Dynamics LectureNotes

Dynamics
(Fall Semester 2021)
Prof. Dr. Dennis M. Kochmann
Mechanics & Materials Lab

Institute of Mechanical Systems
Department of Mechanical and Process Engineering
ETH Zürich
e3
B
e3=e3
R B
e3 =e3
A C
e3A=e3C
y
n n
e2 B
e2
R
e2 B
CM + e2 A
CM + e2 A
+ +
O O e1 R
j e1=e1
B A j e1=e1
B
A A
nutation
Copyright c 2021 by Dennis M. Kochmann

Dynamics Prof. Dr. Dennis M. Kochmann
Fall 2021 (version November 8, 2021) ETH Zürich
These lecture notes cover the concepts and most examples discussed during lectures.
They are not a complete textbook, but they do provide a thorough introduction to all course
topics as well as some extra background reading, extended explanations, and various examples
beyond what can be discussed in class.
Especially in times of covid-19, you are strongly recommended to visit the (virtual or in-person)
lectures and to take your own notes, while studying these lecture notes alongside for support and
further reading. The weekly exercise sessions will help deepen the acquired knowledge and
practice problem solving techniques.
The lecture topics covered during each week are announced at the beginning of the semester in
the course syllabus, so you are welcome read up on the topics ahead of time.
Contents
0 Preface 1
1 Single-Particle Dynamics 3
1.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Kinematics in a Cartesian reference frame . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Constrained motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.3 Kinematics in polar coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.1.4 Kinematics of space curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1.5 A brief summary of particle kinematics . . . . . . . . . . . . . . . . . . . . . 14
1.2 Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.1 Newton’s axioms and balance of linear momentum . . . . . . . . . . . . . . . 15
1.2.2 Work–energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.2.3 Balance of angular momentum . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.4 Particle impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.3 Summary of Key Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2 Dynamics of Systems of Particles 45

2.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2 Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.2.1 Balance of linear momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.3 Particle Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.3.1 Collision of two particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.3.2 Particles of variables mass: mass accretion and mass loss . . . . . . . . . . . 59
3 Dynamics of Rigid Bodies 64

3.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.1.1 Rigid-body kinematics in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.1.2 Rigid-body kinematics in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2.3 Moment of inertia tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.2.4 Angular momentum transfer formula . . . . . . . . . . . . . . . . . . . . . . . 97
3.3 Collision of Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.4 Non-Inertial Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.4.1 Active and passive rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.4.2 Rotating frames of reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.4.4 Balance of angular momentum, Euler equations . . . . . . . . . . . . . . . . . 133
3.5 Application: Spinning Tops and Gyroscopes . . . . . . . . . . . . . . . . . . . . . . . 141
i
3.5.1 General description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

3.5.2 Axisymmetric bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4 Vibrations 152
4.1 Lagrange Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.2 Mechanical Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.3 Single-Degree-of-Freedom Vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.3.1 Definitions and equation of motion . . . . . . . . . . . . . . . . . . . . . . . . 158
4.3.2 Free vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.3.3 Forced vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.4 Multi-DOF Vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.4.1 Equations of motion for multi-DOF vibrations . . . . . . . . . . . . . . . . . 172
4.4.2 Free, undamped vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
4.4.3 Damped and forced vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.4.4 Modal Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
5 Dynamics of Deformable Bodies 193

5.1 Dynamics of Systems with Massless Deformable Bodies . . . . . . . . . . . . . . . . 193
5.2 Dynamics of Deformable Bodies with Non-Negligible Mass . . . . . . . . . . . . . . . 198
5.3 Waves and Vibrations in Slender Rods . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.3.1 Longitudinal wave motion and vibrations . . . . . . . . . . . . . . . . . . . . 204
5.3.2 Torsional wave motion and vibrations . . . . . . . . . . . . . . . . . . . . . . 211
5.3.3 Flexural wave motion and vibrations . . . . . . . . . . . . . . . . . . . . . . . 212
Appendix 218
A What is a tensor? 218

A.1 What is a vector? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
A.2 What is a matrix? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
A.3 What is a tensor? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
A.4 Brief summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Index 223
Glossary 226
ii
0 Preface
These course notes summarize the contents of the course called Dynamics, which constitutes the
third part of the series of fundamental mechanics courses taught within ETH Zürich’s Mechanical
and Civil Engineering programs. Following Mechanics 1 (which introduced kinematics and stat-
ics) and Mechanics 2 (which focused on concepts of stresses and strains and the static deformation
of linear elastic bodies), this course studies the time-dependent behavior of mechanical systems –
from individual particles to systems of particles to rigid and, ultimately, deformable bodies. We will
discuss how to describe and understand the time-dependent motion of systems (generally referred
to as the kinematics), followed by the relation between a system’s motion and those externally
applied forces and torques that cause the motion (generally referred to as the kinetics). Besides
presenting the underlying theory, we will study numerous examples to highlight the usefulness of the
derived mathematical relations and assess their practical relevance. It is assumed that participants
have a proper understanding of the contents of Mechanics 1 and 2 (so we will keep repetitions to
a minimum) as well as of Analysis 1 and 2 (so we can exploit those mathematical concepts in our
derivations). Where necessary, we will introduce new mathematical and physical principles along
the way to the extent necessary.
At the end of each section, you will find a Summary of Key Relations: a concise table with the most
important equations derived and required to solve the exercise, homework and exam problems. The
sum of all those tables will also serve as the formula collection you will be allowed to consult during
the final exam of this course (as the only available supporting material). It is therefore advised to
familiarize oneself with those summary tables during the course (e.g., by using those to solve the
exercise problems), so that they form a familiar reference by the time of the exam. In case this
is your first encounter of mechanics in English, you may find the Glossary (starting on page 235)
at the end of these course notes helpful, which include German translations of the most important
terminology as well as a brief description of those terms. Most terms highlighted in blue throughout
the notes can be found in the glossary (just click on those in the PDF version).
Before we dive into dynamics, let us recap on just about one page and in a cartoon fashion the key
concepts of mechanics to be expanded here:
Mechanics 1 concentrated on static (i.e., non-moving) systems and established that the resultant
force R and the resultant torque M acting on a body (or a system of bodies) vanish in equilibrium:
n
X n
X
R= Fi = 0, M= ri × Fi = 0.
i=1 i=1
For the below example of “Tauziehen”, we must have F1 = −F2 for the teams to be in equilibrium.
F1 F2
1
Mechanics 2 went a step further and showed that inner forces are responsible for deformation, and
constitutive relations were discussed which link the deformation of a body (described by strains)
to the causes of deformation (described by stresses). Throughout, inner and external forces were
assumed to be in static equilibrium, so that the above relations of mechanical equilibrium (vanishing
resultant forces and torques) still applied.
N N
F1 -N -N F2
l + Dl
Finally, this course in Dynamics (which is essentially Mechanics 3 ) will address scenarios in which
the resultant forces and torques are no longer zero:
n
X n
X
R= Fi 6= 0, M= ri × Fi 6= 0.
i=1 i=1
In this case, the system is not in static equilibrium anymore but instead will respond to the applied
forces and torques with motion, governed by – among others – Newton’s famous second law,
F = ma.
Our goal will be to extend the concepts from Mechanics 1 and 2 to bodies and systems in motion.
To start simple, we will first discuss the dynamics of particles (i.e., bodies of negligible small size),
which we gradually extend to systems of particles, which then naturally leads to rigid and finally
deformable bodies.
F1 a(t) F2
I apologize in advance for any typos that may have found their way into these lecture notes. This
is a truly evolving set of notes that was initiated in the fall semester of 2018 and has been extended
and improved ever since. Though I made a great effort to ensure these notes are free of essential
typos, I cannot rule out that some have remained. If you spot any mistakes, feel free to send me a
highlighted PDF at the end of the semester, so I can make sure those typos are corrected for future
years. I would like to thank Prof. George Haller and his team for their course slides, some of which
served as the basis for these lecture notes. I am also grateful to Dr. Paolo Tiso for many helpful
discussions and to all those students who have pointed out typos.
I hope you will find the course interesting and these notes supportive while studying Dynamics.
Dennis M. Kochmann
Zürich, November 2021
2
1 Single-Particle Dynamics
We begin simple – by discussing the mechanics of a single particle, which you may be familiar
with from physics courses. We will review and explain the key concepts for our needs, as this will
serve as the basis for our later discussion of the mechanics of a rigid or deformable body. We
also use this introductory chapter to lay out our notation and terminology.
A particle is an idealized view of a (small) object whose size and shape have a negligibly influence
on its motion. Of course, no object is negligibly small in reality. However, when the object’s shape
and size are such that they do not significantly affect the object’s motion, then we may safely assume
that the total mass m of the object is concentrated in a single point (without a particular shape
or size), and that its motion can be described as a translation through space without considering
rotations of the object about any of its axes.
As an example, consider the flight of a golf ball over a long distance. To a good approximation,
the golf ball is a point moving through 3D space and – unless one wants to study the intricate fluid
mechanics around the ball’s surface – the exact shape and size of the ball is of little relevance. Its
flight can be described by its time-dependent position only. By contrast, consider throwing a book
up in the air. Depending on whether or not you give the book an initial spin (and about which axis
you spin it), the book’s tumbling motion can change dramatically. This is a scenario in which the
book cannot be viewed as a particle of irrelevant size and shape, but its rotation can significantly
affect its motion, and its size and shape influence how it tumbles.
A key consequence of the above assumptions is that the motion of a particle can be described
uniquely by translational degrees of freedom only, and we can safely neglect any rotational degrees
of freedom. In the following, we will focus on a single particle and study its motion through space
(which we call its kinematics), and we will discuss how its motion depends on the forces acting
on the particle (which we call the kinetics).
1.1 Kinematics
1.1.1 Kinematics in a Cartesian reference frame
We describe the positions of particles in space within a fixed Carte- x2

sian reference frame C in 3D, defined by an orthonormal basis of
vectors {e1 , e2 , e3 } and a fixed origin O. Consequently, any position m
in space can uniquely be described by the coordinates {x1 , x2 , x3 }.
We denote the position of a particle by r ∈ R3 , writing r (t)
e2
e3 O e1 x1
3
X
r= xi ei = x1 e1 + x2 e2 + x3 e3 . (1.1)
i=1 x3
Here and in the following, we use bold symbols to denote vector (and later matrix and tensor)
quantities, while scalar variables are never bold. In the above 3D case, x1 , x2 , and x3 are the three
3
independent degrees of freedom (DOFs) that describe the particle’s position. In general:
A particle in d dimensions has d degrees of freedom. (1.2)
As the position of the particle changes with time t, we are interested in the trajectory r(t), which
assigns to each time t a unique position r ∈ Rd . The motion of a single particle is hence described
by its position r(t), from which we derive its velocity and acceleration as, respectively,1
dr dv d2 r
v(t) = (t) = ṙ(t) ⇒ a(t) = (t) = (t) = v̇(t) = r̈(t) (1.3)
dt dt dt2
Dots always denote derivatives with respect to time. The speed of particle is v(t) = |v(t)|, and
likewise the scalar acceleration is a(t) = |a(t)|.
In most of our discussion throughout this course, we use a fixed Cartesian reference frame. We
may hence exploit the fact that the basis C = {e1 , e2 , e3 } is time-invariant (i.e., ėi = 0). Such a
fixed reference frame is called an inertial frame, and it allows us to write
d
X d
X d
X
r(t) = xi (t)ei ⇒ v(t) = ẋi (t)ei ⇒ a(t) = ẍi (t)ei (1.4)
i=1 i=1 i=1
It is important to keep in mind that this only holds as long as ei (t) = ei = const., since otherwise
we would require additional terms in the above time derivatives. Since the basis vectors form an
orthonormal basis, the ith component of vector r is
xi = r · ei for i = 1, 2, 3. (1.5)
For a particular basis C, the above vectors can also be expressed in their component form as
     
x1 (t) ẋ1 (t) ẍ1 (t)
[r(t)]C = x2 (t) ⇒ [v(t)]C = ẋ2 (t) ⇒ [a(t)]C = ẍ2 (t) . (1.6)
x3 (t) ẋ3 (t) ẍ3 (t)
As a notation convention, we write
[r]C (1.7)
to denote the components of a vector r in the Cartesian frame C. Since for the most part, we
only work with a single – fixed Cartesian – frame of reference, we may safely drop the subscript C
for convenience and simply write [r], [v], etc., whenever only a single, unique inertial coordinate
system is being used. This differentiation between frames will become important later in the course
when discussing moving reference frames.
Having established the above relations between position, velocity and acceleration, we are in place
to describe the motion of a particle through space. In most problems only partial information
is available (e.g., the position r(t) is known but the velocity and acceleration are not, or only the
acceleration is known as a function of velocity, i.e., a = a(v) or as a function of position, i.e.,
a = a(x)). In such cases, the mathematical relations derived in Examples 1.1 and 1.2 below are
helpful in finding the yet unknown kinematic quantities.
1
All boxed equations in these lecture notes become part or the formula collection (see, e.g., Section 1.3).
4
Example 1.1. Finding a particle trajectory from a time-dependent acceleration
If the acceleration a(t) of a particle is a known function of time (starting from an initial time t0 ),
then integration with respect to time gives
t v(t)
dv
Z Z
a(t) = ⇔ a(t) dt = dv ⇒ a(t) dt = dv, (1.8)
dt t0 v(t0 )
so that, with v(t0 ) being the initial velocity at time t0 ,

Z t
v(t) = a(t) dt + v(t0 ). (1.9)
t0
Integrating once again with respect to time (and starting from the initial position r(t0 )), we arrive
at
Z t
dr
v(t) = ⇔ v(t) dt = dr ⇒ r(t) = v(t) dt + r(t0 ). (1.10)
dt t0
As an example, consider gravity g accelerating a particle in the negative x2 -

direction. Releasing a particle at t = 0 from the initial position r(0) and with
initial velocity v(0) gives v(0)
a(t) = −ge2 = const. ⇒ v(t) = −gte2 + v(0) (1.11) x2 r(0)
x1 g
and
x3
gt2
r(t) = − e2 + v(0)t + r(0). (1.12)
2
Notice that, because the acceleration acts only in the x2 -direction, the velocity in the perpendicular
x1 - and x3 -directions remains constant throughout the flight (and hence remains the same as in
the initial instance).
————
Example 1.2. Finding a particle trajectory from a velocity- or position-dependent

acceleration (or from a position-dependent velocity)
In case of drag acting on a particle, the particle experiences a negative acceleration. For example,
viscous drag due to a surrounding medium at low speeds results in a negative acceleration linearly
proportional2 to the particle velocity, i.e., a = −kv (drag counteracts the particle motion, and the
resistance grows linearly with the particle speed through a viscosity constant k > 0). In such a
case, we know a = a(v), i.e., the acceleration as a function of velocity.
2
At low speeds (to be correct, at low Reynolds numbers) the viscous drag on a particle due to a surrounding
medium is typically linearly proportional to the particle speed. This is so-called Stokes’ drag, in 1D a = −kv. At
higher Reynolds number, Rayleigh drag results in a deceleration that is quadratically proportional to the particle
speed, in 1D a = −kv 2 . We here limit ourselves to the former case of linear drag, though the latter results in a similar
kinematic problem that can be solved in an analogous fashion.
5
For simplicity, consider the case of 1D motion, i.e., r = xe and v = ve, a = ae in a known
direction e. Here, we know a = a(v) = −kv. Separation of variables in this case lets us find the
velocity via
Z v(t) Z t
dv dv dv
a(v) = ⇔ = dt ⇒ = dt = t − t0 . (1.13)
dt a(v) v(t0 ) a(v) t0
For the specific case of linear viscous drag, i.e., a(v) = −kv, we thus obtain
1 v(t) dv
Z
− = t − t0 ⇒ v(t) = v(t0 ) exp [−k(t − t0 )] . (1.14)
k v(t0 ) v
The above can be generalized for 3D motion. Writing a = a(v) = −kv for each Cartesian compo-
nent (i = 1, 2, 3) becomes ai = −kvi . Separation of variables in this case lets us find each velocity
component via
Z vi (t)
dvi dvi
ai = ⇒ = t − t0 ⇒ vi (t) = vi (t0 ) exp [−k(t − t0 )] . (1.15)
dt vi (t0 ) −kv i
A similar strategy can be applied if the velocity is known in the form v = v(x). Again exploiting
the definition of v and using separation of variables, we find (in 1D) that
Z x(t) Z t
dx dx
v = v(x) = ⇒ = dt = t − t0 . (1.16)
dt x(t0 ) v(x) t0
Alternatively, if the acceleration is known as a function of position, i.e., a = a(x) (e.g., when a
rocket launch from earth is considered, where the gravitational acceleration is a function of the
rocket’s height above ground), then
Z v(x) Z x
dv dv dx dv
a = a(x) = = = v ⇒ v dv = a(x) dx. (1.17)
dt dx dt dx v(x0 ) x0
Integration yields v = v(x), so that (1.16) can again be applied to find x = x(t).
————
Example 1.3. Particle on a circular trajectory
As a simple closing example, consider a particle attached to a string of fixed length r > 0 and
rotating in the x1 -x2 -plane around the coordinate origin. The particle position is given by
r(t) = r cos ϕ(t)e1 + r sin ϕ(t)e2 (1.18)
with a known time-dependent angle ϕ(t). The particle velocity and speed follow as, respectively,
v = ṙ = r [− sin ϕ(t)e1 + cos ϕ(t)e2 ] ϕ̇(t) ⇒ v = |v| = rϕ̇(t), (1.19)
and its acceleration is
a = v̇ = −r [cos ϕ(t)e1 + sin ϕ(t)e2 ] ϕ̇2 (t) + r [− sin ϕ(t)e1 + cos ϕ(t)e2 ] ϕ̈(t). (1.20)
————
6
1.1.2 Constrained motion
An unconstrained particle moving freely in d dimensions has a total of d degrees of freedom (DOFs),
as it can translate in each of the d directions (while rotational DOFs are neglected). In many
scenarios, the motion of a particle is kinematically constrained (e.g., if the particle is moving on
a rigid ground, or if it is attached to a rigid rod or rope – in such cases, the particle can move but
its trajectory is restricted). Generally speaking:
In case of k kinematic constraints, a particle has (d − k) degrees of freedom. (1.21)
When dealing with constraints, it is beneficial to choose the frame of reference and the associated
coordinates wisely to minimize complications and efforts (see Example 1.4 below).
Although we will discuss this point in more detail later, we point out that any kinematic constraint
requires a constraint force (or reaction force) to enforce that the particle stays on the restricted
trajectory. For examples, see Example 1.4(c) and (d).
Example 1.4. Unconstrained and constrained motion
m x2=0 n m
m r
Äx3 j
x2 x1 O O r m
N
x1
(a,b) free motion in 2D (c) particle sliding on the ground (d) particle rotating around point O in 2D/3D
(a) A particle moving freely through space is unconstrained and therefore has d translational
DOFs in d dimensions: r = x1 e1 +. . .+xd ed (as pointed out before, particles have a negligible
size, so that no rotational DOFs are considered).
(b) A particle moving in 2D (which is a common assumption in many of the following exam-
ples) can be uniquely described by two translational DOFs only, e.g. r = x1 e1 + x2 e2 .
(c) A particle sliding on a flat ground can only move within the 2D plane of the ground
(which is possibly inclined). A single constraint (k = 1) in 3D spaces leads to the particle
having (3 − 1) = 2 DOFs. If the ground has a unit normal vector n (and the coordinate
origin O lies in the ground plane), then we must have r(t) · n = 0. (This scalar equation
implies that the particle indeed has only two independent DOFs.) In such cases, it is wise
to choose a coordinate system that is aligned with the ground, e.g., as shown above with
e2 ||n such that x2 = 0 and r = x1 e1 + x3 e3 . As an important consequence (to be discussed
later), a constraint or reaction force N from the ground onto the particle is needed to
keep the particle on the ground. Its magnitude is yet unknown, but we know that such a force
must exist to prevent the particle from falling into the ground. This will become important
when dealing with forces on particles, and we should keep in mind that any constraint always
implies the existence of (possibly hidden) reaction forces.
7
(d) A particle moving around a fixed point O on a rigid link is constrained in its motion,
since the distance to point O, |r(t) − rO |, must remain constant at all times. In practice, this
is realized by a rigid bar or rope which imposes the kinematic constraint (k = 1). To keep
the particle on its trajectory, a constraint or reaction force is needed, which is the force
in the rigid link.
In 3D, the particle has (3 − 1) = 2 DOFs, and it is moving on a sphere of constant radius
|r(t) − rO |. The particle’s position can be described, e.g., by two spherical angles (ϕ, θ).
In 2D, the particle has (2 − 1) = 1 DOF, so the particle’s motion is described by only a single
independent DOF. It is moving on a circle of constant radius |r(t) − rO |, described, e.g., by
the angle ϕ with the x1 -axis (as in Example 1.2).
If the rigid link was replaced by an elastic spring, the particle could move in all three
directions. The spring will generate a force, if it is stretched or compressed, but it does not
restrict the motion of the particle kinematically. The particular arrangement may still make
it convenient to use polar coordinates (r, ϕ) in 2D, or spherical coordinates (r, ϕ, θ) in 3D,
to describe the particle’s position in space; yet, this is a choice of convenience and does not
imply a kinematic constraint.
————
1.1.3 Kinematics in polar coordinates
In many problems, it will be convenient to not use Cartesian coordinates to describe the motion of
a particle. For example, for a particle rotating around a fixed pole, it may be more convenient to
use polar coordinates in 2D or spherical coordinates in 3D.
As an example, consider the rotation of a particle on a spring around a fixed m
hinge in 2D, as shown on the right. Rather than using {x1 (t), x2 (t)} as the two
independent DOFs to describe the particle position, we can alternatively use, r
e.g., polar coordinates {r(t), ϕ(t)}, where r denotes the distance from the
hinge and ϕ the angle with respect to the a fixed axis. After all, it is our choiceej
to introduce d reasonable, independent DOFs to describe the particle motion
erj
in d dimensions.
The polar description in terms of {r(t), ϕ(t)} comes with an orthonormal basis {er (t), eϕ (t)}, which
– unlike in a Cartesian frame of reference – is time-dependent. Such a moving frame is called a
non-inertial frame. We can still use the kinematic relations (1.3) between position, velocity and
acceleration vectors, but we must be careful when formulating those quantities in component form,
since we no longer have ėi = 0. Let us discuss the example of kinematics in polar coordinates in
detail in the following.
We describe the position of a particle in 2D by polar coordinates r(t) (the distance from the origin)
and ϕ(t) (the angle measured counter-clockwise from the x1 -axis), such that
r(t) = r(t) cos ϕ(t)e1 + r(t) sin ϕ(t)e2 . (1.22)
8
Such a description is convenient, e.g., when describing the motion of a particle attached a (massless)
stretchable spring whose length is r(t) and whose orientation is ϕ(t) (as shown on the right).
Noting that both r and ϕ depend on time, the velocity of the particle is
dr
v= = (ṙ cos ϕ − rϕ̇ sin ϕ)e1 + (ṙ sin ϕ + rϕ̇ cos ϕ)e2 , (1.23)
dt
where we dropped the explicit dependence of r and ϕ on t for conciseness.
Similarly, the acceleration follows as
a = r̈ cos ϕ − 2ṙϕ̇ sin ϕ − rϕ̇2 cos ϕ − rϕ̈ sin ϕ e1

(1.24)
+ r̈ sin ϕ + 2ṙϕ̇ cos ϕ − rϕ̇2 sin ϕ + rϕ̈ cos ϕ e2 .

Recall that all those vectors were expressed based on a Cartesian reference frame C, so that in that
frame we also write the vector components as

r cos ϕ ṙ cos ϕ − rϕ̇ sin ϕ
[r]C = , [v]C = , (1.25)
r sin ϕ ṙ sin ϕ + rϕ̇ cos ϕ
and
r̈ cos ϕ − 2ṙϕ̇ sin ϕ − rϕ̇2 cos ϕ − rϕ̈ sin ϕ

[a]C = . (1.26)
r̈ sin ϕ + 2ṙϕ̇ cos ϕ − rϕ̇2 sin ϕ + rϕ̈ cos ϕ
Alternatively, we may define a basis that rotates with the particle. To this end we introduce a
co-rotating frame R, whose basis vectors point in the radial (r) and perpendicular circular (ϕ)
directions at any time t. Specifically we define
er (t) = cos ϕ(t)e1 + sin ϕ(t)e2 , eϕ (t) = − sin ϕ(t)e1 + cos ϕ(t)e2 (1.27)
or, in component form with respect to the Cartesian frame C,

cos ϕ − sin ϕ
[er ]C = , [eϕ ]C = . (1.28)
sin ϕ cos ϕ
It is important to note that the new basis {er , eϕ } is time-dependent. Further, notice that the basis
is orthonormal since er · eϕ = 0 and |er | = |eϕ | = 1.
We may now express any vector x with respect to this co-rotating basis R:
x = xr er + xϕ eϕ with xr = x · er , and xϕ = x · eϕ . (1.29)
For example, the velocity vector v can be written as v = vr er + vϕ eϕ , vr vt

v
whose components are obtained by a projection of (1.23) onto the v2 v
basis vectors (1.27): e2 v1 er
ej j
er · v = vr = ṙ and eϕ · v = vϕ = rϕ̇. (1.30) e1
9
We here exploit that the co-rotating basis is orthonormal, i.e., er · eϕ = 0 and er · er = eϕ · eϕ = 1,

so that v · er = (vr er + vϕ eϕ ) · er = vr and v · eϕ = (vr er + vϕ eϕ ) · eϕ = vϕ .
The velocity vector in the rotating polar frame R hence becomes
v = ṙer + rϕ̇eϕ (1.31)
In words, the velocity vector at each instance of time has a component ṙ into the radial outward
er -direction and a component rϕ̇ in the tangential, circular eϕ -direction.
Analogously, the components of the acceleration vector a in the co-rotating R-frame are
er · a = ar = r̈ − rϕ̇2 and eϕ · a = aϕ = 2ṙϕ̇ + rϕ̈, (1.32)
so that
a = r̈ − rϕ̇2 er + (2ṙϕ̇ + rϕ̈) eϕ

(1.33)
We have thus identified the radial and tangential components of the velocity and acceleration
vectors. When formulating all vectors in the rotating, non-inertial frame R, the components of the
velocity and acceleration vectors are
r̈ − rϕ̇2

ṙ
[v]R = , [a]R = . (1.34)
rϕ̇ 2ṙϕ̇ + rϕ̈
It is important to realize that these components apply in a rotating, non-inertial frame of reference.
As a consequence, we generally have
ar 6= v̇r and aϕ 6= v̇ϕ , (1.35)
because these lack the additional terms that stem from differentiating the basis vectors with respect
to time. Therefore, keep in mind that the kinematic relations (1.3) apply to the position, velocity
and acceleration vectors – and not their components.
To introduce some further definitions, ω = ϕ̇ is usually referred to as the angular velocity [rad/s],
while ω̇ = ϕ̈ is known as the angular acceleration [rad/s2 ] of the particle.
We close by noting that the same theoretical treatment can be applied to cylindrical coordinates
and spherical coordinates in 3D, the details of which are skipped here (but can be found in dynamics
textbooks; see, e.g., this website for spherical coordinates).
Example 1.5. Particle rotating on a taut, inextensible string
Consider a particle that is attached to a taut, inextensible3 string of constant length l, which is
rotating about a fixed point O in 2D. The kinematics of the particle is best described using polar
coordinates. In this case we have r(t) = l = const. so that ṙ = 0 and r̈ = 0. Consequently,
v(t) = lϕ̇(t)eϕ (t), a(t) = −lϕ̇2 (t)er (t) + lϕ̈(t)eϕ (t), (1.36)
3
Taut means that the string always remains under tension (is not slack). For an inextensible string, i.e., one
which cannot stretch, this means that the string maintains a constant length.
10
i.e., at any instance of time the particle velocity is tangential to its circular motion, while the
acceleration contains both radial and tangential components. If the rotation is of constant angular
velocity, i.e., ω = ϕ̇ = const., then v = l|ω| and
v2
v(t) = lωeϕ (t), a(t) = −lω 2 er (t) = − er (t), (1.37)
l
i.e., the velocity is of constant magnitude and only changes its direction as the particle moves on
the circular trajectory (always being perpendicular to the string). The angular acceleration also
has a constant magnitude and points radially inward (accelerating the particle in a centripetal
fashion towards the center of rotation). It grows quadratically with angular velocity.
————
1.1.4 Kinematics of space curves
Having discussed Cartesian and polar coordinates, we here introduce a general description of a 3D
particle trajectory, which is sometimes more convenient and known as the description of a space
curve. As in the polar scenario discussed above, we do not use Cartesian coordinates to describe
the particle trajectory but we introduce a convenient non-inertial frame of reference.
To this end, we introduce the unique coordinate s(t) which measures the path length traveled
by the particle since time t = 0. For example, if a particle is travelling on a straight line in the
x1 -direction (starting at x1 (0) = 0), then s(t) = x1 (t). If the particle is traveling on a circular
trajectory of constant radius R, then s(t) = Rϕ(t) with ϕ(t) being the angle traveled since time
t = 0. For general trajectories, the description naturally becomes more complex.
The position of a particle is now described as
r = r(s) with s = s(t). (1.38) v

t
For convenience, let us introduce the unit tangent
vector along the particle trajectory as
dr s(t )
et = , (1.39) t =0
ds
which may be re-interpreted by introducing the speed
v = |v| = ṡ (1.40)
and noticing that

