Solid Mechanics and Its Applications PDF
Solid Mechanics and Its Applications PDF
Solid Mechanics and Its Applications PDF
Wolfgang H. Müller
An
Expedition to
Continuum
Theory
Solid Mechanics and Its Applications
Volume 210
Series editor
G. M. L. Gladwell, Waterloo, Canada
The fundamental questions arising in mechanics are: Why?, How?, and How much?
The aim of this series is to provide lucid accounts written by authoritative researchers
giving vision and insight in answering these questions on the subject of mechanics as
it relates to solids.
The scope of the series covers the entire spectrum of solid mechanics. Thus it
includes the foundation of mechanics; variational formulations; computational
mechanics; statics, kinematics and dynamics of rigid and elastic bodies: vibrations
of solids and structures; dynamical systems and chaos; the theories of elasticity,
plasticity and viscoelasticity; composite materials; rods, beams, shells and
membranes; structural control and stability; soils, rocks and geomechanics;
fracture; tribology; experimental mechanics; biomechanics and machine design.
The median level of presentation is to the first year graduate student. Some texts
are monographs defining the current state of the field; others are accessible to final
year undergraduates; but essentially the emphasis is on readability and clarity.
Wolfgang H. Müller
An Expedition to Continuum
Theory
123
Wolfgang H. Müller
Institute of Mechanics
Technical University of Berlin
Berlin
Germany
Just to make it clear from the very start: This is not a monograph for the specialists
who are looking for a compendium on continuum theory. Rather I attempt to fill a
niche with this book, which started to widen after our government in all their
infinite wisdom decided to terminate the diploma syllabi at German Universities
and replace them by bachelor and master modules instead:
True students of engineering have a notorious aversion to mathematical
abstraction. They prefer a quick and dirty way of solution, in particular when
modeling the behavior of advanced materials in complex technical systems. The
usual, highly formalized textbooks on continuum theory do not really support such
an approach. Even their last resorts, for example consulting the user manuals of the
all-time-favorite finite element codes, are blocked, because the same cryptic sym-
bols are lurking there. On the other hand, students of physics face another problem,
which is due to the way physics is commonly taught. They hear much about discrete
systems, in particular in mechanics and in thermodynamics. The concept of fields is
usually not presented before they attend classes on electrodynamics or quantum
mechanics. Finally, the education of both groups of students has in common that
usually no difference is made between the laws of nature (the balances of mass,
momentum, energy, etc.) and constitutive equations. Both are usually well mixed to
form a hodge-podge of recipes. The best examples are the NAVIER–STOKES equations.
Moreover, every subject of physics, i.e., mechanics, thermodynamics, electrody-
namics, etc., is usually taught separately without emphasizing the connections and
the similarities. This is definitely not what we need when developing modern
technologies, which only thrive because of their multiphysics interaction.
This is where continuum theory can help. It provides a bridge between the
various subjects, by working out a common structure and by emphasizing the
common roots. In this context constitutive equations form a most essential joint.
This is where this book sets in. It is the result of two teaching modules of four
contact hours per week each. These modules are currently taught at TU Berlin, in
particular in the master course Physical Engineering Science. The exercises
compiled in this book play an important role in the teaching: Approximately two
of the four hours per week are reserved for a seminar, where each student presents
one of the various problems. Moreover, further problems are worked out in written
form every week. Thus the students study the subject matter on a continuous basis,
throughout the semester. They are forced to learn for their future job and not just
vii
viii Preface
for a final examination, after which they may forget everything immediately.
Another side effect of this hard way of learning is to make sure that the students do
not study singular aspects of continuum theory and ignore the rest. And, finally,
the students learn from each other so that collectivism can be good for something
after all.
Another important aspect of this book is to educate students in tensor calculus
both in abstract as well as in index notation. Future engineers and physicists must
be able to perform calculations in their daily practice, which is why the latter
method is particularly important.
Finally, I would like to thank all those who helped me with this book. First of
all, Prof. Ingo Müller and Priv.-Doz. Dr. Wolf Weiss, who contributed to various
sections that were not included in the German edition of this book. Further credits
are due to Messrs. B. Emek Abali, M.Sc., Dipl.-Ing. Benjamin Schmorl, and Felix
Reich, M.Sc. The latter contributed in particular Exercises 13.2.3, 13.4.1, and
13.4.2. Several students found typos in the book: Messrs. Heinrich Grümmer,
Wilhelm Hübner, Andre Klunker, Christian Seidel, and Oliver Stahn. Finally,
I would like to thank Springer Engineering in the Netherlands, in particular
Ms. Nathalie Jacobs and Ms. Cynthia Feenstra.
And now we start
An Expedition to
Continuum Theory
ΔS ≥ 0
∇ ⋅σ = − ρ f
1 Prologue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...... 1
1.1 Some Remarks Regarding the Purpose of this Book . ...... 1
1.2 A Reminder of Scalars, Vectors, and Tensors. . . . . . ...... 6
1.2.1 Example I: Glass Fiber Optics—Pressing
of a Spherical Lens into a Bushing . . . . . . . ...... 7
1.2.2 Example II: Composites—Thermal Stresses
Around Fibers . . . . . . . . . . . . . . . . . . . . . ...... 10
1.2.3 Example III: Cracks . . . . . . . . . . . . . . . . . ...... 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...... 14
2 Coordinate Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 A Personal Remark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 First Definitions and Notions in Index Notation . . . . . . . . . . 16
2.3 Vector Interpretation of the Metric . . . . . . . . . . . . . . . . . . . 23
2.4 Co- and Contravariant Components. . . . . . . . . . . . . . . . . . . 27
2.5 Co- and Contravariant from the Perspective of Vectors . . . . . 34
2.6 Physical Components of Vectors and Tensors . . . . . . . . . . . 37
2.7 A Touch of Differential Geometry . . . . . . . . . . . . . . . . . . . 42
2.8 Would You Like to Know More? . . . . . . . . . . . . . . . . . . . . 45
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
ix
x Contents
Abstract This introductory chapter explains the scope and structure of this book
and why it may be useful for the reader. Some basic notions and concepts of
continuum theory are introduced. As a motivation for studying continuum theory
several engineering problems are presented, which will eventually be solved in the
following chapters.
‘‘What!? Yet another book on continuum theory?,’’ one may ask. ‘‘How is this one
different from the established literature?’’ In this context we must first of all explain
what this book does not want to be: It certainly does not cover all there is to know in
continuum theory, not even rudimentarily. Rather the intention is to give a first
impression of what continuum theory can do for the engineer who has to solve
technical problems involving solids, fluids, or gases. It is by no means a compen-
dium for a specialist or an advanced student. Its clientele are beginners instead. The
various fields of application of continuum theory will be illustrated to suit their
needs, and we will explain to them how continuum theory ‘‘works,’’ at least in
principle and, to a certain degree, also in detail. One of our intentions is to create
awareness for certain notions, such as universal balance equations in contrast to
material-specific constitutive equations and, by making this distinction, show how
relatively simple problems for solids, fluids, and gases can be solved analytically.
Robert HOOKE was born on July 18, 1635 in Freshwater on the Isle of
Wight. He died on March 3, 1703 in London. He was a contemporary of
NEWTON. However, other than Sir Isaac, he was interested in physics and
in biology. His experiments with springs led him to more or less verbal
forms of the law named after him and which he summarized in an
anagram as follows: CEIIINOSSSTTUU. Deciphered this means ‘‘ut
tensio sic vis,’’ what we may translate as ‘‘like the distension so the
force.’’ (painting by Rita GREER, Oxford University)
Claude Louis Marie Henri NAVIER was born on February 10, 1785 in
Dijon and died on August 21, 1836 in Paris. In 1819 he became a
professor of mechanics at the École des Ponts et Chaussées. In 1831 he
succeeded CAUCHY at the École Polytechnique. He worked in various
fields relevant to mechanical engineering, such as elasticity theory and
fluid mechanics. Being an ‘‘offspring’’ of FOURIER he contributed much
to the theory of FOURIER series and suggested an equation for fluid
friction, the NAVIER-STOKES-model.
George Gabriel STOKES was born on August 13, 1819 in Skreen, County
Sligo, Ireland and died on February 1, 1903 in Cambridge, England.
After a detour to Bristol College he starts studying mathematics in
Cambridge at Pembroke College. He graduates in 1841 in the Mathe-
matical Tripos Examination with highest honors as a Senior Wrangler. In
1849 STOKES becomes one of the successors to NEWTON’s famous Luca-
sian Chair at the University of Cambridge.
Pierre Louis DULONG was born on February 12, 1785 in Rouen and died
on July 19, 1838 in Paris. He was a French physico-chemist. He also
studied at the famous École Polytechnique in Paris and became a
professor at the school in 1820. Moreover, he became a member of the
Académie des Sciences in 1823 and assumed the position of a secretary
in 1832. Chemists live dangerously: While working with nitrogen-tri-
chloride DULONG tragically lost his eye and three fingers.
Alexis Thérèse PETIT was born on October 2, 1791 in Vesoul and died on
June 21, 1820 in Paris. He was a French physicist and studied, just like
DULONG, at the École Polytechnique where he became successor to Pierre-
Simon LAPLACE. He discovered ‘‘his law’’ in 1819 together with DULONG.
Moreover, being a French patriot, he was also a strong supporter of the wavy
nature of light and strongly opposed NEWTON’s belief of a corpuscular
theory.
4 1 Prologue
Chapter 8 deals with a relatively abstract question: Do balance laws and con-
stitutive equations keep their form if transferred from an observer at rest to an
arbitrarily moving one? In mathematical terms this problem can be analyzed by
using so-called EUCLIDEAN transformations. We will touch upon an almost philo-
sophical principle according to which a genuine law of nature must keep its form,
independently of the frame of reference.
The geometrician EUCLID was born around the year 360 B.C., presumably
in Athens. He died around 280 B.C., presumably in Alexandria. His great
heritage is a set of textbooks, the so-called ‘‘elements’’ in which he
compiled the knowledge about mathematics of his times. The most
famous anecdote about EUCLID tells us about his encounter with the
pharaoh PTOLEMY I who, being a lazy politician, asked him if there was a
shortcut to geometry other than the tedious lore of the ‘‘elements.’’ EUCLID
answered laconically that there is no royal road to knowledge.
Carl Henry ECKART was born on May 4, 1902 in St. Louis, Missouri. He died
on October 23, 1973 in La Jolla, California. He was a multi-talented
American scientist who contributed to theoretical physics as well to geology
and oceanography. For example, in quantum mechanics he provided proof
that HEISENBERG’s and SCHRÖDINGER’s point of view are equivalent to each
other. He can also be considered to be one of the fathers of irreversible and
relativistic thermodynamics. However, on top of that he was a competent
administrator of academic affairs which shows that, surprisingly, science and business can mix
after all. (Photo public domain, credit SIO Archives/UCSD)
In the final Chap. 13 we shall learn how the realm of continuum physics can be
extended to electromagnetic fields. The emphasis is on a rational presentation of
fundamental principles: What are the foundations of MAXWELL’s equations, how
can the occurring fields be measured, at least in principle, and how are they linked
to each other? Moreover, the question regarding the frame indifference of the
equations and the transformation properties of the electromagnetic fields will be
posed, which had already been answered before in context with the thermo-
mechanical fields. This will lead us to the beginnings of relativistic field theories.
At the very end of the chapter we will present simple constitutive equations and
couple electro-magnetic to mechanical phenomena.
In order to consolidate the acquired knowledge many exercises have been
added to the text. In part they pick up and examine statements of the surrounding
text in detail because they had not been proven before. However, some of them
also require deeper thinking since their solution requires a broader understanding
of the subject matter. Clearly, the reader will gain maximum profit by solving all
of the exercises. However, a more superficial reading is also possible by recog-
nizing the meaning and relevance of the given solution. Moreover, at the end of
each chapter an overview to additional literature on the current topics is presented
together with a preliminary discussion of each reference: Would you like to know
more?
Finally, for the entertainment, relaxation, and inspiration of the reader pictures
and interesting details of the life of each scientist who was mentioned in the text
have been added.
In what follows we present some motivation of what makes learning continuum
theory worth while for the engineer.
Jean Baptiste Joseph Baron de FOURIER was born on March 21, 1768 in
Auxerre (Bourgogne, France). He dies on May 16, 1830 in Paris. After
having survived the chaos of the French revolution (luckily FOURIER was not
born as a nobleman) FOURIER starts as a student at the École Normale in
Paris. His teacher, the great LAGRANGE, considers him among the top sci-
entists of Europe at that time. It is, therefore, not surprising that he becomes
successor to LAGRANGE’s chair for analysis and mechanics at the École
Polytechnique in 1797. He is also one of the scientists who accompanied Napoleon during his
military campaign to Egypt. This acquaintance certainly helped getting him the title of a baron.
6 1 Prologue
Many fields of continuum theory are no scalars and cannot simply be described by
a single ‘‘number.’’ Rather they have a ‘‘direction’’ and show an ‘‘orientation.’’ For
their quantification more than just one number is required. Such quantities are
known as vectors or, even more general, as tensors. Typical vectors mentioned in a
beginner’s class on engineering mechanics are the position, x, its time derivatives,
the velocity, t, and the acceleration, a, or the force vector, F, as well as the specific
force,1 f. Other mechanics-related vectors are the angular velocity, x, or the
torque, M. It is intuitively clear that the latter vectors are associated with some
rotation or, in other words, they have both a direction and an orientation—in
contrast to the vectors previously mentioned which had only a direction. There-
fore, such objects are called axial or polar vectors. Moreover, in thermodynamics
we encounter the heat flux, q, and the temperature gradient, grad T. And in
electrical engineering use is made of the electric field, E, the dielectric dis-
placement, D, the polarization, P, or the magnetic fields, H and B. As in mechanics
some of these vectors are of a polar nature. However, due to lack of intuition and
everyday experience it is hard to tell which ones. Finally, we should also think of
vectors related to the geometry of a body, such as the unit surface normal, n, the
directed surface element, dA, the unit tangent, s, or the directed line element, ds. In
a more general manner of speech we may also refer to vectors as first order tensors.
Scalar quantities, like mass density, q, temperature, T, or the electric potential, U,
may analogously be called tensors of zeroth order. We shall later add some pre-
cision to this nomenclature, which for the time being is purely heuristic.
Well known mechanical tensors of second order, or ‘‘tensors’’ for short, are the
stress,r, the linear strain, e, or the velocity gradient, grad t. In HOOKE’s law for
anisotropic materials we even encounter a tensor of fourth order—the stiffness
matrix, C. Constitutive equations of electromagnetism even use third order tensors
with a physical meaning, for example the piezo-electric matrix of coefficients,
e. However, as we shall see, tensors can also be used to shed a different light on a
well-established mathematical concept like the vector-product. As we shall see it
can be rewritten in terms of an axial third order tensor, the LEVI-CIVITA-symbol.
Note that whenever we wish to emphasize the absolute vector or tensor char-
acter of a physical quantity, which is independent of a coordinate system, the
coordinate base or, even more general, the observer, we denote the corresponding
quantity by a bold letter, e.g., by F in the case of a force. This is quite customary in
the contemporary scientific literature and we have already followed this custom in
the text above. Other symbolic ways of writing is to underline2 the symbol, e.g., by
writing F, or to use an arrow, i.e., ~ F. However, for the solution of a specific
engineering problem it is in most cases necessary to refer to a suitable set of
1
The adjective ‘‘specific’’ refers to a quantity per unit mass. An example for a specific force is
the well-known gravitational acceleration, g.
2
Second order tensors are sometimes underlined twice.
1.2 A Reminder of Scalars, Vectors, and Tensors 7
coordinates, i.e., to specify the corresponding vector base and to represent all
physical quantities in this very base by means of components. Therefore in this
book the reader will, first, be introduced to important notions and concepts of
continuum theory and, second, learn the appropriate engineering mathematics for
the solution of technical problems. The following examples were taken from daily
engineering practice and illustrate the importance of a deeper understanding of
how to evaluate vector and tensor relations in arbitrary coordinate systems.
For an optical link a spherical lens ([ 1 mm) made of glass or sapphire is pressed
into a cylindrical bushing of a slightly smaller diameter ([ 0.995 mm) (cf.,
Fig. 1.1). This leads to building up a pressure along the equators of the sphere and
the corresponding deformation will lead to tensile stresses in its interior.
However, brittle materials such as glass or sapphire react extremely sensitive to
tensile stresses. Even if these stresses are not high enough to crush the sphere
immediately, the phenomenon of subcritical crack growth will lead to failure
eventually: Under the influence of tensile stress water vapor, which is always
present in the environment, will preferably diffuse to the tips of microcracks in the
glass. There it will loosen the atomic bonds and stepwise increase the crack length.
In the very moment where the first microcrack reaches a critical size, the sphere of
glass will fail. Indeed, it was shown experimentally that fracture starts in the
equatorial region of the sphere. This is where one would intuitively expect the
highest tensile stresses: Fig. 1.2. The engineer must now provide an answer to the
sphere of glass
metal bushing
1.000 mm
0.995 mm
8 1 Prologue
following question: How is the misfit in diameters or, in other words, how is the
clearing tolerance related to an acceptable malfunction rate in view of the lifetime
warranty of the whole system?
In order to answer this question the stress distribution in the sphere must be
known first. Basically it can be calculated as follows. In a first step a suitable
differential equation for the stresses needs to be established. As we shall explain in
more detail below this equation can be obtained from the static balance of
momentum:
div r ¼ 0 ð1:2:1Þ
The symbol ‘‘div’’ denotes a differential operator known as divergence. We
shall see later that it entails a certain differentiation with respect to position. Eq.
(1.2.1) now needs to be complemented by a suitable constitutive law for the stress
tensor. For brittle matter HOOKE’s law may serve as a suitable model. It connects
the stress r and the strain e linearly via the stiffness tensor C:
r ¼ C e: ð1:2:2Þ
The symbol ‘‘ ’’ stands for a double scalar product. We shall see later what its
precise meaning is. At this point it suffices to say that Eq. (1.2.2) is rather general
and holds for arbitrary anisotropic, linear elastic materials. In the case of an
isotropic linear-elastic material the above-mentioned equation simplifies consid-
erably: The ominous double scalar product after the stiffness matrix can be
evaluated and rewritten in terms of the LAMÉ parameters k and l:
r ¼ ktrðeÞ1þ2le: ð1:2:3Þ
1.2 A Reminder of Scalars, Vectors, and Tensors 9
Gabriel LAMÉ was born on July 22, 1795 in Tours (France). He died on
May 1, 1870 in Paris. In 1813 he enters the École Polytechnique in
Paris and graduates from that school in 1817. This is followed by
further studies at the famous École des Mines, which he finishes with
another degree in 1820. In the same year LAMÉ moves to Russia and
becomes a professor and engineer at the Institut et Corps du Génie des
Voies de Communication in St. Petersburg. In 1832 he returns to Paris,
founds an engineering firm together with French colleagues and finally
accepts a chair in physics at his first alma mater.
In this equation ‘‘1’’ denotes the unit tensor and ‘‘tr’’ is an operator known as the
trace of a tensor. We shall see later that the LAMÉ parameters, k and l, are related
to Young’s modulus, E, and to the shear modulus, G, as follows:
GðE 2GÞ
k¼ ; l ¼ G: ð1:2:4Þ
3G E
If the stress r in Eq. (1.2.1) is substituted by (1.2.3) a set of partial differential
equations of second order in space for the displacement vector, u, results. This is
because the strain tensor, e, contains first order derivatives in u. For a concrete
solution it is necessary to specify boundary conditions. For example: The normal
stress in the equatorial region of the sphere must be equal to the pressure, p0 , due
to the misfit. Moreover, the stresses in the interior of the sphere must remain finite.
Thomas YOUNG was born on June 13, 1773 in Milverton, Somersetshire and
died on May 10, 1829 in London. In Anglo-Saxon countries his name is
inseparably connected to the modulus of elasticity. It is highly illuminating
to read YOUNG’s own words when he introduces the modulus: ‘‘The modulus
of elasticity of any substance is a column of the same substance, capable of
producing a pressure on its base which is to the weight causing a certain
degree of compression as the length of the substance is to the diminution of
its length.’’ His commissioner—the British Admiralty—responded immedi-
ately: ‘‘Though science is much respected by their Lordships and your paper is much esteemed,
it is too learned … in short it is not understood.’’ (cited after [2])
Thus the problem is well-posed and a mathematician could turn to the proof of a
unique solution. However, this would be insufficient for the engineer. For an
explicit calculation of the stresses it is first of all necessary to choose a suitable
coordinate system. In the present case spherical coordinates r, u, # would do
nicely. The coordinate system is placed in the center of the sphere of glass (cf.,
Fig. 1.3). The letter r denotes the radial distance in the sphere of glass; u and #
refer to the so-called azimuthal and polar angles, respectively.
Before a solution can be obtained, Eqs. (1.2.1) and (1.2.2) need to be rewritten
in spherical coordinates. This requires two things. First, one must be able to
express physical quantities, such as stress, strain, and displacements correctly.
Second, it is required to perform derivatives in spherical coordinates, for example
10 1 Prologue
ϑ
r
x2
ϕ
x1
to specify the abstract operation ‘‘div,’’ as well as the relation between strain and
displacement which, as indicated above, also contains a spatial derivative. How to
do this will be described in Chaps. 2–4.
Adrien Marie LEGENDRE was born on September 18, 1752 in Paris where he
also died on January 10, 1833. His life falls into the heydays of the French
Revolution. In fact he strongly supported the new ideals that originated during
those days of change. For example, he was engaged in the definition of
rational units independent of the weight or the length of the currently acting
monarch. In 1791 he becomes a member of the corresponding committees in
the French Academy and starts producing logarithmic tables in 1792. Note
that he did not do this all by himself. At certain times he had more than 80 (!)
assistants, which clearly shows that revolutions can be good for something.
Materials reinforced with (for example) carbon or silicon carbide fibers are used
more and more frequently in advanced lightweight engineering constructions.
Fig. 1.4 shows a typical hexagonal unit cell of such a composite material. A
potential reliability problem originates from the thermal stresses in these materials,
which are due to the Coefficients of Thermal Expansion (CTEs) of the various
matrices and fibers that are being used and which can be considerably different.
The resulting stresses (also known as thermal eigenstresses) can be considerable
1.2 A Reminder of Scalars, Vectors, and Tensors 11
SiC-fibers
metal matrix
y
α matrix > αfiber
Fig. 1.5 Left generating radial cracks in fiber-matrix composition, e.g., by thermal stresses; right
optical micrograph of radial cracks generated by a lithium niobate-lithium disilicate double
crystal in a glass matrix after Serbena and Zanzotto [3]
and will eventually lead to cracking of the material. For example, if the CTE of the
matrix is greater than the CTE of the fiber the matrix will shrink upon the fiber
during cooling and tensile stresses will be generated in the matrix perpendicularly
to the radial direction (cf., Fig. 1.5, left). These stresses can lead to the formation
and growth of radial cracks (see Fig. 1.5, right). On the other hand, if the CTE of
the matrix is smaller, radial stresses along the circumference of the fiber will form
and the matrix will debond from the fiber (see Fig. 1.6).
In order to quantify the thermal stresses the compound is first idealized by the
following system shown in Fig. 1.7: A cylinder ‘‘1’’ (the fiber) is inserted in a
hollow cylinder ‘‘2’’ (the matrix). Both cylinders have different LAMÉ parameters
k1, l1 and k2, l2, respectively. Their CTEs are also different, a1 and a2. An
interaction between the various fibers of the composite is neglected in this simple
model.3
3
A more careful investigation shows that this assumption is reasonable up to a fiber volume of
40 %.
12 1 Prologue
Fig. 1.6 Left detachment of a fiber from the matrix by thermal stresses; right Interfacial
debonding at a monazite/fiber interface after Chawla et al. [1]
2 R1
2 R2
As in the previous case the computation of the stresses relies upon the balance
of momentum of elastostatics, Eq. (1.2.1). In the present case it is suitably eval-
uated in polar coordinatesr, #, and z. As a constitutive equation we will once more
use HOOKE’s law, although it needs to be extended to cover thermal expansion.
According to DUHAMEL and NEUMANN we write:
r ¼ ð3k þ 2lÞa½T TR 1 þ ktrðeÞ1þ2le: ð1:2:5Þ
In this equation T and TR denote the current and the reference temperature,
respectively. The latter can be interpreted as the manufacturing temperature at
which the composite is typically free of stress.
Note that we have to distinguish between two regions filled with different
materials and a borderline at which both materials meet. Clearly, there must be a
transition between the mathematical solutions in both regions. More specifically,
we proceed as follows. The differential equations resulting by combination of Eqs.
(1.2.1) and (1.2.5) are first solved for a hollow cylinder. We shall see that in the
resulting expressions two constants of integration will occur. This solution is then
1.2 A Reminder of Scalars, Vectors, and Tensors 13
applied to each of the two regions separately so that four constants of integration
result. These are determined from suitable boundary conditions. One can argue
from first principles that the radial stresses at all boundaries must be continuous as
well as finite in all points of the continuum. Moreover, the displacement at the
inner borderline must be continuous, at least as long as the fiber and the matrix
does not debond.
Moreover, the same remarks as in the case of spherical coordinates hold, which
were made in context with the sphere of glass in a bushing. More explicitly, in
order to obtain an explicit solution it is now necessary to represent operators like
the divergence as well as all occurring tensors in cylindrical coordinates.
In the previous examples the notion of cracks has been mentioned already. From a
mechanics point of view we may, more generally, talk about stress concentrators
instead. The first two-dimensional mathematical model of such a concentrator and,
in the limit, of a sharp crack in a brittle solid was suggested and detailed by the
Englishman A.A. GRIFFITH.
14 1 Prologue
x2 σ x2
σ
x1 x1
-c +c
σ
σ
Alan Arnold GRIFFITH was born on June 13, 1893 in London and
died on October 13, 1963. He graduated from the University of
Liverpool. From his early days on he was very much interested in
avionics and the problems of material fatigue and fracture encoun-
tered therein. In 1920 he publishes his famous paper on brittle
fracture which already contains the fracture criterion that later car-
ried his name. However, it contained a faulty factor, which Griffith
corrected soon—but without providing details. It took a long time
until in the sixties another paper was published which proved his
correction.
References
Capt. Terrell: Chekov, are you sure these are the correct
coordinates?
Chekov: Captain, this is the garden spot of Ceti Alpha Six!
Typically scientists working in continuum theory split into two extremely hostile
ideological factions: the supporters of symbolic tensor calculus and the friends of
index notation. The former emphasize the absolute character laws of nature and
constitutive equations should have. In other words these laws should be stated
independently of an observer and, therefore, independently of a coordinate system.
As to whether this is possible, at least in principle, or, in other words, as to whether
the tensorial relations that force the laws of nature and constitutive relations into a
We will initially start from an index point-of-view that circumvents the notion of a
(unit) vector base. In this spirit we consider a three-dimensional Cartesian coor-
dinate grid consisting of three straight lines, x1, x2, x3, which are orthogonal to
each other. For brevity we shall denote them by xk and note that a Latin case index
runs from 1 to 3. Moreover, we consider other three-dimensional coordinate lines
which may be curvilinear and denote them by zi. Clearly points in space should be
identifiable by either set of lines. In other words invertible relations of the fol-
lowing kind must hold:
zi ¼ zi ðx1 ; x2 ; x3 Þ zi ðxk Þ and xk ¼ xk z1 ; z2 ; z3 xk zi : ð2:2:1Þ
Mathematically speaking such invertible relations are also known as isomor-
phisms. The distance s between two points (1) and (2) with the corresponding
ð1Þ ð2Þ
Cartesian coordinates xi and xi can easily be calculated following PYTHAGORAS:
s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 2
ð1Þ ð2Þ ð1Þ ð2Þ ð1Þ ð2Þ
s¼ x1 x1 þ x2 x2 þ x3 x3 : ð2:2:2Þ
Note that, in general, the corresponding formula does not hold in curvilinear
coordinates, zi:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1Þ ð2Þ2 ð1Þ ð2Þ2 ð1Þ ð2Þ2
s 6¼ z1 z1 þ z2 z2 þ z3 z3 : ð2:2:3Þ
However, if the coordinate points (1) and (2) are infinitesimally close, it
becomes possible to derive a relation similar to Eq. (2.2.2). We first define:
ð1Þ ð2Þ
ð1Þ ð2Þ
dxi ¼ xi xi or dzk ¼ zk zk ð2:2:4Þ
and obtain for the total differential by using Eq. (2.2.1)2:
oxi 1 oxi 2 oxi 3
dxi ¼ dz þ 2 dz þ 3 dz : ð2:2:5Þ
oz1 oz oz
This can be expressed in shorthand notation:
oxi k
dxi ¼ dz ; ð2:2:6Þ
ozk
if we agree to sum up automatically from 1 to 3 (or up to 2 for planar problems)
whenever an index appears twice in a product. In the present case this concerns the
index k. This is known as EINSTEIN’s summation rule in the literature. We also refer
to k as a bound index in contrast to free indices, i.e., those that do not appear twice
(in the present case the letter ‘‘i’’). The infinitesimal distance ds between the two
infinitesimally close points (1) and (2) can now be calculated if we insert
Eqs. (2.2.4) and (2.2.6) into (2.2.2):
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffi oxi oxi k j
ds ¼ dxi dxi ¼ dz dz : ð2:2:7Þ
ozk oz j
Albert EINSTEIN was born on March 14, 1879 in Ulm (South Germany)
and died on April 18th, 1955 in Princeton. He is certainly the most
eminent scientist of the twentieth century. Similar to NEWTON or
MAXWELL he enriched physics by many fundamental discoveries from
different fields. The development of General Relativity and the tensor
calculus that was used therein for describing space–time is probably his
most popular contribution. However, this did not win him the NOBEL
price. On the contrary: This award was given to him for something much
less ‘‘obscure,’’ namely for his interpretation of the photo-electric effect.
x3
ϑ = const.
(r-line)
r = const.
(ϑ -line)
x2
z
x2
x2= const.
ϑ (x1 -line)
r x1
x1 = const.
x1
(x2 -line)
x2
z3 u ¼ arctan ; x3 ¼ r cos # ¼ z1 cos z2 :
x1
ð2:2:15Þ
Discuss the shape of coordinate lines of a constant radial distance,
azimuthal as well as polar angle. Show that these lines are perpendicular to
each other. Moreover, show that the components of the metric tensor read:
0 1
1 0 0
gkj ¼ @ 0 r 2 0 A: ð2:2:16Þ
0 0 r 2 sin2 #
As a trivial example for an application of the formula for the line element
(2.2.9) we calculate the circumferential length U of a circle CR of radius R. In this
case we have dr ¼ 0 and dz ¼ 0 and obtain:
I Z2p
U¼ ds ¼ Rd# ¼ 2pR: ð2:2:17Þ
CR 0
As a more complex example for Eq. (2.2.14) we consider the situation shown in
Fig. 2.2: The objective is to determine the length L of the diagonal in a rectangle of
height H and width 2pR. Obviously this can be obtained by using the Pythagorean
theorem:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L ¼ H 2 þ ð2pRÞ2 ; ð2:2:18Þ
which has nothing to do with Eq. (2.2.14). However, now transform the rectangle
into a three-dimensional object, namely the mantle of a cylinder, as shown in
Fig. 2.2. This way the former diagonal is also transformed into a three-dimensional
curve.
2.2 First Definitions and Notions in Index Notation 21
⇒
L
2π R
It makes sense to calculate the length of the curve with cylindrical coordinates.
We first note that the radial distance of the curve on the mantle does not change:
r ¼ R ¼ const: ð2:2:19Þ
Consequently Eq. (2.2.14) becomes:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dr ¼ 0 ) ds ¼ R2 d#2 þ dz2 : ð2:2:20Þ
Now we assume that the height z changes linearly with the polar angle #:
z ¼ A# þ B: ð2:2:21Þ
Of course we have:
zð# ¼ 0Þ ¼ 0 and zð# ¼ 2pÞ ¼ H; ð2:2:22Þ
and therefore the constants A and B become:
H
A¼ and B ¼ 0: ð2:2:23Þ
2p
If we insert this in Eq. (2.2.20) we obtain:
#Z¼2p sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
H 2
H 2
H
dz ¼ d# ) s ¼ R þ d# ¼ R þ 2p; ð2:2:24Þ
2p 2p 2p
#¼0
Use the equation to show that the length of the equatorial circumference
as well as the length of any great circle of a sphere of radius R is given by:
U ¼ 2pR: ð2:2:26Þ
22 2 Coordinate Transformations
r
L
⇒ L
α
β
We now return to the problem of calculating the distance between two infinites-
imally close points, which was already investigated component-wise in Eq. (2.2.6).
In the absolute language of vectors we denote the infinitesimal distance between
both points by dx. Moreover, e1 , e2 , and e3 (or short ei ; i ¼ 1; 2; 3) stand for three
Cartesian unit vectors satisfying so-called conditions of orthonormality:
ei ej ¼ dij ; ð2:3:1Þ
where the so-called KRONECKER symbol a.k.a. unit tensor has been used:
1; if i ¼ j
dij ¼ ð2:3:2Þ
0; if i 6¼ j:
If we now observe the relations (2.2.1) between the coordinates xi and zk we
may write:
oxi k
dx ¼ dxi ei ¼ dz ei ; ð2:3:3Þ
ozk
where Eq. (2.2.6) has been used again and EINSTEIN’s summation rule has been
extended to expressions related to coordinate lines and vectors. Reshuffling terms
in (2.3.3) yields:
oxi
dx ¼ dzk gk ; gk ¼ ei : ð2:3:4Þ
ozk
24 2 Coordinate Transformations
Figure 2.5 illustrates for the planar case (which is simpler to draw) how we
have to interpret the new defined vectors gk . The figure shows how a tangent
vector to the line z1 ¼ const:, namely g2 , is obtained:
For the case of three dimensions all arguments hold analogously. Therefore we
may say that, in general, gk denotes tangent vectors to the lines z j ¼ const:
However, note that these are not necessarily unit vectors. We now use them to
calculate the following scalar product and rearrange terms slightly:
oxi oxj oxi oxj
gk gl ¼ k
ei l
ej ¼ k l ei ej
oz oz oz oz
ð2:3:7Þ
oxi oxj oxi oxi
¼ k l dij ¼ k l gkl ;
oz oz oz oz
where the KRONECKER symbol of Eq. (2.3.2) has been used. Note that the effect of
the KRONECKER symbol consists of replacing one of its bound indices in the
remaining expression of a product with its other index. In order to prove this
statement all sums must be expanded first. Then all of the occurring terms can be
simplified by observing Eq. (2.3.2).
We conclude that the scalar product between the two tangent vectors yields the
components of the metric tensor. The basic definition of the scalar product of two
vectors involves the cosine of the angle they enclose. Therefore non-diagonal
components of the metric must vanish, if the curvilinear coordinates are orthog-
onal as, for example, in the case of cylindrical or spherical transformations.
Consequently, their explicit calculation is unnecessary, albeit possible, as for
example demonstrated in Eq. (2.2.12)2.
As a specific example we consider the case of plane polar coordinates for which
the tangent vectors to the coordinate lines can be calculated explicitly. We use Eqs.
(2.2.11) in context with (2.3.4)2 to obtain:
ox1 ox2
g1 ¼ 1
e1 þ 1 e2 ¼ cos # e1 þ sin # e2 er ;
oz oz
ð2:3:8Þ
ox1 ox2
g2 ¼ 2 e1 þ 2 e2 ¼ r sin # e1 þ r cos # e2 re# :
oz oz
In this equation we have introduced the commonly used unit vectors er and e#
of polar coordinates. They are shown in Fig. 2.6. In particular, the second chain of
equations shows that the tangent vectors gi do not necessarily need to be
normalized.
e1
26 2 Coordinate Transformations
Use these results to reconfirm the expression for the metric tensor shown
in Eq. (2.2.16). Also confirm that the tangent vectors can be linked to the unit
vectors er , e# , and eu of Fig. 1.3 as follows:
g1 ¼ e r ; g2 ¼ r e # ; g3 ¼ r sin # eu : ð2:3:13Þ
2.3 Vector Interpretation of the Metric 27
x-system
z-system x-system
A2 z-system
(z)
A
(x) 2 A
L2 (x)2 A2 L2
A (z)
A
β
l 2 l3 L1
L1
A
l1 (z)1 A1
β α (z) α
α β
A A
(x)1 (x)1
l1 l2
ozi oxk
Ai ¼ Ak ; Ai ¼ Ak : ð2:4:1Þ
ðzÞ oxk ðxÞ ðzÞ ozi ðxÞ
2.4 Co- and Contravariant Components 29
x2
z2 L2
L1
β z1 α
x1
l1 l2
For the proof we consider the systems x and z in Fig. 2.8 and conclude that the
coordinate transformation of Eq. (2.2.1) can be written explicitly as (also see
Exercise 2.3.1):
Next we multiply in Eq. (2.4.1) Ai by the metric gni , observe Eq. (2.2.8) and
ðzÞ
obtain:
oxl oxl ozi oxl oxk
gni Ai ¼ n i
Ak ¼ n dlk Ak ¼ n Ak ¼ An : ð2:4:5Þ
ðzÞ oz oz oxk ðxÞ oz ð xÞ oz ðxÞ ðzÞ
Here we have used the chain rule after the second equality sign (or, figuratively
speaking, ‘‘cancelled out’’ ozi ). This generates a KRONECKER symbol, dlk , first and
then Eq. (2.4.1)2 was observed. Of course, the KRONECKER symbol is nothing else
but the unit matrix in component form, i.e., we may write:
oxl oxk
n
dlk Ak ¼ n Ak or dlk Ak ¼ A l ; ð2:4:6Þ
oz ðxÞ oz ðxÞ ðxÞ ðxÞ
which, consequently, transforms the index l in Eq. (2.4.5) into the index k, or vice
versa. Eq. (2.4.1) was applied once more after the last equality sign of Eq. (2.4.5).
This time, however, for the covariant components An .
ðzÞ
We conclude that by means of the covariant components glk of the metric tensor
it becomes possible to convert the contravariant index k into a covariant one,
l. This process is also known as contraction in textbooks on tensors: Multiplication
with the covariant metric components glk lowers the index k. However, it is also
possible to raise indices. To this end we now introduce the inverse to glk by:
ozl ozk
glk ¼ ; ð2:4:7Þ
oxp oxp
and may write:
Ai ¼ gij Aj ; Ai ¼ gij Aj : ð2:4:8Þ
ðzÞ ðzÞ ðzÞ ðzÞ
The first equation shows that contraction of an expression with gij will raise the
covariant index j to a contravariant index i.
In fact, we may manipulate the partial derivatives in Eq. (2.4.1) as if they were
fractions, i.e., write:
oxk i ozi
Ak ¼ A; Ak ¼ Ai : ð2:4:9Þ
ðxÞ ozi ðzÞ ðxÞ oxk ðzÞ
All of this is a consequence of the chain rule. If, for example, we multiply
Eq. (2.4.1)1 by the expression oxm =ozi , we obtain:
i.e., by renaming the indices m ! k exactly Eq. (2.4.9)1. Equation (2.4.9)2 can be
validated in an analogous manner.
It was mentioned in context with Eq. (2.2.9) that the metric tensor allows to
calculate distances with curvilinear coordinates. We will now show that it can also
be used to determine the length of a vector. For this purpose we start from the
basic definition for the length of a vector, namely with the scalar product. If the
vector is represented by Cartesian components we may write:
pffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A ¼ A A ¼ ðAi ei Þ ðAj ej Þ ¼ Ai Aj ei ej
ðxÞ ðxÞ ð xÞ ð xÞ
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffi ð2:4:11Þ
¼ Ai Aj dij ¼ Ai Ai ;
ð xÞ ð xÞ ðxÞ ðxÞ
where use was made of Eqs. (2.3.1/2.3.2) and the properties of the KRONECKER
symbol. By observing Eq. (2.4.9)1 and the basic definition of the metric tensor
(2.2.8) this yields:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
oxi k oxi l oxi oxi k l rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A¼ A A ¼ A A ¼ gkl Ak Al ; ð2:4:12Þ
ozk ðzÞ ozl ðzÞ ozk ozl ðzÞ ðzÞ ðzÞ ðzÞ
Note that bound indices in the sense of the summation convention can be used
only once (consequently we have to distinguish between k and l). Moreover, the
32 2 Coordinate Transformations
index calculus allows us to check easily if the summation convention has been
applied correctly: Bound indices in a tensor equation always have to appear in
pairs, i.e., one of them is covariant and the other one is contravariant (see, for
example, the index k in gkl connecting to Ak ). This property becomes also evident
ðzÞ
in the following third alternative equation for the length of a vector:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
oxi k ozl oxi ozl k
A¼ k
A Al ¼ A Al
oz ðzÞ oxi ðzÞ ozk oxi ðzÞ ðzÞ
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffi ð2:4:14Þ
¼ dlk Ak Al ¼ Al Al :
ðzÞ ðzÞ ðzÞ ðzÞ
Analogously to the case of vectors co- and contravariant components can also
be introduced for tensors. If we consider the absolute tensor quantity B, which
could represent the stress tensor r or the strain tensor e (say), we can write
analogously to Eq. (2.4.1):
ozi oz j oxk oxl
Bij ¼ Bkl ; Bij ¼ i j Bkl ;
ðzÞ oxk oxl ðxÞ ðzÞ oz oz ðxÞ
ð2:4:15Þ
ozi oxl oxk oz j
Bi j ¼ j
Bkl ; Bj i ¼ i Bkl :
ðzÞ oxk oz ðxÞ ðzÞ oz oxl ðxÞ
For obvious reasons the components in the last two Eq. of (2.4.15) are also
known as mixed components of the tensor B. As in the case of vectors all indices
can be raised and lowered by means of the co- and contravariant metric compo-
nents of Eq. (2.4.8). For example:
Bij ¼ gik gjl Bkl ; Bij ¼ gik gjl Bkl : ð2:4:16Þ
ðzÞ ðzÞ ðzÞ ðzÞ
And we may treat the derivatives like ordinary fractions following Eq. (2.4.9):
oxk oxl ij ozi oz j
Bkl ¼ B ; B kl ¼ Bij ;
ðxÞ ozi oz j ðzÞ ðxÞ oxk oxl ðzÞ
ð2:4:17Þ
oxk oz j i ozi oxl j
B kl ¼ i B j; Bkl ¼ B i:
ð xÞ oz oxl ðzÞ ðxÞ oxk oz j ðzÞ
Again the proof is based on successive application of the chain rule. Observe
that bound indices always appear in co-/contravariant pairs.
2.4 Co- and Contravariant Components 33
x1 ¼ r ; x2 ¼ 0 ; x3 ¼ z; x1 ¼ r; x2 ¼ 0; x3 ¼ 0: ð2:4:19Þ
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Explain the difference between coordinate lines and the position vector.
The symbols rxx , rxy , rxz , ryy , ryz , rzz denote the components for plane stress
in Cartesian coordinates x1 ; x2 ; x3 x; y; z. Moreover, use Eqs. (2.4.15)2 and
(2.2.11) and derive corresponding expressions for the covariant components of
the stress tensor in cylindrical coordinates. How do the co-/contravariant
expressions for the stress tensor fit into the world of MOHR’s circle in 2D ?
34 2 Coordinate Transformations
By using the base vectors gk from Eq. (2.3.4)2 we may write for an arbitrary
vector A:
oxi
A ¼ Ak g k ¼ A k ei : ð2:5:1Þ
ðzÞ ðzÞ ozk
On the other hand we have:
A ¼ Aj e j : ð2:5:2Þ
ð xÞ
oxi ozk
Ai ¼ Ak k
) Ak ¼ Ai : ð2:5:3Þ
ðxÞ ðzÞ oz ðzÞ oxi ðxÞ
ozl
gl ¼ ej ð2:5:4Þ
oxj
2.5 Co- and Contravariant from the Perspective of Vectors 35
ozl
A ¼ gl A l ¼ e j Al : ð2:5:5Þ
ðzÞ oxj ðzÞ
ozl oxi
Ai ¼ Al ) Al ¼ Ai : ð2:5:6Þ
ðxÞ oxi ðzÞ ðzÞ ozl ðxÞ
A ¼ Al g l ¼ Ak g k : ð2:5:7Þ
ðzÞ ðzÞ
For the scalar product of the two sets of base vectors, gl and gk , we find:
l
oz oxi ozl oxi
gl gk ¼ ej e i ¼ ej ei
oxj ozk oxj ozk
ð2:5:8Þ
ozl oxi ozl oxi ozl l
¼ d ji ¼ ¼ ¼ d k :
oxj ozk oxi ozk ozk
If we recall Eq. (2.3.7), i.e., the relation gk gl ¼ gkl for the scalar product and
the analogous condition:
l k
l k oz oz ozl ozk
g g ¼ ej ei ¼ ej ei
oxj oxi oxj oxi
ð2:5:9Þ
ozl ozk ozl ozk lk
¼ dji ¼ g ;
oxj oxi oxj oxj
we find by using Eq. (2.5.7) after scalar multiplication by gm :
!
m l m k
g g A l ¼ g A gk ) Am ¼ gml Al ; ð2:5:10Þ
ðzÞ ðzÞ ðzÞ ðzÞ
or by gm :
!
l
gm Al g ¼ g m Ak g k ) A m ¼ gmk Ak : ð2:5:11Þ
ðzÞ ðzÞ ðzÞ ðzÞ
Note that we have run across these formulae before in Eq. (2.4.8).
36 2 Coordinate Transformations
Thus the vector A can be represented in the Cartesian base ej (see Eq. 2.5.2) as
well as in the skew-curvilinear bases gk and gl , which are not normalized
(Eq. 2.5.7). The same holds for tensorial quantities. As an example we consider
the tensor of second order of Eq. (2.4.15), B. The following representation is valid
in the Cartesian base:
B ¼ Bkl ek el ð2:5:12Þ
ð xÞ
B ¼ Bkl gk gl ; B ¼ Bkl gk gl ;
ðzÞ ðzÞ
ð2:5:13Þ
B ¼ B l gk g ; B ¼ Bk l gk gl ::
k l
ðzÞl ðzÞ
The notion of the tensor product or dyad, i.e., the symbol ‘‘’’ deserves an
explanation. Even though it is as necessary as a Mercedes star it is customarily
used in the literature, the main reason being to distinguish a product between
vectors (or tensors) in absolute notation from the scalar and the vector product,
which are identifiable by the symbols ‘‘’’ and ‘‘’’. Mathematically speaking, two
vectors (first order tensors), A and B, are mapped onto a number (zeroth order
tensor) by writing A B, onto another (axial) vector by A B, and onto a second
order tensor by A B. Just like the scalar or vector product ‘‘’’ can also be
introduced axiomatically by defining a corresponding algebra. However, this will
not be detailed any further in this book and the reader is referred to the more
mathematically oriented literature cited below.
2.5 Co- and Contravariant from the Perspective of Vectors 37
Exercises 2.4.6 and 2.4.7 have shown that the co-/contravariant components of
vectors and tensors do not necessarily all have the same physical dimension, for
example that of a stress. We have to concede that in engineering terms co-/
contravariant components of a vector or a tensor are, in general, rather unphysical
quantities.
However, in the case of orthogonal coordinate transformations whose coordi-
nate lines are perpendicular to each other it becomes possible to recover the
physical notion of Cartesian components by introducing so called physical com-
ponents. The key to their definition lies in the fact that the metric tensor of
orthogonal coordinate transformations is diagonal:
2 3 2 11 3
g11 0 0 g 0 0
gm gl ¼ 4 0 g22 0 5; gm gl ¼ 4 0 g22 0 5: ð2:6:1Þ
0 0 g33 0 0 g33
In that case the length of a vector (cf., Eqs. 2.4.12/2.4.13) can be expressed as a
sum of quadratic terms:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u 3 u 3
uX uX
A¼ t gii ðA Þ ¼ t
i 2
gii ðAi Þ2 : ð2:6:2Þ
ðzÞ ðzÞ
i¼1 i¼1
Note that physical components can also be defined for tensors of second and
higher order:
pffiffiffiffiffipffiffiffiffiffi pffiffiffiffiffiqffiffiffiffiffi
B hiji ¼ gii gjj Bij ¼ gii gjj Bij : ð2:6:5Þ
ðzÞ ðzÞ
How is this result related to Eq. (2.6.7)? Use the following trigonometric
theorems to answer this question:
1 1
sin2 # ¼ ½1 cosð2#Þ; cos2 # ¼ ½1 þ cosð2#Þ;
2 2 ð2:6:9Þ
sinð2#Þ = 2sin# cos#:
40 2 Coordinate Transformations
1
R ij ¼ rij r kk dij : ð2:6:11Þ
ð xÞ 3ðxÞ
ð xÞ
and:
3 3 3
r2y ¼ R ij R ji ¼ gir gjs R rs R ji ¼ gir gjs R ij R rs : ð2:6:18Þ
2ðzÞ ðzÞ 2 ðzÞ ðzÞ 2 ðzÞ ðzÞ
and:
3
r2y ¼ Rhiji Rhiji : ð2:6:20Þ
2ðzÞ ðzÞ
Richard Edler VON MISES was born on April 19, 1883 in Lemberg (now
Ukraine) and died on July 14, 1953 in Boston. From 1909 to 1918 he was a
professor of applied mathematics in Straßburg (now France) where he
investigated problems of solid and fluid mechanics, aerodynamics, statistics,
and probability theory. During WW I he was on the Austrian side where he
built and flew his own battle plane. After the war he went to Berlin until the
Nazis forced him into exile in 1933, first to Istanbul and then on to Harvard.
42 2 Coordinate Transformations
It should be noted that some authors use the following relation for the double
scalar product instead:
ei ej ::ðek el Þ ¼ dik djl : ð2:6:23Þ
Also note that all of these complications can be avoided if the index notation is
used exclusively from the very start.
Two tangent vectors, s1 and s2 , can now be defined analogously to Eq. (2.3.4)2:
oxi ðzc Þ
sa ¼ ei ; a ¼ 1; 2: ð2:7:2Þ
oza
Obviously their components in (three-dimensional) Cartesian coordinates are
then given by:
oxi ðzc Þ
sia ¼ : ð2:7:3Þ
oza
In this case it does not matter if we write the index i at the top or at the bottom
of the symbol s because it refers to a Cartesian representation. However, the
positioning of the index a does matter: Just as in the case of index k from Eq.
(2.3.4)2 it signals covariance.
Note that just like the case of gk the tangent vectors sa are no unit vectors. They
still need to be normalized in order to define a unit vector n normal to the surface:
s1 s2
n¼ : ð2:7:4Þ
js1 s2 j
Moreover, just like in Eq. (2.5.4), it is possible to define dual tangent vectors sb
by:
ozb
sb ¼ ej : ð2:7:5Þ
oxj
Because of the chain rule the following orthogonality conditions hold (cf.,
(2.5.8)):
sb sa ¼ dba : ð2:7:6Þ
We are now in a position to define co- and contravariant surface metrics in
complete analogy to the Eqs. (2.2.10), (2.3.7), (2.4.7), and (2.5.8/2.5.9):
o2 x i
bab ¼ ni : ð2:7:8Þ
oza ozb
This definition deserves a comment: Calculus teaches us that the extreme values
of a curve y ¼ yð xÞ within the plane are governed by the second derivative
d2 y d2 x, i.e., the local curvature in a point x. This explains the second derivatives
in Eq. (2.7.8). Moreover, a plane surface is, of course, not curved. Therefore we
44 2 Coordinate Transformations
expect that the curvature tensor vanishes. And, indeed, the scalar product (note the
summation implied by the index i) in Eq. (2.7.8) or, in other words the projection
of the curvature onto the normal vector, becomes zero in this case. Moreover, note
that by virtue of Eq. (2.7.3) we may write for the curvature in Eq. (2.7.8) as well:
Show that both vectors are orthogonal to each other. Are they related to
the base vectors g1 , g2 , g3 of Exercise 2.3.1? Depict them on the surface of a
2.7 A Touch of Differential Geometry 45
sphere together with surface coordinate lines. Are the tangent vectors nor-
malized? Show by using Eq. (2.7.7) that the surface metric gab is given by:
2
R 0
gab ¼ : ð2:7:14Þ
0 R2 sin2 #
Calculate its inverse gab . Show by using Eq. (2.7.4) that the Cartesian
components of the unit normal ni to the sphere are given by:
n1 ¼ sin # cos u ; n2 ¼ sin # sin u ; n3 ¼ cos #: ð2:7:15Þ
Use them to calculate the curvature tensor based on the definition shown
in Eq. (2.7.8). Finally show that the mean curvature is given by Km ¼ 1=R.
Try to interpret the minus sign by using terms like ‘‘convex’’ or ‘‘concave.’’
The book by Schade and Neemann [1] is a real treasure chest of mathematical
formulae (which makes it easier to read since it is written in German) for true
disciples of the index calculus. In particular one should look at Sects. 4.2.2 and
46 2 Coordinate Transformations
4.2.4 for the concepts of ‘‘metric,’’ and ‘‘co-/contravariance.’’ The books by Itskov
[2] and Bertram [3] insist on a mathematically more stringent approach and
emphasize the absolute tensor calculus. Particularly worth reading in context with
the present section are Chap. 1 and Sect. 3.2 (for differential geometry) in the first
and Sects. 1.1 and 1.2 in the second book. Tensor algebra and tensor analysis are
also treated in concise form in Irgens [4], Chap. 12, in index as well as in absolute
notation, and also in Liu [5], Appendix A.1.
In general the tensor concepts presented so far have been known for a long
time. Consequently it is also worth while to study the ‘‘classics.’’ In this context
the article by Ericksen [6] in the Encyclopedia of Physics, Sects. I, II, and (in
parts) III should be mentioned first. Moreover, the book by Green and Zerna [7] is
to be recommended, in particular Sects. 1.1 to 1.10. Finally Chaps. 1, 12 and,
notably, Sect. 13 (with many exotic coordinate transformations) in the book by
Flügge [8] should be pointed out.
Several notions, such as Mohr’s circle or yield stress, were used in this section
without further explanations. In this context it may be useful to study textbooks on
strength of materials, e.g., Hibbeler [9], Sects. 10.3 and 10.7, or Gross et al. [10],
Sects. 2.2.3 and 3.3.
References
In the previous chapter we have already specified a few important notions of tensor
calculus, namely the metric tensor and co-/contravariant components including
their intuitive interpretation as parallel and orthogonal projections onto the coor-
dinate axes. In the Chap. 4 we will consider derivatives with respect to skew and
curvilinear coordinates. We will explain how these can be introduced quite nat-
urally as a part of tensor calculus. However, before that we will motivate how
derivatives of fields in arbitrary coordinate systems arise. For that reason we
present in this chapter the first ‘‘ingredient’’ for the solution of continuum related
problems, the so-called balance equations, where spatial derivatives play an
essential role.
Our primary objective is the following one: In continuum thermo-mechanics we
wish to determine five fields, namely mass density, q ¼ q ^ðx; tÞ,1 the three
1
We distinguish a function of current position and time from its value by a circumflex.
−
A
V
components of the velocity vector, t ¼ ^tðx; tÞ, and temperature, T ¼ T^ ðx; tÞ, in all
material points, x, of a body, V(t), and at all times, t. Initially we will assume that
these five fields are described in a Cartesian frame x ¼ xi ei . From an atomistic
point of view the three fields can be interpreted as follows: The sum of the mass of
all the molecules, dm, divided by the volume of the material point, dV, gives the
mass density, q ¼ q ^ðx; tÞ. Similarly, if we sum up the momenta of all the mole-
cules in dV, i.e., the product of their individual masses and velocities, and divide
the result by the total mass, we obtain the macroscopically observable velocity
field, t ¼ ^tðx; tÞ, of that very ‘‘point.’’ And, finally, by computing the ‘‘average’’
kinetic energy of all the molecules in the volume we obtain a measure of the
intensity of the macroscopically invisible erratic atomic movement. This is the
(imprecise) kinetic interpretation of temperature, T ¼ T^ ðx; tÞ.
‘‘point’’ can be neglected. Only then the concept of fields becomes physically
relevant: If we consider real volumes that are only a few Ångströms wide, very
few atoms will be inside and, what is more, they will leave and enter the volume
over extremely short time periods. Thus it does not make sense to compute the
local mass density as outlined above. In fact, one must ask, how small the real
volume element can be so that it can be treated on the basis of continuum theory.
Note that, just like space, appropriate smallness of time is another big issue in
continuum theory. Scientists love arguing about the ‘‘limits of the continuum
approach,’’ for example, when they talk about nanostructures. However, they
rarely give a satisfying general answer, maybe, because this question can only be
answered on a case-by-case basis.
Interestingly the fields of continuum theory are treated as if they were math-
ematical functions in space and time: They are integrated with respect to an
arbitrarily small environment, they are differentiated without hesitation, etc. In
what follows we shall learn about the corresponding mathematical tools. However,
we must realize that mathematics has nothing to do with reality and we should
never forget that it is nothing more but a neat ‘‘language’’ conceived by the human
brain to model reality.
Obviously, in order to determine the five fields, five equations are required, the
so-called field equations. The foundation of these equations are the balance of
mass (a scalar relation), the balance of momentum (a vector equation with three
components), and the balance of energy (a scalar equation).
Moreover, we have to acknowledge the possibility that the fields are discon-
tinuous within the body. In other words, they jump when crossing a certain
boundary. The corresponding situation is illustrated in Fig. 3.1: A material body
(i.e., a region that always consists of the same material particles) is split into two
halves, denoted by V , by an open surface, A. Note that these quantities can be
time-dependent. Very often we assume this implicitly and do not acknowledge it in
the symbol. So we just write A and not A(t). Moreover, we use the symbol A for
the (open) surfaces of V . In general these are time-dependent as well. Their
outward normal unit vectors are denoted by n whereas, for the sake of distinction,
the normal on the separating surface A is characterized by the symbol e. Moreover,
the symbol L ¼ oA is used for the closed line bordering the surface A. Finally, the
outward normal to that line, and which is also a tangent to the surface A, was
denoted by m.
Osborne REYNOLDS was born on August 23, 1842 in Belfast (Ireland) and
died on February 21, 1912 in Watchet, Somerset (England). He graduated
from Cambridge University in mathematics. In 1868 he became the very
first Professor of Engineering at the University of Manchester. He stayed
there until his retirement in 1905. He is mostly known because of his
pioneering work in fluid mechanics. Particularly noteworthy among his
findings is the REYNOLDS’ number, which characterizes the transition
between laminar and turbulent flow and made him immortal.
50 3 Balances (in Particular in Cartesian Systems)
Note that the singular surface, A, can move at its own speed, independently of
the movement of the surrounding material particles. In other words, A is not
necessarily a material surface. Consequently, the two volumes V separated by
A are not necessarily material and may consist of different particles as time goes
on. Recall that at the singular surface the primary fields and their derivates will
‘‘jump,’’ i.e., they are discontinuous. As an example of such a situation consider
the interface between liquid water and its vapor. If the temperature rises the liquid
water will vaporize: Water molecules will leave the liquid and enter the gaseous
region, i.e., the matter in both subvolumes is clearly not conserved. They are open
systems. Moreover, the interface between a fiber and the surrounding matrix of a
composite material may be considered as a singular surface: Although both regions
do not show macroscopically visible motion, there are density gradients. The
matter in these systems does not change, they are material systems. It is a judg-
ment call to decide as to whether the interface in both examples is a material
singular surface or not. This depends on the physical accuracy our modeling
requires. For example there could be very thin coatings on the fibers or a
‘‘chemical reaction’’ takes place in a thin transition region between the fiber and
the matrix. We may wish to assign intrinsic properties to a material singular
surface to model such regions. With a grain of salt we may say that our model is
somewhere between two- and three-dimensions. An example of an immaterial
moving singular surface through which matter is not transported is a shock wave.
Behind and in front of a shock wave the velocity of the material particles is zero.
However, the mass densities, the pressure fields, and, if the shock wave moves
through a solid, certain components of the stress tensor show a huge gradient.
In what follows we shall specialize on material bodies, i.e., V ðtÞ ¼ V þ [ V [ A
will always consist of the same material particles. However, since we do not treat
open systems explicitly, this is a deficit only on first glance: It is true that for open
systems relative convective flows across the open surface of the to-be-balanced
quantity need to be added to global balance equations. Moreover, the velocities in
REYNOLDS’ transport theorem as stated below need to be changed from material
velocities to non-material ones of the surface of the open system. However, we shall
see that the global balances are only a vehicle leading us to the final result namely
local equations: In regular points of the continuum we will obtain partial differ-
ential equations for the fields and their derivatives. They hold in a material point
and, therefore, they will no longer carry information regarding the movement of the
system boundaries, open or not. Moreover, in singular points, i.e., points on the
singular surface we will obtain so-called jump conditions. These do carry infor-
mation about the state of the singular surface, including its movement.
