrosion also may occur, further roughening the conduit interior.
Extending our scope to more vigorous flows than Eq. (6.12) and to pipes with
rough surfaces, we can see the dependence of the friction factor on ReD and relative
roughness, /D, in Figure 6.3, which is known now as the Moody diagram [5]. This
diagram is based on data presented by Nikurasde [2] and Colebrook [3], the latter of
whom curve-fit these data to find an implicit correlation for the friction factor:
2
1 2.51
f 2 log . (6.14)
3.7 D Re D f

The data fit by this correlation are within ±3–5% of Eq. (6.14). Moody developed
this diagram to simplify finding f; the plot acts as an explicit replacement for the implicit
Eq. (6.14), eliminating the need for iterative root finding for the friction factor. How-
ever, using such a diagram is not possible when calculating with a computer, so several
researchers have reported different correlations that are explicit in f. For example,
2
0.9
1 6.97
f 2 log [6] (6.15)
3.7 D Re D

TABLE 6.1
Approximate Absolute Roughness Values for
Common New Surfaces [4]
Material (mm)
drawn tubing 0.00013
steel 0.0038
cast iron 0.022
concrete 0.025–0.25
plastic, glass 0.0 (effectively smooth)
260 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.3 Moody diagram for Darcy friction factor dependence on Reynolds number
and relative roughness.

and
2
1.108 0.966
1 5.472
f 1.805 log [7]. (6.16)
4.267 D Re D

Examination of these correlations indicates that, at sufficiently high Reynolds num-

bers, the roughness effect in the first term inside the log functions dominates the
second term. In that “fully turbulent” regime, the friction factor is only a function
of the surface roughness. (Of course, in a smooth pipe, /D 0, that regime never
develops and the friction factor is always Reynolds number dependent.) All of the
correlations are valid in the range 5 × 103 < ReD < 108 and 10–6 < /D < 10–2 and agree
with each other well within the uncertainty of the original experimental data. All in
all, with the scatter in the original data and the lack of clarity in the definition of ,
the calculation of f has significant uncertainty, estimated in various textbooks (e.g.,
[8, 9]) as around ±10%.
The foregoing discussion is based on the assumption of a circular conduit
cross-section, while other cross-section shapes (especially rectangular) are not
uncommon. In these latter cases, we can use the previous friction factor correla-
tions with some alteration and within certain limits. To approximate the response
in a noncircular channel, we replace the round pipe diameter with the hydraulic
diameter,

DH 4 A / Pw . (6.17)

The area (A) is of the conduit cross-section normal to the flow direction, and Pw is
the wetted perimeter, the length of the conduit perimeter in contact with the moving
Mechanical Energy Balance in Fluid Flow 261

TABLE 6.2
Coefficients for Laminar Friction Factors for Noncircular Conduits [9]
Shape DH Parameters C
values Eq. (6.18)
ANNULUS D2 – D1 D1/D2
D1 = inner diameter 0.0001 71.8
D2 = outer diameter 0.01 80.1
0.1 89.4
0.6 95.6
1.0 96.0
RECTANGLE 2ab a/b
a, b = lengths of sides a b 0 96.0
0.05 89.9
0.25 72.9
0.50 62.2
0.75 57.9
1.00 56.9

fluid. In turbulent flows, more of the pressure drop is due to extensive mixing at
different length scales through the cross-section; this mixing is absent from laminar
flow, in which most of the frictional losses are near the wall. Therefore, the simple
replacement of DH for D in existing circular pipe correlations works better with
turbulent flows (Re DH 5 103 ) than in the laminar regime. It works best if the
conduit shape is somewhat close (within a factor of four or so) to an aspect ratio of
one. For laminar flow, the friction factor can be found in the functional form of Eq.
(6.11) as

f C Re DH , (6.18)

where C values are found from the exact solutions of the velocity field in the differ-
ent conduits. Many values for C are presented in [9], from which a few are shown in
Table 6.2.

6.5 INFLUENCE OF BENDS, FITTINGS, AND

CHANGES IN PIPE RADIUS
Eq. (6.8) shows the frictional losses in a pipe flow as a function of only the shear
introduced at the conduit walls. However, in typical piping systems there are other
sources of flow disruptions acting as sinks of mechanical energy. The occurrence of
bends, valves, and abrupt changes of radius in a pipe system increases the extent to
which the energy of the fluid is dissipated and hence increases the magnitude of the
friction loss. The increase in the friction loss caused by the presence of a bend in a
pipe is presented in terms of the increase in the length of a straight pipe, which would
be required to produce the same increase in friction loss. Thus, any bend or fitting has
262 An Introduction to Transport Phenomena in Materials Engineering

an equivalent length, Le, of straight pipe. For example, the presence of a 45° elbow
in a cylindrical pipe is equivalent to increasing the length of straight pipe by 16 pipe
diameters, and thus for a 45° elbow,
Le
16.
D
While the pressure drop over a length of pipe is proportional to 2 , experiments
show the pressure drop through different devices goes like P ~ n , where 1.8 <
n < 2.1 [4]. For simplicity’s sake, we will treat n = 2 and Le/D is a constant for a
given flow disruption. These equivalent lengths have been measured over a range
of ReD > 1000. To apply to slower laminar flows (ReD < 1000), the simple linear
relation

Le Re D Le
(6.19)
D laminar
1000 D

is used.
The values of Le/D for various fittings are listed in Table 6.3, and the friction loss
for a pipe system containing such fittings is calculated as

1 L 1 Le 2
ef f v2 f v . (6.20)
2 D pipe 2 D i
wall friction friction loss through fittings

As noted in [10], these equivalent length values have an uncertainty up to ±25%.

Given that level of uncertainty, estimates made with these models are approximate;
measurements of the pressure drop through a system as a function of the flow rate
must be made to obtain more accurate system performance curves. However, calcu-
lations done with the method presented here are useful for a first pass design to show
trends in system behavior.
Sudden changes in the radius of pipe through which the fluid is flowing, such as
sudden expansions or contractions, are dealt with in terms of the friction loss factor,
K, using which the consequent friction loss is calculated as

1
ef K v 2. (6.21)
2

TABLE 6.3
Values for Le/D for Various Fittings [4]
Fitting Le/D Fitting Le/D
45° standard elbow 16 Gate valve, fully open 13–17
90° elbow, standard radius 30 Gate valve, ¼ closed 35–50
90° elbow, long radius 20 Gate valve, ½ closed 160–200
90° square elbow 57 Gate valve, ¾ closed 900–1200
180° close return bend 50 Globe valve, fully open 340
Mechanical Energy Balance in Fluid Flow 263

For a sudden contraction,

As
K 0.5 1 (6.22)
AL

and for a sudden expansion

2
As
K 1 , (6.23)
AL

where As and AL are the cross-sectional areas of the smaller and larger diameter pipes,
respectively. Eqs. (6.22) and (6.23) apply to turbulent flow, and the velocity used in
Eq. (6.21) is the average value in the smaller diameter pipe.

6.6 STEADY-STATE APPLICATIONS OF THE MODIFIED

BERNOULLI EQUATION
Now that we have models of how frictional losses in conduits bleed mechanical
energy from a moving fluid, we can examine the steady-state behavior of such a fluid
as it flows through a conduit. First, we explore an example in which fluid drains from
a vessel while being replenished, so that the flow rate and the amount of liquid in the
vessel is steady.

Example 6.3
Drainage of a Replenished Ladle
Figure 6.4 shows fluid draining through an orifice in the base of a ladle. The fluid is
being poured into the ladle at the same rate it drains, such that the depth of the fluid,
ho, remains constant. The diameters of the ladle and the orifice are, respectively,
D and d. Assume the liquid is an aluminum alloy with μ = 1.3 × 10–3 kg/ms and
= 2500 kg/m3 and the diameters are D = 2 m and d = 0.1 m. The depth of the vessel
is maintained at ho = 1 m.

Find: What is the mass flow rate of the aluminum into and out of the ladle?
Solution: If plane 1 is located at the liquid-free surface in the vessel and plane
2 is located at the level of the exit of the orifice, the modified Bernoulli
equation in Eq. (6.6) becomes

2 2
P P 2 1
g z2 z1 ef 0.
2 1
2 2

The aluminum at the top surface in the ladle (P1) is at the pressure of the surround-
ing air, Patm. The fluid requires some pressure to push it through the bottom opening
at level 2, but free expansion of the metal jet once it exits the orifice gives P2 = Patm
264 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.4 Fluid draining through an opening in the base of a replenished vessel.

also. From mass balance considerations, the mass flow rate through the orifice equals
m through the vessel:

m 1 D2 4 2 d2 4

or
2
d
1 2 . (6.24)
D

If d << D, as in this example, then 1 2 . The friction loss is from the friction on
the ladle walls and the orifice:

1 ho 2 1
ef f 1 K exit 22 ,
2 D 2
ladle wall orifice

where, from Eq. (6.22), Kexit for a sharp cornered, sudden contraction is 0.5. With
1 2 , the kinetic energy at plane 1 can be neglected in comparison, in which case
the modified Bernoulli equation becomes
2
1
2
g h2 h1 K exit 22 0
2 2
increase decrease frriction loss (6.25)
in KE in PE
Mechanical Energy Balance in Fluid Flow 265

or
2
2
1 K exit g ho 0,
2

so the exit velocity from the orifice is

2 g ho
2 , (6.26)
1 K exit

With rounder corners or larger diameter, the loss coefficient for the orifice decreases
and eventually becomes negligible. At that point,

2 2 g ho ,

the maximum possible value of 2, where the decrease in potential energy feeds
only the increase in kinetic energy from 1 to 2. For this ladle geometry, with friction
in the orifice,

2 g ho 2 9.81m/s2 1m
2 3.62 m/s.
1 K exit 1 0.5

It is interesting to note that this velocity is not a function of the fluid properties at all,
as long as the friction loss on the ladle walls can be neglected. Also, in this configura-
tion, Eq. (6.25) shows that the flow is driven by the potential energy term, which is
used at steady-state to increase the kinetic energy and make up for the frictional loss
in the orifice. The steady mass flow rate of aluminum alloy from this ladle is
2
m 2 d2 4 2500 kg/m 3 3.62 m/s 0.1 m 4 71.0 kg/s.

Example 6.4
Pumping Water from a Lower to a Higher Reservoir
Figure 6.5 shows a piping system with a pump to raise water from one tank to another
several meters higher. The volume flow rate of the water through a smooth pipe of
diameter D = 0.1 m is Q = 6 × 10−3 m3/s. The density and dynamic viscosity of water
are, respectively, = 997 kg/m3 and μ = 8.57 × 10–4 Pa∙s. The pipes are treated as
smooth.

Find: The power required by the pump to move the water.

Solution: From Eq. (6.6),

P2 P1 1 2 2
2 1 g z2 z1 w ef 0.
2
266 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.5 Piping system for Example 6.4 (not to scale).

Levels 1 and 2 are taken as being the top surfaces of the water in the lower and
higher reservoirs, respectively. Sketching a line connecting these two points indicates
that between 1 and 2 there are frictional contributions due to the length of pipes and
several bends, as well as the work addition from the pump. As both surfaces are in
contact with the atmosphere, P1 = P2, and as both surfaces are virtually stationary,
1 2
0 . Eq. (6.6) thus reduces to

w g z2 z1 e f .
pump increase friction (6.27)
work in PE loss

Sketching the flow path from 1 to 2 through the pipe (with several bends) and the
pump, the increase in the potential energy of the water per unit mass is

g z2 z1 9.81 m s2 1.5 m 35 m 15m 211 m 2 / s2 ,

and the friction loss in the pipe due to wall shear, bends, and changes in area is
obtained from Eqs. (6.20) and (6.21) as

1 L Le 2 1 2
ef f 3 K entry K exxit .
2 D pipe D elbow
2

In this case, we treat the average velocity as uniform through the system between
points 1 and 2.
Evaluation of f requires calculation of Re, which, in turn, requires knowledge of
the average flow velocity in the pipe, :

Q Q 4 6 10 3 m 3 s
2 2
0.764 m / s.
A D2 0.1m
Mechanical Energy Balance in Fluid Flow 267

Therefore,

D 997 kg m 3 0.764 m s 0.1m

Re 8.89 10 4 ,
8.57 10 4 kg m s

and thus the flow is turbulent. We can obtain the friction factor from Eq. (6.15):
0.9 2 2
0.9
6.97 6.97
f 2 log 2 log 1.83 10 2.
Re D 88,900

At both the entrance and the exit of the pipe, the reservoirs have much larger cross-
sectional areas than the pipe, so AL >> As for both ends of the pipe. Thus, from Eqs.
(6.22) and (6.23), respectively, K = 0.5 at the entrance (a contraction of area) and
K = 1 at the exit (an expansion):

1 1 2 2
K entry K exit v 2 1 0.5 0.764 m s 0.44 m s .
2 2

The friction loss is then

1 1.5 m +100 m 35m 25 m 15 m

ef = 0.0183 3 31
2 0.1 m
2 2 2
m m m
0.764
6 0.44 10.4 .
s s s

Substitution into Eq. (6.27) gives

2 2 2
w 211 m / s 10.4 m / s 221 m / s 221 J/kg,

which, being a positive quantity, indicates that work is done on the fluid by the pump.
The mass flow rate is

m Q 997 kg m 3 6 10 3 m 3 s 5.98 kg/s,

and thus the power required of the pump, W, is

W wm 221J kg 5.98 kg s 1324 W.

Of this power, 211/221, or 95.3%, is needed to pump the water to a higher elevation,
increasing the potential energy of the water by the required amount at the required
rate, and the remaining 4.7% is dissipated as friction loss. The contribution of the
frictional losses at the entrance and exit of the pipe is negligible.

Just as work must be done on a fluid (by means of a pump) to increase its potential
energy, useful work can be obtained from a fluid (using a turbine) that is flowing
downward under the influence of gravity. In such a case, w is a negative quantity.
268 An Introduction to Transport Phenomena in Materials Engineering

Example 6.5
Using Potential Energy of Water to Drive a Turbine
Water is fed from a reservoir to a turbine through a vertical concrete pipe of diameter
D = 0.5 m and average roughness = 5 × 10–4 m at a mass flow rate of m = 1000 kg/s.
The surface of the reservoir is L = 50 m above the entrance of the turbine, and the
pressure at the reservoir surface and at the turbine exit are both P = 105 Pa. The den-
sity and viscosity of the water are, respectively, = 997 kg/m3 and μ = 8.57 × 10–4
Pa∙s.

Find: Power generated by the turbine.

Solution: With m = 1000 kg/s and D = 0.5 m, the linear flow velocity is

m 1000 kg s
v 2 2
5.11 m / s
D/2 0.25m 997 kg m 3

and thus the Reynolds number is

vD 997 kg s 5.11m s 0.5m

Re D -4
2.98 106 ,
8.57 10 kg m s

which indicates that the flow is turbulent. The friction factor is obtained from Eq.
(6.15) as

2
0.9
1 6.9
f 2 log
3.7 D Re D
2
0.9
1 0.0005m 6.9
2 log ,
3.7 0.5 m 2.98 106

0.9
which gives f = 0.0198. (Note that 3.7 D 6.9 Re D in this case, showing
that we are in the fully turbulent regime where f is only a very weak function of Re D .)
For the path between the reservoir surface and the turbine exit,

1 L 2 1 50 m 2 2
ef f v 0.0198 5.11 m s 25.8 m s .
2 D 2 0.5m

For a constant and uniform m, Eq. (6.5) shows that v1 v2 and Eq. (6.6) is

g h2 h1 w ef 0

or
2
w 9.81 m s2 50 m 25.8 m s 465 J kg.
Mechanical Energy Balance in Fluid Flow 269

The negative sign indicates that the water is transferring work to the turbine. The
power extracted from the water is thus

power wm 465 J kg 1000 kg s 465 kW.

Note that only 1.3% of the potential energy is lost to frictional losses in the pipe.
The friction loss in the turbine is expressed in terms of the efficiency with which
the energy extracted from the water is converted to shaft energy. The turbine effi-
ciency, , is defined as

shaft energy delivered by turbine

.
o water
energy extracted from

Thus, if the turbine has an efficiency of 90%, the shaft work delivered would be

0.9 465 kW 418 kW.

6.7 CONCEPT OF HYDROSTATIC HEAD

The total energy of a fluid at any point in a flow system is

Pm 1 2
mv mgz.
2
flow kinetic potential (6.28)
energy energy energy
Dividing Eq. (6.28) by the mass of the fluid multiplied by gravity, mg, gives

P 1 v2
H z. (6.29)
g 2 g

Eq. (6.29) has the units of length and defines the hydrostatic head, H, of the flow at
a point of interest. The first term in Eq. (6.29) is the pressure head, the second term the
velocity head, and the third term the elevation head. In considering flow between plane
1 and plane 2 in a flow system, the loss of head, – HL, is obtained from Eq. (6.29) as

P2 P1 1 v22 v12
HL H 2 H1 z2 z1 . (6.30)
g 2 g

Eq. (6.29) shows that in ideal flow with no frictional loss or work addition, H is a
constant along the flow direction, so the head loss is zero (HL = 0). However, com-
parison of Eq. (6.30) with Eq. (6.6), for a flow system in which w is zero between
planes 1 and 2, shows that

ef
HL ; (6.31)
g
270 An Introduction to Transport Phenomena in Materials Engineering

hence the loss of head between two points in a flow system is an alternative measure
of the dissipation of energy by friction loss between the two points. Application of
the concept of head loss is restricted to the flow of incompressible fluids in flow
systems in which there is no work transfer into or out of the fluid (e.g., no pumps or
turbines).

6.8 FLUID FLOW IN AN OPEN CHANNEL

Consider a steady flow of fluid in the inclined open channel shown in Figure 6.6.
The channel, of length L and width W, is inclined at an angle from the horizon-
tal, and under conditions of uniform flow, the depth h of the fluid is constant along
the channel.
The gravitational force acting on the fluid in the channel in the direction of flow
equals the frictional force exerted on the fluid by the walls of the channel:

1
hWL g sin o L 2h W f v 2 L 2h W , (6.32)
8

where the wetted perimeter is the sum of the width (W) and twice the height of the
liquid layer (2h). From Eq. (6.17), the hydraulic diameter of the flow, DH, is

4A 4hW
DH , (6.33)
Pw 2h W

FIGURE 6.6 Gravity-induced flow in an open channel.

Mechanical Energy Balance in Fluid Flow 271

substitution of which into Eq. (6.32) gives

1 L
hWLg sin f v 2 4 hW
8 DH

or
1/ 2
2 DH g sin
v . (6.34)
f

For fully turbulent flow, the friction factor is given by Eq. (6.15) when Re DH :
2
1
f 2 log , (6.35)
3.7 DH

where is the absolute roughness of the channel wall. Substitution of Eq. (6.35) into
Eq. (6.34) gives

1/ 2 1
v 8 DH g sin log . (6.36)
3.7 DH

Example 6.6
Flow of Liquid Lead Tapped from a Furnace
Liquid lead, tapped from a blast furnace, runs down an open channel (of width
W = 0.3 m and at an inclination from the horizontal of angle ) into a kettle. The
maximum allowable depth of the moving lead layer is h = 0.1 m. The absolute rough-
ness of the channel wall is = 0.0003 m, and the density and viscosity of liquid lead
are = 10,050 kg/m3 and μ = 1.82 × 10–3 Pa∙s.

Find: The depth of the mass flow rate of lead down the channel, m, over the
range 5° < < 25°.
Solution: We will use the correlation for friction factor for turbulent flow in
a circular pipe, but using the hydraulic diameter as the length scale. That
quantity at the maximum depth of lead is

4A 4hW 4 0.1m 0.3m

DH 0.24 m.
Pw 2h W 2 0.1m 0.3 m

The aspect ratio of this layer is AR h W 0.1 m 0.3 m 1 3, which is close

enough to unity to use the pipe flow correlations. The friction factor, assuming a fully
turbulent flow, is from Eq. (6.15):
2 2
1 1 0.0003m
f 2 log 2 log 0.0207.
3.7 DH 3.7 0.24 m
272 An Introduction to Transport Phenomena in Materials Engineering

From Eq. (6.34), we find the average velocity and mass flow rate as functions of
inclination angle:

1/ 2 1/ 2
2DH g sin 2 0.24 m 9.81m/s2 1/ 2 m 1/ 2
v s
sin 15.1 sin
f 0.0207 s

and

1/ 2
2DH g sin
m vA vhW hW
f
1/ 2
2 0.24 m 9.81m/s2 1/ 2
10, 050 kg/m 3 0.1 m 0.3m sin
0.0207
kg 1/ 2
4550 sin .
s

These results are plotted in Figure 6.7, which shows that the rate of lead transfer
down this inclined channel more than doubles over the range of possible inclina-
tion angles.
To check if the flow really is fully turbulent, we compare the two terms in the
expression inside the log function in Eq. (6.15). For the flow to be in that regime,

0.9 0.9
6.9 1 Re DH
or 1.
3.7 DH Re DH 3.7 DH 6.9

FIGURE 6.7 Average velocity and mass flow rate down inclined open channel in this exam-
ple as a function of inclination angle.
Mechanical Energy Balance in Fluid Flow 273

FIGURE 6.8 Ratio of terms inside log function in friction factor correlation, Eq. (6.15).
0.9
Because the ratio 3.7 DH Re DH 6.9 >> 1, the assumption that the flow is fully turbulent
is correct over the range of possible channel inclination angles.

A plot of this expression over the range of inclination angles is shown in Figure 6.8.
Because the curve is much greater than unity over the entire range, we can safely
assume the flow is fully turbulent and f ≠ f (Re).

6.9 TRANSIENT APPLICATIONS OF THE

MODIFIED BERNOULLI EQUATION
In many cases the purpose of the analysis shown in previous sections is to estimate
the rate of change of liquid levels in a system such as liquid metal filling a mold or
draining from a ladle. Using the latter case as an example, Figure 6.9, we will analyze
the transient liquid behavior in a ladle.
For the round ladle shown, the liquid level begins at h1(t=0) = ho and falls to h1
(t=tf) = 0, where tf is the draining finish time. The outlet of the free jet existing the
ladle is at z2 = h2 = 0. The modified Bernoulli equations for this configuration (assum-
ing w = 0) is written as:

P2 P1 1 2 1 1
v2 v12 g h2 h1 t K1 v12 K 2 v22 0. (6.37)
2 2 2

All the possible frictional losses in the ladle and the orifice discussed in earlier sec-
tions are lumped into the coefficients K1 and K2, respectively, which are assumed to
be not functions of liquid velocity. Rearranging Eq. (6.37) and using h2 = 0, gives

P1 P2 1 2 1 1
g h1 t h2 v2 v12 K1 v12 K 2 v22 . (6.38)
2 2 2
274 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.9 A ladle emptied by draining through an orifice.

The left-hand side of Eq. (6.38) has the terms driving the flow, the pressure difference
and potential energy, which provide the mechanical energy necessary to change the
kinetic energy and overcome frictional losses on the right-hand side.
The two velocities in Eqs. (6.37) and (6.38) are related by the volumetric flow
rate, which is uniform along the path from 1 to 2 if the fluid is incompressible:

Q v1 A1 v2 A2
or

v2 v1 A1 A2 . (6.39)

Substituting Eq. (6.39) into (6.38) and rearranging gives

2 2
P1 P2 1 A1 2 1 A1
g h1 t h2 1 v1 K1 K 2 v12
2 A2 2 A2

or

2 P1 P2 2
2 g h1 t h2 K1 1 A1 A2 1 K2 v12 .

The velocities at planes 1 and 2 are

2 P1 P2 2 g h1 t h2
v1 2
v2 A2 A1 . (6.40)
K1 1 A1 A2 1 K2
Mechanical Energy Balance in Fluid Flow 275

So far this derivation follows the case of the replenished vessel in Section 6.6, except
h1(t) is still an unknown function of time. However, we observe that the liquid veloc-
ity at 1 is the rate of change of the plane 1 height:

dh1
v1 ,
dt
which gives us a differential equation for h1(t):

dh1 2 P1 P2 2 g h1 t h2
C1 h1 (t ) C2 , (6.41)
dt K1 1 A1 A2
2
1 K2

where

2g 2 P1 P2 2 g h2
C1 2
and C2 . (6.42)
K1 1 A1 A2 1 K2 K1 1 A1 A2 1 K2

Using separation of variables, the solution of Eq. (6.41) can be written as

2 1/ 2
C1 h1 C2 t C3 .
C1

Applying the initial condition, h1(t = 0) = h0, we obtain the constant of integration,

2 1/ 2
C3 C1 ho C2 .
C1

The complete solution for the transient liquid level, h1, is thus
2
1 1/ 2 C1 C2
h1 t C1 ho C2 t . (6.43)
C1 2 C1

To find the time, tD, needed to drain the vessel to level h2, we set h1(tf) = h2 and solve
Eq. (6.43) for time, giving

2 1/ 2 1/ 2
tf C1 ho C2 C1 hD C2
C1

or, using the definitions of the constants in Eq. (6.42),

2
2 A1
tD K1 1 1 K2
g A2
1/ 2 1/ 2
P1 P2 P1 P2
g ho h2 g hD 2 . (6.44)
276 An Introduction to Transport Phenomena in Materials Engineering

The transient velocity at the orifice is the time derivative of Eq. (6.43) combined with
the uniform volumetric flow rate, Eq. (6.40):

A1 dh1 A1 A1 1/ 2 C1 t
v2 v1 C1ho C2 (6.45)
A2 dt A2 A2 2

or

A1 2
v2
A2 K1 1 A1 A2
2
1 K2

1/ 2 gt (6.46)
g ho h2 P1 P2 .
2

The results of the previous analysis can be simplified in several ways. If there is
atmospheric pressure on the top surface in the vessel and at the jet exit, then
P1 = P2 = Patm. If the orifice is much smaller than the area of the vessel, then A1/A2
2
>> 1 and A1 / A2 1 K 2 K 2 1 << 1. Applying these simplifications to Eqs.
(6.44) and (6.46) gives

A1 2 ho 1 K 2
tf (6.47)
A2 g

and

2 1/ 2 gt
v2 g ho . (6.48)
1 K2 2

To obtain these solutions, we began with the modified Bernoulli equation (6.6),
which was derived for steady flow conditions. The fully transient version, a deriva-
tion for which can be found in [8, 9], includes an acceleration term:

2 v P2 P1 1 2
ds + v2 v12 g z2 z1 w ef 0. (6.49)
1 t 2

The first term here accounts for accelerations along the length of the path from 1
to 2. We assume this transient effect to be small compared to the other terms. This
assumption means that we treat the flow at any instant in time as a steady flow, i.e.,
we have assumed the flow to be quasistatic. In that case, the first term in Eq. (6.49)
can be neglected.

Example 6.7
Emptying a Ladle of Metal
Revisiting Example 6.3, in which liquid metal is draining from a ladle, we turn our
attention to the same problem except that the metal is not replenished as before.
Mechanical Energy Balance in Fluid Flow 277

In this case, the height of the liquid metal falls from h1(t=0) = ho = 1 m until depleted
at h1 = h2 = 0. The diameter of the ladle and the orifice are, respectively, D = 2 m and
d = 0.1 m. Assume the liquid is an aluminum alloy, with μ = 1.3 × 10–3 kg/ms and
= 2500 kg/m3.

Find: The time to drain the ladle and average mass flow rate during draining.
Compare the latter to the steady m in the replenished ladle in Example 6.3.
Solution: Here we have all the simplifications noted in the paragraph leading
to Eq. (6.47), so Eqs. (6.47) and (6.48) are applicable to this situation. The
time to drain is then
2
D2 4 2 ho 1 K 2 2m 2 1m 1 0.5
tf 221s,
d2 4 g 0.1 m 9.81m 2

where K2 = 0.5 because the sharp contraction in the orifice is the only significant
friction loss in the system.
The average mass flow rate can be found in two ways. The first is to integrate
m t over time from t = 0 to t = tf:

1 tf A2 tf
m m dt v2 dt.
tf 0 tf 0

The flow rate is a function of the velocity, Eq. (6.48), and decreases linearly in time
as the ladle empties, so a simpler approach, now that we have tf, is
2
m h0 A1 h0 D2 4 2500 kg m 3 1 m 1m 4 kg
m 35.55 .
tf tf tf 221s s

This value is half the steady mass flow rate of the replenished vessel with a constant
depth, due to the decreasing potential energy driving the flow.

6.10 SUMMARY
This chapter introduces a simple model for estimating the behavior of a moving fluid
in a conduit by deriving a balance of mechanical energy on that fluid. Bernoulli’s
equation is that balance of the net work done on the fluid (in terms of a pressure
drop), the change in kinetic energy, and the change in potential energy. In its original
form, Bernoulli’s equation represents the transfer of mechanical energy among these
three phenomena, but neglects the loss of mechanical energy due to friction. (Fric-
tion is the mechanism for transforming mechanical energy to heat and appears in a
mechanical energy balance as a sink.) The modified Bernoulli equation gives us the
interaction of friction with the average velocity in the conduit and the net work added
(as a pressure drop or a change in potential energy). Models of frictional effects
are mostly empirical correlations for friction factors (non-dimensional wall shear
stresses) as functions of Reynolds numbers and surface roughness and the energy
278 An Introduction to Transport Phenomena in Materials Engineering

losses caused by flow through bends, fittings, and expansions and contractions in the
conduit. These models are used in steady-state applications, such as pumping metal
between two melting furnaces to mix alloys, using a reservoir of water to drive a
turbine at a lower altitude, and slag flow down a launder. Models of transient system
include draining of vessels and the filling of molds.

6.11 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

6.1 To mix liquid lithium into an aluminum alloy, Li is melted pure in one
furnace (A) and pumped into another for mixing (B). The two furnaces
have argon atmospheres to reduce lithium oxidation and evaporation and,
for safety reasons, are kept in different rooms. The liquid lithium is moved
from A to B by pressurizing furnace A. Assume that the transfer pipe is
always full of liquid Li. (See Figure 6.10.)

DA << DB DA = 0.7 m dP = pipe diameter = 0.03 m Lp = total

pipe length = 100
Li = 500 kg/m Al = 2500 kg/m HA = 1.5 m HB = 3 m
3 3

PA = 7 × 10 Pa
5 PB = 1 × 10 Pa
5 f = 0.02

(a) Neglecting the frictional effects in each furnace (including the inlet and
outlet of the transfer tube), write an expression for KTOT in terms of frictional
losses, e f 1 2 KTOT v p2 and v p is the velocity of the lithium in the pipe. (b)
Assume the melting furnace (A) is replenished continuously and the mixing
furnace (B) is so large that the added lithium has little effect on liquid height.
Find an expression for v p . What physical mechanism(s) drive this flow? What
physical mechanism(s) restrain this flow? (c) Calculate v p from part (b) and the
time to add 300 kg of lithium to the alloy. (d) Briefly describe how you would
work this problem differently to find the lithium transfer time as in part (c) if
the induction melting furnace (A) is not replenished.

6.2 Figure 6.11 shows fluid being poured into a vessel (diameter = D = 0.5 m) at
a rate that maintains the level of the liquid free surface at a constant value,
ho = 1 m. The fluid drains through a tube of length L = 1 m and diameter
d = 0.01 m attached to the bottom of the vessel. Plane 1 is at the free surface,
and plane 2 is at the exit of the tube. Assume a fully turbulent flow through
the tube. Use = 2500 kg/m3 and μ = 0.001 kg/ms. (a) Write the modified
Bernoulli equation, identifying all terms restraining and driving the flow
and all the contributions to frictional loss. (b) Guess a reasonable value
for the friction factor and find the tube exit velocity and friction factor; an
iterative solution may be required. (c) Modify the mechanical energy equa-
tion from part (a), assuming that the exit of the tube is submerged to a depth
of 0.25 m. Explain why this change increases or decreases the exit velocity.
Mechanical Energy Balance in Fluid Flow 279

FIGURE 6.10 Schematic of system for mixing an aluminum lithium alloy as in Problem 6.1.

6.3 Repeat Problem 6.2, but do not replenish the fluid in the vessel. Assume a
friction factor of f = 0.03 and plot the velocity at the tube exit as a function
of time.
6.4 A layer of molten slag of density 2,700 kg/m3 and viscosity 0.3 Pa∙s is
being transferred from one reverberatory furnace to another by flow down
a plane between the two furnaces that is inclined at an angle of 50° to the
vertical. The inclined plane is 5 m in width and 5 m in length, and the mass
flow rate of the slag is 7.5 kg/s. Calculate: (a) the thickness of the layer and
(b) the average linear flow velocity of the slag. Neglect end effects.
6.5 This problem is a follow-on to problem 5.11, which should be done first
to find the pressures in the system. An electrode is melted by dipping it in
a hot slag and the liquid metal, after falling through the slag as droplets,
forms a liquid pool below. The metal exits this reservoir through a small
nozzle at the bottom and this metal stream is atomized. (This spray of liq-
uid metal is then used to form an ingot with a very refined, but somewhat
porous, microstructure.) We are interested in the control of the metal height
in the reservoir and the flow rate. A sketch is in Figure 6.12. (a) Write
the modified Bernoulli equation in the metal, including all expressions for
frictional losses. Ignore any frictional effects in the slag. Label significant
points on Figure 6.12. (b) At steady state, what is the melt rate? (c) If the
melt rate is increased by 50%, what would be the required depth of the
liquid at steady state?

K nozzle = 20 metal = 8900 kg/m

3
slag = 2000 kg/m
3

LM = 0.3 m LS = 0.2 m
µ = 10 –3 kg/ms Patm = 0.101 MPa (on top of slag layer) Area of
exit = Ae = 10 –5 m2
Area of top of metal reservoir = AM = 1 m2
280 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.11 Drainage of a vessel through a tube for Problems 6.2 and 6.3.

6.6 Consider a round jet of liquid leaving a vertical nozzle of radius = Ro at a

velocity Uo. the exit of the nozzle is z = 0 and the positive z-axis extends in
the direction of gravity. (a) Carefully sketch the shape of the liquid stream
below the nozzle in the figure. (If you want a hint, try looking at the water
leaving a faucet. For that case, the shape of the water jet is easier to see if it
is relatively slow.) (b) Using Bernoulli’s equation, find R = f (z) (the radius
of the falling stream as a function of the distance from the nozzle outlet).
Ignore friction losses at outlet. Explain the shape in physical terms.
6.7 Liquid in a riser is maintained at a depth of y = ho while fluid leaves it to fill
a mold of height Hf (see Figure 6.13). The surface area of the riser is much
larger than that of the mold or the cross-sectional area of the pipe. The
mold is a cylinder of diameter D, and the communicating pipe is diameter
d. In the pipe is a filter designed to keep inclusions from entering the mold,
as well as slowing the fill rate. The height of the fluid in the mold is H = 0
at t = 0; the rest of the system is full of liquid metal.

f = 0.03 Lpipe = 1 m (Le/d)elbow = 30 Kf = 2 (loss coefficient for

flow through the filter)
D = 0.2 m d = 0.01 m ho = 1 m Hf = 0.5 m

(a) Write an expression for time to fill the mold as a function of the system
geometry, fluid properties, and loss coefficients. (Do not substitute numbers in
at this point; leave in symbolic form.) (b) Rank the various resistances to flow
Mechanical Energy Balance in Fluid Flow 281

FIGURE 6.12 System for generating liquid metal spray by modifying an electroslag remelt-
ing unit as in Problem 6.5.

by their numerical value. Can any of them be safely ignored in the design of
this process? (c) How long to fill the mold? (d) How long to fill the mold if the
filter is removed? (e) Repeat (a), (c), and (d), letting the height of the fluid in the
riser fall, i.e., ho = ho(t), and assuming the diameter of the riser is DR = 0.17 m.

6.8 We want to feed aluminum alloy from a melting furnace to a mold for direct
chill casting by moving the liquid metal down an inclined trough. (a) What
is the mass flow rate at which metal is required if the cross sectional area of
the cast rectangular ingot is 3 m × 1 m and the casting velocity is 5 mm/s?
(b) If the flow in the trough is fully turbulent, what is the depth of the flow in
the trough and the angle of the trough from the horizontal? The rectangular
trough is 0.3 m wide. Use Al = 2500 kg/m3 and μ Al = 0.001 kg/ms.
6.9 Water flows in a cast iron pipe (D = 0.5 m) with an absolute roughness
of 0.25 mm at an average velocity of 2 m/s. Find: (a) the energy loss per
100 m; (b) pressure drop per 100 m.
6.10 In Figure 6.14 oil is pumped from a lower to a higher reservoir by a pump
with a work input of 150 kW and an efficiency of 80%. Calculate the mass
flow rate of oil between the tanks if the frictional energy loss is 120 J/km.
6.11 A ladle with an inside diameter of 1 m and a depth of 1.5 m has a sharp-
cornered tapping nozzle (0.1 m diameter) in its base. The ladle is initially
filled with a liquid aluminum alloy ( = 2410 kg/m3, μ = 0.003 kg/ms).
Find: (a) time required to drain the ladle; (b) the initial discharge rate (in
kg/s); (c) the discharge rate when the ladle is half empty.
282 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 6.13 Mold filling process in Problem 6.7.

FIGURE 6.14 Pumping oil from lower to higher reservoir in Problem 6.10.

6.12 Calculate the time required to fill the cylindrical mold in Figure 6.15 with
liquid metal from the previous tundish. (Assume the tundish has a very
large surface area compared to the mold.) The free surface level in the
tundish remains constant. Use = 6400 kg/m3 and μ = 0.0017 kg/ms.
6.13 The crucible in Figure 6.16 contains liquid steel and is open to the atmo-
sphere. Find: (a) the initial discharge rate from the full tank; (b) the time
to drop the free surface from 10 m to 2 m; (c) the pressure to which the
crucible would have to be pressurized to halve the time in part (b). Use
= 7500 kg/m3 and μ = 0.007 kg/ms.
Mechanical Energy Balance in Fluid Flow 283

FIGURE 6.15 Tundish and mold in Problem 6.12.

FIGURE 6.16 Drainage of a crucible through a horizontal tube in Problem 6.13.

REFERENCES
1. Reynolds, O., “An experimental investigation of the circumstances which determine
whether the motion of water shall be direct or sinuous, and of the law of resistance in
parallel channels,” Transactions of the Royal Society of London, v. 174, pp. 935–982,
1883.
2. Nikurasde, J., “Laws of flow in rough pipes,” in NACA Technical Memorandum 1922,
National Advisory Committee on Aeronautics, 1950.
3. Colebrook, C. F., “Turbulent flow in pipes, with particular reference to the transition
region between the smooth and rough pipe laws,” Journal of the Institute of Civil Engi-
neers London, v. 11, pp. 133–161, 1939.
4. Engineering Division, Crane Co., Flow of Fluids Through Valves, Fittings, and Pipes,
Technical Paper 410, Crane Co., 1957.
284 An Introduction to Transport Phenomena in Materials Engineering
 7 Equations of Fluid
Motion

7.1 INTRODUCTION
In the previous chapters, we have examined characteristics of static fluids and devel-
oped a coarse model of a mechanical energy balance on a fluid moving in a conduit.
Beginning in this chapter, we will look at fluid behavior in more detail. In developing
models for fluid flow, we use mass conservation and Newton’s second law of motion
to derive relationships between the velocity field and the physical mechanisms driv-
ing and restraining flow. These mass and force balances will be used in later chapters
to describe many different flows and the heat and mass transport in them.
These models for mass and momentum conservation are extremely useful in
understanding the behavior of flow, especially in that each term in the equations rep-
resents some physical phenomenon and each equation shows the interaction of those
phenomena. As we have seen with the diffusion of heat and mass, understanding
these interactions is important for gaining insight into the behavior of the transport
phenomena, even when we do not explicitly solve the equation. Thus, a brief intro-
duction to the derivation of these governing equations and a recognition of their phys-
ical meaning is vital to understanding why fluids behave as they do. Studying fluid
mechanics without this beginning is like studying basic Newtonian physics without
calculus; we could take that approach, but it is an unnecessary obstacle to learning.
The derivation of these governing equations is facilitated by looking at these flows
from a particular point of view. We use the approach, termed the Eulerian point of
view, of observing flow through and forces acting on a small (generally fixed in
space) control volume to derive the governing equations. Eulerian modeling includes
not only the forces and flows interacting with a small control volume, but also the
local accumulation of mass or momentum at a fixed point in space.1 This method is
different from the Lagrangian point of view, which follows the behavior of a small
packet of fluid as it moves through space under the influence of various effects.

7.2 CONSERVATION OF MASS

To model the conservation of mass in a flowing fluid, we begin by examining mass
moving through an infinitesimal Cartesian control volume, fixed in space, with a
differential volume, d = dx dy dz. The mass balance requires that the rate of change
of mass within the control volume be the difference between the rates at which mass
enters and exits the volume:
massaccumulation or depletion rate = rate of mass entering
(7.1)
raate of mass leaving .

DOI: 10.1201/9781003104278-7 285

286 An Introduction to Transport Phenomena in Materials Engineering

Consider the two-dimensional control volume in Cartesian coordinates shown in

Figure 7.1. (We will show the details of the derivations of mass conservation and
(in the next section) the momentum balance in a two-dimensional velocity field, but
will also present the three-dimensional versions at the end of the process.) The rate at
which mass flows across the volume interfaces is a product of the density, the magni-
tude of the velocity normal to the interface, and the area of that interface:

m V A, (7.2)

where m has the units kg/s. Referring to Figure 7.1, we can rewrite Eq. (7.1) as

m
mx my mx dx my dy , (7.3)
t

where the three terms are the mass storage in the volume, the mass flow rate in, and
the mass flow rate out. Alternately, we can write this mass balance in terms of the
change in mass flow rate across the volume in each direction:

m
mx mx dx my my dy .
t
mass mass rate mass rate (7.4)
stored off change of change
in x in y

FIGURE 7.1 Mass flow rates and storage in a two-dimensional Cartesian control volume.
Equations of Fluid Motion 287

The fluid velocity is a vector field, which comprises three components:

V u iˆ v ˆj w k̂ (7.5)

in Cartesian coordinates and, in cylindrical coordinates,

V u rˆ v ˆ w ẑ. (7.6)

Using the Cartesian definition of the mass flow rates across the surfaces at x and y
and of the velocity vector in Eqs. (7.2) and (7.5), we get

mx ux Ax ux dydz and my vy Ay vy dxdz. (7.7)

(We have assumed the control volume has a depth dz into the page.) A Taylor series
approximation (truncated to two terms) can be used to obtain estimates of the mass
flow rates in both directions in close vicinity of positions x and y:

mx dx ux dx Ax dx and my dy vy dy Ax dy

u v
mx dx u x
dx dydz and my dy v y
dy dxdz. (7.8)
x x y y

The mass storage term can be written as

m
d dxdydz . (7.9)
t t t

Combining relationships (7.7)–(7.9) in the mass flow balance (7.4) produces

u v
dxdydz dx dydz dy dxdz.
t x x
y y

Dividing by the differential volume gives

u v
0, (7.10)
t x y

or in three dimensions

u v w
0. (7.11)
t x y z
288 An Introduction to Transport Phenomena in Materials Engineering

The first term in Eqs. (7.10) and (7.11) is the local storage of mass in a constant
volume. The other terms are the change in the mass fluxes ( u, v, w) across the
control volume. In cylindrical coordinates, where the velocity field is represented by
Eq. (7.6), a similar process gets

1 r u 1 v w
0. (7.12)
t r r r z

If the density is constant and uniform, then Eq. (7.11) is reduced to

u v w
0 (7.13)
x y z

in Cartesian coordinates and, in cylindrical coordinates,

1 ru 1 v w
0. (7.14)
r r r z

This approximation holds true in incompressible fluids (such as liquids), where den-
sity is at most a weak function of pressure, and in gases with small pressure differ-
ences. These mass balance equations, Eqs. (7.11) and (7.12), are known as the mass
conservation equation or the continuity equation. The first name is obvious from
our derivation; the second is because this equation represents a constraint on the
momentum equations (in the next section), which keeps the fluid continuous with no
gaps or overlaps.
To illustrate the relationship between the two mass flux terms in Eq. (7.10),
Figure 7.2 shows a volume with velocity in from the left side (u1) and out the right
(u2), with some flow through the top surface at velocity v2 and none through the bot-
tom (v1 = 0). If u1 > u2, then the flow decelerates in x and ∂u/∂x < 0. From Eq. (7.13)
with ∂w/∂z = 0, we see that ∂v/∂y > 0 or v2 > v1. In this case, the flow enters from the
left and splits into two different directions; mass leaving out the top of the volume
(v2 > 0) slows the horizontal velocity. In the opposite case, if u1 < u2 (accelerating

FIGURE 7.2 Control volume illustrating conservation of mass in two dimensions.

Equations of Fluid Motion 289

flow), then ∂u/∂x > 0 and ∂v/∂y < 0, so v2 < v1. To accomplish this increased flow
rate in the x direction, mass must be entrained into the volume through the top with
a downward velocity, v2 < 0.
Beginning with Eq. (7.13) for two-dimensional steady flow (w = 0), we can define
the streamfunction, (x, y), in Cartesian coordinates as

u and v . (7.15)
y x

This quantity is defined thus to satisfy the continuity equation identically:

2 2
u v
0.
x y x y y x x y x y x y y x

The total differential of (x,y) is

d dx dy v dx u dy. (7.16)
x y

We also define a streamline as a line in a velocity field everywhere tangent to the

flow, as shown in the example flow field in Figure 7.3 (a). This tangency condition
means that velocity has no normal component on a streamline and so streamlines are
impermeable. The slope of the streamline can be defined as

dy v
, (7.17)
dx u
where the rise over run is determined by the ratio of velocity components. Eqs. (7.16)
and (7.17) combined gives

v dx u dy d 0,

which shows that the value of the streamfunction is uniform along a streamline.
The fact that no fluid crosses a streamline (a line of constant ) is useful for dis-
playing the distribution of the volume flow rate, Q = V A, in a flow field. Any two
streamlines form a stream tube, and we find that the flow rate is the same across any
line drawn from one streamline to the other. This result is illustrated in Figure 7.3 (b)
by calculating Q across the two lines, A and B. For line A, the only normal velocity
component is u, so
y2 y2

QA u dy dy 2 1 ,
y1 y1
y
For line B, with no horizontal component of velocity,
x1 x2 x2

QB v dx v dx dx 2 1 .
x2 x1 x1
x
290 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 7.3 (a) Streamlines (lines of constant ) in a flow field, showing tangent velocity
vectors. The surface also acts as a streamline, as no flow passes through it. (b) Flow through
a stream tube between two streamlines ( 1 and 2).

FIGURE 7.4 Streamlines in a flow field moving from left to right. The flow first accelerates
as streamlines converge, then decelerates as they diverge.
Equations of Fluid Motion 291

We see that QA = QB, so the volume flow rate does not change along the stream
tube constrained by 1 and 2. The value of the volume flow rate, Q, is also the dif-
ference of the streamfunction values for the two streamlines.
Figure 7.4 shows a family of streamlines representing a flow field moving left to
right. Moving from the left, the streamlines converge then diverge. As they move
closer together, the cross sectional areas of the stream tubes decrease, causing an
increase in the velocities along the tubes. When they then diverge farther down-
stream, those areas increase and the flow decelerates.

7.3 MOMENTUM BALANCE: THE NAVIER–STOKES EQUATIONS

In this section, we apply Newton’s second law of motion to a fluid control volume
to derive a momentum balance equation. This balance and the continuity equation
(7.13) form a basis for a description of the mechanisms controlling fluid motion.
In most early physics courses, Newton’s second law is written as

F ma,

where the mass is constant and a is the acceleration vector. A more general version
is that the sum of the forces causes a change in the momentum vector, , with time:

d
F . (7.18)
dt

We can write the momentum field as

x, y,t mV ,

and its total derivative in two-dimensional Cartesian coordinates as

mV mV mV
d dt dx dy. (7.19)
t x y

The total change in momentum with time is then

d mV dt mV dx mV dy
dt t dt x dt y dt
(7.20)
mV mV mV
u ,
t x y

where u = dx/dt and v = dy/dt. This form of the total time derivative in a flow field
is known as the substantial, or material, derivative, which is usually written for any
transported quantity as
292 An Introduction to Transport Phenomena in Materials Engineering

D
u v . (7.21)
Dt t x y

In Cartesian coordinates, the vector definition can be written in its x and y components:

D mu mu mu mv mv mv
u v iˆ u v ˆj (7.22)
Dt t x y t x y

In each direction, the first term is the local acceleration in the flow. It represents
changes in the fluid velocity with time at a point; it is zero in a steady flow. The next
two terms show how a fluid moving in a velocity field can accelerate or decelerate by
moving from one point in that field to another with a different velocity. They repre-
sent the rate of change of mass flux through a point because there is a velocity gradi-
ent there. To examine this idea more closely, let us focus on the term u mu x î in
Eq. (7.22). The derivative represents the rate of change of the x direction momentum
(mu) with a change in position along the x axis. How fast does that change occur? As
noted earlier, the rate of position change is the velocity, u dx dt , and so the product

mu dx mu mu
u
x dt x t

Is the rate of change of mu as a result of horizontal fluid motion through a nonuni-

form momentum field. If there is a vertical motion, v dy dt, then the x momentum
will change at the rate v mu y as the fluid moves in the vertical direction. Sim-
ilar arguments are made for the time rate of change of the y momentum, mv, due to
change in position. These four terms in Eq. (7.22) represent the advective accelera-
tion, or the change in momentum with position in a flow, which occurs where there
is a nonuniform velocity field, even if that flow is steady.
Having found an expression in terms of the velocity field for the change of
momentum in Newton’s second law, Eq. (7.22), we turn to the sum of the forces.
The forces on a two-dimensional control volume (with depth dz into the page) are
shown in Figure 7.5, and they fall into three types: normal, shear, and body forces.
The normal forces act perpendicular to surfaces, the shear parallel to them, and the
body forces throughout the volume. The net normal force in the x direction, using a
truncated Taylor series, is

xx xx
xx x dx xx x dydz xx x dx xx x dydz dx dydz. (7.23)
x x
x

Similarly, the shear forces in x are

yx yx
yx y dy yx y dxdz yx y dy yx y dxdz dy dxdz. (7.24)
y y
y
Equations of Fluid Motion 293

FIGURE 7.5 Two-dimensional control volume showing forces on a moving fluid.

The body force in the x direction depends on what physical effects act on the
entire volume (as opposed to acting on the surfaces) in the system we are modeling.
In many processes, electromagnetic forces are used to either damp or impel fluid
flow, and gravity is always included. Here, we include only gravity:

FBx mg x g x dxdydz , (7.25)

where the gravity vector is g g x iˆ g y ˆj..

The sum of these forces in the x direction is

Fx iˆ g x dxdydz iˆ,
xx yx

x y (7.26)
normal shear gravity

and a similar process in the y direction will give

Fy ˆj g y dxdydz ˆj.
yy xy

y x (7.27)
normal shear gravity

Applying the substantial derivative of the momentum field, Eq. (7.22), at the location
of the control volume subject to the forces in Eqs. (7.26) and (7.27), and using m =
dxdydz, gives us a new form of Newton’s second law in x and y:

u u u xx yx
u v gx
t x y x y (7.28)
channge in momentum sum of forces
294 An Introduction to Transport Phenomena in Materials Engineering

and
v v v yy xy
u v gy .
t x y y x (7.29)
chaange in momentum sum of forces

To progress from this point, we require constitutive equations for the shear and normal
stresses in two dimensions; these relations are found from those in Section 1.3.1.1
assuming ∂/∂z = 0.

u v u v
xx P 2 yy P 2 xy yx (7.30)
x y y x
A derivation of these equations for a Newtonian fluid is found in Schlichting [1].
Using these relations between forces on the fluid and its velocity gradients, gives us
the momentum balances in the x direction,

u u u P u u
u v 2
t x y x x x y y
(7.31)
v
gx ,
y x

and in the y direction,

v v v P v v
u v 2
t x y y x x y y
(7.32)
u
gy .
x y

For many applications, the temperature variation in a fluid is small enough that the
dynamic viscosity, μ, may be treated as uniform, and the range of pressure is small
enough that the fluid is effectively incompressible ( ≈ constant and uniform). Within
those limitations, we can reduce the equations to a more simplified form:

2 2
u u u 1 P u u
u v gx
t x y x x 2
y2 (7.33)
local convecctive pressure viscous gravity
acceleration friction

and
2 2
v v v 1 P v v
u v 2
gy .
t x y y x y2 (7.34)
local conveective pressure viscous gravity
acceleration friction
Equations of Fluid Motion 295

These equations, expressions of Newton’s second law in terms of fluid velocity

and pressure, are the Navier–Stokes equations. In each direction, the left-hand side
is the time rate of change of momentum. Causing these local and advective accel-
erations are the forces on the right-hand sides of Eqs. (7.33) and (7.34): pressure,
viscous friction arising from shear stress, and gravity.
Extending the previous derivation to three dimensions, using the full constitutive
equations in Section 1.3.1.1, gives us

2 2 2
u u u u 1 P u u u
u v w gx , (7.35)
t x y z x x2 y2 z2

2 2 2
v v v v 1 P v v v
u v w gy , (7.36)
t x y z y x2 y2 z2

and

2 2 2
w w w w 1 P w w w
u v w gz . (7.37)
t x y z z x2 y2 z2

Written in cylindrical coordinates (r- -z), these equations are

u u v u v2 u
u w
t r r r z
(7.38)
2 2
p 1 1 u 2 v u
ru gr ,
r r r r r2 2
r2 z2

v v v v uv v
u w
t r r r z
(7.39)
2 2
1 p 1 1 v 2 u v
rv g ,
r r r r r2 2
r2 z2

and

w w v w w
u w
t r r z
2 2
(7.40)
p 1 w 1 w w
r gz .
z r r r r2 2
z2

The Navier–Stokes equations and the continuity equation govern the flow of a fluid,
but a general solution for the velocity field does not exist. This system consists of
four equations and four unknowns (u, v, w, P) and the momentum equations might
296 An Introduction to Transport Phenomena in Materials Engineering

be solved for the velocity components, if P(x, y, z) is known. Unfortunately, the

mass conservation equation does not include pressure, complicating the solution to
the system. Another difficulty arises in the advective acceleration, which includes
the nonlinear terms u u x iˆ and v v y ˆj . These nonlinearities are an important
reason that configurations that admit solutions are few and only for idealized condi-
tions. However, like other efforts in this text, a combination of exact, approximate,
and scaling results (the “five o’clock solutions” of Section 1.4) may be used to
make estimates about the flow behavior. These efforts may provide enough infor-
mation for an engineering decision or, if these results are not sufficient, they will
inform modeling choices made in subsequent studies using computational fluid
dynamics (CFD).

7.4 BOUNDARY CONDITIONS FOR FLUID FLOW

To attempt any exact, approximate, or numerical solutions to the Navier–Stokes
equations and the continuity equation, boundary conditions for the velocity com-
ponents are required. The two types of conditions considered here are based on the
velocity and the shear behavior at the domain boundaries.
The most common boundary condition for fluids is that of a continuous velocity
between the fluid and whatever material defines its volume. While in most situa-
tions the boundary is a solid, it can be another fluid. In any case, in the limit as the
boundary is approached from both sides, the velocities in the two fluids approach one
another. Thus, a fluid at the macroscopic scale appears to “stick” to whatever bounds
it; this behavior is known as the no-slip condition.

lim ut fluid
lim ut surroundings
, (7.41)
n 0 n 0

where ut is the velocity tangential to the interface and n is the direction normal to
it. In the special case of a solid, stationary boundary, the boundary conditions are
ut fluid 0, the no-slip condition, and

un fluid
0, (7.42)

because the boundary is impermeable and flow does not pass through it.
At the fluid volume boundary, we usually find a continuous shear also. For a New-
tonian fluid, this continuity can be written with the aid of Eq. (1.5) as

ut
interface , (7.43)
n fluid

where the interface shear is proportional to the change of the tangential velocity in
the direction normal to the surface.
At a solid boundary, this shear is not generally known and indeed may be one of
the important outcomes of the analysis, but it can be found once the velocity field
Equations of Fluid Motion 297

is known. At a fluid-fluid boundary, we can equate the continuous Newtonian shear

from both sides

ut ut
interface . (7.44)
n fluid 1
n fluid 2

In some rare cases, these conditions can be modified as a velocity jump at a solid wall
proportional to the shear stress at that boundary:

ut fluid
ut fluid
ut wall
LS . (7.45)
interface n
interface

The coefficient LS is the slip length, a strong function of fluid and wall properties and
flow geometry. If LS ≠ 0, then there is slip at the wall and the fluid is not perfectly
“sticky” there. To define a regime in which there is this velocity discontinuity, we
examine the value of the Knudsen number (Kn) in a flow,

Kn / LC , (7.46)

where is the mean free path of atoms or molecules in the fluid and LC is a charac-
teristic length scale based on the flow geometry. In most engineering applications,
Kn < 0.01 and the no-slip condition is a reasonable approximation. In rarified gases
or in very small spaces, the Knudsen number may be larger, as the mean free path
becomes a significant fraction of the length scale of the flow. If 0.01 < Kn < 0.3, the
use of the slip boundary condition (7.45) is a better model than (7.41). If Kn > 0.3
or so, then the assumptions of a continuum fluid break down and we require for this
regime a different model of flow altogether. Discussions about flows in these regimes
(Kn > 0.01), including estimates of LC and LS, can be found in books on microfluidics
[2, 3] and rarefied gases [4, 5]. In this text, all problems are in the range of Kn << 1,
and so the no-slip condition holds.

Example 7.1
Boundary Conditions on Fluid Layers in a Float Glass Process
Here we consider the boundary conditions on two fluid layers, one on top of the
other, as seen in Figure 7.6. The figure shows float glass processing, in which a layer
of molten glass floats on top of a molten tin pool in pursuit of the production of a flat
sheet of glass.

Find: The boundary conditions for horizontal velocities, uSn(y) and uglass(y), for
the flows in the two fluids.
Solution: For the stationary furnace wall at y = 0, the no-slip condition applies

uSn 0. (a)
298 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 7.6 Schematic of float glass process, with a layer of molten glass on a pool of
liquid tin.

The shear condition is

uSn
wall Sn ,
y y 0

but this equation is not an aid to solving for the velocity field, as we do not know wall
a priori.
At the tin-glass interface (y = ySn), there is no discontinuity in velocity or shear, so
Eqs. (7.41) and (7.44) apply there:

uSn y ySn
uglass (b)
y ySn

uSn uglass
Sn glass . (c)
y y ySn
y y ySn

Finally, at the air-glass interface, the no-slip condition applies,

uglass uair y yglass

,
y yglass

but is of little use unless we are also solving the velocity field in the air. Of more util-
ity is the shear continuity condition,

uglass uair
glass air ,
y y yglass
y y yglass

which can be rewritten, with the values of air and glass viscosities at 1200 oC, as

uglass y air 4 10 5 kg/ms 7

4 10 1.
uair y glass 10 2 kg/ms
Equations of Fluid Motion 299

Because this ratio is so small, we can treat this interface as a free surface for the
glass, which has a zero shear (or zero velocity gradient) at its top limit:

uglass
0. (d)
y y yglass

This result is typical of an interface between a gas and a liquid.

If solving for the flow in the tin and glass layers, then we only need four boundary
conditions, two for each fluid. (We obtain general velocity solutions from the force
balance in each fluid separately; they are coupled through the boundary conditions at
their common interface, y = ySn.) The shear at y = 0 is not known without the velocity
field, and the air does not affect the glass flow, so the four conditions needed are (a),
(b), (c), and (d).

7.5 CHARACTERISTICS OF PRESSURE-DRIVEN

FLOW BEHAVIOR IN A CHANNEL
A pressure-driven flow in a horizontal channel is studied in this section by scaling
the Navier–Stokes equations and continuity. The purpose of the analysis is to identify
the flow structure, the physical mechanisms that determine it, and the relationships
between pressure and average velocity.
The geometry for this study is in Figure 7.7. The inlet to the channel has a velocity
(U∞) uniform across the channel height (H). The steady flow is driven by a constant
and uniform negative pressure gradient along its length (L):

dp PL PH
0, (7.47)
dx L 0

where PH is the high upstream pressure and PL the low pressure at the downstream
outlet (x = L). The flow is two-dimensional and the fluid Newtonian. The channel is
also assumed to be slender, or H/L << 1.
The first step in this scaling analysis is to simplify the Navier–Stokes equations as
much as possible. We begin with Eq. (7.33), eliminating the local acceleration terms

FIGURE 7.7 Conditions for a pressure-driven flow in a horizontal channel.

300 An Introduction to Transport Phenomena in Materials Engineering

because the flow is steady and the gravity term because the channel is horizontal. Eq.
(7.47) replaces the pressure gradient term in the x momentum equation. So far, then,
we have
2 2
u u 1 P u u
u v ,
x y L x2 y2 (7.48)
advective pressure visscous
acceleration friction

where ΔP = PH – PL. Left in the momentum equation (7.48) are advective accelera-
tion, pressure drop, and viscous friction. The subsequent scaling analysis will shed
some light on which physical mechanisms are dominant and under what conditions.
Next, we select the reference values for the parameters that change in the system.
Because the pressure field is already known, the only dependent variables are the
velocity components. For the horizontal velocity, the scale is

uref ~ U . (7.49)

Because the mass flow rate is constant and uniform, U∞ is always the order of mag-
nitude of the velocity along the length of the channel. Nothing is known yet about
the vertical velocity, vref, but we will discover that later. In the flow direction, we are
concerned about changes over the length of the channel, so

xref ~ L. (7.50)

The vertical length scale must be less than the height of the channel,

yref ~ H . (7.51)

but whether it is a smaller order of magnitude than H is not known yet.

