Fundamentals of Physics

Institut für Grundlagen (IFG) & Institute for Flow in Additively

Manufactured Porous Media (ISAPS)

Physics
Basic Knowledge

Prof. Dr. rer. nat. Markus Scholle

Fundamentals of Physics
Contents

0 Introduction 9
0.1 Experiment and theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
0.2 Physical quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
0.3 Measurement uncertainties (measurement statistics) . . . . . . . . . . . . . 11
0.3.1 Measurement errors and uncertainties . . . . . . . . . . . . . . . . . 11
0.3.2 Measurement statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 12
0.3.3 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
0.4 Taylor expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
0.4.1 First-order Taylor expansion . . . . . . . . . . . . . . . . . . . . . . 14
0.4.2 Second and higher order Taylor expansion . . . . . . . . . . . . . . . 15

I Mechanics 19

1 Kinematics 21
1.1 Kinematics of mass points . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.1.1 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.1.2 Cartesian coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.1.3 Polar coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.1.4 Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.1.5 Other coordinate systems . . . . . . . . . . . . . . . . . . . . . . . . 28
1.2 Kinematics of rigid bodies for motion in a plane . . . . . . . . . . . . . . . . 29
1.2.1 Degrees of freedom of rigid bodies in a plane . . . . . . . . . . . . . 30
1.2.2 Essential kinematic quantities . . . . . . . . . . . . . . . . . . . . . . 30

2 Dynamics of mass points 35

2.1 Newton’s laws of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.1.1 Newton’s first law (Galilei’s principle of inertia) . . . . . . . . . . . . 35
2.1.2 Newton’s second law (law of motion) . . . . . . . . . . . . . . . . . . 35
2.1.3 Newton’s third law (action-reaction law) . . . . . . . . . . . . . . . . 35
2.2 Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.1 Active forces and reaction forces . . . . . . . . . . . . . . . . . . . . 36
2.2.2 Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.3 Spring forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.4 Frictional forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.5 Lorentz force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.1 Fall of a body with friction . . . . . . . . . . . . . . . . . . . . . . . 39

3
 Fundamentals of Physics

2.3.2 The simple gravity pendulum . . . . . . . . . . . . . . . . . . . . . . 41

3 Constants of motion 49
3.1 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.1 Examples of energy conservation . . . . . . . . . . . . . . . . . . . . 51
3.2 Momentum and centre of mass . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.1 Examples of conservation of momentum . . . . . . . . . . . . . . . . 52
3.2.2 Example of the centre-of-mass theorem . . . . . . . . . . . . . . . . 54
3.3 Law of conservation of angular momentum . . . . . . . . . . . . . . . . . . . 54
3.3.1 Examples of conservation of angular momentum . . . . . . . . . . . 55

4 Dynamics of rigid bodies for motion in a plane 59

4.1 The Newton-Euler equations of motion . . . . . . . . . . . . . . . . . . . . . 59
4.2 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Energy and the parallel axis theorem . . . . . . . . . . . . . . . . . . . . . . 63
4.3.1 Kinetic energy of rotation . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3.2 The parallel axis theorem . . . . . . . . . . . . . . . . . . . . . . . . 63

II Thermo- and Fluid Dynamics 67

5 Thermodynamics 69
5.1 States of matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Pressure and temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.1 Kinetic model of pressure . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.2 Kinetic model of temperature . . . . . . . . . . . . . . . . . . . . . . 72
5.2.3 Degrees of freedom in molecules . . . . . . . . . . . . . . . . . . . . 73
5.3 Maxwell–Boltzmann distribution . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4 Internal energy and heat capacity . . . . . . . . . . . . . . . . . . . . . . . . 74
5.5 Classification of thermodynamic quantities . . . . . . . . . . . . . . . . . . . 75
5.5.1 Extensive and intensive quantities, specific and molar quantities,
densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.5.2 State variables, state functions and state space . . . . . . . . . . . . 75
5.5.3 Processes and process-dependent quantities . . . . . . . . . . . . . . 76
5.6 The laws of thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.6.1 Temperature (zeroth law) . . . . . . . . . . . . . . . . . . . . . . . . 78
5.6.2 Energy (first law) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.6.3 Entropy (second law) . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.6.4 The Nernst heat theorem (third law) . . . . . . . . . . . . . . . . . . 81
5.7 Selected processes of ideal Gases . . . . . . . . . . . . . . . . . . . . . . . . 81
5.7.1 The T -S-diagram as a means of representation . . . . . . . . . . . . 81
5.7.2 Isotherms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.7.3 Adiabats / isentropes . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.7.4 Isochores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.7.5 Isobars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.8 Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6 A glimpse of fluid dynamics 89

6.1 Historical overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 The “ideal” fluid and its flow . . . . . . . . . . . . . . . . . . . . . . . . . . 90
 Fundamentals of Physics

6.2.1 Bernoulli’s principle . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

6.2.2 Continuity equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3 Navier Stokes equation (2D) for viscous flows . . . . . . . . . . . . . . . . . 95
6.4 Fluid-structure interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.4.1 Interfacial stresses and resultant forces . . . . . . . . . . . . . . . . . 95
6.4.2 Determining flow forces indirectly from the momentum balance . . . 96
6.4.3 Example: The splitting of a free jet by a wedge . . . . . . . . . . . . 96
6.4.4 Buoyancy forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

III Systems 101

7 Variational methods and the Lagrangian formalism 103

7.1 Simple optimisation problems . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.1.1 The lifeguard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.1.2 The fire engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.2 Fermat’s principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.2.1 Transition to the continuum . . . . . . . . . . . . . . . . . . . . . . . 105
7.2.2 The basic concept of the finite element method . . . . . . . . . . . . 107
7.3 The Lagrangian formalism in classical mechanics . . . . . . . . . . . . . . . 107
7.3.1 Hamilton’s principle . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.3.2 Euler-Lagrange equations . . . . . . . . . . . . . . . . . . . . . . . . 108
7.3.3 Variational problems with constraints . . . . . . . . . . . . . . . . . 111
7.4 The Rayleigh-Ritz method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
7.5 Motion under the influence of friction or other factors . . . . . . . . . . . . 115
7.5.1 Lagrangesche Gleichungen zweiter Art . . . . . . . . . . . . . . . . . 115
7.5.2 Strict variational formulations for systems with non-conservative
forces and for continuum mechanics . . . . . . . . . . . . . . . . . . 117

8 Autonomous oscillations 121

8.1 The fundamentals of oscillation analysis . . . . . . . . . . . . . . . . . . . . 121
8.1.1 The characterisation of oscillations . . . . . . . . . . . . . . . . . . . 121
8.1.2 The oscillation equation . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.2 Examples of free undamped oscillations . . . . . . . . . . . . . . . . . . . . 123
8.2.1 Mass-on-spring oscillator . . . . . . . . . . . . . . . . . . . . . . . . 123
8.2.2 LC circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.3 Damped free oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.3.1 Linear damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

9 System dynamics aspects 129

9.1 Forced oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
9.1.1 The general form of the equation of motion . . . . . . . . . . . . . . 129
9.1.2 Possible types of excitation . . . . . . . . . . . . . . . . . . . . . . . 129
9.1.3 The resulting motion and resonance curve . . . . . . . . . . . . . . . 130
9.2 Coupled oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
9.2.1 The double oscillator and its application as a tuned mass damper . . 132
9.2.2 A chain of rotating bodies . . . . . . . . . . . . . . . . . . . . . . . . 135

A Solutions of the Exercises 139

A.0 Regarding measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
 Fundamentals of Physics

A.0.1 Vereinfachte Berechnung der Standardabweichung . . . . . . . . . . 139

A.0.2 Bestimmung der Masse von Weinbeeren . . . . . . . . . . . . . . . . 139
A.0.3 Freitag der 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
A.0.4 Lineare Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
A.0.5 Fehlerfortpflanzung . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
A.1 Regarding kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
A.1.1 Flussüberquerung . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
A.1.2 Und sie bewegt sich doch . . . . . . . . . . . . . . . . . . . . . . . . . 145
A.1.3 Neulich am Hambacher Forst . . . . . . . . . . . . . . . . . . . . . . 146
A.1.4 Flugbahn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
A.1.5 Kinematik starrer Körper . . . . . . . . . . . . . . . . . . . . . . . . 147
A.1.6 Neulich im Parkhaus . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
A.2 Regarding dynamics of mass points . . . . . . . . . . . . . . . . . . . . . . . 148
A.2.1 Bewegung gekoppelter Massen . . . . . . . . . . . . . . . . . . . . . 148
A.2.2 Bewegung gekoppelter Massen mit Schmierung . . . . . . . . . . . . 149
A.2.3 Flaschendrehen-Variante . . . . . . . . . . . . . . . . . . . . . . . . . 150
A.3 Regarding constants of motion . . . . . . . . . . . . . . . . . . . . . . . . . 152
A.3.1 Vollständig inelastischer Stoß . . . . . . . . . . . . . . . . . . . . . . 152
A.3.2 Vollständig elastischer nichtzentraler Stoß . . . . . . . . . . . . . . . 152
A.3.3 Beidseitiger elastischer Anschub . . . . . . . . . . . . . . . . . . . . 154
A.3.4 Wackeliges Pendel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
A.3.5 Schifoan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
A.4 Dynamics of rigid bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
A.4.1 Neulich beim Discounter . . . . . . . . . . . . . . . . . . . . . . . . . . 157
A.4.2 Neulich im Baumarkt . . . . . . . . . . . . . . . . . . . . . . . . . . 158
A.4.3 Massenträgheitsmoment(e) einer T-Anordnung . . . . . . . . . . . . 160
A.5 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
A.5.1 Allgegenwärtige Energie . . . . . . . . . . . . . . . . . . . . . . . . . 160
A.5.2 Systematik in der Thermodynamik . . . . . . . . . . . . . . . . . . . 161
A.5.3 Isobare Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
A.5.4 Kalorimetrie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
A.5.5 Zwei-Kammer-System . . . . . . . . . . . . . . . . . . . . . . . . . . 162
A.5.6 Halber Otto-Zyklus . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
A.5.7 Adiabatenexponent idealer Gase . . . . . . . . . . . . . . . . . . . . 165
A.5.8 Thermodynamischer Prozess . . . . . . . . . . . . . . . . . . . . . . 165
A.6 Fluid dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.6.1 Staupunkt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.6.2 Fallender Wasserstrahl . . . . . . . . . . . . . . . . . . . . . . . . . . 167
A.6.3 Thermo- und Fluiddynamik . . . . . . . . . . . . . . . . . . . . . . . 168
A.7 Regarding variational calculus . . . . . . . . . . . . . . . . . . . . . . . . . . 169
A.7.1 Konkavspiegel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
A.7.2 Rutschender Stab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
A.7.3 Das Seilproblem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
A.7.4 Mehrkörperproblem . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
A.7.5 Perle auf einer dünnen Schnur . . . . . . . . . . . . . . . . . . . . . 172
A.8 Regarding autonomous vibrations . . . . . . . . . . . . . . . . . . . . . . . . 173
A.8.1 Spezielle ungedämpfte Schwingungen . . . . . . . . . . . . . . . . . . 173
A.8.2 Spezielle gedämpfte Schwingungen . . . . . . . . . . . . . . . . . . . 174
A.8.3 Bewegung im aperiodischen Grenzfall . . . . . . . . . . . . . . . . . 175
CONTENTS

A.8.4 Thermodynamik und Schwingungen . . . . . . . . . . . . . . . . . . 177

A.9 Regarding system dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
A.9.1 Drehschwingerkette mit N = 3 . . . . . . . . . . . . . . . . . . . . . 178
A.9.2 Eigenmoden des Zweimassenschwingers . . . . . . . . . . . . . . . . 179
A.9.3 Schwingungen — Gekoppelte Schwingungen . . . . . . . . . . . . . . 180

B Supplements 183
B.1 About Taylor expansion of second and higher order . . . . . . . . . . . . . . 183
B.1.1 Taylor expansion of second order . . . . . . . . . . . . . . . . . . . . 183
B.1.2 Taylor expansion of higher order . . . . . . . . . . . . . . . . . . . . 183
B.1.3 Taylor series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
B.2 About curvilinear coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 185
B.2.1 Spherical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
B.2.2 Natural coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
B.3 Selected mathematical problems . . . . . . . . . . . . . . . . . . . . . . . . 190
B.3.1 Obtaining the solution for the body falling under friction form Eq.
(2.18) for n = 2 (Newton case) . . . . . . . . . . . . . . . . . . . . . 190

7
 CONTENTS

8
Chapter 0

Introduction

Physics is a quantitative science. Its focus is primarily on the acquisition and assessment
of data on the properties that determine a system. This can be done via calculations in
the context of a theory or through measurements as part of an experiment.

0.1 Experiment and theory

Experiment and theory are the two pillars of physics and must meet strict requirements:
Experiments must be reproducible so that their results can be confirmed at any time,
in any place and by any experimenter. Due to the imperfection of every measure-
ment principle, however, this is only possible down to (calculable) measurement
uncertainties.
Theories are measured against the predictions that can be made from them. In many
cases, this requires modelling the problem at hand using idealised assumptions, so
that theoretical predictions, too, are usually fraught with uncertainty.
Theories and simulations are often not appreciated very much, but their importance be-
comes apparent especially in the current situation: the spread of the Covid19 virus can
be predicted by theoretical models. These predictions can then be used to decide which
countermeasures are necessary and appropriate. But also with regard to the development
of the global climate in the following decades, reliable predictions are invaluable in order to
be able to quantitatively assess the necessity and effectiveness of possible countermeasures.
Speaking of disaster prevention: our solar system is teeming with asteroids, and some oc-
casionally cross the Earth’s orbit. Impact events are rare, but devastating on an unimag-
inable scale. One impact 65 million years ago was instrumental in the disappearance of
dinosaurs from this planet, along with 95% of all species! But there were also smaller
impacts that happened not so long ago. The Tunguska event in 1908 was, presumably,
an event of this kind. Thanks to powerful telescopes, we can catalogue more and more
celestial bodies (a credit to experimental physics), and thanks to reliable theoretical mod-
els, we can calculate their orbits over many decades in advance and thus detect a possible
impact on our planet well ahead of time. Now you might think “In this case, wouldn’t it
be better not to know this in advance?” Well, it would, if it were unpreventable – and it
would be unpreventable if we only noticed it shortly before it happens! But if we knew
about the impending impact decades beforehand, all we would have to do is give the as-

9
 0.2. PHYSICAL QUANTITIES

teroid a “little nudge”, perhaps in the form of a massive probe deliberately slamming into
it. The minimal change of course could be sufficient if this measure were taken early on!

0.2 Physical quantities

A physical quantity X is given in multiples of a defined standard, the magnitude of which
is referred to as its unit (of measurement) [X]. The ratio of the quantity to the unit of
measurement, i.e.
X
{X} =
[X]
is called measured value. If we rearrange this equation, we can write:

X = {X}[X] , (1)

for example: F = 21,6N.

All units can be represented as a product of integer powers of base units. In the Interna-
tional System of Units (SI), these are compiled in Table 1.

Physical quantity Unit Name

time s second
length m metre
mass kg kilogram
temperature K kelvin
electric current A ampere
luminous intensity Cd candela
amount of substance mol mole
Table 1: The seven SI base units.

Examples of compound units:

Force 1N = 1kgm/s2
Work 1J = 1Nm = 1kgm2/s2
Electric charge 1C = 1As
Voltage 1V = 1J/C = 1kgm2/As3
For very large and very small measured values, one often uses unit prefixes instead of
powers of ten. These prefixes are placed in front of the respective unit. For example:
• 1km = 1000m
• 1µF = 10−6 F
• 1nm = 10−9 m
Units of time are an exception:
• 1min = 60s
• 1h = 60min = 3600s
• 1d = 24h = · · ·

10
0.3. MEASUREMENT UNCERTAINTIES (MEASUREMENT STATISTICS)

The reasons for not using powers of ten here are of a cultural-historical nature: We owe
the decimal system to the Romans, whose cultural heritage we adopted. However, time
is such an essential quantity for human life that it was measured long before the Roman
Empire, i.e. before the introduction of the decimal system.
Angles are preferably given as radian (= arc length / radius). Actually the unit is 1, but
people often (still) write rad in order to indicate an angle. The commonly used degree is
defined as:
π
1deg = 1° = (2)
180
The principle of dimensional homogeneity applies, i.e. only quantities of the same type
can be added together. For example:
• 10m/s + 72km/h = 30m/s is dimensionally homogeneous.
• 10m/s + 72m2 , however, does not make sense!
The principle of dimensional homogeneity helps identify errors in calculations: If the
principle is violated, you are guaranteed to have miscalculated and usually you will have
a hot lead on how to find the mistake quickly!
Quantities that have the unit 1 are referred to as dimensionless. These are quantities that
are simply represented by numbers. Angles are dimensionless, and naturally also ratios of
two physical quantities of the same dimension. Radicands, logarithms and the arguments
of all functions that are not pure integer powers must be dimensionless.

0.3 Measurement uncertainties (measurement statistics)

0.3.1 Measurement errors and uncertainties
There are many reasons why the actual value xw of the measured quantity can differ from
the measured value x. Generally, their effects are summed up as the error

∆x = x − xw

or, more specifically, as the absolute deviation. The relative error is given by

∆x x − xw
= .
x xw

The error ∆x is a quantity, in which contributions from various causes are summed up. A
distinction is made between the following types of error:
1. Gross errors for which the experimenter themselves is responsible due to lack of care.
2. Systematic errors based on influences not taken into account (hysteresis, distur-
bance variables, . . . ). These reappear in unchanged form when the measurement is
repeated.
3. Random errors which are based on stochastic disturbances (“noise”) and are therefore
unavoidable. However, their influence can be reduced by repeated measurement and
calculation of the mean value.
Since the true value xw is naturally unknown, the error ∆x, too, can never be determined
with certainty, but can only be estimated. The estimated error is called uncertainty.

11
 0.3. MEASUREMENT UNCERTAINTIES (MEASUREMENT STATISTICS)

0.3.2 Measurement statistics

For a quantity x, N independent and usually different measurement data x1 , x2 , · · · , xN
are collected by repeated measurement.
The (relative) frequency is given by
occurrences of xi ∈ [a, b]
p[a, b] = (3)
N
and denotes the proportion of the values located within an interval [a, b] to the total number
of measurements. If the range within which measured values (can) occur is divided into M
equally sized subintervals [an , an+1 ], with an+1 − an = ∆x, the corresponding frequencies
pn := p[an , an+1 ] can be plotted against x in a histogram, as shown in Fig. 1. An essential

pn

n
...
1234 M
Figure 1: A histogram visualises the distribution of measured values.

property of the frequency distribution is its normalisation

M
pn = 1, (4)
X

n=1

which necessarily results from its definition (3). If we let ∆x approach zero and at the
same time N → ∞, then the histogram, as is also indicated in Fig. 1, changes into a
smooth curve p(x) and the normalisation (4) becomes:
ˆ∞
p(x)dx = 1 (5)
−∞

The question arises, what the “true” value of the measured quantity x is. Since it can never
be determined objectively with absolute certainty, one must content oneself with narrowing
it down by specifying the problem to first find the value that most likely corresponds to
the true value, and then specify an interval around this value in which the true value lies
with a high degree of probability.
The best “candidate” for the value that most likely corresponds to the true value is not
necessarily one of the measured values xi , but the (arithmetic) mean value
N
1 X
x̄ = xi , (6)
N i=1

12
0.3. MEASUREMENT UNCERTAINTIES (MEASUREMENT STATISTICS)

whose reliability is given by the standard deviation

v s
N
1 X
u
N
σ=t (xi − x̄)2 = (x − x̄)2 (7)
u
N − 1 i=1 N −1

If the error is of a purely statistical nature, then for a very large number of measurements
the frequency distribution changes to a Gaussian distribution 1 :

1 (x − x̄)2
!
p(x) = √ exp − (8)
2πσ 2σ 2

If this were not the case, one would have an indication of an unrecognised systematic error.
The role of the variance σ is that with x̄ ± σ there is an interval defined in which approx.
68% of all measured values are located. Thus the variance determines the uncertainty of
a single measurement.

In contrast, the uncertainty of the mean value decreases the more often the measurement
is repeated. Here, the root-like relationship
σ
∆x = √ (9)
N

applies and we use the notation x = x̄ ± ∆x.

0.3.3 Linear regression

In physics, there are numerous examples of linear relationships of the form y = ax + b
between two quantities x and y. The experimental validation of such correlations as well
as the identification of the two relevant parameters, the slope a and the y-intercept b, is
one of the standard tasks of the experimenter.

If one starts from a set of value pairs (xi , yi ) obtained through measurements, these do not
result in an exact straight line after they have been entered into a diagram, as exemplified
in Fig. 2, due to measurement errors. The question now is how to find a suitable line of

y
y = ax + b

x
Figure 2: Representation of measured value pairs (xi , yi ) as points in a diagram (blue),
together with the line of best fit (red).

best fit that is as close as possible to the measured values. Its construction “by eye” is
a legitimate but, nevertheless, uncertain method and, above all, not independent of the
subjective view of the person doing it.
1
The mathematical proof of this is yet to be provided.

13
 0.4. TAYLOR EXPANSION

The method of least squares is an objective alternative: In this method, one chooses a and
b in such a way that the sum
N
S (a, b) = [axi + b − yi ]2
X

i=1

becomes minimal. Generally, one can find extrema as the zeros of the first derivatives
with respect to a and b, i.e.
N
∂S
0= =2 xi [axi + b − yi ]
X
∂a i=1
N
∂S
0= =2 [axi + b − yi ]
X
∂b i=1

which, after simple transformations, leads to the system of equations

N N N
! !
a+ xi b = (10)
X X X
x2i x i yi
i=1 i=1 i=1
N N
!
xi a + N b = (11)
X X
yi
i=1 i=1

. Solving this system of equations2 for the quantities sought yields

! ! ! !
N N N N N N N
yi − N x i yi − x2
P P P P P P P
xi x i yi xi yi i
i=1 i=1 i=1 i=1 i=1 i=1 i=1
a= !2 , b= !2 (12)
N N N N
−N x2i −N x2i
P P P P
xi xi
i=1 i=1 i=1 i=1

as ready-made equations for the slope and the y-intercept of the line of best fit.

0.4 Taylor expansion

It often makes sense to simplify complex problems in order to improve their solvability. Be-
sides descriptive physical considerations about the actual relevance of physical effects (e.g.
friction), there is the possibility of a systematic analysis using mathematical means. Its
most important tool is the Taylor expansion, which allows approximation of any function
by simple polynomials.

0.4.1 First-order Taylor expansion

A function f (x) is to be approximated by a first-order polynomial, i.e. a straight line of
the form
p1 (x) = ax + b (13)
around the point x0 , where a and b have yet to be suitably chosen. Looking at the graph,
N 
2
P
z.B. so: Through the linear combination xi · eq. (11) − N · Gl. (10) b is eliminated and one can
i=1
solve for a. The result is then applied to (11) in order to obtain b.

14
0.4. TAYLOR EXPANSION

f (x)
6 @
@
@
@
@
@
f (x0 ) @b`
@
@
@
@
@
@
@
x
-
x0
Figure 3: Function with tangent as a first-order Taylor polynomial.

it becomes clear that the best approximation of this kind is given by the tangent at the
point x0 , cf. Fig. 3, which satisfies the two conditions

p1 (x0 ) = f (x0 )
p′1 (x0 ) = f ′ (x0 )

This means that both its value and its slope coincide with those of the function to be
approximated. After plugging in (13), we obtain

ax0 + b = f (x0 ) ,
a = f ′ (x0 )

as a linear system of equations for a and b, whose solution is a = f ′ (x0 ) and b = f (x0 ) −
f ′ (x0 ) x0 . Applying this to the original approach (13) then yields

f (x) ≈ p1 (x) = f (x0 ) + f ′ (x0 ) (x − x0 ) (14)

as the approximation formula we are looking for.

0.4.2 Second and higher order Taylor expansion

A better approximation could be achieved with a curved line that fits the graph of the
function. In the simplest case, this can be a parabola. Still better results can be achieved
with cubic functions or even higher-order polynomials.
This idea is pursued in Appendix B.1 and will be applied in later chapters.

Exercises
Exercise 0.1: Simplified calculation of the standard deviation
(Solution cf. A.0.1)
Prove that the standard deviation defined according to equation (7) can alternatively be
written as: s
N  2 
σ= x − x̄2 . (15)
N −1
Why is this way of calculation less prone to error when using a pocket calculator than the
original formula (7)? Explain also, why x2 is always greater than x̄2 .

15
 0.4. TAYLOR EXPANSION

Exercise 0.2: Determination of the mass of grapes

(Solution cf. A.0.2)
Twelve grapes are weighed individually with a (relatively simple) kitchen scale, the accu-
racy of which is 1g. Finally, the total mass of all the grapes together is determined. The
results are summarised in the following table:
i 1 2 3 4 5 6 7 8 9 10 11 12 all
mi /g 2 5 3 4 4 3 4 2 4 3 3 4 47
The last column contains the independent measurement of the total mass.
1. As you can see, only the four different values 2g, 3g, 4g and 5g occur. Give their
relative frequencies and make a histogram.
2. Calculate the mean m̄ of the twelve individual measurements.
3. Calculate the corresponding standard deviation σ.
4. Infer the uncertainty ∆m̄ of the mean vand write down the result in the form m̄±∆m̄
(round sensibly and do not forget the unit!).
5. Add the mean m̄, standard deviation σ and a plot of the corresponding Gaussian
distribution function to the histogram. Does the latter represent the frequency
distribution of the measured values well?
6. Alternatively, determine the mean mass m̄∗ of a grape from the measured total mass
and also give the result with the corresponding uncertainty. The latter results from
the display accuracy of the balance.
7. Are the two results from the individual measurements and the measurement of the
total mass consistent with each other? How do you explain the discrepancy? Which
of the two results is more trustworthy?

Exercise 0.3: Friday the 13th

(Solution cf. A.0.3)
The following table (source: Wikipedia) lists all the months on which the 13th is a Friday,
using the years 2001 to 2028 as an example.
2001 2007 2018 April, July
2002 2013 2019 2024 September, December
2003 2008 2014 2025 June
2020 March, November
2009 2015 2026 February, March, November
2004 February, August
2010 2021 2027 August
2005 2011 2016 2022 May
2028 October
2006 2017 2023 January, October
2012 January, April, July
As can be seen from this table, there is at least one and at most three Fridays in each
year, the 13th.
1. Find the three relative frequencies p1 , p2 , p3 for one, two and three occurrences
respectively within a year for the above-mentioned time span.

16
0.4. TAYLOR EXPANSION

2. How often does Friday the 13th occur on average per year?
3. What is the standard deviation σ?
4. If you create a histogram from the frequencies, it looks nothing like a Gaussian
distribution (try it :). What is the reason for this?
◦ This is simply due to the small number of data. If one were to look at a
complete millennium instead of 28 years, the distribution would be quite close
to Gaussian statistics.
◦ The frequency of Friday, the 13th, is not based on chance, but is determinis-
tically predetermined by the Gregorian calendar. Gaussian statistics, on the
other hand, only apply to random events.
◦ Every rule has its exception: Other combinations of weekday and date (e.g. Mo.
the 9th) largely satisfy Gaussian statistics, only Friday the 13th “dances out of
line”. This conspicuousness, which cannot be explained by rational arguments,
is the reason for Friday the 13th’s bad reputation as an unlucky day.

Exercise 0.4: Linear regression

(Solution cf. A.0.4)
It is said that the Covid19 virus spreads exponentially. We check whether this hypothesis
is correct using data from the Robert Koch Institute. The Robert Koch Institute has
announced the following figures for cumulative number of infections for the time period
from 5 March to 15 March at the beginning of the pandemic in 2020:
t/d n
5 400
6 639
7 795
8 902
9 1139
10 1296
11 1567
12 2369
13 3062
14 3795
15 4838
Note that up to this point the German government did not yet adopted measures to
prevent spreading of the infection.
1. Assuming that the time evolution of the number of infections obeys a exponential
relationship n(t) = n0 exp (γt). Then by applying the natural logarithm, one receives
the linear relationship:
ln (n(t)) = ln (n0 ) + γt .

Use the linear regression described in section 0.3.3 to determine ln (n0 ) and γ.
2. Contrast the graph of the function ln (n(t)) = ln (n0 ) + γt with the values from the
Robert Koch Institute and use the comparison to decide whether the assumption of
exponential spread is justified.

17
 0.4. TAYLOR EXPANSION

3. What is the forecast for the cumulative number of infections for 21 March (the
calendar start of spring)?

Exercise 0.5: Error propagation

(Solution cf. A.0.5)
You measure the edge length a of a cube and get the result:

a = (9,35 ± 0,05) cm

1. What is the relative uncertainty of the measurement?

2. From this, determine the area of one side of the cube including the uncertainty.
3. Also determine the volume of the cube, also including the uncertainty.
4. Consider in general the problem that a quantity y is calculated via a relation y = f (x)
from a measured quantity x, which is subject to uncertainty according to x = x̄ ±
∆x. How can we then use the Taylor expansion to infer the uncertainty ∆y of the
calculated quantity y?

18
 Part I

Mechanics

19
Chapter 1

Kinematics

The aim of kinematics is to describe the motion of bodies. This can be the motion of a
single mass point, the motion of several or even many mass points, the change in position
and orientation of a rigid body, the deformation of a solid and even the flow of a fluid.
Depending on the type of mechanical system, different mathematical means must be used
to describe motion. However, it is not the task of kinematics to explain a motion’s cause.
This is, in fact, the task of dynamics.

1.1 Kinematics of mass points

1.1.1 Basic concepts
A body is considered as a mass point as long as its shape and extension do not influence its
motion. Then it is assumed that the body’s mass is essentially concentrated in one single
point, which corresponds to the center of mass of the actual body. in fact the absolute size
of the body is not decisive whether modelling it as a mass point is justified. Intuitively,
you would consider the Earth as a large body. However, if you compare the Earth’s orbit
around the sun, the radius of which (1 AU≈ 150000000 km) is significantly larger than the
Earth’s radius (≈ 6400 km), we can certainly consider the Earth as a mass point.
In three–dimensional space and in order to determine the position of a mass point at a
given point in time t, you need a position vector ⃗r at whose apex, seen from the origin,
the mass point is located. Since the position of the mass point changes as time passes, ⃗r
is a function of t, i.e. ⃗r = ⃗r(t). The path of a mass point is the total of all positions ⃗r(t)
the mass point traverses as time passes, see Fig. 1.1.
The velocity ⃗v (t) is the change of the position per time unit. Figure 1.2 shows how this
quantity is to be determined. Evidently, ⃗r(t + ∆t) − ⃗r(t) is the vector that points from
the mass point’s current position at time t to its future position at time t + ∆t. To obtain
the velocity, you have to divide this vector by the time ∆t that has passed, as is shown in
Fig. 1.2. The vector constructed in this way gives us the mean velocity in the considered
time interval [t, t + ∆t]. We obtain the instantaneous velocity by finding the limit ∆t → 0.
Thus we obtain the velocity by finding the derivative of the position vector with respect
to time, i.e.
d
⃗v (t) := ⃗r (t) = ⃗r˙ (t) , (1.1)
dt
where here and below the derivative with respect to time will be denoted with a dot above

21
 1.1. KINEMATICS OF MASS POINTS

y
6

`x
h
3

 ~
⃗r (t) 







 x
-
Figure 1.1: Position vector and path of a mass point.

y
6

`h
x
`h
x
3

⃗r (t)   *
 
 R
  ⃗ r (t + ∆t)

 ⃗v (t)





 x
-
Figure 1.2: Calculation of the velocity of a mass point.

the respective quantity. Velocity is a vector by definition. Its direction is always tangential
to the path and thus indicates the current direction of motion. Its magnitude

⃗r˙ 2
p
|⃗v | = (1.2)

is called path velocity, sometimes also speed, and indicates the length of the arc travelled
along the path per unit of time. The magnitude and direction of ⃗v characterise the current
state of motion of the mass point.
Especially in dynamics, the change of the state of motion with time is of great importance.
This quantity is called acceleration and is obtained, as shown in Fig. 1.3, by calculating
the difference vector ⃗v (t + ∆t) − ⃗v (t) between the velocities at two points in time close
to each other and dividing it by the time ∆t that has passed, i.e.
⃗v (t + ∆t) − ⃗v (t)
.
∆t
This gives us the mean acceleration in the time interval ∆t. The instantaneous acceleration
is obtained by finding the limit ∆t → 0 and, in this way, leads to a differentiation with
respect to time, i.e.
d
⃗a(t) := ⃗v (t) = ⃗v˙ (t) = ⃗r¨(t) . (1.3)
dt
In contrast to velocity, which is oriented in a direction tangential to the path, the direction
of acceleration, generally, is not subject to any restrictions and thus can point in any
direction, depending on the type of motion.

22
1.1. KINEMATICS OF MASS POINTS

y
6
M
⃗a (t)
bP
qb
ZPP
Z P
Z
`
x
h ~b
Z
Z
Z `P
x
h
Z
Z PPP q
Z
~
Z ⃗v (t + ∆t)
⃗v (t)
x
-
Figure 1.3: Graphical illustration for calculating the acceleration of a mass point. For
better clarity, the two velocity vectors are translated to a point off the curve in order to
construct the acceleration vector there.

The differentiation of the path velocity with respect to time

d d ˙
apath := |⃗v (t)| = ⃗r(t) , (1.4)
dt dt
the so-called path acceleration, is also important. Generally, this scalar quantity is not
identical with the magnitude of the acceleration vector, but is related to it in the following
way:
d d√ 2 2⃗v · ⃗v˙ ⃗v
apath = |⃗v | = ⃗v = √ = · ⃗a (1.5)
dt dt 2 ⃗v 2 |⃗v |
Since ⃗v / |⃗v | is the direction vector of the tangent to the path, we find that apath is the
tangential component of the acceleration. Therefore, it is also called tangential acceleration
and indicates whether the mass point on its path becomes faster or slower. In contrast,
the other part of the acceleration
⃗v
⃗a⊥ := ⃗a − apath (1.6)
|⃗v |

that is always perpendicular to the path is called normal acceleration. It represents a mere
change in the direction of the motion without changing the speed of the mass point.
The third derivative of the position vector with respect to time, i.e.
...
⃗j := d⃗a = ⃗a˙ = ⃗v¨ = ⃗r (1.7)
dt
is called jerk. However, this vector is only rarely used, since the equations of motion in
dynamics contain derivatives of the position vector up to the second order.
The above definitions and statements are valid in general, i.e. they apply to any coordinate
system. A suitable coordinate system should always be chosen with regard to the respective
problem. The most frequently used coordinate systems will be covered in the following,
after a general consideration has been made.