In the next chapters we will first concentrate on bodies without a singular
surface and, in particular, state the global balances of mass and momentum for
these. We will discuss the general structure common to all global balances and
derive a general field formulation. This will allow us to derive local balances in
regular points, in particular for mass and momentum. The local balance of
momentum in regular points will then be used to derive, first, the local and,
3.1 Preliminary Remarks 51
second, the global balance of kinetic energy. We shall realize that this quantity is
not conserved since it contains a production term. This in mind we shall postulate
the balance of total energy, which is a conserved quantity. The local balance of
momentum will finally also be used to obtain local and global balances for the
moment of momentum. All balances will then be formulated for bodies with a
singular surface.
The total mass, M, of a material body, V(t), follows by adding the masses, dM, of
all material points within small sub-volumes, dV. Moreover, the ratio dM=dV
represents the mass density, q ¼ q
^ðx; tÞ, of the material point, which can change in
space and time. Consequently we may write:
Z ZZZ
M¼ dM ¼ q dV: ð3:2:1Þ
M V ðtÞ
The total mass of a material body is (by definition) conserved, i.e., it does not
change with time:
ZZZ
dM d
¼0 ) ^ðx; tÞ dV ¼ 0:
q ð3:2:2Þ
dt dt
V ðt Þ
Rewriting the last expression any further is not possible yet: We have to clarify
how to perform the time derivative since the time variable is present in the inte-
gration boundaries and in the integrand as well. At this point we only note that Eq.
(3.2.2) is the global balance of mass. Note again that mass is a conserved quantity,
it cannot be produced and it obeys a conservation law. In classical physics mass
cannot be created out of nothing nor can it vanish into nowhere. Moreover, a
material system does not allow that mass is transported across its boundaries, i.e.,
mass cannot leave nor enter through the system’s boundaries. This is why the right
hand side of Eq. (3.2.2) is simply zero.
Just like mass the momentum of mass is an additive quantity, too. Thus the total
momentum can be obtained by summing up the contributions dP ¼ dMt ¼ qt dV
of all material points. However, unlike mass momentum is a vector, due to the
velocity field t ¼ ^tðx; tÞ it contains. Consequently, we obtain for the total
momentum of a body:
ZZZ
P¼ q t dV: ð3:2:3Þ
V ðt Þ
Baron Augustin Louis CAUCHY was born on August 21, 1789 in Paris. He
died on May 23, 1857 in Sceaux near Paris. In his early years he was a
scientist in NAPOLEON’s army, just like FOURIER. At the age of 26 he was
already a professor at the École Polytechnique, where he soon established
himself as one of the leading French mathematicians of his time. More
than 780 publications are attributed to him. In the same context, however,
rumor has it that he was not free of committing plagiarism. True or not, in
any case it got him the dubious nickname ‘‘cochon’’ in academic circles.
Note that the so-called traction, t, does not only depend on position and time,
but also on the unit normal, n, of the surface element, dA. CAUCHY showed by using
the so-called tetrahedron argument (cf., Exercise 3.2.1) that the dependence is
linear and the following relation holds to the field of the stress tensor, r:
t ¼ n r: ð3:2:6Þ
3.2 Balances of Mass and Momentum 53
in which we recognize the Cartesian components of the stress tensor. For the
proof also apply and comment on the reaction-principle in the form:
^tðx; t; nÞ ¼ ^tðx; t; nÞ: ð3:2:9Þ
e3
d A1
d A2
dA
e2
e1 d A3
Also interpret Eq. (3.2.7) in terms of a (left sided) scalar product between
the vector n ¼ nk ek and the tensor r ¼ rji ej ei so that the general vector
relation (3.2.6) results.
In honor of its inventor the quantity r is also referred to as the CAUCHY stress
tensor. Sometimes it is also called the current or true stress tensor, since it relates
current forces to current surfaces.
54 3 Balances (in Particular in Cartesian Systems)
As before, the left hand side of this equation cannot be transformed any further
yet, because of the complication involved with the time differentiation. Once again
we note that this equation represents a global balance, this time for the momentum.
Momentum is a conserved quantity. There is no production of momentum.
However, there is a flux of momentum across the boundary, namely T, and a
volume supply of momentum, namely F. One may ask why is F a supply and not a
production? The answer is hidden in the following argument: In principle supplies
can be shielded and, therefore, be controlled. This is not possible with productions.
They develop inside the system and cannot be influenced by an outside observer,
even in principle. Gravity is a concrete example of a supply of momentum. It can
be switched off if we move the system far away from all other masses. This should
be possible, at least in the mind.
Now that we have learned about two specific balances it is time to generalize and
then face the problem of how to perform the time derivatives on their left hand side.
surfaces A and across the periphery of the singular surface A, i.e., the line L oA,
all of which are depicted in the figure. By using the vector field densities / and /,
A L
which are with reference to unit surface and unit length, respectively, we write:
ZZ I
F¼ / n dA / m dl: ð3:3:3Þ
A L
Aþ [A L
The two minus signs are a matter of convention: If the vectors / and / point
A L
into the body their scalar product with the outward normals n and m becomes
negative. In this case, however, we expect w to grow. This will be guaranteed by
the minus signs. As far as P is concerned, we note that production is possible
within the volumes V as well as in the (open) surface A. Therefore we write with
the corresponding volume and surface densities of production p and p:
V A
ZZZ ZZ
P¼ p dV þ p dA: ð3:3:4Þ
V A
V þ [V A
Just like the production the supply can be described by volume and surface
densities s and s:
V A
ZZZ ZZ
S¼ s dV þ s dA: ð3:3:5Þ
V A
V þ [V A
At this point we may ask again why it is useful to distinguish between pro-
ductions and supplies which, from a mathematical point of view, are completely
alike. As already mentioned the reason for the distinction is a physical one:
Supplies can be controlled and suppressed by the experimenter, whereas produc-
tions cannot. As an example of a supply, gravity has already been mentioned,
which can be switched off by performing the experiment in outer space (say).
Another example (in context with the energy balance, see this section below) is
radiation, which can be prevented to enter the system by suitable shielding. An
example of a production is the dissipated power of the stresses within a volume
due to velocity gradients, profanely known as internal friction. Friction cannot
simply be switched off. On the contrary, it will adjust itself during the process.
Thus we obtain by combination of Eqs. (3.3.1–3.3.5):
ZZZ ZZ ZZ I
d d
w dV þ w dA ¼ / n dA / m dl
dt V dt A A L
V þ [V A Aþ [A L
ZZZ ZZ ð3:3:6Þ
þ p þ s dV þ p þ s dA:
V V A A
V þ [V A
56 3 Balances (in Particular in Cartesian Systems)
This equation is much more detailed than Eq. (3.3.2). However, it is compar-
atively unwieldy to use during calculations. The integrals pose a potential prob-
lem. They are often hard to evaluate. Mathematically ‘‘easier’’ are localbalances.
In these the various densities are related differentially. As we shall see they can be
turned into partial differential equations for the aforementioned primary fields, so
that well-established solution methods can be applied. In fact we have to distin-
guish between localbalances inregular and in singularpoints of the body,
respectively, or in other words, points within the (material) volumes, V þ and V ,
and points on the singular surface, A. To obtain them we first of all have to deal
with the time derivatives on the left side of Eq. (3.3.6) and pull them beneath the
integrals, so to speak. For this purpose we need so-called transport theorems which
will be discussed—for volume integrals—in the Sect. 3.4. Finally, it is worth
mentioning that Eq. (3.3.6) simplifies considerably for the case of a material
volume V with the closed surface oV that does not contain a singular surface:
ZZZ ZZ ZZZ
d
w dV ¼ u n dA þ p þ s dV: ð3:3:7Þ
dt V A V V
V oV V
Executing the time derivatives on the left hand side of Eqs. (3.3.6 / 3.3.7) is
problematic because the integrand as well as the integration boundaries are both
time-dependent. In this section we concentrate on the case of the volume integral
shown in Eqs. (3.3.6 / 3.3.7) and anticipate the result. It turns out that in Cartesian
coordinates we may write:
2 3
ZZZ ZZZ ow
d 4 o
w dV ¼ V
þ w ti 5 dV: ð3:4:1Þ
dt V ot oxi V
V þ [V V þ [V
Note that the same equation holds also for a volume without a singular surface.
In this case we simply replace V þ [ V by V. Also note that we have chosen the
Cartesian representation for the velocity field, mainly for computational reasons as
we shall see shortly. However, before we turn to the proof of this relation we will
rewrite the second integral. This is where GAUSS’ theorem comes in.
Vi
In Cartesian coordinates the field g can, but does not have to be, a
component of a vector or a tensor. We proceed to sketch a proof of the
theorem in Cartesian coordinates and will extend it later to arbitrary coor-
dinates (see Sect. 4.5). To this end consider the sketch of Fig. 3.3. Start with
the surface integral on the right hand side of Eq. (3.4.2) and apply it suitably
to the six surfaces of the small cubes shown in the figure.
Now extend the surfaces of each cube to cover its whole volume. Com-
bine adjacent surfaces of the cube and generate partial derivatives while
observing the mean value theorem for integrals. Consider now the limit case
of infinitesimally small cubes and arrive at Eq. (3.4.2). Reconsider each step
of the sequence of mathematical operations and explain why Eq. (3.4.2) is
only valid for continuous fields. Explain that the equation can be generalized
as follows in order to cover also the case of a singular surface, A, dividing the
volume as shown in Fig. 3.1:
ZZZ ZZ ZZ
og
dV ¼ g ni dA ½½g ei dA; ð3:4:3Þ
oxi
V þ [V Aþ [A A
½½g ¼ gþ g : ð3:4:4Þ
58 3 Balances (in Particular in Cartesian Systems)
gþ and g denote the right- and left-sided limits of the field g when
approaching the singular surface.
Now define the so-called del operator or nabla symbol by using the
Cartesian base of unit vectors, ei :
oðÞ
r¼ ei : ð3:4:5Þ
oxi
Show that Eqs. (3.4.2 / 3.4.3) can be written in the following symbolic,
system independent form:
ZZZ ZZ
rg dV ¼ g n dA;
V oV
ZZZ ZZ ZZ ð3:4:6Þ
rg dV ¼ g n dA ½½g e dA
V þ [V Aþ [A A
or:
ZZZ ZZ
r g dV ¼ g n dA;
V oV
ZZZ ZZ ZZ ð3:4:8Þ
r g dV ¼ g n dA ½½g e dA:
V þ [V Aþ [A A
By using the result from Eq. (3.4.3) we can now rewrite Eq. (3.4.1) as follows:
ZZZ ZZZ o w
d V
w dV ¼ dV
dt V ot
V þ [V V þ [V ð3:4:9Þ
ZZ ZZ
þ w ti ni dA ½½wV ti ei dA:
V
Aþ [A A
Note that during the proof of this equations we have implicitly assumed (namely
during the application of the extended GAUSS’ theorem with discontinuities)
that V þ and V are material volumes, in which the fields are continuous.
3.4 Transport Theorem for Volumes 59
We write for the jump ½½wV ti ¼ wþ tþ
i w t
i ¼ w wþ
t i . After the last
V V V V A
where for the sake of brevity the normal component of the velocity of the
immaterial surface A has been introduced:
t ? ¼ t i ei t e: ð3:4:11Þ
A A A
that quantity in a material point may change in time, which explains the volume
integral. Second, the quantity may leave or enter the volume by convective flow.
This corresponds to the surface integral. An influx corresponds to a negative value
of t n and a drain to a positive one. If the velocity and the normal are perpen-
dicular to each other there is no net influx: nothing will leave nor enter the volume.
However, we are not completely satisfied with the intuitive argument yet.
Before we comment on how to treat the time derivative in front of the surface
integral of Eq. (3.3.6) and discuss the corresponding transport theorem, we present
a more formal proof of Eq. (3.4.1). For this purpose we need the concept of the
Lagrangian description also known as referential or material perception. Recall
that on the one hand the motion of a material point as shown in Fig. 3.4 can be
described from the Eulerian (also known as spatial) point of view: At the time t the
body moves across a fixed grid in space, x. According to LAGRANGE we may
alternatively characterize the motion by moving along with the material point:
Fig. 3.5. The material point itself—which can neither be created nor destroyed—is
FiK ¼ o~xi ðXL ; tÞ=oXK ¼ F~ iK ðXL ; tÞ denotes the so-called deformation gradient.
Its purpose is to connect current distances with distances in the reference con-
figuration. Note that we have used capital Latin characters for some of the indices
in the previous formulae. They run from 1 to 3 just like small letters. However,
they are supposed to remind us of the reference configuration. Note that the
deformation gradient possesses both types of indices since it connects the refer-
ence configuration with the current one. Although such a subtle distinction during
the choice of indices is not imperative, it is helpful and provides an additional
means of checking the correctness of tensor equations.
Moreover, the mathematical requirement det F 6¼ 0 can be interpreted as conti-
nuity of particle motion. In order to appreciate this notion, recall that the motion of a
particle must be unique since particles cannot be created nor destroyed. Conse-
quently, the following unique inversions of Eq. (3.4.13) are possible:
2
In this book functions written in Eulerian notation are identified by a circumflex whereas
functions in Lagrangian description are easily spotted by a tilde.
62 3 Balances (in Particular in Cartesian Systems)
x(X , t )
x=~
(Lagrange)
Carl Gustav Jacob JACOBI was born on December 10, 1804 in Potsdam
(Germany) and died on February 18, 1851 in Berlin. He was a child
prodigy, at least in mathematics, ready to go to university at the tender
age of 13. However, Berlin University would not accept students younger
than 16 and so he did private studies on advanced mathematics for four
more years. This allowed him to finish his dissertation in 1825 and his
habilitation thesis the year after, both in Berlin. From 1826 to 1843 he
taught in Königsberg (East Prussia). In 1844 he became a full member of
the Prussian Academy of Science and, not surprisingly, had a nervous
breakdown the very same year. For rehab he first went to Italy to finally retire in Berlin while
enjoying his pension provided to him by the Prussian Crown. Foolishly he engaged himself in
the revolution of 1848 and the payment of his monies was temporarily suspended.
In order to guarantee this property the function of motion ~xi ðXk ; tÞ must be
unique and continuously differentiable, i.e., the following determinant must not
vanish (so-called inverse function theorem):
oxj
oX 6¼ 0: ð3:4:16Þ
K
Here we have used the completely antisymmetric tensor of third order (a.k.a.
the Levi-Civita symbol) in Cartesian components:
8
< þ1; if i; j; k ¼ 1; 2; 3 and cyclic permutations
ijk ¼ 1; if i; j; k ¼ 2; 1; 3 and cyclic permutations ð3:4:18Þ
:
0; else:
3.4 Transport Theorem for Volumes 63
Now identify Fi1 , Fj2 , Fk3 suitably and validate the form
1
6ijk LMN FiL FjM FkN :
Now the first assertion is that the following transformation rule holds for vol-
ume elements:
dV ¼ J dV0 : ð3:4:22Þ
Here dV denotes the volume element in the current and dV0 the volume element
in the reference configuration, respectively. For the proof we choose without loss
of generality dV0 to be a rectangular volume. Consequently it can be described by
a triple product of the following three vectors:
ð1Þ ð1Þ ð2Þ ð2Þ ð3Þ ð3Þ
d X L ¼ d X 1 ; 0; 0 ; d X M ¼ 0; d X 2 ; 0 ; d X N ¼ 0; 0; d X 3 :
ð3:4:23Þ
We check:
ð2Þ
ð1Þ ð3Þ
ð1Þ ð2Þ ð3Þ
dV0 ¼ d X d X d X ¼ d X L LMN d X M d X N
ð3:4:24Þ
ð1Þ ð2Þ ð3Þ
¼ 123 d X 1 d X 2 dX 3 ¼ 1 dV0
64 3 Balances (in Particular in Cartesian Systems)
Equations (3.4.22 / 3.4.27) will now be used to prove Eq. (3.4.1). We assume
that the volume density is given in material coordinates, i.e., w ¼ w~ ðXK ; tÞ and
V V
find:
2 3
ZZZ ZZZ ZZZ ow
d d 4
w dV ¼ w J dV0 ¼ V
J þ w J_ 5 dV0
dt V dt V ot V
V þ [V V0þ [V0 V0þ [V0
2 3 2 3 ð3:4:32Þ
ZZZ ow ZZZ ow
¼ 4 V þ w oti 5 J dV0 ¼ ot
4 V þ w i 5 dV:
ot V oxi ot V oxi
V0þ [V0 V þ [V
In a last step we convert the first term in the integrand from Lagrangian to
Eulerian description:
0 1
ow ow~ ðX; tÞ ow~ X ^ ðx; tÞ; t ow xðX; tÞ; tÞ
^ ð~
@ VA V ¼ V ¼ V
ot ot ot
ot
Lagrange X X
X
ow^ ðx; tÞ o w^ ðx; tÞ
o~xi ðX; tÞ
¼ V þ V
oxi ot X ot
x
t
ow^ ðx; tÞ o w ^ ðx; tÞ
¼ V þ
V
~ti ðX; tÞ
ot oxi
x t ð3:4:33Þ
ow^ ðx; tÞ o w ^ ðx; tÞ
^
¼ V þ
V
~ti Xðx; tÞ; t
ot oxi
x t
ow^ ðx; tÞ o w ^ ðx; tÞ
¼ V þ
V
^ti ð~xðX; tÞ; tÞ
ot oxi
x t 0 1
^
o wðx; tÞ o wðx; tÞ
^ ow ow
¼ V ^ti ðx; tÞ @ þ V ti A
V V
þ :
ot oxi ot oxi
Euler
x t
66 3 Balances (in Particular in Cartesian Systems)
Note that before the first and after the second identity sign the arguments of the
functions were not explicitly shown as is common practice in continuum theory, at
least for non-ambiguous cases. We finally combine the result with the second part
of the argument of the integral in Eq. (3.4.32):
0
ow
1
ow~ ðX; tÞ
@ V þ w oti A V þw ~ ðX; tÞo~ti ðX ðx; tÞ; tÞ
ot V oxi ot V oxi
t
Lagrange X
ow^ ðx; tÞ o w ^ ðx; tÞ
¼
V
þ
V
^ti ðx; tÞ þ w^ ðx; tÞo^ti ðx; tÞ
ot oxi V oxi t
x t
^ ^
o wðx; tÞ o wðx; tÞ^ti ðx; tÞ
oxi
V V
¼ þ
ot
x t
0 1
ow ow
o o
@ V þ w ti A ¼ Vþ w ti ;
ot oxi V ot oxi V
Euler
ð3:4:34Þ
where the expressions after the last identity sign were written in Eulerian manner.
This concludes the proof of the transport theorem for volumes.
We now turn to the problem of how to rewrite the time derivative of the surface
integral from Eq. (3.3.6). As for the case of volume integrals we first state and
discuss the result:
2 3
ZZ ZZ o w ffi
d 4 A þ w t D 2Km t ? 5 dS:
w dS ¼ ;D ð3:5:1Þ
dt A ot A A A
A A
Greek indices run from 1 to 2. The capital letters Z C denote the two surface
coordinates of the material point in the reference configuration,3 i.e., the termi-
nology was chosen analogously to the symbol X from Sect. 3.4. The local velocity
t of the singular surface is now decomposed w.r.t. the two tangential vectors, s1
A
and s2 , and the normal vector, e, of the corresponding point on the surface:
t ¼ t D sD þ t ? e or t i ¼ t D siD þ t ? ei : ð3:5:3Þ
A A A A A A
Recall (cf., Sect. 2.7) that the components siD and ei of all three vectors refer to
a (global) Cartesian coordinate system. If we include time among the variables we
find similarly to Eq. (2.7.3) that:
o~xi ðZ C ; tÞ
siD ¼ : ð3:5:4Þ
oZ D
otD oZ D o2 xk
t D;C ¼ A
þ CDCR t R ; CDCR ¼ : ð3:5:5Þ
A oZ C A oxk oZ C oZ R
CDCR denotes the so-called CHRISTOFFEL symbols. Both concepts anticipate
results from Chap. 4, where we will discuss in great detail how to differentiate
vectors and tensors in skew curvilinear coordinates, albeit for three dimensions. At
this point it may suffice to say that the CHRISTOFFEL symbols vanish for Cartesian
coordinate transformations. In other words, for a moving singular plane the
covariant derivative of the velocity will reduce to a partial one. This in turn, is a
3
This is why we have used capital Greek characters for the indices.
68 3 Balances (in Particular in Cartesian Systems)
direct analogue of the partial spatial derivative of Eq. (3.4.1). In this case the mean
curvature, Km , defined in Eq. (2.7.12) vanishes.
In order to prove Eq. (3.5.1) a few auxiliary equations are in order. A (directed)
surface element results from a vector product between two non-collinear vectors
ð1Þ ð2Þ
d x and d x . We use a Cartesian base, observe the chain rule in context with Eq.
(3.5.2) as well as the definition for the tangent vectors of Eq. (3.5.4) and obtain:
ð1Þ ð2Þ oxj oxk ð1ÞC ð2ÞD
dSi ¼ ijk dx j dx k ¼ ijk dZ dZ
oZ C oZ D ð3:5:6Þ
ð1Þ ð2Þ
¼ ijk sCj skD dZ dZ C D
¼ dS ei :
ð1Þ ð2Þ
Obviously d x and d x are tangential to the surface. Therefore their vector
product points in the direction of the unit normal e. This was explicitly
acknowledged after the last equal sign so that we now possess a relation that links
the current surface element to the time-independent coordinates of the reference
configuration. This will be quite beneficial during differentiation. We obtain:
ð1Þ ð2Þ
dS ¼ ijk ei sCj skD d Z C
dZ D
: ð3:5:7Þ
In order to make calculations even easier we align the mesh of the reference
ð1Þ ð2Þ
configuration and the line elements d Z and d Z as follows:
ð1Þ ð1Þ ð2Þ ð2Þ
C 1 D 2
d Z ¼ d Z ; 0 ; d Z ¼ 0; d Z : ð3:5:8Þ
o oxj otj
s_ Cj ¼ ¼ A
: ð3:5:10Þ
ot oZ C oZ C
If we now observe (3.5.3) this turns into:
The coefficients KDC can be related to the surface metric—cf., Eqs. (2.7.7)—and
derivatives of the tangent vectors, which are orthogonal to the normal:
oej osk o2 x k
C
¼ gDR ek RC sDj ¼ gDR ek C R sDj ¼ gDR bCR sDj : ð3:5:14Þ
oZ oZ oZ oZ
This result and application of the chain rule allows us to rewrite Eq. (3.5.11):
!
o t D oZ D o2 x ot?
j k j
s_ C ¼ A
þ t R
s D þ A
ej gDR bCR sDj t ?
oZ C oxk oZ C oZ R A oZ C A
ð3:5:15Þ
ot?
¼ t D;C sDj þ A C ej gDR bCR sDj t ? :
A oZ A
We are now in a position to evaluate the second time derivative of Eq. (3.3.6)4:
ZZ ZZ
d ð1Þ ð2Þ
~ Z i ; t dS ¼ d
w ~ Z i ; t ijk ei s j sk d Z 1 d Z 2
w 1 2
dt A dt A
AðtÞ A0
ZZ o w ZZ
ð1Þ ð2Þ
¼ A
dS þ w ijk ei s_ 1j sk2 þ s1j s_ k2 d Z 1 d Z 2
ot A
Að t Þ A0
ZZ o w ZZ ffi ð1Þ ð2Þ
¼ A
dS þ w ijk t D;D gDR bCR t ? s1j sk2 d Z 1 d Z 2 :
ot A A A
Að t Þ A0
ð3:5:16Þ
In order to arrive at this result use has been made of the fact that whenever
symmetric expressions, such as ei ej or sCj skD , are multiplied by the antisymmetric
tensor ijk the result is simply zero.5 If we now compare the second term in the
second integral with the definition (2.7.12) of the mean curvature we finally arrive
at Eq. (3.5.1).
4
A0 refers to the area of the singular surface in the reference configuration.
5
For a proof one should simply expand the products.
70 3 Balances (in Particular in Cartesian Systems)
We now insert the equations for the transport theorems of volumes and surfaces
from Sects. 3.4 and 3.5 into Eq. (3.3.6) and separate quantities that refer to the
subvolumes V and their surfaces A as well as to the singular surface A:
ZZZ o w ZZ
ZZZ
V
dV þ w t þ u n dA p þ s dV
ot V A V V
V þ [V Aþ [A V þ [V
0 1
ZZ o w ffi
¼ @ A
þ w t D;D 2Km t ? þ u D;D ½½wV tA e p þ s A dA:
ot A A A L A A
A
ð3:6:1Þ
In the process the following integral theorem was used in which another
covariant derivative occurs:
I ZZ
u m dl ¼ u D;D dA since u i ¼ u D siD þ u ? ei and mi ¼ m R siR : ð3:6:2Þ
L L L L L
L S
We postpone its proof and evaluation until the concept of a covariant derivative
has been established in detail. Eq. (3.6.1) represents the most general balance for
the situation depicted in Fig. 3.1. Of course, this relation may also be applied to a
(material) point and its volume dV. This way we obtain local balances for fields.
From a mathematical point of view they are easier to treat, namely in terms of
partial differential equations and jump conditions at boundaries. However, we have
to distinguish between material points within the regular volume and on the sin-
gular surface.
In the limit of an infinitely small volume we conclude that the integrand must
vanish:
3.7 General Balances in Regular and Singular Points 71
ow
V o
þ w tj þ / j ¼ p þ s : ð3:7:2Þ
ot oxj V A V V
This is the general local balance in regular points of a body. If we recall GAUSS’
theorem in its invariant form (3.4.8) we may write alternatively:
o w
V
þr w tþ / ¼ p þ s: ð3:7:3Þ
ot V A V V
This is an aesthetically pleasing equation, but not much more. In contrast to the
form shown in Eq. (3.7.2) it is not very useful during calculations, especially if a
specific skew curvilinear coordinate system is concerned. We shall learn in Chap. 4
how to specify the del operator and the covariant derivative in arbitrary coordinates.
For the time being we content ourselves with the explicit Cartesian way of writing
shown in Eq. (3.7.2).
In order to derive the corresponding equation for the case of a singular point we
consider the limit process shown in Fig. 3.6: A point on the singular surface is
enclosed from both sides by a pillbox. We apply the general balance (3.6.1) to this
very volume and, in a first step, shrink its height to zero. Then all the volume
related contributions in the equation will vanish. In a second step the surfaces A
and A will be contracted into a point so that the normals n and e are collinear.
From the remaining surface contributions we obtain the general local balance in
singular points which, for obvious reasons, is also know in the trade as the jump
condition:
ow ffi hh ffi ii
A D D p
þ w t ;D 2Km t ? þ u ;D þ s ¼ e wV t t ? e þ uA :
ot A A A L A A A
ð3:7:4Þ
A
e
72 3 Balances (in Particular in Cartesian Systems)
In Sect. 3.2 the global balances for mass and momentum have already been
introduced, at least for the case of a material volume without a singular surface.
We now use Eqs. (3.2.2 / 3.2.10) in combination with the transport theorem
(3.4.12)1 and obtain:
ZZZ
ZZZ
oq o oq
þ qtj dV ¼ 0 , þ r ðq tÞ dV ¼ 0 ð3:8:1Þ
ot oxj ot
V ðt Þ V ðt Þ
and:
ZZZ
oqti o
þ qti tj rji qfi dV ¼ 0 ,
ot oxj
V ðt Þ
ZZZ
ð3:8:2Þ
oq t
þ r ðq t t rÞ q f dV ¼ 0:
ot
V ðt Þ
If we observe the general balance structure shown in Eqs. (3.4.1 / 3.4.32) the
balances of mass and momentum read in regular points:
oq o oq
þ qtj ¼ 0 , þ r ð q tÞ ¼ 0 ð3:8:3Þ
ot oxj ot
and:
oqti o oq t
þ qti tj rji ¼ qfi , þ r ðq t t r Þ ¼ q f : ð3:8:4Þ
ot oxj ot
Equation (3.8.3) is also known as continuity equation in the literature. We shall
now present a few alternative forms of these equations. Application of the product
rule to (3.8.3)1 yields:
oq oq otj
þ tj þ q ¼ 0: ð3:8:5Þ
ot oxj oxj
If we switch between Eulerian and Lagrangian variables as outlined in Sect. 3.4
the first two terms can be combined as follows:
dq d^ qðx; tÞ o^
qð~xðX; tÞ; tÞ o^ qðx; tÞ o~xj ðX; tÞ
¼ ¼ þ : ð3:8:6Þ
dt dt ot ox x j ot
t X
mass density of Eq. (3.8.6) is just an example. Frequently material time derivatives
are also denoted by a dot:
dð Þ
ðÞ : ð3:8:8Þ
dt
In fact, we have already tacitly used this form in context with Eq. (3.4.25).
Mathematically speaking the sequence of relations shown in Eq. (3.8.6) is nothing
else but forming a total differential w.r.t. the variables x and t. Without any
reference to Eulerian or Lagrangian coordinates we could have written instead:
oqðx; tÞ oqðx; tÞ
dqðx; tÞ ¼ dt þ dxj : ð3:8:9Þ
ot x oxj t
‘‘Division’’ by dt yields:
dq oq oq
¼ þ tj ; ð3:8:10Þ
dt ot oxj
since velocity is nothing else but a change of position with time. However, note
that Eq. (3.8.6) adds another flavor and, maybe, allows for a more subtle under-
standing of motion of matter both from the view of the comoving as well as of an
external observer. Furthermore note that in the case of a Lagrangian representation
we have:
dq d~ qðX; tÞ o~ qðX; tÞ
¼ ; ð3:8:11Þ
dt dt ot X
i.e., total (or material) and partial time derivative are identical.
Application of the product rule to Eq. (3.8.4)1 while observing Eq. (3.8.3)
yields:
dti orji dt
q ¼ þ q fi , q ¼ r r þ q f: ð3:8:14Þ
dt oxj dt
This equation is the epitome of NEWTON’s law of motion, which in high school
physics is frequently summarized by the slogan ‘‘force equals mass times accel-
eration.’’6 Note that dt=dt is nothing else but the acceleration, a, and the forces are
decomposed into surface and volumetric parts, ðr rÞ and ðq f Þ, respectively.
Now that we have introduced the material time derivative, a reflection on the
global balance equation in context with REYNOLDS’ transport theorem is in order: It
is sometimes required to consider a balance for a specific density w instead of the
volume density w. Clearly w ¼ qw holds and so we rewrite (3.3.7) for the case of
V V
material volumes without a singular surface:
ZZZ ZZ ZZZ
d
qw dV ¼ / n dA þ p þ s dV: ð3:8:15Þ
dt A V V
V oV V
Since q dV ¼ dM and the total mass M of the volume is time independent (in
contrast to V ðtÞ) we may circumvent REYNOLDS’ transport theorem by a simple
substitution:
ZZZ Z Z Z
d d dw
qw dV ¼ w dM ¼ dM ¼ qw_ dV: ð3:8:16Þ
dt dt dt
V M M M
qw_ ¼ r / þ p þ s : ð3:8:17Þ
A V V
6
From the viewpoint of balance laws it would make more sense to say ‘‘mass times acceleration
equals force’’ and to distinguish strictly between cause and effect as NEWTON did (see Chap. 8).
3.8 Local Balances of Mass and Momentum in Regular Points 75
The balance of momentum shown in Eq. (3.8.14) is the origin of two other
mechanical concepts: kinetic energy and moment of momentum. In this section we
shall look into kinetic and other forms of energy relevant to continuum physics.
For this purpose we transform the vector equation (3.8.14)1 by scalar multiplica-
tion with ti into a scalar relation:
dð12ti ti Þ o rji ti oti
q ¼ rji þ q f i ti ,
dt oxj oxj
ð3:9:1Þ
dð12t2 Þ
q ¼ r ðr tÞ r : rt þ q f t:
dt
In the latter formula we have introduced the so-called double scalar product
‘‘:’’ for a neighboring index. It is defined in a very specific way by double sum-
mation (see the indices i and j in the previous equation where i is the neighboring
76 3 Balances (in Particular in Cartesian Systems)
index).7 The result is now integrated w.r.t. the (regular) material volume V(t) and
GAUSS’ theorem is observed:
ZZZ ZZ
dð12ti ti Þ
q dV ¼ rji ti nj dA
dt
V ðt Þ oV
ZZZ ZZZ ð3:9:2Þ
oti
rji dV þ q fi ti dV:
oxj
V ðtÞ V ðt Þ
The integral on the left hand side is transformed by using the substitution
already mentioned in context with Eq. (3.8.16): Recall that the mass of a material
body does not change [see Eq. (3.2.2)], in other words it is time-independent in
contrast to its volume. Also recall that dM ¼ q dV. In context with Eq. (3.8.16)
this leads to:
ZZZ ZZ ZZZ
d 1 oti
qti ti dV ¼ rji ti nj dA rji dV
dt 2 oxj
V ðt Þ oV V ðt Þ
ZZZ ZZZ
d q 2
þ q fi ti dV , t dV ð3:9:3Þ
dt 2
V ðtÞ V ðtÞ
ZZ ZZZ ZZZ
¼ n r t dA r : rt dV þ q f t dV:
oV V ðt Þ V ðtÞ
Thus we have found another balance equation, namely one for the kinetic
energy, whose volume density is given by q2t2 .8 This equation is known as the
work-energy equation in the mechanics community. It states that the (temporal)
change of kinetic energy of a system is equal to the sum of the power provided by
the forces applied to the body either on its surface or to its volume (the first and the
last term after the equal sign) minus the power loss due to non-conservative forces,
i.e., friction (the second term). If we make use of the jargon established in context
with balance equations and the remarks of Sect. 3.2 we may say alternatively that
the temporal change of kinetic energy is given by the sum of the non-convective
flow of kinetic energy across the surface and a corresponding volumetric supply,
namely [according to CAUCHY’s theorem of Eq. (3.2.6)]:
ZZ ZZZ ZZ ZZZ
n r t dA þ q f t dV t t dA þ q f t dV: ð3:9:4Þ
oV V ðt Þ oV V ðt Þ
7
Frequently the stress tensor is symmetric and then we may as well write for the production term
rij oti =oxj r rt where the double scalar product for non-neighboring indices has been
used, which will be introduced in Eq. (3.10.1).
8
It is easily verified that the specific kinetic energy (energy per unit mass) is given by t2 =2.
3.9 Local Balances of Energy in Regular Points 77
9
Note that this convention was not applied to the traction vector, t, probably for historic reasons.
78 3 Balances (in Particular in Cartesian Systems)
du oqj oti du
q ¼ þ rji þ qr , q ¼ r q þ r : rt þ q r: ð3:9:7Þ
dt oxj oxj dt
Just like kinetic energy internal energy is not conserved either. Except for the
sign the corresponding production is equal to the production of kinetic energy.
In context with Eqs. (3.9.5–3.9.7) a caveat is in order: They hold in this (rel-
atively simple) form only if the matter in the material volume does not possess any
intrinsic moment of momentum, a.k.a. spin. We proceed to discuss this subtle
point further below. At the end of this section it is only fair to remember that the
development of the concept of various energy forms as well as the idea of energy
conservation cost mankind several centuries of contemplation, intellectual strug-
gle, as well as mutual personal animosity. The first ideas of momentum versus
kinetic energy, or vis viva go back to LEIBNIZ. Then the industrial revolution took
over, and it became important to think about energy generation as well as energy
conversion. In this context mankind expected several scientists to do their duty,
James Prescott JOULE, Robert Wilhelm MAYER, and Hermann von HELMHOLTZ, to
mention just a few of them.
Julius Robert (von) MAYER was born on November 25, 1814 in Heil-
bronn (Germany) and he also died there on March 20, 1878. He was a
German physician and physicist, known for his pioneering work in the
establishment of the mechanical equivalent of heat. MAYER studied
medicine at the University of Tübingen. In 1842 he published a paper in
the journal Annales de Chimie, in which he gave a value for the
mechanical equivalent of heat. His figure was based on the rise of
temperature in paper pulp that was stirred by a horse-powered mech-
anism. MAYER was also the first to state the principle of conservation of
energy, most notably for biological phenomena as well as for physical
systems. He was fascinated by the concepts of heat and the conversion
of thermal and mechanical energy. Being a physician by trade, he
measured temperatures whenever he could in order to find evidence for his archaic ideas about
energy and heat: During a trip to Java he has ample opportunity to take the temperatures of his
patients. He states his findings (in a nowadays politically extremely incorrect language) as
follows: ‘‘for a negro, lazy and idly laying in the cabana 37; the same, however, sitting idly in
the sun 40.20; the same, however, working in the sun 39.75.’’ From this fascinating obser-
vation he concludes that heat is converted to mechanical work. However, his contemporaries, in
particular JOULE, thought nothing of him and his findings. So he was at best ignored if not
belittled by the scientific community. This did not really add to his psychological well-being:
He attempts to commit suicide in 1850 after the sudden death of two of his children. In those
days society was not too sensitive or even patient with the mentally ill, and so he is sent to a
lunatic asylum right away. He returns to Heilbronn three years later, now truly broken. How-
ever, in his later years he is finally given credit for his work. In 1867 the local king awards him a
medal, the Ritterkreuz 1st Class, which came with the privilege of a nobility title. From now on
MAYER could call himself VON MAYER. In fact, this is an enormous advantage for the bearer of
such an ordinary German name, as the author of this book can tell by personal experience.
3.9 Local Balances of Energy in Regular Points 79
ð3:10:2Þ
In engineering mechanics the vector products n r x x n r x t and
x q f on the right hand side are known as moments of forces or just moments, for
short. More specifically the first expression represents the moment of tractions and
the second one the moment of volume force density. Therefore the vector product
x q t on the left side of the equation must consistently be referred to as moment
(density) of linear momentum. It is sometimes—imprecisely—also called angular
momentum. It obviously obeys a balance equation and, as known from engineering
mechanics, its temporal change is dictated by the moments exerted by the forces.
However, we also see that it is not a conserved quantity: The balance contains a
80 3 Balances (in Particular in Cartesian Systems)
production term, which vanishes only if the stress tensor is symmetric so that
ijk rjk ¼ 0. Indeed, the stress tensor is symmetric for most engineering materials
and applications. In fact, we have tacitly assumed symmetry when we studied
MOHR’s circle in Exercises 2.4.6 and 2.6.3. However, there are materials which
have an ‘‘intrinsic moment of momentum,’’ invisible in terms of force couples
applied at certain distances, at least on the continuum scale. This additional internal
degree of freedom is known as spin. It is also a vector denoted by the symbol
s. Liquid crystals may serve as an example of materials with spin. The situation is
similar to the case of energy, where it is possible to convert macroscopically visible
kinetic energy into microscopic motion, i.e., a change of temperature or internal
energy: Spin can be converted into macroscopically visible angular momentum or
vice versa. Now total energy, i.e., the sum of internal energy and (translational as
well as intrinsic rotational) kinetic energy, is conserved and so is total angular
momentum, which is the sum of moment of momentum and spin. We write:
ZZZ ZZ
d
q si þ ijk xj tk dV ¼ mli þ ijk xj rlk nl dA
dt
V ðt Þ oV
ZZZ
þ q ijk tj fk þ li dV ,
V ðt Þ
ZZZ ZZ ð3:10:3Þ
d
qðs þ x tÞ dV ¼ n ðm þ r xÞ dA
dt
V ðtÞ oV
ZZZ
þ qðx f þ lÞ dV:
V ðt Þ
The vector s denotes the specific spin, m is the so-called surface couple-stresstensor,
and l refers to the vector of specificspinsupply, a.k.a. specific body couple density.
The local form of the balance of total angular momentum in regular points reads:
d si þ ijk xj tk o mli þ ijk xj rlk
q ¼ þ q ijk tj fk þ li ,
dt oxl ð3:10:4Þ
dð s þ x t Þ
q ¼ r ðm þ r xÞ þ qðx f þ lÞ:
dt
This settles a red-herring discussion frequently started in the mechanics com-
munity10: Is the principle of angular momentum independent of NEWTON’s law of
motion or not? The answer is very clear and simple: Yes and no! No, if we believe
10
The argument was started by two eminent mechanics professors with a strong disposition and
admiration for mathematics and clearly geared toward libeling physicists as numbskulls: see the
paper (in German) and books by Truesdell [18, 19] as well as the book (also in German) by Szabó
[16]. Until today many mechanics professors join the clamor of the Boeotians in a sycophant
manner even without being able to give an explanation of what the problem really is.
3.10 Local Balances of Angular Momentum 81
where i denotes a factor converting mass into weight. In fact these three formulae
are nothing else but Eq. (3.2.10) if we only remind ourselves that the notion of a
stress tensor was introduced fifty years later by CAUCHY. But this is just the first set
of EULER’s laws of mechanics and, indeed, he emphasizes that moments have to be
considered as well: ‘‘§. 28. Cum igitur
elemento d M, quod in puncto z concipi-
mus, primo applicata fit vis ¼ d M dddt2x ’’ secundum directionem IA agens, ex ea
nullum nascitur momentum pro hoc axe; … unde pro axe IA summa omnium
momentorum erit
Z ffi Z ffi
ddy ddz
þ zd M 2
yd M 2
¼ i S ’’:
dt dt
11
Scattering experiments make some particle physicists believe that the electron is a true point.
However, it does have a (quantized) spin of ±h, h ¼ 1:055 1034 Js being the normalized
PLANCK constant (note the units of moment of momentum), which could easily be interpreted in
terms of moment of momentum if the electron were only a rotating distributed mass.
82 3 Balances (in Particular in Cartesian Systems)
The other two components are deduced similarly and we conclude (up to the
sign) that this second set of EULER’s laws of mechanics agrees indeed with
(3.10.2)2 if we do not explicitly specify the moments in terms of forces just like
EULER did. He only has to say the following about them in §. 27.: ‘‘… quamobrem
designemus ista momenta, quae ex omnibus viribus sollicitantibus pro ternis ax-
ibus IA, IB, IC nascuntur, litteris S, T, V, ita ut his quantitatibus per i multiplicatis
summae omnium momentorum elementarium, quas singulae vires acceleratrices
suppeditant aequari debeant.’’ This concludes his argument and, to a certain
degree, he makes it sounds like a conclusion from the balance of linear momentum
if it were not for two things: First, he does not link the moments to the applied
forces in an explicit manner and, second, he puts all of his six equations inde-
pendently side-by-side: ‘‘§. 29. Hac igitur ratione sex nacti sumus aequationes,
quas hic coniunctim conspectui exponamus
R R R
I. d M dddt2x ¼ i P IV. z d M dddt2y y d M dd dt2z ¼ i S
R d d y R R
II. d M d t2 ¼ i Q V. x d M dd dt2z z d M dddt2x ¼ i T
R R R
III. d M d t2 ¼ i R
ddz
VI. y d M dddt2x z d M dddt2y ¼ i U’’.
It is also interesting to note that the typo in the last equation12 is not mentioned
in the pertinent literature. Rather it was tacitly corrected in Szabó [16], p. 30.
After these philosophical remarks we now subtract the balance for the moment
of momentum from the balance for the total angular momentum and obtain the
balance of spin for a material volume:
ZZZ ZZ ZZZ ZZZ
d
qsi dV ¼ mli nl dA þ ijk rjk dV þ qli dV ,
dt
V ðtÞ oV V ðt Þ V ðt Þ
ZZZ ZZ ZZZ ZZZ ð3:10:5Þ
d
qs dV ¼ n m dA þ r dV þ q l dV:
dt
V ðt Þ oV V ðt Þ V ðtÞ
12
The typo actually appears twice in §. 28. and §. 29. of EULER’s work so that we may suspect
that it was incorrectly written down in his personal notes.
3.10 Local Balances of Angular Momentum 83
Recall the general form of a balance equation in regular points shown in Eq.
(3.7.2). Table 3.1 makes it easy to reconstruct all of the balances shown so far.
The table needs two more lines of entry, namely for the balances of entropy as
well as of electric charge. We will get back to that in Chaps. 12 and 13. It should
also be noted that the entries for the various types of energy are valid for bodies
without intrinsic moment of momentum, i.e., spin. We will reconsider them in
Chap. 8 after angular velocity has been introduced as a kinematic quantity.
In contrast to Sect. 3.10 we start with a table which, if used in context with Eq.
(3.7.4), directly leads to the balances for mass, momentum, energy, and angular
momentum in singular points.
Note that, a singular surface is basically a mathematical model for a transition
zone in a volume showing a very steep gradient. It is sometimes necessary to
assign intrinsic properties to this structure. Therefore most of the entries in
84 3 Balances (in Particular in Cartesian Systems)
Table 3.2 are in perfect analogy to the entries in the previous table showing
volume properties.13 Examples are a mass density, q, as well as related mechanical
A
properties of the singular surface, such as momentum, kinetic energy, or moment
of momentum. It can also have its own internal energy (‘‘temperature’’) or spin.
Such properties may become important when modeling soap bubbles or (more
engineering-like) rubber membranes. However, the transition between the wall of
a pressure vessel and the surrounding gas or fluid is very steep and not associated
with any mass.
A few other surface properties are less intuitive and, therefore, deserve a
comment. For example, as we shall see soon, the jump condition for the
momentum dictates certain requirements regarding the continuity of the stress
tensor. In Chap. 9 we will discuss which components are affected when we speak
about boundary and interface conditions. Moreover, note that r is known as the
L
tensor of surface tensions. Recall that the production terms for kinetic and internal
energy can be derived by scalar multiplication of the balance of momentum and
suitable rearrangement. The situation is analogous to the procedure outlined in
context with Eq. (3.9.1) for the volumetric kinetic energy. The production terms
for spin and moment of momentum become clear after vector multiplication of the
balance of momentum similarly as in context with Eq. (3.10.1).
As a specific example of how to use Table 3.2 we consider the balance of mass.
The various entries lead to:
oq ffi
A
þ q t a;a 2Km t ? ¼ ½½ qð t tA ? eÞ e: ð3:12:1Þ
ot A A A
If the singular surface has no mass of its own, in other words if q ¼ 0; this
A
simplifies to:
½½qðt tA ? eÞ e ¼ 0: ð3:12:2Þ
If the singular surface does not move, in other words if t ? ¼ 0, we obtain:
A
þ
½½qt e ¼ 0 , ðqtÞ e ¼ ðqtÞ e: ð3:12:3Þ
Intuitively speaking, this means that matter entering from one side has to leave
the singular surface on the other: Mass cannot simply disappear. This is another
possible version and interpretation of the equation of continuity. If the singular
surface moves with the surrounding matter the expression in brackets shown in Eq.
(3.9.2) vanishes and an identity results. Mass is neither entering nor leaving.
13
We have used index notation in Table 3.2 since it makes it easier to distinguish operations
referring to the volume and to the surface, respectively.
3.12
Mass q 0 0 0
A
(linear) momentum qti r Di 0 q f i
AA L A A
2
qA
q u þ12 t2 L L A q f it iþr
A A A A A A A
Moment of momentum q ijk xj t k ikl xk r Dl ijk r Dj s kD q ijk xj f k
A A L L L A A
Spin q si m Di ijk r Dj s kD q li
A A L L L A A
Angular momentum qðs i þijk xj t k Þ m Di þikl xk r Dl 0
A A A L L q ijk xj f k þ l i
A A A
85
86 3 Balances (in Particular in Cartesian Systems)
pþ ¼ p : ð3:12:6Þ
14
In German.
References 87
References
1. Euler L (1775) Nova methodus motum corporum rigidorum determinandi. Novi Commentarii
Academiae Petropolitanae, pp 208–238
2. Greve R (2003) Kontinuumsmechanik: Ein Grundkurs für Ingenieure und Physiker. Springer,
Berlin
3. Haupt P (2002) Continuum mechanics and theory of materials, 2nd edn. Springer, Berlin
4. Irgens F (2008) Continuum mechanics. Springer, Berlin
5. Liu I-S (2010) Continuum mechanics. Springer, Berlin
6. McBride AT, Javili A, Steinmann P, Bargmann S (2011) Geometrically nonlinear continuum
thermomechanics with surface energies coupled to diffusion. J Mech Phys Solids
59:2116–2133
7. Moeckel GP (1974) Thermodynamics of an interface. ARMA 57:255–280
8. Müller I (1973) Thermodynamik Die Grundlagen der Materialtheorie. Bertelsmann
Universitätsverlag, Düsseldorf
9. Müller I (1985) Thermodynamics. Pitman Advanced Publishing Program, Boston
10. Müller I (1994) Grundzüge der Thermodynamik mit historischen Anmerkungen, 1st edn.
Springer, Berlin
11. Müller WH, Ferber F (2008) Technische Mechanik für Ingenieure, 4. aktualisierte Auflage,
Carl Hanser, München
12. Müller I, Müller WH (2009) Fundamentals of thermodynamics and applications. Springer,
Berlin
13. Müller WH, Muschik W (1983) Bilanzgleichungen offener mehrkomponentiger Systeme I:
Massen- und Impulsbilanzen. J Non-Equilib Thermodyn 8:29–46
14. Muschik W, Müller WH (1983) Bilanzgleichungen offener mehrkomponentiger Systeme II:
Energie und Entropiebilanz. J Non-Equilib Thermodyn 8:47–66
15. Schade H (1970) Kontinuumstheorie strömender Medien. Springer, Berlin
16. Szabó I (1977) Geschichte der mechanischen Prinzipien. Birkhäuser, Basel
17. Truesdell C, Toupin R (1960) The classical field theories. In: Flügge S (ed) Encyclopedia of
physics, vol III/1, Principles of classical mechanics and field theory. Springer, Berlin,
Göttingen, Heidelberg
18. Truesdell C (1968) Whence the law of moment of momentum. In: Essays in the history of
mechanics. Springer, Berlin
19. Truesdell C (1969) Rational thermodynamics. McGraw-Hill, New York
Chapter 4
Spatial Derivatives of Fields
The general balance equations (3.7.3/3.7.4) in combination with Tables 3.1 and 3.2
clearly show that it is necessary to investigate how to calculate spatial derivatives
of scalar fields like mass density, vector fields, such as velocity and, finally, tensor
fields like stress in arbitrary coordinate systems.
We start with the gradient of an arbitrary scalar field f ¼ ~f ðxi Þ ¼
ðxÞ ðxÞ
~f ðxi ðz j ÞÞ ¼ ^f ðz j Þ1 w.r.t. a Cartesian coordinate system. By means of the chain
ðxÞ ðxÞ
o ~f ðxi Þ o ^f ðz j Þ of of
ð xÞ ozk ð xÞ ð xÞ ozk ðxÞ
¼ or ¼ for short: ð4:1:1Þ
oxi oxi ozk oxi oxi ozk
Scalar functions are independent of the coordinate frame, and therefore:
^f z j ¼ ^f z j or f ¼ f : ð4:1:2Þ
ðxÞ ðzÞ ðxÞ ðzÞ
transforms like the components of a covariant vector field. This quantity is called
the gradient of the scalar field f.2
0 l 1
o Ai o Ai ffi oA
ðxÞ ozk ðxÞ ozk o oxi l ozk oxi @ ðzÞ ozl o2 xs n A
¼ ¼ A ¼ þ A : ð4:2:1Þ
oxj oxj ozk oxj ozk ozl ðzÞ oxj ozl ozk oxs ozk ozn ðzÞ
In this equation use was made of the chain rule, the product rule, as well as Eq.
(2.4.9). On this basis we may write:
oAl oAi
ðzÞ ozl o2 xs n ozl oxj ðxÞ
þ A ¼ : ð4:2:2Þ
ozk oxs ozk ozn ðzÞ oxi ozk oxj
If we compare this result with Eq. (2.4.15) we must conclude that the quantity
on the left hand side, i.e.:
2
Note that the symbol f for the scalar field does not contain any information about the coordinate
system that was used. It is to be understood in an absolute sense just like the symbol A denotes an
absolute vector. The contravariant components of the gradient in Eq. (4.1.3) can be obtained by
multiplication with the metric glk.
4.2 Spatial Derivatives of Vector Fields 91
oAl
l ðzÞ ozl o2 xs
A ;k ¼ þ Clkn A n ; Clkn ¼ ð4:2:3Þ
ðzÞ ozk ðzÞ oxs ozk ozn
form the components of a mixed tensor field, namely the gradient of the vector
A. The quantity A l ;k defined by Eq. (4.2.3) is also known as the covariant derivative
ðzÞ
of the contravariant components of the vector A. The quantities Clkn are known as
CHRISTOFFEL symbols. Note that they are symmetric w.r.t. the lower indices.
sinhð2z1 Þ
C111 ¼ C122 ¼ C212 ¼ C221 ¼ ;
coshð2z1 Þ cosð2z2 Þ
ð4:2:8Þ
sinð2z2 Þ
C112 ¼ C121 ¼ C211 ¼ C222 ¼ :
coshð2z1 Þ cosð2z2 Þ
o2 zn oxi ozn o2 xi
þ ¼ 0: ð4:2:10Þ
ozk oxi ozt oxi ozk ozt
4.2 Spatial Derivatives of Vector Fields 93
We conclude that:
oAl oAi
ðzÞ oxi oxj ðxÞ
Ctkl A t ¼ : ð4:2:11Þ
ozk ðzÞ ozl ozk oxj
If we compare this with Eq. (2.4.15) we must conclude that the expression
oAl
ðzÞ
A l;k ¼ Ctkl A t ð4:2:12Þ
ðzÞ ozk ðzÞ
transforms like the components of a covariant tensor field. This is just another
representation of the gradient of A. It is also known as the covariant derivative A l;k
ðzÞ
A l ;l ¼ r A or A l ;l ¼ divA: ð4:2:16Þ
ðzÞ ðzÞ
94 4 Spatial Derivatives of Fields
Recall the result from Sect. 4.1 according to which we may write for the gradient
of an arbitrary scalar field f ¼ f f w.r.t. a Cartesian and an arbitrary skew
ðxÞ ðzÞ
of ozk of oð Þ
rf ¼ k
ei ¼ ei ) rðÞ ¼ ei : ð4:3:5Þ
oz oxi oxi oxi
The del operator, which allowed to form the gradient of a scalar, can now be
used to obtain the LAPLACE operator when applied to a scalar. In Exercise 4.2.7 it
was shown that the LAPLACE operator can be identified as the divergence of a
contravariant vector field. In Cartesian coordinates we may write:
ffi
o of o2 f
r ðrf Þ ¼ ei ej ¼ ei ej
oxi oxj oxi oxj
ð4:3:6Þ
o2 f o2 f
¼ dij ¼ :
oxi oxj oxi oxi
3
In absolute notation it is also customary to write grad f instead of rf.
96 4 Spatial Derivatives of Fields
Note that the differentiation of the base vectors ei yields zero. Consequently we
obtain the well known result in Cartesian coordinates for three dimensions:
o2 ðÞ 1 oðÞ 1 o2 f o2 f
DðÞ ¼ þ þ 2 2þ 2: ð4:3:11Þ
or 2 r or r o# oz
4
or div grad ().
4.3 Invariant Notation of Spatial Derivatives of Scalar Fields 97
The product rule and the definition (4.2.3)2 for the CHRISTOFFEL symbols were
used during the various transformations. It follows that:
oBnp oB
ðzÞ oxk ozn ozp ðxÞij
þ Cnlr Brp þ Cplr Bnr ¼ l : ð4:4:2Þ
ozl ðzÞ ðzÞ oz oxi oxj oxk
We conclude in perfect analogy to the transformation rule for tensors of second
order, see Eq. (2.4.15), that the combination
o B np
ðzÞ
B np ;l ¼ þ Cnlr B rp þ Cplr B nr ð4:4:3Þ
ðzÞ ozl ðzÞ ðzÞ
represents the components of a mixed tensor of third order, namely the gradient of
the second order tensors B. We may also say that this equation represents the
covariant derivative of the contravariant components of the second order tensor
B. Note that it is also customary in mathematics to denote the partial derivatives in
such equations by a comma:
p
Bnp ;l ¼ Bnp ;l þ Cnlr Brp þ Clr Bnr : ð4:4:4Þ
ðzÞ ðzÞ ðzÞ ðzÞ
This way of writing shows very nicely, which corrections are required to turn a
partial derivative into a covariant one or, in other words, how to create an invariant
tensor expression. The corrections are given by the CHRISTOFFEL symbols which are
suitably combined with both indices of the second order tensor. The ‘‘corrections’’
vanish for ‘‘straight’’ Cartesian coordinate systems. The equation also implies a
hierarchy of derivatives. According to the results of the previous sections we may
write for scalars (i.e., zero order tensors), vectors (i.e., first order tensors), and
tensors that are higher than second order:
98 4 Spatial Derivatives of Fields
f ;l ¼ f ;l ;
ðzÞ ðzÞ
n
A ;l ¼ A n ;l þ Cnlr A r ; ð4:4:5Þ
ðzÞ ðzÞ ðzÞ
p q
C npq ;l ¼ C npq ;l þ Cnlr C rpq þ Clr C nrq þ Clr Cnpr ; etc:
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
This is a consequence of the product rule and the definition (4.2.3)2 for the
CHRISTOFFEL symbols, just like Eq. (4.4.1). We conclude that:
oBnp oB
ðzÞ oxk oxi oxj ðxÞij
Crnl Brp Crpl Bnr ¼ l n p : ð4:4:7Þ
ozl ðzÞ ðzÞ oz oz oz oxk
In full analogy to Eq. (2.4.17) we realize that the combination
oBnp
ðzÞ
Bnp;l ¼ Crpl Bnr Crnl Bpr ð4:4:8Þ
ðzÞ ozl ðzÞ ðzÞ
Finally note that the following formulae hold for the covariant derivative of a
mixed tensor:
4.4 Spatial Derivatives of Tensors 99
o B np
ðzÞ
B n p;l ¼ þ Cnlr B r p Crpl B n r ;
ðzÞ ozl ðzÞ ðzÞ
ð4:4:10Þ
o B pn
ðzÞ
B p n ;l ¼ Crpl B r n þ Cnlr B p r
ðzÞ ozl ðzÞ ðzÞ
As before each index gets ‘‘its own’’ CHRISTOFFEL symbol. These ‘‘correction
terms’’ come with a plus sign for contravariant indices and a minus sign in case of
covariant ones.
Moreover, note that Eqs. (4.4.2) and (4.4.3) imply that B nl ;l are the components
ðzÞ
Exercise 4.4.2: The covariant derivative for mixed tensors of second order
Provide a proof of Eq. (4.4.10), first, in analogy to the sequence of
transformation steps shown in Eqs. (4.4.1/4.4.2) and, second, quasi in an
indirect manner, by application of the rules for raising and lowering indices
by means of the metric (see Sect. 2.4) when applied to Eqs. (4.4.3/4.4.8) and
by observing (4.4.9).
or: ffi
i j
t r k ¼ t i ;l r j k þ t i r j k;l : ð4:4:14Þ
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
;l
Recall the proof of GAUSS’ theorem from Exercise 3.4.1. It was based on Cartesian
coordinates and finally led to Eq. (3.4.2). Then this equation was rewritten in
absolute tensor form, Eq. (3.4.6), by means of the del operator. However, we never
clarified the details and explained how to transform GAUSS’ theorem in co-/con-
travariant notation (say) and, to begin with, we also never explained how the
volume element dV or the surface element dA are evaluated in skew curvilinear
coordinates. The solution is as follows. First, the volume element:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dV ¼ det gij dV ; dV ¼ dz1 dz2 dz3 : ð4:5:1Þ
ðzÞ ðzÞ
The proof is related to the arguments presented in context with Eq. (3.4.22).
The volume element in curvilinear coordinates is given by a triple product5 of
three line segments:
ð1Þ
ð2Þ ð3Þ
oxi oxj oxk ð1Þ ð2Þ ð3Þ
dV ¼ d x d x d x ijk l m n d z l d z m d z n ð4:5:2Þ
oz oz oz
5
The issue how the volume element transforms during reflections is not addressed in this
section. See Sect. 8.4 for more details.