To find an expression for vref, other references are applied to the continuity equa-
tion (7.13):
U vref H
~ or vref ~ U . (7.52)
L H L

Because (H/L) << 1, vref << U∞, suggesting that the vertical velocity has little role in
the overall flow behavior. Applying that result to the steady y momentum equation
(7.34), the vertical pressure gradient, ∂P/∂y, can also be shown to be negligible com-
pared to ∂P/∂x.
Using the previous reference values, the order of magnitude of each term in Eq.
(7.48) can be estimated:

U H U P U U
U , U ~ , , ,
L L H L L2 H2 (7.53)
A A ~ P, F, F
Equations of Fluid Motion 301

with the three terms representing the advective acceleration (A), the pressure gra-
dient (P), and the viscous friction (F). Both advective acceleration terms are always
(U 2 L ), so they may be combined into one term, which is also (U 2 L ). Also, we
compare the orders of magnitude of the two viscous friction terms to evaluate their
relative importance:
2
viscous friction from x gradient in shear U U H
~ ~ 1. (7.54)
viscous friction from y gradient in shear L2 H 2
L

This ratio has a very small value, indicating that 2 u x 2 2

u y 2 , so the former
term is definitely negligible. However, the latter term is only known to be much
larger than the former; only further analysis will tell if that shear term is relevant
compared to the other mechanisms.
Given these evaluations of the advective acceleration and viscous friction terms,
the x momentum equation can be rewritten as

U2 P U
~ , .
L L H2 (7.55)
A ~ P, F

In this system, the flow is driven by a constant and uniform pressure gradient and is
restrained by inertia and friction; without the pressure field, there is no flow under
any condition, so we assume that it is the dominant mechanism always important in
the flow. Hence, we divide Eq. (7.55) by the pressure term

U2 U L
~ 1, .
P PH 2 (7.56)
A ~ P, F

Eq, (7.56) has a nondimensional quantity for each term, all of which are (1) or less.
(If we discover one term had an order of magnitude greater than unity, then the pres-
sure is not the dominant mechanism and we should rethink that assumption.)
So, which term or terms balance the pressure gradient? To answer that question,
we examine two extreme limits, in which the pressure is balanced either by accel-
eration or friction. In the first case, we assume the pressure and acceleration are the
same order of magnitude, with the friction term being much smaller than both.

P U L
~ 1 and 1. (7.57)
U2 P H2

Using the pressure drop in the first part of that result in the friction term,

U L U L L
~ ~ 1. (7.58)
P H2 U2 H2 U H H
302 An Introduction to Transport Phenomena in Materials Engineering

The Reynolds number in this case is defined as Re H U H/ , and so

H
Re H 1 (7.59)
L

is the condition under which friction may be neglected and the advective acceleration
is balanced by the pressure gradient. This condition can occur with a combination of
a high velocity, low viscosity, wide channel (high ReH), and/or a short channel (low
L). In this regime, the viscous effects at the walls have not penetrated significantly
into the flow in the channel.
The other limit for Eq. (7.56) is that the acceleration term has no influence, and
pressure and friction have the same order of magnitude. In that case,
1
U L P H
1 ~ or ~ Re H . (7.60)
P H2 U2 L

Using that expression for the pressure drop, the acceleration can be neglected if

U2 H
~ Re H 1, (7.61)
P L

which is true for a combination of a low velocity, high viscosity, narrow channel (low
ReH), and/or a long channel (long L). Here the viscous effects control the flow field to
the centerline. The negligible advective acceleration terms suggest that the flow on a
particular streamline does not change velocity downstream.

FIGURE 7.8 Nondimensional pressure drop vs. ReH (H/L) from a scaling analysis of pres-
sure-driven flow in a horizontal channel.
Equations of Fluid Motion 303

Figure 7.8 is a plot of the relationship of the vigor of the flow, represented by
ReH (H/L), and the pressure drop in the channel. For low values of ReH (H/L), the
mechanical work added by the pressure gradient is dissipated by the viscous friction,
and there is no acceleration (local or advective) in the flow. This condition is the fully
developed case applied to internal flows in Chapter 8. For higher ReH (H/L), the fric-
tion is negligible and the pressure feeds flow acceleration. When ReH (H/L) ~ (1),
the flow behavior is less clear, but the sketch at least gives an order of magnitude
estimate of the relationship between ΔP and U∞.
At this point, we have accomplished our goals for this analysis. We found the
dominant physical mechanisms in different flow regimes and the pressure drop-
velocity relationships there. It is now possible to calculate ReH (H/L) when beginning
such a problem and to know which flow regime dominates, and so to know how to
simplify the momentum equations before seeking a more detailed solution. Several
possible solutions will be introduced in Chapters 8 and9.

7.6 SUMMARY
In this chapter, we derive equations of mass conservation and momentum balance
in terms of the velocity and pressure fields. These continuity and Navier–Stokes
equations are the bases of more detailed analysis of fluid motion than in previous
chapters. Possible boundary conditions for velocity are introduced, where the no-slip
and free surface conditions are the most commonly used. A scaling analysis of these
equations in a channel flow demonstrates how to begin extracting useful information
from these equations. Such analyses may be used to gain insight into the flow behav-
ior for its own sake or to guide further, more detailed modeling.

7.7 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

7.1 For the scaling analysis worked out in Section 7.5 for the pressure-driven
flow in a horizontal channel, calculate the ReH (H/L) parameter for liquid
aluminum and molten polyethylene and comment on the balance of physi-
cal mechanisms in the flow of each. Assume H = 0.01 m, L = 0.3 m, and
U∞ = 0.2 m/s. Use: Al = 2400 kg/m3, μAl = 0.003 kg/ms; PE = 1500 kg/m3,
μapp,PE = 300 kg/ms.
7.2 Initially a fluid is at rest (quiescent) between two infinite flat plates sep-
arated by a distance . The size of the plates relative to gap means that
∂/∂x = ∂/∂x = 0, a one-dimensional flow. No pressure gradient or body forces
are applied. At t = 0, the upper plate is instantaneously accelerated to a
constant horizontal velocity, u = V. Due to the no-slip condition at the top
plate, the fluid at y = matches that velocity, but a finite time is required
for momentum to diffuse into the rest of the fluid layer. The left side of Fig-
ure 7.9 shows this early response. Eventually, at much later times, the steady-
state Couette flow described in Section 1.3.2 is achieved. We would like to
304 An Introduction to Transport Phenomena in Materials Engineering

know, to an order of magnitude, how long does it take to reach that steady
state? (a) Apply the simplifications described earlier to the continuity equa-
tion (7.13) and solve for v(y). (b) Apply the simplifications described earlier
to the x momentum equation (7.35) to reduce it to the minimum number of
remaining physical phenomena. How many terms are left, and which physi-
cal mechanisms do they represent? (c) For each variable parameter, define
a reference quantity characteristic of this problem. (Assume that the flow
reaches steady state shortly after the effect of the moving upper plate reaches
the lower one.) (d) Scale the reduced Navier–Stokes equation from (b) to find
an estimate for the time to steady state. How does that time depend on V? (e)
For liquid aluminum ( = 2400 kg/m3 and μ = 0.003 kg/ms) and air ( = 1.1
kg/m3 and μ = 1.8 × 10 –5 kg/ms), estimate the time to steady state.

FIGURE 7.9 Startup of Couette flow in Problem 7.2. The figure on the right is steady state
(t ∞).

7.3 A very long duct of height 2 contains a fluid initially at rest. Motion in the
fluid is started by a sudden change in its velocity from zero to a uniform
value, Uo, in the x direction. Because there is no leading or trailing edge,
there is no dependence of velocity on x. There are no body forces on the
fluid. (a) Write the Navier–Stokes equations and the continuity equations
and simplify as much as possible from what is already known about the sys-
tem. What is the value for the velocity in the y direction? Using this value,
and other appropriate approximations, simplify the momentum equation
in the x direction. What physical phenomena are left? (b) Show the order
of magnitude of the terms left in part (a) by using appropriate reference
quantities, including Px for the pressure gradient and t for time. Divide the
remaining terms by the one physical phenomenon always important in this
flow. (c) Find an estimate for Px(t) using the scaling analysis in part (b) at
short times. What is the physical effect that can be neglected at short times?
Under what conditions is it reasonable to neglect that physical mechanism
(i.e., what does “at short times” mean)? (d) Perform the analysis and answer
the questions in (c) for long times. (e) Sketch and label Px(t) vs. t from start-up
to steady state.
7.4 A liquid metal flows into an initially empty, vented channel in a die-casting
process. (The vent allows air to escape, keeping the pressure in the mold
at Patm.) We are interested in a quick estimate of how long it takes to fill a
Equations of Fluid Motion 305

mold of length L when the pressure change from the inlet, x = 0, to the lead-
ing edge of the metal front, x = xf (t), is kept constant through the process.
This pressure drop is ΔP = Pin – P(xf ), where P(xf ) = Patm and Pin is the
constant inlet pressure. (a) Assuming a fully developed two-dimensional
flow (∂/∂x = ∂/∂z = w = 0) everywhere, show that the x momentum equation
(7.35) can be reduced to

dp
,
y dx

Assuming a Newtonian fluid, write this force balance equation in terms of

velocity. (b) Remembering that u = dxf /dt, pick scales for velocity (Uref ),
time (tref ), and length (xref and yref ). In the mold-filling process, the pressure
gradient can be written as

dp P( x f ) Pin P
.
dx x f (t ) x f (t )

(c) Scale the previous momentum equation and find an expression for the
front position (xf ) as a function of time. (d) Sketch the functions xf and dxf /
dt as functions of time and explain their behavior. Write an estimate for the
filling time (when xf = L).

FIGURE 7.10 Liquid flowing into a mold cavity.

7.5 Write the constitutive equations for a shear-thinning, power-law fluid to

model a polymer in an injection-molding process. Repeat parts (a) through
(e) from Problem 7.4 for the power-law fluid and compare and contrast the
Newtonian (n=1) and power-law (n=0.7) results.
7.6 Perform the analysis in Problem 7.4, except to find ΔP(t) when a constant
injection flow rate, Uin, is maintained. (Eliminate the xf (= Uin t) from the
final result to get the proper function of time.)
7.7 Perform the analysis in Problem 7.6, except with a shear-thinning, power-
law fluid. Compare and contrast the Newtonian (n=1) and power-law
(n=0.7) results.
306 An Introduction to Transport Phenomena in Materials Engineering

NOTE
1. A simple analogy is different peoples’ points of view of the university experience. During
their education, students see this and other courses from a Lagrangian perspective, while
instructors (who once saw it that way) now view it as a Eulerian control volume through
which students are continually entering and leaving.

REFERENCES
1. Schlichting, H., Boundary-Layer Theory, 7th ed. (trans. Kestin), McGraw-Hill, 1979.
2. Nguyen, N. T., and S. T. Wereley, Fundamentals and Applications of Microfluidics, 2nd
ed., Artech House, 2006.
3. Tabeling, P., Introduction to Microfluidics, Oxford University Press, 2006.
4. Kogan, M. N., Rarefied Gas Dynamics, Springer, 1969.
5. Ivchenko, I. N., S. K. Loyalka, and R. V. Thompson, Jr., Analytical Methods for Prob-
lems of Molecular Transport, Springer, 2007.
 8 Internal Flows

8.1 INTRODUCTION
In the previous chapter, the general equations of motion, the momentum (Navier–
Stokes) and continuity equations, were derived. In this chapter, the discussion is
limited to application of these relationships to internal flows, which move between
geometric constraints. In terms of the results of the scaling analysis of Section 7.5,
we examine the case of small values of ReH(H/L), which suggests that any effects of
an entrance into the conduit may be safely neglected.1 These flows are fully devel-
oped, as they do not change in the direction of fluid motion. This chapter will develop
solutions for several simple, Newtonian, laminar flows in parallel plate channels or
cylindrical tubes, then discuss methods for modeling the passage of fluid through
porous media. In addition, several solutions with non-Newtonian fluids will also be
derived.

8.2 SIMPLIFICATIONS OF EQUATIONS OF MOTION

FOR INTERNAL FLOWS
To begin modeling internal flows, we consider the case of motion in a passage, as in
Figure 7.7, and that in the extreme case of ReH(H/L) << 1. That scaling result showed
that the advective acceleration and changes of viscous friction in the flow direction
are negligible compared to the pressure and friction changes perpendicular to the
flow. This observation allows us to model this regime by assuming fully developed
flow in the x direction, in this case meaning that there is no velocity change in the flow
direction, ∂(velocity)/∂x ≈ 0. Applying that condition to the continuity equation for
two-dimensional flow, Eq. (7.13),

u v
0.
x y

We can integrate the remaining term with respect to y and apply the no-slip condition
(v = 0) at either wall to show that:

v
dy 0 v C,
y

so that everywhere

v 0. (8.1)

DOI: 10.1201/9781003104278-8 307

308 An Introduction to Transport Phenomena in Materials Engineering

This result eliminates the y momentum equation (7.36). The x momentum equation
(7.35) is simplified by the fully developed assumption and Eq. (8.1):

2 2
u u u 1 P u u
u v gx (8.2)
t x y x x2 y2

or, changing the friction term to a more general form,

u P
g.
t y x (8.3)
local friction pressure gravity
accelerationn

This momentum equation can be applied generally to what are seen as one-dimen-
sional flows, but this chapter will only treat steady cases, so ∂u/∂t = 0 here and, for a
one-dimensional geometry,

P
g 0.
y x (8.4)
friction pressure gravity

This force balance will be the basis of our analysis of several flows in this chapter.
As an initial estimate of the interactions of these phenomena, we choose uref as
the reference velocity, and xref ~ L (the length of the channel) and yref ~ H (the height
of the channel) as length scales in the flow direction and perpendicular to the flow,
respectively. Scaling Eq. (8.4), we see that the orders of magnitude of the three terms
are written for a Newtonian fluid as

uref P
~ , g.
H2 L (8.5)
friction ~ pressure , gravity

The viscous friction force always restrains the flow, which is driven by a combination
of pressure gradient and gravity. In one extreme case, friction is balanced only by
pressure, and gravity is negligible, so

P H2 g L
uref ~ and 1. (8.6)
L P

At the other extreme, gravity drives the flow against the friction, and pressure has no
influence:
g H2 g L
uref ~ and 1. (8.7)
P
This simple analysis produces estimates and the functional forms for the flow veloc-
ities, depending on the dominant driving force.
Internal Flows 309

8.3 SHEAR-DRIVEN FLOW BETWEEN FLAT PARALLEL PLATES

The first case to which the force balance, Eq. (8.4), and no-slip boundary condi-
tions are applied is the Couette flow from the discussion of Newtonian fluids in Sec-
tion 1.3.1, seen again in Figure 8.1. Here the flow is between two horizontal (gx = 0),
parallel flat plates, and infinite in extent in the x and z directions. The bottom plate is
held stationary, u(y = 0) = 0, the top is moving at velocity u(y = H) = V, and the fluid
at both is subject to the no-slip condition. There is no pressure gradient in the fluid,
dp dx 0.
Given these assumptions, the static force balance on this flow, Eq. (8.4), reduces
to only shear:

0. (8.8)
y

For Newtonian fluids, this relation can be written in terms of the horizontal velocity, u,

u
0. (8.9)
y y

Integrating once produces

u
C1* and u C1 y C2 ,
y

if we let C1 C1* / and integrate again. Applying the lower boundary condition,

u( y 0 ) 0 C1 (0) C2 or C2 0.

FIGURE 8.1 Schematic of Couette flow, driven by shear at the no-slip, bounding plates.
310 An Introduction to Transport Phenomena in Materials Engineering

The upper boundary gives

u( y H ) V C1 ( H ) or C1 V H,

and so
u V y H . (8.10)

The shear stress found from the Newtonian constitutive relation is uniform across
the layer:
u
yx V H . (8.11)
y

The average velocity between the two plates is found by integrating the velocity
profile across that gap.
1 H
u u dy (8.12)
H 0

H
1 H y V H V 1 2 V
u V dy y dy y (8.13)
H 0 H H2 0 H2 2 0 2

The mass and volume flow rates between the plates are

m uA V 2 HW and Q uA m V H W 2, (8.14)

where W is the width into the page.

From these relationships, we see that an increased velocity (V) drives higher
shear stress and flow rates. Higher viscosity (μ) increases the shear stress, but does
not affect the velocity field here, which only depends on the boundary conditions.
A thicker layer (higher H) decreases the velocity gradient and so the shear stress, but
also increases the flow rate because of the larger cross-sectional area (H W). Finally,
the linear velocity field gives a uniform shear stress distribution across the fluid layer.
The procedure described earlier is typical for modeling the shear stress and veloc-
ity distributions with Navier–Stokes equations in fully developed flows. The steps for
one-dimensional, steady flows are as follows:

1. Start with the static force balance in Eq. (8.4) and apply approximations
based on the geometry;
2. Substitute a constitutive equation, e.g., Eq. (1.5) or (1.15), for in the force
balance;
3. Integrate the differential equation twice to obtain the general solution for
velocity; and
4. Evaluate the two integration constants from two boundary conditions.

In the following sections, this procedure will be applied to other types of fluid flow.
Internal Flows 311

8.4 PRESSURE-DRIVEN FLOW BETWEEN FLAT PARALLEL PLATES

The Couette flow in Section 8.3, with its linear velocity profile, is driven by shear
due to the plates moving relative to each other, with no slip occurring at the plates. In
contrast, a steady, fully developed fluid flow between two stationary, horizontal flat
plates (Figure 8.2) requires another force to drive the flow. In this case, the impetus
for fluid motion is provided by an imposed uniform pressure gradient. To have a pos-
itive velocity, u > 0, the pressure must decrease in the positive x direction. With this
geometry and a uniform driving force, symmetry dictates that the maximum value
of u occurs at the plane midway between the plates. Because du/dy = 0 at this plane,
from Eq. (1.5) the shear stress, yx, is also zero there. Because of symmetry, it is con-
venient to locate y = 0 at the plane midway between the plates and consider that the
plates are a distance 2H apart (see Figure 8.2).
With these restrictions, the force balance in Eq. (8.4) is reduced to

friction pressure
yx dP (8.15)
.
y dx

Integrating once in the y direction gives

dP
yx y C1* . (8.16)
dx

Assuming a Newtonian fluid gives us an expression for the shear stress,

u dP u 1 dP
y C1* or y C1 .
y dx y dx

FIGURE 8.2 Geometry for flow between parallel horizontal flat plates.
312 An Introduction to Transport Phenomena in Materials Engineering

Integrating again gives the general solution to Eq. (8.15):

1 dP 2
u y C1 y C2 . (8.17)
2 dx

The boundary conditions are no shear at the centerline plane of symmetry,

u
yx (y 0) 0,
y y 0

and no slip at the upper plate,

u( y H) 0.

The integration constants in Eq. (8.17) are found by applying these boundary condi-
tions to Eq. (8.17).

u 1 dP
0 (0) C1 C1 0
y y 0
dx

1 dP 2 H 2 dP
u y H 0 H C2 C2
2 dx 2 dx

The velocity field is then

2
H 2 dP y
u( y ) 1 . (8.18)
2 dx H

As the static pressure P decreases linearly with x, dP/dx is negative and hence
P/L = – dP/dx, where ΔP = PU – PD is the pressure drop occurring over the distance
L between the upstream and downstream positions.
2
H2 P y
u( y) 1 (8.19)
2 L H

The maximum velocity is at the centerline,

H2 P
u(0) umax , (8.20)
2 L

and integrating Eq. (8.19) gives the average velocity over the channel, u :

2 3 H
1 H 1 H H2 P y H2 P y 1 y
u u dy 1 dy
H 0 H 0 2 L H 2 L H 3 H
o
Internal Flows 313

H2 P
u . (8.21)
3 L
This definition of average velocity is such that u 2 H equals the area under the curve
for u(y) between y = -H and y = H (divided by 2H), or, using symmetry, the area
u H is proportional to the area under the curve for u(y) between y = 0 and y = H
(divided by H). Note that the maximum and average velocities just derived are the
same functional form as the reference velocity in the scaling analysis in Eq. (8.6).
These velocities are plotted in Figure 8.3.
The mass flow rate in this situation is

P H2
m uA 2 HW . (8.22)
3 L
The mass flow rate is directly proportional to the pressure drop, which is the potential
field driving motion. It is inversely proportional to (μL), representing the friction in
the fluid and the length over which friction must be overcome. The dependence on
plate spacing is m~ H3, from the average velocity and the cross-sectional area.
The shear stress in the fluid is found from Eq. (8.16):

dP P
yx y y, (8.23)
dx L
and plotted in Figure 8.3. Here we see that the shear stress is a linear function of y,
with a maximum value of (–ΔP H/L) at the plates. The shear is negative because it is
the force restraining the fluid being driven by the pressure gradient.

FIGURE 8.3 Velocity and shear stress fields for fully developed flow driven by a uniform,
negative pressure gradient between two horizontal flat plates.
314 An Introduction to Transport Phenomena in Materials Engineering

Example 8.1
Fluid Flow Model for Tape Casting Ceramic Slurries
As an extended example of the application of the methods in Sections 8.3 and 8.4,
we examine here the tape casting process. Originally devised to make thin sheets of
titanium dioxide for electrical capacitors [1], it is still used to manufacture capacitors
as well as multilayered ceramics for circuits and electronic packages and zirconia
membranes for solid oxide fuel cells [2]. One important aspect of this process is its
ability to orient solid grains in the plane of the sheet. A discussion of the mechanisms
of such alignment is found in [3] and the importance of such a general fiber orienta-
tion distribution (or texture) in the sheet on its magnetic, thermal, piezoelectric, and
structural properties is described in a review by Messing et al. [4]. The basic process
configuration is shown in Figure 8.4, in which an aqueous slurry primarily loaded
with ceramic particles (but also including binders, plasticizers, and dispersants) is
spread onto a moving carrier. The following analysis is taken from Kim et al. [5],
which also includes the effect of beveled blades.
The tape casting process forms a wet tape, which is subsequently dried and fired.
The shrinkage of the tape during drying and firing and its final microstructure are
functions of the final wet tape thickness (Hf) formed in the tape casting process far
from the blade exit.

Find: The final tape thickness, Hf, as a function of fluid properties, geometry,
and process parameters.
Solution: The first step in modeling the fluid flow in this process is to ask:
what moves the slurry from the reservoir to cover the carrier? There are
two driving phenomena: pressure and shear. If the carrier is not moving,
the gauge pressure at the bottom of the reservoir will push the slurry out
under the doctor blade, restrained by friction under the blade and on the
carrier (as in Section 8.4). In the case with a small reservoir height but a

FIGURE 8.4 Schematic of tape casting process. The velocity field under the blade shown
here assumes purely shear-driven flows.
Internal Flows 315

moving carrier, the wet ceramic sheet is formed primarily by the shear on
the lower surface of the tape dragging the slurry out of the reservoir (as in
Section 8.3).

The force balance for the steady one-dimensional flow under the blade on a hori-
zontal carrier is Eq. (8.4), with shear and pressure as the only important terms.

P
y x

The pressure gradient is assumed uniform in the flow direction and is based on
the pressure difference under the blade from the reservoir to the exit. Integrating the
force balance once in the y direction gives us an expression for the shear in the flow.

P
y C1*
x

Assuming a Newtonian fluid, this equation can be written in terms of the velocity
gradient:

u 1 P
y C1 .
y x
Integrating once again,

1 P 2
u y C1 y C2 .
2 x

Applying the no-slip boundary conditions at the moving carrier and the stationary
doctor blade gives us the constants of integration:

1 P 2
u 0 V C2 , u h 0 H C1h V ,
2 x
V 1 P
and C1 H.
H 2 x
The velocity field under the blade is
2
H2 P y y y
u y V 1 . (8.24)
2 x H H H

The hydrostatic pressure at the bottom of the reservoir is PU = Patm + ghs and at the
blade exit is PD = Patm. The pressure gradient is then

P PU PD Patm Patm ghs ghs

x L L L
316 An Introduction to Transport Phenomena in Materials Engineering

and the velocity field as a function of height is

2
ghs H 2 y y y
u y V 1 .
2 L H H H

The characteristic velocity for the pressure-driven component is the coefficient of the
first term, V
P
gh H / 2 L , and for the shear-driven term is VS = V. Nondimen-
s
2

sionalizing our expression for velocity, we obtain an expression for velocity field:

u 2
1 ,
V
pressure shear
where = y/H and the parameter, , is the ratio of the pressure and shear character-
istic velocities:
VP ghs H 2 1
.
VS 2 L V

If << 1, then the shear forces dominate the flow and the solution is the Couette
flow in Section 8.3; only the pressure-driven flow solution in Section 8.4 remains
if >> 1. Figure 8.5 illustrates the transition in velocity field from the Couette to
pressure-driven flow solutions as goes from small to large values.
These velocity profiles are present up to the blade exit, at which point the top
boundary condition switches from zero velocity (no slip) to zero shear (a free sur-
face). Because the only shear on the fluid is then from the carrier, the momentum
imparted by that shear quickly diffuses upwards through the wet tape until it has a
uniform velocity, u(y) = V, far downstream from the blade. We know then that the
mass flow rate of the ceramic slurry in the downstream wet tape is

m V HfW ,

where W is the width of the tape. This mass flow rate is the same as what comes out
from under the blade at the average velocity, u

m u HW V HfW .
at exit far downstream

Equating these mass flow rates, we see that the ratio of the average velocities is
the ratio of tape thicknesses:

u Hf
.
V H

The average velocity under the blade, u , is found from the integral over the tape
thickness there:

1 H g hS H 2 V
u u dy .
H 0 12 L 2
Internal Flows 317

FIGURE 8.5 Velocity profiles under the doctor blade for different flow regimes. (a) Shear-
driven and mixed flows, ≤ 1. As approaches zero, the profile becomes purely Couette
flow (Section 8.3). (b) Pressure-driven flows, >> 1, where the profile at high is a close
approximation of the parabolic profile in Section 8.4.

Written in nondimensional form, the average velocities far downstream and at the
blade exit (the last two equations) are set equal to each other:

u 1 Hf
.
V 6 2 H
Figure 8.6 shows this function and how it asymptotes to the shear-dominated solution
at low values of Π and to pressure driven at Π >> 1.
318 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.6 Nondimensionalized average velocity at blade exit and final wet tape thick-
ness ( = u /V = Hf/H) as a function of the pressure to shear ratio, .

FIGURE 8.7 Effect of flow on fiber orientation during tape casting in a Π << 1 case.
Schematic is based on results reported in [6].

Using these relations, we can find a dimensional expression for the final tape
thickness:
g hS H 3 H
hf .
12 LV 2

Effect on tape structure: The previous analysis provides a simple model for esti-
mating the wet tape thickness in tape casting, but also can be used to gain insight on
the texture of that wet tape. That orientation develops due to the process fluid flow,
but changes little in downstream processes (e.g., drying, firing). In the case in which
the solid ceramic particles are high aspect ratio fibers (ℓ/d >> 1), as in Figure 8.7,
Internal Flows 319

the velocity gradients rotate the fibers so that their long axes align with the direction
of the flow. The drag on each part of a fiber in the flow direction is proportional to the
local velocity squared. In a uniform velocity field, the drag is uniform and the fiber
is carried with the flow without change in orientation. In a nonuniform velocity field,
the drag varies along the length of the fiber, which variation imparts a moment to the
solid, the net effect of which is rotation of the fiber. That moment and rotation ceases
when the fiber is stably aligned with the local flow. This effect can be seen in a sche-
matic representing an experiment with fibers in transparent corn syrup (Figure 8.7).

8.5 FLUID FLOW IN A VERTICAL CYLINDRICAL TUBE

Flow in a cylindrical tube is another common configuration, and we use the same
approach as in the previous section to find a solution for flow rates and shear stress.
To illustrate the effect of gravity in non-horizontal tubes, this mechanism is also
included. In Figure 8.8, a vertical tube is shown, in which the sign convention is
that a positive axial velocity, w > 0, is upwards against gravity in the positive z
direction. The fluid velocity in contact with the inner tube wall (r = R) is zero and,
by symmetry, w has its maximum value at the tube axis (i.e., at r = 0). The flow is
driven by an imposed pressure gradient along the pipe and is resisted by the shear
stress, rz, arising from friction at the no-slip boundary at the outer wall; gravity
may assist or resist motion depending on the flow direction. We assume a steady,
fully developed (∂/∂z = 0) flow, and Figure 8.8 shows the pipe in a cylindrical coor-
dinate system (z, r, ). If the flow is also axisymmetric (∂/∂ = 0), the continuity
equation (7.14) is

1 ru 1 v w
0 or ru C, (8.25)
r r r z

So the radial velocity u = 0 everywhere, given the no-slip condition at the outer sur-
face (r = R). These results and simplifications are applied to the z direction momen-
tum equation in cylindrical coordinates, Eq. (7.40):

w w v w w
u w
t r r z
(8.26)
2 2
p 1 w 1 w w
r gz
z r r r r2 2
z2

or

p w
0 r gz
z r r r (8.27)
pressure friction gravity
320 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.8 The geometry for a one-dimensional flow in a cylindrical tube.

Multiplying by and integrating over r,

w 1 dp 2
r g r C1 . (8.28)
r 2 dz
The integration constant C1 = 0 because symmetry gives ∂w/∂r = 0 at the centerline.
For a Newtonian fluid,

w 1 dp
g r. (8.29)
r 2 dz
A second integration then gives

1 dP 2
w g r C2 . (8.30)
4 dz
Applying the no-slip condition, w(r = R) = 0,

1 dP 2
C2 g R (8.31)
4 dz
Internal Flows 321

and thus
2
R2 dP r
w g 1 . (8.32)
4 dz R

Defining the uniform pressure gradient in terms of the pressure difference ΔP =

PU – PD over a length of pipe L,

2
R2 P r
w g 1 . (8.33)
4 L R

(Note that our sign conventions in Figure 8.8 has the upstream pressure at the bottom
and the downstream at the top.) The axial velocity thus has a parabolic shape, as
shown in Figure 8.9, which also shows the linear variation of rz across the diameter
of the tube. The maximum value of w, which occurs at r = 0 is

R2 P
wmax g . (8.34)
4 L

The average value of the flow velocity is defined as

2 R R2 P R r2
w R2 w dA w r dr d g 1 rd
A 0 0 2 L 0 R2
R
R2 P 1 2 1 r4 R2 P R2
g r g .
2 L 2 4 R2 0
2 L 4

Finally,

R2 P
w g , (8.35)
8 L

which is half the value of wmax. The volume flow rate Q is then

R4 P
Q w R2 g ; (8.36)
8 L

this relationship is known as the Hagen–Poiseuille equation. The mass flow rate is

R4 P
m w R2 g . (8.37)
8 L
322 An Introduction to Transport Phenomena in Materials Engineering

If we nondimensionalize the profiles of axial velocity, Eq. (8.33), and shear stress,
Eq. (8.29), we obtain these forms:
2
w 1 r
w* 1 1 (8.38)
gR 2 4 R

and
1 r
T* rz
1 , (8.39)
gR 2 R
where
P
,
gL

a ratio of the applied pressure drop along the tube and the hydrostatic head. With
these expressions, plotted in Figure 8.9, we can see that the shear stress and velocity
profiles are only functions of the nondimensional radius, r/R, and .
In this form, it is easy to see that there is no flow (w = 0) and so no shear ( rz = 0)
when = 1, or the imposed pressure drop is positive (ΔP > 0) and the same absolute
value as the hydrostatic head. In that case, the pressure drop is imposed to keep the fluid
from draining through the tube. If the pressure difference increases (larger ), we see
in Figure 8.9 that the velocity increases as ( −1) and the shear stress opposing the flow
changes with (1− ). The reverse is true as the pressure drop is imposed so that PD > PU
(ΔP and < 0) and is steadily more and more negative. If | | >> 1 and so −1 ≈ , then
the hydrostatic contribution is negligible, regardless of the flow direction.

Example 8.2
Introducing Chlorine Gas into a Metal Bath
Zinc is removed from liquid lead by reaction with chlorine gas to form ZnCl2. Fig-
ure 8.10 shows the chlorine gas being bubbled gently through a bath of liquid lead
at 400 oC. The graphite bubbling tube has an internal diameter of 1.5 mm and is
immersed to a depth of 1 m in the lead. The atmospheric pressure at the liquid surface
is 101.3 kPa. At 400 oC,

Pb
10, 560 kg/m 3 Cl2
2.87 10 5 kg/ms.

Find: The static pressure in the tube at the level of the surface of the bath
required to give a bubbling rate of 0.2 kg Cl2/hour.
Solution: The first step is to write the proper relationship among the material
properties, process geometry, desired flow rate, and required pressure at the
top of the tube. Assuming a laminar flow, Eq. (8.37) gives that relationship,
here rearranged as
8m L
Ptop Pbottom g L,
R4
Internal Flows 323

FIGURE 8.9 Radial profiles of nondimensional (a) axial velocity and (b) shear stress in a
cylindrical pipe, over a range of imposed pressure drop-hydrostatic head ratios ( ).
324 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.10 Schematic of system for bubbling chlorine gas through liquid lead to capture
zinc dissolved in the bath.

where the pressures are at the top and bottom of the inside of the tube. To push
the chlorine out of the tube, the pressure needed at the top (Ptop) is a sum of the
three terms on the right-hand side of this equation, representing pressure at the
exit, the pressure drop due to friction in the tube, and the weight of the column
of gas.
The pressure at the bottom of the tube is the hydrostatic pressure under the weight
of the lead to the depth of the tube:

Pbottom Patm Pb gL 101, 000 Pa 10, 600 kg/m 3 9.81 m/s2 1 m

Pbottom 205, 000 Pa.

To find the value of the next two terms, the density of the gas is needed. Treating
chlorine gas as an ideal gas of molecular weight (MW) 0.07091 kg/gmol, the density
of the gas is

kg
205, 000 Pa 70.91
1 kg P MW kgmol
2.60 kg/m 3 .
V RT kg m 2
8.314 2 673K
s K kgmoll
Internal Flows 325

Plugging these values into the previous expression for the top pressure gives:

kg Cl 2
8 0.2 2.87 10 5 Pa s 1 m
hr
Ptop 205,000 Pa
s kg
3600 2.60 3 (0.00075m )4
h m
kg m
2.596 9.81 1m .
m3 s2

Note that the sign of the mass flow rate is negative, as the mean velocity is in the
negative z-direction.

Ptop 205, 000 Pa 4, 900 Pa 25 Pa = 210, 000 Pa

So the pressure needed at the top of the tube is a little over two standard atmo-
spheres. Most of this pressure is required to overcome the high pressure due to the
hydrostatic head in the lead at the tube exit (205 kPa). The pressure drop in the tube
(4.9 kPa) and the effect of the weight of the gas (0.025 kPa) are very small in com-
parison due, respectively, to the low viscosity and density of the gas compared to
liquids. This result should be expected because the quantity ≈ 8200 in this case,
showing that the hydrostatic term in the tube is negligible compared to the pressure
drop required by the lead.
We have assumed that the analysis in this section applies to the conditions in this
example, particularly that the flow is laminar. For laminar flow in a tube, the Reyn-
olds number based on the diameter and the mean velocity must be below approxi-
mately 2100. Writing ReD in terms of the mass flow rate and calculating its value,

wD 4m 4 0.2 kg/hour
Re D 1640,
D 3600 s/hour 0.0015m 2.87 10 5 kg/ms

we see that the flow is indeed laminar.

8.6 CAPILLARY FLOWMETER

In the preceding derivations, fluid flow was considered between horizontal plates,
and in a vertical cylindrical tube. In both cases, both the shear stress and velocity
of the fluid are proportional to ΔP/L – g cos . In the case of horizontal flow, = 90°
and only ΔP/L appears in the expressions for and w; in the case of inclined flow, ≠
90° and cos ≠ 0, so the expressions for and w contain both terms.
The significance of the terms ΔP/L – g cos can be appreciated from an exami-
nation of the capillary flowmeter shown in Figure 8.11. This simple device is a tube,
narrowed over a length L to a capillary tube of inside diameter d inclined at an angle
to the vertical, with a U-tube manometer containing a fluid of density m, placed
across the length of capillary. The diameter of the capillary section is much less than
326 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.11 Schematic of capillary flowmeter.

the diameter of the rest of the tube so that practically all the resistance to flow over
the manometer is due to friction there. When a second fluid of density f, immiscible
with the manometer fluid, flows through the capillary tube, a difference in height h
occurs between the levels of the fluid in the two arms of the manometer.
If Po is the static pressure in the flowing fluid at the entrance to the capillary and
PL is the static pressure in the fluid at the point of exit from the capillary, the pressure
at the level x at the bottom of the manometer, Px, is found from consideration of the
left arm of the manometer,

Px pressure due to the weight of the column of manometer flluid of height h1 h

pressure due to the weight of the column of second fluid of height h2 PL

h1 h m g h2 f g PL .

From consideration of the right arm of the manometer, that same pressure can be
written as
Px pressure due to the weight of the column of manometer flluid of height h1

pressure due to the weight of the column of second fluid of height

h h2 Lcos Po

h1 m g h h2 Lcos f g Po .
Internal Flows 327

Equating these expressions for pressure at the bottom of the manometer and rear-
ranging gives

PL Po h m f g L cos f g,

and hence the pressure difference measured at the manometer, h ( m – f) g, is

h m f g PL Po L cos f g

or
h m f g P
f cos g. (8.40)
L L

(The pressure difference is defined as in Eq. (8.33), where PL = PD and Po = PU.)

Thus, the pressure difference measured at the manometer includes the decrease
in the static pressure across the capillary and the contribution due to the influence of
gravity on the fluid, and consequently, use of the flowmeter does not require knowl-
edge of the angle of inclination of the capillary.

Example 8.3
Measuring Volume Flow Rate Using a Capillary Flowmeter
Consider the use of the capillary flowmeter shown in Figure 8.11 to measure the vol-
ume flow rate of water. The manometer fluid is an organic liquid, which is immiscible
with water. When the water flows through the capillary (L = 0.1 m and d = 0.005 m),
the height difference between the levels of the manometer fluid is h = 0.030 m.
The thermophysical data are

m 1154 kg/m 3 f 998 kg/m 3 f 9.44 10 4 kg/ms.

Find: The volume flow rate of water, Q.

Solution: The pressure difference measured by the height difference in the
U-tube, h, is

kg m kg
m f gh 1154 998 9.81 0.03 m 45.9
m3 s2 m s2

and, using Eq. (8.40),

m f gh kg 1 P
45.9 459 kg / m 2 s2 f g cos .
L m s2 0.1 m L

From Eq. (8.36), modified to include the inclination angle (g cos for the inclined
pipe rather than g for a vertical one), the volume flow rate is
328 An Introduction to Transport Phenomena in Materials Engineering

4
P R4 kg 0.0025 m
Q f g cos 459 2 2
L 8 ms 8 9.44 10 4 ms/kg

Q 7.46 10 6
m 3 /s.

The use of this relation for flow in the capillary tube assumes laminar flow. To con-
firm this assumption, we find the Reynolds number based on the average flow veloc-
ity. That velocity is

6
Q 7.46 10 m 3 /s
w 0.38 m/s,
R2 0.0025m
2

and Red is

f wd 998 kg/m 3 0.38 m/s 0.005m

Re d 2008.
f 9.44 10 4 kg/ms

Because Red < Red,trans (≈ 2100 for flow in a tube), we see that the flow is indeed laminar.

8.7 NON-NEWTONIAN INTERNAL FLOWS

So far we have concentrated our effort on Newtonian fluids with their simple con-
stitutive relation between shear stress and velocity gradient (or strain rate). In this
section, we explore the flow of non-Newtonian fluids (introduced in Section 1.3.1.2)
in some of the simple, one-dimensional geometries in previous sections. The first
two examples are with power-law fluids and the third with a Bingham plastic. More
such solutions are found in books on rheology (e.g., [7, 8]) and in some transport
phenomena texts [9, 10].

8.7.1 sHear-driven Flow oF a power-law Fluid

We start with the same conditions as the shear-driven flow of a Newtonian fluid
between flat parallel plates (Section 8.3). Figure 8.1 shows the geometry and bound-
ary conditions and the appropriate static force balance is

0. (8.8)
y
For a power-law fluid,
n
u
0 k , (8.41)
y y y
Internal Flows 329

which, when integrated over the layer thickness, is

n n
u u
C1* k dy k . (8.42)
y y y

1/ n
C1* u
C1
k y
Integrating again gives

u C1 y C2 ,

and applying the boundary conditions of no slip on the bottom plate and u = V on
the top,
y
u V . (8.43)
H

This velocity solution is plotted in Figure 8.1 and is the same as for Newtonian flu-
ids, Eq. (8.10), as are the mass and volume flow rates, Eq. (8.14). The shear stress is
different:
n
V
yx k . (8.44)
H

For shear thinning fluids, the shear is less sensitive to V and H than in Newtonian
fluids; for shear thickening, the opposite is true.

8.7.2 pressure-driven Flow oF a power-law Fluid

Once again we revisit a solution from Section 8.4, in which we use the power-law
constitutive relation, Eq. (1.15), to find the flow driven by a uniform pressure gradient
between two stationary, parallel flat plates. For a power-law fluid, the force balance

dP
(8.15)
y dx
becomes
n
u dP
k . (8.45)
y y dx

In the Couette flow example in Section 8.7.1, the velocity gradient is positive, so its
presence under the non-integer exponent in Eq. (8.45) is not a mathematical diffi-
culty. However, in this case, the velocity gradient is negative for y > 0. To address this
issue, we rewrite Eq. (1.15) as
330 An Introduction to Transport Phenomena in Materials Engineering

n 1
u u u
yx app k (8.46)
y y y

so the shear stress and velocity gradient have the same sign (as expected) and the
apparent viscosity is always positive. Eq. (8.45) can be written anew as

n 1
u u dP
k ,
y y y dx

which integrates to
n 1
u u dP
k y C1* .
y y dx

Because the velocity gradient is negative for y > 0,

u u
y y

and
n 1 n
u u u dP
k k y C1* . (8.47)
y y y dx

Applying the centerline symmetry condition

u
0
y y 0

to Eq. (8.47) gives C1* = 0. Rearranging Eq. (8.47) gives

1 1
1 1
u 1 dP n 1 P n
y n
yn ,
y k dx k L

which after integration produces

1
1 n
1 P n n
u y n
C2 .
k L n 1

The integration constant is found by applying the no-slip condition at y = H, and the
complete velocity solution is
Internal Flows 331

1 n
1/ n 1 n
1 P n y n
u H n
1 . (8.48)
k L n 1 H

The average velocity across the layer and the mass flow rate are:
1/ n 1 n
1 P n
u H n (8.49)
k L 2n 1
and
1/ n 1 2n
1 P 2n
m W H n , (8.50)
k L 2n 1

where W is the width of the channel into the page and the cross-sectional area is
(2HW).
Eq. (8.50) gives us the relationship between the driving force, ΔP, and the flow
rate, m:

m ~ P1/ n .

For a shear thinning fluid (n < 1), the exponent on ΔP is greater than one, so the mass
flow rate increases faster than any change in pressure drop; that is, the fluid becomes
progressively easier to push through the channel as the flow rate increases. The oppo-
site is true with shear thickening fluids (n > 1).
If we define a normalized velocity as

1 n
u 1 2n y n
1 , (8.51)
u 1 n H

we can examine the effect of shear thinning and thickening behavior on the velocity
profile shape compared to the Newtonian case in Figure 8.3. Velocity profiles for a
range of values for the flow behavior index of power-law fluids from n = 0.1 to n = 10
are compared in Figure 8.12. As n decreases from the Newtonian case (n = 1), the
profile at the centerline is more and more blunted and the velocity gradient steepens
near the walls. With the highest shear stress at the walls, μapp is lowest there, whereas
the relatively low stress around y = 0 results in a high viscosity and low velocity gra-
dient there. The opposite effect is seen in a shear thickening fluid.

Example 8.4
Finding the Pressure Drop Necessary to Extrude a Polymer through a Circular Die
Most of the pressure drop in a polymer extrusion process is in the narrowest pas-
sage for a fluid, that is, in the die through which the polymer leaves the extruder. In
this case, we are interested in extruding a high-density polyethylene (HDPE) fiber
332 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.12 Velocity profiles of power-law fluids between parallel flat plate driven by
uniform pressure gradient over a range of flow behavior index values (0.1 < n < 10).

FIGURE 8.13 Polymer extrusion through and exiting a circular die.

through the circular die shown in Figure 8.13. Assuming a fully developed flow in the
die and treating HDPE as a power-law fluid, we would like to know the pressure drop
over the die length L from point 1 to 2 necessary to maintain a volumetric flow rate of
Q = 10–4 m3/s. For comparison to the viscosities of more common fluids (e.g., water),
it is interesting to calculate the apparent viscosity profile in the flow.

R 0.005m L 0.025m n 0.4 k 20 kPa s

In this cylindrical geometry, the static force balance is

P 1
r rz
z r r
or, for a power-law fluid,
n
P 1 w
rk .
L r r r
Internal Flows 333

The solution, with no slip at r = R and symmetry at r = 0, is

1 1 n
1 n
n 1 P n r n
w R n
1 .
1 n 2k L R

Find: (a) The pressure drop to maintain the specified flow rate and (b) apparent
viscosity as a function of radial position.
Solution: (a) The volumetric flow rate is related to the pressure drop through the
average velocity across the die, so we begin by finding an expression for w.
1
2 R 1 n
1 n 1 P n
w w r dr d R n
R2 0 0
1 3n 2k L

The volumetric flow rate is then

1
1 3n
2 n 1 P n
Q w Ac w R R n
.
1 3n 2k L

Rearranging this equation for the pressure drop,

n
n 1 3n 2k L
P Q .
n R1 3n

Using the data given, we can calculate the pressure drop:

0.4 0.4
4 m3 1 3(0.4) 2 20,000 Pa s0.4 0.025m
P 10 1 3( 0.4)
s (0.4) 0.005m
m1.2 Pa s0.4
0.0251 1.25 1.15 108 3.63 MPa = 36.3 bar.
s0.4 m1.2

This result is useful not only for selecting an extruder that can generate that pressure
but also for selecting die material and designing the fixture to withstand this internal
pressure.

(b) The apparent viscosity is a function of the velocity gradient in the die:
1
1
1 6 0.4 1
w 1 P n 1 3.63 10 Pa
r n
r 0.4
r 2k L 2 20,000 Pa s0.4 0.025m
1

8 r 0.4 1
7.94 10 .
m s
334 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.14 Variation of apparent viscosity and velocity gradient for flow of HDPE
through a circular die in Example 8.4.

The apparent viscosity for a power-law fluid is defined by Eq. (8.46):

1 0.4 1
n 1
w 0.4 8 r 0.4 1
app k 20, 000Pa s 7.94 10
r m s
1.5
r
0.0914 Pa s .
m

Figure 8.14 shows both the velocity gradient and apparent viscosity across the die.
The gradient is highest at the outer wall (r = R), where the viscosity is lowest, as we
expect in a shear thinning fluid. As r 0, the velocity gradient goes to zero due to
the system’s symmetry, which causes the apparent viscosity there to approach infin-
ity. As noted in Chapter 1, the flow field is predicted well at the wall, but is less good
near the center line, where the viscosity really should approach the zero shear viscos-
ity. Because the shear is so small in this region, the use of a power-law model in this
case has little effect on the important result (the required pressure drop).

8.7.3 pressure-driven Flow oF a bingHaM plastic

Another type of non-Newtonian fluid requires a minimum critical shear stress ( c)
below which no deformation occurs and above which it flows according to one of
the many models presented in Chapter 1. A Bingham plastic is a fluid that does not
deform below c and behaves like a Newtonian fluid above that critical shear stress
[8]. For one-dimensional flows,

u
yx c for yx c . (8.52)
y
Internal Flows 335

and

u
0 for yx c . (8.53)
y

Some fluids have a nonlinear response past the critical shear stress, and the Her-
schel–Bulkley model may be used [8]:
n
u
yx c k for yx c , (8.54)
y

The approach to solving flow fields in fluids with a minimum critical stress for motion
is the same as previous cases, but with a few complications. Taking pressure-driven
flow between the two flat parallel plates (a distance 2H apart) as an example (Fig-
ure 8.15), the motion of a Bingham plastic is divided into two sections, one with high
enough shear stress to cause deformation and one in which there is no relative motion
(slug flow). Starting with the familiar static force balance for this configuration,

yx P
, (8.15)
y x

the constitutive relation in Eq. (8.52) is inserted and the procedure developed in pre-
vious sections is applied. Integrating twice and applying the no-slip condition at the
walls (y = ±H) and zero shear at the centerline (due to symmetry), gives the velocity
field in the sheared region as
2
1 P y y
u( y) H2 1 c
H 1 , (8.55)
2 L H H

FIGURE 8.15 Nondimensional velocity profiles for the flow of Bingham plastic between
parallel flat plates, showing regions of deformation (|y| > Hc) and slug flow (|y| < Hc).
336 An Introduction to Transport Phenomena in Materials Engineering

and the velocity at the critical shear is

2
1 P Hc Hc
uc H2 1 c
H 1 . (8.56)
2 L H H

This velocity is uniform in what is called the slug flow region around the centerline,
|y| ≤ Hc, where the position of the transition between the two regimes is

P L
Hc c c . (8.57)
dx P

In the low stress, slug flow region where |y| < Hc, there is no relative motion, so the
velocity has uniformly the value from Eq. (8.56). Eq. (8.58) shows the normalized
velocity profiles ( ) vs. the normalized height ( ),

1 2
1 c 1 where 1 c (8.58)
2

and the core velocity, c, is given by Eq. (8.59):

1 2
c 1 c c 1 c where c . (8.59)
2

The normalized variables are defined by

L y Hc c L
u , , and c . (8.60)
P H2 H H H P

Figure 8.15 shows the steep velocity gradients near the walls, where the shear stress
is high enough to initiate deformation. The flat regions near the center indicate stress
lower than critical.
When c is high, the pressure gradient driving the flow is weak compared to the
shear required to begin deformation, so the slug flow covers most of the channel and
the maximum flow velocity is low. A flow with lower values of c is easier to deform
because the pressure gradient is stronger relative to the critical shear stress. In that
case, more of the channel has deforming fluid and the maximum velocity increases.
As c 0, the solution approaches the behavior of a Newtonian fluid derived in
Section 8.4.

8.8 FLOW THROUGH POROUS MEDIA

8.8.1 resistance to Flow
A porous medium is a rigid, solid structure with a complimentary, continuous system
of void space through which a fluid may move. Interest in flow through such structures
Internal Flows 337

began with the study by Henry Darcy of flows in the water system in Dijon, France
[11], and has found wide application in the field of groundwater flows. Applications
in materials processing include infiltration of metal into ceramic preforms or poly-
mers into mattes of carbon fiber (to make metal-matrix or carbon-fiber-reinforced
composites, respectively), the movement of liquid metal through dendritic arrays
during alloy solidification, and the flow of water into a mold during slip casting of a
ceramic slurry.
The structure of the solid in a porous medium, or packed bed, may be as regular
as uniformly sized spheres or fibers, or may be made of particles of much more irreg-
ular shapes and over a distribution of sizes. The structure may also be made of such
particles and then sintered to give it some rigidity. Whatever the size and shape of the
solid and the voids, we assume that their length scale is much smaller than the size
of the entire bed (as in Figure 8.16). Also, the porosity of the bed is continuous along
its length and the structure is evenly distributed over the cross-section so that there is
a spatially uniform resistance to flow. The latter feature prevents a preferential flow,
or channeling, in some areas of the bed. These assumptions are made to describe the
simplest of porous media.
An example of a cross-section of a bed is shown in Figure 8.16, which has an area
normal to the bulk flow of Abed, there is a much smaller area through which the fluid
can flow (the cross-sectional area of the voids, Avoid). The void fraction of the bed,
, is defined as

Avoid Abed . (8.61)

Measurement of the volume flow rate of fluid through the bed, Q, allows the calcula-
tion of two average velocities: the superficial velocity, us , calculated as

us Q Abed

and the actual average velocity of the fluid through the voids, u ,

u Q Avoid .

From Eq. (8.61) the two velocities are related as

Q u Abed us Abed

or

u us . (8.62)

Given the geometric complexity of a packed bed, with its many twisty passages
mostly alike, we immediately see that the motion of the fluid in it will be much more
complicated than the relatively simple flows discussed earlier in this chapter. The
338 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.16 Example of structure of porous medium, showing the difference in length
scale of the entire packed bed (on left) and the detailed structure (on right), assuming
ℓ/L << 1.

fluid will flow through the bed’s pores, constantly changing direction and speed at
small length and time scales. The question arises, how much detail is required to
model and understand the behavior of the fluid? As usual, the appropriate level of
modeling depends on its purpose. In many engineering applications, it is sufficient to
consider only the gross behavior of the flow, and to ignore the details at the smallest
scale. A distinguished researcher in the field has written about this approach that “the
decision to assume away (to smooth out) the complicated features of the actual phe-
nomena is in fact an admission of defeat in the face of Nature” [12]. While an under-
standing of the flow at the fine, microstructural scale is interesting and in many cases
useful, here we will satisfy ourselves with coarse models of the system as a whole
and refer the reader to more advanced references when interest or need arises [e.g.,
13]. In choosing this path, we are not admitting defeat, but resisting mission creep
(however interesting) and focusing on simpler models that frequently are enough to
obtain information useful for system design and performance prediction. Our models
of Nature should only be as complicated as our purpose for making them requires.
To predict the fluid behavior in a porous medium, we treat the resistance to flow
as a volumetric drag force, operating uniformly throughout the bed and not just at
the no-slip boundaries. To start, experiments have shown that the average velocity
through a bed is directly proportional to the pressure gradient along the flow direc-
tion and inversely proportional to the fluid dynamic viscosity:

1 P
us ~ .
x
Introducing the permeability of the porous medium, K, with units m2, as a propor-
tionality constant, we can write what is known as Darcy’s law:

K P K P
us . (8.63)
x L
Internal Flows 339

This expression can be rearranged to show a balance between the driving pressure
gradient and the resisting volumetric drag force in the packed bed:

P us
FD . (8.64)
x K

This volumetric drag force, FD, which is known as Darcy drag, can be added to the
steady, fully developed x momentum equation (8.3):

us P us
g .
t y x K (8.65)
local shear pressure gravitty Darcy drag
acceleration

Here we assume that us is the local velocity, averaged over a small volume in the
entire bed (but still much larger in scale than the fine structure). This representation
allows the local velocity, us (y), to vary with position based on the effect of conditions
at the boundary of the porous medium. (It also is more general, in cases not consid-
ered here, with spatially varying permeability or other body forces.)
As an example, we examine the problem of a steady, one-dimensional, pres-
sure-driven flow between two parallel, horizontal plates; this configuration is the
same as in Figure 8.2, except that the domain is filled with a porous medium of uni-
form permeability. Writing the forces resisting flow on the left and driving the flow
on the right, the static force balance is

us P
g.
y K x (8.66)
shear Darcy drag pressure gravity

First, we perform a scaling analysis on this static force balance to obtain estimates
of average velocity and for conditions under which different phenomena may be
neglected. Assuming a Newtonian fluid and no gravity effects, the orders of magni-
tude of the terms in Eq. (8.66) may be estimated as

uo uo
, ~ Px .
H 2
K (8.67)
shear Darcy drag pressure

We have chosen uo as the unknown reference velocity, H as the bed-level length scale
in the y direction, and Px as the pressure gradient in the flow direction. Because the
pressure is the only effect driving the flow, it is always important. Dividing Eq. (8.67)
by Px gives
uo uo
2
, ~ 1.
H Px K Px (8.68)
shear Darcy drag pressure
340 An Introduction to Transport Phenomena in Materials Engineering

There are two extremes for this force balance: the first has (shear) ~ (pressure) and
the second (drag) ~ (pressure). For the first case (without Darcy drag),

uo
shear ~ 2
~ 1 ~ pressure. (8.69)
H Px

From Eq. (8.69) we have

H 2 Px
uo ~ , (8.70)

which is expected because, with no Darcy drag, this case should have the solution
found in Eq. (8.18) from Section 8.4. Substituting Eq. (8.70) into the drag term, we
have the conditions under which the drag in the bed can be safely neglected com-
pared to shear at the boundaries:

uo H2
Darcy drag ~ ~ 1, (8.71)
K Px K

which requires a very large permeability compared to the size of the channel.
In the second case, the shear in the fluid layers is negligible compared to the vol-
umetric drag caused by fluid moving through the porous medium.

uo
Darcy drag ~ ~ 1 ~ pressure (8.72)
K Px

The estimate for velocity is then

K Px
uo ~ , (8.73)

which is valid when the shear is negligible:

uo K
shear ~ ~ 1. (8.74)
H 2 Px H2

This extreme holds when the permeability is very small. Eq. (8.66) is reduced to Dar-
cy’s law, Eq. (8.63), which gives the velocity directly, without solving a differential
equation. The velocity is spatially uniform because it is entirely controlled by Darcy
drag in the packed bed; the effect of the no-slip walls at y = ± H is confined to a very
small region.
Again, beginning with a rearrangement of Eq. (8.66) for a Newtonian fluid,

2
us us H 2 dP
, (8.75)
y2 K dx
Internal Flows 341

we use = us /uo, where uo is defined by Eq. (8.70), and = y/H to nondimensionalize

the force balance. Eq. (8.75) is thus reduced to

2
D2 1 D2 H 2 K , (8.76)

and the boundary conditions are symmetry at the centerline and no slip at the
boundaries:

us
y 0 0 and us y H 0 1. (8.77)
y y 0 0

The general solution to Eq. (8.76) is

C1 e D C2 e D
C3 ,

and the complete solution, plotted in Figure 8.17, after applying the boundary con-
ditions is

1 eD e D
1 . (8.78)
D2 eD e D

FIGURE 8.17 Nondimensional velocity profiles, Eq. (8.78), in a parallel plate channels
filled with porous media of various D H K . At low D, the porous medium has almost
no effect and the solution matches the simple parallel channel solution (dotted line); at high
D, the profile approaches slug flow with the effect of the boundary limited to very near the
plates.
342 An Introduction to Transport Phenomena in Materials Engineering

The dimensional version of Eq. (8.78) is

K exp y K exp y K
us y 2
1 . (8.79)
H exp H K exp H K

The trend in the dimensionless velocity, Eq. (8.78), plotted in Figure 8.17 over a
range of D, is that suggested by the scaling analysis. With D H K << 1, the perme-
ability is very high and the bed offers scant resistance to flow, so the solution matches
the parallel plate result of Eq. (8.18). With D H K >> 1, the permeability is very
small, Darcy drag dominates the flow resistance, and the velocity profile is very close
to slug flow (a uniform velocity field).

8.8.2 eFFect oF porous Media structure on Flow

While we can write force balances and their solutions for the macroscopic velocities
in a porous medium, one aspect that must be addressed is the connection between the
flow resistance (or permeability) as a function of the structure of the medium. The
average axial velocity w along a horizontal pipe (Eq. (8.35) with no gravity term)
is written as
R2 P
w .
8 L
This relation can be adapted to handle arbitrarily shaped passages thus

Rh2 P
w K1 (8.80)
L
or
Rh2 P
ws K1 , (8.81)
L

where K1 is an empirical constant of proportionality. The hydraulic radius, Rh, of the

passage is defined as the ratio of the void volume (Vvoid) and the solid surface area
wetted by the fluid (S ):
Vvoid
Rh . (8.82)
S
This definition approximates the effect of the porous medium geometry through use
of a simple quantity into which is lumped more complex information about that
geometry.
With a bed of volume Vbed and porosity , the volume of the voids is

Vvoid Vbed
Internal Flows 343

and the volume of the solid is Vs = (1 – )Vbed. Thus, if S’ is the total surface area of
the solid, then S0, the total surface area of the solid per unit volume of solid, is

S
So (8.83)
1 Vbed
and

Vbed
Rh . (8.84)
So 1 Vbed So 1

Substitution of Eq. (8.84) into Eq. (8.81) gives

3
K1 P
ws 2
.
L So2 1

For laminar fluid flow through the packed bed, K1 has been evaluated as 1/4.2, so that

3
1 P
ws 2
. (8.85)
4.2 L So2 1

Equation (8.85) is known as the Blake–Kozeny equation. Comparison with Darcy’s

law in Eq. (8.63), shows that the Blake–Kozeny equation gives us a model for the
permeability,
3
1
K 2
. (8.86)
4.2 So2 1

The permeability is thus determined only by the porosity of the bed and geometry of
the solid structure.
Again, transition from laminar to turbulent flow begins at a critical value of the
Reynolds number, which is defined for packed beds as follows:

wD ws 2 Rh 2 ws
Re . (8.87)
So 1

The length scale used, D = 2 Rh from Eq. (8.84), reflects the size of the passages
through which fluid flows, not the size of the bed. For packed beds, the factor of 2 in
Eq. (8.87) is omitted and a new Reynolds number, Rec, is defined as

ws
Re c , (8.88)
So 1
344 An Introduction to Transport Phenomena in Materials Engineering

where the passage length scale is Rh. Flows are laminar through the packed bed of
Rec < 2 and the transition from laminar to fully turbulent flow occurs in the range
2 < Rec. < 1000. Turbulent flow requires the introduction of the Darcy friction factor,
which is obtained from Eq. (6.9) as

2
D P 4 P
f . (8.89)
L 1 So 1 L ws2
w2
2

For fully turbulent flow, the friction factor has a constant value of 0.292, and for
Rec > 1000,

P 0.292 So 1 ws2
3
. (8.90)
L 4

Eq. (8.90) is known as the Burke–Plummer equation. Note that in this regime, the
drag goes like the square of the more vigorous velocity, compared to the linear drag
of Darcy’s law, Eq. (8.63). The variation of ws with ΔP/L over the entire range of
Reynolds numbers is given by the sum of Eqs. (8.85) and (8.90) as

2
P 4.2 So2 1 ws 0.292 So 1 ws2
3 3
. (8.91)
L 4

Equation (8.91) is simplified by dividing both sides by ws2 1 So 4 3

and com-
bining with Eqs. (8.88) and (8.89) to get

16.8
f 0.292, (8.92)
Re c

which again shows that the friction factor is a function only of the Reynolds num-
ber. Thus, for a required value of ws through the packed bed, Rec is obtained from
Eq. (8.88), f from Eq. (8.92), and ΔP/L from Eq. (8.91). At very low Reynolds
numbers, the term 16.8/Rec in Eq. (8.92) is much larger than 0.292 and thus the
Blake–Kozeny contribution to Eq. (8.91) predominates, and at very large Reynolds
numbers, 16.8/Rec << 0.292, and thus the Burke–Plummer contribution controls
flow behavior.
As an example of a particular packed bed geometry, let us consider one spherical
particle in a bed with uniform radius (R) and diameter (dp), in which the ratio of the
surface area to the volume, So, is

4 R2 3 6
So , (8.93)
4 R3 3 R dp
Internal Flows 345

substitution of which into Eq. (8.91) gives

2
P 151 1 ws 0.438 1 ws2
. (8.94)
L d p2 3
dp 3

Equation (8.94) is known as Ergun’s equation, the overall pressure drop through
a bed of packed spheres. Substitution of Eq. (8.93) into (8.88) and (8.89) gives,
respectively,

ws d p
Re c
6 1

and

2d p P 3
f .
31 L ws2

When working with packed beds of spherical particles, it is convenient to redefine

the Reynolds number and the friction factor:

ws d p
Re E 6 Re c (8.95)
1

4d p P 3
fE 6f . (8.96)
1 L ws2

Dividing Ergun’s equation, Eq. (8.94), by ws2 1 dp 3

then gives

151
fE 0.438, (8.97)
Re E

which is plotted in Figure 8.18.