−1
dr dr dt ds dr v dr v
= = = ⇒ et = = (1.41)
ds dt ds dt dt v ds v
In other words, et (t) is always tangential to the trajectory of the particle and it points into the
direction of the instantaneous velocity v at any point on the trajectory (and it has unit length).
11
This allows us to write the velocity vector in general as
v = ṡet (1.42)
In order to arrive at the acceleration vector, we differentiate with respect to time, arriving at
a = s̈et + ṡėt . (1.43)
To find ėt , let us first realize that, since |et | = 1,
et · et = 1 ⇒ 2et · ėt = 0 ⇒ ėt ⊥ et , (1.44)
i.e., ėt must always be oriented perpendicular to et and hence to the particle trajectory. This
motivates the introduction of the principal normal vector en as the unit vector pointing into
the direction of ėt . In order to determine the magnitude of ėt , we inspect two times t and t + dt,
separate by an infinitesimal time increment dt, and compare their et -vectors, as shown schematically
below.
dj et(t+dt)
r
t+dt djen
et(t+dt) dj
ds
et(t)
t et(t)
From geometry, we conclude that (approximately)

et (t + dt) − et (t) dϕ en
ėt = = , (1.45)
dt dt
where we exploited that both et (t) and et (t + dt) are unit vectors of length 1, and the angle dϕ
is so small that et (t + dt) − et (t) is approximately the length of the arc of a circle of radius 1 and
angle dϕ. We then replace dϕ via the relation
ds = ρ dϕ, (1.46)
where ds is an infinitesimally small segment of the trajectory (approximated as a circular arc), and
ρ denotes the radius of curvature of the so-called osculating circle4 . It is related to the local
curvature κ via
κ = 1/ρ. (1.47)
4
The osculating circle of any point on a trajectory is the circle going through that point and having the same
curvature as the trajectory at that point. For a more intuitive interpretation, imagine you are driving a vehicle on a
curved path when suddenly the wheels lock in their current orientation. Unless that orientation is straight, the car
will leave the path and continue in a circle. That circle is the osculating circle and its radius the radius of curvature.
12
For example, for a trajectory in 2D described by x2 = f (x1 ), the radius of curvature is obtained
from
1 f 00 df d2 f
=p 3 with f0 = , f0 = . (1.48)
ρ 1 + (f 0 )2 dx1 dx21
For a circular motion with constant radius R, we obviously have ρ = R. For rectilinear motion on
a straight line ρ → ∞. If the particle travels on a quadratic parabola, then the curvature depends
on position following (1.48), e.g.,
p 3
1 + (2cx1 )2
x2 = f (x1 ) = c x21 ⇒ ρ= . (1.49)
2c
Insertion of (1.46) into (1.45) results in
et (t + dt) − et (t) dϕ en 1 ds en ṡ v
ėt = = = = en = en , (1.50)
dt dt ρ dt ρ ρ
which finally yields the acceleration vector in the space curve frame as
v2
a = s̈et + en (1.51)
ρ
As already seen for circular motion (cf. (1.37)), the second term represents the centripetal accel-
eration which accelerates any particle with non-zero speed and finite curvature radius towards the
center of the (instantaneous) osculating circle. Note that (1.51) lets us re-interpret the curva-
ture κ = 1/ρ as the magnitude of acceleration experienced by a particle traveling with unit speed
v = ṡ = 1 = const. along the space curve. Shown below are examples of osculating circles at various
points on a trajectory in 2D.
r r
r
r
To arrive at a 3D formulation, one may define the (unit) binormal vector
eb = et × en , (1.52)
so that {et , en , eb } forms an orthonormal triad for each point on the space curve. However, the
acceleration a has a vanishing component in this third direction, so (1.51) remains valid.
13
Example 1.6. Circular motion
For a particle moving on a circular path with constant radius R, the length of the path traveled
up to time t ≥ 0 (assuming that ϕ = 0 at time t = 0) is s(t) = Rϕ(t). Applying (1.42) leads to the
velocity
v = ṡet = Rϕ̇et , (1.53)
and comparison with (1.31) shows that et = eϕ (which makes sense: the eϕ -vector defined in (1.28)
always points into the tangential direction of the circular motion). Also, v = ṡ = R|ϕ̇| = R|ω|.
Next, applying (1.51) with ρ = R = const. yields

(Rϕ̇)2
a = Rϕ̈ et + en = Rϕ̈ et + Rϕ̇2 en . (1.54)
R
Comparison with (1.33) shows that en = −er , which could also have been expected (er was defined
radially outwards, while en is defined such that it points towards the center of curvature).
We note that for a general 2D motion

p expressed in polar coordinates (r, ϕ) with varying radius r(t)
and angle ϕ(t), we have ds = dr + r dϕ2 , so the above derivation becomes significantly more
2
complex (e.g., et and en will no longer align with eϕ and −er , respectively).
————
1.1.5 A brief summary of particle kinematics
The kinematics of a particle in d dimensions is described by d independent DOFs, the exact choice
of which can be made by convenience. In the presence of k kinematic constraints, that number
reduces to (d − k) DOFs. Irrespective of the choice of description, we always have the vector
relations
dr dv d2 r
v= , a= = , (1.55)
dt dt dt2
whereas their components depend on the chosen reference frame. For example, in
• Cartesian coordinates in d dimensions:

Xd Xd
v= ẋi ei , a= ẍi ei (1.56)
i=1 i=1
• polar coordinates in 2D:

a = r̈ − rϕ̇2 er + (2ṙϕ̇ + rϕ̈) eϕ

v = ṙer + rϕ̇eϕ , (1.57)
• space curve description:

v2
v = ṡet , a = s̈et + en (1.58)
ρ
Importantly, from the above three only the Cartesian description with a fixed origin is an inertial
frame, while the polar and space curve descriptions involve non-inertial reference frames.
14
1.2 Kinetics
So far, we have described the motion of a particle in terms of its position, velocity, and acceleration
(and we have shown how those are linked to each other). All those considerations were linked
purely to the geometry of the system and of the particle’s motion, without caring about what
caused the motion. All of that is known as the kinematics of a system (and it is quite analogous
to the kinematic description of deformation in Mechanics 2, where, e.g., the displacements and
strains were linked through kinematic relations to describe the deformation of bodies). In order
to link a particle’s motion to the causes of its motion – viz., any applied forces – we need to go a
step further and study the kinetics of a particle. The latter is summarized by the following three
famous axioms, which were first published by Sir Issac Newton in 1687.
1.2.1 Newton’s axioms and balance of linear momentum
Before presenting Newton’s axioms, let us introduce the linear momentum vector P of a particle
as the product of its velocity vector v and the particle mass m, i.e.,
P = mv (1.59)
A particle’s linear momentum always points in the direction of its velocity. With this definition,
we are in place to discuss Newton’s three axioms, which are schematically shown below. We point
out that all three axioms apply only in an inertial frame (i.e., in a non-moving frame – we will
get back to this point later in Section 3.4).
I II III
x (t) F
m
v = const. m F
F
F = ma R -R
I. Newton’s first axiom:

P
In the absence of any force acting on a particle (so the resultant force F = i Fi vanishes),
its linear momentum is conserved (i.e., remains constant) over time:
X d
F = Fi = 0 ⇒ (mv) = 0 ⇒ P = mv = const. (1.60)
dt
i
Note that, here and in the following, we often simply write F for the applied (resultant/net)
force5 acting on a particle (i.e., this is the sum of all individual forces being applied, and we
do not explicitly write out the sum over all applied forces Fi ).
5
In Mechanics 1, the symbol R was introduced for the resultant force acting on a body (i.e., the sum of all applied
forces). Here, we prefer to use F , since we will reserve R for frictional forces that appear in many examples (also,
the majority of textbooks uses F ).
15
II. Newton’s second axiom:

The rate of change of linear momentum equals the applied (net) force:
X d
Fi = Ṗ = (mv) (1.61)
dt
i
We will refer to this relation as linear momentum balance (abbreviated as LMB). It is

one of the key equations we will use frequently throughout the course. Simply put, to change
the linear momentum of a particle, it takes a force.
Let us consider three special cases. First, in case of a static problem (no particle motion, so
v = 0), Newton’s second axiom reduces to static equilibrium:
X
special case statics: v = 0 ⇒ Fi = 0 (1.62)
i
The case of equilibrium in statics is hence recovered as a special case. Second, consider a
particle of constant mass (i.e., ṁ = 0), in which case Newton’s second axiom becomes
X
special case constant mass: ṁ = 0 ⇒ Fi = ma (1.63)
i
To be correct, Newton – who did not consider particles with changing mass – formulated his
second axiom as F = ma, which is equivalent to the above if m = const. Third, in case of
no applied forces, we automatically recover the first axiom since
X d
special case zero forces: Fi = 0 ⇒ (mv) = 0 ⇒ P = mv = const. (1.64)
dt
i
III. Newton’s third axiom:

When a particle exerts a force on a second particle, the second particle simultaneously exerts
a force equal in magnitude and opposite in direction onto the first one:
Action = Reaction (1.65)
This axiom was already introduced in Mechanics 1 and is simply restated here. There is no
change needed in the context of dynamics.
It is important to note that, in the third example (III) in the above figure we have F 6= R in
general (unlike in statics) because, by Newton’s second axiom,
(
a = 0 (in statics),
ma = F − R ⇒ F − R = 0 only if (1.66)
m = 0 (for massless bodies).
The latter case will become important in some examples later: whenever the mass of a body is
negligibly small (e.g., the mass of a string compared to the mass of a heavy particle attached
to the string), then the net force on the string must vanish – as in statics.
16
Example 1.7. Sliding block on a frictionless surface

Consider a particle of mass m sliding on a frictionless horizontal
ground in the x1 -direction, such that the particle kinematics are sim-
ply described by
r(t) = x1 (t)e1 , v(t) = ẋ1 (t)e1 , a(t) = ẍ1 (t)e1 . (1.67)
The constraint x2 (t) = 0 entails a reaction force N = N e2 to keep

the particle on the constrained path, i.e., on the ground (without
such a force, gravity would accelerate the particle downwards so that
x2 (t) 6= 0). Since there is no friction, no horizontal force acts.
When it comes to the kinetics of the sliding particle, linear momentum balance hence reads
ma = mẍ1 e1 = (N − mg)e2 . (1.68)
Note that linear momentum balance can also be interpreted component-wise:

P
Pi Fi,1 = mẍ1 ⇒
0 = mẍ1
(1.69)
i Fi,2 = mẍ2 N − mg = 0
so that
ẍ1 = 0 ⇒ v1 = ẋ1 = const. and N = mg. (1.70)
In other words, since no forces act in the x1 -direction, linear momentum in the x1 -direction is
conserved and hence the velocity in the x1 -direction remains constant at all times:
mẋ1 = const. ⇒ v1 = const. (1.71)
————
Example 1.8. Sliding block on an inclined surface
A block of mass m is sliding down an inclined surface (under an angle

α). The block is subjected to gravity acting vertically down. The
sliding friction is characterized by a kinetic friction coefficient µ,
so that during sliding R = µN (where R and N denote the magni-
tudes of the friction and normal forces, respectively) with R oriented
opposite to the velocity of the sliding particle (as long as |v| =
6 0).
Here, it is wise to align the coordinate system with the inclined surface (as shown). In this inclined
frame of reference, the particle’s kinematics is simply described by its position r(t) = x1 (t)e1 , so
that its velocity is v(t) = ẋ1 (t)e1 and acceleration a(t) = ẍ1 (t)e1 .
When it comes to the kinetics, linear momentum balance reads

P
i Fi,1 = −R + mg sin α = mẍ1 ,
P (1.72)
i Fi,2 = N − mg cos α = 0,
17
so that
R = µN = µmg cos α ⇒ ẍ1 = g(sin α − µ cos α). (1.73)
Notice that only in the special case sin α − µ cos α = 0 – which is equivalent to µ = tan α – the
particle is sliding at a constant velocity ẋ1 = const. since ẍ1 = 0. If µ > tan α, then ẍ1 < 0 so that
the particle will decelerate and eventually come to rest. If µ < tan α, then ẍ1 > 0 so the particle
will accelerate downward and never come to rest.
Recall for comparison that within statics we have |R| < µ0 |N | with µ0 the coefficient of static
friction. In dynamics, sliding friction is characterized by the kinetic friction coefficient µ and
the relations
v
R = −µ|N | such that |R| = µ|N |, (1.74)
|v|
and R always points into the direction opposite to v. Note that µ is not a property of a material,
but it depends on the combination of the two materials in contact and we usually have µ ≤ µ0 .
Therefore, the force needed to overcome static friction (|R| = µ0 |N |) is typically larger than the
force required to maintain sliding friction (|R| = µ|N |).
————
Example 1.9. Projectile motion (throwing a particle)
Throwing a particle of mass m from the ground (x = 0 at time t = 0) with an initial velocity
v0 = v0 cos α e1 + v0 sin α e2 at time t = 0 (1.75)
results in a parabola of flight. Let us find the motion of the flying ball in 2D.
x2 v g
0 x2,max
a
x1
x1,max
For a free flight in 2D, the kinematics are given by
r(t) = x1 (t)e1 + x2 (t)e2 , (1.76)
and velocity and acceleration follow by differentiation with respect to time.
Kinetics is governed by the balance of linear momentum in the two coordinate directions, which
yields the accelerations whose integration with respect to time provides the velocity and position:
P
P i Fi,1 = 0 = mẍ1 ⇒
ẋ1 = c1
⇒
x1 = c1 t + c0
(1.77)
F
i i,2 = −mg = mẍ 2 ẋ2 = −gt + d1 x 2 = − g2 t2 + d1 t + d0
18
Using the initial conditions to find the constants of integration yields the balls’ motion as
g
x1 (t) = v0 t cos α, x2 (t) = − t2 + v0 t sin α. (1.78)
2
The particle hits the ground again at point x1,max = x1,max e1 when
g 2v0 sin α
x2 (t1,max ) = − t21,max + v0 sin α t1,max = 0 ⇔ t1,max = , (1.79)
2 g
so that
2v02 sin α cos α
x1,max = x1 (t1,max ) = . (1.80)
g
Note that the maximum distance is achieved when the ball is thrown under an angle of α = π/4
(45◦ ) when x1,max = v02 /g.
The maximum height during flight is reached when

v0 sin α
ẋ2 (t2,max ) = −gt2,max + v0 sin α = 0 ⇔ t2,max = , (1.81)
g
so that
v02 sin2 α
x2,max = x2 (t2,max ) = . (1.82)
2g
————
Example 1.10. Particle on a string
In Example 1.5 we showed that the acceleration of a particle following a circular trajectory of radius
l at constant angular velocity ω is given by
a = −lω 2 er , (1.83)
where er is the unit vector pointing radially outward.

As a simple example, consider a particle of mass m attached to a l
string of fixed length l, which remains taut at all times and rotates at
a constant angular velocity ω. In this scenario, if we neglect gravity,
w S
the only force acting on the particle is the force S in the string. Linear er
momentum balance in this case becomes m
S = ma = −mlω 2 er . (1.84)
Consequently, the force S in the string must always point radially inward (along the rope), and it
has a constant magnitude of |S| = mlω 2 .
————
19
1.2.2 Work–energy balance
For many systems it is convenient to invoke so-called energy principles rather than working with
the momentum balance laws. The main idea is to compare two states of the particle rather than
considering the entire trajectory of the particle in between those two states. Starting from the
balance of linear momentum, we integrate along a particle’s trajectory r(t) over time to arrive at
Z r2 Z r2 X
d X d
(mv) = Fi ⇒ (mv) · dr = Fi · dr. (1.85)
dt r1 dt r1
i i
dr
Note that with dr = dt dt = v dt and assuming m = const., this leads to
r2 t2 v(t2 )
d dv |v(t2 )|2 |v(t1 )|2
Z Z Z
(mv) · dr = m · v dt = m v · dv = m −m , (1.86)
r1 dt t1 dt v(t1 ) 2 2
where we identify m|v|2 /2 as the kinetic energy T of the particle. Therefore, the change in
kinetic energy when going from r1 to r2 equals the work W12 performed on the particle by external
forces Fi , which is the work–energy balance for a particle of constant mass:
t2 r2
m
Z X XZ
T (t2 ) − T (t1 ) = W12 , T (t) = |v(t)|2 , W12 = Fi · v dt = Fi · dr (1.87)
2 t1 r1
i i
Example 1.11. Frictional sliding

Consider a particle of mass m with initial velocity v0 slipping in
the x1 -direction over a flat surface with kinetic friction coefficient
g
m v0 v=0
µ. Besides gravity and normal forces acting on the particle in the m
x2 -direction (like in the frictionless Example 1.7), the particle here
experiences additionally a frictional force R = −µmge1 (as long x1(t1) x1(t2)
as v 6= 0), which opposes the 1D particle motion. When does the Dx
particle come to a stop?
x2
mg
v
This is an ideal example for using the work–energy balance, since we x1 R
are only interested in the initial state and final state of the particle. N
The kinetic energy at the start (initial velocity ẋ1 = v0 ) and end (final velocity ẋ1 = 0) is,
respectively,
1
T (t1 ) = mv02 , T (t2 ) = 0. (1.88)
2
Linear momentum balance in the x2 -direction gives N = mg so that R = −Re1 with R = µN =
µmg. Therefore, the work done on the particle between times t1 and t2 is
Z t2 Z t2 Z x1 (t2 )
W12 = [(N − mg)e2 − Re1 )] · ẋ1 e1 dt = −R ẋ1 dt = −µmg dx1 . (1.89)
t1 t1 x1 (t1 )
20
Note that N and mg do not perform any work on the particle, as those forces are acting perpen-
dicular to the motion of the particle.
Defining the travel distance until the particle comes to a halt as
∆x = x1 (t2 ) − x1 (t1 ), (1.90)
we finally obtain from the work–energy balance
1 v02
W12 = −µmg ∆x = T (t2 ) − T (t1 ) = 0 − mv02 ⇔ ∆x = . (1.91)
2 2µg
The complete initial kinetic energy of the particle is hence consumed by the frictional force over the
distance ∆x. (As a sanity check, in case of negligible friction, µ → 0, the travel distance correctly
approaches ∆x → ∞, as linear momentum is conserved in the absence of a friction force, so the
particle maintains a constant velocity and never comes to rest).
————
The above sliding friction example is a non-conservative system. To understand this, let us
repeat the above example by first sliding on a frictional surface from r1 to r2 , and then pushing
the particle back with the same initial velocity so it moves back from r2 to r1 (over a distance
∆x = |r2 − r1 |). The work done on the entire path (noting that R = −µmge1 when going from 1
to 2, and R = µmge1 on the return path, as shown in the schematic below) is
Z r2 Z r1 Z x1 (t2 ) Z x1 (t1 )
W12 + W21 = R · dr + R · dr = −µmge1 · dx1 e1 + µmge1 · dx1 e1
r1 r2 x1 (t1 ) x1 (t2 )
Z x1 (t2 ) Z x1 (t1 )
= −µmg dx1 + µmg dx1 = −µmg ∆x + µmg(−∆x) = −2µmg ∆x < 0,
x1 (t1 ) x1 (t2 )
i.e., the particle is losing kinetic energy due to friction (like in Example 1.11). Recall from the
previous example that normal and gravitational forces do not perform work in this scenario, so
they are omitted here in the work calculation.
mg mg
v g v
R N x2 N R
m r1 r2 x1 m r1 r2
m m
Dx Dx
By contrast, consider now a particle subjected to gravity, being moved from r1 up to r2 and back
down to r1 (without frictional losses). For example, we kick the particle up a hill with some initial
velocity and, once it has reached its maximum height, we let it slide down the hill to the point from
where we started – all without friction. It is important to notice that the normal force N from
21
the ground does not perform any work on the particle since N ⊥ dr (the same applies, e.g., if the
particle is attached to a massless string, where the force in the string is always perpendicular to
the particle motion; see the schematic below). In the absence of friction, the only force performing
work on the particle is hence that due to gravity. The total work performed on the particle hence
evaluates to
Z r2 Z r1 Z r2 Z r1
W12 + W21 = F · dr + F · dr = −mge2 · dr + −mge2 · dr
r1 r2 r1 r2
Z h2 Z h1 Z h2 Z h2 (1.92)
= −mg dx2 − mg dx2 = −mg dx2 + mg dx2 = 0,
h1 h2 h1 h1
where we defined h1 = r1 · e2 and h2 = r2 · e2 as the altitudes (x2 -coordinates) of points r1 and

r2 , respectively. Overall, the particle apparently conserves its kinetic energy during the cycle; we
conclude that gravity is a conservative force.
r2 F
g g F
mg v j v
massless
r1 h2 mg
N x2
h1 x1 m F
The latter example defines a conservative system: the energy of a particle is constant when
returning to its initial point along an arbitrary closed path. In mathematical terms, the work done
by a conservative force F along any closed path Γ (starting at some point r1 and ending at the
same point) vanishes, i.e.,
I
WΓ = F · dr = 0. (1.93)
Γ
This implies that the path itself is irrelevant to the work done by a conservative force F , only the
initial and end points matter. This in turn makes us conclude that we should not even need to
carry out the integration along a (non-closed) path from r1 to r2 in order to calculate the work
done by force F . Instead, there must be a way to evaluate the work by only considering the two
end points (since the path is irrelevant anyways). This is indeed the case for conservative forces,
since they can be derived from a potential energy V (r) such that
dV
F =− (1.94)
dr
We note that vector derivatives are carried out component-wise, i.e.,

 
d − dV / dx1
dV X dV
V = V (r) ⇒ F =− =− ei ⇔ [F ] = − dV / dx2  . (1.95)
dr dxi
i=1 − dV / dx3
22
Examples of a conservative force are
• Gravity: for a gravitational acceleration g, the potential energy is
V (r) = −mr · g and F = − dV / dr = mg. (1.96)
Specifically, for g = −ge3 this gives V (r) = mgx3 and F = −mge3 .

• Elastic spring: for an elastic spring stretched in the xi -direction with unstretched spring
position x0 and stiffness k, the potential energy reads
1
V (r) = k(xi − x0 )2 and F = − dV / dr = −k(xi − x0 )ei (1.97)
2
Since we know that for any conservative force F there exists a potential energy V (r) which is
unique for each point r, the work W12 performed by the conservative force between points r1 and
r2 now becomes
Z r2 Z r2 Z V (r2 )
dV
W12 = F · dr = − · dr = − dV = V (r1 ) − V (r2 ). (1.98)
r1 r1 dr V (r1 )
In other words, the work along any path between two points is given by the (negative) potential
energy difference:
W12 = V (r1 ) − V (r2 ) (1.99)
This is a convenient shortcut to computing the work W12 for any conservative forces. If mixed
forces act, some of which being conservative and some non-conservative, then we can compute the
work done by the conservative forces in the above fashion, whereas the work done by any non-
conservative force is generally computed by integration (recall that the work W12 in (1.87) was
additive in case of several forces Fi ).
In the special case of a conservative system, i.e., one where only conservative forces act, we know
that the total work is given by
W12 = V (r1 ) − V (r2 ), (1.100)
with V being the total potential energy. Substituting the work–energy balance (1.87) for W12
immediately leads to
T (t2 ) − T (t1 ) = V (r1 ) − V (r2 ). (1.101)
This indicates that conservative systems obey the conservation of energy (i.e., the total energy
including kinetic and potential contributions is conserved along any continuous path):
T (t2 ) + V (r2 ) = T (t1 ) + V (r1 ) ⇔ T + V = const. (1.102)
Summary:
Let us quickly recap. The work–energy balance (1.87) holds for both conservative and non-
conservative systems (the work done by external forces must always balance the change in kinetic
23
energy of a particle). For the special case of conservative forces, the work W12 can conveniently
be calculated as a potential difference that is path-independent, cf. (1.99). Finally, if all externally
applied forces are conservative, then we are dealing with a conservative system and we can take
a shortcut by exploiting the conservation of energy, cf. (1.102). For a non-conservative force, the
work W12 is obtained by integrating along the path, cf. (1.87).
Examples of a non-conservative force are
• Friction: in case of sliding friction with a kinetic friction coefficient µ > 0, the frictional
force acting on a particle is
v
R = −µ|N | for |v| =
6 0. (1.103)
|v|
As discussed above, the direction of the friction force R depends on the direction of the
particle velocity, which is why it cannot derive from a potential V (r) that depends only on
position (the path matters as well) – hence, the friction is not conservative.
• Viscous drag: viscous drag results, e.g., from a particle moving through a fluid. At low
speeds (low Reyonlds number), the resulting drag force on the particle is linearly proportional
to the particle velocity:
Fd = −kv (1.104)
with a viscosity k > 0. Obviously, this force depends on v and hence on the particle’s
trajectory, which is why the drag force cannot derive from a potential V (r) and is therefore
non-conservative.
Example 1.12. Maximum height during flight parabola (revisited)
Let us reconsider a particle of mass m, thrown with an initial velocity v0 under an angle α and
subjected to gravity. In Example 1.9 we solved for the trajectory of the particle by applying linear
momentum balance and integrating the kinematic equations. Here, we revisit the problem, using
the work–energy balance.
The starting point is characterized by kinetic energy T1 = T (t1 ) and zero potential energy6 , whereas
the point of maximum altitude x2,max is the point where the vertical velocity component of the
particle is zero (at this point the particle is no longer moving upward and will in the next instance
start to move downward, so the instantaneous vertical velocity is zero). Since no forces act in the
horizontal direction, the horizontal velocity component is constant during flight and hence agrees
with the initial horizontal velocity v = ẋ1 = v0 cos α. Therefore, the kinetic and potential energies
of initiation (state 1) and at the maximum height (state 2) are, respectively,
1 1
T1 = mv02 , V1 = 0, T2 = m(v0 cos α)2 , V2 = mgx2,max . (1.105)
2 2
6
The work–energy balance considers only differences between states, which is why the absolute altitude of the
particle is irrelevant and only the distance in altitude matters. For that reason, we can choose the zero altitude level
(the coordinate origin) arbitrarily and here pick the initial point.
24
Conservation of energy for this conservative systems yields directly

1 1 v02 v 2 sin2 α
mv02 + 0 = mv02 cos2 α + mgx2,max ⇔ x2,max = (1 − cos2 α) = 0 . (1.106)
2 2 2g 2g
As a variation, assume that the particle is accelerated vertically up-

wards from rest by being released from a compressed spring
x2
(stiffness k). With the spring energy V (∆x) = k2 (∆x)2 for a com- Dx
pression by ∆x, the maximum height during flight becomes k g
k
T1 = 0, V1 = (∆x)2 , T2 = 0, V2 = mgx2,max (1.107)
2
so that the conservation of energy yields the maximum height as
k k(∆x)2
(∆x)2 = mgx2,max ⇔ x2,max = . (1.108)
2 2gm
We thus obtain the maximum height quickly and without having to integrate the kinematic relations
as in Example 1.9. Note that the same solution applies to frictionless sliding up on a foundation
of arbitrary inclination.
————
Example 1.13. Car in a looping
We consider a car of mass m which enters a circular looping of g

radius R with an initial horizontal velocity v0 . We assume that the
R
car engine is off (i.e., the foot is off the gas pedal) and friction is
negligible. Further, we assume the car is small compared to R, so it j
may be approximated as a particle. v0
What is the minimal entry velocity v0 required for the car to not fall
off? What is the maximal force acting on the driver?
mg
Rcosj
Let us first use the work–energy balance, which here in the absence j et
of friction reduces to the conservation of energy, in order to find the
minimum velocity v0 such that the car reaches the top of the looping.
er N
We know the speed v0 at entry (at ϕ = 0, which we define as the altitude 0 so that the potential
energy vanishes). Since the only other force that does work on the car is gravity, we can use the
conservation of energy to determine the car’s speed v(ϕ) under any other angle ϕ ∈ [0, 2π) within
the looping by equating the total energy:
1 1
mv02 = mv 2 (ϕ) + mgR(1 − cos ϕ), (1.109)
2 2
which we solve for the speed
q
v(ϕ) = v02 − 2gR(1 − cos ϕ). (1.110)
25
As can be expected, the velocity is minimal when cos ϕ = −1, i.e., at ϕ = π, where
q
vmin = v(π) = v02 − 4gR. (1.111)
Apparently, a necessary condition for the car to achieve the looping is hence
p
v02 − 4gR > 0 or v0 > 2 gR. (1.112)
If this condition is not satisfied, the car will never reach the top and simply slide down backwards
after having reached its maximum altitude.
In order to determine under what conditions the car does not fall off, we need to go one step further.
During the looping, while there is no friction force, there is a normal force N (see the schematic
above) which keeps the car on the ground (this is the reaction force due to the constrained motion
on the circular looping). At any location in the looping, we can calculate N from the system
kinetics. Falling off implies that the reaction force N becomes negative – a solution with N < 0
would imply that the reaction force no longer pushes the car off the ground, but the reaction force
N would be required to pull the car towards the ground (to prevent it from falling off). Since the
car is freely sliding on the ground, there is no such attractive force and the car would indeed fall
off. Therefore, the sufficient condition for the car to not fall off is
N (ϕ) ≥ 0 ∀ ϕ ∈ [0, 2π). (1.113)
Reaction force N is most easily obtained from the balance of linear momentum in polar coordinates;
the radius r = R = const., only ϕ is a free coordinate. As derived earlier for a circular path (see
Section 1.1.3), we know that in polar coordinates
v = vt et = Rϕ̇et ⇒ v(ϕ) = Rϕ̇(ϕ). (1.114)
Moreover, from a free-body diagram we obtain the balance of linear momentum in the radial
direction with ar = −Rϕ̇2 = −(Rϕ̇)2 /R = −v 2 /R as
v 2 (ϕ)
mar = −N (ϕ) + mg cos ϕ ⇒ −m = −N (ϕ) + mg cos ϕ (1.115)
R
so that, inserting (1.110) for v(ϕ),
mv 2 (ϕ) mv02 mv02

N (ϕ) = +mg cos ϕ = −2mg(1−cos ϕ)+mg cos ϕ = +mg(3 cos ϕ−2). (1.116)
R R R
Preventing lift-off means
mv02
N (ϕ) ≥ 0 ∀ ϕ ∈ (0, 2π] ⇔ + mg(3 cos ϕ − 2) ≥ 0 ∀ ϕ ∈ (0, 2π]. (1.117)
R
The worst-case scenario is reached when cos ϕ = −1 or ϕ = π (at the maximum height, as could
have been expected), from which follows the sufficient condition to prevent lift-off :
mv02 p
+ mg(3 cos π − 2) ≥ 0 ⇔ v0 ≥ 5gR. (1.118)
R
26
√
Note that this condition is stricter than the necessary condition (1.112), hence v0 ≥ 5gR ensures
both that the car reaches the top and does not fall off.
The maximum N -force on the car and hence on the driver is attained 1
v/v0
when cos ϕ = 1 or ϕ = 2π, i.e., when the car has completed the
looping. In that case,√using as an example the minimum required
entry velocity of v0 = 5gR, we obtain from (1.116)
N/Nmax
m5gR
N (2π) = + mg(3 cos(2π) − 2) = 6mg. (1.119)
R 0 p 2p
That is, the driver experiences an acceleration of 6g upon completion of the looping. As a rule of
thumb, an average person loses consciousness at about 5g. To avoid such high accelerations, typical
roller coasters never have circular loopings. Recall that we showed (see Section 1.1.4) the relation
for the particle acceleration
v2
a = ṡet + en (1.120)
ρ
for a general trajectory through space (not necessar-

ily following a circular motion). If one wants the cen- 1.2
Cornu spiral
tripetal acceleration, the second term in (1.120), to not (a=n=1)
grow significantly, one must choose a varying radius of 1.0
looping
curvature ρ. For example, noting that v 2 decreases lin-
early with altitude across the looping (by the conser- 0.8
vation of energy), we may choose ρ to decrease linearly

0.6
from the entry point of the looping until its top (and
then ascend in a symmetric fashion). The result is a so-
0.4
called Euler spiral (also known as the Cornu spiral
or clothoid) whose curvature increases linearly with
0.2
path length. This design is typically used for roller
coasters in amusement parks as well as for the transi-
tion from straight to curved segments on highways. 0.2 0.4 0.6 0.8 1.0 1.2 1.4
A possible parametrization (only mentioned without proof here) is

t t
τ n+1 τ n+1
Z Z
r(t) = a sin dτ e1 + a cos dτ e1 , (1.121)
0 n+1 0 n+1
resulting in a trajectory of curvature κ = tn /a; so the choice n = 1 results in a linearly increasing

curvature.
Finally, note that we can also evaluate the support reactions at the base of a looping, assuming
that it is clamped to its foundation. The looping itself is not moving at all (only the car in contact
with the looping is). Therefore, we have a static system where the classical equilibrium relations
apply with a moving normal force N (ϕ).
27
Drawing a free-body diagram of the (circular) looping without the car, which
is assumed under an angle ϕ, and applying the static balance of forces and
torques yields:
R
mv02