4.5 Integral Theorems Revisited 101
ox
dV ¼ ijk 1 2 3 dz1 dz2 dz3 det ki d V
oz oz oz oz ðzÞ
rffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffi
By expansion it follows that the three components of the surface element vector
in Cartesian coordinates can be expressed in terms of the skew curvilinear coor-
dinates by three JACOBI determinants:
ffi
ox2 ox3 ox3 ox2 oðx ; x Þ
dA1 ¼ 1 oz2
1 2 dz1 dz2 21 23 dz1 dz2 ;
ð xÞ oz oz oz oðz ; z Þ
ffi
ox3 ox1 ox1 ox3 oðx ; x Þ
dA2 ¼ 1 2
1 2 dz1 dz2 31 21 dz1 dz2 ; ð4:5:9Þ
ð xÞ oz oz oz oz oðz ; z Þ
ffi
ox1 ox2 ox2 ox1 oðx ; x Þ
dA3 ¼ dz1 dz2 11 22 dz1 dz2 :
ð xÞ oz1 oz2 oz1 oz2 oðz ; z Þ
102 4 Spatial Derivatives of Fields
We now consider the surface element itself and find by using Eqs. (4.5.8) and
(8.3.35):
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
oxj oxk oxr oxs
dA ¼ ijk 1 2 irs 1 2 dz1 dz2
oz oz oz oz
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð4:5:10Þ
oxr oxr oxs oxs oxs oxs oxr oxr 1 2
¼ dA ¼ dz dz ¼ det g ab d A
oz1 oz1 oz2 oz2 oz1 oz2 oz1 oz2 ðzÞ
where:
These two equations are the direct analogues to the expressions for volume
elements shown in Eqs. (4.5.1). We conclude that:
oðx ; x Þ 2 oðx ; x Þ 2 oðx ; x Þ 2
det gab ¼ 21 23 þ 31 21 þ 11 22 : ð4:5:12Þ
oðz ; z Þ oðz ; z Þ oðz ; z Þ
Use these results to reconfirm the equations for the normal vectors shown
in Eqs. (2.7.15/2.7.18). Finally prove Eq. (4.5.12).
4.5 Integral Theorems Revisited 103
We now assume that the symbol g in Eq. (3.4.3) stands for a scalar f ¼ f ¼ f ;
ð xÞ ðzÞ
ARCHIMEDES was born around 287 in Syracuse, Sicily, and died there around
212 B.C. Rumor has it that he found the law of buoyancy when his king,
HIERO of Syracuse, asked him whether his new crown was pure gold, as
commissioned, or contained an addition of cheap copper or silver. ARCHI-
MEDES measured the volume of the crown by dipping it into water and
watching the rise of the water level and measuring the buoyancy. He
compared that to the case when a piece of gold of the same weight was
immersed. They say that the successful solution of the task made him run-
ning around in the streets, stark naked and shouting, out of joy, Eureka!,
which is Greek for I’ve got it! ARCHIMEDES is considered the greatest engineering scientist of
ancient times. He made some progress with the rectification of the circle and the determination
of the number p. He was also expelled from the Greek Academy of Science when he examined
the volume of geometrical objects by experiment instead mathematically by pure thought. When
Syracuse was conquered by the Romans he was slain by soldiers on the beach while hunched
over some circles drawn in the sand.
104 4 Spatial Derivatives of Fields
We now turn to the pending proof of Eq. (3.6.2). Mutual insertion of the
equations yields:
I I ffi I I
/ m dl ¼ / D siD þ / ? ei m R siR dl ¼ D
/ gDRmR dl ¼ / D mD dl ð4:5:19Þ
L L L L L
L L L L
if we only observe Eq. (2.7.7). The last step in this equation is the basis for a 2D
analogue to Eq. (4.5.16) if the singular part is ignored. We simply identify A r ! /D ;
ðzÞ L
n r ! mD ; A ! L; V ! S and write without any hesitation:
ðzÞ
I ZZ
/D mD dl ¼ /D ;D dA: ð4:5:20Þ
L L
L S
4p
rji d A j ¼ 0; 0; R3 q0 g : ð4:5:22Þ
ð xÞ 3
A
where p0 and q0 denote the pressure and the density on the ground. Use
results from Exercise 4.5.1 and show that the mass of the displaced air is
given by:
4.5 Integral Theorems Revisited 105
ZZZ
Ma ¼ qðR; #; uÞ dV
V
ð4:5:24Þ
4pq0 exp gqp00 h hgq R
gq R
gq R
i
¼
3
0
cosh 0 sinh 0 ;
gq 0
p0 p0 p0
p0
where h denotes the height of the center of the balloon. Calculate now the
total force acting on the balloon surface and prove the surprising result (x3 is
pointing ‘‘upward’’):
ZZ ZZ
rji d A j ¼ pðR; #; uÞ d A i ¼ ð0; 0; Ma gÞ: ð4:5:25Þ
ð xÞ ð xÞ
A A
References
1. Bertram A (2008) Elasticity and plasticity of large deformations, 2nd edn. Springer, Berlin
2. Einstein A (1983) Über die spezielle und die allgemeine Relativitätstheorie. Wissenschaftliche
Taschenbücher 59. 21. Auflage. Vieweg, Braunschweig
3. Eisenhart LP (1947) An introduction to differential geometry with use of the tensor calculus.
Princeton University Press, Princeton
4. Ericksen JL (1960) Appendix. Tensor fields. In: Flügge S (ed) Encyclopedia of physics,
volume 3/1 principles of classical mechanics and field theory. Springer, Berlin
5. Flügge W (1972) Tensor analysis and continuum mechanics. Springer, New York, Berlin
106 4 Spatial Derivatives of Fields
6. Green AE, Zerna W (1968) Theoretical elasticity, 2nd edn. Dover Publications, Inc, New York
7. Itskov M (2007) Tensor algebra and tensor analysis for engineers with applications to
continuum mechanics. Springer, Berlin, New York
8. Liu I-S (2010) Continuum mechanics. Springer, Berlin, New York
9. Schade H, Neemann N (2009) Tensoranalysis, 3. überarbeitete Auflage. de Gruyter. Berlin,
New York
Chapter 5
Balance Equations in Skew Curvilinear
Coordinate Systems
Abstract We now return to the balances of Chap. 3 and rewrite them for arbitrary
coordinate systems. We start with the balances in regular points and, in this
context, with the simplest one, namely the balance of mass, which is a scalar
equation. We then move on to the more complex ones for momentum, energy, as
well as total angular momentum, and specify them for cylindrical and spherical
coordinates. As before we follow both ways and present the balances in index form
as well as symbolically. The chapter ends with a discussion of the jump conditions
and of global balances in arbitrary coordinates.
There are two sides of the balance sheet—the left side and the
right side.
On the left side, nothing is right, and on the right side, nothing
is left !
Answer by UBS to the journalist Dirk MAXEINER after the
resignation of Ingrid MATTHÄUS-MAIER, CEO of KfW Bank
oti
ð xÞ
tl ;l ¼ ð5:1:3Þ
ðzÞ oxi
we finally obtain for the balance of mass in arbitrary coordinates from Eqs. (3.8.3/
3.8.5) in combination with Table 3.1:
oq oq ot j oq oq
ðxÞ ð xÞ ðxÞ ðzÞ k ðzÞ
þt j þq ¼0 ) þt þ q tk ;k ¼ 0 ð5:1:4Þ
ot ðxÞ oxj ð xÞ oxj ot ðzÞ ozk ðzÞ ðzÞ
By using the definition of cylindrical coordinates (2.2.11), Eq. (2.2.13) for the
metric, and the definition (2.6.3) for the physical components of a vector we find
that:
1
tr ¼ r_ ¼ thri ; t# ¼ #_ ¼ th#i ; tz ¼ z_ ¼ thzi : ð5:2:1Þ
r
The definition (4.2.3) for the covariant derivative of contravariant vector
components in combination with the CHRISTOFFEL symbols for cylindrical coordi-
nates shown in Eq. (4.2.5) yields:
dqðr; #; zÞ oq oq dr oq d# oq dz
¼ þ þ þ
dt ot or dt o# dt oz dt
ð5:2:5Þ
oq oq 1 oq oq
¼ þ thri þ th#i þ thzi
ot or r o# oz
if we only observe Eq. (5.2.1). Thus Eq. (5.2.4) turns into:
ffi
dq othri 1 oth#i othzi thri
¼ q þ þ þ : ð5:2:6Þ
dt or r o# oz r
The right hand sight of this equation is nothing else but the divergence of the
velocity written in cylindrical coordinates and it is this very quantity which dic-
tates how the density of a material particle changes in time.
and:
dti
ðzÞ
q ¼ r ji; j þ q f i : ð5:3:6Þ
ðzÞ dt ðzÞ ðzÞ ðzÞ
Note that whenever expressions involving the stress tensor appear in the next
sections we shall not assume that it is necessarily symmetric. The presented
expressions can, of course, be simplified if this assumption is made and non-polar
media are considered. In order to transform the balance of momentum into
cylindrical coordinates we start from the general formula, Eq. (5.3.2), and obtain
by using the equations for the Christoffel symbols (4.2.5) and the relations for the
velocity from Eq. (5.2.1) for the r-component:
ffi
o_r o_r o_r o_r
q þ r_ þ #_ þ z_ r #_ 2 þ
ot or o# oz
rr #r zr
ð5:4:1Þ
or or or ## 1 rr r
þ rr r ¼ qf ;
or o# oz r
for the #-component:
112 5 Balance Equations in Skew Curvilinear Coordinate Systems
!
o#_ o#_ _ o#_ o#_ 2 _
q þ r_ þ # þ z_ þ r_ # þ
ot or o# oz r
ð5:4:2Þ
orr# or## orz# 1 r#
2r þ r# r ¼ qf # ;
or o# oz r
and for the z-component:
ffi
o_z o_z o_z o_z
q þ r_ þ #_ þ z_ þ
ot or o# oz
ð5:4:3Þ
orrz or#z orzz 1 rz
r ¼ qf z :
or o# oz r
r ji ; j ¼ q f i : ð5:5:1Þ
ðzÞ ðzÞ ðzÞ
r ji ; j ¼ 0: ð5:5:2Þ
ðzÞ
In other words: The divergence of the stress tensor vanishes [cp., the remarks in
context with Eq. (4.4.11)]. This is one of the relations that has already been
mentioned in Chap. 1.
If velocities and time derivatives are neglected in Eqs. (5.4.1–5.4.3) or (5.4.4) the
static balance of momentum in (physical) cylindrical coordinates can immediately
be read off. However, as an example of how to use absolute tensor notation we
shall derive these equations differently. First we note that the del operator in
cylindrical coordinates reads:
5.6 Balance of Momentum (Regular Form) 115
oðÞ 1 oð Þ oð Þ
rðÞ ¼ er þ e# þ ez : ð5:6:1Þ
or r o# oz
This is easy to prove either by using Eq. (4.3.3) in combination with Eqs. (2.3.8)
and (2.5.8) or by combining Eqs. (4.3.4) and (4.3.10). This operator must now by
applied to the (symmetric) stress tensor, which we also decompose w.r.t. the unit
base er , e# , ez . Clearly this requires us to use physical coordinates for the tensor
components since the unit base vectors have no units. We write:
rhiji ei ej rhrri er er þ rhr#i er e# þ þ rhzzi ez ez ; ð5:6:2Þ
and must now evaluate r r. However, by doing so, we have to perform the
differentiations very carefully. Note that not only the tensor components need to be
differentiated with respect to r, #, and u but also some of the base vectors. In fact
from Eq. (2.3.8) we must conclude that:
oer oer oer
¼ 0; ¼ e# ; ¼ 0;
or o# oz
ð5:6:3Þ
oe# oe# oe# oez oez oez
¼ 0; ¼ er ; ¼ 0; ¼ 0; ¼ 0; ¼ 0:
or o# oz or o# oz
Thus we expect contributions from differentiations of two unit vectors. Hence
by observing the product rule we finally obtain from Eq. (5.6.2):
ffi
orhrri 1 orh# ri orhzri rhrri rh##i
rr¼ þ þ þ er
or r o# oz r
ffi
orhr#i 1 orh##i orhz#i 1
þ þ þ þ ðrhr#i þ rh# ri Þ e# : ð5:6:4Þ
or r o# oz r
ffi
orhrzi 1 orh#zi orhzzi 1
þ þ þ þ rhrzi ez
or r o# oz r
oðÞ 1 oð Þ 1 oðÞ
r ð Þ ¼ er þ e# þ eu : ð5:6:5Þ
or r o# r sin # ou
Use it and prove that the LAPLACE operator is given by:
DðÞ ¼ r ðrðÞÞ
o2 ðÞ 1 o2 ðÞ 1 o2 ðÞ 2 oðÞ cot # oðÞ ð5:6:6Þ
¼ þ þ þ þ 2 :
or 2 r 2 o#2 r 2 sin2 # ou2 r or r o#
How does this compare to Eq. (4.2.19)? Now argue that the stress tensor
in spherical coordinates can be decomposed as follows:
rhiji ei ej rhrri er er þ rhr#i er e# þ þ rhuui eu eu : ð5:6:7Þ
Use the del operator to compute the divergence of that expression and
show that:
orhrri 1 orh#ri 1 orhuri
rr¼ þ þ
or r o# r sin # ou
2rhrri rhuui rh##i þ rh# ri cot # orhr#i 1 orh##i
þ er þ þ
r or r o#
1 orhu#i 2rhr#i þ rh#ri þ rh##i rhuui cot #
þ þ e# ð5:6:8Þ
r sin # ou r
orhrui 1 orh#ui 1 orhuui
þ þ þ
or r o# r sin # ou
2rhrui þ rhuri þ cot # ðrh#ui þ rhu#i Þ
þ eu :
r
However, before that prove the following auxiliary formulae:
oer oer oer oe# oe#
¼ 0; ¼ e# ; ¼ sin # eu ; ¼ 0; ¼ er ; ð5:6:9Þ
or o# ou or o#
oe# oeu oeu oeu
¼ cos # eu ; ¼ 0; ¼ 0; ¼ ½sin # er þ cos # e# :
ou or o# ou
5.7 Balances of Energy for Regular Points in Arbitrary Coordinate Systems 117
By inserting the corresponding entries in Table 3.1 in Eq. (3.7.2) we obtain the
local regular balance for the total energy in a Cartesian coordinate system:
ffi ! ffi !
o 1 o 1
q uþ ti ti þ q uþ ti ti tj
ot ðxÞ ðxÞ 2 ðxÞ ðxÞ oxj ðxÞ ðxÞ 2 ðxÞ ðxÞ ðxÞ
ffi ! ð5:7:1Þ
o
¼ q
j þ r ji t i þ q f i t i þ r :
oxj ðxÞ ðxÞ ðxÞ ðxÞ ðxÞ ðxÞ ð xÞ
In order to rewrite it for the z-system it should be mentioned that the (invariant)
scalar product t i t i of the kinetic energy part can be written as:
ðxÞ ðxÞ
Note that the rules of transformation for co- and contravariant vectors as well as
the chain rule have been used. The result is not too surprising since a scalar
product represents an invariant. Also note the following alternative formulae to Eq.
(5.7.2), which are somewhat lengthier:
ozk ozl
ti ti ¼ tk t l ¼ gkl t k t l
ðxÞ ðxÞ oxi ðzÞ oxi ðzÞ ðzÞ ðzÞ
ð5:7:3Þ
oxi oxi
t i t i ¼ k t k l t l ¼ gkl t k t l
ðxÞ ðxÞ oz ðzÞ oz ðzÞ ðzÞ ðzÞ
because just like mass density the specific internal energy, u, is a scalar quantity
u ¼ u . The second expression on the left hand side of Eq. (5.7.1) corresponds to
ðxÞ ðzÞ
chosen a covariant notation for the velocity (note that because of Eq. (4.4.9) we
have gij ;k ¼ 0):
ffi ! ffi !
o 1 1
q u þ t i t i t j ¼ q u þ t k t k t i gli : ð5:7:6Þ
oxj ðxÞ ðxÞ 2 ðxÞ ðxÞ ðxÞ ðzÞ ðzÞ 2 ðzÞ ðzÞ ðzÞ
;l
The expressions on the right hand side of Eq. (5.7.1) are manipulated similarly.
For the divergence of the heat flux vector it is possible to write:
oq j !
ð xÞ r rm
¼ q ;r ¼ g q m ¼ grm
;r q m þ grm q m;r
oxj ðzÞ ðzÞ ðzÞ ðzÞ ð5:7:7Þ
;r
rm rm
¼0þg q m;r ¼g q m;r
ðzÞ ðzÞ
depending upon whether the heat flux vector is interpreted as a co- or contravariant
object. Moreover, the scalar product between stress tensor and velocity results in a
vector, which can be written the co- or contravariant way:
ozr ozs oxi k ozr s ozr
r ij t i ¼ r k
t ¼ dk r t k ¼ r ts
ðxÞ ðxÞ oxj oxi ðzÞrs oz ðzÞ oxj ðzÞrs ðzÞ oxj ðzÞrs ðzÞ
oxj
) r t s ¼ r r ji t i ;
ðzÞrs ðzÞ oz ðxÞ ðxÞ
ð5:7:8Þ
oxj oxi rs ozk oxj oxj
r ji t i ¼ r s r t ¼ r dks r rs t ¼ r r rs t
ðxÞ ðxÞ oz oz ðzÞ oxi ðzÞk oz ðzÞ ðzÞk oz ðzÞ ðzÞs
ozr
) r rs t s ¼ r ji t i ;
ðzÞ ðzÞ oxj ðxÞ ðxÞ
if we only apply the transformation rules for co- and contravariant vectors and
tensors as well as the chain rule sufficiently often. Thus the second term on the
right hand side of Eq. (5.7.1), which is the divergence of a vector, can either be
written as the covariant derivative of a contravariant or a covariant vector field:
! !
o
q þ r t ¼ q r þ r rs t
oxj ð xÞi ðxÞji ðxÞi ðzÞ ðzÞ ðzÞs
;r
! ! ð5:7:9Þ
¼ q r þ grl gsm r lm t s ¼ grl q l þ r lm t m :
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
;r ;r
k
ti f i ¼ tk f ¼ tk f k ¼ gkl t k f l ¼ gkl t k f l : ð5:7:10Þ
ðxÞ ðxÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Exercise 5.7.1: Rewriting the left hand side of the energy balance
Recall the definition of the material time derivative first introduced in
context with the mass density in Eq. (3.8.6) as well as the mass balance
(3.8.3). Show that the left hand side of the energy balance (5.7.1) can
alternatively be written as:
ffi ! ffi !
o 1 o 1
q uþ ti ti þ q uþ ti ti tj
ot ðxÞ ðxÞ 2 ðxÞ ðxÞ oxj ðxÞ ðxÞ 2 ðxÞ ðxÞ ðxÞ
ffi ffi ð5:7:13Þ
d 1 d 1 i
q uþ ti ti q uþ tit :
ðxÞ dt ðxÞ 2 ðxÞ ðxÞ ðzÞ dt ðzÞ 2 ðzÞ ðzÞ
Use further arguments from the previous section to show that the local
balance of kinetic energy can be rewritten as:
0q 1 0q 1
o @ðzÞ ð zÞ
t i t iA þ @ t i t i t jA
ot 2 ðzÞ ðzÞ 2 ðzÞ ðzÞ ðzÞ
;j ð5:7:18Þ
ffi
¼ r ji t i þ q f k t k r ji t i ; j :
ðzÞ ðzÞ ;j ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Use these results to discuss the pros and cons of the following coordinate
independent form of the First Law of thermodynamics:
ou
q þ q t ru ¼ r q þ r : ðrtÞ þ q r: ð5:7:26Þ
ot
Derive the following equations for the divergence of the heat flux in
physical cylindrical as well as in physical spherical coordinates by using the
del operator and, alternatively, the covariant derivative:
oqhri 1 oqh#i oqhzi qhri
rq¼ þ þ þ ð5:7:27Þ
or r o# oz r
and:
oqhri 1 oqh#i 1 oqhui 2 cot #
rq¼ þ þ þ qhri þ qh#i : ð5:7:28Þ
or r o# r sin # ou r r
5.8 Balances of Angular Momentum for Regular Points 123
In the previous sections it was explained in great detail how to derive a balance
equation for scalar and vector quantities in arbitrary coordinate systems, (z), once
the corresponding balance has been established in a Cartesian system, (x). We
therefore recall the balance of total angular momentum from Sect. 3.10:
ffi ffi
d s i þ ijk j t k
x o m li þ ijk j r lk
x !
ðxÞ ðxÞ ðxÞ ðxÞ ð xÞ ðxÞ ðxÞ ðxÞ
q ¼ þ q ijk t j f k þ l i ð5:8:1Þ
ð xÞ dt oxl ðxÞ ðxÞ ðxÞ ðxÞ ðxÞ
and, by applying the same rules as before, i.e., partial derivatives turn into
covariant ones, covariant and contravariant indices must be appropriately placed,
and free indices must agree on both sides of a tensor equation, we arrive imme-
diately at:
ffi
i
d s þ xjtk ijk ffi !
ðzÞ ðzÞ ðzÞ ðzÞ
li i j lk i j k i
q ¼ m þ jk x r þ q jk t f þ l ; ð5:8:2Þ
ðzÞ dt ðzÞ ðzÞ ðzÞ ðzÞ
;l ðzÞ
ðzÞ ðzÞ ðzÞ ðzÞ
which is just one way of writing this equation among many other co-/contravariant
variations. Similarly the balance of moment of momentum and of spin read [cf.,
Eqs. (3.10.1) and (3.10.6)]:
ffi
d ijk x
j tk ffi
ðzÞ ðzÞ ðzÞ
q ¼ ijk x j r lk ijk r jk þ ijk x j q f k ;
ðzÞ dt ðzÞ ðzÞ ðzÞ
;l
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
ð5:8:3Þ
i
ds
ðzÞ
q ¼ m li ;l þ ijk r jk þ q l i :
ðzÞ dt ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
8
< þ1r ; if i; j; k ¼ r; #; z and cyclic permutations
ijk ¼ 1r ; if i; j; k ¼ z; #; r and cyclic permutations ð5:8:4Þ
:
0; else
and
8
< þr2 sin
1
# ; if i; j; k ¼ r; #; u and cyclic permutations
ijk ¼ r2 sin
1
# ; if i; j; k ¼ #; r; u and cyclic permutations ð5:8:5Þ
:
0; else;
respectively. Show that this can be rewritten as:
8
< þ1; if i; j; k ¼ r; #; z (or uÞ and cyclic permutations
hijki ¼ 1; if i; j; k ¼ z ðor uÞ; #; r and cyclic permutations ð5:8:6Þ
:
0; else:
Write the last result in terms of scalar and vector products between the
corresponding unit base vectors er ; e# , and ez=u :
and
In what follows we will only investigate how the fields of mass, velocity, etc. jump
when passing through a singular surface with no intrinsic properties. This is mainly
due to lack of space. For an extensive discussion the reader is referred to Müller
[7], Sect. 3.2 or to the paper by Moeckel [6]. This means that the mass density per
unit surface, q, in the general Eq. (3.10.1) is equal to zero as well as all the other
A
surface related quantities on the left hand side. Thus we find in combination with
Table 3.2 in Cartesian coordinates for the jump of mass density:
"" ffi ##
q t i t? e i e i ¼ 0; t ? ¼ t i e i ð5:9:1Þ
ð xÞ ð xÞ A ð xÞ ðxÞ A ðxÞ ðxÞ
of momentum density:
"" ! ##
q ti tj t ? ej rji ej ¼ 0; ð5:9:2Þ
ðxÞ ðxÞ ðxÞ A ð xÞ ðxÞ ð xÞ
and, for the sake of brevity, only for the jump of the internal energy:
126 5 Balance Equations in Skew Curvilinear Coordinate Systems
"" ffi ##
qu ti t ? e i þ qi ei ¼ 0 ð5:9:3Þ
ðxÞ ðxÞ ð xÞ A ðxÞ ð xÞ ðxÞ
The scalar, vectors, tensors, and scalar products appearing in these relations can
now easily be transformed into an arbitrary skew curvilinear coordinate system
according to the previously explained transformation rules. Consequently, possible
co-/contravariant notations read for mass:
"" ffi ##
q ti t? e i
e i ¼ 0; t ? ¼ t ie i ð5:9:4Þ
ðzÞ ðzÞ A ðzÞ ðzÞ A ðzÞ ðzÞ
A
for momentum:
"" ffi ##
i j j ji
qt t t? e r e j ¼ 0; ð5:9:5Þ
ðzÞ ðzÞ ðzÞ A ðzÞ ðzÞ ðzÞ
and, again for the sake of brevity, only for the internal energy:
"" ffi ##
q u t i t ? e i þ q i e i ¼ 0: ð5:9:6Þ
ðzÞ ðzÞ ðzÞ A ðzÞ ðzÞ ðzÞ
For rewriting the global balances of mass, momentum, and energy we shall first
investigate GAUSS’ theorem (3.4.7) in arbitrary coordinate systems:
ZZZ o g i ZZ ZZ
ðxÞ
dV ¼ g i n i dA gi e i dA: ð5:10:1Þ
oxi ðxÞ ðxÞ ðxÞ ðxÞ
V Aþ [A A
This allows rewriting the transport theorem of Eq. (3.4.12) in the following
form:
5.10 The Transport Theorem for Volume Integrals in Arbitrary Coordinate Systems 127
2 3
ZZZ ZZZ o wV !
d 6 ðzÞ 7
wV dV ¼ 4 þ wV t i 5 dV
dt ðzÞ ot ðzÞ ðzÞ
V þ [V V þ [V ;i
ZZZ o wV ZZ ZZ "" ##
ðzÞ
¼ dV þ wV t ði zÞ i dA wV t i e i dA ð5:10:3Þ
ot ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
V þ [V Aþ A
ZZZ o wV ZZ ZZ "" ##
ðzÞ i
¼ dV þ wV t ðzÞ i dA wV t ? dA
ot ðzÞ ðzÞ ðzÞ A
V þ [V Aþ A
References
It has already been mentioned in Chap. 3 that the main objective of a thermo-
mechanical continuum theory consists of computing five fields, namely mass
density, velocity, and temperature at all times and in all points of a body. To this
end constitutive equations are required in addition to the balances of mass,
momentum, and energy.
Indeed, the balance equations turn into field equations of continuum thermo-
mechanics only if we clarify the dependence of the stress tensor, of internal
energy, of the heat flux, of the specific volumetric force, and of the heat supply in
C ijkl denotes the so-called stiffness matrix, a tensor of forth order, which charac-
ð xÞ
0 1
oui ouj
1 ð xÞ ð xÞ
e ij ¼ @ þ A: ð6:2:2Þ
ð xÞ 2 oxj oxi
material particle of the body from its reference position which, in Cartesian
coordinates, is given by the vector Xi :
u i ¼ x i Xi : ð6:2:3Þ
ð xÞ
Moreover, e ij denotes the (symmetric) tensor of strains of inelastic origin, for
ð xÞ
and l , which may depend on temperature, are the so-called LAMÉ constants.
ð xÞ
Siméon Denis POISSON was born on June 21, 1781 in the small town of
Pithiviers (France) and died on April 25, 1840 in Paris. He was born
into a poor family. Consequently, during his youth he had hardly any
opportunity to acquire much more than elementary skills in reading and
writing. However, his real talents finally emerged: He attempted to
study mathematics and physics and passed the entry exam at the
famous École Polytechnique in Paris in 1798 with highest honors. After
that it did not take him very long to become one of the leading figures
at the French Academy of his time.
neglecting thermal expansion) and the kinematic relations for small strains
and obtain the LAMÉ-NAVIER differential equations for the displacements for
the anisotropic as well as for the isotropic case:
o2 u k ! o2 u o2 u j
k
ð xÞ ð xÞ ð xÞ
C ijkl ¼0; kþl þl ¼ 0: ð6:2:10Þ
ð xÞ oxi oxl ð xÞ ð xÞ oxj oxk ð xÞ oxk oxk
x1
1
Thanks for suggesting this problem are due to Dr. Wolf Weiss from the Weierstrass Institut in
Berlin.
134 6 Constitutive Equations in Arbitrary Coordinate Systems
Now evaluate HOOKE’s law from Eq. (6.2.6) without thermal strains with
the ansatz, conclude that C1, C2, C3 are true constants and show that:
u 1 ¼ C1 x1 ; u 2 ¼ C2 x1 ; u 3 ¼ C 3 x1 : ð6:2:13Þ
ð xÞ ð xÞ ð xÞ
Combine this result with the boundary conditions for the displacement to
obtain:
u0
u1¼ x1 ; u 2 ¼ 0; u3¼0 ð6:2:14Þ
ð xÞ l ð xÞ ð xÞ
and:
!
u0 u0 u0
r 11 ¼ k þ2 l ; r 22 ¼ k ; r 33 ¼ k : ð6:2:15Þ
ð xÞ ð xÞ ð xÞ l ð xÞ ð xÞ l ð xÞ ð xÞ l
Exploit the LAMÉ-NAVIER equations with this ansatz, solve the resulting
ordinary differential equation to prove that:
G¼l ð6:2:19Þ
and:
s ¼ G tan c Gc; ð6:2:20Þ
where c denotes the shear angle indicated in Fig. 6.2 (right). Comment on
the differences between this and the previous line of arguments. In particular
explain why some of the flanks in Fig. 6.2 (left) are unloaded whereas in
Fig. 6.2 (right) they are not, i.e., discuss as to whether the problem can easily
be solved for a block of finite size in x1 direction.
The strain tensor in covariant notation can be found by applying Eqs. (4.2.11/
4.2.12) and Eq. (6.2.2):
1
e ij ¼ u i;j þ u j;i : ð6:2:22Þ
ðzÞ 2 ðzÞ ðzÞ
Exercise 6.2.3: HOOKE’s law and the strain tensor in contravariant and
other notations
Repeat all the steps required to get from Eqs. (6.2.6) to (6.2.21), and from
Eqs. (4.2.11/4.2.12) and (6.2.2) to Eq. (6.2.22). Also explain and verify the
validity of the following alternative equations for the (trace of the) strain
tensor:
136 6 Constitutive Equations in Arbitrary Coordinate Systems
rs
e ¼ gri gsj e ; e r ¼ gri e ; e i s ¼ gsj e ;
ðzÞ ðzÞ ij ðzÞ j ðzÞ ij ðzÞ ðzÞ ij
pffiffiffiffiffiqffiffiffijjffiffi pffiffiffiffiffi
qffiffiffiffiffi
pffiffiffiffiffi
qffiffiffiffiffi
pffiffiffiffiffipffiffiffiffiffi
ehiji ¼ gii g e ¼ gii gjj e i j ¼ gii gjj e i j ¼ gii gjj e ij ;
ðzÞ ij ðzÞ ðzÞ ðzÞ
rj
e i i ¼ gij e ¼ gjr e ; e j j ¼ gji e ¼ gjr e rj
;
ðzÞ ðzÞ ij ðzÞ ðzÞ ðzÞ ij ðzÞ
ð6:2:23Þ
and for HOOKE’s law:
ij
r ¼ k gkl e lk gij þ 2 l e ij
3 k a DTgij ;
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
r i j ¼ k gkl e j
lk di þ 2 l e i
j
3 k a DTdi j ; ð6:2:24Þ
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
i i
r i j ¼ k gkl e lk d j þ 2 l e i j 3 k a DTd j :
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Exercise 6.2.5: HOOKE’s law and the strain tensor in spherical coordinates
Use the metric tensor and the CHRISTOFFEL symbols in spherical coordi-
nates from Eqs. (2.2.16) and (4.2.6), and evaluate the general Eqns. (6.2.21/
6.2.22) to show that:
rhiji ¼ k ehlli dhiji þ2l ehiji ; i; j 2 ðr; u; #Þ ð6:2:28Þ
and:
ouhri 1 ouh#i 1
ehrri ¼ ; eh##i ¼ þ uhri ;
or r o# r
1 ouhui 1 cotð#Þ
ehuui ¼ þ uhri þ uh#i ;
r sinð#Þ ou r r
1 1 ouhri 1 ouhui
ehrui ¼ uhui þ ; ð6:2:29Þ
2 r sinð#Þ ou r or
1 1 ouhri 1 ouh#i
ehr#i ¼ uh#i þ ;
2 r o# r or
1 1 ouhui cotð#Þ 1 ouh#i
ehu#i ¼ uhui þ :
2 r o# r r sinð#Þ ou
Observe Eq. (3.4.17) and show that the Jacobian is given by:
J 1 þ ekk : ð6:2:33Þ
Recall the result from Exercise 3.8.1 that the current mass density can be
calculated from the mass density of the reference state by means of the
Jacobian. Show that for small deformations Eq. (3.8.12) can be rewritten as:
q ¼ q0 ð1 ekk Þ: ð6:2:34Þ
Use Eq. (6.2.8) and show that the mass density of a tensile bar can be
obtained from:
q q0 Dl
¼ ð1 2mÞ : ð6:2:35Þ
q0 l
By how many percent does the density of a bar made of steel decrease
after an elongation of 5 %?
Viscous fluids subjected to small or medium (shear) velocity gradients are fre-
quently modeled by the constitutive equation of NAVIER–STOKES. It reads in a
Cartesian base x:
0 1
otk oti otj
ð xÞ ð xÞ ð xÞ
r ij ¼ p dij þ k dij þ l @ þ A: ð6:3:1Þ
ð xÞ ð xÞ ð xÞ oxk ð xÞ oxj oxi
The scalars k and l , which may depend on temperature and density, are known
ð xÞ ð xÞ
The scalar constitutive term p is known as the pressure, which—as one should
ð xÞ
p ¼ p; k ¼ k; l ¼ l :
ð xÞ ðzÞ ð xÞ ðzÞ ð xÞ ðzÞ
k l
If these relations are inserted in Eq. (6.3.1) the term oz oz
oxi oxj can be extracted as a
common factor and the constitutive law of NAVIER–STOKES in covariant notation
w.r.t. the free indices k and l in an arbitrary base z results:
p r
r kl ¼ gkl þ k t ;r gkl þ l t k;l þ t l;k : ð6:3:4Þ
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Use the results from Exercises 6.2.4/6.2.5 to prove the validity of the
following forms of the NAVIER–STOKES constitutive equation:
rhiji ¼ pdhiji þk dhlli dhiji þ2l dhiji ; ð6:3:6Þ
where i; j 2 ðr; #; zÞ for cylindrical coordinates and i; j 2 ðr; u; #Þ for
spherical coordinates and the symmetric part of the velocity gradient is given
by:
1
d kl ¼ t k;l þ t l;k : ð6:3:7Þ
ðzÞ 2 ðzÞ ðzÞ
m ~
R
p¼ T: ð6:4:4Þ
V NAvo l M
142 6 Constitutive Equations in Arbitrary Coordinate Systems
~ ¼ R 103 kg ¼ 8:314 J :
R ð6:4:7Þ
mol K mol
This is the ideal gas constant frequently used in school. Note that Eq. (6.4.1) is
more general than Eq. (6.4.2). It holds locally or, in other words, it can also be
applied to a vessel with a heterogeneous gas filling. Thus gradients of density and
temperature may exist. Moreover, note that the term ideal gas equation does not
refer in any way whatsoever to the phenomenon of friction. Friction is allowed and
accounted for by a different term as shown in Eq. (6.3.1). Thus the ideal gas law
may also be used in context with frictional flow of gases, at least as a first
approximation, as an analytic constitutive relation for the pressure.
Ludwig BOLTZMANN was born on February 20, 1844 in Vienna and died
on September 5, 1906 in Duino near Trieste. He studied physics at the
University of Vienna where he became a scientific assistant in 1867. In
1869 he accepted a professorship for theoretical physics in Graz, fol-
lowed by vocations to the University of Munich, Vienna, and Leipzig.
Finally, in 1895 he became the successor to the chair of Josef STEFAN in
Vienna. BOLTZMANN suffered from severe psychological problems. This is
also reflected in a series of events following his vocation to a chair in
theoretical physics in Berlin, which he would accept on one day and
reject the day after—not only once, but repeatedly in several reiterations,
simply because he felt unapt to cope with the duties of the new office. KAISER WILHELM finally put
an end to all this and withdrew the offer. BOLTZMANN’s most important contribution to science
was probably the statistical interpretation of thermodynamics and, in particular, the statistical
interpretation of entropy. We must realize that the notion of the atomistic nature of matter was
revolutionary in BOLTZMANN’s days. Consequently, BOLTZMANN had many famous adversaries, in
particular those of the positivistic Viennese school, such as Wilhelm OSTWALD, but—surpris-
ingly—also one of the fathers of quantum mechanics, namely Max PLANCK, who acted through
his assistant ZERMELO. Rumor has it that the scientific discussions were so painful to BOLTZMANN
that he finally committed suicide. But, maybe, he was just a thoroughbred Viennese, who lived
his life according to Georg KREISLER’s song: Der Tod, das muss ein Wiener sein.
Finally we will present a few alternative ways of writing Eqs. (6.4.1) and
(6.4.2), which can frequently be found in thermodynamics textbooks. They are
based on the definition of the number of moles as the ratio between the number of
particles, N, and AVOGADRO’s number:
6.4 The Ideal Gas Law 143
N
m¼ ð6:4:8Þ
NAvo
and an atomistic gas constant, which is known as the BOLTZMANN constant:
J
k ¼ l R ¼ 1:38 1023 : ð6:4:9Þ
K
Simple algebraic manipulations lead to:
pV ¼ NkT; pV ¼ mNAvo kT: ð6:4:10Þ
This section deals with the constitutive equation for the scalar field of the internal
energy. It is also known as the caloric equation of state, since it is related to the
heat storage of a material as we shall see soon. We first turn to gases: In general
the internal energy of a gas is a non-analytic function of (at least) two variables
and it depends, just like the gas pressure, on the (local) mass density, q, (or on its
inverse, the specific volume, t) and on the temperature, T:
u ¼ ~uðq; T Þ or u ¼ ^uðt; T Þ: ð6:5:1Þ
James Prescott JOULE was born on December 24, 1818 in Salford near
Manchester and died on October 11, 1889 in Sale, Greater Manchester. He
was born into a family of brewers and continued to work in the family
tradition together with one of his brothers. However, he also began to
study mathematics and science in 1834, both as a hobby and for the
benefit of the family enterprise. In 1837 he installed his own chemical lab
which was also financially supported by various brewery organizations
later. His greatest scientific achievement was the experimental determi-
nation of the so called mechanical heat equivalent at a time where ther-
modynamic notions like heat and internal energy just started to emerge. In his later years his
health deteriorated and he was more and more troubled by financial disasters. The latter
problem was solved, at least in part, when Queen VICTORIA awarded this famous son of the
British Crown a pension in 1878.
Recall the ideal gas relations (6.4.1) and (6.5.2). Use them to show that for
ideal gases we have:
R R R
h ¼ ð f þ 1Þ T þ u0 ; c p ¼ ð f þ 1Þ ; c p ¼ c t þ : ð6:5:18Þ
M M M
Obviously the specific heat of an (ideal) gas at a constant pressure is
greater than the one at a constant volume. Standard textbooks on thermo-
dynamics interpret this observation by saying that the piston which keeps a
gas under a constant pressure will be lifted when adding heat. This is an
additional amount of work, which is not required in the situation of a gas in a
vessel of fixed size. In the latter case the supplied heat would directly and
completely affect the internal energy alone. Obviously this line of arguments
is rather intuitive and it is rather difficult to judge its range of validity.
However, it should be mentioned in this context that by combining pdV
thermodynamics with the notion of entropy and the Second Law it can be
shown that, in general, the specific heat at a constant pressure is always
greater than the specific heat at a constant volume.
f ¼ 32 in this case. If the gas consists of bi-atomic molecules two rotational degrees
of freedom must be added so that rotational motions perpendicular to the atomic
bond are acknowledged. Consequently this leads to f ¼ 52. Finally, if molecules of
three or more atoms are involved three degrees of freedom for translational as well
as rotational motion result so that f ¼ 6=2 ¼ 3.
With a little imagination this kind of argument can easily be extended to the
case of solids: The solid is envisioned as a three-dimensional system of mass
points (the atoms) that are connected to each other by nonlinear springs (repre-
senting the bonding forces). Such a solid has three translational degrees of freedom
and (because of the 3D spring arrangement) three degrees of freedom of potential
energy as well. Each of them we assign 12 MR T and, consequently, the specific heat
must be:
R
c¼3 : ð6:5:19Þ
M
This is DULONG-PETIT’s rule. However, there is a catch in our line of arguments:
In contrast to the case of a gas we did not specify what is kept constant when
measuring the specific heat of a solid, its volume or the pressure acting on it, or
…? We start over again and just like the gas the internal energy of a solid depends
on two variables, one of which is temperature. Now recall Eq. (6.2.34): It shows
that the mass density can be determined from the trace of the strain tensor (for
small deformations). This is why it is useful to replace the mass density in the
constitutive equation for the specific internal energy by the strain tensor (com-
ponents) instead:
u ¼ ~uðe; T Þ: ð6:5:20Þ
Thus DULONG-PETIT’s rule reads more precisely:
Def: ou R
ce ¼ ¼ 3 : ð6:5:21Þ
oT e M
That this is really the specific heat at a constant strain can be shown with
statistical mechanics arguments by using simple atomistic models for solids.
Frequently the range of validity of DULONG-PETIT’s rule is shown in a high school
experiment which will be discussed in-depth in the following exercise.
148 6 Constitutive Equations in Arbitrary Coordinate Systems
Eduard GRÜNEISEN was born on May 26, 1877 in Giebichenstein near Halle
(Germany) and died on April 5, 1949 in Marburg (Germany). At the age of
17 he studied physics in Halle and Berlin and obtains his doctoral degree in
1900 under the scientific guidance of WARBURG and PLANCK. In 1911 he
becomes a professor at the Physikalisch-Technische Reichsanstalt (today’s
Federal Institute for Materials Research and Testing), advances to
departmental manager in 1919, and moves in 1927 to the University of
Marburg, where he stays until the end of his life. GRÜNEISEN worked pre-
dominantly on equations of state for solids. How- ever, one of his tasks was
also the examination of medical students in physics, which can be very disillusioning for a true
physicist to say the least. According to an anecdote of the author’s father he was very pleased to
hear that at least one of the physicians-to-be knew what a differential quotient was and that
velocity was defined by one.
For this purpose start from the global balance for the total energy of Eq.
(3.9.5) and apply it to the material volume V(t) with surface oV consisting of
the iron lump and of the water (see Fig. 6.3). Assume that the heat exchange
is exclusively between the water and the iron: The container allows for no
other heat exchange and is adiabatically sealed. Argue that under these
circumstances we have:
ZZ ZZ
dV
nj qj dA ¼ 0 ; nj ti rji dA ¼ p0 ; ð6:5:22Þ
dt
oV oV
where p0 is the (constant) pressure of the surrounding air. Now integrate the
energy balance w.r.t. time from the beginning to the end of the internal heat
exchange. Which additional assumptions will then lead to:
6.5 The Internal Energy of Gases and Solids 149
technik-ethik.de/ 4.21
wasser_eigenschaften. 4.2
html 4.19
4.18
4.17
4.16
4.15
0 10 20 30 40 50 60 70 80 90 100
temperature [°C]
GRÜNEISEN derived the following formula for the difference between the two
specific heats of a solid:
It was FOURIER’s great deed to realize that the heat flux is proportional and opposite
in direction to the temperature gradient. He was one of the first to recast these
verbal statements into mathematics. Following him we write in a Cartesian system:
oT
ð xÞ
qi¼j : ð6:6:1Þ
ð xÞ ð xÞ oxi
6.6 FOURIER’s Law of Heat Conduction 151
Exercise 6.6.1: Direction of the heat flux and the temperature gradient
Consider a wall of thickness d whose left and right side are kept at
temperature levels T1 and T2 , respectively. Assume that T2 [ T1 . Determine
the direction of the temperature gradient and of the heat flux vector. Use the
expression for the total heat flux in the First Law in global form shown in Eq.
(3.9.6) and confirm the rule that ‘‘heat flows from hot to cold.’’ How could
the analysis be used to obtain a numerical value for the heat conductivity?
References
Abstract We now combine balance and constitutive equations and obtain field
equations for fluids and solids, all in Cartesian coordinates. They are used to pose
and to solve initial-boundary value problems for simple geometries and to reach
the primary goal of continuum theory, namely the determination of the five fields
for mass density, velocity, and temperature in each point of a material body and at
all times.
The iron fist of the real, inside the velvet glove of airy
mathematics.
Consider the situation shown in Fig. 7.1: An adiabatically sealed cylinder (mass
mc ) with an adiabatic piston of cross-sectional area A contains an ideal gas of mass
mg under the initial pressure ps (s = start). Initially this pressure is not completely
counterbalanced by the weight of the piston, mp g, in combination with an external
pressure p0 : ps 6¼ mp g A þ p0 .
In other words, it is initially necessary to fix the piston at a starting position, zs ,
so that it does not move. If the fixtures are detached the piston will start moving
and turns the gas into turbulent motion. However, due to internal friction the
motion will eventually come to a standstill.
Our objective is to determine the height ze (e=end) at which the piston will
finally stop as well as the corresponding gas temperature Te and pressure pe ,
mp
g
mg
∂V
7.2 Globally Stated Problems Involving Control Volumes 155
respectively. The latter is very easy to compute since a purely mechanical problem
is involved: In the end equilibrium of forces must prevail and thus:
mp g
mp g p0 A þ pe A ¼ 0 ) pe ¼ p0 þ : ð7:2:1Þ
A
The term ‘‘adiabatic’’ has been used several times in context with the problem
statement. Within the scope of a beginner’s course on thermodynamics one would
be tempted to use the adiabatic relations from Exercise 6.5.3 in order to determine
the remaining unknowns. Thus we start with Eq. (6.5.29), connect it with the
(constant) mass mg of the gas and conclude that:
ffij1
ps
ts ¼ mVsg ¼ Az s Ve
m g ; te ¼ m g ¼ m g
Aze
) ze ¼ zs pe
ffij1 ð7:2:2Þ
ps
) ze ¼ zs p 0 þ mp g :
A
The final temperature follows from the ideal gas law, which can be applied to
the initial and final states of equilibrium:
ffij1
j
ps=e zs=e A ¼ mg MR Ts=e ) Te ¼ Ts ppes zzes ¼ Ts pe
ps
mp gffij1 ð7:2:3Þ
p þ j
) Te ¼ T s 0 p s A :
Brook TAYLOR was born on August 18, 1685 in Edmonton and died on
December 29, 1731 in London. He obtained his mathematical training at
St. John’s College in London and is known as an enthusiastic admirer of
NEWTON. From 1712 onwards he published several papers in the Philo-
sophical Transactions of the Royal Society, on the motion of projectiles
and the shape of liquid surfaces. The famous TAYLOR expansion was
established in 1715 and can be found in Proposition 7 of his paper
Methodus Incrementorum Directa et Inversa.
Nevertheless the results shown in Eqs. (7.2.2/7.2.3) are dubious for several
reasons. The main point of criticism is related to the fact that for large initial pressure
differences—in other words for a very heavy piston—it cannot be avoided that
turbulence in the gas will set in and that the resulting thermodynamic process is
highly irreversible. However, under such circumstances the adiabatic relations of
Exercise 6.5.3 do definitely not hold. Unfortunately they form the backbone in our
previous line of arguments. On the other hand it is to be suspected that the equations
represent the situation almost correctly for small pressure differences. This is why
we expand the results in TAYLOR series using a smallness parameter Dp:
mp g
ps p0 þ þ Dp: ð7:2:4Þ
A
If the series resulting from Eqs. (7.2.2/7.2.3) is truncated after the linear term
we obtain:
156 7 A First Glance on Field Equations
! !
1 Dp j 1 Dp
ze zs 1þ ; Te T s 1 : ð7:2:5Þ
j p0 þ mAp g j p0 þ mAp g
Note that the proper choice of a control volume was crucial in order to arrive at
this simple result: If part of the envelope is positioned (for example) on the inner
side of cylinder instead, Eq. (7.2.6) would not hold since the gas will exchange
heat with the cylinder during the process. Moreover, there is no radiation supply in
this problem:
ZZZ
qrdV ¼ 0: ð7:2:7Þ
V ðt Þ
1
The latter effect is best known from the air pump which heats up during fast compression (i.e.,
decrease of air volume).
7.2 Globally Stated Problems Involving Control Volumes 157
ni ¼ ð0; 0; 1Þ so that each of its material points shows the same velocity
ti ¼ ð0; 0; dz=dtÞ. Thus:
ZZ ZZ ZZ
dz dz dz dðzAÞ dV
ti ni dA ¼ dA ¼ dA ¼ A ¼ ¼ : ð7:2:9Þ
dt dt dt dt dt
oV ðtÞ oV ðtÞ oV ðtÞ
In order to rewrite the power supply of the volumetric force in the balance of
energy we note that gravitation is a (static) conservative force. In general con-
servative forces can be obtained from a spatial derivative of a scalar field, the
potential u, as follows:
ou
fi ¼ : ð7:2:10Þ
oxi
Consequently the power assigned to a (static) conservative specific force is
given by:
ZZZ ZZZ ZZZ
ou dxi du
qti fi dV ¼ q dV ¼ q dV
oxi dt dt
V ðtÞ V ðtÞ V ðtÞ
Z Z ZZZ ð7:2:11Þ
du d d
¼ dm ¼ u dm ¼ qu dV;
dt dt dt
M M V ðt Þ
and, therefore, we obtain for the case of gravity near the surface of the Earth,
u ¼ gz:
ZZZ ZZZ ZZZ
d d
qti fi dV ¼ qp gz dV qg gz dV
dt dt
V ðtÞ Vp Vg ð t Þ
0 1
ZZZ ZZZ
Bd d C ð7:2:12Þ
¼ g@ qp z dV þ qg z dV A
dt dt
Vp Vg ðtÞ
dzcp dzcg d ffi
¼ g mp þ mg mp gzcp þ mg gzpg :
dt dt dt
Vp and Vg ðtÞ denote the (current) volumes of the piston and of the enclosed gas,
respectively. Note that only the latter volume must be considered as time-
dependent. The piston has already been idealized as a rigid body. Moreover, zcp and
zcg denote the current positions of the center of gravity in vertical direction for the
piston and for the gas, respectively. Note that the center of gravity is defined as
follows:
RRR RRR
qp z dV qg z dV
Vp Vg ðtÞ
zcp ¼ ; zcg ¼ : ð7:2:13Þ
mK mg
158 7 A First Glance on Field Equations
This equation can be integrated between the beginning and the end of the
process, i.e., times ts and te , respectively:
ZZZ
q u dV þ p0 Azi þ mp gzcp ðti Þ þ mg gzcg ðti Þ
V ðts Þ
ZZZ ð7:2:15Þ
¼ q u dV þ p0 Aze þ mp gzcp ðte Þ þ mg gzcg ðte Þ:
V ðte Þ
The contributions from kinetic energy vanish because initially and at the end
the system is at rest. For the difference of the internal energies we may write
according to Eqs. (6.5.2/6.5.20/ 6.5.21):
Z Z Z
q u dV q u dV¼ q ce ðTe Ts Þ dV
V ðt e Þ V ðts Þ Vp [Vc
Z
R
þ qf ðTe Ts Þ dV ð7:2:16Þ
M
Vg
R
¼ ce mp þ mc þ f mg ðTe Ts Þ;
M
if we assume that the initial and final temperature fields are homogeneous and the
piston and the cylinder are made of the same material, i.e., they possess the same
specific heat ce . Note that the constants of Eq. (6.5.2) drop out during the sub-
traction since masses are conserved during the process. Moreover, the geometry
requires that:
h i
zcp ðte Þ zcp ðts Þ ¼ 2 zcg ðte Þ zcg ðts Þ ¼ ze zs ; ð7:2:17Þ
where ze and zs denote the bottom position of the piston initially and at the end,
respectively. Thus Eq. (7.2.15) yields:
R
ce mp þ mc þ f mg ðTe Ts Þ
M
ð7:2:18Þ
1
¼ mp þ mg g þ p0 A ðze zs Þ:
2
Furthermore recall that equilibrium of forces must be guaranteed in the end: Eq.
(7.2.1). The initial and the final pressure in the gas, ps and pe , obey the ideal gas
law (6.4.1) applied to a homogeneous state:
7.2 Globally Stated Problems Involving Control Volumes 159
R R
ps Vs ¼ mg Ts ; pe Ve ¼ mg Te : ð7:2:19Þ
M M
Because Ve ¼ Aze it follows that:
ps A mp g þ p0 A
Ts ¼ zs ; Te ¼ ze : ð7:2:20Þ
mg MR mg MR
Decoupling of Eqs. (7.2.18/7.2.20) finally yields:
ffi
m þm
mp þ 12 mg g þ p0 A þ ps Af 1 þ pmg c fcRe
ze ¼ ffi M
zs ;
mp þmc ce 1
f þ mg R mp g þ p0 A þ mp þ 2 mg g þ p0 A
M
ffi ð7:2:21Þ
1 mp þmc ce
mp g þ p0 A m p þ 2 m g g þ p 0 A þ p s Af 1 þ mg fM R
Te ¼ ffi Ts :
ps A mp þmc ce 1
fþ mg R mp g þ p0 A þ mp þ mg g þ p0 A
2
M
How are these results related to the quasistatic argument based on the adiabatic
relation resulting in Eq. (7.2.5)? In order to find out we have to omit in the last two
equations all the quantities that were irrelevant back then. These involve the
(gravitational) mass mg of the gas and the specific heat ce of the piston and of the
cylinder. We then obtain:
!
mp g þ p0 A þ ps Af f Dp
ze ¼ zs 1 þ zs ;
ðf þ 1Þ mp g þ p0 A f þ 1 mAp g þ p0
! ð7:2:22Þ
mp g þ p0 A mp g þ p0 A þ ps Af 1 Dp
Te ¼ Ts 1 Ts
ps A ðf þ 1Þ mp g þ p0 A f þ 1 p0 þ mAp g
In other words the approximate results (7.2.5) obtained by using the adiabatic
relations are re-derived:
! !
1 Dp j 1 Dp
) ze zs 1 þ ; Te T s 1 : ð7:2:24Þ
j p0 þ mAp g j p0 þ mAp g
We now consider a few special cases and start with the exact relations (7.2.21).
In the limit of an infinitely heavy piston, i.e., for mp ! 1 we find:
g
lim ze ¼ 0; lim Te ¼ Ts þ zs : ð7:2:25Þ
mp !1 mp !1 ce
160 7 A First Glance on Field Equations
mp g þ p0 A þ ps Af 1 f mg MR
ze ¼ zs zs þ Ts ;
ðf þ 1Þ mp g þ p0 A fþ1 f þ 1 m p g þ p0 A
ð7:2:26Þ
1 mp g þ p0 A þ ps Af f 1 mp g þ p0 A
Te ¼ Ts Ts þ zs :
ps A fþ1 fþ1 f þ 1 mg MR
This means that even an infinitely heavy piston cannot compress the gas
completely because the temperature of the gas becomes infinitely large since
internal energy cannot be absorbed neither by the piston nor by the cylinder, both
of which were assumed to have no heat storage capacity.
Exercise 7.2.1: Failure of the safety latch between two gas vessels
Consider the situation shown in Fig. 7.2: The gas in the vessel on the right
(start volume V2s , mass m2 , molecular weight M2 ) is initially subjected to a
much higher pressure ps2 than the gas in left vessel (volume V1s , mass m1 ,
molecular weight M1 , pressure ps1 ).
Both gases have the same temperature T s at the start. After failure of the
safety latch the piston separating both vessels starts moving. Turbulent
motion is induced in both gases until friction has turned all kinetic energy
into heat and stationary, homogeneous conditions prevail again. Assume that
the piston is permeable to heat whereas the chamber walls are not (i.e., they
are adiabatic) so that the temperature at the end of the process is the same in
both chambers: T1e ¼ T2e ¼ T e . For simplicity both gases are assumed to be
monatomic. In the beginning and at the end they can be described by the
ideal gas law. Proceed analogously to the previous arguments in this section
and show that:
m1 MR1 þ m2 MR2
T e ¼ T s; pe1 ¼ pe2 ¼ pe ¼ T s;
V tot
m1 m2 ð7:2:28Þ
M1 M2
V1e ¼V tot
m1 m2 ; V2e ¼V tot
m1 m2 ; V tot
¼ V1s þ V2s :
M1 þM 2 M1 þM2
7.2 Globally Stated Problems Involving Control Volumes 161
l1(t) l2(t)
A
V1(t) V2 (t)
p1 <<p2 p2
We will first deal with the case of Cartesian coordinates x and combine the local
static balance of momentum (without volumetric forces) with HOOKE’s law, i.e.,
Eqs. (5.5.2) and (6.2.6):
0 1
o r ji ou k ou j ou i
ðxÞ ðxÞ ðxÞ ð xÞ
¼ 0; r ji ¼ k dji þ l @ þ A 3 k a DTdij : ð7:3:1Þ
oxj ðxÞ ðxÞ oxk ðxÞ oxi oxj ðxÞ ðxÞ
Moreover, we assume that the temperature does not depend on position and
obtain:
o2 u j l o2 u i
ð xÞ ðxÞ ð xÞ
þ ¼0; i; j ¼ 1; 2; 3: ð7:3:2Þ
oxi oxj k þ l oxj oxj
ðxÞ ðxÞ
These are three coupled partial differential equations of second order for the
three unknown displacements u i . They are already field equations in continuum
ð xÞ
mechanics terms because they only contain derivates of the primary field
‘‘velocity’’ and no further unknowns (LAMÉ’s elastic constants are assumed to be
known). These relations are also known as NAVIER-LAMÉ’s equations after the
names of their discoverers. In general they can only be solved numerically after
suitable boundary conditions have been chosen. However, for a few simple cases it
is possible to solve them analytically. Examples are one-dimensional states of
stress, such as the beam under tension or the simple shear from Exercises 6.2.1 and
6.2.2. In this context it is advisable to use the so-called semi-inverse method. In
other words we anticipate the solution and insert a simple but useful ansatz as
shown in Eqs. (6.2.11) and (6.2.18) into the originally complex partial differential
162 7 A First Glance on Field Equations
equations and prove that the solution of the resulting simplified differential
equations does not suffer from internal contradictions. If such contradictions occur
the suggested ansatz was too simple, i.e., simply ‘‘wrong’’ and a more complex
ansatz must be tried instead: In the extreme it may become necessary to look for a
purely numerical solution.
It is interesting to note that the first term in Eq. (7.3.2) represents the gradient of
o2 u j ou j
ðxÞ ðx Þ
a scalar, i.e., the divergence of the displacement vector: oxi oxj ¼ oxo i oxj . Indeed, we
have already shown in Sect. 4.1 that such an expression transforms like the
components of a covariant vector field. The same results if we start directly from
the equations in the z-system corresponding to (7.3.2). Moreover, it is worth
mentioning that the derivatives in the second term of Eq. (7.3.2) look like a
LAPLACE operator applied to the single components of the displacement vector. We
must investigate how this term can be transformed into an arbitrary coordinate
system z since, more precisely, the LAPLACE operator was not introduced as an
operator acting on a vector but on a scalar instead (see Exercise 4.2.7). In order to
clarify such issues we start from relations corresponding to Eq. (7.3.2) but in a z-
system, as they result from Eqs. (5.5.2), (6.2.21/ 6.2.22):
r ji ;j ¼ 0;
ðzÞ
ð7:3:3Þ
r ji ¼ k u k
;k g ji
þ l g ik u j jk
;k þg u
i
;k 3 k a D T gji :
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
jk
Dui¼ g u i : ð7:3:7Þ
ðzÞ ðzÞ ;kj
Explain how the first equation can also be derived from the following
definition of the covariant derivative of a mixed tensor of second order:
o uk ;j
k k ðzÞ
u ;ji ¼ u ;j ¼ þ Ckir ur ;j Crji uk ;r : ð7:3:8Þ
ðzÞ ðzÞ ;i ozi ðzÞ ðzÞ
planar and all material particles of the front cross-section are turned by the
angle #0 w.r.t. the origin of the beam axis (cf., Fig. 7.3).