1.1.2 Cartesian coordinates

When describing a system in Cartesian coordinates, first three directions that are mutually
perpendicular in space are designated as x-, y- and z-direction, see Fig. 1.4. If, from the

23
 1.1. KINEMATICS OF MASS POINTS

z
6 h
x

⃗r y

z
⃗ez
6
x sL
⃗
ey
b - y -x
⃗ex

Figure 1.4: Cartesian coordinates, position vector and basis vectors.

current position ⃗r of the mass point to be considered, we drop the perpendicular onto the
xy-plane, then we can obtain from its length the z coordinate. From the projected point L
the perpendicular can be dropped onto the x- or y-axis to obtain the coordinates y and x.
Thus one could represent the position vector ⃗r by the three coordinates x, y and z:
 
x
⃗r =  y  (1.8)
 
z

(you probably know it from school in this form). However, a different representation is
common and useful, based on the use of basis vectors: According to the laws of linear
algebra, the position vector
1 0 0
       
x
⃗r =  y  = x  0  + y  1  + z  0 
       
z 0 0 1
can be represented as a linear combination of the three vectors
1 0 0
     

⃗ex =  0  , ⃗ey =  1  , ⃗ez =  0  .

     
0 0 1
which are called unit vectors or basis vectors. The resulting representation
⃗r = x⃗ex + y⃗ey + z⃗ez (1.9)
may look bulky compared to (1.8), but it also has advantages. In particular, analogous
representations also apply in curvilinear coordinates (e.g. cylindrical or spherical coordi-
nates).
The three basis vectors ⃗ex , ⃗ey and ⃗ez are vectors of length 1 that point into the three
respective coordinate directions and are, therefore, perpendicular to each other.
By finding the time derivative we obtain the velocity of the mass point, taking into account
the invariability of the unit vectors over time:
⃗v = ⃗r˙ = ẋ⃗ex + ẏ⃗ey + ż⃗ez , (1.10)

24
1.1. KINEMATICS OF MASS POINTS

In this way, each component of the velocity vector results from the time derivative of the
respective coordinate. In line with this, the acceleration of the mass point results in:

⃗a = ⃗v˙ = ⃗r¨ = ẍ⃗ex + ÿ⃗ey + z̈⃗ez . (1.11)

The path velocity is q

|⃗v | = ẋ2 + ẏ 2 + ż 2 ,
and the path acceleration is
d ẋẍ + ẏ ÿ + ż z̈
apath = |⃗v | = p 2 .
dt ẋ + ẏ 2 + ż 2

1.1.3 Polar coordinates

In the case of motion on a plane, i.e. two-dimensional motion, the current position of the
mass point can also be characterised by the distance r from the origin and by the direction
of the position vector, which is given by its angle φ to the x–axis, see also Fig. 1.5a.

y6 m y6 m
i
y i
y


r
 ⃗
e
⃗eφ  r
φ x
-
XX
y
XX φ

x
-

(a) (b)

Figure 1.5: Polar coordinates (a) and corresponding unit vectors (b).

Two unit vectors are required to represent a given vector in the coordinate system. For
this purpose, we use the unit vector ⃗er , which points from the origin to the mass point,
and the unit vector ⃗eφ that is perpendicular to it, see also Fig. 1.5b. If the vector ⃗er is
represented by the Cartesian unit vectors introduced in the previous section, we obtain
the linear combination
⃗er = cos φ⃗ex + sin φ⃗ey , (1.12)
which is based on the relations ⃗er · ⃗ex = cos φ and ⃗er · ⃗ey = sin φ. The vector ⃗eφ can
be easily constructed if we take into account that it makes an angle with the x–axis 90°
greater than ⃗er . So in (1.12), you just need to replace φ with φ + 90° in order to obtain

⃗eφ = − sin φ⃗ex + cos φ⃗ey . (1.13)

Both vectors ⃗er and ⃗eφ form an orthonormal basis, i.e. they satisfy the identities

⃗er · ⃗er = 1 ,
⃗eφ · ⃗eφ = 1 ,
⃗er · ⃗eφ = 0 ,

25
 1.1. KINEMATICS OF MASS POINTS

which will prove useful later. In particular, all vectors can be represented by a linear
combination of these two basis vectors. The easiest to represent is the position vector,
which according to Fig. 1.5 can be written as

⃗r = r⃗er .

Now we have to find the velocity by differentiating with respect to time, according to

d
⃗v = ⃗r˙ = (r⃗er ) = ṙ⃗er + r⃗e˙ r . (1.14)
dt

It is essential to note that the basis vector ⃗r˙ , which always follows the mass point (see also
Fig. 1.5b), is variable in time. Plugging in its definition (1.12) and taking into account
(1.13) leads to

d d
⃗e˙ r = (cos φ) ⃗ex + (sin φ) ⃗ey = φ̇ [− sin φ⃗ex + cos φ⃗ey ] = φ̇⃗eφ (1.15)
dt dt
and then, after applying this to (1.16), we obtain the velocity

⃗v = ⃗r˙ = ṙ⃗er + rφ̇⃗eφ . (1.16)

The two components of the velocity are often also called radial velocity vr = ṙ and az-
imuthal velocity or circumferential velocity vφ = rφ̇. The quantity ω = φ̇ is also called
angular velocity.
The path velocity is then
√ q q q
|⃗v | = ⃗v · ⃗v = vr2 + vφ2 = vr2 + r2 ω 2 = ṙ2 + r2 φ̇2 .

Differentiating (1.16) again with respect to time yields the acceleration as

d
⃗a = ⃗v˙ = (ṙ⃗er + rφ̇⃗eφ ) = r̈⃗er + ṙ⃗e˙ r + ṙφ̇⃗eφ + rφ̈⃗eφ + rφ̇⃗e˙ φ , (1.17)
dt

where, again, we have to replace ⃗e˙ r in accordance with (1.18) and additionally consider

d d
⃗e˙ φ = (− sin φ) ⃗ex + (cos φ) ⃗ey = −φ̇ [cos φ⃗ex + sin φ⃗ey ] = −φ̇⃗er , (1.18)
dt dt
so we can write the acceleration as a linear combination

⃗a = ⃗v˙ = ⃗r¨ = r̈ − rφ̇2 ⃗er + [rφ̈ + 2ṙφ̇] ⃗eφ ,

h i
(1.19)

of the unit vectors ⃗er and ⃗eφ , as required. Finally, the path acceleration is:

d ṙr̈ + rṙφ̇2 + r2 φ̇φ̈

apath = |⃗v | = .
dt ṙ2 + r2 φ̇2
p

Example: Uniform circular motion

If we choose the centre of the circle as the origin of the coordinate system, as is shown in
Fig. 1.6, then a circular motion is characterised by a constant distance r = R from the
origin. Therefore, polar coordinates prove to be particularly useful for the description of

26
1.1. KINEMATICS OF MASS POINTS

i
y

7

y
6 r =

R


φ

-x

Figure 1.6: circular path.

such a motion, since these can be represented in the form

r = R = const
φ = φ (t)

where the azimuth angle φ is the only variable. In contrast, the description in Cartesian
coordinates is as follows

x = R cos φ (t)
y = R sin φ (t)

and has two variables. The form above represents a general circular motion. A uniform
circular motion is a circular motion in which the azimuth angle increases linearly with
time, i.e.
φ = φ0 + ω0 t (1.20)
applies. The coefficient ω0 in (1.20) is called angular velocity or also angular frequency.
Since an angular increase of 2π corresponds to a full revolution, ω0 is related to the
revolution period T via the simple relation ω0 T = 2π.
In this example, the velocity is determined by applying this to the equation (1.16) as:

⃗v = ⃗r˙ = ṙ⃗er + rφ̇⃗eφ = ω0 R⃗eφ ,

which yields a constant path velocity:

|⃗v | = ω0 R .

The acceleration results from (1.19) as:

⃗a = ⃗v˙ = ⃗r¨ = r̈ − rφ̇2 ⃗er + [rφ̈ + 2ṙφ̇] ⃗eφ = −ω02 R⃗er .

h i

1.1.4 Cylindrical coordinates

Cylinder coordinates are considered to be the natural extension of polar coordinates for
spatial motion.

27
 1.1. KINEMATICS OF MASS POINTS

z
6 h
x

⃗r y

z

r sL
b φI -x

Figure 1.7: Cylindrical coordinates

In contrast to Cartesian coordinates, where the position of a point is determined by three

lengths, in cylindrical coordinates we use two lengths and one angle. In accordance with
Fig. 1.7, the z–coordinate is determined by dropping the perpendicular from the mass point
onto the xy–plane, just like in the Cartesian case. The position of the point projected
onto the xy–plane can then be written in polar coordinates r and φ.
Based on this, the position vector pointing to the mass point m can be represented by

⃗r = r⃗er + z e⃗z (1.21)

with the two unit vectors ⃗er and ⃗ez being defined as before. In contrast, a given arbitrary
vector ⃗u is represented
⃗u = ur⃗er + uφ⃗eφ + uz ⃗ez (1.22)
in all spatial directions perpendicular to each other.
For the velocity of a mass point, then the following equation applies:

⃗v = ⃗r˙ = ṙ⃗er + rφ̇⃗eφ + ż⃗ez

And, accordingly, for the acceleration:

⃗a = ⃗v˙ = ⃗r¨ = r̈ − rφ̇2 ⃗er + [rφ̈ + 2ṙφ̇] ⃗eφ + z̈⃗ez .

h i

There is also an example for these coordinates, but in line with the “learning by doing”
principle in the form of an exercise :)

1.1.5 Other coordinate systems

There are many ways to set a point in space, not all of which can be listed here. However,
two others that are frequently used should be briefly described here. But no detailed
elaboration will be provided.

Spherical coordinates
When using spherical coordinates, the position of a point in space is described by one
length and two angles. First of all, r is the distance of the point in space from the origin.

28
1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

z f
v
6
r y

R
ϑ
r
a φ I -x

Figure 1.8: Spherical coordinates

ϑ is the angle between the z-axis and the position vector directed towards the point in
space. Finally, the angle φ is defined in the same way as with the cylindrical coordinates
in the previous section, i.e. as the angle between the projection of the position vector onto
the xy-plane and the x-axis. By giving r, ϑ and φ, every point in space is thus uniquely
defined. For details on spherical coordinates we refer at Wikipedia or Appendix B.2.1.

The GPS coordinates used in determining the position of a point on the Earth’s surface
are very similar to spherical coordinates, at least conceptually.

Natural coordinates

When describing the motion of a mass point, the polar, cylindrical and spherical coordi-
nates shown in the previous sections can prove to be advantageous over the supposedly
simple Cartesian coordinates despite the greater complexity of the resulting velocity and
acceleration vectors, if these are better fitted to the motion. An optimal fit is given when
a coordinate line coincides with the path of the mass point to be considered. In this case,
it is sufficient to indicate one coordinate to determine the current position of the mass
point, namely the arc length s covered up to the current time t, i.e.

⃗r = ⃗r (s) . (1.23)

However, the path ⃗r = ⃗r (s) is usually unknown a priori, unless, for example, one consid-
ers a motion guided on a rail. Here, too, there are unit vectors that satisfy the so-called
Frenet–Serret formulas and form the Frenet frame. In fact, these coordinates are particu-
larly useful for guided motion.

More details about natural coordinates together with an expample can be found in Ap-
pendix B.2.2.

1.2 Kinematics of rigid bodies for motion in a plane

So far, only mass points have been considered. Real bodies, however, have a spatial
dimension, a possibly quite complex shape and, in most cases, also deformation properties.
If one can neglect the latter, one speaks of a rigid body.

29
 1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

1.2.1 Degrees of freedom of rigid bodies in a plane

We first distinguish between position and orientation. In order to determine the position
of a rigid body, which has a volume expansion, one chooses, for example, a representative
material point of the body and considers the position vector pointing to it. This repre-
sentative point could be a marker, but usually it is the body’s centre of mass, along with
the position vector ⃗rs pointing to it.
However, this is not sufficient to specify the body’s position, because with the centre of
mass fixed, rotation around it is still possible. Therefore, we additionally require the
specification of an angle of rotation φ relative to a fixed reference orientation of the body.
Although a body generally has a rather complex shape, in the following we will use a
simple rod of mass m, length l, cross-sectional area A and material density ϱ. Here, we
distinguish a√rod from a cuboid, cylinder
√ or other body by the fact that the “slenderness
parameter” A/l is very small, i.e. A/l ≪ 1. Fig. 1.9 shows a very simple example of

P
r′ φ
⃗r
S
⃗rs
x
Figure 1.9: Determining the position of the centre of mass and orientation of a body as
well as the position of a material point P.

a rotating rod. The position vector ⃗rs points here towards the centre of the rod and the
angle of rotation φ is chosen as a reference orientation in relation to a rod in a horizontal
position.

1.2.2 Essential kinematic quantities

Since, in the case of motion in a plane, the position vector is determined by two components
xs , ys , together with the angle of rotation there are three degrees of freedom. It’s different
with motion in space: In 3D space, the position vector has three components and rotation
is possible around three independent axes, resulting in 6 degrees of freedom. The motion
of rigid bodies in space will be covered in Applied Mechanics and Kinematics of Robots,
and will not be pursued further here.
Just as with the mass point, we can find the velocity of a rigid body by differentiation of
the position vector ⃗rs that points to the centre of mass:

⃗vs = ⃗r˙s . (1.24)

Similarly, the time derivative of the angle of rotation φ yields the angular velocity

ω = φ̇ (1.25)

as the increase in the angle per time. Accordingly, the acceleration of the centre of mass
is given by
⃗as = ⃗v˙ s = ⃗r¨s (1.26)

30
1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

and the angular acceleration

α = ω̇ = φ̈ (1.27)
as an analogous quantity of the rotational motion.
However, these two quantities only describe a global motion of the entire body. If we
consider an arbitrary material point of the body, for instance the point P indicated in
Fig. 1.9, its individual motion may be different from the motion of the center of mass.
In order to determine the position vector ⃗r of an arbitrary material point P of the body
at a distance r′ from the centre of mass (Fig. 1.9), the following relationship results from
pure geometric considerations (vector addition):

⃗r = ⃗rs + r′ cos φ⃗ex + r′ sin φ⃗ey , (1.28)

and its velocity is obtained by derivation w.r.t. time:

⃗v = ⃗r˙s − r′ sin φφ̇⃗ex + r′ cos φφ̇⃗ey

= ⃗r˙s + r′ φ̇ [− sin φ⃗ex + cos φ⃗ey ] . (1.29)

Above formulas are later referred for computing other global properties of the rigid body
like e.g. momentum, angular momentum and energy.

Exercises
Exercise 1.1: River crossing
(Solution cf. A.1.1)
A river with flow velocity v0 and width l is crossed by a swimmer whose velocity relative
to the water is constant v1 , where v1 < v0 . For the description of this motion we choose
a Cartesian coordinate system with x-axis parallel to the flow direction and coordinate
origin at exactly the point of the shore on this side where the swimmer jumps into the
water.
1. We first assume that the swimmer is always oriented in a perpendicular direction to
the shore on the other side. In doing so, s/he is driven sideways by the flow. Which
total velocity ⃗v then results from the superposition of her/his own swimming motion
with the flow?
2. Infer from (1) the position vector ⃗r and represent its Cartesian components in the
form

x(t) = · · ·
y(t) = · · ·

as functions of time.
3. After what time t1 does the swimmer reach the shore on the other side?
4. How large is the distance x1 the swimmer is drifted in total?
5. The swimmer now changes her/his strategy by not orienting her/himself perpen-
dicularly to the shore on the other side as before, but instead changes direction of
swimming by an angle α with respect to the original swimming direction, and in such
a way that s/he also swims a little against the direction of the flow. What changed
swimming time t1 results then and by what distance x1 is s/he then drifted?

31
 1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

6. How should α be chosen so that the distance x1 that the swimmer is drifted is
minimal?
7. How should α be chosen so that the swimmer arrives at the other side of the river
in the shortest possible time t1 ?

Exercise 1.2: Earth moving

(Solution cf. A.1.2)
The astromomical unit is defined as the mean radius of the Earth’s orbit and is: 1AU =
149597987.7 km. A sidereal year, 1a, comprises 365 days, 6 hours, 9 minutes and 9.54
seconds. We neglect the eccentricity of the Earth’s orbit in the following and consider it
to be approximately circular.
1. Find the (mean) orbital velocity |⃗v | of the Earth in its orbit around the Sun with an
accuracy of three valid digits.
2. Determine the magnitude and direction of the acceleration of the Earth’s orbital
motion, also with an accuracy of three valid digits.

Exercise 1.3: Hambach Forest

(Solution cf. A.1.3)
An environmental activist takes a bath in what is assumed to be a circular lake near
the Hambach Forest. Having just arrived in the middle of the lake, he sees an RWE
employee running to the shore in a rage, who makes it very clear to him by shouting “Get
out of there!” that he is looking for confrontation. Since the confrontation would in all
probability not be limited to a purely verbal discourse and the rather lean environmental
activist sees himself at a clear disadvantage compared to the athletic appearing RWE
employee, he seeks his salvation in escape.
The radius R of the lake is given. We assume that the RWE-employee runs with a speed
v0 four times as fast on the shore as the activist can swim in the water, but cannot run
faster than the latter.
1. Why would the simplest of all strategies, swimming in a straight line to the point
exactly opposite the pursuer’s position, not allow the activist to make his escape?
2. Develop a better strategy that allows escape. Validate your strategy by calculating
the advantage the activist has over the pursuer when he reaches the shore. Note that
the pursuer would reverse his running direction as soon as it would be advantageous
for him.

Exercise 1.4: Trajectory

(Solution cf. A.1.4)
Let the trajectory of a rocket taking off from the ground be given in polar coordinates by
r(t) = R exp(γt)
φ(t) = ωt
given with earth radius R and positive constants γ and ω.
1. Find the velocity ⃗v . Show that the direction of motion always makes the same angle
α with the ground and determine α.
2. Find the acceleration ⃗a. How must ω and γ be matched so that ⃗a has no component
in the radial direction (⃗er -direction)?

32
1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

3. Make a sketch of the orbital curve.

Exercise 1.5: Kinematics of a rigid body y6

(Solution cf. A.1.5) v0
6
As shown on the left, a rod (length l) is pulled up a wall by a rope
A
with constant velocity v0 at one end. Initially (t = 0) the rod is
lying on the ground.
@
@
@
@
1. Find the centroid position ⃗rs in Cartesian coordinates xs (t) @@
@ rS
and ys (t).
@
@@
2. Obtain the velocity ⃗vs (t) and the acceleration ⃗as (t). ⃗rs @ @
3. Determine the angle φ(t), the angular velocity ω(t), and the
@
@
φ@@ x
@B -
angular acceleration α(t).
@

Exercise 1.6: In the parking garage

(Solution cf. A.1.6)
You drive by car in a parking garage up to the top. The path curve in three-dimensional
space which you follow thereby is called a screw path. This is sketched in the following:
z
6

y


b -x

The motion is considered in cylindrical coordinates as a uniform helical motion resulting

by superposition of a uniform circular motion (radius R) in the xy-plane,

r = R,
φ = ωt ,

with a uniform linear motion in z- direction, i.e.:

z = v0 t .

1. Compute the velocity ⃗v and the speed |⃗v |.

2. Compute the acceleration ⃗a.
3. Let R = 5m be given as well as the time T = 15s you need to get one floor higher in
the parking garage and finally h = 3.5m the vertical distance between two consecu-
tive levels of the parking garage. Calculate the value of the speed |⃗v |.

33
 1.2. KINEMATICS OF RIGID BODIES FOR MOTION IN A PLANE

34
Chapter 2

Dynamics of mass points

So far, we have only described the motion of a body, without discussing the motion’s cause.
One could also say that, strictly speaking, we haven’t discussed any physics yet, we have
just got the mathematical tools ready!
Aristotle already had the idea that motion is caused by forces, but he assumed a false
connection between force and motion by supposing that a motion ends abruptly as soon
as there is no more force acting on the body in question. The fact that a motion once
initiated continues even if the driving force is no longer present, can be seen from the fact
that, for example, a moving car does not come to a standstill (at least not immediately!)
simply by taking one’s foot off the accelerator. In later times, the ideas of Aristotelian
mechanics were replaced by the theory of impetus, which, however, also has weaknesses.
Galilei and Newton were able to formulate a theory of motion that is coherent in itself and
has endured to this day, even though relativity and quantum mechanics reveal the limits
of classical mechanics.

2.1 Newton’s laws of motion

Newton’s laws of motion are the basis of dynamics. All other laws can be derived from
them.

2.1.1 Newton’s first law (Galilei’s principle of inertia)

A free mass point that is not subjected to any external force will move in a straight line
at constant speed, i.e.
⃗r˙ = ⃗v0 = const .

2.1.2 Newton’s second law (law of motion)

As soon as a mass point is subjected to a resultant force F⃗ = i F⃗i ̸= 0 it is accelerated,
P

i.e. its state of motion changes according to

m⃗r¨ = F⃗ . (2.1)

2.1.3 Newton’s third law (action-reaction law)

The forces that two bodies exert on each other are of equal magnitude and opposite to
each other (actio=reactio), i.e. the force F⃗21 , that a body 1 exerts on a body 2 is related

35
 2.2. FORCES

to the force F⃗12 which, conversely, body 2 exerts on body 1:

F⃗21 = −F⃗12 . (2.2)

The reaction principle applies both to remote interactions, i.e. bodies that influence each
other through force fields, and to bodies that are in direct contact with each other. In the
latter case, this leads to the well-known free body diagram, which is used in both statics
and dynamics. How the principle of intersection is applied in statics will be covered
especially in Applied Mechanics.

2.2 Forces
2.2.1 Active forces and reaction forces
The total force F⃗ from Newton’s second law is composed additively of two fundamentally
different parts:
F⃗ = F⃗e ⃗r, ⃗r˙, t + F⃗z
 
(2.3)

namely the sum of all active forces F⃗e ⃗r, ⃗r˙, t and the sum of all reaction forces F⃗z . While
 

the active forces are usually determined from the outset by the position and velocity
of the mass point in question and may also depend on time in a defined way, reaction
forces are forces caused by a rigid guidance of the mass point (rope or guide rail) and are
always oriented perpendicularly to the guide. The direction of a reaction force is, therefore,
determined by geometry, while its magnitude is to be treated as an unknown that can only
be evaluated by solving Newton’s equations of motion resulting from Newton’s second law.

2.2.2 Gravity
Gravity is one of the simplest active forces, as it can be given at the Earth’s surface to a
good approximation as
F⃗G = m⃗g (2.4)
with |⃗g | ≈ 9.81 m/s2 and in the direction of the Earth’s centre. Hence it is constant over
small distances, i.e. it depends neither on the location nor on the velocity and not even
explicitly on time.
On larger length scales, gravity is given by Newton’s law of gravitation, which can be
written in spherical coordinates, with the centre of the Earth as the origin, as

R 2
F⃗G = −mg 2 ⃗er (2.5)
r
where R is the radius of the Earth. Gravity thus depends on the location. This is important
for space travel, for example.

2.2.3 Spring forces

In the following, “spring” serves as a collective term for any kind of elastic structure whose
deformation is linearly related to the force acting on it. Formally, this means that there
is a relationship of the form F = c∆l, where ∆l is the elongation of the spring and F the
pulling force required for this. The constant c is called stiffness (the characteristic of the
spring). Other common formula symbols are k or D.

36
2.2. FORCES

l
|
r1 C OC
A
A
A c
]
J
J ⃗
J
A
C
A
J C
A
 y
i

A
J C
J C ⃗r2
JCC
JC         3

Figure 2.1: Regarding the spring force.

Despite the initially simple linear relationship between force and deformation, the resulting
force law looks somewhat more complicated due to the three-dimensionality of space. To
show this, we consider a spring stretched between two mass points ⃗r1 and ⃗r2 (see Fig. 2.1).
For the formulation of the force law, the current length of the spring is required, which
is obtained from the difference position |⃗r2 − ⃗r1 | between the two position vectors. The
difference |⃗r2 −⃗r1 | − l of the current length to the length l of the relaxed spring determines
the force, whose magnitude is thus

F⃗21 = c ||⃗r2 − ⃗r1 | − l| . (2.6)

The direction of this force is, in turn, parallel to the connecting vector ⃗r2 − ⃗r1 from mass
point 1 to mass point 2. If we also take into account that a stretched spring causes an
attractive interaction, the formula for the spring force exerted on point ⃗r2 is then
⃗r2 − ⃗r1
F⃗21 = −c [|⃗r2 − ⃗r1 | − l] . (2.7)
|⃗r2 − ⃗r1 |

2.2.4 Frictional forces

Frictional forces always counteract the current movement, i.e.

F⃗R ⃗v
=− . (2.8)
F⃗R |⃗v |

The force’s magnitude depends on the respective friction law. The following three laws of
friction are particularly widespread. These are discussed in more detail below:

Coulomb friction
Coulomb friction or dry friction occurs on contact surfaces between solids and follows the
law
F⃗R = µN (2.9)
with µ being the coefficient of friction and N the normal force between the bodies. In
particular, the magnitude of Coulomb friction does not depend on the velocity. The vector
of Coulomb friction results from (2.8) as

⃗v
F⃗R = −µN , (2.10)
|⃗v |

37
 2.2. FORCES

and thus depends on the direction of the velocity. Since the path velocity |⃗v | can be
obtained via the Pythagorean equation, this is indeed a non-linear relationship. Another
unpleasant consequence from a mathematical point of view is that when reversal points
are reached (e.g. in oscillatory motion), the frictional force abruptly changes direction and
can therefore be discontinuous in time, which makes it necessary to divide the motion into
several sections when calculating with Coulomb friction.

Stokes friction
Stokes friction or lubricant friction is a consequence of the friction in a fluid with which
the body under consideration is in contact. This type of friction does not only occur in
lubricated contacts between solids, but also in the motion of bodies that are completely
immersed in a liquid. The magnitude of this friction depends linearly on the velocity, i.e.

F⃗R = k |⃗v | , (2.11)

where k is a constant depending on the problem. Using (2.8), Stokes friction can thus
simply be written vectorially as

⃗v
F⃗R = −k |⃗v | = −k⃗v . (2.12)
|⃗v |

This law only applies as long as the flow in the fluid in questions remains laminar.
From a mathematical point of view, Stokes friction is the most pleasant of the three forms
of friction discussed here, as it is also linear in vectorial representation.

Newtonian friction
Newtonian friction or air friction is experienced by a body that moves through a fluid
(liquid or gas), provided that the resulting flow caused in the fluid is turbulent. The
resulting friction law is
F⃗R = k |⃗v |2

and thus proportional to the square of the velocity. The constant of proportionality k is
often written as ϱcw A, where ϱ is the density of the fluid, A is the cross-sectional area of
the body and cw is the drag coefficient, which depends on the body’s shape. Vectorially
and in line with (2.8), this leads to:

⃗v
F⃗R = −k |⃗v |2 = −k |⃗v | ⃗v .
|⃗v |

2.2.5 Lorentz force

A body that is moving in a magnetic field of flux density B
⃗ and has the electric charge q
experiences the Lorentz force
F⃗L = q⃗v × B⃗.

Like a frictional force, this active force is velocity-dependent. However, in contrast to a

frictional force, it is always perpendicular to the direction of motion (see Fig. 2.2). It
therefore does not cause any path acceleration, but merely a deflection of the body.

38
2.3. EXAMPLES

Figure 2.2: Lorentz force acting on a particle with negative charge in a magnetic field.

2.3 Examples
2.3.1 Fall of a body with friction
An object of mass m in a fluid medium (gas or liquid) is dropped, as shown in Fig. 2.3.
In addition to its gravity force FG = mg, it is also subjected to a velocity-independent

FR
6

~
m
x ?

FG
?
Figure 2.3: Gravity and frictional force in the case of deceleration

frictional force
FR = k |v|n−1 v
where the exponent n can have the value n = 1 (Stokes friction, e.g. a small aerosol particle
in air) or n = 2 (Newtonian friction, e.g. a parachutist), depending on the parameters of the
problem considered. Thanks to the problem being one-dimensional, the vector character
of the related quantities need not to be considered: characterising the object’s position
simply by its falling distance x = x (t) is sufficient! We also neglect buoyancy forces,
which is justified as long as the mass density of the body’s material is much larger than
the density of the surrounding fluid (which ist certainly fulfilled for a gas). Furthermore,
since there are no constraints to consider, Fig. 2.3 is already the free–body diagram of the
problem. Then Newton’s second law can be written as

↓: FG − FR = mv̇ ,

which, after plugging in the forces and carrying out simple transformations, leads to the
differential equation
k
v̇ + |v|n−1 v = g .
m
If we also assume that the motion is always downwards, we can leave out the lines of the
absolute value in the friction term, i.e. |v| = v. Thus the differential equation above can
be written as:
k
v̇ + v n = g . (2.13)
m
A differential equation typically has an infinite number of solutions. For full determination
of the solution an initial condition is required, depending on how the motion starts at t = 0.

39
 2.3. EXAMPLES

If we assume that the motion starts at rest, the initial condition is:
v(0) = 0 . (2.14)

Free fall without friction

If no friction is present, i.e. k = 0, the differential equation simplifies to:
v̇ = g ,
allowing for direct integration:
ˆt
v(t) = gdt′ = gt .
0
Thus, the velocity is linearly increasing with time without any limit, which ist unrealistic!
However, it could be a sufficiently good solution at the beginning of the motion when the
influence of the friction force is yet small.

Asymptotic velocity
After a sufficiently long time, the friction force reaches the same value as the gravity force,
inhibiting any further acceleration of the body. Thus we assume:
lim v = v∞ ,
t→∞
lim v̇ = 0 ,
t→∞
simplifying the equation (2.13) of motion to:
k n
v = g,
m ∞
allowing for determining the asymptotic velocity as:
1/n
mg

v∞ = . (2.15)
k

Full solution of the equation of motion

We can use the so-called separation of variables as a method to solve this equation: For
this purpose, we first rewrite the differential equation as
dv k
= g − vn (2.16)
dt m
and treat the infinitesimals dv and dt as if they were ordinary quantities, in order to bring
all terms with v or dv to one side and all terms with t or dt to the other side. Finally, we
obtain the form
dv k
n n
= dt (2.17)
v∞ − v m
Integration of (2.17) within the bounds 0 and t yields
ˆv(t)
dv k
n − vn
= t (2.18)
v∞ m
v(0)

which is the implicit solution of the differential equation. In the following we consider the
initial condition (2.14) for a body released without initial velocity. In order to solve the
integral above, the two cases n = 1 and n = 2 must be distinguished.

40
2.3. EXAMPLES

Solution for Stokes friction

In the case of Stokes friction, n = 1, we obtain

ˆv(t)
k dv v (t)
 
t= = − ln 1 − , (2.19)
m v∞ − v v∞
0

which can be solved for v (t) and leads to

k
  
v (t) = v∞ 1 − exp − t . (2.20)
m
Since the differential equation (2.13) is linear for n = 1, it can also be solved via the
characteristic equation (“exponential approach”) instead of by separation of variables.

Solution for Newtonian friction

In the case of Newtonian friction, n = 2, we preferably use a handbook of mathematics and
table of formulas to find the integral in (2.18). One possibility would be a partial fraction
decomposition as demonstrated in Appendix B.3.1. Maybe by using a list of integrals the
goal can be achieved more easily. Regardless of the method, one arrives at the final result:
s 
gk 
v (t) = v∞ tanh  t , (2.21)
m

where tanh(x) = [exp(x) − exp(−x)] / [exp(x) + exp(−x)] denotes the tangent hyperbolic
function.

Visualisation of the solutions related to the different cases

If we now compare the results for Stokes friction (2.20) with those for Newtonian friction
(2.21), we see that, in both cases, there is a constant asymptotic velocity p v∞ , which
according to (2.15) is v∞ = mg/k in the case of Stokes friction, and v∞ = mg/k in the
case of Newtonian friction. The temporal dynamics of this process, which can be roughly
characterised by the time constant (the reciprocal of the factor before t in the argument
of exp or tanh)
, Stokes friction
(
m/k
τ= √ √
m/ gk , Newton friction
however, is determined in the case of Stokes friction by the mass and the friction constant
alone, whereas in the case of Newtonian friction the acceleration due to gravity also plays
a role. Finally, x(t) is obtained straight forwardly by integration of v(t) with respect to
time, which is not shown here in detail, but the result is plotted a diagram: Fig. 2.4
shows both the temporal course of the velocity and that of the distance x(t) travelled in
comparison to each other and to the free fall v(t) = gt.

2.3.2 The simple gravity pendulum

A simple gravity pendulum is a system in which a mass point is attached to an inextensible
massless cord, the upper end of which is fixed, see also Fig. 2.5.
Since the motion of the pendulum bob takes place on a circular arc due to the kinematic
constraint by the cord, polar coordinates are suitable for describing the motion, whereby

41
 2.3. EXAMPLES

v
v∞
1.00
0.75 without friction
0.50 Stokes friction
0.25 Newton friction
t
0
0 1 2 3 4 5 τ
x 1
x = gt2
v∞ τ 2
4
3
2 Stokes friction
1 Newton friction
t
0
0 1 2 3 4 5 τ

Figure 2.4: Representation of a fall with friction in a v-t-graph (above) and in an x-t-
graph (below). Comparison of Stokes friction (blue), Newton friction (red) and the free
fall without friction (green).

Figure 2.5: Simple gravity pendulum: sketch (left) and free-body diagram (right).

the point of suspension is chosen as the origin. Here, the angle φ is the only variable that
changes with time, while the distance from the origin is constant and corresponds to the
length of the cord, i.e. r = l. We first consider the illustration in Fig. 2.5 and represent
the pendulum bob isolated from its surroundings. From the resulting free-body diagram
we obtain all the forces that the pendulum bob is experiencing, namely its weight force
mg⃗ex and the force S ⃗ = −S⃗er of the cord. The former is an active, and therefore known,
force. The latter is a reaction force that results from the kinematic constraint by the cord
and whose magnitude is initially unknown. In contrast to statics, however, the sum of
both forces is not zero. Instead, the two forces cause acceleration in line with Newton’s
equation of motion

mg⃗ex − S⃗er = m⃗r¨ (2.22)

42
2.3. EXAMPLES

where the acceleration ⃗r¨ has, according to (1.19) the form

⃗r¨ = r̈ − rφ̇2 ⃗er + [rφ̈ + 2ṙφ̇] ⃗eφ = −lφ̇2⃗er + lφ̈⃗eφ ,

h i

considering the kinematic constraint r = l. Hence, Newton’s equation of motion becomes:

mg⃗ex − S⃗er = −mlφ̇2⃗er + mlφ̈⃗eφ .