A1 = −N sin ϕ = − + mg(3 cos ϕ − 2) sin ϕ,
R j N
mv02

A2 = N cos ϕ = + mg(3 cos ϕ − 2) cos ϕ, (1.122)
R
MA A1
mv02 A2

MA = N R sin ϕ = + mg(3 cos ϕ − 2) R sin ϕ.
R
By comparison, if we had only placed the particle inside the looping and fixed it at an angle ϕ
under static conditions (so only −mge3 is acting on the foundation), the support reactions would
have been
A1 = 0, A2 = mg, MA = mgR sin ϕ. (1.123)
The moving car results is significantly altered support reactions, which must be considered when
desining the support of the looping.
————
Example 1.14. Elastic slingshot
Consider a slingshot composed of a linear spring (stiffness k, unde-
formed length l0 ) which is stretched by moving a particle of mass m Dx
by ∆x to the left on a frictionless ground, as shown on the right. If m r1 m=0 r2
the particle is released in that position from rest, what is its speed v v
when it reaches the unstretched spring length at position r2 ?
l0
The problem can best be solved by exploiting the conservation of
energy in this conservative system (no friction forces, gravity is per-
pendicular to the motion of the particle, so only the spring force must
be considered, which is conservative).
Comparing the two states’ kinetic and potential energies,
q 2
k m
T (t1 ) = 0, V (r1 ) = 2 2
(∆x) + l0 − l0 , T (t2 ) = v 2 , V (r2 ) = 0, (1.124)
2 2
leads to
r q
k 2 2
T (t1 ) + V (r1 ) = T (t2 ) + V (r2 ) ⇒ v= (∆x) + l0 − l0 . (1.125)
m
We note that the same result could have been obtained from application of the balance of linear
momentum. However, that would have been significantly more complex since the spring force
varies nonlinearly with position (and so does the acceleration on the mass to be integrated for the
velocity). The work–energy principle is therefore a convenient shortcut.
————
28
Example 1.15. Frictional sliding up a slope

Consider a particle of mass m, which is sliding up an inclined slope
(angle ϕ), starting with an initial speed v0 and experiencing friction
with the ground. What distance ∆x does the particle travel up the g
slope?
m v0
Since we again know the initial state (speed v0 , potential energy
chosen to vanish) and only care about the final state (speed 0, al- m Dx
titude h = ∆x sin α), this example can beneficially be solved by the a
work–energy balance. However, we encounter both conservative forces
(gravity) and non-conservative forces (friction).
In this case the work–energy balance states that:
1 gravity
0 − mv02 = W12
friction
+ W12 (1.126)
2
with
Z ∆x
friction gravity
W12 = −R dx = −µmg cos α ∆x, W12 = V1 − V2 = 0 − mgh. (1.127)
0
Altogether, with h = ∆x sin α we arrive at

1 v02
mv 2 = mg∆x sin α + µmg cos α ∆x ⇒ ∆x = . (1.128)
2 0 2g(sin α + µ cos α)
Note that for the special case α = 0 we recover the solution ∆x = v02 /2µg for horizontal sliding
from Example 1.11, whereas for α = 90◦ we recover the solution of the maximum height of a free
vertical flight (without normal force and hence without friction) previously derived in Example 1.9
as ∆x = v02 /2g.
————
1.2.3 Balance of angular momentum
With the balance of linear momentum, we derived in Section 1.2.1 the dynamic version of force
equilibrium in statics. We saw that, under dynamic conditions, the sum of all forces is no longer
necessarily zero, and that a non-zero resultant force leads to an acceleration of the particle. We
further saw that the proportionality between applied force and particle acceleration is the particle’s
mass. Following Newton’s axioms, in order to change the motion of a particle, we must apply a
force due to the particle’s inertia which aims to maintain a constant velocity. The heavier the
particle, the larger the required force to impose a certain acceleration of the particle.
We next seek a principle that replaces the equilibrium of all torques 7 in statics, which will lead us
to the balance of angular momentum.
7
Here and in the following, we use the term “torque” rather than “moment” (in German both correspond to
“Moment”) to avoid confusion with “momentum” (“Impuls” in German).
29
Let us first introduce the angular momentum HB of a particle P

x3
with respect to an arbitrary point B as the vector F (t)
P
HB = rBP × P with linear momentum P = mvP , (1.129)
rP P = mvP
rBP
where we denote the particle velocity by vP for clarity. O x2
To interpret the angular momentum vector introduced above, we rec- rB vB

x
ognize that, by definition, HB is perpendicular to the plane spanned 1 B
by rBP and the particle velocity vP .
Further, taking the norm of the above definition shows that

|HB | = m|rBP × vP | = mvP dBP , where dBP is the distance between P dBP
vP HB
the linear momentum vector (and hence vP ) through particle P and
point B (as sketched on the right). rBP
B
Inserting rBP = rP −rB and therefore
P ṙBP = vP −vB and using linear
momentum balance (i.e., Ṗ = i Fi ), we obtain
d X
ḢB = (rBP × P ) = ṙBP × P + rBP × Ṗ = (vP − vB ) × P + rBP × Fi . (1.130)
dt
i
We next introduce the resultant torque (i.e., the sum of all torques) with respect to B as
X
MB = rBP × Fi . (1.131)
i
We insert (1.131) into (1.130) and exploit that vP × P = vP × mvP = 0, so that we finally arrive
at the angular momentum balance (which we abbreviate as AMB):
MB = ḢB + vB × P (1.132)
This equation is the analog to the vanishing sum of all torques in the static equilibrium case.
Let us consider four special cases:

• First, we automatically recover the static case: without motion, we have ḢB = 0 and P = 0,
so that (1.132) reduces to
MB = 0, (1.133)
so the sum of all torques must vanish for static equilibrium (as we know from Mechanics 1).
• An often used special case occurs when point B is fixed or if B and P move in parallel:
if vB = 0 ∧ vB ||vP ⇒ vB × P = 0 ⇒ MB = ḢB . (1.134)
This is analogous to linear momentum balance: an applied external torque results in a change
of the particle’s angular momentum (like an applied force results in a change of its linear
momentum). As a general guideline, we should aim for this special case whenever possible by
choosing B wisely (viz. either as a fixed point or one that is moving parallel to the particle)
to simplify the relations.
30
• If no external torque acts on the system (MB = 0) with a fixed point B (or parallel
motion of P and B), then we arrive at
ḢB = 0 ⇒ HB = rBP × mvP = const. ⇒ |rBP | |v| = rBP vP = const. (1.135)
Therefore, the product of distance rBP and particle speed remains constant over time.
This is known as the conservation of angular momentum (analogous to the conservation
of linear momentum by Newton’s first axiom).
• Finally, for a massless body, m = 0 leads to P = 0 so that HB = 0 = const. and ḢB = 0,

so
X
if m = 0 ⇒ Fi = 0, MB = 0, (1.136)
i
where we included both LMB and AMB for a comparison. That is, for massless bodies we
recover the static equilibrium relations of a vanishing net force and torque.
Special case 2D:
Consider a particle of mass m rotating around a fixed point B in a 2D plane with instantaneous
angle ϕ(t) and distance r(t).
Using co-rotating polar coordinates, the above 3D
formulation here reduces to
m
ej e3 r
P
HB = rBP × P = rer × (mvP )
= rer × m(ṙer + rϕ̇eϕ ) B er vP
j
= mr2 ϕ̇(er × eϕ ) = mr2 ϕ̇e3 ,
where e3 denotes the out-of-plane unit vector.

Applying the balance of angular momentum (1.132) with an applied in-plane torque MB = MB e3
leads to
d
mr2 ϕ̇e3 = m(r2 ϕ̈ + 2rṙϕ̇)e3 = mr(rϕ̈ + 2ṙϕ̇)e3 .

MB = MB e3 = (1.137)
dt
Therefore, we arrive at
MB = MB e3 ⇒ MB = mr(rϕ̈ + 2ṙϕ̇) = maϕ r. (1.138)
In many subsequent examples, we consider a rotation at a fixed distance r(t) = R = const., e.g.,
when the particle is attached to a rotating disk or to a rigid string or stick. In such cases the above
relations simplify since ṙ = 0, so that
if r(t) = R = const. ⇒ MB = IB ϕ̈ with IB = mR2 (1.139)
We call the constant IB the moment of inertia (of a particle of mass m with respect to a point
B in 2D, rotating at a distance R).
31
Angular vs. linear momentum balance:
The above 2D formulation lets us reinterpret the balance of angular momentum (AMB) in com-
parison to the balance of linear momentum (LMB).
Recall that LMB set into relation the force applied to a
particle and the resulting acceleration of the particle via F
F = ma, with the mass m acting as the proportionality
constant (affecting the inertia of the particle as a resistance
L
against changes of its velocity). vP m
Let us rewrite (1.139) as
j(t) B R Fj
MB = IB ϕ̈ = mR2 ϕ̈. (1.140)
Let us further assume that a force F is applied at a distance

massless
L perpendicular to the massless support frame to introduce support
the external torque MB , so MB = F L.
The tangential force component Fϕ accelerating (or decelerating) the particle’s circular motion can
be obtained from AMB on the massless support frame, for which the equilibrium of torques yields
F L = Fϕ R. Overall, we obtain
d dvϕ
F L = Fϕ R = mR2 ϕ̈ ⇔ Fϕ = mRϕ̈ = m (Rϕ̇) = m = maϕ , (1.141)
dt dt
where vϕ and aϕ denote the tangential velocity and acceleration of the particle, respectively (and
dvϕ / dt = aϕ since r = const.). We see that AMB is indeed nothing else but an alternative state-
ment of LMB (in fact, we did derive AMB from LMB). The force on the particle is related to
its acceleration via its mass. Here, we convert the force into a torque and the constant of pro-
portionality between torque and acceleration must then involve the lever arm. Finally, converting
acceleration into angular acceleration – the moment of inertia IB is nothing else but the resultant
constant of proportionality between the torque and the angular acceleration (when rotating about
a fixed point).
Example 1.16. Rotating particle in air

Consider a particle of mass m that is rotating at constant distance R around
a fixed point. The surrounding air results in a viscous drag force opposing the
Fdrag m
particle’s motion (assumed linearly proportional to the particle’s velocity with .
viscosity k > 0). If gravity is acting perpendicular to the plane of rotation (as R v =Rj
shown), then the only force acting on the particle is the drag force j
Fd = −kv ⇒ MB = rBP × Fd = −Rkve3 . (1.142)
g
For a particle rotating at constant radius R, we know that v = vt = Rϕ̇ (see Section 1.1.3), so that
MB = −kR2 ϕ̇e3 . (1.143)
32
Finally, applying the balance of angular momentum for a 2D rotation at constant radius R yields
k
MB = mR2 ϕ̈ ⇒ ϕ̈ +
ϕ̇ = 0. (1.144)
m
This last form is the so-called equation of motion of the particle, a differential equation to be
solved for its time-dependent degree of freedom ϕ(t), given appropriate initial conditions.
————
Example 1.17. Massless pendulum with an attached mass
We consider a particle of mass m attached to a rigid mass- F
less string of length R subject to gravity. Here, angular g F
momentum balance is preferable since it avoids having to
B
j massless
introduce and calculate the force F in the string. Using mg
angular momentum balance around B avoids the force in
the string and yields
m
-F
g
mR2 ϕ̈ = −mgR sin ϕ. ⇔ ϕ̈ +
sin ϕ = 0. (1.145)
R
For small amplitudes (|ϕ| 1), we may use a Taylor expansion about ϕ = 0, viz.
ϕ2
sin(0 + ϕ) = sin 0 + ϕ cos 0 − sin 0 + O(|ϕ|3 ) = ϕ + O(|ϕ|3 ) (1.146)
2
to simplify the equation above of motion into a linear ordinary differential equation (ODE):
g
ϕ̈ + ϕ ≈ 0. (1.147)
R
It is important to keep in mind that this last step is valid only for small angles |ϕ| 1.
Note that the same equation of motion, of course, could also have been obtained from linear
momentum balance in the tangential direction:
maϕ = −mg sin ϕ and aϕ = Rϕ̈ ⇒ mRϕ̈ = −mg sin ϕ, (1.148)
which again leads to (1.145).
Finally, if the pendulum experiences drag due to the surrounding medium, then we may introduce
the same drag force as in Example 1.16, leading to the equation of motion of a particle pendulum
subjected to gravity and air drag:
k g k g
ϕ̈ + ϕ̇ + sin ϕ = 0 ⇒ ϕ̈ + ϕ̇ + ϕ ≈ 0. (1.149)
m R m R
This is a good example of the assumption of massless bodies (such as the string here). As discussed
before, for a massless body all inertial terms are neglected, so that one recovers the static equilibrium
equations
X
if a particle or body is massless ⇒ Fi = 0, MB = 0. (1.150)
i
————
33
Example 1.18. Circling mass on a string pulled inward in 2D

A particle of mass m is circling around a hole, attached
to a taught string of changing but known length l(t). The frictionless plane
string is held steady at a length l(0) until time t = 0. It is
F(t) m
then pulled inward. The force needed to pull the particle
B l(t) v(t)
inward is unknown (it involves, among others, frictional
forces in the hole). Known are the initial length l(0),
the initial particle speed v(0), and the changing length
l(t). Assuming that the particle is sliding on a frictionless ~
F(t)
ground, what is the particle speed v(t)?
Using the balance of angular momentum around the hole (point B) is ideal here, since there are
no applied torques to consider in this case (neglecting frictional forces), i.e., MB = 0. Since also
vB = 0, we have
ḢB + vB × P = MB ⇒ ḢB = 0 ⇒ HB = const. (1.151)
i.e., angular momentum is conserved. Inserting the definition of the angular momentum vector
leads to
HB = rBP × P = rBP × mvP ⇒ rBP (t) × mvP (t) = const. (1.152)
Using polar coordinates and rBP (t) = l(t)er (t), we realize that
rBP × vP = l(t)er (t) × [vr (t)er (t) + vϕ (t)eϕ (t)] = l(t)vϕ (t)er (t) × eϕ (t) = l(t)vϕ (t)e3 (1.153)
at all times t (with e3 being the out-of-plane unit vector). Therefore, the conservation of angular
momentum, assuming a constant mass m, yields
l(0)
rBP (0) × v(0) = rBP (t) × v(t) ⇒ l(0) vt (0) = l(t) vt (t) ⇔ vt (t) = vt (0). (1.154)
l(t)
Finally, exploiting that vt (t) = l(t)ω(t), the angular velocity ω(t) = ϕ̇(t) scales as
2
l(0)
ω(t) = ω(0) (1.155)
l(t)
Note that in this example energy is not necessarily conserved and the work done by the pulling
force is unknown, so the work–energy balance does not help in this scenario.
————
34
Example 1.19. Heavy mass held on a rotating chair

w
The conservation of angular momentum is most easily visualized on
a spinning chair. (Try this out at home, if you have a rotating desk m
chair!) Let us consider an approximately massless person sitting on a
rotating chair that is spinning at a constant angular velocity ω0 (with-
out any frictional losses or air drag, we may assume that no torques
act on the person, so they would spin at ω0 forever by angular mo- r
mentum conservation – analogous to a particle forever continuing at
constant velocity if no forces act by linear momentum conservation).
Let the person hold in their hands a heavy object, approximated as
a particle of mass m, at a distance r0 from the axis of rotation. How
does the angular velocity change, if the distance is changed from r0
to r?
Exploiting the conservation of angular momentum (in the absence of any torques that could accel-
erate or decelerate the rotation), we know from Example 1.18 that
r 2
0
r02 ω0 = r2 ω ⇒ ω= ω0 . (1.156)
r
Therefore, bringing m closer to the body (i.e., closer to the axis of rotation) will increase the angular
velocity, while moving it further away will decrease the angular velocity. Note that since a force is
needed to move the mass m (and the force does work on the mass), energy is not conserved.
————
Example 1.20. Handle brake

The shown construction is used to control the vertical motion of a L
heavy particle of mass m. The particle is attached to a rigid, massless l F
rope coiled around a massless disk (inner radius ri , outer radius ro ).
The downward motion of the heavy particle can be controlled by m ri
ro
applying a force F to the shown massless handle brake, resulting in massless
kinetic friction (coefficient µ) between the brake lever and disk.
What force F must be applied to let the particle move downwards at g m

a constant speed vc ?
This is a classical problem which in statics was solved by drawing free-body diagrams and applying
the equilibrium relations of vanishing net forces and torques on each subsystem. Here, we proceed in
exactly the same manner – but we replace the static equilibrium equations by the balance equations
of linear and angular momentum.
We start with free-body diagrams of all three system components, the heavy particle, the disk, and
the brake, as shown below. For each system we can formulate a balance law of convenience: LMB
for the particle m in the shown x-direction, AMB for the disk to avoid the support reactions A
and B at the center, and again AMB for the brake lever for the same reason (avoiding the support
35
reactions H and V ). Note that for each massless component we do not need to account for inertial
effects.
L N
l S
F R
H ro j
R ri
N B
V
A massless m
massless
S x
mg
For the massless lever arm, static equilibrium of torques gives

L
0 = Nl − FL ⇒ N =F . (1.157)
l
Note that friction always counteracts the motion, so that – knowing the mass will move downwards
and hence the disk will rotate clockwise, we drew the friction force R in the above free-body diagram
in the correct direction to counteract the motion of the rotating disk. Similarly, we know that N
is positive in the shown direction, since a compressive force pressed the lever against the disk.
Therefore, we may write, assuming kinetic friction between the lever and the disk,
L
R = µN = µF , (1.158)
l
without further worrying about the directions of the forces.
For the massless disk we obtain from static equilibrium of torques

ro L ro
0 = Sri − Rro , ⇒ S=R = µF . (1.159)
ri l ri
Finally, for the heavy particle, linear momentum balance reveals
L ro F L ro
mẍ = mg − S = mg − µF ⇔ ẍ = g − µ . (1.160)
l ri m l ri
Note that, for convenience,
P we here define the x-coordinate to point downwards (so that, when
formulating mẍ = F
i i all forces pointing into the x-direction downward are positive, while
upward forces are appear as negative). This choice is arbitrary and we could as well have written
mẍ = −mg + S with x defined as positive upwards.
For the mass to move downwards at a constant velocity vc , we need its acceleration to vanish, i.e.,
F L ro mg l ri
ẍ = 0 ⇔ g−µ =0 ⇒ F = = Fcrit . (1.161)
m l ri µ L ro
36
However, this force cannot be applied right away. Otherwise, the mass would never start to move.
If the system is first considered without any applied force (F = 0), then the particle is falling freely
with ẍ = g. Then, the motion is described by
gt2
ẍ(t) = g ⇒ ẋ(t) = gt ⇒ x(t) = . (1.162)
2
The velocity vc is reached after a time period tc :
vc
ẋ(tc ) = vc ⇔ tc = . (1.163)
g
Therefore, a constant downward speed of vc is achieved by applying the force Fcrit from time tc on
(neglecting any transient effects).
————
1.2.4 Particle impact
Particle impact is an interesting example of applying the momentum balance laws derived so far.
It does not require much new physics, but it is worth discussing this example in detail, as it will
lead us to simple relations about the particle speeds before and after impact.
Let us consider a particle of mass m, which is impacting a rigid wall with an initial velocity v− .
The wall is significantly heavier and larger than the particle, so we may treat it as a rigid object
with constant zero velocity. The collision event between the particle and the wall hence results
in a change in the velocity of the particle, which we would like to analyze. During the collision,
the wall exerts a time-dependent force F (t) onto the particle (initially being zero until the particle
first touches the wall at time t− , then increasing to a maximum, and finally decaying to zero again
as the particle separates from the wall after impact at time t+ > t− , as shown below).
t- t0 t+ period of
n compression
-evn F
m v(t0) =vtet vt period of
F=Fen F F ^
P
v(t+) comp
restitution
v (t-) m
vt m ^
P rest
vn
steps during a particle-wall collision
t- t0 t+ t
Since the balance equation of linear momentum must hold at every instance of time during the
collision, we may integrate it over the time period of the collision, denoted by [t− , t+ ], to obtain
mv(t+ ) t+
d
Z Z
(mv) = F (t) ⇔ d(mv) = F (t) dt ⇒ d(mv) = F (t) dt. (1.164)
dt mv(t− ) t−
37
If we denote the integrated force received by the particle by the impulsive force P̂ , then
Z t−
mv(t+ ) − mv(t− ) = P̂ = F (t) dt. (1.165)
t+
Note that P̂ [N s] does not have the units of a force, but it is a force integrated over time (so,
roughly speaking, the ‘accumulated force’ applied to the particle during a specific period).
In order to better understand the nature of this impulsive force, we decompose the particle collision
into a period of compression [t− , t0 ) and a period of restitution [t0 , t+ ], as schematically shown
below, such that
Z t0 Z t+
P̂ = P̂comp + P̂res , P̂comp = F (t) dt, P̂rest = F (t) dt, (1.166)
t− t0
where time t0 is taken as the time at which the particle velocity normal to the wall vanishes before
changing its sign (the normal velocity changes direction, so there must exist such time t0 at which
the normal velocity component is zero).
Let us assume that the wall is frictionless, so we know that there is no frictional force and,
consequently, the force on the particle must be acting perpendicular to the wall at all times. For
convenience, we introduce a local coordinate system such that in 2D the basis vectors {en , et } point
in the direction normal and tangential to the wall, respectively, and en ⊥et . This allows us to write
F (t) = F (t)en and consequently
P̂ = P̂ en and P̂comp = P̂comp en , P̂rest = P̂rest en . (1.167)
Now formulating the momentum balance (1.165) for each of the two phases leads to
mv(t0 ) − mv(t− ) = P̂comp en , mv(t+ ) − mv(t0 ) = P̂rest en . (1.168)
Inserting v = vn en +vt et and exploiting that en ·et = 0 lets us conclude that the tangential velocity
component must stay constant during the entire collision:
vt (t− ) = vt (t0 ) = vt (t+ ) (1.169)
Furthermore, the normal components in (1.168) must satisfy
mvn (t0 ) − mvn (t− ) = P̂comp , mvn (t+ ) − mvn (t0 ) = P̂rest . (1.170)
Recall that we had defined t0 such that vn (t0 ) = 0 (since t0 marks the turning point for the particle
velocity during the collision). and was the definition for time t0 . (1.170) thus reduces to
−mvn (t− ) = P̂comp , mvn (t+ ) = P̂rest . (1.171)
These are two equations with three unknowns, P̂comp , P̂rest , and vn (t+ ) (while we assumed that
the initial particle velocity v(t− ) = v− is known, so that vn (t− ) = v(t− ) · en is known as well).
38
Consequently, one additional equation is required to solve the above system of equations. To this
end, we introduce the coefficient of restitution e as
P̂rest
e= ⇒ P̂rest = e P̂comp , e ∈ [0, 1]. (1.172)
P̂comp
e characterizes how much of the impulsive force “injected into” the particle during the initial
compression phase of the collision is available for recovery of the particle and separation of particle
and wall during the subsequent restitution phase. If e = 1, the impact is purely elastic and
the particle is fully restored during the restitution phase and separates from the wall. For the
perfectly plastic case e = 0, no restitution occurs and the particle remains at the wall (there is
no impulsive force driving the particle to separate from the wall). For intermediate cases 0 < e < 1,
often referred to as partially elastic, the particle is only partially restored.
Inserting (1.172) into (1.170) finally yields

−mvn (t− ) = P̂comp , mvn (t+ ) = P̂rest = e P̂comp ⇒ vn (t+ ) = −e vn (t− ) (1.173)
As a consequence, the particle’s velocity changes sign (as could have been expected) and it also
changes its magnitude unless e = 1. Overall, the particle velocity during a frictionless impact thus
changes according to
v(t− ) = vn en + vt et ⇒ v(t+ ) = −evn en + vt et . (1.174)
The coefficient of restitution e generally depends on the particle’s material properties. Linear elastic
materials have e = 1 (e.g., a rubbery bouncy ball), whereas 0 ≤ e < 1 implies that some of the
kinetic energy of the particle before impact is used to permanently deform or change the particle,
thus dissipating energy (e.g., a hacky sack is an example for e ≈ 0, changing its shape and showing
little to no rebound).
Let us finally investigate the change in angle of the particle velocity during impact. While the
tangential component of the velocity vector remains the same for a frictionless impact, the normal
component changes depending on the value of e. As shown schematically below, this affects the
direction of the particle’s flight in the following manner. For a purely elastic impact, the angle of
the incoming particle (measured against the normal) is also the angle of the rebounding particle
(with the normal) since |vt (t− )/vn (t− )| = |vt (t+ )/vn (t+ )|. In case of a perfectly plastic collision,
the particle retains no normal component after impact (vn (t+ ) = 0) and hence slides down the
wall with v(t+ ) = vt et . For any case 0 < e < 1, the particle rebounds with an angle between the
incoming angle and the vertical wall since |vt (t+ )/vn (t+ )| = |vt (t− )/evn (t− )| ≥ |vt (t− )/vn (t− )|.
vt vt
vt
-vn -evn
vt vt vt
vn vn vn
e=1 e=0 0 <e < 1
39
As a concluding remark, notice that the kinetic energy of the particle always changes during impact,
unless the collision is purely elastic, since (with 0 ≤ e ≤ 1)
m 2 2
mh 2 2
i
T (t− ) = v (t− ) + vt (t− ) and T (t+ ) = evn (t− ) + vt (t+ ) ≤ T (t− ). (1.175)
2 n 2
Example 1.21. Particle dropped vertically onto a rigid ground

Consider a particle of mass m dropped vertically onto a rigid ground,
released from rest at an initial height h− and rebounding to a
final height h+ . Let us find out how we can use the difference in
height (and later the rebound period) to determine the coefficient of g m
restitution e.
Since the particle is dropped in the normal direction onto the ground, h-
only normal velocity components are to be considered and v · et =
h+
vt = 0 at all times t. Except for the purely elastic case (for which
e = 1 and the particle bounces back to the initial height), we cannot
e
use the conservation of energy for the entire experiment. However,
we can use it for each flight of the particle (from release to impact,
and from rebound to the final height).
This gives for the flight from release at h− to the ground with impact velocity v− = v(t− )
m 2 p
mgh− = v− ⇒ |v− | = 2gh− (1.176)
2
and for the flight from rebound with velocity v+ = v(t+ ) to the maximum rebound height h+
m 2 p
v+ = mgh+ ⇒ |v+ | = 2gh+ . (1.177)
2
Since v− and v+ are the normal velocities immediately before and after the impact, they are related
by the coefficient of restitution:
s
p p h+
|v+ | = e|v− | ⇒ 2gh+ = e 2gh− ⇔ e= . (1.178)
h−
Hence, the coefficient of restitution can be obtained from the heights before and after impact.
As a variation, let us find out how the coefficient of restitution can alternatively be determined
from the flight times of the particle in the air between two ground contacts. For a particle of mass
m, flying upwards with some initial velocity v = v0 e2 (and only gravity is acting on the particle,
so F = −mge2 ), the flight starting from x2 = 0 is described by
gt2
mẍ2 (t) = −mg ⇒ ẋ2 (t) = −gt + v0 ⇒ x2 (t) = − + v0 t. (1.179)
2
The time t∗ from launching the particle with v0 to reaching the ground again is therefore given by
gt2∗

gt∗ ! 2v0
x2 (t∗ ) = − + v0 t∗ = t∗ − + v0 = 0 ⇒ t∗ = (or t∗ = 0). (1.180)
2 2 g
40
Furthermore, note that – by the conservation of energy during the particle flight – the particle
speed at launch is the same as the particle speed when it reaches the ground again.
Now consider the particle of mass m dropped onto the rigid ground. As shown above, the particle
reaches the ground with |v− | and leaves the ground again with |v+ | = e|v− |. By the above energy
conservation argument, the particle again has velocity e|v− | when it reaches the ground the next
time. Hence, the upward particle velocity after rebound varies from one rebound to the next
according to
v0,i+1 = ev0,i , (1.181)

m g
By inserting the latter into (1.180), we obtain the following relation for
the duration of the particle flight between two subsequent rebounds: ev 0,i v0,i+1 v0,i+1 v0,i+2 v0,i+2
t∗,i+1 v0,i+1
= = e. (1.182) t*,i+1 t*,i+2
t∗,i v0,i
In conclusion, we have found two ways to measure the coefficient of restitution for a particle–wall
impact. We can either measure the change in height or the change in flight times between rebounds
to find the coefficient of restitution e. Note that the above is independent of the particle mass m.
Finally, recall that e is a material property that defines (losely speaking) how much of the kinetic
energy is consumed by a permanent change of the particle by, e.g., plasticity or damage.
————
Example 1.22. Car roll-over
A car of mass m and cross-section a × b slides across a
frictionless ground horizontally with an initial velocity v0 x2 b g
m
v0
against a shallow curb, resulting in an impulsive collision x1 a
with the curb. At what velocity v0 does the car start to roll
m=0
over?
We need to make a few approximations and assumptions to solve this problem with those methods
available to us. To this end, let us approximate the car as a particle of mass m, located at the
center of mass of the car. We assume that during the collision an impulsive force F̂ acts onto the
car at the bottom of the curb, directed towards the center of mass (i.e., acting onto the particle).
Since the curb is shallow, its height is negligible and we may assume that force F acts (and the
care trips over) a fixed point B, as shown below.
We consider the balance of angular momentum around the curb at the fixed point B (so vB = 0):
MB = ḢB + vB × P ⇒ MB = ḢB . (1.183)
For a short collision time period [t− , t+ ] the above can be integrated as
Z t+
HB (t+ ) − HB (t− ) = MB dt and MB = rBP × (N − mg). (1.184)
t−
Note that this relation is analogous to (1.165), which was derived based on linear momentum
41
balance during a collision. The impulsive force F̂ plays no role here since it goes through point B
and hence produces no torque with respect to B.
We further assume that the time period of the collision is so short (and the impulsive force so
dominant), that the normal and gravity forces remain approximately constant during [t− , t+ ]. This
in turn leads to the conclusion that the angular momentum remains constant during the collision
in this problem:
Z t+ Z t+
HB (t+ ) − HB (t− ) = MB dt = rBP × (N − mg) dt ≈ 0. (1.185)
t− t−
mg mg v+
am v0 am
B B
N F N F
b b
before the collision immediately after the collision
The angular momentum before the collision is

b a a
HB (t− ) = rBP × mv0 = − e1 + e2 × mv0 e1 = −mv0 e3 , (1.186)
2 2 2
and the angular momentum immediately after √ the collision (when the car is assumed to have a
speed v+ rotating about point B with radius a2 + b2 /2, as shown above) is
√
a2 + b2
HB (t+ ) = rBP × mv+ = − mv+ e3 . (1.187)
2
Conservation of angular momentum thus yields the velocity v+ immediately after the impact as
√
a ! a2 + b2 a
−mv0 e3 = − mv+ e3 ⇒ v+ = v0 √ . (1.188)
2 2 a + b2
2
To determine whether or not the car rolls over, we need to consider the car’s motion after it hits the
curb. We here assume for simplicity that, immediately following the collision, the car will perform
a rotation around point B and the car remains attached to point B during the rotation (as shown
below). The question is now, given an initial velocity V+ , how far will the car rotate? Under the
action of gravity (and by the conservation of energy in this conservative system), the velocity will
decrease as the car rolls over and the center of mass gains altitude. We thus determine if the center
of mass has a non-zero speed when it has rotated right above the curb vertical above point B – if
the velocity is positive, the car will continue to rotate onto the curb; if the velocity has gone to
zero before that point, then the car will rotate back onto the street.
42
After the collision, the car’s rotation is a conservative rigid-body motion, so that we may use the
conservation of energy. Right after the collision we have
a m 2
E(t1 ) = V (t1 ) + T (t1 ) with V (t1 ) = mg , T (t1 ) = v , (1.189)
2 2 +
whereas
√ the instance when the center of mass is right above point B (at an altitude of h =
a2 + b2 /2 above B) is characterized by
√
a2 + b2 m 2
E(t2 ) = V (t2 ) + T (t2 ) with V (t2 ) = mgh = mg , T (t2 ) = vcrit . (1.190)
2 2
As discussed, a roll-over occurs if vcrit ≥ 0, so we insert the limiting

case vcrit = 0. Conservation of energy along with v+ from (1.188)
thus gives x2
x1 a/2 h
√ B
a m 2 a2 + b2 2 a2
mg + v+ = mg with v+ = v02 2 , (1.191) critical state for roll-over
2 2 2 a + b2
and solving for the critical initial velocity leads to
s
b2
p
v0 = g 2 2
a +b −a 1+ 2 . (1.192)
a
We close by verifying two special cases: if b a (a very long and flat car), then v0crit → ∞ so an
extremely large initial speed v0 is required for a roll-over. By contrast, for a b (a very high and
thin car) v0crit → 0, so a very tall and narrow car will roll over already at very low initial speeds.
————
1.3 Summary of Key Relations
At the end of our discussion of the dynamics of a single particle, we summarize all key relations –
for a single particle of constant mass m and position r(t) – in the following box. This summary
box (together with all other such summary boxes that we will add at the end of each topic that we
discuss in the following) will grow into the formula collection to be used during exams.
43
kinematics of a single particle:

dr dv
v= , a=
dt dt
velocity and acceleration components in polar coordinates (r, ϕ):
v = ṙer + rϕ̇eϕ , a = (r̈ − rϕ̇2 )er + (2ṙϕ̇ + rϕ̈)eϕ
velocity and acceleration components using the space curve description:
v2 dr v ėt ρ
v = ṡet , a = s̈et + en with et = = , en = = ėt
ρ ds v |ėt | v
balance of linear momentum for a particle of mass m:

X d
Fi = Ṗ = (mv)
dt
i
kinetic friction:
v
R = −µ|N | , |R| = µ|N |
|v|
work–energy balance for a particle of constant mass m:

t2 r2
1 XZ XZ
T (t2 ) − T (t1 ) = W12 , T (t) = m|v(t)|2 , W12 = Fi · v dt = Fi · dr
2 t1 r1
i i
for a conservative force:
dV
F =− ⇒ W12 = V (r1 ) − V (r2 )
dr
conservation of energy for a conservative system:
T + V = const.
balance of angular momentum of a particle with respect to point B:
MB = ḢB + vB × P with HB = rBP × P , P = mvP
special case of a rotation in 2D around a fixed point B at a distance R:
MB = IB ϕ̈ with IB = mR2
particle collision with a frictionless rigid wall:
vn (t+ ) = −e vn (t− ), vt (t+ ) = vt (t− )
44
2 Dynamics of Systems of Particles
Having discussed the dynamics of a single particle, we proceed to consider a system of n particles at
positions ri (t), (where i = 1, . . . , n), described in a Cartesian, inertial reference frame. Systems of
multiple particles are not only of importance in many practical situations, but the relations derived
here will also form the basis for our subsequent discussion of the dynamics of bodies of finite size.
As for a single particle, we first discuss the kinematics and then the kinetics of systems of particles.
2.1 Kinematics
We begin by defining the total mass M and the position

of the center of mass rCM (t) of a system of n particles at x2
positions ri (t) and with masses mi (for i = 1, . . . , n) as
mi
n n
X 1 X
M= mi , rCM (t) = mi ri (t) (2.1)
i=1
M
i=1
ri
Note that, while the total mass M remains constant over O x1

time, the center of mass rCM (t) is instantaneously changing rCM
+
CM
based on the changing positions of the individual particles.
Also, the center of mass generally does not coincide with x3
the location of any particle.
In an unconstrained system, all particles can move freely, so the system has a total of d × n DOFs
in d dimensions8 . In most mechanical systems, particles are connected by means of, e.g., rigid rod
or stretchable springs. Like for a single particle, we can impose constraints on the particle motion
and, in particular, we can constrain the motion of particles relative to each other by introducing
kinematic constraints (as already discussed for a single particle in Section 1.1.2).
An important kinematic constraint is a rigid link between particles i and j, which enforces that
|ri (t) − rj (t)| = const. for all times t. If all particles are rigidly connected, i.e., if |ri (t) − rj (t)| =
const. for all i, j = 1, . . . , n at all times t, then we speak of a rigid system of particles. In simply
terms, in a rigid system the distances between all particles remain constant.
A rigid link |ri − rj | = const. is an example of a so-called a holonomic constraint, which can most
generally be expressed as
f (r1 , . . . , rn , t) = 0. (2.2)
That is, a kinematic constraint which depends only on the particle positions and time is called a
holonomic constraint. We further say two constraints are independent, if the two relations (2.2)
associated with the two constraints are linearly independent.
8
For us, all particles in a particle system are distinguishable (unlike, e.g., in atomic particle ensembles). Therefore,
the total number of DOFs is d × n.
45
In an extension of our observation for a single particle, we state:
If a system of n particles has r independent holonomic constraints,

(2.3)
then the system in d dimensions has (n × d − r) degrees of freedom.
As already discussed for a single particle, each constraint produces a constraint force (or con-
straint forces): in order to maintain the kinematic constraint, forces are required (and shall not be
forgotten when drawing free-body diagrams). Summarized below are four examples of holonomic
kinematic constraints.
Example 2.1. Kinematic constraints
mj j
xi
j mi
R ro ri
mi
rj xj xj
ri mi mi xj mj
xi mj xi mj
(a) (b) (c) (d)
(a) A rigid link (e.g., a rigid bar) is a kinematic constraint defined by |ri (t) − rj (t)| = const.
The constraint results in a force in the bar and hence in constraint forces acting onto the two
particles parallel to the bar, which prevent the particles from increasing or decreasing their
distance. In 3D this system has 2 × 3 − 1 = 5 DOFs.
(b) Two particles connected by an inextensible rope (i.e., a rope of constant length) also
adhere to a kinematic constraint. If we assume that the particles move only vertically, then
it is convenient to introduce only vertical DOFs, such as xi and xj shown in the graphic. In
principle, we can introduce those arbitrarily up or down, and the rotation angle of the pulley
clockwise or counterclockwise. It is oftentimes convenient to introduce them in a fashion that
is ‘kinematically reasonable’, as done here (if the left particle moves down, the pulley rotates
counter-clockwise, and the right particle moves up). Yet, this is not required.
In the shown example of two particles connected by a rope over a pulley of radius R, the two
vertical positions of particles i and j and the rotation angle of the pulley are connected via
xi (t) = xj (t) = R ϕ(t). (2.4)
The constraint results in a force in the rope, leading to forces acting onto the two particles
(and the pulley). It is important to define the constraints consistently with the DOFs. In
example (b) shown above, xi points downwards, while xj points upwards; therefore xi = xj .
If we instead chose both xi and xj to point downwards, then would we instead have xi = −xj .
When we consider that both particles can only move vertically, then each particle has 1 DOF,
so that with 1 constraint the system has only a single DOF (2 × 1 − 1 = 1).
46
(c) Two particles connected to two inextensible ropes wrapped around a rigid cylinder with
inner and outer radii ri and ro , respectively, satisfy the kinematic constraints (again, using
the specific DOF definitions of the above schematic)
ro
xi (t) = ro ϕ(t), xj (t) = ri ϕ(t) ⇒ xi (t) = xj (t) . (2.5)
ri
Again, when we consider that both particles can only move vertically, then the system has
only a single DOF.
(d) The last schematic shows two particles that are connected by an inextensible rope, as in case
(b). Therefore, they must satisfy the kinematic constraint
xi (t) = xj (t), (2.6)
when we introduce the DOFs conveniently, as shown above. That is, rather than introducing
horizontal and vertical components for each particle position, we exploit that we know their
direction motion, so that xi and xj can conveniently be introduced as shown to admit the
simple constraint xi = xj . Of course, one could also introduce alternative DOFs and change
the constraint accordingly. As before, this system has only a single DOF.
We re-iterate that the choice of the definition of the DOFs in the above examples is arbitrary.
As already mentioned, we could as well have defined the translation DOFs xi and xj in opposite
directions, the same applies to the rotational DOFs – as long as we formulate the constraints
appropriately. E.g., if in case (b) both xi and xj pointed upwards, then we had xi = −xj = −Rϕ.
Or if in case (b) ϕ was defined clockwise (and xi , xj as shown), then we had xi = xj = −Rϕ. We
may hence choose the DOFs arbitrarily, but it makes sense to define DOFs in agreement with known
or anticipated motion. For example, in cases (b) and (c) we drew all DOFs based on the assumption
that the pulleys were to rotate counterclockwise. Again, this is arbitrary (and the solution would
be exactly the same if defined otherwise), as long as we define all constraints consistently (and the
same applies to the system kinetics, discussed next in Section 2.2).
————
We point out that not every connection between particles is necessarily a kinematic constraint.
For example, when two particles are connected by an elastic spring, then that spring does indeed
induce a force onto the particles, as soon as they change their relative distance (and hence stretch or
compress the spring). However, the particles are still free to move in a kinematically unconstrained
fashion. So, two particles in 3D still retain their 2×3 = 6 DOF, even though connected by a spring.
-F
mi mj mi F mj
47
2.2 Kinetics
For a system of particles i = 1, . . . , n that are connected (e.g., through

rigid links, deformable springs, ropes, etc.), we decompose all forces in the
system into internal forces (i.e., forces stemming from particle-particleF1ext
interactions) and external forces (i.e., forces applied externally to m1 m2
int
individual particles in the system ). F13 ext F2
ext
F3
-F13
int
int
As we generally assume that particle links (e.g., rigid bars or elastic m3 F24
springs) are massless, internal forces always act as force pairs between F34int -F int
24
particles (same magnitude, opposite direction), as shown on the right. -F34int m
When denoting the force acting on particle i due to particle j by Fij , 4
ext
Newton’s third law (action = reaction) for a massless link requires F4
Fij = −Fji for all i 6= j. (2.7)
Therefore, the net force acting on particle i is
X X
Fi = Fiext + Fij = Fiext + Fiint where Fiint = Fijint (2.8)
j6=i j6=i
is the total internal force on particle i, and Fiext denotes the external force applied to particle i.
When summing over all internal forces acting on the system of particles, we observe that the internal
forces cancel pairwise, since
 
Xn n X
X n
X X X
Fiint = Fijint =  Fijint + Fijint 
i=1 i=1 j6=i i=1 j>i i>j
    (2.9)
n
X X X n
X X X
=  Fijint + Fjiint  =  Fijint + (−Fijint ) = 0.
i=1 j>i j>i i=1 j>i j>i
Consequently, the sum over all forces on all particles is the same as the sum over all external forces
only (which will be beneficial in the formulation of the balance laws):
n
X X n
X
Fiint =0 ⇒ Fi = Fiext (2.10)
i=1 i i=1
In the following, we will extend the balance laws previously derived for a single particle to a system
of particles.
2.2.1 Balance of linear momentum
Assuming that the number of particles remains constant as does the total mass (i.e., M = const.),
we start with the balance of linear momentum for each particle in an inertial frame, i.e.,
d
Fi = (mi ṙi ) for i = 1, . . . , n. (2.11)
dt
48
Summing the above linear momentum balance over all particles, using (2.10), gives
n n n n
X X X d d 1 X d
Fi = Fiext = (mi ṙi ) = M mi ṙi = M vCM , (2.12)
dt dt M dt
i=1 i=1 i=1 i=1
where
n n n n
1 X 1 X 1 X 1 X
vCM = mi ṙi = mi vi and aCM = mi r̈i = mi ai (2.13)
M M M M
i=1 i=1 i=1 i=1
denote, respectively, the velocity and acceleration of the center of mass of the system of particles.
Realizing that the total linear momentum of a system of particles is given by
P = M vCM (2.14)
(2.12) gives us the linear momentum balance for a system of n particles as
n n
X d X
Fiext = Ṗ = (M vCM ) ⇒ Fiext = M aCM (2.15)
dt
i=1 i=1
This allows us to interpret the system of particles, loosely speaking, as a “super-particle” having
mass M and whose motion is described by the position rCM (t) of its center of mass. Note that
these relations apply not only for rigid collections of particles but for general systems of n particles,
irrespective of the (symmetric) particle interactions, since the inner forces do not enter the equation.
The above relations of linear momentum balance are analogous to Eqs. (1.61) and (1.63), respec-
tively, for a single particle – here replacing the particle mass m by the total mass M of the system
of particles, and replacing the particle velocity by the velocity of the center of mass of the system
of particles.
Of course, we may still also – additionally – apply the balance of linear momentum equations
(1.61) and (1.63) to each particle individually, if needed. However, when formulating balance laws
for an individual particle, we may not drop the internal forces (since those only cancel in the above
relations for the full system).
Example 2.2. Falling dumbbell

Consider a dumbbell, i.e., two particles of masses m1 and m2 connected by
a rigid, massless bar. When the system is subjected to gravity as the only m
external forces, linear momentum balance on the whole system reads 1 CM
+ m2
(m1 + m2 )aCM = m1 g + m2 g ⇒ aCM = g, (2.16)
so that integration with suitable initial conditions leads to +

gt2 g
rCM (t) = − + vCM (0)t + rCM (0). (2.17)
2
49
Therefore the dumbbell will accelerate downwards under the action of gravity and thereby, as
expected, behave like a “super-particle” of mass m1 + m2 (cf. our derivation in Example1.1 for a
single particle). However, when we consider linear momentum balance for each particle individually,
we obtain for particle 1 and 2, respectively,
m1 a1 = m1 g + Fs , m2 a2 = m2 g − Fs , (2.18)
where Fs is the constraint force (acting onto each particle parallel to the rigid link). Therefore,
while the system as a whole is undergoing a free fall, accelerated by g, the two particles generally
experience different accelerations (a1 , a2 6= g in general), which depend on the current orientation
of the dumbbell and are more complex to calculate. As expected, adding up the two individual
LMB equations in (2.18) yields the system LMB in (2.16).
————
Starting with the work–energy balance for each individual particle, which we derived in Eq. (1.87),
we now sum the kinetic energies of all n particles as well as the work done on all n particles. This
leads to the work–energy balance for a system of particles as
n n Z ri (t2 )
1X X
T (t) = mi |vi (t)|2 , W12 = Fi · dri ⇒ T (t2 ) − T (t1 ) = W12 (2.19)
2 ri (t1 )
i=1 i=1
Hence, the work–energy balance for a system of particles is completely analogous to a single particle
– the kinetic energy and the work are replaced by those of the system.
We consider the following special cases:
• As before, in case of a conservative force, W12 = V1 − V2 can be derived from a potential V

for each force. If all forces are conservative, then the conservative system is governed by
the conservation of the total energy, i.e.,
T + V = const. (2.20)
for the whole system of n particles.
• If the link between two particles is rigid, one can show that the internal forces perform no
work. To see this, notice that a rigid link between particles i and j implies
d
|ri (t)−rj (t)| = const. ⇒ |ri −rj |2 = 0 ⇒ (ri −rj )·(ṙi − ṙj ) = 0. (2.21)
dt
As discussed before, the interaction force between particles i and j is always parallel to the
distance vector: Fij ||rij = rj − ri . Therefore, from (2.21) it follows that
rij ⊥ (ṙi − ṙj ) ⇒ Fij ⊥ (vi − vj ). (2.22)
50
Looking at the work done by the interaction forces Fij and Fji = Fij onto the two particles
i and j, we realize that
Z ri (t2 ) Z rj (t2 ) Z t2 Z t2 Z t2
Fij · dri + Fji · drj = Fij · vi dt + −Fij · vj dt = Fij · (vi − vj ) dt = 0.
ri (t1 ) rj (t1 ) t1 t1 t1
(2.23)
Hence, the forces arising from a rigid link in sum do not perform any work.
• We can extend the previous point to a rigid system: if a system of particles is rigid, all
internal forces perform no work. This can easily be established by summing over all particles:
n Z
X ri (t2 ) n XZ
X t2 n XZ
X t2
Fiint · dri = Fij · vi dt = Fij · (vi − vj ) dt = 0. (2.24)
i=1 ri (t1 ) i=1 j6=i t1 i=1 j>i t1
Therefore, all interaction forces together perform no work on the rigid system of particles.
Note that this allows us to replace the total particle force Fi in (2.19) by only the external
force Fiext . Keep in mind that this is true only for rigid systems. If, e.g., two particles are
connected by an elastic spring, the spring may perform (conservative) work on the particles, as
evidenced by the change in potential energy when the distance between the particles changes.
We highlight the important fact that, even though the linear momentum of a system of particles
was simply P = M vCM , the analogous does not hold for its kinetic energy, i.e.,
n
X mi M
T = |vi |2 6= |vCM |2 . (2.25)
2 2
i=1
Physically speaking, this is because the center of mass’ translation alone does not capture all kinetic
energy in the system (an additional contribution to the kinetic energy comes from the rotation of
the system about the center of mass, which is not captured by vCM ).
Like for linear momentum balance, we start with the balance law of angular momentum for each
individual particle (see Eq. (1.132)). The total angular momentum is then given by the sum over
all particles, i.e.,
n
X
HB = rBPi × Pi with Pi = mi vPi . (2.26)
i=1
Following the analogous derivation as for a single particle (cf. Eq. (1.130)), we now obtain
n
X n
X n
X n
X
ḢB = ṙBPi × Pi + rBPi × Ṗi = (vPi − vB ) × Pi + rBPi × Fi
i=1 i=1 i=1 i=1
n n n n (2.27)
X X X X
= vPi × Pi − vB × Pi + rBPi × Fi and Pi = P = M vCM .
i=1 i=1 i=1 i=1
51
Note that only the external forces survive summation in the last term (all internal forces cancel
since they are pairs of equal magnitude and opposite direction and same lever am with respect to
point B). Further, we may use that vPi ||Pi since Pi = mi vPi .
This leads to the angular momentum balance for a system of n particles:
MBext = ḢB + vB × P with P = M vCM (2.28)
with the resultant torque of all external forces,

n
X
MBext = rBPi × Fiext . (2.29)
i=1
As already discussed for a single particle, the term vB × P vanishes, if (i ) point B is fixed (vB = 0)
or (ii ) if B moves parallel to the center of mass (vB ||vCM ). It is therefore convenient to pick one of
those two options when choosing point B.
Example 2.3. Particles on a rotating disk

We consider a massless disk that is rotating with angular velocity
ω(t) = ϕ̇(t) around its center B. n particles of masses m1 , . . . , mn are
rigidly attached to the disk at fixed radii r1 , . . . , rn . In this case the
above formulation reduces to (using co-rotating polar coordinates, as
in the case for a single particle; see Section 1.2.3):
j massless
B
n n
ri m
X X
HB = rBPi × Pi = ri eri × (mi vPi )
i
i=1 i=1
n n (2.30)
X X
= mi ri2 ωe3 = mi ri2 ϕ̇e3 ,
i=1 i=1
where we used that vPi = ri ωeϕi and eri × eϕi = e3 . Hence, for this special case the balance of
angular momentum and the moment of inertia IB become, respectively,
n
X
MBext = IB ϕ̈, with IB = mi ri2 . (2.31)
i=1
If the only force acting on the system is gravity (besides the support reactions at the center), then
the total external moment is (assuming the coordinate origin at point B)
n n
!
X X
MBext = rBPi × mi g = mi rBPi × g = M rCM × g. (2.32)
i=1 i=1
This shows that, if the particles are arranged such that their center of mass is point B, then gravity
does not produce an external moment on the system of particles and hence does not lead to a
rotation of the disk.
————
52
Example 2.4. Moment of inertia of a thin wheel and of a slender rod

Let us consider the special case of n particles dis-
M
tributed uniformly along the circumference of an j j
otherwise massless disk of radius R. We keep the
B
n®¥
B
total mass M constant while increasing the num- R R
ber of particles, so that the mass of each particle
mi
becomes mi = M/n. This leads to massless
n n n
X X M 2 X 1
IB = mi ri2 = R = M R2 = M R2 . (2.33)
n n
i=1 i=1 i=1
This shows that the result is independent of the number of particles and, in particular, also holds
for the case n → ∞, which corresponds to a wheel whose mass is distributed evenly around the
perimeter (e.g., a bike’s tire rim). We have thus derived the 2D moment of inertia of a thin wheel
of radius R and mass M , rotating about its center, as IB = M R2 .
Analogously, we consider a slender rod of total mass M and length L, which is rotating about one
ofits end points. Following the above example, we interpret the rod as a distribution of particles,
each of mass mi = M/n, which are equally spaced at distances l = L/n, so that the ith particle is
located a distance ri = il = iL/n from the center of rotation. This leads to
n n 2 n
M L M L2 X 2
mi M
X X
2
IB = mi ri = i = i . (2.34)
n n n3
i=1 i=1 i=1 n®¥
l
Inserting the square sum formula j j L
ri
n
X n(n + 1)(2n + 1)
i2 = (2.35)
6
i=1
finally gives the 2D moment of inertia of a slender rod of mass M and length L, rotating about one
of its end points:
n(n + 1)(2n + 1) (n + 1)(n + 2) M L2
IB = M L2 = M L2 ⇒ lim IB = . (2.36)
6n3 6n2 n→∞ 3
————
Example 2.5. Particle on a rope attached to a rotating disk
The construction shown on the right consists of three particles of m
equal masses mi = m, two of which are rigidly attached to a rotating g
massless disk of radius R, the third one is connected to a rigid rope j
wrapped around the disk. The system is released from rest under
R
the influence of gravity.
m
What is the downward acceleration of the particle on the rope? m
We describe the kinematics of the system by the two DOFs x and ϕ, shown below. To find
the acceleration of the particle on the rope, we need to use linear momentum balance on that
53
particle only. This end, we introduce free-body diagrams for both components of the system (see
Example 1.20). Linear momentum balance on the particle then reads
mẍ = mg − S. (2.37) g mg
For the rotating disk, angular momentum balance is advantageous to S H j
avoid the support reactions. With moment of inertia IB = 2mR2 ,
V
this gives
x S mg
2
2mR ϕ̈ = SR + mgR sin ϕ − mgR sin ϕ = SR, (2.38)
mg
where we observe that the external moment resulting from gravity acting on the two masses on the
disk cancels by symmetry for any rotation angle: r1 × (−mg)e2 + r2 × (−mg)e2 = 0 since r1 = −r2
(where r1 and r2 denote the vectors from the center of rotation to the two particles). This confirms
our derivation in (2.32) (the center of mass of the two particles here rests at the center of the disk).
Finally, we need a kinematic constraint to solve the system of (so far) two equations with three
unknowns (ẍ, ϕ̈ and S). The rigid rope implies that
x(t) = Rϕ(t) ⇒ ẍ(t) = Rϕ̈(t). (2.39)
The three equations (2.37), (2.38), and (2.39) can be solved for
g
mẍ = mg − S = mg − 2mRϕ̈ = mg − 2mẍ ⇒ . ẍ = (2.40)
3
Hence, the particle accelerates downwards with only a third of the gravitational acceleration (be-
cause it additionally needs to overcome the inertia of the rotating particles on the disk).
————
Example 2.6. Two particles connected by a rigid rope
Two particles of masses m1 and m2 are connected by a rigid rope
that revolves around a massless disk of radius R. Considering g R
massless
gravity, what are the accelerations of the two particles and what is
the reaction force on the disk’s hinge?
m1
Using the kinematic constraint x1 = x2 = x (as drawn) we can find m2
the solution from linear momentum balance on the two particles and
angular momentum balance on the disk (which is massless):
m1 ẍ = m1 g − S1 , m2 ẍ = S2 − m2 g, (S1 − S2 )R = 0. (2.41)
j
Solving the three equations leads to the particle acceleration H
m1 − m2
V
ẍ = g. (2.42)
S1
rCM S2
m1 + m2
As a quick sanity check, we verify that for equal masses m1 = m2 , x
the system is in static equilibrium and ẍ = 0. Further, if m1 m2 x
then ẍ ≈ g, while for m1 m2 we have ẍ ≈ −g. m1g m2g
54
The reaction forces on the hinge are obtained from linear momentum balance as H = 0 and
m1 m2
V = S1 + S2 = m1 (g − ẍ) + m2 (g + ẍ) = (m1 + m2 )g + (m2 − m1 )ẍ = 4g . (2.43)
m1 + m2
Notice the difference to the static case where V = (m1 + m2 )g, which is recovered if m1 = m2 .
Let us also verify the balance of linear momentum on the global (complete) system. We know that
X
(m1 + m2 )aCM = Fiext = −(m1 + m2 )ge2 + V e2 + He1 . (2.44)
i
The location and acceleration of the center of mass, respectively, can be written as
m1 (−xe2 ) + m2 (xe2 ) m1 − m2 m1 − m2
rCM = =− xe2 ⇒ aCM = − ẍe2 . (2.45)
m1 + m2 m1 + m2 m1 + m2
Note that there is, in general, also an e1 -component of the location of the center of mass (which
depends on the two masses m1 and m2 as well as on the radius of the disk). However, both particles
move only vertically, so the e1 -component of the location of the center of mass does not change
over time, and we chose to place our coordinate system such that the e1 -component is zero. In
addition, the second derivative of the constant e1 -component with respect to time vanishes, so the
acceleration aCM remains unaffected irrespective of the coordinate origin.
Insertion of the center of mass’ acceleration aCM into (2.44) gives

m1 − m2
−(m1 + m2 ) ẍe2 = −(m1 + m2 )ge2 + V e2 + He1 ⇒ H = 0. (2.46)
m1 + m2
We can easily verify that ẍ from (2.42) and V from (2.43) satisfy this relation. Alternatively, if
only one of those two unknowns had been determined, we could have used (2.46) to find the missing
one.
We can also exploit the conservation of energy of the global system since this is a conservative
system (the support forces in the hinge do no work). Choosing the potential energy with respect
to the initial configuration (so the initial kinetic and potential energy is zero) gives
m1 2 m2 2 m1 − m2
−m1 gx + m2 gx + ẋ + ẋ = 0 ⇔ 2gx = ẋ2 . (2.47)
2 2 m1 + m2
Differentiating both sides of this equation with respect to time (and assuming that ẋ 6= 0 during
motion) yields directly
m1 − m2 m1 − m2
2g ẋ = 2ẋẍ ⇒ ẍ = g. (2.48)
m1 + m2 m1 + m2
Finally, note that each particle does not conserve its energy, since the force in the rope performs
work on each particle. That is, even though the complete system conserves its energy, each particle
individually does not in general.
————
55
2.3 Particle Collisions
2.3.1 Collision of two particles
In an extension of the particle-wall impact of Section (1.2.4), let us consider two particles of masses
m1 and m2 , colliding at initial velocities v1 and v2 , respectively. We seek the velocities of the two
particles after the collision.
We follow an analogous procedure here as for the particle-wall impact in Section (1.2.4). There we
showed that linear momentum balance for a particle of mass mi and velocity vi , subjected to an
impulsive force over a time interval [t− , t+ ], leads to
Z t+
mi vi (t+ ) − mi vi (t− ) = P̂i with P̂i = Fi (t) dt, (2.49)
t−
which we may apply to each particle now (i = 1, 2). Furthermore, as for the single particle we
decompose the collision into a period of compression [t− , t0 ) and a period of restitution
[t0 , t+ ] such that any impulsive force P̂ is decomposed into
Z t0 Z t+
P̂i = P̂i,com + P̂i,res , P̂i,comp = Fi (t) dt, P̂i,rest = Fi (t) dt. (2.50)
t− t0
period of
Fn compression
m1 m2 v1(t0)=v2(t 0) v1(t+) period of
^
P restitution
comp
v1(t-) v2(t-) v2(t+)
t t
n v1 v2 m1 m2 m1 m2 ^
P
v1 v2
n rest
steps during a two-particle collision t

t- t0 t+
While the collision process (schematically shown above) is analogous to that of a particle impacting
a wall, we emphasize a few key differences. First, we may no longer assume that the velocity of
any particle at time t0 vanishes. Instead, we assume that v1 (t0 ) = v2 (t0 ) = v(t0 ); i.e., the two
particles are first coming closer, than attain a common velocity v at time t0 , and then – depending
on the coefficient of restitution – may separate again. Second, there is now an impulsive force acting
onto each of the two particles and, by Newton’s third axiom, they are of the same magnitude but
opposite sign: F1 = −F2 and therefore P̂1 = −P̂2 (and the same holds true for the compression
and restitution phases).
Applying linear momentum balance to each particle for the compression and restitution phases
separately (and writing P̂1 = P̂ and P̂2 = −P̂ for convenience) leads to
m1 v(t0 ) − m1 v1 (t− ) = P̂comp , m2 v(t0 ) − m2 v2 (t− ) = −P̂comp ,

(2.51)
m1 v1 (t+ ) − m1 v(t0 ) = P̂rest m2 v2 (t+ ) − m2 v(t0 ) = −P̂rest .
56
For a frictionless collision, momentum is again exchanged only in the normal direction 9 (n ), while
no interaction forces are generated in the tangential direction (t ), i.e., P̂ = P̂ en and hence
v1t (t+ ) = v1t (t− ), v2t (t+ ) = v2t (t− ) . (2.52)
Having found the tangential velocity components, we proceed to evaluate (2.51) in the normal
direction to find
m1 v n (t0 ) − m1 v1n (t− ) = P̂comp , m2 v n (t0 ) − m2 v2n (t− ) = −P̂comp ,