ϑ0
d = 2R MT
ur ¼ 0; u# ¼ f ðzÞ; uz ¼ 0: ð7:3:10Þ
Convert the ansatz into physical components and insert it in combination
with HOOKE’s law of Eqs. (6.2.26/ 6.2.27) in the static balance of momentum
(5.6.4). Neglect volumetric forces. Show that all but one component of
the momentum balance are identically satisfied. Verify by integration of the
remaining ordinary differential equation and adjustment of the solution to the
boundary condition that:
z
uh#i ¼ r #0 ð7:3:11Þ
l
and that only the following strain and stress components are different from
zero:
r r
eh#zi ¼ #0 ; rh#zi ¼ l #0 : ð7:3:12Þ
2l l
This yields:
ot ot op ! o2 t
i i k
ð xÞ ðxÞ ðxÞ ðxÞ
q þq t j ¼ þ kþl
ðxÞ ot ðxÞ ðxÞ oxj oxi ð xÞ ðxÞ oxi oxk
2
ð7:4:2Þ
o t i
ð xÞ
þl þ q f i:
ðxÞ oxj oxj ðxÞ ðxÞ
Note that the first term after the equal sign is the gradient of a scalar, namely of
the pressure field. We have shown in Sect. 4.1 that gradients of scalars transform
like the components of a covariant vector field. The second term in Eq. (7.4.2) can
also be interpreted as the gradient of a scalar, i.e., the divergence of the velocity
o2 t j ot j
ðxÞ ðx Þ
vector: oxi oxj ¼ oxo i oxj and the third term is essentially a LAPLACE operator acting on
the single components of velocity. How it transforms has already been investigated
in Exercise 7.3.1. All of this must be observed during transformation of Eq. (7.4.6)
into the z-system. Also note that very similar arguments have been used in context
with the NAVIER-LAMÉ equations.
x2
V
h
υi = (υ1,υ 2 ,υ 3)
.
x3 x1
Fig. 7.4 Sketch of the plan-parallel fluid flow between two large plates
oti oti op 0 1 ffi o2 tk o2 t i
q þ qtj ¼ þ l þ l þl þ q fi : ð7:4:11Þ
ðxÞ ot oxj oxi 3 oxi oxk oxj oxj
It was STOKES who suggested in 1845 (see [2] that the bulk viscosity
vanishes, i.e., k ¼ 23 l. Whilst this may be true for ideal monatomic gases it
is certainly not the case for more complex gas molecules. Also note [3] that
the combination l0 þ 13 l is (sometimes) called bulk viscosity as well. Shear
viscosity is typically measured by using a so-called cone-on-plate viscosim-
eter. Consult the literature as well as the internet and find out how this is done
exactly. The bulk viscosity, however, is difficult to measure. Follow up on
some references in Gad-el-Hak [2] and find out what is known about it so far.
Now define the mechanical pressure by:
1
pme ¼ rkk ð7:4:12Þ
3
168 7 A First Glance on Field Equations
a ¼~
q~ k grad p þ g D~
t: ð7:4:15Þ
Under which circumstances is this equivalent to Eqs. (7.4.2/7.4.11)? What
assumptions are required? Ziegler (p. 198) calls the material
parameter g
dynamic or absolute viscosity and quotes g ¼ 103 Ns m2 for water at
20 C. Is that a new type of viscosity when compared to k and l? Is there
also a static viscosity?
In context with viscous fluids the term NEWTONian fluid is also frequently
used. In which way is a NEWTONian fluid ‘‘different’’ from a NAVIER-STOKES
fluid?
We will now transform Eq. (7.4.2) into the z-system. To this end we start from
the balance of momentum (5.3.2) in contravariant form:
0 i 1
ot
ðzÞ
q @ þ ti ti ;j A ¼ rji ;j þ q f i ð7:4:16Þ
ðzÞ ot ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
Thus:
0 1
oti op
ðzÞ ðzÞ
q@ þ t t ;j A ¼
j i
gji þ k tr ;rj gji
ðzÞ ot ðzÞ ðzÞ oz j ðzÞ ðzÞ
ð7:4:18Þ
þ l t ;rj g þ t ;rj g þ q f i :
j ri i rj
ðzÞ ðzÞ ðzÞ ðzÞ ðzÞ
We now return to the problem of the dropping piston from Sect. 7.2.2 However,
this time we want to study the dynamic problem, i.e., the oscillatory motion of the
piston which, due to friction, will slowly be damped down. The semi-inverse
ansatz for this problem reads:
z_ ðtÞ
t1 ¼ 0; t2 ¼ 0; t3 ¼ x3 : ð7:5:1Þ
zðtÞ
It is worth several comments. First, as weird as it may sound, it can be moti-
vated from HUBBLE’s law of astrophysics. According to that the universe expands
such that the recessional velocity, t, of a far-away galaxy increases proportionally
to its distance, x, from Earth:
t ¼ H0 x: ð7:5:2Þ
2
Credit must be given to Prof. Ingo Müller (TU Berlin) and to Dr. Wolf Weiss (WIAS Berlin)
who brought the arguments in this section to my attention.
170 7 A First Glance on Field Equations
mg NAvo lH Mm e
mR
q ¼ qðtÞ ¼ ¼ ; pðtÞ ¼ T ðtÞ; ð7:5:4Þ
V ðtÞ AzðtÞ AzðtÞ
where the ideal gas law of Sect. 6.4 in its various forms has been used.
This, however, means that the mass balance is identically satisfied and gives us no
further information as we shall convince ourselves now. The first term in Eq. (3.8.7)1
yields:
dq d NAvo lH Mm NAvo lH Mm dz
¼ ¼ : ð7:5:5Þ
dt dt AzðtÞ Az2 ðtÞ dt
And if we observe the ansatz (7.5.1) the second one reads:
otj NAvo lH Mm z_ ðtÞ
q ¼ ; ð7:5:6Þ
oxj AzðtÞ zðtÞ
which is just the negative of (7.5.5). We now turn to the balance of momentum in
global form (3.2.10), apply it to the piston, which as assumed to behave like a rigid
body, and obtain for the three terms:
7.5 The Semi-Inverse Method Applied to Dynamic Gas Flow 171
ZZZ
d d2 z
qp t3 ðx3 ¼ zðtÞ; tÞ dV ¼ mp ;
dt dt2
Vp
ZZ ZZZ ð7:5:7Þ
4 z_ ðtÞ
nj rj3 dA ¼ p l A; qp f3 dV ¼ mp g:
3 zðtÞ
oVp Vp
In the second equation we neglected the pressure p0 acting on the upper side of
the piston. The traction on the bottom side of the piston, n ¼ ð0; 0; 1Þ, results
from the viscous gas, which was modeled as a NAVIER-STOKES fluid with k ¼ 23 l:
Eqs. (7.4.10) and (7.5.1). If we finally replace p by the ideal gas equation and take
the homogeneity of the mass density into account, cf., Eq. (7.5.4), the equation of
motion for the piston results:
d2 z m RTe ðtÞ 4 z_ ðtÞ
mp 2 ¼ lA mp g: ð7:5:8Þ
dt zðtÞ 3 zðtÞ
This is the first of two ordinary differential equation for the two unknowns
z(t) and T(t). Note that it resembles the equation for the harmonic oscillator but
with a nonlinear damping term. However, the damping is still proportional to
velocity as known from standard textbooks. The equation is obviously nonlinear
and must be solved numerically. In this context it is useful to define dimensionless
variables:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e ð 0Þ
RT T ðt Þ zðt Þ
t ¼
2
t; T ðtÞ ¼ ; zðtÞ ¼ : ð7:5:9Þ
mp z ð0Þ T ð0Þ z ð 0Þ
The only non-vanishing term on the right side is the working term. If we insert
our ansatz in Eq. (7.4.10) and ignore bulk viscosity we find:
ZZZ
oti e
m RTðtÞ dz 4 A dz 2
rji dV ¼ þ l : ð7:5:12Þ
oxj zðtÞ dt 3 zðtÞ dt
Vg
172 7 A First Glance on Field Equations
Fig. 7.5 Falling of a piston in an adiabatically sealed cylinder (case of strong damping)
Fig. 7.6 Falling of a piston in an adiabatically sealed cylinder (case of small damping in
comparison to the solution with the adiabatic equation)
piston simply oscillates like a mass on a harmonic spring without friction: Fig. 7.6
(left).
This again can be compared with an analysis of the situation on the basis of the
adiabatic relation shown in Eq. (6.5.29). In this case we find the following dif-
ferential equation for the motion of the piston:
d2 z fþ1
¼ m zj C2 ; j¼ : ð7:5:16Þ
dt2 f
The temperature follows algebraically and not from a differential equation:
All of this is very true, indeed. However, words help at most some philosophers
but certainly not the engineers, especially if the ‘‘dynamical considerations’’ are
neither revealed nor rationally explained. It is therefore not surprising that a vast
amount of literature on this problem started to emerge immediately, and even
today the discussion is not over yet. From the literature we only mention two
papers: Crosignani et al. [5] and Gislason [6]. Their final formulae can directly be
compared with ours even though their derivation is totally different and essentially
based on kinetic theory.
For the solution of the dynamic problem we make exactly the same assumptions
as in the previous example of the falling piston. Thus the ansatz for the velocities
in both vessels (1) and (2) reads (cf., Fig. 7.2 for the meaning of certain symbols):
ð1Þ l_ð1Þ ðtÞ ð1Þ ð2Þ l_ð2Þ ðtÞ ð2Þ l_ð1Þ ðtÞ ð2Þ
t1 ¼ x 1 ; t1 ¼ x1 x ;
ð Þ
l ðt Þ
1 ð
l l ðt Þ
1Þ l lð1Þ ðtÞ 1 ð7:5:18Þ
ð1Þ=ð2Þ
t2=3 ¼ 0:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e ð1Þ ð0Þ
RT ð1Þ T ð1Þ ðtÞ ð2Þ T ð2Þ ðtÞ lð1Þ ðtÞ
t ¼ t; T ðtÞ ¼ ð1Þ ; T ðtÞ ¼ ð1Þ ; lðtÞ ¼ ð7:5:22Þ
mp l 2 T ð 0Þ T ð 0Þ l
This corresponds exactly to Eqs. (6–8) from the paper by Crosignani et al.
(1995) or Eqs. (8–10) from Gislason [6] with one exception: In our derivation the
shear viscosities were constant whereas in the cited papers a square root depen-
dence on temperature was assumed. This is basically due to the aforementioned
strong linkage of their derivation to the kinetic theory of gases, which also predicts
the square root dependence of viscosity on temperature.
If these equations are solved numerically the results are very similar to those
shown in Fig. 7.5, provided the friction due to viscosity is strong enough. It should
ð1Þ ð2Þ
also be noted that if T goes down T goes up and vice versa: If the gas is
compressed in one of the cylinders the gas in the other cylinder must expand. This
in turn reminds us of an air pump, which gets warmer if compressed. Typically the
latter behavior is described with the adiabatic equation. Hence similarly to Eq.
(7.5.16) we obtain:
j ð2Þ j
d2 l mð1Þ l mð2Þ T ð0Þ 1l
¼ : ð7:5:24Þ
dt2 lð0Þ lð0Þ 1 lð0Þ T ð1Þ ð0Þ 1 lð0Þ
We start by deriving the heat conduction equation for a gas (at rest). To this end
we start from the First Law in local regular form according to Eq. (6.5.5) in
Cartesian coordinates:
du i oq i ot i
ð xÞ ð xÞ ð xÞ
q ¼ þ r ji þ q r; ð7:6:1Þ
ðxÞ dt oxi ðxÞ oxj ðxÞ ðxÞ
We now observe the constitutive equation for the internal energy of an ideal
gas, Eq. (6.5.3), FOURIER’s law (6.6.1), the constitutive equation for a frictionless
fluid (6.3.2), and the balance of mass, Eq. (6.5.7):
dT o2 T dt
ðxÞ ð xÞ ð xÞ
q ct ¼ j q p þq r : ð7:6:2Þ
ð xÞ dt ðxÞ ox2i ð xÞ ð xÞ dt ðxÞ ðxÞ
dT o2 T
ð xÞ ðxÞ
q ct ¼ j þq r : ð7:6:3Þ
ðxÞ dt ðxÞ ox2i ðxÞ ðxÞ
However, for isobaric processes we find by using the ideal gas law (6.4.1) and
the relation (6.5.18)3 for the specific heat at a constant volume:
dT o2 T
ð xÞ ð xÞ
q cp ¼ j þq r : ð7:6:4Þ
ðxÞ dt ðxÞ ox2i ðxÞ ðxÞ
Consequently, the differential equation takes the same form in both cases:
0 1
oT oT o2 T
ðxÞ ðxÞ ð xÞ
q ct=p @ þ t i
A¼ j þq r : ð7:6:5Þ
ð xÞ
ot ðxÞ oxi ðxÞ ox2i ðxÞ ðxÞ
Frequently the gas is at rest so the second term on the left side vanishes. This is
the classical equation of heat conduction, namely a parabolic partial differential
equation for the temperature field. Parabolic equations lead to an artifact: They
predict an infinitely fast propagation of disturbances. This means that a temper-
ature peak applied to the center of a large beam (the initial condition) could
immediately be felt everywhere in space independently of the beam size. This is a
consequence of the constitutive laws, in particular of FOURIER’s law. However, the
effect is irrelevant for most engineering applications and the heat conduction Eq.
(7.6.5) serves well as a field equation for temperature.
7.6 The Heat Conduction Equation 177
Josiah Willard GIBBS was born on February 11, 1839 in New Haven,
Connecticut, where he also died on April 28, 1903. He was a theoretical
physicist and is well known for his fundamental work on thermodynamics.
It is fair to say that GIBBS was probably the first American physicist of
worldwide reputation. During his early years he visited the strongholds of
European physics to carry his newly acquired knowledge back home to the
U.S. After that he decided to stay preferably in his home town. Rumor has it
that he also hated to attend scientific congresses.
178 7 A First Glance on Field Equations
By taking into account the dependencies of Eq. (7.6.7) we can expand the time
derivatives
by means of the chain rule and relate the resulting factors to dT=dt and
deij dt:
os ou os ou 1
T ¼ ; T ¼ rji : ð7:6:11Þ
oT e oT e oeij T oeij T q0
Mutual differentiation of both equations and exchanging the sequence of
derivatives yields:
os 1 orji
¼ ð7:6:12Þ
oeij T q0 oT e
If this is inserted into Eq. (7.6.11)2 we obtain:
ou T orji 1
¼ þ rji : ð7:6:13Þ
oeij T q0 oT e q0
This relates the second partial derivative in Eq. (7.6.9) to the stress tensor
(which in thermodynamic terms corresponded to the thermal equation of state for
the case of gases (cf., Sect. 6.4). Moreover, the first partial derivative is already
known: According to Eq. (Sect. 6.5.21) it is related to the specific heat of solids at
a constant strain which, in principle, can be measured, see Exercise 6.5.2. And
therefore:
dT oqj orji deij
q0 ce ¼ þT : ð7:6:14Þ
dt oxj oT e dt
If we now apply FOURIER’s law (6.6.1) the heat conduction equation for solids
results in a form similar to Eq. (7.6.5):
oT oT o2 T orji deij
q0 ce þ ti ¼j 2 þT ð7:6:15Þ
ot oxi oxi oT e dt
For a solid at rest we must choose ti ¼ 0. If we then insert HOOKE’s law in the
form (6.2.6) we find:
oT o2 T dekk
q0 ce ¼ j 2 3ka T : ð7:6:16Þ
ot oxi dt
7.6 The Heat Conduction Equation 179
References
8.1 Introduction
In Chap. 5 we have already extensively dealt with the question how to rewrite
equations of balance in arbitrary, curvilinear coordinate systems. In conclusion we
may say that the balances have the same form in all coordinate system. Additional
coordinate dependent quantities do not occur during a change of coordinate sys-
tem. In Eq. (3.7.3) we have already stated the generic most general form of a
balance: It contains a local time derivative of the quantity to-be-balanced, a
convective and a non-convective flux, a supply and a production term. Indeed, this
structure stays the same, no matter which coordinate system is being used: If for
practical reasons we wish to obtain the equations of balance in a non-Cartesian
coordinate system, we transform each term, which is easily written down in
Cartesian coordinates, from the Cartesian system onto a curvilinear one. This
transition can be made explicit by using the co-/contravariant equations for mass,
momentum, and energy from Chap. 5. Note again that during the transformation
process no additional quantities of a coordinate dependent nature arise, which
otherwise would need to be interpreted, for example, as an additional supply,
which only occurs because of the choice of special coordinates.
In fact, this is not too surprising since a ‘‘decent’’ law of nature must not depend
on the choice of coordinate system. First, its form should be invariant against a
change of coordinates and, second, a change of coordinates must not result in the
creation of new fluxes, supplies, or productions that only exist in certain coordinate
systems. Moreover, the same remark holds analogously for ‘‘decent’’ constitutive
equations.
Note that so far all of our changes of coordinate systems have one thing in
common: We transform the spatial positions, but in a time independent manner.
We are going to change that in this chapter. The coordinate system will now be
switched in a time-dependent manner. In other words, the new coordinate system is
going to ‘‘move’’ w.r.t. the old one. In this context it is customary to refer to a
change of observers. The old coordinate system is a.k.a. the reference frame and
the new one as the moving frame. As we shall see this is going to bring in a whole
new quality to our former arguments. In fact, it will become immediately clear that
this is due to the time derivative in Eq. (3.7.3), which so far could be ignored
completely!
Initially we want to assume naively that the balances for mass, momentum, and
energy were initially verified experimentally by an observer at rest. Such a resting
frame of reference is also known as a Galilean inertial system. We want to answer
the question if and how the balance laws are affected if we change from the inertial
system to an ‘‘arbitrarily’’ moving one. The most general change of observer
within the framework of classical physics allows that, first, the origins of both
systems separate at an arbitrary velocity. Second, it is possible that the coordinate
axes of both systems turn against each other at an arbitrary speed or rather angular
velocity.
Oij′ (t )
b( t )
ej
Now these notions need to be converted into mathematical terms. The Cartesian
unit base of the resting, hence time-independent inertial system is denoted by ej . It
represents an observer who is doing measurements within the inertial frame. In
contrast to that the moving, i.e., time dependent Cartesian unit base e0i ðtÞ represents
the so-called Euclidean observer. In the simplest case both fixate a position P in
space at the same time. This point could be, for example, a material point. The
moving coordinate frame can be (rigidly) attached to a moving material point but
is does not have to be. From Fig. 8.1 we conclude:
or rather:
xj ¼ O0ij ðtÞ x0i b0i ðtÞ ; ð8:2:4Þ
These relations are visualized in Fig. 8.1. Note that the rotation matrix also gets
a dash: Eq. (8.2.2) shows that it is responsible for converting the Cartesian
components xj of the inertial frame into components of the moving system with the
dashed base (summation w.r.t. the second index). However, in Eq. (8.2.4) it is used
to transform components of the moving frame (see the two terms shown within the
parentheses) to quantities of the inertial system (summation w.r.t. the first index).
Alternatively to the component-wise representations of Eqs. (8.2.2) and (8.2.4) we
may also write:
x0 i ¼ O 0 0
ij ðtÞ x j þ b0 i ðtÞ: ð8:2:8Þ
ðx0 Þ 0 ðx ;xÞ ðxÞ ðx Þ
x0 ¼ x þ b )
x0i g0 i ðtÞ ¼ x j ðtÞ g j þ b0i ðtÞ g0 i ðtÞ; x0 i g0i ðtÞ ¼ x j ðtÞ g j þ b0 i ðtÞ g0i ðtÞ:
ðz0 Þ ðz0 Þ ðzÞ ðzÞ ðz0 Þ ðz0 Þ ðz0 Þ ðz0 Þ ðzÞ ðzÞ ðz0 Þ ðz0 Þ
ð8:2:15Þ
Use Eq. (2.5.8) and show that:
where the mixed co-/contravariant forms of the rotation matrix are defined as
follows:
O0 i j V ¼ g0 i ðtÞ g j O0 i
j
¼ g0 i ð t Þ g j
ð8:2:17Þ
ðz0 ;zÞ ðz0 Þ ðzÞ ðz0 ;zÞ ðz0 Þ ðzÞ
where:
O0 ij ¼ g0 i ðtÞ g j O0 ij
¼ g0 i ð t Þ g j
ð8:2:19Þ
ðz0 ;zÞ ðz0 Þ ðzÞ ðz0 ;zÞ ðz0 Þ ðzÞ
186 8 Observers and Frames of Reference in Classical Continuum Theory
O0 0 i j (same question for O0 0 ij )? Discuss the pros and cons of all the different
ðz ;zÞ ðz ;zÞ
forms. Can the upper dash in the symbols O0 0 ij , b00 i , x00 i be omitted? Show
ðz ;zÞ ðz Þ ðz Þ
ox0l 0 oz j
b00 i ¼ 0i
b0 l , b00 l b0l , x j ¼ x l , x l xl , and so on.
ðz Þ oz ðx Þ ðx Þ ðzÞ oxl ðxÞ ðxÞ
Does this make O0 a tensor of second order in the sense of Eq. (2.4.15)?
And, finally, rewrite Eqs. (8.2.10–8.2.14) in curvilinear coordinates.
Now turn to the relations (8.2.6/8.2.7) in invariant notation. Why is there
no dash on the distance vector b? In the same context comment on the
formula
e0i ¼ O0 ej ð8:2:21Þ
and on the following statement translated from the book by Greve [8]: ‘‘The
apparent contradiction between the equations (Note: He is referring to Eqs.
(8.2.1)1 and (8.2.6)1) can be explained that on the one hand side the equa-
tions refer to different objects, namely to the vectors themselves as geometric
objects, and on the other hand side to their representation in different
coordinate systems in form of number triples.’’
In future calculations we will make use of the following useful relations that
follow from Eqs. (8.2.2) and (8.2.4):
ox0i oxi
¼ O0ij ; ¼ O0ji : ð8:2:22Þ
oxj ox0j
Note that Eq. (8.2.2) can also be applied to the case when switching from one
inertial system to another. Such a change of observer is referred to as a GALILEAN
transformation in classical continuum theory. It is characterized by a time-inde-
pendent rotation matrix. Moreover, both origins of the bases move uniformly away
from each other in a straight direction. We write:
be eliminated so that relations for the positions as determined from two Euclidean
observers can be derived. They look completely alike to Eqs. (8.2.2) or (8.2.10).
A0 ¼ A , A0i1 i2 ... il ¼ e0i1 ej1 e0i2 ej2 . . . e0il ejl Aj1 j2 ... jl ,
Ai1 i2 ...il ¼ g0i1 gj1 g0i2 gj2 . . . g0il gjl A j1 j2 ...jl : ð8:3:3Þ
ðz0 Þ ðz0 Þ ðzÞ ðz0 Þ ðzÞ ðz0 Þ ðzÞ ðzÞ
1
For the determinant of Eq. (8.3.1) we could simply write det(O0 ) = ±1. This is a consequence
of Eqs. (8.2.5/8.2.7), which follows from the fact that the inverse of a rotation matrix is given by
its transposed. However, there are didactic reasons why we do not write it that way in Eq. (8.3.1):
In Chap. 13 we will introduce so-called world tensors in complete analogy to Eq. (8.3.1/8.3.2).
However, in contrast to the Euclidean transforms the corresponding world transforms are not
normalized.
188 8 Observers and Frames of Reference in Classical Continuum Theory
In contrast to that the distance vector d between two (material and simulta-
ð1Þ ð2Þ
neous) points of space, P and P , is objective as we can see easily:
ð2Þ ð1Þ ffi
ð2Þ ð1Þ 0 0 ð2Þ ð1Þ
di ¼ x i x i ) di ¼ x i x i ¼ Oij x j x j ¼ O0ij dj
0 0
ð8:3:8Þ
where:
8.3 Objective Tensors and Kinematic Applications 189
Moreover:
Consequently velocity is not an objective vector. This is not too surprising since
the position and the displacement vector were not objective either. Suggestively
speaking, velocity is nothing else but a momentary change of position, in other
190 8 Observers and Frames of Reference in Classical Continuum Theory
t0 ¼ t; ð8:3:19Þ
but a change of position is not.
Moreover, Eq. (8.3.18)2 deserves another comment: If it were only for the first
term after the equals sign, velocity would be an objective vector, too. tj denotes
the components of the velocity of the point P w.r.t. the Cartesian base of an inertial
system. By means of the matrix O0ij these components are turned onto the coor-
dinate system of the Euclidean observer, so to speak. Correspondingly, O0rj xj
x0r b0r are the components of the position vector x in the Euclidean system. The
quantity X0ir is characteristic for a non-inertial system, and this is why we add a
dash to the symbol. The matrix X0ir vanishes for transformations between inertial
systems. This becomes evident by looking at the Galilean transformation (8.2.23):
Due to the time-independence of the rotation matrix the matrix of angular
velocities X0ir vanishes by its very definition for a Galilean observer: Eq. (8.3.17).
The matrix of angular velocities is a 3 9 3 matrix. However, it is not completely
populated and contains only three independent components because X0ir is anti-
symmetric, as can be seen from Eq. (8.3.16)2:
Tullio LEVI-CIVITA was born on March 29, 1873 in Padua and died on
December 29, 1941 in Rome. He was a mathematician down to the
very bottom of his heart. However, first he accepted a professorship for
mechanics in Padua in 1898 and obtained a chair for mathematics in
Rome much later in 1918. Being Jewish the fascists kicked him out of
office in 1938. We owe him important contributions to tensor analysis.
Rumor has it that he was the ‘‘inventor’’ of the covariant derivative. It
is certain though, that he provided the mathematical framework which
allowed Albert EINSTEIN to create his theory of general relativity.
8.3 Objective Tensors and Kinematic Applications 191
it is guaranteed that
8
< þ1 i; j; k ¼ 1; 2; 3 and cyclic permutations
0ijk ¼ 1 if i; j; k ¼ 2; 1; 3 and cyclic permutations ð8:3:28Þ
:
0 else
in the transformed system as well. Moreover, consider the following two
vectors:
a ¼ a e1 ; b ¼ b e2 ð8:3:29Þ
and show by using Eq. (8.3.24) that:
192 8 Observers and Frames of Reference in Classical Continuum Theory
c ¼ ab e3 : ð8:3:30Þ
Now use the transformation rules for true vectors:
c0 ¼ þabe03 : ð8:3:34Þ
Finally confirm with the results of this exercise the equations shown in
(8.3.21). In particular confirm that the factor 1=2 is correct. Use and prove in
the same context the following auxiliary formula:
ijk klm ¼ dil djm dim djl : ð8:3:35Þ
If we now wish to establish a vector relation for the velocity in analogy to the
transition from Eqs. (8.2.1) to (8.2.2) we must write:
_
t0 ¼ t x x0 þ b: ð8:3:37Þ
Note that the vector of angular velocity does not carry a dash just like the
distance vector b in the vector relation (8.2.1)1. And just like the matrix X0ij the
vector of angular velocity is a quantity that indicates the presence of a non-inertial
system (which is what the dashed system in general represents) if it does not
vanish. Like all vectors we can span the vector of angular velocity x w.r.t. the base
ei (with the components xi ) or in the base e0i (with the components x0i ). The latter
has already been done in Eq. (8.3.36). By its very definition, shown in Eq. (8.3.21),
the vector of angular velocity is an axial quantity and transforms like:
The correctness of the first two terms after the equal sign in Eq. (8.3.37) is
easily confirmed in analogy to the quantity O0ij xj from Eqs. (8.3.36) and (8.2.1/
8.2.2), and by observing the results from Exercise 8.3.2. The last term in Eq.
(8.3.36) becomes obvious if we recall that the base e0i is time dependent:
0
b_ ¼ b0i e0i ¼ b_ i e0i þ b0i e_ 0i : ð8:3:39Þ
Moreover, we may write:
0 0
e_ 0i ¼ O_ ij ej ¼ O_ ij O0kj e0k ¼ X0ik e0k ¼ 0ikl x0l e0k ð8:3:40Þ
and also:
b0i e_ 0i ¼ 0ikl x0l b0i e0k ¼ 0ilk x0l b0k e0i : ð8:3:41Þ
Thus:
0
0
b_ ¼ b_ i þ 0ilk x0l b0k e0i : ð8:3:42Þ
Note that Eq. (8.3.37) is the transformation rule for the velocity typically found
in elementary mechanics textbooks. In the same context strange looking formulae
like
dð Þ d0 ð Þ
¼ þ x ðÞ ð8:3:43Þ
dt dt
are often presented, where the dash at the differentiation symbol is supposed to
indicate a time derivative in the moving system. Such definitions and consider-
ations are not required here. Be that as it may, the final result is the same.
In contrast to the gradients of displacements, gradients of velocity are not
objective. In order to show that explicitly we differentiate Eq. (8.3.36) w.r.t. the
position:
ot0i otk oxl otk
¼ O0ik 0irs x0r dsj ¼ O0ik O0jl þ 0ijr x0r : ð8:3:44Þ
ox0j oxl ox0j oxl
This is a very important and satisfying result in context with the NAVIER–STOKES
constitutive relation, which guarantees that it keeps its form in the moving frame.
But what about acceleration? We start from Eq. (8.3.18)2 by identifying:
194 8 Observers and Frames of Reference in Classical Continuum Theory
ai t_ i
::
) a0i ¼ O0ij aj þ O_ 0ij tj þ X_ 0ir ðx0r b0r Þ þ X0ir O_ 0rj xj þ X0ir O_ 0rj tj þ b0i ð8:3:46Þ
::
¼ O0ij aj þ 2X0ir ðt0r b_ 0r Þ X0ir X0rs ðx0s b0s Þ þ X_ 0ir ðx0r b0r Þ þ b0i :
It is customary to refer to
• 2X0ir t0r b_ 0r as CORIOLIS acceleration,
• X0ir X0rs x0s b0s as centrifugal acceleration,
• X_ 0ir x0r b0r as EULER acceleration and
• €b0 as relative acceleration.
i
What are the names of the various acceleration terms in this equation
(observe the signs)? Is there also a centripetal acceleration or how would it
look like?
We now turn to the question how the balance of mass in its form (3.8.3)1 behaves
under Euclidean transformations. Judging by the results of Sect. 5.1 we expect that
it will have the same form in the Euclidean system as in the inertial system and we
may simply replace all quantities by dashed ones. However, this is just a guess and
we should be cautious since the Euclidean transformation is time dependent and
the balance of mass contains a time derivative that might cause trouble. In other
words without further calculations we cannot be sure of a potential impact of the
time derivative. Hence we argue as follows:
First of all we shall assume that the mass of a material particle is an objective
scalar, at least within the framework of classical physics:
dV 0 ¼ dV; ð8:4:2Þ
if interpreted as the absolute value of a scalar triple product of three (infinitesi-
ðiÞ
mal) distances d x :
ð1Þ ð2Þ ð3Þ
dV ¼ d x d x d x : ð8:4:3Þ
This makes sense since because of Eq. (8.3.8) the distance vector between two
points is objective. Note that a volume element defined by a pure scalar triple
product would be an axial scalar due to the vector product and this is why we take
its absolute value. If we now observe the definition q ¼ dm=dV we conclude that
mass density is an objective scalar:
196 8 Observers and Frames of Reference in Classical Continuum Theory
q0 ¼ q: ð8:4:4Þ
Next we consider the material time derivative of an objective scalar, in par-
ticular of the mass density, q. We recall that by definition we may write in
Lagrangian or in Eulerian representation (cf., Sect. 3.8):
dq oqðX; tÞ oq oq
q_ ¼ ¼ þ ti : ð8:4:5Þ
dt ot ot oxi
The Lagrangian representation is easily converted:
oqðX; tÞ oq0 ðX; tÞ
¼ , q_ ¼ q_ 0 : ð8:4:6Þ
ot ot
Thus the material time derivative of an objective scalar is also objective. The
proof is a little more cumbersome in Eulerian representation. In that case we find
for the first part of the material time derivative:
oqðx; tÞ oq0 ðx0 ; tÞ ox0i oq0 oq0 oq0
¼ 0 þ ¼ 0 O_ 0ij xj þ b_ 0i þ
ot x oxi t ot ot x0 oxi ot
0
ð8:4:7Þ
oq oq 0
¼ 0 X0ik x0k b0k þ b_ 0i þ ;
oxi ot
The same holds for the unit normal vector. It results from dA by dividing it by
the surface area jdAj. The latter is always positive and an objective scalar, just like
the volume element, because the size of an area should not depend on the observer,
at least not in classical physics [note that this property also follows from Eq.
(8.5.3)]. Hence we find:
0
dA ¼ dAj ) n0 ¼ detðO0 ÞO0 nj : ð8:5:4Þ
i i ij
According to Eq. (3.2.7) the traction is a linear function of the unit normal. For
this reason we require that it is an axial tensor of first order:
Thus in case of surface forces we shift our preference for objective (and not
axially objective) force-related quantities from the traction vector to the stress tensor.
Finally recall the transformation rule for the acceleration according to Eq.
(8.3.46) as well as the balance of momentum in the form (3.8.14)1, from which the
mass balance has already been extracted, and the validity of which has been
established in an inertial system, so to speak. Hence the momentum balance
assumes the following form in a Euclidean system:
or0ji h :: i
q0 t_ 0i ¼q 0 0
f þ q 0
2X0
ðt 0
_ 0 Þ X0 X0 ðx0 b0 Þ þ X_ 0 ðx0 b0 Þ þ b0 :
b
i ik k k il lk k k ik k k i
ox0j
ð8:5:9Þ
We must conclude that the form of the balance of momentum in a Euclidean
system differs considerably from the one in an inertial system: In contrast to the
transformation behavior of the mass balance all of a sudden system dependent
terms appear and the form invariance is at stake. However, a simple trick comes to
the rescue: The various accelerations are combined to form a new specific volu-
metric force:
0
f^0 i ¼ fi0 þ 2X0ik t0k b_ 0k X0il X0lk x0k b0k þ X_ 0ik x0k b0k þ €bi : ð8:5:10Þ
Thus we may write:
or0ji
q0 t_ 0i ¼ þ q0 f^0i ; ð8:5:11Þ
ox0j
which corresponds exactly to the form of the balance of momentum in the inertial
frame. To this extend we may talk of form invariance of the balance of momentum
under the presence of system depending quantities. However, the whole story gets a
mystical touch if we begin to refer to the various accelerations multiplied by the
8.5 The Balance of Momentum in a Moving Coordinate System 199
mass density as fictitious forces after moving them to the right hand side of the
momentum balance. The adjective ‘‘fictitious’’ is an extremely unfortunate choice
of a word. On the contrary! These ‘‘fictitious forces’’ are extremely real. We can
convince ourselves immediately of their presence during a car accident. And what
is more, in their effect they cannot be distinguished at all from the protagonist of the
force density fi0 , which is gravity. In the latter case we got used to saying that its
cause is the attraction between gravitational masses. But we neither explain why
gravitational masses attract each other nor what a gravitational mass really is, at
least within the framework of classical continuum physics. However, we can
compare gravitational masses with each other by placing them on the bowls of a
beam scale. In fact this phenomenological comparison works in all gravitational
fields, on Earth, on the Moon, etc., and will lead to the same result everywhere.
q_t0 ¼ r0 r þ q f 2x t0 x ðx x0 Þ x_ x0 þ €b : ð8:5:12Þ
Why do the symbols q, r, and f carry no dashes? Note that the expressions
2qx t0 , qx ðx x0 Þ, qx_ x0 , and q0 €b are known as CORIOLIS
force, centrifugal force, EULER force, and force of relative translation,
respectively. They are all inertia forces.
Instead of using the term fictitious forces one should rather talk of inertia forces
(cf., Meriam (1978), p. 201), since this is exactly what these system dependent
forces are: Masses seem to have a certain perseverance to remain in their original
state of motion. In other words they are ‘‘inert’’ and in order to accelerate them or
to change the course of their motion forces are required. This was NEWTON’s great
discovery. At the beginning of his famous book Philosophiae Naturalis Principia
Mathematica (1726)2 (Mathematical Principles of Natural Philosophy) he distin-
guishes in the Definitiones two kinds of forces. On the one hand side there is the vis
insita, in other words a force proper to matter, which we call inertia force today
(with the exception of the sign):
2
The Latin citations stem from the third edition of NEWTON’s book and can be found in the most
carefully edited two volumes by Koyré et al. [9]. When compared to the first edition of the
Principia (1687) we notice considerable differences in the wording, in terms of alterations as well
as additional comments. This is an indication of NEWTON’s lifelong struggle with his findings, and
it also shows how his understanding grew steadily over time. Moreover, note, that all but one
translation of the original Latin text stem from the book of Chandrasekhar (1995). The translation
of the ‘‘hypotheses non fingo’’ passage is from Cohen and Whitman (1999).
200 8 Observers and Frames of Reference in Classical Continuum Theory
Definitio III. Materiae vis insita est potentia resistendi, qua corpus unum-
quodque, quantum in se est, perseverat in statu suo vel quiescendi vel movendi
uniformiter in directum. (Definition III. The vis insita, or innate force of matter, is a
power of resisting, by which every body, as much as in it lies, continues in its
present state, whether it be of rest, or of moving uniformly forwards in a right line.)
On the other hand side there is the vis impressa (in other words: an ‘‘applied’’
force, which comes from outside and which belongs on the right hand side of the
momentum balance):
Definitio IV. Vis impressa est actio in corpus exercita, ad mutandum ejus statum
vel quiescendi vel movendi uniformiter in directum. (Definition IV. An impressed
force is an action exerted upon a body, in order to change its state, either of rest, or
of uniform motion in a right line.)
It becomes particularly fascinating when NEWTON explains in his fifth definition
the notion of a centripetal force for the first time, since his centripetal force is not
what we mean by that term today, i.e., not qx ðx x0 Þ:
Definitio V. Vis centripeta est, qua corpora versus punctum aliquod, tanquam
ad centrum, undique trahuntur, impelluntur, vel utcunque tendunt. (Definition V.
A centripetal force is that by which bodies are drawn or impelled, or any way tend,
towards a point as to a centre.)
He continues to explain and says: Hujus generis est gravitas, qua corpora
tendunt ad centrum terrae; vis magnetica, qua ferrum petit magnetem; …. Lapis,
in funda circumactus, a circumagente manu abire conatur; & conatu suo fundam
distendit, eoque fortius quo celerius revolvitur; &, quamprimum dimittitur, avolat.
Vim conatui illi contrariam, qua funda lapidem in manum perpetuo retrahit & in
orbe retinet, quoniam in manum ceu orbis centrum dirigitur, centripetam appello.
(Of this sort is gravity, by which bodies tend to the centre of the Earth; magnetism,
by which iron tends to the loadstone; …. A stone, whirled about a sling,
endeavours to recede from the hand that turns it; and by that endeavour, distends
the sling, and that with so much the greater force, as it is revolved with the greater
velocity, and by which the sling continually draws back the stone towards the
hand, and retains it in its orbit, because it is directed to the hand as the centre to the
orbit, I call the centripetal force). Thus NEWTON’s concept of a centripetal force is
not a kinematic one. To him a centrifugal force is linked to physical causes, which
we summarize today in the term qf (for gravity and, in context with magnetism,
the LORENTZ force) and/or r r (for the tension in the string of a sling to which a
stone is attached).
Then NEWTON states his First Law:
Lex I. Corpus omne perseverare in statu suo quiescendi vel movendi uniformiter
in directum, nisi quatenus illud a viribus impressis cogitur statum illum mutare.
(Law I. Every body continues in its state of rest, or of uniform motion in a right
line, unless it is compelled to change that state by forces impressed upon it.)
On first glance one might think that NEWTON’s First Law is just a special case of
his Second Law, if applied to a case where no forces are present:
Lex II. Mutationem motus proportionalem esse vi motrici impressae, & fieri
secundum lineam rectam qua vis illa imprimitur. (Law II. The change of motion is
8.5 The Balance of Momentum in a Moving Coordinate System 201
proportional to the motive force impressed; and is made in the direction of the
right line in which that force is impressed.)
However, in its final consequence NEWTON’s First Law implies the existence of
an inertial system, where the body is at rest or moves uniformly along a straight
line due to lack of forces. Moreover, in his Second Law we recognize the simplest
form of a balance: A quantity (namely ‘‘motus’’) changes in time (‘‘mutatio’’) due
to an applied propelling force (‘‘vi motrici impressae’’). The former is the effect
and the latter the cause. Also note that NEWTON explains a few pages before that his
‘‘motus’’ is exactly that what we call momentum today:
Definitio II. Quantitas motus est mensura ejusdem orta ex velocitate et quan-
titate materiae conjunctim. (Definition II. The quantity of motion is a measure of
the same, arising from the velocity and quantity of matter conjointly.)
Frequently the question is asked as to whether the force (in other words the
right hand side of the Second Law) is defined by the left hand side, i.e., the change
of momentum or, in other words, for a constant mass by a change of position, i.e.,
geometry. This is not so. In this context force is considered as a ‘‘primitive’’
quantity, and the Second Law serves by no means as an equation for its definition.
Thus it is only fair to ask how a measurement standard for a force can be defined.
The answer to this is an evasive one: In the end the standard must be based on a
deeper understanding of the origin of force, for example on an atomic or subatomic
level. This holds for the volumetric force density qf , in particular gravity but also
other inertia forces, as well as for surface force related terms like r r. In the
latter case we have constitutive relations for the stress tensor and the underlying
materials theory (which is not subject of this book, at least not in detail) in mind.
But again, a profound understanding of these concepts also takes place on various
scales, on the macro-, the meso-, and on the micro-level.
When NEWTON spoke of forces the attraction of masses was definitely on his
mind, which he so aptly described by his famous law of gravitation. Rather at the
end of his Principia (in the Scholium Generale after Propositio XLII) NEWTON
muses that so far he had not been able to find the cause for the existence of gravity:
Rationem vero harum gravitatis proprietatum ex phaenomenis nondum potui
deducere, & hypotheses non fingo. Quicquid enim ex phaenomenis non deducitur,
hypothesis vocanda est; & hypotheses seu metaphysicae, seu physicae, seu qual-
itatum occultarum, seu mechanicae, in philosophia experimentali locum non
habent. I have not as yet been able to discover the reason for these properties of
gravity from phenomena, and I do not feign hypotheses. For whatever is not
deduced from the phenomena must be called a hypothesis; and hypotheses,
whether metaphysical or physical, or based on occult qualities, or mechanical,
have no place in experimental philosophy. In this philosophy particular proposi-
tions are inferred from the phenomena, and afterwards rendered general by
induction.) In the spirit of his laconic phrase ‘‘hypotheses non fingo’’ force is for us
nothing else but a primitive quantity.
Besides CORIOLIS force and Co. it is customary in mechanics to call the (neg-
ative) temporal change of momentum an inertia force as well. This allows
rewriting NEWTON’s Second Law in the form that ‘‘the sum of all forces including
202 8 Observers and Frames of Reference in Classical Continuum Theory
inertial ones is equal to zero.’’ Thus all of the computational techniques we have
been introduced to in statics can usefully be transferred to dynamics. As conve-
nient as this may be, it is conceptionally counterproductive: That way the
important aspect that the additive quantity momentum obeys a balance law is
completely lost. The dubious method is due to D’ALEMBERT [4], and I am inclined
to say that it is probably more of a reaction of a member of the Académie
Française to the dominant Anglo-Saxon way of mechanics than a deep insight.
Let us now turn to NEWTON’s Third Law according to which to each force (the
‘‘action’’) there belong a responding force (the ‘‘reaction’’):
Lex III. Actioni contrariam semper & aequalem esse reactionem: sive corporum
duorum actiones in se mutuo semper esse aequales & in partes contrarias dirigi. (Law
III. To every action there is always opposed an equal reaction: or, the mutual actions
of two bodies upon each other are always equal, and directed to contrary parts.)
Ernst Waldfried Josef Wenzel MACH was born on February 18, 1838 in
Chirlitz-Turas, Moravia, today Brno, and he died on February 19, 1916 in
Vaterstetten near Munich. As a true Austrian multi-talent MACH was
physicist, philosopher, psychologist, and theoretician of the sciences.
Above all his name is well known from the MACH number used in
supersonic aviation. However, he is also one of the most influential
representatives if not the founder of empirio-criticism. In psychology he
prepared the way for the so-called gestalt psychology and gestalt theory.
Note that NEWTON does not say anything about how fast this reaction builds up,
it is just ‘‘there:’’ The Earth is pulling the moon, and the moon strikes back,
instantaneously. Moreover, in some mechanics textbooks (e.g., Hauger, Schnell,
Gross, 1993) one can find the following interesting comment of D’ALEMBERT’s
inertia force: ‘‘This force is no force in the sense of NEWTON since it does not come
with a reaction force (it violates the axiom action = reaction!).’’ This may be so,
but NEWTON also never claimed this to be the case either. Rather he clearly had the
imposed forces of his First and Second Law in mind (mark his words ‘‘Vis im-
pressa est actio…’’ in Definitio IV and ‘‘… a viribus impressis…’’ in Lex I as well
as ‘‘… vi motrici impressae…’’ from Lex II) and not the temporal change of
momentum representing the quantity known as inertia force today. He emphasizes
8.5 The Balance of Momentum in a Moving Coordinate System 203
this in his book by means of several examples, e.g., a stone pressing on a finger
and vice versa or a horse pulling a stone that prevents the horse’s motion since it is
pulling back. Nothing can be found at this place about CORIOLIS force & Co. nor
does NEWTON claim that they obey the action = reaction principle. This shows
even more that D’ALEMBERT’s approach has a certain aftertaste.
It is only fair to pose the question what the origin of inertia accelerations really
is. To say that they are a consequence of correct differentiation of spatial coordinate
transformations w.r.t. time is not a truly satisfying answer. It was the Viennese
physicist and natural philosopher Ernst MACH who attempted to give an answer,
albeit a sibyllic one. Einstein [6] picks up MACH’s ideas and says: ‘‘Analogously, I
seek in vain for a real something in classical mechanics (or in the special theory of
relativity) to which I can attribute the different behaviour of bodies considered with
respect to the reference-systems K and K0 .1 Newton saw this objection and
attempted to invalidate it, but without success. But E. Mach recognised it most
clearly of all, and because of this objection he claimed that mechanics must be
placed on a new basis. It can only be got rid of by means of a physics which is
conformable to the general principle of relativity, since the equations of such a
theory hold for every body of reference, whatever may be its state of motion.’’ In
the footnote 1 of the text he becomes even more explicit: ‘‘1 The objection is of
importance more especially when the state of motion of the reference-body is of
such a nature that it does not require any external agency for its maintenance, e.g. in
the case when the reference-body is rotating uniformly.’’ In his German paper of
(1914) [5] he says clearly what the physical origin of the forces arising during
rotation is, unfortunately without a translation. However, there is an English
equivalent to that text in Einstein [7]: ‘‘Can gravitation and inertia be identical?
This question leads directly to the General Theory of Relativity. It is not possible to
regard the earth as free from rotation, if I conceive of the centrifugal force, which
acts on all bodies at rest relatively to the earth, as being a ‘‘real’’ field of gravitation,
or part of such a field? If this idea can be carried out, then we shall have proved in
very truth the identity of gravitation and inertia. … According to Newton, this
interpretation is impossible, because by Newton’s law the centrifugal field cannot
be regarded as produced by matter, and because in Newton’s theory there is no
place for a ‘‘real’’ field f the ‘‘Koriolis-field’’ type.’’ In short: The fictitious forces
have their origin in the gravitational action of far away masses.
Jean Bernard Léon FOUCAULT was born on September 18, 1819 in Paris
and also died there on February 11, 1868. From 1829 on he attended first
the Collège Stanislas in Paris. Due to his laziness and rude behavior he
was advised to leave school so that he had to get his further education
from a private teacher. He quit medical studies due to insurmountable
aversion of dissection and dedicated himself to physics from then on as
an autodidact. In this field he showed an enormous experimental talent
and so he presented publicly in 1851 the famous pendulum named after
him. Around 1850 he determined the speed of light by using a cheval
mirror construction. In 1855 he invented a typewriter, and so on. His dislike for medicine finally
took vengeance when he tragically became almost blind and mute and died of aphasia.
204 8 Observers and Frames of Reference in Classical Continuum Theory
Hans THIRRING was born on March 23, 1888 in Vienna where he also died
on March 22, 1976. He studied mathematics, physics and—most use-
ful—gymnastics at the University of Vienna until 1910. In 1911 he
became an assistant at the Institute for Theoretical Physics of the uni-
versity, where he also obtained his Ph.D. in 1911 and presented his
habilitation thesis in 1915. In 1921 he became an associate professor and
in 1927 finally full professor. Although he did mostly theoretical work
he was also open to technology. For example, he invented a method for
the production and playback of talkies and started his own business
where he exploited that technology. However, in 1938 the Nazis forced
him into ‘‘early retirement:’’ His interest in ‘‘Jewish physics,’’ such as relativity, his friendship
with EINSTEIN and FREUD and his notorious pacifism provided enough reasons. After the war he
became the Dean of the Department of Philosophy at the University of Vienna.
This is what THIRRING [12, 13] did in order to explain the existence of inertia
forces as the effect of far away rotating masses. In other words he tried to ‘‘verify’’
MACH’s idea starting from EINSTEIN’s field equations. He considered a thin-walled
hollow sphere of homogeneous mass density (total mass M, radius a). An observer
is located in its center turning at a constant angular velocity x about the x03 -axis.3
For this kind of mass distribution he determined approximately the four-dimen-
sional metric tensor (cf., Sect. 13.11) by solution of EINSTEIN’s field equations.
Once the metric was known he used the geodesic equation and studied the motion
of a test mass-point inside of the hollow sphere. Using the nomenclature of Eq.
(8.5.9) he derived the following formulae:
3
Strictly speaking THIRRING’s analysis also allows the hollow sphere to rotate with an angular
velocity different from x.
8.5 The Balance of Momentum in a Moving Coordinate System 205
ffi ffi
GM _0 GM
€x01 ¼ 2x 1 þ x 2 þ x 2
1 þ x0 ; ð8:5:13Þ
4pc2 a 4pc2 a 1
ffi ffi
0 GM _0 GM
€x2 ¼ 2x 1 þ 2
x1þx 1þ x0 ; €z0 ¼ 0:
4pc2 a 4pc2 a 2
G ¼ 6:67 1011 m/kg s2 denotes the gravitational constant and c ¼ 3 108 m/s
is the speed of light. The terms on the right hand side correspond w.r.t. their
dependence of the angular velocity and the coordinates exactly to the CORIOLIS- and
to the centrifugal acceleration—up to relativistically small corrections. We will
explore this in detail in the following exercise.
e2
e ′2
P(t)
x2
e1′
x ′2 x ′1
α (t )
x1 e1
2 3 2 3
cos a sin a 0 0 x 0
O0ij ¼ 4 sin a cos a 0 5; X0ij ¼ 4 x 0 0 5; ð8:5:14Þ
0 0 1 0 0 0
where the absolute value of the angular velocity x ¼ a_ has been introduced.
Use the result to show that:
Moreover, show that we obtain for the various parts in the equation of
motion (8.5.12) (all components are w.r.t. the non-inertial system e0i ):
x0 ¼ x01 ; x02 ; x03 ; t_ 0 ¼ €x01 ; €x02 ; €x03 ; x t0 ¼ x _x02 ; x_ 01 ; 0 ; ð8:5:16Þ
x ðx x0 Þ ¼ x2 x01 ; x02 ; 0 ; x_ x0 ¼ x_ x02 ; x01 ; 0 ; b ¼ ð0; 0; 0Þ:
Insert this into the equations of motion (8.5.12) and demonstrate that in
Cartesian coordinates of the non-inertial system we have (while neglecting
mechanical stresses and imposed volumetric forces):
ac ¼ x ðx x0 Þ: ð8:5:20Þ
In order to obtain specific numbers, specialize Eq. (8.5.10) to the case of a
wheel-like station rotating about its ‘‘hub’’ w.r.t. the fixed stars with a
constant angular velocity x0 :
x ¼ x 0 e3 : ð8:5:21Þ
How big is x0 and how strong is the corresponding CORIOLIS force acting
on an astronaut who is moving radially forward from the center at pedestrian
speed?
There are several ways to derive the balance of kinetic energy in a moving system.
One possibility is scalar multiplication of the balance of momentum in the moving
frame, Eq. (8.5.12), by velocity t0 and to perform a few algebraic manipulations as
follows:
1
q t02 ¼ r0 ðr t0 Þ r : r0 t0 þ q f x ðx x0 Þ x_ x0 þ €b t0 :
2
ð8:6:1Þ
Obviously, there are additional supplies of kinetic energy to the well known
one, f t0 , namely the power of the centrifugal acceleration, the EULER accelera-
tion, and the relative acceleration.
Alternatively we may start from Eq. (8.5.9), multiply it by t0 and rewrite. The
dashes characteristic of the moving frame allow to distinguish very clearly which
components we refer to:
ffi
1 or0ji t0i ot0
q0 t0i t0i ¼ 0 r0ji i0
2 oxj oxj
h :: i
þ q0 fi0 X0il X0lk x0k b0k þ X_ 0ik x0k b0k þ b0i 2X0ik b_ 0k t0i :
ð8:6:2Þ
208 8 Observers and Frames of Reference in Classical Continuum Theory
u0 ¼ u; r 0 ¼ r: ð8:6:4Þ
Moreover, the heat flux is assumed to behave like a Euclidean vector:
The heat flux density q ¼ ^qðx; t; nÞ is (due to its dependence of the axial normal
vector) a rare example of an axial objective tensor of zeroth order, i.e., an axial
scalar:
ot0i o 0 oxk
r0ji 0 ¼ O0js O0it rst Oil tl þ X0ir x0r b0r þ b_ 0i
oxj oxk ox0j
ffi
otl otl otl
¼ O0js O0it rst O0il þ X0ir O0rk O0jk ¼ rkl þ O0rs O0it rst X0ir ¼ rkl :
oxk oxk oxk
ð8:6:9Þ
Thus Eq. (8.6.3) reads in a Euclidean system:
oq0k 0 oti
0
q0 u_ 0 ¼ þ r ji þ q0 r 0 : ð8:6:10Þ
ox0k ox0j
Hence the form of the total energy balance is determined. By addition of Eqs.
(8.6.2) and (8.6.10) we obtain:
ffi
1 oq0 or0ji t0i
q0 u0 þ t0i t0i ¼ 0k þ
2 oxk ox0j
h :: i
þ q0 fi0 X0il X0lk x0k b0k þ X_ 0ik x0k b0k þ b0i 2X0ik b_ 0k t0i þ r0 :
ð8:6:11Þ
In Sect. 8.4 the (material) time derivative of a scalar quantity, namely of the mass
density q, has already been discussed: Eq. (8.4.6). However, in Chap. 10 and 11,
dedicated to fluids with memory and to an introduction to plasticity, we will
encounter in the constitutive equations time derivatives of Euclidean tensors of
higher order, in particular of the stress tensor. Consequently we must pose the
question how they transform during change of frames. In particular we expect the
constitutive equations to keep their form during the change. Moreover, no frame
specific quantities should occur. We now turn to that problem and investigate the
time derivative of an objective tensor of first order, i.e., of a Euclidean vector, first:
210 8 Observers and Frames of Reference in Classical Continuum Theory
James Gardner OLDROYD was born in April, 1921 in Bradford and died on
November 22, 1982. He first went to Bradford Grammar School and then
attended Trinity College at the University of Cambridge. After his
graduation he worked for the Ministry of Supply during World War II.
After the war he went to the research laboratories of Courtaulds. In 1953
he became a professor of mathematics at the University of Wales in
Swansea. In 1965 he moved to Liverpool University and became Head of
Department of Applied Mathematics and Theoretical Physics in 1973
until his death. His main research was dedicated to the visco-elastic
behavior of Non-NEWTONIAN fluids.
We will now show that its material time derivative is also not objective. For this
purpose we first apply the operator dt d ð Þ to Eq. (8.7.6) and obtain while
observing Eq. (8.7.3):
Clifford Ambrose TRUESDELL III was born on February 18, 1919 in Los
Angeles and died on January, 14, 2000 in Baltimore. He first studied
physics and mathematics at Caltech. In 1943 he obtained a Ph.D. from
Princeton. From 1944 until 1946 he worked at M.I.T.’s Radiation Lab-
oratory and then until 1950 for the Naval Research Laboratory in
Washington, D.C. In 1950 he became a professor for mechanics at
Indiana University and, finally, in 1961 professor for ‘‘Rational
Mechanics’’ at the Johns Hopkins University. His most important
achievement are most certainly the various monographs by means of
which he established a rational mathematically based way of thinking in
mechanics as well as thermodynamics as a fundamental principle.
Marius Sophus LIE was born on December 17, 1842 in Nordfjordeid and
died on February, 18, 1899 in Christiania (today Oslo). In Christiania he
studied the sciences, obtained a teacher’s degree in 1865, and dedicated
himself from 1868 onward completely to mathematics. A stipend allowed
him to go abroad and so he met the famous German mathematician Felix
KLEIN. This acquaintance resulted in several joined papers. In 1872 he
became a professor in Christiania and in 1886 successor to KLEIN’s chair
in Leipzig. LIE’s health and psyche were quite delicate. He suffered from
pernicious anemia, mental breakdowns and fought regularly with his
colleagues about priority issues. In 1894 he was awarded a personal professorship by the
Norwegian government in Christiania. However, he returned not before 1898 as a very sick
man.
212 8 Observers and Frames of Reference in Classical Continuum Theory
Finally in this chapter we want to touch briefly upon the question regarding the
form invariance of the constitutive equations mentioned in Chap. 6. In the case of
HOOKE’s law we start from Eq. (6.2.6). In order to change the frame of reference
we assume that LAMÉ’s constants, the coefficient of thermal expansion, and the
temperature are Euclidean scalars. Moreover, it follows from the orthogonality
relations (8.2.5) that the KRONECKER symbol transforms like a Euclidean tensor of
second order:
By means of the transformations properties of the linear strain tensor and of the
stress tensor shown in Eqs. (8.3.14) and (8.5.6) we conclude that the form of
HOOKE’s law for a Euclidean observer is the same as in an inertial system:
Similarly we have form invariance for the NAVIER-STOKES law of Eq. (6.3.1) if
we, first, assume that the pressure is a Euclidean scalar. In fact it has to be, because
it is the trace of the stress tensors and in view of Eqs. (2.6.11), (6.3.2), and the
transformations rule for the KRONECKER symbol shown above. Second, we observe
the transformation property of the symmetric velocity gradient according to Eq.
(8.3.45). Thus:
In view of all that and by observing that the mass density is a Euclidean scalar
[cf., Eq. (8.4.4)] it follows that the ideal gas law from Eq. (6.4.1) is also form
invariant. We have already mentioned that the internal energy is a Euclidean scalar:
Eq. (8.6.4). This in turn is consistent with the representation of the caloric equation
of state for the ideal gas in Eq. (6.5.3). As far as FOURIER’s law of heat conduction in
8.8 A Remark on the Form Invariance of Constitutive Equations 213
its form (6.6.1) is concerned it is also form invariant based on Eq. (8.2.22) and if we
consider the coefficient of thermal conductivity as a Euclidean scalar:
oT 0
q0i ¼ j0 : ð8:8:4Þ
ox0i
In all the usual textbooks on continuum mechanics the problem of a change of the
frame of reference is a major point of discussion, cf., for example, Greve [8] Sect.
1.4, Bertram [2] in Sect. 4.3 (with additional comments on the Principle of
Material Objectivity (PMO), which is not covered in this book), Becker and
Bürger [1] in Sect. 1.5, Liu [10], Sects. 1.7 and 3.2 (on the PMO), and in Truesdell
[14] on pages 22 pp. and on the P.M.O. on pages 39 pp. The cited text passages
also contain remarks on objective time derivatives. In the handbook article by
Truesdell and Toupin [16] the problem of a change to a Euclidean frame is treated
in Sect. 143. A discussion of Galilean frame invariance can be found in Sect. 171.
The problem of the so-called material frame-indifference (a synonym for the
P.M.O) is extensively outlined in Sect. 19 of Truesdell and Noll [15] and also put
into historical context. Even today the notion of material objectivity is subject to
fiery controversies and often confused with the concept of objective quantities
under Euclidean transformations. More details can be found in the paper by
Bertram and Svendsen [3], where also many other references on the subject matter
were compiled.
For those interested in the true wording of NEWTONIAN mechanics the various
original editions of the Principia are the prime source of information. In fact it is
worthwhile to compare the first edition of 1687 with the third one, which shows
considerable differences due to numerous comments that NEWTON added in his
later years (see the annotated version by Koyré et al. [9]). Here one can also read
about NEWTON’s ideas on relative motion, absolute space and absolute time, all of
which culminates in a famous gedankenexperiment, which is known in the liter-
ature as NEWTON’s bucket. The philosophy behind all this is explained in the reprint
of MACH’s book [11].