This vector equation can be decomposed into its components as usual, whereby it makes
sense here to decompose into a radial direction (⃗er –direction, “↘”) and into an azimuthal
direction (⃗eφ –direction, “↗”), i.e.

↘: mg cos φ − S = −mlφ̇2 , (2.23)

↗: −mg sin φ = mlφ̈ . (2.24)

This system of equations consisting of two equations for two unknown functions φ (t) and
S (t) can be written as
g
φ̈ + sin φ = 0 (2.25)
l
S = mg cos φ + mlφ˙2 (2.26)

Obviously the force S = S(t) of the cord can be obtained from (2.26) once φ = φ(t)
has been determined by solving (2.25). However, solving the differential equation (2.25)
is a challenge that should not be underestimated, since it is a second-order, non-linear
differential equation.

Approximate solution of the equation of motion (2.25)

In section 0.4, we have learned about the Taylor expansion of a function, and in the
Appendix B.1.2 this concept has been extended to higher order. By applying the Taylor
expansion to the sine function, its Taylor expansion around the zero point is [6]:
1
sin φ ≈ φ − φ3 + · · ·
6
For very small angles, where “small” is to be understood in the sense of radians, we can
thus approximate the sine of an angle by the angle itself, accepting an error of the order
of φ3 /6. For an angle of 30°, which is φ = π/6 radians, this leads to an estimated error
of π 3 /64 ≈ 2,4%, at only 10°, in radians φ = π/18, we obtain an estimated error of just
π 3 /183 /6 ≈ 0,89‰. Thus, depending on the required accuracy, it may be legitimate to
approximate sin φ with φ in (2.25), which leads to the simplified differential equation
g
φ̈ + φ = 0 . (2.27)
l
In contrast to the original equation (2.25), this equation is linear with respect to the sought
quantity φ, which means that its solutions can be superposed as desired, i.e. each linear
combination of different solutions is another solution. A legitimate method of finding
solutions is to guess correctly. In this case, we are looking for a function that, after
being differentiated twice, reproduces itself except for the sign and the prefactor. The
two trigonometric functions sin(· · · ) and cos(· · · ) meet the requirement of reproducing
themselves after being differentiated twice, except for the sign. The time t must not stand
alone in the arguments of these functions, as it has a unit. In line with the general rule

43
 2.3. EXAMPLES

that the arguments of transcendental functions must always be dimensionless, t must be

combined with a yet unknown constant ω0 to obtain a dimensionless quantity, ω0 t. Thus,
we insert φ = A sin(ω0 t) in Eq. (2.27) and obtain:

g
 
− ω02 A sin(ω0 t) = 0
l

which is fulfilled (at any time t and for A ̸= 0) if and only if

r
g
ω0 = .
l
One can easily convince oneself that φ = B cos(ω0 t) is a second solution of (2.27). Due to
linearity, however, the combination of both,

φ(t) = A sin(ω0 t) + B cos(ω0 t) , (2.28)

with any given constants A and B is also a solution of (2.27). In fact, it can be mathe-
matically proven that there are no further solutions, so that (2.28) is indeed the general
solution of the differential equation (2.27). Finally, the two constants A and B can be
determined by considering the initial conditions of the motion, i.e. by specifying the initial
displacement and the initial angular velocity. If, for example, the motion starts from a rest
state, i.e. φ(0) = 0, then the derivative of (2.28), i.e. φ̇(t) = ω0 A cos(ω0 t) − ω0 B sin(ω0 t),
must be zero für t = 0, which leads to the condition ω0 A = 0 and to A = 0. In this case,
B is the initial displacement φ0 , so that the solution in this special case is

φ(t) = φ0 cos (ω0 t) . (2.29)

Exact solution of the equation of motion (2.25)

The equation of motion can be transformed into an integrable form as follows: By multi-
plying with φ̇ and using the chain rule dt
d
f (x(t)) = f ′ (x)ẋ, (2.25) implies

φ̇φ̈ +ω02 sin φφ̇ = 0

|{z} | {z }
d d
dt
[φ̇2 /2] dt
[− cos φ]

and finally
d 1 2
 
φ̇ − ω02 cos φ = 0 .
dt 2
In this form, the equation of motion can be integrated easily, which leads to
1 2
φ̇ − ω02 cos φ = C (2.30)
2
with C being the constant of integration. The constant C can be determined from the
initial state. For example, if the pendulum is released from rest (φ̇ (0) = φ̇0 ) at an initial
angle φ (0) = φ0 , the constant is then calculated as
1
C = φ̇ (0)2 − ω02 cos φ (0) = −ω02 cos φ0 .
2
Plugging in (2.30) yields
1 2
φ̇ = ω02 [cos φ − cos φ0 ] , (2.31)
2

44
2.3. EXAMPLES

which is a first-order differential equation. Solving for the angular velocity yields

dφ √ √
= φ̇ = ± 2ω0 cos φ − cos φ0 .
dt
The sign depends on the pendulum’s direction of motion. In the motion’s first phase from
the right starting point to the left reversal point, the negative sign applies. The above
differential equation can be solved by separation of variables. We obtain

ˆφ ˆt
dφ′ √ √
− √ = 2ω0 dt′ = 2ω0 t
cos φ − cos φ0
′
φ0 0

and thus the pendulum’s equation of motion in the form

ˆφ0
1 dφ′
t = t (φ) = √ √ , (2.32)
2ω0 cos φ′ − cos φ0
φ

instead of in the usual form φ = φ (t). Unfortunately, there is no antiderivative of the

integrand that can be represented by elementary functions. It is a so-called elliptic integral.
The corresponding antiderivative can be found either in mathematics specialist literature
[1] or by computer-aided methods.
Fig. 2.6 shows the solution for the amplitude φ0 = π/3, i.e. 60°, together with the ap-
proximate solution (2.29). The comparison shows that it is not so much the shape of the

0.5
φ/φ0

−0.5

−1

−1.57 −0.79 0 0.79 1.57

ω0 t

Figure 2.6: Motion of the simple gravity pendulum at a maximum displacement of 60°
(blue) and 90° (green) in comparison with the approximate solution for very small dis-
placements (red).

motion’s graph that is different, but rather that the pendulum’s period of oscillation is
longer in the exact solution than in the approximate solution for small displacements. The
period of oscillation T can be determined from (2.32) as

√ ˆ 0
+φ
2π dφ′
T = 2t (−φ0 ) = √ .
ω0 cos φ′ − cos φ0
−φ0

45
 2.3. EXAMPLES

π π π π 2π
φ0
6 4 3 2 3
ω0 T
1.0174 1.0400 1.0732 1.1803 1.3729
2π
Table 2.1: Relative period of oscillation T /T0 = ω0 T / (2π) for various displacements.

Table 2.1 shows the relative period of oscillation T /T0 = ω0 T / (2π) for various displace-
ments. We see that for displacements up to 30°=π/6 the increase of the period of oscillation
is below 2%, but becoming rapidly larger for larger displacements. For a displacement of
120°=2π/3 the increase of the period of oscillation is already about 37%.

Exercises
Exercise 2.1: Motion of coupled masses
(Solution cf. A.2.1)  m
r s 1s
A body (mass m1 ) mounted frictionless on rollers  @@@@@@@@@@@@
is pulled over a table by a massless rope which is 
deflected by a roller and attached to a weight (mass x
m2 ). The mass moments of inertia of all rollers,
their friction and the elongation of the rope are m2
negligible. The movement starts from rest at the
position x = 0. x?
1. Formulate the two Newtonian equations of motion for the two masses.
2. From the equations of motion, determine the force with which the rope is tensioned.
3. Determine the motion x = x(t) by integrating one of the two equations of motion
twice. Note that the movement starts at t = 0 from rest at x = 0.

Exercise 2.2: Motion of coupled masses with lubrication

(Solution cf. A.2.2)  m1
A body (mass m1 ) is supported by a rope, which s
is deflected by a pulley and is attached to a weight  @@@@@@@@@@@@

(mass m2 ). over an oil-smeared table, whereby a 

x
Stokes frictional force of the amount |F⃗R | = k ẋ has
to be considered. In contrast the mass moment of
inertia of the pulley, its friction and the elongation m2
x
of the rope are not to be taken into account. ?
1. Cut the system free and set up the equations of motion of both masses m1 and m2 .
2. Obtain a single equation for x (second order differential equation) from both equa-
tions by eliminating the rope force.
3. Determine the constant velocity ẋ = v∞ being reached asymptotically.
4. Let l0 be the distance from the mass m1 to the edge of the table. What time t1
elapses until the body reaches the edge of the table under the assumptions that (i)
it starts from rest and (ii) by the time it reaches the edge it has already (factually)
reached the limiting velocity determined in task part 3?

46
2.3. EXAMPLES

Note: You do not have to solve the differential equation obtained in part 2! It is
sufficient to integrate it once according to time (certain integration from 0 to t1 ) and
then insert all known results.

Exercise 2.3: Circular motion with dry friction

(Solution cf. A.2.3)
A string is fixed in the middle of a table top with f
v
a glass of mass m attached to its other end. The 
7
y r = R

glass with a (path) velocity v0 with the string taut 6
is set in motion in a tangential direction, whereby 
φ -x
it is forced by the string into a circular path with
radius R. Due to a Coulomb friction force, whose
amount is |F⃗R | = µmg with coefficient of friction
µ, the movement gradually slows down until it
comes to a standstill.

1. What is the angular velocity ω0 at the beginning of the motion?

2. Cut the glass free and determine both components of Newton’s equation of motion
using polar coordinates (compare gravity pendulum). These equations can easily be
solved for the angular acceleration φ̈ and the rope force S, i.e. you can write them
in the form

φ̈ = · · · (2.33)
S = ··· (2.34)

State the equations in this form.

3. Solve the equation of motion (2.33) obtained earlier for the angle φ = φ(t) by
integrating twice after time and taking into account the initial conditions φ(0) = 0
(initial angle) and φ̇(0) = ω0 (initial angular velocity). Give the solution for φ(t)
depending on given quantities (m, µ, g, v0 , R) and of course time t.
4. Determine the force S = S(t) with which the string is tensioned. Again, present the
result exclusively as a function of given quantities and time.
5. After what time t1 does the glass come to a standstill?
6. At what angle φ1 does the glass come to a standstill?

47
 2.3. EXAMPLES

48
Chapter 3

Constants of motion

A major challenge in dynamics is the temporal variability of numerous quantities, which

makes solving Newton’s equations of motion more difficult. All the more valuable is the
identification of specific quantities that remain constant throughout a motion, the so-
called constants of motion. This leads to the definition of new quantities, in particular,
energy, momentum and angular momentum. The advantage of such quantities is obvious:
Regardless of how complicated a motion may be, it is possible to use such quantities to
infer changes in other quantities without having to calculate the motion in detail. However,
there are also limits to applicability, and preconditions.

3.1 Energy
The inner product of Newton’s equation of motion (2.1) with the mass point’s velocity,

m⃗r˙ · ⃗r¨ = F⃗ ⃗r, ⃗r˙, t · ⃗r˙ ,

 

and using the chain rule according to

d m 2
 
m⃗r˙ · ⃗r¨ = m⃗v · ⃗v˙ = ⃗v = Ṫ , (3.1)
dt 2

allows for writing the left-hand side of the equation as a time derivative of the quantity
m 2
T = ⃗v (3.2)
2
which is called the kinetic energy of the mass point. In contrast, we define the dot product

P = F⃗ · ⃗v = F⃗ · ⃗r˙ (3.3)

as the power of the force acting on the mass point. Thus (3.1) takes the distinctive form

Ṫ = P (3.4)

of a balance equation for the kinetic energy, which states that the system’s kinetic energy
changes due to the power of the active forces, namely of both the external forces and the
interaction forces. This is the most general form of the law of conservation of energy.

49
 3.1. ENERGY

Under certain conditions, the power in (3.4) can be interpreted as the temporal changes
of another form of energy. However,

d
F⃗ · ⃗r˙ = − U (⃗r) (3.5)
dt

must be satisfied for this purpose, with the scalar function U (⃗r) being the potential or
potential energy of the force F⃗ . In Cartesian coordinates and in accordance with the chain
rule, (3.5) can also be written in the form

d ∂U ∂U ∂U
Fx ẋ + Fy ẏ + Fz ż = − U (x, y, z) = − ẋ − ẏ − ż .
dt ∂x ∂y ∂z

For this identity to be satisfied for any motion ẋ, ẏ, ż,

∂U ∂U ∂U
Fx = − , Fy = − , Fz = −
∂x ∂y ∂z

must necessarily apply, or in vector notation:

F⃗ = −grad U = −∇U . (3.6)

It is by no means self-evident that such a potential exists. Dependence of the force on

the velocity, as is the case in frictional forces or in the Lorentz force, already results
in the non-existence of a potential. Forces that have a potential are called conservative
forces. Gravity, spring forces and attractive forces between electric charges are examples
of conservative forces:
1. The potential energy pertaining to the weight force m⃗g must be a linear function,
since the weight force is constant. Evidently,

⃗g · ⃗r
 
U = −m⃗g · ⃗r = mg − = mgh (3.7)
g
| {z }
h

satisfies all conditions, where h is the height in the sense of the projection of the
position vector onto the fall line.
2. In the one-dimensional case, the potential energy of a spring is simply the negative
antiderivative of the pertaining force F = −C(x − l0 ), i.e.

C
U= (x − l0 )2 . (3.8)
2
Due to the fact that energy is a scalar, it has the same form in the three-dimensional
case, whereby the elongation x − l0 of the spring in relation to the relaxed length l0
can simply be expressed in the form |⃗r2 − ⃗r1 | − l0 , i.e.:

C
U= (|⃗r2 − ⃗r1 | − l0 )2 . (3.9)
2

3. By definition, the electric voltage U is an energy per charge q, so qU is the potential

energy of a charge in an electric field.

50
3.2. MOMENTUM AND CENTRE OF MASS

If now in a mass point system all active forces are conservative, then the balance (3.4) for
the kinetic energy can be rewritten as
d
[T + U ] = 0 ,
dt
which is integrable and yields the conservation law
T + U = E = const
for the total mechanical energy, consisting of potential energy and kinetic energy.
Reaction forces are also non-conservative, but since these are basically oriented perpendic-
ularly to the direction of motion and thus to the velocity vector, they do not contribute to
the balance (3.4), so that the energetic treatment of problems is also applicable to guided
motion. The best example of this is the gravity pendulum discussed earlier.

3.1.1 Examples of energy conservation

Gravity pendulum In the case of the gravity pendulum discussed in section 2.3.2,
the conservation of energy was already exploited without the term “energy” appearing
there. Nevertheless, the cleverly established equation (2.30) is nothing other than the
conservation of energy, except for one pre-factor.

Elastic boost A body (mass m) is initially pressed with a force F against a spring
(stiffness C) that is attached to a wall, and then released. The expanding spring accelerates
the trolley until it is completely relaxed. Then the trolley is released and continues its
motion with the terminal velocity ve .

initial state acceleration phase final state

x(t) =? x = ve t

F
S S

Due to statics, the length l0 − xa = F/C of the spring is shortened via F + C(xa − l0 ) in
the initial state. Thus the potential energy of the spring is Ua = C(xa − l0 )2 /2 = F 2 /(2C).
If we ignore friction, we can find the terminal velocity ve according to
m
0 + Ua = E = ve2 + 0
2
which leads to
F
ve = √
mC
without having to consider the acceleration process and without having to solve the equa-
tion of motion (differential equation).

3.2 Momentum and centre of mass

The quantity p⃗ = m⃗v = m⃗r˙ is called the momentum of a mass point. Newton’s equation
of motion (2.1) can therefore also be written as
p⃗˙ = F⃗ . (3.10)

51
 3.2. MOMENTUM AND CENTRE OF MASS

In this context, force can also be interpreted as a current of momentum.1

When the momentum is conserved, the force disappears: F⃗ = 0 leads directly to p⃗ =
p0 = const, which is, in the end, nothing other than a reproduction of the Galilean law of
inertia.
The law of conservation of momentum is much more interesting in the case of several
interacting mass points: For example, taking into account p⃗i = mi⃗vi , the following applies
for two mass points m1 and m2 :

p⃗˙1 = F⃗e1 + F⃗12 , (3.11)

p⃗˙2 = F⃗e2 + F⃗21 , (3.12)

where F⃗ei is the external force acting on the i-th mass point and F⃗ij is the interaction
force that the j-th mass point is exerting on the i-th mass point. Taking into account
Newton’s third law, the sum of both equations yields:
d
p1 + p⃗2 ] = F⃗e1 + F⃗e2 + F⃗12 + F⃗21
[⃗
dt | {z } | {z } | {z }
p
⃗ ⃗e
F 0

and thus the law of conservation of momentum is

p⃗˙ = F⃗e , (3.13)

which states that the total momentum of a mass point system only changes due to the
sum of all external forces.
In the special case that there are no external forces present or that their sum is zero,
conservation of momentum applies:

p⃗˙ = ⃗0 =⇒ p⃗ = p⃗0 = const . (3.14)

Besides the law of conservation of energy, the law of conservation of momentum is another
first integral of the equations of motion, since it has only first derivatives with respect
to the position and no second derivatives, as in the original equations of motion. When
conservation of momentum applies, problems can be simplified significantly, as is shown
in the following examples:

3.2.1 Examples of conservation of momentum

1. Completely inelastic collision: See exercise A.3.1.
2. Deriving the rocket equation
The rocket principle is based on the fact that a missile ejects a part of its mass
with as much momentum as possible against the flight direction, so that, due to the
conservation of momentum, an opposite momentum is transferred to the rocket and
thus provides a forward thrust. From this basic principle, we can easily derive a
simple equation to find the terminal velocity of the rocket: Suppose that the rocket
ejects matter of mass ∆m in a time interval ∆t, with a relative velocity −v0 (as seen
from the rocket), which corresponds to a velocity of v − v0 from the point of view
of an observer resting on the ground, with v being the velocity of the rocket. The
1
Note the analogy of the relation Q̇ = I between electric charge and electric current.

52
3.2. MOMENTUM AND CENTRE OF MASS

Figure 3.1: On the rocket principle.

latter increases by ∆v after ejection of the matter. If there are no external forces to
be taken into account, the conservation of momentum requires the equality

(m + ∆m) v = m (v + ∆v) + ∆m (v − v0 )

of the rocket’s momentum before the ejection of matter (mass m + ∆m) and the
sum of the rocket’s momenta after the ejection on the one hand, and of the matter
ejected on the other hand. This equation can be solved for
∆m
∆v = v0 .
m
In the limit ∆t → 0, ∆v changes into dv and ∆m into −dm (negative change in
mass!), which, after integration over the entire flight and assuming a launch from
rest, finally yields the terminal velocity
ˆvE ˆmE
dm mA
 
vE = dv = −v0 = −v0 [ln (mE ) − ln (mA )] = v0 ln .
m mE
0 mA

This relationship, which is called Tsiolkovsky’s rocket equation, represents the ter-
minal velocity in dependence on the velocity of the propulsion jet and on the ratio of
the initial mass mA to the final mass mE . In contrast, neither the type of propulsion
nor the exact sequence of the flight are relevant.

The centre-of-mass theorem

The total mass and the centre of mass (centre of gravity) of a system of mass points are
given by
N N
1 X
m= ⃗rs =
X
mi , mi⃗ri
i=1
m i=1
as is well-known. Using the transformation
N N N
d X d 1 X
" # " #
p⃗ = ˙
mi⃗ri = mi⃗ri = m mi⃗ri = m⃗r˙s
X

i=1
dt i=1
dt m i=1

the momentum is written as a product of the total mass and the velocity of the centre of
mass. The law of conservation of momentum (3.13) then has the alternative form

m⃗r¨s = F⃗e . (3.15)

53
 3.3. LAW OF CONSERVATION OF ANGULAR MOMENTUM

This means that a system’s centre of mass moves in the same way as a single mass point
that has the total mass of the system and is influenced by the sum of all external forces.
Especially in isolated systems on which no external forces act, Eq. 3.15 can be integrated,
implying a uniform motion of the centre of mass:
⃗rs (t) = ⃗rs0 + ⃗vs0 t .
If we move to an alternative reference frame that performs the same motion as that of the
centre of mass (Galilean transformation), we can even reach the case
⃗rs = ⃗0 .

3.2.2 Example of the centre-of-mass theorem

It is a widespread idea that the Moon is orbiting a stationary Earth. In fact, however,
this contradicts the centre-of-mass theorem, because the Earth-Moon system’s common
centre of gravity, which is located at a distance of
mM · 380000km 1 380000
s = MS ≈ = 380000km ≈ km ≈ 4600km
mE + mM 1 + mM /mE 82
from the centre of the Earth, would describe a circle of radius s within one month, as
shown in Fig. 3.2. In fact, the centre-of-mass theorem requires that the centre of gravity
is a fixed point of motion and not the centre of the Earth, which leads to the conclusion
that both bodies orbit their common centre of gravity.

⃗vM Moon ⃗vM Moon

S S
⃗vE
M M
Earth Earth

(a) (b)

Figure 3.2: Motion in the Earth-Moon system: the naive idea that the moon orbits the
Earth’s centre of gravity being at rest, (a) is in contradiction with the law of conservation
of momentum and the centre-of-mass theorem, while the assumption that both bodies
orbit the common centre of gravity (b) is consistent with these conservation laws.

3.3 Law of conservation of angular momentum

If, for a mass point, we calculate the cross product of the position vector ⃗r with Newton’s
equation of motion, we obtain
m⃗r × ⃗r¨ = ⃗|r × F⃗ ,
{z }
(3.16)
⃗
M

54
3.3. LAW OF CONSERVATION OF ANGULAR MOMENTUM

*
 y

  y
i
P 

Pφ

P PP
PP
P
q
P
x



6
?
Figure 3.3: Circular and spiral motion.

where M⃗ is the torque of the force with respect to the origin. The left-hand side can, in
turn, be written as the time derivative
d  d h
m⃗r × ⃗r¨ = m ⃗r × ⃗r˙ − m ⃗|r˙ {z
× ⃗r}˙ = m⃗r × ⃗r˙
 i
dt dt
0

of a quantity
⃗ = m⃗r × ⃗r˙ = m⃗r × ⃗v
L (3.17)
which is called the angular momentum of the system. Overall, the law of conservation of
angular momentum is written in the form

⃗˙ = M
L ⃗ . (3.18)

In the special case that there are no external and internal torques, the angular momentum
is a constant of motion, i.e. L
⃗ =L ⃗ 0 = const.

3.3.1 Examples of conservation of angular momentum

1. Circular and spiral motion
If you attach a body of mass m to a rope which is passed through a pipe, as is shown
in Fig. 3.3, you can force the body into a circular motion with a chosen radius or
into a spiral motion with a time-varying radius by pulling out or easing the rope at
the lower end of the pipe after a single push, as long as the motion is fast enough
for the rope force to be much greater than the weight force. Represented in polar
coordinates, the following applies, as usual,

⃗r = r⃗er
⃗r˙ = ṙ⃗er + rφ̇⃗eφ

for the position vector and the velocity. Here, r = r (t) is prescribed by the kinematic
guidance of the rope, while the angle φ = φ (t) is the unknown variable to be
determined from the equations of motion. Unless no other force acts from outside,
the rope force S⃗ always points in the negative ⃗er –direction and that the resulting
torque ⃗r × S is zero with respect to the origin. Since otherwise no force and thus
⃗
no other torque has to be taken into account, the system is torque-free and the
conditions for conservation of angular momentum are given, i.e.
⃗ = m⃗r × ⃗r˙ = mr⃗er × [ṙ⃗er + rφ̇⃗eφ ] = mr2 φ̇⃗ez = L
L ⃗ 0 = const ,

55
 3.3. LAW OF CONSERVATION OF ANGULAR MOMENTUM

where L
⃗ 0 = mr2 ω0⃗ez results from the initial data r0 = r (0), ω0 = φ̇ (0) of the
0
motion. By solving the above equation, the angular velocity is represented as:
2
r0

ω (t) = φ̇ (t) = ω0
r (t)
in dependence on the radius r (t). If, for example, the rope reaches half of its length,
the angular frequency is quadrupled. It is irrelevant whether the rope is pulled in
quickly or slowly, since only the initial and final states are important.
2. Kepler’s second laws of planetary motion

Exercises
Exercise 3.1: Fully inelastic collision
(Solution cf. A.3.1)
Two bodies with masses m1 and m2 and velocities ⃗v1 and ⃗v2 collide at a time t = tK and
merge together to form a body of mass m1 + m2 , having the velocity ⃗v .
1. Make a descriptive sketch of this process. Note that both the speeds of the two
bodies and the directions of motion can be different.
2. How do you calculate the final velocity ⃗v after the collision, given the masses and
velocities of the two bodies before the collision?
3. Determine the difference T − T ′ of the kinetic energy before and after the collision.
Explain the result.
4. Calculate final velocity ⃗v and energy loss T − T ′ for the following data: m1 = 3g,
m2 = 2g, ⃗v1 = [3⃗ex − 2⃗ey ] m/s and ⃗v2 = [4⃗ex + 1⃗ey ] m/s.

Exercise 3.2: Fully elastic non-central collision

(Solution cf. A.3.2)
A ball with mass m1 and initial velocity ⃗v1 collides with a second, stationary ball with mass
m2 completely elastically, i.e. without loss of mechanical energy. However, the collision
occurs non-centrally, so that the two spheres move in different directions after the collision,
furthermore both spheres are smooth. In order to describe this process appropriately, the
coordinate system is adapted as follows:
y ⃗v1 ′ y

m1
x β x
α α ⃗v2 ′

⃗v1 m2 m2
m1

Let the x-axis be defined by the connecting line of the centres of both balls at the moment
of collision. Since, due to the assumption of smooth balls, only normal forces act during
the collision, the movement of the second ball after the collision can only take place in the
x-direction, i.e. ⃗v2 ′ = v2′ ⃗ex .
Given both masses m1 and m2 , the entry angle α of the first ball and its orbital velocity

56
3.3. LAW OF CONSERVATION OF ANGULAR MOMENTUM

v1 = |⃗v1 |, determine the exit angle β and the orbital velocity v1′ = |⃗v1 ′ | of the first ball
and the orbital velocity v2′ of the second ball. Proceed as follows:
1. Determine the x-component and the y-component of the law of conservation of mo-
mentum.
2. Formulate the law of conservation of energy for this system.
3. Solve the system of three equations obtained from the previous parts of the problem
according to the three unknowns v1′ , v2′ and β.
4. If the second sphere becomes larger and more massive, the limiting case of the “fixed
wall” finally results for m2 → ∞. What are then the results for v1′ and β?

Exercise 3.3: Elastic push in two directions

(Solution cf. A.3.3)
Between two masses moving frictionlessly on a horizontal plane in the ratio m2 /m1 = 3/7,
a spring with stiffness C is compressed to half its relaxed length l0 (sketch A). Starting
from this, in the initial phase (sketch B) both masses are accelerated by the relaxing spring
until the spring is fully relaxed and detaches from the masses, which then move away with
constant velocities v1 , v2 in the final phase of movement (sketch C).

l0 /2 A x2 B v1 C v2
x1
m1 m2

1. What is the initial (state A) potential energy stored in the spring?

2. Set up Newton’s equations of motion for the initial phase B for both masses, based
on their x-coordinates x1 , x2 (solution not required!).
3. Determine the final velocities v1 and v2 as a function of given quantities by making
use of two independent conservation laws.
4. Calculate v1 and v2 for the following data: m1 = 140g, C = 3N/cm, l0 = 10cm.

Exercise 3.4: Shaky pendulum

(Solution cf. A.3.4)
A gravity pendulum consisting of a massless rope of length l0 and a pen- m1
x
dulum bob of mass m2 is attached to a carriage of mass m1 moving on
horizontally mounted rails. The deflection of the pendulum from its equilib-
l0
rium position is given by the angle φ with respect to the vertical. Initially,
φ
the carriage is at rest and the pendulum deflected by φ = φ0 . Friction is m2
irrelevant.
1. What movement does the carriage make as the pendulum bob approaches the lowest
point of its path and why?
• The carriage moves to the left, caused by the swing of the pendulum mass.
• The carriage remains at rest because the weight force acts only in the vertical
direction.
• The carriage moves to the right, as both x-position of the system centre of
gravity and horizontal component of momentum are conserved.

57
 3.3. LAW OF CONSERVATION OF ANGULAR MOMENTUM

2. At what path velocity v2,max does the pendulum bob pass through the lowest point
of its path? Consider your statement about the motion of the carriage according to
part (1) and use the associated and another conservation law.
3. Calculate v2,max for the data m1 = 2,1kg, m2 = 700g, l0 = 50cm and φ0 = π/3.

Exercise 3.5: Skiing

(Solution cf. A.3.5) 7a
7


A skier (mass m) is skiing down a hemispherical hill- R
top (radius R). He starts at the time t = 0 at the y
6 
summit (φ = π/2), approximately from rest. The


friction is negligible. φ -x

a) What is the speed of the skier at any position given by the angle φ?
b) Determine the normal force (constraint force) acting on the skier as a function of his
current position. as a function of the skier’s current position, i.e. FN = FN (φ).
c) From the result of problem part (b), deduce that the skier takes off at a certain
angle φ = φ1 . Determine φ1 .

58
Chapter 4

Dynamics of rigid bodies for

motion in a plane

In principle, a rigid body can be understood as a system of an infinite number of mass

points that occupy fixed positions relative to each other. In this sense, their dynamics can
in principle be described within the framework of standard Newtonian mechanics, without
any need to formulate new axioms. However, we will avoid Newton’s equations of motion
and use momentum balance and angular momentum balance instead.

4.1 The Newton-Euler equations of motion

For the sake of simplicity, we choose a rigid body with a particularly simple shape, namely
a rod of mass m and length l, as is shown in Fig. 4.1. The equations of motion will be

P
r′ φ
⃗r
S
⃗rs
x
Figure 4.1: Determining the position of the centre of mass and orientation of a body as
well as the position of a material point P.

derived using this model body as an example, but all considerations can ultimately be
transferred to bodies of any shape. As was already discussed in section 1.2, the collective
motion of the body is defined by a position vector ⃗rs with respect to the centre of mass,
in the case of the rod with respect to its centre, and by a rotation angle φ in relation
to a reference orientation (e.g. a rod in horizontal position), more precisely, by their first
derivatives, i.e. the velocity ⃗vs = ⃗r˙s of the centre of mass and the angular velocity ω = φ̇
and the second derivatives, i.e. the acceleration ⃗as = ⃗r¨s of the centre of mass and the
angular acceleration α = ω̇ = φ̈. However, these two quantities describe only a global
motion of the entire body. If we want to conceive the body as a system of mass points

59
 4.1. THE NEWTON-EULER EQUATIONS OF MOTION

and derive the corresponding equations of motion, we will actually need the position and
motion of each individual material point.
In order to first determine the position vector and velocity of any material point P of
the rod shown in Fig. 4.1 at a distance r′ from its centre of mass, one arrives at the
following relationships by means of geometric considerations, the laws of vector addition
and subsequent differentiation:

⃗r = ⃗rs + r′ cos φ⃗ex + r′ sin φ⃗ey (4.1)

⃗v = ⃗r˙s + r′ φ̇ [− sin φ⃗ex + cos φ⃗ey ] (4.2)

Now we consider a small piece of the rod from the position r′ of point P to r′ + dr′ , i.e.
a small piece of length dr′ that is adjacent to point P, and treat it as a mass point. Its
momentum is calculated as:

p = dm⃗v = ϱAdr′ ⃗vs + ωr′ (− sin φ⃗ex + cos φ⃗ey ) = ϱA⃗vs dr′ + (· · · ) r′ dr′
d⃗
 

To determine the total momentum of the rod, we have to sum up all the small momentum
contributions d⃗
p, which leads in the limit dr′ → 0 to the integral

ˆl/2 ˆl/2
p⃗ = ϱA⃗vs dr′ + (· · · ) r′ dr′ = m⃗vs (4.3)
−l/2 −l/2

and thus to a surprising insight: Although each point of the rod can have a different
velocity, some faster, some slower, in the end all differences compensate one another, so
that the total momentum is simply composed of the total mass m and the velocity ⃗vs of
the centre of mass, as is the case with a mass point.
As a second quantity that is very important in dynamics, we consider the angular mo-
mentum as defined by (3.17) for a mass point. Thus the angular momentum of the small
piece of rod can be calculated from r′ to r′ + dr′ by

dL
⃗ =dm⃗r × ⃗v
=ϱAdr′ ⃗rs + r′ (cos φ⃗ex + sin φ⃗ey ) × ⃗vs + ωr′ (− sin φ⃗ex + cos φ⃗ey )
   

=ϱA⃗rs × ⃗vs dr′ + (· · · ) r′ dr′ + ϱAr′2 dr′ ω (cos φ⃗ex + sin φ⃗ey ) × (− sin φ⃗ex + cos φ⃗ey ) ,
| {z }
(cos2 φ+sin2 φ)⃗ez =⃗ez

and the total angular momentum, again, via integration as

ˆl/2 ˆl/2 ˆl/2

′ ′ ′
⃗ = ϱA⃗rs × ⃗vs
L dr + (· · · ) r dr + ϱA r′2 dr′ ω⃗ez = m⃗rs × ⃗vs + Jω⃗ez (4.4)
−l/2 −l/2 −l/2
| {z }
=:J

whereby a new quantity appears here, namely the moment of inertia:

ˆl/2
l3 1
J = ϱA r′2 dr′ = ϱA = ml2 (4.5)
12 12
−l/2

of the rod. If the rod is replaced by another body, the moment of inertia J changes
according to the shape of the body.

60
4.2. EXAMPLE

The decisive factor is that the rod’s angular momentum, in contrast to the linear momen-
tum, is not calculated in exactly the same way as it is for a mass point, because that
would be only the first term m⃗rs × ⃗vs in equation (4.4), which is called orbital angular
momentum. The orbital angular momentum is complemented by the intrinsic angular mo-
mentum, which is also called spin and whose magnitude can be calculated as the product
Jω of the moment of inertia and the angular velocity. The latter relationship applies not
only to rods but also to bodies of any shape, if J is the moment of inertia corresponding
to the respective body.
The fundamental equations for the dynamics of the rod can now be derived from the law
of conservation of momentum (3.13) and the law of conservation of angular momentum
(3.18) by plugging in the previously determined quantities (4.3) and (4.4):
F⃗ =p⃗˙ = m⃗r¨s
⃗˙ = d m⃗rs × ⃗r˙s + Jω⃗ez = m ⃗r˙s × ⃗r˙s +m⃗rs × ⃗r¨s + J ω̇⃗ez
h i
⃗ =L
M
dt | {z }
0

The first equation is nothing other than the well-known Newtonian equation of motion,
which is thus just as valid for rigid bodies as for mass points. If we use these equations
and replace m⃗r¨s with F⃗ in the penultimate term of the second equation, we obtain:
J ω̇⃗ez = M
⃗ − ⃗rs × F⃗

Here, M⃗ is the sum of all moments acting on the body, calculated with respect to the origin
of the chosen coordinate system. The other term, ⃗rs × F⃗ , is the moment of the resultant
force acting on the body, viewed from the origin of the chosen coordinate system. The
difference
M⃗s = M⃗ − ⃗rs × F⃗
is thus the moment calculated with respect to the centre of mass (index s), see also the
corresponding equations in engineering mechanics. The moment is naturally a vector, but
it can be reduced to a scalar for the motion in a plane (here, too, compare with engineering
mechanics).
Combining all this, we now obtain Newton’s equation of motion and Euler’s equation of
motion (simplified):
m⃗r¨s =F⃗ (4.6)
J φ̈ =Ms (4.7)
These are the two key fundamental equations with which we can describe any motion of
a rigid body in a plane. Motion in space is a subject of the course “Applied Mechanics 2”
and is therefore not discussed here.