(2.53)
m1 v1n (t+ ) − m1 v n (t0 ) = P̂rest m2 v2n (t+ ) − m2 v n (t0 ) = −P̂rest .
These are four equations for five unknowns (the two final velocities, the velocity at t0 , and the
impulsive forces during the compression and restitution periods). One additional equation is needed.
Like for the particle–wall impact, we exploit the definition of the coefficient of restitution,
P̂rest
e= , e ∈ [0, 1] (2.54)
P̂comp
to solve the above system of equations, which yields
m1 v1n (t− ) + m2 v2n (t− ) + e m2 [v2n (t− ) − v1n (t− )]

v1n (t+ ) = (2.55)
m1 + m2
and

v2n (t+ ) = (2.56)
m1 + m2
As before, we distinguish three cases:
• elastic collision: the impulsive force is fully recovered. In this case P̂rest = P̂comp and the
solution of the complete system of equations yields
m1 v1n (t− ) + m2 v2n (t− ) + m2 [v2n (t− ) − v1n (t− )] m1 − m2 n 2m2

v1n (t+ ) = = v1 (t− )+ v n (t− ).
m1 + m2 m1 + m2 m1 + m2 2
and
m1 v1n (t− ) + m2 v2n (t− ) + m1 [v1n (t− ) − v2n (t− )] 2m1 m2 − m1 n
v2n (t+ ) = = v1n (t− )+ v (t− ).
m1 + m2 m1 + m2 m1 + m2 2
For example in the special case of equal masses m1 = m2 , we conclude v1n (t+ ) = v2n (t− ) and
v2n (t+ ) = v1n (t− ).
9 n t
Note that here and in the following we use superscripts and to avoid confusion with the subscripts 1 and 2.
57
• plastic collision: the impulsive force during the restitution phase vanishes. In this case
P̂rest = 0 and
m1 v1n (t− ) + m2 v2n (t− )
v1n (t+ ) = v2n (t+ ) = . (2.57)
m1 + m2
In this case the two masses continue with the same normal velocity (whereas their tangential
velocities may still be different). Also, in case of equal masses, the resulting normal velocity
is simply the average of the two initial velocities.
• partially elastic collision: collisions with 0 ≤ e ≤ 1 represent a partially elastic collision
(only some of the compressive impulse force is recovered during the restitution phase) and
relations (2.55) and (2.56) apply.
Note that we can rearrange the relations (2.55) and (2.56) for the general case 0 ≤ e ≤ 1 and use
conservation of momentum on the global system for the total time period of collision (i.e., adding
the compression and restitution phases in (2.53)):
m1 v1n (t+ ) + m2 v2n (t+ ) = m1 v1n (t− ) + m2 v2n (t− ) (2.58)
This is legitimate since no external forces act on the system of particles (only inner interaction forces
are present during the collision). This altogether leads to a re-interpretation of the coefficient of
restitution,
v2n (t+ ) − v1n (t+ )
e=− (2.59)
v2n (t− ) − v1n (t− )
as the ratio of the relative normal particle velocities after and before the collision. Solving (2.58)
and (2.59) for the normal velocities v1n (t+ ) and v2n (t+ ) after the collision again leads to (2.55) and
(2.56).
Finally, note that particle collisions are not conservative in general. Energy conservation only
applies for elastic collisions (e = 1). Otherwise, energy is dissipated (i.e., kinetic energy is consumed
by the permanent deformation of the particles).
————
Example 2.7. Particle collision with a rigid wall (revisited)
The collision of a single particle of mass m with a rigid wall (discussed in Section 1.2.4) can be
interpreted as a special case of a two-body collision, taking m1 = m and m2 → ∞ as well as
v2 (t) = 0 for all times t. Equation (2.55) here specializes to
v1n (t+ ) = lim = −e v1 (t− ), (2.60)
m2 →∞ m1 + m2 v2 (t− )=0
while we have for a frictionless wall

v1t (t+ ) = v1t (t− ). (2.61)
These are the same relations already derived in Section 1.2.4.
————
58
Example 2.8. Collision with a particle at rest

Consider two particles of masses m1 and m2 which are colliding elastically. m2
is initially at rest, while m1 hits with a normal velocity of v0 . What are the
m1 v0 m2
velocities after the collision?
Using v1n (t− ) = v0 and and v2n (t− ) = 0 as well as e = 1, relations (2.55) and (2.56) reduce to
m1 − m2 2m1
v1n (t+ ) = v0 , v2n (t+ ) = v0 . (2.62)
m1 + m2 m1 + m2
Therefore, particle 2 after the collision will always have a velocity in the same direction as the
initial velocity of particle 1 (thus moving away from the collision). The orientation of the velocity
of particle 1 after the collision depends on the two masses. If m1 > m2 , then both masses will
continue in the same direction. If m1 < m2 , then m2 will bounce back in the opposite direction.
In case of equal masses, we obtain
v1n (t+ ) = 0 v2n (t+ ) = v0 , (2.63)
i.e., the complete linear momentum is transferred from particle 1 to particle 2. Notice that, inde-
pendent of the two masses, the kinetic energy is conserved since m21 [v1n (t+ )]2 + m22 [v2n (t+ )]2 = m21 v02 .
————
2.3.2 Particles of variables mass: mass accretion and mass loss
Having discussed systems of particles, we are also in a position to study the kinetics of a particle
with a changing mass m(t). (Recall that so far we have always assumed the mass of each particle
remains constant over time.) This scenario is in fact a special case of the above two-particle collision.
If a particle is continuously accumulating mass by picking up infinitesimally small particles from
the surroundings, we speak of mass accretion. The opposite case is mass loss where a particle is
continuously losing mass by ejecting infinitesimally small particles to its environment.
The process of mass accretion is important when particles gain

or lose mass during their motion, e.g., when a rocket is propelled Dm
by exhausting gas, or when droplets grow while moving through a vm(t)
saturated medium. Consider a particle of mass m(t) and velocity
v(t), which continually acquires (i.e., accretes) mass.
v(t)
time t
Assume that, at time t, over a small time increment ∆t an additional
mass ∆m is gained by a particle which has an initial velocity v(t) and
m(t)
mass m(t), while the small additional mass ∆m has the initial velocity
time t +Dt
vm (t). Let us compare the linear momentum before and after mass
accretion, P (t) and P (t + ∆t), respectively. We start with the initial
linear momentum at time t, which is the sum of the linear momentum v(t+Dt)
of the particle and of the additional mass increment:
m(t +Dt)
P (t) = m(t)v(t) + ∆m vm (t). (2.64)
59
After the small increment ∆t, the particle has linear momentum P (t + ∆t) = m(t + ∆t)v(t + ∆t).
Since the time increment ∆t is small, we may apply a Taylor expansion (truncated to quadratic
order), so that the linear momentum at time t + ∆t is approximated by
P (t + ∆t) = m(t + ∆t)v(t + ∆t) = m(t) + ṁ(t)∆t + O(∆t2 ) v(t) + v̇(t)∆t + O(∆t2 )

(2.65)
= m(t)v(t) + ∆t [ṁ(t)v(t) + m(t)v̇(t)] + O(∆t2 ).
The rate of change of linear momentum is hence (dropping the t-dependence for brevity)
P (t + ∆t) − P (t) mv + ∆t (ṁv + mv̇) + O(∆t2 ) − (mv + ∆m vm )

Ṗ = lim = lim
∆t→0 ∆t ∆t→0 ∆t (2.66)
= ṁv + mv̇ − ṁvm = ṁ(v − vm ) + mv̇,
where we used that lim∆t→0 ∆m vm /∆t = ṁvm . Further, note that v̇ = a.

P
From Newton’s second axiom we know that i Fi = Ṗ . Therefore, equating the rate of linear
momentum change to the applied forces yields
X X
Fi = ma + ṁ(v − vm ) ⇔ ma = Fi + ṁvrel with vrel = vm − v, (2.67)
i i
which is also known as the rocket equation (the reason will become apparent in Example 2.9).
Of course, in the case ṁ = 0 the rocket equation reduces to the classical form of Newton’s second
law.
While the above derivation applies for mass accretion, we can extend
it to the case of mass loss. We start with a linear momentum of v(t)
P (t) = m(t)v(t) (2.68) time t

m(t)
and, after an infinitesimal time increment ∆t, end up with
Dm
vm
P (t + ∆t) = m(t + ∆t)v(t + ∆t) + ∆m vm , (2.69)
time t +Dt
where we have ejected a small particle of mass ∆m at velocity vm , so
that ṁ < 0. Performing the analogous Taylor expansion as in (2.66) v(t+Dt)
again leads to (2.67). m(t +Dt)
Therefore, the rocket equation (2.67) applies equally to mass accretion (ṁ > 0) and mass loss
(ṁ < 0), and of course also for ṁ = 0.
Example 2.9. Rocket motion (mass loss)
A rocket is moving into the x2 -direction and propelled by hot exhaust gas leaving the rocket
downward at a constant velocity vrel = c (relative to the rocket), i.e.,
vm − v = ce2 with c < 0. (2.70)
60
Since the hot gas is leaving the rocket, the rocket is losing mass Fdrag
over time (ṁ < 0). Therefore, linear momentum balance in the x2 -
direction is written as the rocket equation:
F2 = mẍ2 − ṁc ⇒ mẍ2 = F2 + ṁc, (2.71) m(t)

v
where F2 is the externally applied force in the vertical direction (which
mg g
usually contains the affects of gravity and drag, i.e., F2 = −mg − x
2
Fdrag ).
Note that ṁ < 0 and c < 0, so that ṁc > 0 is indeed accelerating the x1
rocket upwards. The term ṁc > 0 therefore represents the thrust
(perceived as an additional upward force). The lift-off condition is vm
|ṁc| > mg, since the initial drag force vanishes at zero initial velocity.
————
Example 2.10. Lifting a heavy rope off the ground (mass accretion)
Consider a long and heavy rope (or chain) of total mass m0 and total length l, which is initially
lying on the ground. A force F is applied to lift the rope gradually off the ground, such that it
is moving upwards (in the x3 -direction) with a constant acceleration a0 . What is the required
magnitude of force F ?
Let us denote by x3 (t) the current length of the rope that has lifted
off the ground at time t. We will use x3 (t) to conveniently describe
F
}
the motion of the rope. g
The known constant acceleration can be integrated along with the x3 m(t)
initial conditions that the rope is lying statically on the ground ini-
tially, i.e., x3 (0) = 0 and ẋ3 (0) = 0. Hence, the 1D kinematics of the
rope is given by
a0 2
ẍ3 (t) = a0 = const. ⇒ ẋ3 (t) = a0 t ⇒ x3 (t) = t . (2.72)
2
To calculate the force F , we invoke the balance of linear momentum, using its specific form of the
rocket equation since the mass of the moving rope changes over time. Specifically, at a height of
x3 and time t, the mass m(t) that has lifted off the ground at time t and is moving is given by
m0 m0 a0 2
m(x3 ) = x3 ⇒ m(t) = t (2.73)
l 2l
Using linear momentum balance in this scenario implies the kinetic relation
m(t)ẍ3 (t) = F (t) − m(t)g + ṁ(t) (vm − v) , (2.74)
where the last term stems from the changing mass. Since the part of the chain on the ground
remains at rest while the other part is moving at a constant velocity of ẋ3 , we have
m0 a0
vm − v = 0 − ẋ3 = −ẋ3 and also ṁ(t) = t. (2.75)
l
61
Overall, this leads to
F (t) = m(t)ẍ3 (t) + m(t)g + ṁ(t)ẋ3 (t). (2.76)
Inserting the known velocity and acceleration from (2.72) into this equation yields
m0 a20 2 m0 a0 g 2 m0 a20 2
F (t) = t + t + t
2l 2l l (2.77)
m0 a0 2 x3
= t (3a0 + g) = m0 (3a0 + g).
2l l
Therefore, the force F must grow linearly with chain length x3 . As a sanity check, note that under
static conditions (no chain motion, so a0 = 0), the force F must only balance gravity for a given
chain length, which is recovered here as a special case (F = mg = m0 gx3 /l).
————
The dynamics of systems of particles is an extension of single-particle dynamics. Importantly,

internal forces do not enter LMB and AMB, which may be written in terms of the motion of the
center of mass. The following box, which becomes part of our formula collection, summarizes all
key relations.
62
center of mass and total mass of a system of n particles:

n n
1 X X
rCM (t) = mi ri (t), M= mi
M
i=1 i=1
balance of linear momentum for M = const.:

n n n
X
int
X d X
Fi = 0, Fiext = Ṗ = (M vCM ) ⇒ Fiext = M aCM
dt
i=1 i=1 i=1
balance of linear momentum for a particle of varying mass m(t):

X
Fi = ma + ṁ(v − vm )
i
conservation of linear momentum:

n
X Xn
if Fiext = 0 ⇒ P = mi vi = const.
i=1 i=1
work–energy balance:
n n Z ri (t2 )
1X X
T (t2 ) − T (t1 ) = W12 with T (t) = mi |vi (t)|2 , W12 = Fi · dri
2 ri (t1 )
i=1 i=1
in the special case of rigid connections:

n Z
X ri (t2 )
W12 = Fiext · dri
i=1 ri (t1 )
balance of angular momentum with respect to point B:

n
X
MBext = ḢB + vB × P with P = M vCM , HB = rBPi × Pi .
i=1
coefficient of restitution for a two-particle collision:
v2n (t+ ) − v1n (t+ ) P̂rest

e=− =
v2n (t− ) − v1n (t− ) P̂comp
particle velocities after a two-particle collision:

v1n (t+ ) = , v1t (t+ ) = v1t (t− )
m1 + m2
v2n (t+ ) = , v2t (t+ ) = v2t (t− )
m1 + m2
63
3 Dynamics of Rigid Bodies
All examples and principles so far have relied on the assumption that the bodies of interest were
so small that they could be approximated as particles. This unfortunately does not apply to many
or most engineering applications. Consider, e.g., a wheel rolling down a hill – a particle has no
rotational DOFs, so rolling and sliding make no difference to a particle, yet there obviously is a
difference in the dynamics of a sliding vs. rolling tyre. Similarly, throw a book up in the air and
observe the influence of rotations about all three axes during the book’s flight – this cannot be
described by the flight of a particle. In such cases, we need to account for the size and shape of
the object of interest; we hence need to transition from a particle to a body. To begin simple, in
this section we only consider rigid bodies. A rigid body is undeformable (i.e., it can move freely
but cannot change its shape or size).
With a bit of imagination, we picture a rigid body as a collection
of infinitely many particles (or material points), packed together
infinitely close and having infinitesimally small masses. We may
mi
hence consider a system of n particles with masses ∆mi (i =
CM
1, . . . , n), and pass to the limit n → ∞ and ∆mi → 0 (at fixed +
relative particle distances) – the limit is a continuous body B. n®¥
If the distance between particles (or material points) remains mi®0 M
constant (consider rigid links between all pairs of particles), then
we have a rigid body.
In the following, we will revisit the dynamics of systems of particles for this special case
R of a
continuous body. In this limit, all discrete sums over particles turn into integrals ( ni=1 → B ), so
P
we will adjust our kinematic and kinetic definitions and relations accordingly.
3.1 Kinematics
The total mass of a rigid body B is defined as the limit of summing over infinitely many particles
of infinitesimal mass, or as the integral over the body:
X n Z Z
M = n→∞lim ∆mi = dm = ρ dV, (3.1)
∆mi →0 i=1 B B
where
∆m dm
ρ = lim = (3.2)
∆V →0∆V dV
denotes the mass density of body B. In the following, we will usually assume a spatially constant
mass density (ρ(x) = ρ = const.), although we may also consider cases of an inhomogeneous mass
distribution.
The center of mass follows analogously as

n
1 X 1 1
Z Z
rCM (t) = n→∞
lim ri (t) ∆mi ⇒ rCM (t) = r(t) dm = r(t)ρ dV (3.3)
∆m →0
M M B M B
i i=1
64
The kinematic description of continuous bodies is more complex than that of particles. Instead
of assigning the positions ri , velocities vi , and accelerations ai of n independent particles, we
now introduce continuous fields including, e.g, the position vector field r(x, t), the velocity field
v(x, t) and, analogously, the acceleration field a(x, t), which all depend on the position x ∈ B and
time t. In principle, each point of the body B could undergo a different motion through space
and time – however, points in a rigid body are physically coupled by the kinematic constraints of
rigidity of the body.
The motion of a rigid body as a whole – unlike that of particles – involves both translations and
rotations. Therefore, the motion of a rigid body in
2D is described by 3 degrees of freedom, including translations in the two in-plane directions,

e.g., x1 and x2 , and an in-plane rotation ϕ about the out-of-plane x3 -axis.
3D is described by 6 degrees of freedom, including translations in the three xi -directions as

well as three rotations about the three xi -axes.
It is important to note that every point P on the body is still described uniquely by its position
vector rP (t). One of the key tasks of rigid-body kinematics is to answer how the trajectories of
different points on a body are related to each other, as the body is moving through space and time.
For example, we may seek relations between the kinematic variables at two distinct points on the
body.
Consider a rigid body B that undergoes both translation and rotation,
and let us go from a point A ∈ B to a point B ∈ B, whose motion is
linked by x3
rB = rA + rAB rB = rA + rAB , vA A B
ṙB = ṙA + ṙAB ⇒ vB = vA + ṙAB , (3.4)

r̈B = r̈A + r̈AB aB = aA + r̈AB . rA
rAB
O
The term rAB and its time derivatives must be related to how the rigid x2
body is moving through space. Note that if the rigid body B is purely vB
rB
translating, then all points are undergoing the exact same motion, so x1
B
that vA (t) = vB (t) and aA (t) = aB (t) and the time derivatives of rAB
vanish. Consequently, the terms ṙAB and r̈AB must be associated with
a rotation of body B.
3.1.1 Rigid-body kinematics in 2D
Assume that a rigid body B is moving in 2D around a

w fixed point O (vO = 0). The velocity vector of any point
vA w(t)
A A on the body must lie in the 2D plane. In addition,
e3 rA
vB
since the body is rigid, we know from multiple-particle
O e rB
e1 2 B dynamics (see Section 2.2.2) that we must have
rOA · (vA − vO ) = rOA · vA = 0 ⇒ rOA ⊥vA .
65
We may thus write the (yet unknown) velocity of a point A on body B as
vA = ωA × rOA with some angular velocity vector ωA = ωA e3 . (3.5)
Since this holds for any arbitrary point A ∈ B, we may pick another point B ∈ B, whose velocity
may be written as
vB = ωB × rOB with ωB = ωB e3 . (3.6)
Since points A and B belong to the same rigid body B, their kinematics are coupled by
|rAB | = const. ⇒ rAB · (vA − vB ) = 0. (3.7)
When inserting (3.5) and (3.6) into (3.7), we obtain
0 = rAB · (ωA × rOA − ωB × rOB ) = (rOB − rOA ) · (ωA × rOA − ωB × rOB )

(3.8)
= rOB · (ωA × rOA ) + rOA · (ωB × rOB ) = (ωA − ωB ) rOB · (e3 × rOA ) .
Since the above needs to hold for any points A and B ∈ B, we conclude that
ωA = ωB = ω. (3.9)
That is, at any instance of time t, there is a unique angular velocity ω(t) for every point on the
rigid body, and every point on the body B rotates around O with the same angular velocity ω(t).
Now, consider two points A and B whose velocities are
vA = ω × rOA and vB = ω × rOB ,
so that
vAB = vB − vA = ω × (rOB − rOA ) = ω × rAB .
Remarkably, this implies that the relative velocity between A and B is independent of O, as given
by the so-called velocity transfer formula:
vB = vA + ω × rAB (3.10)
In words, every point in B is rotating around every other point in B with the same unique
(instantaneous) angular velocity vector ω(t). Therefore, if we know ω and the velocity at any
point within B, then we can uniquely determine the velocity at any other point of the same body
B.
Moreover, notice that we do not even need to assume that body B is rotating around a fixed point
O ∈ B. The above derivation also holds if point O is moving, since changing (3.5) and (3.6) to
vA = vO + ω × rOA , vB = vO + ω × rOB (3.11)
has no effect on (3.8) and (3.10) nor any of the conclusions drawn above, because vA − vB remains
unaffected and unchanged.
66
Now that point O does not need to be fixed, this brings up a question: can the motion of a body B
(that is rotating and translating through space) always be interpreted as a pure rotation around a
stationary point Π such that vΠ = 0. If so, it may be beneficial to locate point Π in order to more
easily calculate the velocity (and later acceleration) of points on the body.
To this end, assume that we know the velocity of an arbitrary point P ∈ B as vP = vP eP along
with the instantaneous angular velocity ω = ωe3 of body B. We may identify the location of our
sought center of rotation Π as follows. If Π exists, we must be able to write
vP = ω × rΠP ⇒ ω × vP = ω × (ω × rΠP ) . (3.12)
Exploiting that for a 2D rotation rΠP · e3 = 0, we have ω × (ω × rΠP ) = −ω 2 rΠP , so that
1
ω × vP = ωe3 × vP = −ω 2 rΠP = ω 2 rPΠ ⇒ rPΠ = e3 × vP (3.13)
ω
This tells us where point Π is located relative to point P. If we repeat the analogous construction
starting from a second point Q ∈ B, then by analogy
! 1 ! 1
rΠ = rP + rPΠ = rQ + rQΠ ⇒ rP + e3 × vP = rQ + e3 × vQ . (3.14)
ω ω
This relation, however, holds for any choice of P, Q ∈ B. To see this, we rearrange the above into
1 1 1
rP − rQ = e3 × vQ − e3 × vP ⇔ rQP = − e3 × (vP − vQ ). (3.15)
ω ω ω
Next, using the velocity transfer formula vP = vQ + ω × rQP gives
1
rQP = − e3 × (ωe3 × rQP ) = −e3 × (e3 × rQP ) = −(−rQP ), (3.16)
ω
which is obviously satisfied for any points P and Q on body B. Therefore, for any body B in 2D
there exists a unique point Π with respect to which each point on B undergoes a pure rotation. Note
that this point Π does not need to lie within B (no such assumption was made), and the location
of this point may change with time (so it is only instantaneous).
Point Π is called the instantaneous center of rotation (ICR) which may change with time
and which may lie inside or outside body B. The ICR is further characterized by the fact that its
velocity vanishes, i.e.,
vΠ = vICR = 0 (3.17)
We remark that relation (3.13) also allows us to identify the ICR graphically. If the velocity vector
vP at a point P is known, then we know from (3.13) that
rPΠ ⊥vP . (3.18)
That is, Π must lie on a ray perpendicular to the velocity vP . For example, if we only know the
(non-parallel) velocity directions of two distinct points A and B, then the ICR can be found as the
intersection of the two rays going through A and B and being perpendicular to the two respective
velocity vectors (see Example 3.1 below).
67
Example 3.1. Ladder sliding down a wall

A ladder of length l is leaning against a rigid wall under an angle ϕ,
as shown. Imagine that the ladder can slide frictionlessly, which
results in the ladder sliding down under the action gravity. Since the A P=ICR+
two end points A and B can move only vertically and horizontally,
j
respectively, the ICR of the ladder’s motion can easily be found by .
a graphical construction: by drawing the two lines perpendicular to vA P
the velocities and finding their intersection, which is point Π, the ICR. l
vP
To be quantitative, we use the velocity transfer formula with vΠ = 0, e2
which gives the two velocities as e1 B vB
vA = vΠ + ω × rΠA = ϕ̇e3 × (−l sin ϕ)e1 = −ϕ̇ l sin ϕ e2 ,

and
vB = vΠ + ω × rΠB = ϕ̇e3 × (−l cos ϕ)e2 = ϕ̇ l cos ϕ e1 . (3.19)
We can also derive the relation between the two speeds for any angle ϕ (assuming ϕ̇ 6= 0) as
vA |vA |
= = tan ϕ. (3.20)
vB |vB |
In addition, we can find the velocity of any point P on the ladder as
vP = ϕ̇|rP − rICR |, (3.21)
whose direction can be constructed graphically, knowing that vP ⊥(rP − rICR ), as shown above.
When graphically finding the ICR for each angle ϕ ∈ [0, π/2], we
obtain the so-called centrode as the line connecting all ICRs during
the course of the ladder sliding from the vertical to the horizontal +
configuration. The centrode is shown on the right (red line). +
+
+
The velocity transfer formula thus allows us to find the velocity of
any point on the ladder if, e.g., the angular velocity ϕ̇ = ω is known. +
Consider, e.g., that points A and B carry each a particle of mass m +
and the ladder (which is assumed massless and rigid) is released from +
rest close to the vertical position. Then ϕ̇ = ω can be calculated
by using the conservation of energy for frictionless sliding (we will +
come back to this problem when discussing energy principles for rigid
bodies).
————
Example 3.2. Slider crank mechanism
A slider crank mechanism transforms linear motion (e.g., of a piston) into circular motion (e.g.
of a wheel), or the other way around. It is frequently used, e.g., in engines. Shown below is an
example, in which the linear motion of point C is causing a circular motion of point B.
68
Points A and B, and B and C are each connected by B

rigid, massless rods linked through hinges. w1 w2
If the instantaneous velocity vC of point C in the l
shown configuration (with known angles α, β) is 2l
known, what is the velocity of point B? e2 a A
e1
e
To solve this problem, we could in principle use the 3 a
b
velocity transfer formula to determine vB , starting vC
from the known velocity vC . C
Unfortunately, the angular velocities of the two rods are unknown (and ω1 those are needed for
the velocity transfer formula). Furthermore, note that the two rods have, in general, two different
angular velocities, since points A, B and C do not lie on the same rigid body. Here, we use the
following strategy.
We may use the velocity transfer formula twice for point B, viz.
vB = vC + ω1 × rCB and vB = vA + ω2 × rAB , (3.22)
where ωi denotes the unknown angular velocity of rod i. The known distance vectors are
rCB = 2l cos β e1 + 2l sin β e2 , rAB = −l cos α e1 + l sin α e2 . (3.23)
The two equations in (3.22) provide two expressions for the sought velocity of point B, but either
equation contains one unknown angular velocity – hence, we need to find those angular velocities.
From the problem definition, we see that
vC = −vC e1 , vA = 0, ω1 = β̇e3 , ω2 = −α̇e3 . (3.24)
Note the correct signs of the angular velocities due to the definition of α and β (e.g., ω2 > 0 implies
that angle α becomes smaller, hence α̇ < 0, whereas ω1 > 0 implies β̇ > 0). Inserting the distance
vectors (3.23) and the velocity and angular velocity vectors (3.24) into (3.22) and equating the two
expressions for vB , i.e., imposing vC + ω1 × rCB = vA + ω2 × rAB , gives
−vC e1 + β̇e3 × (2l cos β e1 + 2l sin β e2 ) = −α̇e3 × (−l cos α e1 + l sin α e2 ) , (3.25)
which simplifies to
−vC e1 + 2lβ̇ (cos β e2 − sin β e1 ) = lα̇ (cos α e2 + sin α e1 ) . (3.26)
This in fact implies two scalar equations:
2lβ̇ cos β = lα̇ cos α and − vC − 2lβ̇ sin β = lα̇ sin α. (3.27)
Note that the first of these equations could also have been obtained from geometry directly since
2l sin β = a + l sin α ⇒ 2lβ̇ cos β = lα̇ cos α. (3.28)
69
Equations (3.27) can be solved for
2β̇ cos β
(3.29)
α̇ = ⇒ vC = − lα̇ sin α − 2lβ̇ sin β = −2lβ̇ (sin β + cos β tan α) ,
cos α
so that
vC 2 cos β −vC
β̇ = − ⇒ α̇ = . (3.30)
2l (sin β + cos β tan α) cos α 2l (sin β + cos β tan α)
Indeed, as drawn we should obtain ω1 = β̇ < 0 and ω2 = −α̇ > 0 for vC > 0.
Finally, the sought velocity of the joint at B is obtained from either of the two expressions in (3.22),
e.g.,
cos β (cos α e2 + sin α e1 )
vB = ω2 × rAB = lα̇ (cos α e2 + sin α e1 ) = −vC . (3.31)
cos α (sin β + cos β tan α)
————
Example 3.3. Two-dimensional rolling disk
Consider a disk of radius R that is rolling without sliding. Due to the kinematic constraint of
rolling, the rolled length Rϕ must match the distance x traveled by the disk, hence x = Rϕ, where
ϕ is measured clockwise in our example:
T 2Rw
CM x Rw
e2 +
A
+
j Rj
e1
e3 A A=P
Since the disk’s center of mass undergoes pure translation, we have xCM = xe1 = Rϕe1 and
consequently vCM = Rωe1 where ω = ϕ̇. The top point of the disk is moving with
vT = vCM + (−ω)e3 × Re2 = Rωe1 + (−Rω)(−e1 ) = 2Rωe1 (3.32)
and the bottom point in contact with the ground is moving with
vA = vCM + (−ω)e3 × (−R)e2 = Rωe1 − Rωe1 = 0. (3.33)
Hence the contact point A is also the instantaneous center of rotation (A = Π). The velocity of
point A after a rotation by ϕ is given by
vA = vCM + (−ω)e3 × [−R sin ϕ e1 − R cos ϕ e2 ] = Rωe1 + Rωe3 × [sin ϕ e1 + cos ϕ e2 ]
(3.34)
= Rωe1 + Rω [sin ϕ e2 − cos ϕ e1 ] = Rω [(1 − cos ϕ)e1 + sin ϕ e2 ]
so that
p p
vA = |vA | = Rω (1 − cos ϕ)2 + sin ϕ2 = Rω 2(1 − cos ϕ). (3.35)
70
As expected, the velocity vA vanishes periodically whenever ϕ = n · 2π with n ∈ Z (i.e., whenever

A is the bottom point and hence the instantaneous center of rotation in contact with the ground).
Its motion and velocity are shown below:
A
j
vA
2Rw
Rw
j
Note that the velocity of any other point P on the disk can easily be obtained both graphically and
analytically as
vP = vA + ω × rAP = (−ω)e3 × rAP . (3.36)
It is important to emphasize that the relation (3.36) applies only to points P that are on the disk
(e.g., the transfer formula cannot be used to compute the velocity between a point on the disk and
a point that is not on the disk but, e.g., on the ground).
————
3.1.2 Rigid-body kinematics in 3D
A rotation in 3D is uniquely defined by an axis of rotation (whose orientation w

we define by a unit vector eω with |eω | = 1) and an angle of rotation ϕ around
the axis of rotation in the sense defined by the orientation of eω (using the
rPA vA
right-hand rule). Consequently, a rotational motion of a body in 3D is defined P A
.
.
by an angular velocity vector rOP
rOA
ω(t) = ω(t)eω (t) with ω(t) ∈ R, |eω (t)| = 1. (3.37) O
We note that for general rigid-body motions in 3D eω (t) changes with time, so
we say that eω defines the instantaneous axis of rotation (IAR). IAR
The extension to 3D of the kinematic relations derived in Section 3.1.1 is now relatively straight-
forward. Instead of the instantaneous center of rotation (ICR), we now have an instantaneous axis
of rotation (IAR), around which the body is instantaneously rotating with an angular velocity ω(t).
Rotations in 2D are included as a special case for which eω (t) = e3 = const.
Pick any point O that lies on the IAR and consider a rotation with angular velocity ω = ωeω . For
any point A ∈ B we can identify an associated point P ∈ B which is the closest point on the IAR
to A (see the schematic above). Then, we may interpret the 3D rotation as an instantaneous 2D
rotation of point A around point P in the plane perpendicular to eω and going through A and P.
Hence, we may re-use the velocity transfer formula derived for 2D, and the velocity of point A is
vA = ωeω × rPA . (3.38)
71
Note that eω × rOP = |rOP |(eω × eω ) = 0, so that we may equivalently write
vA = ωeω × (rOP + rPA ) = ω × rOA . (3.39)
This shows that, in case of a 3D rotation with angular velocity vector ω, the velocity of any point
A ∈ B can be expressed as vA = ω × rΠA (like in 2D) with Π being any point that lies on the
instantaneous axis of rotation.
As a new complication in 3D, the IAR need not have vanishing velocity but it can translate with
a velocity vΠ = vΠ eω (since in any plane perpendicular to eω the body undergoes a pure rotation,
the IAR cannot translate in the plane – but, of course, it is free to move parallel to eω , which does
not affect the rotations). Because the body is rigid, every point on the IAR must move with the
same velocity vΠ . Hence, points on the IAR may move only parallel to the IAR. This, however,
does not alter any of our prior derivations for the following reason. Take vA = vΠ + ω × rΠA
and vB = vΠ + ω × rΠB , with vΠ = vΠ eω . Subtraction of the two equations eliminates vΠ – like
vO vanished in the 2D derivation (cf. Eq. (3.11) and its discussion). As a consequence, all of the
expressions derived above, especially (3.10), also hold in 3D, i.e., we may still apply the velocity
transfer formula
vB = vA + ω × rAB (3.40)
This shows that, in 3D as in 2D, every point on B is rotating around every other point on B with ω.
Also, if the velocity vA of a point A∈ B as well as ω is known, then the vector pointing from A
to the axis of rotation (yielding the vector from A to P in the above schematic) is given by (3.13)
(which is unaffected by any non-zero vΠ = vΠ eω since eω × vΠ = 0), so that
1
rAP = eω × vA , (3.41)
ω
which may again aid in the graphical construction of points on the IAR.
Having established the velocity transfer expression (which is identical in 2D and 3D), we can also
apply the very same transfer relation to the accelerations of two points A and B on the same body
B as follows. Starting with vB = vA + ω × rAB , differentiation with respect to time yields
d
aB = v̇A + (ω × rAB ) = aA + ω̇ × rAB + ω × ṙAB
dt (3.42)
= aA + ω̇ × rAB + ω × (vB − vA ) = aA + ω̇ × rAB + ω × (ω × rAB ) .
Altogether, the velocity transfer formula and acceleration transfer formula for two points
A and B on the same rigid body, valid in 2D and in 3D, have been obtained as, respectively:
vB = vA + ω × rAB
(3.43)
aB = aA + ω̇ × rAB + ω × (ω × rAB )
Thus, if we know the velocity (acceleration) at any point A ∈ B along with the angular velocity
vector ω, we can uniquely calculate the velocity (acceleration) at any other point B ∈ B.
72
Example 3.4. Rolling cone