Consider a perfectly close-packed bed of spherical particles of diameter dp.
Recalling the face-centered cubic crystal structure, the spheres are in contact with
one another along the diagonal of a face of the unit cell. The length of this diagonal
is thus 2d, and the length of the side of a unit cell is 2 d. The volume of the unit
cell (the bed volume) is thus 2 2 d 3. The FCC unit cell contains four spheres, each
3
of volume 4 d 2 3, and thus the bed porosity is

3
2 2 d 3 16 d 2 3
3
0.26,
2 2d
346 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.18 Variation of Darcy friction factor with Reynolds number for fluid flow in a
packed bed of spherical particles.

which is the minimum possible void fraction for packed beds of uniformly sized
spheres. Less well-packed beds have larger porosities, between 0.36 and 0.40 for
spheres large enough to have weak interparticle forces [14].
Real packed beds are usually not made up of uniformly sized spheres; they typ-
ically contain particles that are not necessarily spherical and have a range of sizes.
One of the simplest means of dealing with the variation in particle size is to define
the specific area mean diameter, d p and is used because it is relatively easy to
measure, as

1
dp ,
wt% n
d p,n

where (wt%)n is the weight percentage of the population of particles that contains
particles of diameter dp. In practice, (wt%)n is determined as the weight percentage
of a sample of the bed that is retained between two sieves and the diameter is taken
as being the average of the sieve sizes. Deviation from spherical shape is quantified
by the shape factor . For a particle of volume V and surface area A, the diameter ds
of the sphere that has the same volume is

1/ 3
ds 6V .

The surface area of this equivalent-volume sphere, As, is thus

2
As = d s
Internal Flows 347

and the shape factor of the particles of volume V and surface area A is defined as

A As .

Thus, for a cube of side a,

1/ 3 2/3
ds 6 a3 , As 6 a3 ,
2/3
and 6 a2 6 a3 1.24.

For nonuniformly sized particles,

So 6 dp ,

and Ergun’s equation becomes

2 2
P 151 1 ws 0.438 1 ws2
. (8.98)
L d p2 3
dp 3

Example 8.5
Pressure Drop in Reducing Gas Flow through the Upper Shaft
of an Iron Blast Furnace
The upper section of an iron blast furnace is essentially a packed bed full of spheri-
cal pellets of metallurgical coke, flux, and iron oxide. This section is a cylindrical
shaft through which reducing gas flow ascending very quickly (meters per second)
while the solid descends much more slowly. The temperature is assumed uniform
at 1200 K over much of the shaft. We will assume here that the shaft has a uniform
cross-sectional area, the pellets all have the same diameter, and the volume has a uni-
form distribution of pellets and gas flow. While none of these are particularly good
assumptions, they are reasonable enough for our goal, which is to obtain a first-order
estimate of the overall pressure drop through the section.

for reducing gas : 2 x 10 5 kg/ms 0.2 kg/m 3 ws 3 m/s

for bed geometry : L length 20 m D diameter 10 m dp 0.01m 0.5

Find: (a) The value of K/R2 over a range of pellet diameter, 0.001 m < dp <
0.1 m. Plot K/R2 vs. dp, and comment on the applicability of Darcy’s law
over that range of pellet size. Assume the bed-scale reference length is
R = Dh/2. (b) Find and plot the pressure drop required to push that reducing
gas flow through the section.
Solution: (a) First, let us assume that we are operating in the low Rec regime,
so the Blake–Kozeny model is valid. There the permeability of the bed is
defined by Eqs. (8.86) and (8.93):
348 An Introduction to Transport Phenomena in Materials Engineering

1 2
1 d p2 2

K 2 2
.
4.2 So2 1 4.2 36 1

Figure 8.19 shows the variation of K/R2 over the requested range of pellet sizes.

2 2
K 1 3
dp 1 0.53 dp
d p2 0.000132 m 2

R2 4.2 36 1 2
R 4.2 36 1 0.5 2
5m

For dp = 0.01 m, K/R2 = 1.3 X 10–8 << 1, verifying that the low Rec range is valid
and Darcy’s law and the Blake–Kozeny model are reasonable for this flow.

(b) Given this result, we can find the pressure drop using the Blake–Kozeny
equation, which is Darcy’s law with the permeability defined as a function
of surface area and porosity.
2 2
151 1 ws L 151 1 0.5 2 10 5 kg/ms 3 m/s 20 m
P
d p2 3
d p2 0.5
3

0.363 kg m/s2
.
d p2

The required pressure drop through the upper shaft of this furnace is approxi-
mately 3630 Pa.

FIGURE 8.19 Variation of K/R2 and pressure drop in packed bed in shaft of an iron blast
furnace over the range of pellet sizes.
Internal Flows 349

8.9 FLUIDIZED BEDS

If the flow velocity of fluid passing upward through a packed bed is increased (which
requires a higher pressure drop across the bed), eventually a point is reached at which
the pressure drop times the cross-sectional area of the bed equals the weight of the bed:

P P
s g 1 . (8.99)
x L

At this point, if the particles are not connected mechanically, they are suspended
in the upward flow and the packed bed becomes a fluidized bed. The relationship
between log ΔP and logus is shown in Figure 8.20. The section AB of that curve is
the variation of ΔP with us for Darcy flow through a packed bed. Between points B
and C, the bed expands slightly and rearranges itself to minimize its resistance to
flow, and at C it becomes entirely fluidized. Further increases in us do not influence
the pressure drop across the bed, but do cause an increase in the porosity of the bed,
which means that the fluidized bed expands. Eventually, point D is reached, at which
has increased to unity, and us is the terminal free-fall velocity of each of the parti-
cles in the fluid. Further increase in us causes the particles to be elutriated (i.e., blown
out of the container), and line DE is the variation of ΔP with u in an empty tube.
Point C is the point of minimum fluidization, and the porosity of the bed in this
state, mf, is both the maximum porosity of the packed bed and the minimum poros-
ity of the fluidized bed. At the point of minimum fluidization, combination of Eqs.
(8.98) and (8.99) gives

FIGURE 8.20 Variation of the pressure drop with superficial flow velocity for fluid flow in
a packed bed and a fluidized bed.
350 An Introduction to Transport Phenomena in Materials Engineering

2 2
P 151 us 1 0.438 us2 1
s g 1 , (8.100)
L dP2 3
dP 3

and multiplication of Eq. (8.100) by d 3

P
/ 1
2
gives

2 2
151 1 us dP 0.438 us dP dP3 s g
3 3 2
. (8.101)

The quantity in brackets in Eq. (8.101) is Red for fluid flow past a sphere of radius d p ,

us dP
Re d .

The term on the right-hand side of Eq. (8.101) is a dimensionless number, the value
of which is determined by the size and density of the solid particles and by the phys-
ical properties of the fluid. This ratio is the definition of the Galileo number,
2
dP2 buoyancy
Ga dP s g 2
~ , (8.102)
viscous drag

and thus Eq. (8.101) can be written as

2
151 1 0.438
3
Re d 3
Re 2d Ga. (8.103)

Wen and Yu [15] have determined that, at the point of minimum fluidization,
2
1 mf
3
11 and 3
14. (8.104)
mf mf

The approximate nature of Eq. (8.104) can be seen by considering spherical particles
for which = 1; in that case, the first expression gives mf = 0.38 and the second
mf = 0.41. Thus, at minimum fluidization, we can use a value of 0.4 for the bed
porosity, mf. Substitution of Eq. (8.104) into (8.103) and solution for the Reynolds
number at minimum fluidization gives
1/ 2
Re d , mf 1134 0.0408Ga 33.7 (8.105)

Wen and Yu found empirically that between points C and D in Figure 8.20, the Reyn-
olds number and the porosity of the bed are related as
4.7
Ga 18 Re d 2.70 Re d
1.687
. (8.106)
Internal Flows 351

At point D, the flow velocity is the terminal velocity of a sphere, as shown in Chapter 9:
1/ 2
16 dP s g
ut . (8.107)
3f

The friction factor for flow over a sphere depends on the particle Reynolds number:

Example 8.6
Behavior of a Fluidized Bed
A bed initially of porosity = 0.35, consists of uniformly sized spherical particles of
alumina of diameter dp = 10–3 m. The bed is fluidized by an upflow of air. The air has
= 1.18 kg/m3 and μ = 1.85 × 10–5 kg/ms and alumina has = 3990 kg/m3.

Find: (a) Galileo number; (b) Re, superficial air velocity, and bed porosity at
minimum fluidization; (c) the variation of with us between points C and
D in Figure 8.20.
Solution: (a) From Eq. (8.102), the Galileo number is

3
dP3 s g 10 3 m 3990 kg/m 3 1.18 kg/m 3 9.81m/s2
Ga 2 2
1.35 10 5.
5
1.85 10 kg/ms

(b) From Eq. (8.105), the Reynolds number at minimum fluidization is

1/ 2
Re d , mf 1134 0.0408 1.35 10 5 33.7 47.8.

The superficial velocity of the air at minimum fluidization is obtained from the Reyn-
olds number as
5
Re d 47.8 1.85 10 kg/ms
us , mf 0.750 m/s.
dP 1.18 kg/m 3 10 3 m

The bed porosity at minimum fluidization is obtained from Eq. (8.106) as

1
1
1.687 4.7
18 Re d 2.70 Re 1.687
d
4.7 18 47.8 2.70 47.8
mf 0.435,
Ga 135,000

which is close to the values estimated in Eq. (8.104).

352 An Introduction to Transport Phenomena in Materials Engineering

.
Internal Flows 353

and

1.18 kg/m 3 11.1 m/s 10 3

m
Re d 5
706.
1.85 10 kg/ms

Because Red > 500, Eq. (8.110) is not valid in this case either.
Finally, we turn to the highest Reynolds number range, 500 ≤ Red ≤ 2 × 105, in
which the friction factor is f = 1.76, which, with Eq. (8.108), gives

ut 10.0 m/s

and

1.18 kg/m 3 10.0 m/s 10 3

m
Re d 5
638.
1.85 10 kg/ms

At last we have a self-consistent result, where the calculated Reynolds number is

in the assumed range. With this result, the bed porosity at the point of elutriation is
obtained from Eq. (8.106) as
1
1
1.687
7 4.7
18 Re d 2.70 Re1.687
d
4.7 18 638 2.70 638
1.03.
Ga 135, 000

Because > 1, we set the porosity at its maximum value of = 1. The variation
of with us in the range of 0.75 m/s < us < 10 m/s, is obtained from Eq. (8.106) and
is plotted in Figure 8.21. As in Figure 8.20, minimum fluidization conditions in Fig-
ure 8.21 are marked as point C and the elutriation as point D.

FIGURE 8.21 Variation of fluidized bed porosity with average velocity and Red .
354 An Introduction to Transport Phenomena in Materials Engineering

8.10 SUMMARY
The behavior of simple internal fluid flows is determined by the geometry of the
containing duct, the magnitude of the force causing the flow, and the properties of
the fluid. At low enough flow velocities laminar flow occurs and is characterized by
lamellae of fluid sliding over one another, with no macroscopic mixing of the fluid
in directions normal to the direction of flow. The fluid in contact with the walls of a
containing duct is generally stationary, and velocity gradients are established in the
fluid in directions normal to the flow.
The shear stresses, velocity distributions, and hence average linear velocities and
volume and mass flow rates are determined solutions to the continuity and momen-
tum equations. These governing equations for flow are simplified by the fully devel-
oped and other approximations and then integrated twice to yield the functions of
shear stress and velocity with position in the flow. The two integration constants
are obtained from boundary conditions for the particular type of flow. Applications
shown here include combinations of shear-, pressure-, and gravity-driven flows in
Cartesian and cylindrical conduits. These flows are found in both Newtonian and
non-Newtonian fluids, and in packed and fluidized beds.

8.11 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

8.1 A shaft of outer diameter D is rotating at revolutions per second inside an

outer, cylindrical casing with a thin film of lubricant (viscosity μ and thick-
ness ) in the annulus between the shaft and the casing. (a) Write an expres-
sion that relates the torque, T, required to overcome the friction in the fluid
to D, r, μ, and . (b) Calculate the torque when D = 2.5 cm, = 0.1 mm, r =
1000 rpm, and μ = 3 × 10–3 Pa∙s. (c) Show that Couette flow is a reasonable
approximation.
8.2 The shaft in Problem 8.1 is rotated in the hole, and it is found that a torque
of 1 N∙m is required to sustain a speed of rotation of 60 rpm. (a) Calculate
the viscosity of the lubricant in the annulus between the shaft and the block.
Ignore any end effects. (b) Assume a linear velocity gradient through the
lubricant and show why that approximation is reasonable.
8.3 Two immiscible incompressible fluids are in horizontal layers between two
parallel plates. Both plates are stationary. The two layers between the plates
of length L, width W, and vertical separation 2 are subject to a uniform
pressure gradient dp/dx. Fluid 1 in the lower layer is more dense and more
viscous than fluid 2 in the upper layer, and the two layers have the same
thickness, . (a) Sketch the velocity fields in fluids 1 and 2. (b) Simplify the
Navier–Stokes equations and solve for u1(y) and u2(y). (c) Write expressions
for the mass flow rate in each layer.
8.4 Sketch the probable orientation patterns (textures) of long fibers in the reser-
voir and under the doctor blade in tape casting for cases of Π >> 1 or Π << 1.
Internal Flows 355

Explain the physical mechanisms that control the texture in both cases.
Assume that the initial orientation distribution of the fibers in the reservoir
is random.
8.5 A capillary flowmeter is designed to measure the volumetric flow rate of
air. The diameter of the capillary is 0.5 mm and the maximum value of h
that can be read on the manometer is 1 m. Calculate the minimum length
of capillary at which, with a reading of h = 1 m, the flow through the
capillary is still laminar. The manometric fluid is carbon tetrachloride
( = 1600 kg/m3) and the density and viscosity of air at 298 K are, respec-
tively, 1.19 kg/m3 and 1.86 × 10 –5 Pa∙s.
8.6 When air at 300 K is flowing through a horizontal tube of inner diameter
1 cm, the local flow velocity measured at a radius of 3 mm for the center-
line of the tube is found to be 1 m/s. Calculate the rate of pressure drop in
the flowing air; the average linear velocity of the flow; the Reynolds num-
ber; and the shear stress exerted on the tube by the air. For air at 300 K,
= 1.187 kg/m3 and μ = 1.85 × 10 –5 Pa∙s.
8.7 For laminar flow in a tube of circular cross-section of radius R, by what
factor would R have to be increased to double the average linear velocity?
to double the volume flow rate?
8.8 Flow through a stationary annular extrusion die can be approximated as
flow in the space between two concentric cylinders. Assume the flow is
fully developed, axisymmetric, steady, and that it is only driven by a uni-
form axial pressure gradient. There are no significant body forces and the
Newtonian fluid has constant and uniform properties. (a) Using a cylindri-
cal coordinate system, simplify the z-momentum equation as far as pos-
sible, giving reasons for neglecting terms. Identify the physical mechanisms
in the remaining terms. (b) Write the appropriate boundary conditions for
velocity and state the reason for each. (c) Solve the system for the velocity
field and sketch u(r).
8.9 A Newtonian fluid falls under gravity in the cylindrical tube in Figure 8.24.
The flow is axisymmetric (∂/∂ = 0), steady, and fully developed, with no
imposed pressure gradient. (a) Using the cylindrical continuity equation,
find a numerical value for the radial velocity component, u. (b) Simplify
the z-direction momentum balance down to two terms, using assumptions
above and the value of u from part (a). (c) Pick reference scales for axial
velocity (wref ) and radial direction (rref ). Scale the two terms left in the
momentum equation and find an estimate for the maximum velocity in the
tube. (d) Starting with the two terms in part (b), find the exact solution for
the radial dependence of velocity along the tube.
8.10 A velocity field, u(y), and average velocity for steady flow of a power-law
fluid between two stationary flat plates (where the height of cavity is 2 )
are given in Section 8.7. Use these solutions to model the injection molding
of a polymer, where the polymer is injected into a thin cavity (Figure 8.24)
with dimensions 2 = 2 mm (height of cavity), w = 20 mm (width of cav-
ity), and L = 40 mm (length of cavity). The properties of the polymer are
n = 0.4, k = 10,000 Ns n/m2 at T = 125 oC and n = 0.6 at T = 250 oC. The
356 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.22 Fluid being pushed through an annular area, as in an extrusion

process.

FIGURE 8.23 Vertical tube with flow due to gravity.

temperature dependence of the consistency index is k A exp E RT ,

where E = 10 kcal/g mole and R is the universal gas constant. (a) Find the
consistency index at 250 oC. (b) At 250 oC, how long will it take to fill the
cavity if the volumetric flow rate (Q = u A) is constant at 100 mm3/s? Sketch
Internal Flows 357

Pinlet P x f as a function of time. (c) If Pinlet P x f is maintained,

could the volumetric flow rate be constant throughout the process? If not,
explain why not and sketch the real Q vs. t. (Hint: the channel is filling

dP P xf Pinlet P dx f
during the real process.) (d) Using ,u , and
dx x f (t ) xf dt

the previous expression for mean velocity, find xf(t). (e) From the solution in
(d), plot Q vs. t at 125 oC and 250 oC, assuming in both cases that the fill time
is the same as in part (b). What are the inlet pressures at each temperature?
8.11 Repeat Problem 8.11, assuming a Newtonian liquid metal with = 2500
kg/m3 and μ = 0.003 kg/ms. Perform all analyses at only one temperature,
T = 700 oC.
8.12 Microfluidics deals with controlling channel flow on a length scale from
approximately 10 to 1000 microns. In these systems, the Reynolds numbers
tend to be order one or less, so the flow is laminar and fully developed. One
application is a membraneless filter. Solute in two immiscible fluids can
have different mass diffusivities and so can be separated based on differ-
ent diffusion rates from one solute-bearing fluid to another. Figure 8.25 is
a schematic showing an H-shaped microfluidic channel in which solutions
A and B are merged into a straight channel and then divided into down-
stream two channels. A is an aqueous buffer and B is an aqueous binary
colloidal suspension of small and large particles. The idea is that the small
particles are more mobile than the large particles, therefore they can diffuse
into the buffer solution faster. By the time the fluids reach the exit channels
the buffer solution is rich in small particles while the colloidal suspension
is mostly composed of large particles. In this problem, you will predict the
velocity profiles and mass flow rates in both fluids. This information can
then be used to predict transport between the two fluids and to design a
filter and the filtering process. A one-dimensional, steady, fully developed
flow in a rectangular channel gives a momentum equation of
2
u dP
,
y2 dx

FIGURE 8.24 Injection of liquid into a die approximated by two stationary, parallel plates.
358 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 8.25 Membraneless filter design using microfluidic channel and two immiscible
fluids.

Internal Flows 359

REFERENCES
1. Howatt, G. N., R. G. Breckenridge, and J. M. Brownlow, “Fabrication of thin ceramic
sheets for capacitors,” Journal of the American Ceramics Society, v. 30, pp. 237–242,
1947.
2. Jabbar, M., R. Bulatova, A. I. Y. Tok, C. R. H. Bahl, E. Mitsoulis, and J. H. Hattel,
“Ceramic tape casting: A review of current methods and trends with emphasis on rhe-
ological behavior and flow analysis,” Materials Science and Engineering B, v. 212,
pp. 39–61, 2016.
3. Moriana, A. D., and S. Zhang, “Lead-free texture piezoceramics using tape casting:
A review,” Journal of Materiomics, v. 4, pp. 277–303, 2018.
4. Messing, G. L., S. Poterala, Y. Chang, T. Frueh, E. R. Kupp, B. H. Watson III, R. L.
Walton, M. J. Brova, A.-K. Hofer, R. Bermejo, and R. J. Meyer Jr., “Texture-engineered
ceramics – Property enhancements through crystallographic tailoring,” Journal of Mate-
rials Research, v. 32, pp. 3219–3241, 2017.
5. Kim, H. J., M. J. M. Krane, K. P. Trumble, and K. J. Bowman, “Analytical fluid flow
models for tape casting,” Journal of the American Ceramics Society, v. 89, pp. 2769–
2775, 2006.
6. Kim, H. J., K. J. Bowman, and K. P. Trumble, “Visualization of tape casting fluid flow
and alignment of anisometric particles,” Proceedings of the 9th International Ceramic
Processing Science Symposium, G. L. Messing, S. Hirano, and L. Gauckler (eds.),
Wiley, 2006.
7. Rosen, S. L., Fundamental Principles of Polymeric Materials, 2nd ed., John Wiley &
Sons, 1993.
8. Chhabra, R. P., and J. F. Richardson, Non-Newtonian Flow and Applied Rheology: Engi-
neering Applications, 2nd ed., Elsevier, 2008.
9. Bird, R. B., W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, 2nd ed., John
Wiley & Sons, 2006.
10. Dantzig, J. A., and C. L. Tucker III, Modeling of Materials Processes, Cambridge Uni-
versity Press, 2001.
11. Darcy, H. P. G., Les fontaines publiques de la ville de Dijon, Victor Dalmont, 1856.
12. Bejan, A., Convection Heat Transfer, 2nd ed., Wiley-Interscience, 1994.
13. Nield, D. A., and A. Bejan, Convection in Porous Media, 4th ed., Springer, 2013.
14. Scott, G. D., and D. M. Kilgour, “The density of random close packing of spheres,” Brit-
ish Journal of Applied Physics (Journal of Physics D), v. 2, pp. 863–866, 1969.
15. Wen, C. Y., and Y. H. Yu, “Fluid particle technology,” Chemical Engineering Progress
Symposium Series 62, 1966.
 9 External Flows

9.1 INTRODUCTION
Up to this point, the solutions to the Navier–Stokes equations in Chapter 7 have
been restricted to internal, mostly laminar flows. These flows are bounded by a
conduit and dominated by pressure and drag forces, with no significant local or
advective acceleration. In this chapter, we examine external flows, in which fluid
moves over an entirely submerged body, with no other surface constraining the
path of the flow. After a simple solution of a fully developed flow on an incline,
external flow over a semi-infinite flat plate is examined and solutions giving the
velocity, wall shear stress, and the extent of the region affected by that shear are
found. A more complete description of turbulence is presented next, including
turbulent flow on a flat plate and the transition to that state from laminar motion.
We then describe the flow past various bodies, including solid cylinders, spheres,
and other shapes, and the behavior of gas bubbles and liquid drops moving in
different fluids.

9.2 FULLY DEVELOPED FLOW DOWN AN INCLINED PLANE

A simple example of external flow is motion of a liquid layer down an inclined plane
(or, launder), as in Figure 9.1, which is a common method of transferring molten
metal or slag from one vessel to another. Upstream of the inclined plane, the liquid
has no vertical velocity but is entrained by the accelerating flow down the plane. The
fluid accelerates under gravity down the incline until it quickly reaches a terminal
velocity, being restrained by the viscous shear at the lower boundary. At that point
the flow does not change along the plate until the flow is disturbed a small distance
upstream of the end of the plane. More details on the entrance and exit regions of the
flow can be found in [1].
As the flow accelerates due to gravity in the entrance region, mass conservation
thins the layer. Once terminal velocity is reached, there is no further change in veloc-
ity in the x direction, and so mass conservation keeps the layer thickness constant, ≠
f(x). In this fully developed region with no pressure gradient, the steady Navier–
Stokes equations in Eqs. (7.35)–(7.37), can be simplified:

2
u
g cos 0, (9.1)
y2

where is defined in Figure 9.1 as the angle of the launder from the vertical.
There is a no-slip condition on the launder surface (y = 0) and a no-shear condition
between the gas and liquid (y = ). Thus, u in the liquid increases from zero at y = 0
to a maximum value at the free surface.

360
36 0 DOI: 10.1201/9781003104278-9
External Flows 361

FIGURE 9.1 Schematic of flow down an inclined plane.

Integrating Eq. (9.1) over the layer thickness gives

u
g cos y C1 , (9.2)
y

and, as the shear stress at the free surface is zero ( y

0), the integration constant is

C1 g cos . (9.3)

Then, as

u
g cos y (9.4)
y

for a Newtonian fluid, a second integration and application of the second boundary
condition, u = 0 at y = 0, gives

2 2
g cos y 1 y
u . (9.5)
2

The maximum value of u is at the free surface (y = ),

2
g cos
umax u y . (9.6)
2
362 An Introduction to Transport Phenomena in Materials Engineering

The shear stress is the maximum at the wall (y = 0),

u
max g cos . (9.7)
y y 0

This relationship shows the balance between the wall shear and the component of
the gravity force parallel to the plane. The velocity and shear stress distributions are
shown in Figure 9.2.
To complete the solution for this flow, one more variable is needed. Generally, for
a fixed geometry and fluid, either the liquid layer thickness, , or the mass flow rate,
m, is specified. The mass flow rate is related to the previous solution, Eq. (9.5), by
the average velocity across the layer:
W g cos W 1 2
u W u dA u dy dz y y dy.
A 0 0 0 2
2
g cos
u (9.8)
3

From Eq. (9.8), we can find the relationship between mass flow rate and layer
thickness:

2 3
g cos W
m u W . (9.9)
3

We have developed these solutions with the assumption that the flow is laminar and
that the layer thickness is uniform. The actual flow regime for this geometry is cor-
related by experiment and analysis to the Reynolds number, defined with the average

FIGURE 9.2 Velocity and shear profiles in liquid layer moving down an inclined plane.
External Flows 363

velocity (u ) and the hydraulic diameter, Dh. This diameter, defined in Section 6.4, is
based on the cross-sectional area (A) normal to the flow direction and the perimeter
of the channel in contact with the fluid (the wetted perimeter, PW):

4A 4 W 4
Dh 4 ,
PW W 2 1 2 W

where /W << 1. So,

u Dh m Dh 4m
Re . (9.10)
W W

At very low values of Re, the flow is laminar and the surface is smooth. With more
vigorous flows, the surface develops gravity-driven waves, at first of long wavelength
and low amplitude, but later less periodic and with a higher amplitude. The underly-
ing flow is still laminar. Experiments show the smooth to wavy transition occurs in
the Re = 20–30 range and the transition to turbulence occurs in the Re = 1200–2500
range [2, 3]. The previous laminar solution is still a reasonable approximation into
the wavy regime, perhaps up to the Re ≈ 1000 [3]. This Re value suggests a break-
down of this solution for metals and slags at layer thicknesses on the order of a
millimeter, so this laminar flow solution is not valid for many common applications.
For the higher Re typical of a launder (e.g., ~107 for Example 6.6), the flow is always
turbulent and Section 6.8 should be consulted for a more practical solution.

9.3 FLOW OVER A HORIZONTAL FLAT PLANE

A schematic of a steady flow of fluid over a flat plate is shown in Figure 9.3. Far
away from the plate the fluid is flowing in the x direction with a uniform freestream
velocity U∞. The streamlines for the flow impinging on the plate show the uniform
U∞ for x < 0 and U∞(x, y) past the leading edge of the plate (x ≥ 0). This flow field
can be found far from the plate by solving the Navier–Stokes equations dominated
by a balance of pressure and advective acceleration and assuming that the fluid has
zero viscosity (μ ≈ 0). Early attempts to solve the Navier–Stokes equations focused
on such inviscid flows, but the predictions of total drag did not match experimental
results in most cases. Although these solutions prevented flow through surfaces,
Eq. (7.42), they did not enforce the no-slip condition, Eq. (7.41), on the plate. How-
ever, when a real fluid flows over a plate, it does experience the influence of viscous
drag caused by the requirement of no slippage. In 1905, Ludwig Prandtl proposed
an idea to address this paradox. He split the flow into two separate regions, the first
in the free stream where the inviscid solution for U∞(x, y) holds. The other region,
very thin compared to the length scale of the body, is a layer on the surface in which
viscous effects play a significant role and which incorporates the no-slip condition at
the surface [4, 5]. It is in this region where velocity varies from u = 0 at the surface
(y = 0) to u = U∞(x) at y = , as shown in the detail of Figure 9.3. The viscous drag
causes a velocity gradient in the y direction in what is referred to as a boundary
364 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.3 Flow over a flat plate (beginning at x = 0). The main schematic shows an
inviscid flow field changing over an inclined plate from a uniform freestream velocity to a
nonuniform, accelerating one (∂U∞/∂x > 0). The inset shows the velocity gradient inside the
very thin, viscous friction-dominated boundary layer, where it merges with the freestream
velocity field.

layer on the plate, and that layer has a thickness (x). Prandtl suggested matching
the inviscid velocity solution at the surface, here U∞(x, y = 0). The assumption is that
(x)/x << 1, so that the growing boundary layer is thin enough to not affect the outer,
inviscid solution. Because velocity in the layer, u(x,y), transitions smoothly into the
freestream, U∞(x, y = 0), the thickness (x) is arbitrarily located at u(y)/U∞ = C, where
C is a constant typically chosen as 0.99. The interesting history of this revolutionary
development in the science of fluid mechanics can be found in [6, 7].
Questions that arise include how thick is the boundary layer, how does that thick-
ness vary with the properties of the fluid and distance down the plate, and what is the
magnitude of the viscous drag force exerted by the fluid on the flat plate? To answer
these questions, we begin with the steady, two-dimensional (w = 0 and ∂/∂z = 0) con-
tinuity and momentum equations, Eqs. (7.13), (7.35), and (7.36):

u v
0, (9.11)
x y

2 2
u u P u u
u v ,
x y x x2 y2
(9.12)
advectiive pressure viscous
acceleration friction
External Flows 365

and

2 2
v v P v v
u v . (9.13)
x y y x2 y2
advective pressure viscous
acceleration friction

The properties of the boundary layer can be obtained from Eqs. (9.11)–(9.13), if
some assumptions are made as to the relative magnitudes, and hence the relative
importance, of the individual terms in the equations; these necessary assumptions
require physical intuition and experience. In addition to being able to account for vis-
cous effects near a surface, a significant advantage of separating the free stream from
the flow in the boundary layer is the simplification of the Navier–Stokes equations in
that region. To address the simplest case, here we examine flow over a semi-infinite,
horizontal flat plate and assume that the growing layer is always very thin relative
to the distance from leading edge of the plate, /x << 1. (There is an abundance of
experimental evidence of the slender nature of this layer.)
Using this description of a steady two-dimensional boundary layer governed
by Eqs. (9.11)–(9.13), we can scale these equations inside the boundary layer. The
boundary layer thickness is the region in which viscous friction plays a role, so the y
direction is scaled as yref ~ , which is an as-yet-unknown function of the x-direction
scale, xref ~ x. The horizontal velocity varies inside that region from zero at the sur-
face to U∞(x) at the edge of the boundary layer (y = ), so uref ~ U∞(x). The change
in the vertical velocity from the surface to the free stream is also an unknown, vref.
Applying these reference values to the continuity equation, we obtain

U vref
~ or vref ~ U . (9.14)
x x

Because /x << 1, we know that vref << U∞. That there is a positive vertical velocity
out of the boundary layer is not surprising, given the discussion of the continuity
equation surrounding Figure 7.2. Eq. (9.11) shows that ∂u/∂x < 0 (a flow decelerat-
ing due to the wall friction) gives rise to a ∂v/∂y > 0 and a positive vertical velocity
throughout the boundary layer.
The steady x momentum equation is scaled as

U U P U U
U , vo ~ , , . (9.15)
x x x2 2

advective pressure viscous

acceleration gradient friction

(The pressure gradient is known from the solution far from the surface.) Both accel-
eration terms are always the same order of magnitude, U 2 x , and so can be com-
bined. The ratio of the two friction terms is
366 An Introduction to Transport Phenomena in Materials Engineering

2
U U
1,
x2 2
x

so the first term is negligible. Finally, an integration of the Bernoulli equation (6.2)
gives an expression for the pressure gradient near the surface,

1 P U
U . (9.16)
x x

Compiling all these results gives a reduced momentum equation in the x direction:

2
u u U u
u v U . (9.17)
x y x x2

In the case where there is zero pressure gradient in a flow over a flat plate,

1 P U
U 0. (9.18)
x x

Scaling Eq. (9.17) and using Eq. (9.18) shows a balance of advective acceleration
and viscous friction,

U2 U U
deceleration ~ ~ 2
~ friction or 1. (9.19)
x x

The left-side term is the deceleration of the flow in the boundary layer caused by
the right-side term, the shear between the free stream and the surface. Rearranging
this equation, we find an expression of the boundary layer thickness as a function of
distance downstream from the leading edge,

x
~ ~ x Re x 1/ 2 , (9.20)
U

where the Reynolds number is defined as Rex = U∞x/μ. The boundary layer thick-
ness is a penetration depth of the diffusion of momentum, analogous to the thermal
and species penetration depths of heat and mass diffusion in earlier chapters. In lami-
nar flows, momentum is advected into the boundary layer from the free stream, and it
diffuses out of the fluid into the surface due to friction retarding flow at the wall. The
boundary layer thickness is useful to understand flow behavior and can help estimate
the shear stress distribution along the flat plate, thus

u U Re1x/ 2 U2
w ~ ~ U ~ . (9.21)
y y 0
x Re x
External Flows 367

After Ludwig Prandtl developed the idea of the boundary layer in 1905, his student
Heinrich Blasius found (in his 1908 doctoral dissertation) a mathematical solution
for these equations on a horizontal flat plate with a uniform free stream, using what
is called a similarity solution [8, 9]. (This solution is described in detail in the fluid
mechanics texts, such as [1, 10, 11].) Blasius reduced these partial differential equa-
tions to one ordinary differential equation, which he solved using a numerical tech-
nique. He found an expression for the boundary layer thickness,

5.0 x Re x 1/ 2 , (9.22)

that is, ~ x1/2, and, at any value of x, the layer thickness decreases with decreasing
kinematic viscosity, v, or increasing U∞. These functional dependencies are the same
as the scaling results. Eq. (9.22) is referred to as an exact solution1 for the thickness
of the boundary layer (although it does require a numerical solution of the ordinary
differential equation). Blasius also found the wall shear stress,

U2
w 0.332 . (9.23)
Re x

We note in these solutions that the shear stress is very high near the leading edge of
the plate (x = 0), but decreases rapidly with the boundary layer growth downstream.
The solution breaks down at x = 0 because the simplification /x << 1 is no longer
valid. Near a real leading edge, significant diffusion of momentum occurs both nor-
mal to the surface and in the direction of the flow, as the velocity gradient in the flow
direction is no longer negligible.

9.4 MOMENTUM INTEGRAL SOLUTION FOR BOUNDARY

LAYER ON A HORIZONTAL FLAT PLATE
The thickness of the boundary layer and the shear stress at the wall also can be
obtained by applying the integral method to the two boundary layer equations (9.11)
and (9.17). This approach, introduced by Theodore von Kármán, will also give us an
approximate expression for the velocity profile in the boundary layer [12, 13]. We
use the same conditions as in the Blasius solution: Newtonian flow over a horizontal
flat plate, with a uniform free stream and no pressure gradient. Figure 9.4 shows a
schematic of the velocity profile inside a boundary layer.
We start by integrating the continuity equation (9.11) over the height of the bound-
ary layer (y = 0 to ):

v u v u
dy 0 and dy dy. (9.24)
0 y x 0 y 0 x

The left-hand side of the latter equation integrates directly,

u
v v 0 dy, (9.25)
0 x
368 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.4 Schematic of velocity profile in boundary layer on a flat plate.

which gives us an expression for the vertical velocity at the edge of the boundary
layer, remembering that there is no velocity at the surface.
The integration of the momentum boundary layer equation (9.17), with zero pres-
sure gradient, is
2
u u u
u v dy dy.
0 x y 0 y2 (9.26)
advective accelerration friction

We begin with the second acceleration term by integrating it by parts,

u v
v dy uv 0
u dy,
0 y 0 y

evaluating the (uv) term thus,

u u
uv 0 u v u 0 v 0 U dy U dy,
0 x 0 x

by using u( ) = U∞ and Eq. (9.25) for v( ). This work changes the second term in Eq.
(9.26) to

u u u u
v dy U dy u dy u U dy. (9.27)
0 y 0 x 0 x 0 x

Combining the two acceleration terms gives

u u u
u v dy 2u U dy. (9.28)
0 x y 0 x
External Flows 369

We can further simplify the integral thus,

u u u2 (x)
2u U dy U dy U u u 2 dy.
0 x 0 x x x

Using Leibnitz’s rule,2 we have an expression for integration of the advective accel-
eration terms:

u u d
u v dy U u u 2 dy. (9.29)
0 x y dx 0

The right-hand side of Eq. (9.26) is somewhat simpler:

2
u u u wall
dy . (9.30)
0 y2 y 0
y0

Reconstructing the full equation from Eqs. (9.29) and (9.30), we have the momentum
integral equation for a flat plate boundary layer with no pressure gradient:
d u
U u u 2 dy v .
dx 0 y0 (9.31)
advective acceleration friction
Up to this point the only approximations made are those to produce the boundary
layer equations, but now we must estimate the shape of the velocity profile, u(y), in
order to extract (x) and wall(x) from the momentum integral equation. We assume a
functional form of the velocity field, taking care to match the boundary conditions at
the surface and at the edge of the boundary layer.
We define nondimensional variables for u and y,
u y
and . (9.32)
U ( x)

Although other choices are possible, we assume a parabolic shape for ( ):

2
C1 C2 C3 . (9.33)

The first two boundary conditions are that the velocity is zero at the surface (no slip)
and the freestream velocity (U∞) at the edge of the layer.

u y 0 0 0 0 and u y U 1 1 (9.34)

As the velocity outside the boundary layer is uniform, we can deduce that the shear
stress goes to zero at the edge of the layer, so

u
0 as y or 0. (9.35)
y 1
370 An Introduction to Transport Phenomena in Materials Engineering

Application of these conditions to Eq. (9.33) gives the velocity profile:

2
2 u y y
2 or 2 . (9.36)
U x x

Of course, the velocity profile is not entirely known, as we still have the undiscovered
boundary layer thickness, (x). To find this unknown function, we use our assumed
profile in the momentum integral equation. Rearranging Eq. (9.32) gives

u U and y ,

which we insert into Eq. (9.31) to yield

d 1 U
U2 2
d , (9.37)
dx 0
0

where

1
2 2
2 and d .
0
0 15

The integral in Eq. (9.37) is a pure number, the value of which is due to the chosen
profile. Choosing another shape would change this value, but not the functional form
of the solution.
Continuing the solution of the momentum integral equation, we find the differen-
tial equation for the boundary layer thickness,

2U 2 d 2 U
.
15 dx
deceleration friction

Separating variables,

d 15 U dx,

and applying the initial condition, x 0 0, gives,

30 U x 5.48 U x 5.48 x Re x 1 2 , (9.38)

the form of which matches the scaling analysis result and the coefficient is roughly
10% higher than the “exact” Blasius solution. Figure 9.5 compares different velocity
solutions, where the second-order polynomial used here is a better fit to the similarity
solution than the cubic profile.
External Flows 371

FIGURE 9.5 Velocity profiles across a zero-pressure gradient boundary layer on a horizon-
tal flat plate. Blasius’s similarity solution is compared to the integral analyses using second-
and third-order polynomials.

Plots of (x) and developing velocity profiles are found in Figure 9.6. Note that
( ) is the same at every downstream position (see Figure 9.5), but u(y) changes to
adjust to the growing boundary layer. With the same velocities at both edges of the
boundary layer, the three plotted profiles have the same shape but are stretched more
vertically as x increases.
Fluid moves through the free stream at an unchanging upstream velocity (uni-
form U∞) until it reaches the upper edge of the boundary layer. (Remember that
the definition of a boundary layer is the region with a velocity different from the
free stream due to friction at the surface.) We see that the velocity (see the thicker
arrows in Figure 9.6 (b)) at a constant height (the dotted horizontal line) continuously
decreases with increasing distance from the leading edge, retarded by viscous drag
in the boundary layer, (i.e., du/dx < 0 in the boundary layer). Consequently, there is
a positive (if small) vertical component of flow within the boundary layer, in compli-
ance with the continuity equation (9.11).
The wall shear stress can also be found using our solutions for u(y) and (x).

u 2 U C1 U U2
wall Re x C1 (9.39)
y0 x Re x

The constant C1 is from the solutions for the boundary layer thickness and is 0.365
for the integral solution and 0.336 for the Blasius solution.
372 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.6 Results for zero-pressure gradient flow over a horizontal flat plate. (a) Comparison
of boundary layer thickness and wall shear stress from Blasius solution and integral analysis. (b)
Velocity profiles at several locations downstream from leading edge. The thicker velocity vectors
show the deceleration at a given height.

Example 9.1
Boundary Layer Thickness and Drag Force on a Flat Plate
Consider the flow of air and water over a flat plate (L = 0.5 m, W = 0.25 m), each at
a velocity of U∞ = 0.5 m/s.

water : w 997 kg/m 3 w 8.57 10 4 kg/ms w w / w 8.60 10 7 m 2 /ss

air : a 1.18 kg/m 3 a 1.85 10 5 kg/ms a a / a 1.57 10 5 m 2 /s

External Flows 373

Find: For each fluid, (a) boundary layer thickness and (b) drag force on the plate.
Solution: (a) At the trailing edge of the plate (i.e., x = L = 0.5 m), the Reynolds
number is
Re L U L .

Therefore, for water,

Re Lw U L 0.5 m s 0.5m 8.6 10 7 kg ms 291, 000

and, for air,

Re La U L 0.5 m s 0.5m 1.57 10 5 kg ms 15, 900.

At the trailing edge, Eq. (9.38) gives

w L 5.48 0.5m 291, 000 5.08 10 3 m

for water and, for air,

a L 5.48 0.5 m 15, 900 2.17 10 2 m.

Thus, for a given geometry and flow velocity, the larger kinematic viscosity of air
causes the boundary layer in air to grow more rapidly than that in water.
We should also note that the assumption was made that /x << 1 in a boundary
layer. This ratio here is shown to be for water

w L 5.08 10 3 m
1.02 10 2 m 1
L 0.5 m

and for air

a L 2.17 10 2 m
4.34 10 2 m 1.
L 0.5 m

In both cases, the ratio is small enough that the assumption is at least internally
consistent.

9.4.1 entry lengtH at entrance to a pipe

In Chapter 8, we assumed the flows to be fully developed and unchanging over the
length of the internal passage. However, if the flow enters the passage with a uniform
velocity profile, then there must be a region between the inlet (x = 0) and some point
downstream (x = xFD) where the flow is fully developed. It is in this region we have
developing or partially developed flow. The flow in this entry length at the channel
entrance may be treated with a boundary model. A schematic of the developing flow
in the entry length is shown in Figure 9.7.
374 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.7 The developing velocity field in the entry length of a channel. The flow transi-
tions from uniform (at x = 0) to the fully developed solution in Eq. 8.19 (at x > xFD).

How does the flow transition from uniform velocity to fully developed? As the
uniform flow enters the channel, the fluid is stopped at the walls (y = ±H) by the
no-slip condition. The layer near the wall is slowed (but not stopped) by viscous fric-
tion between the layer and the wall. A layer adjacent to the first is then also slowed
(but not as much), and so on until the viscous effects peter out and the flow is inviscid
nearer the centerline. This inviscid core velocity is unaffected by the viscous action
in the developing boundary layer, so we have a core velocity, Uc(x), uniform in the y
direction.
Because the fluid near the walls becomes progressively slower downstream, the
core velocity must increase so that the average velocity in this steady flow remains
uniform in the channel. In fact, the velocity at the centerline (always in the core)
must increase from u x 0, y 0 U at the inlet to u x xFD , y 0 32 U ,
where the fully developed solution becomes valid. (This fully developed value is
found from remembering that the average velocity, Eq. (8.22), is always U∞ and the
centerline velocity is found in Eq. (8.21).)
An integral solution by Sparrow [14] assumes a velocity profile in the entry region
boundary layer as
2
u x, y y y
2 , (9.40)
Uc x E x E x

where E(x) is the boundary layer thickness in the entry length. This profile is used
in the application of the integral method to the boundary layer equations (9.11) and
(9.17), and the core velocity is related to the pressure gradient by the Bernoulli equa-
tion as in Eq. (9.18), where U∞ = Uc(x). The results of this analysis are laid out in
[14, 15] as

E x U
3 1 (9.41)
H Uc x
and

x H 3 U U Uc
9 c 2 7 16 ln , (9.42)
Re H 40 U Uc U
External Flows 375

FIGURE 9.8 Development of inviscid core velocity, Uc(x), and boundary layer thickness,
(x), in a channel entry length.

where ReH = U∞H/μ. The position of the beginning of the fully developed region
(xFD) is predicted by this method to be

xFD H
0.026. (9.43)
Re H

The solutions presented here for channel flow are shown in Figure 9.8. For laminar
flow into a circular pipe, Langhaar [16] derived a similar relation,

zFD D
0.05, (9.44)
Re D

where D is the pipe diameter and z the axial direction.

The prediction of the extent of the entry region is important when estimating the
total flow behavior in a channel. If xFD/L << 1 (where L is the total channel length),
then almost all the channel experiences fully developed flow and the solutions in
Chapter 8 suffice. If a significant portion of the channel (or even all of it) is entry
flow, then the solutions here are necessary to calculate global results, such as average
wall shear stress or the value of a pressure gradient to maintain a desired flow rate.

9.5 TURBULENT FLOW

We have been referring to turbulence and turbulent flow based on a painfully brief
description of this flow regime given in Section 6.2. There, a limited, qualitative
observation was made that a turbulent flow is more disorderly and more vigorous
376 An Introduction to Transport Phenomena in Materials Engineering

(higher Re) than laminar flows. This weak description has been a useful start, and we
will extend it in this section, although we still only scratch the surface of this com-
plicated and interesting subject.
In materials processing, there are many laminar flows (e.g., slip and tape casting
of ceramic slurries, most polymeric flows, droplet impact during coating, and many
solidification processes after the initial pour), but these are important exceptions to
the observation that turbulence is ubiquitous. Any vigorous or large-scale flow (e.g.,
pouring and stirring of metals and slags, mixing of ceramic slurries, and various
spray processes) has a good chance to be turbulent. Observation in the natural world
and extensive experimentation have provided evidence that very many flows are not
regular, ordered, or smooth, as we assume when analyzing laminar flows. We can
have a qualitative notion that if a flow is vigorous enough (e.g., Figure 9.9) it may
be very turbulent, but vigor is not the only characteristic of turbulent flow. Laminar
flows may indeed have large-scale fluctuations, may be transient, and may appear at
first glance to be disorderly. So how can we tell the difference between the two flow
regimes? To help distinguish between the two different regimes, we first will rely
heavily on descriptions in [17, 18] to provide a set of measurable characteristics of
turbulent flows and then consider the transition to turbulence and the behavior of a
turbulent boundary layer along a flat plate.

9.5.1 cHaracteristics oF turbulent Flows

Although a popular notion of turbulence is qualitatively familiar to anyone who has
flown in an airplane, a comprehensive definition pertaining to fluid flow is difficult.
Turbulent flow is chaotic and irregular in that local velocities and pressure vary with

FIGURE 9.9 Flow of liquid copper from a launder into a tundish. Is the flow turbulent?
Source: (Photo by Mathew John M. Krane)
External Flows 377

FIGURE 9.10 Fluctuating local velocity of glycerin measured with multicolor particle
shadow velocimetry [19, 20] in a turbulent pipe flow. The u component is in the primary flow
direction, while v is perpendicular to it.
Source: [Unpublished data courtesy of M. H. Krane.]

time and position and do so in a non-periodic manner. The variation of the local
velocity, u, at a fixed location, xo, in a turbulent flow is illustrated in Figure 9.10. We
can define velocity, u, as the sum of u time-averaged over some period, ,

1
u u xo , t dt (9.45)
0

and a fluctuating component, u t :

u xo , t u xo u xo , t . (9.46)

The trace of measured velocity in Figure 9.10 shows two components (u and v) at a
particular location in a pipe. The mean velocity is constant if averaged over enough
time, but in general, it may be transient, as in laminar flow. The time average of the
disturbance is, by definition, zero:

1
u u t dt 0. (9.47)
0
378 An Introduction to Transport Phenomena in Materials Engineering

These disorderly measured disturbances are evidence of the irregular nature of

turbulence. To illustrate this characteristic, we recall one flow visualization exper-
iment from [17], which has water flowing in a channel past a vertical wire, which
generates a row of hydrogen bubbles when a current passes through it. The bub-
bles are seen in the sketch in Figure 9.11 (a) at a time just after a current pulse; at
such a small time since their generation, the bubbles have been advected only a
short distance downstream and the paths are still mostly horizontal. However, as
they move farther downstream, this pattern quickly degrades. The bubble profiles
across the channel take on different patterns at different positions downstream, e.g.,
Figure 9.11(b)–(e). In a laminar flow, the result would have been nearly identical
for each profile, determined by the velocity profile in Eq. (8.19) and Figure 8.3. If we
collect all of these individual turbulent profiles and superimpose them, Figure 9.11
(f), we get an idea of an average profile shape, with a steep gradient at the wall and
a much flatter pattern in the channel center. If we were to insert a probe into the
flow, we would obtain a velocity history qualitatively similar to that in Figure 9.10.
If we look at the frequency spectrum of these fluctuations, we would find it to be
a wide and continuous range of frequencies, showing that the fluctuating part of
the velocity in Eq. (9.46) is not periodic but irregular. This irregularity shows the
practical impossibility of exact repeatability of local velocity and pressure fields in
turbulent flows.3
The fluctuating velocity component in Eq. (9.46) is apparent in the observed irreg-
ular behavior just described. The introduction of the total local horizontal velocity,
as the sum of the time-averaged and fluctuating components, into the x momentum
equation (7.35) gives

u u u u u u u u
u u v v w w
t x y z
2 2 2
(9.48)
P P u u u u u u
x x2 y2 z2

or

u u u u u u
u u u u
t t x x x x
u u u u u u u u
v v v v w w w w (9.49)
y y y y z z z z
P P 2
u 2
u 2
u 2
u 2
u 2
u
.
x x2 x2 y2 y2 z2 z2
External Flows 379

FIGURE 9.11 Schematic of flow visualizations in a water channel. (a) Hydrogen bubbles
generated by current pulse as they leave the wire. (b)–(e) Snapshots of bubble positions
downstream for four different pulses. (f) Ensemble of (b)–(e) to show an average velocity
profile.
Source: (Sketches based on data from [17].)
380 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.11 Continued

We can rewrite the terms nonlinear in disturbances (underlined earlier) as

u u u
u v w
x y z
u v w
uu u uv u uw u (9.50)
x x y y z z
u v w
uu uv uw u ,
x y z x y z

with the last term equal to zero because of continuity. Substituting Eq. (9.50) into
(9.49), and taking a time average of the entire equation, as in Eq. (9.45) and (9.47),
produces what is known as the Reynolds averaged Navier–Stokes (RANS) equations.
Similar analysis gives the y and z components also.

u u u u
u v w uu uv uw
t x y z x y z
(9.51)
2 2 2
P u u u
x x2 y2 z2
External Flows 381

v v v v
u v w vu vv vw
t x y z x y z
(9.52)
2 2 2
P v v v
y x2 y2 z2

w w w w
u v w wu wv ww
t x y z x y z
(9.53)
2 2 2
P w w w
z x2 y2 z2

These equations are the same as the Navier–Stokes equations derived in Chapter 7,
except they are written in terms of the time-averaged velocities. The three (under-
lined) terms in each equation are time averages of velocity disturbance products.
These terms are usually moved to the right-hand side of the equations and paired with
the viscous friction terms, thus

u u u u
u v w
t x y z
(9.54)
P u u u
uu uv uw .
x x x y z z

The form of the viscous friction terms has its origin in Newton’s law of viscosity,
relating shear stress and mean velocity gradients. Eq. (9.54) treats the disturbance
terms as a diffusive effect due to shear stress; hence, they are typically known as
“Reynolds stresses.” The terms arise originally from the advective acceleration terms,
so they can also be thought of as the improved advective transport of momentum due
to the fluctuations in the velocity field, which agrees with the qualitative description
of the effect of the eddies given later.
Having introduced the velocity disturbances into the momentum equations, we
have expanded the number of dependent variables without more equations to keep
the problem well-posed. To solve these momentum equations, we must have some
information about the form of the Reynolds stress terms. There are no obvious
choices to solve this closure problem, but many approximations have been devel-
oped, e.g., in [21–23]. Some of these methods have produced some agreement with
experiments, but none solve the problem in every case, and dealing with these terms
is an ongoing topic of research.
If we examine a turbulent flow, we can see vortices (eddies) in the flow structure.
The largest eddies in the stream are, at most, roughly the same size as the part of
the flow subject to shear (the thickness of a boundary layer or the width of a chan-
nel or a jet). These larger eddies move high-momentum fluid toward regions with
low momentum and vice versa. This mixing throughout the flow tends to flatten the
velocity profile, as in Figure 9.11 (f), moving momentum down the mean velocity
382 An Introduction to Transport Phenomena in Materials Engineering

gradients more efficiently than the momentum diffusion only due to friction in lami-
nar flow. The increase in the rate of momentum transfer is the reason turbulent flows
are described as diffusive. Although, given the actual mechanism just described, they
may more properly be seen as advective, as the momentum is transported by the
moving fluid with which it is associated. So, momentum is advected into the bound-
ary layer from the free stream, and it diffuses to the wall as in laminar flow, but it is
also carried from the freestream to the surface by advection. Either way, these eddies
and the mixing they cause are the reasons turbulent flow has significantly enhanced
transport behavior compared to laminar flow.
Inspection of a turbulent flow shows that, in addition to eddies the size of the flow,
there are many smaller disturbances swirling in and among the largest structures. The
range of sizes varies continuously from the largest eddies to ones that may be micro-
scopic. The large eddies are generated when the mean flow has enough momentum to
amplify any disturbance in the flow. This unstable condition occurs at high Reynolds
numbers, where the flow inertia is much larger than the damping viscous friction. If
we use Lℓ as the largest length scale (from both the flow size and the largest eddies),
we define Re L UL . The eddies are in themselves unstable and tend to break
down into eddies with smaller sizes. This process of spawning smaller and more
plentiful structures continues until the eddy inertia is small compared to the friction,
or Re Le ULe << 1 (where Le is the eddy size). These tiny, spinning structures
exhibit solid body rotation, with no internal velocity gradients.
At this point in the discussion, we are more or less professionally obligated to
quote a bit of doggerel by L. F. Richardson, in which he pithily described the energy
cascade [24]:

Big whirls have little whirls which feed on their velocity,

And little whirls have lesser whirls, and so on to viscosity.

Mechanical energy moves from the mean flow (driven by shear, pressure, or buoy-
ancy) through a series of progressively smaller and more stable eddies, until the vor-
tices reach a size at which the energy is dissipated as heat through viscous friction.
This energy cascade, with a continuous range of disturbance frequencies, is associ-
ated with the continuous range of eddy sizes.
So, how small is that viscous limit? A scaling analysis of the overall flow and the
flow in the smallest “whirls” suggests that the ratio of eddy sizes is

Ls L Re L3 4 . (9.55)

The smallest size in a given flow structure, Ls, known as the Kolmogorov length
scale, is the diameter of the smallest eddy. For higher Re flows, the range of eddy
diameters and frequencies is very wide; it narrows as the flow Reynolds number
decreases. We see again that a high Reynolds number is necessary for a turbulent
flow to exist. There must be enough mechanical energy provided to the bulk flow to
sustain its transfer down through a wide range of lengths and frequencies; otherwise,
disturbances are easily damped and the flow remains laminar. To illustrate the range
of eddy sizes as a function of Re, schematics of the two jets of the same size are
External Flows 383

FIGURE 9.12 Schematics of two submerged jets of the same size, but different Reynolds
numbers. The higher Re jet (a) has significantly finer eddy structure than the lower Re
jet (b).
Source: (From Turbulence [17], copyright by Education Development Center, Inc. (EDC).
Used with permission and with all other rights reserved.)

found in Figure 9.12. They have different Re, but the largest feature size inside both
jet structures is slightly smaller than half the jet width, which is the same in both
cases. However, the jet with Re = 20,000 has a much finer structure than the Re = 400
flow. The larger length scales are then based on the flow geometry, not the Reynolds
number, while the size of the smallest eddies is based on the geometry and Re.

9.5.2 transition and turbulent Flow over a Flat plate

Turbulence is initiated by the instability of a flow, where the Reynolds number is
large enough that the fluid’s inertia breaks down the main flow into a cascade of ever
smaller eddies until they are small enough that their mechanical energy is dissipated
into heat. But how does an otherwise laminar flow break down into such disorder in
the first place? In this section, we describe qualitatively the transition to turbulence in
a horizontal flat plate boundary layer and model the drag on such a plate.
Figure 9.13, suggested by a figure in [25], shows a flow development that is lam-
inar at the leading edge, turbulent at the trailing edge, and in between experiences
a series of events that break down the former regime to produce the latter. Near
the leading edge, the laminar boundary layer is thin and any flow disturbances in it
are readily damped. At some distance from the leading edge, the Reynolds number,
384 An Introduction to Transport Phenomena in Materials Engineering

Rex = Uox/ , has reached a critical value at which some disturbances are able to
grow rather than decay. The result is production of Tollmien–Schlichting waves in the
mean flow direction, which are velocity fluctuations with growing amplitude as the
fluid moves downstream. These waves are two-dimensional, with no component in
the spanwise direction. Once their amplitude reaches 1–2% of the freestream veloc-
ity (Uo), they become unstable and form three-dimensional vortices with a significant
spanwise component. Beginning in the high shear region near the wall, the vortices
stretch and break apart into many smaller eddies in a way similar to the process
described in Section 9.5.1. The field of these smaller structures then suddenly spawns
apparently random turbulent spots, with a high turbulence intensity and a tendency to
grow and combine with others as they move downstream. This appearance, mixing,
and melding of turbulent spots develops into a fully turbulent flow.
Frequently, Rex,crit, the point at which transition begins, is given as a single value
with the implication that the flows changes from laminar to turbulent rapidly. How-
ever, we can see from our discussion of Figure 9.13 that flows do not typically flip
from laminar to fully turbulent at one point, but usually transition over a region in the
flow. We should also point out that there are ways to encourage or delay the onset of
turbulence. Rough surfaces and turbulence in the free stream can reduce the critical
value by introducing or amplifying disturbances. As the surface becomes rougher,
it introduces more disturbances into the flow at the point of maximum friction. The
uniform flow outside the flat plate boundary layer may also have what is called free
stream turbulence, and the intensity of that turbulence will influence the initiation
and development of transition. The greater the amplitude of the velocity fluctuations
entering the boundary layer, the sooner the laminar boundary layer flow will break

FIGURE 9.13 Development of flow structure on a horizontal flat plate with uniform
freestream velocity, from laminar to turbulent flow. The upper sketch is the side view and the
lower figure is the top view (after [25]).
External Flows 385

down. In the limit of practically no freestream turbulence, Rex,crit in a zero-pressure

gradient flow over a flat plate has been measured as Rex,crit ≈ 2.8 × 106 [26]. At small
values of free-stream turbulence (up to ≈ 5%), experiments by Mayle [27] showed
values of Rex,crit in the range of 4 × 104 to 108 with a dependence of
54
Re x ,crit free stream turbulence intensity . (9.56)

In Section 9.4, Eq. (9.39) gives the local shear stress on the surface of the plate at a
distance x from the leading edge as

U2
o 0.365 (9.57)
Re x

from the integral solution for laminar flow. The local Darcy friction factor at x, fx, is
thus obtained from its definition,

fx 8 o U2 , (6.9)

and Eq. (9.57) as

8 U2
fx 0.365 2.92 Re x 1/ 2 , (9.58)
U2 Re x

where Rex is the local Reynolds number at x.

As the drag force, Fdrag, exerted by the laminar boundary layer on one side of a flat
plate of width W and length L is given as

W L 1/ 2 L
2 U 1/ 2
Fdrag o dx dz 0.365 U W x dx 0.730 U 2 WL Re L1/ 2 , (9.59)
0 0 0

the friction factor averaged over the length of the plate, fL , is obtained from Eq.
(9.59) as

8 Fdrag A
fL 5.84 Re L1/ 2 . (9.60)
U2

To estimate turbulent boundary layer growth, we can use the integral method as
outlined in Section 9.4. However, we will need an approximate mean velocity profile
across the boundary layer. As we have seen, the time-averaged velocity field does not
have the same shape as the parabolic laminar profile, but experimental data suggests
that a one-seventh power law is a reasonable curve fit in the turbulent regime:
1/ 7
u y
. (9.61)
U
386 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.14 Transition from laminar flow to turbulent flow in the momentum boundary
layer on the surface of a horizontal flat plate (not to scale). The transition regime stretches
over 3 x 105 < Rex < 3 x 106.

This velocity profile is shown in the downstream profile in Figure 9.14.

The use of Eq. (9.61) as the mean velocity profile in the development of Momen-
tum Integral Equation, Eq. (9.31), gives

d o
10.3 , (9.62)
dx U2

where o is the shear stress exerted on the plate. Although the expression

u
o
y y 0

is valid at the surface of the plate, it gives an infinite shear stress if we use the one-
seventh mean profile. The existence of a thin laminar sublayer in contact with plate
means that the one-seventh power law does not hold as y approaches zero, and thus
Eq. (9.61) should not be used to obtain o in Eq. (9.62). A curve fit to a model of
that laminar sublayer at the wall can be expressed in terms of the Darcy friction
factor:

8 o 1/ 4
fD 0.080 Re (9.63)
U2
External Flows 387

or

o fD 1/ 4
0.010 Re , (9.64)
U2 8

which, in combination with Eq. (9.62), gives

1/ 4
d 1/ 4 U 1/ 4
10.3 0.010 Re 0.103
dx

or
1/ 4
1/ 4 d U
0.103 dx. (9.65)
dx

If the laminar boundary layer at the front end of the plate is ignored, integration of
Eq. (9.65) from x = = 0 to x and gives

1/ 5
U x 1/ 5
0.194 x 0.194 Re x x. (9.66)

Thus, a turbulent boundary layer thickness increases as x4/5, in contrast with the lam-
inar dependence, ~ x1/2.
As noted earlier, the transition on a flat plate is effected over a range of Rex, and a
sketch of the growing boundary layer thickness is shown in Figure 9.14. The thick-
ness of the turbulent boundary layer increases at a faster rate than the laminar and
the turbulence causes a flattening of the velocity profile and increases the shear stress
exerted on the plate.
From Eqs. (9.64) and (9.66), the local friction factor, fx, is obtained as

1/ 4 1/ 5 1/ 4
U U x U
f x 0.080 0.080 0.194 x 0.121 Re x 1/ 5 . (9.67)

The drag force on the plate over length L and width W is

W L 1/ 5 L
1 0.121 U
Fdrag U 2 f x dx dz U2 W x 1/ 5
dx
0 0
8 8 0 (9.68)
2 1/ 5
0.0188 U WL Re L ,

and hence the average friction factor for the area W L is

8 Fdrag
fL 0.151 Re x 1/ 5 . (9.69)
WL U2
388 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.15 Variation of the average friction factor with Reynolds number for fluid flow
over a flat plate of length L. The laminar curve is from Eq. (9.60), and the turbulent is Eq.
(9.69). In this case, the transition region begins at Rex,crit = 5 x 105.