4.2 Example
As a simple example of uniaxial rotation (in combination with translation), consider a
cylinder rolling down an inclined plane, see Fig. 4.2 (left).
This is a simultaneous motion between translation, which is given by the displacement xs
of the centre of mass, and rotation, which is given by the angle φ. Assuming that there is
no slippage, we must take into account the kinematic bond between the two motions
xs
φ= (4.8)
R

61
 4.2. EXAMPLE

Figure 4.2: Position plan and free-body diagram of a cylinder rolling down an inclined
plane.

which directly results from the definition of the angle in radians. After isolating the
cylinder from its surroundings, we obtain the free-body diagram in Fig. 4.2 (right), which
leads to the equations of motion (or the static equilibrium for the y-direction) as

↙: mg sin α − H = mẍs ,
↖: N − mg cos α = 0 ,
↶
S: HR = J φ̈ .

If we multiply the first equation with R and add the third equation, we obtain the equation

mgR sin α = mRẍs + J φ̈ ,

where we first eliminated the unknown adhesive force H. Now we have yet to take into
account the kinematic bond (4.8) in order to replace φ̈ with ẍs /R, which finally leads to
J
 
mgR sin α = mR + ẍs .
R
Here you can see that this is a motion with a constant acceleration
mR2
ẍs = g sin α
mR2 + J
so that integrating the equation of motion twice finally yields the solution
mR2 t2
xs (t) = g sin α + v0 t + x0
mR2 + J 2
where v0 is the initial velocity and x0 the initial position.
How inert the cylinder actually is, now depends on J and thus on the nature of the
cylinder: For a solid cylinder, J = mR2 /2 and thus ẍs = 2g sin α/3 apply, whereas a
hollow cylinder of the same mass has a moment of inertia of J = mR2 , which leads to
the smaller acceleration ẍs = g sin α/2. Comparing the motions of a solid and a hollow
cylinder rolling down an inclined plane is one of the classic experiments in mechanics!

62
4.3. ENERGY AND THE PARALLEL AXIS THEOREM

4.3 Energy and the parallel axis theorem

4.3.1 Kinetic energy of rotation
Using equation (3.3), we defined power as a product of force and velocity. The following
considerations show that an analogous definition of power can also be found for rotational
motion: For this purpose, consider a rod of length 2r that is set into a rotary motion by
means of a pair of forces (two forces F acting perpendicularly on the rod’s two ends in
opposite directions), as is shown in Fig. 4.3. According to (3.3), power is calculated as

F
v
r
ω
S

v
F
Figure 4.3: A moment generated by a pair of forces drives a rotary motion.

P = 2F v, where v is the magnitude of the velocity of the two points where the forces are
applied. The latter is given by the well-known equation v = ωr, where ω is the angular
velocity of the rotary motion. On the other hand, M = 2F r is the resultant moment of
the force pair, so that we can write

P = 2F v = 2F ωr = |{z}
2F r ω
M

and thus finally

P = Mω . (4.9)
The analogy to equation (3.3) is unmistakable! However, one can push this analogy even
further by expressing the moment by means of Euler’s equation of motion (4.7) using
M = Jω, which leads to
d J 2
 
P = M ω = Jω ω̇ = ω
dt, 2
where the term in square brackets is the kinetic energy of the rotary motion, i.e.:
J 2
T = ω . (4.10)
2

If we now consider a combined motion of translation and rotation, their parts add up as
follows:1
m J
T = ⃗vs 2 + ω 2 . (4.11)
2 2
Here, ⃗vs is, as always, the velocity of the centre of mass.

4.3.2 The parallel axis theorem

Fig. 4.4 shows an example of a rod rotating around a point P with the angular velocity ω.
Point P, in turn, is at a distance r from the centre of mass S. In the approach we have
been using so far, the motion of the body is a combination of a translational motion of
1
We waive the proof!

63
 4.3. ENERGY AND THE PARALLEL AXIS THEOREM

r P

Figure 4.4: Deriving the parallel axis theorem based on energy-related considerations.

the centre of mass on a circular path with radius r around P and the tangential velocity
vs = ωr, and the rod’s spin around its centre of mass. According to (4.11), we then obtain
the kinetic energy
m 2 J 2 1h i
T = ⃗vs + ω = J + mr2 ω 2 , (4.12)
2 |{z} 2 2 | {z }
(ωr)2
J′

which shows that one can also formally regard the motion simply as pure rotation2 , pro-
vided that the moment of inertia J with respect to the centre of mass is replaced with
another moment of inertia, namely

J ′ = J + mr2 . (4.13)

This important relationship is called the parallel axis theorem. It can also be used for
finding the moment of inertia of composite bodies, as is the task in, for example, A.4.3.

Exercises
Exercise 4.1: Recently at the discounter . . .
(Solution cf. A.4.1)
At the discount store, a customer places a can of sauer- y

6
kraut (radius R, mass m, moment of inertia J = mR /2) 2
~
 -x
across the treadmill at the checkout. The belt jerks from  m m
rest at a speed v0 . The motion of the can is conveniently 
? v 0

?
described in a reference frame that moves along with the treadmill. In it, the belt is at
rest while the can’s initial speed is v0 (sketch A). Subsequently, the can slows down due to
friction and begins to rotate at an accelerated rate (sketch B) until the angular velocity
and centre of gravity velocity are in a suitable relationship for slip-free rolling (sketch C).
y y y
(A) (B) (C)

ω =0 v0 ω v ω1 v1
R
x µ x x

1. State the relation in which the angular velocity ω1 and the velocity v1 must be in
relation to each other for slip-free rolling in the final state C.
2
Which is no doubt easier!

64
4.3. ENERGY AND THE PARALLEL AXIS THEOREM

2. In the transition phase B there is a Coulomb friction force FR = µN with friction

coefficient µ and normal force N . Cut the can free and set up Newton’s and Euler’s
equations of motion (solution not required!).
3. Justify why in the transition phase B the total angular momentum (with respect to
the origin of the coordinates) is a conservation quantity and use it together with the
result from part 1 to determine the final angular velocity ω1 and the final centre-of-
mass velocity v1 .
Note: As an alternative to the angular momentum theorem, you can also solve the
equations of motion from part (b), which requires only a few more calculation steps.

Exercise 4.2: Recently at the DIY store

(Solution cf. A.4.2)
You place a board with an inclination α on a magazine trolley (total mass m1 ) and a rolling
mattress (mass m2 , radius R, moment of inertia J = m2 R2 /2) on top. Unfortunately, you
attach it insufficiently so that it comes loose and rolls down the board.

S2
y

ω
x
S2 ⃗v1
h S1 S1
⃗v2
α α

1. While the mattress rolls, the magazine trolley moves in the opposite direction. How
does this explain itself? Choose one of the following alternatives:
◦ Conservation of kinetic energy. The latter is initially zero, so that of the trolley
is negative to compensate for the positive kinetic energy of the mattress.
◦ Conservation of the horizontal component of the momentum, equivalent to the
invariability of the x-coordinate of the total centre of mass.
◦ Conservation of the vertical component of the momentum, equivalent to the
invariability of the y-coordinate of the total centre of mass.
2. Kinematic no-slip constraint: If the trolley were fixed, the relationship |⃗v2 | = Rω
between the velocity ⃗v2 of the centre of gravity S2 and the angular velocity ω of
the rolling mattress would apply. However, since the trolley is moving, its velocity
⃗v1 = v1⃗ex must be added to that of the pure rolling motion (vectorial) and the fol-
lowing applies . . .
◦ ⃗v2 = (v1 − ωR sin α) ⃗ex + ωR cos α⃗ey ◦ ⃗v2 = (v1 − ωR cos α) ⃗ex + ωR sin α⃗ey
◦ ⃗v2 = (v1 − ωR sin α) ⃗ex − ωR cos α⃗ey ◦ ⃗v2 = (v1 − ωR cos α) ⃗ex − ωR sin α⃗ey
3. Substitute the form of ⃗v2 chosen in (b) into the conservation law chosen in part
(a) and obtain from it a representation of v1 as a function of m1 , m2 , R, α and the
angular velocity ω.
4. Assuming conservation of total energy and using previous findings, determine the
angular velocity ω of the rolling mattress just before it touches the ground. Calculate

65
 4.3. ENERGY AND THE PARALLEL AXIS THEOREM

m1 = 20kg, m2 = 40kg, R = 30cm, α = 45°, as well as height difference h = 80cm

and gravitational acceleration g = 9.8m/s2 .
5. Finally, also give the final velocity v1 of the magazine trolley.

Exercise 4.3: Moment(s) of inertia of a T-arrangement

(Solution cf. A.4.3)
Two identical rods with respective mass m and length l are joined to a T-shaped structure,
see following image:
y

x
S1
S2

Here S1,2 denotes the partial centre of mass of the respective rod.
1. Where is the overall centre of mass of the arrangement? Give its coordinates and
indicate it in the picture above.
hint: The centre of mass can also be determined without a calculation formula using
common sense!
2. Use the parallel axis theorem to determine the moment of inertia Jz for a rotation
around the axis through the centre of mass oriented in the z direction.
Note: For a single rod rotating about its centre of gravity axis perpendicular to the
rod axis, the moment of inertia is ml2 /12.
3. What moment of inertia Jx results for a rotation about the x axis?
4. And because it is so nice, finally calculate the moment of inertia Jy for a rotation
around the centre of gravity axis oriented in the y direction.

66
 Part II

Thermo- and Fluid Dynamics

67
Chapter 5

Thermodynamics

Thermodynamics has a special role among the various disciplines of physics, as it pri-
marily describes the material and energy transfer between systems of different natures.
As an overarching discipline, thermodynamics can be considered a science that acts as
an interface to other disciplines, especially to chemistry. The old term theory of heat is
outdated and does not do justice to the importance of thermodynamics.
Thermodynamics is often divided into classical phenomenological thermodynamics and
statistical physics, but it is difficult to draw a clear line between the two — which is
why this classification will not be used below. The first chapter on this subject area
deals with statistical models and shows an attempt to relate the known phenomena of
thermodynamics to the laws of mechanics.

5.1 States of matter

An essential feature of thermodynamics is its relation to the internal structure of matter.
Since it has been known that matter is made up of atoms or molecules, attempts have
been made to understand their properties by means of microscopic models. These include,
in particular, the observed different states of matter.
Long before the existence of atoms was known for certain, people in ancient Greece held
the view that the world consisted of four elements: Earth, Water, Air and Fire. This
sounds naive from today’s point of view, but if we consider “elements” in this context not
in the sense of chemical elements, but rather as states of matter, then this classification
does make sense:
Earth corresponds, in this context, to the solid state of matter, where atoms/molecules
are arranged in a crystalline structure in which each atom/molecule occupies a reg-
ular place. A solid, therefore, has a defined shape in the unstressed state.
Water stands for the liquid state of matter, where atoms/molecules maintain direct
contact with each other due to interaction forces, but without having fixed neigh-
bours. This mobility of the atoms/molecules in relation to each other causes the
formlessness of liquids, but their permanent direct contact with each other explains,
in a way, why a liquid occupies a certain volume at a given pressure and temperature.
Air represents the gaseous state of matter, where the interaction between atoms/-
molecules is not sufficient to ensure cohesion of these particles. As a result, the

69
 5.2. PRESSURE AND TEMPERATURE

atoms/molecules move in all directions and are randomly distributed, which explains
why a gas tends to occupy every volume available.
Fire can be regarded as a descriptive paraphrase of the plasma state of matter. In the
plasma state, collisions have sufficient energy to ionise the colliding atoms/molecules,
which results in a mixture of positive ions and free electrons. In a plasma, these
particles move at a high velocity and at a great distance from each other, as in a
gas, but there are electrical forces of attraction between the electrons and the ions,
so that they strongly influence each other in their motion. External influence can be
exerted via electromagnetic fields. In addition, recombinations take place in which
electrons are captured by ions, whereby light is emitted, which explains the luminous
phenomena of the plasma.
Since there are commonalities between the four states of matter, they are also summarised
under the following generic terms:
Condensed matter is characterised by the cohesion of the building blocks of matter.
This includes both solids and liquids.
Fluids are matter that do not have a defined body shape. These are both liquids and
gases, but also plasmas in a broader sense.
Continua is the collective term for all states of matter.

5.2 Pressure and temperature

Pressure and temperature have long been known as phenomenological physical quantities.
The true nature of gases is to be determined by means of a kinetic model that describes
the motion of the atoms/molecules of a gas.

5.2.1 Kinetic model of pressure

In the following, we consider the model of an ideal gas. This is characterised by:
• A statistically distributed, disordered motion of the atoms/molecules and
• hardly any significant interaction of the atoms/molecules with each other due to
collisions.
Based on these assumptions, we consider a model system of N atoms/molecules of respec-
tive mass m in a cube of edge length a, see Fig. 5.1 (left). The atoms/molecules are both
spatially distributed statistically and have statistically distributed velocities (regarding
magnitude and direction). We assume that the average distance between the particles is
much greater than their size, so that there are very few collisions between the particles.
Therefore, we can assume an undisturbed motion of the particles and consider each par-
ticle individually and independently of the motion of the others. Furthermore, we ignore
the effect of gravity on the motion of the particles.
According to Galilei’s principle of inertia (Newton’s first law), each particle maintains its
state of motion, i.e. it moves at a constant velocity until it finally collides with one of the
walls, as shown in 5.1 (middle). We assume that the collision is completely elastic. In
addition, we use the advantage that we can freely choose the coordinate system by placing
the x-axis in the direction of the normal and the y-axis parallel to the wall, but in such a
way that the xy-plane coincides with the plane in which the path of the particle lies, see

70
5.2. PRESSURE AND TEMPERATURE

a
a
p⃗1
x
x
a
p⃗2

Figure 5.1: A simple kinetic model of an ideal gas. N statistically distributed particles in
a cube of edge length a (left), elastic collision of a particle with the wall of the container
(middle) and the illustration of its path in a plane (right).

5.1 (right). Because then we can make use of the results from exercise A.3.2: since the wall
is immobile, the energy of the particle is conserved during the collision process and, after
the impact, the particle moves away along a path that is the flipped direction of incidence
with respect to the x-axis. This means that in the momentum vector, the x-component is
inverted by the collision and the momentum transfer in x-direction is given by

∆p = (⃗
p2 − p⃗1 ) · ⃗ex = 2mvx , (5.1)

where vx is the x-component of the velocity. Since the total momentum is conserved due
to the absence of external forces, this momentum difference is transferred to the wall. In
line with Newton’s equation of motion in the alternative form (3.10), this leads to a force
acting on the wall, which, averaged over time, results accordingly from the relation

dpx ∆p 2mvx
F̄x = = = , (5.2)
dt ∆t ∆t
being related to the momentum transfer per time interval ∆t between two collisions.
However, the latter still has to be determined in order to be able to express F̄ by known
quantities. Let us assume that, after colliding with the wall, the particle reaches the
opposite wall unaffected by all other particles, is reflected there and then travels back to
the original wall with the same velocity vector, again unaffected by all other particles.
Next, the particle collides again with the original wall, and we obtain ∆t (similarly to
exercise A.1.1) as:
2a
∆t = , (5.3)
vx
so that, by applying this to (5.2), the (average) force F̄x can be determined as

2mvx vx mvx2
F̄x = = . (5.4)
2a a
This, however, is just the force that a single atom/molecule exerts on the wall through
collisions. If we now consider N atoms/molecules, their individual contributions add up
to a total force
N N 2
mvx,i mvx2
F̄x,ges = F̄x,i = =N (5.5)
X X
,
i=1 i=1
a a
where, in the last step, we have the mean value
N
1 X
vx2 = v2
N i=1 x,i

71
 5.2. PRESSURE AND TEMPERATURE

of the square of the velocity in x-direction.. Now we assume that the velocities are uni-
formly distributed in all spatial directions. We can therefore say that

1h 2 i 1
vx2 = vy2 = vz2 = vx + vy2 + vz2 = ⃗v 2
3 3

applies, independently of any coordinate system, so that (5.5) can also be written in the
direction-independent form
2N m⃗v 2
F̄n = , (5.6)
3a 2
which correlates the normal force acting on any wall with the average translational kinetic
energy.
The question arises, however, whether one should not consider a physically more appro-
priate quantity as an alternative to the normal force acting on the wall of the cube, i.e. a
quantity that no longer depends on the individual geometry of the container. A decisive
step in this direction is considering the particle density

N
n= , (5.7)
V

i.e. the number of particles per volume V = a3 , instead of the total number of particles.
Then we can rewrite the fraction N/a in equation (5.6) as N/a = nV /a = nA, with A = a2
being the face of the cube. With these substitutions and after dividing by A, equation
(5.6) becomes
F̄n 2n m⃗v 2
p := = . (5.8)
A 3 2

5.2.2 Kinetic model of temperature

Experimental determination of the temperature The reaction of a known sub-
stance (expansion, phase transition) allows conclusions to be drawn about the temperature
or its change. A very elegant measurement method is given by the thermocouple.

Empirical equation of state for ideal gases: Based on comprehensive experiments

by Boyle-Mariotte, Gay-Lussac and Charles, the following empirical relationship was found
between pressure p, volume V , temperature T and amount of substance ν:

pV = νRT (5.9)

The amount of substance ν = N/Na is the number of atoms/molecules N of a substance

divided by the Avogadro constant NA = 6.022140857 · 1023/mol, and R = 8.3144598 J/mol K
is the universal gas constant.

Comparison of the experimental findings with the theoretical micro-model

When we compare the above equation of state for ideal gases (5.9) with the fundamental
equation of the kinetic theory of gases derived from model considerations, pV = N m⃗v 2 /3,
a consistent picture emerges if we write

1
νRT = N m⃗v 2 (5.10)
3

72
5.2. PRESSURE AND TEMPERATURE

f =3 f =5 f =6

Figure 5.2: Degrees of freedom in the motion of atoms (left), of linear molecules (middle)
and of angled/3D molecules (right).

and thus correlate the temperature with the average kinetic energy of the atoms/molecules.
This correlation becomes more evident when we substitute the amount of substance accord-
ing to ν = N/Na , divide by 2N/3 and define the ratio kB := R/NA = 1.38064852·10−23 J/K
as the Boltzmann constant. (5.10) can then be written as:

3 1
kB T = m⃗v 2 (5.11)
2 2
Here we can see that the temperature is nothing other than the average kinetic energy of a
gas atom, distributed over its three degrees of freedom. Thus we succeeded in understand-
ing another macroscopic quantity within the framework of a kinetic micro-model. Since
the unit of temperature has historically been the kelvin, but that of energy the joule, we
need the Boltzmann constant kB as a conversion factor between the two units.

5.2.3 Degrees of freedom in molecules

An atom has actually only f = 3 degrees of freedom, which are assigned to the translational
motions in three independent spatial directions. It is theoretically possible that the atom
can rotate around its three independent axes, but the moment of inertia J is negligibly
small for this type of movement, since the mass of the atom is approximately concentrated
in its nucleus. The final reason why these degrees of freedom aren’t induced is of a
quantum-mechanical nature and won’t be discussed here.1

A molecule, by contrast, has a moderate moment of inertia and therefore also rotational
energy. The general three-dimensional molecule2 or the two-dimensional angled or ring-
shaped molecule has three independent rotational axes and a moderate moment of inertia
and thus, together with the three translational degrees of freedom, a total of f = 6
degrees of freedom. If the molecule consists of two atoms, however, there is no rotation
around the molecular axis and N = 5 degrees of freedom remain, see Fig. 5.2. Linear
polyatomic molecules such as CO2 or cyanogen (NCCN), . . . can, however, also have
bending vibrations.

In all cases, the available kinetic energy is statistically distributed over all available f
degrees of freedom, so that, for molecules, we have to replace equation (5.11), which is
related to the kinetic energy, with the more general form

f
kB T = Ekin . (5.12)
2

From the equation above, we also find that 12 kB T is the kinetic energy per molecule per
degree of freedom.
1
The reason lies in the large energy gap between the rotational ground state and the first excited state.
This amount of energy cannot be provided by the kinetic energy of the atoms at moderate temperatures.
It is only at very high temperatures that these “frozen” degrees of freedom begin to “unfreeze”.
2
e.g.: methane, ammonia, sulfur hexafluoride, . . .

73
 5.3. MAXWELL–BOLTZMANN DISTRIBUTION

5.3 Maxwell–Boltzmann distribution

If we regard temperature as the average energy per molecule and degree of freedom, we
can then infer the average velocity of a molecule from the temperature. However, a mean
value does not say much about how the velocities of different molecules are distributed
around this mean value. This requires some statistical considerations, which are not made
here. The result of such statistical considerations is the Maxwell-Boltzmann distribution.

3
2
!
mv 2
r
m

2
p(v) = v exp −
2
, (5.13)
π kB T 2kB T

which states the relative frequency with which molecules have the respective (path) velo-
city v. Fig. 5.3 shows this distribution at two different temperatures using the example of
nitrogen.

f /10−3 [s/m]
300K
1 N2

0.5 1200K

v/[m/s]
0
0 500 1000 1500 2000
Figure 5.3: Maxwell Boltzmann distribution of velocities at two different temperatures
using the example of nitrogen.

5.4 Internal energy and heat capacity

The statistical motion of atoms/molecules goes unnoticed, therefore the total energy as-
sociated with this motion is called internal energy U . Its amount can be comfortably
determined from the product of the average kinetic energy and the number of particles N ,
in accordance with (5.12), i.e.:

f f
U = N Ekin = N kB T = νRT , (5.14)
2 2
where, in the last step, we have again made use of the equality N kB = νR. The essential
statement here is that there is a proportional relationship between internal energy and
temperature3 . The proportionality factor between the two quantities,

f f
CV = N kB = νR , (5.15)
2 2
is called heat capacity at a constant volume. Using this, we can write (5.14) in a more
compact form as
U = CV T . (5.16)
3
similar to, for example, the relationship between the stored electrical charge in a capacitor and the
applied voltage.

74
5.5. CLASSIFICATION OF THERMODYNAMIC QUANTITIES

5.5 Classification of thermodynamic quantities

5.5.1 Extensive and intensive quantities, specific and molar quantities,
densities
There is an important classification in thermodynamics, namely the distinction between
extensive and intensive quantities:
Extensive quantities are quantities depending on the amount of material or the size of
the system, e.g. quantities that increase or decrease when the system becomes larger
or smaller. Examples of extensive quantities are mass, volume, amount of substance,
energy, heat capacity, . . .
By contrast,
Intensive quantities remain the same when the system size is changed. These include
typical compensation variables such as temperature and pressure, but also ratios of
extensive quantities such as concentration and density.
Each extensive quantity X can be assigned a specific quantity x = X/m by reference
to the mass m of the system. Alternatively, we can also use the amount of substance
ν as a reference quantity, which leads to the corresponding molar quantity xm = X/ν.
Finally, we can relate an extensive quantity to the volume, which leads to the associated
density ϱx = X/V of this quantity. Densities, specific and molar quantities are intensive
quantities.

Specific and molar heat capacity

The specific and the molar heat capacity, i.e. cV = CV /m and cV,m = CV /ν, are of
particular importance, because these quantities are substance-dependent. For the model
of the ideal gas, we obtain

f ν
cV = R, (5.17)
2m
f
cV,m = R, (5.18)
2
which yields the same values, especially for the molar heat capacity, for all gases whose
molecules have the same number of degrees of freedom.

Dulong-Petit law: Even though all considerations so far have been related to gases,
they can be easily transferred to condensed matter. For monatomic solids (e.g. metals),
Dulong and Petit found a molar heat capacity of cV,m = 3R, which corresponds to a
number of f = 6 degrees of freedom. Later, the quantum-mechanical analysis of such
solids led to the same result.

5.5.2 State variables, state functions and state space

The state of a thermodynamic system is determined by a certain number of independent
quantities, which are called state variables. In a broader sense, these quantities, and all
other quantities that are also determined by specifying the state variables, are referred to
as state functions and the functional dependencies that link different state functions with
each other are called equations of state. The latter individually depend on the respective
system.

75
 5.5. CLASSIFICATION OF THERMODYNAMIC QUANTITIES

p, V V
−dx V2 V1
Figure 5.4: Compression of a gas and the representation of this process in a p-V -diagram.

Just as in mechanics the position of a mass point in real space is determined by specifying
coordinates, the state of a thermodynamic system can be represented as a point in a
space, the so-called state space, as is shown in the lower diagram of Fig. 5.4. In the
present example, a gas trapped in a cylinder with a movable piston is represented by
pressure p and volume V .

5.5.3 Processes and process-dependent quantities

When the state of a system changes, this is referred to as a process. A process is represented
by a curve in the state space, see again Fig. 5.4, analogous to the path of a mass point in
mechanics.4
p

isochoric

adiabatic
¸
work done W = pdV

isothermal
V
V2 V1
Figure 5.5: A cycle.

If we have specified the starting point and the end point of a process, there are numerous
different process routes through the state space. Again, we use the gas shown in Fig. 5.4
as an example. The gas is trapped in a cylinder with a movable piston and is to be
compressed by a piston movement to the left:
• If the piston is moved slowly enough, so that a possible heating/cooling of the gas due
4
However, this analogy is not complete: While time plays an important role in the motion of a mass
point, especially the question of when the mass point is located at a certain point of the trajectory, such
a question is largely irrelevant for a thermodynamic process.

76
5.5. CLASSIFICATION OF THERMODYNAMIC QUANTITIES

to heat exchange with the surroundings by means of heat conduction is prevented,

and thus the temperature remains constant, the process is categorised as isothermal.
• If the piston is moved fast enough, so that there is no time for heat exchange with the
surroundings, or if the system is thermally insulated, such a process step is referred
to as adiabatic or also isentropic 5 . However, since the temperature changes in the
process, at the end of the process step one does not arrive at the same point in
the state space as after an isothermal process, but, for example, at a point of the
same volume (as in an isothermal process) and a higher/lower pressure - see also
Fig. 5.5. In order to reach the same final state as in the isothermal process, a further
process step is required. It is obvious that this would be an isochoric process, i.e. a
process in which the volume does not change, but the pressure changes accordingly
with temperature equalisation. As a result, the complete adiabatic-isochoric process
reaches the same final state as the isothermal process.
In principle, one can also pass through each step of the process in the opposite direction. In
the present example, isothermal compression can also be regarded as isothermal expansion
and the three steps, i.e. adiabatic compression, isochoric temperature equalisation and
isothermal expansion, can be combined to form a cycle. A cycle is a process in which, after
completion, all state functions return to their initial value. In the present example, these
are the state variables p and V as well as the dependent state variable T . In order to move
the piston, a force F = pA must be exerted at a given cross-section A and, ultimately,
when moving the piston by the distance −dx, the work done is:

δW = F (−dx) = −pAdx = −pdV . (5.19)

Thus the total work done in a process step γ between two states (1) and (2) is:
ˆ
W12 = − pdV , (5.20)
γ

which corresponds to the area under the curve in the p-V -diagram. Now, in the present
example, the work done in an isothermal process is less than the work done in an adiabatic-
isochoric process (see also Fig. 5.5), which makes sense when we consider that in adiabatic
compression there is an increased temperature and thus an increased pressure. In thermo-
dynamics, work done is therefore not a state function, because it does not depend solely
on the initial and final states, but on the process itself. Quantities like this are called
process-dependent variables. In the differential representation, these variables are given a
δ instead of a d in order to distinguish them from state functions.
We can gain an interesting insight if we again consider the adiabatic-isochoric-isothermal
cycle: Since less work comes out of the system from the isothermal expansion than was
previously expended in the adiabatic compression, a total amount of mechanical work is
lost in the cycle that is equal to the area enclosed by the cycle in Fig. 5.5. We can assume
that this work leaves the system in the form of heat dissipated to the surroundings. This
assumption can neither be proven nor disproven, but following Robert Mayer’s considera-
tions it led to the formulation of the first law of thermodynamics, which identifies heat as
another form of energy and postulates the conservation of total energy. What is interesting
in this context is the fact that an irreversible conversion of mechanical energy into heat
takes place here without mechanical friction being involved.
5
One often refers rather to entropy, which is a state function, than to heat, which is process-dependent.

77
 5.6. THE LAWS OF THERMODYNAMICS

Another insight is that heat Q, just like work W , is a process-dependent variable and thus
not a state function. Therefore, for infinitesimal changes of state, we write δQ and not
dQ for the heat.

5.6 The laws of thermodynamics

5.6.1 Temperature (zeroth law)
Introduction of temperature as a fundamental physical quantity: If a system A is in
thermal equilibrium with a system B and B is in thermal equilibrium with a system C,
then A is also in thermal equilibrium with C. The state function that coincides in these
systems is the temperature, which is scalar, intensive and the same everywhere in the
system.

Example: A thermometer is itself a system and shall be referred to as B. If B shows

the same temperature for a system A as for a system C, it can be concluded that A and
C will also be in thermal equilibrium with each other when they are brought into contact.
This law was formulated only after the other three laws. But since it is the basis of
thermodynamics, it was later referred to as the “zeroth” law.

5.6.2 Energy (first law)

Every system has an internal energy U (extensive property). This property can only
change by transferring energy in the form of work W and/or heat Q across the boundary
of the system, i.e.:
dU = δQ + δW (5.21)

Here, δW is the infinitesimal change in the work Instead of volume work, equivalent other
types of work that are also extensive properties can be used. For example, in a magnetic
system in a magnetic field H, increasing the magnetic moment m of a sample does the
extensive work δW = Hdm.

The energy of an isolated system remains unchanged. Different forms of energy can there-
fore be converted into each other, but energy can neither be created out of nothing nor can
it be destroyed. For this reason, a perpetual motion machine of the first kind is impossible
(a system cannot perform work without another form of energy being supplied and/or
without reducing its internal energy).

Thus, the energy of an isolated system is always constant and a perpetual motion machine
of the first kind, i.e. a machine that creates energy from nothing, is not possible!

A restriction on the conversion of heat into work only arises from the second law of
thermodynamics.

The first law of thermodynamics for the ideal gas

If we specifically use δW = −pdV for gases in general according to (5.19) and if we use
p = νRT /V according to the equation of state for ideal gases (5.9), the following applies
then to the work:
T
δW = −νR dV .
V

78
5.6. THE LAWS OF THERMODYNAMICS

If we apply this equation and the form (5.16) of the internal energy to (5.21), the first law
of thermodynamics for ideal gases then takes the form
T
CV dT = δQ − νR dV ,
V
where CV is the heat capacity as defined in (5.15). Solving this equation for δQ yields

T
δQ = CV dT + νR dV , (5.22)
V
which allows us to find the heat input during a process from a state (1) to a state (2)
according to ˆ ˆ
T
Q12 = CV dT +νR dV . (5.23)
γ γ V
| {z }
T2 −T1

There is no generally valid solution to the remaining integral of T /V on the right-hand

side of the equation, since the course of the process decides how T changes in relation to
V . We consider two special cases:
Isochorical process: Here, dV = 0 applies, so that (5.23) simplifies to:

Q12 = CV [T2 − T1 ] .

Isobaric process: From the equation of state (5.9) it follows that T /V is constant. Then
it follows from (5.23) that

T2 T1
 
Q12 = CV [T2 − T1 ] + νR V2 − V1 = [CV + νR] [T2 − T1 ] ,
V2 V1 | {z }
Cp

which can be written in the simple form Q12 = Cp [T2 − T1 ] based on the definition
of the heat capacity Cp = CV + νR at a constant pressure.
The comparison of both examples shows that heat is an “inconvenient” variable due to
its process dependence. We will show below that there exists an “alternative measure of
heat” in the form of a state function.

Integration of the first law and entropy

We again consider the first law for ideal gases (5.22) and divide it by the temperature, i.e.

δQ dT dV
= CV + νR ,
T T V
which leads to an equation that is integrable. This means that, due to the generally valid
mathematical identity (chain rule)

d X 1 X dX
   
ln = =⇒ d ln =
dX X0 X X0 X
with a given constant X0 , its right-hand side can be taken as the differential dS of a new
state function
T V
   
S = CV ln + νR ln , (5.24)
T0 V0

79
 5.6. THE LAWS OF THERMODYNAMICS

which is called entropy. Due to its definition, which does not only apply to ideal gases but
is generally valid,
δQtotal
dS = , (5.25)
T
entropy can be regarded as “weighted” heat. However, δQtotal = δQe + δQi is not merely
the heat δQe supplied to the system externally6 , but the sum of this heat and the heat
δQi produced internally in the system7 . Its special significance lies in the fact that, in
contrast to heat, entropy is a state function and thus, in a way, a “better measure of heat”
than heat itself.
However, entropy – like internal energy – cannot be measured directly. Rather, it can be
accounted for indirectly via the first law (5.21), taking into account (5.25), i.e. it can be
determined by solving the first law for dS,

dU p
dS = + dV (5.26)
T T
and then integrating the equation. Here, the form (5.19) of the work that is valid for any
gas was used. For ideal gases, this leads to the form (5.24) of the entropy, as shown above.

5.6.3 Entropy (second law)

In general, we distinguish between reversible and irreversible processes. Reversible pro-
cesses differ from irreversible processes in that from a reversible process, a real process
emerges again via time reversal - the process is thus reversible. The second law of thermo-
dynamics links the question of reversibility with the change in entropy: It states that, in
any process, the entropy of an isolated system never decreases. If the entropy remains the
same (dS = 0), the process is reversible. If it increases (dS > 0), the process is irreversible.
As a consequence, thermal energy cannot be converted into other types of energy to any
degree.