References
4. d’Alembert JL (1967) Traité de dynamique dans lequel les lois de l’équilibre & du
mouvement des corps sont réduites au plus petit nombre possible, & démontrées d’une
manière nouvelle, & ou l’on donne un principe général pour trouver le mouvement de
plusieurs corps qui agissent les uns sur les autres, d’ une manière quelconque. David, Paris
5. Einstein A (1914) Die formale Grundlage der allgemeinen Relativitätstheorie.
Sitzungsberichte der Königlich Preußischen Akademie der Wissenschaften (Berlin),
pp 1030–1085
6. Einstein A (1920) Relativity. The special and the general theory. A popular exposition, 3rd
edn. Methuen & Co. Ltd., London, p. 68
7. Einstein A (1921) A brief outline of the development of the theory of relativity. Nature
106(2677):782–784
8. Greve R (2003) Kontinuumsmechanik—Ein Grundkurs für Ingenieure und Physiker.
Springer, Berlin, New York
9. Koyré A, Cohen IB, Whitman A (1972) Isaac Newton’s Philosophiae Naturalis Principia
Mathematica, 3rd edn (1726), vol I/II with variant readings. Cambridge University Press,
Cambridge
10. Liu I-S (2010) Continuum mechanics. Springer, Berlin, New York
11. Mach E (1976) Die Mechanik—historisch-kritisch dargestellt. Wissenschaftliche
Buchgesellschaft Darmstadt
12. Thirring H (1918) Über die Wirkungen rotierender ferner Massen in der Einsteinschen
Gravitationstheorie. Physikalische Zeitschrift, pp 33–39
13. Thirring H (1921) Berichtigung zu meiner Arbeit: ‘‘Über die Wirkungen rotierender ferner
Massen in der Einsteinschen Gravitationstheorie’’. Physikalische Zeitschrift, pp 29–30
14. Truesdell C (1966) The elements of continuum mechanics. Springer, Berlin, New York
15. Truesdell C, Noll W (1965) The non-linear theories of mechanics. In: Flügge S (ed)
Encyclopedia of physics, volume III/3. Springer, Berlin, Göttingen
16. Truesdell C, Toupin R (1960) The classical field theories. In: Flügge S (ed.) Encyclopedia of
physics, vol III/1, Principles of classical mechanics and field theory. Springer, Berlin,
Göttingen
Chapter 9
Problems of Linear Elasticity
9.1 Introduction
The previous sections were dedicated to the mathematical foundations for treating
problems of continuum physics namely, in particular, to the balances of mass,
momentum, and energy in local and in global form. In combination with suitable
constitutive relations, more specifically HOOKE’s law and the equations for a
NAVIER-STOKES-FOURIER material, we have already solved the resulting field
equations for very simple geometries, such as the one-dimensional tensile bar or
parallel plate flow. The following chapters are dedicated to more advanced
problems. In fact in Chap. 13 we shall even go beyond thermo-mechanics and
present the equations of electromagnetism for continuous matter.
Consider the circular disc shown in Fig. 9.1 (left) rotating at a constant angular
speed x. Its outer radius is denoted by R. Our objective is to determine the internal
stresses resulting from the rotation or, in other words, the stresses that arise in the
material while counterbalancing the centrifugal forces. We assume that the disc is
‘‘very thin’’ so that stresses will only develop within its plane. In other words, most
of the nine components of the stress tensor in cylindrical coordinates will vanish
and only rhrri , rh##i , and rhr#i will remain.
Situations where the stress state is considerably reduced due to geometrical
constraints occur frequently in solid mechanics. For obvious reasons the present
case is known as the state of plane stress.
Consequently, the static balance of momentum in cylindrical coordinates,
Eq. (5.6.1), consists only of two components:
orhrri 1 orhr#i rhrri rh##i
þ þ ¼ qf hri ;
or r o# r ð9:2:1Þ
orhr#i 1 orh##i 2
þ þ rhr#i ¼ qf h#i :
or r o# r
For the specific body forces on the right hand side we insert the centrifugal
acceleration. From high school physics it is known that they are proportional to
the distance r from the center of rotation:
We now assume that the disc shows linear-elastic material behavior. Conse-
quently, the stress–strain relation is given by HOOKE’s law. In particular, HOOKE’s
law in cylindrical coordinates has already been analyzed in Eq. (6.2.26):
ω
σ<ϑϑ >
x2 σ<rr > e2 ω
e´2 e´1
r
ϑ ϑ
e´3 = e 3 e1
R x1
rhrri ¼ k ehrri þ eh##i þ ehzzi þ 2l ehrri ;
rh##i ¼ k ehrri þ eh##i þ ehzzi þ 2l eh##i ;
ð9:2:3Þ
rhzzi ¼ 0 ¼ k ehrri þ eh##i þ ehzzi þ 2l ehzzi ;
rhr#i ¼ 2l ehr#i ; rhrzi ¼ 0 ¼ 2l ehrzi ; rh#zi ¼ 0 ¼ 2l eh#zi :
Due to the state of plane stress some of the components of the stress tensor
could already be identified as zero. In particular we now eliminate the ‘‘annoying’’
3D strain component ehzzi with Eq. (9.2.3)3 from Eq. (9.2.3)1,2. Note that ehzzi is not
necessarily zero. Only stress components related to the cylindrical axis z are zero.
Moreover, the shear strains ehrzi and eh#zi must vanish due to the state of plane
stress. Thus, we obtain:
rhrri ¼ k ehrri þ eh##i þ 2l ehrri ;
rh##i ¼ k ehrri þ eh##i þ 2leh##i ;
ð9:2:4Þ
2l
rhr#i ¼ 2lehr#i ; k ¼ k:
k þ 2l
We finally use certain relations from Eq. (6.2.27) for the remaining components
of the strain tensor in cylindrical coordinates and find:
ffi
ouhri 1ouh#i uhri
rhrri ¼ ðk þ 2lÞ þk þ ;
or r o# r
ffi
ouhri 1 ouh#i
rh##i ¼ k þ ðk þ 2lÞ þ uhri ; ð9:2:5Þ
or r o#
ffi ffi
1 ouhri ouh#i
rhr#i ¼ l uh#i þ :
r o# or
If the relations (9.2.2) and (9.2.5) are inserted into the balance of momentum
(9.2.1) a coupled system of partial differential equations for the displacements uhri
and uh#i results, which is not easy to solve. However, we have not used all of our
knowledge regarding the mathematical form of the displacements yet: Each material
point of the disc rotates at a constant angular velocity at all times. For reasons of
symmetry the radial part of the displacement should therefore not depend on the
angle. For the same reason there must be no displacement in angular direction. Thus,
it seems natural to propose the following ansatz for the displacements:
uhri ¼ f ðr Þ; uh#i ¼ 0: ð9:2:6Þ
f ðr Þ is an unknown function which is yet to be determined from a solution of a
differential equation. We have encountered this strategy before in Exercises 7.3.2
and 7.4.1 as well as in Sect. 7.5. It is known as the semi-inverse method and widely
applied in continuum theory: The structure of the solution to a problem is antic-
ipated by an intelligent guess, which is still general enough so that internal con-
tradictions do not occur. If we now insert Eq. (9.2.6) into Eq. (9.2.5) we obtain:
218 9 Problems of Linear Elasticity
f ðr Þ
rhrri ¼ ðk þ 2lÞf 0 ðr Þ þ k ;
r ð9:2:7Þ
f ðr Þ
rh##i ¼ k f 0 ðr Þ þ ðk þ 2lÞ ; rhr#i ¼ 0
r
and from Eqs. (9.2.1)1 and (9.2.2)1:
fh ðr Þ ¼ Dr a ; fp ðr Þ ¼ Cr sþ1 ð9:2:10Þ
it follows that:
qx2
a1 ¼ 1; a2 ¼ 1; s ¼ 2; C¼ ð9:2:11Þ
8ðk þ 2lÞ
and therefore:
B qx2
uhri f ðr Þ ¼ Ar þ r3 : ð9:2:12Þ
r 8ðk þ 2lÞ
The remaining constants A and B are determined from suitable boundary
conditions. First of all it is intuitively clear that the displacement cannot become
singular in any material point r of the disc. Since the disc is not hollow, the point
r = 0 is part of the system, and therefore:
B ¼ 0: ð9:2:13Þ
Moreover, the traction vector thii must be continuous on the free outer rim of the
disc. This is a consequence of the local balance of momentum for (non-moving)
singular points, cf., Sect. 5.9, Eq. (5.9.5). The traction can be obtained from the
stresses:
thii ¼ rhjii nhji : ð9:2:14Þ
nhji denotes the unit normal vector of the singular surface. This means in our case:
thri ¼ rhrri ; th#i ¼ 0; ð9:2:15Þ
9.2 The Rotating Disc 219
1 1
2μ
2 λ* + 3 μ
R r R r
Fig. 9.2 Radial dependence of the stresses in a disc rotating at a constant angular velocity
Now all of the constants in (9.2.12) are known. In summary the stresses and the
radial displacement of Eqs. (9.2.6/9.2.7) are given by:
ffi r 2 ffi
2k þ l r 2
rhrri ¼ r0 1 ; rh##i ¼ r0 1 ;
R 2k þ 3l R
ffi
1 2k þ l 2 2 qx2 R2 2k þ l r 2
r0 ¼ qx R ; rhr#i ¼ 0; uhri ¼ r :
4 2k þ 3l 8ðk þ 2lÞ 2k þ 3l R
ð9:2:18Þ
Figure 9.2 shows how the stresses develop as functions of the radius r. Note
that in contrast to the radial stress rhrri the hoop stress rh##i does not reduce to zero
at the outer radius R of the disc.
calculate the displacement uhzi including the proper sign by using Eq. (6.2.27)3.
Interpret the sign. Recall that the strain component ehrzi has to vanish due to the
assumption of plane stress. Now calculate ehrzi directly from Eq. (6.2.27)5 and
show that an inconsistency arises. Discuss possible alternative boundary
conditions. For example, investigate the case that the average of shear stresses
vanishes for a disc of height d z þ d:
Finally we will show the validity of the equations for the centrifugal force in
cylindrical coordinates (9.2.2) by using the general results of Sect. 8.5. We start
from Eq. (8.5.9) in Cartesian form and note immediately that in the present sit-
uation of a coordinate system circling centrally on a rotating disc w.r.t. the inertial
frame we have:
b0i ¼ 0; ð9:2:19Þ
since the origin of both systems obviously does not move. Thus, we obtain for the
rotation matrix according to Eq. (8.2.3) by computing the scalar products between
the unit vectors as shown on the right of Fig. 9.1:
2 3
cos # sin # 0
O0ij ðtÞ ¼ 4 sin # cos # 0 5: ð9:2:20Þ
0 0 1
The change of the angle per unit time or, in other words, the angular velocity is
given by:
x ¼ #_ ¼ const. ð9:2:21Þ
9.2 The Rotating Disc 221
Hence by Eq. (8.3.17) we find for the matrix of angular velocities and for its
time derivative:
2 32 3
sin # cos # 0 cos # sin # 0
0 6 76 7
Xij ¼ x4 cos # sin # 0 54 sin # cos # 05
0 0 0 0 0 1
2 3 ð9:2:22Þ
0 1 0
6 7 0
¼ x4 1 0 0 5 ) X_ ij ¼ 0:
0 0 0
Moreover, we must emphasize that a material point on stationary rotating disc
does not move w.r.t. the comoving observer:
t0i ¼ 0: ð9:2:23Þ
Thus Eq. (8.5.9) reduces to:
2 32 32 0 3
0 0 1 0 0 1 0 x1
orji 0 0 0 26 76 76 0 7
¼ q Xil Xlk xk ¼ qx 4 1 0 0 54 1 0 0 54 x2 5
ox0j
0 0 0 0 0 0 x03
2 03 ð9:2:24Þ
x1
6 7
¼ qx2 4 x02 5:
0
Next we transform mutatis mutandis the position vector x0 (within the plane) to
polar coordinates as indicated in Exercise 2.4.5:
Thus the result shown in Eqs. (9.2.1/9.2.2) is confirmed incl. the sign on the
right hand side.
Consider the situation shown in Fig. 9.3: A long, thick-walled, circular, hollow
cylinder of inner radius Ri and outer radius Ro , the pipeline, is subjected to a (high)
internal pressure pi and to a (low) external pressure po . We want to determine the
resulting stresses within the cylinder wall.
Due to its huge length it makes sense to assume that all the relevant fields do
not depend on the axial coordinate z. Moreover, we assume that the displacement
in z direction either vanishes or is given by a constant. We say that the cylinder is
in a state of plane strain.
222 9 Problems of Linear Elasticity
pi Ri
rhrri Ri ¼ pi ; rhrri Ro ¼ po : ð9:3:5Þ
By solving the resulting linear system of equations for A and B, and by inserting
the solution into Eq. (9.3.4) we find:
R2i pi R2o po R2i R2o 1
rhrri ¼ ð p i p o Þ ;
R2o R2i R2o R2i r 2
R2i pi R2o po R 2 R2 1
rh##i ¼ 2 2
þ ð pi po Þ 2 i o 2 2 ; ð9:3:6Þ
Ro R i R o Ri r
k R2i
rhr#i ¼ 0; rhzzi ¼ ðpi po Þ:
k þ l Ro R2i
2
Next we assume that the cylinder is a thin-walled structure with a wall thickness
t, so that we may write:
1 1
Ro ¼ Ri þ t; R2o R2i þ 2Ri t; r ðRo þ Ri Þ ¼ Ri þ t: ð9:3:7Þ
2 2
If these relations are inserted in Eq. (9.3.6) we obtain after expanding w.r.t. the
smallness parameter t=Ri :
1 Ri
rhrri ¼ ðpi þ po Þ; rh##i ¼ ðpi po Þ;
2 t
ð9:3:8Þ
k Ri
rhzzi ¼ ðpi po Þ:
k þ l 2t
Derive Eq. (9.3.8) and discuss the signs. Under which circumstances do
we observe tension or compression?
Eq. (9.3.8) is also known as the elementary pressure vessel stress formulae.
Obviously for the case of a thin-walled cylinder the radial stress is given by the
mean value of the inner and of the outer pressure. This is intuitively clear and it is
also indicated by the negative sign that it is a compressive stress. In order to
interpret the expressions for the hoop stress and for the axial stress, it is helpful to
assume that the outer pressure is smaller than the inner one. Under such circum-
stances both are of tensile nature which, in case of the hoop stress, is also
immediately evident. Moreover, it is remarkable that the hoop stress is more than
twice as large as the axial one.
The consequences of this observation is known to anyone who has already
prepared breakfast sausages in his life or at least watched someone doing it: When
the sausage filling starts swelling and the casing finally bursts, the resulting fissure
runs (mostly) in axial direction. In other words, the casing is torn apart by the
higher hoop stresses.
224 9 Problems of Linear Elasticity
σ <zz>
Consider the cartoon drawings shown in Fig. 9.4. Use them for calcu-
lating the hoop stress rh##i : First, determine the force F required to maintain
the cylindrical shell in equilibrium and show that:
F ¼ 2pdr: ð9:3:9Þ
Second, calculate the surface area on which this force is acting to con-
clude that:
rp
rh##i ¼ ; ð9:3:10Þ
t
where p denotes the inner pressure in the cylinder. Proceed analogously to
find the following expression for the stress rhzzi :
rp
rhzzi ¼ : ð9:3:11Þ
2t
Compare both results with Eq. (9.3.8). Which value of POISSON’S ratio
guarantees identical results for rhzzi and why?
We now return to the problem that has already been addressed in Chap. 1, i.e., the
calculation of the thermal Stresses around a fiber in a composite material. In this
context we refer to Fig. 1.7, which shows an idealized model of a fiber-reinforced
material. Recall that, in general, the fiber (index 1) as well as the matrix (index 2)
possess different elastic and thermal properties, identified by k1 , l1 , a1 , and k2 , l2 ,
a2 , respectively. In what follows we will also use the following expressions for the
so-called compressibility, 3k1 ¼ 3k1 þ 2l1 and 3k2 ¼ 3k2 þ 2l2 , respectively.
Due to the length of the fibers it makes sense to assume that the presented
system is in a state of plane strain. Moreover, in general, each fiber the length of
each fiber will be much greater than its distance to the next fiber, i.e., R2 R1 .
9.4 Thermal Stresses in Fiber Reinforced Composites 225
We can immediately read off the solution for the non-vanishing stresses and
displacements from Eqs. (9.3.4) and (6.2.24/6.2.26):
9
r1hrri ¼ 3k1 a1 ðT TR Þ þ 2ðk1 þ l1 ÞA1 2l1 Br21 >>
=
1 B1
rh##i ¼ 3k1 a1 ðT TR Þ þ 2ðk1 þ l1 ÞA1 þ 2l1 r2 0 r R1 ;
>
1 1 B1 >;
rhzzi ¼ 3k1 a1 ðT TR Þ þ 2k1 A1 ; uhri ¼ A1 r þ r
9 ð9:4:1Þ
r2hrri ¼ 3k2 a2 ðT TR Þ þ 2ðk2 þ l2 ÞA2 2l2 Br22 >>
=
r2h##i ¼ 3k2 a2 ðT TR Þ þ 2ðk2 þ l2 ÞA2 þ 2l2 Br22 ; R1 r R2 ;
>
>
r2 ¼ 3k a ðT T Þ þ 2k A ; u2 ¼ A r þ B2 ;
hzzi 2 2 R 2 2 hri 2 r
where it has been observed that, first, the thermal stresses need to be taken into
account, which becomes possible by means of the generalized HOOKE’s law from
Eq. (6.2.24). Second, one needs to distinguish two circular regions with different
material parameters. This increases the number of relevant equations and makes
each of them longer and more cumbersome. For this reason we refrain from
presenting the corresponding relations for the shear stresses (which have to vanish
anyway) and for the axial stresses.
Obviously it is necessary to determine four constants of integration, namely A1,
A2, B1, and B2. For this purpose we need four conditions. Following the remarks of
the previous sections these are given by:
• a regularity requirement:
r1hrri \1; ð9:4:2Þ
r¼0
• and, finally, continuity of displacement, i.e., a perfect fit of both cylinders along
the inner interface:
u1hri ¼ u2hri : ð9:4:4Þ
r¼R1 r¼R1
The last relation can immediately be related to the four constants A1, A2, B1, and
B2 by using Eq. (9.3.3).
This solves the problem, at least in principle. However, in practice the calcu-
lation of the stresses in their most general form leads to rather unwieldy expres-
sions, which most likely will only be of interest to composite specialists. Thus, we
shall study two special cases in the following exercises.
226 9 Problems of Linear Elasticity
Discuss the sign of the stresses. What results in case of a hole and what for an
incompressible fiber? Use the internet and find materials data for typical fiber
reinforced composites. Calculate and plot the stresses that arise after the
composite is cooled down from fabrication to room temperature.
atmosphere. A few missing tiles can be detrimental as we had to learn the tragic
way from the Columbia catastrophe.
As indicated ceramics are also extremely resistant against corrosion, and this is
most important for the chemical industry. Moreover, they have excellent tribo-
logical properties (also at high temperatures). In other words, they are extremely
resistant against wear while showing highest durability. This recommends
ceramics for special applications, e.g., as heavy-duty bearings.
However, ceramics also have a well known disadvantage: They are very brittle,
i.e., they break easily, and react quite sensitively to impact and tensile stresses
because, unlike metals, they do not yield by plastic flow. In order to quantify this
statement we compare the fracture toughness of various technically important
materials: Table 9.1.
The resistance of a material against (brittle) fracture is characterized by the so-
called fracture toughness or KIc -value for short. Without going into the details we
may say that the higher the toughness the more resistant against fracture the
material will be. Indeed, high strength steel has a much higher KIc -value than
brittle cast iron (say). Note that the KIc of a typical ceramic, for example Alumina
(or Al2O3 in chemical terms), is still much lower than that cast iron, namely almost
by an order of magnitude. This makes sense since ceramics are typically very
brittle and quite susceptible to fracture.
However, if a certain amount of Zirconia particles (ZrO2) of micrometer size
are added to Alumina powder we observe that the fracture toughness of the sin-
tered compound has grown considerably, even beyond the KIc value of cast iron.
With a grain of salt we may claim that we have invented the ‘‘ceramic steel.’’
Naturally the question arises why an addition of Zirconia leads to such a dra-
matic increase of the fracture toughness. The reason for the effect is hidden in the
pronounced polymorphic nature of ZrO2. The crystal structure of polymorphic
solids depends on temperature and on the applied pressure. At atmospheric pres-
sure and above 1,480 K (ca. 1,200 C) Zirconia exists in tetragonal form. Below
that it has a monoclinic crystal lattice. Moreover, the volumes of the unit cell of
these two crystal structures are very different. The monoclinic phase is approxi-
mately 3 % more voluminous than the tetragonal1 one. Thus, if a piece of Zirconia
1
Strictly speaking, during the tetragonal to monoclinic phase transition the volume of Zirconia
increases and we observe a shear as well: The right-angled, tetragonal unit cell transforms into a
slightly inclined monoclinic configuration. In other words the angle b of Fig. 9.5 is not quite 90.
Consequently shear strains will arise and a more complex state of stress will result. However, in
our calculations we will ignore this effect.
228 9 Problems of Linear Elasticity
a a
β a
b
tetragonal monoclinic
1100
p(T)
900
700
monoclinic
500
0 10 20 30 40 50
pressure / kbar
9.5 Transformation Toughened Ceramics 229
σ tetragonal
Al2O3 ZrO2
tetragonal ZrO2
Al2O3
monoclinic ZrO2
σ
This is exactly what should happen if ZrO2 particles are sintered into a Al2O3
matrix. After cooling below 1,480 K the Zirconia particles are basically ready to
switch from their tetragonal phase into the monoclinic one. However, this is not
automatically possible due to the increase in volume, which makes it necessary to
push the matrix aside. This, however, creates a counter pressure which is high
enough to stabilize the ZrO2 particles in their tetragonal phase way down below
room temperature: Fig. 9.7 (left).
Now if a crack enters such a metastable system under the influence of an
external load, cf., Fig. 9.7 (right), it reduces in his vicinity the stabilizing effect of
the matrix on the Zirconia particles. In the neighborhood of the crack the particles
will then spontaneously transform from their tetragonal into the monoclinic phase.
During the process they will increase their volume, push the matrix aside, and
generate a compressive zone around the crack. In order to move through that zone
an extra amount of energy is required, which in macroscopical terms results in an
increased fracture toughness of Zirconia containing ceramics.
In order to guarantee an optimal increase of fracture toughness, it is obviously
necessary that the Zirconia particles are in a metastable, tetragonal state and do not
transform before the crack is running through the material. It is therefore of
interest to predict at which temperature a Zirconia particle of given size will
transform in an undamaged matrix of given stiffness. In what follows we will try to
provide an estimate for that temperature.
To this end we idealize the situation and consider a spherical Zirconia particle,
denoted by ‘‘1,’’ of radius R1 in an infinitely large hollow sphere of matrix, denoted
by ‘‘2,’’ cf., Fig. 1.7. In order to calculate the stresses in that system we make use
of the balance of the static balance of momentum in its form (5.5.2) to (5.6.4) in
combination with HOOKE’s law from Eq. (6.2.28), and the strain tensor from Eq.
(6.2.29). In other words, all equations are written in spherical coordinates.
Moreover, for reasons of symmetry it is reasonable to assume that:
uhri ¼ f ðrÞ; uhui ¼ 0; uh#i ¼ 0: ð9:5:1Þ
If this semi-inverse ansatz is inserted we obtain for the strain tensor:
230 9 Problems of Linear Elasticity
f ðrÞ f ðrÞ
ehrri ¼ f 0 ðrÞ; eh##i ¼ ; ehuui ¼ ;
r r ð9:5:2Þ
ehrui ¼ 0; ehr#i ¼ 0; ehu#i ¼ 0;
and for the stress tensor:
f ðrÞ
rhrri ¼ ðk þ 2lÞf 0 ðrÞ þ 2k ;
r
f ðrÞ ð9:5:3Þ
rh##i ¼ rhuui ¼ 2ðk þ lÞ þ kf 0 ðrÞ;
r
rhrui ¼ 0; rhr#i ¼ 0; rhu#i ¼ 0;
and from the radial component of the balance of momentum (the other two
components are identically satisfied):
f 0 ðrÞ f ðrÞ
f 00 ðrÞ þ 2 2 2 ¼ 0: ð9:5:4Þ
r r
This ordinary differential equation is solved by using the power function
(9.2.10)1. Thus we obtain for the displacement in radial direction:
B
uhri ¼ Ar þ ð9:5:5Þ
r2
and, consequently, for the stresses different from zero:
B B
rhrri ¼ ð3k þ 2lÞA 4l ; rh##i ¼ rhuui ¼ ð3k þ 2lÞA þ 2l : ð9:5:6Þ
r3 r3
As before A and B denote two constants of integration.
Explain in detail the various steps required in order to obtain Eqs. (9.5.5/
9.5.6) for the displacements and stresses for the case of perfect spherical
symmetry. Moreover, recall the definition of the displacement:
u i ¼ zi Z i ð9:5:7Þ
ðzÞ
with the current and with the reference positions zi and Z i , respectively.
Evaluate this relation for the case of spherical coordinates and derive an
expression for the current radial distance r as a function of the radial distance
R in the reference configuration by using Eq. (9.5.5). In the same context
discuss the question if it is necessary to distinguish between r and R in Eqs.
(9.5.5/9.5.6) for small displacements and strains.
9.5 Transformation Toughened Ceramics 231
These equations need to be written down for a solid sphere made of (mono-
clinic) Zirconia, denoted by an index 1 of radius Rm in the reference configuration,
and for a hollow sphere made of matrix material, denoted by the index 2, with an
inner radius suitable for a tetragonal solid sphere of Zirconia of radius Rt in the
reference configuration, respectively. In order to obtain concise formulae that still
allow for a modeling of the effect, we will assume that the outer radius of the
matrix sphere is infinitely large. This leads to the following equations containing
four unknown constants of integration A1 , A2 , B1 , and B2 (also see the analogous
cylindrical problem depicted in Fig. 1.7):
ffi
B1 B1
r1 ¼ 1 þ A1 þ 3 R; r1hrri ¼ ð3k1 þ 2l1 ÞA1 4l1 3 ;
R R
ð9:5:8Þ
B1
r1h##i ¼ r1huui ¼ ð3k1 þ 2l1 ÞA1 þ 2l1 3 ; 0 R Rm
R
and:
ffi
B2 B2
r2 ¼ 1 þ A2 þ 3 R; r2hrri ¼ ð3k2 þ 2l2 ÞA2 4l2 ;
R R3
ð9:5:9Þ
B2
r2h##i ¼ r2huui ¼ ð3k2 þ 2l2 ÞA2 þ 2l2 ; Rt R R2 ! 1:
R3
In order to determine the four constants four boundary conditions are required.
These include requirements for
• regularity in the origin at r ¼ 0, for example:
r1hrri \1; ð9:5:10Þ
r¼0
• continuity of the traction vector at the transition between Zirconia sphere and
the matrix:
r1hrri ¼ r2hrri ; ð9:5:11Þ
r¼Rm r¼Rt
• a perfect fit; in other words, the cavity must be extended and the solid sphere of
monoclinic Zirconia must be compressed until the volume expansion due to the
phase transition of roughly f ¼ 3 % has been compensated:
r1 jr¼Rm ¼ r2 jr¼Rt ; ð9:5:12Þ
If these four equations are evaluated with Eqs. (9.5.8/9.5.9) and if, in addition,
the balance of mass is observed:
232 9 Problems of Linear Elasticity
4p 4p
q R3 ¼ q R3 ; ð9:5:14Þ
3 t t 3 m m
with the mass densities qt and qm of the tetragonal and of the monoclinic Zirconia
in the reference configuration, respectively, we obtain with the smallness
parameter:
qt qm
f¼ [0 ð9:5:15Þ
3qm
immediately:
B2 3k1 þ 2l1
A2 ¼ B1 ¼ 0; ¼ f;
R3t 3k1 þ 2l1 þ 4l2
ð9:5:16Þ
4l2
A1 ¼ f:
3k1 þ 2l1 þ 4l2
Thus the stresses of interest are given by:
4l2 ð3k1 þ 2l1 Þ
r1hrri ¼ r1h##i ¼ r1huui ¼ f;
3k1 þ 2l1 þ 4l2
4l ð3k1 þ 2l1 Þ R3t
r2hrri ¼ 2 f; ð9:5:17Þ
3k1 þ 2l1 þ 4l2 R3
2l2 ð3k1 þ 2l1 Þ R3t
r2h##i ¼ r2huui ¼ f:
3k1 þ 2l1 þ 4l2 R3
Note that the sphere of Zirconia is subjected to an isotropic homogeneous state
of stress, i.e., under a ‘‘pressure’’ that can be linked to the phase diagram shown in
Fig. 9.7. In order to compute the transition temperature T of the Zirconia
embedded in the matrix we put:
1
r1hrri þ r1h##i þ r1huui ¼ pðT Þ; ð9:5:18Þ
3
where pðT Þ is the straight line shown in Fig. 9.7 that divides the region of
tetragonal phase from the one of monoclinic stability. From measurements it is
known that:
102 GPa
pð T Þ ¼ T þ 4:89 GPa;
3:02 K ð9:5:19Þ
k1 ¼ 31 GPa; l1 ¼ 66 GPa; k2 ¼ 102 GPa; l2 ¼ 110 GPa:
Hence we find:
T ¼ 128 K: ð9:5:20Þ
9.5 Transformation Toughened Ceramics 233
Confirm in detail the solution for the stresses shown in Eq. (9.5.17) and
calculate the constants of integration by means of the boundary and jump
conditions (9.5.10–9.5.13). In particular study the arguments that led to the
continuity condition (9.5.12).
Also verify the transition temperature of Eq. (9.5.20) and explain why we
may use the experimental result of Fig. 9.7 for the transformed sphere of
Zirconia in our model. Would this also work with a Zirconia inclusion of
arbitrary shape?
And exactly this has been observed: It has been shown experimentally that
depending on their size Zirconia particles embedded in Alumina need to be cooled
down up to temperatures of liquid nitrogen. It turns out that the smaller the
particle, the lower the transformation temperature, i.e., the more difficult it
becomes to accomplish the phase transition. Nevertheless, we may argue that our
estimate concerns notably small particles since the sphere of Zirconia was
embedded in an infinitely large matrix. However, the presented method can be
extended to particles embedded in a matrix of finite size and the afore-mentioned
size effect can be predicted from the resulting equations.
where e0ij stands for all non-elastic strain contributions. The latter are, for
example, thermal strains, for which we may write in the isotropic case:
examples of e0ij . Use this relation and state the corresponding HOOKE’s law.
Show that it leads to the same results as given by Eq. (9.5.17) by evaluating
the condition
u1hri ¼ u2hri ; ð9:5:24Þ
ri ri
and by using the inner radius ri of the hollow matrix as a measure of distance.
Discuss the pros and cons of this method.
In this section we will address the problem of the broken sphere in a bushing
shown in Figs. 1.1 and 1.2. In contrast to the transformed Zirconia sphere this
problem does not show perfect spherical symmetry any more. However, there is
some symmetry left, namely w.r.t. the polar axis of the sphere. Hence it makes
sense to look for a solution for the stresses, strains, and for the displacements by
starting from the static balance of momentum, Hooke’s law, and kinematic con-
ditions in spherical coordinates as presented in Exercises 5.4.2 and 6.2.5. As we
shall see the semi-inverse method can still be used, although the ansatz will be
more complex and basically consist of products of unknown functions of the radius
r and of the polar angle #. The azimuthal angle u will not occur because of the
symmetry with respect to the polar axis. During the solution we will discover a
special class of mathematical functions, which are very useful when tackling
problems of spherical symmetry, the so-called LEGENDRE polynomials. In fact, this
whole section is inspired by the paper of Hiramatsu and Oka [3] who, unlike us, in
the end specialized the solution the case of a sphere symmetrically compressed at
the poles.
In this spirit we assume static conditions, neglect body forces, and allow no
dependence on u so that the balance of momentum of Eq. (5.4.7) reduces to:
orhrri 1 orhr#ri 1
þ þ 2rhrri rh##i rhuui þ rhr#i cot # ¼ 0;
or r o# r
orhr#i 1 orh##i 1
þ þ 3rhr#i þ rh##i rhuui cot # ¼ 0; ð9:6:1Þ
or r o# r
orhrui 1 orh#ui 1
þ þ 3rhrui þ 2rh#ui cot # ¼ 0:
or r o# r
We now turn to HOOKE’s law including the strains and the displacements. After
simplifying the kinematic relations of Eq. (6.2.29) by omitting derivatives w.r.t. u
and inserting the strains into Eq. (6.2.28) we find that:
9.6 Compression of a Sphere 235
ffi
ouhri ouh#i 1
rhrri ¼ kD þ 2l ; rh##i ¼ kD þ 2l þ uhri ;
or o# r
ffi
2l ouh#i 1 ouhri
rhuui ¼ kD þ uhri þ uh#i cot # ; rhr#i ¼ l uh#i ;
r or r o#
ffi ffi
l ouhui ouhui 1
rh#ui ¼ uhui cot # ; rhrui ¼ l u h ui ;
r o# or r
ð9:6:2Þ
where the following abbreviation has been used:
1 o 2 o
D¼ 2 r uhri sin # þ ruh#i sin # : ð9:6:3Þ
r sin # or o#
We now insert Eq. (9.6.2) into Eq. (9.6.1)1,2 and a system of coupled partial
differential equations for uhri and uh#i results:
oD 2l oX 2l
ðk þ 2lÞ X cot # ¼ 0;
or r o# r ð9:6:4Þ
1 oD oX X
ðk þ 2lÞ þ 2l þ 2l ¼ 0;
r oh or r
where a second abbreviation has been used:
ouh#i uh#i 1 ouhri
2X ¼ þ : ð9:6:5Þ
or r r o#
Equation (9.6.1)3 will later serve us to determine uhui and is ignored for the time
being. By cross-differentiation w.r.t. r and # and mutual addition and subtraction,
Eq. (9.6.4) can be decoupled and the following second order partial differential
equations for D and X result:
o2 D 2 oD 1 o2 D cot # oD
þ þ þ 2 ¼ 0;
or 2 r or r 2 o#2 r o#
ð9:6:6Þ
o2 X 2 oX 1 o2 X cot # oX 1
þ þ þ 2 X ¼ 0:
or 2 r or r 2 o#2 r o# ðr sin #Þ2
00
fn00 f0 g g0
r2 þ 2r n ¼ n þ cot # n ;
fn fn gn gn
00 0
00
ð9:6:8Þ
F
2 n F n G n G0n 1
r þ 2r ¼ þ cot # :
Fn Fn Gn Gn sin2 #
The dashes refer to differentiation w.r.t. the corresponding argument. Since
r and # are independent of each other the left and right hand side of Eq. (9.6.8)
must be constant. For reasons that will become evident shortly we choose this
constant as nðn þ 1Þ where n ¼ 0; 1; 2; . . .. We attempt a solution by power
functions of the form fn ðrÞ r a and Fn ðrÞ r b for
the differential
equations on the
left hand side and find that a1;2 ¼ b1;2 ¼ 12 n þ 12 . Hence:
Bn bn
fn ðrÞ ¼ An r n ; Fn ðrÞ ¼ an r n : ð9:6:9Þ
r nþ1 r nþ1
Recall that infinitely many choices of n are possible. Thus, we have found
infinitely many solutions of power functions in r, positive as well as negative ones.
If n had not been chosen as integer the r-dependence of the displacements would
not result in power functions and thus contradict the idea of a power series. In fact
an infinite sum of functions with positive and negative powers is known as a
LAURENT series. With a grain of salt it can be considered as an extension of the well
known TAYLOR series including also contributions that become singular for r ¼ 0.
As we shall see such power functions in r are perfectly sufficient in order to solve
the axially symmetric boundary value problem. It will not lead to contradictions.
This is in agreement with our afore-mentioned principle that the semi-inverse
ansatz should always be kept as simple as possible. Finally note that the minus in
front of the second terms in (9.6.9) is arbitrary. We are just following the con-
vention established in Hiramatsu and Oka [3].
We now turn to the right hand sides of Eq. (9.6.8) and obtain the following
ordinary differential equations of second order:
d 2 Pn dPn
þ cot # þ nðn þ 1ÞPn ¼ 0: ð9:6:11Þ
d#2 d#
We will investigate the specific form of these polynomials in Exercise 9.6.1. If
we differentiate (9.6.10)1 by # it is easy to show that the second differential
equation is solved by dPn =d#. Thus, the general solutions for D and X read:
1 ffi
X 1 ffi
X
n Bn n bn dPn
D¼ An r nþ1 Pn ; X¼ an r : ð9:6:12Þ
n¼0
r n¼0
r nþ1 d#
However, D and X are not independent but linked to each other via Eq. (9.6.4).
Thus, if we insert (9.6.12) into (9.6.4)1 (say) and observe Eq. (9.6.11), we finally
find that:
1 ffi
k þ 2l X An n Bn 1 dPn
2X ¼ r þ : ð9:6:13Þ
l n¼0 n þ 1 n r nþ1 d#
The term n ¼ 0 in this equation does not really present a singularity problem
since we shall see in Exercise 9.6.1 that dP0 =d# ¼ 0. The solutions shown in Eqs.
(9.6.12)1 and (9.6.13) will now help us to find the general solution for the dis-
placements uhri and uh#i . We use the definitions (9.6.3) and (9.6.5) to find
uncoupled differential equations for both displacements. First for uh#i :
ffi
o2 uh#i 1 o2 uh#i 4 ouh#i cot # ouh#i 1 1
þ þ þ þ 2 uh#i
or 2 r 2 o#2 r or r 2 o# r2 sin2 # ð9:6:14Þ
oð2XÞ 1 oD 3
¼ þ þ ð2XÞ:
or r o# r
As before we look for solutions of the product type:
Dn
fnhom ðrÞ ¼ Cn r n1 : ð9:6:16Þ
r nþ2
238 9 Problems of Linear Elasticity
The radial part is solved by fn ðrÞ r a ) a1;2 ¼ 12 ðn þ 12Þ and the angular
one is satisfied by putting gn ¼ dPn =d#. Thus, we conclude that:
X1 ffi
n Fn dPn
u h ui ¼ En r þ nþ1 : ð9:6:23Þ
n¼1
r d#
It is interesting to note that the solutions for perfect radial symmetry from Eqs.
(9.5.1) and (9.5.5) can be obtained from the terms for n ¼ 0 in Eqs. (9.6.17/9.6.19/
9.6.23) after neglecting the rigid body part. It is now only a matter of algebra to
obtain the stresses by inserting the formulae for the displacements into Eq. (9.6.2):
X1
ðn2 n 3Þk þ ðn þ 1Þðn 2Þl ðn2 þ 3n 1Þk þ nðn þ 3Þl Bn
rhrri ¼ An r n
n¼0
2n þ 3 2n 1 r nþ1
Dn
þ nðn 1Þ2lCn rn2 ðn þ 1Þðn þ 2Þ2l nþ3 Pn ;
r
X1
ðn þ 3Þk ðn 2Þl ðn 2Þk ðn þ 3Þl Bn
rh##i ¼ An r n
n¼0
2n þ 3 2n1 r nþ1
Dn
þ 2lnCn rn2 þ 2lðn þ 1Þ nþ3 Pn
r
X 1
ðn þ 3Þk þ ðn þ 5Þl ðn 2Þk þ ðn 4Þl Bn
þ An r n
n¼1
ðn þ 1Þð2n þ 3Þ nð2n 1Þ r nþ1
2
Dn d Pn
þ 2lCn r n2 2l nþ3 ;
r d#2
X1
ðn þ 3Þk ðn 2Þl ðn 2Þk ðn þ 3Þl Bn
rhuui ¼ An r n
n¼0
2n þ 3 2n 1 r nþ1
Dn
þ 2lnCn rn2 þ 2lðn þ 1Þ nþ3 Pn
r
X 1
ðn þ 3Þk þ ðn þ 5Þl ðn 2Þk þ ðn 4Þl Bn
þ An r n
n¼1
ðn þ 1Þð2n þ 3Þ nð2n 1Þ r nþ1
Dn dPn
þ 2lCn r n2 2l nþ3 cot #;
r d#
X1
nðn þ 2Þk þ ðn2 þ 2n 1Þl ðn2 1Þk þ ðn2 2Þl Bn
rhr#i ¼ An r n þ
n¼1
ðn þ 1Þð2n þ 3Þ nð2n 1Þ r nþ1
Dn dPn
þ 2lðn 1ÞCn r n2 þ 2lðn þ 2Þ nþ3 ;
r d#
X1 ffi ffi
Fn d2 Pn dPn
rh#ui ¼l En r n1 þ nþ2 cot # ;
n¼1
r d#2 d#
X1 ffi
Fn dPn
rhrui ¼l ðn 1ÞEn r n1 ðn þ 2Þ nþ2 :
n¼1
r d#
ð9:6:24Þ
240 9 Problems of Linear Elasticity
Use a suitable mathematical textbook, for example, Butkov [4], and show
that the LEGENDRE differential equation (9.6.11) can be solved by polynomials
following RODRIGUES’ generating formula:
1 dn n
Pn ð xÞ ¼ x2 1 ; x ¼ cos #: ð9:6:25Þ
2n n! dxn
Use this formula to determine and plot the first five LEGENDRE polynomials:
ffi ffi
3 2 1 5 2 3
P0 ¼ 1; P1 ¼ x; P1 ¼ x ; P3 ¼ x x ;
2 3 2 5
ffi ffi ð9:6:26Þ
35 4 6 2 3 63 4 10 2 5
P4 ¼ x x þ ; P5 ¼ x x x þ :
8 7 35 8 9 21
Moreover, show that the LEGENDRE polynomials are orthogonal to each other:
Zþ1
2
Pm ðxÞPn ðxÞ dx ¼ dmn ð9:6:27Þ
2m þ 1
1
We now need to take care of the six times infinitely many constants of inte-
gration An, Bn, Cn, Dn, En, and Fn in the general solutions for the stresses and for
the displacements. As usual this can be done by adjusting them to the requirements
of our problem or, in other words, to the boundary conditions. First, all of the
stresses and displacements should not be singular at any point. This concerns
notably the origin r ¼ 0 which is part of a solid sphere. Therefore, we require that:
Bn ¼ 0; Dn ¼ 0; Fn ¼ 0; n ¼ 0; 1; 2; . . .: ð9:6:30Þ
Second, recall that the traction vector must be continuous at the outer radius, R,
of the sphere. Since no loads are applied in angular direction we have:
rhr#i r¼R ¼ 0; rhrui r¼R ¼ 0: ð9:6:31Þ
rhrri X1
¼1þ ð2n þ 1Þ ð4n2 2n 3Þa þ ð2n2 n 1Þb q2
p n¼1
q2ðn1Þ a2n
þ n 8nðn þ 1Þa þ ð4n2 þ 4n 1Þb P2n ;
ð8n2 þ 8n þ 3Þa þ ð4n2 þ 2n þ 1Þb
rh##i X1
r m m
p ¼ p0 sin #0 ; ; a¼
q¼ ; b¼ ;
R ð1 þ mÞð1 2mÞ 1þm
ð9:6:38Þ
P2n1 ðsin #0 Þ P2nþ1 ðsin #0 Þ
a2n ¼ :
sin #0
We now consider the special case of a sphere subjected to a (constant) line force
q0 along its equator, i.e., the case #0 ! 0 so that pR sin #0 pRd# ! q0 . L’HO-
PITAL’s rule is used to calculate the following relevant limit for a numerical
evaluation of the stress formulae shown in Eq. (9.6.37):
P2n1 ðxÞ P2nþ1 ðxÞ
lim an ¼ lim ¼ ð4n þ 1ÞP2n ð0Þ
x!0 x!0 x
ð9:6:39Þ
ð1Þnþ1 ð2nÞ!ð4n þ 1Þ
¼ :
22n ðn!Þ2
In the same context it is helpful to know the following relations for LEGENDRE
polynomials:
n
P00n ðxÞ ¼ ðn 2Þx2 Pn ðxÞ ð2n 3ÞxPn1 ðxÞ þ ðn þ 1ÞPn2 ðxÞ
ð1 x 2 Þ2
n
½Pn1 ðxÞ xPn ðxÞ
;
1 x2
1 d2 P2n 1 00
¼ P2n ðxÞ ð1 x2 Þ xP02n ðxÞ ;
2 d#2 2
1 dP2n 1 n
cot # ¼ xP02n ðxÞ; P0n ðxÞ ¼ ½Pn1 ðxÞ xPn ðxÞ
; x ¼ cos #:
2 d# 2 1 x2
ð9:6:40Þ
Figure 9.8 shows all the stresses as a function of (normalized) radius. POISSON’s
number for glass was used, i.e., m ¼ 0:256, and #0 ¼ 0. As expected, compressive
stresses are dominant. However, there are also zones of tensile stress. Consider for
example the angular stress r## in the equatorial plane, i.e., at # ¼ 90 . The cor-
responding values are all positive and almost constant w.r.t. r. However, a thorough
examination shows that for the limit r ! R the tensile stress turns rapidly into a
compressive one. Nevertheless this explains the tensile fracture shown in Fig. 1.2.
Prove Eqs. (9.6.39/9.6.40). Use the results to program the formulae for the
stresses shown in Eq. (9.6.37). Discuss the transition from a pressure load in
the equatorial zone to a line force along the equator.
9.7 The GRIFFITH Crack Model 245
The first two-dimensional model of a sharp crack in a brittle solid was proposed
and mathematically analyzed during the 20 s of the last century by A. A. GRIFFITH.
As indicated in Fig. 9.8 GRIFFITH considers a sequence of ellipses with
decreasing minor axis. It was mentioned in Sect. 2.3 that these ellipses can
mathematically be described by elliptic coordinates, ðz1 ; z2 Þ. Recall that they are
related to Cartesian coordinates ðx1 ; x2 Þ as follows:
x1 ¼ c cosh z1 cos z2 ; x2 ¼ c sinh z1 sin z2 ;
ð9:7:1Þ
z1 2 ½0; 1Þ; z2 2 ½0; 2pÞ:
and we realize that the major and the minor axes, a and b, are given by:
a ¼ c cosh z1 ; b ¼ c sinh z1 : ð9:7:3Þ
Obviously the ellipse degenerates to a slit of infinite sharpness in the limit
z1 ! 0, which is known as the GRIFFITH crack. The crack length is given by the
parameter 2c. We are interested in the stresses developing in the immediate
neighborhood in front of the crack tip at x1 ¼ c þ r if a biaxial tensile stress r is
applied at infinity in x1 as well as in x2 direction (cf., Fig. 9.9). Thus we have:
r
c þ r ¼ c cosh Dz1 ;
1: ð9:7:4Þ
c
If the hyperbolic function is expanded in a TAYLOR series it follows that:
rffiffiffiffiffi
1 r
Dz ¼ 2 : ð9:7:5Þ
c
z1 = 1
x1 = c
z =π
2 z1 = 0 z2 = 0
x1
z2 = 2π
z2 = 3π/ 2
246 9 Problems of Linear Elasticity
σ σ
x2 x2
x1 x1
-c +c
σ
σ
In order to calculate the stresses in the vicinity of the GRIFFITH crack we will
first determine the stresses around an elliptical hole in a most general manner. The
periphery of the hole is characterized by z1 ¼ a ¼ constant. Then, in a second step,
we will study the limit case z1 ! 0, i.e., the crack problem, cf., Fig. 9.10.
The stresses around an elliptic hole can be characterized best by working with
elliptic coordinates from the very beginning on. As a first step in that direction we
specialize the general form of the static balance of momentum from Eq. (5.5.1) to
such coordinates. By taking the definition of the covariant derivative for tensors in
account, Eqs. (4.4.3/4.4.5), it follows in two dimensions:
or11 or12
1
þ 2 þ r11 2C111 þ C212 þ r12 3C112 þ C222 þ r22 C122 ¼ 0;
oz oz
ð9:7:6Þ
or12 or22
1
þ 2 þ r11 C211 þ r12 3C212 þ C111 þ r22 2C222 þ C112 ¼ 0:
oz oz
From Eq. (9.7.1) we find the coefficients of the metric tensor by observing Eqs.
(2.2.8) and (2.4.7) (also see Exercise 4.2.4):
ffi
c2 coshð2z1 Þ cosð2z2 Þ 0
gij ¼ ;
2 0 coshð2z1 Þ cosð2z2 Þ
1
! ð9:7:7Þ
2 coshð2z1 Þcosð2z2 Þ 0
kl
g ¼ 2 1
:
c 0 coshð2z1 Þcosð2z2 Þ
Recall that the elliptic coordinates ðz1 ; z2 Þ are orthogonal so that only the
components along the diagonal of the metric tensor are different from zero.
Moreover, it is curious to note that the diagonal metric coefficients are identical.
Eq. (4.2.4) allows now to calculate the corresponding CHRISTOFFEL symbols (also
see Exercise 4.2.4):
9.7 The GRIFFITH Crack Model 247
sinhð2z1 Þ
C111 ¼ C122 ¼ C212 ¼ ;
coshð2z1 Þ cosð2z2 Þ
ð9:7:8Þ
sinð2z2 Þ
C112 ¼ C211 ¼ C222 ¼ :
coshð2z1 Þ cosð2z2 Þ
By means of these relations and the following auxiliary formulae easily con-
firmed by differentiation:
1 og11 1 og11
2C111 ¼ ; 2C112 ¼ : ð9:7:9Þ
g11 oz1 g11 oz2
and by observing the definition of physical tensor components in Eq. (2.6.5), the
two-dimensional balance of momentum in contravariant form of Eq. (9.7.6) can
alternatively be written as:
orh11i orh12i
þ þ C111 rh11i þ 2C112 rh12i þ C122 rh22i ¼ 0;
oz1 oz2
ð9:7:10Þ
orh12i orh22i
þ þ C211 rh11i þ 2C111 rh12i þ C222 rh22i ¼ 0:
oz1 oz2
It is easily verified by inserting that the following expressions for the stresses
into satisfy the momentum balance identically:
¼ ;
r ½cosh ð2z1 Þ cos ð2z2 Þ
2
rh12i sin ð2z2 Þ½coshð2z1 Þ coshð2aÞ
¼ ; ð9:7:11Þ
r ½cosh ð2z1 Þ cos ð2z2 Þ
2
rh22i sinhð2z1 Þ½coshð2z1 Þ þ coshð2aÞ 2 cosð2z2 Þ
¼ :
r ½cosh ð2z1 Þ cos ð2z2 Þ
2
It is straightforward to proof that these relations fulfill the requirements for a
hole that is free of tractions. In other words the traction vector vanishes at positions
z1 ¼ a ¼ constant:
rh11i rh12i
¼ 0; ¼ 0: ð9:7:12Þ
r r
248 9 Problems of Linear Elasticity
Confirm the last statement by starting from Eq. (9.7.11). Moreover, show
that the components rh11i and rh12i are the relevant components of the traction
vector at the periphery of the hole. Write down and evaluate the jump
conditions for the flux of momentum.
We will now investigate the behavior of the stress component rh22i in front of a
GRIFFITH crack by using the equations shown in (9.7.11). The stress r is tensile, i.e.,
positive. Thus we conclude from Eq. (9.7.11)3 that the stress rh22i leads to an
opening of the plate material along the x1-axis. In other words, the flanks of the
GRIFFITH crack are driven apart.
By means of a TAYLOR expansion and by observing Eqs. (9.7.11)3 and (9.7.5)
we conclude that:
rh22i 1
lim z1 ¼Dz1 1 ; ð9:7:14Þ
a!0 r Dz
z2 ¼0
and thus:
KI pffiffiffiffiffiffi
rh22i pffiffiffiffiffiffiffiffi ; KI ¼ r p c: ð9:7:15Þ
2p r
The quantity KI is also known as stress intensity factor of a GRIFFITH crack in an
infinite plane subjected to biaxial tension. As its name indicates, KI characterizes the
pffiffi
intensity of the 1= r singularity characteristic of the stresses in front of the crack tip.
Obviously the opening mode stress becomes infinitely large at the crack tip. Thus, we
should suspect that a GRIFFITH crack cannot be stable. However, this is an artifact of
the linear elastic analysis. In fact, it is not the stresses that decide about stability or
instability of a crack. Rather it is the release rate of elastic energy in competition with
the energy required to form new surfaces that controls crack instability and propa-
gation. It can be shown that the elastic energy release rate is proportional to the square
of the stress intensity factor. Catastrophic fracture occurs if a critical value of the
stress intensity factor is reached, the so-called fracture toughness KIc , which we have
already introduced phenomenologically in Sect. 9.5.
9.7 The GRIFFITH Crack Model 249
References
Consider in analogy to Fig. 7.4 the plane-parallel shear flow between two infinitely
large plates as shown in Fig. 10.1. However, this time the upper plate at position
y ¼ H is fixed and the lower one at y ¼ 0 moves at a time-dependent speed VðtÞ,
which is a big difference when compared to the stationary case of Exercise 7.4.1.
The switch of plates has computational reasons only. Note that the time-dependent
speed adds a new quality to the problem: We are no longer exclusively interested
For the solutions to the problems presented in this chapter I would like to give credit to Prof.
Dr. W. Dreyer and Dr. W. Weiss of the Weierstrass Institut in Berlin and also Prof. emer.
Dr. rer. nat. Dr. h.c. I. Müller, formerly with Technische Universität Berlin.
(
υi = υx ,υ y ,υz )
x
V ()
z t
kg Ns
q ¼ 103 3
; l ¼ 102 2 : ð10:1:4Þ
m m
kg Ns
q ¼ 103 ; l ¼ 5:6 ; k ¼ 0:0427 s: ð10:1:5Þ
m3 m2
Obviously, the relaxation time is very short. In other words, the propagation of
a disturbance is rather fast. In the case of water the relaxation time is even smaller
by several decades. Thus in that case the influence of a finite speed of propagation
of disturbances should be (negligibly) small.
In combination with the ansatz (10.1.1) and Eqs. (10.1.2/10.1.3) for the y and
z components of the balance of momentum shown in Eq. (3.8.14)1, we conclude
that the pressure can only be a function of position x and of time t—independent of
the constitutive relations. On the other hand the x component of the balance of
momentum reads:
ot op orxy ot orxy op
q ¼ þ ) q ¼ : ð10:1:6Þ
ot ox oy ot oy ox
According to the ansatz (10.1.1) and the constitutive Eqs. (10.1.2/10.1.3) the
function on the left can only depend on y (at all times t). However, as it was just
pointed out, the function on the right can only depend on x (at all times t). Hence
both must be equal and we conclude that they can only be a constant. Thus,
physically speaking, the pressure gradient in flow direction x must be constant. But
since no pressure gradient was applied this constant must be zero and we have:
ot orxy
q ¼ : ð10:1:7Þ
ot oy
If we insert Eq. (10.1.2) for Newtonian fluids the afore-mentioned parabolic
Partial Differential Equation (PDE) results:
ot o2 t l
¼ D 2 ;D ¼ : ð10:1:8Þ
ot oy q
On the other hand we obtain by combination of Eqs. (10.1.3) and (10.1.6) for
MAXWELL fluids:
ot o2 t o2 rxy
q ¼l 2k : ð10:1:9Þ
ot oy oyot
254 10 Selected Problems for Newtonian and Maxwellian Fluids
If we now differentiate the field equation (10.1.6) w.r.t. time and recall that the
fluid is incompressible the following hyperbolic equation is finally obtained:
o2 t 1 ot 2
2o t D
þ ¼ c ; c2 ¼ : ð10:1:10Þ
ot2 k ot oy2 k
The quantity c has the dimension of a velocity. By using the data from above
we find that:
1=2 1=2
D l m
c¼ ¼ ¼ 0:36 : ð10:1:11Þ
k qk s
o2 u o2 u o2 u ou ou
A 2
þ 2B þ C 2 þ a þ b þ cu ¼ f : ð10:1:12Þ
ox oxoy oy ox oy
In older textbooks it is customary to use the following quantity d for a
classification of the PDE:
d ¼ AC B2 : ð10:1:13Þ
We say that the PDE is of the elliptic type if d [ 0, of parabolic type if
d ¼ 0; and of hyperbolic type if d [ 0: Use that scheme and show that Eqs.
(10.1.8) and (10.1.10) represent PDEs of the parabolic and of the hyperbolic
type, respectively. Interpret c as a wave propagation speed (of which phys-
ical quantity?). Moreover, consult modern mathematics textbooks or the
internet and find alternative classification schemes for PDEs.
In order to solve Eqs. (10.1.8) and (10.1.10), respectively, we need initial and
boundary conditions. The boundary conditions read in both cases:
tð0; tÞ ¼ V ðtÞ; tðH; tÞ ¼ 0; t0 ð10:1:14Þ
and the initial conditions:
ot
tðy; 0Þ ¼ 0; ðy; 0Þ ¼ 0; t ¼ 0; ð10:1:15Þ
ot
where the second one is only required for the MAXWELL case, which led to a PDE
of second order in time. As plate velocity VðtÞ we choose the step function shown
in Fig. 10.2:
10.1 Some Comments on Modeling, in Particular Modeling 255
t=0 t
ffi
0; t0
V ðt Þ ¼ ð10:1:16Þ
V0 ; t[0
For t ! 1 stationary conditions must result. These have already been analyzed
in Exercise 7.4.1 and we may write:
y
ts ð yÞ ¼ V0 1 where ts ¼ lim tðy; tÞ: ð10:1:17Þ
H t!1
If ts ðyÞ is split off from the full solution the calculations simplify considerably.
Thus we define a reduced velocity:
wðy; tÞ ¼ tðy; tÞ ts ð yÞ ð10:1:18Þ
and obtain the following PDEs:
ow o2 w o2 w 1 ow o2 w
¼ D 2 and 2 þ ¼ c2 2 : ð10:1:19Þ
ot oy ot k ot oy
Note that the initial and boundary conditions change as well:
wð0; tÞ ¼ 0; wðH; tÞ ¼ 0 ð10:1:20Þ
and:
ow
wðy; 0Þ ¼ ts ð yÞ; ðy; 0Þ ¼ 0: ð10:1:21Þ
ot
In the next two chapters we will explain in detail that the parabolic and the
hyperbolic PDE lead to a totally different time behavior of the evolving parallel
flow. In fact the parabolic equation carries in itself the artifact of infinite propa-
gation of disturbances which, from a technological point-of-view, may or may not
be important. We will also see that a numerical solution in form of a FOURIER series
is required, which unless converted into a D’ALEMBERT type of expression must be
cut off after a finite number of terms for practical reasons. This may bring in other
artifacts that we must judge very critically. The problem of a numerical solution
will become even more evident in Sect. 10.4, which deals with the transient as well
as with the stationary behavior of gas clouds, i.e., models for expanding stars if not
for the whole universe. We should always keep in mind that a good model should
capture as many aspects of reality as possible. However, models will never be able
256 10 Selected Problems for Newtonian and Maxwellian Fluids
to mimic all aspects of reality simultaneously. In the end, the final answers can
only be obtained by asking questions to mother nature, as some physicists so aptly
put it, i.e., by performing experiments. Models are work of art, like paintings, and
therefore only feeble substitutes for our material world.
1 v_ w00
¼ ¼ const. ¼ k2 ; ð10:2:2Þ
Dv w
since the left side depends only on time t and the right side only on position y: The
minus sign can be motivated physically: We expect that the motion will be damped
down when time increases. We do not expect excitation. The general solution of
the two Ordinary Differential Equations (ODEs) reads:
v_ þ Dk2 v ¼ 0 ) v ¼ exp Dk2 t ð10:2:3Þ
and:
The constants An must be adjusted to meet the (first) initial condition from Eq.
(10.1.21):
y X 1 np
wðy; 0Þ ¼ V0 1 ¼ An sin y : ð10:2:8Þ
H n¼1
H
We realize that the disturbance spreads infinitely fast from the lower plate at
y ¼ 0 to the position y ¼ H since:
tðy; tÞ 6¼ 0; 8y 2 ½0; H Þ; 8t [ 0: ð10:2:12Þ
A graphic representation of the solution at different times is shown in Fig. 10.3.
For the distance H we chose the value 1 mm.
258 10 Selected Problems for Newtonian and Maxwellian Fluids
Fig. 10.3 Temporal development of the velocity profile for a NAVIER-STOKES fluid
10.2 Transient Channel Flow of a NAVIER-STOKES-Fluid 259
y 2X
1
1 np
1 ¼ sin y ð10:2:13Þ
H p n¼1 n H
represents the FOURIER series of the function f ðnÞ shown in Fig. 10.4.