Equation (9.69) is valid in the range 5 x 105 < Rez, < 107. A more complex analysis
[25] gives

0.455
fx 2
(9.70)
ln 0.06 Re x

and

0.523
fL 2
(9.71)
ln 0.06 Re x

as the local and average friction factors for a smooth flat plate at any turbulent Reyn-
olds number. The variation of fL with ReL in both regimes is shown in Figure 9.15.

Example 9.2
Effect of Plate Length in Flow Direction on Overall Drag
A flat rectangular plate of dimensions 1 m x 0.5 m is towed at a velocity of U = 10 m/s
through water. Two configurations are possible: (i) towing in the direction parallel to
the longer side of 1 m and (ii) towing in the direction parallel to the shorter side of
0.5 m. For water: = 997 kg/m3 and μ = 8.57 × 10–4 kg/ms.

Find: Which configuration causes the greater dissipation of power?

Solution: Configuration (1): L = 1 m and W = 0.5 m
External Flows 389

At the trailing edge of the plate, x = L,

UL 997 kg/m 3 10 m/s 1m

Re L 1.16 10 7.
0.000857 kg/ms

The transition from laminar to turbulent boundary flow begins at that location on the
plate at which Rex = 5 × 105:

5 10 5 0.000857 kg/ms
x 0.0430 m.
997 kg/m 3 10 m/s

The boundary layer is thus laminar over the first 4.3% of the plate, which is a small
enough percentage that the layer may be approximated as turbulent over the entire
plate. Thus, from Eq. (9.69), the average friction fraction for the entire plate is

fL 0.151 Re L1/ 5 0.00582

and the drag force on one side of the plate is

1 1 2
Fdrag fL WL U2 0.00582 0.5m 1 m 997 kg/m 3 10 m/s 36.3 N.
8 8

The power dissipated by friction on both sides of the plate is thus

power 2 Fdrag U 2 36.3 N 10 m/s = 725 W.

Configuration (2): L = 0.5 m and W = 1 m

In this configuration, at the trailing edge

UL 997 kg/m 3 10 m/s 0.5 m

Re L 5.81 106
0.000857 kg/ms

and the boundary layer is laminar over the first 8.6% of the plate, which is still small
enough to be ignored. The average friction factor for the entire plate is now

fL 0.151 Re L1/ 5 0.00669,

and hence

1 1 2
Fdrag fL WL U2 0.00669 0.5m 1 m 997 kg/m 3 10 m/s 41.7 N,
8 8

and energy is dissipated at the rate

power 2 Fdrag U 2 41.7 N 10 m/s = 834 W.

390 An Introduction to Transport Phenomena in Materials Engineering

Thus, towing the plate in case (2) requires more power dissipation than does case (1).
This result can be understood in terms of the variation of the local shear stress, o,
with distance x along the plate. Equation (9.69) gives the variation of the local fric-
tion factor with x, and hence from a combination of Eqs. (6.9) and (9.69) we obtain
the variation of the local shear stress with x as

1 U2
o fL U2 0.151 Re x 1/ 5 C1 x 1/ 5
.
8 8

The average shear stress over the length of plate from the leading edge (x = 0) to
trailing (x = L) is obtained from
W L L
1 1 1/ 5 C2 1 / 5
o o dx dz C1 x dx L .
A 0 0
WL 0 W

Thus, because o decreases with increasing x, the overall drag force acting on a plate
of constant area (A = WL) decreases as L increases (and hence W decreases). The
shear is higher nearer to the leading edge where the velocity gradients are larger.

9.6 FLOW PAST SUBMERGED BLUFF OBJECTS

We are also interested in the flow past and drag on objects other than a flat plate.
On a flat plate parallel to the flow, the entire drag is due to friction at the wall, but
that is not the case for flow with bluff bodies. The total drag force exerted on most
structures, when they and the surrounding fluid have a relative velocity, has two
components. The first is friction drag arising from the shear stress on the surface. The
other force is form (or pressure) drag, which is due to the pressure difference caused
by changes in the freestream velocity induced by the body’s shape and the possible
formation of a wake. An overall drag coefficient can be defined as

CD FD U2 A 2 , (9.72)

where FD is the total drag force and A is the projected area of the body normal to
the freestream flow. This parameter has the same form as the Fanning friction factor
defined in Chapter 6, but CD is based on the entire drag force (friction and form) and
the frontal, not the surface, area.
Figure 9.16 (a) shows a cylinder in crossflow (i.e., the fluid flow is in a direc-
tion normal to the cylinder axis). As the fluid free stream approaches the cylinder,
its velocity decreases and falls to zero at the forward stagnation point. A laminar
momentum boundary layer begins at that point of impingement on the cylinder sur-
face, under conditions of a negative pressure gradient in the flow direction. This
favorable condition of decreasing pressure increases the fluid velocity, and eventu-
ally a point is reached on the cylinder surface at which the velocity is at a maximum
and the pressure gradient has increased to zero. Beyond that point, the pressure gra-
dient along the surface is positive and acts to retard the flow near the cylinder. Due to
External Flows 391

FIGURE 9.16 A cylinder in crossflow: (a) Fluid flow; influence on separation point by
(b) laminar flow and (c) turbulent flow.

this adverse pressure gradient, the velocity gradient at the wall decreases as the fluid
moves downstream. Eventually, the flow momentum in the layer near the surface
is insufficient to overcome the influence of the rising pressure, causing the velocity
gradient (and shear stress) at the surface to fall to zero. Downstream of this point,
the direction of the flow is reversed close to the cylinder and the boundary layer is
detached from the surface. This phenomenon is known as flow separation, and down-
stream from this separation point a wake is formed. This wake has a relatively low
pressure and is an important source of form drag.
This description and the sketch in Figure 9.16 (a) assumes a moderate Reynolds
number (based on cylinder diameter) with a laminar boundary layer on the surface.
In the extreme case of a very low ReD, a creeping flow occurs in which the separation
392 An Introduction to Transport Phenomena in Materials Engineering

point is at the cylinder’s trailing edge (180o from the forward stagnation point) and
so there is no wake region. In this case, the drag on the structure is almost entirely
due to the skin friction; the flow around the cylinder is dominated by viscous forces.
With an increase in ReD to values above 1, the situation changes to the case described
earlier, in which a laminar boundary layer develops but separates around 80o from
the leading edge, Figure 9.16 (b). This separation and the development of a wake
increases the form drag on the cylinder, which now dominates the total drag; the
overall CD becomes almost constant with ReD. The wake at lower ReD is laminar,
albeit with large-scale vortices being shed regularly behind the body. In more vig-
orous flows, the disturbances in the wake are distributed over a much wider range
of sizes and the wake becomes turbulent. A transition from laminar to turbulent
flow in the boundary layer begins around ReD ≈ 3 x 105. The mixing in the turbulent
boundary layer is more effective, transferring momentum between the free stream
and the cylinder, thereby delaying its separation compared to the laminar regime, as
shown in Figure 9.16 (b). This effect reduces the size of the wake and the form drag
suddenly as ReD passes the critical value, but then the overall drag slowly increases
with Reynolds number to a value insensitive to the strength of the flow (around
ReD ≈ 107). As in flow on a flat plate, turbulence can be induced in less vigorous
flows by surface roughness and freestream turbulence. The measured variation of
the drag coefficient on a cylinder and a sphere over a range of Reynolds numbers is
shown in Figure 9.17.

FIGURE 9.17 Variation of the drag coefficient with Reynolds number for a cylinder in
crossflow and a sphere. The data points are experimental measurements compiled in [28].
The curve for the cylinder connects the data, while the curve for the sphere is based on cor-
relations in Eqs. (9.76)–(9.81).
External Flows 393

In a similar way, the drag on a falling (or rising) sphere may be found from anal-
ysis and experiment. The progression of flow behavior is from the creeping laminar
flow with no wake (at very low ReD), through moderate Reynolds numbers with a
laminar boundary layer and separation on the upstream face of the sphere, and finally
to high ReD values with a turbulent boundary layer and a separation point moved to
the downstream side of the body.
An analytical solution exists for low Reynolds number flow, based on the balance
of buoyancy and drag forces on a steadily rising or falling sphere. The buoyancy
force is

4
FB p f g R3 f p g, (9.73)
3

and, in the special case of creeping flow (ReD < 1), the drag force is

FD 6 R vt . (9.74)

(The fluid is denoted by the subscript f and the moving particle by p; vt is the terminal
velocity.) This balance, known as Stokes’ law, gives a terminal velocity at low ReD of

2 f p gR 2
vt , (9.75)
9 f

and, from the definition of the drag coefficient in Eq. (9.72),

CD 24 Re D . (9.76)

Eq. (9.76) is valid when the flow is friction dominated (ReD < 0.1) and is within 10%
of the experimental data up to when ReD ≈ 1. In the range 2 < ReD< 500, we can use
the correlation

18.5
CD , (9.77)
Re 3D/ 5

and, for 500 < ReD <3 × 105,

CD 0.44. (9.78)

In this highest ReD range, CD is independent of ReD because the separation point is
stable and most of the drag is due to the low pressure wake rather than the surface
shear stress. Around ReD = 3 × 105, the boundary layer transitions to turbulence and
separation moves to the back side of the sphere. In the brief transition range (≈3 × 105
to 4 × 105), the drag coefficient drops drastically,
394 An Introduction to Transport Phenomena in Materials Engineering
External Flows 395

Note that the area used here is not the surface area, but the area of the projection of the
body in the flow direction. At this point, we must remember that the drag coefficient
is a function of the Reynolds number, which is also a function of the sphere radius:

f vt 2 R
Re R .
f

If we assume that the small size and speed of the sphere will result in creeping flow,
we can start by guessing that the drag coefficient is from Eq. (9.76). Using that rela-
tion and the definition of ReR, we can write

4 1 12 f 1
R3 f p g f vt2 R2
3 2 f vt R

or, rearranging for the sphere radius,

1/ 2 1/ 2
9 vt f 9 0.00125 m s 0.006 kg ms
R 33 m
2 f p g 2 7160 kg m 3 3980 kg m 3 9.81m s2

This result is reasonable if the Reynolds number for this radius and terminal velocity
is very small,

f vt 2 R 7160 kg m 3 0.00125 m s 2 3.3 10 5 m

Re R 0.097 1,
f 0.0061kg ms

so the selection of the creeping flow drag coefficient is valid. (If the ReD had been
greater than unity, then other CD correlations should be tried and checked for valid-
ity in this manner.) We have assumed here that the flow is quiescent, but upflows in
the ladle will shorten the time or decrease the minimum size alumina sphere while
downdrafts will have the opposite effect.

Table 9.1 shows CD values of more complicated shapes at specific Reynolds numbers.
Some interesting comparisons may be made among these shapes and to Figure 9.17.
For instance, we see that the cylinder (when ReD > 1000) has a drag coefficient just
over 1. In this range, the flow separates on the front of the cylinder, so it and the con-
vex rod (entry a) have similar CD, while turning that shape around so it catches the
wind (entry b) about doubles the drag for the same conditions and frontal area. Also,
the solid hemisphere (entry f) is similar to the sphere in Figure 9.17, but the flow on
the flat face of the same body (entry g) has triple the drag of the sphere.
Another solid structure of interest is the equiaxed dendrite found in myriad
metal casting processes [30]. For a metal with a cubic atomic structure, a dendrite
will have six-fold symmetry, as seen in Figure 9.18. The rate at which these unat-
tached solid structures sink (or, less often, float) relative to the liquid metal flow
helps determine the distribution of microstructure and mechanical properties in a
cast part. The Re range for these settling dendrites, based on the relative velocity
396 An Introduction to Transport Phenomena in Materials Engineering

TABLE 9.1
Drag Coefficients for Different Bluff Shapes (compiled in [29])
Shape L/D Frontal area ReD CD
a long, semicircular convex rod LD >103 1.2
b long, semicircular concave rod LD >103 2.3
c rectangular flat plate 1 LD 6 × 105 1.15
2 LD 6 × 105 1.16
5 LD 6 × 105 1.2
10 LD 6 × 105 1.22
30 LD 6 × 105 1.62
∞ LD 6 × 105 1.98
d long square section (flow on face) LD >103 2.0
e long square section (flow on edge) 2 LD >103 1.0
f solid hemisphere (flow on curved face) πD2/4 103 0.38
g solid hemisphere (flow on flat face) πD2/4 103 1.17
h circular disk (flow on flat face) πD2/4 106 1.12

FIGURE 9.18 Schematic of an equiaxed dendrite with six-fold symmetry.

Source: (courtesy of J. Dantzig)
External Flows 397

between the solid and liquid (typically ul us < 1 mm/s) and a length scale on the
order of dendrite size (typically < 1 mm), is usually much less than 1, with signifi-
cant skin friction drag and not much of a wake. The characterization of the drag on
such a structure, especially because its complex shape grows and changes during
processing, is more complicated than the simpler shapes from earlier. Typically
values of the drag coefficient are between 50% and 90% of spheres with equivalent
volume [31].
All of the examples of drag given earlier are for flow over rigid bodies. Many
examples of non-rigid bodies appear in materials processing, such as liquid ceramic
droplets in thermal sprays or rising oxygen bubbles through liquid blister copper dur-
ing fire refining. In these cases, depending on the relative viscosities and velocities of
the two fluids, the fluid in a bubble or droplet may flow and its surface may deform.
The fundamentals of the fluid mechanics of flows in and around deformable bodies
are found in [32].
To give a few simple examples, let us look at spherical gas bubbles and liquid
droplets. These particles may keep this spherical shape (within a 10% deviation) up
to ReD ≈ 500. The transition from a sphere occurs at higher Reynolds numbers for
fluids with higher surface tension. For a liquid sphere in a gas medium, the viscos-
ity ratio, = μp/μf, is very large. (The particle viscosity is μp and the external fluid
viscosity is μf.) As >>1, the flow over the sphere does not move the internal fluid
through shear at the boundary. In this condition, the body can be treated with the CD
correlations for a rigid sphere. However, systems with two immiscible fluids and
moderate values of , or a gas bubble rising in a liquid ( << 1), do have internal
circulation. An empirical correlation for drag coefficients as a function of ReD and
has been found [32]:

2
3.05 783 2142 1080
CD Re D0.74 . (9.82)
60 29 4 3

This correlation is plotted in Figure 9.19, which shows that the fluid motion inside
the particle at low to moderate decreases the drag coefficient compared to a rigid
sphere. While the outside flow must provide the energy to mobilize the fluid in the
particle, separation of the external flow is delayed on the moving surface, reducing
the form drag.
Depending on the strength of the fluid’s surface tension, deformable spherical par-
ticles tend to transform to spherical caps somewhere around 150 < ReD < 600. This
cap is approximately the shape of a slice off the top of a sphere, with the rounded
edge on the upstream side. Clift et al. [32] give a correlation for all ReD at which caps
may exist:

8 8 2 3
CD . (9.83)
3 Re D 1
398 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.19 Variation of drag coefficient for fluid spheres and spherical caps with ReD and
the viscosity ratio = μp/μ f.

As seen in Figure 9.19, for ReD > 100, CD ≈ 8/3 and is insensitive to the viscosity
ratio and ReD.

9.7 SUMMARY
When a fluid flows over a submerged plate, the no-slip boundary condition at the
surface causes a fluid velocity gradient. The local flow velocity varies from zero
at the plate to the freestream velocity of the fluid. This phenomenon forms on the
plate a momentum penetration depth, or boundary layer, the upper edge of which is
where the local flow velocity reaches the freestream value. Scaling analysis shows
a boundary layer thickness proportional to the square root of the distance along the
plate from the leading edge, and which increases with increasing kinematic viscosity
and decreasing freestream velocity of the fluid. The fluid slows when it enters the
momentum boundary layer due to viscous drag between the fluid and the plate, and
the local shear stress exerted on the plate increases with increasing viscosity and
freestream velocity and decreases with increasing boundary layer thickness. Conse-
quently, the local shear stress on the plate is proportional to the inverse of the square
root of distance from the leading edge. The total shear force exerted on the plate is
obtained by integrating the local shear stress over the area of the plate. These results
are confirmed by integral and similarity solutions of boundary layer flow. The idea
of a boundary layer is used to define the entry length of a conduit, before the flow is
fully developed, as in Chapter 8.
External Flows 399

Turbulent flows are characterized by vigor, high Reynolds numbers, and disorder.
Measurements show a wide spectrum of the frequencies of velocity and pressure
fluctuations, which is related to the size of vortices (or, eddies) in the flow. These
structures vary in size from the spatial extent of the flow down to so small that the
eddies only move by solid body rotation. The mixing associated with the wide range
of fluctuation frequencies and vortex sizes is the cause of the more effective advec-
tion of heat, species, and momentum than the pure diffusion in laminar flow. The
initiation and growth of disturbances during transition to turbulent flow is described
and we see that this change happens over a length of the surface.
Flow over bluff bodies is examined, with more detail to cylinders in crossflow.
The point of flow separation and size and behavior of the wake affect the cylinder
drag coefficient differently over a wide range of Reynolds numbers. Experimen-
tal drag coefficient values are tabulated and discussed for several shapes, includ-
ing equiaxed metal dendrites. Behavior of gas bubbles and liquid droplets are also
discussed.

9.8 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

FIGURE 9.20 Schematic of falling film on a plate rising from a liquid reservoir.
400 An Introduction to Transport Phenomena in Materials Engineering

9.1 A plate is withdrawn from a Newtonian liquid at velocity Uo. A film of liq-
uid of thickness is drawn upwards with the plate. Assume that the liquid
flow down the plate in Figure 9.20 is one-dimensional and steady state.
(a) Simplify the momentum and continuity equations based on assumptions
given and any reasonable ones you make. (b) Write two boundary condi-
tions for the velocity at y = 0 and y = ; briefly state the physical reason
for each. (c) Nondimensionalize the simplified momentum equation and
2
boundary conditions with u Uo, y , and A g U o. (d) Solve
the nondimensional momentum equation and apply the boundary condi-
tions to obtain . (e) What value of A = Acrit gives u( ) = 0? Sketch plots
of y vs. u(y) for A = Acrit, A > Acrit, and A < Acrit, making sure you indicate
the direction of the velocities in each sketch.
9.2 Repeat the analysis in Section 9.4, using the integral analysis to estimate
boundary layer thickness, shear stress at the wall, and velocity profile,
assuming a sine function for velocity in the boundary layer.
9.3 Repeat the analysis in Section 9.4, using the integral analysis to estimate
boundary layer thickness, shear stress at the wall, and velocity profile,
assuming a linear function for velocity in the boundary layer.
9.4 (a) Using the momentum integral equation for flow over a flat plate with
zero pressure gradient,

d 1
U2 2
d w
,
dx 0

find the boundary layer thickness for turbulent flow, assuming a velocity
profile of u/U∞ = (y/ )1/7 = 1/7. Show that this profile gives an infinite shear
stress at the wall. (Because this profile does give an infinite shear at the
wall, use w = 0.0233 U∞2 ( /U∞ )1/4 in the momentum integral equation.)
(b) Neatly plot on the same graph and compare /x vs Rex for 5 × 102 < Rex
< 5 × 105, using the laminar solution found in Section 9.4 and for 105 < Rex
< 108 for the turbulent solution found in part (a).
9.5 Water flows over a submerged square flat plate of area 1 m2 at a freestream
velocity of 0.3 m/s. Calculate: (a) the local shear stress on the plate at
x = 0.5 m; (b) the total shear force on one side of the plate; (c) the rate of
transport of viscous momentum at x = 0.5 m and y = half the momentum
boundary thickness at x = 0.5 m; and (d) the rate of transport of advective
momentum at x = 0.5 m and y = half the momentum boundary thickness at
x = 0.5 m. For water, = 995 kg/m3 and μ = 6.92 x 10 –4 Pa∙s.
9.6 A semi-infinite flat plate (begins at x = 0 and extends infinitely in the posi-
tive x direction) initially has no freestream velocity. At t =0, a uniform,
steady freestream flow in the positive x direction suddenly begins. Perform
a scaling analysis to discover the behavior of the boundary layer thickness
from start-up to steady state. Obviously, there is no boundary layer at t = 0
and the flow approaches steady state as t ∞. (The freestream velocity
is steady for t ≥ 0, but the flow in the boundary layer may or may not be.)
Assume a two-dimensional flow, with no pressure gradient, no body force,
External Flows 401

and a very thin boundary layer. (a) Assume xref ~ x and tref ~ t. Select ref-
erence values for uref and yref. Are any unknown? (b) Scale the continuity
equation to find vref. (c) Identify the terms in the x momentum equation.
Simplify and scale this equation and identify the three physical phenomena
left. (d) At early times, identify a two-term balance from the result in part
(c). Find an expression for , the boundary layer thickness. Under what con-
ditions may we neglect the physical mechanism we neglected as t 0. (e)
At late times, identify a two-term balance from the result in part (c). Find
an expression for , the boundary layer thickness. Under what conditions
may we neglect the physical mechanism we neglected as t ∞. (f) At early
times, sketch (x) at three different times. At late times, sketch (x).
9.7 (a) Describe the energy cascade in fully turbulent flow, beginning with the
source of mechanical energy and finishing with the final state of that energy.
Mention any relevant length scales in the flow. (b) Which of these charac-
teristics are applicable to fully turbulent flow? Irregular flow; mechanical
energy generation; flow with apparently enhanced diffusion; low values of
Re; intermittent flow.
9.8 The viscosities of experimental molten glasses are being determined by
measuring the terminal velocity of a platinum sphere falling through a col-
umn of molten glass. When dropped into a column of a standard glass of
viscosity 10 kg/ms and density 2500 kg/m3, the measured terminal veloc-
ity is 0.0258 m/s, and when dropped into an experimental glass of density
3000 kg/m3, the measured velocity is 0.0168 m/s. Calculate the viscosity of
the experimental molten glass. The density of platinum is 21,500 kg/m3.
9.9 Small glass spheres of density 2620 kg/m3 are allowed to fall through liquid
CCl4 of density 1590 kg/m3 and viscosity 9.58 X 10 –4 Pa∙s. Calculate: (a) the
maximum diameter of sphere for which the flow obeys Stokes’ law and (b)
the terminal velocity that a sphere of this diameter attains.
9.10 A steel ball of radius 0.01 m is dropped through a molten glass to determine
the viscosity of the glass. The density of the glass is half the density of the
steel ball and the ball attains a terminal velocity of 0.03 m/s. Calculate: (a)
the kinematic viscosity of the glass; (b) the radius of the largest steel ball
that can be used along with Stokes’ law to determine the kinematic viscos-
ity of the glass; (c) the terminal velocity of this largest steel ball.
9.11 An alumina-coated steel ball suspended on a rod in liquid aluminum flow-
ing in a trough is used as a flow measuring device (Figure 9.21). The drag
on the sphere is balanced by its buoyancy and by the force along the rod.
(a) Draw a free body diagram on the sphere. (b) Write expressions for the
drag and the aluminum velocity as functions of . (c) If Al = 2500 kg/m3,
steel = 8000 kg/m , μ Al = 0.003 kg/ms, R = 0.015 m, L = 0.25 m, L’ = 0.15 m,
3

and = 30 , what is UAl? (d) What is the maximum velocity that can be
o

measured with this apparatus?

9.12 A photochemically active, viscous polyimide solution is used as a photo-
resist during semiconductor fabrication and commonly is spread on silicon
wafers by a spin coating process. The polymer is deposited on the silicon
disk (Figure 9.22) and spun slowly until it has a more or less uniform thick-
402 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 9.21 Drag on a sphere used to measure liquid flow velocity.

ness, o. The angular velocity, Ω, then is increased quickly by few orders

of magnitude and the outward radial velocity is driven by the centripetal
acceleration. Through most of the process, fluid bleeds off the edge of
the substrate, thinning the layer while it maintains uniformity ( ≠ (r)).
Assume the flow is axisymmetric, the layer is thin ( << R), there is no
radial pressure gradient, and there is a uniform initial thickness, (t=0) = o.
For this system, zr u z w r and scaling analysis can show that
u z w r. (a) Write the radial momentum conservation equation (in
cylindrical coordinates) and reduce it to two terms, giving reasons for ne-
glecting terms and identifying those retained. (b) Use v=Ω r for the angular
velocity in result from part (a) and find the general solution for the radial
velocity u=u(r,z). Assume that the fluid layer thickness ( ) is independent
of r. (c) Write and explain the radial velocity boundary conditions at z =
and at z = 0. Apply those conditions to the result from part (b) to find the
solution to part (a). (d) Use the result from part (c), u(r,z), in the continuity

FIGURE 9.22 Schematic of spin coating process.

External Flows 403

equation to find the vertical velocity, w=w(r,z). (e) How is the result in part
(d), w(r,z), related to the rate of change of the thickness, d /dt? Using the
answer to that question, find = (t), what is the value of d /dr? Sketch
2 2
/ o vs. dimensionless time, 4 o 3 t. For easiest control of the
thickness, where on that curve should you operate? (f) For μ = 0.03 kg/ms,
o = 1 mm, f = 1 μm (final thickness), and = 5000 rpm, find the final time
(when (tf) = f). Is this value on the part of the curve selected in part (e)?
(This problem is based on solutions in [33] and suggested by [34]).

NOTES
1. Because the mathematical solution for velocity in the layer asymptotically approaches the
freestream value, the exact edge of the boundary layer is unclear and subject to arbitrary
definition. While the coefficient 5.0 is used to represent the “exact” thickness, the veloc-
ity is not quite the freestream value there (u/U∞ ≈ 0.992). As noted earlier, the edge is
frequently set where u/U∞ = 0.99, but at this value the coefficient is not 5.0 but 4.92. This
minor discrepancy illustrates that we should recognize that the “exact” value used here is
a convenient rounding (within 2%) of the coefficient based on the arbitrary selection of
u/U∞ = 0.99 as the border with the free stream.

B( x ) f d B dA dB
2. dt fdt f A f B
A( x ) x dx A dx dx
3. One can think of the nonrepeatability of local velocity fields as a more literal twist on the
pre-Socratic philosopher Heraclitus’s famous observation that one “never steps in the same
river twice.”

REFERENCES
1. Fox, R. W., and A. T. McDonald, Introduction to Fluid Mechanics, 4th ed., Wiley, 1992.
2. Chang, H.-C., “Wave evolution on a falling film,” Annual Review of Fluid Mechanics, v.
26, pp. 103–136, 1994.
3. Fulford, G. D., “The flow of liquids in thin films,” Advances in Chemical Engineering,
v. 5, pp. 151–236, 1964.
4. Prandtl, L., Über Flüssigkeitsbewegung bei sehr kleiner Reibung, Verhandlinger 3.
Internationalen Mathematiker-Kongress, pp. 484–491, 1904. (in German)
5. Prandtl, “Fluid motion with very small friction,” NACA TM 452, 1928. (English transla-
tion of [4])
6. Tani, I., “History of boundary layer theory,” Annual Review of Fluid Mechanics,
pp. 87–111, 1977.
7. Anderson, J. D., Jr., “Ludwig Prandtl’s boundary layer,” Physics Today, pp. 42–47,
December 2005.
8. Blasius, H., “Grenzschichten in Flüssigkeiten mit kleiner Reibung,” Zeitschrift für
Angewandte Mathematik und Physik, v. 56, pp. 1–37, 1908.
9. Blasius, H., “The boundary layer in fluids with little friction,” NACA TM 1256, 1950.
(English translation of [8])
10. Munson, B. R., D. F. Young, and T. H. Okiishi, Fundamentals of Fluid Mechanics, 2nd
ed., Wiley, 1994.
11. White, F., Fluid Mechanics, 8th ed., McGraw-Hill, 2015.
12. von Kármán, T., “Über laminare und turbulent Reibung,” Zeitschrift für Angewandte
Mathematik und Mechanik, v. 1, pp. 233–252, 1921.
404 An Introduction to Transport Phenomena in Materials Engineering
10 Convection Heat Transfer

10.1 INTRODUCTION
When a material at a hot temperature, TH, is in contact with another at a cold tem-
perature, TC, and both are stationary, heat flows from the former to the latter, as
with conduction heat transfer in Chapters 2 and 3. In such a case, the colder body is
heated as shown in the time progression in Figure 10.1 (a), where heat diffuses into
it, affecting a growing region of thickness T, the thermal penetration depth. The heat
flow rate from the surface by conduction goes like

q ~ TH TC T ,

and thermal resistance between the hot surface and the unaffected cold region is
R ~ T kA. The T and resistance increase with time, while the heat flow rate drops,
as in Figure 10.1 (c).
The conduction heat transfer sketched in Figure 10.1 (a) assumes that the
cold material is stationary relative to the hot body. However, in convection heat
transfer there is relative motion, and the heat diffused across the interface is
immediately swept away, Figure 10.1 (b). Not only does this action keep thin the
hot layer that builds up to insulate the hot surface, the relative motion continu-
ously brings cold material closer to the surface, maintaining a high heat flow rate,
Figure 10.1 (c).
This combination of transporting heat by being attached to moving fluid (advec-
tion) and by diffusion (conduction) is what defines convection heat transfer. For
example, transferring a hot slab of steel from a furnace to a stack of cold slabs in
the yard outside a mill transfers heat almost entirely by advection, as the heat travels
only because the slab is moved. However, once stacked, the slab cools, by conduction
to the slabs on which it is stacked and by convection to the air surrounding it. The
convection process is described in Figure 10.1 (b); it is a combination of conduction
into a thin layer between the hot steel surface and cold air and the advection that
moves that heated air away from the surface and replaces it with fresh, cold air.
The magnitude of the convection heat transfer depends on many factors, such as the
properties of the fluid, the geometry of the solid surface, and the velocity of the fluid
relative to that of the solid.
Two distinct convection regimes, forced convection and free convection, can be
identified from the mechanism that drives the flow. Forced convection occurs when
the flow is induced by causes external to the moving fluid, such as pumps or fans,
or by motion of the solid surface. Free, or natural, convection happens in buoyancy-
induced flows. The driving force there is variations in fluid density, usually caused by
local differences in temperature and/or composition.

DOI: 10.1201/9781003104278-10 405

406 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.1 Comparison of heat transfer from a hotter surface to a colder fluid by
(a) conduction and (b) convection. (c) Comparison of transient response of heat flow rate by
conduction and convection.
Convection Heat Transfer 407

Example 10.1
A Qualitative Comparison of Cooling by Conduction, Forced Convection, and
Natural Convection
In April of 1970, the crew of Apollo 13 (Jim Lovell, Fred Haise, and Jack Swigert) were
on their way to what would have been the third landing of men on the moon, when a
severe malfunction in their spacecraft endangered the crew and ended the mission. In
order to conserve battery power on the voyage home, the crew shut down almost all their
electronics (which had helped heat the spacecraft) and the air circulation fans. When the
crew were not moving around, there was almost no air movement, which led to some
interesting heat-transfer problems when sleeping. Lovell describes it in his memoir:

Approaching Haise in his right-hand couch, Lovell shook him awake. The am-
bient temperature in the command module was by now, Lovell guessed, down
in the low forties or high thirties (Fahrenheit). Around the sleeping Haise, how-
ever, a thin layer of nearly body-temperature air had formed. In the absence of
gravity, which meant the absence of (natural) convection, the warm air would
not be lighter than the cold air around it, and thus would not rise and drift away.
Helping Haise out of his couch, Lovell scattered the atmospheric blanket
his junior pilot had spent the last three hours creating, and sent him down to
the LEM. The commander then climbed onto his own couch, wrapped his arms
about himself, and curled up against the chill that his own body heat had not
had time to change [1].

The heated layer around Haise was formed while he slept entirely by conduction.
The air near his skin was warmer than the ambient temperature in the spacecraft,
and therefore less dense, but being so far from the earth or the moon, there was not
enough gravity to move it by natural convection. The slightest air or body motion
would have easily swept the heated layer away by forced convection, and did when
he awoke, but, without circulating fans on, we must conclude that Haise must have
been sleeping rather soundly and without significant movement.

In this chapter, we begin by deriving an advection-diffusion equation for heat

transfer, and defining the nondimensional heat transfer coefficient (the Nusselt
number) and the Prandtl number, a fluid property. This new thermal energy
balance is combined with the Navier–Stokes equations for momentum from Chap-
ter 7 to describe heat transfer in external and internal flows. The goal of such analysis
is to find approximate correlations for convection heat transfer coefficients as func-
tions of flow conditions, geometry, and fluid properties. We also discuss how such
correlations might be found through approximate analysis and experimentation.
In addition to the study of heat transfer by forced and natural convection in sin-
gle-phase fluids, we will briefly treat boiling heat transfer. This two-phase mode of
convection cools a surface more efficiently than single-phase heat transfer due to the
latent heat absorbed by the liquid-vapor phase change. It finds wide application in the
fast quenching of metals as part of heat treat processes.
408 An Introduction to Transport Phenomena in Materials Engineering

10.2 GENERAL ENERGY EQUATION WITH

ADVECTION AND DIFFUSION
In Chapter 1, we introduced the rate equation for convection heat transfer between
a solid surface and a surrounding fluid, here written in terms of the heat flow from
the surface:

q hAs Ts T , (10.1)

where Ts is the surface temperature and T∞ that of the ambient fluid. This relation-
ship between heat flow and overall temperature difference defines the proportionality
constant, the heat transfer coefficient, h. This quantity is useful for defining the rela-
tionship of the conduction across the solid–fluid interface and the advection of heat
away from it and will be used to model that heat flow.
On the fluid side of the interface, we equate the convection rate equation to the
heat flow from the solid. The no-slip condition prevents fluid motion at the surface,
V y 0 0, leaving only conduction as the mode of heat transfer there, so

Tf
q y 0 h Ts T kf . (10.2)
y y 0

Here the subscript f indicates the fluid and the surface is y = 0. A typical temperature
gradient on the fluid side of a hot surface is shown in Figure 10.2. Rearranging Eq.
(10.2), we see that the heat transfer coefficient is an expression of the temperature
gradient in the fluid at the interface,

kf Tf
h . (10.3)
Ts T y
y 0

Eq. (10.3) can be nondimensionalized by kf and a length scale, L, which is based on

the given flow geometry:

hL L Tf
NuL . (10.4)
kf Ts T y
y 0

The Nusselt number, NuL, is defined as the dimensionless temperature gradient at a

surface with convection. The form of the Nusselt number is similar, but not the same
as the Biot number (Bi) introduced in Section 2.6; the two parameters are based on
fundamentally different physical mechanisms. The latter includes the solid thermal
conductivity because Bi is based on the thermal resistance in the solid. The NuL is
based on the fluid temperature gradient at the interface and so uses the fluid thermal
conductivity. Also, the length scale in Bi is from the internal conduction resistance,
while L in NuL is a characteristic of the external flow field. These two important
dimensionless parameters and their physical meaning should not be confused.
Convection Heat Transfer 409

FIGURE 10.2 Temperature gradient in a fluid near a wall (y = 0+) and T(y=0) = Ts.

FIGURE 10.3 Energy transfer rates in an infinitesimal control volume with advection, con-
duction, generation, and storage of thermal energy.

To find the temperature field necessary to calculate the heat transfer coefficient
in Eq. (10.3), an energy conservation equation is required. Figure 10.3 shows the
energy transfer rates in and out of an infinitesimal control volume in a flow field, and
the normal and shear forces.
410 An Introduction to Transport Phenomena in Materials Engineering

The total energy rate balance on the control volume in Figure 10.3 is written as

U in U out U gen Ustor . (2.19)

The storage term is the same as in the energy equation for conduction,

E cT
Ustor dxdydz . (10.5)
t t

The model of enthalpy, E, in Eq. (10.5) is reasonable if the material may be treated
as single phase and incompressible (valid for liquids and solids, and for gases under
conditions with small pressure changes).
The energy transfers across the surfaces are due to conduction and advection. The
conduction contribution is developed in Chapter 2:

T T
U in U out k k dxdyydz . (2.24)
cond x x y y

Thermal energy is also advected across the control volume boundaries in Figure 10.3.
Again, assuming little change in density with pressure (which allows dU ≈ dE) and
using Taylor series at x + dx and y + dy as in Eqs. (2.20)–(2.22):

U in U out Ein Eout

adv adv

eu x
Ax eu x
eu x
dx Ax dx
x

ev y
Ay ev y
ev y
dy Ay dy
y

eu dxdydz ev dxdydz
x y

ucT vcT dxdydz . (10.6)

x y

The specific enthalpy for incompressible substances is e E m cT .

Different forms of energy may be transformed into thermal energy by many
mechanisms. These transformations can be treated as generation terms, if we formu-
late our energy equation as being for the conservation of thermal energy. One gener-
ation example is the work done on the fluid in a control volume by normal and shear
forces in Figure 10.4. Multiplying these forces by their parallel velocity components,
we have the work transfer rates through the four sides of the infinitesimal control
Convection Heat Transfer 411

FIGURE 10.4 Energy transfer rates in an infinitesimal control volume by shear and normal
forces.

volume, as shown in [2]. Adding these four contributions gives the generation of heat
due to mechanical energy dissipation in two dimensions:

xx u xy v xx v yx u
U gen,mech dxdydz. (10.7)
x x y y

Using the constitutive equations for a two-dimensional, incompressible, Newtonian

flow introduced in Section 1.3.1,

u v
xx P 2 yy P 2 (1.11)
x y
and

u v
xy yx , (1.12)
y x

and neglecting kinetic energy effects (see [2] for details) transforms Eq. (10.7) to

2 2 2
u v u v
U gen,mech 2 , (10.8)
y x x y
412 An Introduction to Transport Phenomena in Materials Engineering

where is known as the dissipation factor. This function is a consequence of the

viscous friction and is the rate at which mechanical energy is degraded to thermal
energy by the irreversible nature of the fluid flow. We should notice that, no matter
the signs of the velocities or gradients, is always positive. Other mechanisms of
thermal heat generation not caused by the flow field include Joule heating (electrical
work to heat) or chemical reactions (endothermic or exothermic), and these can be
written as in Chapter 2,

U gen,other q dxdydz . (2.26)

Substituting Eqs. (2.24), (2.26), and (10.5)–(10.8) into Eq. (2.19) produces an advec-
tion-diffusion equation for thermal energy:

cT cuT cvT T T
k k q. (10.9)
t x y x x y y

For constant and uniform properties, we can simplify this equation to

2 2
T T T T T q
u v .
t x y x 2
y 2 c c (10.10)
storage advection conduction viscous heat gen
dissipation

We have derived the thermal energy equation in two-dimensional form, and the
three-dimensional form is given in Appendix II.
The boundary conditions for the energy equation are the same as for conduction,
as shown in Section 2.5. As before, we will assume a surface with a specified tem-
perature, a specified heat flux, Eq. (2.41), or a specified heat transfer coefficient,
Eq. (2.43). Also, we will recognize again that these models of thermal interactions
between a body and its environment are simplifications of reality and must be appre-
ciated in that light.
The thermal energy equation (10.10) has storage and advection of heat, diffusion
of heat by conduction, and heat generation for a single phase. This form should look
somewhat familiar. If we examine the x momentum equation,

2 2
u u u u u 1 P
u v 2 2
gx ,
t x y x y x (7.33)
accelerattion viscous pressure gravity
local advective friction

we see a strong similarity between the two balances. In Eq. (7.33), the local accel-
eration is a storage of momentum and we have momentum advection, as well as the
diffusion of momentum through viscous friction. The pressure gradient and gravity
are sources of momentum. Eqs. (7.33) and (10.10) are both in a standard form of an
advection-diffusion equation:
Convection Heat Transfer 413

storage advection diffusion source

2 2
u v S. (10.11)
t x y x2 y2

Because these two equations are of the same form, we expect some similarities in
behavior, subject to the boundary conditions. In the absence of generation terms, the
similarities are governed by the ratio of the thermal diffusivity, , and the momentum
diffusivity, . That ratio is

momentum diffusivity k c
Pr, (10.12)
thermal diffusivity c k

where Pr is the Prandtl number.1 This dimensionless parameter shows us the relative
speed of transfer of heat and momentum by diffusion in a given fluid.
Unlike many other dimensionless parameters we discuss in this text, Pr is a mate-
rial property of the fluid, so we can examine values for a variety of interesting sub-
stances (Table 10.1). First, we notice that the order of magnitude of the specific heat
of most materials is in a range between (102) and (103) J/kgK, so the order of
magnitude of Pr is primarily governed by the relative values of the more widely rang-
ing quantities, viscosity, μ, and thermal conductivity, k. Gases tend to have Pr just
less than one, as their thermal diffusivities are slightly higher than their kinematic
viscosities. Water is somewhat more viscous relative to its thermal diffusivity, and
5 ≤ Pr ≤ 8 between its freezing and boiling points. Liquid polymers are quite different
due to their complex molecular structures; they have low thermal conductivities and
high effective viscosities and so behave as a fluid with Pr >> 1. (Magmas, while not
engineering fluids, also have large Pr; being molten ceramic mixtures, they have low
k and high μ.) On the other hand, liquid metals have a kinematic viscosity on the
order of magnitude as water, but possess a much higher thermal diffusivity, so their
Pr << 1.

10.3 ADVECTION IN RIGID, MOVING MEDIA

The simplest case of advection heat transfer is a moving isothermal body (e.g., the
steel slab mentioned in Section 10.2), but perhaps the next simplest is a process in
which a rigid body with internal conduction moves out of a hot region and sheds
heat to a colder environment as it does so. Examples of this configuration include
polymer strands leaving an extruder and metal rods undergoing continuous induction
hardening. In many cases, we can approximate the heat transfer as occurring in one

TABLE 10.1
Estimates of Prandtl Numbers for Several Materials Classes
Pr ≈ 0.01 0.5–1 5–8 > 100
materials liquid metals gases water polymers, magma
414 An Introduction to Transport Phenomena in Materials Engineering

dimension (the direction of motion, or the axial direction) and treating heat losses in
perpendicular directions as heat sinks. In order for this approximation to be valid, the
heat flow in the body must be oriented so that it is mainly in the axial direction. If
the heat flow in the direction of motion is much greater than the directions normal to
motion, then this one-dimensional approximation is reasonable.
If the moving body can be modeled as one-dimensional, then we can define a
control volume over which we can perform an energy balance in order to derive a
conservation equation for thermal energy in terms of temperature. Figure 10.5 shows
such a control volume, defined by length dx, cross-sectional area Ac, and perimeter p,
and in which thermal energy is transferred by conduction (qx) and advection in the x
direction. The amount of energy that is brought into the control volume at location x
by bulk solid motion is m ex VAc cT x , where ex is the specific enthalpy at x and
V is the speed of the moving body. The rate at which energy is advected out of the
volume at (x + dx) can be different and is written as m ex dx VAc cT x dx . Also,
heat can be generated in the volume (q) and also lost to the ambient by convection
from the surface. (This convection loss moves normal to the axial motion. To use
a simple one-dimensional model, we treat convection as a heat sink, a “negative
generation,” rather than as a boundary condition in y or z for a two-dimensional or
three-dimensional model.) For the rest of this derivation, we will assume mass flow
rate, m, material properties, heat transfer coefficient (h), and geometry do not change
along the direction of motion (x).
The first law of thermodynamics (energy is conserved) for this control volume can
be written as the usual energy rate balance:

Uin Uout U gen U stor , (2.19)

where heat flows in the x direction by conduction and advection, and the process is
steady state.

Uin q x m ex qx VAc cT x
U gen q Ac dx h p dx (T T )
(10.13)
U out qx dx m ex dx qx dx VAc cT x dx
U stor 0.

FIGURE 10.5 Control volume (in gray) for analysis of moving rigid body.
Convection Heat Transfer 415

Note that only the part of the body that actually exchanges heat with the environ-
ment is counted in the perimeter, p. Writing a Taylor series expansion of the specific
enthalpy at (x + dx), we obtain

dex
ex dx ex dx higher order terms.
dx

Using this expression to find the difference between the energy advected in and out
of the control volume, we get

dT
m (e x dx ex ) VAc c dx. (10.14)
dx

Similarly, we can find the change in the diffusion heat transfer over dx:

d dT
qx dx qx k Ac dx. (10.15)
dx dx

Putting these relations into Eq. (2.18), we get the energy conservation equation that
describes the temperature along the length of the moving body.

d VAc cT d dT
kAc hp T T q Ac
dx dx dx (10.16)
advection conductiion convection heat
loss generation

It is useful to look carefully at this energy equation to remind ourselves of the

physical phenomena that govern it. Eq. (10.16) is equivalent to Eq. (10.10) for a
steady, one-dimensional flow. The first term is the change in the thermal energy of
a mass as it moves through space, i.e., it is the rate of change of thermal energy
advected as it passes a specified point in space. The second term represents the dif-
fusion of thermal energy along the length of the body due to the axial temperature
gradient. This conduction term happens independent of the body motion and its mag-
nitude. The third term is the convection heat loss from the outer surface to the envi-
ronment and the final term is heat generated inside the body. For uniform properties
and cross-sectional area, the energy balance in Eq. (10.16) reduces to

dT d dT hp
cV k T T q.
dx dx dx Ac (10.17)
advection conduction connvection heat
loss generation
The boundary conditions in x for this equation are found by considering the physical
configuration. As the body leaves a constant temperature source (e.g., a furnace or
extruder), it likely has the temperature of that device. As the body loses heat to the
environment, eventually it will cool to the ambient temperature. These boundary
conditions are written as
416 An Introduction to Transport Phenomena in Materials Engineering

T (x 0) To and T (x ) T .

At this point, we have an equation that can be solved for temperature as a function of
11 parameters: T(x, k, Ac, , c, V, h, p, T∞, To, q). It would simplify matters to reduce
the parameter space and produce a more generalized result by nondimensionalizing the
equation and boundary conditions, using carefully chosen reference values. We start
here by dropping the heat generation term. (This step is not necessary, but most appli-
cations do not have this effect and it does simplify the following procedure.) We pick
references for the temperature difference, T(x) – T∞, and for the axial coordinate, x, thus

T T
and x l. (10.18)
To T

The characteristic length, ℓ, is initially unknown, and its form will be chosen to
reduce the number of parameters. Inserting Eq. (10.18) into Eq. (10.17) and rear-
ranging, we get

kAc To d2 VAc c To d
2 2
hp To 0, (10.19)
l d l d

a second-order differential equation for = f( ). We nondimensionalize the equation

by dividing by the coefficient of the first term, and setting ℓ = (kAc/hp)½ and Pe = Vℓ/ :

d2 V d
2
0, (10.20)
d d

we have reduced the problem to one of finding temperature, , as a function of the

axial coordinate, , and the Peclet number, Pe. Introducing the Peclet number as
the ratio of the speed of advection, V, and an effective thermal diffusion speed, /ℓ,
the energy equation and boundary conditions are then written as

d2 d
2
Pe 0 0 1 0.
d d (10.21)
axial axiial convection
conduction advection loss

The general solution for Eq. (10.21) is = exp(-B ) where

B ( Pe 2) 2 1 ( Pe 2). (10.22)

We can also find approximate solutions for different physical situations, depending on
which terms (i.e., which physical phenomena) dominate the energy conservation equa-
tion. There are three interesting cases that are simplifications of the general case [3].

1 Pe 0: In a slow-moving solid (low V), or one with a large thermal diffu-

sivity ( ), the advection term is negligible, and relatively little heat is carried
Convection Heat Transfer 417

by the motion of the body. The heat flow is then a balance of conduction
along the body and the convection heat loss:

d2 1
2 1 0 1 0 1 1 0, (10.23)
d

and the solution is 1 = e- (i.e., B = 1). This result is the classic solution for
a stationary fin of infinite length [4].
2 Pe ∞: In this situation, the advection term completely overwhelms the
other effects and almost all of the heat moving downstream in x is carried
by bulk motion. While conduction always occurs, its effect is negligible. If
the body is moving fast enough (Pe ∞), there is not time even for convec-
tive losses to occur and the conservation equation (10.21) is:
d 2
0. (10.24)
d
Integrating this equation and applying the condition at = 0, we find 2 = 1
(B = 0), implying that there is simply no time to lose any thermal energy to
the environment, there is effectively no drop in temperature. This result is
interesting as a limit, but is not applicable to systems that do have a measur-
able heat loss.
3 Pe large, but not infinite: In this case, advection is much more important
than axial conduction (as shown earlier), but is still balanced by heat loss to
the ambient. Most practical materials processing problems fall into this case.

d 3
Pe 3 0 3 0 1 (10.25)
d

The solution for this regime is 3 = e- /Pe (B = 1/Pe), which is simpler than
the general case. The validity of these three approximate cases can be seen
in comparison to the exact solution in Figure 10.6.

FIGURE 10.6 Coefficient (B) in exponent in solutions of general and special cases of a one-
dimensional advection-diffusion problem.
418 An Introduction to Transport Phenomena in Materials Engineering

Example 10.2
Cooling of a Steel Sheet Exiting a Heat Treat Furnace
A steel slab emerging from a soaking furnace at 1400 K cools as it moves away from the
furnace. We would like to roll the slab when its temperature is no lower than 1350 K. As
a first approximation, assume that all the heat loss is by radiation and that the average
temperature of the slab is 1375 K for the purpose of estimating a radiation heat transfer
coefficient. The material property data, processing parameters, and geometry are:
k = 31 W/mK c = 1200 J/kg K = 7800 kg/m3 = 3.3 × 10–6 m2/s
To = 1400 K T∞ = 300 K V = 0.05 m/s = 0.7
d = 0.05 m (slab thickness) w = 2 m (slab width)

Find: The approximate distance, L, the roll stand should be from the furnace
to roll at T(L) ≥ 1350 K.
Solution: First, we estimate the “radiation heat transfer coefficient,” taking the
slab surface temperature as the average of the furnace temperature and the
rolling temperature (T = 1375 K),

hrad TS2 T 2
TS T . (1.38)

8 2 2
hrad (0.7)(5.67 10 W/m 2 K 4 ) 1375 K 300 K 1375 K 300 K

132 W/m 2 K

Because it is a thin, wide slab (d << w), the perimeter to cross-sectional area ratio
will be Ac/p = wd/2(w+d) ≈ d/2 = 0.025 m and ℓ is found to be

1/ 2
1/ 2 1/ 2
ksteel Ac ksteel d 31W mK 0.05m
m
0.0767 m.
hrad p 2hrad 2 132 W m 2 K

FIGURE 10.7 Geometry of a steel slab cooling by radiation after exiting a furnace.
Convection Heat Transfer 419

Note that ℓ is not the length to the first roll, but the characteristic length based on
thermal properties and heat loss. Calculating the Peclet number to check if the previ-
ous approximations hold, we find

V 0.05 m s 0.0767 m
Pe 6
1160,
3.3 10 m2 s

which matches Case 3. Using that solution, we have

L 1
exp exp or L Pe ln
Pe Pe

1350 K 300 K
L 0.0767 m 1160 ln 4.15m.
1400 K 300 K

So the rolling stand should be approximately 4 m from the exit of the furnace. The
temperature profile (dimensional and nondimensional) from the furnace (x = 0) to the
roll stand (x = L) is shown in Figure 10.8.
As a check that the temperature is practically uniform through the thickness d,
we can calculate the Biot number, where the length scale is the half-thickness, d/2,

hrad d 2 132 W m 2 K 0.05 m 2

Bi 0.106 1.
ksteel 31W mK

This result shows that the assumption of a one-dimensional temperature field along
the sheet axis is reasonable.

FIGURE 10.8 (a) Temperature profile in slab from furnace to roll stand. (b) Nondimensional
temperature profile.
420 An Introduction to Transport Phenomena in Materials Engineering

10.4 EXTERNAL FORCED CONVECTION

10.4.1 Forced convection FroM a Horizontal Flat plate
The thermal energy equation (10.10) and the continuity and momentum equations
(7.13) and (7.33) hold the promise of estimating heat transfer coefficients by predict-
ing temperature fields in space and time when heat is transported by advection and
diffusion. Such solutions also are useful for understanding the patterns of transport as
heat is moved through a domain and especially carried to and from a surface. Unfor-
tunately, closed-form solutions for T(x, y, z, t), even approximate ones, are thin on
the ground and so most problems of practical interest may eventually require some
sort of numerical treatment (depending on how much detail is required for a specific
problem).
One type of problem that does admit a limited number of mathematical solutions
is the thermal boundary layer on a horizontal flat plate. We will use this configuration
with its considerable simplifications as a starting point for understanding convection
heat transfer, especially the thermal interaction at the fluid–solid boundary, which is
characterized by the surface heat transfer coefficient, h.
Figure 10.9 shows the geometry under consideration, with an isothermal surface,
T x, y 0 TS . The flow field is as described in Section 9.4 and Figure 9.6, with
freestream velocity parallel to the plate, V U î . Also, there is neither pressure gra-
dient nor heat generation, and the flow is two-dimensional and steady. A hydrody-
namic boundary layer, in which the velocity ranges from zero at the no-slip surface
(y = 0) to U∞ at y = (x), develops as wall friction bleeds momentum from the flow.
At the same time, the isothermal surface is a different temperature from the incoming
fluid at T(y ≥ T) = T∞ and so a thermal boundary layer forms between the surface and
the unaffected fluid.

FIGURE 10.9 Schematic of general case of convection over an isothermal, horizontal flat
plate. The temperature and velocity profiles are shown in their respective boundary layers.
The relative thickness of the two boundary layers is shown as / T > 1, but in general also
may be / T < 1.
Convection Heat Transfer 421

The primary goals here are (i) to understand the physical mechanisms important
to the fluid flow and heat transfer and (ii) to estimate the surface heat flux and heat
transfer coefficient. To do so, we begin with a scaling analysis of the governing equa-
tions inside the boundary layers. The analysis in Section 9.4 gives results for the flow
field:

vref ~ U and ~ x Re x 1/ 2 , (10.26)

x

based on the boundary layer equation with no pressure gradient,

2
u u u
u v , (10.27)
x y x2

and the reference values

uref ~ U yref ~ . (10.28)

The thermal boundary layer thickness ( T) is as yet unknown, but we might find an
order of magnitude estimate if we scale the energy equation in that near-wall region,
0 ≤ y ≤ T. The thickness of the thermal boundary layer will help us find an expression
for the heat transfer coefficient but, as we will see later, T is not usually the same as
. In a steady condition without heat generation, the energy equation is

2 2
T T T T
u v 2
. (10.29)
x y x y2
advection conduction

Applying reference values from Eqs. (10.26) and using T T ref

~ TS T T
as a reference for the temperature difference, we get

T T T T
U , U ~ 2
, 2
. (10.30)
x x T T x

Note that the y-direction reference is defined for this equation as yref ~ T, because
the thermal boundary thickness is the distance over which the temperature varies.
2
The ratio of the two conduction terms on the right-hand side is T x which we
assume is very small. Except near the plate’s leading edge, x = 0, this assumption is
supported by experimental observation, and we conclude that there is no significant
conduction in the flow direction.
Using this result, we drop the x-direction conduction term in Eq. (10.29),

2
T T T
u v , (10.31)
x y y2
422 An Introduction to Transport Phenomena in Materials Engineering

which scales as

T T T
U , U ~ 2
.
x x T T (10.32)
advection advection conducttion
in y in x in y

The relationship between the advection terms depends on the as-yet-unknown / T

ratio. Because we do not know its value, we examine thermal behavior at its limits,
/ T >> 1 and / T << 1.

10.4.1.1 Scaling in the Thermal Boundary Layer for / T >> 1

We start with / T >> 1, as shown in Figure 10.10. We see that the maximum velocity
in the thermal boundary layer shown, u y T , is much less than the freestream
value, so U∞ is a poor velocity reference there. We scale the velocity gradient inside
the thermal layer thus,

u uref U
~ ~ , (10.33)
y T

where the reference velocity for each layer is chosen to be the velocity at its edge. So,

T
uref ~ U , (10.34)

FIGURE 10.10 Temperature and velocity profiles near an isothermal flat plate where the
thermal boundary layer is much smaller than the velocity layer ( T 1).
Convection Heat Transfer 423

which is much less than U∞. If we use Eq. (10.34) to rescale the continuity equation,
we find a new reference for the vertical velocity also:

u v uref vref T T T
0 or ~ and vref ~ uref ~U . (10.35)
x y xref yref x x

So we have the proper velocity scales inside the thermal boundary layer and we
re-examine the heat balance, Eq. (10.31).

T T T
uref , vref ~ 2
x T T

T T T T T T
U , U ~ 2
x x T T

T T T T T
U , U ~ 2
x x T (10.36)
advection advection conduction
in y in x in y

The two advection terms have the same order of magnitude everywhere, so they
can be combined:

T T T
U ~ 2
.
x T (10.37)
advection ~ conduction

This result shows the fundamental mechanisms of convection heat transfer in the
thermal boundary layer. Heat diffuses normal to the surface into the layer, and then
is swept away by the advecting fluid. Rearrangement of Eq. (10.37) gives us an esti-
mate of T,

T ~ x Re x 1 2 Pr 13
. (10.38)

Comparing the two boundary layer thicknesses, we see that their ratio is

x Re x1 2
~ 12 13
~ Pr1 3 . (10.39)
T x Re x Pr
424 An Introduction to Transport Phenomena in Materials Engineering

We had assumed that / T >> 1, so Eq. (10.39) tells us that this assumption is valid
if Pr >> 1.
One use for the previous relation is the estimation of the heat transfer coefficient,
which we find from scaling the surface boundary condition.
dT f
q h TS T kf
dy 0

kf dT f kf TS T kf
h ~ ~
TS T dy 0
TS T T T

kf Re1x 2 Pr1 3
h~ ~ kf (10.40)
T x

The local Nusselt number is then

hx
Nux ~ ~ Re1x 2 Pr1 3 . (10.41)
kf

A sketch of h(x), Figure 10.11, shows that h ~ x-1/2 and is infinite at the leading edge
(x = 0) because the boundary layer thickness is predicted to go to zero there, sug-
gesting an infinite temperature gradient. (Of course, in reality h remains finite at the
leading edge. There the boundary layer approximations fail, especially that /x and
T/x << 1, and diffusion of both heat and momentum in the flow direction are signif-
icant.) Eq. (10.40) also shows that, because h ~ Re1x 2 ~ U 1 2 , the surface heat transfer
coefficient increases as the flow becomes more vigorous. (This result certainly fits

FIGURE 10.11 Heat transfer coefficient on an isothermal flat plate: h as functions of dis-
tance downstream from the leading edge and freestream velocity.
Convection Heat Transfer 425

with common experience: if you are outside on a warm day, do you want a slow or
fast wind to cool yourself?) The exponent ½ also indicates that doubling the flow
velocity does not double the heat transfer rate and that the return on increasing veloc-
ity diminishes with more vigorous flow and farther downstream.
We began by assuming that T 1, which we showed occurs where Pr >> 1.
However, this analysis actually is applicable for much lower values of Pr, indeed all
the way to Pr ~ (1). The solution then is used successfully for fluids such as water
and gases.

10.4.1.2 Scaling in the Thermal Boundary Layer for / T << 1

If we repeat the previous analysis with T 1, where heat diffuses faster than
momentum, the boundary layers flip their relative positions, as seen in Figure 10.12.
In this case, we still scale the energy equation over the thermal boundary layer, but
because u ≠ U∞ only very close to the surface (compared to T), we simplify the anal-
ysis by assuming u = U∞ everywhere. The scaling analysis for this case proceeds in
the same way as before and gives

kf Re1x 2 Pr1 2
h~ ~ kf (10.42)
T x

and

hx
Nux ~ ~ Re1x 2 Pr1 2 . (10.43)
kf

Solutions for T ~ (1) and T 1cases are worked out in detail in homework
problems at the end of the chapter.

FIGURE 10.12 Temperature and velocity profiles near an isothermal flat plate where the
thermal boundary layer is much larger than the velocity layer ( T 1).
426 An Introduction to Transport Phenomena in Materials Engineering

10.4.1.3 Integral Analysis of Forced Convection

in a Thermal Boundary Layer
For more detailed results for forced convection from a flat plate, we can apply the
integral analysis method to the energy equation near the wall, relying on the inte-
gral analysis of Section 9.4 for velocity fields. There we used the nondimensional
velocity and the spatial coordinate normal to the surface, u U and y , and
found that

2
2 and 5.48 x Re x 1 2 . (10.44)

With the flow field known, we can tackle the boundary layer energy equation
(10.31). A second-order polynomial is a reasonable approximation of the shape of
the temperature profile.

T T 2
A B C, (10.45)
TS T

where the nondimensional distance from the surface in the thermal boundary
layer is

y T . (10.46)

The boundary conditions on the temperature field are a uniform temperature at the
wall (TS) and at the boundary layer edge (T∞) and zero flux between the layer and the
freestream.

T y 0 TS 0 1
T y T T 1 0
T
0 0
y y 1
T

Applying these conditions to the profile in Eq. (10.45), we have the nondimen-
sional temperature profile,

2
2 1. (10.47)

Now we only lack information for T(x) to complete the solution. To find it, we inte-
grate Eq. (10.31) from y = 0 to y = T:

T T T 2
T T T
u dy v dy dy. (10.48)
0
x
0
y
0
y2
Convection Heat Transfer 427

The right-hand side (the conduction term) is

T 2
T T T T
2
dy . (10.49)
y y y 0
y 0
0 T

Using integration by parts, we find that the second term on the left-hand side of Eq.
(10.44) is

T T T v T u T v
v dy v y T T T dy T dy T dy, (10.50)
0 y 0 y 0 x 0 y

v u T u
remembering that continuity gives and v y T dy . The
y x 0 x
complete left-hand side of the integrated equation can now be written as

T dT u
u T T dy,
0 dx x

giving the energy integral equation (EIE):

d T T
u T T dy . (10.51)
dx 0 y0
downstream change in heat flux
heat
a advection from surface

To find the unknown T, we substitute the assumed velocity and temperature profiles,
Eqs. (10.44) and (10.47), into the EIE:

d 1 TS T
U T TS T d
dx 0 T 0

or

d T 1 2
U d . (10.52)
dx 0 T

Defining the function T f Pr , we find a relationship between the two

nondimensional distances from the surface ( and ),

y y T
, (10.53)
T
428 An Introduction to Transport Phenomena in Materials Engineering

and so

2 2 2 2 2
2 2 1 2 2 1.