Examples:
1. The adiabatic movement of the piston (dS = 0) in the system shown in Fig. 5.4
can just as well take place in the opposite direction and is thus reversible. Since
this process neither exchanges heat with the surroundings nor produces heat in the
system, i.e. δQges = 0, in accordance with (5.25) the entropy does not change either.
2. Typical examples of irreversible processes are equalisation processes, in which an
equilibrium is established through the exchange of an extensive property. Fig. 5.6a,
for example, shows pressure equalisation between two containers that are filled with
gas at different pressures and connected through a thin tube. The pressure equalisa-
tion is brought about by a material flow V̇ , while in the case of the thermal contact
between two systems of different temperatures shown in Fig. 5.6b, a temperature
equalisation takes place, namely through the intangible heat exchange between the
two subsystems. Both cases demonstrate that the flow – whether material or non-
material – always takes place from the higher level of pressure or temperature to
the lower level of the quantity in question, and never vice versa. It can be clearly
explained why the entropy of the entire system increases in both cases: In the case
6
δQe is what we have referred to as δQ so far.
7
This contribution is also called irreversible heat.

80
5.7. SELECTED PROCESSES OF IDEAL GASES

p2 V̇ p1 T2 Ṡ T1

p∗2 p∗1 T2∗ T1∗

(a) (b)

Figure 5.6: Analogy between pressure equilibrium and temperature equilibrium: In (a),
two systems with different pressures p2 > p1 reach the equilibrium state p∗2 = p∗1 through
material exchange V̇ , while in (b) two systems with different temperatures T2 > T1 reach
the equilibrium state T2∗ = T1∗ through the exchange of heat or entropy Ṡ.

of the heat transfer shown in Fig. 5.6b, each infinitesimally small portion of heat δQ
that passes from the left to the right reservoir causes a reduction of the entropy in
the left reservoir by δQ/T2 , but at the same time an increase of the same by δQ/T1 ,
so that
δQ δQ δQ
dS = − = (T2 − T1 ) > 0
T1 T2 T1 T2
applies, which is in accordance with the second law of thermodynamics. In the case
of the pressure equalisation shown in Fig. 5.6a, each infinitesimally small volume of
gas dV that passes from the left to the right reservoir causes an expansion which,
according to (5.24), likewise leads to an increase in entropy.

5.6.4 The Nernst heat theorem (third law)

When the temperature approaches absolute zero, the heat capacity approaches zero.
For this reason, the absolute zero of temperature is experimentally unattainable.

5.7 Specific process steps for ideal gases and their graphical
representation
In the following, we will assume an ideal gas as the working medium and rule out both
material exchange with the surroundings and chemical reactions.

5.7.1 The T -S-diagram as a means of representation

Various diagrams can be used to represent thermodynamic processes. The p-V -diagram
shown in Fig. 5.7 is characterised by the fact that the area under the curve is, in accordance
with (5.20), the work done by the system. On the other hand, by reversing the definition
of entropy (5.25) one can see that
δQ = T dS (5.27)

81
 5.7. SELECTED PROCESSES OF IDEAL GASES

represents the process-dependent heat δW , analogously to (5.19), as the product of the

intensive compensation variable T and the differential of an associated extensive property
S. In this respect, it makes sense to represent a process alternatively in a T -S-diagram,

p T

V S
V2 V1 S2 S1
Figure 5.7: Representation of a process in a p-V -diagram (left) and in a T -S-diagram
(right).

see Fig. 5.7, because then the area under the curve represents the heat input into the
system, according to
ˆS2
Q12 = T dS . (5.28)
S1

To transfer an existing representation in a p-V -diagram into a T -S-diagram, we only need

to know the relationships T = T (p, V ) and S = S (p, V ). In the case of an ideal gas, the
equation of state (5.9) and the relationship (5.24) are used for this purpose.

5.7.2 Isotherms
An isothermal process is a change of state in which the temperature remains the same.
This is typically achieved by heat equalisation with a temperature bath. According to the
equation of state (5.9), this means that for ideal gases the product pV remains constant
and thus
p0 V0
p= (5.29)
V

applies. Thus isotherms in the p-V -diagram are hyperbolas, as is shown in Fig. 5.8.

p
T

V S
Figure 5.8: Isotherms in the p-V -diagram (left) and in the T -S-diagram (right).

82
5.7. SELECTED PROCESSES OF IDEAL GASES

5.7.3 Adiabats / isentropes

In the reversible adiabatic change of state, in which heat is neither added nor generated
internally in the system, the entropy is a constant. According to (5.24), for an ideal gas
T V T1 V1
       
CV ln + νR ln = S(V, T ) = S1 = CV ln + νR ln ,
T0 V0 T0 V0
applies. After a few simple transformation steps (division by CV , laws of ln), this equation
simplifies to
T νR V
   
ln + ln = 0.
T1 CV V1
Finally, applying the exponential function leads to the relationship
γ−1
T V1

= (5.30)
T1 V
between the volume and the temperature. The equation has an adiabatic exponent, which
is defined as
Cp νR
γ= =1+ . (5.31)
CV CV
However, we are also looking for the relationship between volume and pressure. This can
be easily obtained if we take into account the equation of state for ideal gases (5.9) and
replace the temperature in (5.30) with pV /νR or T1 = p1 V1 /νR and thus obtain

p
T

V S
Figure 5.9: Adiabats in the p-V -diagram (left) and in the T -S-diagram (right).

γ
p V1

= (5.32)
p1 V
as the relationship between pressure and volume by which the form of the adiabats in the
p-V -diagram is determined, see Fig. 5.9 (left). On the right, the adiabats are represented
in a T -S-diagram in the form of vertical lines.

5.7.4 Isochores
Isochoric changes of state are those at constant volume V = V0 . In a p-V -diagram,
isochores are vertical lines. For ideal gases, the linear relationship
νR
p= T (5.33)
V0
between pressure and temperature follows from the equation of state (5.9). Furthermore,
the entropy (5.24) takes the shortened form
T
 
S = CV ln , (5.34)
T0

83
 5.8. CYCLES

which can be easily solved for the temperature:

T S
 
= exp , (5.35)
T0 CV
so that in the T -S-diagram, isochores take the form of exponential functions.

T
p

V S
Figure 5.10: Isochores in the p-V -diagram (left) and in the T -S-diagram (right).

5.7.5 Isobars
Isobars are defined by the property p = p0 , so that they appear as horizontal lines in the
p-V -diagram. In the T -S-diagram, they take the form of exponential functions due to the
relationship T = T0 exp (S/Cp ).

T
p

V S
Figure 5.11: Isobars in the p-V -diagram (left) and in the T -S-diagram (right).

Other relationships:
νR
V2 − V1 = (T2 − T1 )
p0
T2
 
S2 − S1 = Cp ln
T1
Q12 = Cp (T2 − T1 )

5.8 Cycles
A cycle is a thermodynamic process whose initial and final states are identical. The aim
is to divide such processes into simple sub-processes on a model basis. We consider a
Carnot cycle. As is shown in Fig. 5.12, a Carnot cycle consists of four steps, namely
isothermal compression, followed by adiabatic compression, then isothermal expansion
and finally adiabatic expansion. In the p-V -diagram (Fig. 5.12a), the area enclosed by

84
5.8. CYCLES

p
T

¸
heat Q = T dS
¸ T2
work done W = pdV T1
S
S2 S1
V (b)

(a)

Figure 5.12: Carnot cycle

the process is the work gained. In the T -S-diagram (Fig. 5.12b), in which the process
can be represented particularly easily as a rectangle, the enclosed area corresponds to the
heat lost in the process. In accordance with the first law, this is exactly equal to the
work gained. One can also clearly see in the T -S-diagram Fig. 5.12b that the heat made
available to the process in the form of work (pink area) and the total heat supplied to the
process (pink and grey area) are in the ratio

(T2 − T1 )(S1 − S2 ) T2 − T1
η= = , (5.36)
T2 (S1 − S2 ) T2

to each other, which is called the Carnot efficiency. It can also be seen that the Carnot
cycle ensures the optimal yield of a heat engine, since between a given minimum and
maximum temperature the rectangle describes the maximum possible area. In addition,
this process is reversible in contrast to, for example, the cycle shown in Fig. 5.5, which
has an irreversible process step with the isochore. It can thus be stated that the Carnot
efficiency (5.36) is the best possible efficiency that a heat engine can achieve.
However, the above also means that heat can never be completely converted into work and
that the usable fraction of heat depends on the fact that two heat reservoirs of different
temperatures must be available. Therefore, a perpetual motion machine of the second kind
cannot exist either, which would be a machine that gains work by removing heat from a
single heat reservoir.

Exercises
Exercise 5.1: Ambient Energy
(Solution cf. A.5.1)
We consider a cubic metre of air under normal conditions, i.e. at temperature 20°C and
a pressure of p = 1013hPa. For simplicity, we assume that air is a diatomic ideal gas
consisting of 80% nitrogen and 20% oxygen.
1. What quantity of substances ν are contained in the cubic metre of air?
2. What is the root mean square velocity of the nitrogen molecules and the oxygen
molecules (find the masses of their molecules by research)?

85
 5.8. CYCLES

3. What is the kinetic energy of all the air molecules together?

4. Can this energy be harnessed?

Exercise 5.2: Systematics in thermodynamics

(Solution cf. A.5.2)
Thermodynamics is characterised by a very consistent terminology in which specific quanti-
ties, molar quantities and densities have a clearly defined meaning. Consider the following
examples:
1. What is meant by specific volume? What name and unit is used for this quantity
and how is it related to the (mass) density ϱ?
2. What is the molar volume of an ideal gas at T = 300K and p = 1 · 105 Pa?
3. What is the internal energy density, what is its name and how is it calculated for an
ideal gas as a function of density ϱ, specific heat capacity cV and temperature T ?

Exercise 5.3: Isobaric expansion

(Solution cf. A.5.3)
An ideal gas expanding by heating
drives a piston (mass m1 , cross- V, ν, CV A m2
m1 µ
section A) which pushes a body of
Q
mass m2 over a rough plane
(coefficient of friction µ). Let the initial volume V0 , the final volume V1 and the amount
of substance ν of the gas, as well as its heat capacity CV for fixed volume, also be given.
1. During the process, the body moves slowly at a constant speed. What force does
the piston consequently exert on the body?
2. What is the consequence for the pressure p in the cylinder? To illustrate the process,
draw a p-V diagram.
3. By how much do temperature T and internal energy U increase during the process?
4. What work W01 does the piston do?
5. How much heat Q01 must therefore be supplied in total?
6. Show that from the above results a relation of the form Q01 = Cp (T1 − T0 ) be-
tween temperature increase T1 − T0 and supplied heat Q01 follows and determine the
constant Cp , which is called heat capacity for constant pressure.

Exercise 5.4: Calorimetry

(Solution cf. A.5.4)
In a thermally insulated container there is m1 = 100 g of water with specific heat capacity
c1 = 4187 J/kg·K at a temperature of 21°C. Three 2€ coins (total mass m2 = 25g, specific
heat capacity c2 = 800 J/kg·K) of unknown temperature are added. After some time, one
notices a heating of the water by 2.5K. What was the initial temperature of the three
coins?

Exercise 5.5: Two-chamber system

(Solution cf. A.5.5)

86
5.8. CYCLES

p1 , V1 , T1 p1 , V1 , T2

A container is initially divided by a movable membrane into two chambers of equal size,
each with a volume V1 . The left chamber is filled with an ideal gas of temperature T1 =
300K at a pressure p1 , the right chamber with the same gas at the same pressure, but
with a higher temperature T2 = 1200K (upper picture). The molar heat capacity of the
gas is cV,m = 5R/2 (R=universal gas constant).
1. Which quantities of substances ν1 , ν2 are in the left and right chamber, respectively?
Give the resulting ratio ν2 /ν1 as a numerical value.
2. What are the heat capacities CV 1 , CV 2 of the gas fillings of the two chambers?
3. Gradually, temperature equalisation occurs internally by heat conduction, whereby
the membrane shifts without forces until, in equilibrium, temperature and pressure
are equal in both chambers (picture below).

p3 , V3 , T3 p3 , V4 , T3

What is the equilibrium temperature T3 if one disregards the heat capacities of the
container and the membrane? hint: Using one of the main theorems, consider which
extensive state variable remains constant during the equilibration process.
4. Determine the two volumes V3 and V4 of both chambers and the pressure p3 .
5. Finally, calculate the change in entropy due to the temperature equilibrium and
check that your result is in accordance with the second law.
Note: Use V0 = V1 and T0 = T1 as reference variables.

Exercise 5.6: One half of an Otto-cycle

(Solution cf. A.5.6)
Nitrogen enclosed in a cylinder (volume V1 = 2.5l, temperature T1 = 300K) m2
keeps a piston (cross-section A = 98cm2 , mass m1 = 6kg) and a body attached
to it (mass m2 = 44kg) in equilibrium.
1. What is the pressure p1 in the cylinder? Use g ≈ 9.8m/s2 as the rounded
acceleration due to gravity and p0 = 1 · 105 Pa as the external pressure.
2. What is the amount of substance ν1 of nitrogen in the cylinder? (R ≈
8.31J/mol K).
3. Now a small amount of lead azide (PbN6 ) is electrically ignited at the m1
bottom of the cylinder. Its detonation releases another 0.1mol of nitro-
gen and a quantity of heat ∆Q = 16kJ. What is the temperature T2 A
and pressure p2 immediately after detonation? Assume cV,m = 5R/2 as
the molar heat capacity for nitrogen.
4. Subsequently, the gas expands adiabatically and moves the piston up-
wards until the pressure has fallen back to the initial value p1 (equilib-
rium, inertia is neglected). What is the volume V3 and temperature T3
of the gas at this moment? PbN6
5. Sketch the two process steps in a p-V diagram. 

Exercise 5.7: Adiabatic exponent of ideal gases

87
 5.8. CYCLES

(Solution cf. A.5.7)

The adiabatic exponent γ, which according to (5.32) determines the form of adiabates
in the p-V -diagram, can be determined according to (5.31) from heat capacity CV (at
constant volume) and amount of substance ν. On the other hand, CV for ideal gases again
depends on the number of degrees of freedom. Show by substitution that you end up with
a formula for calculating γ that contains only the number of degrees of freedom of the
molecular motion and nothing else. What values does γ assume for argon, nitrogen and
ozone?

Exercise 5.8: Thermodynamic cycle

(Solution cf. A.5.8)
An ideal gas with temperature T1 , amount of
substance ν and heat capacity CV enclosed p, V, T, ν, CV A
in a cylinder with inner volume V1 is isochor-
Q
ically heated to a temperature T2 .
In a second process step, the gas drives a piston through adiabatic expansion. The process
ends as soon as the pressure has dropped to its value at the beginning of the process.
1. For illustration, plot the whole process on a p-V diagram.
2. What are the initial pressure p1 and pressure p2 after isochoric heating?
3. How much heat Q12 is supplied?
4. What is the volume V3 and temperature T3 of the gas at the end of the process, i.e.
after adiabatic expansion?
5. How much work W23 is done by the piston?

88
Chapter 6

A glimpse of fluid dynamics . . . or

“soft matter”, as it is called now

There is hardly a subdiscipline of physics that is more afflicted with prejudice than fluid
dynamics, which was also shamefully neglected by the physics community in the past,
especially in Germany. Yet this field is full of surprises and unsolved problems. With
the emergence of biophysical problems, the community realised that physics could not do
without this subdiscipline after all. So the physicists devoted themselves to it again, under
the new label of ”soft matter physics” in order to clearly distinguish themselves from the
engineering sciences — needless to say what I personally, who am as much an engineer as
a physicist, think of that . . .
The following chapter can be regarded as a rudimentary-minimalist introduction to the
topic that can hardly do justice to the thematic diversity. Furthermore, it is based on
a simplified approach from the perspective of thermodynamics. The classical and more
precise approach via mechanics would certainly go beyond the scope of this lecture and is
postponed to a later semester!

6.1 Historical overview

Fluid dynamics is one of the oldest subdisciplines of physics. The works of Archimedes and
Sextus Iulius Frontinus in antiquity can be considered relevant contributions due to their
technological importance for shipbuilding and water supply. In the Renaissance, Leonardo
da Vinci dealt with flow processes. Later, relevant contributions followed from Galileo
Galilei, Evangelista Torricelli, Blaise Pascal, Edme Mariotte, Isaac Newton and Daniel
Bernoulli. But with regard to the close intertwining of physics and mathematics, the works
of Leonhard Euler and Jean-Baptiste le Rond d’Alembert can be seen as groundbreaking
for the subsequent development of fluid dynamics, because the resulting potential theory,
beyond the theory of frictionless and vortex-free flows, also allowed the calculation of
electromagnetic fields and is a prime example of how physics and mathematics inspire
each other. Since that time, fluid dynamics and physics are inextricably linked.
Despite the success of potential theory, physicists soon came up against its limitations,
which became clear from certain paradoxes such as d’Alembert’s paradox. The conse-
quent extension of Euler’s theory with regard to viscosity by Claude Louis Marie Henri
Navier and George Gabriel Stokes, leading to the Navier Stokes equations, presented

89
 6.2. THE “IDEAL” FLUID AND ITS FLOW

mathematics with new challenges that were to engage generations of scientists to this
day. The pioneering work by William Froude on the buoyancy of ships, Ernst Mach on
supersonic aerodynamics, Lord Rayleigh on hydrodynamic instability, Vincent Strouhal
on vibration stimulation through vortex shedding as well as Hermann von Helmholtz on
vortex dynamics and scientific meteorology, was followed by the groundbreaking works
of Osborne Reynolds and Ludwig Prandtl, and later Andrei Nikolajewitsch Kolmogorow,
which formed the basis of boundary layer theory and turbulence theory and, in particular,
led to an indispensable contribution to a deeper understanding of viscous flows based on
advanced mathematical methods.
Starting from the mid-20th century, flow measurement technology and computational fluid
dynamics progressed to the point where they could be used for finding solutions to practical
problems, but there are still unsolved problems and open questions to this day.

6.2 The “ideal” fluid and its flow

A fluid is a form of matter whose components do not have a fixed arrangement in relation
to each other. This is the case in liquids, gases and plasmas. From a mechanics perspective,
the shear stresses, which are created during shearing of the material as its reaction to the
distortion, plays an important role here. In this sense, one speaks of an ideal or inviscid
fluid when there is no shear stress created at all during shearing.

6.2.1 Bernoulli’s principle

Similar to a solid, a fluid can be regarded as an accumulation of mass points, with the
difference that there are no rigid bonds between them. As is shown in Fig. 6.1, we now
consider a small material volume V and its movement within a time interval [t1 , t2 ]. We

Figure 6.1: a small material volume in a steady flow at different points in time t1 and t2 .

can imagine this as marking a small part of the fluid, for example with dye or suspended
particles. The set of all points in space that the volume passes over during its movement
is called a flow conduit, which can also be seen as a bundle of the trajectories of all the
volume’s material points.
We will now use the findings from thermodynamics to identify the laws that determine
the movement. In particular, we will be using the first law (5.21), which has the form

90
6.2. THE “IDEAL” FLUID AND ITS FLOW

dU = δQ + δW , i.e. the change in internal energy results from the total balance of heat
supplied and work done. According to (5.27), the heat supplied can be expressed by
δQ = T dS, whereas the work δW is composed of the following contributions:
Compression −dV : This contribution was already discussed in the context of a gas
trapped in a cylinder and is to be evaluated as −pdV in line with (5.19), with p
being the pressure in the fluid.
Lift dz: If the gas is lifted in the gravitational field, the work mgdz is done (or is released
if dz is negative). Here, m is the mass of the material volume V .
Momentum change d⃗ p: In general, the velocity and the direction of movement changes,
and so does the momentum of the material volume. According to (3.4), a power
P = Ėkin must be used for this process, which according to (3.1) can also be written
as P = m⃗v · ⃗v˙ = ⃗v · d⃗p/dt. Looking at it from the thermodynamics perspective
and taking into account the well-known relationship dW/dt = P between work and
power, we obtain ⃗v · d⃗
p for the work that is required to change the momentum of the
material volume by d⃗ p.
In total, we obtain
δW = −pdV + mgdz + ⃗v · d⃗
p (6.1)
as the work input. Using the product rule, we also carry out the following transformation:

dU + pdV = d [U + pV ] −V dp (6.2)
| {z }
dH

In thermodynamics, the quantity H = U + pV is known as enthalpy and is an alternative

measure of a system’s internal energy. For ideal gases, the enthalpy is H = Cp T and
thus solely depends on the temperature. This is not exactly true for other fluids, but still
applies approximately.
If we summarise everything discussed so far, the 1st law takes the form

dH − T dS = V dp + mgdz + ⃗v · d⃗
p (6.3)

for material volumes in a flow. However, we still have to eliminate the dependence on the
individually chosen volume. To do this, we divide the equation by the volume’s mass m,
which leads to the “field form”
1
dh − T ds = dp + gdz + ⃗v · d⃗v (6.4)
ϱ
with specific instead of extensive quantities, namely the specific volume V /m = 1/ϱ, which
is nothing other than the inverse of the mass density ϱ, the specific enthalpy H/m = h,
the specific entropy S/m = s and the specific momentum p⃗/m = ⃗v , which is nothing other
than the flow velocity.
We can only benefit from the differential relationship (6.4) if it is integrable. The two
last terms on the right-hand side of the equation are unproblematic in this respect, but
the compression term dp/ϱ could become an issue: If we again take an ideal gas as an
example, thermodynamics provides a general relationship between pressure and density in
the form of the ideal gas law (5.9). After dividing this equation by m, it can be written as
p R
= T (6.5)
ϱ M

91
 6.2. THE “IDEAL” FLUID AND ITS FLOW

with R being the universal gas constant, M the molar mass of the respective gas and T
the temperature. For now, this brings in to play another unknown field, namely the tem-
perature, but under certain model assumptions it is possible to eliminate this dependence.
For example, if we take the isothermal process as a starting point, i.e. T = const, we will
obtain a proportionality ϱ ∼ p between pressure and density. In the case of a adiabatic
process, however, ϱγ ∼ p applies. Here, γ is the adiabatic exponent, which for air can be
quantified very well as 1.4. In general, we assume a simple functional relationship of the
form
ϱ = ϱ(p) . (6.6)
In such a case, we speak of Barotropie. In this case, the integral
ˆ
1
P (p) := dp , (6.7)
ϱ
exists and is also referred to as a pressure function. In the special case of an incompressible
flow (ϱ = const), the pressure function is simply P = p/ϱ.
Another problematic term in (6.4) is T ds. In order to eliminate it, we restrict ourselves
to purely reversible and thus loss-free processes, so that ds = 0 applies, in line with the
second law of thermodynamics. Thus (6.4) becomes integrable and, by integration of the
flow conduit, we obtain
⃗v22 ⃗v 2
+ P (p2 ) + gz2 − h2 = 1 + P (p1 ) + gz1 − h1 (6.8)
2 2
as the, from a thermodynamics point of view, most general form of Bernoulli’s equation.
Better known is the special case of an incompressible and at the same time isenthalpic
flow, where due to the assumed reversibility the latter is already given at a constant
temperature, which leads to the classical form of Bernoulli’s equation:
⃗v22 p2 ⃗v 2 p1
+ + gz2 = 1 + + gz1 . (6.9)
2 ϱ 2 ϱ
The advantage of this equation, obviously, is its simplicity, because it is a simple algebraic
relationship between pressure and velocity in a flow. However, its validity is subject to
certain conditions, which you have to make sure are observed in any case. In addition, the
applicability of this form of Bernoulli’s equation is limited to steady, i.e. time-independent,
flows. This is due to the underlying classical thermodynamics, which is in principle limited
to equilibrium states.
In summary, under the conditions of a
• steady
• incompressible
• loss-free
• isothermal
flow, we can say that the classical Bernoulli equation (6.9) applies, albeit only along
streamlines.
Historically, an alternative derivation of Bernoulli’s equation in terms of a first integral
of the flow’s momentum balance leads to a more general form. However, this would go
beyond the scope of this lecture, as further basic knowledge would have to be acquired
beforehand.

92
6.2. THE “IDEAL” FLUID AND ITS FLOW

Example: Torricelli’s law

We consider a liquid flowing from an orifice of a large, open container, as shown in 6.2. In

(1)

(2)
v2 =?

Figure 6.2: Outflow from a large open container due to gravity.

order to apply Bernoulli’s equation to this case, a streamline would first have to be located,
but in this case it is sufficient to assume that such a streamline exists, i.e. we assume
that there are streamlines connecting the surface (”point” 1) with the outlet (“point” 2).
Furthermore, additional assumptions are required to justify the use of Bernoulli’s equation:
In order to guarantee that the flow is steady, i.e. time-independent, the liquid level must
not drop noticeably, which means that v1 can be assumed to be ≈ 0. This is achieved
when the cross-sectional area of the container is significantly larger than that of the outlet.
Furthermore, since the density of the air is much smaller than that of the fluid, we can
ignore differences in air pressure between the surface and the outlet and assume that
p1 = p2 = p0 , where p0 is the air pressure. If we plug all this into Bernoulli’s equation
(6.9), i.e.
⃗v22 p0 02 p0
+ + gz2 = + + gz1 ,
2 ϱ 2 ϱ
and solve for v2 , we obtain
q
v2 = 2g(z1 − z2 ) = 2gh , (6.10)
p

which is known as the Torricelli’s law.

6.2.2 Continuity equation

In Newtonian mechanics of mass points and rigid bodies, the conservation of mass is a
given property of the objects under consideration and therefore did not have to be verified.
In fluid dynamics, however, this no longer applies automatically, but must be guaranteed
by the fulfilment of an equation, the continuity equation.
We consider a spatially fixed test volume V within a fluid flow that is represented by
the velocity field ⃗v (⃗x, t), as shown in Fig. 6.3. Then the mass of the fluid inside the
volume is changed by the flow across the boundary ∂V . For an infinitesimal element dS
of the surface area and its corresponding normal vector ⃗n, the volume of fluid passing
the boundary per time from the outside to the inside is given by −⃗v · ⃗n dS, and the mass
passing the boundary per time is given by −ϱ⃗v · ⃗n dS. Integration over the total area thus
yields the integral mass balance

dm {
ṁ = =− ϱ⃗v · ⃗n dS . (6.11)
dt
∂V

93
 6.2. THE “IDEAL” FLUID AND ITS FLOW

Figure 6.3: Test volume V in a flow.

Graphically speaking, this means nothing other than that the change in mass is given by
the total balance of the fluid flow across the boundary of the control volume. In the steady
case ṁ = 0 and for incompressible flows ϱ = const, the continuity equation simplifies to
{
⃗v · ⃗n dS = 0 . (6.12)
∂V

Constriction of a tube (Venturi nozzle)

We consider a steady and incompressible flow through a constricted section of a tube, see
Fig. 6.4. We choose the tube section between the cross sections A1 and A2 as a control

Figure 6.4: The Venturi nozzle as an example of the combined application of the continuity
equation and Bernoulli’s equation.

volume for the continuity equation (6.12). For the sake of simplicity, we also assume that
both the pressures p1 , P2 and the velocities v1 , v2 are constant over the respective cross
sections. In this way, the continuity equation (6.12) takes the form
−v1 A1 + v2 A2 = 0 , (6.13)
where the different signs result from the direction of the respective normal vectors of the
cross sections (these always point outwards) compared to the direction of flow. Simply
put, we can say that an inflow is always evaluated negatively and an outflow positively.
Solving (6.13) for v2 at a given velocity v1 of the inflow yields the velocity in the constricted
section as
A1
v2 = v1 . (6.14)
A2
Thus the flow velocity is higher in the constriction.
In a second step, we can now also find the pressure by means of Bernoulli’s equation (6.9).
If the direction of flow is assumed to be horizontal, z1 = z2 applies and we obtain
⃗v22 p2 ⃗v 2 p1
+ = 1 + ,
2 ϱ 2 ϱ

94
6.3. NAVIER STOKES EQUATION (2D) FOR VISCOUS FLOWS

which we can solve for the quantity we are looking for after we have substituted v2 ac-
cording to (6.14): " #
v12 A21
p2 = p1 − ϱ −1 . (6.15)
2 A22
Since the factor in the square brackets is positive because of A2 < A1 , it follows, surpris-
ingly, that p2 < p1 , i.e. a negative pressure is created! This counterintuitive phenomenon
is used, for example, in the vacuum ejector.

6.3 Navier Stokes equation (2D) for viscous flows

When limited to two dimensions, i.e. ⃗x = x⃗ex + y⃗ey as well as ⃗v = u⃗ex + v⃗ey , the Navier
Stokes equation broken down into components [10] is:
" #
∂u ∂u ∂u ∂p ∂2u ∂2u
 
ϱ +u +v =− +η + 2 + ϱgx (6.16)
∂t ∂x ∂y ∂x ∂x2 ∂y
" #
∂v ∂v ∂v ∂p ∂2v ∂2v
 
ϱ +u +v =− +η + + ϱgy (6.17)
∂t ∂x ∂y ∂y ∂x2 ∂y 2

These equations are supplemented by the local form of the continuity equation, which is
in this case
∂u ∂v
+ = 0. (6.18)
∂x ∂y
For now, we will refrain from deriving these equations, as well as from discussing them
further. The theory of viscous flows will be a topic in physics in the 6th semester.

6.4 Fluid-structure interaction

A fluid is usually in contact with solid bodies, e.g. container walls, pipe walls or the bodies
around which the fluid flows. Forces are also acting via such contacts.

6.4.1 Interfacial stresses and resultant forces

⃗v
σ τ
τ
⃗v σ
⃗t

Figure 6.5: Stresses along a body contour that is in contact with a fluid, after the object
has been isolated from its surroundings.

To determine the flow forces, we must first isolate the system from its surroundings along
the contact surface, as is shown in Fig. 6.5. As a result, normal and shear stresses appear,

95
 6.4. FLUID-STRUCTURE INTERACTION

which can also be combined to a vectorial quantity, the stress vector ⃗t. In order to obtain
a force from the stress vector, integration over the entire contact surface Af-s is required:
x
F⃗f-s = − ⃗t dS , (6.19)
Af-s

where the negative sign applies if ⃗n is the outer normal vector as seen from the fluid.

6.4.2 Determining flow forces indirectly from the momentum balance

Provided that volume forces on the fluid can be ignored and temporal changes are slow or
non-existent, we can calculate the force in the integral momentum balance [10]. If friction
is neglected, the corresponding equation has the form
x
F⃗f-s = − [ϱ⃗v (⃗v · ⃗n) + p⃗n] dS . (6.20)
Af-f

The main feature of this indirect calculation method is the integration over the areas Af-f ,
which are not the direct contact surfaces between the object and the flow, but just the
inflow and outflow areas. This makes it possible to limit the consideration to regions where
the velocity field is known, without the need to calculate the complete flow.

6.4.3 Example: The splitting of a free jet by a wedge

′ y′
AK x * """
"
"3
A" " T 3"
"
" T"
3
" "
6y
"
" AK δ ""
"
v0 AU"
"
- "
" α
- 6 "
x-
- h0 "
b
- b
- ? b
b
b
b
b b
b b
b b
b b
b b
s
b b
b b
s
bbs
Figure 6.6: The splitting of a free jet by a wedge.

A plane jet of water (of constant density ϱ) symmetrically strikes a wedge with the tip
angle 2α, see Fig. 6.6. Far ahead of the wedge, we have the constant velocity v0 , the jet
thickness h0 and the jet width b. Due to friction, a velocity profile is formed at the end of
the wedge, which is given by
 ′
πy
v(y ′ ) = v0 sin (6.21)
2δ
with an as yet unknown film thickness δ. We want to find the film thickness as well as the
force acting on the wedge, assuming a constant ambient pressure p0 .

96
6.4. FLUID-STRUCTURE INTERACTION

1. To find the film thickness δ, we use the continuity equation (6.12), which in this case
has the form
ˆδ
−v0 h0 b + 2 v(y ′ )bdy ′ = 0 .
0
As before, the negative sign of the first term results from the antiparallelism of the
normal vector and the flow direction. After plugging in (6.21) and cancelling v0 and
b, we obtain
ˆδ  ′ δ
4δ 4δ
 ′
πy πy
h0 = 2 sin dy ′ = − cos =
2δ π 2δ 0 π
0
and finally
π
δ= h0 . (6.22)
4

2. The force acting on the wedge is calculated via (6.20)1 The pressure acting on all
sides is disregarded from the outset. Thus ⃗v (⃗v · ⃗n) is the relevant contribution. Note
that this is a vector! At the inward flow, ⃗v (⃗v · ⃗n) = −v02⃗ex applies, at the upper and
lower outward flow, however,

⃗v (⃗v · ⃗n) = v(y ′ )2 [cos α⃗ex ± sin α⃗ey ] ,

is valid. This clearly shows that the contributions annihilate in the vertical direction
(y-direction), so that (6.20) ultimately leads to

ˆδ ˆδ
 
πy ′ 2
 
F⃗f-s = ϱv02 h0 b⃗ex − 2ϱ v(y ′ )2 bdy ′ cos α⃗ex = ϱv02 b h0 − 2 cos α sin dy ′  ⃗ex
 
2δ
0 0

Note that sin2 x = [1 − cos(2x)] /2 and

ˆδ 2
πy ′

cos dy ′ = 0 ,
δ
0

so that, taking into account the film thickness (6.22), the force results as:

π
 
F⃗f-s = ϱv02 b [h0 − cos αδ] ⃗ex = ϱv02 h0 b 1 − cos α ⃗ex . (6.23)
4
The resulting force is positive in any case.

6.4.4 Buoyancy forces

We consider a body completely immersed in a fluid at rest with mass density ϱ. The
body occupies the volume V and has the mass m, as shown in Fig. 6.7(a). Two forces are
acting on it, namely the weight force m⃗g as the active force and the buoyancy force F⃗A as
the reaction force. Since both forces are different from each other, we cannot necessarily
expect that there is a static equilibrium, i.e. the body moves upwards or downwards in an
1
Because friction is present, one would actually have to resort to a more general form of momentum
balance here. However, if the viscosity is sufficiently low, we can accept the slight error from neglecting
the velocity gradient.

97
 6.4. FLUID-STRUCTURE INTERACTION

(a) (b)

F⃗A

ϱV ⃗g
(c) (d)

Figure 6.7: Deriving Archimedes’ principle from Newton’s third law

accelerated manner, depending on which force dominates. The situation will be different
if, as shown in Fig. 6.7(b), the body is hypothetically hollowed out completely and the
resulting cavity is filled with the same fluid as the one that surrounds the body: Then
nothing changes in the buoyancy force F⃗A acting on the body, because it results solely
from the surrounding fluid, while the weight force is now ϱV ⃗g . The crucial point, however,
is that the system is put into a state of static equilibrium by the fictitious manipulation
described above. F⃗A + ϱV ⃗g = ⃗0, so that

F⃗A = −ϱV ⃗g (6.24)

necessarily follows. This has become known as Archimedes’ principle: The buoyancy force
opposing the weight force corresponds in its magnitude to the weight force of the fluid
displaced by the body.
The total force acting on the original body in Fig. 6.7(a) is then:

F⃗ = m⃗g + F⃗A = (m − ϱV ) ⃗g . (6.25)

The above considerations inevitably lead to consequences for the immersion depth of
floating bodies: Any body floating on the surface of a fluid, see Fig. 6.8, is obviously in
static equilibrium, i.e. the buoyancy force ϱV ′ g acting on it must be identical to its weight
force mg, so that the volume of the fluid displaced by the body is
m
V′ = . (6.26)
ϱ
Here, the material the body is made of can have a higher density than the fluid, as long
as the body has a favourable shape that occupies at least the above volume V ′ .