A rigid cone of length L and end radius R is anchored by a joint x3
at height R and rolling on the ground without slipping with
.
f
an angular velocity φ̇ around the x3 -axis (as shown). What is
the (instantaneous) velocity of point C within the Cartesian
reference frame, for a given orientation φ of the cone? O
R
The motion of the cone is complex as it involves two rotations: x2
f L
on the one hand, it rotates around the x3 -axis with known
angular velocity φ̇; simultaneously, it rotates about its own R
x C
central axis as it rolls over the ground with unknown angular 1
velocity.
We will solve this problem in two ways to illustrate the concepts at play. First, we will solve the
problem in a general fashion, using the fixed Cartesian coordinate system shown above with a given
angle φ within the x1 -x2 -plane (Approach I). Second, we will re-solve the problem using a more
convenient coordinate system that rotates with the cone (Approach II).
x3
Approach I:
.
f w
.
w3=f
We know that we can always describe the motion of a rigid w1 a
body in 3D by a pure rotation about the IAR with an angular w2
velocity vector ω, plus possibly a translation that can only be O f
parallel to the IAR. Therefore, if we know any point Π on the R a
IAR, we may write the velocity of any point P on the cone as x2
f L
vP = vΠ + ω × rΠP . (3.44) R
x1 IAR
Unfortunately, we do not yet know the location of the IAR C rAC
nor the angular velocity vector ω. A
First, we know that point O is stationary: vO = 0. Second, we have learned from the rotating disk
(Example 3.3) that point A also has a vanishing velocity, vA = 0, since it is the ICR of the rolling
end of the cone. The velocity transfer formula between A and C tells us that
vA = vO + ω × rOA ⇒ ω × rOA = 0 ⇒ ω || rOA . (3.45)
Now we know at least the orientation of ω. Further, applying the velocity transfer formula again
for any other point P on the line connecting points A and O yields
vP = vO + ω × rOP = ω × rOP = 0 because rOP || rOA || ω. (3.46)
From this it follows that all points on the line connecting points A and O have a zero velocity.
Hence, the line through OA is the IAR of the cone (and, since vO = 0, the IAR has no parallel
velocity). We may draw the general conclusion:
If two distinct points on the same (rotating) body have zero velocity,
(3.47)
then the body’s IAR goes through those two points.
73
Now that we know the orientation of the IAR, we may exploit that we know the angular velocity
component around the x3 -axis, viz.
ω · e3 = φ̇, (3.48)
so that
 
ω1
ω = ω1 e1 + ω2 e2 + φ̇e3 or [ω]C = ω2  (3.49)
φ̇ Figure 3.10
with yet unknown components ω1 , ω2 ∈ R.

z
Since we know the direction of ω, we may write
rOA ˙
ω = ωeω with eω =
|rOA |
0
⇡ ↵
and R
2 y
e' !
rOA = L cos φ e1 + L sin φ e2 − Re3 . (3.50) er
C 00
x ! z = ˙ ez
Evaluating the components of vector rOA / |rOA | for angles
!r R
0 ≤ φ, α ≤ π2 (as drawn) and defining tan α = R/L, gives R
A
   
cos φ cos α ω1
!
[ω]C = ω  sin φ cos α  = ω2  , (3.51)
− sin α φ̇
which we may solve for

 
− cos φ cot α
φ̇
ω=− and [ω]C = φ̇  − sin φ cot α  . (3.52)
sin α
1
Now that we know ω, the velocity of point C follows from the velocity transfer formula, e.g.
vC = vA + ω × rAC = ω × rAC with rAC = Re3 + R(sin φ e1 − cos φ e2 ) (3.53)
so that
vC = φ̇ (− cos φ cot α e1 − sin φ cot α e2 + e3 ) × [Re3 + R(sin φ e1 − cos φ e2 )]
(3.54)
= Rφ̇ (cos φ − sin φ cot α)e1 + (cos φ cot α + sin φ)e2 + cot α e3 .
This completes the problem. The velocity of any other point on the cone can be obtained analo-
gously. Finally, keep in mind that these are all instantaneous velocities only valid in the current
configuration of the cone. For example, as the cone rotates, point A will lift off the ground and
assume a non-zero velocity while the new point in contact with the ground (A’) will define the new
IAR as the line connectin OA’, etc. Point C’s velocity will also change and, e.g., turn to zero once
point C touches the ground.
74
Approach II:
We can solve the above problem also in a more convenient
x3=x~3
fashion (the above is more general and therefore a helpful
~ w
.
.
reference; the following uses a simpler approach). f w3=f
~ a
We know that the cone rotates simultaneously around the ~ w1=-j
x2
.
e3 -axis with angular velocity φ̇ and around its central axis O
with an orientation R a
x2
f
eM
1 = cos φ e1 + sin φ e2 (3.55) L
j
and a yet unknown angular velocity ϕ̇ (where ϕ denotes x1 IAR
rAC
the angle of rotation of the cone about its central axis of
C
A x~1
symmetry in the shown orientation).
We may use kinematics to relate the angles of rotation. Since the cone is rolling in a frictionless
manner, the cone must satisfy the kinematic constraint
φ L = ϕR ⇒ φ̇ L = ϕ̇R. (3.56)
In other words, a rotation by φ implies that the cone’s end cross-section has traveled a distance φL
on the ground, and the motion of the center of mass of a disk was shown to move as ϕR (we chose
the orientation of ϕ so it is positive). In our chosen coordinate system, the total angular velocity
vector is thus
L M
ω = φ̇ e3 + (−ϕ̇)eM M
1 = φ̇ e3 − φ̇ e1 = φ̇ e3 − φ̇ cot α e1 with tan α = R/L. (3.57)
R
Inserting (3.55) leads to
 
− cos φ cot α
ω = φ̇ e3 − φ̇ cot α (cos φ e1 + sin φ e2 ) ⇒ [ω]C = φ̇  − sin φ cot α  , (3.58)
1
which is the same angular velocity found in Approach I above, see Eq. (3.51). The velocity vC
follows as in Approach I from applying the velocity transfer formula once again.
————
In Approach II of the above example we exploited one important observation, viz. that rotation
vectors are additive. That is, if a body is simultaneously rotating about multiple axes with unit
vectors gi and angular velocities ωi (i = 1, . . . , n), then the total angular velocity vector is
n
X
ω= ωi gi . (3.59)
i=1
So, we may add the individual angular velocity vectors to arrive at the total angular velocity vector
(as done, e.g., in Approach II of Example 3.4).
75
Example 3.5. Double-rotating funfair ride
A classical example of two simultaneous rotations is a funfair ride, in which a small circular cart is
rotating with a relative angular velocity ω2 around a joint P which itself is rotating with angular
velocity ω1 around a fixed pole O, as shown below. This case is simple since both rotations are in
the plane (hence the concept of an instantaneous center of rotation in 2D applies).
w2
P
+ A
P
w2
O
+ =
w1 +
P
w1
w1+w2
The combined rotation produces a velocity field on the cart that can indeed be described as a
rotation with angular velocity ω1 + ω2 around the ICR Π (which is neither point O nor point P);
instead it can be obtained graphically as shown above. The velocity of any other point A on the
body then follows as a rotation around Π with ω1 + ω2 . When using the velocity transfer formula,
we obtain the velocity of A as
vA = vO + ω1 × rOP + (ω1 + ω2 ) × rPA . (3.60)
Shown below is an analogous scenario with counter-rotating angular velocities (assuming ω1 > 0
and ω2 < 0).
P
+
P
+
w2
A
w2
O
+ =
w1 +
P
w1
w1+w2
76
It is important to point out that ω2 here is the relative angular velocity around the rotating pole P.
That is, ω2 is the angular velocity generated by a motor that is sitting on the arm OA and rotating
with ω1 around O. The absolute angular velocity of the cart is, as demonstrated above, ω1 + ω2 .
Hence, if the absolute angular velocity of the rotating cart (around its ICR) was given as, say, ω ∗ ,
then (ω1 + ω2 ) in (3.60) would have to be replaced by ω ∗ .
————
Example 3.6. Sliding ladder revisited

Let us revisit Example 3.1 of the ladder sliding down a wall with
some modifications. Assume that points A and B carry particles,
each of mass m, which slide without friction on the walls under the m
action of gravity. If the massless rod of length l is initially at rest A P
+
close to the vertical position, what are the accelerations of points A j g
and B, and what is the force in the rigid rod as a function of the
angle ϕ?
vA l
In order to find the accelerations of the two particles (or the force in e2 B
the rod, which is then obtained from linear momentum balance also e1 vB
e3 m
requiring the force in the rod) we could use the acceleration transfer
formula. Unfortunately, we however do not yet know the acceleration
of any point in the system.
In Example 3.1 we already found the two velocities as
vA = −ϕ̇ l sin ϕ e2 , vB = ϕ̇ l cos ϕ e1 . (3.61)
To find the yet unknown angular velocity ω = ϕ̇, we may use the conservation of energy, since the
system is conservative. Comparing the energy at some angle ϕ with angular velocity ϕ̇ with the
initial state at rest at ϕ ≈ 0 having only the potential energy of mass A at altitude l, we have
m 2 m 2
mgl = vA + vB + mgl cos ϕ. (3.62)
2 2
Inserting the two velocities from (3.61) leads to
m 22 2 m
mgl = ϕ̇ l sin ϕ + ϕ̇2 l2 cos2 ϕ + mgl cos ϕ, (3.63)
2 2
so that the angular velocity of the system is
g
ω 2 = ϕ̇2 = 2 (1 − cos ϕ). (3.64)
l
In order to calculate the accelerations and (ultimately) the inner force in the rod, we use the
acceleration transfer formula:
aB = aA + ω̇ × rAB + ω × (ω × rAB ) (3.65)
77
with
rAB = l sin ϕ e1 − l cos ϕ e2 , and ω = ω e3 , ω̇ = ω̇ e3 (3.66)
so that
aB = aA + ω̇l (sin ϕ e2 + cos ϕ e1 ) − ω 2 l (sin ϕ e1 − cos ϕ e2 ) . (3.67)
We now exploit that the directions of motion of the two masses are known, so we may define
aA = −aA e2 , aB = aB e1 . (3.68)
Insertion into (3.67) and decomposing the vector equation into its two components in the e1 - and
e2 -directions yields
aA = ω̇l sin ϕ + ω 2 l cos ϕ, aB = ω̇l cos ϕ − ω 2 l sin ϕ. (3.69)
The yet unknown angular acceleration ω̇ can be obtained by time differentiation of (3.64), giving
d 2 d h g i g g
ω = 2 (1 − cos ϕ) ⇒ 2ω ω̇ = 2 sin ϕ ω ⇒ ω̇ = sin ϕ (3.70)
dt dt l l l
(assuming that ω 6= 0 during the sliding of the ladder). Finally, inserting ω̇ and ω 2 from (3.64) and
(3.70), respectively, into (3.69) yields the sought accelerations as
aA (ϕ) = g 1 + 2 cos ϕ − 3 cos2 ϕ ,

aB (ϕ) = g (3 cos ϕ − 2) sin ϕ. (3.71)
From these two, the force in the rigid, massless rod can be calculated
from the balance of linear momentum applied to either of the two mg
masses. From free-body diagrams on each of the two masses we obtain H A
the linear momentum balance in the respective direction of motion as
j S
aA (ϕ) − g aA l
mg
maA = S cos ϕ + mg ⇒ S=m (3.72)
cos ϕ j B
S aB
or
aB (ϕ) V
maB = −S sin ϕ ⇒ S = −m . (3.73)
sin ϕ
Inserting the above accelerations aA and aB from (3.71) shows that both equations indeed result
in the same inner force in the rod, viz.
S(ϕ) = mg (2 − 3 cos ϕ) . (3.74)
As a quick sanity check, the force in the rod is S = −mg when ϕ = 0, i.e., the rod is under a
compressive load that equals the weight of mass A. Once the ladder slides, the force in the rod
grows continuously and turns positive. When ϕ = π/2 (the ladder has fully come down), then
S = 2mg, so the rod is under tension. The force vanishes in between when ϕ = arccos(2/3) = ϕ0
(in this configuration aA (ϕ0 ) = g, aB (ϕ0 ) = 0, and S(ϕ0 ) = 0).
The two accelerations as well as the force in the rod, along with the angular velocity and acceleration
of the system, are shown below for ϕ ∈ (0, π/2) (all quantities are normalized as labeled).
78
w/(g/l)1/2 aA/g
.
w/(g/l)
j
p/2
S/mg aB/g
j
p/2
Let us finally note that we can also formulate linear momentum balance on the complete system,
which can be used to compute, e.g., the two reaction forces from the ground, since
X
M aCM = Fiext = −2mge2 + V e2 + He1 .
i
with
1
M aCM = 2m (maA + maB ) = m (−aA e2 + aB e1 ) , (3.75)
2m
which leads to
V = m(2g − aA ), H = maB . (3.76)
The latter two could, of course, also have been obtained from linear momentum balance on the two
individual masses (using the respective either component not used above to compute the force in
the rod).
————
3.2 Kinetics
Having discussed the kinematics of rigid bodies in 2D and 3D, it is time to review the kinetic balance
laws of rigid bodies, which follow quite analogously from those derived for systems of rigidly linked
particles (see Section 2).
The total linear momentum vector of a body B is defined as

N
X Z Z
P = lim vi ∆mi = v dm = ρv dV. (3.77)
N →∞
i=1 B B
∆mi →0
By invoking the definition of the center of mass of a rigid body, cf. (3.3), we also have
Z
1
P =M v dm = M vCM . (3.78)
M B
79
Hence, as already discussed for systems of particles, we may interpret a rigid body as a “super-
particle” with an effective mass M and an effective velocity vCM .
As shown for systems of particles, the sum of all inner forces cancels in a rigid body (where all
material points are linked by pairwise rigid constraints), so that the linear momentum balance
for a rigid body reads
X d
Fiext = P. (3.79)
dt
i
If the total mass M of B remains constant (which we usually assume), then the above becomes
X
Fiext = M aCM if M = const. (3.80)
i
Note that in the following we usually consider only externally applied forces, so that we may omit
the superscript “ext” for convenience (and for analogy with the static case).
If no external forces act on a rigid body, we observe the conservation of linear momentum:
d d
P = (M vCM ) = 0 ⇔ M vCM = const. (3.81)
dt dt
The balance equation of angular momentum around a point B is, in principle, a straight-forward
extension of the relations derived in Section 2 for systems of particles. However, the devil is in the
detail, as we will see soon (and relations become more complex here).
The angular momentum vector of a rigid body with respect to a point B is defined as
n
X Z Z
HB = lim rBPi × (∆mi vi ) = (r − rB ) × v dm = (r − rB ) × ρv dV. (3.82)
N →∞
i=1 B B
∆mi →0
Taking the analogous limit – from a system of rigidly connected particles to a rigid body – yields
the balance of angular momentum for rigid bodies as
MB = ḢB + vB × P , (3.83)
where we need to consider only the resultant torque due to external forces acting on the rigid body
(for the same reasons discussed for systems of particles).
Unfortunately, calculating the integral definition of HB in (3.82) is in practice non-trivial in general

and calls for some discussion. Assuming that point B is part of body B, we can use the velocity
transfer formula to expand the above integral into
Z Z
HB = (r − rB ) × v dm = (r − rB ) × [vB + ω × (r − rB )] dm
ZB ZB (3.84)
= (r − rB ) dm × vB + (r − rB ) × [ω × (r − rB )] dm.
B B
80
Note that the first integral can be re-written as

Z Z Z
(r − rB ) dm × vB = r dm − rB dm × vB = M (rCM − rB ) × vB . (3.85)
B B B
Therefore, if we choose point B to be either a fixed point (so that vB = 0) or the center of mass of
body B (i.e., B = CM), then (3.85) vanishes and
Z
HB = r × (ω × r) dm with r = r − rB . (3.86)
B
The above may be reinterpreted as an integral of r × (ω × r) over body B, where r is now measured
in a coordinate system whose origin has been moved to point B (by moving all Cartesian axes in
parallel). By using the vector identity r × (ω × r) = (r · r)ω − (r · ω)r, we obtain
r × (ω × r) = (r · r)ω − (r · ω)r = [(r · r)I − r ⊗ r] ω, (3.87)
where we used the tensor/dyadic product10 . Note that ω is constant for the entire body B, so that
– when inserting (3.87) into (3.86) – ω can be taken outside the integral, so
Z Z
HB = [(r · r)I − r ⊗ r] dm ω = IB ω with IB = [(r · r)I − r ⊗ r] dm. (3.88)
B B
IB is the moment of inertia tensor. In component form, the above becomes

   
ω1 x1
2 2 2  
[r × (ω × r)]C = (x1 + x2 + x3 ) ω2 − (x1 ω1 + x2 ω2 + x3 ω3 ) x2 

ω3 x3
 2 (3.89)
x2 + x23 −x1 x2 −x1 x3
  
ω1
= −x1 x2 x21 + x23 −x2 x3  ω2  ,

−x1 x3 −x2 x3 x21 + x22 ω3
where we denoted the components of the angular velocity and position vectors in Cartesian coor-
dinates as, respectively,
[ω]C = (ω1 , ω2 , ω3 )T , [r]C = (x1 , x2 , x3 )T . (3.90)
Thus, if B = CM or vB = 0 and Cartesian coordinates xi are measured with respect to point B,

then the angular momentum vector in 3D can be expressed as
x22 + x23 −x1 x2 −x1 x3

 
Z
HB = IB ω with [IB ]C =  −x1 x2 x21 + x23 −x2 x3  ρ dV (3.91)
B −x1 x3 −x2 x3 x21 + x22
It is important to Keep in mind that the integration in (3.91) must be carried out with respect to
point B (following our definition of r), i.e., the coordinate origin must be B for integration.
10
The tensor product (or dyadic product) is defined as u ⊗ v = uv T . It yields a second order tensor whose
components (u ⊗ v)ij = ui vj form a square matrix.
81
When we insert (3.91) into (3.83) and recall (3.85), we arrive at the general form of angular
momentum balance for a rigid body with respect to a point B on the body:
d
MB = ḢB + vB × P = IB ω + M (rCM − rB ) × vB + vB × P (3.92)
dt
The above form of AMB simplifies significantly, if we choose point B wisely (and we are free to
choose, as long as B is on the body). Note the following special cases:
• if B is the center of mass, then rCM − rB = 0 and vB ||P so that

d
MB = (IB ω) if B = CM (3.93)
dt
• if B is a fixed point, then vB = 0 so that, again,
d
MB = (IB ω) if vB = 0 (3.94)
dt
• if the frame of reference is chosen such that I˙B = 0, then the above two cases reduce to
MB = IB ω̇ if B = CM or vB = 0, and if I˙B = 0 (3.95)
The integral definition of IB in (3.91) depends on the chosen coordinate system and is in general
time-dependent. In the following, we will first consider only cases in which I˙B = 0 so that ḢB =
IB ω̇. The case I˙B 6= 0 will be discussed in Section 3.4.
Recall that we had previously already encountered the 2D version of (3.95) for particles (see
Eq. (1.139)), viz.
MB = IB ω̇ = IB ϕ̈, (3.96)
and for a particle of mass m rotating at a distance R around B we obtained IB = mR2 . The above
presents the more general expression for rigid bodies in 3D.
To interpret relation (3.95), let us revisit our discussion for a single particle – shown below on
the left is the scenario of a single particle on a massless support. As established in Section 1.2.3,
AMB is indeed a reformulation of LMB, linking the torque to the angular acceleration of the
particle (like LMB links forces to accelerations), and IB was introduced back then as the constant
of proportionality in between: MB = IB ϕ̈.
MB MB j(t)
vP m vP
j(t) B R j(t) B B
massless
support
rigid body rigid body MB
82
Here, we have arrive at the analogous version for rigid bodies in 3D, Eq. (3.95). We can picture a
rigid body as a collection of infinitely many infinitesimally small material points that rotate about
a given axis. Comparing a particle (shown above on the left) to a rigid body (shown in the middle),
the principle is the same: for a rigid body, AMB tells us how an applied torque MB is related to
the collective angular acceleration ϕ̈ of all material points on the rigid body, whose shape and size
define the constant IB . As a novelty here, Eq. (3.95) is a vector/tensor equation, which comes from
the nature of 3D rigid bodies: depending on the axis about which they rotate, the value of IB is
different. For example, the sketch on the right shows the same rigid disk, now rotating about a
different axis. The mass inside the rigid body is distributed differently with respect to the two
shown axes (and the velocity profiles across the body are also different), so IB must be different
when rotating about either of the two shown axis.
Consider the shown disk rotates about each of its three principal axes, and let us assume that there
is a unique IB,i for each axis (i = 1, 2, 3). When we rotate with ωi about each axis with a torque
MB,i , we arrive at
MB,1 = IB,1 ω̇1 
MB,1
 
IB,1
 
ω̇1
MB,2 = IB,2 ω̇2 ⇔ MB,2  =  IB,2  ω̇2 , (3.97)
MB,3 = IB,3 ω̇3 MB,3 IB,3 ω̇3
| {z } | {z } | {z }
MB IB ω̇
so we arrive at MB = IB ω̇, which is Eq. (3.95). Unfortunately, the full story is more complicated
and IB is not always diagonal (which we will discuss in Section 3.2.3). Yet, (3.97) gives a quick
intuitive understanding for the reason of having a vector-valued relation in 3D.
Example 3.7. Moment of inertia tensor of a cylinder
Consider a cylinder of length L and radius R with a constant mass density ρ. By symmetry, the
center of mass is at the geometric center of the cylinder (i.e., the mid-point of the central axis of
rotation). Let us calculate the moment of inertia tensor of the cylinder with respect to its center
of mass, using a coordinate system with the x3 -axis aligned with its axis of rotational symmetry
(as shown below) and writing x1 = r cos ϕ and x2 = r sin ϕ to ease integration over the circular
cross-section:
 2
sin ϕ + x23 /r2 − cos ϕ sin ϕ − sin ϕ x3 /r

Z R Z 2π Z L/2
[ICM ] = ρ r2  − sin ϕ cos ϕ cos2 ϕ + x23 /r2 − cos ϕ x3 /r dx3 r dϕ dr.
r=0 ϕ=0 x3 =−L/2 − sin ϕ x3 /r − cos ϕ x3 /r 1
(3.98)
Exploiting that sin ϕ and cos ϕ as well as sin ϕ cos ϕ integrate to zero x2
over [0, 2π), we observe that all off-diagonal entries evaluate to zero.
With the total mass M = ρπR2 L, the remaining components evaluate
to
M 2 2
12 (L + 3R ) 0 0

R L
[ICM ] =  0 M 2 2
12 (L + 3R ) 0 . (3.99) x1
0 0 M R2
2
x3
83
As can be expected, the moment of inertia I33 against rotation about the x3 -axis, i.e., about the
cylinder’s axis of rotational symmetry, is much smaller than when rotating the body about the
length of the cylinder (I33 I11 , I22 when R L).
Since the derived [ICM ] is diagonal, the three non-zero entries are also the principal values of tensor
ICM , and our coordinate axes align with the principal directions of the cylinder. The above diagonal
form of ICM (which is a consequence of our choice of the coordinate system) is advantageous since
it decouples the equations of angular momentum balance. This is a consequence of
HCM = ICM ω = I11 ω1 e1 + I22 ω2 e2 + I33 ω3 e3 and I˙CM = 0, (3.100)
so that, with MCM = 3i=1 MCM,i ei ,

P
MCM = ḢCM = ICM ω̇ ⇒ MCM,i = I(ii) ω̇i for i = 1, 2, 3. (3.101)
This confirms that the torque [MCM ]i needed to alter the angular velocity ωi about the xi -axis is
directly proportional to Iii .
In the special case of a slender cylindrical rod with R L, the components of the moment of
inertia tensor further simplify to
M 2 M L2 M R2
I11 = I22 = (L + 3R2 ) ≈ , and I33 = . (3.102)
12 12 2
Notice that I33 is the moment of inertia of a disk of mass M and radius R in 2D rotating about
its center of mass, while I11 = I22 is the moment of inertia of a slender rod of mass M and
length L in 2D rotating about its center of mass.
Thus, for a slender rod of length L and mass M , rotating about its
center of mass in 2D, we have
L
M L2 CM
MCM = ICM ϕ̈ with ICM =
12
(3.103) j
while for a disk or cylinder of radius R and mass M , rotating about
CM
its center of mass in 2D, we have R j
M R2
MCM = ICM ϕ̈ with ICM = (3.104)
2
ICM , i.e., the moment of inertia around the center of mass in 2D is often referred to as the cen-
troidal moment of inertia.
84
3.2.3 Moment of inertia tensor
As the moment of inertia tensor IB is crucial for the balance of angular momentum of rigid
bodies,
P we will discuss some of its properties in more detail. Recall that a second-order tensor
T = 3i,j=1 Tij ei ⊗ ej in 3D is generally characterized by its components Tij (which form a 3 × 3
matrix) as well as the associated basis B = {e1 , e2 , e3 }. Unlike a matrix (which is defined only by
components and has no basis), the basis of a tensor is essential, since it allows us to apply certain
mathematical concepts (e.g., we can compute principal axes and associated eigenvalues, or we can
apply coordinate transforms to derive the components of the tensor in any other basis). All tensor
relations derived for, e.g., the stress, strain, and area moment of inertia tensors in Mechanics 1
and 2 apply here.
As shown above, the components of the moment of inertia tensor in a Cartesian reference frame
C (whose origin lies at the center of mass of body B) with respect to a point B that is either the
center of mass or not moving are given by
 2
x2 + x23 −x1 x2 −x1 x3

Z
[IB ]C =  −x1 x2 x21 + x23 −x2 x3  ρ dV. (3.105)
B 2
−x1 x3 −x2 x3 x1 + x2 2
Irrespective of the shape of the rigid body under considerations, we can draw the following conclu-
sions from the above form of the moment of inertia tensor:
• Symmetry: by definition the moment of inertia tensor is symmetric: IB = IBT .
• Additivity: since the integral over body B can be split additively into separate integrals over
various subdomains B i ⊂ B, the moment of inertia tensor of a composite body is additive.
That is, ICM of a composite body B = B 1 ∪ B 2 (and B 1 ∩ B 2 = ∅) is the sum of the individual
contributions, i.e.,
Z Z Z
IB = r × (ω × r) ρ dV = r × (ω × r) ρ dV + r × (ω × r) ρ dV = IB1 + IB2 . (3.106)
B B1 B2
This is equally true when considering hollow objects or bodies with inner holes, where IB
equals the complete moment of inertia tensor (of the solid body) minus the moment of inertia
tensor of the removed parts. It is important to note though that all added contributions IBi
must be with respect to the same point B.
For example, for a hollow cylinder (inner radius Ri , outer radius Ro ) the non-zero compo-
nents of the moment of inertia tensor from Example 3.7 are
Mo 2 Mi 2 e2
I11 = I22 = (L + 3Ro2 ) − (L + 3Ri2 ), Ro
12 12
(3.107) e1 = - Ri
Mo Ro2 Mi Ri2 2
I33 = − , Mo/i = ρπRo/i L.
2 2
As a special case, let us take the limit of a thin-walled hollow cylinder of radius R, thickness
t R, and length L. In this case we may assume Ri = R − 2t and Ro = R + 2t as well as,
85
for a homogeneous body of mass density ρ, M0 = ρπRo2 L and Mi = ρπRi2 L. In the limit of a
thin cylinder (t R), insertion of the radii and masses into (3.107) leads to
ρπRtL 2 ρπRtL M 2
2L + 3(4R2 + t2 ) tR → (2L2 + 12R2 ) = (L + 6R2 ),

I11 = I22 =
12 12 12
ρπRtL tR
I33 = (4R2 + t2 ) → ρ 2πRtL R2 = M R2
2
(3.108)
with the total (approximate) mass M = 2πR t L of the thin cylinder. As a sanity check,
notice that I33 agrees with the moment of inertia derived in Example 2.4 for a thin wheel in
2D, while I11 = I22 correctly reduces to the moment of inertia of a slender bar in the limit of
a long and slender cylinder (R L), cf. (3.103).
• Symmetry axes: if the xi -axis is an axis of rotational symmetry, then all off-diagonal terms
of I with i-subscript vanish (see, e.g., the I13 ,I23 -components in Example 3.7). The same
applies if xi stands normal on a symmetry plane (so the origin lies in the symmetry plane)
(see, e.g., the I12 = I21 -component in Example 3.7).
For example, the cylindrical cone shown on the right has rotational
symmetry about the x1 -axis, so I12 = I13 = I21 = I31 = 0. Also, x3
the x2 -axis is perpendicular to the x1 -x3 -plane, which is a symmetry
plane of the cone, so I21 = I23 = I12 = I32 = 0. Finally, the x2
analogous applies to the x3 -axis (being perpendicular to the x1 -x2 B
symmetry plane) so that I31 = I32 = I13 = I23 = 0. The only non-
zero components in the shown coordinate frame are hence I11 , I22
L
and I33 . Note that the same also applies, e.g., if the origin lies at
R
the center of mass of the cone (it generally holds as long as one of
the Cartesian coordinate axes is aligned with an axis of rotational x1
symmetry of the cone).
• Reduction to 2D: for 2D problems in the x1 -x2 -plane, the x3 -axis x3

is naturally perpendicular to a plane of symmetry, so that I13 =
I23 = 0. Further, the rotation is confined to the plane, so ω = ωe3 .
Therefore, in this reference frame C
w

I11 I12 0
  
0 0
 x2
w(t)
[HCM ]C = I12 I22 0   0  =  0  . (3.109)
0 0 I33 ω I33 ω x1
Hence, for 2D problems the angular momentum balance (if B = CM or vB = 0, and if I˙33 = 0)
becomes
Z Z
MB = I33 ω̇, with I33 = (x21 + x22 ) dm = r2 ρ dV. (3.110)
B B
I33 is often referred to as the axial moment of inertia.