The integral in Eq. (10.52) is then

1
I d 5 . (10.54)
0 30 6

(This equation has been simplified by eliminating the 2 term because < 1.) The
EIE is now
d T 2
U ,
dx 6 T
advection conduction

which reduces to a differential equation for T with a boundary condition at the lead-
ing edge:

d T 12
T T x 0 0 T 24 x Pr Re x .
dx 6 U Pr

Using Eq. (10.44) for the velocity boundary layer thickness, we can obtain a function

12 1
T
24 x Pr Re x 5 2
12
Pr .
30 x Re x 4

Solving this equation for gives

0.928 T
(10.55)
Pr1 3

and

T 5.08 x Re x1 2 Pr 13
. (10.56)

With T (x), we write the dimensional form of the temperature profile near the wall as

2
y y
T ( x, y ) TS T 2 1 T . (10.57)
T T
Convection Heat Transfer 429

The point of this analysis is to understand the energy and momentum transport near
the plate, and also to obtain a heat transfer coefficient on the surface. Again, the
energy balance at y = 0 is

T
qS h TS T kf ,
y0

where temperature gradient in the fluid can be written

Tf TS T 2 TS T
. (10.58)
y 0 T 0 T

The heat transfer coefficient is then

kf
h 2k f T 0.393 Re1x 2 Pr1 3 (10.59)
x

and

Nux hx k f 0.393 Re1x 2 Pr1 3 . (10.60)

This result is the same as the scaling analysis, except now we have a value for the
coefficient that scaling analysis cannot produce. As noted, this value depends on
the assumed profiles through their influence on the shape factor, e.g., Eq. (10.54).
“Exact” similarity solutions like those found by Blasius for the flow field [5] were
developed for forced convection heat transfer on the same isothermal plate modeled
by Pohlhausen [6], which analysis produced

Nux hx k f 0.332 Re1x 2 Pr1 3 , (10.61)

within 16% of Eq. (10.60). Figure 10.13 shows temperature profiles predicted by the
integral solution at several positions along the expanding thermal boundary layer. At
a specific height (y = yo), we see that temperature increases as fluid moves down-
stream. That increase and the expansion of the boundary layer is due to the conduc-
tion from the plate. The boundary layer thermal resistance increases and the rate of
heat transfer into the fluid slows with increasing x.
So far the results we have are for local heat transfer coefficients, h(x), and we can
see that variation downstream in Eq. (10.59) and Figure 10.11. The average heat
transfer coefficient, h , is obtained by averaging h(x) between x = 0 and x = L, where
L is the length of the plate, as

1 L
h h x dx, (10.62)
L 0
430 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.13 Temperature profiles and boundary layer development in flow over a hot flat
plate. A horizontal line in the boundary layer shows how the temperature at a specific height
increases farther downstream.

which, for an isothermal flat plate, is found from Eq. (10.59):

12
0.393 k U L k
h Pr1 3 x 12
dx 0.786 Re L1 2 Pr1 3 , (10.63)
L 0 L

which is twice the local value at the end of the plate, h(x = L). The definition of the
average Nusselt number is then

hL
Nu L C Re L1 2 Pr1 3 , (10.64)
k

with C = 0.786 from the integral solution (10.60) and C = 0.664 from the exact solu-
tion (10.61).
While information from an expression of the local heat transfer coefficient
is always interesting and is frequently useful, h(x) is not always available mathe-
matically. Many correlations of heat transfer coefficients (or Nusselt numbers) are
based on the average h because the experiments from which they are drawn were not
designed to measure spatially detailed local temperature and heat flow data.

Example 10.3
Calculation of Heat Transfer Coefficients and Heat Loss from Isothermal
Flat Plate under Laminar Flow
Consider air at 20 °C flowing over a flat plate at U∞ = 5 m/s. The plate is 1 m in length
(L), 0.5 m in width (W), and is held at the uniform temperature TS = 100 °C. The
values of the physical properties of the fluid to be used in convection heat transfer
Convection Heat Transfer 431

problems are those for the film temperature, Tf, which is the arithmetic average of TS
and T∞,

Tf TS T 2.

In the present example, Tf = (100 °C + 20 °C)/2 = 60 °C, at which temperature the

properties for air are

Pr = 0.708 = / = 1.89 10 5 m 2/s k = 0.0284 W/mK

Find: (a) Whether the velocity boundary layer on the plate is laminar or turbu-
lent; (b) and T as functions of distance from the leading edge; (c) local
heat transfer coefficient and heat flux as functions of x; (d) the fraction of
the total heat flow from the first 0.5 m of plate length, and the fraction from
the last 0.5 m of plate length; (e) a plot T(x) at y = 0.002 m, 0.005 m, and
0.008 m.
Solution: (a) In Section 9.5.2, we suggested that a transition to turbulence in
this geometry might begin at Recrit ~ 5 X 105. Beyond that point, the flow
is usually either in transition or fully turbulent. In the present example, the
Reynolds number at the trailing edge of the plate at x = L = 1 m is

U L 5 m s 1m
Re L 2.65 105 Re crit .
1.89 10 5 m 2 s

Therefore, the boundary layer does not begin transition to turbulence and is lami-
nar over the entire plate.
(b) From Eq. (9.38), the laminar velocity boundary layer, , is

12
x 5.48 x Re x 1 2 5.48 U x 12

12
5m s 12
5.48 x 0.0107 m1 2 x 12
.
1.89 10 5 m 2 s

This variation is shown in Figure 10.14 (a). From Eq. (10.55),

0.928 0.928
T x 13
x 0.0107 m1 2 x 12
0.0111 m1 2 x 12
,
Pr Pr1 3

which is also shown in Figure 10.14 (a). The Prandtl number value just below 1
causes the thermal boundary layer to develop slightly more rapidly than the momen-
tum boundary layer.
432 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.14 Results for air over a flat plate. (a) The variation with downstream distance
of the momentum and thermal boundary layer thicknesses and the local heat transfer coef-
ficient; (b) fluid temperature variation with x at three values of y (the dotted horizontal lines
in (a)).

(c) From Eq. (10.59),

12
h x 0.393k Pr1 3 U x 12

12
13 5m s 12
0.393 0.0284 W mK 0.708 x
1.89 10 5 m 2 s
5.12 W m 3 2 K x 12
.

This function is shown in Figure 10.14 (a). It can be noted that h(x) ∞ as x 0.
This artifact arises from Eq. (10.58), in which T y y 0 as x and T approach
zero. The local heat flux from the plate is thus
Convection Heat Transfer 433

qS x h x TS T 5.12 W m 3 2 K x 12
80 K 409 W m 3 2 x 12
.

(d) We have the Reynolds number at the trailing edge of the plate, from part (a),
which we use to find the average plate Nusselt number from Eq. (10.64),

12
Nu L hL k 0.786 Re L1 2 Pr1 3 0.786 2.65 105 (0.708)1 3 360.

Thus,

hL Nu L k L 360 0.0284 W mK 1m 10.2 W m 2 K

and

qS hL WL TS T 10.2 W m 2 K 0.5 m 1m 100 o C 20 o C 409W,

that is, the total rate at which heat flows from the entire plate to the air is 409 W.
To compare the heat loss rate from the two halves of the plate (one closer to the
leading edge and one closer to the trailing edge), we begin by giving a more general
definition of the average heat transfer coefficient than found in Eq. (10.62). Here we
integrate between two arbitrary points on the plate, x1 and x2:

12
1 x2 0.393 U x2
h h x dx k Pr1 3 x 12
dx
x2 x1 x1 x2 x1 x1

12
U x12 2 x11 2
0.786 k Pr1 3 .
x2 x1

Thus, for the first half of the plate, starting at the leading edge, x1 = 0 and x2 = 0.5:

12
W 5m s 0.5 m 0
hLE 0.786 0.0284 5 2
0.7081 3
mK 1.89 10 m s 0.5 m 0

14.4 W m 2 K

and, for the area ending at the trailing edge, x1 = 0.5 and x2 = 1.0,

12
W 5m s 1m 0.5m
hTE 0.786 0.0284 5 2
0.7081 3
mK 1.89 10 m s 1 m 0.5m

6.0 W m 2 K .
434 An Introduction to Transport Phenomena in Materials Engineering

Note that the average of these two values, (14.4 W/m2K + 6.9 W/m2K)/2 = 10.22 W/m2K,
is the average for the entire plate. So, the rate of heat loss from the first 0.5 m is

qS , LE hLE W L / 2 TS T 14.4 W m 2 K 0.5 m 0.5m 100 o C 20 o C

2899 W

and from the last 0.5 m is

qS ,TE hTE W L / 2 TS T 6.0 W m 2 K 0.5 m 0.5m 100 o C 20 o C

120 W.

Thus, 71% of the heat loss is from near the leading edge, while 29% is from near
the trailing edge. This result is expected from the shape of the h(x) curves in Fig-
ure 10.11 and Figure 10.14 (a). Adding more area to the end of the plate does in-
crease the overall heat flow but is a case of diminishing returns.
(e) The variations of the temperature of the air with distance downstream (x) at
constant y values are found using our temperature solution in Eq. (10.57), using T
found in part (b) above. The T(x) function for constant y values of 0.002 m, 0.005 m,
and 0.008 m are shown in Figure 10.14 (b). At y = 0.002 m, the air enters the ther-
mal boundary layer at x = 0.03 m and thereafter is progressively heated by convec-
tion from the plate. At y = 0.005 m, the air enters the thermal boundary layer at
x = 0.22 m and at y = 0.008 m entrance occurs at x = 0.52 m.

We discussed in Section 9.5.2 that, as the fluid moves farther downstream on a flat
plate, it reaches a point at which the upstream laminar flow begins to break down
and transition to turbulence ensues. For these flows, analytical solutions for temper-
ature and velocity fields are practically non-existent. Our approach is based on the
Reynolds–Colburn analogy [7] between the diffusion of heat and momentum, which
suggests that

Nux Cf 2
, (10.65)
Re x Pr 1 12.7 Pr 2 3 1 Cf 2

where the constants on the right-hand side are fitted to experimental data by Gnieliski
[8] and are valid in the range of 0.6 < Pr <105. The skin friction in turbulent flat plate
flow in Eq. (9.71) is
0.455
Cf . (10.66)
ln 2 0.06 Re x

A simpler, power-law expression for Nux has been fit by Reynolds et al. [9] to data in
a much tighter range of Pr:

Nux 0.0291 Re 4x 5 Pr 0.6 , (10.67)

Convection Heat Transfer 435

which is valid for gases and other fluids with Pr ~ (1).

A comparison of the behavior of the laminar Nux, Eq. (10.60), and the two for
turbulent flow, Eqs. (10.65) and (10.67), is seen in Figure 10.15. The plot assumes a
simple gas (Pr = 0.7) and a critical Reynolds number of Rex,crit = 5 × 105.
Figure 10.15 suggests a severe discontinuity in h and so q at a specific loca-
tion on the surface. This sudden change is not unexpected as there is a significant
qualitative alteration in the nature of the flow from laminar to turbulent. In reality,
this alteration is not so abrupt. As we saw in Figure 9.14, the transition region from
laminar and fully turbulent flow occurs over a range of Rex values. If the transition
region is not a large portion of the total plate, then using the turbulent correlation here
through the transition region can give a reasonable estimate of the average heat loss
rate on a plate, even though it may poorly predict local values of h near the nominal
transition point.
Recent work by Lienhard [10] evaluates these laminar and turbulent models and
proposes a correlation that combines them and an empirical model built from meas-
urements in the transition region, building on earlier work by Churchill [11]. Lien-
hard uses an approximation of Rex,crit from Mayle [12] based on measured freestream
turbulence, but that level of detail is beyond the scope of this text. Here we assume
values of the critical Reynolds number for transition and direct readers to dig deeper
into the literature if they need a better estimate. Once we have an educated guess for
Rex,crit, Lienhard suggests a correlation that fits measured heat transfer from a flat
plate in the laminar, transition, and turbulent regions,

15
12
5 10 10
Nux Nulam
x Nuxtrans Nuxturb . (10.68)

FIGURE 10.15 Nusselt number correlations on a flat plate over laminar, transition, and
fully turbulent flow regimes, assuming Rex,crit = 5 x 105. Use of only the laminar and turbulent
models shows a discontinuity at the Rex,crit, which is smoothed out by Lienhard’s correlation,
Eq. (10.68) [10].
436 An Introduction to Transport Phenomena in Materials Engineering

Nulam
x is found from the laminar model, Eq. (10.60), and the turbulent Nusselt num-
ber is from Eq. (10.65) or (10.67). The transition region heat transfer is estimated by

c
Nuxtrans Nulam
x ,crit Re x Re x ,crit . (10.69)

Churchill originally suggested c = 3/2, but Lienhard accounts for a dependence on

critical Reynolds number,

c 0.9922 log10 Re x ,crit 3.013. (10.70)

(For Rex,crit = 5 × 105, c = 2.64.) Based on data used to construct these correla-
tions, they are valid for 0.7 < Pr < 257, 4 × 103 < Rex < 4.3 × 106, and a freestream
turbulence less than 5%. The composite correlation, Eq. (10.68), is plotted in
Figure 10.15, showing a smoother and more gradual change from laminar to tur-
bulent conditions.

Example 10.4
Convection Heat Transfer from Isothermal Flat Plate with Laminar
and Turbulent Flow
Consider air flowing over a flat plate, under the same conditions as in Example 10.3,
except with twice the freestream strength, U∞ = 10 m/s, and twice the plate length,
L = 3 m. We assume that Rex,crit = 5 × 105.

Find: (a) Point on plate where transition to turbulence begins; (b) plot h(x) for
the entire plate.
Solution: (a) At the trailing edge (x = L),

U L 10 m s 3m
Re L 5 2
1.59 106 Re x,crit ,
1.89 10 m s

which indicates a laminar to turbulent flow transition occurs on the plate. At xcrit,

U xcrit 10 m s xcrit
Re x ,crit 5 105 .
1.89 10 5 m 2 s

or

xcrit 0.945 m.

Thus, the boundary layer is laminar over approximately the first third of the plate and
is turbulent or in transition over the rest.
Convection Heat Transfer 437

(b) Up to x = xcrit = 0.945 m, the flow is laminar and Eq. (10.60) is useful:
12
k U
hlam x 0.396 Re1x 2 Pr1 3 0.396 k Pr1 3 x 12
x
12
10 m s 13 12
0.396 0.0284 W mK 5 2
0.708 x
1.89 10 m s
32 12
7.229 W m K x .

This function is shown in Figure 10.16 along with the variation of h with x in the
turbulent boundary layer.
45
k U
hturb x 0.0291 Re 4x 5 Pr 0.0291k Pr x 15
x
45
10 m s 15
0.0291
0 0.0284 W mK 0.708 x
1.89 10 5 m 2 s
27.9 W m 9 5 K x 15
.

These equations and Lienhard’s composite correlation, Eq. (10.68), are shown in
Figure 10.16. The composite function matches the laminar solution for h(x) almost

FIGURE 10.16 Local heat transfer coefficients for air flow over an isothermal flat plate,
from laminar, transition, and turbulent correlations. Here flow is laminar for the first third of
the plate, in transition for the middle section, and fully turbulent only near the trailing edge.
The plate average coefficient is found from integrating Lienhard’s correlation from the lead-
ing edge to x = L.
438 An Introduction to Transport Phenomena in Materials Engineering

to xtrans. After transition begins, Lienhard’s function shows a rise in h, smoothly ap-
proaching and then matching the fully turbulent solution. The plotted plate average
heat transfer coefficient is calculated from the Lienhard’s correlation.

An integral analysis of laminar boundary layer flow in the same configuration as

earlier, except with a surface with a uniform heat flux, produces results similar to
Eqs. (10.59) and (10.60):
k
h 0.453 Re1x 2 Pr1 3 and Nux 0.453 Re1x 2 Pr1 3 , (10.71)
x
where the only difference is a slightly higher constant in the expressions. Using the
convection rate equation, we can find an expression for the temperature distribution
on the plate:

qS qS
TS x T 2.21 x Re x 1 2 Pr 13
~ x1 2 . (10.72)
h x k

Given the x dependence in Eq. (10.72), we see that the temperature difference across
the boundary layer, (TS – T∞), is zero at the leading edge and increases as the square
root of the distance downstream. For the transition and fully turbulent regimes, the
correlations for a uniform wall temperature found earlier are also valid for the uni-
form heat flux case. Lienhard’s composite correlation, Eq. (10.68), can also be used
with Eq. (10.71) for the local laminar Nux .

Example 10.5
Temperature Distribution on Flat Plate with a Uniform Heat Flux Due to Forced
Convection
In Example 10.3, we considered air flowing over a flat plate with a uniform tempera-
ture and found a total heat loss rate from the surface of 409 W. In this example, we
consider the same conditions except that the heat flow is uniform over the plate, so

qS qS A qS WL 409 W 0.5 m 2 818 W m 2 .

Find: (a) The hottest temperature on the plate; (b) the average temperature of
the plate.
Solution: (a) In Example 10.3, the flow is entirely laminar, so the temperature
distribution on the plate is found from Eq. (10.72). As the surface temperature
begins at T∞ and rises downstream, the hottest part of the plate is x = L = 1 m:

qS
TS ( L ) T 2.21 L Re L1 2 Pr 13
k
818 W m 2 12 13
2.21 1 m 2.65 105 0.708 139 o C.
0.0284 W mK
Convection Heat Transfer 439

The trailing edge temperature is thus

TS ( L ) 139 o C T 139 o C 20 o C 159 o C.

(b) To find the average temperature of the plate, we integrate Eq. (10.72) over
the length L.

1 L qS
TS ,ave T 2.21 x Re x 1 2 Pr 13
dx
L 0 k
12
2.21 qS L
T x1 2 dx
P 1 3 kL
Pr U 0

12 L 12
2.21 qS 2 32 1.47 qS
T x T L1 2
Pr1 3 kL U 3 0 Pr1 3 k U
12
1.47 818 W m 2 1.89 10 5 m 2 s 12
20 o C 13
1m
0.708 0.0284 W mK 5m s

20 C 92.6 o C
o

TS ,ave 113 o C

10.4.2 Forced convection correlations in otHer geoMetries

The previous section introduced the idea of the heat transfer coefficient and demon-
strated different ways to estimate its value as functions of geometry, flow conditions,
and fluid properties. The procedure has been to find the temperature field, use it to
predict the surface temperature gradient, and then calculate h(x) and h (and so Nu(x)
and Nu L ). But, for most practical problems, mathematical solutions cannot be found
more precisely than scaling analysis and often not even that well.
Another purpose of the analysis in the previous section is to understand the inter-
actions among the physical mechanisms, so to suggest a form for

h f U , T ,fluid properties,geometr or Nu f Re, Pr . (10.73)

But how do we relate the heat transfer coefficient in the Nusselt number to the
flow conditions and geometry without reliable mathematical solutions? Numerical
analysis is also useful in many cases, but, given the variability of and reliance on
experimental data inherent in most turbulent flow and other models, some careful
measurements of the surface thermal conditions may be the most accurate approach.
To find the form of the functions suggested in Eq. (10.73), we need a method to
measure the heat transfer coefficient. Of course, the procedure is not to measure it
directly, but to measure other quantities and use them in heat transfer models, such as
440 An Introduction to Transport Phenomena in Materials Engineering

kf Tf
h or h q Ts T , (10.74)
Ts T y y 0

for conduction or convection at the wall, respectively. We should measure TS, T∞, and
the fluid temperature, Tf (y), in the boundary layer or measure qS , TS, and T∞, and use
those data to calculate the heat transfer coefficient.
For a given geometry, we suspect from the results of flat plate boundary layer
analysis that

h f U , T , k, , , c, L , (10.75)

where L is the characteristic length scale of the geometry. Correlating these seven
independent variables (in this case) requires many experiments, and it is much
easier with fewer variables. The procedure is to collect the physical parameters
into dimensionless groups, ideally by performing a scaling analysis to find their
proper forms and functional relationships. Here, we can reduce the order of the
problem represented by Eq. (10.75) from six to two by writing the nondimensional
function,

NuL f Re L , Pr . (10.76)

Collecting data for this function is simpler, as we could perform experiments in, say,
air to fix the Pr, vary the flow velocity to vary the Reynolds number, and finally find
the Nusselt number by calculating h values from measurements mentioned earlier.
That process produces a Nu-Re relationship for a given Pr and geometry, and can
be repeated with different fluids to understand the effect of the Prandtl number. Of
course, experiments with other fluids are more complicated, more expensive, and
messier than using air. Sometimes the data are collected only with either air or water
and then correlated with Prn, where n might be 1/3, based on flat plate boundary
layer results from earlier sections, at least for Pr ~ O(1) or greater. This process is
illustrated in Figure 10.17.
The direct way of collecting data for the calculation of a heat transfer coefficient
is to measure the surface and ambient temperatures and applied surface heat fluxes.
The most common method of measuring temperature is reading a voltage drop over
a thermocouple, which is made of two wires of dissimilar metals joined at one end
[13]. The voltage difference is correlated to the unknown temperature. The junc-
tion is attached to or buried inside a material to obtain its temperature; good ther-
mal contact must be maintained to provide a good measurement. Other techniques
include fiber optic probes, thermochromic liquid crystals, infrared thermography,
and interferometry. To measure heat flow rates, the easiest method is to generate
thermal energy by Joule heating (as mentioned in Chapter 2) in the body cooled
by the fluid. The heat flow rate, the average surface temperature, and the ambient
temperature will be enough to find the heat transfer coefficient from the convection
rate equation. It is important to note that, depending on the complexity of the flow
Convection Heat Transfer 441

FIGURE 10.17 Example of convection correlation developed from measurements. (a)

Measured temperature and heat flux data are used with Eq. (10.74) to find h = f(U∞) for a
specific geometry and fluid. (b) Plot of Nu/Prn = f(Re) calculated from data in (a).

and the geometry, the uncertainty in the heat transfer correlations assembled from
measurements is frequently reported in a range of 5–20%, occasionally rising to
much higher values. Reviews of temperature and heat flow measurements, as well
as different direct and indirect methods of calculating the heat transfer coefficient,
are found in [13, 14].
As noted earlier, these measurements are necessary because only in a very few
geometries do momentum and energy conservation equations admit mathemat-
ical solutions to find forced convection correlations. We found such solutions for
flat boundary layers in laminar flow, with isothermal, Eq. (10.60), and isoflux, Eq.
(10.71), surfaces. A mix of experimental data and mathematics produces heat trans-
fer correlations for the same geometry in turbulent flow, Eq. (10.65), and also over
both regimes and a wide range of Reynolds numbers, Eqs. (10.68)–(10.70).
The Nusselt number for forced convection from a horizontal cylinder to a fluid
flowing in a direction normal to the axis of the cylinder is correlated with ReD and Pr:

Nu D C Re D Pr1 3 . (10.77)
442 An Introduction to Transport Phenomena in Materials Engineering

TABLE 10.2
Empirical Data for NuD in Eq. (10.77) for Flow over a Cylinder [16]
ReD range C m
0.4–4 0.989 0.330
4–40 0.911 0.385
40–4000 0.683 0.466
4000–40,000 0.193 0.618
40,000–400,000 0.027 0.805

The empirical values of C and m to be used in Eq. (10.77) depend on the Reynolds
number as listed in Table 10.2. The length scale for both Nu and Re is the cylinder
diameter, D. This correlation is averaged over the entire surface area of the cylinder,
while local heat transfer coefficients have been estimated by integral analysis by
Khane et al. [15]. Those integral solutions, however, are only valid up to the separa-
tion point; in the wake there is no generally applicable solution.
A single unified correlation for a wide range of conditions (ReD Pr > 0.2) was
developed by Churchill and Bernstein [17]:

0.62
Nu D 0.3 14
Re1D2 Pr1 3 FD Re D . (10.78)
23
1 0.4 Pr

For ReD Pr > 0.2 and ReD < 10,000 (slow, laminar flow over the cylinder),

FD ReD 1, (10.79)

and, for a more general result (ReD Pr > 0.4),

45
58
FD ReD 1 Re D 282, 000 . (10.80)

This general correlation matches well the experimental data in [17]. The largest devi-
ation from data occurs in ReD ranges where the flow pattern suddenly shifts (e.g., 7
× 104 < ReD < 105), which is where the separation point changes rapidly to a loca-
tion farther from the leading stagnation point. The authors suggest that, except in
that range, changes in flow patterns cause only weak transitions in value or slope of
Nu D Re D and the given correlation is sufficient. For liquid metals, the Reynolds
and Prandtl numbers are expected to have the same exponents (as in our previous
analysis for Pr<<1), as shown in a correlation by Ishiguro et al. [18]:

0.413
Nu D Re D Pr 1 Re D Pr 100. (10.81)
Convection Heat Transfer 443

Finally, a correlation for a circular cylinder in crossflow with an isoflux surface was
found by Perkins and Leppert [19] for 40 < ReD < 105 and 1 < Pr < 300:

14
Nu D 0.30 Re1D2 0.10 Re 2D 3 Pr 0.4 s , (10.82)

where μ∞ is the viscosity at the freestream temperature (T∞) and μs is evaluated at the
cylinder surface temperature.
In addition to circular cylinders, we also have access to correlations for heat trans-
fer between gases and rods with different cross-sections, shown in Table 10.3. The
constants in this table are for the correlation

m
Nu D C Re D . (10.83)

TABLE 10.3
Constants for Correlation in Eq. (10.83) for Forced Convection to Gas
Moving over Noncircular Rods, Normal to Their Primary Axis [20]
Solid Shape ReD Range C m
(a) 5 × 103–10 × 103 0.222 0.588

(b) 2.5 × 103–8 × 103 0.160 0.699

(c) 5 × 103–100 × 103 0.675 0.092

(d) 4 × 103–15 × 103 0.205 0.731

(e) 5 × 103–19.5 × 103 0.144 0.638

(Continued)
444 An Introduction to Transport Phenomena in Materials Engineering

TABLE 10.3 (Continued )

Solid Shape ReD Range C m

(f) 5 × 10 –100 × 10
3 3 0.138 0.638

(g) 2.5 × 103–15 × 103 0.224 0.612

(h) 3 × 103–15 × 103 0.085 0.804

Example 10.6
Measurement of Air Velocity Using a Hot Wire Anemometer
Consider a horizontal platinum wire of length L = 0.1 m and diameter D = 0.0005 m,
which is heated by the passage of an electric current, I. The purpose of the wire
(called a hot wire anemometer) is to measure the velocity normal to it. Here the air-
stream is at T∞ = 25°C and the resistance heating of the wire is such that its surface
temperature, Ts, is 300 °C. The properties of air at the film temperature of (25 °C +
300 °C)/2 = 163 °C are

Pr = 0.701 = 0.812 kg / m 3 = 3.02 10 5 m 2 / s k = 0.0354 W/mK.

The specific resistance of the wire varies with temperature as

9 11 o
e 1.05 10 m + 3.93 10 m C Ts

and thus the electrical resistance of the wire varies with temperature

9 11 o
Ts L Ts L 4 1.05 10 m + 3.93 10 m C Ts L
e e
R 2
.
Ac D2 D2

At Ts = 300 oC, this relationship gives a total resistance for the wire of R = 0.0596 .

Find: (a) An expression for U∞ = f(I); (b) a value for U∞ when I is measured as
30 A, and (c) plot the function from part (a).
Solution: (a) The form of a correlation for flow normal to a circular cylinder
(the hot wire) is taken from Eq. (10.77). An expression for the velocity is
taken from that correlation as
Convection Heat Transfer 445

1m 1m
Nu D hD
U .
D C Pr1 3 D kair C Pr1 3

From the rate equation for convection on a cylinder,

q h DL Ts T .

The velocity can be written as

1m 1m
q I2R
U ,
D kair C Pr1 3 L Ts T D kair C Pr1 3 L Ts T

where all the heat convecting away from the wire is generated by the electrical
resistance.
(b) To plot U∞ vs. I, we first need values for C and m from Table 10.2, which rely
on ReD. Because we do not know the freestream velocity yet, the Reynolds number
is difficult to find, so we will assume ReD to be in the 40–4000 range (C = 0.683,
m = 0.466). This assumption will be checked once the velocity is found. Using these
values, the U∞(I) function is plotted in Figure 10.18, and the measured value at I = 30
A is U∞ = 81.4 m/s. The velocities from this model in the range of 0.24 m/s < U∞ <
240 m/s (from currents 7.7 A< I < 38.6 A) give the range of ReD = 4 to 4000, so the
choices of C and m above are valid for I = 30 A.

FIGURE 10.18 Velocity vs. measured current for hot wire anemometer.

There are many processes that include the heating or cooling of spherical (or
near-spherical) droplets, including the production of steel shot (for peening) and vari-
ous powders. For heat transfer by forced convection from a sphere, the correlation [21]

14
Nu D 2 0.4 Re1D2 0.06 Re 2D 3 Pr 0.4 s (10.84)
446 An Introduction to Transport Phenomena in Materials Engineering

was developed for Prandtl numbers in the range 0.7 to 380 and Reynolds numbers in
the range 3.5 × 104 to 8 × 104, where all fluid properties except μs are evaluated at T∞.
This correlation works for Prandtl numbers of order 1 and higher; for flows of liquid
metals (Pr <<1), we can use the correlation from Witte [22],

12
Nu D 2 0.386 Re D Pr , (10.85)

for flow over a sphere in the range 3.6 × 104 < ReD < 2 × 105. For heat transfer by
convection from freely falling liquid drops in air, Ranz and Marshall [23] have pro-
posed for 0 < ReD < 200

Nu D 2 0.6 Re1D2 Pr1 3 . (10.86)

In these correlations for flow over spheres, the calculation of the Reynolds numbers
must be done with the velocity of the flow over the spheres; that is, the relative veloc-
ity of the fluid and the sphere, as noted in Section 9.6. As the sphere becomes small
enough that drag dominates its motion and it is carried along at the fluid velocity,
then the relative velocity approaches zero. In that case, the Nusselt numbers in Eq.
(10.84)–(10.86) go to 2, which accounts for a conduction-driven heat loss from the
sphere.
A common cooling/heating configuration is that of a flow impinging on a target
surface. As seen in Figure 10.19, that flow suddenly decelerates along the jet axis as
it approaches the surface. As the fluid turns the corner, it first accelerates along the
plate outward from the stagnation point. This wall jet then decelerates by friction on
the surface and by entrainment of otherwise quiescent fluid from above. In general,
we can count on higher heat transfer coefficients nearer a leading edge of a boundary
layer, compared to farther downstream. We see from Figure 10.19 that the boundary
layer in this configuration starts at the plate center, so the average distance from

FIGURE 10.19 Streamlines of a two-dimensional slot jet impinging on a plate below.

Convection Heat Transfer 447

the stagnation point to other surface position is shorter than if the flow was entirely
parallel to the plate. These effects will give a much higher plate average heat transfer
coefficient than that in parallel flow; further increase in the heat transfer coefficient
near the stagnation point is caused by the downward momentum there, which thins
the boundary layer.
There are many configurations for single jets and multiple jet arrays onto various
surfaces, and their heat transfer behavior is discussed in reviews by Martin [24], Webb
and Ma [25], and Zuckermann and Lior [26]. One significant distinction among these
configurations is between submerged and free jets. A free jet is a liquid flowing into a
gas environment with a sharp boundary between the two phases. Submerged jets, by
contrast, are any fluid flowing into an environment of the same fluid. As two simple
examples, we limit our consideration to cases of submerged jets flowing through a
single slot nozzle and a single round nozzle, Figure 10.20.
The empirical heat transfer correlation for a single two-dimensional slot jet
impinging on a finite flat plate, Figure 10.20 (a), is of a familiar form,

Nu Dh G Re mDh Pr 0.42 , (10.87)

except that G where C and m are complicated functions of the geometry [24]:

1
43
3.06 1 L 1 H
G and m 0.695 3..06 . (10.88)
LW H W 2.78 2W 2W

The use of this correlation should be limited to 3000 < Re < 90,000, 2 < H/W < 10, 4
< L/W < 20, and a Prandtl number around 1. Both the Nusselt and Reynolds number
are based on the hydraulic diameter of the slot,Dh 4 Area Perimeter 2W .
A similar correlation is found in [24] for a single round nozzle, Figure 10.20 (b),

Nu D G F Re D Pr 0.42 , (10.89)

where the geometric function is

D 1 1.1 D R
G (10.90)
R 1 0.1 H D 6 D R

and the dependence on the vigor of the jet is

12
F Re D 2 Re1D2 1 0.005 Re 0D.55 . (10.91)

This correlation is based on experiments in gases in the ranges 2000 < ReD < 400,000,
2 < H/D < 12, and 2.5 < R/D < 7.5. Both the Nusselt and Reynolds number are based
on the nozzle diameter of the nozzle, D.
448 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.20 Submerged jets impinging on target plates. (a) A single slot nozzle, with a
stagnation line at x = 0. (b) A single round nozzle, with a stagnation point at r = 0.

The previous correlations are both based on experiments in gases, so the Prandtl
number exponents in Eqs. (10.87) and (10.89) are speculative. For a wider range of
Pr, other correlations are reviewed in [25]. For the single slot case [27],
0.664
Nu L 4.33 Re1W2 0.157 ReW
0.664
L W W Pr1 3 , (10.92)

and, for a single round nozzle on to a circular target [28],

2
Nu D 32.4 Re 0.D523 0.266 Re 0D.828 2 R D 8 D 2R Pr1 3 . (10.93)

The latter correlation is valid for 2000 < ReD < 40,000, R/D > 4, and H/D < 7.

Example 10.7
Design of Convection Oven for Polymer Curing
We want to cure continuously a polymer sheet by passing it through a convection
oven in which hot air at Tair is blown through a single slot nozzle onto the top surface
(and only the top surface) of the sheet. The sheet enters the oven at Tin and must exit
at Tout. To maintain profitability, the sheet velocity (VS) must stay above 0.01 m/s. The
Convection Heat Transfer 449

longest we can make the furnace is L = 20W and the velocity of the impinging air
(Vair) can be controlled between 0.5 m/s and 5 m/s.
The material properties, geometry, and conditions are:

air = 0.812 kg/m 3 vair = 3.02 10 5 m 2 / s k air = 0.0354 W/mK

Prair = 0.701 W = 0.1 m H = 0.3 m

We will assume that the velocity of the polymer sheet is slow enough compared to
the air speed over it to have only a small effect on the heat transfer.

Find: (a) The values of the geometric parameters in the proper correlation
for Nusselt number in the configuration. (b) Calculate the Reynolds and
Nusselt numbers for the air velocities at each end of the possible range. (c)
Plot h as a function of Vair over the possible air velocity.

Solution: (a) Because the fluid is air and it is a single slot configuration, we
select Eqs. (10.87) and (10.88), and begin by calculating the parameters in
those equations:

Dh W
L/W W/W
H/W

The geometric functions are then:

3.06 3.06
G 0.119
LW H W 2.78 20 3 2.78

and
1
43
L H
m 0.695 3.06
2W 2W
1
43
20 3
0.695 3.06
2 2
1
0.695 14.8 0.627

(b) Low Re: ReDh = Vair Dh/ = 0.5 m/s (0.2 m)/3.02 × 10–5 m2/s = 3310
High Re: ReDh = Vair Dh/ = 5 m/s (0.2 m)/3.02 × 10 –5 m2/s = 33,100

Substituting these values into the correlation:

0.627 0.42
Low ReDh: Nu Dh G Re mDh Pr 0.42 0.119 3310 0.701 16.5
450 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.21 An oven with a moving polymer sheet cured by heating from an impinging jet.

FIGURE 10.22 Average heat transfer coefficient as function of velocity of air jet in convec-
tion oven.

0.627 0.42
High ReDh: Nu Dh 0.119 33,100 0.701 70.0.

(Note that H/W, L/W, and both of the ReDh values are in the proper ranges for this
correlation.)
Based on Eq. (10.87), an expression for the heat transfer coefficient on the sheet
as a function of the jet velocity is

m
kair V 2W
h G air Pr 0.42 ,
2W air

which is plotted in Figure 10.22.

Convection Heat Transfer 451

10.5 INTERNAL FORCED CONVECTION

If a conduit’s inner surface temperature is different than the temperature of a fluid
flowing in it, heat transfer occurs between the fluid and surface, giving rise to spatial
variations in the fluid temperature, both along the flow direction and over the con-
duit’s cross-sectional area. Because the local fluid temperature varies with position, it
is useful to define the concept of a mean temperature of the fluid, also known as the
mixing cup or bulk temperature. This quantity is found at any location, z, downstream
of the entrance of a conduit, if a thin slice of fluid (as in Figure 10.23) is removed
and is adiabatically mixed to eliminate any internal temperature gradients. The mean
temperature, T , can be defined as

w cT dAc
Ac
T , (10.94)
w c dAc
Ac

where w (the axial velocity) and T vary over the cross-section Ac. For a cylindrical
pipe (dA = r dr d ) and uniform fluid properties,

2 R R
wT rdr d wT rdr
0 0 0
T 2
, (10.95)
R
w rdr d w R2
0 0

where the average velocity in a horizontal pipe is, from Eq. (8.36),

R 2 dP
w .
8 dz

Expressions for T depend on the radial dependence of velocity,

2
R2 dP r
w 1 , (8.34)
4 dz R

and temperature (found later).

FIGURE 10.23 Volume element for consideration of energy balance for fluid flow in a
cylindrical pipe.
452 An Introduction to Transport Phenomena in Materials Engineering

Figure 10.23 shows a fluid flowing at a mass flow rate, m, through a horizontal,
cylindrical pipe of diameter D and length L. The mean temperature of the enter-
ing fluid is Tin and it exits at a higher Tout . The balance of energy rates in the pipe
includes heat picked up by the fluid from the surface and the temperature rise from
Tin to Tout and is written as

qS m c Tout Tin .
(10.96)
heat rate from rate of enthalpy
pipe surface risse in flow

Eq. (10.96) is a general result for the temperature rise in a flowing fluid in any con-
duit with the total surface heat input rate of qS. This relationship is unaffected by
conduit geometry and is independent of the thermal boundary conditions on the inner
surface.
We seek to find the relationships among the surface temperature, heat flux, and
mean fluid temperature along the conduit under different thermal boundary condi-
tions. To start, we apply the same approach as in Eq. (10.96), finding a differential
energy balance on an element of differential length, dz, in Figure 10.23. To begin, we
assume a uniform surface heat flux, so that

dT qS p
qS dAS qS p dz m c dT or , (10.97)
dz mc

where p is the perimeter of the conduit wall.

A cylindrical pipe has a uniform perimeter (p = πD), so integration of Eq. (10.97)
from T z 0 Tin to T z gives

qS D
T z Tin z (10.98)
mc

for a uniform qS . This linear mean temperature rise along the flow direction is shown
in Figure 10.24 (a). To find the surface temperature variation for a uniform heat flux
input, we use

qS
qS h TS z T z or TS z T z , (10.99)
h

which defines the surface heat transfer coefficient for internal flow. Expressions for
h are discussed later.
We can also consider the case of a uniform surface temperature and seek the sur-
face heat flux along the length. Defining the difference between the surface and mean
fluid temperatures as

TS TS T , (10.100)
Convection Heat Transfer 453

FIGURE 10.24 Surface (TS) and mean (T ) fluid temperatures along a pipe length (z). (a) Uniform
surface heat flux. (b) Uniform surface temperature. LE,th is thermal entry length for uniform
flux case.

we show that this difference changes along the length as

d TS d TS T dT . (10.101)

Combining Eq. (10.97) and

qS z h TS T z , (10.102)

we find

qS D h TS D
dT dz dz (10.103)
mc mc

or

d TS h D
dz. (10.104)
TS mc

Integrating from z = 0, where Tin = (TS – Tin ), to an arbitrary distance downstream

(z) where ΔTS(z) = [TS – T (z)] and noting that the heat transfer coefficient averaged
over that length is
z
hz h z dz,
0
454 An Introduction to Transport Phenomena in Materials Engineering

gives
TS z Dh
ln z (10.105)
Tin mc
or
Dh
TS z TS T z Tin exp z . (10.106)
mc

The exponential variation of T z with a uniform TS is shown in Figure 10.24 (b).

Eq. (10.96) can be written as

qS m c Tout Tin mc TS Tin TS Tout

(10.107)
mc Tin Tout .

From Eq. (10.105),

Dh
mc z, (10.108)
ln Tin Tout

substitution of which into Eq. (10.107) gives

Tin Tout
qS Dhz h A Tlm , (10.109)
ln Tin Tout

in which the log mean temperature difference, ΔTlm, is defined as

Tin Tout
Tlm . (10.110)
ln Tin Tout

So far, we have relationships between surface and mean fluid temperatures for two
ideal boundary conditions, but we lack information about heat transfer coefficients
in Eqs. (10.99) and (10.109). We will take an approach similar to finding heat trans-
fer coefficients in external flow: first, we look to analytical solutions for specific
cases in laminar flow, and then we appeal to empirical correlations for more com-
plex conditions. To start, we recall the case of steady (∂/∂t = 0) and axisymmetric
(∂/∂ = 0) flow in a cylindrical pipe with no swirl (v = 0). The velocity field is fully
developed (∂w/∂z = 0, u = 0). (The velocity is written in cylindrical coordinates,
V u rˆ v ˆ w ẑ.) For a hot surface, both the surface temperature, TS, and the mean
temperature of the fluid, T , increase as the fluid moves down the pipe, and, after
some entry length, a fully developed temperature profile is attained, in which the ratio

TS z T r, z
TS z T z
Convection Heat Transfer 455

is a unique function of r/R and is independent of z. In this relationship, T(r, z) is the

local fluid temperature. A heat transfer coefficient is defined in Eq. (10.99) and the
no-slip condition at the inner surface means conduction is the only mode of radial
heat transfer there (r = R), so

T
q k h TS T z . (10.111)
r R

Thus, the heat transfer coefficient can be defined as

k T r R
h . (10.112)
TS T

Because TS T TS T is a function only of radius, the partial derivative

TS T R T
(10.113)
r R TS T R
TS T r R

has a unique value independent of z. If we assume a uniform q , Eqs. (10.112) and

(10.113) show that h, q , and TS T are all uniform. Thus, TS and T increase at the
same rate with increasing z, or

TS T
. (10.114)
z z

Because TS T TS T is not a function of z,

TS T
0,
z TS T
then

TS T TS T TS T TS T
2 2
0,
TS T z z TS T z z

which, by comparison with Eq. (10.114), shows that

TS T T
. (10.115)
z z z
The thermal energy equation then can be written as

T T
w r .
z r r r (10.116)
advection radial
downstream conduction
456 An Introduction to Transport Phenomena in Materials Engineering

Combination of Eqs. (8.34) and (8.36) gives the radially dependent axial velocity
field, w(r),
2
w r 2w 1 r R , (10.117)

and substitution into Eq. (10.116) gives

2
r T T
2w 1 r . (10.118)
R z r r r

Integration twice over the radius, and applying a symmetry condition at r = 0 and a
uniform heat flux condition at r = R, gives the radial temperature distribution,

4 2
w T R2 r r
TS T 4 3 . (10.119)
z 8 R R

The difference between the surface and mean temperature and the temperature gradi-
ent at the wall are obtained from using Eq. (10.95) and (10.119):

w T 2
TS T 0.229 R (10.120)
z

and

T w T R
. (10.121)
r R z 2

Substitution of Eqs. (10.120) and (10.121) in the definition of the heat transfer coef-
ficient, Eq. (10.112),

h k 2 0.229R and NuD hD k R 0.229 R 4.36,

which is valid for laminar flow in a cylindrical pipe with a uniform surface heat
flux. Similar analysis for the case of a uniform surface temperature gives NuD = 3.66.
We see that for laminar thermally and hydrodynamically fully developed internal
flow, the heat transfer coefficient is not a function of position or flow conditions.
Table 10.4 shows similar behavior for other conduit cross-sections, assuming lami-
nar, fully-developed flow, with the NuD based on the hydraulic diameter.
These NuD values are valid only with a fully developed temperature field, which
is achieved only some distance from the conduit entrance. The entry length of pipe
required for the establishment of a fully developed temperature profile is analogous
with the entry length required for establishment of fully developed hydrodynamic
Convection Heat Transfer 457

TABLE 10.4
Forced Convection Heat Transfer in Laminar Fully Developed Flow in Pipes
of Different Cross-Sections. The Nusselt Numbers Are Based on the
Hydraulic Diameter, Dh = 4A/p
Geometry NuDh (UHF) NuDh (UST)
(a) 4.36 3.66

(b) 3.09 2.98

(c) 3.86 3.34

(d) 3.02 3.39

(AR = 2)

(e) 2.93 4.44

(AR = 4)

(f) 2.90 5.60

(AR = 8)

(g) 3.00 2.35

(equilateral)
458 An Introduction to Transport Phenomena in Materials Engineering

flow shown in Figure 9.7. For laminar flow in a cylindrical pipe of diameter D, the
entry length required for establishment of a fully developed temperature profile is
approximately

L E , th 0.05 D Re D Pr . (10.122)

In turbulent cylindrical pipe flow, to a first approximation the flow becomes hydrody-
namically fully developed after an entry length of between 10 and 60 diameters and
becomes thermally fully developed after an entry length of between 20 and 40 diam-
eters. An empirical correlation referenced in [29] gives an average Nusselt number
and heat transfer coefficient for the entire length (L) of a pipe of diameter D, includ-
ing the entry length. The correlation is for a tube of uniform surface temperature and
approaches the laminar, fully developed relation as the entry length becomes less and
less of the entire pipe length (D/L 0):

hD 0.0668 D L Re D Pr
Nu D 3.66 23
. (10.123)
k 1 0.04 D L Re D Pr

As always, the situation is more complicated with turbulent flows. For a pipe with a
smooth surface, a simple, commonly used Nusselt number correlation is the Dittus–
Boelter equation:

hD
Nu D 0.023 Re 4D 5 Pr n , (10.124)
k

for 2500 < ReD < 107 and 0.7 < Pr < 120. The Prandtl number exponent is n = 0.4
if the surface heats the fluid and n = 0.3 if the surface is colder than the fluid. The
uncertainty in this correlation can be as high as 25%, but it can be applied to both
uniform flux and temperature conditions. Another set of correlations [30] also may
be used with both boundary conditions, each over a different range of Pr:

Nu D 0.0214 Re 4D 5 100 Pr 0.4 0.5 Pr 1.5 10 4 Re D 5 106

(10.125)
Nu D 0.012 Re 0D.87 280 Pr 0.4 1.5 Pr 500 3 103 Re D 106.

These relations have less than 5% difference between them and experimental meas-
urements, as does another correlation from [31]:

Nu D 5 0.015 Re aD Pr b 0.1 Pr 105 10 4 Re D 106

(10.126)
a 0.88 0.24 4 Pr b 13 0.5 exp 0.6 Pr .

The Dittus–Boelter correlation, Eq. (10.124), is within about 7% of the latter two
relations for turbulent flow at moderate Pr (gases) and has a deviation up to 27% at
Convection Heat Transfer 459

Pr = 100, with the greatest separation at the high end of the Reynolds number ranges.
For Pr << 1, as for liquid metals, the relations

0.8
Nu D 5.0 0.025 Re D Pr Re D Pr 100 (uniform temperature) (10.127)

and

0.827
Nu D 4.82 0.0185 Re D Pr Re D Pr 100 (uniform heat flux) (10.128)

give reasonable agreement with experiments [32].

Example 10.8
Heating Water Flowing through a Pipe with a Uniform Heat Input
Water flows through a 10 m length of pipe with a 0.05 m inner diameter at a mass
flow rate of m = 0.01 kg/s. We want to heat the water, which enters the pipe at a mean
temperature Tin = 22 °C.

Find: (a) The uniform heat flux, qS , from the pipe to produce a mean exit water
temperature of Tout = 72 °C at the exit from the pipe; (b) the surface tem-
perature at the pipe exit.
Solution: (a) The average temperature of the water in the pipe is (22 °C + 72 °C)/2
= 47 °C, at which temperature the properties are

= 5.79 10 4 Pa s = 989 kg / m 3 c = 4180 J / kg K

k = 0.637 W/m K n = 5.9 10 7 m 2 /s Pr = 3.79

Rearranging Eq. (10.98) to get the uniform surface heat flux, qS qS AS ,

mc 0.01 kg s 4180 J kgK

qS Tm,out Tm,in 72o C 22o C 1330 W m 2 .
DL 0.05 m 10 m

(b) To find the surface temperature, we must first find the heat transfer coefficient
based on the flow conditions. The Reynolds number for the flow is

4 0.01kg s
Re D VD 4m D 440.
0.05 m 5.79 10 4 Pa s

Because ReD < ReD,crit ≈ 2100, the flow is laminar. From Eq. (10.122), the entry
length required for the establishment of a fully developed temperature profile is

L E , th 0.05 D Re D Pr 0.05 0.05m 440 3.79 4.17 m,

460 An Introduction to Transport Phenomena in Materials Engineering

and so, the thermal entry length takes 42% of the total length. In the fully developed
region (z > 4.17 m), the surface to bulk fluid temperature difference, (TS – Tm), is
uniform. From Table 10.4, NuD = hD/k = 4.36 and so

h 4.36 0.637 W mK 0.05 m 55.5 W m 2 K .

Thus, through the region of the pipe with a fully developed temperature profile,

TS Tm q h 1330 W m 2 55.5 W m 2 K 24.0 o C;

hence, the surface temperature at the tube exit is

TS,out q h Tm,out 24 o C + 72 o C = 96 o C,

which is the maximum surface temperature.

Example 10.9
Heating Water Flowing through a Pipe with a Uniform Inner Surface Temperature
Consider the same flow situation as in Example 10.8 except that the surface tempera-
ture is maintained at 96 °C uniformly over the pipe’s inner surface. (Note that this value
is the maximum surface temperature in Example 10.8, which has a uniform heat flux.)

Find: The total rate of heat transferred to the water.

Solution: In this case, from Eq. (10.108),

mc Tin mc T Tm,in
h ln ln S
DL Tout DL TS Tm,out
0.01kg s 4180 J kgK 96o C 22o C
ln
0.05 m 10 m 96o C 72o C
W
30 ,
m2 K

and from Eq. (10.109),

W 74o C 24o C
qS h A Tlm 30 0.01 m 10 m 418 W.
m2K ln 72o C 24o C

10.6 NATURAL CONVECTION HEAT TRANSFER

Natural convection2 flows are induced by body forces, which act at every point
throughout a fluid, as opposed to forces that act on areas (e.g., pressure or shear) that
Convection Heat Transfer 461

drive forced convection. The body forces treated here arise from spatial variation of
density in a fluid, which is, in turn, caused by nonuniform temperature or composi-
tion fields, i.e., = (T, C).
The energy equation for natural convection is the same as for forced convection,

2 2
T T T T
c u v k , (10.129)
x y x2 y2

for a steady flow with uniform properties. The conservation equations governing
fluid behavior are continuity,

u v
0, (10.130)
x y

and momentum (here shown only in the vertical, x direction):

2 2
u u u u P
u v 2 2
gx . (10.131)
x y x y x

Here, gravity acts in the negative x direction, g g x iˆ g iˆ . These equations are

almost the same as forced convection, with the exception of the two last terms in
Eq. (10.131). These we will transform first into a function of temperature, coupling
momentum directly to Eq. (10.129).
We begin with the two Boussinesq approximations as a way to treat the variations
in density that drive the flow. First, density is assumed everywhere uniform, except
where it appears in the buoyancy term. Second, Δ / << 1, that is, there is only a
relatively small change in density. With the latter assumption, we can write = (T)
as a Taylor series in the vicinity of T∞ and truncated to two terms:

T T T . (10.132)
T T

The pressure gradient is known from hydrostatics:

P
g. (10.133)
x

Using the definition of the thermal expansion coefficient,

1
, (10.134)
T
462 An Introduction to Transport Phenomena in Materials Engineering

and rearranging Eq. (10.132) gives

p
g g g T T , (10.135)
x

which replaces the last two terms in Eq. (10.131):

2 2
u u u u
u v 2
g T T . (10.136)
x y x y2
advecttive acceleration viscous friction buoyancy

Examining Eqs. (10.129) and (10.136), we clearly see that there is a fundamental dif-
ference in how these equations interact compared to forced convection. With a forced
flow, the velocity can be solved first from the fluid equations, then the temperature
from the thermal energy balance. In contrast to that one-way coupling of momentum
and energy equations, natural convection systems have two-way coupling, that is,
energy and momentum must be solved simultaneously.
In cases with multicomponent fluids, the buoyancy can also be a function of com-
position, if it is nonuniformly distributed. Then, = (T, C) and Eq. (10.135) can be
written as

p
g g T T S C C , (10.137)
x

1
where the solutal expansion coefficient is S .
C

10.6.1 natural convection FroM an isotHerMal vertical Flat plate

The physics of natural convection are illustrated by examining the steady flow near a
semi-infinite vertical flat plate immersed in a stagnant fluid, as in Figure 10.25. The
plate is isothermal at TS, which is greater than the ambient temperature far from the
wall, T∞. The hot wall heats and so expands the fluid near it and the lighter material
rises up the wall. A boundary layer forms beginning at the leading edge (x = 0) and
grows downstream (which, just to be clear, is up and against gravity).
We begin with scaling the governing equations to find estimates and functional
dependence for velocities, temperature, boundary layer thickness, and heat transfer
coefficient. We start with these reference values in the thermal boundary layer:

xref ~ x, yref ~ T, and Tref ~ TS T , (10.138)

and scale the energy equation as

Tref Tref Tref Tref

uref , vref ~ 2
, 2
xref yref xref yref
Convection Heat Transfer 463

FIGURE 10.25 Boundary layers formed on an isothermal vertical flat plate by natural con-
vection in a fluid with Pr ≈ 1.

or

T T T T
uref , vref 2
~ 2
, 2
. (10.139)
x T x T

This equation can be simplified in two ways. First, we can compare the two conduc-
tion terms,

x direction conduction T T
~ 2 2
~ T x 1,
y direction conduction x T

which, given that T/x << 1 in the thin boundary layer, shows that the conduction
along the plate (x direction) is negligible compared to conduction across the layer (y
direction). The second simplification is based on a result from continuity,

uref vref T
~ or vref ~ uref . (10.140)
x T x
464 An Introduction to Transport Phenomena in Materials Engineering

Using Eq. (10.140) in Eq. (10.139), the advection terms are

T T T T
uref , vref 2
or uref , uref
x T
x x T

so

T T T
uref , uref ~ 2
.
x x T T

At this point, we must make some assumption about the order of magnitude of the
ratio of the thermal and momentum boundary layers, T/ , as we saw in the scaling
of the forced convection case. For the sake of simplicity, we will assume that T/ ~
1 and that Pr ~ 1. In this case, the two thermal advection terms are the same order of
magnitude, and

T T
uref ~ 2
x T
advection conduction
and
x
uref ~ 2
. (10.141)
T

We see again that heat conducts into the fluid at the surface and is then swept away by
advection in the thermal boundary layer. Using these results to scale the x direction
momentum gives
2
uref u
~ 2
, g T
x
advection friction buoyancy
and

x 1 x
2
~ 2 2
, g T. (10.142)
T
x T
advection friiction buoyancy

When rearranging Eq. (10.142) we can discover a vigor parameter (similar to the
Reynolds number discussed in this and earlier chapters) which can be used to char-
acterize many natural convection flows. This parameter is the Rayleigh number,

g TS T x3
Ra x , (10.143)
Convection Heat Transfer 465

which appears in the scaled momentum equation thus,

4 2
x 1 x x 1
~ , 1. (10.144)
T Ra x Pr T Ra x
advection fricttion buoyancy

Because we have assumed Pr ≈ 1 and T/ ~ 1, Eq. (10.144) becomes

T ~ x Ra x 1 4 (10.145)

and so
x
uref ~ 2
~ Ra1x 2 . (10.146)
T
x

We see that T ~ x1/4 and uref ~ x1/2, so both quantities increase downstream. While
velocity at any given y value decreases farther from the leading edge in forced con-
vection, the characteristic velocity rises in natural convection. In the former case,
friction at the wall bleeds momentum from the freestream, while, in the latter,
momentum is added continuously to the thermal boundary layer by heat from the
wall expanding the fluid.
Scaling the thermal boundary condition at the vertical wall will give us estimates
for the heat transfer coefficient.

T TS T
k ~k ~ h TS T
y y 0 T

k
h~ ~ k Ra1x/ 4 x (10.147)
T

A similar but much expanded procedure for Pr << 1 and Pr >> 1 is found in [3,
33], and the results are presented in Table 10.5.

TABLE 10.5
Results of Scaling Analysis in the Thermal Boundary Layer
in Natural Convection on an Isothermal, Vertical Flat Plate
Pr << 1 Pr >> 1

T 14 14
x Ra x Pr x Ra x
u
12 12
Ra x Pr Ra x
x x
h k k
14 14
Ra x Pr Ra x
x x
Nux 14 14
Ra x Pr Ra x
466 An Introduction to Transport Phenomena in Materials Engineering

To improve on these order of magnitude estimates of boundary layer behavior,

integral solutions have been found. Over a much wider range of Pr, Bejan [33] shows
integral solutions for Nusselt numbers for laminar flow near a vertical flat plate:
For Pr >> 1:
14
hx 3 q3
Nux Ra1x 4 where
k 8 q 1 q 12 q 2

5 2 q 12
Pr q
6 q 2
For Pr << 1:
14
hx 3 q2
Nux 2
Ra1x 4 where
k 8 2q 1

2
5 q2
Pr
3 q 1

Another treatment of an isothermal, vertical flat plate is based on experimental data

over a wide range of Rayleigh numbers in the laminar regime [34]. The heat transfer
coefficient here is averaged over a plate of height, H.

0.67 14
Nu H 0.68 49
RaH Pr (10.148)
9 16
Pr 0.671

So far, we have concentrated on laminar flows, but a transition to turbulence will

occur somewhere above the leading edge, if the flow is vigorous enough. Compar-
ison of many experimental data sets in [35, 36] shows that the last laminar flow
observed on a vertical flat plate is approximately Rax ≈109 Pr. So we can define a
different vigor parameter for natural convection flows, the Grashof number,

Ra x g TS T x3
Grx 2
, (10.149)
Pr

and conclude from the data that transition to turbulence is underway around Grx ≈
109. For a correlation for heat transfer at the isothermal wall in the transitional and
turbulent (in addition to laminar) regimes (RaH < 1012), Churchill and Chu suggest
this equation from experimental data:

0.387 16
Nu H 0.825 8 27
RaH Pr . (10.150)
9 16
Pr 0.671
Convection Heat Transfer 467

10.6.2 natural convection FroM otHer geoMetries

As in forced convection, we have used mathematical solutions for flow and heat
transfer near a flat plate to elucidate the mechanisms that control the behavior in
natural convection. Given that such solutions are only found for a few simple config-
urations, once again we rely on the results of careful experiments to correlate surface
heat transfer coefficients to geometries and the conditions that drive buoyant flows.
From experimental observations, the correlation of NuD with RaD for free convec-
tion from a horizontal cylinder is given by

NuD C RaDn (10.151)

and the values of C and n to be used depend on the Rayleigh number as listed in
Table 10.6.
A comprehensive correlation, including data for a wide range of conditions (up to
RaD ≈ 1012), is reported [37] as
2

0.387 16
Nu D 0.60 8 27
RaD Pr . (10.152)
Pr 9 16 0.671

For a horizontal cylinder in liquid metal (Pr << 1) [38],

14
Nu D 0.53 RaD Pr . (10.153)

Natural convection from a sphere was examined by Yuge [39], who found

0.392 14
Nu D 2 12
RaD Pr . (10.154)
Pr
Another interesting configuration is buoyancy-induced flows caused by heating or
cooling horizontal, finite flat plates. In Figure 10.26, we see four possible configura-
tions. With a heated surface facing upwards or a cooled surface facing downwards,

TABLE 10.6
Correlation Constants for Natural Convection over
a Horizontal Cylinder
RaD C n
10 –10
–10 –2 0.675 0.058
10–2–102 1.02 0.148
102–104 0.850 0.188
104–107 0.480 0.250
107–1012 0.125 0.333
468 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.26 Horizontal plates with buoyancy-induced flows (a) away from and (b)
toward the surfaces. The ambient temperature in each sketch is either less than TH or more
than TC.

fluid is pulled in from the sides of the plate by the formation of a plume moving away
from the surface, driven by the temperature difference between surface and ambient.
If we define a characteristic length for an arbitrary horizontal, plate surface as

surface area
L , (10.155)
surface perimeter

we have correlations for heat transfer,

Nu L 0.54 Ra1L 4 105 RaL 107

(10.156)
Nu L 0.15 Ra1L 3 107 RaL 1010 .