98
6.4. FLUID-STRUCTURE INTERACTION

m
m

V′
V′

Figure 6.8: Displacement of a fluid by floating bodies.

Exercises
Exercise 6.1: Stagnation point
(Solution cf. A.6.1)
Consider an incompressible flow around a body with flow
velocity v0 , density ϱ0 and pressure p0 far away from the
body. Then there is always a streamline flowing directly
into a point of the body contour, the so-called stagnation
point. What is the pressure there if frictional losses are
neglected?

Exercise 6.2: Waterfall

(Solution cf. A.6.2)
Liquid (filling height h, density ϱ) runs 6
out of a circular opening (diameter D) at
h
the bottom of a container without fric-
tion. Influenced by gravity, the jet ta-
pers with increasing distance of fall x. ?
The liquid level in the container does not ?
???
???
v0
change noticeably during this process.
a) Determine the velocity v0 at the
outlet i.e. at x = 0. For simplic-
ity, assume constancy and unidi- x
? - d = d(x)
rectionality (= uniform vertical di- ?
???
?
rection) of the velocity over the re- v(x)
spective cross-section of the beam.
b) Specify the velocity v of the beam
as a function of x.
c) Determine the beam diameter d as a function of the distance of fall x.

Exercise 6.3: Thermodynamics meets fluid mechanics

(Solution cf. A.6.3)
In a cylinder (cross-sectional area A0 ) there is a liquid (density ϱ0 ) and above it a gas
column (height h0 ), bounded by a piston (Figure A). Pressure p0 and temperature T0

99
 6.4. FLUID-STRUCTURE INTERACTION

correspond to those of the environment. An opening at the bottom of the cylinder is

initially closed by a valve.
(A) (B) (C) (D)
A1
p0 ,T0
F
A1 F F

p0 ,T0 h0

p1 ,T1 h1 p1 ,T0 p1 ,T0 p0 ,T0

h2
v1 A2
ϱ0 ϱ0 ϱ0 ϱ0 v2

1. In the first step, the gas is adiabatically compressed (adiabatic exponent γ), with a
given force F acting on the piston in the final state (B). What is then the gas pressure
p1 (let the weight of the piston be negligible)? Then calculate the gas temperature
T1 and the new height h1 of the gas column.
2. Then let the gas cool down to the ambient temperature T0 while maintaining the
force F (Figure C). What is the height h2 of the gas column then?
3. Finally, open the valve (Figure D) so that a jet of liquid (cross-sectional area A2 )
escapes through the overpressure while its level in the cylinder decreases with velocity
v1 . Determine v1 and jet velocity v2 assuming stationary flow and neglecting gravity.

100
Part III

Systems

101
Chapter 7

Variational methods and the

Lagrangian formalism

The following chapter is to be regarded as a bonus chapter in the sense that it deals with
a new mathematical tool and also provides a new perspective. Although every physical
problem can be solved just as well with the existing mathematical tools and within the
framework of a conventional approach, the Lagrangian formalism has, above all, inspired
modern physics and contributed significantly to its development. For the engineer, varia-
tional methods are primarily a useful tool and allow an efficient and systematic treatment
of many problems in mechanics.

7.1 Simple optimisation problems

Although they are by no means mathematically more sophisticated than differential equa-
tions, variational problems demand an unfamiliar way of thinking from most readers. For
this reason, this chapter aims to provide a motivational approach to the calculus of vari-
ations using simple examples. Fermat’s principle, which describes the optical path in an
optical system, will be useful for this purpose. The subsequent transition to Hamilton’s
principle using mathematical analogy will then appear all the more natural.

7.1.1 The lifeguard

A lifeguard (R) standing on the shore is trying to reach a non-swimmer as quickly as
possible (N) who is about to drown (see Picture 7.1). However, it must be noted that on
land, he can move at a speed that is higher by the factor of n than in water. Without the
speed difference, the shortest path between points R and N, i.e. the straight line connecting
the two points, would also be the fastest. But since the speed is reduced in water, it may
be more advantageous to minimise the path through the water. For this purpose, the
lifeguard would have to run to the point LN at which the perpendicular drawn from the
non-swimmer meets the coastline, and then swim the shortest possible distance dN . Both
possible paths, i.e. the shortest total distance and the shortest path through the water,
are indicated in green in the picture 7.1.

The actual fastest path, however, which is indicated in red in the picture 7.1, lies between
the two paths mentioned earlier, as will be shown below: If the lifeguard first heads for

103
 7.1. SIMPLE OPTIMISATION PROBLEMS

LN s -N
p *d

6

  

 x

dE β

α q


 -

 dN

v = v2


l










v = v1


dR

?

Figure 7.1: Initial situation for the lifeguard and three possible paths to save the drowning
person.

a point on the coastline that is at a distance

q x from the perpendicular LN of the non-
swimmer, he will have to run the distance d2E + (l − x)2 , where dE denotes his initial
distance from the coastline and l the distance of his own perpendicular LR from the
perpendicular
q of the non-swimmer, LN , . He will have to cover the remaining distance,
i.e. dN + x , swimming at a speed v0 , so that the time until he reaches the drowning
2 2

person is
1 1q 2
q
t = t (x) = d2E + (l − x)2 + dN + x2 .
nv0 v0
Finding the position x at which t becomes minimal corresponds mathematically to finding
an extremum, i.e.
 
dt 1  1 l−x x
0= = − q +q ,
dx v0 n d2 + (l − x)2 d2 + x 2
E N

which, using the two auxiliary angles α and β (see Picture 7.1), can be written more
shortly as
1 1
 
0= − sin α + sin β
v0 n
or even better as
sin α
= n. (7.1)
sin β

7.1.2 The fire engine

A team of firefighters are to extinguish a fire. Before that, however, the empty tank
of the fire engine has to be filled up with water, which can be done at a nearby river
(see Picture 7.2). It is assumed that the fire engine can move through any part of the flat
terrain at a given speed (v0 ). Which point on the riverbank should the firefighters head
for, so that as little time as possible will be lost? This problem can be solved even more
easily than the previous one, if we imagine the position of the fire reflected over the river
course. In this case, reaching the reflected position requires the same amount of time as
reaching the actual fire. However, since the shortest path between two points is a straight

104
7.2. FERMAT’S PRINCIPLE

river fire
g -g
kQ
Q ]
J (b) 
3
7

Q J 
Q  
Q J  
Q J  
Q
Q J 
Q 
 J 
Qβ
Q
J 
}
ZZα 
J
(c)
Z
Z
 J
Z J
 Z J
 Z J
 Z J
Z

 Je
(a)
Z
fire engine

Figure 7.2: The situation for the team of the fire brigade and possible routes for the fire
engine.

line, the point we are looking for is the intersection of the river course with the connecting
line between the initial position of the fire engine and the reflected position of the fire.
This consideration implies, in particular, the equality

α=β

of the two angles.

7.2 Fermat’s principle

The two geometric laws that were derived from minimum-time problems, as was shown in
the previous section, are identical with the law of refraction for light rays passing through
an interface as well as the law of reflection on a surface, which are both known from
geometrical optics. This analogy was noticed as early as the 18th century by the French
mathematician Pierre de Fermat, who thereupon formulated his well-known principle on
the optical path of light in optical systems:

Fermat’s principle: With a given starting and end point, the light always follows the
optical path that is characterised by a minimisation of the light’s travel time compared to
all other possible optical paths.

Obviously, the lifeguard from 7.1.1 behaves in the same way as a ray of light entering from
an optically thinner medium into an optically denser medium. In this way, Snell’s law of
refraction
sin α
=n (7.2)
sin β
can be derived directly from Fermat’s principle. Likewise, the condition α = β for the
firefighters in 7.1.2 is nothing other than the law of reflection known from optics. In fact,
the entire field of geometrical optics can be attributed to Fermat’s principle.

7.2.1 Transition to the continuum

A ray of light from outer space enters the Earth’s atmosphere at an angle at point P1 .
Until it reaches the ground at point P2 , it passes through air layers of various densities,

105
 7.2. FERMAT’S PRINCIPLE

where the refractive index n is a function of the height. Therefore, the optical path is
not a polygonal chain, as in the previous examples, but has a curvature, as shown in
Fig. 7.3a. Here, we speak of atmospheric refraction or, in case of a light beam from one
point of the Earth’s surface to another one of a Fata Morgana, cf. Exercise ??. It is not

c c c
P2 c P2
n(x, y) c
c q ∆yk
∆xk
c
nk
c
P1 (a) P1 (b)
c c
x0 x1 · · · xk xk+1 · · · xN xN +1
Figure 7.3: A ray of light passing through an inhomogeneous medium (a) and the discreti-
sation of the problem (b).

easy to calculate the optical path for any given shape of a curve. We simplify the problem
by replacing the continuous system with a discrete model, see Fig. 7.3b, in which the
atmosphere is thought to consist of N + 1 layers, each with a constant refractive index
nk = n(xk ) and with an interface at the positions x1 , · · · , xN . Then the optical path will
be a polygonal chain again, i.e. a continuous sequence of straight lines. This polygonal
chain is completely determined by the y-values yk at the N positions xk of the interfaces.
Thus the optical path can be calculated as
s
N N
∆yk
q  2
ct = ∆x2k + ∆yk2 = nk 1 + ∆xk (7.3)
X X
nk
k=0 k=0
∆xk

with the definitions ∆xk = xk+1 − xk and ∆yk = yk+1 − yk . With the limit N → ∞
we then get to the continuum. In the process, ∆yk /∆xk becomes dy/dx = y ′ (x) and the
sum becomes an integral of x. With a given starting point ⃗x1 = (0, h) and end point
⃗x2 = (l, 0), the optical path of the light ray passing through the Earth’s atmosphere can
then be written as
ˆl q
ct = n (x, y(x)) 1 + y ′ (x)2 dx (7.4)
0

where the optical path is determined by the function y = y(x) as ⃗r(x) = (x, y (x)).
Fermat’s principle requires that y(x) must be chosen in such a way that ct becomes
minimal for specified initial and final values y(0) and y(l). With this, the problem turns
out to be mathematically more sophisticated than the examples discussed earlier: here, ct
is no longer a simple function, but a functional, i.e. a mapping that assigns a number to a
function. This is usually denoted by square brackets, i.e. ct = ct[y(x)] or simply ct = ct[y].
Using the definition of the function
q
L x, y, y ′ := n (x, y) 1 + y ′2 , (7.5)

106
7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

we can transform the above functional into the more general and clear form
ˆx2
ct [y] = L y (x) , y ′ (x) dx . (7.6)

x1

Many variational problems can be written in this or, at least, a similar form . The
functional character is indicated by writing t[y] in square brackets, i.e. to indicate the fact
that t depends on the function y as a whole, and not on the current value y(x).

7.2.2 The basic concept of the finite element method

The example of atmospheric light transmission discussed in the previous section is partic-
ularly well suited for understanding the concept of finite elements: The integral form of
Fermat’s principle (7.4) resulted from the process of finding the limit of the discrete form
(7.3). The finite element method simply takes the opposite approach: Starting from a
variational formulation of the problem in question, a continuous function is approximated
by a piecewise linear (or, optionally, quadratic or cubic) function, which leads to an alge-
braic sum that has a form similar to (7.3). Its variation with respect to the finitely many
parameters of the function yields an algebraic system of equations.

7.3 The Lagrangian formalism in classical mechanics

The question is whether the laws of mechanics, too, can be derived from a variational
principle. At least for systems with exclusively conservative forces, this was achieved
using Hamilton’s principle.

7.3.1 Hamilton’s principle

Consider a conservative system with N degrees of freedom, whose configuration is given
by so-called generalised coordinates qi = qi (t), such as lengths or angles. The Cartesian
coordinates considered so far are special cases of generalised coordinates, as are polar,
cylindrical and spherical coordinates. We also need to know the total kinetic energy

T = T (qi , q̇i )

of the system as well as the total potential energy

U = U (qi ) .

This makes it possible to define a function

L (qi , q̇i ) := T (qi , q̇i ) − U (qi ) , (7.7)

the so-called Lagrangian. Finally, we assume that the initial configuration qi (t1 ) and the
final configuration qi (t2 ) of the system under consideration are predefined at the points
in time t1 and t2 . The real process q (t) that takes place within the specified time interval
is then characterised by Hamilton’s principle

ˆt2
L (qi , q̇i ) dt = extremum
t1

107
 7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

q2
6
qi (t2 )
b
q qi (t)
q


b
qi (t1 ) δqi (t)
q1
-
Figure 7.4: The real process qi (t) (red) and the virtual process qi (t)+δqi (t) with predefined
initial and final states.

compared to other, fictitious, or as we usually say: virtual processes. Graphically speaking,

one would therefore have to – substitute all possible processes with predefined initial and
final states into the variational problem above and look for an extremum in order to
evaluate which one is the real-world process. But since there are an infinite number of
possible process sequences, this is of course not feasible. Instead, as will be shown below,
we will derive conditions necessary for Hamilton’s principle to be satisfied, which have the
form of familiar differential equations.

7.3.2 Euler-Lagrange equations

Usually, extrema are found by finding the zeros of the first derivative. This is valid at
least for simple functions, i.e. for maps / mappings that assign a number to another
number. Hamilton’s principle, however, includes a mapping that assigns a number to
N functions qi (t). Such a mapping is called a functional and its derivative is called a
functional derivative. Since functional analysis and the calculus of variations are not part
of standard engineering mathematics, a detailed derivation is required and will be given
below.
We assume that the real-world process characterised by Hamilton’s principle is given by
qi (t). We now add a given variation δqi (t), which leads to a virtual process qi (t) + δqi (t)
that is not characterised by Hamilton’s principle, as is shown in Fig. 7.4 using the case of
two degrees of freedom N = 2 as an example. According to the assumption, the initial
and final states are not to be varied, i.e.

δqi (t1 ) = 0 , (7.8)

δqi (t2 ) = 0 . (7.9)

Furthermore, we restrict ourselves to arbitrarily small variations and expand the La-
grangian for the virtual process in a Taylor series, i.e.

N 
∂L ∂L

L (qi + δqi , q̇i + δ q̇i ) = L (qi , q̇i ) + δqi + δ q̇i + · · · (7.10)
X

i=1
∂qi ∂ q̇i
N 
∂L d ∂L d ∂L
    
= L (qi , q̇i ) + δqi + δqi + · · · ,
X
δqi −
i=1
∂qi dt ∂ q̇i dt ∂ q̇i

where, in addition, the chain rule of differentiation was used in the last step. The neglected
terms are at least quadratic in the variations δqi and thus arbitrarily small compared to
the leading terms. If we now calculate the integral of (7.10) over the entire time interval

108
7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

between t1 and t2 , we can drop

ˆt2
d ∂L ∂L t2
 
δqi dt = δqi =0
dt ∂ q̇i ∂ q̇i t1
t1

because the variation is zero at the starting point and at the end point according to (7.8,
7.9). What remains, on the other hand, is an equation of the form

Ivirt = Ireal + IVar + Iremainder , (7.11)

where the definitions

ˆt2
Ivirt := L (qi + δqi , q̇i + δ q̇i ) dt
t1
ˆt2
Ireal := L (qi , q̇i ) dt
t1
ˆt2 X
N 
∂L d ∂L
 
IVar := − δqi dt
i=1
∂qi dt ∂ q̇i
t1

are used for better clarity, while all residual terms not taken into account are summed up
under Iremainder . Now the integral representing the real-world process Ireal is extremal, i.e.
minimal or maximal.
1. If there is a minimum, Ivirt > Ireal must apply, i.e. Ivirt − Ireal = IVar + Iremainder > 0
and thus Iremainder > −IVar . From this, in turn, it follows that IVar ≥ 0, because if we
assumed IVar , ≤ 0 then Iremainder > |IVar | would necessarily follow from Iremainder >
−IVar . However, since Iremainder depends quadratically on the variation, but IVar
depends on it only linearly, reducing the variation by the factor M will lead to
Iremainder being reduced by the factor M 2 , whereas IVar will just be reduced by the
factor M , so that the inequality above becomes Iremainder > M |IVar |. Since we
can choose M to be arbitrarily large, IVar = 0 must necessarily apply for the above
inequality to remain true. The case of IVar > 0, in turn, can be excluded by reversing
the sign of the variation δqi , i.e. δqi → −δqi . Because then the sign of IVar will be
reversed as well, so that the case of IVar > 0 will lead back to the case of IVar < 0,
which we have already excluded.
2. In the case of a maximum, i.e. Ivirt < Ireal and thus Ivirt −Ireal = IVar +Iremainder < 0,
−Iremainder > IVar applies. Here, the assumption IVar ≥ 0 leads to the inequality
−Iremainder > |IVar | , and the reduction of the variation by the factor M leads to
the modified inequality −Iremainder > M |IVar |. For this inequality to remain valid
also for factors M that are arbitrarily large, IVar = 0 ein. must apply. Again, if we
reverse the sign in the variation, i.e. δqi → −δqi , the alternative assumption IVar < 0
can be led back to the case of IVar > 0, which we have already excluded.
As a consequence of the considerations above,
ˆt2 X
N 
∂L d ∂L
 
IVar = − δqi dt = 0 (7.12)
i=1
∂qi dt ∂ q̇i
t1

109
 7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

q
H

H

H
C, l0

H

H

H
x 
H
q
? m
Figure 7.5: Schematic representation of a mass-on-spring oscillator under the influence of
gravity.

is a necessary condition that must be satisfied for an extremum. However, since the
variation δqi can be chosen arbitrarily except for the initial and final states (7.8, 7.9),
it follows that the expressions in the square brackets vanish. The proof is indirect: If
even just one of the terms in square brackets is not zero for a particular i, the associated
variation δqi can then be chosen in such a way that it has the same sign at any time.
Then, however, the above integral would have a positive value, which is a contradiction.
Thus the opposite assumption is false and the statement

∂L d ∂L
 
− =0 , i = 1, · · · , N (7.13)
∂qi dt ∂ q̇i

is proved. This is a system of N differential equations of up to second order, for N

unknown functions qi (t). These equations are called the Euler-Lagrange equations of the
system. They are equivalent to Newton’s equations of motion, which, however, will not
be proved here.

Example: Mass-on-spring oscillator Consider a mass m suspended from a spring

(spring constant C, relaxed length l0 ) as an example of a simple system with only one
kinematic degree of freedom q1 = x, where we only consider longitudinal movement in the
vertical direction, see the schematic representation in Fig. 7.5.
To formulate the Lagrangian, we are going to need the kinetic and the potential energy
m 2
T = ẋ ,
2
C
U = −mgx + (x − l0 )2 ,
2
The rest is like a “recipe”, starting with the Lagrangian

m 2 C
L (x, ẋ) = T − U = ẋ + mgx − (x − l0 )2 . (7.14)
2 2
Using
∂L ∂L
= mg − C (x − l0 ) , = mẋ ,
∂x ∂ ẋ
we obtain the Euler-Lagrange equation as

∂L d ∂L
 
0= − = mg − C (x − l0 ) − mẍ ,
∂x dt ∂ ẋ

110
7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

which can be rearranged to

mẍ = mg − C (x − l0 ) .
One can see that this equation is identical to the Newtonian equation of motion.
Using the substitution X := x − l0 − g/ω02 and the definition
s
C
ω0 := ,
m
the equation above can be written in the form

Ẍ + ω02 X = 0 .

This is a differential equation of a well-known form, namely that of a harmonic oscillation.

Its solution is also well-known and has the form X = A cos(ω0 t) + B sin(ω0 t), where the
constants A and B are obtained from the individual initial conditions.

7.3.3 Variational problems with constraints

In many cases, motion is not free but guided, for example by ropes or rails. In the process,
the respective guide provides a kinematic constraint, i.e. a defined relationship between
the coordinates of the motion, which can be written mathematically using a function F
of the generalised coordinates:
F (qi , t) = 0 . (7.15)
Using this equation, we can implicitly define the area or curve in space on which mass
points can move. As an example, we consider a circular motion in a plane, with r0 being
the radius and (x0 , y0 ) the centre of the circle. In Cartesian coordinates, by

(x − x0 )2 + (y − y0 )2 = r02 ,

the circle is represented which can also be written in the form (7.15) as

(x − x0 )2 + (y − y0 )2 − r02 = 0 .

We thus obtain the function F by means of comparison as

F (x, y) = (x − x0 )2 + (y − y0 )2 − r02 .

In most cases, the choice of the function F is not unique.

q In the present case, we can
describe the circle in the alternative form |y − y0 | = (x − x0 )2 − r02 and therefore use
q
F̃ (x, y) = |y − y0 | − (x − x0 )2 − r02 instead of the above function F . But of course one
always chooses a rather simple representation, so that one would prefer the previously
defined function F to F̃ . The choice of coordinates also plays a role here. A circular
motion around the origin would be represented rather in polar coordinates r φ, since in
this case the kinematic constraint simply is r = r0 . This, in turn, can be brought into the
form (7.15) as r − r0 = 0 and the function defining the constraint reads:

F (r) = r − r0 .

The fact that the function F in (7.15) may even be explicitly time-dependent, takes into
account the possibility of designing the rigid guide, in the general case, to be flexible and
thus variable in time.

111
 7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

For a given process qi (t), a constraint of the form (7.15) restricts the degree of freedom of
the variation δqi (t) in the following way: Using the Taylor expansion, we obtain
N
∂F
0 = F (qi + δqi , t) ≈ F (qi , t) + δqi + · · · ,
X

i=1
∂qi

which leads to
N
∂F
δqi = 0 , (7.16)
X

i=1
∂qi
taking into account F (qi , t) = 0 for the reference process and ignoring the higher-order
terms. This shows that only N −1 variations can be chosen freely from the N variations δqi ,
while the remaining variation results from the others in accordance with (7.16). Without
loss of generality, we consider all δqi for i = 1, · · · , N − 1 as variations that can be chosen
freely and δqN as the variation that depends on the former.1 This constraint of the
variation does not affect the derivation of the Euler-Lagrange equations in section 7.3.2
up to equation (7.12). From then on, however, the dependence (7.16) prevents equations
of motion being obtained. This can be remedied by resolving this dependence by means
of a suitable linear combination: To this end, we multiply (7.16) by a factor λ = λ(t),
the so-called Lagrange multiplier, that can be freely chosen at the beginning. Then we
integrate over the time interval [t1 , t2 ] and add the result to (7.12), which leads to :

ˆt2 X
N 
∂L d ∂L ∂F
  
IVar = − +λ δqi dt = 0 . (7.17)
i=1
∂qi dt ∂ q̇i ∂qi
t1

Since this relationship is valid for any Lagrange multiplier λ, we use a simple trick: We
choose λ in such a way that the equation
∂L d ∂L ∂F
 
− +λ =0 (7.18)
∂qN dt ∂ q̇N ∂qN

is satisfied automatically.2 , whereby (7.17) is reduced to

ˆt2 NX
−1 
∂L d ∂L ∂F
  
IVar = − +λ δqi dt = 0 . (7.19)
i=1
∂qi dt ∂ q̇i ∂qi
t1

The subtle but important difference from the original form (7.17) is the fact that the
dependent variation δN has been eliminated (by means of a linear combination) and that
now only those variations δq1 , · · · , δqN −1 occur that can be chosen freely and independently
of each other. But then we can follow the same line of argument that ultimately led to the
Euler-Lagrange equations in section 7.3.2: Since the remaining variations can be chosen
freely, also with respect to their sign, each summand in (7.19) must vanish, i.e.

∂L d ∂L ∂F
 
− +λ = 0, (7.20)
∂qi dt ∂ q̇i ∂qi

This initially applies to i = 1, · · · , N −1, but after taking into account (7.18) it is ultimately
valid for all i = 1, · · · , N . In this way, we have obtained the equations of a guided motion,
1
For this sake, ∂F/∂qN ̸= 0 must apply, which can be certified by permutation of the order of the qi .
2
This can easily be done by solving (7.18) for λ!

112
7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

thanks to the introduction of an additional unknown λ. That is not at all surprising,

because the equations (7.20) and the constraint (7.15) form a system of N + 1 equations.
Nevertheless, leaving mathematics aside, the question arises as to what the physical mean-
ing is of the Lagrange multiplier λ. We can easily find the answer to this question if we
don’t look from a mathematics perspective at the fact that equation (7.18) is satisfied by
choosing a suitable λ, but rather remind ourselves that a zero on the right-hand side of
an equation can be regarded as a requirement for an equilibrium. Thus λ has a physical
function, namely that of establishing an equilibrium. In the conventional approach within
the framework of Newton’s equations of motion, this is achieved by means of the normal
force, i.e. the reaction force perpendicular to the guide. Thus it is clear that λ can be
associated with this normal force3 , i.e. physically it can be interpreted as a reaction of the
system to the constraint.

Implementation of the constraint into Hamilton’s principle: If we compare the

Euler-Lagrange equations (7.13) of the free motion with those of a motion with a con-
straint, i.e. (7.20), we can see that equations (7.13) become (7.20) when the substitution
L → L + λF is performed. In other words: If we replace the Lagrangian (7.7) of the free
motion with the modified form

L (qi , q̇i , λ, t) := T (qi , q̇i ) − U (qi ) + λF (qi , t) , (7.21)

the dynamics of the system can ultimately be attributed again to a form of Hamilton’s
principle with the modified Lagrangian (7.21), in which all variables can formally be varied
freely, including the Lagrange multiplier λ: Its associated Euler-Lagrange equation is
nothing other than the constraint (7.15), while the variation with respect to the generalised
coordinates δqi reproduces the system of equations (7.20).
In standard textbooks, the topic of variational problems with constraints is covered very
briefly, by more or less postulating the form (7.21) of the Lagrangian. Here, in contrast,
it was shown how, through a systematic consideration, the Lagrange multiplier λ “arises”
out of a mathematical (elimination of a dependent variation by means of a linear com-
bination) or physical (realisation of an equilibrium by means of a compensating normal
force) necessity. This approach is equivalent to the original considerations of Lagrange in
his textbook from 1788 [4]4 .

Example: As a simple example, we consider a body (mass m) that is moving down an

inclined plane without friction and under the influence of gravity. The plane is inclined
against the horizontal by the angle α (Fig. 7.6). For the body,
mh 2 i
T (ẋ, ẏ) = ẋ + ẏ 2 ,
2
U (y) = mgy ,
F (x, y) = y + (tan α) x − h ,

is valid, where h is the height of the body’s centre of mass S if the body starts from the
initial position x = 0. The Lagrangian with constraint is then
mh 2 i
L (x, y, ẋ, ẏ, λ) = ẋ + ẏ 2 − mgy + λ [y + (tan α) x − h] .
2
3
Of course, λ may differ by a factor from the normal force, but in any case there is a connection!
4
see also the article [3] on the historical development of the Lagrangian formalism

113
 7.3. THE LAGRANGIAN FORMALISM IN CLASSICAL MECHANICS

mg

α
x
Figure 7.6: Body on an inclined plane.

From this, the Euler-Lagrange equations are obtained according to (7.13) as

δx : 0 = −mẍ + λ tan α , (7.22)
δy : 0 = −mÿ − mg + λ , (7.23)
δλ : 0 = y + (tan α) x − h . (7.24)
If we differentiate the third equation (7.24) twice with respect to the time, we obtain
ÿ = − (tan α) ẍ and can substitute this into equation (7.23), which then becomes
0 = m (tan α) ẍ − mg + λ .
The sum of this equation and the equation tan α (7.22) which was previously multiplied
by tan alpha yields
  λ
0 = −mg + 1 + tan2 α λ = −mg + ,
cos2 α
which we can solve for the Lagrange multiplier
λ = mg cos2 α .
If we substitute this into the other two equations, we obtain the simple equations
mẍ = mg cos α sin α ,
mÿ = −mg sin2 α
which can be directly integrated twice, with the result of
1
x = x0 + vx0 t + g cos α sin αt2 ,
2
1
y = y0 + vy0 t − g sin2 αt2 .
2
The four integration constants can be determined from the initial conditions as x0 = 0,
y0 = h, vx0 = vy0 = 0, i.e.
1
x = g cos α sin αt2 ,
2
1
y = h − g sin2 αt2 .
2
According to (7.20), the constraint force (normal force) is given as F⃗z = λ tan α⃗ex + λ⃗ey .
Thus its magnitude is
p λ
Fz = F⃗z = λ tan2 α + 1 = = mg cos α .
cos α

114
7.4. THE RAYLEIGH-RITZ METHOD

7.4 The Rayleigh-Ritz method

Up to now, when formulating the Lagrangian and subsequently the Euler-Lagrange equa-
tions, we have been using Hamilton’s principle merely as an elegant means to an end
in order to ultimately obtain Newton’s equations of motion in a different way, namely by
means of energy expressions. But as an alternative mathematical formulation of dynamics,
Hamilton’s principle also has some advantages that have not been used so far. How these
advantages can be used and transformed into a calculation method that is both elegant
and effective, is demonstrated in Appendix ??.

7.5 Motion under the influence of friction or other factors

7.5.1 Reformulating Newton’s equations as “Lagrange’s equations of the
second kind”
The starting point of our considerations are the well-known Newtonian equations. For
the sake of simplicity, we restrict ourselves to just one mass point and use the Cartesian
coordinates x1 = x, x2 = y und x3 = z. We now divide the total force F⃗ acting on the
mass point into three amounts:
F⃗ = −∇U (xi , t) + λj ∇Fj (xi , t) + R
⃗
X

These are the sum of all conservative forces, which can be represented as the negative
gradient of a potential energy U (xi , t) , the sum of all constraint forces that are caused
by kinematic constraints of the form Fj (xi , t) = 0, and finally the sum of all frictional
forces R,
⃗ whose form will be discussed in more detail later. Newton’s equations of motion
m⃗r¨ = F⃗ , which can alternatively be written as F⃗ − m⃗r¨ = ⃗0 (D’Alembert’s formulation of
the equations of motion), are then formulated for each component:
∂U X ∂Fj
− + λj + R1 − mẍ1 = 0 , (7.25)
∂x1 j
∂x1
∂U X ∂Fj
− + λj + R2 − mẍ2 = 0 , (7.26)
∂x2 j
∂x2
∂U X ∂Fj
− + λj + R3 − mẍ3 = 0 . (7.27)
∂x3 j
∂x3

Without frictional forces, the system could be described by a Lagrangian of the form
m 2 
L = L (xi , ẋi , λj , t) = ẋ1 + ẋ22 + ẋ23 − U (xi , t) + λj Fj (xi , t) ,
X
2 j

which is not possible in case of friction. Nevertheless, we can formally calculate the Euler–
Lagrange terms relating to the variations with respect to x1 , x2 , x3 as
∂L d ∂L ∂U X ∂Fj
 
− =− + λj − mẍ1 , (7.28)
∂x1 dt ∂ ẋ1 ∂x1 j
∂x1
∂L d ∂L ∂U X ∂Fj
 
− =− + λj − mẍ2 , (7.29)
∂x2 dt ∂ ẋ2 ∂x2 j
∂x2
∂L d ∂L ∂U X ∂Fj
 
− =− + λj − mẍ3 , (7.30)
∂x3 dt ∂ ẋ3 ∂x3 j
∂x3

115
 7.5. MOTION UNDER THE INFLUENCE OF FRICTION OR OTHER FACTORS

which can still be a useful tool for the mathematical reformulation of the Newtonian
equations of motion (7.25–7.27), which, after plugging in (7.28–7.30), are now

∂L d ∂L
 
− + R1 = 0 (7.31)
∂x1 dt ∂ ẋ1
∂L d ∂L
 
− + R2 = 0 (7.32)
∂x2 dt ∂ ẋ2
∂L d ∂L
 
− + R3 = 0 . (7.33)
∂x3 dt ∂ ẋ3
Note that these equations are not to be regarded as the Euler–Lagrange equations resulting
from Hamilton’s principle, but are merely a reformulation of the Newtonian equations of
motion. This representation has certain advantages over the classical formulation, since
at least all conservative forces occurring in the system can be taken into account through
potential energies, which is a simplification. However, the above form is restricted to
Cartesian coordinates. To extend the formalism to generalised coordinates, we will use
another trick: Taking into account the fact that frictional forces usually have the form

r˙
 ⃗
⃗r˙

⃗ = −f
R
⃗r˙

(see also Chapter ...) and using the formula

∂ ˙ ∂ q 2 ẋi ẋi
⃗r = ẋ1 + ẋ22 + ẋ23 = q = ,
∂ ẋi ∂ ẋi ẋ1 + ẋ2 + ẋ3
2 2 2 ⃗r˙

the three components of the frictional force can be expressed by

∂
D ⃗r˙ ,
 
Ri = − (7.34)
∂ ẋi
where D is the antiderivative of the function f and is called dissipation function. For
a Stokesian frictional force R = −k⃗r, for example, f ⃗r = k ⃗r˙ is valid and therefore
˙ ˙

⃗
1 2 1
D ⃗r˙ = k ⃗r˙ = k⃗r˙ 2 . applies. The decisive factor here is that D is a scalar just
 
2 2
like the Lagrangian L, i.e. a one-component quantity that remains invariant to a rotation
of the coordinate system and won’t change even if coordinates other than Cartesian are
used. Following this, the reformulated Newtonian equations (7.31–7.33), which can now
be written as
∂L d ∂L ∂D
 
− = ,
∂xi dt ∂ ẋi ∂ ẋi
can also be formulated using any given, i.e. generalised coordinates qi , so that we can write
∂L d ∂L ∂D
 
− = . (7.35)
∂qi dt ∂ q̇i ∂ q̇i
The proof is waived here. In this context, we speak of Lagrange’s equations of the sec-
ond kind. It should be emphasised once again that these equations were not obtained
from a variational principle, but are merely a reformulation of Newton’s equations of mo-
tion, albeit quite a useful one that allows us to transfer the advantages of the Lagrangian
formalism to the Newtonian equations of motion: We have only to determine scalar func-
tions (kinetic energy, potential energy, the dissipation function and, where applicable,

116
7.5. MOTION UNDER THE INFLUENCE OF FRICTION OR OTHER FACTORS

@
@
@
@
@q
A A A A A A qm q q
@







@ C, l0 k
@ -
@ x
@
Figure 7.7: Schematic representation of a dampened mass-on-spring oscillator.

constraints) that depend on the generalised functions and their first time derivatives, but
do not depend on the second derivatives. We don’t have to find the direction of a force,
and we don’t need to isolate the system from its surroundings either.
Of course, this applies not only to a single mass point, but also to a system of mass points,
and also to rigid bodies with rotational degrees of freedom.