As an example, let us revisit the moment of inertia of a slender bar in 2D (length L, density
ρ, cross-sectional area A) rotating about its center of mass in the plane (with x1 pointing along
86
the beam’s length and x3 being the out-of-plane axis, the origin is at the center). Applying
the definition of I33 from (3.110) again leads to the moment of inertia of a slender rod about
its center of mass:
Z L/2
(L/2)3 ρAL3 M L2
I33 = x21 ρ A dx1 = 2 ρA = = with M = ρAL. (3.111)
−L/2 3 12 12
We note that for a 2D rotation about the center of mass, one may also define
I33 = rg2 M, (3.112)
where M is the mass of the body and rg is the so-called radius of gyration. By way of
interpretation, the rigid body’s rotational inertia is equivalent to that of a particle of mass
M rotating at a √radius rg . For the above example of a slender bar rotating about its CM, we
see that rg = L/ 12.
• Relation to area moments: The moment of inertia tensor bears great similarity to the
area moment of inertia tensor introduced in Mechanics 2. For example, take the axial
moment I33 in (3.110) and consider the case of a uniform mass density ρ and thickness t. In
this case, the above reduces to
Z Z Z
2 2 2
I33 = ρ (x1 + x2 )t dA = ρt r dA = ρtIP , where IP = r2 dA (3.113)
B B B
is the polar moment of area already introduced in Mechanics 2 (where and x-y-coordinate
system was used, as shown for beam bending).
x2 x2
M F M
x1 x2 x1 x1 F
x3 x3
x1 x2
rotation vs. bending about the x1-axis rotation vs. bending about the x2-axis
Similarly, if we consider a thin plate of uniform thickness t and density ρ that is rotating with
ω in 3D with the x3 -axis pointing out-of-plane (so x3 ≈ 0) as shown in the two schematics
above, then we observe that the non-zero components in the moment of inertia tensor in
(3.109) become
Z Z
I11 = (x22 + x23 ) dm = ρt x22 dA = ρt Ix ,
ZB ZB
I22 = (x21 + x23 ) dm = ρt x21 dA = ρt Iy , (3.114)
B Z B Z
I12 = I21 = −x1 x2 dm = −ρt x1 x2 dA = ρt Cxy ,
B B
87
where Ix and Iy are the area moments of inertia and Cxy is the deviatoric moment known
from Mechanics 2. Therefore, the smaller (e.g.) the bending resistance about the x1 -axis
(characterized by Ix ), the smaller the rotational inertia about the x1 -axis (characterized by
I11 ). If Ix < Iy , then I11 < I22 . The above schematic illustrates the analogy between mass
and area moments of inertia.
• Principal axes: working with the general form of IB in (3.105) can be cumbersome, since it
implies a coupling of the three equations of angular momentum balance in 3D in general (if
all components of [IB ]C are non-zero, then IB ω = (I11 ω1 + I12 ω2 + I13 ω3 )e1 + . . . yields three
equations in which ω1 , ω2 and ω3 are coupled). Example 3.7, by contrast, had a diagonal
moment of inertia tensor, which decouples the equations of angular momentum balance (in
this case IB ω = I11 ω1 e1 + I22 ω2 e2 + I33 ω3 e3 decouples the angular velocities). We now
understand that this was the case Example 3.7, because the x3 -axis was aligned with the axis
of rotational symmetry (and because of the symmetries of the body).
Analogous to the stress, strain, and area moment of inertia tensors introduced in Mechanics 2,
we can always find a (principal) coordinate system Cˆ in which the moment of inertia tensor
is diagonal. This special system aligns with the principal axes of the body or, equivalently,
with the principal directions {ê1 , ê2 , ê3 } of the tensor. In the principal axes system with
basis {ê1 , ê2 , ê3 }, the moment of inertia tensor assumes the form
ˆ 
I1
[IB ]Cˆ =  Iˆ2 , (3.115)
Iˆ3
where the diagonal entries {Iˆ1 , Iˆ2 , Iˆ3 } are the tensor’s eigenvalues. Rotation of the moment
of inertia tensor can also be accomplished by the transformation rules (and, graphically, using
Mohr’s circle) as derived in Mechanics 2. As discussed above, axes of rotational symmetry
and axes perpendicular to planes of symmetry are always principal axes.
In practice, it is recommended to always align the coordinate system with the principal axes
or axes of symmetry, so one can work with a diagonal moment of inertia tensor and decouple
the angular moment components, whenever possible.
• Parallel axes theorem (Steiner’s theorem): we have seen that the components of IB
within a given reference frame C generally depend both on the geometry of the body and on
the reference point B. Since we cannot tabulate the components of I for a given geometry
for all possible points B, one usually finds only the components of ICM in textbooks. It will
hence be convenient to have a simple formula that tells us the relation between IB and ICM
for any arbitrary point B ∈ B.
Recall that IB was generally defined by the integral in (3.105) with respect to a reference
frame C having the coordinate origin at point B (assuming that vB = 0):
Z
IB = [(r · r)I − r ⊗ r] ρ dV with r = r − rB . (3.116)
B
If we shift the origin from B ∈ B to the CM of B by the distance vector
∆x = rB − rCM (3.117)
88
while keeping all coordinate axes parallel (i.e., we only move the origin
by translation without changing the orientations of the coordinate
axes, resulting in a new frame C),˜ then we may write for the new
frame of reference _ CM
r = r − rB = r − rCM + rCM − rB = r̃ − ∆x (3.118)
x2 + ~ r
Dx
_
with B _ r
x1
r̃ = r − rCM . (3.119)
Insertion into the integral yields
Z
IB = [(r̃ − ∆x) · (r̃ − ∆x) I − (r̃ − ∆x) ⊗ (r̃ − ∆x)] ρ dV
ZB Z
= [(r̃ · r̃) I − r̃ ⊗ r̃] ρ dV + [(∆x · ∆x) I − ∆x ⊗ ∆x] ρ dV
B Z B (3.120)
− [2(r̃ · ∆x) I − (r̃ ⊗ ∆x + ∆x ⊗ r̃)] ρ dV
B
= ICM + [(∆x · ∆x) I − ∆x ⊗ ∆x] M,
where the last integral vanished because ∆x = const. and
Z Z Z Z
r̃ρ dV = (r − rCM )ρ dV = r ρ dV − rCM ρ dV = M rCM − rCM M = 0, (3.121)
B B B B
so that
Z
[2(r̃ · ∆x) I − (r̃ ⊗ ∆x + ∆x ⊗ r̃)] ρ dV
B
Z Z Z (3.122)
=2 r̃ρ dV · ∆x I − r̃ρ dV ⊗ ∆x − ∆x ⊗ r̃ρ dV = 0.
B B B
Altogether, we have thus arrived at
IB = ICM + [(∆x · ∆x) I − ∆x ⊗ ∆x] M, (3.123)
T
which in component from, with [∆x]C = (∆x1 , ∆x2 , ∆x3 ) , evaluates to
(∆x2 )2 + (∆x3 )2
 
−∆x1 ∆x2 −∆x1 ∆x3
[IB ] = [ICM ] + M  −∆x1 ∆x2 (∆x1 )2 + (∆x3 )2 −∆x2 ∆x3  (3.124)
−∆x1 ∆x3 −∆x2 ∆x3 (∆x1 )2 + (∆x2 )2
This relation is the so-called parallel axis theorem, also known as Steiner’s theorem (or
Hyugens’ theorem). This is analogous to Steiner’s theorem already introduced in Mechanics 2
for the area moments of inertia.
For 2D rotations in the plane, Steiner’s theorem simply re-
CM
duces to
L/2
IB = ICM + M (∆x)2 B
where ∆x = |rB − rCM | is the shifted distance in the plane.
89
For example, consider a rod (R L) that is rotating about its end point B in 2D. Here,
2
M L2 M L2 L M L2
ICM = ⇒ IB = ICM + M (∆x)2 = +M = . (3.125)
12 12 2 3
As can be expected, the moment of inertia is larger when rotating about the rod’s end point
than when rotating about its center of mass. This indicates, by definition, that the axial
moment of inertia is always minimal with respect to the center of mass (IB ≥ ICM for all
B ∈ B).
As another example, a disk rotating about its instantaneous center of

rotation A, while rolling on the ground, has the moment of inertia CM
M R2 3 R
IA = ICM + M R2 = + M R2 = M R2 . (3.126)
2 2 A
If, by contrast, IA is known, we can also use Steiner’s theorem to find ICM = IA − M R2 .
• Examples: Table 1 (on the next page) lists example moment of inertia tensors for a few
frequently used geometric bodies with respect to the shown frames of reference.
Example 3.8. Moments of inertia in 2D
The rigid body shown on the right consists of a rod (length L, mass L
m3 ) and two point masses (m1 and m2 ) attached to its end points. m1 B
m2
The combined moment of inertia around point B is easily obtained
by exploiting the additivity of the moment of inertia: m3
2
rod 1 2 m3 L2 L
IB = IB + IB + IB = + (m1 + m2 ) . (3.127)
12 2
The second example is a rigid body in the shape of a circular disk (outer
A radius R) with three symmetrically arranged holes (each of radius r, dis-
tance from center to center a, as shown). The disk has a uniform mass
+ density ρ and uniform thickness t.
R The moment of inertia with respect to the center of mass (point C) is
C
+
a + 1
r IC = ICfull disk − ICholes = mfull disk R2 − 3 ICM
hole
+ mhole a2
2
1 2 2 1 2 2 2 2 πρt 4
R − 3r2 (r2 + 2a2 ) .

= (πR tρ)R − 3 (πr tρ)r + (πr tρ)a =
2 2 2
If the moment of inertia of the same disk with respect to point A is needed, the parallel axis theorem
helps us shift the moment of inertia to
IA = IC + mdisk R2 = IC + (πR2 tρ − 3πr2 ρt)R2 . (3.128)
————
90
x2 x2
r r
h x1 h x1
x3 x3
(a) solid cylinder: (b) hollow cylinder:

 3 2
  2 2

M 3r + h M 6r + h
[ICM ] =  6r2  [ICM ] =  6r2 
12 6
3r3 + h2 2
6r + h 2
x2 x2
r r
x1 x1
x3 x3
(c) solid sphere: (d) hollow sphere:
   
2M r2 1 2M r2 1
[ICM ] =  1  [ICM ] =  1 
5 3
1 1
x2 x2
a r
c
b h
x1 x1
O
x3 x3
(e) solid cuboid: (f) solid cone
 2 2
  2 2

M b +c M 3r + 12h
[ICM ] =  a2 + c2  [IO ] =  6r2 
12 20
a2 + b2 3r2 + 12h2
Table 1: List of moment of inertia tensors of frequently used bodies of total mass M .
91
Example 3.9. Moment of inertia of a cuboidal block

x2
Consider a block of extensions a, b and c with a uniform mass density x3 x2
ρ. If we compute the moment of inertia tensor with respect to point
x3
O shown on the right, then our coordinate axes are not aligned with
c
the symmetry axes of the body (none of the three symmetry planes of x1
the body falls onto xi = 0 for any of the coordinate axes). Therefore, CM
b
we may expect that IB is not diagonal.
O a x1
In fact, we have, e.g.,
Z cZ bZ a
a2 b2 M
Z
IO,12 = − x1 x2 ρ dV = −ρ x1 x2 dx1 dx2 dx3 = −ρ c = − ab (3.129)
B 0 0 0 2 2 4
with the total mass M = ρabc. If, by contrast, we move the coordinate origin to the center of mass
(CM) while keeping all axes parallel (as shown), then we see
Z Z c/2 Z b/2 Z a/2
ICM,12 = − x1 x2 dm = −ρ x1 x2 dx1 dx2 dx3 = 0. (3.130)
B −c/2 −b/2 −a/2
The latter can also be verified by Steiner’s theorem since (using ICM,12 = 0)
a b M
IO,12 = ICM,12 − M ∆x1 ∆x2 = −M − − = − ab. (3.131)
2 2 4
With respect to the center of mass, the axes are aligned with the symmetries of the cuboid, which
is why [ICM ] is diagonal and we have
M 2 M 2 M 2
ICM,11 = (b + c2 ), ICM,22 = (a + c2 ), ICM,33 = (a + b2 ). (3.132)
12 12 12
Note that, if the cuboid has two very short sides (e.g., a, b c and writing c = L), then we recover
the moment of inertia tensor for a slender rod around the x1 - and x2 -axes, since in this case
M 2 M 2 M 2
ICM,11 ≈ L , ICM,22 ≈ L , ICM,33 = (a + b2 ). (3.133)
12 12 12
Here, ICM,33 is the moment of inertia of a rectangular body in 2D rotating about its center
of mass.
————
92
Example 3.10. Pulley system
The system of two pulleys shown below is used to lift a heavy mass mw by applying a force F .
The pulleys have masses M and m, and radii R and r, as indicated, and the inextensible ropes are
approximately massless. What is the upward acceleration of mw for a given force F ?
F r
S1
m
r Mg S2
F S1
R y R
ICR
M S3
g f
S3
mW x
mWg
We begin by drawing free-body diagrams (see above); noting that the ropes are assumed massless,
the forces in the ropes (labeled Si ) are considered as shown.
We describe the kinematics of the system have the shown degrees of freedom x(t), y(t), φ(t) and
ϕ(t). We recognize that the ICR of the larger pulley is on the right, as shown. Further exploiting
the inextensibility of the rope leads to the kinematic constraints
rϕ = 2Rφ, y = Rφ, y = x, (3.134)
and hence
ẍ = Rφ̈, rϕ̈ = Rφ̈ = 2ẍ, ÿ = ẍ. (3.135)
Thus, having four degrees of freedom and three independent kinematic constraints implies that we
can reduce the kinematic description to a single degree of freedom only, e.g., x(t).
Next investigating the kinetics of the system, linear momentum balance on the heavy weight leads
to
mw ẍ = S3 − mw g. (3.136)
To avoid the reaction forces on the small pulley as well as force S2 , we use angular momentum
balance for the two pulleys (around its center for the small pulley, and around the ICR for the large
pulley):
mr2 3M R2
ICM ϕ̈ = ϕ̈ = (F − S1 )r, and IICR φ̈ = φ̈ = S1 2R − (S3 + M g)R. (3.137)
2 2
93
Inserting the kinematic relations from (3.135) leads to

3M
mẍ = F − S1 , and ẍ = 2S1 − (S3 + M g) and mw ẍ = S3 − mw g, (3.138)
2
which can be solved for the sought acceleration
4F − 2(mw + M )g
ẍ = . (3.139)
3M + 4m + 2mw
As a sanity check, equilibrium in the static case (assuming that all accelerations are zero) leads
to ẍ = 0 and hence 4F = 2(mw + M )g (i.e., in equilibrium the force F must only balance gravity
acting on the heavy weight and the larger pulley, while the smaller pulley is fixed to the cealing
and hence has no impact). Also, if the pulleys were massless (m = M = 0), then ẍ = −g + 2F/mw .
————
Example 3.11. Planetary gear (epicyclic gear)
Shown on the right is a planetary gear system. The outer ring gear rP
of radius rR is fixed. The inner sun gear of radius rS rotates about its P
sun gear
center of mass S. Its mass and centroidal moment of inertia are mS MT
and IS , respectively. The mass and centroidal moment of the planet rR S
gears (each of radius rP ) are mP and IP , respectively. All gears are rS

rolling on each other without slipping. If a torque MT is applied to P P
the sun gear, what is its angular acceleration ω̇S , if the planet gears planet gears
are free to move without load? Gravity may be neglected. fixed ring gear
Let us first describe the kinematics of the system. Shown on the left is
w the sun gear (fixed center of mass S) with only one of the planet gears
(center of mass P). Since the center of mass of the sun gear is hinged and
sun since the outer ring is fixed (cf. Example 3.3), we know that
w vS = 0, vC = 0, (3.140)
where the latter follows from rolling without slipping of the planet gear
on the outer ring gear. Contact points A and B of the two gears must be
moving with the same velocity for rolling without slipping, so
planet
vA = vB . (3.141)
This allows us to use the velocity transfer formula, leading to
!
vB = vC + ωP × rCB = vA = vS + ωS × rSA ⇒ ωP × rCB = ωS × rSA . (3.142)
Note that the exact location of the planet gear is irrevelant since points S-A-B-C always lie on a
straight line (of some orientation e). Therefore, we may write, e.g., rCB = 2rP e and rSA = −rS e,
which gives (with e3 denoting the out-of-plane basis vector)
rS
ωP e3 × 2rP e = ωS e3 × (−rS e) ⇒ 2rP ωP = −rS ωS ⇔ ωP = − ωS . (3.143)
2rP
94
Next, we inspect the kinetics of the system which

is driven by the torque M = MT e3 applied to the
sun gear. To this end, we draw free-body diagrams
of the sun gear and one of the planet gears (shown
on the right). For reasons of symmetry, we may
assume that the contact forces Fi between the sun
gear and each planet gear are the same (as drawn). MT
Rolling without slipping means that the gears in
contact may exchange both normal and tangential
forces (for which we must have |Ft | ≤ µ0 |Fn | to
avoid slipping for smooth surfaces; note that in reality
the gears have teeth, so slipping is prevented anyways).
Angular momentum balance on the sun wheel gives
MT − 3Ft rS = IS ω̇S . (3.144)

Angular momentum balance on the planet wheel is best formulated around point C, so that the
contact forces at point C vanish from the equation which becomes, using (3.143),
rS rS
−Ft · 2rP = IC ω̇P ⇒ −Ft · 2rP = −IC ω̇S ⇔ Ft = IC ω̇S , (3.145)
2rP 4rP2
with IC = IP + mP rP2 by Steiner’s theorem. Insertion into (3.144) finally yields the sought angular
acceleration:
rS2 4rP2 MT
MT − 3IC ω̇S = IS ω̇S ⇔ ω̇S = . (3.146)
4rP2 4rP2 IS + 3(IP + mP rP2 )rS2
The angular acceleration of the planet gear is obtained as

rS 2rP rS MT
ω̇P = − ω̇S = − 2 . (3.147)
2rP 4rP IS + 3(IP + mP rP2 )rS2
Based on the above results, we can also determine, e.g., the velocity of points P by the velocity
transfer formula, viz. vP = vC + ωP × rCP = ωP × rCP .
————
Example 3.12. Maximum acceleration of a car

Consider a car with either front or rear drive (i.e., only one axis is g
being driven by the engine), whose simplified version is shown on CM
+
the right. The car has a total mass M and the shown approximate rear front
h
extensions. For each of the two cases (front vs. rear drive), what is the
maximum acceleration of the car so the wheels roll without slipping? l/2 l/2
To identify the kinetics of the vehicle, we start with a free-body diagram (see below).
95
For the case of a rear drive, we have R2 ≈ 0 (no torque is driving

Mg
the front wheel, and we may assume the moment of inertia of
e2
e1 the wheels is small) and R1 6= 0 (the rear wheel is driven by
e3 a torque). Then, linear momentum balance for the car on the
CM
+ xCM,1 ground yields
h R1 R2 M ẍCM,1 = R1 , M ẍCM,2 = 0 = N1 + N2 − M g. (3.148)
N1 l/2 l/2 N2
Balance of angular momentum around the center of mass of the
car contributes the equation
l l 2h
ICM ω̇ = 0 = −N1 + N2 + hR1 ⇔ N1 − N2 = R1 , (3.149)
2 2 l
where we used that the car is not supposed to rotate about its CM, so ω̇ = 0 (i.e., no wheel is
allowed to get off the ground). For rolling without slipping we must have
|R1 | ≤ µ0 |N1 | ⇒ R1 ≤ µ0 N1 (3.150)
with the shown directions of the forces. From (3.148) and (3.149) we obtain
2h Mg h R1 µ0 M g
2N1 = M g + R1 ⇒ N1 = + R1 ≥ ⇔ R1 ≤ , (3.151)
l 2 l µ0 2(1 − µ0 h/l)
where we assumed 1 − µ0 h/l > 0. Insertion into linear momentum balance finally yields
R1 µ0 g
ẍCM,1 = ≤ = ẍmax,rear . (3.152)
M 2(1 − µ0 h/l)
Note that we may also check that N2 ≥ 0 to avoid lift-off of the front wheel of the car, which is
satisfied as long as (using the limiting case R1 = µ0 N1 as a worst-case scenario)
Mg
N2 = M g − N1 ≥ 0 ⇒ Mg − ≥0 ⇒ l ≥ 2µ0 h. (3.153)
2(1 − µ0 h/l)
Of course, we can repeat the entire procedure for a front-driven car, in which case R1 ≈ 0 and
R2 6= 0. Without showing the full derivation (which is analogous to the above), the maximum
acceleration in this case is obtained as
R2 µ0 g
ẍCM,1 = ≤ = ẍmax,front . (3.154)
M 2(1 + µ0 h/l)
A comparison shows that ẍmax,front < ẍmax,rear , so the front-driven car can tolerate less acceleration
before slipping occurs (when driving forward).
————
96
3.2.4 Angular momentum transfer formula
In our previous definition of IB we always assumed that point B was on the body. Let us also
consider the case that B is an arbitrary point. In this case we can exploit the center of mass
relations (writing rCB = rB − rCM for conciseness) to find that
Z Z
HB = (r − rB ) × v dm = [r − (rCM + rCB )] × v dm
ZB B
Z Z
= (r − rCM ) × v dm − rCB × v dm = HCM − rCB × [vCM + ω × (r − rCM )] dm
B Z B Z B
= HCM − rCB × vCM dm − rCB × ω × (r − rCM ) dm

B
ZB
= HCM − rCB × M vCM − rCB × ω × (r − rCM ) dm = HCM − rCB × M vCM ,
B
where we re-used relation (3.121). Next, inserting P = M vCM , we find for any arbitrary point B
that
HB = HCM − rCB × P = HCM + P × rCB . (3.155)
Likewise, for another point A we have
HA = HCM + P × rCA . (3.156)
Subtracting the two hence leads to
HB − HA = P × (rCB − rCA ) = P × rAB ⇒ HB = HA + P × rAB (3.157)
This is the angular momentum transfer formula for arbitrary points A and B, which do not
need to lie on the same body B. This has an important consequence: if we know HA for any
point A on the body, we can use (3.157) to find HB for any arbitrary other point B that is not
necessarily on the body. We highlight this principle in Example 3.13 below.
Example 3.13. Angular momentum transfer relations for a rotating disk
Consider a disk of radius R and mass M , which is rolling without slipping on the ground. We
previously derived the angular momentum of the disk in this scenario with respect to the center of
mass as (cf. (3.104))
M R2
HCM = ICM ω = ω. (3.158)
2
We further used Steiner’s theorem to show that angular momentum balance with respect to point A
on the disk in contact with the ground becomes (cf. (3.126))
3M R2
HA = IA ω = ω. (3.159)
2
97
The angular momentum transfer formula (3.157) confirms this since
HA = HCM + P × (rA − rCM ) (3.160)
so that, with vCM = vA + ω × (rCM − rA ) and vA = 0, we write

HA = ICM ωe3 + M vCM × (rA − rCM ) = ICM ωe3 + M [ωe3 × (rCM − rA )] × (rA − rCM )
3M R2 (3.161)
= ICM ωe3 + M ωe3 |rCM − rA |2 = ICM + M R2 ωe3 =

ωe3 .
2
Writing HA = IA ωe3 , this agrees with (3.160) as expected.
It is essential to recall that Steiner’s theorem only applies to points on the body but not e.g. to points
in contact with the body. The angular momentum transfer formula (3.157) gives us a convenient
way to involve points that are not on the body. For example, let us pick point A0 which is the point
on the ground instantaneously in contact with the disk (as shown in the above schematic). For this
point we have
HA0 = HCM + P × (rA0 − rCM ) = HCM + P × (rA − rCM ) = HA . (3.162)
Therefore, we can also calculate the angular momentum of the disk with respect to contact point
A0 on the ground, and it agrees with the one with respect to contact point A on the disk.
mg
Be careful when using angular momentum balance with re- A
spect to these two points! While the externally applied M g
torque (due to gravity on an inclined slope) is constant for e2
R x
point A, one must carefully define which point A0 is chosen e3 e
on the ground. Angular momentum balance generally reads 1
A'' R
N
MA0 = ḢA0 + vA0 × P , (3.163)
A' a
where
MA0 = (rCM − rA0 ) × mg with g = g (sin α e1 − cos α e2 ) (3.164)
in the shown reference frame, and
P = M vCM = M ẋe1 , (3.165)
where x denotes the disk’s travel distance down the slope, as shown.
We now need to make a choice. If we choose point A0 to be the moving point on the ground always
in contact with the disk (i.e., at each instance of time pick A0 as the contact point instead of a fixed
point in space), then
rCM − rA0 = Re2 , vA0 = ẋe1 6= 0, MA0 = −M gR sin α e3 = const., (3.166)
so that angular momentum balance (3.163) with HA0 = HA from (3.162) and P = M ẋe1 gives
−M gR sin α e3 = ḢA0 + vA0 × P = ḢA + ẋe1 × P = ḢA . (3.167)
98
Hence, the torque and the angular momentum with respect to A0 remain constant over time.
By contrast, if we choose a fixed point A00 on the ground, then the body moves away from A00 over
time. In this case,
rCM − rA00 = xe1 + Re2 , vA00 = 0, (3.168)
and
MA00 = (rCM − rA0 ) × (mg + N ) = −M g(R sin α + x cos α)e3 + xN e3 , (3.169)
where N = N e2 is the normal force from the ground onto the disk. The angular momentum
transfer formula here yields
HA00 = HCM + P × (rA00 − rCM ) = ICM ωe3 + P × (rA0 − rCM − xe1 ) = HA0 − P × xe1 . (3.170)
When exploiting HA = HA0 and P k e1 , angular momentum balance (3.163) with respect to the
fixed point A00 becomes
−M g(R sin α + x cos α)e3 + xN e3 = ḢA . (3.171)
If we further exploit linear momentum balance (in the direction normal to the ground), we obtain
N = M g cos α ⇒ −M gR sin αe3 = ḢA , (3.172)
so that (3.171) again reduces to (3.167). Hence, we arrive at the same angular momentum balance
when using point A on the disk, point A’ on the ground moving with the disk, and a fixed point
A” on the ground.
Finally, note that we could alternatively formulate angular momentum balance with respect to the
disk’s center of mass, which would avoid the complications stemming form the angular momen-
tum discussed above. Yet, this would necessarily involve the unknown tangential force R on the
ground (responsible for the rolling motion), so one would need an additional equation (e.g., linear
momentum balance.)
————
Example 3.14. Angular momentum balance with respect to the ICR
In many applications, like the rolling disk in Example 3.13, we may want to formulate angular
momentum balance with respect to the instantaneous center of rotation (ICR) of a body in 2D.
In this case we need to exercise caution since that point may not lie on the body (recall, e.g., the
sliding ladder of Example 3.1).
Angular momentum balance with respect to any point, therefore also with respect to the ICR,
reads
MICR = ḢICR + vICR × P = ḢICR , (3.173)
99
since the instantaneous velocity of the ICR vanishes by definition (vICR = 0). As the ICR does not
necessarily lie on the moving body, we invoke the angular momentum transfer formula to write
HICR = HCM + P × rCM−ICR , (3.174)
where rCM−ICR = rICR − rCM is the vector from the bodies center of mass to the ICR. We further
know that
HCM = ICM ω and P = M vCM = M ω × rICR−CM = −M ω × rCM−ICR , (3.175)
where we exploited that the velocity of any point on the body in 2D can be formulated as a pure
rotation about the center of mass. Insertion of (3.175) into (3.174) yields
HICR = ICM ω − (M ω × rCM−ICR ) × rCM−ICR . (3.176)
For a 2D rotation in the x1 -x2 -plane we know ω = ωe3 and ω⊥rCM−ICR and can further exploit
(3.109) to arrive at
2
d 2

HICR = ICM + M rCM−ICR ωe3 ⇒ MICR = ICM + M rCM−ICR ωe3 (3.177)
dt
with rCM−ICR = |rCM−ICR |. For example, in the case of a rolling disk of radius R, we have
rCM−ICR = R = const., so that angular momentum balance with respect to the ICR becomes
MICR = ICM + M R2 ω̇,

(3.178)
which agrees with are findings in Example 3.13. Note that, even though (3.178) looks like a simple
application of Steiner’s theorem, it is not in general, since the ICR does not lie on the body, so
Steiner’s theorem cannot be applied. What we have shown here is that it can indeed be applied
even to the ICR in 2D, as long as the distance rCM−ICR remains constant.
————
Like for particles, we can also formulate the work–energy balance law for rigid bodies. The total
kinetic energy of a rigid body B is obtained by integration:
1 2 1
Z Z
T = |v| dm = |vC + ω × (r − rC )|2 dm
B 2 2 B
(3.179)
1 1
Z Z Z
2 2
= |vC | dm + vC · ω × (r − rC ) dm + |ω × (r − rC )| dm.
2 B B 2 B
If we assume that C ∈ B and that point C is either fixed (vC = 0) or the center of mass (C = CM),
then the second term above vanishes so that
1 1 1 1
Z Z
T = |vC |2 dm + |ω × (r − rC )|2 dm = M |vC |2 + ω · IC ω, (3.180)
2 B 2 B 2 2
100
where the latter expression follows from (ω × r) · (ω × r) = [r × (ω × r)] · ω and using the definition
of HC , cf. Eq. (3.86). Hence, the kinetic energy of a rigid body B can be expressed solely
through the kinematics of a point C ∈ B, viz.
1 1
T = M |vC |2 + ω · IC ω if vC = 0 or C = CM (3.181)
2 2
We see that T has a translational kinetic energy contribution Ttrans = 12 M |vC |2 as well as a
rotational kinetic energy contribution Trot = 12 ω · IC ω.
Again taking the limit of infinitely many material points (or particles) stacked into body B, the
work–energy balance (2.19) for a rigid body becomes
XZ ri (t2 ) XZ t2
T (t2 ) − T (t1 ) = W12 , W12 = Fiext · dri = Fiext · vi dt, (3.182)
i ri (t1 ) i t1
where ri and vi = ṙi denote, respectively, the position and velocity of the point of attack of the
external forces Fiext . Further, we exploited that in a rigid body the internal forces perform no work
(as discussed for rigid links within systems of particles).
As before, if an external force is conservative, then we can exploit that the force derives from a po-
tential V and the work done by the force can be expressed as a potential difference (see Section 2.2.2
for a detailed discussion). If both conservative and non-conservative forces act simultaneously on
the body, then we may again exploit the additive structure of W12 . Specifically, we may still use
the difference in potential energy to compute W12 = V1 − V2 for each conservative force in the
system, while we only need to integrate the work done by those forces that are not conservative, so
non-cons. .
T (t2 ) + V (t2 ) − [T (t1 ) + V (t1 )] = W12
If all external forces are conservative, then the system is conservative and the overall (kinetic plus
potential) energy is conserved. The conservation of energy of a rigid body in the special case
of only conservative forces implies
T + V = const. (3.183)
Note that if gravity acts on a homogeneous rigid body B with center of mass CM, then the potential
energy due to gravity can conveniently be expressed as V = M ghCM ,
Z Z
V = g · rρ dV = g · rρ dV = M g · rCM , (3.184)
B B
where M is the total mass and rCM denotes the location of CM (the shape and current orientation
of the body are irrelevant here). For a 2D problem, the above simply becomes V = M ghCM with
hCM being the altitude of CM above the chosen zero level.
101
Example 3.15. Rolling disk on an inclined slope