1
f ð nÞ ¼ ðH ½n 2nH Þ for 2nH\n\ð2n þ 2ÞH ð10:2:14Þ
H
and
1
f ðnÞ ¼ ðH þ ½n þ 2nH Þ for 2ðn þ 1ÞH\n\ 2nH; ð10:2:15Þ
H
where n ¼ 0; 1; 2; . . .. Interpret this function as the periodic continuation of
the straight line 1 Hy :
We now turn to the second PDE shown in (10.1.19). For our discussion it is most
helpful to consider a special case first, a.k.a. the non-dissipative wave propagation:
We neglect the term ow=ot and obtain the classical wave equation:
o2 w 2
2o w
¼ c ; ð10:3:1Þ
ot2 oy2
which will now be solved in combination with the aforementioned initial and
boundary conditions. By using the Eq. (10.2.1) it follows that:
1 €v w00
¼ ¼ k2 ; ð10:3:2Þ
c2 v w
and consequently:
260 10 Selected Problems for Newtonian and Maxwellian Fluids
v þ c2 k 2 v ¼ 0
€ ) v ¼ C sinðcktÞ þ D cosðcktÞ; ð10:3:3Þ
and:
for the solution of the PDE (10.3.1), with An ¼ AC and Bn ¼ AD. The remaining
coefficients will be adjusted to the initial conditions from Eq. (10.1.21). We obtain
in complete analogy to Eq. (10.2.9) of the NAVIER–STOKES case:
X
1 np 2V0
Bn sin y ¼ ts ð yÞ ) Bn ¼ ð10:3:7Þ
n¼1
H np
and
X
1
np np
cAn sin y ¼0 ) An ¼ 0: ð10:3:8Þ
n¼1
H H
2V0 X
1
1 np np
wðy; tÞ ¼ cos ct sin y ; ð10:3:9Þ
p n¼1 n H H
If we define:
n ¼ y ct; ð10:3:13Þ
it becomes obvious that the two sums in Eq. (10.3.12) correspond to FOURIER series
of the saw tooth function shown in Fig. 10.4. Thus by replacing the FOURIER series
we obtain the following compact form of the solution:
y 1
tðy; tÞ ¼ V0 1 ½f ðn Þ þ f ðnþ Þ ð10:3:14Þ
H 2
with:
1
f ð n Þ ¼ ðH ½n 2nH Þ for 2nH \ n \ 2ðn þ 1ÞH; ð10:3:15Þ
H
and:
1
f ð n Þ ¼ ðH þ ½n þ 2nH Þ for 2ðn þ 1ÞH \ n \ 2nH; ð10:3:16Þ
H
where n ¼ 0; 1; 2; . . .: This way of writing is known as D’ALEMBERT’s solution for
the wave equation in the literature.
In order to convince ourselves that the solution really describes a propagating
wave that starts as a disturbance at y ¼ 0 and moves at a finite speed c ¼
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffi
D=k ¼ l=qk into a fluid at rest, we study the following scheme:
H ct = 0 H 2H 3H 4H 5H 6H 7H 8H
y= 2 4 4 4 4 4 4 4 4
H
n ¼ 2H 4 ct
2H
4 4
0 H4 2H
4 3H
4 4H
4 5H
4 6H
4
2H 2H 3H 4H 5H 6H 7H 8H 9H 10H
nþ ¼ 4 þ ct 4 4 4 4 4 4 4 4 4
2 3 4
f ðn Þ 4 4 4 34 24 14 0 1
4
2
4
2 1
f ðnþ Þ 4 4
0 14 24 34 44 3
4
2
4
mðH=2;tÞ 0 0 0 1 1 1 1 0 0
V0
Obviously, the velocity tðH=2; tÞ of the fluid at the center of the fluid, y ¼ H=2;
is equal to zero until ct ¼ H=2: At that time t jumps to V0 6¼ 0; where it remains
until the time ct ¼ 6H=4 has been exceeded. This corresponds to the amount of
time it takes until the reflection (i.e., the negative) of the first wave at the wall
y ¼ H has reached the position y ¼ H=2 and eliminates the original disturbance
from y ¼ 0 completely. The temporal development at all positions y is shown in
the sequence of plots in Fig. 10.5.
With the same methods we will now tackle the general MAXWELL case with
damping. In other words, we will solve the second PDE from (10.1.19) with the
initial and boundary conditions shown in Eqs. (10.1.20) and (10.1.21). Separation
of variables according to Eq. (10.2.1) yields:
262 10 Selected Problems for Newtonian and Maxwellian Fluids
Fig. 10.5 Temporal development of the solution for a propagating wave w/o damping
1 k€v þ v_ w00
¼ ¼ k2 ð10:3:17Þ
D v w
and thus:
1 1
€v þ v_ þ k2 v ¼ 0; ð10:3:18Þ
c2 D
10.3 Transient Channel Flow of a MAXWELL Fluid 263
c2
m2 þ m þ c2 k2 ¼ 0; ð10:3:20Þ
D
which has two solutions:
s ffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
c2 c2 2 2 2 1 n2 p2
m ¼ c k ¼ 1 1 4k2 2 c2 ; ð10:3:21Þ
2D 2D 2k H
i.e.,
v ¼ Aþ
n expðmþ tÞ þ An expðm tÞ: ð10:3:22Þ
Here we have already made use of the solution of the second remaining ODE
from (10.3.17). As in the previous case we must write:
Note that the constant A has been chosen as 1. The relation (10.3.25) will now
be adjusted to meet the initial conditions shown in Eq. (10.1.21). We obtain two
equations for the unknown coefficients A n:
X
1 np y
Aþ
n þ An sin y ¼ ts ð yÞ ¼ V0 1 ð10:3:26Þ
n¼1
H H
and:
X
1 np
mþ Aþ
n þ m An sin y ¼ 0: ð10:3:27Þ
n¼1
H
or:
m 2V0 mþ 2V0
Aþ
n ¼ ;A ¼ : ð10:3:29Þ
mþ m pn n mþ m pn
These formulae are inserted in Eq. (10.3.25) and the representation (10.1.18) is
observed. Hence the following equation for the velocity field results:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
y 2V0 t X 1
1 t n2 p2
tðy; tÞ ¼ V0 1 exp cosh 1 4k2 2 c2 þ
H p 2k n¼1 n 2k H
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!!
n2 p2 t n2 p2 np
1 4k2 2 c2 sinh 1 4k2 2 c2 sin y :
H 2k H H
ð10:3:30Þ
We will now investigate the solution somewhat further and simplify appro-
priately. In particular, we are interested to know as to whether a change in velocity
at the lower plate will propagate infinitely fast, as in the NAVIER-STOKES case, or
moves at a finite speed, as in the previously discussed case of a traveling wave.
First note that Eq. (10.3.30) will lead to solutions of the wave type only if:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2n p 2
2 2 n2 p2 n2 p2
1 4k 2
c ¼ i 4k2 2 c2 1 with 4k2 2 c2 1: ð10:3:31Þ
H H H
For n ¼ 1 and with the data for polyisobutylene in decahydronaphthalene it
follows that:
p2 2 Dk 9:4 103 m2
4k2 2
c ¼ 4p2 2 ¼ 1: ð10:3:32Þ
H H H2
This relation is satisfied only if H 9:7 cm: Otherwise we suspect that the
velocity tðy; tÞ at an arbitrary point y is different from zero for t [ 0: In what
follows we restrict ourselves to the case H\9:7 cm: With the identities
coshðiaÞ ¼ cosðaÞ; sinhðiaÞ ¼ i sinðaÞ ð10:3:33Þ
the solution for the velocity field (10.3.30) becomes:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
y 2V0 t X 1
1 t n2 p2
tðy; tÞ ¼ V0 1 exp cos 4k2 2 c2 1
H p 2k n¼1 n 2k H
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!!
1 t n2 p2 np
þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sin 4k2 2 c2 1 sin y :
4k2 n p c2 1
2 2 2k H H
H2
ð10:3:34Þ
10.3 Transient Channel Flow of a MAXWELL Fluid 265
This relation can be simplified further for the following special case. First, we
have:
8
3 2
>
< 9400 for H ¼ 1 mm
kD 9:4 10 m
4p2 2 ¼ ¼ 94 for H ¼ 1 cm ð10:3:35Þ
H H2 >
:
0:94 for H ¼ 10 cm
Consequently the term ‘‘1’’ in the square roots of Eq. (10.3.34) can be neglected
for H 1 mm. Thus by the addition theorem shown in (10.3.11) and the following
one:
1
sinðaÞ sinðbÞ ¼ ½cosða bÞ cosða þ bÞ ð10:3:36Þ
2
it follows that:
(
y V0 t X 1
1 hnp i hnp i
tðy; tÞ ¼ V0 1 exp sin ðy ctÞ þ sin ðy þ ctÞ
H p 2k n¼1
n H H
)
1 H X1 1 hnp i hnp i
þ pffiffiffiffiffiffi 2
cos ðy ctÞ cos ðy þ ctÞ :
2p kD n¼1 n H H
ð10:3:37Þ
In this special case the FOURIER series can be rewritten by means of the function
f ðnÞ introduced in Exercise 10.2.1, and a quadratic function gðnÞ that is defined
piecewise as follows:
1 6 3 2
gð nÞ ¼ 2 ½n 2nH þ 2 ½n 2nH ð10:3:38Þ
12 H H
for 2nH\n\2ðn þ 1ÞH and
1 6 3 2
gð nÞ ¼ 2 þ ½n þ 2nH þ 2 ½n þ 2nH ð10:3:39Þ
12 H H
for 2ðn þ 1ÞH \ n \ 2nH: It can be expanded into the following FOURIER
series:
1X 1
1 np
gð nÞ ¼ cos n : ð10:3:40Þ
p2 n¼1 n2 H
Fig. 10.6 Temporal development of the velocity profile for a MAXWELL fluid
10.3 Transient Channel Flow of a MAXWELL Fluid 267
y V0 t ffi H
tðy; tÞ ¼ V0 1 exp f ð n Þ þ f ð nþ Þ þ p ffi ffi ffi ffi ffiffi ð gð n Þ gð nþ Þ Þ :
H 2 2k kD
ð10:3:41Þ
Figure 10.6 shows the result of a numerical evaluation of the various terms in
Eq. (10.3.37) for the velocity profile of transient plan-parallel flow of a MAXWELL
fluids at various times t: It becomes evident that from a certain point in time the
profile cannot be distinguished from the stationary one any more.
Exercise 10.3.2: FOURIER series for the D’ALEMBERT solution of the transient
flow of a MAXWELL fluid
Prove Eq. (10.3.40) and explain concisely its relevance for obtaining the
solution shown in Eq. (10.3.41).
with better quantitative models of reality we have to keep in mind that they are
also just models after all. In fact, classical theories allow us frequently to get first
useful insights of what the problem is all about. An example of this kind is the
application of classical continuum concepts to the problem of an expanding or
contracting ‘‘sphere of gas’’ which may serve as a (coarse) model for the stars or
even for the whole universe.
We will assume perfect spherical symmetry, i.e., no angular dependence. This
means that the velocity field in spherical coordinates is given by:
thri ¼ thri ðr Þ; th#i ¼ 0; thui ¼ 0: ð10:4:1Þ
Thus the balance of mass (5.2.10) reduces to:
oq oq othri 2q
þ thri þ q þ thri ¼ 0: ð10:4:2Þ
ot or or r
Before we turn to the balance of momentum we reduce the form of the NAVIER-
STOKES stress tensor. If we insert the ansatz (10.4.1) into Eqs. (6.3.6/6.3.9) we find
that all shear components vanish and obtain for the normal stresses:
othri thri
rhrri ¼ p þ ðk þ 2lÞ þ 2k ; ð10:4:3Þ
or r
othri thri
rh##i ¼ rhuui ¼ p þ k þ 2ðk þ lÞ :
or r
Note that for a self-gravitating sphere the specific volume force acts only in
r-direction:
fhri ¼ fhri ðr Þ; fh#i ¼ 0; fhui ¼ 0: ð10:4:4Þ
Thus, if we insert these relations in the balance of momentum shown in
Eq. (5.4.7), and if we observe once more Eq. (10.4.1), we see that the # and u
components are identically satisfied. However, the r component reads:
2
othri othri o thri 2 othri 2thri op
q þ th r i ðk þ 2lÞ 2
þ 2 ¼ þ q fhri:
ot or or r or r or
ð10:4:5Þ
Equations (10.4.2) and (10.4.5) are two equations for the two unknowns, q and
thri : However, they are not field equations yet, unless we specify the pressure and
the body force in terms of the unknowns. We start with the body force. According
to NEWTON’s law of gravity we find for an infinitesimal contribution to the grav-
itational acceleration between a mass element qðr ÞdV and a test mass M outside of
the sphere (cf., Fig. 10.7 left):
10.4 Expanding and Contracting Stars and Universes 269
M ρ (r )dV
G
x2 ρ (r )dV
ζ
r
ρ (r ) dV
x x
ζ
df <r >
ϑ r
M r
rϑ
Fig. 10.7 Test mass M outside and within radially heterogeneous spherical regions
kgs
nometry in combination with Eq. (4.5.13)2 leads to:
ZR Zp Z2p
qðr Þ r 2 sin # ðr r cos #Þdud#dr
fhri ¼ G : ð10:4:7Þ
ðr 2 þ r 2 2rr cos #Þ3=2
r ¼0 #¼0 u¼0
With suitable substitutions (see the following exercise) the integrations can
relatively easily be performed and the following astonishing result is obtained:
ZR
mð R Þ
fhri ¼ G ; mðRÞ ¼ 4p qðr Þ r 2 dr ; r R: ð10:4:8Þ
r2
r¼0
Perform the integrations shown in Eq. (10.4.7). To this end use the fol-
lowing substitutions and integral:
Z
r ð1 ayÞdy ya
y ¼ cos #; a ¼ ; 3=2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð10:4:9Þ
r 2
ð1 þ a 2ayÞ 1 þ a2 2ay
270 10 Selected Problems for Newtonian and Maxwellian Fluids
Also use Fig. 10.7 (left) in combination with the sine and cosine rule to
explain how Eq. (10.4.7) was obtained. Why is it possible to orient the test
mass M the way it is shown in the figure? What happens to the gravitational
forces dfh0ri i.e., the pull within planes of constant angle # perpendicular to the
vertical axis?
Consider now the situation shown on the right hand side of Fig. 10.7: The
test mass M is situated within the hole of a radially heterogeneous spherical
shell. Show that the total gravitational pull can be expressed by the following
integral:
ZR Zp Z2p
qðr Þ r 2 sin # cosðf þ #Þdud#dr
fhri ¼ G : ð10:4:10Þ
r 2 þ r 2 2rr cos #
r¼0 #¼0 u¼0
Perform the integrations by using the substitutions of Eq. (10.4.9) and the
following formula:
Z
ð1 byÞy þ ð1 y2 Þb 1 þ by 1
3=2
dy ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; b ¼ ; ð10:4:11Þ
ð1 þ b2 2byÞ 2
b 1 þ b 2by 2 a
and show that the test mass is free of forces, even for the case of a radially
heterogeneous, hollow sphere.
Now consider a homogeneous sphere of a constant mass density q0. Argue
that the gravitational acceleration of a test mass at radial position r is given
by:
ffi 4p
3 Gq0 r if 0 r R
f ¼ fhri er ; fhri ¼ ð10:4:12Þ
G mrð2RÞ ; mðRÞ ¼ 4p3 q0 R3 if R r\1:
Recall the statement made in context with Eq. (7.2.10) according to which
conservative forces can be obtained by differentiation of a potential, u. Show
that the following formulae hold in the present case:
(
4p
Gq0 r 2 þ C1 if 0 r R
f ¼ ru; u ¼ uðr Þ ¼ 6 mðRÞ ð10:4:13Þ
G r þ C2 if R r\1;
where the del operator is given by Eq. (5.6.5). Determine both constants, C1
and C2, from the requirement that the potential is continuous at the sphere’s
surface, i.e., at R, and vanishes at infinity so that:
( h 2 i
12G mðRRÞ 3 Rr if 0 r R
u ¼ uðr Þ ¼ : ð10:4:14Þ
m ð RÞ
G r if R r\1
10.4 Expanding and Contracting Stars and Universes 271
Recall that the average value of the Earth’s radius is ca. 6370 km. Show
that the average mass density of Earth is therefore given by
3g kg
qE ¼ 5500 3 ð10:4:16Þ
4pGR m
and that it would take roughly 2500 s, i.e., 42 min to send a person to the
other side of the globe by free frictionless fall through a radial hole. Why is
the acceleration of this person during the fall always less or equal to 1 g or,
in other words, why are the accelerations harmless? What is the maximum
velocity that could be reached? Moreover, find arguments why a consider-
able portion of the Earth must consist of iron.
As it was pointed out in Sect. 6.4 the pressure is given by the so-called thermal
equation of state as a function of density and temperature, p ¼ pðq; T Þ. Thus we
conclude that Eq. (10.4.5) can be rewritten as follows:
2
othri othri o thri 2 othri 2thri
q þ thri ðk þ 2lÞ þ 2
ot or or 2 r or r
ð10:4:17Þ
op oq op oT m ðr Þ
þ þ ¼ Gq 2 :
oq or oT or r
If we now use the thermal equation of state for an ideal gas, Eq. (6.4.1), this
turns into:
2
othri othri o thri 2 othri 2thri
q þ thri ðk þ 2lÞ þ 2
ot or or 2 r or r
ð10:4:18Þ
R oq oT m ðr Þ
þ T þq ¼ Gq 2 :
M or or r
Similar to the arguments in Sect. 7.5 we are now going to turn the system of
coupled PDEs, Eqs. (10.4.2) and (10.4.18), into a system of ODEs by choosing a
suitable semi-inverse ansatz. Basically we claim that the density and temperature
fields are homogeneous but time-dependent. We assume perfect spherical
272 10 Selected Problems for Newtonian and Maxwellian Fluids
symmetry for the velocity field or, in other words, the angular components vanish.
Moreover, the radial velocity follows HUBBLE’s law with a time-dependent HUB-
BLE’s constant:
thri r_
q ¼ qðtÞ; T ¼ T ðtÞ;
¼ H ðtÞ; th#i ¼ thui ¼ 0: ð10:4:19Þ
r r
In the equation for the radial velocity we have used the explicit form for its
physical component shown in Eq. (5.2.7)1. It can be integrated easily:
0 ~t¼t 1
Z
r ðtÞ ¼ r ð0Þ exp@ H ð~tÞd~tA; 0 r ðtÞ RðtÞ; 0 r ð0Þ Rð0Þ: ð10:4:20Þ
~t¼0
q
d
dH 4p Gqð0Þ
þ 3
qH ¼ 0; 2 þ c
þH q ¼ 0; c¼ : ð10:4:25Þ
dt dt 3 H 2 ð0Þ
10.4 Expanding and Contracting Stars and Universes 273
Fig. 10.8 Mass density, HUBBLE’s constant, and outer radius over time for c ¼ 0:4 (see text)
Fig. 10.9 Mass density, HUBBLE’s constant, and outer radius over time for c ¼ 0:51 (see text)
RRR
energy is not conserved due to the term r : rt dV: This term can be inte-
V ðt Þ
grated for the case of perfect spherical symmetry by using the ansatz from
Eq. (10.4.19) in combination with Eqs. (10.4.3), (4.5.13)2, and (5.7.21)2. The final
result reads:
ZZZ
r : rt dV ¼ ½p þ ðk þ 2lÞH ðtÞH ðtÞ 4p R3 ðtÞ: ð10:4:27Þ
V ðt Þ
ZZ Interestingly the same result, but with the opposite sign, is obtained for the term
n r tdA; if we only observe Eq. (4.5.14)2. Thus the two terms involving the
oV
stress
RRR q 2 tensor cancel each other. The general formula for the kinetic energy
2t dV can also easily be integrated:
V ðtÞ
ZZZ
q 2 3
Ekin ðtÞ ¼ t dV ¼ mðRÞR2 ðtÞH 2 ðtÞ;
2 10
V ðt Þ ð10:4:28Þ
4p 4p
mðRÞ ¼ qðtÞR3 ðtÞ
qð0ÞR3 ð0Þ:
3 3
Moreover, as we shall prove later, the result for the gravitational field shown in
Eq. (10.4.12) can be used to prove the following formulae:
10.4 Expanding and Contracting Stars and Universes 275
ZZZ
d 3 m2 ðRÞ 3 m2 ð R Þ
f t dV ¼ G ; Epot ðtÞ ¼ G : ð10:4:29Þ
dt 5 RðtÞ 5 Rð t Þ
V ðtÞ
We may interpret the potential energy as the total gravitational energy stored in
a sphere of gas of perfect radial symmetry. Moreover, we conclude that the bal-
ance of kinetic energy can be rewritten in the following form:
d
Epot ðtÞ þ Ekin ðtÞ ¼ 0: ð10:4:30Þ
dt
In other words, the mechanical energy is surprisingly conserved under the present
assumptions even though the gas was treated as a viscous NAVIER-STOKES fluid:
Epot ðtÞ þ Ekin ðtÞ ¼ Epot ð0Þ þ Ekin ð0Þ: ð10:4:31Þ
The definitions for kinetic and potential energy can now be used to interpret the
factor 2c originally defined in Eq. (10.4.25)3. A simple calculation shows that it is
nothing else but the ratio between the total gravitational and the kinetic energy
initially assigned to the sphere of gas:
Epot ð0Þ Gmð0Þ
¼2 3
2c: ð10:4:32Þ
Ekin ð0Þ R ð0ÞH 2 ð0Þ
If the initial amount of kinetic energy is very large, i.e., for small values of c;
gravity is not strong enough and the sphere will continue to expand, and vice
versa. It is exactly that kind of behavior we see in Figs. 10.8 and 10.9. A threshold
value is reached when both energies are equally strong, i.e., c ¼ 12. It is also quite
instructive to ask under which circumstances the expanding star will come to a
standstill. In that case the kinetic energy must vanish and Eq. (10.4.31) yields:
2c
Epot ð0Þ þ Ekin ð0Þ ¼ Epot ðtÞ ) Rð t Þ ¼ Rð 0 Þ : ð10:4:33Þ
2c 1
Since the radius must always be positive (plus infinity is also acceptable) we
conclude that in order to reach a state of rest it is required that c 12. Recall that in
the original definition c was only required to be positive. Therefore choices
0 c\12 will lead to continuously expanding stars or, in other words, a never-
ending blast.
We finish our discussions with a proof of Eq. (10.4.29). The starting points of
our line of arguments form Eqs. (10.4.12)2 and (10.4.14), which we apply to the
case of a gravitating expanding star of current radius RðtÞ and total mass mðtÞ ¼
276 10 Selected Problems for Newtonian and Maxwellian Fluids
4p
3qðtÞR3 ðtÞ; thus acknowledging Eq. (10.4.19)1 for a time-dependent density and a
time-dependent external radius. Only the fields within the star, i.e., 0 r RðtÞ are
relevant for the proof and, consequently, we write:
4p 2p
fhri ¼ GqðtÞr; u ¼ GqðtÞ 3R2 ðtÞ r 2 : ð10:4:34Þ
3 3
Recall that due to the perfect spherical symmetry this is the only non-vanishing
component of the body force. Eq. (10.4.13)1 allows us to write:
o
f ¼ ru , fhri ðr; tÞ ¼ uðr; tÞ: ð10:4:35Þ
or
These results will now be used to obtain the total potential energy of the
gravitating mass. To begin with, the total potential energy is given by:
ZZZ
1
Epot ðtÞ ¼ qu dV: ð10:4:36Þ
2
V ðt Þ
which holds only in the resent case of the perfectly symmetric star. The auxiliary
formula is easy to proof, if we start from Eq. (10.4.34)2 and differentiate:
ou 2p 3 4p
¼ Gq_ 3R2 r 2 4pGqRR_ ¼ GqH R2 r 2 ;
ot 3 2 3 ð10:4:40Þ
ou 4p
¼ Gq r;
or 3
where Eqs. (10.4.21) and (10.4.23)1 have been used. This can now be integrated
over the total volume of the star and we find:
10.4 Expanding and Contracting Stars and Universes 277
ZZZ ZRðtÞ
ou ou 2 ð4pÞ2 2 5
q dV ¼ 4p q r dr ¼ Gq HR ; ð10:4:41Þ
ot ot 15
V ðt Þ r¼0
where Eq. (10.4.19)3 has been used. Now we are ready to perform the remaining
integration in Eq. (10.4.36):
ZZZ Z ðtÞ
r¼R
1 4p2 2
Epot ðtÞ ¼ qu dV ¼ Gq 3R2 r 2 r 2 dr
2 3
V ðt Þ r¼0 ð10:4:42Þ
2
ð4pÞ 3 m2 ð R Þ
¼ Gq2 R2 ¼ G :
35 5 R
This concludes the cumbersome proof of Eq. (10.4.29) without any artificial
tricks as they are frequently used, e.g., by calculating the work required for
transporting spherical shells of mass successively from infinity to form the total
volume of the expanding star.
Study the alternative proofs of Eq. (10.4.29) shown on the World Wide
Web (for example). Also show that the result still holds for radially and time
dependent mass densities q ¼ qðr; tÞ for 0 r RðtÞ:
From the proof above we have learned that the balance of kinetic energy
is unaffected by the dissipative stress terms (cf., Eq. (10.4.27) and the cor-
responding text). Thus the sum of kinetic and potential energy does not
change over time and the corresponding balance decouples completely from
the balance of internal energy. However, all of this does not mean that the
temperature stays constant, too: The total internal energy of the sphere of gas
is conserved (in the case of vanishing bulk viscosity), but the sphere expands,
and so the temperature should decrease over time since the internal energy is
distributed over a greater volume. We will explore this in more detail: Show
the local balance of internal energy, Eq. (3.9.7), results in the following ODE
for the dimensionless temperature, T ¼ TðtÞ=Tð0Þ:
278 10 Selected Problems for Newtonian and Maxwellian Fluids
dT 3 ð3k þ 2lÞH ð0Þ
q
¼
qT þ H H: ð10:4:43Þ
dt f qð0ÞR=MT ð0Þ
Assume that the star can be treated as an ideal gas, and use Eq. (10.4.3) to
evaluate the stresses. Note the different signs of the two terms in the
parentheses. What is their effect regarding the change in temperature? Show
that the dissipative term in Eq. (10.4.43) vanishes for the case of vanishing
bulk viscosity, l0 ¼ 0: Interpret the ansatz (10.4.1) as pure volumetric
deformation and relate it to the ‘‘compressibility factor’’ 3k þ 2l in
Eq. (10.4.43) (in the same context also discuss the analogue for the Hookean
solid shown in Eq. (6.2.6)2).
Neglect the mass of the atmosphere and show that we may write for the
gravitational field of the Earth:
g Dr Dr 2
fhri ¼ 2
¼ g 1 2 þ 3 2 ; Dr ¼ r rE :
ð1 þ DrrE Þ
rE rE
ð10:4:44Þ
g ¼ G mr2E ¼ 9:81 m
s2 stands for the gravitational acceleration at ground level,
E
mE and rE denote the mass and the radius of the Earth, respectively. Use the
stationary momentum balance and the ideal gas law to show that the mass
density of an isothermal atmosphere decreases exponentially:
" !#
1 GmE
qðr Þ ¼ qðrE Þ exp CE 1 ; CE ¼ R ; Dr ¼ r rE :
1 þ Dr
rE M TrE
ð10:4:45Þ
Expand into a TAYLOR series and rediscover the classical expression for
the isothermal barometric equation:
g g
qðr Þ qðrE Þ exp R Dr , pðr Þ pðrE Þ exp R Dr :
MT MT
ð10:4:46Þ
Explain why the pressure at ground level must equal the weight of the
atmosphere per unit surface, which is given by the following integral:
10.4 Expanding and Contracting Stars and Universes 279
Z1
pð r E Þ ¼ qðrE Þ fhri dr: ð10:4:47Þ
r¼rE
Fig. 10.10 Gas mass and gas mass density for r-2 varying gravity
Zr0 ¼r
dm
mðr Þ ¼ 4p qðr 0 Þ r 02 dr 0 ) ¼ 4pqðr Þ r 2 : ð10:4:53Þ
dr
r0 ¼0
We also assume that the planet is isothermal and obtain the following system of
coupled ODEs:
10.4 Expanding and Contracting Stars and Universes 281
dm dq G mðr Þ
¼ 4pqðr Þ r 2 ; þ R q 2 ¼ 0: ð10:4:54Þ
dr dr M T r
This system can be uncoupled by eliminating the mass. We obtain the following
non-linear second order ODE for the mass density:
2
d2 q dq dq 4p G
rq 2 r þ2q þ R rq3 ¼ 0: ð10:4:55Þ
dr dr dr MT
Note that the differential equations hold for all material points of the stationary
gas cloud. Due to the assumed perfect spherical symmetry this means that they
hold within 0 r R where R denotes the outer radius of the cloud. However, the
point r ¼ 0 will cause trouble since the ODEs degenerate at that point. Moreover,
Exercise 10.4.5 taught us that we should be prepared for a semi-infinite domain,
i.e., the case R ! 1: Consequently, a ‘‘natural’’ length scale parameter for nor-
malization is not available. Therefore we choose an arbitrary radius r0 2 ð0; 1Þ
and use it to define a dimensionless radius r ¼ r=r0 : Moreover, we expect a finite
value qð0Þ for the mass density at r ¼ 0 and define a dimensionless density by
ðr Þ ¼ q=qð0Þ: Thus Eq. (10.4.55) reads:
q
2
d2 q
q
d q
d
r q 2 r
þ2q þ 3Cr q 3 ¼ 0; 0 r \1 ð10:4:56Þ
dr dr dr
where:
Gm0 4p
C¼ R ; m0 ¼ qð0Þr03 : ð10:4:57Þ
M Tr0
3
The mass-like quantity m0 was introduced only formally and has no immediate
physical meaning, unlike its analogue in Eq. (10.4.45)2. The boundary conditions
to be observed in context with Eq. (10.4.56) read:
q
d
ð0Þ ¼ 1;
q ¼ 0: ð10:4:58Þ
dr r¼0
The first one follows from the definition of the normalized mass density, and the
second one from Eq. (10.4.54)2 since there is no mass at r ¼ 0. In fact the vicinity
of r ¼ 0 presents a problem during the numerical solution of the non-linear
boundary value problem (10.4.56/10.4.58). Therefore we solve it for r 2 ½e; 1Þ,
with e ¼ 104 and explore the situation for r 2 ½0; e analytically by using the
following quadratic ansatz:
ðr Þ ¼ a þ b r þ c r 2 :
q ð10:4:59Þ
If this is inserted in (10.4.55/57) we find that:
C C 2
a ¼ 1; b ¼ 0; c ¼ ) ðr Þ ¼ 1
q r : ð10:4:60Þ
2 2
282 10 Selected Problems for Newtonian and Maxwellian Fluids
Fig. 10.11 Mass density and mass distribution in a sphere of gas (see text)
The radial development of mass density and mass is shown in Fig. 10.11.
Obviously mass is accumulated linearly at large distances. In view of the last
equation we conclude that the mass density varies like r 2 . If such an ansatz is
inserted in the ODE (10.4.55) we find that:
2C 1
ðr Þ ¼
q ; r
1: ð10:4:62Þ
3 r 2
This leads us to conclude that stability between internal pressure and gravity is
possible for a gas cloud of infinite size containing infinite mass. To hear the word
infinite twice is a bit disturbing and we should try to improve the model, especially
since it is hard to believe that the mass of a gas planet like Jupiter is infinite.
Maybe a rigid core of sufficiently high mass mc ¼ 4p 3
3 qc rc with a (constant) mass
density qc and a radius rc would increase the gravitational pull so much that the
mass density would decrease faster and, after integration, lead to a finite amount of
gas distributed within an infinitely large spherical shell? As a matter of fact we
have already studied such situations in Exercise 10.4.5. There we have noticed that
a constant gravitational field g reaching out to infinity would indeed lead to an
10.4 Expanding and Contracting Stars and Universes 283
atmosphere of finite mass: Eq. (10.4.50). However, the gravitational field of a solid
core is not constant. Rather it follows NEWTON’s law of gravity and decreases like
r 2 . This results in an atmosphere of infinite mass: Eq. (10.4.51). However, in
view of Figs. 10.101,2 we may argue that the accumulated mass seems to saturate
initially. The density and the pressure at the onset of saturation are governed by the
factor A ¼ expðCE Þ and thus extremely small. Consequently we may consider
the corresponding position as the ‘‘extension’’ of the planet incl. its ‘‘atmosphere’’
and disregard the later increase of mass as an artifact of the model. In fact, we
neglected the contribution of the atmosphere’s mass to gravity in order to obtain
the closed-form result shown in Eq. (10.4.51). If we now include the mass of the
atmosphere the stabilizing effect should even be more pronounced. However, this
case can only be studied numerically. We proceed to discuss the details.
We define a dimensional radius in a ‘‘natural’’ manner by r ¼ r=rc : However, in
the case of mass density we have a choice, q ðr Þ ¼ q=qc or qðr Þ ¼ q=qðrc Þ: We
will choose the latter way of normalization. However, then the differential equa-
tion for the normalized density differs slightly from Eq. (10.4.56):
2
d2 q
dq q
d
2 r
r q þ2q þ 3aCcr q 3 ¼ 0; 1 r \1 ð10:4:63Þ
dr dr dr
where:
Gmc 4p qðrc Þ
Cc ¼ ; mc ¼ q r3 ; a¼ ; ð10:4:64Þ
R
M Trc
3 c c qc
Solving the ODE (10.4.63/10.4.65) for the density and then performing the
integration for the mass shown in (10.4.66), both numerically, is, in principle, a
feasible way, at least for small normalized radii r . However, we would like to
study the behavior at very large values of r and find out as to whether the mass
saturates. Thus it is much more advisable to eliminate the mass density from
Eqs. (10.4.54) and to derive the following ODE for the mass instead:
d2 m 2 G m dm
¼ 0; rc r\1: ð10:4:67Þ
dr 2 r MR T r 2 dr
284 10 Selected Problems for Newtonian and Maxwellian Fluids
Two boundary conditions are required. The first stems from the fact that at the
radius rc of the core, the mass of the rigid core is present:
4p
mðrc Þ ¼ q r 3
mc : ð10:4:68Þ
3 c c
The second boundary condition follows from Eq. (10.4.54)1:
dm 3mc qðrc Þ
¼ : ð10:4:69Þ
dr rc rc qc
Fig. 10.12 Mass distribution in a gas planet with a solid core (see text)
¼ Ar a ;
m ¼ Br b :
q ð10:4:73Þ
After insertion we find the following relations:
aA b
¼ r baþ3 ; ¼ r a1 : ð10:4:74Þ
3 qðqrc Þ B Cc A
c
The constancy and independence of radius of the left hand sides must still be
guaranteed if r ! 1. Thus we conclude that:
a ¼ 1; b ¼ 2: ð10:4:75Þ
286 10 Selected Problems for Newtonian and Maxwellian Fluids
This means that the mass grows linearly1 and the density decreases quadrati-
cally with increasing distance:
2 2 1
¼
m r ; ¼
q : ð10:4:76Þ
3Cc qðqrc Þ r
Cc 2
c
Thus the situation is very similar to that discussed in context with Eq. (10.4.51).
The last row of double-logarithmic plots in Fig. 10.12 (for qðrc Þ=qc ¼ 0:5) shows
the development of mass together with the asymptotic evolution for Cc ¼ 5 (left)
and Cc ¼ 35 (right). The stabilizing influence of and increasing value of Cc , i.e.,
increasing core mass is clearly visible. However, in the end the mass diverges
unless Cc ! 1:
Show that the mass within the rigid core grows like
c ðr Þ ¼ r 3 ;
m 0 r 1: ð10:4:77Þ
Also determine the constants A and B in Eq. (10.4.73). Gather information
from the World Wide Web and find out more about the core-aggregation
hypothesis used to explain the genesis of gas giants like Jupiter or Saturn.
1
Note that the third order increase of mass (cf., Exercise 10.4.3) was reduced drastically due to
the gravitational action of the atmosphere.
10.5 Would You Like to Know More? 287
Finally, those who wish to learn more about the stationary or dynamic behavior
of stars and gas planets are recommended to consult the monograph by
Kippenhahn et al. [3] and the literature cited therein.
References
In what follows we consider a hollow sphere made of metal (Fig. 11.1), the
pressure vessel, which is subjected to a very high internal pressure. In fact, the
pressure is supposed to be high enough to induce plastic deformation in the metal.
Recall Exercise 2.6.5 where it was mentioned that the VON MISES equivalent stress
rMises must reach the following threshold for this purpose:
pi
Ri
3
r2Mises R ij R ij ¼ r2y ð11:1:1Þ
2 ðxÞ ðxÞ
Recall the results from Sect. 9.5. The linear-elastic solution for the non-vanishing
stresses and strains of a completely spherically symmetric problem in physical
spherical coordinates reads:
B B
rhrri ¼ 3kA 4l ; rh##i ¼ rhuui ¼ 3kA þ 2l ;
r3 r3 ð11:2:1Þ
B
uhri ¼ Ar þ ; 3k ¼ 3k þ 2l:
r2
We now observe the boundary conditions of the problem, i.e., the radial stress
must equal the internal pressure p at radius Ri , and the outer pressure vanishes at
radius Ro , and find:
11.2 The Radially Symmetric Solution 291
ðRo Þ3 1 1 Ro 3
ð Þ þ1
rhrri ¼ p r 3 ; rh##i ¼ rhuui ¼ p 2 r 3 ;
Ro
Ri 1 Ro
Ri 1
ffi ð11:2:2Þ
p ð1 þ mÞR3o 1
uh r i ¼ ð1 2mÞr þ R 3 ; Ri r Ro :
E 2r 2 o
1
Ri
Verify Eqs. (11.2.1/11.2.2) and state all assumptions required for their
derivation. Use results of previous exercises for that purpose. In particular
show that:
1 B
3kA ¼ p 3 ; 3kA ¼ 4l 3 : ð11:2:3Þ
Ro
1 R o
Ri
We will now evaluate the VON MISES flow condition with these results. We start
by calculating the stress deviator. Equation 2.6.11 allows us to write in spherical
coordinates:
2 3
Ro 3 2 0 0
p ð Þ
Rhiji ¼ r3 4 0 1 0 5; ð11:2:4Þ
2 Ro 1
Ri 0 0 1
since:
1 1
r þ rh##i þ rhuui ¼ 3kA ¼ p 3 : ð11:2:5Þ
3 hrri R o
1
Ri
This result will now be inserted into the flow condition from Eq. (2.6.20). We
obtain:
" #2
ðRro Þ3
2 3 3 2 1
rMises ¼ Rhiji Rhiji ¼ p 3 1þ
2 2 Ro
1 2
Ri
3 ðRo Þ3
) rMises ¼ p r3 ¼ rMises ðp; r Þ:
2 Ro
1
Ri
We first answer the question at which radial position r the VON MISES equivalent
stress will assume a maximum for a given pressure p. This happens at r ¼ Ri
where the nominator in Eq. (11.2.6) assumes its maximum. This is where the
critical value for the material-dependent flow stress ry will be reached first. Thus
from Eq. (11.2.6) it follows for the smallest pressure pmin required for first plastic
flow:
292 11 Introduction to Time-Independent Plasticity Theory
rMises ðp ¼ pmin ; r ¼ Ro Þ ¼ ry
3 ffi 3 ffi 3 ð11:2:7Þ
2 Ri Ro 2 R
) pmin ¼ ry 1 ¼ ry 1 i :
3 Ro Ri 3 Ro
From Eq. (11.2.2)3 the corresponding displacements at the inner and at the outer
radius can be determined:
ffi R
pmin 1 þ m Ro 3 i
umin
hri ¼ u hri ð p ¼ p min ; r ¼ R i Þ ¼ 1 2m þ R 3
Ri E 2 Ri o
1
Ri
ffi
2ry R 3 1þm
¼ ð1 2mÞ o þ Ri ;
3E Ri 2
ð11:2:8Þ
min pmin h 1þm
i R
o
uhri ¼ uhri ðp ¼ pmin ; r ¼ Ro Þ ¼ 1 2m þ R 3
Ro E 2 o
1
Ri
3
ry Ri
¼ ð1 mÞ Ro :
E Ro
Note that for the case of total spherical symmetry the flow condition can
alternatively be written as follows:
ry ¼ rh##i rhrri : ð11:2:9Þ
For a proof we start from Eq. (11.2.1) and write:
1 1 1
rhkki ¼ rhrri þ rh##i þ rhuui ¼ rhrri þ 2rh##i ¼ 3kA: ð11:2:10Þ
3 3 3
Thus the stress deviator in spherical coordinates reads according to Eq. (11.1.2):
2 3
2 0 0
1
Rhiji ¼ rh##i rhrri 4 0 1 0 5; ð11:2:11Þ
3
0 0 1
and because of Eq. (11.1.1) we finally find that:
3 3 2 4 1 1
r2Mises ¼ Rhiji Rhiji ¼ rh##i rhrri þ þ
2 2 9 9 9 ð11:2:12Þ
2 2
¼ rh##i rhrri ¼ ry :
R 3
i
3 r
ry ¼ p 3 ; ð11:2:14Þ
2 Ri
1 Ro
which is identical to Eq. (11.2.6). We now turn to the next problem, namely the
mathematical description of the development and growth of the plastified zone due
to the steady increase of the pressure beyond the critical initial value pmin . For
reasons of symmetry we expect that a plastified spherical shell grows radially
starting from the inside at r ¼ Ri to the position r ¼ q. We would like to calculate
the required pressure pq and the stress distribution within the hollow sphere. To
this end we first note that the sphere remains in a linear elastic state in the region
beyond r ¼ q. The solution shown in Eq. (11.2.1) can be used to compute the
corresponding stresses:
~ 4l B ~ þ 2l B
rhrri ¼ 3kA ; rh##i ¼ rhuui ¼ 3kA ;
r3 r3 ð11:2:15Þ
~r þ B
uhri ¼ A ; 3k ¼ 3k þ 2l; q r Ro :
r2
The pressure along the outer radius is still zero, thus:
B ~ ¼ A :
4l ¼ 3kA ð11:2:16Þ
R3o
By inserting this result into Eq. (11.2.15) for the stresses we obtain:
ffi 3 ffi
Ro Ro 3
rhrri ¼ A 1 ¼ A 1 ;
r r
ffi ð11:2:17Þ
3
1 Ro
rh##i ¼ rhuui ¼ A þ1 :
2 r
The unknown constant A can be determined from the requirement that at the
position r ¼ q plastic flow conditions have just been reached. The situation shows
still perfect spherical symmetry so that the VON MISES flow criterion in the form
shown in Eq. (11.2.9) can be used:
3
3 Ro 3 2 q
ry ¼ rh##i rhrri ¼ A ) A ¼ r y : ð11:2:18Þ
2 q 3 Ro
Note that rhrri is a negative, i.e., compressive stress, whereas rh##i and rhuui are
of tensile nature. The displacements within the elastic region can now also be
obtained from Eq. (11.2.15)3:
~ B A B 1 Ro 3
uhri ¼ A þ 3 r ¼ þ 4l 3 r
r 3k Ro 4l r
ð11:2:20Þ
1 1 Ro 3
¼A þ r;
3k 4l r
2ry q 3 1 þ m Ro 3
uhri ¼ 1 2m þ r; q r Ro ; ð11:2:21Þ
3E Ro 2 r
since:
E E
3k ¼ 3k þ 2l ¼ ; l¼ : ð11:2:22Þ
1 2m 2ð 1 þ m Þ
However, that is not all there is. In order to determine the stresses in the
plastified region Ri r q we argue as follows. In this region we will also have
perfect spherical symmetry. Thus:
rhrri 6¼ 0; rh##i ¼ rhuui 6¼ 0; rhr#i ¼ rhrui ¼ rh#ui ¼ 0 ð11:2:23Þ
and:
rhrri ¼ f ðr Þ; rh##i ¼ rhuui ¼ gðr Þ: ð11:2:24Þ
Recall the balance of momentum in spherical coordinates from Eq. (5.6.4).
Obviously the u and # components are identically satisfied. The r component
reduces to:
orhrri 1
þ 2rhrri rh##i rhuui ¼ 0; ð11:2:25Þ
or r
The expression shown in the parentheses of the right hand side of this differ-
ential equation is equal to the yield stress ry according to the VON MISES flow rule
of Eq. (11.2.9). Thus we can immediately integrate:
rhrri ¼ 2ry ln ðr Þ þ C: ð11:2:27Þ
The constant of integration, C, follows from the continuity relation for the
traction at the interface to the elastic zone at r ¼ q [cf., Eq. (11.2.19)]:
11.2 The Radially Symmetric Solution 295
3 ffi 3
2 q Ro
rhrri r¼q ¼ 2ry ln ðqÞ þ C ¼ ry 1
3 Ro q
ffi 3 ð11:2:28Þ
q 2 q
) rhrri ¼ ry 2 ln þ 1 ; Ri r q:
r 3 Ro
Note that the radial stress is always compressive and becomes even more
negative with decreasing distance r (cf., Fig. 11.2 top, left). By using the VON
MISES flow criterion it follows within the region Ri r q:
ffi 3
q 2 q
rh##i ¼ rhuui ¼ ry þ rhrri ¼ ry 1 2 ln 1 : ð11:2:29Þ
r 3 Ro
Note that the angular stresses are tensile at radial distances r close to q (cf.,
Fig. 11.2 top, right). They increase with increasing distance r. In order to find a
relation for the pressure pq required to initiate plastic flow up to the radius q, we
simply evaluate Eq. (11.2.28) at r ¼ Ri :
ffi 3
q 2 q
rhrri Ri ¼ ry 2 ln þ 1 ¼ pq ð11:2:30Þ
Ri 3 Ro
since due to the continuity of the traction at that position the radial stress must be
equal to the negative pressure. Consequently:
Fig. 11.2 Stress distributions before and after pressure relief Ri/Ro = 0.3, q/Ro = 0.6
296 11 Introduction to Time-Independent Plasticity Theory
ffi 3
q 2 q
pq ¼ ry 2 ln þ 1 : ð11:2:31Þ
Ri 3 Ro
We will now reduce the pressure pq down to zero again. This will lead to a
redistribution of the stresses. During this process equilibrium of forces and con-
tinuity of the traction at the interfaces must be satisfied. We claim that in the end
the stresses are as follows
(a) within the inner plastified region:
ffi 3
2 pq Ri r
rhrri ¼ ry 1 þ 3 ln \0;
3 pmin r Ri
ffi
2 3 r pq 1 Ri 3
rh##i ¼ rhuui ¼ ry þ 3 ln 1þ ; Ri r q;
3 2 Ri pmin 2 r
ð11:2:34Þ
2 q 3 pq Ri 3 R 3
rhrri ¼ ry i ;
3 Ri pmin r R o
ð11:2:35Þ
2 q 3 pq 1 Ri 3 R 3
rh##i ¼ rhuui ¼ ry þ i ; q r Ro :
3 Ri pmin 2 r Ro
11.2 The Radially Symmetric Solution 297
Obviously the contraction of the outer region compresses the inner one:
Fig. 11.2, second row. Compressive residual stresses dominate, which is advan-
tageous since they strengthen the whole structure. This technique is actually used
for the benefit of pressure vessels and known as autofrettage.
In order to verify the results for the residual stresses due to the autof-
rettage process proceed as follows. First specify all continuity conditions at
the interfaces and boundaries. Use the solution shown in Eqs. (11.2.34/
11.2.35) and show that they are all satisfied. Second, analyze the differential
equations for static equilibrium of forces in both regions and prove that they
are identically satisfied by the solution for the stresses. In this context make
use of the perfect spherical symmetry.
Hill [1] claims in his famous textbook on plasticity that the solution can
alternatively be obtained by subtraction of the elastic stress field resulting
from Eq. (11.2.2) for the choice p ¼ pq from the stresses shown in Eqs.
(11.2.32) and (11.2.33). Confirm this statement and explain why it is legit-
imate to calculate the stresses in this manner after pressure relief?
e_ pl _ o/ :
kl ¼ k ð11:3:3Þ
orkl
σ - σ y,0 β σ - σ y,0 γ
σ y,0
dσ dσ
dε el
α
α α
ε - ε el,0 ε pl
ε el,0 ε
ε pl dε pl
ε pl d ε pl (ε -ε el,0)
ry is the current yield stress an must not be confused with the initial one, ry;0 .
The difference is shown in Fig. 11.3. Only for the case of ideal plasticity, where
there is no hardening, both are equal. In addition, we must guarantee that the
so-called consistency condition is satisfied. This is a capricious expression for the
fact that the flow condition is identically satisfied at each incremental loading step
or at each point in quasi-time:
o/ o/
/_ ¼ 0 ) /_ r; ry ¼ r_ ij þ r_ y 0: ð11:3:6Þ
orij ory
In particular, we obtain for the case of VON MISES plasticity, Eq. (11.1.1), and by
observing the definition for the stress deviator shown in Eq. (11.1.2):
o/ o/ oRrs
1
¼ ¼ Rrs dri dsj dni dnj drs ¼ Rij ð11:3:7Þ
orij oRrs orij 3
deformation is dominant. For its proper characterization one first eliminates the
elastic strain from the curve. Thus only the non-linear stress-strain part remains
starting from the initial yield stress ry;0 to an arbitrary (current yield stress) value
r ¼ ry . Consequently we obtain a plot of r ry;0 versus e eel;0 (Fig. 11.3,
center) and we define:
ry;0
E¼ : ð11:3:9Þ
eel;0
Figure 11.3 (center) also shows what happens if we start unloading at a stress
value above the point of initial yield. We then return to zero stress level, basically
along a straight line parallel to the HOOKEAN branch. The abscissa is intersected at a
strain epl 6¼ 0: This is the remaining plastic strain after monotonous loading and the
total strain e is simply the sum of the elastic and of the plastic strain parts
e ¼ eel þ epl , in other words the integrated, one-dimensional form of Eq. (11.3.1).
As indicated in Fig. 11.3 (center) we obtain for a neighboring point an increment
of plastic strain depl . The corresponding elastic (reversible) strain increment is
el
given el;0by de ¼ E dr. If we now subtract in each point of the curve
e e ; r ry;0 the elastic bit E r ry;0 from the ordinate value we can
‘‘renormalize’’ and generate the curve e eel;0 E r ry;0 ; r ry;0 Þ ¼
pl
e ; r ry;0 . For this purpose we make use of Eq. (11.3.9). Thus we obtain the
third viewgraph in Fig. 11.3 (right), which shows stress exclusively as a function
of plastic strain. This allows us to define the (plastic) tangent modulus correctly as
the slope of the function shown in Fig. 11.3 (right). Note that there is a connection
to the slope b of the function shown in Fig. 11.3 (center):
dr r_
ET tanðcÞ ) ET ¼ pl ¼ pl and
de _
e
dr dr depl ET ET E ð11:3:10Þ
tanðbÞ el ¼ ¼ ¼ :
de þ depl dr=depl 1 þ ET =E E þ ET
1 þ dr de el
=
Thus the difference between tanðbÞ and tanðcÞ becomes obsolete if ET =E 1.
If we now observe that:
1 1
Rrs rrs ¼ Rrs Rrs þ rnn drs ¼ Rrs Rrs þ rnn Rrr ¼ Rrs Rrs ð11:3:11Þ
3 3
and assume that the power dissipated due to plastic deformation in the one-
dimensional case is equal to that in 3D (recall that r ry is the currently applied
one-dimensional tension during plastic deformation, i.e., the current yield stress):
r e_ pl ry e_ pl ¼ rij e_ pl
ij ; ð11:3:12Þ
11.3 The PRANDTL-REUSS Equations 301
This relation shows us indirectly that we are dealing with a quasistatic, i.e.,
time-independent theory. A last potential time- (or better rate-) dependence is
contained in the quantity k, _ which we shall identify now. We start from the
consistency condition (11.3.6) and insert successively the newly derived Eqs.
(11.3.3/11.3.7/11.3.8/11.3.14):
o/ o/
2r
0¼ r_ ij þ r_ ¼ Rij r_ ij y r_ ¼ Rij Cijkl e_ ij e_ pl
ij
orij ory 3
2r o/ 4E _ kl k_ 4ET r2 :
y r_ ¼ Rij Cijkl e_ kl k_ k_ T r2y ¼ Rij Cijkl e_ kl kS
3 orkl 9 9 y
ð11:3:15Þ
_
Consequently it follows for the unknown function k:
Rrs Crstu e_ tu Rrs Crstu Rkl e_ tu
k_ ¼ ) e_ pl
kl ¼ : ð11:3:16Þ
Rop Copmn Rmn þ 49ET r2y Rop Copmn Rmn þ 49ET r2y
If we insert this in Eq. (11.3.2) we obtain the following equation, which con-
nects stress and strain rates:
!
Cijrs Rrs Ruv Cuvkl
r_ ij ¼ Cijkl e_ kl : ð11:3:17Þ
Rop Copmn Rmn þ 49ET r2y
Equation (11.3.19) must be solved incrementally during each loading step while
equilibrium of forces, i.e., the incremental static balance of momentum is observed:
or_ ji
¼ 0: ð11:3:20Þ
oxj
Thus we are confronted with a mathematical problem that—in a certain way—
bears a certain similarity to the LAMÉ-NAVIER equations from Sect. 6.2. The
equations can be simplified if we assume isotropic linear-elastic behavior. To this
end we write:
Cijkl ¼ kdij dkl þ l dik djl þ dil djk ð11:3:21Þ
and find (recall that the trace of the deviator vanishes, Rrr ¼ 0, as well as its
symmetry, Rij ¼ Rji ):
Cijrs Rrs Ruv Cuvkl ¼ kdij Rrr Ruv Cuvkl þ l Rij þ Rji lðRkl þ Rlk Þ
¼ 4l2 Rij Rkl ¼ Rop Copmn Rmn ¼ kRoo Rmn þ lðRnm þ Rmn Þ Rmn
4
¼ 2lRmn Rmn ¼ lr2y :
3
ð11:3:22Þ
Thus:
! 0 1
9l2 Rij Rkl 3lR R
e_ kl ¼ @Cijkl A e_ kl :
ij kl
r_ ij ¼ Cijkl ð11:3:23Þ
ð3l þ ET Þr2y 1 þ E3lT r2y
Note that according to VON MISES we may also write for the yield stress [see Eq.
(11.1.1)], where the right hand side is a.k.a. equivalent stress:
rffiffiffiffiffiffiffiffiffiffiffiffiffi
pl 3
ry e ¼ Rij Rij req : ð11:3:24Þ
2
A similar relation can be found for the equivalent plastic strain rate by means
of Eqs. (11.3.7/11.3.10/11.3.14/11.3.13):
2
e_ pl _
e pl
¼ _ 2 o/ o/ ¼ k_ 2 Rkl Rkl ¼ k_ 2 2r2 ¼ 2 r_ r2 ¼ 3 e_ pl 2
k
kl kl y 2 2 y
orkl orkl 3 3 ET r y 2
rffiffiffiffiffiffiffiffiffiffiffi ð11:3:25Þ
2 pl pl
) e_ pl ¼ e_ kl e_ kl :
3
Note that the factor 23 is exactly the inverse of the one shown in Eq. (11.1.1) for
the equivalent plastic strain rate. This expression can easily be integrated w.r.t.
‘‘time:’’
Z Z rffiffiffiffiffiffiffiffiffiffiffi
pl pl 2 pl pl
eeq ¼ e_ dt ¼ e_ kl e_ kl dt: ð11:3:26Þ
3
11.3 The PRANDTL-REUSS Equations 303
Use the PRANDTL-REUSS relation (11.3.16) and show that the trace of the
plastic strain rates vanishes:
tr e_ pl ¼ 0: ð11:3:29Þ
What does this have to do with incompressibility? Specialize the PRANDTL–
REUSS relation to the case of the plastifying hollow sphere subjected to an
internal pressure from Sect. 11.2 and show that within the region Ri r q:
ffi ffi
2 ou_ hri u_ hri 1 ou_ hri u_ hri
e_ pl
hrr i ¼ ; _
e pl
h##i
¼ _
e pl
huui ¼ : ð11:3:30Þ
3 or r 3 or r
304 11 Introduction to Time-Independent Plasticity Theory
To this end recall that it was assumed that the material of the sphere was
ideally plastic:
ET ¼ 0 ð11:3:31Þ
and kinematic relations of the form (6.2.29) must be used but for the total
strain rates. Confirm the general expression shown in Eq. (11.3.28) by
evaluating it with Eq. (11.3.29). Moreover, show that the equivalent plastic
strain for the pressurized hollow sphere is given by:
tðpq Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffi tZðpq Þ rffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z 2 2 2
2 pl pl 2
epl
eq ¼ e_ hkli e_ hkli dt ¼ e_ hrri þ e_ h##i þ e_ pl
pl pl
huui dt
3 3
tðpmin Þ tðpmin Þ
t ðp q Þ
Z ffi
2 ou_ hri u_ hri
¼ epl
eq ¼ dt; Ri r q:
3 or r
tðpmin Þ
ð11:3:32Þ
How can this result be evaluated further? Recall HOOKE’s law for the
isotropic case and show by assuming additive decomposition of the strain
rates according to Eq. (11.3.1) that:
rhiji ¼ keel
hkki dhiji þ 2lehiji
el
h i h i ð11:3:33Þ
¼ k ehkki eplhkki d h iji þ 2l e hiji e pl
hiji :
Calculate the trace of this equation and derive the following differential
equation for the radial component of the displacement:
duhri uhri 1 2m
þ2 ¼ rhrri þ 2rh##i ; Ri r q: ð11:3:34Þ
dr r E
Recall that complete spherical symmetry was assumed and observe the
kinematic relations (6.2.29). Integrate this equation by using the expressions
for the stresses within the plastic region, Eq. (11.2.33), and show by
adjusting the resulting constant of integration to the solution for the radial
displacement in the elastic region from Eq. (11.2.21) that within Ri r q:
ffi ffi 3 3
2ð1 2mÞry 3 1 m q3 q r
uh r i ¼ 1 þ ln r: ð11:3:35Þ
3E 2 1 2m r Ro q
11.4 Would You Like to Know More? 305
References
1. Hill R (1998) The mathematical theory of plasticity, Oxford classic texts in the physical
sciences. Clarendon Press, Oxford
2. Betten J (2001) Kontinuumsmechanik—Elastisches und inelastisches Verhalten isotroper und
anisotroper Stoffe, 2nd edn. Springer, Berlin, Heidelberg
3. Khan AS, Huang S (1995) Continuum theory of plasticity. Wiley, New York, Chichester,
Brisbane, Toronto, Singapore
4. Houlsby GT, Puzrin AM (2006) Principles of hyperplasticity—an approach to plasticity
theories based on thermodynamic principles. Springer, London
5. Maugin GA (1992) The thermomechanics of plasticity and fracture, Cambridge texts in
applied mathematics. Cambridge University Press, Cambridge
Chapter 12
Entropy
Anonymus
Entropy and the Second Law of thermodynamics belong to the concepts of science
with mystical quality. On the one hand side this may be due to the fact that entropy
and the corresponding entropy production are relatively abstract quantities, for
which we have no gut feeling, as in the case of mass, velocity, or even internal
energy and heat flux. On the other hand side the concept of entropy is related to a
very fundamental, if not frightening property of our physical world, namely to its
irreversibility and mortality. Clearly, this attracts very quickly the attention of
philosophers, prophets, esoterics, and other witch doctors.
Of course, we shall not follow their line of reasoning. On the contrary: In this
chapter we will show that entropy is a very useful tool for the rational engineer, a
tool that can be understood and mastered easily, if it is presented within a proper
mathematical setting. Entropy will help us, first, to reduce the number of calori-
metric measurements considerably, which are required to determine the depen-
dence of the internal energy on density and temperature. Second, entropy poses
constraints regarding the possible form and dependence of constitutive equations,
like the heat flux and the stress tensor on the state variables. Third, entropy puts us
in a position to quantify the degree of irreversibility of a physical process.
Note that the mechanical pressure is given by the sum of the thermodynamic
pressure, p, the state variable, so to speak, and of the dynamic pressure, which
covers all irreversible isotropic stress related parts (cf., also Exercise 7.4.2). We
find that:
du oqj oti oti
q ¼ ð p þ pÞ þ Rji : ð12:1:2Þ
dt oxj oxi oxj
Nicolas Léonard Sadi CARNOT was born on June 1, 1796 in Paris where
he also died on August 24, 1832. In 1812 he enrolls at a very young age
as a student at the École Polytechnique. However, in 1814 he quits to
become a military engineer. But now his republican convictions lead to
problems with his army colleagues. Thus in 1819 he tenders his (pre-
liminary) resignation and dedicates his life to science, in particular to the
theoretical foundations of (steam) engines and their maximum effi-
ciency. This culminates in his famous paper Réflexions sur la puissance
motrice du feu et sur les machines propres à développer cette puissance.
By the end of the year 1826 CARNOT reenters the military. However, in
1828 his health forces him to retire for good. In June 1832 he is taken ill
with scarlet fever and he finally dies at age 36 during a cholera epidemic.