Example: Mass-on-spring oscillator with Stokes friction

We consider a mass–spring oscillator again and add according to Fig. 7.7 a linear damper,
i.e. a component that induces a frictional force in line with Stokes’ law

Ri = −k ẋi .

According to (7.34) we now require a function D = D (ẋi ) whose derivative with respect to
ẋi will give us, bar the sign, the i–th component of the frictional force. This is obviously
given for D (ẋi ) = k ẋ21 + ẋ22 + ẋ23 /2, which can be written without coordinates as
 

k ˙2
D= ⃗r . (7.36)
2
This scalar quantity can be represented depending on any generalised coordinates. In the
present case, we have just one coordinate x, so that simply D = k ẋ2 /2 applies. Formulating
the Lagrangian equations of the second kind (7.35) with the Lagrangian (7.14) and the
dissipation function (7.36) thus yields

mg − C (x − l0 ) − mẍ = −k ẋ ,

After rearranging the equation, we recover the Newtonian equation of motion for a mass-
on-spring oscillator with Stokes friction:

mẍ = mg − C (x − l0 ) − k ẋ .

7.5.2 Strict variational formulations for systems with non-conservative

forces and for continuum mechanics
The reformulation of Newton’s equations of motion that was presented in the previous
section is without a doubt elegant, but it has the weakness of not being at the basis of a
true variational principle. Possible alternative methods are listed in Appendix ??.
Another field that needs to be developed further is that of variational principles in contin-
uum mechanics. Initial success in the field was achieved relatively early on [2]. In recent

117
 7.5. MOTION UNDER THE INFLUENCE OF FRICTION OR OTHER FACTORS

times, due to a systematic analysis based on fundamental symmetries, scientists were able
to formulate firm rules [7] that in the case of non-dissipative systems even reveal a general
pattern for the Lagrangian (actually: the Lagrangiandensity). But to this day, there are
open questions that continue to make this topic the subject of current research [8, 5, 9].