In the opposite case, the heated or cooled fluid spills over the edge of the plate and
that motion draws more fluid from the ambient to the center of the plate, forming a
weak impingement region there. A correlation for that system is

Nu L 0.27 Ra1L 4 105 RaL 1010 . (10.157)

There are many other published correlations for a wide variety of shapes, and we
cannot include them all here. As a first-order approximation of natural convection
heat transfer from an arbitrary shape, Lienhard [40] developed a correlation

Nu l 0.52 Ra1l 4 (10.158)

The length scale, , is the vertical length of the unseparated boundary layer on the
shape.
Convection Heat Transfer 469

10.7 BOILING HEAT TRANSFER

The convection phenomena discussed earlier in this chapter are limited to heat trans-
fer in a single-phase fluid medium. In this section, we introduce convection heat
transfer with boiling, that is, a liquid-to-vapor phase change caused by heat addition
at a surface, when the surface temperature is greater than the saturation tempera-
ture of the liquid. Figure 10.27 shows two common configurations. In part (a), pool
boiling occurs in a liquid that has no liquid motion except that caused by the heat
addition and the phase change, while, in part (b), forced convection boiling has an
imposed liquid flow across the hot surface. This flow increases the rate of surface
heat transfer and changes the pattern of vapor production, with progressively more

o
FIGURE 10.27 Adding heat to cause boiling, which occurs if TS TSAT
P 1atm
100 C
in water. (a) Saturated pool boiling. (b) Successive regimes of subcooled forced convection
boiling in a pipe.
470 An Introduction to Transport Phenomena in Materials Engineering

vapor fraction as the fluid picks up more heat from the surface. When the bulk mass
of the liquid is already at the saturation temperature (which depends on the local
pressure), it may boil (saturation boiling); the bulk liquid may also boil if there is
enough thermal energy applied to first raise the fluid temperature near the heated
surface up to and above the saturation temperature (subcooled boiling). In this sec-
tion, we introduce heat transfer with liquid-vapor phase change by concentrating on
saturated pool boiling.
The liquid-vapor system has different vapor bubble morphologies that depend
on the surface temperature and heat flux, and these bubble shapes have a profound
effect on the heat transfer rate. One way to map out these different boiling regimes is
to examine the boiling curve for a particular fluid (here we will restrict our consid-
eration to water). The boiling curve for water is shown in Figure 10.28, in which the
excess temperature, Te TS – Tsat , is plotted against the surface heat flux, qS . This
curve assumes the bulk liquid is at the saturation temperature, Tsat, and is quiescent.
First, we begin with the usual way of exploring the boiling curve, by starting at
qS 0 Te , raising the heat flux, and finding the corresponding excess temperature
at each qS . This “heat-flux controlled” boiling experiment was first performed by
[41]. At the lowest heat fluxes, region (i) in Figure 10.28, thermal energy is not added

FIGURE 10.28 The pool boiling curve for liquid water at 1 atm, where excess temperature
ΔTe= TS – TSAT. The path is heat flux–controlled, with a jump with constant flux from nucleate
to film boiling.
Convection Heat Transfer 471

quickly enough to the liquid to produce vapor and all the heat drives a single-phase
natural convection (NC) flow, and

qS hNC Te , (10.159)

where hNC is found from the appropriate correlations, such as in Section 10.6.
As the heat flux is increased, the surface reaches the onset of nucleate boiling
(ONB). At isolated points, usually in small surface defects, vapor bubbles form as
heat from the surface is transported through the liquid to a liquid-vapor interface,
causing local evaporation. Most often, this process initiates in small crevices that
the liquid never completely fills when it first floods the surface. More heat is added
over time and more liquid vaporizes, enlarging the bubbles. Once a bubble is large
enough that its buoyancy is sufficient to overcome the surface tension binding it
to the solid, it detaches from the surface and rises in the liquid pool. As the bub-
ble moves up, nearby colder liquid that was around the growing bubble, rushes
in to take its place on the hot surface, thereby quenching it locally. This cycle of
liquid-to-vapor transformation and surface quenching occurs with great rapidity
and greatly enhances the heat transfer rate compared to the single-phase natural
convection at lower heat flux. This isolated bubble regime, Figure 10.29 (a), occurs
at the lower heat flux range of the nucleate boiling regime (NB), region (ii) in

FIGURE 10.29 Nucleate boiling: (a) isolated bubble production at lower surface heat fluxes;
(b) vapor column and jet production at higher heat fluxes; (c) patches just before boiling crisis.
(d) Film boiling above the critical heat flux.
472 An Introduction to Transport Phenomena in Materials Engineering

Figure 10.28. With an even higher heat flux, ΔTe continues to increase (albeit at a
progressively slower rate) and more and more bubble nucleation sites are activated
and produce columns of bubbles, Figure 10.29 (b). This higher vapor production
rate enables the coalescence of vapor into multiple patches on the surface, Fig-
ure 10.29 (c), and for isolated bubbles to merge into columns of vapor rising from
patches on the surface. With the formation of these patches, more of the surface is
covered in vapor, increasing the thermal resistance between the hot surface and the
liquid-solid interface.
Rosenhow [42] correlated experimental data for saturated pool boiling in the
nucleate boiling regime for a horizontal flat plate:

12 3 3
g L V cL Te 1
qS CS3 f L h fg . (10.160)
h fg PrLn

The variable g L v increases the surface heat flux with more bubble
buoyancy and decreases with lower surface tension ( ). These trends contribute to
faster heat flow because buoyancy drives bubble detachment and surface tension
makes it more difficult. With easier detachment, the bubbles leave the surface
more often, increasing qS and the frequency of the surface quenching effect as
bubbles leave. Another grouping, cL Te h fg , known as the Jakob number, is the
ratio of the sensible heating of the liquid from its bulk temperature (Tsat) to the
surface temperature (TS) and the latent heat (hfg) required to vaporize the liquid.
The leading constant, Cs-f, depends on the surface-fluid combination, as shown in
Table 10.7.
Increasing the surface heat flux further, the vapor patches eventually cover most
or all of the surface, fundamentally changing the heat transfer path from the surface
to the liquid-vapor interface. During nucleate boiling, that path is primarily through
the liquid, but that path is no longer available. The heat must transfer, primarily by
conduction and radiation, through the vapor blanket, which has a very high thermal
resistance. Because we are fixing qS and increasing Rsurface, the excess temperature
rises suddenly and significantly. With water at 1 atm, that jump in surface tempera-
ture may be as much as 1000 oC, with which many surfaces may not survive contact.

TABLE 10.7
Values of Cs,f for Various Surface-Fluid Combinations; n = 1
Surface-Fluid Combination Cs,f
Water – copper: scored 0.0068
Water – copper: polished 0.0130
Water – stainless steel: chemically etched 0.0130
Water – stainless steel: polished 0.0060
Water – brass 0.0060
Water – nickel 0.0060
Water – platinum 0.0130
Convection Heat Transfer 473

This qS at which this phenomenon occurs is known as the critical heat flux (CHF),
which, with the reasonable assumption at 1 atm of L >> v, is
14
g L
qCHF 0.149 v h fg 2
. (10.161)
v

Once the qS > qCHF , the boundary condition jumps over to region (iii) in Figure 10.28,
film boiling, which is characterized by a continuous surface vapor layer, which forms
at and above the critical heat flux. This layer both insulates and lubricates the surface,
requiring a ΔTe much higher than that needed for nucleate boiling and disconnecting
the liquid from a no-slip condition at the surface.
For film boiling on a horizontal cylinder, a combination of analysis and experi-
mentation gives an expression for the heat transfer:
14
hFB D vg L v h*fg D3
Nu D, FB 0.62 , (10.162)
kv v kv TS TSAT

where D is the diameter and h*fg is the latent heat of vaporization modified to include
the sensible heat required to raise the liquid temperature from Tsat to TS,FB,

h*fg h fg 0.4cv Te . (10.163)

At this point, we have seen the effect of increasing the heat flux from zero, with the
heat transfer to the liquid changing regime from natural convection to nucleate boil-
ing, then at the critical heat flux a sudden change to film boiling. This description of
a heat flux–controlled process with boiling is useful to familiarize us with its differ-
ent regimes, but the most common application for boiling in materials processing is
quenching. Here, we begin with a surface at a high temperature (let us assume high
enough to be in the film boiling regime) and the excess temperature falls continuously
to zero through several boiling regimes. With the very high heat transfer rates in boil-
ing, parts are cooled very quickly during a quench. The fast cooling of an aluminum
alloy or nickel superalloy after solutioning avoids unwanted secondary solid-phase
precipitation, and in steel the high cooling rate enables martensite production.
The progression of boiling regimes during a quench begins with film boiling when
the part is put in the quenchant. As the surface temperature drops to the Leidenfrost
point (or minimum heat flux), the film begins to break up and move into the transition
region. There a mixture of patches of vapor and of vigorous nucleate boiling, moving
around on the surface in an unpredictable way; their motion is highly unstable and
transient. This behavior also explains why the heat transfer coefficient is so high dur-
ing transition and why it has the highest cooling rate. (This unpredictability makes
it difficult to design devices to operate in this excess temperature range.) Once the
excess temperature drops to the point of the maximum heat flux, the surface reverts
to nucleate boiling and then single phase natural convection, and cooling until the
surface is near the bulk temperature of the liquid.
474 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 10.30 The pool boiling curve for liquid water at 1 atm, where excess temperature
ΔTe= TS – Tsat. The quenching path is temperature-controlled, with an isothermal jump from
film transition boiling to nucleate transition boiling.

10.8 SUMMARY
Convection is heat transfer between a solid and fluid with a non-zero relative veloc-
ity. The primary physical mechanisms in convection are conduction of heat into and
through the fluid near the solid surface and the advection of heat toward or away from
that surface. The interaction at the surface is defined by the heat transfer coefficient,
which is proportional to the temperature gradient at the surface in the fluid; the Nus-
selt number is the nondimensional temperature gradient. A useful fluid property in
convection is the Prandtl number, the ratio of momentum and thermal diffusivities.
External forced convection is heat transfer at a surface with unbounded fluid flow
on one side, where that flow is driven by an imposed shear or pressure gradient
unrelated to the temperature field. To understand the mechanisms of convection, we
modeled flow over a semi-infinite flat plate, defining a thermal boundary layer as the
region in which the temperature ranges from the surface to the freestream values.
Scaling analysis in that region produces functions for thermal boundary layer thick-
ness, the heat transfer coefficients, and Nusselt number in terms of the Reynolds and
Prandtl numbers and distance from the plate leading edge. The scaling analysis also
gives a relationship between Pr and the ratio of the two boundary layer thicknesses.
More detailed analyses of the flat plate boundary layers are performed using integral
Convection Heat Transfer 475

analysis. However, in most realistic geometries and flow configurations, no math-

ematical solution is available (short of numerical analysis), but we can use the flat
plate solutions as a guide to how we treat these situations using experimental data.
Some experimental correlations of measurements are shown, using as correlation
parameters the nondimensional variable found from scaling, e.g., Nu = f(Re,Pr).
Internal forced convection occurs when a fluid at some inlet temperature interacts
with the walls of a conduit at a different temperature. The heat transfer coefficient
is redefined as using a different temperature difference than external flows because
there is no freestream value; instead we use the local mean temperature in the con-
duit, which changes along the flow direction. Heat transfer coefficient correlations
are presented here from analysis and experiments.
Natural convection is driven by forces dependent on the density field, as it is
induced by the effect of spatial variations in temperature (or composition) on the
local density. Scaling analysis on the coupled thermal and momentum boundary lay-
ers on a vertical flat plate in a gravity field produces a new vigor parameter similar to
the Reynolds number in forced flow. The Rayleigh number, Ra, includes the effect of
buoyancy due to temperature differences. The heat transfer correlations in the form
of Nu = f(Ra, Pr) are given from the scaling analysis and experimentation. Finally,
the topic of boiling heat transfer is presented, examining the many boiling regimes
in stagnant pools. These regimes each have different vapor-liquid interface morphol-
ogies that control the heat flux–temperature different correlation. Using the boiling
curve, which shows these different regimes in heat flux–temperature space, related
experimental correlations are discussed. Quenching is also discussed in terms of the
temperature path on the boiling curve.

10.9 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

10.1 A cylindrical cobalt rod is cold-drawn through a die at a velocity of

V = 0.5 m/s. The resulting diameter is D = 0.002 m and the heat genera-
tion due to plastic deformation raises the temperature of the rod to 430 oC
as it leaves the die. The rod is cooled by an air flow at U = 1 m/s and 30 oC
perpendicular to it. (a) Estimate the heat transfer coefficient on the rod.
(b) If the rod temperature must be lower than 130 oC when it is drawn
again, how far apart should the dies be? Air: = 0.8 kg/m3, c = 1000 J/kg
K, k = 0.03 W/m K, μ = 2.0 × 10–5 kg/ms. Cobalt alloy: = 8900 kg/m3,
c = 450 J/kg K, k = 67 W/m K.
10.2 We wish to continuously heat treat a nickel superalloy rod by induction
hardening as it moves down a production line. The initial condition at
x = 0 is that the metal temperature is at ambient. (a) For Vrod = 0.01 m/s
and Uair = 0.2 m/s (normal to rod), calculate the heat transfer coefficient
and the Peclet number. (b) Solve Eq. (10.16) for the temperature along the
moving rod, except with heat generation. (c) Using the solution in part (b),
estimate the required uniform volumetric heat generation rate in the rod
476 An Introduction to Transport Phenomena in Materials Engineering

if we want it heated from 30 oC to 900 oC over a distance L = 10 m. Air:

= 0.8 kg/m3, c = 1000 J/kg K, k = 0.03 W/m K, μ = 2.0 X 10 –5 kg/ms.
Nickel alloy: = 8500 kg/m3, c = 450 J/kg K, k = 10 W/m K.
10.3 Repeat the scaling analysis of Section 10.4.1, except assume that / T ~
O(1). Show that the estimates for T(x), h(x), and Nu(x) where / T >> 1 are
still valid.
10.4 Repeat the scaling analysis of Section 10.4.1, except assume that / T << 1.
Show that the estimates for T(x), h(x), and Nu(x) in Eqs. (10.42) and
(10.43) are correct.
10.5 For a steady-state, fully developed, laminar, pressure-driven flow in a
channel, we have continuity and momentum equations that reduce to
2
u dP
v 0 0 2
.
y dx

Imagine this channel with side walls held at To, but the friction in the fluid
generates heat. This viscous dissipation acts like a source term in the
energy equation:
2 2 2
T T T T T u
u v .
t x y x2 y2 c y

(a) Assuming the channel is H in height and L long (and H/L << 1) and that
the flow and heat transfer are steady state, which three terms in the energy
equation can we neglect? (b) Scale the momentum equation and the reduced
energy equation with: uref ~ Umax (unknown); ΔTref ~ (Tmax – To) (unknown); xref
~ L; yref ~ H; and the pressure gradient, dP/dx is a known quantity. (c) Find
an expression for Umax by scaling the previous momentum equation, and use
that result to scale the three terms in the energy equation. (d) Because we
have a fully developed flow regime, we know that ReH H/L << 1 (Section 7.5).
Assuming the Prandtl number is order one or less, what term can we neglect in
the energy equation? (e) Use the two remaining terms to find an estimate for
the (Tmax – To). Identify these last two significant terms and explain the balance
of heat flow based on that result.

10.6 In Section 10.4, we showed that an estimate of the thermal boundary layer
behavior in a zero-pressure gradient flow on an isothermal flat plate can
be found from scaling to be

T ~ x Re x 1/ 2 Pr 1/ 3
for Pr O(1).

Using this result (which is the same for an isoflux case, i.e., one with a
uniform heat flux), we perform further scaling analyses to find (a) q”(x) for
an isothermal surface and (b) Tw(x) for an isoflux surface. (c) Sketch these
two functions vs. x and explain their behavior. (Hint: Use the rate equation
for conduction, applied to the stationary fluid at the wall.)
Convection Heat Transfer 477

10.7 In this problem, consider the steady-state behavior of a Newtonian fluid

in a rotating viscometer. Because the gap width is much smaller than
the radius, we can treat the flow as an example of Couette flow where
u(y) = (Uo/ )y and v(y) = 0. (a) Identify the physical phenomena associated
with each of the terms in this differential equation.

2 2 2
T T T T T u
u v
t x y x2 y2 c y

(b) The inner side (y = 0) of the gap is thermally insulated while the outer side
is maintained at a constant temperature, To. Find the steady-state temperature
distribution across the gap, T(y), caused by viscous dissipation. Sketch T(y)
and u(y).

10.8 A liquid metal (T∞ = freestream temperature, Pr << 1) flows over a uniform
temperature (TS) flat plate, which takes energy out of the fluid. A thermal
boundary layer develops near the surface, which causes heat flux distribu-
tion qW (x). (x is the distance from the leading edge.) Because Pr << 1, the
velocity boundary layer thickness is very small compared to the thermal
boundary layer thickness ( << T ), so we can treat the velocity as uniform
throughout the thermal layer (u=U∞, everywhere). T(y) should be para-
bolic. (a) Write the normalized boundary conditions (in terms of and )
for the energy equation at y = 0 (on the plate) and y = T (at the edge of the
thermal boundary layer).

T T y
and
TW T T

(b) Use these boundary conditions (and one other) to find a second-order
profile for temperature: ( C1 2 C2 C3 ). (c) Use the energy integral
d t T
equation, u T T dy , and this profile to find T = f(x,
dx 0 y0
Rex, Pr). (d) Find h(x), the heat transfer coefficient as a function of distance
downstream from the leading edge. (e) Find qW (x), the heat flux as a function
of distance downstream from the leading edge.

10.9 Repeat the analysis in Example 10.6 with a liquid aluminum alloy as the
fluid. Use these data: Al = 2400 kg/m3, μAl = 0.003 kg/ms, cAl = 950 J/kg
K, k Al = 100 W/mK.
10.10 Air at 3 m/s and 15 oC is used to cool a square, hot, horizontal molded
plastic surface 0.5 m square with a surface temperature of 140 oC. (a) What
is the heat transfer rate from the surface if the air flows parallel to the
surface? Use the average heat transfer coefficient for a flat plate bound-
ary layer with a uniform temperature. (b) To speed up production of these
478 An Introduction to Transport Phenomena in Materials Engineering

parts, it is proposed to cool the plates using a single slot nozzle placed
0.1 m from the surface and centered over the surface. The velocity and
temperature from the nozzle are the same as for the boundary layer flow. If
we want to double the average heat transfer rate by replacing the system in
(a) with the nozzle, what nozzle width (W) should we use? (NB: You can’t
solve for W explicitly; the equation is too complicated. Any root finding
method will do here. The simplest method is to plot h vs. W over W = 0
to 0.5 m and find where the new heat transfer coefficient equals twice the h
from part (a). You can also use a bisection method.) Air data: Prair = 0.707,
kair = 0.0263 W/mK, air = 1.16 kg/m3, μair = 1.85 × 10 –5 kg/ms.
10.11 You are asked to quench Ni superalloy bars with a square cross-sectional
area from 1200 oC to 300 oC in a furnace by forcing Ar gas over them at
100 oC. The velocity of the argon is VAr = 4 m/s. (a) Use a lumped capaci-
tance method (is that valid?) to predict the cooling behavior of the parts,
which are L = 0.2 m long and d = 0.02 m on a side. Assume that the gas
flow is perpendicular to the axis of the cylinders. Plot T(t). (b) It is sug-
gested that increasing the gas pressure will increase the heat transfer rate.
Evaluate this suggestion by calculating how long the bars take to cool to
300 oC for the cases of P = 1 atm, P = 3 atm, and P = 5 atm. Plot the cool-
ing curves (T vs. t) for these three cases and discuss whether this change
in pressure is helpful. (It is reasonable in this pressure range to assume
that only density is affected by pressure.) Argon gas: = 0.68 kg/m3 (at 1
atm), c = 520 J/kg K, k = 0.034 W/mK, μ = 4.3 × 10 –5 kg/ms. Superalloy:
= 8500 kg/m3, c = 450 J/kg K, k = 10 W/m K.
10.12 A refrigerated truck trailer (L long, H high, and H/2 wide) gains heat from
the environment by forced convection while moving at a speed of U∞.
Assume that: (i) no significant radiation heat transfer occurs; (ii) all signifi-
cant heat transfer takes place on the top and the two long sides of the trailer;
(iii) those surfaces can be treated as flat plates; (iv) the inside and the outside
of the trailer walls are isothermal and the inside wall is maintained at Tin.
(a) Find the total rate of heat gain into the cooled interior of the trailer, using
a simple resistance network and using a convection coefficient averaged over
the active surface. (This result gives us the load on the refrigeration unit to
maintain Tin inside the trailer.) Assume significant conduction resistance in
the trailer walls. (b) Estimate the outside surface temperature of the trailer.
Air: Prair = 0.707, kair = 0.0263 W/mK, air = 1.16 kg/m3, μair = 1.85 × 10 –5 kg/
ms. Wall: kwall = 1.0 W/mK, wall thickness = t = 0.05 m. Other conditions:
U∞ = 30 m/s, L = 10 m, H = 4 m, Tin = −5 oC, T∞ = 30 oC.
10.13 A hot wire anemometer is used to measure velocity of an air stream.
The hot wire is an isothermal cylinder in cross flow. A certain amount of
power (qo) is dissipated by the wire due to electrical resistance, and the
temperature of the wire is measured. (a) What is the heat transfer coef-
ficient in terms of the power provided, the wire geometry, and wire and air
temperatures? (Use convection rate equation.) (b) Using the appropriate
correlation for convection heat transfer, find an expression for the velocity
of the air flow in terms of the heat flux provided, the wire geometry, the
Convection Heat Transfer 479

wire and air temperatures, and any other appropriate parameters. Do not
substitute numbers into this result yet. (c) For the parameters and mea-
surements given next and the expression you found in part (b), what is the
velocity of the air flow?

qo = 4 W dwire = 0.5 mm Lwire = 5 mm Twire – T∞ = 10 oC

air = 1 kg/m cair = 1000 J/(kg K) kair = 0.03 W/(m K) μair = 2 × 10–5 kg/(m s)
3

wire = 8500 kg/m cwire = 440 J/(kg K) kwire = 14 W/(m K)

3

10.14 Water moves over a cylinder that is perpendicular to the flow under
the following conditions: U∞ = 2 m/s, D = 0.01 m (cylinder diameter),
L = 0.3 m (cylinder length). (a) Find the Reynolds, Prandtl, and Nusselt
numbers for this configuration. (b) Find the total rate of heat loss from
the cylinder. Cylinder: kcyl= 15 W/mK, Tcyl = 90 oC. Water: = 2.0 × 10 –6
m2/s, µ = 0.002 kg/ms, = 1.4 × 10 –7 m2/s, kwater= 0.6 W/mK, T∞ = 10 oC,
water = 1000 kg/m .
3

10.15 A strand of polyethylene (diameter = D = 0.01 m) exits an extruder at 120 oC

into a 25 oC air environment. (a) What is the heat transfer coefficient from
the strand to the environment? (b) Assuming no significant conduction
along the axis of the strand (high Pe), we can use the lumped capacitance
method for estimating the temperature drop in the polymer with distance
from extruder. In that case, x = Vpt, where x = 0 is the extruder exit and
ip is the polymer speed. It must enter a roll stand downstream at no less
than 100 oC. For the conditions listed next, what is the maximum distance
from the extruder to the roll stand, L? HDPE: = 960 kg/m3, c = 2260
J/kg K, k = 0.33 W/m K. Air: = 0.8 kg/m3, c = 1000 J/kg K, k = 0.03
W/m K, μ = 2.0 × 10 –5 kg/ms. Air speed = U∞ = 1 m/s; polymer speed,
Vp = 0.02 m/s.
10.16 We want to cure a polymer sheet by passing it through a convection oven
in which hot air is blown through an array of round nozzles onto the top
surface (and only the top surface) of the sheet. The sheet enters the oven
at 30 oC and must exit at 350 oC. To maintain profitability, the velocity of
the sheet (Vs) must not go below 0.01 m/s. The longest we can make the
oven is 4 m, and the velocity of the air impinging on the sheet (Va) can
be controlled between 0.5 m/s and 5 m/s. The average heat transfer coef-
ficient on the sheet can be correlated from

hD
NuD 0.5 Re 2 / 3 Pr 0.42 ,
k

where the Reynolds number is also based on nozzle diameter. Polymer:

kpoly = 0.2 W/mK, cpoly = 1500 J/kg K, poly = 1000 kg/m3, To = 30 oC,
Tend = 350 oC, d = 0.01 m (sheet thickness), w = 0.3 m (sheet width), and
D = 0.005 m (nozzle diameter) (a) Calculate Pe for Va = 0.5 m/s and for
Va = 5 m/s. (b) Plot the necessary furnace air temperature (Ta) to get
Tend = 350 oC as a function of velocity of the air (0.5 m/s < Va < 5 m/s). This
480 An Introduction to Transport Phenomena in Materials Engineering

curve gives you the operating states for your oven to cure the polymer at
the given sheet speed and size of oven.
10.17 A long horizontal stainless steel rod of 50 mm diameter is annealed at
1000 K, and is removed from the furnace and transported to a hot forming
process. The room in which the transfer takes place has wall and ambient
temperatures of 300 K. (a) What is the initial heat loss rate (qNC) due to
natural convection? (b) What is the initial heat loss rate (qrad) due to radia-
tion ( = 0.8 and = 5.67 × 10 –8 W/m2K4)? Which mode of heat transfer is
dominant at that initial temperature? (c) Find the effective radiation heat
transfer coefficient for this rod at the beginning of cooling. What is the
overall heat transfer coefficient (keeping in mind that natural convection
and radiation act in parallel)? Assuming a lumped capacitance approxima-
tion and that the initial overall heat transfer coefficient holds throughout
the cooling (a conservative assumption), find the maximum transit time
allowed if the forming must be done above 850 K. Steel: ksteel = 25 W/mK,
steel = 0.0000052 m /s
2

10.18 A heat treating furnace (outer dimensions: 6 m long, 3 m wide, 6 m tall)

loses heat from its surface to the environment by natural convection.
Thermocouples on the top, outer surface give an average temperature of
365 K, while the average outer surface temperature for the four sides is
340 K. If the ambient temperature in the plant is 300 K, calculate the total
rate of energy loss (not heat flux) from the furnace.
10.19 A square horizontal surface (L = 0.3 m) of an unknown material (t = 0.01 m
thick) is heated from below by an attached resistance heater with a power
input of 100 W. Thermocouples are located in the center of the material on
the top surface exposed to ambient air (To = 300 K) and on the bottom sur-
face in contact with the heater. The top thermocouple reads 350 oC and the
bottom 360 oC. (a) Sketch streamlines showing the flow pattern in the air
around the surface. (b) Assuming that natural convection is the only heat
transfer mode on the top of this surface, what is the thermal conductivity
of the unknown material? (c) Is the assumption in part (b) reasonable? Air
at T = 325 oC: = 1.1 kg/m3, μ = 2.1 × 10 –5 kg/ms, = 1/(325 K), k = 0.03
W/mK, c = 1000 J/kg K.
10.20 A long 2-mm diameter wire passes an electric current and reaches a sur-
face temperature of 120 °C when submerged in boiling water at 1 atm
pressure. Calculate the boiling heat transfer coefficient.
10.21 Calculate the nucleate pool boiling heat transfer coefficient for water
boiling at atmospheric pressure on the outer surface of a platinum-plated
10-mm diameter tube maintained 10 °C above the saturation temperature.
10.22 A tube (diameter = d) through which water is pumped is wrapped around
a cylindrical furnace (diameter = D) to cool the outside furnace walls. The
tube is wrapped around the furnace n = 10 times. The heat that must be
dissipated is q = 1.5 kW and it can be treated as being evenly distributed
over the tube surface. If we desire the temperature rise in the water to be
less than 10 oC, what is the maximum surface temperature of the tube?
Assume thermally and hydrodynamically fully developed flow. Water:
Convection Heat Transfer 481

= 1000 kg/m3, c = 4000 J/kg K, k = 0.6 W/m K, μ = 3.0 X 10 –4 kg/ms.

Geometry: d = 0.02 m, D = 0.4 m.
10.23 A 20 mm × 20 mm tungsten carbide tool is used to machine steel. During
the process, the tool heats up, which can lead to carbon diffusion out of
the tool and the consequent degradation of the tool’s life. To combat this
heating, we would like to keep the tool below 800 oC under a frictional
heat load of 40 W. To cool the tool, a square (2 mm × 2 mm) flow pas-
sage is cut into its back and nitrogen gas at −50 oC is pumped through it.
Assume that the passage is 90 mm long. (a) Assuming a constant tempera-
ture body (T = 800 oC), what flow rate of nitrogen is necessary to achieve
the cooling necessary? (b) Assuming a uniform heat flux in the passage,
what flow rate is necessary to keep the surface temperature at the outlet
below 800 oC?

NOTES
1. Pr is named for Ludwig Prandtl, the discoverer of the boundary layer.
2. Also known as free convection or buoyancy-induced flow.

REFERENCES
1. Lovell, J., and J. Kluger, Lost Moon: The Perilous Voyage of Apollo 13, Houghton Mif-
flin, 1994, pp. 248–249.
2. Burmeister, L. C., Convective Heat Transfer, 2nd ed., Wiley-Interscience, 1993.
3. Bejan, A., Heat Transfer, John Wiley & Sons, 1993.
4. Kraus, A. D., A. Aziz, and J. Welty, Extended Surface Heat Transfer, Wiley-Interscience,
2011.
5. Blasius, H., “Grenzschichten in Flüssigkeiten mit kleiner Reibung,” Zeitschrift für
Angewandte Mathematik und Physik, v. 56, pp. 1–37, 1908.
6. Pohlhausen, E., “Der Wärmeaustausch zwischen festen Körpen und Flüssigkeiten mit
kleiner Reibung und kleiner Wärmeleitung,” Zeitschrift fur Angewandte Mathematik
und Mechanik, v. 1, pp. 115–121, 1921.
7. Colburn, A. P., “A method of correlating forced convection heat transfer data and a com-
parison with fluid friction,” Transactions of the American Institute of Chemical Engi-
neers, v. 29, pp. 174–206, 1933.
8. Gnielinski, V., “New equations for heat and mass transfer in turbulent pipe and channel
flow,” International Chemical Engineering, v. 16, pp. 359–386, 1976.
9. Reynolds, W. C., W. M. Kays, and S. J. Kline, “Heat transfer in the incompressible tur-
bulent boundary layer: I-constant wall temperature,” NASA Memorandum 12-1-58W,
1958.
10. Lienhard, J. H. V, “Heat transfer in flat-plate boundary layers: A correlation for laminar,
transitional, and turbulent flow,” ASME Journal of Heat Transfer, v. 142, pp. 061805–1–
14, 2020.
11. Churchill, S. W., “A comprehensive correlating equation for forced convection from flat
plates,” AIChE Journal, v. 22, pp. 264–268, 1976.
12. Mayle, R. E., The Role of Laminar-Turbulent Transition in Gas Turbine Engines (Paper
91-GT-261), International Gas Turbine and Aeroengine Congress and Exposition,
ASME, 1991.
482 An Introduction to Transport Phenomena in Materials Engineering
Convection Heat Transfer 483
11 Mass Transfer in Fluids

11.1 INTRODUCTION
Mass transport in flowing fluids occurs by diffusion down concentration gradients
and by bulk flow of the fluid, and in considering mass transport from a solid surface
to a surrounding fluid, we rely on the analogy with heat transfer from a solid to a
flowing fluid. Unfortunately, quantitative knowledge of the diffusion coefficients in
systems in which mass transport is occurring is significantly less than quantitative
knowledge of the thermal diffusivities in systems in which heat transfer is occurring,
and thus although the mathematics of mass transport in a large number of situations
can be developed by analogy with heat transfer, numerical calculation is limited to
systems for which diffusion data are available.

11.2 MASS AND MOLAR FLUXES IN A FLUID

If a fluid containing species i at a uniform mass concentration i (kg/m3) is flowing
with the x-direction velocity u (m/s), the rate of mass flux of i due to bulk motion of
the fluid, as observed by a stationary observer, is i u (kg/m 2s). However, from the
point of view of an observer who is also traveling at the velocity u, the rate of mass
flux of i is zero. If the concentration of i in the flowing fluid is not uniform, i also is
transported through the fluid by a diffusive flux in the x direction at the rate

ji, x D i
kg m 2 s ,
x
This expression is the rate of mass flux of i in the x direction observed by an
observer traveling at the velocity u. However, in this case, the stationary observer
sees the transport of i due to the bulk motion of the fluid and the transport of i due to
diffusion within the fluid as the sum of ji and i u.
The local mass average velocity, u, of a binary fluid of density containing com-
ponents A and B is defined by

u AuA B uB . (11.1)

In this expression, u is the rate at which mass passes through a unit cross-
sectional area perpendicular to the direction of flow, and uA and uB are, respectively,
the averaged values of the velocities of the two components, A and B, as observed by
a stationary observer. The local mass average velocity is thus the absolute velocity
of the center of mass of a volume element of the fluid. Alternatively, Eq. (11.1) can
be written as
u AuA B uB (11.2)

484
48 4 DOI: 10.1201/9781003104278-11
Mass Transfer in Fluids 485

in which A = A/ and B = B/ are the mass fractions of A and B in the mix-

ture. The existence of concentration gradients in the mixtures gives rise to diffusion
fluxes, which in the x direction are

A B
j A, x D and jB, x DB .
x x

Thus, the absolute mass fluxes in the x direction are

n A, x AuA j A, x Au kg m 2 s (11.3)

and

nB, x B uB j B, x Bu kg m 2 s . (11.4)

for A and B, respectively, where u is the x-component of the local mass average
velocity u and uA and uB are the x-components of the A and B velocities. The sum of
the absolute mass fluxes is thus

n A, x nB, x AuA B uB , (11.5)

which, from Eq. (11.1), written for the x-component of the local mass average veloc-
ity, gives

n A, x nB, x u. (11.6)

From Eqs. (11.3) and (11.4),

j A, x A uA u (11.7)

and

j B, x B uB u , (11.8)

the sum of which is

j A, x j B, x AuA B uB u 0. (11.9)

Mass transport can also be considered on a molar basis, in which case the local molar
average velocity, u, is defined by

Cu C A u A C B uB , (11.10)
486 An Introduction to Transport Phenomena in Materials Engineering

where CA and CB are the molar concentrations of A and B in the fluid, and C = CA +
CB. In this case Cu is the rate at which matter passes through unit cross-sectional area
perpendicular to the direction of the flow in units of (gmol/m2s). Although the veloci-
ties uA and uB are identical in Eqs. (11.1) and (11.10), the local molar average velocity
v has the same value as the local mass average velocity u only when the molecular
weight of A is equal to the molecular weight of B. Alternatively, Eq. (11.10) can be
written as

u X AuA X B uB , (11.11)

in which XA and XB are the mole fractions of A and B in the fluid. In this case concen-
tration gradients in the fluid give rise to molar diffusion fluxes in the x direction of

CA
j*A, x D (11.12)
x
and
CB
jB* , x DB .
x
The absolute molar fluxes in the x direction are thus

N A, x C AuA j*A, x C A u* (11.13)

and
N B, x C B uB jB* , x C B u* , (11.14)

and the total absolute molar flux is

N A, x N B, x C A u A C B uB Cu* (11.15)

and

j*A, x jB* , x C A u A C B uB Cu* 0. (11.16)

11.3 EQUATIONS OF DIFFUSION WITH ADVECTION

IN A BINARY MIXTURE A-B
Figure 11.1 shows a control volume of dimensions dx and dy that is fixed in space
relative to a stationary observer and through which a binary mixture of A and B is
flowing with the molar average velocity u. A mass balance for A requires that

rate of change of concentration of A in the control volume

net rate at which A enters and leaves the control volume.
Mass Transfer in Fluids 487

FIGURE 11.1 Control volume, fixed in space relative to a stationary observer, through
which a binary mixture of A and B is flowing with the molar average velocityV uiˆ vjˆ.

The rate of change of the concentration of A in the control volume is

CA
dxdydz ,
t
and the net rate of transport of A into the control volume in the x and y directions is

C A u A dx dydz C A v A dy dxdz .
x y

The mass balance for A is thus

CA
C AuA C A vA , (11.17)
t x y

and, similarly, the mass balance for B is

CB
C B uB C B vB . (11.18)
t x y
The sum of Eqs. (11.17) and (11.18) is

C A CB
C A u A C B uB C A v A C B vB ,
t x y
488 An Introduction to Transport Phenomena in Materials Engineering

which, from Eq. (11.10), gives

C
Cu* Cv* . (11.19)
t x y

Eq. (11.19) is thus the rate of change with time of the molar concentration in the
stationary control volume.
The molar concentration of A in the flowing fluid is a function of time and position

CA C A x, y,t ,

and therefore

CA CA CA
dC A dt dx dy,
t x y
or

dC A CA C A dx C A dy
, (11.20)
dt t x dt y dt
which is the substantial derivative of CA,

DC A CA CA CA
u v . (11.21)
Dt t x y

If, in Eq. (11.21), u = u* and v = v*, then

DC A CA CA CA
u* v* (11.22)
Dt t x y

is the rate of change, with time, of the molar concentration of A in the fluid from the
point of view of an observer traveling at the velocity V . Substitution of Eq. (11.13)
into Eq. (11.17) gives

CA
j*A, x C A u* j*A,y C A v*
t x y
(11.23)
u* v* j*A, x j*A, y CA CA
CA u* v* ,
x y x y x y

and substitution of Eq. (11.23) into Eq. (11.22) gives

DC A u* v* j*A, x j*A, y
CA . (11.24)
Dt x y x y
Mass Transfer in Fluids 489

Similarly,

DC B u* v* jB* , x jB* , y
CB . (11.25)
Dt x y x y

From Eq. (11.16),

j*A, x jB* , x j*A, y jB* , y 0,

and thus the sum of Eqs. (11.24) and (11.25) is

DC u* v*
C . (11.26)
Dt x y

For a fluid of constant density, the continuity equation gives

u* v*
0, (7.13)
x y

and hence, for such a fluid, in which D is constant, Eq. (11.24) becomes

DC A j*A, x j*A, y 2
CA 2
CA
D . (11.27)
Dt x y x2 y2

11.4 EQUIMOLAR COUNTERDIFFUSION

For steady-state, one-dimensional counterdiffusion of A and B in a binary gas mixture,

C AuA C B uB ,

and thus from Eq. (11.16), u* = 0. Also, at steady state ∂CA/∂t = 0 and for one-
dimensional diffusion (∂CA/∂y = 0), Eq. (11.27) gives

d 2C A
0. (11.28)
dx 2

Given that the molar concentration of A in a mixture can be found from the ideal gas
law

nA pA
CA , (11.29)
V RT
490 An Introduction to Transport Phenomena in Materials Engineering

or, from Eq. (11.29),

d 2 pA
0. (11.30)
dx 2

Integrating Eq. (11.30) twice gives

pA C1 x C2 (11.31)

[i.e., the partial pressures of A and B are linear functions of x]. Applying the bound-
ary conditions,
pA 0 at x 0
pA ,
pA L at x L

gives the constants for Eq. (11.31).

x
pA pA L pA 0 pA 0 . (11.32)
L

The gradient of the partial pressure of A is thus

dp A pA L pA 0
,
dx L

which corresponds to a concentration gradient of

dC A pA L pA 0
,
dx RT L

and hence the molar flux of A due to diffusion is

dC A D pA 0 pA L
j*A D . (11.33)
dx RT L

11.5 ONE-DIMENSIONAL STEADY-STATE DIFFUSION

OF GAS A THROUGH STATIONARY GAS B
In this case u* ≠ 0, and hence, for one-dimensional, steady-state diffusion
(∂CA/∂t = v* = ∂u*/∂x = 0, Eqs. (11.23) and (11.27) give

dC A dj*A d 2C A
u* D . (11.34)
dx dx dx 2
Mass Transfer in Fluids 491

As the gas B is stationary, uB = 0, and hence from Eq. (11.14),

jB* , x C B u*

or

D dC B
v*x J B* ,x C B , (11.35)
C B dx

substitution of which into Eq. (11.34) gives

1 dC B dC A d 2C A
, (11.36)
C B dx dx dx 2

Or, from Eq. (11.29),

1 dpB dp A d 2 pA
. (11.37)
pB dx dx dx 2

As the total pressure of the gas, P, is constant and pA + pB = P,

dp A dpB ,

and thus Eq. (11.37) becomes

2
1 dpB d 2 pB
. (11.38)
pB dx dx 2

From the identity

2
d 1 dy 1 d2 y dy 1 dy 1 d2 y dy
,
dx y dx y dx 2 dx y 2 dx y dx 2 y dx

Eq. (11.38) is equivalent to

d 1 dpB
0.
dx pB dx

Integrating this expression twice and applying the boundary conditions,

pB 0 at x 0
pB
pB L at x L
492 An Introduction to Transport Phenomena in Materials Engineering

gives pB(x):

pB x pB L
ln ln . (11.39)
pB 0 L pB 0

With pA = P – pB,

P pA x P pA L
ln ln . (11.40)
P pA 0 L P pA 0

These partial pressures are shown in Figure 11.2 for the case of pB(0) = 0.1 and
pB(L) = 0.9. The partial pressure, and hence concentration of B, increase with higher
*
x/L, and the gradient of B gives rise to a diffusion flux of B from right to left, jB, x ,
which, from Eq. (11.35), is exactly balanced by the transport of B from left to right
by the bulk flow of the gas mixture, C B u*.
The absolute molar flux of A is given by Eq. (11.13) as

N A, x C AuA j*A, x + C A u* ,

which from Eq. (11.35) can be expressed as

dC A D dC B D dpA pA D dpA
N A, x D CA
dx C B dx RT dx pB RT dx
(11.41)
D P dpA
.
RT P pA dx

FIGURE 11.2 Variation with position of gas partial pressures when A diffuses through
stationary gas B.
Mass Transfer in Fluids 493

From Eq. (11.40),

d ln P pA 1 P pA L
ln
dx L P pA 0

or

dp A P pA P pA L
ln ,
dx L P pA 0

substitution of which into Eq. (11.41) gives

DP P pA L
N A, x C AuA ln , (11.42)
RT L P pA 0

which at steady state has a constant value. Thus, in Figure 11.2, as CA decreases with
increasing x, Eq. (11.42) shows that uA increases with increasing x.
Diffusion of a gas A through a stationary gas B is realized in Stefan’s apparatus
shown in Figure 11.3. In this arrangement, a pool of liquid A is held at the bottom
of a vertical tube. A fixed composition mixture of gas B and vapor A flows through
a horizontal tube. The temperature and total pressure within the system are constant.
If B does not dissolve in liquid A, the partial pressure of A in the gas mixture at the
liquid A-gas interface is the saturated vapor pressure of liquid A at the temperature
T, designated pA(0) at zo, and the partial pressure of A at zL (pA(L)) is determined by the

FIGURE 11.3 Stefan’s apparatus for measuring the interdiffusion coefficient in the binary
gas A-B.
494 An Introduction to Transport Phenomena in Materials Engineering

composition of the gas mixture flowing in the horizontal tube. Thus, as pA = XAP, Eq.
(11.40) gives the variation of XA with position in the vertical tube as

1 XA z z0 1 XA L
ln ln . (11.43)
1 XA 0 zL z0 1 XA 0

Steady state has been achieved and evaporating liquid A is replenished continuously
to maintain the distance zL – z0. Measurement of the rate of evaporation of liquid A
then allows determination of D from Eq. (11.42).

Example 11.1
Measurement of a Diffusion Coefficient of Mn Vapor in Ar Gas
When liquid Mn at 1600 °C evaporates and diffuses into pure Ar in an apparatus at
which zL – z0 is maintained constant at 0.02 m, the steady-state rate of loss of mass of
liquid Mn is measured to be 3.21 x 10–4 kg/m2s. The atomic weight of Mn is 0.05494
kg/g mol, and at 1873 K, the saturated vapor pressure of liquid Mn is 5370 Pa. We
will treat this system as one-dimensional in the z direction.

Ar
Find: Calculate DMn .
Solution: The flux of Mn vapor from the melt surface is

3.21 10 4 kg m 2 s 3
N Mn,z = 5.84 10 g mol m 2 s ,
0.05494 kg Mn gmol Mn

Ar
and with a total pressure of P = 101,325 Pa, DMn is obtained from Eq. (11.42) as

Ar N Mn, z RT L
DMn
P
P ln
P pMn
5.84 10 3 gmol Mn m 2 s 8.3144 J gmol K 1873 K 0.02 m
101,325 Pa
101, 325 Pa ln
101,3325 Pa 5370 Pa
4
3.30 10 m2 s .

From Eq. (11.35), with Mn = A and Ar = B

Ar
DMn dCAr d ln CAr d ln pAr
w* Ar
DMn Ar
DMn ,
CAr dz dz dz
Mass Transfer in Fluids 495

and from Eq. (11.39)

d ln pAr 1 pAr L
ln .
dz L pAr 0
Thus,

Ar
DMn pAr L 3.3 10 4 m 2 /s 101,325 Pa
w* ln ln
L pAr 0 0.02 m 101,325 Pa 5370 Pa

4
9.0 10 .
s

From Eqs. (11.12) and (11.13),

dCMn
N Mn,z D CMn w* .
dz

The highest value of CMn in the gas is

pMn 0 5370 Pa
CMn 0.345 gmol / m 3 ,
RT 8.3144 J/gmol K 1873 K

and thus the highest value of CMnw* is 3.1 × 10–4 (gmol/m2s), which is only 5.3% of
the total flux of 5.84 × 10–3 (gmol/m2s). The small contribution of bulk flow to the
transport of Mn is because the concentration of Mn vapor in the mixture is very much
smaller than that of Ar. In the previous calculation of w*, as pAr(0) approaches P, w*
Ar
approaches zero. With very small values of w*, an approximate value of DMn can be
obtained from Fick’s first law as

Ar
* Ar dCMn DMn dpMn
jMn DMn ,
dz RT dz

which, assuming a linear variation of pMn with z, gives

*
Ar RT L jMn
DMn
pMn L pMn 0

8.3144 J gmol K 1873 K 0.02 m 5.84 10 3 gmol m 2 s

5370 Pa
4 2
3.39 10 m /s,

which is 2.7% higher than the estimation including the weak bulk transport on Mn.
496 An Introduction to Transport Phenomena in Materials Engineering

11.6 SUBLIMATION OF A SPHERE INTO A STATIONARY GAS

Consider a sphere of solid A of radius Rs subliming into a stationary gas B. Because
the surrounding gas is quiescent, species A moves through B only by diffusion. The
rate of transport through the stationary B in the radial direction at a radius r (< Rs),
WA, in units of gmol/s is

WA 4 r2 NA,

and hence, from Eq. (11.41),

WA DAB P dp A
NA .
4 r2 RT P p A dr

In the gas, the boundary conditions are no gaseous A far from the sphere (pA = pA,2=
0 at r = r2 = ∞) and the partial pressure of A at temperature T on the sphere surface is
pA = PA,1 at r = r1 = Rs. With these conditions, we integrate from r1 to r2,

WA 1 1 DAB P P p A,2
ln .
4 r1 r2 RT P p A,1

The molar flux of A from the surface of the sphere (r1 = Rs) is

4 Rs DAB P P
WA ln . (11.44)
RT P pA

However, sublimation causes the radius of the sphere to decrease, and equating the
molar flux with rate of loss of moles from the sphere of radius Rs gives

dn A 4 dRs3 2 dRs
WA m 4 m Rs ,
dt 3 dt dt

in which m is the molar density of solid A. Combination of the preceding equation

and Eq. (11.44) gives

DAB P P
Rs dRs ln dt.
m RT P PA

Integrating this equation with the initial condition of Rs (t=0) = a0 gives

2DAB P P
Rs2,o Rs2 ln t.
m RT P PA
Mass Transfer in Fluids 497

So the time required for sublimation to cause the sphere to disappear is

Rs2,o m RT
t . (11.45)
2DAB P ln P / P pA

Example 11.2
Sublimation of a Naphthalene Sphere
A naphthalene mothball of radius Rs,o = 0.005 m sublimes into stagnant air with tem-
perature T = 20 °C and an air pressure of P = 101.3 kPa. For naphthalene at 293 K,
air
po = 30 Pa, DNap = 8 x 10–6 m2/s, and m = 5470 gmol/m3.

Find: (a) The initial rate of sublimation; (b) the time required for complete
sublimation of the mothball.
Solution: (a) From Eq. (11.44),
air
4 Rs,o DNap P P
WNap ln
RT P p
6
4 0.005m 8 10 m 2 s 101.3 kPa 101.3 kPa
= ln
8.3144 J gmol K 293 K 101.3 kPa 0.03 Pa
9
= 6.2 10 gmol / s.

(b) From Eq. (11.15), the time required for the ball to disappear due to sublima-
tion is
2
0.005m 5470 gmol m 3 8.3144 J gmol K 293K
t
6
2 8 10 m 2 s 101.3 kPa ln 101.3 kPa / 101.3 kPa 0.03 kPa

6.94 105 s 8 days

The variation of the radius of the mothball with time is shown in Figure 11.4.

FIGURE 11.4 Variation with time of the radius of the mothball subliming into stagnant air.
498 An Introduction to Transport Phenomena in Materials Engineering

11.7 FILM MODEL

Consider Figure 11.5, which shows a mixture of gas B and vapor A flowing over the
surface of solid A at the temperature T. If the partial pressure of B in the mixture is
less than the saturated vapor pressure of solid A at the temperature T, then A sub-
limes from the solid and is transported into the bulk gas. Phase equilibrium between
the solid and the gas mixture requires that the partial pressure of A at y = 0 be the
saturated vapor pressure of solid A at the temperature T, p°A(T), and thus the concen-
tration of A in the gas phase at y = 0 is

CA 0 pA T RT .

If the concentration of A in the bulk gas phase is CA,∞, the variation of CA with dis-
tance away from the surface can be expected to be as shown in Figure 11.6 (a), and
the molar flux of A in the y direction is from Eq. (11.13).
In Eq. (11.13), the contribution of the bulk flow to the molar flux of A is deter-
mined by the magnitudes of CA and v*. If the product CA v* is small in comparison
with N A , treatment of molar transport of A from the solid to the bulk gas is simplified
by treating transport as only diffusion through a laminar flow parallel to the solid sur-
face. In Figure 11.6 (b), the molar flux at y = 0 is determined by DA and the tangent to
the concentration profile at y = 0. The thickness, L, of the estimated film is the value
of y at which the tangent line reaches the value CA = CA,∞. For y > L, the flowing gas
in the well-mixed region beyond the layer is considered to have no concentration,
temperature, or velocity gradients. From Eq. (11.42), the steady-state molar flux of A
through a film of thickness L is

DA P pB L CDA XB L
NA ln ln . (11.46)
RT L pB 0 L XB 0

FIGURE 11.5 Mixture of A vapor and B gas flowing over the surface of solid A.
Mass Transfer in Fluids 499

FIGURE 11.6 (a) Concentration profile of A vapor in the gas phase in Figure 11.5; (b)
approximate and actual composition profiles and definition of estimated film thickness; (c)
approximations of local flow velocity, temperature, and composition profiles in gas film near
the surface of solid A.

Defining the logarithmic mean of XB(L) and XB(0) as

XB L XB 0
XB lm
ln X B L XB 0

and substituting it into Eq. (11.46) gives the molar flux of A through the film as

CDA X A 0 X A L
NA . (11.47)
L X B lm
When XA(0) and XA(L) have small values, (XB)lm has a value close to unity and hence

CDA
NA XA 0 XA L , (11.48)
L
which approaches Fick’s first law, diffusing A down the linear concentration gradient
from y = 0 to L.
As L is a function of the velocity and physical properties of the flowing fluid, the
difficulty of estimating its value in mass transport systems limits the extent of the
practical application of the film model. It is nonetheless a useful tool in understand-
ing mass transport in a fluid to and from a condensed phase over which the fluid is
500 An Introduction to Transport Phenomena in Materials Engineering

flowing. If the temperature of the solid, Ts, differs from the temperature of the bulk
gas, T∞, the temperature profile through the film is as shown in Figure 11.6 (c), which
also shows the velocity profile through the film.

11.8 CATALYTIC SURFACE REACTIONS

Figure 11.7 shows a gas mixture A-B flowing over a flat surface on which a hetero-
geneous catalytic chemical reaction either produces or consumes the A. We assume
that the reaction consumes A by means of a first-order, irreversible chemical reaction
(one in which the chemical reaction rate, r, is proportional to the concentration of A
at the reaction site). Thus,

r kr C A 0 , (11.49)

in which r is also a molar flux with units (gmol/m2s), and kr, the reaction rate con-
stant, has the units (m/s). Steady-state conditions require that the rate at which A
arrives at y = 0 by diffusion through the film of thickness L be equal to the rate at
which it is consumed by the chemical reaction at y = 0:

dX A
r NA D AC . (11.50)
y 0 dy y 0

(The negative sign in the equality arises because the diffusion flux of A is drawn in
the negative y direction.)
Steady-state diffusion of A in the film requires that

dN A d2 XA
0 or DAC 0.
dy dy 2

FIGURE 11.7 A-B gas mixture flowing over a flat surface on which a catalytic reaction
consumes the species A.
Mass Transfer in Fluids 501

Integration of the latter twice and applying Eqs. (11.49) and (11.50) gives the con-
centration profile as

kr X A 0
XA XA L L y .
DA

Thus, at y = 0,

XA L
XA 0 . (11.51)
1 kr L / DA

From Eqs. (11.50) and (11.51), the molar flux of A is

kr C X A L
NA . (11.52)
1 kr L / DA

Here we explore the behavior of Eqs. (11.51) and (11.52) represented by kr 0 and
kr ∞.
Case 1: kr 0. From Eq. (11.51),

XA 0 XA L

and, from Eq. (11.52),

NA kr C X A L .

Case 2: kr ∞. The composition at the interface, y = 0, goes to zero at the high

reaction rate:

XA 0 0,

and the mass transfer rate is

NA CDX A L L.

In case 1, with low reaction rates relative to transport across the layer, we have diffu-
sion moving gas A to the surface so fast that the reaction cannot keep up with supply.
Hence, A concentration in the layer builds up until the diffusive flux decreases to the
level of the reaction rate. The opposite effect is true in case 2, which is limited by
the diffusive flux because A is consumed as soon as it reaches the surface. We refer
to these extremes as reaction controlled and diffusion controlled, respectively. For
intermediate values of kr, the process is affected by both the surface reaction and
diffusion and is referred to as a mixed controlled process. The effect of the reaction
rate constant on the concentration profile of A in the film is shown in Figure 11.8.
502 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 11.8 Possible concentration profiles of A in the film on the surface of the solid A
shown in Figure 11.7.

11.9 DIFFUSION AND CHEMICAL REACTION IN STAGNANT FILM

Figure 11.9 shows gas B flowing over the surface of solid A. In this case, A sublimes
from the surface and diffuses into the film and B diffuses through the film toward the
surface. Gas B and vapor A react to form the solid AB according to

A B AB,

which is a second-order irreversible homogeneous reaction for which

rA = rB kr C AC B kr C 2 X A X B g mol m 3s . (11.53)

In a two-dimensional control volume of dimensions dx and dy within the film, the

steady-state molar balance requires that the rate at which A (or B) enters minus the
rate at which A (or B) leaves be equal to the rate at which A (or B) is consumed by
chemical reaction in the control volume. Thus, for A,

NA NA dxdz r dxdydz ,
y y y

which leads to

dN A
rA ; (11.54)
dy
Mass Transfer in Fluids 503

FIGURE 11.9 Control volume in a film on the surface of a solid A into which gas B and
vapor A are transported by diffusion.

similarly for B,

dN B
rB . (11.55)
dy

Thus, from Eqs. (11.53) and (11.54),

dN A d2 XA
D AC 2
kr C 2 X A X B (11.56)
dy dy

and from Eqs. (11.53) and (11.55),

dN B d2 XB
DBC 2
kr C 2 X A X B (11.57)
dy dy

The equation
2
d2 XA d 1 XA
DA DB
dy 2 dy 2

is nonlinear and its solution requires numerical analysis. However, it was seen in the
preceding section that with a large enough reaction rate constant, the concentration
of the reactant can be decreased to zero at the reaction site. In the present case, with
a large enough value of kr, A and B are transported by diffusion to the reaction zone
504 An Introduction to Transport Phenomena in Materials Engineering

at y = Y, where fast chemical reaction decreases their concentrations to zero. Thus,

from Eqs. (11.56) and (11.57),

d2 XA
0 for 0 y Y (11.58)
dy 2

and

d2 XB
0 for Y y L. (11.59)
dy 2

Integration of Eq. (11.58) with the boundary conditions

XA 0 at y 0
XA
0 at y Y

gives

y
XA XA 0 1 , (11.60)
Y

and integration of Eq. (11.59) with the boundary condition

XB L at y L
XB
0 at y Y
gives

y Y
XB XB L . (11.61)
L Y

At the reaction site y = Y, the ratio of the molar fluxes of A and B must be equal to
the ratio of the stoichiometry coefficients in the chemical reaction; that is, as rA = rB,
Y occurs where N A = N B. From Eq. (11.60),

dX A
NA D AC DACX A 0 Y . (11.62)
dy

and from Eq. (11.61)

dX B
NB DBC DBCX B L Y.
dy
Mass Transfer in Fluids 505

Thus,
D A CX A 0 DB CX B L

Y L Y
or

L
Y = , (11.63)
1 DB X B L / DA X A 0

substitution of which into Eq. (11.62) gives the molar flux of A in the film as

D A CX A 0 DB X B L
NA 1 . (11.64)
L DA X A 0

In the absence of a chemical reaction and the case in which XA decreases to zero at
y = L, the molar flux of A is given by

NA DA CX A 0 L,

and thus when a fast chemical reaction occurs, increasing the molar concentration
gradient of A from -XA(0)/L to – XA(0)/Y, the flux of A is increased by the factor

DB X B L
1 .
DA X A 0

The concentration gradients occurring when a fast reaction occurs in the film are
shown in Figure 11.10.

FIGURE 11.10 Concentration profiles in the film shown in Figure 11.9, when chemical
reaction between A and B in the film is rapid enough to decrease the concentrations of both
species to zero at y = Y.
506 An Introduction to Transport Phenomena in Materials Engineering

Example 11.3
Fuming Liquid Iron into Air
When air flows over the surface of liquid iron at 1873 K, iron vaporizes, diffuses
through the stagnant film, and reacts with oxygen gas to form a fume of solid FeO.
The diffusion coefficient for gaseous iron in nitrogen at 1873 K is 6 x 10–5 m2/s and
the observed rate of fume formation is 1.25 x 10–2 kg/m2s.

Find: At what distance from the surface of the liquid iron is the fume formed?
Solution: On the assumption that the fuming rate is controlled by diffusion
of iron in the film, the answer is obtained from Eq. (11.63). The saturated
vapor pressure of liquid iron is

45, 390 K
ln pFe 1.27 ln T 35.45
T
(with pressure in Pascals and temperature in Kelvin); at 1873 K, pFe = 5.18 Pa. The
molar density of an ideal gas at 1873 K and 101.3 kPa is

n P 101,300 Pa
C 6.51 gmol / m 3 .
V RT 8.3144 J gmol K 1873 K

With a molecular weight of FeO of 0.07185 kg/gmol, iron is consumed by the

fuming reaction at the rate

1.25 10 2 kg/m 2 s
0.174 gmol Fe m 2 s .
0.07185 kg gmol

The mole fraction of iron in the gas phase at the surface of the liquid iron is

5.18 Pa
5.11 10 5 ,
101,300 Pa

and thus Y is obtained from Eq. (11.64),

5
6 10 m 2 s 6.51 gmol / m 3 5.11 10 5
2
0.174 gmol m s
Y

as Y = 1.15 x 10–7 m. So, the reaction rate is so fast that it occurs just above the
surface of the solid.

11.10 MASS TRANSFER AT LARGE FLUXES AND

LARGE CONCENTRATIONS
In the previous sections, we assumed that the product CAv* in Eq. (11.13) was
small enough to be ignored, in which case the molar flux of A could be expressed
Mass Transfer in Fluids 507

by Eq. (11.48). Consider the case in which a significant applied advective flux is
superimposed on the diffusion flux. An example of such a case occurs when the
solid A in Figure 11.5 is porous, and in addition to gas B flowing past the solid in
the x direction, gas B is forced through the pores in the y direction. When the gas
B emerges from the pores it is saturated with A and the molar flux of A in the y
direction is given by Eq. (11.13) as

NA j*A C A v* . (11.13)

In this expression,

C A v* C A C C A v A C B vB X A N A, y N B, y ,

and, because the total molar flux NT , y N A, y N B, y, Eq. (11.13) can be written as

dX A
NA j*A X A NT DAC X A NT . (11.65)
dy

At steady state in the film, dN A dy 0, or

d2 XA NT dX A
2
0. (11.66)
dy DAC dy

Letting r NT DAC allows us to write Eq. (11.66) as

d2 XA dX A d dX A
2
r 0 or rX A 0,
dy dy dy dy

which have the solutions

dX A dX A
0 and rX A 0.
dy dy

From these two equations, we can show that

XA C1 and X A C2 ery ,

summation of which gives

XA C2 ery C1

as the general solution to Eq. (11.66). Applying the boundary conditions,

XA 0 at y 0
XA
XA L at y L,
508 An Introduction to Transport Phenomena in Materials Engineering

gives

exp NT y / CDA 1
XA XA 0 XA L XA 0 . (11.67)
exp NT L / CDA 1

*
Thus, j A in Eq. (11.65) is obtained as

dX A exp NT y / CDA
j*A D AC NT X X 0 XA L . (11.68)
dy exp NT L / CDA 1

Substitution of Eqs. (11.67) and (11.68) into Eq. (11.65) gives

XA L XA 0
NA NT X A 0 , (11.69)
exp NT L / CDA 1

which in the limit of NT 0 approaches

CDA
NA XA L XA 0 . (11.70)
L

For a system in which L = 0.02 m, DA = 3 x 10–5 m2/s, XA(0) = 0.1, XA(L) = 0.01, and
the A-B mixture is an ideal gas at 50 °C and 101.3 kPa,

P 101,300 Pa
C 37.72 gmol / m 3
RT 8.3144 J gmol K 323 K

and Eq. (11.69) gives

0.09
NA NT 0.1 ,
exp 17.67NT 1

which is shown as a log-log plot in Figure 11.11. With decreasing NT , N A asymptot-

ically approaches the value of 5 x 10–3 gmol/m2s (logN A = −2.29), which is the same
as that obtained from Eq. (11.70):

CDA 37.73gmol m 3 3 10 5 m 2 s
NA XA L XA 0 0.01 0.1
L 0.02 m
3 gmol A
5 10 .
m2s
Mass Transfer in Fluids 509

FIGURE 11.11 Variation of the molar flux of A with total molar flux during mass transfer at
large fluxes and large concentrations.

When the flux of B in the y direction through the porous A is decreased to zero, the
flux of A is given by Eq. (11.42) as

DP P PA L 3 10 5 m 2 s 101,300 Pa 1 0.01
N A, y ln ln
RT L P PA 0 8.3144 J gmol K 313 K 0.02 m 1 0.1

5.57 10 3 gmol m 2 s,

and thus the line drawn in Figure 11.11 has a physical significance only at values
of N A greater than 5.57 × 10–3 g mol/(m2s). The concentration gradients across the
film predicted by Eq. (11.67) for NT = 0.1, 0.05, and 0.01 (gmol/m2s) are shown in
Figure 11.12.

11.11 INFLUENCE OF MASS TRANSPORT ON

HEAT TRANSFER IN STAGNANT FILM
Figure 11.13 shows a mixture of hot gas B and vapor A in contact with the surface
of a cooler solid A. The difference in temperature across the film causes conduction
heat transfer from the bulk gas to the solid surface and the concentration gradient of
A in the film causes the mass transport of A by diffusion from the surface to the bulk
gas. At finite temperatures, the atoms of A have finite enthalpies, and thus the molar
flux of A through the film is also an enthalpy flux, which contributes to the total heat
flux and thus influences the temperature gradient across the film. The enthalpy flux
caused by the molar flux of A is given by N A H A, in which HA is the enthalpy of A
510 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 11.12 Influence of the total molar flux on the concentration profile of A in the film
on the surface of porous solid A.

FIGURE 11.13 Temperature and concentration profiles in the film arising from simultane-
ous mass and heat transport.
Mass Transfer in Fluids 511

vapor per gram mole. An energy balance at steady state on the control volume shown
in Figure 11.13 is thus

net heat flux by conduction + net enthalpy flux 0

or

qy qy dxdz NAHA NA HA y
dxdz 0.
y y y y y

Dividing by dxdydz and allowing dz to approach zero gives

dq y dH A
NA 0
dy dy

or, for a constant thermal conductivity, k, and a constant heat capacity of A, cp,A,

d 2T N A c p,A dT
0,
dy 2 k dy

which has the solution

T C2 ery C1

in which

r N A c p,A k .