If we now observe the balance of mass in Eq. (6.5.7) this result can be rewritten
as follows:
du dt oqj oti oti
q þp ¼ p þ Rji : ð12:1:3Þ
dt dt oxj oxi oxj
We now assume that the specific internal energy depends on only two variables,
as we did previously in Sect. 6.5. However, this time and for reasons which will
become immediately obvious, we choose two ‘‘mechanical’’ quantities as vari-
ables, namely the thermodynamic pressure and the specific volume:
ffi ffi
du ouffiffi dp ouffiffi dt
u ¼ uðp; tÞ ) ¼ þ : ð12:1:4Þ
dt opffit dt otffip dt
Thus the expression in parentheses on the left hand side of Eq. (12.1.3)
becomes:
ffi ffi !
du dt ouffiffi dp ouffiffi dt
þp ¼ ffi þ þp : ð12:1:5Þ
dt dt op t dt otffip dt
The objective is now to write this expression as a single total differential. This
is always possible thanks to a mathematical theorem due to EULER, which is known
as the method of the integrating factor. Recall that EULER considered the following
differential equation:
dy
f ðx; yÞ þ gðx; yÞ ¼ 0; ð12:1:6Þ
dx
310 12 Entropy
where f ðx; yÞ and gðx; yÞ are arbitrary, continuously differentiable functions of the
variables x and y. He now multiplied the equation with a currently unknown
function lðx; yÞ—the so-called integrating factor—, such that:
dy
lðx; yÞf ðx; yÞ þ lðx; yÞgðx; yÞ ¼ 0; ð12:1:7Þ
dx
and required that:
ou ou
¼ lðx; yÞf ðx; yÞ; ¼ lðx; yÞgðx; yÞ: ð12:1:8Þ
ox oy
EULER called the function uðx; yÞ the potential. This was done for good reasons
since then Eq. (12.1.7) can be transformed into the following form:
ou ou
dx þ dy du; ð12:1:9Þ
ox oy
which can immediately be integrated. From a mathematical point-of-view finding
the integrating factor may present a certain problem. In principle we could find it
by solving the following PDE:
olf olg
¼ ; ð12:1:10Þ
oy ox
which follows from Eq. (12.1.8) if SCHWARZ’ theorem is observed. However, it is
impossible to find a general solution to this PDE. On the other hand for a specific
choice of f and g we only need one particular solution lðx; yÞ. Thus special
functional forms of lðx; yÞ are tried, for example:
x
l ¼ lð xÞ; l ¼ lð yÞ; l ¼ lðxyÞ; l ¼ l ; etc: ð12:1:11Þ
y
These are useful if they convert the PDE of Eq. (12.1.10) into an ordinary
differential equation. Obviously this method is successful only if the functions
f ðx; yÞ and gðx; yÞ are explicitly known. However, this is unfortunately not the case
for our physics-based problem (12.1.5)—with the exception of the ideal gas, of
course. We shall now stop our mathematical excursion and apply EULER’s tech-
nique as far as possible to the solution of (12.1.5). To this end we identify:
ffi ffi
ouffi ouffi
x ! p; y ! t; f ! ffiffi ; g ! ffiffi þp: ð12:1:12Þ
op t ot p
We multiply Eq. (12.1.5) by the integrating factor 1=T ¼ 1=T ðp; tÞ and denote
the corresponding potential by the symbol s ¼ sðp; tÞ:
" ffi ffi ! #
ds 1 ouffiffi dp ouffiffi dt
¼ þ þp : ð12:1:13Þ
dt T opffit dt otffip dt
12.1 Entropy as a Balanceable Quantity 311
Exercise 12.1.1: The integrating factor for the case of an ideal gas
Show that all of these choices will lead to contradictions with the exception
of the third one for b ¼ 1. Thus we conclude for mathematical reasons
that:
pt ¼ aT: ð12:1:18Þ
Recall now the thermal equation of state for ideal gases (6.4.1), which was
established empirically by the experiments of BOYLE, MARIOTTE, GAY-LUS-
SAC, and AMONTONS. Conclude that the integrating factor is indeed (absolute)
temperature and the remaining factor a is given by:
R
a¼ : ð12:1:19Þ
M
Edme MARIOTTE was born in Til-Châtel ca. 1620 and died in Paris on
May 12, 1684. He was at least as devout as BOYLE. More specifically, he
was a priest and Prior of St. Martin sous Beaune. He rediscovered
BOYLE’s law and he extended it by noting that, at a constant pressure,
the volume of a gas grows with temperature. BOYLE had not made that
observation or, at least, he did not mention it. In that early age of
natural science most scientists did not limit their attention to a single
field of research and so MARIOTTE was also a keen meteorologist and
physiologist. He discovered the circulation of the Earth’s water supply
and the ‘‘blind spot’’ of the eye.
In view of Eq. (3.7.3) and the balance of mass (3.8.3)1, Eq. (12.1.15) must be
interpreted as the local balance of entropy in regular points:
oqs o qj qj oT p oti oti
þ q s ti þ ¼ 2 þ Rji : ð12:1:20Þ
ot oxj T T oxj T oxi oxj
Alternatively we may integrate Eq. (12.1.15) w.r.t. a material system, observe
Eq. (3.8.16) and GAUSS’ theorem (3.4.2). This will lead to the following global
balance:
ZZZ ZZ
d qj
q s dV ¼ nj dAþ
dt T
V ðtÞ oV ðtÞ
ZZZ ð12:1:21Þ
qj oT p oti oti
2 þ Rji dV:
T oxj T oxi oxj
V ðt Þ
Obviously the first term after the equal sign represents the non-convective flux
of entropy. The second term is the entropy production. There is no volume supply
of entropy because we cannot control the development of the heat flux, of tem-
perature, of the stress deviator, or of the velocity gradient within the body (see
Sect. 3.3 where the difference between a production and a supply was explained).
All of these will simply develop inside the system after initial and boundary
conditions have been specified.
The ‘‘derived’’ entropy balances (12.1.15, 12.1.20 and 12.1.21) resulted basi-
cally from the assumption that the thermodynamic process is governed by two
mechanical state variables, the thermodynamic pressure p and the (specific) vol-
ume t. Two comments are now in order. First, we may use the integrating factor
T ð p; tÞ as a state variable and replace one of the two mechanical variables by a
thermodynamic quantity, namely the temperature T. Thus possible sets of state
variables are either ð p; T Þ or ð t; T Þ. These choices have certain advantages, which
we shall learn to appreciate during the rational setup of measurements required to
determine the internal energy of real materials in the vicinity of thermodynamic
equilibrium. Of course, in order to give a physical meaning to such a substitution,
we must establish rules of measurement for the temperature. In the case of pressure
and volume we did not even mention these since they are easy to define. However,
how to measure temperature is neither a trivial task nor immediately evident.
Second, it is by no means obvious that the specific internal energy depends only
on two state variables. This choice may be sufficient if processes are concerned
that are close to thermodynamic equilibrium. Indeed, the local state of a one-
component material in thermodynamic equilibrium can be described completely
by only two variables, for example volume and pressure. But there are more
complicated materials and processes associated with them than that.
In fact, this simple choice of variables is pertinent to the so-called p dt-ther-
modynamics, which is, in many cases, perfectly sufficient for quantification of the
performance of fast operating engineering machines (in this context also see the
314 12 Entropy
remarks in Sects. 7.2 and 7.5). It is surprising that it works, because during such
processes there will certainly be gradients of pressure, velocity, or temperature,
which might be important for the description of a current state and of the whole
process. After all, note that these processes cannot simply be reversed: They have a
certain direction in time and reversing them is an extreme, sometimes impossible
effort. In other words, they are irreversible. Indeed, there are technically important
processes that cannot be described within the framework of p dt-thermodynamics.
Heat conduction problems are just one example.
It seems obvious that, in non-equilibrium, the internal energy must depend on
additional variables beyond pressure and volume (or temperature). However, the
question is which variables this should be, what kind of dependencies are relevant
and, finally, how can these dependencies be assessed experimentally, in particular
in context with non-mechanical, diva-like quantities, such as temperature. Indeed,
from a mathematical point-of-view it would be possible without major problems to
generalize and extend the aforementioned procedure of the integrating factor to
more than two variables. However, this could easily degenerate into a physically
unmotivated, rigid formalism, totally useless for engineering applications.
Two phenomenological methods have been established in the literature in order
to avoid such problems. The first one is known as Thermodynamics of Irreversible
Processes, or TIP for short. In fact we just presented some of its arguments
following the early paper of ECKART [1]. One of TIP’s essential features is the so-
called hypothesis of local thermodynamic equilibrium: Even if in the considered
processes velocity and temperature gradients, i.e., non-equilibrium phenomena, are
present, it is sufficient to characterize the state of a material particle and the state
functions required for its description (e.g., its specific internal energy) just by
pressure and by specific volume (or alternatively by temperature), at least if we
consider a heat conducting, viscous gas or liquid.1 In a moment we shall see how
the entropy relations (12.1.15 / 12.1.20 / 12.1.21) of TIP can be used to find
constraints regarding the possible mathematical form of constitutive quantities like
the specific internal energy, the heat flux, and the stress (deviator).
Thus in TIP entropy is introduced not for sheer academic pleasure. Rather
entropy provides us a service and helps to reduce the amount of measurements
required to determine constitutive relations. Moreover, as we shall see, it also
allows for a quantitative interpretation of the state of order or rather disorder of a
system and quantifies how much it takes to reverse a given thermodynamic pro-
cesses. This is particularly useful when characterizing technical systems, where
waffling and hand-waving would not do us any good.
1
In order to describe the behavior of solids we have to introduce strain or its trace. Moreover, it
may be advisable to work with other than linear strain measures. However, we shall not go into
details here.
12.1 Entropy as a Balanceable Quantity 315
All the quantities in that equation, the specific entropy, s, the entropy flux
vector, /j , the volume supply of entropy, z, and also the entropy production, r, are
(unknown) constitutive functions of a suitable set of state variables. In other
words, the entropy flux is not a priori given by qj T and the volume supply is not
automatically equal to q r=T (r: specific radiation density). Rather these results
might follow under certain assumptions from exploitation of the entropy principle.
The second important requirement during the exploitation consists of the entropy
production being positive-semidefinite for all thermodynamically admissible
processes:
316 12 Entropy
r 0: ð12:1:23Þ
We may call this the Second Law of thermodynamics. Finally, the third
requirement concerns the continuity of the entropy flux at heat conducting walls at
rest:
/j ej ¼ 0: ð12:1:24Þ
Start from Eqs. (12.1.5 / 12.1.13) and combine them to the so-called
GIBBS’s equation:
T ds ¼ du þ p dt; ð12:1:25Þ
where the specific entropy, s, the specific internal energy, u, and the pressure,
p, are functions of two state variables, namely ð p; tÞ, or ð p; T Þ, or ð t; T Þ,
respectively. In particular, choose the thermal and the caloric equations of
state for the ideal gas according to Eqs. (6.4.1) and (6.5.2), and integrate Eq.
(12.1.25) between a reference state ‘‘0’’ and the current state. Show that the
specific entropy of the ideal gas is given by:
R T R t R T R p
s s0 ¼ f ln þ ln ; s s0 ¼ ðf þ 1Þ ln ln
M T0 M t0 M T0 M p0
R p R t
s s0 ¼ f ln þ ðf þ 1Þ ln ;
M p0 M t0
ð12:1:26Þ
depending on the choice of state variables each version is particularly suited
for describing a given problem.
We have just learned that the concept of entropy will put us in a position to restrict
the possible form of constitutive relations. Before we explore this any further we
shall see that entropy also offers a quantitative measure to characterize the state of
(dis-)order in a systems as well as the degree of (ir-)reversibility of a process. We
shall demonstrate this by means of simple examples.
First, we shall consider the vessel shown in Fig. 12.1, which is divided by a
sliding wall into two volumes of equal size, V. One half is filled with an ideal gas,
12.2 Entropy as a Measure of (Dis-)Order and (Ir-)Reversibility 317
the other one is empty. At the start of the process the gas is characterized by an
initial pressure ps and an initial temperature Ts. According to the thermal equation
of state for ideal gases (6.4.1) we may then write for its specific volume:
R Ts
ts ¼ : ð12:2:1Þ
M ps
We now remove the slide. Turbulent flow will set in and after some time the ideal
gas will come to rest again due to internal friction. This process is obviously
irreversible: The gas will not ‘‘by itself’’ reorganize and occupy its original space. We
want to calculate by how much entropy has grown after this process. The mass of the
gas is constant and therefore we may as well calculate the change of the specific
entropy and make use of Eq. (12.1.26). In principle we can use any of the three
representations. We simply replace symbols referring to the current state by quan-
tities characteristic of the end of the process and the 0-state by the corresponding
quantities at the start. The latter ones are prescribed, the former ones still need to be
determined. To this end we apply the global energy balance (also see the remarks in
Sect. 7.2 and in Exercise 7.2.1) to the total volume and find that in the end:
Te ¼ Ts ; ð12:2:2Þ
i.e., the temperature does not change. We may interpret this result to that effect that
on the one hand side the ideal gas will cool down if its volume is increased. On the
other hand side the internal friction will heat it up. Both effects are equally strong.
This is also known as the JOULE-THOMSON effect in the literature. The ideal gas law
(6.4.1) allows us now to calculate the final pressure:
1
pe ¼ ps : ð12:2:3Þ
2
Thus the specific entropy will increase. Equation (12.1.26)2 tells us by how
much:
R
se ss ¼ f lnð2Þ: ð12:2:4Þ
M
The increase of entropy makes immediate sense due to the apparent irrevers-
ibility of the process. Intuitively we also expect a decrease of the amount of order.
Indeed, the amount of order must decrease because after removal of the slide the
gas has twice as much space than before. This also explains the factor of 2 in Eq.
(12.2.4). We can also interpret the effect probabilistically: BOLTZMANN developed
the following famous formula according to which entropy can be calculated from
318 12 Entropy
S ¼ k lnðW Þ; ð12:2:5Þ
where the probability W is related to the number of possible realizations of system
states.
We will now extend the argument slightly and imagine that the second half of the
vessel is now also filled with an ideal gas of the same type. For simplicity we assume
that the sliding wall is permeable to heat. In other words the temperatures in both
sections will initially be equal and given by Ts. The pressure in the second half differs
by a positive factor a from the other one and is initially equal to aps. We pull the slide,
both gases will mix turbulently due to the difference in pressure. After a while they
will come to rest due to internal friction. As before we ask by how much the state of
order, i.e., entropy, has changed. As before it is intuitively clear that the state of order
must have decreased, i.e., entropy must have increased, because we do not observe
that the gases will rearrange spontaneously into two regions with different pressures.
For the computation it is, first, important to note that the masses in both halves
of the vessel are different: If we apply the ideal gas law to the initial state in both
sections, we conclude that the mass in the section with the pressure a ps must differ
by a factor a from the mass m in the other half. If we apply the global energy
balance to the total chamber we obtain the result (12.2.2) again, i.e., the temper-
ature does not change as a consequence of the mixing. The final pressure is
obtained from the ideal gas law:
1þa
pe ¼ ps : ð12:2:6Þ
2
Note that for a ¼ 0 the previous case is recovered. Equation (12.1.26)3 results
in the following change of entropy if the two different masses are taken into
account:
h i
R 2 2a
DS ¼ Se Ss ¼ m ln þ a ln : ð12:2:7Þ
M 1þa 1þa
It is not immediately obvious that the expression is always positive for all
values 0 a\1: The practical engineer shuns and despises the general mathe-
matical proof and solves the problem graphically: As can be seen in Fig. 12.2 the
12.2 Entropy as a Measure of (Dis-)Order and (Ir-)Reversibility 319
should not shatter our belief in the usefulness and interpretability of entropy. In
order to gain further reassurance we will now tend to two more complex examples.
The first one concerns the calculation of the entropy difference for the heavy
oscillating piston under the influence of gravity from Sect. 7.2. For simplicity we
shall neglect the mass of the gas, mg, the external pressure, p0, and the specific
heat, ce, of the piston in Eq. (7.2.22). Under these circumstances we obtain:
mp g þ ps Af mp g þ pp Af
ze ¼ zs ; Te ¼ Ts : ð12:2:11Þ
ðf þ 1Þmp g ðf þ 1Þpp A
This can be inserted into Eq. (12.1.26)1 to compute the difference between the
specific entropies, which due to the homogeneity of the situation is proportional to
the difference in entropy. We define the following positive factor:
mp g
0x ¼ \1; ð12:2:12Þ
ps A
and find:
h i
xþf xþf1
DS ¼ Se Ss ¼ Nk f ln þ ln : ð12:2:13Þ
fþ1 fþ1x
Again it is hard to see that the difference is always positive for all possible
values of x, independently of the choice of f. As before, we provide a ‘‘proof’’ by
means of a graphical solution, which is shown in Fig. 12.3.
We realize that the curves are always positive and almost coincide for the three
possible choices of f. Moreover, the entropy difference is equal to zero for x ¼ 1:
This makes sense because in this case the piston will not move at all. Consequently
a slight difference between the weight of the piston and the inner pressure will lead
to almost no increase of entropy. This confirms the conclusions from our previous
analyses by means of the adiabatic equation in from Sects. 7.2 and 7.5. Recall that
the adiabatic equation (6.5.29) is a consequence of pdt-thermodynamics, which as
explained above is reversible a priori.
The second example analyzes the growth of entropy after the turbulent mixing
process of two ideal gases separated in two chambers that are initially at different
pressure but the same temperature. The details of the situation are described in
Exercise 7.2.1 and the results are presented in what follows.
Use the results compiled in Eq. (7.2.29) from Exercise 7.2.1, define the
following positive, dimensionless parameters
V2s N2
0x ¼ \1; 0y ¼ \1 ð12:2:14Þ
V1s N1
in combination with Eq. (12.1.26)1 and show that the difference in entropy is
given by:
1þx yð 1 þ xÞ
DS ¼ Se Ss ¼ N1 k ln þ y ln ; ð12:2:15Þ
1þy xð 1 þ yÞ
We now start from the global balance of entropy shown in Eq. (12.1.22), assume
that the entropy flux is given by qj T, and that there is no entropy supply due to
radiation. Then it follows that:
322 12 Entropy
ZZZ ZZ ZZZ
d qj
q s dV þ nj dA ¼ r dV 0: ð12:3:1Þ
dt T
V ðtÞ oV ðtÞ V ðt Þ
First, we consider a system with an adiabatic hull for which we may write by
definition:
qj 0 on oV ðtÞ: ð12:3:2Þ
Then it follows from Eq. (12.3.1) that the entropy can only grow after an
internal (irreversible) process in the system has ended. Note that we do not need to
know the details of the internal process. Rather we may simply write:
ZZZ ZZZ Zte ZZZ
DS ¼ Se Ss ¼ q s dV q s dV ¼ r dVdt 0: ð12:3:3Þ
V ðte Þ V ðts Þ ts V ðtÞ
In the previous section we have already provided a few specific examples for
this kind of situation. Moreover, we have calculated the entropy production,
Zte ZZZ
R¼ r dVdt; ð12:3:4Þ
ts V ðt Þ
explicitly. We now turn to systems that are not adiabatically sealed. However, the
temperature on their surface is assumed as constant and the state of stress on the
surface is given by a pressure, which is constant as well, for the whole duration of
the process:
T ¼ T0 ¼ const: and rij ¼ p0 dij ¼ const: on oV ðtÞ: ð12:3:5Þ
Equation (12.3.1) leads us to conclude that:
ZZZ ZZ ZZZ
d
T0 q s dV qj nj dA ¼ T0 r dV 0: ð12:3:6Þ
dt
V ðt Þ oV ðtÞ V ðtÞ
We eliminate the heat flux with the energy balance (3.9.5), which we may rewrite
for the present case as [in particular observe the relations (7.2.8) and (7.2.11)]:
ZZ
d
U þ Ekin þ Epot þ p0 V ¼ qj nj dA ð12:3:7Þ
dt
oV ðtÞ
with:
ZZZ ZZZ ZZZ
q 2
U¼ q u dV; Ekin ¼ t dV; Epot ¼ qu dV: ð12:3:8Þ
2
V ðtÞ V ðtÞ V ðtÞ
12.3 Properties of the Global Entropy Inequality: The Concept of Availability 323
This is easily confused with the so-called GIBBS free energy (sometimes also
termed free enthalpy), which is defined as follows:
ZZZ
G¼ q ðu Ts þ ptÞ dV: ð12:3:13Þ
V ðt Þ
Note the subtle difference: In here the local temperature and pressure field
inside of the system must be used and not the corresponding values on the surface
of the system, which are assumed not to change during the process. Nevertheless,
chemical engineers often claim in a sloppy manner that the GIBBS free energy in an
open system assumes a minimum.
Frequently chemists also consider closed systems, i.e., sealed reaction vessels
of a fixed volume. If we again neglect kinetic and potential energies during a
process performed with such a system, Eq. (12.3.10) simplifies to:
ZZZ
A¼ q ðu T0 sÞ dV: ð12:3:14Þ
V
Erroneously this expression is often confused with the so-called HELMHOLTZ free
energy, which is defined as follows:
ZZZ
F¼ q ðu TsÞ dV; ð12:3:15Þ
V
324 12 Entropy
and despite its obvious difference in comparison with Eq. (12.3.14) we often hear
chemical engineers say that the HELMHOLTZ free energy of a closed systems must
turn into a minimum.
As we shall realize now the concept of entropy, in combination with the GIBBS
equation, can be used to reduce the amount of calorimetric measurements required
for determination of the specific internal energy as a function of two state variables
considerably. For calorimetric measurements it is most appropriate to include
temperature in the pair of variables. We therefore choose specifically the pair
ð T; tÞ. Indeed, one does not measure u ¼ uðT; tÞ directly. Rather it is determined
from its derivatives, as already indicated by the remarks in Sect. 6.5. We write:
ffi ffi
ou ffi ouffi
u ¼ uðT; tÞ ) duðT; tÞ ¼ ffiffi dT þ ffiffi dt: ð12:4:1Þ
oT t ot T
Thus we will obtain the specific internal energy for viscous, heat-conducting
gases and fluids by (numerical) integration up to an additive constant, if we only
determine the derivatives on the right hand side on the second equation
experimentally:
Z ffi Z ffi
ou ffiffi ouffiffi
uðT; tÞ ¼ dT þ dt þ const: ð12:4:2Þ
oT ffit ot ffiT
In fact, the derivatives can be related to the specific heats from Sect. 6.5. These
are known by measuring the change in temperature after energy in form of heat has
been added to the system in a controlled manner. Indeed, the first derivative in
Eq. (12.4.2) is nothing else but the specific heat at a constant volume, which has
already been introduced in Eq. (6.5.12). We will relate the second derivative with
the specific heats at a constant volume and at a constant pressure in combination
with the pressure, i.e., the thermal equation of state. In order to learn how, we start
from the First Law for ‘‘slow processes’’ according to Eq. (6.5.10) and replace dt
by the thermal equation of state [p ¼ pðT; tÞ ) t ¼ tðp; T Þ], so that:
ffi ffi
ot ffiffi ot ffiffi
dtðp; T Þ ¼ ffi dT þ ffi dp: ð12:4:3Þ
oT p op T
For dp ¼ 0, the left hand side obviously represents the specific heat provided at
a constant pressure, i.e., during isobaric processes. Simple algebraic manipulations
allow us now to write:
ffi ffi ffi
ouffiffi ot ffiffi ouffiffi cp ct
cp ¼ ct þ þp ) ¼ ot ffiffi p: ð12:4:5Þ
otffiT oT ffip ot ffiT oT p
This equation shows explicitly that the internal energy can indeed be determined if
only the thermal equation of state and the specific heats at a constant volume and at a
constant pressure are known. Note that the specific heats must be known for every set
of data t, T. In other words, until now there is no other way but measuring the specific
heats for every pair by means of calorimetry applied to every gas or fluid. This is a
substantial effort and from the experimental point-of-view by no means trivial,
because we have to pay attention that the heat is really supplied to the substance of
interest and does not simply vanish in the container, the environment, etc.
However, if we now make use of the concept of entropy in combination with
the GIBBS equation, the amount of measurements required can be reduced dra-
matically. We start from Eq. (12.1.25) and write:
1
dsðT; tÞ ¼ ½duðT; tÞ þ p dt )
ffi ffi
ffiT ffi ð12:4:7Þ
os ffiffi osffiffi 1 ou ffiffi ouffiffi
dT þ dt ¼ dT þ þp dt :
oT ffit otffiT T oT ffit otffiT
By comparing the corresponding terms on both sides we find that:
ffi ffi ffi ffi
os ffiffi 1 ou ffiffi osffiffi 1 ouffiffi
¼ ; ¼ þp : ð12:4:8Þ
oT ffit T oT ffit otffiT T ot ffiT
We differentiate the first expression w.r.t. t and the second one w.r.t. T.
According to SCHWARZ’ theorem for continuously differentiable functions the
sequence of both derivatives does not matter. We assume that entropy and internal
energy do have that property and conclude that:
ffi ffi
ouffiffi op ffiffi
¼ p þ T ffi : ð12:4:9Þ
ot ffiT oT t
In view of Eq. (12.4.5) we conclude that it is no longer required to know and
measure both
ffi specific heats for all specific volumes at a constant temperature,
because ouffi can already be determined completely from the thermal equation of
ot T
state p ¼ pðt; T Þ; which we assume to be known.
326 12 Entropy
However, there is more to be concluded from the concept of entropy and from
the GIBBS equation. If we differentiate Eq. (12.4.9) by T we find that:
ffi ffi Z ffi
o2 u o2 p ffi oct ffiffi o2 p ffiffi
¼ T 2 ffiffi ) c ðt; T Þ ¼ T dt þ f ðT Þ: ð12:4:10Þ
oT t ot ffiT oT 2 ffit
t
oTot
This shows that the dependence of the specific heat at a constant volume, ct , of
the specific volume results by integration from the known thermal equation of
state. Thus we conclude that it is sufficient to measure ct at one specific volume t
and all temperatures T in order to fix the last unknown function f ðT Þ, which
depends only of the temperature, T.
Finally in this section we turn to the right hand side of Eq. (12.1.15). In
particular we will now examine the entropy production, r, for which we may write
according to Eq. (12.1.23):
qj oT oti oti
Tr ¼ p þ Rji 0: ð12:4:11Þ
T oxj oxi oxj
The requirement of the positive-semi-definite of the inequality is guaranteed
within the framework of TIP by relating so-called fluxes (in our case the heat flux,
the dynamic pressure, and the stress deviator) in a linear manner to the so-called
driving forces (in our case the temperature gradient, the divergence of velocity,
and the deviatoric parts of the velocity gradients):
oT oti oti otj 2 otk
qj ¼ j ; p ¼ k ; Rji ¼ l þ dij : ð12:4:12Þ
oxj oxi oxj oxi 3 oxk
In the last relation it has been guaranteed that the stress deviator is symmetric
and trace free. We have already encountered the heat conduction coefficient, j, the
(quasi) bulk viscosity, k, and the shear viscosity, l, in Eqs. (6.6.1) and (6.3.1).
According to the entropy principle they must all be positive parameters that
potentially depend on temperature. In summary, the entropy principle puts us in a
position to reduce the possible form of constitutive relations considerably.
On the other hand Eq. (12.4.11) can also be used to find out how large the local
entropy production is, i.e., to determine the intensity of local irreversibility. Of
course, for this purpose we must know the constitutive equations for the heat flux
and for the stress deviator as well as the temperature and the velocity field for the
corresponding system as functions of time and space. The latter may be hard to
achieve, in particular, if turbulent flow is concerned, as we discussed in the
Examples of Sects. 12.2 and 12.5. Interestingly it was possible to calculate the
total entropy production, i.e., the dissipation integrated w.r.t. space and time in
closed form [see Eq. (12.3.4)] just from the initial and final state of the system.
Both were homogeneous states of equilibrium. Specific results are compiled in
Eqs. (12.2.4, 12.2.7, 12.2.9 / 12.2.13).
However, for stationary processes it is sometimes also possible to compute r in
closed form. An example of this is provided by the stationary parallel flow
12.4 Reduction of the Constitutive Equations for a Viscous Heat-Conducting Fluid 327
between plates from Exercise 7.4.1. Without anticipating the proof required in the
exercise we note that velocity field for laminar flow in Cartesian coordinates is
given by:
x
2
ti ¼ V ; 0; 0 : ð12:4:13Þ
h
Then the stress deviator and the velocity gradient can be calculated from the
NAVIER–STOKES constitutive relation:
0 1 0 1
0 lVh 0 0 Vh 0
@ A ot i @
Rji ¼ l h 0 0 ;
V
¼ 0 0 0 A: ð12:4:14Þ
oxj
0 0 0 0 0 0
This allows us already to determine the mechanical part of the entropy
production:
2
oti V
Rji ¼l : ð12:4:15Þ
oxj h
In order to obtain the thermal one we need to turn to the heat conduction
equation which follows from the First Law (7.6.1) and solve it for the present case.
There is no radiation r, the heat flux is given by FOURIER’s law, and the time
derivative of the specific internal energy vanishes completely because of sta-
tionarity and the ansatz (7.4.7) for the velocity:
du ou ou
¼ þ ti ¼ 0: ð12:4:16Þ
dt ot oxi
Since the internal energy depends on the density and on the temperature it can
only be a function of height, x2 . However, t2 is equal to zero and thus:
2
d2 T V
j 2 ¼ l : ð12:4:17Þ
dx2 h
Obviously isothermal conditions contradict this differential equation. We
assume that the upper as well as the lower plate are kept at a constant temperature
level, T0. It is easily verified that the corresponding solution for the temperature
must be:
l x2 x2
T ¼ T0 þ V 2 1 : ð12:4:18Þ
2j h h
Now we are in a position to determine the local entropy production r from Eq.
(12.4.11):
" #
l V2 2x2 2 V 2
Tr ¼ 1 þ1 l 0: ð12:4:19Þ
4j T h h
328 12 Entropy
Note that in contrast to the mechanical bit the thermal part of the entropy
production depends on position. In order to find out at which height the local
entropy production assumes a maximum, it is advisable to introduce the following
parameters:
x2 l V2 l V 2
0 x ¼ 1; 0 a ¼ \1; r0 ¼ : ð12:4:20Þ
h 2j T0 T0 h
Thus we obtain a dimensionless temperature:
T ð xÞ
T ðxÞ ¼ 1 þ axð1 xÞ; ð12:4:21Þ
T0
and a dimensionless entropy production:
rð xÞ 1 þ a2 þ axðx 1Þ
ðxÞ
r ¼ : ð12:4:22Þ
r0 ½1 þ axð1 xÞ2
Figure 12.5 shows the latter function vs. x for a ¼ 0:5; 1:5; and 10. It is
obviously symmetric, as it should, due to the choice of the same temperatures at
the lower and at the upper plate. It assumes a minimum in the center, and it tends
to maximum values when we get closer to the plates. This seems reasonable
because this is where the temperature gradient is particularly strong, whereas in the
center it is equal to zero. The greater a the more the entropy production differs
from the constant reference value r0. This is also easy to explain since increasing
values of a reinforce the impact of the plate temperatures, which homogenizes
entropy production.
It is interesting to note that the factor a introduces to the entropy production two
material parameters that one would intuitively expect behind dissipation, the shear
viscosity and the heat conduction parameter. However, note that although the
constitutive relations for the viscous stresses and for the heat flux are both linear,
the corresponding parameters add a highly non-linear touch to entropy production.
A first introduction to the concept of entropy from the standpoint of the CARNOT
cycle and engineering thermodynamics of discrete systems can, for example, be
found in [6], Chapters 5 and 6, [7], Chapter 4 or [8], Chapter 4. The continuum
theoretical aspects of the entropy principle are outlined in [4], Chapter 4, and in
[5], Chapter 5. The latter book can also be consulted in order to learn to appreciate
CARATHÉODORY’s contributions to entropy (Section 5.1.3.3), and to hear more about
the notion of availability in a wider context (Section 7.2.2). The traditional ways of
the chemical engineers and the physical chemists, who think in terms of mini-
mizing HELMHOLTZ and GIBBS functions, are outlined, for example, in [9], Chapter
3. The monograph of [10] presents a very detailed exposition of nearly all aspects
of the notion of entropy, from a continuum or theoretical point-of-view, for gases,
fluids, and solids. This book also provides further information on availability:
Sections 7.5 / 7.6.
The principles of statistical mechanics and the entropy of the so-called great
canonical ensemble are explained in the books by Münster [2] and by Tolman [11].
The latter monograph also quickly introduces aspects of quantum mechanics. The
kinetic theory of gases and the corresponding notion of entropy, as introduced by
MAXWELL and, in particular, BOLTZMANN, can be studied best by consulting the
bible on this topic, namely the book by Chapman and Cowling [12] including
BOLTZMANN’s famous H-Theorem. Another valuable source in the same context is
the book by Becker [13], in particular Chapter 2.
The classic concepts of TIP are presented in the books by de Groot [14] or de
Groot and Mazur [15]. Becker [13] gives a first introduction: Chapter 7. The
methods of so-called Rational Thermodynamics, i.e., methods that go beyond that
of TIP, can be found in the book by Truesdell [16].
References
1. Eckart C (1940) The thermodynamics of irreversible processes I. The simple fluid. Phys Rev
58(3):267–269
2. Münster A (1969) Statistical thermodynamics, vol 1. Springer, Berlin, First English Edition
3. Haupt P (2002) Continuum mechanics and theory of materials, 2nd edn. Springer, Berlin
4. Müller I (1973) Thermodynamik. Die Grundlagen der Materialtheorie. Bertelsmann
Universitätsverlag, Düsseldorf
5. Müller I (1985) Thermodynamics. Pitman Advanced Publishing Program, Boston
6. Çengel YA, Boles MA (1998) Thermodynamics: an engineering approach, 6th edn. McGraw
Hill, Boston
7. Müller I (1994) Grundzüge der Thermodynamik mit historischen Anmerkungen, 1st edn.
Springer, Berlin
8. Müller I, Müller WH (2009) Fundamentals of thermodynamics and applications. Springer,
Berlin
9. Moore WJ (1963) Physical chemistry, 4th edn. Longmans Green and Co Ltd, London
10. Müller I, Weiss W (2005) Entropy and energy—A universal competition. Springer, Berlin
330 12 Entropy
11. Tolman RC (1979) The principles of statistical mechanics. Reprint of the original edition of
1938. Dover Publications, Inc., New York
12. Chapman S, Cowling TG (1939) The mathematical theory of non-uniform gases. Cambridge
at the University Press, Cambridge
13. Becker R, Leibfried G (eds) (1967) Theory of heat, 2nd edn. Springer, Berlin
14. de Groot SR (1960) Thermodynamik irreversibler Prozesse. BI Hochschultaschenbücher 18/
18a. Bibliographisches Institut, Mannheim
15. de Groot SR, Mazur P (1984) Non-equilibrium thermodynamics. Dover Publications Inc.,
New York
16. Truesdell C (1969) Rational thermodynamics. McGraw-Hill, New York
Chapter 13
Fundamentals of Electromagnetic Field
Theory
I am an expert of electricity.
My father occupied the chair of applied electricity
at the state prison.
W. C. FIELDS
very simple constitutive equations or study preferably problems for which the
motion of matter is of no importance. Such cases are perfectly covered by MAX-
WELL’s equations in a form that can be found, for example, in the renowned book
by Becker [1], §53 on pg. 216, Eq. (53.2). Becker says: ‘‘… and thus we obtain the
four fundamental equations, of remarkably symmetrical structure1:
1 oD 4p g
ðIÞ curl H ¼ þ ; ðIIÞ div D ¼ 4pq;
c ot c ð13:1:1Þ
1 oB
ðIIIÞ curl E ¼ ; ðIVÞ div B ¼ 0
c ot
as the final form of the Maxwell equations for media at rest.’’ It is noteworthy that
these relations hold for matter at rest. Of course the obvious question arises how
the general equations read that hold if matter is moving. In what follows we
attempt to clarify these issues. Moreover, we will emphasize and explain why two
‘‘electric’’ fields, i.e., Eðx; tÞ and Dðx; tÞ, and two ‘‘magnetic’’ fields, i.e., Bðx; tÞ
and Hðx; tÞ, must be distinguish and are required in electromagnetism. In other
words we will put a strong emphasis on independent measurement instructions for
these four fields and all the other ones.
At this point it should already be mentioned that we will define and use the
electric field Eðx; tÞ and the magnetic induction Bðx; tÞ as primary fields from the
very beginning on. However, the fields Dðx; tÞ and Hðx; tÞ of Eq. (13.1.1), which
are known in the pertinent literature on electrodynamics as electric displacement
and magnetic field, will not be used before we talk extensively about the con-
servation of charges and currents. More appropriately these fields should be
referred to as charge and current potential in matter. We will denote them by the
symbols D und H in order to distinguish them from the general charge and current
potentials D and H, respectively.
1
Becker uses the symbols q and g for the (true) electric charge density and the (true) current
density of free charge carriers. Both symbols have been used in this book before but in a different
context. In what follows we will use the symbols qf and jf instead. Also note that Becker does not
use SI units, which explains the factor 4p and the speed of light symbol, c, in his equations.
13.1 Preliminary Remarks 333
Hendrik Antoon LORENTZ was born on July 18, 1853 in Arnheim and died
on February 4, 1928 in Haarlem. In 1870 he turned to the University of
Leiden to study mathematics and physics. After that he returned to his
hometown and became a teacher in 1871. This gave him ample time to
work on his dissertation, which he finished in 1875. In 1878 he finally
became a professor for theoretical physics at the University of Leiden,
where he stayed for the rest of his life. His scientific achievements are
mostly based in theoretical physics and concern, in particular, the elec-
tromagnetism of light and of matter. For his explanation of the ZEEMAN-
effect he was awarded the Nobel prize in 1902 together with ZEEMAN.
We shall later derive the connection between Dðx; tÞ and D ðx; tÞ on the one
hand side and Hðx; tÞ and H ðx; tÞ on the other. This will allow for a conversion of
one quantity into the other. Clearly, this must be possible: In the end all notations
must be equivalent since there is only one theory of electromagnetism.
Moreover, we will learn that there is a connection between the fields Eðx; tÞ and
Dðx; tÞ on the one hand and Bðx; tÞ and Hðx; tÞ on the other, the so-called MAXWELL-
LORENTZ aether relations. Many scientists have worked on the setup of the mathe-
matical framework preferred in this book, and we shall mention them in due course.
Finally note that we will frequently write Ei , Di , Bi , Hi , etc. By these symbols
we refer to the components of the corresponding electromagnetic vector fields in a
334 13 Fundamentals of Electromagnetic Field Theory
Cartesian frame at rest. This in turn indicates that we must also speak about the
transformation properties of all of these fields as well as of MAXWELL’s equations
during change of observers.
Consider the situation illustrated in Fig. 13.1. Similarly to the general balance for a
volume field density w, integrated w.r.t. a material volume without a singular
V
surface as shown in Eq. (3.3.7), we present the following general balance for a
vector flux density ci , defined for an open material surface, S,2 which is not
dissected by a singular line, L:
ZZ I ZZ
d
ci ni dS ¼ /i si dl þ ðpi þ si Þ ni dS: ð13:2:1Þ
dt
S oS S
The vector /i denotes the flux of ci across the closed periphery (‘‘line’’) oS in
complete analogy to the nomenclature established in Sect. 3.3. Moreover pi and si
are vector densities of production and supply (per unit area) of the quantity ci
defined on the surface S.
We will now anticipate some results from the following sections and argue in a
formal manner without providing precise definitions for measuring the electro-
magnetic fields Ei and Bi : Theoretical investigations, observations of natural
phenomena in combination with well-defined lab experiments by diligent scientists
and discoverers like FARADAY, MAXWELL, or LORENTZ have shown that in the case of
the magnetic flux we must relate quantities of Eq. (13.2.1) as follows:
ci ! Bi ; /i ! Ei þ ðt BÞi ; pi ! 0; si ! 0: ð13:2:2Þ
2
For didactic reasons, which will become clearer in Sect. 13.4, we denote the open surface by
the symbol S and not like in Sect. 3.3 by the generic symbol for surfaces, A.
13.2 The Conservation Law for the Magnetic Flux 335
∂S
The magnetic flux is a conserved quantity, which explains the vanishing pro-
duction density. Moreover, the combination
E i ¼ Ei þ ð t B Þ i ð13:2:3Þ
is also known as the electromotive intensity. In honor of LORENTZ the part ðt BÞi
is also called LORENTZ force density. Thus we arrive at the following global con-
servation law for the magnetic flux, a.k.a. FARADAY’s law of induction:
ZZ I
d
Bi ni dS ¼ Ei þ ðt BÞi si dl: ð13:2:4Þ
dt
S oS
Therefore the total magnetic flux through a closed surface is conserved in time
and must be equal to a constant. However, this constant must be zero, because at
some point the magnetic field had to be switched on, and before that the magnetic
flux was equal to zero. Thus we conclude that:
ZZ
Bi ni dA ¼ 0 ð13:2:6Þ
oV
and rewrite this result by means of GAUSS’ theorem into a volume integral. By
doing so we tacitly assume that the magnetic flux densities shows no disconti-
nuities within the volume V:
ZZZ
oBi
dV ¼ 0: ð13:2:7Þ
oxi
V
336 13 Fundamentals of Electromagnetic Field Theory
The usual arguments lead to the corresponding local equation in regular points:
oBi
¼ 0: ð13:2:8Þ
oxi
This is already the fourth of MAXWELL’s equations (13.1.1). We now return to
FARADAY’s law of induction in Eq. (13.2.4). In a first step we transform the left side
of the equations by means of a transport theorem for open surfaces which will be
proven in the Exercise 13.2.2:
ZZ ZZ ffi I
d oBi oBk
Bi ni dS ¼ þ ti ni dS þ ðB tÞi si dl: ð13:2:9Þ
dt ot oxk
S S oS
The second part in the surface integral vanishes because of the previously
derived MAXWELL equation (13.2.8), if we assume continuity of the magnetic flux
density. Moreover the line integral cancels with the second term on the right side
of Eq. (13.2.4) because of the anti-symmetry B t ¼ t B of the vector
product, so that we obtain:
ZZ I
oBi
ni dS þ Ei si dl ¼ 0: ð13:2:10Þ
ot
S oS
Hence we obtain another local relation, namely the third of MAXWELL’s equa-
tion from the set (13.1.1):
oBi
þ ðr EÞi ¼ 0: ð13:2:12Þ
ot
dS
x3 dx
x2
∂S
x1
ZZ ZZ ZZ
d
ci ðx; tÞ dSi ¼ ½ci ðxðX; tÞ; tÞ dSi þ ci ðx; tÞ ½dSi ðxðX; tÞ; tÞ
dt
SðtÞ S S
ð13:2:15Þ
by using the directed surface element in Lagrangian description:
dSi ¼ ni dS; x ¼ xðX; tÞ: ð13:2:16Þ
In a second step show that:
ffi
otj otm
½dSi ðxðX; tÞ; tÞ ¼ þ dij dSj : ð13:2:17Þ
oxi oxm
To this end express the directed surface element by a vector product
ð1Þ ð2Þ
between two infinitesimal co-moving line elements dx, dx:
ð1Þ ð2Þ
ð1Þ ð2Þ
dAi ¼ d x d x ¼ ijk d x j d x k ; ð13:2:18Þ
i
Exercise 13.2.3: Extension of the transport theorem for vector surface flux
densities including singular lines
Now prove the following extension of the transport theorem shown in Eq.
(13.2.14) if the open surface contains a singular line, L:
ZZ ZZ ffi
d oci oc
ci ni dA ¼ þ ti k ni dA
dt ot oxk
S [Sþ S [Sþ
Z Z ð13:2:21Þ
þ ðc tÞi si dl ð½½c wÞi ti dl:
l [lþ L
13.2 The Conservation Law for the Magnetic Flux 339
L
S- e
l- t S+
l+
A
Fig. 13.3 A singular surface intersecting with an open surface thus creating a singular line
To this end observe the notation shown in Fig. 13.3 and start the proof for
an open surface S ¼ S [ Sþ [ L by application of Eq. (13.2.14) to both
(regular) surfaces S . Assume that the singular surface A moves indepen-
dently of the material points of S with a velocity w. Explain in detail the
presence of the jump term on L in Eq. (13.2.21). Why is the jump propor-
tional only to w and not to t or w t (say)?
Finally in this section it should be pointed out explicitly that S and oV can be
material structures, moving along with matter dispersed by electromagnetic fields.
Then the quantity t in the transport theorem (13.2.14) is nothing else but the local
velocity of the material points. Interestingly the two local MAXWELL equations
(13.2.8/13.2.12) are independent of this motion and, consequently, the fields Ei and
Bi are also measurable quantities with respect to a co-moving observer. In what
follows we will look into this more deeply.
Just like mass the electric charge is another fundamental, so-called primitive
quantity, which cannot be reduced any further or related to other even more
fundamental entities. Introducing this notion is useful when explaining and
quantifying certain natural phenomena. This does not mean that we have truly
explained them or provided the ultimate understanding. All we do is to define a
rational frame, which is as far-reaching as possible, self-contained, free of con-
tradictions, mathematically expressible, and capable of predictions which, in turn,
can be verified by experiments.
340 13 Fundamentals of Electromagnetic Field Theory
In principle we can measure the amount and the direction of an electric field
generated, for example, by the Styrofoam ball, by placing a test change of known
strength, Q, at a certain point in space and measuring the amount and the direction
of the resulting force F. We use this information to compute the following
expression:
F
E¼ : ð13:3:1Þ
Q
We now turn to a second electric phenomenon and consider an iron magnet at
rest. In its vicinity we scatter iron filing chips and observe how these start rear-
ranging. We interpret this behavior by saying that they follow the orientation of the
field lines of the magnetic flux density, B. We now investigate the behavior of an
electron beam carrying a certain electric charge, Q, in the vicinity of a magnet. In
other words a ‘‘swarm’’ of test charges is passing by at a certain speed. We observe
that the beam does not follow a straight line. Rather it will curve and the electrons
are deflected along circles away from their original straight path. The cause for this
deflection must be a force and, by observation, we conclude that the greater the
speed of the electrons, the greater the curvature and, thus, the greater the force.
13.3 Electric Charges, Currents, Electric Field Density 341
Moreover, the force is obviously perpendicular to the velocity vector. And, finally,
the force vector is perpendicular to the magnetic field lines indicated by the filing
chips. Now we use all these facts and derive a rule for measuring the magnetic flux
density, i.e., the amount and the direction of the vector B from the following
measurable quantities, namely force vector, F, amount of moving electric charge,
Q, and their velocity vector, t:
F
¼ t B: ð13:3:2Þ
Q
If an electric field E and the magnetic induction B act simultaneously, the total
force per unit charge is given by:
F
¼ E þ t B E: ð13:3:3Þ
Q
We have already used this combined quantity in Eq. (13.2.3) and interpreted it
as the flux of magnetic induction across the periphery of the surface. Of course, our
tests are more or less idealized gedankenexperiments. However, they are helpful if
we want to get used to electromagnetic fields, for which we have developed no
intuitive feel during our biological evolution unlike mechanical quantities, such as
force or mass. Clearly, in technical practice E and B are not measured that way.
oU
Ei ¼ : ð13:3:6Þ
oxi
The unit of the electric potential is Volt (V). Show by using the other
results from this exercise that:
J J V Js Vs
dim ½U ¼ ¼ V, dim ½E ¼ ¼ ; dim ½B ¼ ¼ : ð13:3:7Þ
C Cm m Cm2 m2
Recall that an electric current, I, is measured in units of Ampere (A). In terms
of physics it is nothing else but a flow of charge per unit time. Conclude that:
C J V J
dim ½I ¼ ¼ A, dim ½E ¼ ¼ ; dim ½B ¼ : ð13:3:8Þ
s Cm m Am2
The second set of MAXWELL’s equations are based on the conservation of charge
within a material volume divided into two regions, oV þ and oV , by a singular
surface, A (cf., Fig. 13.4). Within V þ and V the density of charge (a scalar per
unit volume like mass density) is denoted by q (in C=m3 ). On the singular surface
the density of charge is called q (per unit area in C=m2 ). Thus in the nomenclature
A
of Eq. (3.3.6) the total charge is given by:
ZZZ ZZ
Q¼ qðx; tÞ dV þ qðx; tÞ dS ð13:4:1Þ
A
V þ [V A
The total electric charge will change in time when electric currents leave the
control volume through the surfaces Aþ [ A [ oA. We describe these directed
quantities by non-convective current density vectors j (in C=ðm2 sÞ) and j (in
L
C=ðmsÞ), which are distributed across the surface and the line, respectively. Thus
we arrive at the following conservation law for charge in form of a conserved
balance:
2 3
ZZZ ZZ
d d
Q¼J, 4 qðx; tÞ dV þ qðx; tÞ dS5
dt dt A
V þ [V A ð13:4:2Þ
ZZ I
¼ ji ni dA j i mi dl:
L
Aþ [A oA
13.4 Conservation of Total Charge 343
+
e A
−
A
V
The minus signs on the right side are pure convention. However, they can be
motivated as follows: The total charge within a material volume will decrease if
charge is removed from the inside by currents crossing the surface.
Also note that there is no (volumetric) supply or production of charge. This
statement is based on experience. In mathematical terms it means (also see the
corresponding remarks in Sect. 3.3) that:
p ¼ 0; p ¼ 0; s ¼ 0; s ¼ 0: ð13:4:3Þ
V A V A
S− e S+
∂S
A
344 13 Fundamentals of Electromagnetic Field Theory
ZZ I ZZ Z
d
Di ni dS ¼ Hi þ ðD tÞi si dl ji ni dS j i mi dl: ð13:4:5Þ
dt L
S oS Sþ [S L
The quantities D and H are also frequently termed dielectric displacement and
magnetic field. Note that according to Eq. (13.4.2) the charge Q is described by
two scalar densities w.r.t. the volume or the surface. It was balanced over a
material volume V ¼ V þ [ V [ A. The fields of the volumetric charge density, q,
may jump, i.e., be discontinuous across the (open) singular surface A. However,
the current J has been related to vector current densities per unit area or per unit
length. The vector fields for the current per unit area behave discontinuously when
crossing from A to Aþ , i.e., oA.
Analogous remarks hold in context with Eq. (13.4.5) and Fig. 13.5. Moreover,
note that the open singular line L represents a part of the periphery oA. If we now
extend the open surfaces Sþ and S to cover Aþ and A they form a closed volume
oV together with the line L ! oA, which is then also closed. We will use this
argument quite soon. Finally recall that the singular lines as well as the singular
surface may have their own velocity field, w, independently of the velocity of the
material particles. In other words they are not necessarily material objects.
Equation (13.4.5) can also be interpreted asRRfollows: The temporal change of
the flow resulting from the charge potential, D Di ni dS, over the open surface S
H S
is given by flux of this quantity, oS Hi þ ðD tÞi si dl across the hull oS as well
RR R
as by supplies from current densities ji ni dA and j i mi dl, over the open surface
Sþ [S L L
S ¼ Sþ [ S [ L, respectively.
As previously indicated we can also specialize
H the second
equation to the case
of a closed surface. Then the integral oS Hi þ ðD tÞi si dl vanishes because
oS ! 0. Moreover we have:
ZZ ZZ ZZ ZZ Z I
! ; ! ; ! ; ð13:4:6Þ
S oV Sþ [S oV þ [oV L oA
Hans Christian ØRSTED was born on August 14, 1777 in Rudkøbing and
died on March 9, 1851 in Copenhagen. At the age of twelve he assisted
already in the pharmacy of his father. This brought him in close contact
with the sciences and in particular with chemistry. Consequently, he
studied chemistry and pharmacy at the University of Copenhagen where
he also obtained a professorship for chemistry and physics in 1806.
ØRSTED discovered various things: In 1819 he was the first to isolate
piperidine, an organic solvent. Then in 1820 he discovered the magnetic
effect of electric currents on compass needles. In 1825 he managed to extract aluminum for the
first time in the history of mankind. He was also an active free mason and a very good friend of
the fairy tale writer Hans Christian ANDERSEN.
13.4 Conservation of Total Charge 345
In view of Eq. (13.4.4) we must also conclude that the balance of charge, Eq.
(13.4.2), is identically satisfied. Equations (13.4.4/13.4.5) represent the first and the
second of MAXWELL’s equations in integral form. In terms of physics they must be
interpreted as global conservation laws for electric charge and current. In the
pertinent literature on electrodynamics Eq. (13.4.4) is also known as COULOMB’s law
(in words, the electric charge is the source of the dielectric displacement) and Eq.
(13.4.5) as ØRSTED-AMPÈRE’s law: Around 1820 ØRSTED and AMPÈRE had discovered
experimentally that an electric current (more precisely the densities j and j ;
L
respectively) is always accompanied by a magnetic field (more precisely the field
H). The corresponding rules of measurement will be investigated in the next section.
At this point we should already ask why the fields D and H were introduced to
begin with and why the conservation law shown in Eq. (13.4.2) was not sufficient for
further analysis. The reason is a practical one: It is possible to connect the fields D and
E as well as the fields H and B by simple algebraic relations, at least in suitable frames
of reference, the so-called LORENTZ systems. In contrast to that the fields q and j are
connected to the fields E and B by complex differential relations. We will get back to
that in context with the so-called MAXWELL-LORENTZ aether relations.
Finally we will derive the local MAXWELL equations in regular points from Eqs.
(13.4.4/13.4.5). If we assume that there is no singular surface intersecting with the
volume V, we can apply GAUSS’ theorem in its usual form rewrite Eq. (13.4.4)
immediately:
ZZZ ZZZ
oDi oDi
dV ¼ qðx; tÞ dV ) ¼ q: ð13:4:8Þ
oxi oxi
V V
André Marie AMPÈRE was born on January 20, 1775 in Polémieux near
Lyon and died on June 10, 1836 in Marseille. Rumor has it that he did not
attend school but got his education in a purely autodidactic manner. From
1799 to 1801 he served as a mathematics teacher at the École Centrale in
Lyon. In 1802 he published a mathematical paper on game theory,
another note on theoretical mechanics, and a treatise on partial differ-
ential equations. The latter got him a membership in the French Academy
of Sciences. However, his interest in mathematics tired down and AMPÈRE
turned to science instead. After a short fling with chemistry he started in
1820 to perform physical experiments with live wires and investigated
their effects on magnetic needles, probably inspired by ØRSTED’s experiments. He came up with
the hypothesis that magnetism of any kind has its origin in electric currents and that, in fact,
electric currents generate magnetic fields. As a final consequence he had to assume that the
molecules of every magnet produced a small circular current.
If there is a singular surface, A, the pill box argument from Sect. 3.7 applies and
we conclude that in a point of this surface characterized by the unit normal e (cf.,
Fig. 13.4) the following relation holds:
346 13 Fundamentals of Electromagnetic Field Theory
½½Di ei ¼ q : ð13:4:9Þ
A
Moreover, for regular open surfaces the following simplified version of Eq.
(13.4.5) results:
ZZ I ZZ
d
Di ni dA ¼ Hi þ ðD tÞi si dl ji ni dA: ð13:4:10Þ
dt
S oS S
By application of the transport theorems of Eq. (13.2.14) we find for the left
side:
ZZ ZZ ffi I
d oDi oDk
Di ni dA ¼ þ ti ni dAþ ðD tÞi si dl: ð13:4:11Þ
dt ot oxk
S S oS
where wi is the velocity of the singular line, which is not necessarily a material
one, and the remarks and the nomenclature in context with Eq. (3.4.11) apply.
ð13:4:14Þ
Interpret this equation in terms of fluxes, supplies, and productions. Then
recall the transport theorem from Exercise 13.2.3 and use it to confirm the
following local equations:
oci ocl
þ ti þ ðr ðc tÞÞi ¼ pi þ si ; e ½½c t þ / ½½c t ? ¼ p þ s
ot oxl A
ð13:4:15Þ
13.4 Conservation of Total Charge 347
for regular and singular points, respectively. Note that in order to find Eq.
(13.4.15)1 contributions from singular lines must simply be disregarded. For
a proof of Eq. (13.4.15)2 consider a small closed loop around L (similar to the
pillbox argument presented in Chap. 3. Then recall that m ¼ e t, e t ¼ 0,
e e ¼ 0, and t ? ¼ w e. Make use of (and prove) the vector identity
A
a ðb cÞ ¼ ða bÞ c. Finally, assume that the vectors of singular pro-
duction and supply are tangential to the singular surface.
Finally apply REYNOLDS’ transport theorem for volumes to Eq. (13.4.2) and
show that in regular points:
oq o
þ ðji þ qti Þ ¼ 0: ð13:4:18Þ
ot oxi
Exercise 13.4.3: An application of the jump condition for the electric field
Assume that the electrons in a metal, i.e., an electrically conductive
material can basically be moved freely. Consider now a charge q in front of a
metallic object. At which angle do the electric field lines E impinge on the
surface of that object? To find the answer decompose the vector E into a
normal and a tangential part as follows:
348 13 Fundamentals of Electromagnetic Field Theory
E ¼ En n þ Et t: ð13:4:19Þ
Recall Sect. 13.3 and interpret the vector E in terms of a force acting on the
free electrons in the metal. Conclude in which direction they should move and
assume that the force is too weak to pull the electrons out of the metal surface.
In a second step consider a point charge in front of an infinitely large,
plane metallic surface and sketch the electric field lines in front and behind
the plate. What do we conclude about the electric field behind the wall
considering the jump condition (13.4.17) and the pervious result regarding
the angle of entry of the electric field lines on a metallic surface? In this
context explain the principle of a FARADAY cage, i.e., the complete shielding
of electromagnetic fields within a closed metallic grid. Is this a way to
protect yourself from poisonous, carcinogenic electro-smog or distracting
cell phone calls?
Paul Adrien Maurice DIRAC was born on August 8, 1902 in Bristol and
died on October 20, 1984 in Tallahassee. He contributed fundamentally
to quantum mechanics, quantum electrodynamics, to the theory of the
electron, and to the properties of antimatter. Being British to the core
DIRAC always kept a stiff upper lip and spoke only if absolutely required.
This is confirmed by many anecdotes, for example: When DIRAC made a
rare error in an equation on the blackboard during a lecture one day, a
courageous student raised his hand: ‘‘Professor DIRAC,’’ he declared, ‘‘I
do not understand equation 2.’’ When DIRAC continued writing, the stu-
dent, assuming that he had not been heard, raised his hand again and
repeated his remark. Again DIRAC merely continued writing … ‘‘Pro-
fessor DIRAC,’’ another student finally interjected, ‘‘that man is asking a question.’’ ‘‘Oh?’’ DIRAC
replied. ‘‘I thought he was making a statement.’’
q0 I0
function, i.e., q ¼ Q0 dðr Þ. The scalar product can easily be evaluated by making
use of the orthogonality of the base vectors in spherical coordinates. We obtain the
following relation for the component Dhri of the charge potential:
ZZ ZZ ZZZ ZZ
Di ni dA Dhri dA ¼ Q0 dðr Þ dV ) Dhri dA ¼ Q0
oV oV V oV ð13:5:2Þ
Q0
) Dhri ¼ :
4pr 2
Thus for a known amount of charge, Q0 , and distance, r, Eq. (13.5.2) constitutes
a rule of measurement for the charge potential or, more precisely, for its radial
component. Note that for the time being we have no information about the two
other components. However, as we shall see below they can both be rescaled to
zero. Moreover, note that extended distributions of charge can, in principle, be
considered as an ensemble of point charges, and we may apply the principle of
superposition for vectors to obtain the corresponding total D field.
We now consider a time-independent, i.e., stationary electric current in an
infinitely long wire at rest. We are interested in measuring the resulting cylin-
drically symmetric magnetic field H around the wire, which is also time-inde-
pendent. In cylindrical coordinates we may write:
H ¼ Hhri er þ Hh#i e# þ Hhzi ez : ð13:5:3Þ
As we shall find out below the radial component can be rescaled to zero. This
corresponds to experience according to which currents may only produce rota-
tional fields. Thus:
Hhri ¼ 0: ð13:5:4Þ
Moreover, the line current is infinitely long. Thus the z direction does not matter
and we may put:
Hhzi ¼ 0: ð13:5:5Þ
350 13 Fundamentals of Electromagnetic Field Theory
Due to the cylindrical symmetry the magnetic field can only depend on the
distance r from the wire (see Fig. 13.6):
Hh#i ¼ Hh#i ðr Þ: ð13:5:6Þ
Finally we use a DIRAC delta function to express the line current as follows:
j ¼ I 0 dð r Þ e z : ð13:5:7Þ
If we insert this into ØRSTED-AMPÈRE’s law from Eq. (13.4.5) (s e# , n ez )
we obtain:
I ZZ I ZZ
I0
0¼ Hi si dl ji ni dA ) Hh#i dl ¼ I0 dðr Þ dA ) Hh#i ¼ :
2pr
oS S oS S
ð13:5:8Þ
This equation constitutes a rule of measurement for the magnetic field for a
given current, I0 , and a given distance, r. It allows us to determine the current
potential H from other, more primitive quantities. If there are several currents the
superposition principle must be applied.