Exercises
Exercise 7.1: Concave mirror
(Solution cf. A.7.1)
r
6
 d(r)
q !!
!!
!!
!!
!!
ta
!!
```
F aa` ` f
aa`````
aa ```
aa
aa
a

A concave mirror collects light beams from a distant light source, which are in fact incident
parallel to the optical axis, in a focal point – but only if the mirror has a precisely defined
shape d(r)! A spherical shape, for example, leads to an aberration, the spherical aberration,
in which the rays further out are deflected too much. Using Fermat’s principle, determine
the ideal mirror shape d(r) for a given focal length f . Proceed as follows:

a) Justify the statement that Fermat’s principle requires that the light paths of all
incident rays from the focal plane to the focal point must be of equal length.

b) Determine the length of the light path of any ray from the focal plane as a function
of the axial distance r and the mirror shape d(r).

c) Infer the mirror’s shape d(r).

Exercise 7.2: Sliding rod

(Solution cf. A.7.2) y
6
A rod (length l, mass m, moment of inertia J = ml2 /12) slides
under the effect of gravity (acceleration due to gravity g) fric- CC
tionlessly down a wall. CC
CC
a) How do the coordinates of the centre of gravity xs and ys
@
φCC@ @
CC @ S
depend on the angle φ? yx c
s
b) Determine the kinetic energy, potential energy and La-
CC @ @
XX CC XX @
grange function as functions of φ and φ̇.
-x
CC XX
@
@X XX
xsx
@
c) Find the Euler–Lagrange equation.
CC @

Exercise 7.3: Sliding rope

(Solution cf. A.7.3)

118
7.5. MOTION UNDER THE INFLUENCE OF FRICTION OR OTHER FACTORS
m, l0

A massive rope (mass m, length l0 ) slides frictionless
from a table top. The movement starts at time t = 0
from rest, with exactly one half of the rope still on
the table and the other half hanging down from the
table. In the following neglect the curvature of the
x
short piece of the rope at the edge.
?

1. Determine the potential and kinetic energy of the rope and give the Lagrange func-
tion L as a function of x and ẋ.
2. Determine the Euler–Lagrange equation.
3. Show that the Euler–Lagrange equation has solutions of the form:
g g
r   r 
x(t) = A exp t + B exp − t
l0 l0
and determine the outstanding constants A and B for the above initial conditions.

Exercise 7.4: Multi-body dynamics

(Solution cf. A.7.4)
A roller (mass m1 , radius R, moment of inertia J1 = m1 R2 /2) is held on a vertical wall
by an inextensible rope which is attached to a weight (mass m2 ) via a pulley.
k
 a -
q
JJ The moment of inertia of the pulley is negligible. For t = 0 the
initial configuration is given by y = y0 , z = z0 and φ = 0. During
y
J
z J the movement of the roller, there is no slip, while losses due to
friction are not to be taken into account.
J
? J
m2 Jr? a) Express the lowering z of the weight and the angle of rotation
φ φ of the roller by the generalised coordinate y.
HH

b) Establish the Lagrange function L = L(y, ẏ) of the system.
m1 , J1
c) Find the Euler-Lagrange equation.

Exercise 7.5: Bead on a thin string

(Solution cf. A.7.5)
Let a bead (mass m) move frictionlessly on a A  d - d -

2 2
thin string fixed at two fixed points A and B, A
@sc - sc


forcing the bead to a movement on an ellipse
A x @

A A B

 2  2 A

x y
+ = 1, A

a b A y? l
|

A

with large and small semi-axis a and b, respectively.
a) Establish the Lagrange function L = L (x, y, ẋ, ẏ, λ) of the system. Let λ be the
Lagrange multiplier, by which the guidance of the motion on an ellipse is taken into
account.
b) Calculate the Euler-Lagrange equations.
c) How to calculate major and minor semi-axis a and b of the ellipse, if rope length l
and distance d between points A and B are given quantities?

119
 7.5. MOTION UNDER THE INFLUENCE OF FRICTION OR OTHER FACTORS

120
Chapter 8

Autonomous oscillations

Oscillation phenomena are periodic, i.e. repetitive processes. A system is able to oscillate
autonomously, if it has
1. a stable equilibrium and
2. the property of inertia.
The stable equilibrium has the effect that every displacement from the equilibrium position
is counteracted by a restoring “force”. As soon as the equilibrium position is reached,
the property of inertia causes the system to move beyond this position in the opposite
direction. The fact that such conditions are often present explains the frequent occurrence
of oscillation phenomena and their technological importance.

8.1 The fundamentals of oscillation analysis

8.1.1 The characterisation of oscillations
A function q(t) is called periodic, if it satisfies the condition
q(t + T ) = q(t) (8.1)
for all t with the period T . Examples of periodic functions are: q(t) = q0 sin(2πt/T )
and q(t) = q0 cos(2πt/T ), and also signals with a triangle, sawtooth or square waveform.
Oscillations that have a sine / cosine waveform are referred to as harmonic 1 , whereas
oscillations that have other waveforms are called anharmonic.
The reciprocal of the period, which is called the frequency
1
f= , (8.2)
T
as well as the angular frequency
2π
ω = 2πf =
, (8.3)
T
are often used instead of the period. Their practical significance lies primarily in the fact
that a harmonic oscillation can be written quite shortly in the form:
q(t) = q0 cos (ωt + φ0 ) . (8.4)
1
Since all types of periodic signals can be represented as superpositions of sine and cosine functions by
means of a Fourier analysis, these functions are of particular importance.

121
 8.1. THE FUNDAMENTALS OF OSCILLATION ANALYSIS

Here, q0 is the amplitude and φ0 is the phase or phase position.

For anharmonic oscillations, too, an amplitude can be generally defined as:
1
q0 = [max q(t) − min q(t)] . (8.5)
2
and also a middle position
1
qm = [max q(t) + min q(t)] , (8.6)
2
which is always zero in the case of the harmonic oscillation. In anharmonic oscillations,
however, the middle position is not necessarily the zero position. The latter is the position
of equilibrium: The quantity q is usually defined in such a way that the equilibrium position
corresponds to q = 0.
Fig. 8.1 shows an anharmonic oscillation in a q-t-diagram, indicating the quantities T , q0
and qm .

q(t)

q0
qm
t

q0

Figure 8.1: Visualisation of the motion of an undamped anharmonic oscillator, indicating

the quantities T , q0 and qm .

A harmonic oscillation can also be written elegantly in complex notation by means of

q0 h iωt+iφ0 i
q(t) = q0 cos (ωt + φ0 ) = e + e−iωt−iφ0
2
1 h iφ0 iωt i h i
= q0 e e + q0 e−iφ0 e−iωt = ℜ q̂eiωt , (8.7)
2
where the complex amplitude
q̂ = q0 eiφ0 (8.8)
contains information about both the actual amplitude and the phase. It is common to
drop the real number and simply use

q(t) = q̂eiωt . (8.9)

8.1.2 The oscillation equation

The question arises as to what form a system’s equation of motion (more generally: the
time evolution equation) must have for the system to be able to oscillate harmonically.

122
8.2. EXAMPLES OF FREE UNDAMPED OSCILLATIONS

q
H

H

H
k, l0

H

H

H
x 
H
q
? m
Figure 8.2: Schematic representation of a mass-on-spring oscillator with gravity.

To answer this question, we differentiate (8.9) twice with respect to the time and find
q̈ = −ω 2 q̂eiωt and thus the relationship

q̈ + ω 2 q = 0 , (8.10)

which is referred to as the oscillation equation for a free undamped harmonic oscillation. If
we succeed in bringing the evolution equation of a dynamic system into the form (8.10) by
rearranging the equation, we will have identified the system’s behaviour as an undamped
oscillation and we can read off the frequency directly (identification of parameters).

8.2 Examples of free undamped oscillations

8.2.1 Mass-on-spring oscillator
Consider a mass m suspended from a spring (stiffness k, relaxed length l0 ) as an example
of a simple system with only one kinematic degree of freedom q1 = x, considering only
longitudinal motion in the vertical direction, see the schematic representation in Fig. 8.2.

As the Newtonian equation of motion, we find

mg − k (x − l0 ) = mẍ , (8.11)

which yields the static equilibrium solution x0 = l0 +mg/k for the rest state x = x0 = const
of the oscillator. As a consequence,
mg
q = x − x0 = x − l0 − (8.12)
k

is identified as the displacement from the equilibrium position. If we apply the substitution
(8.12) to the equation of motion (8.11), we obtain the simplified differential equation
kq = mq̈, which can also be written as

k
q̈ + q = 0. (8.13)
m
|{z}
ω2

We identify this equation as the oscillation equation (8.10) and without the need to solve
the equation, we can state that the system’s
p behaviour is characterised by harmonic os-
cillations with the angular frequency ω = k/m.

123
 8.3. DAMPED FREE OSCILLATIONS

L C

Figure 8.3: Diagram of an LC circuit.

8.2.2 LC circuit
The simple circuit shown in Fig. 8.3, consisting of a coil with the inductance L and a
capacitor with the capacitance C, is known as an LC circuit. In order to understand its
function, we first consider the functional principles of its components: For the voltage UC
dropping across the capacitor, the relationship

Q
UC = , (8.14)
C

applies depending on the charge Q stored in the capacitor, while a voltage

UL = LI˙ , (8.15)

is dropping across the coil due to induction based on the change in the current I. According
to Kirchhoff’s voltage law,
Q
0 = UL + Uc = LI˙ + , (8.16)
C
is valid. As a consequence, only the relationship between the charge and the current,

Q̇ = I (8.17)

, known as the continuity equation, will be required to make equation (8.16) an equation
for one unknown, namely
Q
LQ̈ + = 0,
C
which, after dividing by L, can also be written as

1
Q̈ + Q=0 (8.18)
LC
|{z}
ω2

and thus in the form of the differential equation for oscillations (8.10). √In this way, the
angular frequency of the resulting oscillation can be identified as ω = 1/ LC.

8.3 Damped free oscillations

Real oscillators are always affected by damping factors. In general, damping appears in the
system’s evolution equation as an additional term that depends on the temporal change q̇
of the oscillation coordinate, i.e. on the generalised velocity.

124
8.3. DAMPED FREE OSCILLATIONS

8.3.1 Linear damping

In the (mathematically) simplest case, the damping in the evolution equation depends
linearly on the generalised velocity, which suggests an extension of the differential equation
(8.10) to
q̈ + 2δ q̇ + ω02 q = 0 , (8.19)
with a damping coefficient δ. As an example of such a case, we consider a mass-on-spring
oscillator with an additional damper (Fig. 8.4 on the left) an LC circuit that has an

R
@
@
@
@
@q
A A A A A A qm q q
L C
@







@ k d
@ -
@ x
@
Figure 8.4: A damped mass–on–spring oscillator and an LCR circuit as its electrical
analogue.

additional ohmic resistor (Fig. 8.4 on the right), which is called an LCR circuit.
In the mechanical damper, a piston is moving through a liquid and experiences a frictional
force d·ẋ that is proportional to the velocity, so that the corresponding Newtonian equation
of motion is
−kx − dẋ = mẍ . (8.20)
After dividing by m, this leads to an equation of the form (8.19), with ω0 = k/m and
p

2δ = d/m.
In the case of the LCR circuit, Ohm’s law must be used for finding the voltage dropping
across the resistor:
UR = RI = RQ̇ . (8.21)
This is supplemented by Kirchoff’s law (8.16), which yields
Q
0 = UL + UR + Uc = LQ̈ + RQ̇ + , (8.22)
C
and ultimately leads to:
R 1
Q̈ + Q̇ + Q = 0. (8.23)
L
|{z} LC
|{z}
2δ ω02

In this way, we have again obtained a differential equation of the form (8.19).
To solve the differential equation (8.19), we slightly modify the solution of the equation
for the undamped oscillation q̂ exp(iω0 t) and use the exponential approach
q(t) = q̂ exp (λt) (8.24)
with λ ∈ C as a reasonable generalisation. We can see that the undamped oscillation
would correspond to the special case λ = ±iω0 . Substituting (8.24) into (8.19) yields
h i
λ2 + 2δλ + ω02 q̂ exp (λt) = 0 ,

125
 8.3. DAMPED FREE OSCILLATIONS

which is satisfied only if the expression in the square brackets is zero, i.e.:

λ2 + 2δλ + ω02 = 0 (8.25)

This, however, is simply a quadratic equation for λ with the two solutions:
q q
λ1,2 = −δ ± δ 2 − ω02 = −δ ± i ω02 − δ 2 . (8.26)
| {z }
=:ω

We can see that it depends on the ratio of the damping coefficient δ to the natural frequency
ω0 whether the values λ1,2 are real or complex, which has a significant effect on the
character of the solution. For this reason, a case-by-case analysis is necessary:
1. The underdamped case is the case of not too high damping, δ < ω0 , for which
exp (λ1,2 t) = exp (−δt) exp (±iωt) applies and thus, after forming the real part, (8.24)
can be written more conveniently as

q(t) = e−δt [A cos(ωt) + B sin(ωt)] , (8.27)

with the constants A and B that have to be determined from the initial conditions.
2. The overdamped case, by contrast, occurs when damping is high, δ > ω0 , which
implies the overall solution

q(t) = | {z } +Be
Ae λ1 t λ2 t

dominant
to be a superposition of two decaying exponential functions, where the function
decaying more slowly determines the long-term behaviour of the system.
3. The case of critical damping, δ = ω0 , characterises the
q transition from the under-
damped to the overdamped case. As in this case ω = ω02 − δ 2 = 0 applies, this is
where the periodic motion (oscillation) is completely suppressed by the damping for
the first time. Therefore, the solution of the equation of motion for this case can
be obtained by applying the limit ω → 0 to (8.27). Note that with the Regel von
L’Hospital, it follows that
sin(ωt) t cos(ωt)
   
lim = lim =t
ω→0 ω ω→0 1
and thus (8.27) becomes:
q(t) = e−δt [A + Bt] . (8.28)

The direct comparison of the overdamped case with the critically damped case is also of
interest: Since q
|λ1 | = δ − δ 2 − ω02 < δ
always applies, the oscillation in the critically damped case is decaying faster than in the
overdamped case. This means that too much damping is counter-productive if you want
to equilibrate a system that is out of equilibrium as quickly as possible, as is the case with,
for example, the damper of a car (“shock absorber”). In fact, the critically damped case
ensures the fastest possible return to the system’s resting position.
The constants A and B, which can be determined by the respective initial conditions,
appear in all three cases. This is the subject of several exercises. All three cases are shown
in Fig. 8.5 for the initial conditions q(0) = q0 and q̇(0) = 0.

126
8.3. DAMPED FREE OSCILLATIONS

q(t)
q0
critically damped

overdamped
t

underdamped

Figure 8.5: Visualisation of the motion of an oscillator with linear damping and the initial
conditions q(0) = q0 and q̇(0) = 0, for the three different cases.

Exercises
Exercise 8.1: Special undamped oscillations
(Solution cf. A.8.1)
In the lecture q(t) = A cos(ω0 t) + B sin(ω0 t) was found as a general solution of the os-
cillation ODE q̈ + ω02 q = 0 with open constants A and B. In the following we call q the
(generalised) displacement and v = q̇ the (generalised) velocity, even if angles or com-
pletely different physical quantities can be hidden behind them.
1. Determine A and B for the case that the oscillator starts at time t = 0 with a
deflection q0 from rest, i.e. with vanishing (generalised) velocity. Write down the
final result for q(t) again. What velocity does the oscillator reach when passing
through the equilibrium position?
2. Determine A and B for the case that the oscillator passes the equilibrium position
at time t = 0 with a (generalised) velocity v0 = q̇0 . What is the maximum deflection
reached by the oscillator?

Exercise 8.2: Special damped Oscillations

(Solution cf. A.8.2)
For the damped oscillation (Stokes damping), given by the equation of motion q̈ + 2δ q̇ +
ω02 q = 0 with δ < ω0 (oscillation case), one canqgive the general solution in the form
q(t) = exp(−δt) [A cos(ωt) + B sin(ωt)] with ω = ω02 − δ 2 and open constants A and B.
1. Determine A and B for the case where the oscillator starts from rest at time t = 0
with a displacement q0 . Then write down the final result for q(t) again. Sketch the
motion.
2. Determine A and B for the case where the oscillator crosses the equilibrium position
at time t = 0 with a (generalized) velocity q̇0 . What is the maximum displacement
reached by the oscillator and after what time? Sketch the motion.

Exercise 8.3: To the aperiodic limit case

(Solution cf. A.8.3)
We consider the damped oscillation again in the aperiodic limit case δ = ω0 , for which the
general solution transitions to q(t) = exp(−δt) [A + Bt] with open constants A and B.

127
 8.3. DAMPED FREE OSCILLATIONS

1. Determine A and B for the case where the oscillator crosses the equilibrium position
at time t = 0 with a (generalized) velocity v0 = q̇0 . At what time t = tm does the
oscillator reach its maximum displacement q(tm ) and what is the latter? Sketch the
motion.
2. Determine A and B for the case where the oscillator starts at time t = 0 with a
displacement q(0) = q0 > 0 and a (generalized) velocity q̇(0) = v0 < −δ · q0 . At what
time t1 is the equilibrium position first reached? At what later time t2 is maximum
deflection reached for the last time and what is its magnitude? Sketch the motion.

Exercise 8.4: Thermodynamic and Vibrations

(Solution cf. A.8.4)
An ideal gas enclosed in a vertical cylinder with adiabatic exponent γ carries pa
a freely moving piston with mass m and cross section A. At equilibrium, the
gas column has height h, and x denotes the vertical displacement of the piston x
from the equilibrium position. The external pressure pa and the gravitational m
acceleration g are also given.
1. What value p0 does the internal pressure take at the equilibrium position A
x = 0?
2. What is the pressure change p − p0 when the piston is displaced x as- p
suming adiabatic change of state? h
3. Show that for very small x the relationship between p − p0 and x is γ
approximately linear, i.e. p − p0 ≈ ax and determine the factor a.
hint: As is often the case, (1 + ε)γ ≈ 1 + γε helps here too!
d) What follows from (c) for the resultant force on the piston?
e) Show that Newton’s equation of motion of the piston for small x has the form of an
oscillation equation. Determine the angular frequency ω of the oscillation.

128
Chapter 9

System dynamics aspects

This chapter, like the previous chapter 8, deals with oscillations. However, not with au-
tonomous oscillations, i.e. oscillations of systems that are left to themselves, but with
oscillations that are caused by external influences. This raises questions of system dy-
namics, specifically the question of how a system responds to a particular excitation. As
always, advanced considerations such as these can, unfortunately, only be covered briefly
due to the limited amount of time.

The interaction of oscillating systems with each other will also be discussed in this chapter.
For this purpose, we will be using all the tools from the Mathematics I course: this means
that, besides calculating with complex numbers, we will be dealing with systems of linear
equations and the associated matrices.

9.1 Forced oscillations

9.1.1 The general form of the equation of motion
A damped oscillation inevitably comes to rest after a long time, but before this can
happen, it must have been initiated. This can be done either by means of a sophisticated
mechanism that is maintaining the oscillation by removing energy from a reservoir in
order to balance damping loss (self-sustaining oscillation), or by means of an external
force (forced oscillation). We only consider the latter case, which can be represented by
the extended differential equation

q̈ + 2δ q̇ + ω02 q = f (t) , (9.1)

where the inhomogeneity f (t) stands for the external force acting on the system and is
referred to as excitation.

9.1.2 Possible types of excitation

Any (decent) function can be represented as a superposition of harmonic functions by
means of a Fourier transformation! Therefore, it is sufficient to analyse a harmonic exci-
tation f (t) = fˆ exp(iωt) and thus solve the equation:

q̈ + 2δ q̇ + ω02 q = fˆ exp(iωt) . (9.2)

129
 9.1. FORCED OSCILLATIONS

f (t) f (t) f (t)

∞

t
t t

Figure 9.1: Three possible ways to excite a system that is able to oscillate: stepwise (left),
pulse (middle) and harmonic excitation (right).

9.1.3 The resulting motion and resonance curve

With a continuous excitation, the motion depends on the individual initial conditions –
as is the case with any motion. However, on assuming that after a sufficiently long time
a form of motion is established that is sustained, we use the approach

q(t) = q̂ exp(iωt) (9.3)

for a harmonic oscillation whose frequency is the same as the excitation frequency. If we
substitute this into the differential equation (9.2), we obtain:

−ω 2 + 2iδω + ω02 q̂ exp(iωt) = fˆ exp(iωt)

h i

This can easily be solved for the complex amplitude:

fˆ
q̂ = (9.4)
ω02 − ω 2 + 2iδω

Thus we have proved that the excited system oscillates at the excitation frequency, but
with a modified amplitude and phase, which is taken into account by means of a complex
factor. The amplitude varies depending on the excitation frequency. For the limiting case
of ω → 0, we obtain q̂ = fˆ/ω02 as the displacement when the external force does not change
or changes very slowly. Based on this, we can also write the relationship (9.4) as

ω02 fˆ
q̂ = (9.5)
ω02 − ω 2 + 2iδω ω02

and thus identify a complex factor by which the resulting amplitude differs from the static
amplitude. Specifically, we consider the magnitude of this factor, which is called the
transmissibility:
ω02
V =q
2 (9.6)
ω02 − ω 2 + 4δ 2 ω 2
Fig. 9.2 shows its graph as a function of the excitation frequency ω for different degrees
of damping. Likewise, the argument ψ of this complex factor, given as
2δω
tan ψ = (9.7)
ω02 − ω2,

is of importance, as it reflects the phase difference of the forced oscillation with respect
to the oscillation’s excitation. The phase difference approaches zero for ω → 0 (in-phase
oscillation); for ω = ω0 it is exactly π/2 and for very large values of ω, it asymptotically
approaches the value π (oscillation in phase opposition). Fig. 9.2 shows the transmissibility

130
9.1. FORCED OSCILLATIONS

V ψ

π
1 2

ω ω
ω0 ω0

Figure 9.2: √
Resonance curves for different degrees of damping. From bottom to top:
δ/ω0 = 1, 1/ 2, 1/2, 1/3, 1/5.

as a function of the excitation’s angular frequency for different degrees of damping, the so-
called resonance curves, along with the corresponding graphs of the phase difference, called
the phase response. In cases of low damping, the resonance curves show a pronounced
maximum in the vicinity of the natural angular frequency ω0 of the undamped oscillator.
However, to accurately determine the resonance frequency, we have to find the zero of the
first derivative, i.e.:
dV ω02 4ω ω 2 − ω02 + 8δ 2 ω


0= =− q
dω
2 3
2 ω02 − ω 2 + 4δ 2 ω 2
This equation can only be satisfied if the numerator becomes zero, which means that
h i
4ω ω 2 − ω02 + 2δ 2 = 0 ,

which inevitably leads to the position of the resonance maximum1 at:

q
ωres = ω02 − 2δ 2 , (9.8)

so it’s just below the natural frequency ω0 when damping is low, i.e. δ ≪ ω0 , but signifi-
cantly far away from it as soon as the degree of damping is of the same order of magnitude
√
as the natural frequency. Once the damping coefficient δ exceeds the critical value ω0 / 2,
a resonance maximum even no longer exists.
√
The peak of the resonance, provided that it exists (i.e. δ < ω0 / 2), is found by substi-
tuting (9.8) into (9.6) and we obtain:

ω 2
Vres = q 0 (9.9)
2δ ω02 − δ 2

We can see that this quantity, which describes the factor by which the excited oscillation in
resonance is higher than its excitation, can have arbitrarily large values for correspondingly
small values of the damping coefficient. Particularly low damping can lead to an oscillation
amplitude of a destructive nature, which is also called “resonance disaster”.
1
In addition, the second zero at ω = 0 indicates that the resonance curve always starts with a horizontal
tangent. Depending on the degree of damping, we have either a minimum or a maximum here.

131
 9.2. COUPLED OSCILLATIONS

9.2 Coupled oscillations

9.2.1 The double oscillator and its application as a tuned mass damper
A mass damper has the function of suppressing unwanted vibrations of an object. For
this purpose, it builds up a counter vibration that absorbs an externally excited vibration.
The kind of mass damping discussed below is called passive mass damping.

Figure 9.3: Schematic representation of a double oscillator with harmonic excitation

through the displacement of the suspension point.

In order to understand the concept, we consider a suspended double oscillator as a model

system, as is shown in Fig. 9.3. The mass m1 reflects the system’s inertia and the spring
c1 the system’s stiffness, while the mass m2 and the spring c2 together have the function
of the mass damper.
The Newtonian equations of motion for this system are:

m1 ẍ1 = −c1 (x1 − xe ) + c2 (x2 − x1 ) , (9.10)

m2 ẍ2 = −c2 (x2 − x1 ) , (9.11)

where x1 and x2 are the displacements of the respective masses from their equilibrium
position and xe is the displacement of the suspension point,

xe (t) = x̂e exp (iΩt) , (9.12)

by which the oscillation is excited in the first place.

Finding the natural frequencies

First we have to find the frequencies at which the system is able to oscillate autonomously,
i.e. without external excitation. To this end, we need to show that, for the equations (9.10)
and (9.11) with xe = 0, there are solutions of the form:

x1 (t) = x̂1 exp (iωt) , (9.13)

x2 (t) = x̂2 exp (iωt) . (9.14)

132
9.2. COUPLED OSCILLATIONS

The second derivative yields ẍ1,2 = −ω 2 x̂1,2 exp (iωt), and substituting this into the equa-
tions of motion (9.10) and (9.11) leads to
h i
−m1 ω 2 x̂1 + c1 x̂1 − c2 (x̂2 − x̂1 ) exp (iωt) = 0 , (9.15)
h i
−m2 ω 2 x̂2 + c2 (x̂2 − x̂1 ) exp (iωt) = 0 . (9.16)

After multiplying by exp (−iωt), this results in a system of equations for the amplitudes
x̂1,2 , which can be written as

c1 + c2 c2
 
ω − 2
x̂1 + x̂2 = 0 (9.17)
m1 m1
c2 c2
 
x̂1 + ω 2 − x̂2 = 0 (9.18)
m2 m2

after simple rearrangements. It is even better to use the matrix form:

! ! !
c1 +c2 c2
ω2 − x̂1 0
m1 m1 = , (9.19)
c2
m2 ω2 − c2
m2 x̂2 0

Since this system of equations is homogeneous, there is always the trivial solution x̂1,2 ,
which corresponds to the equilibrium position. Non-trivial solutions can only exist if the
determinant of the system’s matrix vanishes. Thus it follows that

ω2 − c1 +c2 c2
c1 + c2 c2 c22
  
0= c2
m1 m1
c2 = ω2 − ω2 − −
m2 ω2 − m2 m1 m2 m1 m2
c1 + c2 c2 c1 c2
 
=ω − 4
+ ω2 +
m1 m2 m1 m2
c1 + c2 c1 + c2 c1 + c2
( )
c2 c2 2 c2 2 c1 c2
    
=ω −2
4
+ ω +
2
+ − + −
2m1 2m2 2m1 2m2 2m1 2m2 m1 m2
| {z }
 h i2
c1 +c2 c
ω2 − 2m1
+ 2m2
2
  
= ω 2 − Ω21 ω 2 − Ω22 (9.20)

with the squares of the natural frequencies:

s
c1 + c2 c1 + c2
2
c2 c2 c1 c2
Ω21,2 = + ∓ + − ,
2m1 2m2 2m1 2m2 m1 m2
s
c1 + c2 c1 + c2
2
c2 c2 c22
= + ∓ − + . (9.21)
2m1 2m2 2m1 2m2 m1 m2

We can see from the second line that the radicand is always positive and, therefore, the
squares of the natural frequencies are always real, while the first line shows us that the
squares are positive and thus the two natural frequencies themselves are also positive.

Knowing the natural frequencies is important, because in the mass damping problem
below we need to make sure that the excitation frequency is different from the natural
frequencies in order to prevent a resonance disaster.

133
 9.2. COUPLED OSCILLATIONS

Steady-state solution of the equations of motion with excitation

We now consider exclusively the established system of two bodies that are oscillating with
their individual amplitudes, but at the same frequency Ω of the external excitation, i.e.
we are looking for solutions of the form

x1 (t) = x̂1 exp (iΩt) (9.22)

x2 (t) = x̂2 exp (iΩt) (9.23)

with the respective amplitudes x̂1,2 that we still have to determine. The second derivative
yields ẍ1,2 = −Ω2 x̂1,2 exp (iΩt), and substituting this into the equations of motion (9.10)
and (9.11) leads to
h i
m1 Ω2 x̂1 − c1 (x̂1 − x̂e ) + c2 (x̂2 − x̂1 ) exp (iΩt) = 0 , (9.24)
h i
m2 Ω2 x̂2 − c2 (x̂2 − x̂1 ) exp (iΩt) = 0 . (9.25)

After multiplying by exp (−iΩt) and rearranging the equations, we obtain a system of
linear equations for x̂1,2 that has the form

c1 + c2 c2 c1
 
Ω2 − x̂1 + x̂2 = − x̂e , (9.26)
m1 m1 m1
c2 c2
 
x̂1 + Ω2 − x̂2 = 0 . (9.27)
m2 m2
If we solve the second equation for x̂1 , i.e.
m2 c2
 
x̂1 = − Ω2 − x̂2 , (9.28)
c2 m2
substitute this into the first equation and then apply the method of completing the square
(9.20), we obtain

c1 + c2
( )
m2 c2 c22 c1
 
− Ω −
2
Ω − 2
− x̂2 = − x̂e (9.29)
c2 m1 m2 m1 m2 m1
| {z }
(Ω2 −Ω21 )(Ω2 −Ω22 )

and thus, ultimately, the solution:

c1 c2
 
− Ω2
m1 m
2
x̂1 =
x̂e (9.30)
Ω2 − Ω21 Ω2 − Ω22
c1 c2
m
1 m2
x̂2 =
x̂e (9.31)
Ω2 − Ω21 Ω2 − Ω22

Fig. 9.4 shows both amplitudes as a function of the excitation frequency Ω. While the
amplitude of the mass damper m2 is always unequal to zero, the amplitude x̂1 of the
system’s mass m1 does become zero, namely at
c2
r
Ω= . (9.32)
m2

134
9.2. COUPLED OSCILLATIONS

Figure 9.4: The amplitudes of the two masses as functions of the excitation frequency.

Using this relationship, the mass damper can be tuned in such a way that it completely
absorbs all vibrations of a given frequency Ω. Tuned mass damper can be found, for
example, in cars and in high-voltage power lines, but also in the design of earthquake-
resistant buildings, as seismic waves mostly have a dominant frequency. Undamped passive
mass dampers like this are less suitable for systems that are subjected to interferences of
a diverse frequency spectrum. In such cases, the mass damper can be equipped with a
damping mechanism or an active regulator.

9.2.2 A chain of rotating bodies

φk−1 φk φk+1

Jk−1 Jk Jk+1
ck−1 ck ck+1

Figure 9.5: A chain of rotating bodies.

We consider a linear arrangement of rotating bodies along an axis. The bodies are linked
by means of thin flexible torsion bars, see Fig. 9.5. Jk denotes the moment of inertia of
the k-th rotating body and φk its displacement from the resting position. In addition, ck
is the torsional stiffness of the flexible bar between the k-th and the (k + 1)-st body.
Euler’s equation of motion for the k-th link of the chain is then:

Jk φ̈k = ck (φk+1 − φk ) + ck−1 (φk−1 − φk ) (9.33)

We have to formulate boundary conditions for the first (k = 0) and the last link (k = N +1).
These could be:
1. The chain is clamped at both ends: φ0 = 0 and φN +1 = 0
2. Both ends are free: c0 = 0 and cN = 0
3. The chain is clamped at one end and free at the other end: φ0 = 0 and cN = 0.

135
 9.2. COUPLED OSCILLATIONS

In each case, (9.33) provides a system of second-order differential equations that can be
compactly represented in the following matrix form:
c0 + c1 0
 
−c1 ···
0 0 0
      
J1 ··· φ̈1  −c c1 + c2 −c2  φ1
1
0 0 0
 
J2   φ̈2   ..   φ2 
    
 .  +  0 . 0 =
 
 . 
  ..   ... 
.. −c2 ,

0   ..  

0

0 .   
.. ..
 
. .

−cN −1  φ
 
··· 0 JN φ̈N 
N 0
| {z } 0 −cN −1 cN −1 + cN | {z }
M | {z } φ
K
(9.34)
which yields the equation
M φ̈ + K φ = 0 (9.35)
with M being the mass matrix, K the stiffness matrix and φ = φ(t) the state vector.
However, the latter isn’t a vector in the physical sense, i.e. it doesn’t describe a directed
quantity in three-dimensional space and, for this reason, is not denoted here by an arrow
above the symbol, but by an underscore. Mathematically, it’s an N -dimensional vector
and the corresponding vector space is referred to as the state space.
A system like this with N degrees of freedom is capable of various types of motion. We
search for particular solutions of the system of differential equations (9.35) of the form
φ(t) = φ̂ exp (iωt) , (9.36)
where each of the rotating bodies in the chain is oscillating with an individual amplitude
(and phase), but all bodies are oscillating at the same angular frequency ω. Once we have
found all solutions of this type, which are called the normal modes of the system, we will
be able to represent the general motion of the chain as a superposition of these normal
modes. If we substitute (9.36) into the system of equations (9.35), take into account
φ̈(t) = −ω 2 φ̂ exp (iωt) and then multiply by exp (−iωt), we finally obtain the system of
linear equations h i
K − ω 2 M φ̂ = 0 (9.37)
Like in the case of the double oscillator covered in section 9.2.1, this is a homogeneous
system of equations. As a consequence, non-trivial solutions are possible only if the de-
terminant of the system’s matrix vanishes, i.e.:
h i
det K − ω 2 M = 0 (9.38)
Mathematically, this equation is to be categorised as finding the zeros of a polynomial of
degree N with respect to ω 2 . Accordingly, it has N solutions ωn2 . Their square roots ωn
are the system’s natural frequencies.

Special case: a chain of uniform bodies and torsion bars that is clamped at
both ends
In this case, c0 = c1 = c2 = · · · = cn = c and J1 = J2 = · · · = JN = J apply, leading to:
2 0 0
 
−1 ···
..  1
0 ··· 0
 

−1 2 −1 . 
0 0 1
   
K = c
 .. .. 
M =J

(9.39)
 0 −1 . . 0,

.. ,
 . 0 0

. 0
 . ..
 
 . . −1 ··· 0 1


0 ··· 0 −1 2

136
9.2. COUPLED OSCILLATIONS

so that with the definition

J 2
λ= ω , (9.40)
c

the system of equations for finding the normal modes can be represented in the form

2−λ 0 0
 
−1 ···
0
   
φ̂1
 .. 
 −1 2 − λ −1 .   0
 φ̂
 2
 .  .
   
.. ..
0   ..
 = . (9.41)

 0

−1 . . 
 .
 .
 . ..  φ̂N −1  0
   
 . . −1 
φ̂N 0
0 ··· 0 −1 2 − λ

and the equation for finding the natural frequencies is

2−λ −1 0 ··· 0
..
−1 2 − λ −1 .
.. ..
0 −1 . . 0 = 0. (9.42)
.. ..
. . −1
0 ··· 0 −1 2 − λ

For the complete calculation of the normal modes, we can first determine the N eigenvalues
λn as the solutions of (9.42), which will lead to the natural frequencies (the eigenfrequen-
cies) ωn in accordance with (9.40). Following this, we can find the normal modes that
correspond to the respective natural frequencies by substituting ω = ωn into (9.41), where
one of the amplitudes must be predefined.

The problem will be discussed further in one of the exercises, namely for the case of N = 3.

Exercises
Exercise 9.1: Torsional oscillator chain with N = 3
(Solution cf. A.9.1)
Consider the torsional oscillator chain from section 9.2.2, specifically one consisting of
N = 3 equal rotating bodies with mass moment of inertia J and equal elastic rods with
uniform torsional stiffness c under fixed restraint on both sides.

1. Determine the three natural frequencies ω1 , ω2 , ω3 of the system.

2. Determine the corresponding eigenmodes (at given amplitude φ̂1 ).

3. Plot the three modes in diagrams where you plot x1 (t), x2 (t), x3 (t) together in one
diagram against time t. So make one diagram for each eigenmode.

Exercise 9.2: Eigenmodes of the two-mass oscillator

(Solution cf. A.9.2)

137
 9.2. COUPLED OSCILLATIONS

The two-mass oscillator was discussed in detail but not

yet fully in section 9.2.1. Consider again the case of
free oscillation without external excitation (xe = 0), for
which the natural frequencies Ω1,2 have been calculated
but not the natural modes x̂1,2 . In the following, x̂1 is
taken as given.
1. Compute x̂2 from (9.19) for ω = Ω1 .
2. Compute x̂2 from (9.19) for ω = Ω2 .
3. Give the results calculated before in general for
the special case m1 = 4m2 , c1 = 3c2 .
4. Plot the motion of both masses, i.e. x1 (t) and
x2 (t) in a graph against time, separately for both
frequencies Ω1,2 .

Exercise 9.3: Coupled oscillations

(Solution cf. A.9.3)
x1
Consider the oscillator chain shown on the right, con- x2
sisting of two frictionless moving masses in the ratio c m1 c c
m2
m2 /m1 = 1/2 and three springs of equal stiffness c.
Let x1 , x2 be the displacement of the respective mass from its rest position.
1. Set up the Newtonian equations of motion of the two masses.
2. Determine the two natural angular frequencies ω1 and ω2 , i.e., those frequencies at
which the system can move according to x1 = x̂1 exp (iω1,2 t)and x2 = x̂2 exp (iω1,2 t)
can autonomously oscillate harmonically, depending on c and m2 .
note: The substitution λ = m2 ω 2 /c simplifies the problem a bit.

138
Appendix A

Solutions of the Exercises

A.0 Regarding measurements

A.0.1 Vereinfachte Berechnung der Standardabweichung

Ausgehend von der ursprünglichen Definition (7) folgt durch Verwendung der zweiten
binomischen Formel zunächst:

s s s
N N N  2 
σ= (x − x̄)2 = (x2 − 2xx̄ + x̄2 ) = x − 2x̄x + x̄2
N −1 N −1 N −1

Da für die Mittelwertbildung (6) offensichtlich cx = cx̄ gilt, wenn c eine beliebige Kon-
stante ist, gilt auch 2x̄x = 2x̄x̄ = 2x̄2 . Einsetzen in obige Gleichung liefert die zu be-
weisende Aussage
s
N  2 
σ= x − x̄2 .
N −1

Dass x2 nicht kleiner als x̄2 sein kann, folgt allein schon aus der oben gezeigten Identität
x2 − x̄2 = (x − x̄)2 ≥ 0.

A.0.2 Bestimmung der Masse von Weinbeeren

1. Sortiert man die Tabelle ein wenig um, so ergeben sich folgende Daten

Index n 1 2 3 4
Messwert mn /g 2 3 4 5
Anzahl zn 2 4 5 1
Häufigkeit pn 1/6 1/3 5/12 1/12

und daraus folgendes Histogramm:

139
 A.0. REGARDING MEASUREMENTS

pn

1
12

mn
g
0 1 2 3 4 5 6

2. Als Mittelwert ergibt sich aus den Einzelmessungen: m̄ = 12 g

41
≈ 3,4167g.
3. Die zugehörige Standardabweichung ist:
σ = (2 · 172 + 4 · 25 + 5 · 49 + 192 )/122 /11g ≈ 0,9003g.
p

4. Die Unsicherheit des Mittelwerts ergibt folglich: ∆m̄ = √σ12 ≈ 0,2599g, so dass das
Resultat in der Form m̄ = (3,41 ± 0,26) g sinnvoll dargestellt ist.
5. Ergänzt man das Histogramm um Mittelwert, Standardabweichung und Gaußsche
Verteilungsfunktion, so erhält man:
pn
m̄/g

1
12

mn
g
0 1 2 3 4 5 6
Die Übereinstimmung zwischen Gaußscher Verteilungsfunktion und realer Mess-
wertverteilung ist recht gut, allerdings gewinnt man den Eindruck, dass letztere
doch eine Asymmetrie aufweist, die der Gaußschen Theorie fremd ist. In Anbetra-
cht der geringen Zahl von Messungen (N = 12) kann man aber nicht mit Gewissheit
sagen, ob dies Zufall ist oder ob die Größenverteilung von Weinbeeren einer vom
reinen Zufall abweichenden Gesetzmäßigkeit folgt.
6. Aus der gemessenen Gesamtmasse resultiert die mittlere Masse m̄∗ = 12
47
g ≈ 3,9167g,
während man die Unsicherheit auf Grundlage der Ablesegenauigkeit ±1g mit ∆m̄ =
12 = 12 g ≈ 0,0833g beziffern kann. Somit ergibt sich m̄ = (3,917 ± 0,083) g.
∆m 1 ∗

7. Die beiden Resultate liegen wechselseitig außerhalb der Fehlerschranke des jeweils
anderen Wertes und sind somit nicht miteinander konsistent! Vermutlich erklärt sich

140
A.0. REGARDING MEASUREMENTS

dies mit der Unzulänglichkeit des Messgeräts (Küchenwaage), das möglicherweise

keine korrekte Rundung vornimmt, sondern einfach die Kommastellen abschneidet.
Wenn dem so sein sollte, dann wäre die Messung der Gesamtmasse weniger von
diesem systematischen Fehler betroffen als die Einzelmessungen. Unabhängig davon
ergeben sich bei der Gesamtmessung auch kleinere Unsicherheiten.

A.0.3 Freitag der 13.

1. Dargestellt in einer Tabelle ergibt sich für die betreffenden 28 Jahre:

Index n 1 2 3
Vorkommen im Jahr Nn 1 2 3
Anzahl Jahre zn 12 12 4
Häufigkeit pn 3/7 3/7 1/7

2. Als Mittelwert ergibt sich aus den Einzelmessungen:

3+6+3 12
N̄ = 1p1 + 2p2 + 3p3 = = ≈ 1,7143 .
7 7

3. Die zugehörige Standardabweichung ist:

s √
12 · (5/7)2 + 12 · (2/7)2 + 4 · (9/72 ) 4 14
σ= = ≈ 0,7127.
27 21

4. Erstellt man aus den Häufigkeiten ein Histogramm, so sieht das so rein gar nicht wie
eine Gauß-Verteilung aus. Woran liegt das?

◦ Das liegt einfach nur an der geringen Zahl von Daten. Würde man anstatt
28 Jahren ein komplettes Jahrtausend betrachten, so käme die Verteilung der
Gaußschen Statistik recht nahe.

⊗ Die Häufigkeit des Freitags, des 13. beruht nicht auf Zufall, sondern ist durch
den Gregorianischen Kalender deterministisch vorherbestimmt. Die Gaußsche
Statistik gilt dagegen nur für zufällige Ereignisse.

◦ Jede Regel hat ihre Ausnahme: Andere Tage (z.B. Mo. der 16.) genügen der
Gauß-Statistik weitgehend. Dieser mit rationalen Argumenten nicht erklär-
baren Auffälligkeit hat der Freitag der 13. seinen schlechten Ruf als Unglückstag
zu verdanken.

A.0.4 Lineare Regression

1. Berechnet man zuerst den natürlichen Logarithmus der Zahl der Infizierten ln (n (t))
so ergeben sich folgende Werte:

141
 A.0. REGARDING MEASUREMENTS

t/d n ln (n)
5 400 5,99
6 639 6.46
7 795 6,68
8 902 6,80
9 1139 7,04
10 1296 7,17
11 1567 7,36
12 2369 7,77
13 3062 8,03
14 3795 8,24
15 4838 8,48
Mit den Formeln aus der Vorlesung für die Koeffizienten der linearen Regression
ergibt sich durch Einsetzen der Wertepaare (t| ln (n))

ln (n0 ) = 4, 921
γ = 0, 235 1/d,

dabei laufen die Summen bis N = 11, da insgesamt elf Wertepaare in der Aufgabe
zur Verfügung stehen.

2. Trägt man in einem Diagramm nun die Wertepaare (t| ln (n)) und die berechnete
Regressionsgerade ln (n (t)) = γt + ln (n0 ) auf, so erkannt man qualitativ eine gute
Übereinstimmung.

Kleiner Tipp: Die lineare Regression bieten auch zahlreiche Plotprogramme an, die
häufig auch auf die in der Vorlesung gezeigt Methode der kleinsten Fehlerquadrate
zurückgreifen.

3. Setzt man für t = 21d ein, so errechnet man auf Basis der linearen Regression den
natürlichen Logarithmus der Zahl der Infizierten am 21. März. Wendet man dann
die Exponentialfunktion darauf an, so ergibt sich nach Rundung die Zahl der In-
fizierten zu n (21d) = 19207. Leider sind an diesem Wochenende die Fallzahlen der

142
A.1. REGARDING KINEMATICS

Gesundheitsämter nicht vollständig an das Robert Koch Institut übermittelts wor-

den, wodurch eine Aussage über die getroffenen Maßnahmen, an Hand der vorhan-
denen Daten, nicht möglich ist. 1

A.0.5 Fehlerfortpflanzung
Die Messgröße liegt in der Standarddarstellung a = ā ± ∆a vor mit ā = 9,35cm und
∆a = 0,05cm.
1. Die relative Unsicherheit der Messung beträgt: ∆a/a = 0,05/9,35 ≈ 0,0053 = 0,53%
2. Der funktionale Zusammenhang A = f (a) zwischen Mess- und Zielgröße ist hier
durch f (a) = a2 gegeben. Der Mittelwert der Fläche einer Würfelseite ergibt sich
einfach auf Grundlage des Mittelwertes von a als Ā = f (ā) = ā2 ≈ 87,4225cm2 ,
während die Unsicherheit aus dem Fehlerfortpflanzungsgesetz (= faktisch Satz von
Taylor) als ∆A = f ′ (ā)∆a = 2ā∆a ≈ 0,94cm2 . Damit lautet das Resultat in
korrekter Schreibweise A = Ā ± ∆A:

A = (87,42 ± 0,94)cm2

Man beachte hier die Rundung des Mittelwerts in Korrespondenz mit der Zahl
gültiger Stellen der Unsicherheit!
3. Für das Volumen des Würfels werden alle vorangegangenen Schritte wiederholt,
diesmal aber für V = f (a) mit veränderter Funktion f (a) = a3 . Dies führt auf
V̄ = ā2 ≈ 817,4004cm3 und ∆V = f ′ (ā)∆a = 3ā2 ∆a ≈ 13cm3 und damit:

V = (817 ± 13)cm3

A.1 Regarding kinematics

A.2 Regarding dynamics of mass points

A.3 Regarding constants of motion

A.4 Dynamics of rigid bodies

A.5 Thermodynamics

A.6 Fluid dynamics

A.7 Regarding variational calculus

A.8 Regarding autonomous vibrations

A.9 Regarding system dynamics

1
(https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/Fallzahlen.
html, 22.03.2019, 18:52)

143
 A.9. REGARDING SYSTEM DYNAMICS

144
Appendix B

Supplements

B.1 About Taylor expansion of second and higher order

B.1.1 Taylor expansion of second order
We obtain the second-order Taylor expansion arising from this idea by applying (14)
mutatis mutandis to the derivative f ′ (x) instead of to the function itself, i.e.
f ′ (x) ≈ f ′ (x0 ) + f ′′ (x0 ) (x − x0 )
and then finding the antiderivative on both sides
1
f (x) ≈ f ′ (x0 ) x + f ′′ (x0 ) (x − x0 )2 + C ,
2
in order to obtain the corresponding form of the Taylor polynomial
1
p2 (x) = f ′ (x0 ) x + f ′′ (x0 ) (x − x0 )2 + C (B.1)
2
with a yet unknown constant of integration C. The latter is determined via the condition
p2 (x0 ) = f (x0 )
and leads to C = f (x0 ) − f ′ (x0 ) x0 , which, after applying this to (B.1), finally implicates
the approximation formula
1
f (x) ≈ p2 (x) = f (x0 ) + f ′ (x0 ) (x − x0 ) + f ′′ (x0 ) (x − x0 )2 (B.2)
2
This equation differs from the first-order expansion (14) only by one additional quadratic
term. The example in Fig. B.1 shows the result of a first- and second-order Taylor expan-
sion of a given function. It can be clearly seen that, with the second-order approximation,
a larger range of the function is represented in a satisfactory manner.

B.1.2 Taylor expansion of higher order

The above procedure can be repeated as often as desired in order to obtain higher-order
Taylor polynomials. Using mathematical induction, one can show that the Taylor poly-
nomial of order N is given by
N
1 (n)
f (x) ≈ pN (x) = f (x0 ) (x − x0 )n . (B.3)
X

n=0
n!

145
 B.1. ABOUT TAYLOR EXPANSION OF SECOND AND HIGHER ORDER

f (x) f (x)

0,5 0,5

x x
1 x0 2 1 x0 2
x2
Figure B.1: Taylor expansion of the function f (x) = of first order (left) and second
1 + x2
3
order (right), both w.r.t. operation point x0 = .
2

B.1.3 Taylor series

When the limit N → ∞ is reached, the Taylor polynomial changes into an infinite series,
the Taylor series, which not only approximates the function to be expanded, but even
represents it exactly according to

∞
1 (n)
f (x) = f (x0 ) (x − x0 )n (B.4)
X

n=0
n!

provided that certain conditions are met, which we cannot go into here, however.

Example: exponential function We perform Taylor expansion of f (x) = exp (x)

around x0 = 0. Since the exponential function reproduces itself at each derivative,

f (n) (x0 ) = exp (0) = 1 ,

is valid here for all n. Thus the Taylor series is

∞
xn x2 x3 x4
exp (x) = =1+x+ + + + ···
X

n=0
n! 2 6 24

We will use this now to determine Euler’s number e = exp(1) to within one percent: as is
well known, 5! = 120, so we can truncate the Taylor series after the fifth term:

5
1 1 1 1 1 120 + 120 + 60 + 20 + 5 + 1 163
e = exp(1) ≈ = 1+1+ + + + = = ≈ 2,717
X

n=0
n! 2 6 24 120 120 60

Comparing this result with the actual value e = 2,718281828 · · · shows that the two
numbers coincide in the first two decimal places, as required.

Sine and cosine Since we already know the Taylor series of the exponential function,
we can use Euler’s formula (??) to obtain the Taylor series of the sine and cosine functions

146
B.2. ABOUT CURVILINEAR COORDINATES

around x0 = 0. First,
∞ ∞ ∞
(ix)n (ix)2k (ix)2k+1
exp (ix) = = +
X X X

n=0
n! k=0
(2k)! k=0
(2k + 1)!
∞  k x2k ∞  k x2k+1
= i2 + i i2
X X

k=0
(2k)! k=0
(2k + 1)!
∞ ∞
x2k x2k+1
= (−1)k +i (−1)k
X X

k=0
(2k)! k=0
(2k + 1)!

applies. leads directly to:

∞
x2k+1 x3 x5
sin (x) = (−1)k =x− +
X
− ···
k=0
(2k + 1)! 6 120
∞
x2k x2 x4
cos (x) = (−1)k =1− +
X
− ··· .
k=0
(2k)! 2 24

B.2 About curvilinear coordinates

B.2.1 Spherical coordinates
The three corresponding unit vectors ⃗er , ⃗eϑ and ⃗eφ are, in turn, defined in such a way that
each of the three vectors is pointing in the direction in which the respective coordinate
increases. Since the angular coordinate, in particular, is the same for polar, cylindrical
and spherical coordinates, this also results in an identical unit vector ⃗eφ , which is defined
by the equation (B.5), i.e.
⃗eφ = − sin φ⃗ex + cos φ⃗ey . (B.5)
In contrast, the unit vector ⃗er , which points exactly in the direction of the point in space
to be considered, is different from the vector (1.12) of the same name in polar coordinates,
since the latter is not directed towards the spatial point itself, but towards its projected
point. However, if one goes by length r sin ϑ in the direction defined by (1.12) and then
by r cos ϑ in the z–direction, one reaches the position of the mass point, i.e.

⃗r = r sin ϑ [cos φ⃗ex + sin φ⃗ey ] + r cos ϑ⃗ez = r [sin ϑ cos φ⃗ex + sin ϑ sin φ⃗ey + cos ϑ⃗ez ] .
(B.6)
Thus, the unit vector we are looking for, which points in the direction of the mass point,
can be identified as

⃗er = sin ϑ cos φ⃗ex + sin ϑ sin φ⃗ey + cos ϑ⃗ez . (B.7)

This makes it possible to write (B.8) as

⃗r = r⃗er . (B.8)

The two unit vectors ⃗eφ and ⃗er are mutually orthogonal in accordance with

⃗eφ · ⃗er = [− sin φ⃗ex + cos φ⃗ey ] · [sin ϑ cos φ⃗ex + sin ϑ sin φ⃗ey + cos ϑ⃗ez ]
= − sin ϑ sin φ cos φ + sin ϑ sin φ cos φ = 0 .

147
 B.2. ABOUT CURVILINEAR COORDINATES

In order to be able to represent further vectors, in addition to the already defined unit
vectors ⃗eφ and ⃗er , the required third unit vector is defined via the cross product as

⃗eϑ = ⃗eφ × ⃗er

= [− sin φ⃗ex + cos φ⃗ey ] × [sin ϑ cos φ⃗ex + sin ϑ sin φ⃗ey + cos ϑ⃗ez ]
h i
= cos ϑ cos φ⃗ex + cos ϑ sin φ⃗ey − sin ϑ sin2 φ + cos2 φ ⃗ez
= cos ϑ cos φ⃗ex + cos ϑ sin φ⃗ey − sin ϑ⃗ez .

By definition, this vector is orthogonal to both ⃗eφ and ⃗er and has the magnitude |⃗eϑ | =
|⃗eφ | |⃗er | sin (90°) = 1. Hence the three unit vectors ⃗er , ⃗eϑ and ⃗eφ form an orthonormal
basis, with which any vector can be represented. For the velocity, in particular, applies

d
⃗v = ⃗r˙ = (r⃗er ) = ṙ⃗er + r⃗e˙ r (B.9)
dt

For the time derivative of ⃗er ,

⃗e˙ r =
h i
cos ϑϑ̇ cos φ − sin ϑ sin φφ̇ ⃗ex
h i
+ cos ϑϑ̇ sin φ + sin ϑ cos φφ̇ ⃗ey
− sin ϑϑ̇⃗ez
= ϑ̇⃗eϑ + sin ϑφ̇⃗eφ , (B.10)

applies, and thus

⃗v = ⃗r˙ = ṙ⃗er + rϑ̇⃗eϑ + r sin ϑφ̇⃗eφ . (B.11)

Accordingly, the following applies for the acceleration:

d 
⃗a = ⃗v˙ =

ṙ⃗er + rϑ̇⃗eϑ + r sin ϑφ̇⃗eφ
dt
= r̈⃗er + ṙ⃗e˙ r + ṙϑ̇⃗eϑ + rϑ̈⃗eϑ + rϑ̇⃗e˙ ϑ
+ṙ sin ϑφ̇⃗eφ + r cos ϑϑ̇φ̇⃗eφ + r sin ϑφ̈⃗eφ + r sin ϑφ̇⃗e˙ φ . (B.12)

Now the time derivatives of the unit vectors still need to be expressed by a linear com-
bination of the unit vectors, as has already been done for ⃗e˙ r in (B.13). This must be
complemented with

⃗e˙ ϑ =
h i
− sin ϑϑ̇ cos φ − cos ϑ sin φφ̇ ⃗ex
h i
+ − sin ϑϑ̇ sin φ + cos ϑ cos φφ̇ ⃗ey
− cos ϑ⃗ez
= −ϑ̇⃗er + cos ϑφ̇⃗eφ (B.13)

und

⃗e˙ φ = −φ̇ [cos φ⃗ex + sin φ⃗ey ]

= −φ̇ [sin ϑ⃗er + cos ϑ⃗eϑ ]
= −φ̇ sin ϑ⃗er − φ̇ cos ϑ⃗eϑ , (B.14)

148
B.2. ABOUT CURVILINEAR COORDINATES

in order to, finally, obtain the form we are looking for:

d 
⃗a = ⃗v˙ = ⃗r¨

ṙ⃗er + rϑ̇⃗eϑ + r sin ϑφ̇⃗eφ
hdt i
= r̈ − rϑ̇2 − r sin2 ϑφ̇2 ⃗er
h i
+ 2ṙϑ̇ + rϑ̈ − r sin ϑ cos ϑφ̇2 ⃗eϑ
h i
+ 2ṙ sin ϑφ̇ + 2r cos ϑϑ̇φ̇ + r sin ϑφ̈ ⃗eφ (B.15)

B.2.2 Natural coordinates

We assume that the position vector is according to (1.23) given as a function of the arc
length.
Depending on the current position of the mass point, we can define three mutually orthog-
onal unit vectors, the so-called Frenet frame, as follows: is defined as a tangent vector

⃗t (s) := d⃗r .
ds
By definition, this vector always points in the current direction of motion,
d⃗r ⃗r (s + ∆s) − ⃗r (s)
= lim ,
ds ∆s→0 ∆s
and therefore has the length ⃗t = 1, since in the limit ∆s → 0 there is no longer any
difference between the arc length ∆s between two positions and the length of the secant
⃗r (s + ∆s) − ⃗r (s).
Since ⃗t(s)2 = 1 therefore applies, it further follows that
d ⃗2 d⃗t
0= t = 2⃗t · ,
ds ds
and from this, in turn, that the derivative of ⃗t with respect to s yields a vector per-
pendicular to ⃗t. Its magnitude is the curvature of the path at the mass point’s current
position,
d⃗t
κ := , (B.16)
ds
while its direction defines the second basis vector,
1 d⃗t
⃗n := , (B.17)
κ ds
the so-called normal vector. Note that the above definition assumes a non-vanishing
curvature. In contrast, the direction of the normal is not unique in the case of a linear path.
In such a case, however, one would not use natural coordinates anyway, but Cartesian ones.
The third basis vector, in turn, is given by the cross product of the other two basis vectors
as
⃗b := ⃗t × ⃗n
and is called a binormal vector. Consequently, Any vector can be represented as a linear
combination of the three unit vectors ⃗t, ⃗n and ⃗b. Specifically, for the velocity applies
d⃗r ds
⃗v = ⃗r˙ = = ṡ⃗t ,
ds dt

149
 B.2. ABOUT CURVILINEAR COORDINATES

so that ṡ is, except for the sign, the magnitude of the velocity, i.e. the speed. The
acceleration is calculated as

d  ⃗ d⃗t ds
⃗a = ⃗v˙ = ṡt = s̈⃗t + ṡ = s̈⃗t + κṡ2⃗n
dt ds dt
and is thus naturally divided into a tangential acceleration s̈, which corresponds to the
change of the path velocity (speed) with time, and a normal acceleration κṡ2 , by which
only the change of the direction of motion with time is indicated. A given vector ⃗u, in
turn, can be represented as a linear combination of all three basis vectors, i.e.

⃗u = ut⃗t + un⃗n + ub⃗b . (B.18)

In order to determine its time derivative, we first need the derivatives of all three basis
vectors with respect to the arc length s. We already know from (B.17) that

d⃗t
= κ⃗n (B.19)
ds
applies. For the other two basis vectors one obtains the corresponding relations for the
other two basis vectors by differentiating the orthogonality and normalisation conditions
with respect to the arc length s. In particular, by differentiating ⃗n2 = 1 and ⃗t · ⃗n = 0, one
obtains
d⃗n
⃗n · = 0
ds
d⃗
⃗t · n d⃗t
= − · ⃗n = −κ
ds ds
from which it follows that d⃗n/ds is to be represented as a linear combination

d
⃗n = −κ⃗t + γ⃗b , (B.20)
ds

where the torsion γ is obtained by finding the dot product of the above equation with ⃗b

d⃗n
γ := ⃗b · .
ds
From the definition of the binormal vector, we finally obtain its derivative with respect to
the arc length directly as

d⃗ d⃗t d⃗n h i
b = × ⃗n + ⃗t × = κ⃗n × ⃗n + ⃗t × −κ⃗t + γ⃗b
ds ds ds
⃗ ⃗
= γ t × b = −γ⃗n (B.21)

The three identities (B.19,B.20,B.21) are called Frenet-Serret formulas. Using these for-
mulas, one can finally determine the time derivative of the vector given according to (B.18)
as
⃗u˙ = [u̇t − κun ṡ] ⃗t + [u̇n + (κut + γub ) ṡ] ⃗n + [u̇b + γun ṡ] ⃗b .

Natural coordinates prove particularly useful in the context of guided motion. In the case
of free motion, there is a certain disadvantage in that the path of the motion is a priori
unknown and thus also the vectors ⃗t, ⃗n and ⃗b as well as the curvature κ and the torsion γ.

150
B.2. ABOUT CURVILINEAR COORDINATES

z
6

y
6 
h

?
b -x

R~

Figure B.2: Schraubenbewegung.

Example: accelerated screw motion

Again, we consider the motion along a helical path from section 1.1.4 and formulate it
in natural coordinates. However, here we assume the generalised case of a tangentially
accelerated motion.
Let the geometry of the helical path be defined by the radius R of the circle resulting from
the projection of the path onto the xy–plane and by the vertical distance h between two
successive turns, see Fig. B.2. The simplest way to calculate the arc length s1 that a mass
point travels when completing a turn, is to unwind the helix and thereby convert the path
into a linear ramp. Then we obtain:
s1 
 p
h
 p
2πR s1 = 4π 2 R2 + h2

This allows us to write the dependence of the azimuth angle on the arc length as
s
φ = 2π , (B.22)
s1
while the height coordinate z is related to the arc length via
s
z=h . (B.23)
s1
The distance r from the axis is constant and given as
r=R (B.24)
it is therefore independent of s. The equations (B.22, B.23) und (B.24) legen zusammen
die Abhängigkeit des of the position vector from the arc length as
s
⃗r = ⃗r (s) = r⃗er + z⃗ez = R⃗er + h ⃗ez .
s1
From this the tangent vector is calculated as

⃗t = d⃗r d⃗er dφ h
=R + ⃗ez
ds dφ ds s1
2πR h
= ⃗eφ + ⃗ez .
s1 s1

151
 B.3. SELECTED MATHEMATICAL PROBLEMS

Its derivative with respect to the arc length

d⃗t 2πR d⃗eφ dφ 4π 2 R
= = − 2 ⃗er
ds s1 dφ ds s1
can be identified with κ⃗n. Therefore, we obtain
⃗n = −⃗er
as the path’s normal vector and
4π 2 R 1 1
κ= 2 =
s1 h 2
R
 
1+
2πR
as its curvature. The binormal vector can be calculated as
⃗b = ⃗t × ⃗n = 2πR ⃗ez − h ⃗eφ .
s1 s1
Now the torsion of the path can be calculated as
d⃗n 2πR h d⃗er dφ
 
γ = ⃗b · =− ⃗ez − ⃗eφ ·
ds s1 s1 dφ ds
2πh 2πh
= = 2 2 .
s21 4π R + h2

B.3 Selected mathematical problems

B.3.1 Obtaining the solution for the body falling under friction form
Eq. (2.18) for n = 2 (Newton case)
Computing the integral in Eq. (2.18)
In order to compute the Integral
ˆv(t)
dv
2 − v2
, (B.25)
v∞
0
the integrand is additively decomposed as follows:
1 1 1 1 1
 
= = + . (B.26)
v∞2 − v2 (v∞ − v) (v∞ + v) 2v∞ v∞ − v v∞ + v
Considering this, the Integral is split into two integrals,
ˆv(t) ˆv(t) ˆv(t)
 
dv 1  dv dv 
= + , (B.27)
2
v∞−v 2 2v∞ v∞ − v v∞ + v

0 0 0

each of which is computed separately via reconsidering the already computed integral
(2.19) for the Stokes case:
ˆv(t)
dv v (t)
 
= ± ln 1 ± , (B.28)
v∞ ± v v∞
0
allowing to determine the integral (B.25) as:
ˆv(t) 
v(t)

dv 1 1+ v∞ 1 v∞ + v (t)
 
= ln  = ln , (B.29)
2
v∞ − v 2 2v∞ 1− v(t) 2v∞ v∞ − v (t)
0 v∞

152
B.3. SELECTED MATHEMATICAL PROBLEMS

Solving (2.18)
By inserting the integral (B.29) into (2.18) with v(0) = 0 and n = 0, one obtains:

v∞ + v (t) k
 
ln = 2 v∞ t (B.30)
v∞ − v (t) |p{zm}
gk
m

as implicit form of the solution. In order to obtain v (t) explicitly, the exponential func-
tion exp(· · · ) is applied
 on both
 sides of the equation and then the resulting equation is
q
gk
multiplied with exp − mt [v∞ − v (t)], leading to:
 s  s 
gk  gk 
exp − t [v∞ + v (t)] = exp  t [v∞ − v (t)] .
m m

By bringing all terms with v(t) to the left hand and all other terms to the right hand,
 s   s   s   s 
exp 
gk  gk  gk  gk 
t + exp − t v (t) = exp  t − exp − t v∞
m m m m

and resolving w.r.t. v(t), we finally obtain

q   q 
gk gk
exp mt − exp − mt
s 
gk 
v (t) = v∞ q   q  = v∞ tanh  t (B.31)
gk gk m
exp mt + exp − mt

with the tangent hyperbolic function.

153
 B.3. SELECTED MATHEMATICAL PROBLEMS

154
Bibliography

[1] M. Abramowitz and I. A. Stegun. Handbook of Mathematical Functions: With For-

mulas, Graphs, and Mathematical Tables. Applied mathematics series. Dover Publi-
cations, 1964.
[2] A. Clebsch. Ueber die Integration der hydrodynamischen Gleichungen. J. f. d. reine
u. angew. Math., 56:1–10, 1859.
[3] M. J. Gandler and G. Wanner. Lagrange-Multiplikatoren: Von der Mechanik zur
Optimalen Steuerung. In J. Schröder and A. Klawonn, editors, GAMM Rundbrief
2/2014, Schierling, 2014. Bauer Satz.Druck.Werbetechnik GmbH.
[4] J.-L. Lagrange. Mécanique Analytique. Desaint, Paris, 1788.
[5] F. Marner, M. Scholle, D. Herrmann, and P. H. Gaskell. Competing Lagrangians
for incompressible and compressible viscous flow. Royal Society Open Science,
6(1):181595, Jan 2019.
[6] L. Papula. Mathematische Formelsammlung: Für Ingenieure und Naturwis-
senschaftler. Springer Fachmedien Wiesbaden, 2017.
[7] M. Scholle. Construction of Lagrangians in continuum theories. Proc. R. Soc. Lond.
A, 460(2051):3241–3260, 2004.
[8] M. Scholle and F. Marner. A non-conventional discontinuous Lagrangian for viscous
flow. Royal Society Open Science, 4(2):160447, 2017.
[9] Markus Scholle, Florian Marner, and Philip H. Gaskell. Potential fields in fluid
mechanics: A review of two classical approaches and related recent advances. Water,
12(5), 2020.
[10] J. H. Spurk and N. Aksel. Fluid Mechanics. Springer, 2008.

155