The quantity r is very large if most enthalpy transport across the layer is due to flux
of A and very small if mostly because of conduction heat transfer.
Thus, with the boundary conditions T y 0 T0 and T y L TL , the temper-
ature profile in y is

exp ry 1
T y T0 TL T0 . (11.71)
exp rL 1

Thus, at y = 0, the conduction heat flux is

dT TL T0 N A c p, A TL T0
qcond k k r
dy y 0
exp rL 1 exp N A c p,A L / k 1

and the enthalpy flux due to the molar transport of A across the film from y = 0 to
y = L is

q A N A c p, A TL T0 .
512 An Introduction to Transport Phenomena in Materials Engineering

The total heat flux is thus

1
qtotal qcond qA N A c p, A TL T0 1 . (11.72)
exp N A c p,A L / k 1

A heat transfer coefficient, h, is identified as

qtotal 1
h N A c p,A 1 .
TL T0 exp N A c p,A L / k 1

As by N A (and hence r) approaches zero, Eq. (11.71) goes to the pure conduction case,

y
T T0 TL T0 . (11.73)
L

In comparing Eqs. (11.71) and (11.73) for a positive value of N A (and hence r),

exp ry 1 y
,
exp rL 1 L

in which case, T(y) in Eq. (11.71) is less than in Eq. (11.73). The enthalpy flux in this
case causes the temperature profile in the film to be convex downward.
If a steep temperature gradient exists from the surface of a hot liquid A to a cold
gas B, the A that evaporates from the surface and diffuses toward the bulk gas can
condense within the film, which, by decreasing the value of XA at the point of conden-
sation, increases the gradient of XA from the surface and hence increases the diffusion
flux of A. Furthermore, if the concentration of A vapor in the film is high enough, the
latent heat of condensation of A evolved within the film also can influence the shape
of the temperature profile in the film.

11.12 MASS TRANSFER COEFFICIENT FOR CONCENTRATION

BOUNDARY LAYER ON A FLAT PLATE
As discussed in Sections 10.1 and 10.2, the heat flux (q ) from a solid surface at
temperature TS to a moving fluid at T∞ can be characterized by the heat transfer coef-
ficient, h, by Newton’s law of cooling,

q h Ts T . (11.74)

Remembering that there is no relative fluid velocity at the solid surface (call it y = 0),
we have only conduction into the fluid there,

T
q k ,
y y 0
Mass Transfer in Fluids 513

so the heat transfer coefficient in Eq. (11.74) is a representation of the temperature

gradient in the fluid at the surface:

k T y y 0
h . (11.75)
Ts T

Consider now a fluid B flowing past the surface of solid A, which is slightly soluble
in the fluid B. If the fluid is not saturated with A, mass transport occurs as the disso-
lution of A into a liquid B or the sublimation of A into a gas B. If phase equilibrium
is maintained at the interface between the solid and the fluid (y = 0), there the molar
concentration of A, CA,S, is the solubility limit of A in B, and if the molar concen-
tration of A in the bulk fluid is CA,∞, a local mass transfer coefficient, hm(x) at the
location x on the surface is defined as

N A, y hm C A,s C A, . (11.76)

Alternatively, in terms of mass concentration of A,

n A, y hm A, s A, (11.77)

and hm has the units m/s. In using the same type of physical model to define a mass
transfer coefficient, we assume that fluid B is unaffected by the amount of A in it
and that the shape of the surface is not altered significantly by the release of A into
the passing fluid. As with heat transfer, the no-slip condition at y = 0 means all mass
transport there occurs by diffusion,

A
n A, y j A, y DA ,
y 0 y y 0

and thus
DA A / y y 0
hm . (11.78)
A, s A,

In heat transfer the Nusselt number is defined as a nondimensional heat transfer

coefficient, NuL = hLC/k, in which LC is a characteristic length. The mass transfer
equivalent of the Nusselt number is the Sherwood number, ShL, which is defined as

ShL hm LC DA . (11.79)

We recall the development of boundary layer theory in Chapters 9 and 10 and

draw an analogy between convective heat and mass transfer. The treatment is
started by considering the flat plate with a uniform freestream (velocity, tempera-
ture, and composition) and no pressure gradient, as discussed in Sections 9.3, 9.4,
and 10.4. The dissolution of A into the flowing fluid B causes the formation of a
514 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 11.14 Concentration profiles of A at several locations downstream from the lead-
ing edge in a concentration boundary layer on a flat plate of solid A over which fluid B is
flowing, when A is slightly soluble in B. A horizontal line in the boundary layer shows how
the composition at a specific height increases farther downstream.

concentration boundary layer, shown in Figure 11.14, which is mathematically

similar to the thermal boundary layer shown in Figure 10.9. In Figure 11.14, the
concentration of A in the fluid B in the boundary layer at any value of y near the
plate increases with increasing x due to dissolution of the plate. Consequently,
the gradients A y at the fixed value of y (including at y = 0) decrease with
increasing x. The concentration boundary layer that develops over the plate has a
value at its upper edge of the freestream concentration of A, and as A y y 0
decreases with increasing x, the local rate of mass transport of A from the plate and
hm both decrease farther downstream.
The control volume analysis in Section 10.2 for deriving a thermal energy balance,
including storage, advection, diffusion, and generation is repeated here for solute
conservation in a mixture. That mass balance for two-dimensional molar transport of
constituent A through a Cartesian control volume (as in Figure 10.3) is

CA CA CA j *A, x j *A, y
u* v* .
t x y x y (11.80)
storage advecction diffusion

The equivalent expression for mass transport is

A A A
j A, y j A, y
u v .
t x y x y (11.81)
storage advectionn diffusion
Mass Transfer in Fluids 515

For fluids of constant density and mass diffusion coefficient in a steady flow, the
mass balance is
2 2
A A A A
u v DA 2 2
, (11.82)
x y x y

which is the mass transport equivalent of the momentum and thermal balance equa-
tions, Eqs. (7.33) and (10.10), respectively.
Comparison of the momentum and thermal boundary layers was facilitated in
Section 10.2 by the introduction of the Prandtl number, Pr = / . In a similar manner,
comparison of the momentum and concentration boundary layers is facilitated by the
Schmidt number, Sc, the ratio of viscous momentum transport and mass transfer by
diffusion, defined as

Sc . (11.83)
D

Like the Prandtl number, the Schmidt number governs the ratio of the viscous and
concentration boundary layers. Comparison of the thermal and concentration bound-
ary layers is facilitated by the Lewis number, Le:

Sc
Le . (11.84)
D Pr

Given mathematically identical models of laminar convection heat and mass transfer,
the integral and exact solutions for heat transfer correlations on a flat plate boundary
layer, in Section 10.4, are the same for mass transport, with the replacement of Nu
and Pr with the Sherwood and Schmidt numbers. These solutions for the mass trans-
port advection-diffusion equation in a flat plate boundary layer give

Shx hm x D C Re1x 2 Sc1 3 , (11.85)

where C = 0.393 for the integral analysis and C = 0.332 for the exact solution.
The average mass transfer coefficient, hm, is obtained by averaging hm(x) between
x = 0 and x = L, where L is the length of the plate, as:

1 L D
hm hm x dx 2C Re L1 2 Sc1 3 , (11.86)
L 0 L

which is twice the local value at the end of the plate, hm(x = L). The definition of the
average Sherwood number is then

Sh L hm L D 2C Re L1 2 Sc1 3 , (11.87)

with C defined after Eq. (11.85). This result is valid in a Sc range from (1) and
higher.
516 An Introduction to Transport Phenomena in Materials Engineering

In turbulent flow, assuming Recrit ≈ 5 x 105, the mass transport correlation is

Shx 0.0291 Re 4x 5 Sc 0.6 , (11.88)

for Sc ~ (1), and the average mass transfer coefficient for a plate of length L on
which a transition from laminar to turbulent flow occurs is thus

1 xcrit L DSc1/ 3
hm hm,lam dx hm,turb dx 0.0364Ree 4L/ 5 871 . (11.89)
L 0 xcrit L

Example 11.4
Evaporation from a Water Pond
Cooling water for a power plant is stored in a pond 1000 m in length and 500 m wide.
A dry wind at 300 K blows in a horizontal direction parallel to the 1000 m side of
the pond at a velocity of 2 m/s. The water in the cooling pond is also at 300 K. The
critical Reynolds number for this geometry is Recrit = 5 x 105.

air / air
1.67 10 5 m 2 / s
DHair2O 2.6 10 5 m 2 s

Find: The rate of evaporation from the pond.

Solution: The transition from laminar to turbulent boundary conditions occurs
at xcrit,

v 1.67 10 5 m 2 s
xcrit Re crit 500,000 4.18 m.
u 2m s

Almost none of the boundary layer is laminar (4.18 m << 1000 m), so we will
consider that the entire layer is turbulent. Using this result, we find the average mass
transfer coefficient from Eq. (11.88),

1 L 1 5 DSc1/ 3
hm hm,turb dx 0.0291DSc1/ 3 Re 4L/ 5 0.03664Re 4L/ 5
L 0 L 4 L
1/ 3
1.67 10 5 m 2 s
2.6 10 5 m 2 s 4/5
2.6 10 5 m 2 s 2 m s 1000 m
0.0364
10000 m 1.67 10 5 m 2 s
3
2.37 10 W m2 K .
The mass flux from the surface of the pond is thus

m hm, L A H2 O s H2 O ,
Mass Transfer in Fluids 517

which, with
air pM RT
gives

pHo pH
2O s 2O
m hm, L AMH2 O .
RTs RT

The variation with temperature of the saturated vapor pressure of water (in Pa) is

6679 K
ln pHo 2O 4.65 ln T K 56.97.
T
o o
At 300 K, pH2O ( s ) = 3580 Pa, while the dry air has pH2O ( ) = 0. The mass flow of
water by evaporation from the pond to the moving air is

3 m kg H 2 0
m 2.37 10 1000 m 500 m 0.018
s gmole H 2 0

3580
5 Pa
0
8.134 J gmole H 2 0 K 300 K
kg
31.3 .
s
Maintenance of a constant level in the storage pond thus requires that water be added
at the rate of 2.70 x 106 kg/day (≈ 2700 m3/day). This calculation does not consider the
influence of the absorption of the latent heat of evaporation, which tends to cool the water.

As another example of the analogue between heat and mass transfer, we can use the
correlation for the average convection heat transfer from a sphere for the isothermal
mass transfer from the same configuration:

Sh D 2 0.4 Re1D2 0.06 Re 2D 3 Sc 0.4 , (11.90)

where the Sherwood and Reynolds numbers are based on D, the sphere diameter.
This section emphasizes the ability to apply the analyses and correlations for con-
vection heat transfer to mass transfer from a solid surface to a flowing fluid with a
different composition. Although we have only examined a few specific cases (flat
plate, sphere), similar results from Chapter 10 can be applied to problems of convec-
tion mass transfer.

11.13 SIMULTANEOUS HEAT AND MASS

TRANSFER: EVAPORATIVE COOLING
Figure 11.15 shows gas B at the temperature T∞, flowing over the surface of liquid
A. If the vapor pressure of A in the gas is less than the saturated vapor pressure of
518 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 11.15 Gas B at the temperature T∞, flowing over the surface of evaporating liquid
A at the temperature Ts.

A at the temperature of the surface, TS, A evaporates into the flowing gas. However,
because evaporation is an endothermic process, it causes a decrease in the enthalpy
and temperature of the liquid A. In turn, the difference between T∞ and Ts causes the
transfer of heat from the gas to the liquid, and a steady state is reached when the rate
of heat transfer from the gas to the liquid equals the rate at which the latent heat of
evaporation is absorbed. The heat flux from the gas to the liquid is from Newton’s
law of cooling, q h TS T . With a mass evaporation rate of m A (kg/s) and a
latent heat of evaporation of H A,evap (J/kg), latent heat is supplied at the flux

qevap m A H A,evap hm A,Ts A,T H A,evap .

Thus, at steady state,

qconv h Ts T hm A,Ts A,T H A,evap qevap . (11.91)

From Eqs. (10.59) and (11.85),

13 13
hm DA Sc1 3 DA B B DA2 3 13
B
(11.92)
h kB Pr1 3 kB DA B kB

or, as k = ( c),
23
hm 1 DA 1
23
. (11.93)
h B cB A B cB Le

Furthermore, if vapor A and gas B mix ideally,

A p A M A RT , (11.94)
Mass Transfer in Fluids 519

and substitution of Eqs. (11.92) and (11.93) into Eq. (11.91) gives

M A H A,evap 1
23
poA,TS p A,T
T TS , (11.95)
B cB Le RTs RT

in which p A,T is the vapor pressure of A in the bulk gas stream. In Eq. (11.95), the
properties B, cB, and Le should be evaluated at the film temperature, Tf = (TS+T∞)/2.

Example 11.5
Heat Balance for Cooling of Evaporating Water
Find: The steady-state temperature of liquid water when dry air at 40 °C flows over
it. The data are

M H2O 0.018 kg gmol

H H2O,evap 3.338 105 J kg
cair 1005 J kg K

air 1.133 kg/m 3

6679 K
ln pHo 2O 4.65 ln T K 56.97.
T

Solution: As TS is unknown, the film temperature cannot be calculated yet, and

so, to begin, we will use the properties of dry air at 40 °C and then iterate from the
results for Ts. The Lewis number for moist air is approximately 1.1, in which case
Eq. (11.95) gives

0.018 kg gmol 3.338 105 J kg 1

23
313K TS
1.133 kg/m 3 1005 J kg K RTs 1.1

6679 K TS
exp 4.65 ln 56.97
TS K
0.5596 K 2 6679 K
exp 4.65 ln TS K 56.97 ,
Ts TS

which has the solution Ts = 304 K. Recalculation using a mean temperature of

(304 + 313)/2 = 309 K changes the 0.596 to 0.584, which also gives a solution of
Ts ≈ 304 K, so we have a converged solution.
The steady-state temperature of the liquid is independent of the velocity of the gas,
because from the heat transfer analogy, the mass transfer coefficient, hm, is the same
function of the Reynolds number as the heat transfer coefficient, and so the ratio
520 An Introduction to Transport Phenomena in Materials Engineering

hm/h in Eq. (11.91) is independent of the Reynolds number. At a film temperature of

309 K, the Prandtl number and thermal conductivity of air are, respectively, 0.711
and 0.0266 W/mK and thus, at a Reynolds number of 5 × 105,

13 12
h 0.393 k Pr1 3 Re1 2 0.393 0.0266 W mK 0.711 5 105 6.60 W m 2 K

and

mA h T TS H evap 6.60 W m 2 K 313 K 300 K 3.338 105 J kg

4
2.57 10 kg m 2 s .

11.14 SUMMARY
Measurement of a mass transport flux in fluids is determined by the relative motion
of the observer of the flux. An observer moving with the velocity of the bulk fluid
sees mass transport due to diffusion down concentration gradients in the fluid, and a
stationary observer sees mass transport due to bulk fluid flow and diffusion. Consid-
eration of mass transport from the surface of a solid to a flowing fluid is simplified by
the film model in which it is envisioned that mass transport in a thin film of fluid adja-
cent to the surface occurs solely by diffusion, and that at distances from the surface
greater than the thickness of the film, the fluid is well mixed and has no concentration
gradients. When mass transport occurs from a cool solid to a warm fluid, the mass
flux also carries an enthalpy flux and so alters the temperature gradient in the fluid
and the heat transfer through it.
The local mass transfer coefficient at the interface between a solid and a fluid is
defined in a manner analogous to that used to define the local heat transfer coeffi-
cient, and the mass flux is proportional to the difference between the mass density
of the fluid in contact with the surface and the mass density of the bulk fluid, with
the mass transfer coefficient being the proportionality constant. The variations of
the local mass transfer coefficient with position on the surface, the surface-averaged
mass transfer coefficient, and the thickness of the concentration boundary layer are
derived by analogy with heat transfer by convection from the solid to the fluid. This
analogy leads to the definition of the Sherwood (Sh) number and the Schmidt (Sc)
number, and, in forced convection, mass transport is modeled with the same corre-
lations as the heat transfer correlations and analyses in Chapter 10, except with the
replacement of Nu and Pr with Sh and Sc.

11.15 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

11.1 Two large vessels containing N2-CO2 mixtures at 273 K and a total pressure
of 101.3 kPa are connected by a tube of length 1 m and 0.05 m diameter. If
the partial pressure of CO2 in one tank is maintained constant at 104 Pa and
Mass Transfer in Fluids 521

in the other tank is maintained constant at 5000 Pa, calculate the mass flow
rate of CO2 between the tanks under conditions of equimolar diffusion. The
interdiffusion coefficient is D = 1.44 × 10–5 m2/s.
11.2 A container is cooled by evaporation of a volatile liquid A from a cloth
wrapped around the container that is saturated with the liquid. The evapo-
ration occurs into the air at 50 °C, and at steady state, the temperature of the
container is 8 °C. Assuming a film temperature of 300 K, calculate the inter-
diffusion coefficient DA-air, given the following properties for A: MWA = 0.2
kg/g mol, ΔHevap,A = 50,000 J/kg, and ln poA,liq 1200 K T 12.79
(where pressure is in Pa).
11.3 Consider a thin layer of water lying on the ground under a clear night sky.
The effective temperature of the sky is −50 °C, the emissivity of the surface
of the water is 1.0, and the heat transfer coefficient for transfer of heat from
the air to the water is 30 W/m2K. It can be assumed that there is no transfer
of heat by conduction between the water and the ground below it and that
the night air is dry. The saturated water vapor pressure at 0 °C is 614 Pa
and the latent heat of freezing of water is 3.338 × 105 J/kg. (a) Estimate the
mass transfer coefficient for evaporation of the water. (b) Calculate the air
temperature below which the ground water will freeze.
11.4 A circular disk of diameter 0.05 m supports a thin film of water. When 0.95
W of electric power is supplied, the water reaches a steady-state tempera-
ture of 305 K and water evaporates at the rate of 5 × 10 –3 kg/hr into the
surrounding dry air, which is at a temperature of 295 K. (a) Calculate the
heat transfer coefficient for transfer of heat by convection from the water
to the air. (b) If the relative humidity of the air is increased to 50% at 295
K and the power input is maintained at 0.95 W, does the water temperature
increase or decrease? (c) Does the rate of evaporation increase or decrease?
(d) Does the mass transport–heat transport analogy hold?
11.5 During the calcination of limestone, the rate of decomposition of the
CaCO3 is controlled by mass transport of CO2 from the surface of the solid
to the surrounding atmosphere. On the assumptions that the CO2 diffuses
through a stagnant film of air of thickness 0.001 m above the surface,
and the concentration of CO2 in the calcining furnace is zero, calculate
the rates of decomposition of CaCO3 at 1000 K and at 1100 K. The dif-
fusion coefficients are DCO2-air = 4.71 × 10 –4 m2/s (at 1000 K) and 5.54 ×
10 –4 m2/s (at 1100 K). The standard free energy for the reaction CaO(s) +
O2(g) = CaCO3(solid), based on 1 atm pressure as the standard state, is
G 161, 300 J/g mol 137T J/g mol K .
11.6 Three rectangular shallow trays each of length 0.5 m and containing water
are laid end to end in contact with one another, and a dry airstream at 300
K flows over the three trays in the lengthwise direction. Calculate the elec-
trical power inputs to each of the three trays required to maintain the water
at a temperature of 300 K when the freestream velocity of the air is (a)
5 m/s, and (b) 10.47 m/s. (c) When the freestream velocity is 10.47 m/s,
what would the relative humidity of the air have to be to require a power
input of 50 W to the middle tray in order to maintain the water in the tray
522 An Introduction to Transport Phenomena in Materials Engineering

at 300 K? The saturated water vapor pressure at 300 K is 3580 Pa, and the
latent heat of evaporation of water is 3.338 × 105 J/kg.
11.7 Fine particles of coal are combusted at 1450 K in pure oxygen gas at a
pressure of 101.3 kPa. If the reaction C(s) + O2(g) = CO2(g) is sufficiently
rapid that the concentration of oxygen in the gas phase at the surface of the
particle is zero, and the reaction rate is controlled by diffusion of oxygen in
the gas to the surface of the particle, calculate: (a) the rate of combustion
of a spherical particle with an initial radius of 10 –4 m; (b) the time for the
particle in (a) to be completely combusted. Consider the particle to be pure
carbon of density 2000 kg/m3. The interdiffusion coefficient DO2-CO2= 1.7 ×
10 –4 m2/s at 1450 K, and for steady-state diffusion of A in spherical coordi-
nates,

d 2 dC A
r D 0.
dr dr
11.8 Repeat Problem 11.7, assuming that the combustion reaction obeys the rate
law

r kr CO2 ro ,

where the rate constant is kr = 0.1 m/s, r is the rate of consumption of

oxygen in gmol/m2s, and CO2(ro) is the concentration of oxygen gas at the
surface of the particle.
12 Radiation Heat Transfer

12.1 INTRODUCTION
The transport of heat as thermal radiation is quite different from heat transport by con-
duction or convection. Thermal radiation is part of the electromagnetic spectrum (Fig-
ure 12.1) and hence the propagation of energy by thermal radiation (unlike by other
modes) does not require the presence of matter; the earth is heated by thermal radiation
from the sun transported through the vacuum of space. In the electromagnetic spectrum,
significant thermal radiation is confined to wavelengths in the range 0.1 to 100 μm, which
includes visible light (0.4 to 0.7 μm) and ultraviolet and infrared light at, respectively,
lower and higher wavelengths. Electromagnetic waves are produced by oscillations of
the electrical fields surrounding charges, and thus all matter at a finite temperature emits
thermal radiation caused by atomic scale motions such as the vibration and rotation of
molecules, the vibration of atoms, and certain electronic transitions. Although thermal
emission can be thought of in terms of the propagation of discrete packets of energy,
called quanta or photons, it is more conveniently dealt with as the propagation of elec-
tromagnetic waves of wavelength and frequency , which are related as

cL , (12.1)

where c is the velocity of light in the medium through which the radiation is passing.
The speed of light in a vacuum, cL, is 2.998 X 108 m/s.
A surface at a finite temperature emits thermal radiation over a range of wave-
lengths in all directions and the intensity and power of this radiation vary with these
emission wavelengths and directions.
While emitting radiation, a surface can also receive thermal radiation from some
other source (think about laying on a beach on a sunny day), in which case the sur-
face is said to be irradiated. If the matter being irradiated is opaque to the incoming
radiation, the irradiation is reflected and/or absorbed by the surface, or if the irra-
diated matter is at least partially transparent to the incoming radiation, some of the

FIGURE 12.1 Electromagnetic spectrum

DOI: 10.1201/9781003104278-12 523

524 An Introduction to Transport Phenomena in Materials Engineering

irradiation is transmitted through the matter. The fractions of the irradiation reflected,
absorbed, and transmitted are called, respectively, the reflectivity, absorptivity, and
transmissivity of the surface being irradiated. Thus, three types of flux must be con-
sidered in the treatment of radiation: (i) the emissive flux emanating from a surface
by virtue of the finite temperature of the surface; (ii) irradiation, which is the flux
from another source of radiation incident on the surface; and (iii) the radiosity of the
surface, which is the sum of the intrinsic emission and the reflected irradiation. First,
we will provide more detail of this radiation behavior at surfaces, then we will use
these concepts to develop models of radiation exchange in enclosures.

12.2 INTENSITY AND EMISSIVE POWER

In Figure 12.2 (a) radiation is emitted in all directions and at all wavelengths by the
surface element dA1. Consider the radiation emitted by dA1, in the direction ( , ), at
the wavelength , and passing through the element of area dA2, which is normal to
the direction ( , ), and located at a distance r from the surface element dA1. When
viewed from dA1, the area dA2 subtends a solid angle d defined as

dA2
d , (12.2)
r2
In which the unit of is the steradian (sr).1 The differential area on the surface at
distance r, is

dA2 r 2 sin d d ,

and hence

d sin d d .

The solid angle subtended by a hemisphere placed over dA1 is obtained by integrating
d over the intervals = 0 to π/2 and = 0 to 2π:

2 /2 /2
sin d d 2 cos 0
2 sr.
0 0

In considering the radiation emitted by the surface element dA1, the monochro-
matic or spectral intensity, I ,e, of the emission is defined as

I ,e rate at which radiant energy is emitted

at wavelenggth
in direction ,
per unit area of emitting surface normal to direction ,
per unit solid angle about thatt direction
per unit wavelength interval d about .
Radiation Heat Transfer 525

FIGURE 12.2 (a) Radiation of wavelength emitted by the surface element dA1, in the
direction ( , ) and intercepted by the area dA2; (b) relationship between the area that dA1
projects in the direction and the actual area dA1.

Thus, with wavelength measured in micrometers (1 μm = 10–6 m), spectral intensity

has the units W/(m2 sr μm). As shown in Figure 12.2 (b), the projected surface area
normal to the direction , and seen when viewed from area dA2, is cos dA1, and
hence

dq
I ,e , , . (12.3)
cos dA1 d d
526 An Introduction to Transport Phenomena in Materials Engineering

The influences of wavelength and direction on the intensity of the radiation are sep-
arated by defining

dq
dq
d

as the rate at which radiation of wavelength leaves dA1 and passes through dA2.
Thus, the spectral influence can be factored out and Eq. (12.2) can be written as

dq I ,e , , cos dA1 d (12.4)

or

dq
dq I ,e , , cos sin d d , (12.5)
dA1

in which d q is the rate at which radiation of wavelength leaves dA1 per unit area
(not per unit projected area) and passes through dA2. If I ,e is known as a function of
and , integration of Eq. (12.5) gives

2 /2
dq I ,e , cos sin d d (12.6)
0 0

and, if I ,e is known as a function of wavelength,

q q d (12.7)
0

gives the total radiation heat flux emitted per unit area, which is intercepted by the
hemisphere above the surface 1.

12.2.1 eMissive Flux

The spectral, hemispherical emissive flux, E , of radiation emitted by a surface is
the rate at which radiation of wavelength is emitted in all available directions, per
unit surface area, per unit wavelength d about . Thus, per unit area of surface, this
quantity is obtained from Eq. (12.6) as

2 /2
E q I , , cos sin d d . (12.8)
0 0 ,e

Integrating Eq. (12.8) over all wavelengths gives the total emissive flux of the radi-
ation as

E E d . (12.9)
0
Radiation Heat Transfer 527

If the intensity of the emitted radiation is independent of the direction of emission,

the surface is said to be a diffuse emitter, and for such a surface, Eq. (12.8) becomes
/22
2 /2 1 2
E I ,e cos sin d d I ,e 2 sin
0 0 2 0 (12.10)
I ,e ,

in which I ,e has the units W/m2sr and π has the units sr.
Similarly, for a diffuse emitter, the total emissive flux, obtained by averaging over
all wavelengths, is

E Ie , (12.11)

where Ie is the total intensity of the emitted radiation. Casual inspection of Eq. (12.11)
might raise the question: why is the total heat flux from a diffuse emitter not given
by the radiant heat flux per steradian (the intensity) times the number of steradians in
a hemisphere, 2π? The factor 2 does not occur in Eq. (12.11) because Ie is defined in
terms of the area that dA1 projects in the direction normal to the angle ( , ), and E is
defined in terms of the actual area dA1.

12.2.2 irradiation
In a manner that is directly analogous with the definition of the monochromatic
(spectral) intensity of radiation emitted by a surface, the monochromatic (spectral)
intensity of radiation from some other source that is intercepted by the surface I ,i is
defined as

I ,i rate at which radiant energy is incident to a surfacee

at wavelength
from direction ,
per unit area of intercepting surface normal to direction ,
per unit solid angle about this direction
per unit wavelength interval about .

The spectral irradiation, G , in units of W/(m2μm) is then defined as the rate at which
radiation from all directions is incident on a surface, per unit area of the surface, at
the wavelength per unit wavelength d about :
2 /2
G I ,i , , cos sin d d (12.12)
0 0

and the total irradiation, G, is obtained as

G G d . (12.13)
0
528 An Introduction to Transport Phenomena in Materials Engineering

If the incident radiation is diffuse,

G I ,i (12.14)

and the total irradiation is

G Ii . (12.15)

12.2.3 radiosity
If some of the incident radiation on the surface is reflected by the surface, the radios-
ity of the surface is the sum of the direct emission and the reflected irradiation. The
spectral radiosity, J , is the rate at which radiation leaves a unit area of the surface at
the wavelength per unit wavelength interval d in all directions. Thus,

2 /2
J I ,e r , , cos sin d d , (12.16)
0 0

and the total radiosity is

J J d . (12.17)
0

If the surface is both a diffuse emitter and a diffuse reflector,

J I ,e r (12.18)

and

J Ie r.

12.3 BLACKBODY RADIATION

The large variations in the radiative properties of real surfaces are dealt with by
comparing the properties of real surfaces with those of a blackbody, which is an ideal
surface. For any given temperature, radiation from a blackbody has the highest possi-
ble value of spectral intensity and hence the highest possible spectral emissive power.
A blackbody is so named because it absorbs all incident radiation of all wavelengths2
and is also a diffuse emitter. The spectral emissive flux of a blackbody was worked
out by Max Planck as obtained from Eq. (12.10),

2 h c02
E ,b ,T 5
(12.19)
exp h c0 / kT 1

C1 W
5 2
,
exp C2 / T 1 m m
Radiation Heat Transfer 529

in which h (= 6.6256 × 10–34 J s) is Planck’s constant, k (= 1.3805 × 10–23 J/K) is

Boltzmann’s constant, and C1 (= 3.742 × 108 W μm4/m2) and C2 (= 1.439 × 104 μm
K) are the first and second radiation constants. Eq. (12.19) is known as Planck’s
distribution and is shown for several temperatures in Figure 12.3. This plot shows
some features generally found in radiation emission. For instance, as temperature
increases, the emitted power increases exponentially at every wavelength and the
peak of the distribution shifts to lower wavelengths. We can also see that the emissive
power seems to be a continuous function of wavelength (although that feature is only
true for ideal surfaces).
Setting dE /d = 0 gives the variation with temperature of the wavelength at
which the spectral intensity has its maximum value, max, as

max T = C3 2898 m K. (12.20)

Eq. (12.20) is Wien’s displacement law, which, on the log-log plot of Figure 12.3, is
a straight line with a slope of −1.
The total emissive power of a blackbody at any temperature T is the area under
the curve in Figure 12.3:

Eb E ,b d T 4. (12.21)
0

Eq. (12.21) is the Stefan–Boltzmann law, named after Josef Stefan, who measured
the dependence of the total emissive power on temperature, and Ludwig Boltzmann,
who derived the relationship theoretically; is the Stefan–Boltzmann constant [1].3

FIGURE 12.3 Planck’s distribution of the spectral emissive power of a blackbody.

530 An Introduction to Transport Phenomena in Materials Engineering

The Stefan–Boltzmann law shows that the total flux of heat by radiation is emitted
by a blackbody is a function only of temperature.
The emission from a blackbody in the interval of wavelengths from l to 2 is

2
E ,b d ,
1

and hence the fraction of the total emission from a blackbody at the temperature, T,
occurring in this interval is

2
E ,b d
f 1
.
1 2
T4
With the lower limit of 1 = 0, this fraction is defined as

E ,b d
0
f0 ,
T4
which, from Eq. (12.19), can be written as

C1 d
f0 4 5
. (12.22)
T 0 exp C2 / T 1

Values of f(0 ) are listed for discrete values of T in Table 12.1.

The two polynomials curve-fit to Eq. (12.22) in references [2, 3] may be more
easily used than Eq. (12.22) or Table 12.1 to find f(0- T) as a function of T. Defining

C2 T,

these fitted functions are

3 2 4 6 8
15 1
f0 T 1 4
for 2
3 8 60 5040 272,160 13, 305,600

TABLE 12.1
Blackbody Radiation Functions
T f0 T f0 T f0 T f0
(μm-K) (μm-K) (μm-K) (μm-K)
600 0.00000 3000 0.27322 6000 0.73778 14,000 0.96285
1000 0.00032 3500 0.38290 7000 0.80807 16,000 0.97377
1500 0.01285 4000 0.48085 8000 0.85625 20,000 0.98555
2000 0.06672 4500 0.56429 9000 0.88999 30,000 0.99529
2500 0.16135 5000 0.63372 10,000 0.91415 50,000 0.99890
2898 0.25011 5500 0.69087 12,000 0.94505 100,000 0.999905
Radiation Heat Transfer 531

and
3 nf exp n
15
f0 T 4 4
n 3 n 6 n 6 for 2,
n 1 n

where nf is the number of terms required to achieve a converged solution.

Example 12.1
Calculating Fractions of Blackbody Emissions from the Sun
The sun is modeled as a blackbody radiating at an effective temperature of 5800 K.

Find: The fractions of the sun’s emission that occur in (a) the ultraviolet range
(between wavelengths of 0.01 and 0.4 μm), (b) the visible range (between
wavelengths of 0.4 and 0.7 μm), and (c) the infrared range (between wave-
lengths of 0.7 and 1000 μm).
Solution: (a) With T = 5800 K, T varies from 58 to 2320 μm K in the ultra-
violet range. Thus, from Table 12.1, with T = 58 μm K, f (0 0.01 μm) = 0, and
with T = 2320 μm K, f(0 0.04 μm) = 0.125. Therefore,

f 0.01 0.04 m at 5800 K 0.125 0 0.125

so 12.5% of the emission occurs in the ultraviolet range.

(b) With T = 5800 K, T varies from 2320 to 4060 μm K in the visible range.
Thus,

f0 0.40 m 0.125 and f0 0.70 m 0.491

and

f 0.40 m 0.70 m at 5800 K 0.491 0.125 0.366,

so 36.6% of the emission occurs in the visible range.

(c) From Table 12.1, f(0 1000μm) for T = 5800 K is essentially unity, and hence
the fraction of the sun’s radiation emitted in the infrared range is 1–0.125–
0.366 = 0.509.

12.4 SURFACE PROPERTIES

In previous sections, we have assumed black (ideal) surfaces but recognize that few
surfaces behave ideally. Using the diffuse, perfectly emitting, and perfectly absorb-
ing blackbody as a standard, we can define various surface properties. We begin to
define several properties by comparing actual emissions to that of a blackbody, then
describe the possible fates of radiative heat flow incident on the surface.
532 An Introduction to Transport Phenomena in Materials Engineering

Comparison of the radiation emitted by a real surface with that emitted by a

blackbody gives rise to the definition of the total emissivity of the real surface,
, as

radiation emitted by a real surface at temperature T

.
radiation emitted by a blackbody at the same temperature T

While a blackbody has emissions independent of direction, the diffuse assumption is

not always reasonable on real surfaces, so we are interested in the dependence of on
on a surface. The nature of a possible variation of emission with direction is shown
in Figure 12.4, which shows emitted radiation with a maximum value in a direction
normal to the surface, = 0, and decreases with increasing .
Typical functions of ( ) for conducting and nonconducting materials are shown
in Figure 12.5 (a). For conducting materials, is approximately constant for in
the range 0–40°, after which it increases at higher to a maximum value, and then
quickly decreases to zero at 90°. In contrast, for nonconducting materials decreases
very slowly up to = 70°, after which its rate of decline increases rapidly. The vari-
ations of ( ) for highly polished surfaces of the nonconductors Al2O3 and CuO and
the conductors Al and Cr are shown in Figure 12.5 (b). Figure 12.5 shows that the
hemispherical emissivity, , is not significantly different from the normal emissivity,
n, corresponding to = 0.

FIGURE 12.4 Typical directional distribution of emission from a surface.

Radiation Heat Transfer 533

If we treat a surface as diffuse, ignoring the relatively small amount of emission

near the parallel to the surface ( approaches 90o), we can define a monochromatic,
hemispherical emissivity.

spectral hemispherical emissive power

of surfacee at temperature T E ,T
,T
spectral hemispherical emissive power E ,b ,T
of a blackbody at temperature T

FIGURE 12.5 Typical variations of the total directional emissivities, , with angle of emis-
sion (a) schematic behaviors of surfaces of a conductor and a nonconductor and (b) actual
behavior of surfaces of two oxides (nonconductors) and two metals (conductors).
534 An Introduction to Transport Phenomena in Materials Engineering

Figure 12.6 shows the variation, with wavelength, of the spectral, normal emissiv-
ities of various materials. This plot covers values of wavelength including the visible
and higher temperature infrared ranges, and, with Figure 12.5, shows that electrical
conductors (e.g., metals) typically have much lower emissivities than nonconductors.
This trend applies strictly to surface conditions, as oxidized metal surfaces (generally
poor conductors) have higher than the base metal from which they are formed.
If we integrate over all directions and wavelengths, the total hemispherical emis-
sivity, (T), is only a function of temperature.

total emissive power of radiation from surface at temperature T

T
total emissive power of blackbody radiation at temperature T

E T ,T E ,b ,T d
0
(12.23)
Eb T Eb T

Figure 12.7 shows the corresponding variations, with temperature, of the total
normal emissivities.

FIGURE 12.6 Variations with wavelength of the spectral normal emissivities of several
substances.

Example 12.2
Estimating Total Hemispherical Emissivity of a Ceramic
The spectral hemispherical emissivity of Al2O3 at 1400 K, shown in Figure 12.6, can
be approximated by the curve in Figure 12.8.
Radiation Heat Transfer 535

Find: Estimate the total hemispherical emissivity of Al2O3 at 1400 K.

Solution: (a) From Eq. (12.23),

E ,b ,T d
0
T ,
Eb T

which, using the ( ) curve in Figure 12.8, can be written as

4 m E ,b ,T 11 m E ,b ,T
T 0.2 d 0.95 d
0 Eb T 4 m Eb T

4 m E ,b ,T 11 m E ,b ,T 4 m E ,b ,T
0.2 d 0.95 d d .
0 Eb T 0 Eb T 0 Eb

From Table 12.1,

4 m and T 1400 K : T 5600 m K and f0 4 m 0.701

11 m and T 1400 K : T 15, 400 m K and f0 11 m 0.972

and

1400 K 0.2 0.701 0.95 0.972 0.701 0.398.

FIGURE 12.7 Variations with temperature of the total normal emissivities of several
substances.
536 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.8 Idealized variation with wavelength of spectral normal emissivity of Al2O3.

Turning from the effect of a real surface on emission to its effect on irradiation,
Figure 12.9 shows the processes occurring at an irradiated the surface. If the irradiated
body is transparent, that irradiation (G) may be reflected, absorbed, and transmitted,
and equals the sum of the portions reflected, G , absorbed, G , and transmitted, G :

G G G G . (12.24)

If the medium is opaque to the irradiation (i.e., G = 0), treatment of phenomenon is

greatly simplified and

G G G . (12.25)

As was the case with emissivity, the absorptivity depends on both the wavelength of
the irradiation and the direction of its incidence on the surface, and so several types
of absorptivity can be defined. However, integrating over all wavelengths and all
directions gives the total, hemispherical absorptivity of the surface ( ) as the fraction
of the total irradiation that is absorbed,

G G. (12.26)

Similarly, the total, hemispherical reflectivity of the surface ( ) is defined as the frac-
tion of the total irradiation that is reflected,

G G. (12.27)

If the matter is transparent to the irradiation, the total, hemispherical transmissivity

is defined as the fraction of the total irradiation that is transmitted

G G. (12.28)
Radiation Heat Transfer 537

FIGURE 12.9 Possible fates of surface irradiation.

The sum of Eqs. (12.26), (12.27), and (12.28) gives

1,

or for an opaque medium,

1. (12.29)

These definitions are all for “total, hemispherical” surface properties, which means they
have been averaged over direction and wavelength of incident radiation. Although not
discussed in detail in this text, the use of materials (as for windows) that are transparent
only in certain wavelengths is technologically important. Also the selective nature of
transmissivity of CO2 gas is a good example. It is transparent to visible light (as from
the sun), but much less so in the infrared (as in lower temperature emissions from
Earth’s surface). This phenomenon explains why CO2 is called a “green-house gas.”
Finally, the absorption and reflection processes are usually considered to be sur-
face phenomena, which most often occur only within fractions of a micrometer
beneath the surface of the medium. (Some materials are semi-transparent and irra-
diation is absorbed to a greater depth and is best modeled as having heat generation
through the body and not just as a surface effect.) The irradiation is thus reflected
or absorbed by the surface and the two processes are influenced by the nature of the
surface. Only the absorbed portion of the irradiation transfers the thermal energy or
enthalpy to the medium.

12.5 KIRCHHOFF’S LAW AND THE HOHLRAUM

Figure 12.10 shows an isothermal cavity, enclosed by a blackbody surface at the
temperature, TS, which contains a small body with a real surface. The body emits
radiation at the rate E and absorbs irradiation from the walls of the cavity at the rate
538 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.10 Isothermal cavity with black walls and containing a small body with a less
than ideal surface.

Eb(TS). The net rate of energy transfer from the body is the difference between these
two quantities. At thermal equilibrium the temperature of the body is also Ts, and
hence the net rate of transfer of energy from the body is zero, in which case

E Ts Eb Ts (12.30)

or, with being the total hemispherical emissivity of the body,

Eb Ts Eb Ts

or

(12.31)

that is, the total hemispherical emissivity of a surface within the enclosure is equal
to its total hemispherical absorptivity. Eq. (12.31) is known as Kirchhoff’ s law,
and Eq. (12.26) shows that inasmuch that is always either less than or equal to
unity, E(Ts) cannot be greater than Eb(Ts), which indicates that a blackbody has the
maximum possible emissive power at any temperature. Eq. (12.31) is valid provided
that either the radiation is black or that and are independent of . A surface at
which and are independent of is said to be gray, and a surface for which ,
and , are independent of and is said to be a diffuse-gray surface. The assump-
tion of diffuse-gray behavior greatly simplifies the treatment of radiation exchange
between surfaces.
Radiation Heat Transfer 539

Although a blackbody is hypothetical, its radiation field can be closely approx-

imated by a hohlraum (heated cavity) shown in Figure 12.11. In that figure, the
aperture in the wall of a real isothermal cavity is made small enough (Aaperture
<< Acavity) that radiation entering the cavity through the hole undergoes multiple
reflections from the walls, and only a small fraction of this radiation ever escapes
from the cavity. As a result, any radiation emission in and then escaping from the
cavity is representative of the radiation field within the cavity. Radiant energy
emitted at some location on the cavity wall at the rate Eb undergoes multiple
reflections at other positions on the wall. At the first point of incidence on the
wall, a portion, Eb, is reflected and the remaining portion, Eb, is absorbed.
At the second point of reflection ( Eb) is reflected and, at the point of the nth
reflection, n Eb is reflected. As the points of reflection are randomly located,
the number of reflections that the radiation has undergone before emerging from
the cavity through the hole varies from 1 to infinity, and the emissive power of the
emerging radiation is thus
2 3
E Eb 1 . (12.32)

Writing the sum in parentheses as S gives

2 3
S 1 ,

S 1
S

or

1
S .
1

FIGURE 12.11 Cavity (hohlraum) eventually absorbing all irradiation entering through a
small aperture.
540 An Introduction to Transport Phenomena in Materials Engineering

However, in view of Eqs. (12.29) and (12.31),

1 1
S ,

and hence Eq. (12.32) becomes

E Eb ,

which shows that the hohlraum very closely approximates a blackbody. When one
looks through a small viewport into a larger furnace cavity, the viewport behaves
like a hohlraum and the apparent color of the light seen can be used to estimate the
temperature inside the furnace.

12.6 RADIATION EXCHANGE IN AN ENCLOSURE: VIEW FACTORS

In this treatment of radiation heat transfer during materials processing, we are pri-
marily interested in systems consisting of several surfaces at different temperatures
and radiation exchange among the surfaces is of interest. For example, what fraction
of the radiation emitted by one surface is intercepted by another surface, and as
a result of the radiation exchange, which surfaces are gaining thermal energy and
which are losing energy? Modeling the exchange of radiation between surfaces is
facilitated by introduction of the view factor, which allows calculation of the frac-
tion of the radiation emitted by one surface that is intercepted by another surface.
Figure 12.12 shows the geometrical arrangement of two plane surfaces, one of area
Al at the temperature T1 and the other of area A2 at the temperature T2. The surface
element dAl emits radiation in all directions, and some of this radiation is intercepted
by the surface element dA2. The distance between the two infinitesimal surface ele-
ments is R, and the angles between the straight line joining the elements and the
directions normal to the two surfaces are, respectively, 1 and 2. From Eq. (12.4),
radiation emitted by dA1 is intercepted by dA2 at the rate

dq1 2 I1 cos 1 dA1 d 2 1, (12.33)

in which I1 is the intensity of the radiation leaving surface 1 and 2–1 is the solid angle
subtended by dA2 when viewed from dAl. The area projected by dA2 normal to dAl is
cos 2 dA2 and hence, from Eq. (12.2),

cos 2 dA2
d 2
,
R

substitution of which into Eq. (12.33) gives

cos 1 cos 2
dq1 2 I 2
dA1 dA2 . (12.34)
R
Radiation Heat Transfer 541

FIGURE 12.12 Radiation emitted by surface element, dA1, intercepted by surface element,
dA2.

If surface 1 emits and reflects diffusely, combination of Eqs. (12.18) and (12.34) gives

cos 1 cos 2
dq1 2 J1 dA1 dA2 . (12.35)
R2

Assuming that the radiosity of surface 1, J1, is uniform over the entire surface, the
total rate at which radiation leaves the surface 1 and is intercepted by surface 2 is
obtained by integrating Eq. (12.35) over both total areas (A1 and A2) to get

cos 1 cos 2
q1 2 J1 dA1 dA2 . (12.36)
A2 A1 R2

The total radiation emitted by surface 1 is

A1 J1 , (12.37)

and hence the view factor F12, which is the fraction of the total radiation emitted by
surface 1 that is intercepted by surface 2, is obtained as the ratio of Eq. (12.36) to
Eq. (12.37):

q1 2 1 cos 1 cos 2
F12 dA1 dA2 . (12.38)
A1 J1 A1 A2 A1 R2

Application of the previous derivation to calculation of the fraction of the total radia-
tion emitted by surface 2 that is intercepted by surface 1 gives the view factor F21 as

q2 1 1 cos 1 cos 2
F21 dA1 dA2 . (12.39)
A2 J 2 A2 A2 A1 R2
542 An Introduction to Transport Phenomena in Materials Engineering

Comparison of Eqs. (12.38) and (12.39) shows

A1 F12 A2 F21 . (12.40)

Eq. (12.40) is called the reciprocity relation and is applicable to any pair of radiat-
ing surfaces within an enclosure. We note that the reciprocity relation is strictly a
geometric equation and does not rely on surface properties or temperature.
Within an enclosure formed by N radiating surfaces, all of the radiation emitted
by any one surface is intercepted by the other surfaces and, if the one surface is con-
cave, it intercepts some of its own radiation. Thus, for such an enclosure containing
N surfaces,
N
Fij 1 (12.41)
j 1

applies to each surface. Eq. (9.45) is called the summation rule.

Figure 12.13 shows an enclosure containing three surfaces (N = 3). In this enclo-
sure, N2 (32 = 9) view factors are required, with these being

F11 F12 F13

F21 F22 F23
F31 F32 F33 .

Application of the reciprocity relations gives three equations,

A1 F12 A2 F21
A1 F13 A3 F31
A2 F23 A3 F32 ,

FIGURE 12.13 Enclosure containing three surfaces.

Radiation Heat Transfer 543

and application of the summation rules gives another three equations,

F11 F12 F13 1

F21 F22 F23 1
F31 F32 F33 1.

Thus, only (9 – 6) = 3 view factors need to be determined by equations of the form

given by Eq. (12.39). In general, for an enclosure containing N surfaces, N2 view
factors are required, but as N view factors can be obtained from application of the
summation rule and N(N – 1)/2 can be obtained from the reciprocity relations, only
N2 – N – N(N – 1)/2 = N(N – 1)/2 must be obtained directly.
Consider the simple enclosure shown in Figure 12.14, which comprises a spher-
ical body (surface 1) contained within a spherical cavity (surface 2). With two sur-
faces, four view factors are required. As surface 1 is convex, it does not intercept any
of its own radiation and hence, by inspection,

F11 0

and all of the radiation emitted by surface 1 is intercepted by surface 2, so that by

summation

F11 F12 1.

The reciprocity relation gives

A1 F12 A2 F21 ,

and hence,
A1 A1
F21 F12 .
A2 A2

The summation rule for surface 2 is

F22 F21 1,

which gives

A1
F22 1 .
A2

The general procedure then begins by writing the summation and reciprocity rela-
tions. Any symmetry in the geometry that allows Fij = Fji are noted, as are any view
factors that can be known by simple inspection. Also, analytical expressions for view
factors for several two- and three-dimensional geometries are shown, respectively,
in Table 12.2 and Table 12.3. Such view factors are scattered through the radiation
literature and collected in texts such as [1, 4], and on a public website [5].
544 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.14 Surfaces body contained within a spherical cavity.

TABLE 12.2
View Factors for Two-Dimensional Geometries

F12 F21 1 sin

2

12
1 h h2
F12 1 1
2 w w2
Radiation Heat Transfer 545

TABLE 12.2 (Continued )

A1 A2 A3
F12
2A1

12
2
L L
F12 F21 1
w w

L1 L2 L3 L4
F12
2W1

Example 12.3
View Factors inside a Cylinder
Heat is transferred by radiation inside a cylindrical enclosure of radius R and length
L = 2R shown in Figure 12.15. To model that thermal radiation exchange, we will
need the view factors for the various surface pairs.
 546
TABLE 12.3
View Factors for Three-Dimensional Geometries
12
1 4Z 22
F12 X X2
2 Z12

An Introduction to Transport Phenomena in Materials Engineering

R1 R2 1 Z 22
Z1 Z2 X 1
L L Z12

w h
X and Y
L L
1/ 2
1 X2 1 Y 2 1/ 2 Y
ln Y 1 X2 tan 1
2 2 1/ 2
2 1 X Y 1 X2
F12
XY
1/ 2 X
X 1 Y2 tan 1 X tan 1 X Y tan 1 Y
1/ 2
2
1 Y

L2 L1
H and W
w w
1 1 1/ 2 1
W tan 1 H tan 1 H2 W 2 tan 1
W H 1/ 2
2
H W2
1
F12 W2 H2
W
1 1 W 2 1 H2 W2 1 W2 H2 H2 1 H2 W 2
ln
4 1 W2 H2 1 W2 W2 H2 1 H2 H2 W 2
Radiation Heat Transfer 547

Find: All view factors in the enclosure in Figure 12.15.

Solution: With three surfaces, nine view factors are required,

F11 F12 F13

F21 F22 F23
F31 F32 F33

of which six can be obtained from the reciprocity relations and the summation
rules. As surfaces 1 and 2 in the figure are planar, they cannot see themselves. So,
F11 = F22 = 0 by inspection, and hence only one of the seven remaining view factors
needs to be determined independently. Table 12.3 (a) provides one of these view fac-
tors; here, Z1 = R1/L = 1/2, Z2 = R2/L = 1/2, and X = 6, so

F12 0.172.

In the enclosure, Al = A2 = πR2 and A3 = (2πR)L. From the summation rule,

F13 1 F11 F12 1 0.172 0.828,

and, from reciprocity,

A1 F12 A2 F21 0.172,

because A1 = A2.
From the summation rule,

F23 1 F21 F22 1 0.172 0 0.828.

FIGURE 12.15 Cylindrical enclosure.

548 An Introduction to Transport Phenomena in Materials Engineering

Reciprocity gives

A1 F13 A3 F31

and

A1 R2 1
F31 F13 F13 0.828 0.207.
A3 2 RL 4

By inspection of the symmetry in the geometry,

F31 F32 0.207

and

F33 1 F31 F32 1 2 0.207 0.586.

The final results are:

F11 0 F12 0.172 F13 0.828

F21 0.172 F22 0 F23 0.828
F31 0.207 F32 0.207 F33 0.586.

As the L/R increases, F33, F13, and F23 also increase, and with the exception of F11 and
F22, which remain zero, the other four view factors decrease.

Example 12.4
View Factors inside a Rectangular Six-Surface Enclosure
Heat is transferred by radiation inside the rectangular enclosure shown in Figure 12.16.
To model that thermal radiation exchange, we will need the view factors for the vari-
ous surface pairs.

FIGURE 12.16 Rectangular enclosure with six surfaces.

Radiation Heat Transfer 549

Find: All view factors in the enclosure in Figure 12.16.

Solution: With six surfaces, 36 view factors are required:

F11 F12 F13 F14 F15 F16

F21 F22 F23 F24 F25 F26
F31 F32 F33 F34 F35 F36
F41 F42 F43 F44 F45 F46
F51 F52 F53 F54 F55 F56
F61 F62 F63 F64 F65 F66 .

As all of the surfaces are planar, by inspection we know

F11 F22 F33 F44 F55 F66 0.

From Table 12.3 (b), for opposing rectangular surfaces 1 and 2 (w = 2 m, h = 1 m, and
L = 0.5 m) X = 2/0.5 = 4 and Y = 1/0.5 = 2. Therefore,

F12 F21 0.509.

The view factors are the same by symmetry in the geometry.

Similarly, for surfaces 5 and 6 (w = 1 m, h = 0.5 m, and L = 2 m), X = 1/2 = 0.5
and Y = 0.5/2 = 0.25, and so

F56 F65 0.036.

The last pair of opposing surfaces (3 and 4, where w = 2 m, h = 0.5 m, and L = 1 m)

has X = 2 and Y = 0.5. Their view factors are

F34 F43 0.165.

Table 9.3(c) has the view factors for two flat plates connected at one edge and per-
pendicular to each other. For surfaces 2 and 3, L2 = 0.5 m, L1 = 1 m, and w = 2 m.
Thus, H = 0.5/2 = 0.25 and W = 1/2 = 0.5. Therefore,

F23 0.167

and, by symmetry,

F24 F13 F14 0.167.

For surfaces 2 and 5, w = 1 m, LI = 2 m, and L2 = 0.5 m. Thus, H = 0.5 and W = 4.

Therefore,

F25 0.079
550 An Introduction to Transport Phenomena in Materials Engineering

and, by symmetry,

F45 F46 F36 0.079.

For surfaces 3 and 5, w = 0.5 m, L1 = 2 m, and L2 = 1 m. Thus, H = 1/0.5 = 2 and

W = 2/0.5 = 4. Therefore,

F35 0.084

and, by symmetry,

F45 F46 F36 0.084.

From the reciprocity relation, with Al = A2 = 2 m2, A3 = A4 = 1 m2, and A5 =

A6 = 0.5 m2,

A1
F31 F13 2 0.167 0.334
A3

and, by symmetry,
F32 F41 F42 0.334.

From the reciprocity relation,

A1
F51 F15 4 0.079 0.316
A5

and, by symmetry,
F52 F61 F62 0.316.

From the reciprocity relation,

A3
F53 F35 2 0.084 0.168
A5

and, by symmetry,

F54 F63 F64 0.168.

All the view factors for this example are found in Table 12.4. We will see in later
sections that we do not always need all N2 view factors to solve a specific problem.
We also remind the reader that the N surfaces must always make up an enclosure
with no gaps to the outside world in order to apply the summation relations. We
look to Example 12.5 to see a method to modify an incomplete enclosure.
Radiation Heat Transfer 551

TABLE 12.4
Fij Values
j
i 1 2 3 4 5 6
1 0 0.509 0.167 0.167 0.079 0.079 1.00
2 0.509 0 0.167 0.167 0.079 0.079 1.00
3 0.334 0.334 0 0.165 0.084 0.084 1.00
4 0.334 0.334 0.165 0 0.084 0.084 1.00
5 0.316 0.316 0.168 0.168 0 0.036 1.00
6 0.316 0.316 0.168 0.168 0.036 0 1.00

12.7 RADIATION EXCHANGE AMONG BLACKBODIES

We begin examination of radiation heat exchange among surfaces in an enclosure by
limiting it to black surfaces. We know that these surfaces absorb all incident radia-
tion, and so their radiosity, J, is simply the emissive flux of a blackbody, Eb. (This
assumption will be relaxed in the next section to model more realistic surfaces.)
In Figure 12.17, the rate at which radiation leaves surface 1 and is intercepted by
surface 2, q1 2, is obtained by rearranging Eq. (12.38) as

q1 2 A1 J1 F12 , (12.42)

or, because the surfaces are blackbody radiators,

q1 2 A1 F12 Eb,1 . (12.43)

Similarly, the rate at which radiation leaves surface 2 and is intercepted by surface 1 is

q2 1 A2 F21 Eb,2 . (12.44)

Thus, the net exchange of radiation, q12, is given by

q12 q1 2 q2 1. (12.45)

The net radiation exchange, q12, is thus the net rate at which radiation leaves surface
1 due to its interaction with surface 2, which is also the net rate at which surface 2
receives radiation due to its interaction with surface 1. As blackbodies absorb all
incident radiation,

q12 A1 F12 Eb,1 A2 F21 Eb,2 ,

which, in view of the reciprocity relation, can be written as

q12 A1 F12 Eb,1 Eb , 2 .

552 An Introduction to Transport Phenomena in Materials Engineering

Application of the Stefan–Boltzmann law gives

q12 A1 F12 T14 T24 , (12.46)

where T1 and T2 are the uniform temperatures of surfaces 1 and 2. With N surfaces in
a blackbody enclosure maintained at different temperatures, the net transfer of radia-
tion from surface 1 due to its interactions with the other surfaces is
N N
q1 q1 j A1 F1 j T14 T j4 . (12.47)
j 1 j 1

Although radiation exchange between surface 1 and itself is included in Eq. (12.47),
we note that q11 = 0 because the surface is isothermal.

Example 12.5
Radiation Heat Loss from the Inside of a Heated Crucible
Figure 12.18 shows a cylindrical crucible mostly full of molten copper. This vessel
has a diameter D = 12 cm and depth from the top of the crucible to the liquid metal
surface of L = 3 cm. The crucible is well insulated on the sides and bottom and is
open to the atmosphere at its upper end. The heat is generated in the walls of the cru-
cible and the metal by induction; no heat is lost from the crucible except by radiation
into the enclosure. Above the metal surface, radiation is the only important mode of
heat transfer. Above the metal free surface, that surface, the inner lining of the cru-
cible, and the opening to the environment (a hypothetical surface) form a blackbody
enclosure. The power supplied to the metal leaves the system by radiation and has
sufficient power supplied to maintain the bottom surface (surface 1) at 1250 °C and
the vertical side walls (surface 2) at 1200 °C. The ambient temperature outside the
crucible is 25 °C.

Find: The total power required to maintain surfaces 1 and 2 at the specified
wall temperatures.

FIGURE 12.17 Blackbody radiation leaving surface 1 and intercepted by surface 2.

Radiation Heat Transfer 553

FIGURE 12.18 Well-insulated electrically heated crucible

Solution: As the crucible is well insulated, the total power requirement equals
the heat loss by radiation through the open mouth of the crucible. In order
to make an enclosure, we consider the mouth to be a hypothetical circular
blackbody surface of diameter 12 cm at the ambient temperature 25 °C. The
rate of heat loss from the crucible is thus

q q13 q23

or, from Eq. (12.47),

q A1 F13 T14 T34 A2 F23 T24 T34 .

From Table 12.3 (a), with

1 Z 32
Z1 R L 6 cm 3 cm 2 Z 3 and X 1 2.25,
Z12

then
12
1 2 4 Z 32
F13 X X 0.610.
2 Z12

As surface 1 is planar, F11 = 0, and hence from the summation rule,

F12 1 F11 F13 1 0.610 0.390.

From the reciprocity relation,

2
A1 6 cm
F21 F12 0.390 0.390
9
A2 12 cm 3cm
554 An Introduction to Transport Phenomena in Materials Engineering

and from symmetry,

F21 = F23 0.390.

Thus, the heat loss from the crucible walls out to opening is

q A1 F13 T14 T34 A 2 F23 T24 T34

2 8 4 4
0.06 m 0.610 5.667 10 W/m 2 K 4 1523K 298 K
4 4
0.12 m 0.06 m 0.390 5.67 10 8 W/m 2 K 4 1473K 298 K

2100 W 887 W 2990 W.

(We note here that we only needed to calculate four of the nine view factors to find
the two needed to calculate the heat loss.)
This example illustrates an important aspect of this analysis. The rules only work
for an enclosure; if we do not have one (as in this example), we must make one by
adding fictitious surfaces, like our treatment of the opening. Such hypothetical sur-
faces are treated as black at the temperature of the surrounding environment.
NB: As noted in Chapter 1, before plugging temperature values into models for
radiation heat transfer, they must first be converted from Celsius to Kelvin. Absolute
temperatures are absolutely required for radiation calculations.

12.8 RADIATION EXCHANGE AMONG DIFFUSE-GRAY SURFACES

Treatment of the net exchange of radiation among nonblack surfaces must take into
account the fact that the radiation may experience multiple reflections from and
absorption at all of the surfaces. The various interactions occurring at an opaque
diffuse-gray surface are shown in Figure 12.19.
Radiation leaves the surface at the rate J1A1, and the surface is irradiated at the rate
G1A1. If the radiosity is greater than the irradiation, maintenance of a constant surface
temperature T1 requires that thermal energy be supplied to the surface by some means
other than radiation at the rate q1 such that

q1 A1 J1 G1 . (12.48)

The radiosity of the surface is the sum of its emissive power and the reflected
irradiation,

J1 E1 1G1 , (12.49)

insertion of which into Eq. (12.48) gives

q1 A1 E1 1 1 G1 . (12.50)
Radiation Heat Transfer 555

FIGURE 12.19 Energy balance required for a constant temperature at an irradiated real
surface.

As the surface is opaque,

1 1 1 or 1 1 1,

and hence

q1 A1 E1 1G1 . (12.51)

Alternatively, with a surface emissivity of 1,

E1 1 Eb,1 ,

and from Kirchhoff’s law, Eq. (12.31),

1 1 1 1.

Eq. (12.49) can thus be expressed as

J1 1 Eb,1 1 1 G1

or

J1 1 Eb,1
G1 , (12.52)
1 1

substitution of which into Eq. (12.48) gives

J1 1 Eb,1 Eb,1 J1
q1 A1 J1 . (12.53)
1 1 1 1 / 1 A1

In Eq. (12.53), q1 is the radiative heat flow, (Eb,1 – J1) is the potential for that flow,
and, applying a “radiation Ohm’s law” to this relationship, the term

1 1

1 A1
556 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.20 Exchange of radiation among three diffuse-gray surfaces in an enclosure.

can be thought of as a resistance to the radiation flow at the surface and described as
a surface radiation resistance.
The use of Eq. (12.53) requires that the radiosity J1 be known, but, because an
enclosure with gray surfaces will have multiple reflections, its value is coupled to the
radiosities of the other surfaces in the enclosure. Figure 12.20 shows the exchange
of radiation among three diffuse-gray surfaces in a simple enclosure. The irradiation
received by surface 1 from surfaces 2 and 3 is

A1G1 A2 F21 J 2 A3 F31 J 3 ,

which in view of the reciprocity relations can be written as

A1G1 A1 F12 J 2 A1 F13 J 3

or

G1 F12 J 2 + F13 J 3 .