Finally the following remark should be made in this section on the principles of
point wise measurement of the D and H fields. Our motivation of the spherically
symmetric ansatz for the D field around a point charge was quite heuristically.
This concerns in particular the argument that only the radial component Dhri ðr Þ is
important. The same holds for the cylindrically symmetric ansatz for H around an
infinitely long line current, where it was claimed that only the tangential com-
ponent Hh#i ðr Þ counts.
However, there is a more formal argument, which adds to the credibility of this
heuristic approach. It turns out that the D field is only determined up to the curl of
a vector field Z, i.e., r Z, and the H field is only unique up to a gradient f, i.e.,
rf. For a proof we write:
13.5 Measuring the Charge and Current Potentials 351
D0 ¼ D þ r Z; H0 ¼ H þ rf: ð13:5:10Þ
If we insert these expressions in COULOMB’s or in ØRSTED-AMPÈRE’s law (13.4.4/
13.4.5), MAXWELL’s equations remain unchanged, since we may write for the
corresponding expressions therein:
ZZ ZZ ZZ
0 0
D n dA ¼ D n dA þ r Z n dA; ð13:5:11Þ
oV oV oV
and
I I I
H0 s dl ¼ H s dl þ rf s dl: ð13:5:12Þ
oS oS oS
If we now apply GAUSS’ and STOKES’ theorem, respectively (to this end we
assume that r Z and rf are continuous fields) we conclude that:
ZZ ZZ ZZZ
oZk o2 Zk
r Z n dA ¼ ijk ni dA ¼ ijk dV 0; ð13:5:13Þ
oxj oxi oxj
oV oV V
2
because ijk oxo i Zoxk j
¼ 0 and:
I I ZZ
of o2 f
rf s dl ¼ si dl ¼ ijk dAi 0; ð13:5:14Þ
oxi oxj oxk
oS oS S
2
because ijk o f=oxj oxk ¼ 0. Consequently we choose the fields Z and f such that
we enforce in Eqs. (13.5.1/13.5.3) the following conditions:
q ¼ qf þ qp : ð13:6:1Þ
Polarization charges develop in so-called dielectric materials. These are
induced in the material if it is subjected to an electric field E. The polarization can
be envisaged as indicated in Fig. 13.7: Due to the force action associated with the
presence of an E field the electric charges within the molecules are separated and
electric dipoles are formed. If we now consider within this piece of matter a
material volume V ðtÞ in the continuum sense, the system boundaries oV will ‘‘cut’’
through the dipoles and, speaking from a phenomenological point of view, we
obtain an additional amount of electric charge on top of the true charges within the
volume. In a manner of speech these charges are fictitious. They come into being
only after the cut with the system boundaries and they must be considered as a
quasi surface charge. In order to describe this contribution mathematically we
define the so-called polarization vector, P, (or polarization for short) as the product
between the number density of dipoles, np , measured in 1=m3 with the dipole
moment, p. The dipole moment is nothing else but the product between the amount
of charge separated in a dipole, e (a positive number in units of C), and the
−
13.6 Decomposition of the Total Charge, Polarization, Rewriting COULOMB’s Law 353
distance vector, d (in units of m), from the negative to the positive centers of the
dipole charges (note the analogy to the force couple known from mechanics). Thus
we write:
P ¼ np p ¼ np e d: ð13:6:2Þ
Consequently, the unit of polarization is C=m2 , which corresponds perfectly to
our interpretation of polarization as a surface charge density. The total charge in
Eq. (13.4.1) can now be written as:
ZZZ ZZ ZZZ ZZ ZZ
q f q dA:
Q¼ q dV þ dA ¼ q dV Pi ni dA þ ð13:6:3Þ
A A
V þ [V A V þ [V oV A
The negative sign in context with the polarization vector, P, can easily be
interpreted (cf., Fig. 13.7): If positive dipole moments are dissected by oV, the
remaining volume is negatively charged. Moreover, the scalar product P n is
what counts in the balance, since a dipole tangential to the surface gives no
contribution at all. If we require that there are no singular surfaces within V ðtÞ and
polarization is continuous on the surface, GAUSS theorem can be applied to Eq.
(13.6.3) to reveal that:
ZZZ ZZ
f oPi
Q¼ q dV þ q dA: ð13:6:4Þ
oxi A
V þ [V A
These relations are inserted in COULOMB’s law from Eq. (13.4.4). We find:
ZZ ZZZ ZZ
f q dA
ðDi þ Pi Þ ni dA ¼ q dV þ ð13:6:5Þ
A
oV V þ [V A
or:
ZZZ ZZZ ZZ
o f
ðDi þ Pi Þ dV ¼ q dV þ q dA: ð13:6:6Þ
oxi A
V V þ [V A
We now define the so-called charge potential for matter, D ðx; tÞ, by:
D i ¼ D i þ Pi : ð13:6:7Þ
Consequently, COULOMB’s law can also be written as follows:
ZZ ZZZ ZZ
D i ni dA ¼ qf dV þ q dA: ð13:6:8Þ
A
oV V þ [V A
or:
ZZZ ZZZ ZZ
oD i
dV ¼ qf dV þ q dA: ð13:6:9Þ
oxi A
V V þ [V A
354 13 Fundamentals of Electromagnetic Field Theory
Analogously to charge we now decompose the total current density into the cur-
rent density of the true (or free) charges, polarization current density, and mag-
netization current density:
ji ¼ jfi þ jpi þ jm
i : ð13:7:1Þ
The current of the free charges for an open material surface (for simplicity
without a singular line) is given by:
ZZ
Jf ¼ jfi ni dS: ð13:7:2Þ
S
∂S
13.7 Decomposition of the Total Currents, Magnetization 355
ZZ ZZ
d
p
J ¼ jpi ni dS ¼ Pi ni dS: ð13:7:3Þ
dt
S S
Finally, as illustrated in Fig. 13.8, the closed line oS may cut through magnetic
dipoles. Following an idea by AMPÈRE these magnetic dipoles may be envisaged as
circular loops of electric current on an atomic level. Indeed, a circular wire car-
rying an electric current of intensity I, and enclosing a directed surface, A, gen-
erates a magnetic dipole of strength m ¼ IA. Similarly we may interpret the
magnetic field generated by electrons circling around an atom as elementary
magnetic dipoles. We have such atomic current loops in mind when we talk
phenomenologically about magnetization currents and write:
ZZ I I I
Jm ¼ jm n
i i dS ¼ n m
m s
i i dl ¼ n m
IA s
i i dl ¼ Mi si dl; ð13:7:4Þ
S oS oS oS
where nm is the density of elementary current loops and M the so-called magne-
tization. Thus ØRSTED-AMPÈRE’s law (13.4.5) can be rewritten as follows:
ZZ I
d
ðDi þ Pi Þ ni dA ¼ Hi Mi þ ðD tÞi si dl
dt
S oS
ZZ Z ð13:7:5Þ
jfi ni dA j i mi dl:
L
Sþ [S L
One may ask why the loop currents pierce only through the line oS and not also
through the open surface S. Indeed, this does happen (see Fig. 13.8). However, in
that case the currents first enter and then leave the surface. In total there is no
contribution to the magnetization current. We now use the transport theorem from
Eq. (13.2.14) and obtain:
ZZ ffi Z
oðDi þ Pi Þ oðDk þ Pk Þ
þ ti ni dS þ ½ðD þ PÞ ti si dl
ot oxk
S oS
Z I
ð½½ðD þ PÞ t wÞi ti dl ¼ Hi Mi þ ðD tÞi si dl ð13:7:6Þ
L oS
ZZ Z
jfi ni dA j i mi dl:
L
Sþ [S L
ZZ ffi Z
oD i
þ ti qf ni dS ð½½ðD þ PÞ t wÞi ti dl
S ot
L
I ZZ Z ð13:7:8Þ
¼ H i si dl ji ni dA jLi mi dl:
f
Sþ [S
oS L
In summary: The MAXWELL equations, e.g., in the form for regular points:
oB
r B ¼ 0; þ r E ¼ 0; ð13:8:1Þ
ot
oD
r D ¼ q; þ r H ¼ j þ qt;
ot
represent two groups of partial differential equations for the fields E, B, D, and H,
i.e., twelve unknowns, and four other fields, namely the charge and current den-
sities q and j, respectively, or derivates thereof in form of true charges and cur-
rents, polarization, or magnetization. MAXWELL’s equations consist of two vector
and two scalar relations, i.e., eight equations in total. In short, the whole system is
extremely underdetermined. If, in a first step, we could find a connection between
the fields E and D on the one hand side and B and H on the other, the situation
would improve considerably. In a second step we would ‘‘only’’ have to provide
constitutive equations for the four unknown fields charge density (a scalar) and the
current density (a vector) by relating these (for example) to E and B.
Indeed, there exist such relations between the set of fields E and D and the set
B and H, respectively. They are known as the MAXWELL-LORENTZ aether relations.
However, their specific mathematical form depends on the frame of reference. As
we are about the see they can assume a rather complex algebraic form, if we
choose a Euclidean observer, for example.
13.8 The MAXWELL-LORENTZ Aether Relations 357
However, in an inertial frame they have a particularly simple form. Here they
are simply proportional to each other:
1
D ¼ e0 E; H ¼ B: ð13:8:2Þ
l0
The two parameters e0 and l0 are known as the dielectric constant and the
permeability of the vacuum, respectively. By means of measurements of the kind
described in Sects. 13.3 and 13.5, i.e., in a terrestrial lab system at rest (which to a
good approximation can be considered as an inertial system) it was found that:
As Vs
e0 ¼ 8:85 1012 ; l ¼ 12:6 107 : ð13:8:3Þ
Vm 0 Am
Heinrich Rudolf HERTZ was born on February 22, 1857 in Hamburg and
died on January 1, 1894 in Bonn. He was born into a family of lawyers.
Nevertheless, he dedicates his life to rational thinking. In 1877 he goes to
Munich and begins to study engineering. However, he soon realizes that
his true love belongs to the pure sciences. Thus he turns to Berlin and
studies under the famous physicists HELMHOLTZ and KIRCHHOFF. He finishes
his Ph.D. in the field of electrodynamics with highest honors and becomes
HELMHOLTZ’ assistant in 1880. He gets interested in optical experiments,
investigates the nature of NEWTONian rings and develops for this purpose a
method to compute the stresses and strains of elastic bodies pressed against
each other, the so-called theory of HERTZian contact.
Jean-Baptiste BIOT was born on April 21, 1774 in Paris where he also
died on February 3, 1862. In 1797 he became a professor of mathematics
at the École Centrale in Beauvais, in 1800 a professor of physics at the
Collège de France in Paris, and in 1809 a professor of astronomy.
Together with the already mentioned scientist GAY-LUSSAC he took a ride
on a hydrogen balloon up to a height of 4000 m. BIOT was also interested
in geodesy and improved and extended the measurements of the meridian
which had been performed until then. He also discovered the effect of
birefringence of mica, which explains why the mineral BIOTit was named
after him.
What happens if these relations are inserted into Eq. (13.8.1), is investigated in the
following exercise.
358 13 Fundamentals of Electromagnetic Field Theory
o2 E 2 o2 B
¼ c DE; ¼ c2 DB; with D ¼ r r: ð13:8:5Þ
ot2 2
What kind of waves are these, in which medium do they propagate and how?
Why were 19th century scientists particularly astonished about this discovery?
Exercise 13.8.2: COULOMB’ law in traditional form and the definition of the
unit Ampere (BIOT-SAVART’s law)
Use results from Sect. 13.5, specifically:
Q0
D ¼ Dhri ðr Þ er ; with Dhri ¼ ; ð13:8:6Þ
4pr 2
the MAXWELL-LORENTZ aether relations, and the arguments in context with the
electric field presented in Sect. 13.3 and show that a field generating charge,
Q0 , exerts the following force on a test charge, Q:
1 Q0 Q 1 Q0 Q
F¼ 2
er r: ð13:8:7Þ
4pe0 r 4pe0 r 3
Interpret this result as COULOMB’s law known from high school. Now
consider in an analogous manner two electric currents and the corresponding
magnetic fields. Prove that the amount of force per unit length between two
parallel, infinitely long wires is given by:
l0 I0 I
F¼ : ð13:8:8Þ
2p r
13.8 The MAXWELL-LORENTZ Aether Relations 359
For this purpose follow the arguments of Sects. 13.3 and 13.5. In particular
show that the quantity I0 in Eq. (13.8.8) corresponds to the current of the
wire generating an H field and I represents the current in the wire on which a
force is acting due to the resulting B field in the MAXWELL-LORENTZ aether
relations. Explain how Eq. (13.8.8) is used for the definition of the SI-unit of
electric currents, the Ampere. Equation (13.8.8) is sometimes a.k.a. BIOT-
SAVART’s law. Write the equation in vector form. In which direction does the
force per unit length point?
During the course of their biological evolution humans have developed a certain
gut feeling for the physical meaning of mechanical quantities, like mass, velocity,
or force. Thus it is fair to say that the transformation rules presented in Chapter 8
have a certain natural and intuitive touch. Unfortunately this is not the case with
the electrodynamic fields. Nevertheless we shall try and build upon the rules
established for mechanical quantities as much as possible so that they can be
extended into unknown territory. However, this is only possible to a certain
degree, even if we aim ‘‘only’’ at intuitive clarity and in the end there is no other
way but to postulate transformation properties. Of course, we shall attempt to
motivate them as much as possible.
Charge is, like mass, a primitive quantity, proper to matter, which in principle
can be obtained by counting elementary units. Consequently, it seems reasonable
to classify charge as a Euclidean tensor of zeroth order, in other words, an
objective scalar. Note that we do not treat charge like an axial scalar, because it
has not been observed that reflections of coordinates have an influence on the
behavior of charge. We have already explained in Sect. 13.3 that the electric field,
E, and the magnetic induction, B, are defined in terms of measurement by the force
they exert on resting or moving charges. Consequently, the question how these
fields transform during a change of the frame of reference according to Euclidean
transformations (see Sect. 8.2),
360 13 Fundamentals of Electromagnetic Field Theory
is easily answered, based on the transformation rule for body forces shown in Eq.
(8.5.1): Charge is an objective tensor of zeroth (a Euclidean scalar) and forces are
objective tensors of first order, i.e., Euclidean vectors [see also Eq. (8.5.1)]: Thus
the following transformation rule applies in combination with Eq. (13.3.3):
E 0i ¼ O0ij E j ) Ei0 þ ðt0 BÞi ¼ O0il El þ ðt BÞl : ð13:9:2Þ
Note that the electric field, E, together with the LORENTZ part, t B, transform
like a Euclidean vector. However, the individual parts do not. We know already
that the velocity, t, does not transform like a Euclidean vector: Eq. (8.3.18)2.
Moreover, the vector product transforms like an axial tensor of third order: Eq.
(8.3.27). Thus Eq. (13.9.2) leads us to conclude that the magnetic flux density must
compensate the axial character. However, this is only possible if the corresponding
transformation law contains the quantity detðO0 Þ. Note that this does not neces-
sarily mean that the magnetic induction must transform like an axial Euclidean
vector. The transformation rule could be more complicated, but it must contain the
determinant of the rotation matrix in a suitable manner.3 This is all information
that can be obtained from Eq. (13.9.2) for the LORENTZ force.
Thus we have to motivate the transformation behavior of the magnetic induc-
tion from other sources, which only leaves MAXWELL’s equations. However, we
will avoid the differential form shown in Eq. (13.8.1) and try an intuitive argument
instead. The magnetic flux d/ ¼ Bi ni dS is nothing else but the number of field
lines piercing through the surface element dS. Thus we postulate that it is an
objective scalar. Moreover, the normal vector ni has axial character according to
Eq. (8.5.4)2 and, consequently, the magnetic induction should transform like an
axial Euclidean vector:
However, note that this argument is quite lax and rather intuitive. In the end we
may say that Eq. (13.9.3) is really nothing else but a postulate. This becomes even
more obvious if we start from the global relations (13.2.6/13.2.7) instead and write
for the total flux / through a closed hull:
ZZ ZZZ
oBi
/ ¼ Bi ni dA ¼ dV 0: ð13:9:4Þ
oxi
oV V
The zero on the right hand side makes it impossible to say if B has axial
character or not. It provides no guidance regarding the transformation properties of
/ in terms of other quantities whose transformation behavior is known. If we now
turn to the left side of Eq. (13.9.4) the following can be said: Recall that / is a
3
Further down we shall see that the magnetic field, H, has much more complicated
transformation properties.
13.9 Transformation Properties of the Electro-Magnetic Fields 361
measure for the number of magnetic field lines and, consequently, should be a
regular, non-axial scalar. Then the axial normal vector in the surface integral of
(13.9.4) suggests that B meets the transformation rule of Eq. (13.9.3). However, the
volume integral indicates non-axial transformation behavior provided we assume
that the volume element behaves like a true scalar as required in Eq. (8.4.2). Also
recall that we could have alternatively considered the volume element to result
from a scalar triple product, which would make it axial.
Combination of the relations (13.9.2) and (13.9.3) and observing the transfor-
mation rule for the velocity shown in Eq. (8.3.18) results in the following trans-
formation for the electric field:
h i
Ei0 ¼ O0il El lbn O0jb X0jk x0k b0k þ b_ 0j Bn ð13:9:5Þ
and conclude that the electric field does not transform like an objective vector.
We have already mentioned that charge is an objective scalar. In the same spirit
we now postulate that the (non-convective) electric currents are objective quan-
tities. This makes sense since they are nothing else but charges (on an atomic
level) transported across a surface area. Consequently, we require in analogy to the
heat flux vector of Eq. (8.6.5) that:
This defines the transformation properties of the charge and of the current
potentials. From COULOMB’s law shown in Eq. (13.4.8) it follows from the
objectivity of the charge density that the charge potential must be an objective
vector:
q0 ¼ q; ð13:9:9Þ
the definition of Euclidean transforms (13.9.1), and the (local) form of
COULOMB’s law shown in Eq. (13.4.8)2. Now use Eq. (13.9.6) and the fol-
lowing definition for the total electric current (i.e., the non-convective plus
the convective part):
Ji ¼ ji þ qti ð13:9:10Þ
in context with Eq. (8.3.18)2 and show that:
Ji0 ¼ O0ij Jj þ q X0ik x0k b0k þ b_ 0i : ð13:9:11Þ
Thus the total electric current is not objective either. Finally show by means
of ØRSTED’s law from Eq. (13.4.12) the validity of Eq. (13.9.7) and also of:
h i
Hi0 ¼ detðO0 Þ O0il Hl þ lbn O0jb X0jk x0k b0k þ b_ 0j Dn : ð13:9:12Þ
Because charge is an objective scalar and the distance between two points is
also objective, [see Eq. (8.3.8)] the dipole moment, p ¼ ed, is another objective
vector. The number density of dipoles, np , is an objective scalar. Thus Eq. (13.6.2)
leads us to conclude that the polarization vector must transform like
In order to find the transformation rule for magnetization, recall AMPÈRE’s idea
that a magnetic dipole is given by the expression m ¼ IA. The current I is nothing
else but a charge per unit of time, in other words an objective scalar. The directed
surface A transforms like an axial objective vector. Hence we conclude that:
Then Eqs. (13.6.7) and (13.9.7/13.9.13) imply that the charge potential in
matter is an objective vector:
D 0i ¼ O0ij D j : ð13:9:16Þ
However, the following complicated transformation rule must hold for the
current potential in matter if we only observe its definition (13.7.7) in combination
with Eqs. (13.6.7), (13.9.12/13.9.13/13.9.15) and (8.6.18)2:
h i
H 0i ¼ detðO0 Þ O0il H l þ lbn O0jb X0jk x0k b0k þ b_ 0j D n : ð13:9:17Þ
Finally it should be noted that after the transformation rules for all electro-
magnetic fields have been motivated and mathematically established by Eqs.
(13.9.3/13.9.5–13.9.7/13.9.9/13.9.12), we have implicitly postulated that MAX-
WELL’s equations (13.8.1) assume the same form, without any system specific
terms, in a Euclidean frame moving against an inertial system. However, it is quite
instructive to show this explicitly:
oq oq oq
þ ti : ð13:9:20Þ
ot oxi ot X
Moreover, use the transformation rules for the other electromagnetic fields
established in this section, start from MAXWELL’s equations in an inertial
frame, Eqs. (13.8.1), transform them into a Euclidean frame and show that
they keep their form and system-dependent terms do not enter:
oB0i oB0i 0 oEk
0
¼ 0; þ ijk ¼ 0; ð13:9:21Þ
ox0i ot ox0j
We now turn to the MAXWELL-LORENTZ aether relations from Eq. (13.8.2). It was
mentioned before that their simple mathematical form was established in an
inertial system. If we now transform them by means of Eqs. (13.9.5/13.9.7) and
(13.9.3/13.9.12), respectively, into a Euclidean system we obtain:
h i
D0i ¼ e0 Ei0 þ 0ijp X0jk x0k bk þ b_ 0j B0p ð13:10:1Þ
and:
1 1 h i
Hi0 ¼ B0i þ 2 0ijr X0jk x0k b0k þ b_ 0j Er0
l0 c
h i ð13:10:2Þ
1 0 0
0 0
0
0 _ 0 0 0 0
þ 2 ijr Xjk xk bk þ bj rst Xsu xu bu þ bs Bt ; _ 0
c
respectively. Obviously the MAXWELL-LORENTZ aether relations do not keep their
simple form (13.8.2) during translational and rotational transformations and many
system dependent terms do occur. On first glance this is not too surprising:
A similar problem occurred in context with the balance of momentum, where
centrifugal, CORIOLIS, EULER, and relative acceleration terms occurred after trans-
formation to a Euclidean system all of which are system dependent.
13.10 Transformation Properties of the MAXWELL-LORENTZ Aether Relations 365
This, however, shows that the balance of momentum has the same form in
Galilean systems (i.e., no system-dependent terms) as in the original inertial
system. Thus nineteen century scientists concluded (prematurely) that all Galilean
systems must be inertial systems.
But then electrodynamics appeared on stage: If we specialize the transformation
rules for the MAXWELL-LORENTZ aether relations from Eqs. (13.10.1/13.10.2) to the
special case of a Galilean transformation of Eq. (13.10.3) we obtain:
D0i ¼ e0 Ei0 þ 0ijk Vj0 B0k : ð13:10:5Þ
and:
" 0
# !
1 V2 Vi0 Vj0
Hi0 ¼ 1 2 dij þ 2 B0j þ e0 0ijk Vj0 Ek0 : ð13:10:6Þ
l0 c c
Insert this in Eqs. (8.3.46) and (13.10.1/13.10.2), and show the validity of
Eqs. (13.10.4–13.10.6).
From now on we shall add time to the group of the three spatial coordinates. In
order to make sure that these four coordinates have the same dimensions we
multiply time by the (constant) speed of light, c, and obtain in Cartesian
coordinates:
x ¼ x0 ; x1 ; x2 ; x3 ¼ xA eA ; x0 ¼ ct; A ¼ 0; 1; 2; 3: ð13:11:1Þ
In this equation use was made of a unit vector e0 ‘‘pointing into the future.’’
Moreover, use is made of capital indices, e.g., A, and, if they appear twice in a
product, automatic summation from 0 to 3 is implied. This extends the usual
EINSTEIN summation convention and combines space and time. Of course, a point
in space-time can also be identified by another observer. However, the point itself
is unique and, consequently, a general, invertible space-time transformation of the
following form must exist:
x0 B ¼ ~x0 B xA : ð13:11:2Þ
Examples of such transformations are the previously discussed Euclidean or
Galilean transformations. They possess the property that the time component 0
decouples completely from the space components 1–3. As we shall see this is no
longer the case during a LORENTZ transformation. Moreover, it is noteworthy that
the Euclidean as well as the Galilean transformation are linear w.r.t. the coordi-
nates. This becomes immediately evident from Eqs. (13.9.1) and (13.10.3). We
shall see that the LORENTZ transformation has this property as well. However, in
general a space-time transformation (13.11.2) may also be non-linear.
It is curious to note that we have employed a co-/contravariant notation (in
capital letters) in context with Eqs. (13.11.1/13.11.2): Recall that a distinction
between co- and contravariant in three-dimensional space was impossible. How-
ever, this is not so in 4D, where the space-time metric used to form a four-
dimensional line element is no longer a unit matrix. In fact, the line element can be
expressed in an inertial system by its time and (Cartesian) space coordinates (this
extension now replaces the three-dimensional Cartesian space) as follows:
13.11 Four-Vector Formalism for the Electromagnetic Fields 367
2 2 2
ðdsÞ2 ¼ ðdctÞ2 þ dx1 þ dx2 þ dx3 ¼ gAB dxA dxB ; ð13:11:3Þ
where the space-time metric of the inertial system is given by4:
2 3
1 0 0 0
6 0 1 0 07
gAB ¼ gAB ¼ 6 4 0 0 1
7: ð13:11:4Þ
05
0 0 0 1
Further down we shall motivate why it is reasonable to introduce a four-
dimensional line-element and what it can be used for. At this point all of these
preliminary notes on 4D-transforms are only intended to illustrate the analogies to
the 3D case. In this spirit we now introduce a so-called world tensor, F, of j þ k
order by its transformation behavior in terms of co-/contravariant coordinates, in
complete analogy to Eq. (8.3.1):
0 w 0 0 A1
0 A1 A2 ...Aj
ox p ox ox ox0 Aj oxD1 oxDk C1 C2 ...Cj
F B1 B2 ...Bk ¼ det sign F :
ox ox oxC1 oxCj ox0B1 ox0Bk D1 D2 ...Dk
ð13:11:5Þ
Note that the functional determinant of space-time transformations shown in
Eq. (13.11.2) must not be equal to 1, in contrast to the Euclidean transformation
(13.9.1) [compare also the footnote in context with Eqs. (8.3.1/8.3.2)]. The factor
p in this equation can assume the values 0 and 1. The exponent w is a.k.a. the
weight of the tensor. Moreover, the case of p ¼ 0, w ¼ 0 is referred to as an
absolute tensor and, for p ¼ 1, w ¼ 0, as an axial tensor. For p ¼ 0, w 6¼ 0 we
speak of a so-called relative world tensor of weight w, and for p ¼ 1, w 6¼ 0 of a
relative axial tensor of weight w. As we shall see, the theory of electrodynamics is
governed by absolute world tensors of second order and by relative world tensors
of first and second order. Without going into the mathematical details it should be
noted that a world tensor with a weight different from zero is also often referred to
as a tensor density.
In a first step we now combine the electric field Ei and the magnetic flux density
Bi in an absolute antisymmetric covariant world tensor of second order:
2 3
0 E1 =c E2 =c E3 =c
6 E1 =c 0 B3 B2 7
uAB ¼ 6
4 E2 =c B3
7: ð13:11:6Þ
0 B1 5
E3 =c B2 B1 0
It transforms according to:
4
In the older relativistic literature the stringent application of tensor calculus is avoided and the
imaginary unit, i2=-1 is used in context with the definition of the time coordinate. This renders it
possible to define the 4D-line element in a quasi-Pythagorean way. If we use tensors from the
very beginning on we do not need this concept any more.
368 13 Fundamentals of Electromagnetic Field Theory
ox A ox B
u0 CD ¼ u : ð13:11:7Þ
ox0 C ox0 D AB
In a similar fashion we now combine charge and current density in the fol-
lowing contravariant four-vector:
ox0 B oxA
0 0 0 0
¼ d B
C ; V i ¼ Xij x j b j þ b_ 0i with Vi0 ¼ O0ij Vj ð13:11:12Þ
oxA ox0 C
13.11 Four-Vector Formalism for the Electromagnetic Fields 369
X0ij ¼ O0ik O0jl Xkl with the definition Xkl ¼ O_ km Olm ; ð13:11:13Þ
axial world tensor of forth order of weight 1 (also compare Eq. (8.3.27) for
the 3D case):
0 0 R 0 S 0 T 0 U
0 RSTU 1 ox ox ox ox ox ABCD
¼ det : ð13:11:15Þ
ox oxA oxB oxC oxD
Use this relation in combination with the transformation rule (13.11.7) and
show that the first set of MAXWELL’s equations is form-invariant so that we
may write for an arbitrary coordinate system:
ou0CD
0 ABCD ¼ 0: ð13:11:16Þ
ox0 B
The balance of electric charge (13.4.18) can be rewritten by means of the defi-
nition (13.11.9) for the four-vector as follows:
orA
¼ 0: ð13:11:17Þ
oxA
This equation can be formally solved by introducing the antisymmetric tensor
gAB ¼ gBA of second order with the following property:
ogAB
¼ rA : ð13:11:18Þ
oxB
Obviously gAB must be interpreted as a four-dimensional charge/current
potential (note the derivative!). It is related to the well-known charge and current
potentials Di and Hi by:
2 3
0 cD1 cD2 cD3
6 cD1 0 H3 H2 7
gAB ¼ 64 cD2 H3
7: ð13:11:19Þ
0 H1 5
cD3 H2 H1 0
og0 AB
¼ r0 A : ð13:11:21Þ
ox0 B
Observe the following identity during the proof:
0 B
o ox ox
det ¼ 0: ð13:11:22Þ
oxB ox ox0 S
Now prove the identity by using the LAPLACE expansion theorem for
determinants (in four dimensions):
AB 1
ARST BCDE ACR ADS AET
A1 ¼6 : ð13:11:23Þ
detðAÞ
Specialize to Euclidean transformations according to Eq. (13.11.10) and
show that the transformation rules (13.9.6/13.9.11) result. Finally show by
means of the transformations rule (13.11.9) that the balance of charge
(13.11.17) in an arbitrary system reads:
or0 A
¼ 0: ð13:11:24Þ
ox0 A
Hint: Use the identity (13.11.22) again.
Recall the so-called space-time metric gAB that was already introduced in
Eq. (13.11.4). We require that it transforms like an absolute contravariant world
tensor of second order:
ox0 A ox 0 B CD
g0 AB ¼ g : ð13:12:1Þ
oxC oxD
372 13 Fundamentals of Electromagnetic Field Theory
For reasons that will become clear immediately we calculate the determinant of
this relation and find:
0
0 AB 2 ox
det g ¼ det det gCD : ð13:12:2Þ
ox
In fact, by using Eq. (13.11.4) for the space-time metric of an inertial system we
may also put detðgCD Þ ¼ 1. However, it is more important to note that in view of
the general definition shown in Eq. (13.11.5) the determinant of the contravariant
space-time metric must be a relative world tensor of zeroth order (i.e., a relative
world scalar) of weight w ¼ 2. Loosely speaking the determinant is a scalar
density.
With this relation the MAXWELL-LORENTZ aether relations from Eq. (13.8.2) can
be rewritten. They were originally verified in an inertial system, and thus:
1 12
gCD ¼ det gMN gCA gDB uAB : ð13:12:4Þ
l0
This can easily be verified by an explicit calculation (the square root
1
2
ð det gMN Þ ¼ 1 has been omitted for convenience in the following long
formula):
13.12 Four-Vector Notation of the MAXWELL-LORENTZ Aether Relations 373
2 3 2 32 0 Ec1 Ec2
3
Ec3 2 1 0 0 3
0 cD1 cD2 cD3 1 0 0 0 0
6 cD H2 7 16 76 E1 7
B2 76 07
6 1 0 H3 7 6 0 1 0 0 76 0 B3
76
0 1 0 7
76 E2
c
6 7¼ 6 6 7
4 cD2 H3 0 H1 5 l0 4 0 0 1 0 564c B3 0 B1 754 0 0 1 05
cD3 H2 H1 0 0 0 0 1 E3
B2 B1 0 0 0 0 1
c
2 E1 E2 E3 3
0 c c c
6 E1 7
16 c 0 B3 B2 7 2
7; c ¼ 1 :
¼ 6
l0 6 E
4 c2 B3 0 B1 75 e0 l0
Ec3 B2 B1 0
ð13:12:5Þ
On first glance the determinant in Eq. (13.12.4) seems rather artificial. However,
it is very important because in a world tensor equation the order and the weight must
agree. We postulate that the MAXWELL-LORENTZ aether relations in the form (13.12.4)
are of the caliber of a ‘‘world equation.’’ It is then easily confirmed that tensors of
second order are present on both sides of the equation. Moreover, the weight on the
left hand side is equal to one, because of Eq. (13.11.8), and so is the weight on the
right hand side, because of Eqs. (13.11.7) and (13.12.1/13.12.2), respectively:
ð2Þ 12 ¼ þ1. Only then Eqs. (13.11.7/13.12.19) and (13.12.1/13.12.2) trans-
form Eq. (13.12.4) in a form invariant manner into an arbitrary system:
1 12
g0 LM ¼ det g0 UV g 0 LR g0 MS u0RS ð13:12:6Þ
l0
Our next objective is to find the transformations for which the MAXWELL-LOR-
ENTZ aether relations keep the simple form shown in Eq. (13.8.2) or (13.12.4),
respectively. Then the metric g0 in Eq. (13.12.6) must have exactly the same
components as the metric g in Eq. (13.11.4). If we observe the transformation rules
(13.12.1) we find that:
2 3AB 2 3CD 2 3CD
1 0 0 0 1 0 0 0 1 0 0 0
6 0 1 0 07 ox 0 A ox 0 B 6 07 6 0 0 07
6 7 6 0 1 0 7 6 1 7
6 7 ¼ 6 7 ,6 7
4 0 0 1 05 oxC oxD 4 0 0 1 05 4 0 0 1 05
0 0 0 1 0 0 0 1 0 0 0 1
2 3AB
1 0 0 0
6
C
ox ox 6 0 D 1 0 07
7
¼ 0A 0B6 7 :
ox ox 4 0 0 1 05
0 0 0 1
ð13:12:7Þ
These two formulae describe the transformation from the original inertial
system into the new one and vice versa. The second equation is particularly useful
in order to derive the LORENTZ transformation in a conceptionally simple manner.
We shall present the details shortly. Note that the elements of the transformation
374 13 Fundamentals of Electromagnetic Field Theory
matrices ox0 =ox and ox=ox0 , respectively, represent a total of sixteen unknowns.
However, due to the symmetry of the matrix (13.11.4), Eq. (13.12.7) provides only
ten constraints. Correspondingly, six undetermined parameter will remain in the
unknown transformation. These can be interpreted as the three components of
relative translational velocity, Vi , between both system origins (a.k.a. boost
velocity) as well as the three angles of rotation between the axes (i.e., the inde-
pendent components of the rotation matrix). We proceed to investigate this a little
further. We conclude from Eq. (13.11.4) that g as well as g0 are constant and,
therefore, the unknown coordinate transformation must be linear. Consequently,
we solve the problem in two steps: First, we transform from the non-dashed system
by a boost into an intermediate inertial system ~xB , whose axes are not tilted w.r.t.
the ones of the non-dashed system, albeit related to each other linearly:
xA ¼ aAB~xB þ bA : ð13:12:8Þ
dx0 ¼ a0 0 d~x0
dxA ¼ aAB d~xB ) ð13:12:9Þ
dxi ¼ ai 0 d~x0 :
In order to include the velocity we divide dxi by dx0 and observe Eq. (13.11.1):
Vi dxi ai 0
0¼ 0 : ð13:12:10Þ
c dx a0
2 X 3 2
1 ¼ a00 þ ai 0 : ð13:12:11Þ
i¼1
5
The minus sign in the velocity is arbitrary. In fact some textbooks do not follow this
convention. However, we do and this guarantees consistency with the assumed direction of the
vector b shown in Fig. 8.1, the remarks in context with Eq. (13.10.3) and, finally, with Exercise
13.11.1.
13.12 Four-Vector Notation of the MAXWELL-LORENTZ Aether Relations 375
6
Note that it is important to distinguish between co- and contra variant components of the space-
time vectors and must strictly be observed in the following formulae. Of course there is no
difference between co- and contra variant for Cartesian non-space-time quantities, like the
velocity Vi or the rotation matrix Oij. Consequently, the rule of cross-wise summation (cf., the
remark after Eq. (2.4.13), which holds mutatis mutandis also in 4D) does not hold in the sub-
sequent formulae: Unfortunately this diminishes their beauty.
376 13 Fundamentals of Electromagnetic Field Theory
Vk
x0 0 ¼ c x0 þ c xk ;
c ffi ð13:12:18Þ
0j 0 Vi 0 Vi Vk
x ¼ Oji c x þ dik þ ðc 1Þ 2 xk :
c V
Beside the rule the following auxiliary formulae have been used:
These relations are in complete agreement with Eq. (13.10.3) including the
signs.
7
Spherically, because space is considered to behave isotropically.
13.12 Four-Vector Notation of the MAXWELL-LORENTZ Aether Relations 377
2 2 2 2
f ðxÞ ¼ x0 þ x1 þ x2 þ x3 0: ð13:12:22Þ
One would think that a moving observer should not see a spherical wave but a
somewhat deformed one instead depending on his state of motion. Intuitively we
expect a certain retardation of the light wave in the direction of the velocity Vi of
relative motion. However, this contradicts the result of the MICHELSON-MORLEY
experiment, which we shall briefly describe in what follows. MICHELSON and
MORLEY found out that the speed of light remains constant with an accuracy of
5 kms after turning their interferometer by 90: Fig. 13.9.
The principle of the experiment can be illustrated as follows: We consider two
equally strong swimmers (a metaphor for two light beams) of speed c in a river
(the aether) flowing at speed V. The first one swims along a distance l1, first with
and then against the current. The other one swims along the distance l2 from one
river bench to the other and then returns. Obviously the second swimmer must
swim at a certain angle w.r.t. the direction of the flow of the river in order to arrive
exactly opposite from the starting point. The time difference between the first and
of the second swimmer is approximately given by:
2ðl1 l2 Þ 2l1 l2 V 2
Dt
þ ; ð13:12:23Þ
c c c
where it has been assumed that the flow velocity of the river, V, is much smaller
than the velocity proper, c, of the swimmers.
l2
mono- semi-
chromatic transparent
light source mirror
l1
mirror
detector
378 13 Fundamentals of Electromagnetic Field Theory
We now apply this result to the light beams shown in Fig. 13.9. There will be a
run time difference between them, potentially resulting in interference, in partic-
ular because the distances l1 and l2 cannot be made exactly equal, even if the
mechanical parts are made with highest diligence. Consequently, the potential
impact of the velocity V of the aether is difficult to ascertain. However, if the
apparatus is turned by 90 after the first run, so that l1 and l2 exchange their role,
the run time difference is now:
2ðl2 l1 Þ 2l2 l1 V 2
D^t
þ : ð13:12:24Þ
c c c
Edward Williams MORLEY was born on January 29, 1838 in Newark, New
Jersey and died on February 24, 1923 in West Hartford, Connecticut. He
completed his academic studies at the Williams College in 1860. From
1869 until 1906 he was a professor of chemistry at today’s Case Western
Reserve University. Being a versatile experimentalist he became an ideal
colleague for MICHELSON: In 1887 they conducted the famous experiment
named after them. Like MICHELSON he did not really accept its negative
outcome that denied the existence of the aether. MORLEY worked also on
the exact chemical composition of Earth’s atmosphere, the thermal
expansion of solids and on measuring speed of light in magnetic fields.
2
Thus the total run time difference is given by Dt þ D^t ¼ l1 þl c
2 V
c , which is
proportional to the square of the aether drift. MICHELSON and MORLEY chose distances
of l1
l2 ¼ 11 m. Moreover, we have to keep in mind that Earth circles around the
Sun at the (under terrestrial circumstances) enormous speed of roughly 30 km s .
Nevertheless, this is still small when compared to the speed of light, which is about
13.12 Four-Vector Notation of the MAXWELL-LORENTZ Aether Relations 379
V 2
3 105 km
s . Thus the optical distance is Dd ¼ cðDt þ D^tÞ ¼ ðl1 þ l2 Þ c
2:2 107 m. This has to be related to the wave length k of the incoming light, which
is roughly 500 nm, i.e., Dd=k
0:44. Thus a change of the interference pattern
should clearly be visible. However, the measurements indicated a much smaller
ratio of not more than 0.01. Thus scientists came to the conclusion that the speed of
light in vacuum has the same (maximum) value of 3 105 km s in all frames of
reference, independently of their state of motion.
The experiment has been repeated many times ever since—with the same neg-
ative outcome. It was also taken into account that Earth travels relatively to the center
of our galaxy at the much greater relative speed of 200 km s . Hence even an aether
drift almost ten times as large does not seem to ‘‘impress’’ light waves at all. The
speed of light appears to be a universal constant, independent of the observer, may he
be moving or not. On the other hand, if we are honest, it would indeed be alarming if
the propagation characteristics of something as fundamental as a quantum of light
depended on the state of motion of the observer. This could become a real problem
for the causality chain since we might be able to ‘‘overtake’’ events and, effectively,
could no longer differentiate between the cause and the effect. The mathematical
models we use to explain the physical world would suffer from inner contradictions,
potentially predicting instability and chaos where no chaos has been observed.
ox C ox D 0 A 0 B
dxC dxD ¼ dx dx ð13:12:30Þ
ox0 A ox0 B
Inserting this in Eq. (13.12.29) results in:
ox C ox D
g0AB ¼ gCD : ð13:12:31Þ
ox0 A ox0 B
This is Eq. (13.12.1) in covariant form which had previously been obtained
from the invariance of the MAXWELL-LORENTZ aether relations, by means of which
the LORENTZ transformation was derived. In retrospect it is not surprising that the
postulate of a spherical form for a flash of light in all systems arising by mutual
boost Vi at equal propagation speed c results in the same equations: After all
light is an electromagnetic wave, and if the MAXWELL-LORENTZ aether relations in
simple form hold in all these systems, Exercise 13.8.1 predicts the same wave
equation for all of them.
We now want to include the effect of the electromagnetic fields on the balances of
momentum and of energy. For this purpose we simply add in the balance of
momentum to the (gravitational) bulk forces q fi the electromagnetic contribution
qEi þ ½ðj þ qtÞ Bi and write in regular points:
oqti o
þ qti tj rji ¼ q fi þ qEi þ ½ðj þ qtÞ B i : ð13:13:1Þ
ot oxj
Experiments show that the power of the electromagnetic field (the so-called
JOULE heating or heat production due to electric resistance) is given by the scalar
product between electric current density j þ qt and the electric field E. Thus the
local energy balance reads:
ffi
o 1 o 1
q u þ t2 þ q u þ t2 tj þ qj rji ti ¼ q fi ti þ ðji þ qti ÞEi :
ot 2 oxj 2
ð13:13:2Þ
13.13 Energy and Momentum of the Electromagnetic Field 381
In other words, momentum and energy are no longer conserved due to pro-
duction terms that suddenly arise. However, it is possible to rearrange MAXWELL’s
equations suitably and to combine them with the balances of momentum and
energy so that conservation of energy and momentum is guaranteed. To this end
we start from MAXWELL’s equations for regular points shown in Eq. (13.8.1) and
insert the MAXWELL-LORENTZ aether relations from Eq. (13.8.2):
oBi oBi oEk
¼ 0; þ ijk ¼ 0;
oxi ot oxj
ð13:13:3Þ
oEi oEi 1 oBk
e0 ¼ q; e0 þ ijk ¼ ji þ qti :
oxi ot l0 oxj
We now multiply the second of these equations by l1 Bi and the fourth one by
0
Ei . Both results are then added and the product rule is applied:
o1 1 o 1
E e0 E þ B B þ jki Ek Bi ¼ ðji þ qti ÞEi : ð13:13:4Þ
ot 2 l0 oxj l0
Inserting the MAXWELL-LORENTZ aether relations yields:
o1 o
ðE D þ B H Þ þ ðE HÞi ¼ ðji þ qti ÞEi : ð13:13:5Þ
ot 2 oxi
If we ignore the minus sign, the left hand side is equal to the production density
of energy due to electromagnetic fields [JOULE heating, cf., Eq. (13.13.2)]. If we
eliminate this term we obtain:
ffi
o 1 1
q u þ t2 þ ð E D þ B H Þ
ot 2 2
ffi ð13:13:6Þ
o 1
þ q u þ t2 tj þ qj rji ti þ ðE HÞj ¼ q fi ti :
oxj 2
Thus the production on the right hand side of the energy balance disappears and
additional terms emerge in the temporal derivative and in the divergence term. For
382 13 Fundamentals of Electromagnetic Field Theory
Clearly the left hand side of this equation is, with the exception of the sign, the
production of momentum due to the electromagnetic field from Eq. (13.13.1). If
we combine both equations the following conservation law is obtained:
o
qti þ ðD BÞi þ
ot ð13:13:10Þ
o 1
qti tj rji þ ðE D þ B HÞdij Ei Dj Bi Hj ¼ q fi :
oxj 2
We conclude that D B represents the density of momentum of the electro-
magnetic field. Moreover, in view of the mechanical stress tensor, rij , the term
12 ðE D þ B HÞ dij þ Ei Dj þ Bi Hj formally represents ‘‘electromagnetic stres-
ses.’’ Indeed, it is known as MAXWELL stress tensor.
Recall Eqs. (13.6.7) and (13.7.7), which relate the material dependent polarization
P and the magnetization M with the charge and current potentials D and H to form
the quantities D and H , the charge and current potentials in matter. As their names
indicate the latter are material dependent quantities:
D ¼ D þ P; H ¼ H M ðP tÞ: ð13:14:1Þ
13.14 Simple Electrodynamic Constitutive Equations 383
However, it is also fair to point out that various technical materials do show a
pronounced non-linear behavior as far as polarization is concerned, e.g., Seignette
salt or barium titanate. Then the loading history, in other words, the temporal
development of a rising and decaying E field becomes an issue and hysteresis is
observed. Moreover, the action of the electric field is in general coupled to the
effect of mechanical stresses. We will learn more about this shortly. Before that,
however, we combine Eq. (13.14.1)1 with the MAXWELL-LORENTZ aether relation
(13.8.2)1 and obtain:
D i ¼ Di þ e0 v Ei ¼ e0 ð1 þ vÞEi ¼ e0 er Ei j Ei ; ð13:14:4Þ
where the so-called relative dielectric constant er ¼ 1 þ v has been introduced.
The dielectric constant is frequently used in high school experiments in context
with plate capacitors with dielectric ‘‘fillings.’’ Another expression used for the
same quantity is the term dielectric or relative permittivity in combination with the
symbols j or jij for the corresponding second order tensor used for anisotropic
matter. Note that the symbol eij is sometimes used in the anisotropic case instead.
Clearly this can easily be confused with the mechanical strain, especially if we
consider coupled problems (see further down). For this reason we are not going to
follow that convention and just warn the reader to read the instructions carefully
when using other literature and programs. In summary, in the anisotropic case we
replace Eq. (13.14.4) by:
384 13 Fundamentals of Electromagnetic Field Theory
In terms of the flux-force concept introduced in Sect. 12.4 we may also refer to
the electric field as the ‘‘driving force’’ and to the polarization or to the charge
potential in matter as the ‘‘fluxes.’’ Now surely, it is only fair to refer to the electric
field as a ‘‘driving force’’ because of the force it exerts on charges. However, if we
start coupling mechanical with electric effects things are not so straightforward any
more. Indeed, it is possible to use mechanical means to create a polarization or a
charge potential. If we follow the arguments outlined in Sect. 12.4 strains are the
drivers on the mechanical side and not the stresses, as one would intuitively
expect. Rather the stresses are the fluxes, as we might have guessed by noting that
the traction is the non-convective flux of momentum. Now if both driving forces
are present, i.e., electric fields and mechanical strains act simultaneously, we may
write within a linear theory:
D i ¼ jij Ej þ eijk ejk or Pi ¼ e0 vij Ej þ eijk ejk : ð13:14:6Þ
Lars ONSAGER was born on November 27, 1903 in Kristiania (now Oslo)
and died on October 5, 1976 in Coral Gables (Florida). He is well known
for his work in irreversible thermodynamics and physical chemistry in
particular for his discovery of the reciprocal relations in constitutive
equations and the corresponding kinetic interpretation. This got him the
Nobel Prize in Chemistry in 1968 which in turn resulted in envy and
tirades of hate by many other thermodynamicists. He held the Gibbs
Professorship of Theoretical Chemistry at Yale University and had a
reputation for being an extremely bad teacher. In fact his chemistry
lectures were referred to as Norwegian I and II and this had nothing to do
with his mother tongue. Be that as it may, as a scientist he surely made a considerable con-
tribution to the understanding of the nature of irreversibility.
The quantities eijk are a.k.a. (charge) piezoelectric constants and they form a
tensor of third order. In fact, they allow quantifying the so-called direct piezo-
electric effect: If we squeeze a piezoelectric crystal the originally balanced charges
will be separated and a polarization charge is generated. However, this is only half
the story. HOOKE’s law must be extended as well in order to cover more than just
the mechanical strain of Eq. (6.2.1). We write:
rij ¼ Cijkl ekl ekij Ek : ð13:14:7Þ
This relation describes what is known as the converse piezoelectric effect: By
application of an electric field the charges of the piezosensitive material are
separated and a mechanical strain results which, if constrained, results in a stress.
Interestingly the same piezoelectric coefficients are used in Eq. (13.14.7) to
relate the effects of the electric field to the mechanical stresses. This is a conse-
quence of ONSAGER’s principle of reciprocity. It can be motivated by kinetic
arguments in context with ONSAGER’s regression hypothesis and the so-called
fluctuation-regression theorem, but we will not provide any details here. However,
13.14 Simple Electrodynamic Constitutive Equations 385
what we will do is answer how many different coefficients there are altogether in a
linearized theory for a material of highest degree of anisotropy. We start with the
stiffness matrix Cijkl. Recall that the linearized strain tensor is symmetric by
definition. We also assume that the stress tensor is symmetric. Thus both have six
independent components and this leaves us with a total of 696 independent
coefficients for the stiffness matrix. However, there is still more symmetry if we
consider the elastically stored energy density, w, and its complimentary form, w*.
They are defined as follows:
Z~e¼e Zr~¼r
w¼ rij ð~eÞ d~eji ; w ¼ eij ðr
~Þ d~
rji : ð13:14:8Þ
~e¼0 ~¼0
r
If we now insert HOOKE’s law [i.e., Eq. (13.14.7) without the electric part] the
integration can easily be performed and we obtain:
Z~e¼e Z~e¼e
w ¼ Cijkl ~ekl d~eij ; w ¼ Cijkl ~eij d~ekl : ð13:14:9Þ
~e¼0 ~e¼0
As indicated in Fig. 13.10 we expect for a linear elastic material both strain
energies to be equal, and thus:
Cijkl ¼ Cklij : ð13:14:10Þ
We conclude that we may exchange index pairs and this reduces the amount of
36 independent components to (6 9 6 - 6)/2 ? 6 = 19 = 21. In fact this can be
reduced even further, but only if we reduce the degree of anisotropy of the
material. How to do this is explained (for example) in the books by Nye [2] or Tsai
[3]. We will not explain this any further, and it may suffice to say that for isotropic
materials we end up with two independent elastic constants, e.g., YOUNG’s modulus
and POISSON’s ratio or the two LAMÉ coefficients.
Fig. 13.10 Strain and electric energy densities and their complements
386 13 Fundamentals of Electromagnetic Field Theory
We now insert Eq. (13.14.7)2 without the mechanical part to find that:
~ ~
Z
E¼E Z
E¼E
w ¼ jij E ~ i ; w ¼ jij
~ j dE E~ i dE
~j: ð13:14:12Þ
~
E¼0 ~
E¼0
In the linear case both should be equal and, hence, we conclude that the sus-
ceptibility tensor and the relative permittivity tensor are symmetric and have six
independent components for the highest degree of anisotropy and if the coordinate
system does not coincide with the principal axes of the crystal:
vij ¼ vji ; jij ¼ jji : ð13:14:13Þ
The situation is similar to the case of the matrix of thermal expansion coeffi-
cients: If the principal axes of the crystal and of the coordinate system coincide the
tensor is diagonal with a maximum of three different components for the three
principal directions.
This leaves us with the coupling coefficients or charge piezoelectric constants
eijk . Since the stress and strain tensors are both symmetric, i.e., consist of six
independent coefficients, it follows that eijk ¼ eikj i.e., we have a maximum of
3 9 6 = 18 independent entries. It was Woldemar VOIGT who combined all of
these material coefficients in a matrix scheme, nowadays known as VOIGT’s
notation. It is very popular in engineering and frequently used in finite element
codes. For these reasons we will touch upon it briefly.
In VOIGT’s notation symmetric index pairs (i, j) are replaced by superindices (I),
which are easily spotted by using capital letters. For example, the symmetric stress
tensor consisting of six independent components can be represented by a 6 9 1
matrix, i.e., a column containing these six independent components. Note that one
should be careful when calling this column a vector, because it does not transform
according to Eq. (2.4.1). Note that superindices are typically assigned according to
the convention established in Table 13.1.
Note the factors of two in the strain matrix. Their origin will become clear if we
present the coupled constitutive relations as a matrix equation of the form A x ¼ b,
which arises from the tensor equations (13.14.6/13.14.7). We incorporate the
entries rI , D i , CIJ , eiJ , jij , eI , and Ei in a giant matrix scheme as follows:
C eT
=
e E
σ1 C11 C12 C13 C14 C15 C16 − e11 − e21 − e31 ε1
σ2 C21 C22 C23 C24 C25 C26 − e12 − e22 − e32 ε2
σ3 C31 C32 C33 C34 C35 C36 − e13 − e23 − e33 ε3
σ4 C41 C42 C43 C44 C45 C46 − e14 − e24 − e34 ε4
σ 5 = C51 C52 C53 C54 C55 C56 − e15 − e25 − e35 ε5 .
σ6 C61 C62 C63 C64 C65 C66 − e16 − e26 − e36 ε6
1 e11 e12 e13 e14 e15 e16 κ11 κ12 κ13 E1
2 e21 e22 e23 e24 e25 e26 κ 21 κ 22 κ 23 E2
3 e31 e32 e33 e34 e35 e36 κ31 κ32 κ33 E3
ð13:14:15Þ
vm
M¼ B: ð13:14:17Þ
l0
Thus the current potential in matter is given by:
l0 H ¼ l0 ð1 þ vm ÞH ¼ lr B: ð13:14:18Þ
The constant of proportionality, vm , is a.k.a. magnetic susceptibility. The
combination lr ¼ 1 þ vm is referred to as relative permeability. Note that
molecular models can be used to compute the magnetic susceptibility.
To a first approximation the true current density, jf , also obeys a linear relation:
1
jf ¼ E: ð13:14:19Þ
r
This makes sense because the presence of an electric field is the reason for the
movement of charge carriers. Moreover, its action, i.e., the true current density,
should increase if the resistance to movement, i.e., the specific electric resistance,
r, of the corresponding material decreases. We now multiply this relation by dxi ,
where jdxi j ¼ dl represents the length of an electric line current of cross-section A.
Moreover we relate the electric field to its potential, U:
oU
Ei ¼ ð13:14:20Þ
oxi
and find:
oU oU
A dxi ¼ rjfi A dxi ) ni A dl ¼ rjf A dl: ð13:14:21Þ
oxi oxi
13.14 Simple Electrodynamic Constitutive Equations 389
Georg Simon OHM was born on March 16, 1789 in Erlangen and died on
July 6, 1854 in Munich. At the young age of sixteen he began to study
mathematics, physics, and philosophy at Friedrich-Alexander University
in Erlangen in 1805. However, due to financial straits he had to abandon
school and take a job as a mathematics teacher at a private school in
Switzerland. He returned to Erlangen when he was 22, finished his Ph.D.
work on light and colors in 1811, and then worked for three semesters as a
private docent for mathematics. In order to survive he recommenced
working as a teacher, first (1813) for a school in Bamberg, then in 1817 at
the Dreikönigs Grammar School in Cologne, and in 1826 at a military
school in Berlin. His main research interest belonged to electricity, a hot subject at his times. In
1833 he obtained a slightly better paid position as a docent at the Royal Polytechnic School in
Nuremberg, where he became a director in 1839 and which bears his name today. In 1849 he was
appointed at the University of Munich, first as an associate professor and, finally, in 1852 as a full
professor for experimental physics. What a long way to go for the remaining short amount of life.
References
P
M PETIT, Alexis, 3
MACH, Ernst, 210 Plane strain, see Strain
Mass, balance of, 51, 72–74, 84, 107–109 Plane stress, see Stress, state of
MAXWELL fluid, 259–266 Point
MAXWELL, James, 5 regular, 55, 70–83, 107–125, 312–313,
MAXWELL-LORENTZ aether relations, 356–359, 345–347, 354, 356, 380
364–366, 371–380 singular, 55–56, 71–72, 83–85, 125–126,
MAXWELL’S equation, 331–339, 342–348, 356, 346–347
363–366 material, 48
MAXWELL stress tensor, 382 POISSON, Siméon, 132
Metric, see Tensor, metric POISSON’s number, 132–134, 225, 385
398 Subject Index
Polarization, 352–354, 362, 382–384 singular, 48–50, 55–59, 70–72, 83, 125,
Potential 126, 338, 339, 342–344
charge, 343, 344, 348–352, 353, 361–362, Susceptibility, 383–388
371
chemical, 319
current, 332–333, 343, 348–352, 355–356, T
361–363, 371, 388 Tangent modulus, plastic, 299
electric, 341–342 TAYLOR series, 155, 236
function, 310 TAYLOR, Brook, 155
POYNTING, John, 381 Tensor, 6–14, 17–21, 36–42, 97–100, 187–194
PRANDTL, Ludwig, 298 curvature, 43, 45
PYTHAGORAS of Samos, 16 dielectricity, 383
metric, 18–32, 40, 43–46, 91, 98, 204
piezoelectric constants, 384, 386
R relative permittivity, 383, 386
Radiation density, 77, 119, 208, 209 strain, 8, 32, 131, 135–137, 189, 385
specific, 315 stress, 39, 40, 52, 53, 114, 115, 130, 189,
Reference configuration, 60–61, 63, 67–68, 198, 308
131, 230–231 susceptibility, 383, 385
Relative motion, 59 world, 367, 373
REUSS, Endre, 298 Thermal expansion coefficient, 10, 11, 131,
REYNOLDS, Osborne, 49 132, 150, 226
Thermodynamics, 6, 140–151
first law, 77, 122
S Irreversible Processes (TIP), 314, 326
SAVART, Félix, 359 THIRRING, Hans, 204
Scalar, 6, 90, 117 THOMSON, William, see KELVIN
product, 8, 22, 25, 31, 35, 41, 42, 53, 59, Time derivative, 6, 56, 65, 66, 209–211
75, 79, 83 JAUMANN, 210
SCHWARZ, Hermann, 311 LIE, 212
SCHWARZ, rule of, 100, 310, 311, 325 material, 72, 111, 177, 196, 208, 211, 363,
Semi-inverse method, 133, 161, 163 364
Shear angle, 135 OLDROYD, 212
Spin, 78, 80, 82 TRUESDELL, 212
balance of, 82, 123 Transition condition, 153
Stiffness matrix, 6, 8, 131, 301 Transport theorem, 56–68, 126, 127, 336–339
STOKES, George, 3 REYNOLD’s, 59, 74, 156
STOKES theorem, 336–338 TRUESDELL III Clifford, 212
Strain, 6, 8, 32, 131, 135, 137, 189, 233,
298–300, 385
Strain, equivalent plastic, 302–304 V
Strain tensor, 8, 32, 131, 135–137, VAN DE GRAAFF, Robert, 332
189, 385 Vector, 6–13, 22–26, 53–54, 89–105, 360–362
Stress, 6 angular velocity, 190–192, 205, 369
axial, 221–224 current density, 331–333, 342, 354, 368,
compressive, 223, 294, 295, 297 388
shear, 40, 134, 135, 251–256 field, 58, 90–95, 95, 336
tensile, 7, 132, 133, 223, 248, 294, 295 flux density, 77, 118, 122, 151, 179, 208
thermal, 10–13, 224–226 four, 366–381
yield, 40, 41, 289–297, 299–300, 302 heat flux, 77, 118, 122, 151, 179, 208
Stress intensity factor, 248, 249 objective, 197
Stress, state of, 33, 40, 216, 232, 289, 290 polarization, 352, 353, 362, 382, 383
Stress-strain diagram, 299 POYNTING, 381
Surface product, 6, 36, 63, 68, 79, 191, 197, 360
Subject Index 399
Viscosity Y
bulk/volumetric, 138, 171, 172 Yield criterion, VON MISES, see Flow rule, VON
dynamic, 168 MISES
shear, 138, 165, 167, 252 YOUNG’s modulus, see Modulus of elasticity
VOIGT notation, 387, 388 YOUNG, Thomas, 9
VOIGT, Woldemar, 388
VON MISES equivalent stress, 40, 41, 299