Substitution of this irradiation into Eq. (12.48) gives

q1 A1 J1 F12 J 2 F13 J 3 . (12.54)

Because the surfaces in the enclosure are planar, Fii = 0 and so the summation rule
gives F12 + F13 = 1, and thus Eq. (12.54) can be expressed as

q1 A1 F12 J1 F13 J1 F12 J 2 F13 J 3 A1 F12 J1 J 2 F13 J1 J 3 . (12.55)

For an enclosure containing N surfaces, Eq. (12.55) becomes

N
q1 A1 F1 j J1 Jj . (12.56)
j 2
Radiation Heat Transfer 557

In Eq. (12.56), q1 is again the net heat flow from surface 1 and J1 – Jj is the potential
for some of that flow between surface 1 and surface j, and, hence from the electric
analogy, the term

1
A1 F1 j

is a resistance to the radiation heat flow between surface 1 and surface j, which is
called the geometric radiation resistance. Combination of Eqs. (12.53) and (12.56)
gives the equality between the net heat flow of surface 1 and the sum of all radiation
exchanges with all surfaces in the enclosure:
N
Eb,1 J1
A1 F1 j J1 J j . (12.57)
1 1 / 1 A1 j 2

With a large number (N) of surfaces in the enclosure, N equations are written from
Eq. (12.57) in the form
a11 J1 + a12 J 2 a1N J N C1
a21 J1 + a22 J 2 a2 N J N C2

aN 1 J1 + aN 2 J 2 aNN J N CN ,

in which the coefficients, aij, and the constants, Ci, are determined by the blackbody
emissive powers, the emissivities, and the areas of the N surfaces. In matrix form, this
set of equations is written as

A J C ,

which, on inversion, gives the radiosities as

1
J A C .

Example 12.6
Radiation Heat Transfer inside an Enclosure with Three Gray Surfaces
Figure 12.21 shows the cross-section of a long duct in which the inner wall condi-
tions are T1 = 600 K, 1 = 0.5, T2 = 800 K, 2 = 0.6, and T3 = 1000 K, 3 = 0.7.

Find: The radiosities of the inner surfaces of the duct and the net heat fluxes
at the walls.
Solution: A “long” duct means that we can simplify the analysis by ignoring
edge effects in the direction normal to the page and treat it as two-dimen-
sional. Eq. (12.57) is written for each of the three surfaces as

Eb,1 J1
A1 F12 J1 J 2 A1 F13 J1 J 3 ,
1 1 / 1 A1
558 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.21 Enclosure with three gray surfaces.

Eb,2 J2
A2 F21 J 2 J1 A2 F23 J 2 J3 ,
1 2 / 2 A2

Eb,3 J3
A3 F31 J 3 J1 A3 F32 J 3 J2 .
1 3 / 3 A3

The areas, per meter length of the duct, are A1 = 3 m2, A2 = 4 m2, and A3 = 5 m2. From
the Stefan–Boltzmann law, Eq. (12.21),
8 4
Eb,1 5.67 10 W / m2 K 4 600K 7350 W / m 2

8 4
Eb,2 5.67 10 W / m2 K 4 800K 23, 200 W / m 2

8 4
Eb,3 5.67 10 W / m2 K 4 1000K 56, 700 W / m 2 .

The needed view factors are from Table 12.2 (c),

A1 A2 A3 3m 2 4m 2 5m 2
F12 0.333,
2A1 2 3m 2

A1 A3 A2 3 m2 5 m2 4 m2
F13 0.667,
2A1 2 3m 2

A2 A3 A1 4 m2 5 m2 3 m2
F23 0.75,
2A2 2 4 m2
Radiation Heat Transfer 559

and from the reciprocity relations,

A1 F12 A2 F21 3 m 2 0.333 1m 2 ,

A1 F13 A3 F31 3 m 2 0.667 2 m2 ,

A2 F23 A3 F32 4 m 2 0.75 3m 2 .

We finally rearrange the first three equations of this example into a set of three alge-
braic equations for the radiosities, Ji.

1 1 1 1 1 1
F12 F13 J1 F12 J 2 F13 J 3 Eb,1
1 1 1

1 2 1 2 1 2
F21 J1 F21 F23 J 2 F23 J 3 Eb,2
2 2 2

1 3 1 3 1 3
F31 J1 F32 J 2 F31 F32 J 3 E b ,3
3 3 3

Using the values found earlier, these equations can be solved for the radiosity
from each surface:

J1 24, 900 W / m 2 , J 2 30, 900 W / m 2 and J 3 48, 240 W / m 2 ,

And, from Eq. (12.53), the net fluxes at the three walls are

7348 W/m 2 24,900 W/m 2

q1 52,700 W,
0.5 / 0.5 3m 2

23, 200 W/m 2 30, 900 W/m 2

q2 46,000 W,
0.4 / 0.6 4 m2

56, 700 W/m 2 48, 200 W/m 2

q3 +98,700 W.
0.3 / 0.7 5m 2

All of the heat is supplied to the enclosure through surface 3 and is drained by sur-
faces 1 and 2. The sum of the net fluxes at the three walls is zero (it is, after all, a
steady state problem).

12.9 NOTES ON THE ELECTRICAL ANALOGY

Figure 12.22 shows two planar, diffuse-gray surfaces that exchange radiation. The
idea of a radiation resistance circuit diagram, arising from application of the electric
analogy to the exchange of radiation, is applied to this system. The diagram contains
560 An Introduction to Transport Phenomena in Materials Engineering

nodes representing the blackbody emissive powers of the surfaces and nodes repre-
senting the radiosities of those same surfaces, and (Eb,1 – Eb,2) is the overall potential
for radiative heat transfer from surface 1 to surface 2. The surface resistances, repre-
senting the effect of nonideality of the surfaces, occur between the Eb and J nodes.
The geometric resistances occur between the J nodes. In Figure 12.22, the resistances
occur in series and hence the total resistance to heat flow from 1 to 2, q12, is the sum
of the individual resistances. The total heat flow through the circuit is determined by
analogy with Ohm’s law as

potential
flow
resistance

or

Eb,1 Eb , 2
q12 . (12.58)
1 1 1 1 2
1 A1 A1 F12 2 A2

For large (infinite) parallel plates in which A1 = A2 = A and by inspection F12 = 1,

Eq. (12.58) becomes

A Eb,1 Eb,2
q12 . (12.59)
1 1
1
1 2

FIGURE 12.22 Radiation resistance circuit diagram for the exchange of radiation between
two planar gray surfaces.
Radiation Heat Transfer 561

FIGURE 12.23 Radiation exchange between diffuse gray (a) concentric cylinders and (b)
concentric spheres.

For the long (infinite) concentric cylinders shown in Figure 12.23 (a), with F12 = 1
and Al/A2 = R1/R2, and Eq. (12.58) becomes

A1 Eb,1 Eb,2
q12 .
1 1 2 R1
1 2 R2

Similarly, for the concentric spheres shown in Figure 12.23 (b), with F12 = 1 and
A1 A2 R12 R22 ,

A1 Eb,1 Eb,2
q12 2
.
1 1 2 R1
1 2 R2

As an example of a slightly more complicated network, Figure 12.24 shows

the circuit diagram for the duct from Figure 12.21 and has been solved in Exam-
ple 12.6.
562 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.24 Radiation resistance circuit diagram for the duct shown in Figure 12.21.

FIGURE 12.25 Gray radiation shield placed between two radiating gray surfaces and the
corresponding radiation resistance circuit diagram.

12.10 RADIATION SHIELDS

Radiation shields are (ideally) highly reflective materials that, when placed between
two radiating surfaces, decrease the net radiation transfer between the original sur-
faces. The radiation resistance circuit diagram for such a shield placed between sur-
faces 1 and 2 in Figure 12.22 is shown in Figure 12.25. As the emissivities of the two
Radiation Heat Transfer 563

surfaces of the radiation shield are not necessarily equal, the emissivity of the surface
facing surface 1 is designated 3(1) and the emissivity of the surface facing surface
2 is designated 3(2). The presence of the radiation shield increases the resistance to
radiation exchange from

1 1 1 1 2
A
1 1 A F
1 12 2 A2
surface geometry surface
1 1 2 2

in Eq. (12.59) to

1 1 1 31 1 32 1 1
1 2
,
1 A1 A1 F13 31 A3 32 A3 A3 F32 2 A2
surface geometry surface surface geometry surface
1 1 3 3(1) 3(2) 3 2 2

by adding the surface resistances of the shield and modifying the geometric resist-
ances. As F13 = F32 = 1 and all areas are equal, the radiation transfer is decreased from
that given by Eq. (12.59) to

A1 T14 T24 A1 T14 T24

q12 . (12.60)
1 1 1 2 1 1 1 1 1 31 1 32
1 2 31 32 1 2 31 32

With small values of 3(2) and 3(2) (very reflective surfaces), the total resistance intro-
duced by the radiation shield can be increased significantly.

Example 12.7
Effect of a Radiation Shield on Heat Flow between Two Surfaces
A shield is placed between two surfaces to reduce the radiative heat flow from one
to the other. The radiation shield has emissivities of 3(1) = 3(2) = 0.1, while the two
surfaces (1 and 2) have T1 = 1000 K, 2 = 0.5 and T2 = 300 K, 2 = 0.5.

Find: Determine the influence of inserting the radiation shield on the heat flow
from surface 1 to surface 2.
Solution: Without the radiation shield, Eq. (12.59) gives

4 4
q12 T14 T24 5.67 10 8 W / m 2 K 4 1000K 300K
A 1/ 1 1/ 2 1 1 / 0.5 1 / 0.5 1
2
18, 700 W / m .
564 An Introduction to Transport Phenomena in Materials Engineering

With the radiation shield in place, Eq. (12.60) gives

q12 T14 T24

A 1 1 1 2 1 1
1 2 31 32

4 4
5.67 10 8 W / m 2 K 4 1000K 300K
2560 W / m 2 .
1 0.5 1 0.5 1 1
0.5 0.5 0.1 0.1

Insertion of the radiation shield thus increases the denominator of the q12 calcula-
tions from 3 to 22 and hence decreases q12 by the factor 22/3 = 7.33. The temperature
of the radiation shield is obtained by considering the radiation exchange between
1 and 3.

q13 T14 T34

;
A 1/ 1 1/ 3(1) 1

4
5.67 10 8 W/m 2 K 4 1000 K T34
2
2560 W/m T3 842 K
1 / 0.5 1 / 0.1 1

Performing the same calculation between the shield and surface 2 yields the same
result.

12.11 RERADIATING SURFACES

A reradiating surface in a radiation enclosure is simply one that is sufficiently well
insulated that it can be considered to be an adiabatic wall at which q = 0. This condi-
tion has the consequence that, from Eq. (12.48), and Eq. (12.53),

Ji Gi Eb,i . (12.61)

Thus, in an enclosure, the equilibrium temperature of a reradiating surface is deter-

mined only by its interaction with the other surfaces in the enclosure and is inde-
pendent of its own emissivity. Figure 12.26 (a) shows a three-surface enclosure
in which surface 3 is reradiating, and Figure 12.26 (b) shows the corresponding
circuit diagram. As q3 = 0, the heat transfer from surface 1 equals the heat transfer
to surface 2:

Eb,1 Eb,2
q1 q2 ,
1 1 1 2
RB
1 A1 2 A2
Radiation Heat Transfer 565

FIGURE 12.26 (a) Three-surface enclosure with one reradiating surface; (b) Radiation
resistance circuit network for (a).

where RB is the total resistance in the two-branch heat flow path between node J1 and
node J2. With

1 1 1
R1 , R2 , and R3
A1 F12 A1 F13 A2 F23
566 An Introduction to Transport Phenomena in Materials Engineering

where

1 1 1
+
RB R1 R2 R3

or

1
RB 1
.
1 1
A1 F12
A1 F13 A2 F23

Using this total resistance of the network, we can find the heat flow in at surface 1
and out at 2:

Eb,1 Eb,2
q1 q2 . (12.62)
1 1 1 1 2
1
1 A1 A1 F12 1 / A1 F13 1 / A2 F23 2 A2

Thus, if T1 and T2 are known and q1 and q2 are obtained from Eq. (12.63), J1 and
J2 are obtained from Eq. (12.53) as

Eb,1 J1 Eb , 2 J2
q1 and q2 .
1 1 / 1 A1 1 2 / 2 A2

Because no heat is lost from the system through the reradiating surface (q3 = 0), the
heat flow from J1 to J3 equals that from J3 to J2:

J1 J3 J3 J2
q13 q32 , (12.63)
1 / A1 F13 1 / A2 F23

which allows JR (= J3) to be determined. The temperature of the reradiating surface is

then determined from Eq. (12.61) as

J3 Eb,3 T34 .

Example 12.8
Radiation Exchange in Three-Surface Enclosure with One Reradiating Surface
Consider that surface 3 in the triangular duct shown in Figure 12.24 is reradiating
(and is referred to as surface R) with 3 = R = 0.1, and that the conditions of the other
surfaces are T1 = 600 K, l = 0.5 and T2 = 800 K, 2 = 0.6.

Find: Calculate the radiosities of the three surfaces and the temperature of the
reradiating surface R.
Radiation Heat Transfer 567

Solution: As in Example 12.6,

A1 F12 1 m 2 , A1 F1R 2m 2 , and A2 F2 R 3m 2 .

From Eq. (12.62),

Eb,1 Eb,2
q1 q2
1 1 1 1
1 A1 A1 F12 1 / A1 F1R 1 / A2 F2 R 2 A2

4 4
5.67 10 8 W / m 2 K 4 600K 800K
16,600
1 0.5 1 1 0.6
2 2 2 2
0.5 3 m 1 m 1 / 2m 1 / 3m 0.6 4 m 2

Using the sign convention in the figures that the arrows for qi is always drawn so that
a positive value means heat flows are into the network. Hence, this result means the
heat flow at surface 1 is actually out of the duct and q2 is into it.
For heat flow through surface resistance at 1, Eq. (12.53) gives

4
Eb,1 J1 5.67 10 8 W / m 2 K 4 6000K J1
q1 16, 600 W .
1 1 / 1 A1 1 0.5 0.5 3m 2

From which we find the radiosity there

J1 12, 900 W / m 2 .

For surface 2, similar calculations show

4
Eb,2 J2 5.67 10 8 W / m 2 K 4 0
800K J2
q2 16,600 W ,
1 2 / 2 A2 1 0.6 0.6 4 m 2

giving

J2 26, 500 W / m 2 .

Then, from Eq. (12.63), the heat flow balance at the reradiating surface is

J1 JR JR J2 12, 900 W / m 2 JR JR 26, 500 W / m 2

0 .
1 / A1 F1R 1 / A2 F2R 0.5 m 2 0.333m 2
568 An Introduction to Transport Phenomena in Materials Engineering

Solving this equation, we obtain JR = 17,400 W/m2= Eb,R = (5.67 × 10–8 W/m2K4)
(TR)4, so
TR 745 K .

12.12 SUMMARY
Thermal radiation, which is part of the electromagnetic spectrum, is emitted over a
range of wavelengths in all directions from the surfaces of all bodies at finite temper-
atures. The intensity of the radiation varies with wavelength and direction of emis-
sion, and hence the spectral and directional distributions of the intensity have to be
taken into consideration. The total intensity and total emissive power are obtained by
integration over all wavelengths and all directions.
A surface can receive radiation from another source, in which case it is said to be
irradiated, and that irradiation is reflected from the surface, absorbed by the body,
and/or transmitted through the body. The absorptivity, reflectivity, and transmissivity
are properties of the irradiated body that are, respectively, the fractions of the inci-
dent radiation absorbed, reflected, and transmitted. Radiosity is all of the radiation
leaving a surface, that is, the sum of its emissive power and the reflected irradiation.
A blackbody has an ideal surface that absorbs all incident irradiation and has the
highest possible values of emissive power at any given T. The spectral distribution of
radiation intensity emitted by a blackbody is given by Planck’s distribution. Integra-
tion of that distribution over all wavelengths gives the Stefan–Boltzmann equation,
which states that the total emissive power of a blackbody is proportional to the fourth
power of its absolute temperature.
The emissivity of a real surface is the ratio of the radiation emitted by the real sur-
face to that emitted by a blackbody at the same temperature. The emissivity of a real
surface varies with wavelength and direction of emission and the total hemispherical
emissivity is obtained by integrating over all wavelengths and all directions. A sur-
face at which the emissivity and absorptivity are both independent of wavelength
is said to be gray, and a surface at which the emissivity and absorptivity are inde-
pendent of the direction of the radiation are said to be diffuse. A surface with both
properties is a diffuse-gray surface.
Consideration of radiation exchange among several surfaces at different temper-
atures requires use of the view factor, which is the fraction of the radiosity from one
surface that is intercepted by another. View factors in an enclosure of several surfaces
are related to one another by the reciprocity relations and the summation rule, and
hence in an enclosure formed by N surfaces only N(N-1)/2 view factors need be
obtained by inspection, symmetry, tabulated formulae, or direct integration.
Treatment of the exchange of radiation among blackbody surfaces in an enclosure
is simplified by the fact that they absorb all and reflect none of incident irradiation
and so the radiosity of the surface is simply its emissive power. The treatment of
radiation exchange among non-blackbody surfaces is different because gray surfaces
may experience multiple reflections from and absorptions at all of the surfaces, in
which case the radiosity of each surface is influenced by the radiosities of all of the
other surfaces in the enclosure. Treatment of radiation exchange among gray surfaces
Radiation Heat Transfer 569

is facilitated by an electric analogy in which the potential for heat flow is differences
in surface radiosities and emissions and the radiation heat transfer balances among
surfaces define surface and geometric resistances.

12.13 HOMEWORK PROBLEMS

(Always begin with a clear sketch of the physical problem. Do not plug in values for
variables until absolutely necessary.)

12.1 Calculate: (a) The emissive power of the radiation emerging from a small
aperture on a large isothermal enclosure that is at the uniform tempera-
ture of 1500 K; (b) the wavelength below which 50% of the emission
occurs; and (c) the maximum spectral emissive power and the wavelength
at which the maximum spectral emissive power occurs.
12.2 A shallow pan of water with bottom and sides insulated is placed outside
on a clear night when the effective sky radiation temperature is 230 K. If
the emissivity of the surface of the water (treated as opaque) is 0.9 and the
convection heat transfer coefficient is 25 W/m2K, what is the minimum air
temperature that can be tolerated without freezing the water?
12.3 An opaque horizontal plate, which is insulated on its backside, receives
irradiation at the rate 3000 W/m2, of which 500 W/m2 is reflected. The
plate is at a temperature of 200 °C and has an emissive power of 500 W/m2.
Air at 25 °C flows over the plate and the convection heat transfer coeffi-
cient is 20 Wm2K. Calculate: (a) the emissivity, absorptivity, and radiosity
of the plate; (b) the net heat transfer rate per unit area.
12.4 Solar irradiation of 1000 W/m2 is incident on the top side of a plate the sur-
face of which has a solar absorptivity of 0.9 and an emissivity of 0.1. The
surrounding air is at 20 °C and the heat transfer coefficient is 15 W/m2K.
If the backside of the plate is insulated, what is the steady-state tempera-
ture of the plate?
12.5 A horizontal opaque surface at a steady-state temperature of 50 °C is
exposed to a flow of air at 23 °C with a convection heat transfer coefficient
of 20 Wm2K. The emissive power of the surface is 60 W/m2, the irradiation
incident on the surface is 1000 W/m2, and the reflectivity of the surface is
0.4. Calculate: (a) the absorptivity of the surface; (b) the net heat transfer
by radiation, and (c) the total heat transfer rate to or from the surface.
12.6 The inner top surface and the vertical side walls of a cubical muffle fur-
nace with black walls and sides of length 2 m are at 500 °C and the bottom
inner surface is at 1000 °C. Calculate the rate of heat transfer by radiation
from (a) the bottom to the top inner surface and (b) from the bottom to the
side walls.
12.7 The flat bottom of a cylindrical crucible, which is fitted with a tight lid, is at
500 °C and the inner surface of the lid and the vertical walls are at 400 °C.
The crucible has an inner diameter of 4 cm and is 4 cm in height. If the
surfaces are black, what is the heat flux by radiation from the bottom sur-
face to (a) the inner surface of the lid and (b) the vertical walls?
570 An Introduction to Transport Phenomena in Materials Engineering

12.8 Air flows in a duct, the inner wall temperature of which is 200 °C. A ther-
mocouple placed in the airflow measures a temperature of 113 °C. Calcu-
late the temperature of the gas if the surface emissivities of the duct wall
and the surface of the thermocouple are, respectively, 0.85 and 0.6, and
the heat transfer coefficient for convection at the surface of the thermo-
couple is 125 W/m2K.
In the following problems (12.9–12.14), the surfaces can be considered to
be diffuse-gray.
12.9 A long heat-treating furnace has a rectangular cross-section 3 m in height
and 4 m in width. The surface of the roof is at 1000 °C, the side walls are
at 800 °C, the floor is at 400 °C, and all the surfaces have an emissivity of
0.8. Calculate the radiant heat transfer to the floor of the furnace per meter
of length.
12.10 An electrically heated wire of radius 1 mm, emissivity 0.5, and surface
temperature 500 °C is surrounded by a thin cylinder of radius 1 cm
that is at 25 °C and has an emissivity of 0.6. (a) Calculate the radiation
heat flux from the wire to the outer cylinder per unit length. (b) A thin
cylindrical radiation shield of radius 5 mm and emissivity 0.1 is placed
between the wire and the outer cylinder. Calculate the new radiation heat
flux from the wire to the outer cylinder. (c) Is the radiation heat transfer
from the wire to the outer cylinder influenced by (i) the radius of the
radiation shield or (ii) the degree of eccentricity of the shield? (d) What
radius of radiation shield is required to decrease the radiation heat flux
occurring without a shield by a factor of four? What is the steady-state
temperature of a shield of this radius? At what radius is the temperature
of the shield 600 K?
12.11 An electrically heated copper wire of 2 cm diameter is surrounded by a
thin cylinder of diameter 5 cm. The annular space between the wire and
the cylinder is evacuated, the wire dissipates 50 W per meter of length, the
surface emissivities of the wire and the cylinder are 0.8, and the steady-
state temperature of the cylinder is 300 K. Calculate the steady-state tem-
perature of the surface of the wire.
12.12 A long duct with an equilateral triangular cross-section and sides of width
1 m has the following surface conditions: side 1 at 600 °C with l = 0.6;
side 2 at 400 °C with 2 = 0.4; and side 3 at 200 °C with 3= 0.2. Calculate
the radiation heat fluxes at the three walls.
12.13 Rework Problem 12.12, but change side 3 to be reradiating. Let the other
two sides have the same surface conditions as Problem 12.12 and calcu-
late: (a) the radiation flux between sides 1 and 2; and (b) the temperature
of the reradiating side.
12.14 An electrically heated furnace enclosure is 2 m long, 1 m wide, and 0.5 m
in height. A 30 kW electric heater embedded in the roof produces a tem-
perature of 700 °C at the inner surface of the roof. The emissivity of the
roof is 0.7 and the emissivity of the floor surface is 0.9. The four vertical
walls are reradiating. Calculate the steady-state temperature of the surface
of the floor. (Ignore any heat transport by convection.)
Radiation Heat Transfer 571

FIGURE 12.27 (a) A long duct (extends into paper); (b) a sphere lying on an infinite plate; (c)
a cylinder with shield over top half for r2/r1 = 1.1 and 5; (d) a load inside a furnace, where the
furnace is 3L x 5L and the load is L x L/4.

12.15 Find numerical values for F12 and F21 for these geometries, using only
reciprocity, summation, and inspection (do not use charts or tables).
12.16 A row of cylindrical heaters at T1 = 1700 K is used to maintain a load
at a temperature, while the top furnace wall is kept at 600 K. At steady
state, there is no net heat flow in and out of the load. The tubes have a
uniform center-to-center spacing (S = 0.2 m) and a uniform diameter
(D = 0.1 m). Treat the system as infinite in the direction into the paper
and infinitely repeating to the left and right (i.e., ignore the effect of
the left- and right-side walls of the furnace). Neglect all modes of heat
transfer inside the furnace except radiation. A good approach to this
problem is to think of the heat flows per tube, instead of the entire
extent of the furnace (which we do not know). (a) Find all nine view
factors. This might help:
1/ 2 1/ 2
2
D D 1 S2 D2
F31 F21 1 1 tan .
S S D2

(b) Assume all the surfaces are black. Draw the resistance network. What is
the heat flux lost to the upper surface? What is temperature of the load? (c)
If the surfaces are gray, draw the resistance network. With an emissivity of
0.7 for the wall and load and 0.9 for the tubes, what is the heat flux lost to
the upper surface? What is the temperature of the load?
572 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.28 Radiation heat transfer from a row of heating tubes to a load and a cooled
wall.

12.17 A polymer coats a flat plate of width wp = 0.2 m and the coating is cured
by radiative heat from a heater (width = ws = 0.3 m) that is L = 0.1 m from
the plate. This configuration is long enough into the plate to ignore end
effects normal to the page and to treat the problem as two-dimensional.
The spaces between heater and plate on the right and left side of the sketch
can be treated as one surface in this problem. (a) Find the nine view fac-
tors among the three surfaces (source, panel, and opening). This view fac-
tor from the heat to the plate should help:
1/2 1/2
2 2
wP L w H /L 4 wP L w H /L 4
FHP .
2 wH L

(b) If the spaces between the plate and heater are open to the environment
at 300 K, what is rate at which heat must be supplied to the heater per unit
length (into the page)? What is the temperature of the panel? (c) If the plate
is at the same temperature as in part (b) and the hypothetical surface in is an
actual, insulated surface, what is rate at which heat must be supplied to the
source per unit length? Compare the rates of heat input in parts (b) and (c).

FIGURE 12.29 Curing a coating on a plate (lower surface) by radiative heating from the
upper surface.
Radiation Heat Transfer 573

12.18 A load in a radiant tube furnace is to be maintained at 1000 K, while

the side walls lose to heat to the environment at a rate that maintains
them at 800 K. Heat is added to the system by combustion inside the tube
(surface 1). Assume that the load (surface 2) is reradiating (there is no
externally applied heat other than radiation exchange among it, the wall
and the tube). Treat the top and sidewalls as one surface (3). The depth of
the furnace into the page is d, and the rest of the lengths are W = 0.5 m,
H = 0.325 m, h = 0.25 m, and r = 0.05 m (radius of tube). (a) Find numeri-
cal values for all nine view factors. View factor between the load (2) and
the tube (1):

2r 1 W
F21 tan
W 2h

(b) If all these surfaces are black, what are T1 and q1/d? (c) If the surfaces
are all gray ( 1 = 3 = 0.8, 2 = 0.5), what are T1 and q1/d? (After drawing the
thermal circuit, it might be helpful to write heat balances at the J1 and J3 nodes
and solve for those radiosities.)

12.19 You are using a hemispherical furnace of diameter D, heated on the flat
surface (surface 1) to T1. Enough heat is removed to the outside through
surface 3, the inside of the hemispherical shell, to maintain it at 700 K.
The load is a sphere (surface 2, with diameter d) suspended a height L
above surface 1. We want the load to be at 1200 K. All surfaces are gray.

1
2 2
1
21 1 1
2

FIGURE 12.30 Radiative heating of a load in a radiant tube furnace.

574 An Introduction to Transport Phenomena in Materials Engineering

FIGURE 12.31 Hemispherical furnace maintaining a load at a specified temperature.

(a) Draw the resistance network for the radiation exchange among these three
surfaces. Find values for the nine view factors. (b) Find q1 and T1, the heat
added to the furnace wall and that heater temperature necessary to maintain
the load temperature.

12.20 Based on the spectral emissive flux for a blackbody, show that Wien’s
displacement law is correct.

NOTES
1. The unit steradian (sr) is a solid angle related to the surface area of a sphere. A sphere has
4π sr in the same way a circle has 2π radians.
2. It should not be confused with a surface that appears black to the human eye because that
surface absorbs all incident radiation in the narrow band of visible light. Such an appar-
ently black surface may not necessarily completely absorb all radiation of all wave-lengths.
3. The calculated (and most used) value of the Stefan–Boltzmann constant is = 5.67 × 10–8
W/m2K4, while the measured value is = 5.73 × 10–8 W/m2K4.

REFERENCES
1. Howell, J. R., M. P. Mengüç, and R. Siegel, Thermal Radiation Heat Transfer, 6th ed.,
CRC Press, 2015.
2. Pivovnsky, M., and M. R. Nagel, Tables of Blackbody Radiation Functions, The Mac-
millan Company, 1961.
3. Wiebelt, J. A., Engineering Radiation Heat Transfer, Holt, Reinhart, and Winston, Inc.,
1966.
4. Modest, M. F., Radiative Heat Transfer, 3rd ed., Academic Press, 2013.
5. Howell, J. R., A Catalog of Radiation Heat Transfer Configuration Factors, 3rd ed.,
www.thermalradiation.net/indexCat.html, retrieved online April 1, 2023.
Appendix I
MATH PRACTICE FOR TRANSPORT PHENOMENA COURSE
The problems given in this file are representative of the mathematics knowledge
and skills required of a student taking a course in transport phenomena. Our course
is usually taken by fifth-semester materials engineering undergraduates who have
completed three semesters of calculus plus courses in linear algebra and differential
equations. A selection of problems like these are given as homework in the first
week of the course taught at Purdue, followed by an in-class quiz. The goal of such a
problem set is to call attention to the tools they will need during the semester so they
won’t trip up on them later.

1. Find dy/dx, where c and n are constants:

(a) y = xn
(b) y = c
(c) y = g(x)h(x)
(d) y = g(x)/h(x)
(e) y = t2 + 1/t2
(f) y = (1 + x + x2 + 3x3)-1
(g) y = 5x3 – 2x2 + 1 (evaluate at x = 2)
2. Integrate the following expressions for z:
(a) z = x2 + 3x1/2 + 2 with respect to x
(b) z = y x4 + cos(3x) + A with respect to x
(c) z = 7x3/4 + 3x with respect to x between x = 0 and x = 2
(d) z = cos[A(1-x)] with respect to x
(e) z = A/x
(f) z = A exp(-ax) + Bx1 2 with respect to x; evaluate the integral from
x = 0 to x = 2 for a =1
(g) z = x exp(2x) dx with respect to x (by parts)
3. (a) Evaluate I by integration by parts. Leave it in terms of u and v.
1
du
I v dy
dy
0

(b) Insert the u and v functions into the result from (a) and obtain a
numerical value for I.

v y u y y1/2 1

(c) Insert the u and v functions into the result from (a) and obtain a
numerical value for I.

v y3 u y 2 y1/2 1

575
576 Appendix I: Math Practice for Transport Phenomena Course
Appendix I: Math Practice for Transport Phenomena Course 577
578 Appendix I: Math Practice for Transport Phenomena Course
Appendix II
EQUATIONS OF MOTION AND THERMAL ENERGY BALANCE
Thermophysical properties are assumed to be constant and uniform.

equations oF Motion in cartesian coordinates

Velocity vector: V uiˆ vjˆ wkˆ where iˆ, ˆj , and kˆ are unit vectors.
Continuity:

u v w
0
x y z

Navier–Stokes equations:
x direction:

u u u u p u u
u v w gx
t x y z x x x y z z z
local advective buoyancy pressurre viscous friction
acceleration

y direction:

v v v v p v v
u v w gy
t x y z y x x y z z z

z direction:

w w w w p w w
u v w gz
t x y z z x x y z z z

EQUATIONS OF MOTION IN CYLINDRICAL COORDINATES

Velocity vector: V urˆ v ˆ wkˆ where rˆ, ˆ and kˆ are unit vectors.
Continuity:

1 1
ru v w 0
r r r z

579
580 Appendix II: Equations of Motion and Thermal Energy Balance

Navier–Stokes equations:
r direction:
u u v u v2 u p
u w gr
t r r r z r
2 2
1 1 u 2 v u
ru 2 2 2
r r r r r z2

direction:
v v v v uv v 1 p
u w g
t r r r z r
2 2
1 1 v 2 u v
rv 2 2 2 2
r r r r r z

z direction:
2 2
w w v w w p 1 w 1 w w
u w gz r 2 2
t r r z z r r r r z2

THERMAL ENERGY EQUATION IN CARTESIAN COORDINATES

2 2 2
T T T T T T T q
u v w 2 2 2
t x y z x y z c c
storage advection conduction dissipation heat gen

2 2 2 2 2 2
u v u w v w u v w
2
y x z x z y x y z

THERMAL ENERGY EQUATION IN CYLINDRICAL COORDINATES

2 2
T T v T T 1 T 1 T T q
u w r 2 2 2
t r r z r r r r z c c

2 2 2
1 u v u w v 1 w
r
r r r z r z r
2 2 2
u 1 v w
2 u
r r z
Appendix III
UNIT CONVERSIONS
acceleration 1 m/s2 = 32.8 ft/s2
area 1 m2 = 10.8 ft2
density 1 kg/m3 = 0.0624 lbm/ft2
1 g/cm3 = 1000 kg/m3
energy 1J = 0.000948 Btu
1J = 0.737 ft lbf
force 1N = 0.225 lbf
heat flux 1 W/m2 = 0.317 Btu/(hr ft2)
heat generation rate 1 W/m3 = 0.0966 Btu/(hr ft3)
heat transfer coefficient 1 W/m2K = 0.176 Btu/(hr ft2 oF)
heat transfer rate 1W = 3.41 Btu/hr
kinematic viscosity 1 m2/s = 10.8 ft2/s
latent heat 1 J/kg = 0.000430 Btu/lbm
length 1m = 3.28 ft
1 cm = 0.394 in
1 km = 0.622 mile
mass 1 kg = 2.20 lbm
mass flow rate 1 kg/s = 2.20 lbm/s
power 1W = 0.00132 horsepower
1W = 3.41 Btu/hr
pressure 1 Pa = 1 N/m2
1 torr = 133.3 Pa
1 lbf/in2 (psi) = 6895 Pa
1 atm = 101,325 Pa
1 atm = 14.70 lbf/in2 (psi)
1 bar = 100,000 Pa
specific heat 1 J/kgK = 0.000239 Btu/(lbm oF)
thermal conductivity 1 W/mK = 0.578 Btu/(hr ft oF)
thermal diffusivity 1 m2/s = 10.8 ft2/s
thermal resistance 1 K/W = 0.527 (hr oF)/Btu
velocity 1 m/s = 3.28 ft/s
viscosity (dynamic) 1 Pa s = 0.672 lbm/(ft s)
1 Pa s = 1 kg/ms
1 poise = 0.1 Pa s
viscosity (kinematic) 1 m2/s = 0.0932 ft2/s
1 m2/s = 1000 stokes
volume 1 m3 = 35.3 ft3
1 m3 = 262 US gal
1 m3 = 1000 liter
volume flow rate 1 ft3/s = 0.02832 m3/s
1 US gal/min = 6.309 × 10–5 m3/s

581
Appendix IV
SELECTED THERMOPHYSICAL PROPERTIES

TABLE IV.A
Properties of Solid Metals
k (W/mK)
Material M.P. (K) (kg/m3) c (J/kgK) (m2/s) 0oC 200oC 400oC 800oC 1000oC
Aluminum 933 2702 902 9.7 × 10–5 236 238 228
Beryllium 1550 1850 1809 6.3 × 10–5 218 144 118 83 69
Boron 2573 2500 1090 1.1 × 10–5 31.7 18.3 10.0 5.9 4.8
Cadmium 594 8650 231 5.2 × 10–5 104 99
Chromium 2118 7160 451 2.9 × 10–5 95 86 77 64 62
Cobalt 1765 8862 419 2.7 × 10–5 104 77
Copper
Pure 1356 8933 385 11.6 × 10–5 401 389 378 352 336
Al Bronze (10%Al) 1293 8800 420 1.4 × 10–5 49 54 61
Brass (30%Zn) 1188 8526 382 3.4 × 10–5 110 142 151
Constantan 1493 8921 410 6.3 × 10–6 22 26
(45%Ni)
Germanium 1211 5360 318 3.6 × 10–5 67 40 19
Gold 1336 19300 129 12.7 × 10–5 318 309 299 273 254
Iron
Pure 1810 7870 440 2.3 × 10–5 83 66 53 32 30
Wrought 7850 460 1.6 × 10–5 59 52 45
Armco 7870 447 2.1 × 10–5 75 62 49 29 29
Cast (4% C) 7272 420 1.7 × 10–5 52 40 34 21
Steels
1% carbon 7801 473 1.2 × 10–5 43 42 36 28
1% chrome 7913 448 1.7 × 10–5 52 52 42 33
304 stainless 7900 477 4.0 × 10–6 114 18 21 26
(18% Cr, 8% Ní)
Lead 601 11340 129 2.4 × 10–5 36 33
Lithium 454 534 3560 4.0 × 10–5 79
Magnesium 923 1740 1017 8.8 × 10–5 157 151 147
Manganese 1517 7430 477 2.2 × 10–5 7.7
Molybdenum 2883 255 5.3 5.3 × 10–5 139 131 123 109 103
Nickel
Pure 1726 8900 442 2.3 × 10–5 94 74 65 73 78
Nichrome 1672 8360 430 3.5 × 10–6 12 15 18
(20%Cr)
Platinum
Pure 2042 21450 130 2.6 × 10–5 72 72 74 80 84
60% Pt, 40% Rh 1800 16600 162 1.7 × 10–5 46 54 62 70 74
(Continued)

583
584 Appendix IV: Selected Thermophysical Properties

TABLE IV.A (Continued )

k (W/mK)
Material M.P. (K) (kg/m3) c (J/kgK) (m2/s) 0oC 200oC 400oC 800oC 1000oC
Rhenium 3453 21100 138 1.7 × 10–5 49 45 44 45 46
Rhodium 2233 12450 247 49 × 10–5 151 141 132 119 113
Silicon 1685 2330 691 9.5 × 10–5 168 93 59 30 25
Silver 1234 10500 235 17.3 × 10–5 428 415 399 368 355
Sodium 371 971 1220 11.2 × 10–5 135
Tin 505 5750 227 5.1 × 10–5 68 60
Titanium 1953 4500 510 9.6 × 10–6 22 20 19 21 23
Tungsten 3653 19300 133 6.9 × 10–5 182 152 134 118 114
Uranium 1406 19070 116 1.2 × 10–5 27 30 35 45
Vanadium 2192 6100 486 1.1 × 10–5 31 33 35 40 44
Zinc 693 7140 388 4.4 122 112 102
Zirconium 2125 6570 280 1.2 × 10–5 23 21 21 24 26

Source: Y. S. Toulukian et al., Thermophysical Properties of Matter, 13 volumes, IFI/Plenum, 1970–1977.

TABLE IV.B
Properties of Non-metallic Solids
k (W/mK)
Material M.P. (K) (kg/m ) c (J/kgK)
3 (m2/s) 0oC 200oC 400oC 800oC 1000oC
Aluminum oxide 2323 3970 765 1.3 × 10–5 40 22 13 7.3 6.2
Asbestos 383 816 3.6 × 10–6 0.11
Beryllium oxide 2725 3000 1030 9.2 × 10–5 302 159 93 41 30
Bricks
Common 1600 840 5.2 × 10–7 0.7
Chrome 3000 840 8.7 × 10–7 2.2 2.3 2.4 2.1
Fireclay 2000 960 5.2 × 10–7 1.0 1.0 1.0 1.1 1.1
Magnesite 1130 4.0 3.6 2.8 2.1 1.8
Masonry 1700 837 4.6 × 10–7 0.66
Silica 1900 1.1
Carbon 4073 1950 1.6 1.9 2.2 2.6 2.8
Cement mortar 1860 780 6.2 × 10–7 0.9
Concrete 2300 880 4.9 × 10–7 1
Coal 1370 1260 1.4 × 10–7 0.24
Diamonds
Type I 1000.0 300.0
Type Ila 3500 510 1.4 × 10–3 2650.0 1300.0
Type IIb 1510.0 780.0
Earths
Clay 1500 80 1.1 × 10–6 1.4
Diatomaceous 1.3
Sand 1500 800 2.5 × 10–7 0.3
Appendix IV: Selected Thermophysical Properties 585

TABLE IV.B (Continued )

k (W/mK)
Material M.P. (K) (kg/m3) c (J/kgK) (m2/s) 0oC 200oC 400oC 800oC 1000oC
Glasses
Pyroceram 2600 810 1.9 × 10–6 4.1 3.6 3.2 29.0 2.8
Window 2700 800 3.9 × 10–7 0.84
Ice 920 2000 1.2 × 10–6 2.2
Insulations
Cork, granular 45–120 1900 2–5 × 10–7 0.045
Corkboard 160 1900 1.4 × 10–7 0.043
Cellulose, loose 45 0.038
Feltboard 50–125 0.035
Glass fiber 220 0.035
Glass wool 40 700 1.4 × 10-6 0.038
Kapok 0.035
Magnesia, 85% 270 0.065 0.085
Polystyrene 50 0.025
Rock wool 160 0.040
Rubber, foam 70 0.030
Sawdust 0.059
Vermiculite, 80 0.058
loose
Magnesium oxide 53 29 18 8.8 7.3
Plaster, gypsum 1600 1000 3.8 × 10–7 0.5
Quartz, fused 1.3 1.6 1.9 3.2 3.8
Rocks
Granite 2640 800 1.4 × 10–6 3.0
Limestone 2400 860 1.0 × 10-6 2.0
Marble 2650 1000 1.0 × 10-6 2.7
Sandstone 2200 740 1.7 × 10–6 2.8
Rubber, hard 1170 2000 6.8 × 10–8 0.16
Skin, human 0.37
Snow
Loose 110 0.05
Packed 500 0.19
Teflon 2200 0.35
Woods
Balsa 140 0.55
Cypress 460 0.1
Fir 420 2700 9.7 × 10–8 0.11
Maple or oak 600 2400 1.2 × 10–7 0.17
Pine, white 440 0.11
Pine, yellow 640 2800 8.4 × 10–7 0.15
Plywood 550 1200 1.8 × 10–7 0.12
Wool 200 0.04

Source: Y. S. Toulukian et al., Thermophysical Properties of Matter, 13 volumes, IFI/Plenum, 1970–1977.

586 Appendix IV: Selected Thermophysical Properties

TABLE IV.C
Properties of Gases at 1 Atm Pressure
T (K) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr g / 2 (1/Km3)
Air
200 1.766 1003 0.0181 1.34 × 10–5 0.740 85700 × 104
300 1.177 1005 0.0261 1.85 × 10–5 0.712 13300 × 104
400 0.883 1013 0.0331 2.29 × 10–5 0.703 3630 × 104
500 0.706 1029 0.0395 2.68 × 10–5 0.699 1360 × 104
600 0.589 1051 0.0456 3.03 × 10–5 0.698 616 × 104
700 0.504 1075 0.0513 3.35 × 10–5 0.702 318 × 104
800 0.441 1099 0.0569 3.64 × 10–5 0.704 180 × 104
900 0.392 1120 0.0625 3.92 × 10–5 0.705 109 × 104
1000 0.353 1141 0.0672 4.18 × 10–5 0.709 70 × 104
1200 0.294 1175 0.0759 4.65 × 10–5 0.720 33 × 104
1400 0.252 1201 0.0835 5.09 × 10–5 0.732 17.2 × 104
1600 0.221 1240 0.0904 5.49 × 10–5 0.753 9.9 × 104
1800 0.196 1276 0.0970 5.87 × 10–5 0.772 6.1 × 104
2000 0.177 1327 0.1032 6.23 × 10–5 0.801 3.9 × 104
Argon
200 2.435 523.6 0.0124 1.60 × 10–5 0.674 113000 × 104
300 1.623 521.6 0.0177 2.27 × 10–5 0.669 16700 × 104
400 1.218 521.0 0.0223 .85 × 10–5 0.665 4480 × 104
500 0.974 520.8 0.0264 3.37 × 10–5 0.664 1640 × 104
600 0.812 520.6 0.0301 3.83 × 10–5 0.662 730 × 104
700 0.696 520.6 0.0336 4.25 × 10–5 0.658 375 × 104
800 0.609 520.5 0.0369 4.64 × 10–5 0.655 211 × 104
900 0.541 520.5 0.0398 5.01 × 10–5 0.654 127 × 104
1000 0.487 520.5 0.0427 5.35 × 10–5 0.652 81 × 104
1200 0.406 520.5 0.0481 5.99 × 10–5 0.648 38 × 104
1400 0.348 520.4 0.0535 6.56 × 10–5 0.638 19.7 × 104
1600 0.304 520.4 0.0588 7.10 × 10–5 0.628 11.3 × 104
1800 0.271 520.4 0.0641 7.60 × 10–5 0.617 6.9 × 104
2000 0.244 320.4 0.0692 8.07 × 10–5 0.607 4.5 × 104
CO2
200 2.683 759 0.0095 1.02 × 10–5 0.814 338000 × 104
300 1.789 852 0.0166 1.50 × 10–5 0.68 46500 × 104
400 1.341 939 0.0244 1.94 × 10–5 0.747 16900 × 104
500 1.073 1017 0.0323 2.35 × 10–5 0.740 4100 × 104
600 0.894 1077 0.0403 2.72 × 10–5 0.727 1770 × 104
700 0.767 1126 0.0487 3.06 × 10–5 0.708 880 × 104
800 0.671 1169 0.0560 3.39 × 10–5 0.708 480 × 104
900 0.596 1205 0.0621 3.69 × 10–5 0.716 284 × 104
1000 0.537 1235 0.0680 3.97 × 10–5 0.721 179 × 104
1200 0.447 1283 0.0780 4.49 × 10–5 0.739 81 × 104
1400 0.383 1315 0.0867 4.97 × 10–5 0.754 42 × 104
1600 0.304 520.4 0.0588 7.10 × 10–5 0.628 11.3 × 104
Appendix IV: Selected Thermophysical Properties 587

TABLE IV.C (Continued)

T (K) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr g / 2 (1/Km3)
1800 0.271 520.4 0.0641 7.60 × 10–5 0.617 6.9 × 104
2000 0.244 320.4 0.0692 8.07 × 10–5 0.607 4.5 × 104
CO
200 1.708 1045 0.0175 1.27 × 10–5 0.763 88100 × 104
300 1.138 1051 0.0252 1.78 × 10–5 0.743 13400 × 104
400 0.854 1060 0.0323 2.21 × 10–5 0.727 3650 × 104
500 0.683 1071 0.0386 2.60 × 10–5 0.720 1360 × 104
600 0.569 1084 0.0444 2.94 × 10–5 0.718 613 × 104
700 0.488 1099 0.0497 3.25 × 10–5 0.718 316 × 104
800 0.427 1114 0.0549 3.54 × 10–5 0.718 178 × 104
900 0.379 1128 0.0596 3.81 × 10–5 0.721 108 × 104
1000 0.342 1142 0.0644 4.06 × 10–5 0.720 69 × 104
1100 0.31 1155 0.0692 4.30 × 10–5 0.718 47 × 104
1200 0.285 1168 0.0738 4.53 × 10–5 0.717 32 × 104
Helium
200 0.244 5197 0.115 1.50 × 10–5 0.676 1320 × 104
300 0.1627 5197 0.15 1.99 × 10–5 0.690 219 × 104
400 0.122 5197 0.18 2.43 × 10–5 0.703 61.9 × 104
500 0.0976 5197 0.211 2.83 × 10–5 0.695 23.4 × 104
600 0.0813 5197 0.247 3.20 × 10–5 0.673 10.6 × 104
700 0.0697 5197 0.278 3.55 × 10–5 0.663 5.41 × 104
800 0.061 5197 0.307 3.88 × 10–5 0.657 3.02 × 104
900 0.0542 5197 0.335 4.20 × 10–5 0.652 1.82 × 104
1000 0.0488 5197 0.363 4.50 × 10–5 0.645 1.15 × 104
1200 0.0407 5197 0.416 5.08 × 10–5 0.635 5240
1400 0.0349 5197 0.469 5.61 × 10–5 0.622 2700
1600 0.0305 5197 0.521 6.10 × 10–5 0.608 1530
1800 0.0271 5197 0.57 6.57 × 10–5 0.599 930
2000 0.0244 5197 0.62 7.00 × 10–5 0.587 595
Hydrogen
200 0.1229 13,540 0.128 0.68 × 10–5 0.717 1610 × 104
300 0.0819 14,320 0.182 0.89 × 10–5 0.705 275 × 104
400 0.0614 14,480 0.221 1.09 × 10–5 0.714 77.8 × 104
500 0.0492 14,510 0.256 1.27 × 10–5 0.721 29.2 × 104
600 0.0410 14,540 0.291 1.45 × 10–5 0.724 13.0 × 104
700 0.0351 14,610 0.325 1.61 × 10–5 0.724 6.66 × 104
800 0.0307 14,710 0.360 1.77 × 10–5 0.723 3.69 × 104
900 0.0273 14,840 0.394 1.92 × 10–5 0.723 2.2 × 104
1000 0.0246 14,990 0.428 2.07 × 10–5 0.724 1.38 × 104
1200 0.0205 15,370 0.495 2.36 × 10–5 0.733 6150
Nitrogen
200 1.708 1043 0.0183 1.29 × 10–5 0.734 86500 × 104
300 1.139 1040 0.0260 1.79 × 10–5 0.715 6500 × 104
400 0.854 1045 0.0325 2.21 × 10–5 0.710 3650 × 104
(Continued)
588 Appendix IV: Selected Thermophysical Properties

TABLE IV.C (Continued )

T (K) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr g / 2 (1/Km3)
500 0.683 1057 0.0387 2.59 × 10 –5 0.708 1370 × 104
600 0.569 1075 0.0441 2.93 × 10–5 0.713 620 × 104
700 0.488 1098 0.0493 3.24 × 10–5 0.720 319 × 104
800 0.427 1122 0.0541 3.52 × 10–5 0.730 181 × 104
900 0.380 1146 0.0587 3.79 × 10–5 0.739 110 × 104
1000 0.342 1168 0.0631 4.04 × 10–5 0.747 70 × 104
1200 0.285 1205 0.0713 4.50 × 10–5 0.761 33 × 104
1400 0.244 1233 0.0797 4.92 × 10–5 0.761 17 × 104
Oxygen
200 1.951 906 0.0182 1.47 × 10–5 0.728 87000 × 104
300 1.301 920 0.0267 2.07 × 10–5 0.711 12900 × 104
400 0.976 942 0.0342 2.59 × 10–5 0.713 3480 × 104
500 0.780 971 0.0412 3.05 × 10–5 0.720 1280 × 104
600 0.650 1003 0.0480 3.47 × 10–5 0.725 574 × 104
700 0.557 1032 0.0544 3.85 × 10–5 0.730 294 × 104
800 0.488 1054 0.0603 4.21 × 10–5 0.736 165 × 104
900 0.434 1074 0.0661 4.54 × 10–5 0.738 99 × 104
1000 0.390 1091 0.0717 4.85 × 10–5 0.738 63 × 104
1200 0.325 1116 0.0821 5.42 × 10–5 0.737 29 × 104
1400 0.278 1136 0.0921 5.95 × 10–5 0.734 15 × 104
Water Vapor
300 0.0253 2041 0.0181 0.91 × 10–5 1.03 25 × 104
400 0.555 2000 0.0264 1.32 × 10–5 1.00 4330 × 104
500 0.441 1977 0.0357 1.33 × 10–5 0.96 1280 × 104
600 0.367 2022 0.0464 2.13 × 10–5 0.93 480 × 104
700 0.314 2083 0.0572 2.54 × 10–5 0.93 214 × 104
800 0.275 2148 0.0686 2.95 × 10–5 0.92 106 × 104
900 0.244 2217 0.078 3.36 × 10–5 0.95 58 × 104
1000 0.22 2288 0.087 3.76 × 10–5 0.99 33 × 104

Source: Y. S. Toulukian et al., Thermophysical Properties of Matter, 13 volumes, IFI/Plenum, 1970–1977.

TABLE IV.D
Properties of Saturated Liquids
T (K) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr g / 2 (1/Km3)
Ethyl Glycol
273 1131 2295 0.254 65.1 × 10–3 588 0.0192 × 108
293 1117 2386 0.257 21.4 × 10–3 199 0.173 × 108
313 1101 2476 0.259 9.57 × 10–3 91 0.844 × 108
333 1088 2565 0.262 5.17 × 10–3 51 2.82 × 108
353 1078 2656 0.265 3.21 × 10–3 32 7.18 × 108
373 1059 2750 0.267 2.15 × 10–3 22 15.5 × 108
Appendix IV: Selected Thermophysical Properties 589

TABLE IV.D (Continued )

T (K) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr g / 2 (1/Km3)
Glycerin
253 1288 2143 0.282 134 1020 × 103 0.42
273 1276 2261 0.284 12.1 96 × 103 50
293 1264 2386 0.287 1.49 12.4 × 103 3200
313 1252 2513 0.290 0.27 2.3 × 103 101000
Water
273.2 1000 4205 0.564 1.79 × 10–3 13.4 -21 × 107
280 1000 4197 0.582 1.44 × 10–3 10.4 22 × 107
300 997 4177 0.608 0.857 × 10–3 5.88 366 × 107
320 989 4176 0.637 0.579 × 10–3 3.79 1250 × 107
340 980 4187 0.659 0.423 × 10–3 2.69 2980 × 107
360 967 4204 0.674 0.320 × 10–3 2.00 6250 × 107
373.2 958 4220 0.681 0.282 × 10–3 1.75 8500 × 107
400 937 4241 0.686 0.219 × 10–3 1.35 16100 × 107
450 890 4419 0.673 0.153 × 10–3 1.01 40200 × 107
500 832 4647 0.635 0.118 × 10–3 0.86 77100 × 107
550 756 5272 0.571 0.095 × 10–3 0.88 144000 × 107
600 650 6691 0.481 0.076 × 10–3 1.05 295000 × 107

Source: Y. S. Toulukian et al., Thermophysical Properties of Matter, 13 volumes, IFI/Plenum, 1970–1977.

TABLE IV.E
Properties of Liquid Metals
T (oC) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr
Al 660 a 2385 1080 94 0.00125 0.0144
700 2374 1080 95.4 0.00115 0.0130
800 2346 1080 98.7 0.00095 0.0104
Cu 1083a 8000 495 165.6 0.00450 0.0135
1100 7986 495 166.1 0.00435 0.0130
1200 7906 495 170.1 0.00363 0.0106
Fe 1536a 7015 795 = 0.00580 0.055
Hg −39a 13,600 142 6.78 0.00202 0.0423
0 13,500 142 7.61 0.00168 0.0314
20 13,460 139 8.03 0.00156 0.0270
Li 181a 525 4370 46.2 0.00064 0.0601
200 521 4360 47.2 0.00060 0.0553
400 484 4220 53.8 0.00039 0.0308
Mg 650a 1590 1360 78 0.00130 0.0226
700 1577 1360 81 0.00106 0.0179
800 1551 1360 88 0.00075 0.0126
Ni 1454a 7905 620 15.4 0.00549 -

(Continued)
590 Appendix IV: Selected Thermophysical Properties

TABLE IV.E (Continued)

T (oC) (kg/m3) c (J/kgK) k (W/mK) μ (kg/ms) Pr
Pb 327 a 10,650 152 16.6 0.00262 0.0240
400 10,590 144 18.2 0.00217 0.0172
Si 1410a 2510 1.04 - 0.00094 -
1500 2481 1.04 - - -
Ti 1665a 4240 200 - 0.00520 -
U 1133a 17,900 161 - 0.000608 -
1200 17,810 161 - 0.000614 -
Zn 419a 5800 481 49.5 0.008 0.0260
500 5800 481 54.1 - -
600 5800 481 59.9 - -

Source: amelting temperature

Brandes, E. A. (ed.), Smithells Metals Reference Book, 6th ed., Butterworths & Co. Ltd., 1983.
Index
A concentration, 487
conduction, 68
absorptivity, 536 conservation of mass, 285
acceleration, local, 292 convection, 409
advective, 292 momentum, 293
advection, 27, 405 convection, 22, 27, 405
advection-diffusion equation, 412 forced, 27, 408
age-hardened aluminum, 1 heat transfer, 4, 27, 405
Apollo 13, 407 mass transfer, 7, 512
approximate solutions, 41 natural, 27, 460
Archimedes’ principle, 244 curing Bakelite, 147
average heat transfer coefficient
external flow, 429
internal flow, 453 D
Darcy drag, 339
B Darcy friction factor, 256
Darken’s analysis, 209
Bernoulli equation, 254 decarburization, 206
modified, 255 deposited metal. droplet, 135
Bingham plastic, 334 dezincing lead, 322
blackbody radiation, 30, 31, 528 diffusion and chemical reaction, 164
boiling curve, 470 doping silicon, 7, 216
boiling heat transfer, 6, 469 drag coefficient, 390
Boltzmann-Matano analysis, 216 bluff shapes, 396
Boltzmann’s constant, 529 cylinder, 390
boundary condition equiaxed dendrite, 395
thermal, 71 gas bubbles, 397
velocity, 296 liquid droplets, 397
boundary layer solid sphere, 392
concentration, 512 drunkard’s walk, 34
forced convection, 420 dynamic viscosity, 13
hydrodynamic, 363
natural convection, 463
E
Boussinesq approximation, 461
buoyancy, 244 electromagnetic spectrum, 523
emissive flux, 526
C emissivity, 532
energy cascade, 382
carburization, 193 energy integral equation
chemical vapor deposition, 7 conduction, 133, 147
closure problem, 381 forced convection, 427
concentration equation, 489 entry length, hydrodynamic, 373
multicomponent, 91 thermal, 458
steady state, 55 evaporative cooling, 517
transient, 120 exact solutions, 41
conduction, 22, 23 external flow, 360
radial system, 144 extrusion, 331
constitutive equations, 10, 39
continuity equation, 288
F
continuous casting, 5
continuum approximation, 11 Fanning friction factor, 256
control volume, 2, 67 Fick’s first law, 36, 187, 188, 484
591
592 Index

Fick’s second law, 190 J

film model, 497
firing ceramic, 6 Jakob number, 472
five o’clock solutions, 48
float glass process, 297 K
flow separation, 391
fluidized beds, 349 Kirchoff’s Law
forced convection conduction, 63
cylinder, 442 radiation, 537
flat plate, 420 Knudsen number, 297
heat transfer coefficient, 424, 425, 429
impingement, 446 L
Nusselt number, 424, 425, 429
ladle emptying, 276
other geometries, 443
laminar flow, 251
sphere, 446
latent heating, 160
form drag, 390
Lewis number, 515
Fourier’s law, 23, 55
log mean temperature difference, 454
free stream turbulence, 384
lumped capacitance
friction drag, 390
convection heat loss, 122
friction factor
radiation heat loss, 126
absolute roughness, 259
flat plate, 351
relative roughness, 259 M
mass diffusion, 7, 33, 187
G mean temperature, 451
model, 40
Galileo number, 350 transient, 273
Grashof number, 466 flow in launder, 270
gray body, 30, 31, 554 steady-state, 263
modified Bernoulli equation, 255
H friction loss factor, 262
momentum equation, 462
heat generation, 85 Moody chart, 260
heat transfer coefficient, definition, 27, 440 MSE triangle, 1
external flows, 408, 429, 435, 439
internal flows, 455, 457, 458
natural convection, 463, 467 N
hohlraum, 539 natural convection, 465
Hooke’s law, 10 isothermal flat plate, 460
Navier-Stokes equations, 291, 295
I Newtonian fluid, 31
pressure-driven flow, 311
infinite diffusion couple, 195 shear-driven flow, 11, 13, 309
injection molding, 4, 141 Newton’s law of cooling, 27
integral analysis, 131 non-Newtonian fluids, 17
interdiffusion coefficient, 211 non-Newtonian models, 21, 22
composition dependence, 216 no-slip condition, 12, 296
temperature dependence, 221 nucleate boiling, 471
interface conductance, 63 numerical solutions, 43
interface resistance, 60 Nusselt number, 408, 440
internal flow, 307
internal flow forced convection, 451
O
inviscid flow, 363
iron blast furnace, 347 Ohm’s law, 55
irradiation, 523, 527 onset of nucleate boiling, 471
Index 593

P solidification, 170
general case, 166
Peclet number, 416 thermally resistant mold, 165
Planck’s constant, 529 pure substance, 160
Planck’s distribution, 30, 529 solid-solid phase change, 197
porous media, 337 stagnant film, 502
power-law fluid, 329 Stefan-Boltzmann law, 529
pressure-driven flow, 329 Stefan condition, 164
shear-driven flow, 21, 328 streamfunction, 289
Prandtl number, 413 streamline, 289
pressure, 234, 241
stream tube, 289
absolute, 237
sublimation, 496
atmospheric, 232
summation rule, 543
gauge, 238
surface reaction, 500
hydrostatic, 238
measurement, 235
T
Q Tape casting, 314
tempering glass, 4, 157
quenching, 124, 127
thermal arrest, 160
thermal conductivity, 24, 82
R thermal energy equation, 70, 412
thermal resistance, 54
radiation, 6, 22, 29, 59, 523
radiation exchange, 554 radiation, 59
black bodies, 524 convection, 58
diffuse grey, 551 radial conduction, 55
radiation resistance slab conduction, 55
geometric, 557 thermal resistance networks, 63
surface, 556 time-averaged velocity, 377
radiation shield, 562 Tollmein-Schlichting waves, 384
radiosity, 528 Torricelli barometer, 239
Rayleigh number, 464 transition to turbulence, 383
reciprocity relation, 542 transport phenomena, 1
reflectivity, 536 turbulence, 251, 376
reradiating surface, 564 characteristics, 375
Reynolds averaged Navier-Stokes (RANS), 380 types of flow, 376
Reynolds number, 252 turbulent spots, 384
Reynolds stresses, 381 two-phase boundary
rheology, 18, 53 composition, 164
rising inclusions, 394 temperature, 200

S U
sand casting, 169 units, 50
scaling analysis, 45 U-tube manometer, 239
Schmidt number, 515
self-diffusion coefficient, 213 V
semi-infinite body, 130
sensible heating, 160 velocity jump, 297
shape factors, 100, 101 velocity of light, 523
shear thickening, 18 view factors, radiation, 540
shear thinning, 18
Sherwood number, 513 W
shot peening, 3
sintering, 6 Wien’s displacement law, 529