We consider a circular disk (or cylinder) of mass M and radius R,
rolling without slipping on an inclined slope of angle α. Given an
g
initial rolling velocity v0 , how high does the body roll, and how does v =Rw
the answer vary if the body is sliding frictionlessly instead of rolling?
w
This problem is ideal to solve with the work–energy balance, which h ICR
allows us to compare the initial and final stages. The system is con- a
servative, so we may apply the conservation of energy.
The kinetic energy of the disk is most conveniently obtained either with respect to the center of
mass (CM) or with respect to the instantaneous center of rotation (ICR), both yielding the same
kinetic energy. Specifically, with |vCM | = |ω|R we obtain with respect to CM or ICR, respectively,
1 1 M R2 2 1 3M R2 2
T = M (ωR)2 + ω or T = ω . (3.185)
2 2 2 2 2
These two are obviously the same, as expected. Since we consider rolling without slipping, the
system is conservative. To find the maximum height to which the body will roll if it starts to roll
up the slope with an initial velocity |vCM | = v0 = Rω0 , we consider the potential energy due to
from gravity only, so that
3M R2 2 3M 2 3v02
V = M gh ⇒ ω0 = v = M ghmax ⇔ hmax = . (3.186)
4 4 0 4g
For comparison, we consider the case of a particle of mass m sliding up an inclined slope without
friction and without rolling. We saw in Example 1.15 that in this case
m 2 v02
v = mghmax ⇔ hmax = . (3.187)
2 0 2g
Hence, the travel distance smax = hmax cot α is larger in case of a rolling cylinder as compared to a
frictionlessly sliding one because of the access kinetic energy that the rolling cylinder initially has.
————
Example 3.16. Falling hinged stick
Consider a rod of length L and mass M , which is hinged to rotate M,L

A
g
freely (without friction) at one end (O) while the other end (A) is
free. Starting from the initial position ϕ = ϕ0 , the rod falls downward
w e2 j
under the action of gravity. Assuming that point O remains hinged,
what is the acceleration of point A during the fall? How does it
compare to a freely falling (not hinged) rod of mass M ? e3 e1
O
The acceleration of point A can be obtained from the acceleration transfer formula, viz.
aA = aO + ω̇ × rOA + ω × (ω × rOA ) (3.188)
102
where aO = 0 and rOA = L cos ϕ e1 +L sin ϕ e2 (where we defined ϕ to be positive counter-clockwise

and with respect to the horizontal e1 -axis). When inserting ω = ϕ̇e3 , this gives
aA = ϕ̈e3 × rOA + ϕ̇e3 × (ϕ̇e3 × rOA )
= ϕ̈e3 × (L cos ϕe1 + L sin ϕe2 ) + ϕ̇e3 × (ϕ̇e3 × (L cos ϕe1 + L sin ϕe2 ))
(3.189)
= Lϕ̈(cos ϕe2 − sin ϕe1 ) − Lϕ̇2 (cos ϕe1 + sin ϕe2 )
= −L(ϕ̇2 cos ϕ + ϕ̈ sin ϕ)e1 + L(−ϕ̇2 sin ϕ + ϕ̈ cos ϕ)e2 .
So far we have only used kinematics, but we still need to find ϕ̇ and ϕ̈ as functions of ϕ. The
former can be found by using the conservation of energy (no external forces do any work on the
system; gravity is conservative). Comparing the initial state with ϕ̇0 at ϕ0 gives:
1 L 1 L 1
IO ϕ̇20 + M g sin ϕ0 = IO ϕ̇2 + M g sin ϕ with IO = M L2 . (3.190)
2 2 2 2 3
If the pendulum is released from rest (ϕ̇0 = 0), then
11 L 3g
M L2 ϕ̇2 = M g (sin ϕ0 − sin ϕ) ⇒ ϕ̇2 = (sin ϕ0 − sin ϕ). (3.191)
23 2 L
Next, we need to find the angular acceleration ϕ̈. For example, we may use angular momentum
balance (which is free of the reaction forces when taken around O):
1 L 3g
M L2 ϕ̈ = −M g cos ϕ ⇒ ϕ̈ = − cos ϕ. (3.192)
3 2 2L
Note that the same could also have been obtained from differentiating ϕ̇2 in (3.191) with respect
to time. Inserting the angular velocity and acceleration into aA finally gives
3g 3g
2(sin ϕ0 sin ϕ − sin2 ϕ) + cos ϕ2 e2 .

aA = − cos ϕ [2(sin ϕ0 − sin ϕ) − sin ϕ] e1 − (3.193)
2 2
For example, when starting from the horizontal position (ϕ0 = 0), this reduces to
9g 3g
1 − 3 sin2 ϕ e2 .

aA = cos ϕ sin ϕ e1 − (3.194)
2 2
The magnitude of the acceleration can be rearranged to

r
3g 1
aA = |aA | = (5 − 3 cos 2ϕ),
2 2
while the vertical component aA,2 = aA · e2 is |a|/g
3g a2/g
1 − 3 sin2 ϕ .

aA,2 = −
2
j
Therefore, the acceleration of point A is | − 3g/2| > g upon
release (at ϕ − 0), after which it decreases until it inverts its
sign before reaching the bottom at ϕ = − π2 (see the plot on
the right). Note that there are indeed points that accelerate
faster than with gravity.
————
103
Example 3.17. Planetary gear revisited
As a final example of the work–energy balance of rigid bodies, let us rP

revisit Example 3.11 and solve it by using the work–energy balance. P
sun gear
The total kinetic energy of the gear at any time t is MT
rR S
1 3
T (t) = IS ωS2 (t) + IC ωP2 (t), (3.195) rS
2 2 P
P
where we picked the fixed points S (on the sun wheel) and C (on the planet gears
planet) wheel, so that only rotational energy is to be considered. fixed ring gear
Assuming that the gear is accelerated from rest (so ωS = ωP = 0 at t = 0) and equating the energy
difference to the work done by the constant external torque MT , work–energy balance becomes
Z t
1 3 1 3
T (t)−T (0) = IS ωS2 (t)+ IC ωP2 (t) = MT dϕS ⇒ IS ωS2 (t)+ IC ωP2 (t) = MT ϕS (3.196)
2 2 0 2 2
Here, we needed the work done by an external torque. The latter can be interpreted as the work
done by a tangentially applied force F such that MT = F × rCT and F ⊥rCT so MT = F rCT with
rCT = |rCT |. Exploiting the circular movement with dr = rCT dϕ and F k dr:
Z r2 Z ϕS Z ϕS Z ϕS
W12 = F · dr = F rCT dϕ = F rCT dϕ = MT dϕ = MT ϕS . (3.197)
r1 0 0 0
In general, we note that the work done by a torque MT (t) on a body rotating with ϕ(t) is
Z ϕ(t2 ) Z t2
W12 = MT · dϕ = MT · ω dt (3.198)
ϕ(t1 ) t1
This relation is analogous to the work derived for forces.
Differentiating (3.196) with respect to time and using ωS = ϕ̇S leads to (for ωS 6= 0)

ωP ω̇P
IS ωS ω̇S + 3IC ωP ω̇P = MT ωS ⇔ IS + 3IC ω̇S = MT , (3.199)
ωS ω̇S
which, upon insertion of

rS rS ωP ω̇P rS
ωP = −ωS and ω̇P = −ω̇S ⇒ = =− (3.200)
2rP 2rP ωS ω̇S 2rP
from the kinematics of Example 3.11, yields the same solution for the sun gear’s angular acceleration
as derived there, cf. Eq. (3.147).
————
104
3.3 Collision of Rigid Bodies
In Section 2.3 we discussed collisions of particles.

Here, we extend the relations found there to rigid
w2 bodies. Specifically, we here treat the general
v CM1 ^ P^
P v case of an eccentric, oblique collision. A collision
dCM1S CM2
dCM2S
CM1 S1 S2 is called an eccentric collision if the impulsive
CM2 forces exchanged between the bodies are not on
w1 the line connecting the centers of mass of the par-
ticipating bodies.
By definition, particle collisions were always centric, as particles have negligible dimensions, so any
force must be directed towards the center of mass of the particles. For rigid bodies, that constraint
is no longer in place. A collision is further called an oblique collision if the impulsive force
exchange is at an angle to the line of impact (e.g., shown above are two particles colliding with
initial velocities that are not aligned with the impulsive forces).
To keep things simple, we here and in the follow-
ing assume that the colliding bodies are convex.
tangent
tangent
In this case, every collision – irrespective of the
initial velocities of the bodies – has unique normal
and tangent directions in 2D (and a unique nor- normal
mal direction and tangent plane in 3D). The nor- normal .
.
mal is given by the local (common) normal vector
perpendicular to each body’s surface, while the
tangent direction (or plane) is defined by being
perpendicular to that normal.
Recall that a collision was defined as a contact over a short time window [t− , t+ ], decomposed into
compression and restitution phases. For particles, we used the conservation of linear momentum
during the collision along with the coefficient of restitution to find the particle velocities after the
collision (see Section 2.3).
Here, that concept is extended to rigid bodies, which requires accounting for rotations. For sim-
plicity, let us consider the 2D scenario schematically shown above, where the collision between two
bodies 1 and 2 happens at a point S in the plane, so that there is a pair of forces P (t) and −P (t)
(having the same magnitudes but opposite directions by Newton’s third axiom) which act onto the
two bodies, respectively, at points S1 and S2 , as shown. For an eccentric collision, S1 , S2 and the
centers of mass CM1 and CM2 do not fall onto a straight line (as shown above; the collision is
eccentric only if dCM1 S = dCM2 S = 0).
Like for particles, forces ±P (t) are time-dependent, and we can eliminate the time-dependence
in the equations by considering the complete collision period [t− , t+ ]. To this end, we integrate
the equations of linear momentum balance and (new here) angular momentum balance in time
analogously to Section 2.3 over the complete collision period, again defining an impulsive force as
Z t+
P̂ = P (t) dt. (3.201)
t−
105
Integration of linear and angular momentum balance for body 1 then yields
P̂ = m1 [vCM1 (t+ ) − vCM1 (t− )] , rCM1 S × P̂ = ICM1 [ω1 (t+ ) − ω1 (t− )] (3.202)
where rCM1 S denotes the vector pointing from body 1’s center of mass to the collision point S1 .
Note that we assume the collision is sufficiently short, so we may assume that I˙CM = 0 for both
bodies.
For body 2, the integrated balance laws of linear and angular momentum require
−P̂ = m2 [vCM2 (t+ ) − vCM2 (t− )] , rCM2 S × (−P̂ ) = ICM2 [ω2 (t+ ) − ω2 (t− )] . (3.203)
For the 2D scenario shown above, the conservation of angular momentum for both bodies further
reduces to
−dCM1 S P̂ = ICM1 [ω1 (t+ ) − ω1 (t− )] , dCM2 S P̂ = ICM2 [ω2 (t+ ) − ω2 (t− )] . (3.204)
In case of smooth, frictionless surfaces, the same assumption applies as for particles, viz. that
forces are exchanged and hence momentum is transferred between the two bodies only in the
direction normal to the plane of impact (n ), so that P̂ = P̂ en and the tangential components of
vCM1 and vCM2 remain unchanged during the collision11 :
t t t t
vCM1
(t+ ) = vCM1
(t− ), vCM2
(t+ ) = vCM2
(t− ) (3.205)
By contrast, in the normal direction we have

n n
n n

P̂ = m1 vCM 1
(t + ) − v CM1
(t− ) , −P̂ = m2 v CM2
(t+ ) − v CM2
(t− ) . (3.206)
Equations (3.204) and (3.206) provide four equations for five unknowns: ω1 (t+ ), ω2 (t+ ), vCMn (t+ ),
1
n
vCM2 (t+ ), and P̂ . The missing equation is given by the definition of the coefficient of restitution,
which is expressed – analogously to the case of particles – in terms of the relative normal velocities
of the contact point S (seen as S1 and S2 from body 1 and 2, respectively):
vSn2 (t+ ) − vSn1 (t+ )

e=− (3.207)
vSn2 (t− ) − vSn1 (t− )
This is the (negative) ratio of the relative separation velocities of the contact points immediatley
after and before the collision (cf. (2.59) for particles). The involved velocities of the two contact
points are related to the motion of the two bodies via the velocity transfer formula:
vS1 = vCM1 + ω1 × rCM1 S , vS2 = vCM2 + ω2 × rCM2 S , (3.208)
which in our 2D scenario becomes (as drawn)
vSn1 (t) = vCM

n
1
(t) − dCM1 S ω1 (t), vSn2 (t) = vCM
n
2
(t) + dCM2 S ω2 (t). (3.209)
11 n t
To avoid confusion with the other indices used here, we use superscripts and instead of subscripts for the
normal and tangential directions, respectively.
106
Evaluating the above velocities at times t− and t+ altogether yields a system of equations that can
be solved for the velocities and angular velocities of the two bodies after the collision.
If, by contrast, the surfaces are rough and hence not frictionless so no slipping occurs, then
the two bodies stick together at point S during contact. As a consequence, there is a non-zero
tangential component of P̂ and we can no longer assume that impulsive forces act only in the
normal direction. In this case, we have the equations analogous to (3.206) in both the normal and
tangential directions, viz.,
P̂ n = m1 vCM
n n
−P̂ n = m2 vCM
n n

1
(t+ ) − vCM 1
(t− ) , 2
(t+ ) − vCM 2
(t− ) ,
(3.210)
P̂ t = m1 vCM
t t
−P̂ t = m2 vCM
t t

1
(t+ ) − vCM 1
(t− ) , 2
(t+ ) − vCM 2
(t− ) .
We need one further equation (since we also have the unknown P̂ t -component), which is given by
the no-slip condition
vSt 1 (t+ ) = vSt 2 (t+ ). (3.211)
This again closes the system of equations.
We close by noting that the above scenario applies to general collisions between two rigid bodies,
and the integrated balance equations of linear and angular momentum may vary based on the
particular scenario of investigation. We here outlined the general approach to arrive at the required
governing equations to solve a collision problem, while we do not present general solutions (as done
for particles in Section 1.2.4), since those are specific to a given problem; see, e.g., Examples 3.18
and 3.19 below.
Example 3.18. Hitting a hanging pendulum with a ball

We consider a hanging pendulum of mass M and length L, which has e2
the shape of a long, slender rod and is initially at rest. The pendulum
is hit by a ball (a particle of mass m) with a horizontal velocity v0 O e1
immediately before the collision. As sketched on the right, the ball hits w
the pendulum at point P at a distance rOP from the hinge O. Let us
assume a frictionless collision with a known coefficient of restitution e rOP
and neglect gravity. What are the velocity v+ of the ball and the angular L
velocity ω+ of the pendulum immediately after the collision?
v0 P
+
Since there is no external torque acting on the system with respect to O, m
angular momentum with respect to O is conserved during the collision. M
Equating the angular momentum before and after the collision with respect to O gives
rm (t− ) × Pm (t− ) + IO ω(t− ) = rm (t+ ) × Pm (t+ ) + IO ω(t+ ) (3.212)
where rm and Pm denote, respectively, the position and (linear) momentum of the ball, ω is the
angular velocity of the pendulum, whose moment of inertia tensor with respect to point O is denoted
by IO . We furthermore know that
rm (t− ) = rm (t+ ) = −rOP e2 , Pm (t− ) = mv0 e1 , Pm (t+ ) = mv+ e1 (3.213)
107
and
ω(t− ) = 0, ω(t+ ) = ω+ e3 . (3.214)
Inserting these relations as well as the moment of inertia tensor for a slender rod (e.g. from Exam-
ple 3.7 with R L) transforms (3.212) into
M L2
−rOP e2 ×mv0 e1 = −rOP e2 ×mv+ e1 +IO ω+ e3 ⇒ mrOP v0 = mrOP v+ + ω+ (3.215)
3
As we need a second equation to determine v+ and ω+ , we introduce the definition of the coefficient
of restitution for a frictionless collision from (3.207). Here, we take contact point S2 as point P on
the pendulum, while S1 is the opposite contact point on the ball. This yields
rOP ω+ − v+
e=− ⇒ v+ = rOP ω+ − ev0 . (3.216)
0 − v0
Insertion into (3.215) gives
M L2 M L2

2
mrOP v0 = mrOP (rOP ω+ − ev0 ) + ω+ ⇔ rOP (1 + e)v0 = rOP + ω+ (3.217)
3 3m
and after some rearrangement

2
M L
1+e v0 1 − e 3m rOP
ω+ = 2 ⇒ v+ = rOP ω+ − ev0 = 2 v 0 . (3.218)
rOP

M L M L
1+ 3m rOP 1+ 3m rOP
Note that, if e = 0 (fully plastic collision, cf. Section 1.2.4), then v+ = rOP ω+ , so the ball continues
after the collision at the same speed as the contact point P on the pendulum. As a further sanity
check, we observe that ω+ is always positive for v0 > 0 (the pendulum rotates to the right after
being hit), whereas the sign of v+ depends on the ratios M/m and L/rOP as well as on e.
For example, for an elastic collision (e = 1) and a hit at rOP = 2L/3, we obtain
3M
3 v0 1− 4m
ω+ = 3M L
and v+ = 3M
v0 . (3.219)
1 + 4m 1+ 4m
If the pendulum is heavy (M > 34 m), then the ball bounces back (v+ < 0). If the pendulum is light
(M < 34 m), the ball continues in the same direction at slower speed (0 < v+ < v0 ). In the special
case M = 43 m the ball transfers its complete kinetic energy onto the pendulum and remains at rest
after the collision. Finally, in case of a very heavy pendulum (M → ∞), the pendulum remains at
rest (ω+ → 0), while the ball bounces back with v+ = −v0 , as expected from an elastic collision
with a rigid wall.
We close by noting that the analogous solution can be found when the ball is initially at rest and
hit by a pendulum of an initial angular velocity.
————
108
Example 3.19. Sweet spot of a baseball bat (or door stopper)

The above example of hitting a hanging pendulum is found in practice, e.g.,
when a baseball bat is hit by a baseball at point P. A similar problem arises M
when a swinging door hits a door stopper at point P. Let us reconsider a
rod of mass M and length L, which is hinged at its end point O, as shown ^
P
P+
on the right, and hit at a point P by an impulsive force P̂ perpendicular to
the axis of the rod. As it turns out, there exists a “sweep spot” P for which
point O experiences no reaction force (of interest for baseball players and thoseL CM+
installing door stoppers alike). Let us find this sweet spot. rOP
We begin by formulating the angular momentum balance in 2D about point O w

integrated over the time of the collision. This leads to
R O
1
IO [ω(t+ ) − ω(t− )] = rOP P̂ with IO = M L2 , (3.220)
3
where ω(t+ ) and ω(t− ) are the rod’s angular velocities immediately before and after the impact,
respectively. Next, we could also integrate linear momentum balance, however this would involve
the reaction force in the hinge. Instead, we integrate the balance of angular momentum over the
collision time again, now with respect to the center of mass of the rod. This yields the second
equation
1
ICM [ω(t+ ) − ω(t− )] = rCM P P̂ + rCM O R̂ with ICM = M L2 , (3.221)
12
where R denotes the unknown horizontal reaction force at the hinge. Note that no vertical reaction
force arises since the body changes the velocity of its center of mass only in the horizontal direction
during the collision and no other vertical forces act.
By writing rCM P = rOP − L/2 and rCM O = L/2 by geometry, we arrive at the two equations
M L2 M L2
[ω(t+ ) − ω(t− )] = rOP P̂ , [ω(t+ ) − ω(t− )] = (rOP − L/2)P̂ + LR̂/2. (3.222)
3 12
Now, let us consider the special case in which the reaction force is assumed to vanish (i.e., point P
is the sweet spot such that R̂ = 0). Setting R̂ = 0 and dividing the two equations in (3.222) by
each other (the first by the second) gives
rOP 2
4= ⇔ rOP = L, (3.223)
rOP − L/2 3
i.e., the sweep spot is at exactly 2/3 of the length of the rod. This is the so-called center of
percussion. The center of percussion of a rigid body attached to a pivot/hinge is that point for
which a perpendicular impact will produce no reactive forces at the pivot/hinge (here at point O).
Translational and rotational motions cancel at the hinge when an impulsive force is applied at the
center of percussion. This sweet spot is of importance in practice as it minimizes, e.g., the force
acting on the wrist of a baseball player or the force in the hinge of a door.
————
109
To understand the concept of the center of percussion, consider a long rod hit by an impulsive
force (e.g., due to a collision with an approach ball), as shown below. The collision changes both
the linear and the angular momentum of the rod: because of the impulsive force acting on the rod
in the horizontal direction, the center of mass (CM) of the rod will start moving to the right in
a purely horizontal fashion. If the impulse occurs exactly at CM, then the rod will move without
rotation. By contrast, if the hit occurs above or below CM, the balance of angular momentum (or
conservation of angular momentum for the whole system) will lead to a rotation of the rod. If the
hit happens above the CM, then the rod rotates clockwise; if the rod is hit below its CM, then it
rotates counter-clockwise. The superposition of translation plus rotation defines the overall motion
of the rod.
R=0
+ CM + + +w = +
+ CP + + +
vCM
Now, consider the top point of the rod to be hinged. If this top point due to the collision is
supposed to move to the right or to the left, then there must be a force in the hinge to suppress
that motion – hence, there is a reaction in the hinge. Only if the velocity of the top point is zero
throughout the collision (as shown in the special case in the above schematic), then there is no
reaction force (R = 0). This requires the hit to occur at the special point which leads to rotations
and translations to cancel in the top point of the rod; and this defines the center of percussion.
110
3.4 Non-Inertial Frames
All of the kinetic balance laws derived so far (which trace back to
Newton’s famous three axioms) apply only in an inertial reference
frame. As soon as we switch our description to a moving, non-
inertial frame, those kinetic balance laws no longer apply for an
observer in the moving frame. Therefore, we proceed to establish y
general relations between the kinematics and kinetics of inertial and
non-inertial frames for rigid bodies in 3D – and we will argue that
those equally apply to (systems of) particles. Finally, we also need
to take into account that many of our assumptions made so far cease CM +
to hold in non-inertial frames. For example, as soon as the axis of
rotation of a body is no longer fixed (such as in a moving frame
attached to a rotating spinning top, shown on the right), we cannot +
assume I˙ = 0, so we need to revisit our relations derived, e.g., for
angular momentum balance.
We generally need to exercise caution when formulating component equations (rather than symbolic
vector equations), because ėi 6= 0 in a moving frame in general. As an added complication, we
sometimes have to deal with several different reference frames in the following, so we here introduce
a framework that is sufficiently general to work with arbitrarily many (inertial or non-inertial)
reference frames.
3.4.1 Active and passive rotations
We begin by considering two reference frames (and extend the concept to more frames as needed):
• an inertial frame C with a fixed Cartesian basis {eC1 , eC2 , eC3 } and origin O,
• a non-inertial frame M with a rotating basis {eM M M

1 (t), e2 (t), e3 (t)} and with the same fixed
origin O (we will later discuss moving origins as well).
We assume that both bases are orthonormal, which implies that {eM M M
1 (t), e2 (t), e3 (t)} must be
related to {eC1 , eC2 , eC3 } by a rotation (generally in 3D).
Here and in the following, we use superscripts C and M to denote vectors associate with frames C
and M, respectively (and we can easily define more frames in this fashion). Since in most cases
we use only a single inertial frame, we may also drop the superscripts C when referring to a unique
inertial frame and simply write {e1 , e2 , e3 } for the basis of the fixed Cartesian frame. As before,
we refer to the components of a vector v in the M-frame as [v]M , and the ith component as [vi ]M .
By projecting the basis vectors of the M-frame onto the Cartesian basis vectors of the C-frame,
we establish a relation between the components of a vector v in the two frames (see the schematic
above). Specifically, we start with a basis vector eM
j and its representation
3
X
eM eM C
C
j = j · ei ei , (3.224)
i=1
111
which allows us to express a vector v in both bases as

3
X
v= [vi ]C eCi (3.225)
i=1
and, equivalently,
 
3
X 3
X 3
X 3
X X3
[vj ]M eM eM C
eCi = eM C  C

v= j = [vj ]M j · ei j · ei [vj ]M ei . (3.226)

j=1 j=1 i=1 i=1 j=1
| {z }
!
=[vi ]C
P
+
P
+
v v
[v2]M
[v2]C
C
e2 e2
M
C e1
C
e3M e1M [v3]M
e3 [v3]C [v1]M
[v1]C
components of v in the C-frame components of v in the M-frame
By comparison, the components of vector v in the two frames C and M are related via
3
X
[vi ]C = [TijCM ][vj ]M with [TijCM ] = eCi · eM
j . (3.227)
j=1
[TijCM ] are the components of a rotation matrix that converts vector components in the M-
basis into their corresponding components in the C-basis; therefore, the superscripts are to be read
as T CM = T C←M , i.e., from M to C (but we usually omit the arrow for better visibility and
conciseness).
Note that, if C is the Cartesian basis, then the ith column of the matrix T CM contains simply the
three vector components of eM j (i = 1, 2, 3) in that Cartesian basis.
The inverse relation can be found analogously, starting with

3 3 3 X
3 3
M !
X X X X
eCj eCj eM eM C C M
[vi ]M eM

= · i i ⇒ v= [vj ]C ej = ej · ei [vj ]C ei = i , (3.228)
i=1 j=1 i=1 j=1 i=1
so that
3
X
[vi ]M = [TijMC ][vj ]C with [TijMC ] = eM C
i · ej . (3.229)
j=1
112
Analogous to the prior observation, we conclude that, if C is the Cartesian basis, then the ith row of
the rotation matrix T MC contains the three vector components of eM j (i = 1, 2, 3) in that Cartesian
basis.
Comparing their definitions shows that the two rotation matrices are each other’s transpose:
C T
T
[TijMC ] = eM C M
= [TijCM ]T T MC = T CM

i · ej = ej · ei ⇒ (3.230)
The thus defined rotations are often referred to as passive rotations, as they do not actively move
anything but are merely instantaneous relations between vector components in two different frames
of reference (vector v does not move, we only express its components in different bases).
As a next step, let us introduce active rotations that take a vector v and rotate it onto another
vector u = Rv. Note that this operation is independent of a basis or frame of reference, so R is a
second-order tensor 12 defined by its components Rij along with a particular basis.
e2
C
For example, we may look for the rotation RMC = RM←C that takes
e3
M
the basis vectors {eC1 , eC3 , eC3 } and rotates them onto the basis vectors
e
M
{eM M M
1 , e3 , e3 }. This implies that 1
eM
1 =R
MC C
e1 , eM
2 =R
MC C
e2 , eM
3 =R
MC C
e3 , (3.231) e1C
which are nine equations for the nine unknown components of RMC
(still assuming that both bases are orthonormal, so a unique solution e3C
e2
exists). M
To compute the components of RMC within the basis C, we may exploit the relation
MC
[Rij ]C = eCi · RMC eCj = eCi · eM
j ⇒ MC
[Rij ]C = eCi · eM
j . (3.232)
Hence, we observe that the components of this tensor RMC are identical to those of the transfor-
mation matrix T CM introduced above (which makes sense: we either rotate the vector within the
a given basis, or we look at the same vector within two bases – these two rotations are opposite).
Analogously, we can also find the components of RCM in the C-frame: the relations
eC1 = RCM eM
1 , eC2 = RCM eM
2 , eC3 = RCM eM
3 (3.233)
lead to
CM
T
[Rij ]C = eCi · RCM eCj = eCj · RCM eCi . (3.234)
From the discussion of the rotation matrices above, we may guess that the components Rij CM should
MC . Inserting this relation here results in
be the transpose of Rij
CM
T
[Rij ]C = eCj · RCM eCi = eCj · RMC eCi = eCj · eM i ⇒ CM
[Rij MC
]C = [Rji ]C , (3.235)
12
Second-order tensors, like vectors, P
are defined by components and a basis. While vectorsP (which in fact are
first-order tensors) are written as v = i vi ei , second-order tensors are defined as R = i,j Rij ei ⊗ ej with the
dyadic/tensor product a⊗b = abT = i,j ai bj ei ⊗ej P
P
. Tensors act on vectors as a linear mapping,
P meaning P that Rv is
a vector u and their components are related by ui = j Rij vj within a given basis; i.e., u = i ui ei = i,j Rij vj ei .
The components of a tensor T within a basis C are identified as Tij = eCi · T eCj .
113
which indeed confirms that RCM = (RMC )T . In summary, we have the following coordinate
transformation rules for passive rotations (replacing T by the respective RT components):
3
X 3
X
CM T MC T MC CM
[vi ]C = [Rij ] [vj ]M , [vi ]M = [Rij ] [vj ]C with [Rij ] = [Rji ] = eCi · eM
j
j=1 j=1
(3.236)
while for active rotations we have

T
eM
i =R
MC C
ei , eCi = RCM eM
i with RMC = RCM (3.237)
MC ] = [RCM ] are the same as defined in (3.236), i.e.,

where the same tensor components [Rij C ji C
MC ] = [RCM ] = eC · eM .
[Rij C ji C i j
v [v2]M v
[v1]C
e2 e2
M C [v3]M u e2
[v2]C e1 M
e1 C
e1
[v3]C e3 e3 [v1]M
C M
e3
passive rotation: same vector v active rotation: vector v

measured in two frames is rotated onto vector u
Note that two subsequent opposite (active) rotations applied to a vector v must result in the
identity, i.e.,
! T
v = RCM (RMC v) = (RCM RMC )v ⇒ RCM RMC = RCM RCM = I. (3.238)
This shows that, in fact, any second-order tensor R that describes a rotation must satisfy
RRT = RT R = I ⇒ R−1 = RT and det R = 1. (3.239)
All such rotation tensors form a group, the so-called special orthogonal group SO(d) in d
dimensions. (Without proof, we mention that the condition det R = 1 ensures that R is a rotation
as opposed to a reflection which is characterized by det R = −1.).
As an example, let us identify the three elementary rotations about each coordinate axis eCi .
For example, a rotation about the eC3 -axis by ϕ3 defines the frame M3 as
eM
1
3
= cos ϕ3 eC1 + sin ϕ3 eC2 , eM
2
3
= − sin ϕ3 eC1 + cos ϕ3 eC2 , eM
3
3
= eC3 , (3.240)
114
which leads to the components of the rotation tensor as

 
cos ϕ3 − sin ϕ3 0
[RM3 C ]C =  sin ϕ3 cos ϕ3 0 , (3.241)
0 0 1
so we see that the ith column of RM3 C contains the components of vector eMi . Analogously, the
3
rotations about the eC1 - and eC2 -axes follow as

   
1 0 0 cos ϕ2 0 sin ϕ2
[RM1 C ]C = 0 cos ϕ1 − sin ϕ1  , [RM2 C ]C =  0 1 0 . (3.242)
0 sin ϕ1 cos ϕ1 − sin ϕ2 0 cos ϕ2
Note the flipped sign in RM2 C , which is due to the definition of a positive rotation angle when
rotating about the eC3 -axis. The three rotations are shown below.
j3
j1
e2 e2 j3 e2=e2C M
e2 e2
M C C
M
e1
M
e1 M
j2
e1
C
e =e C M
e1 C
e3=e3
C M
e3 e 1 1
C M e3C
3 e3 M
j2
j1
We can easily verify that inverting any of the above rotations (i.e., replacing ϕi by −ϕi ) immediately
confirms that R−1 = RT for each of the above tensors, since cos(−ϕ) = cos ϕ while sin(−ϕ) =
− sin ϕ.
Any general rotation in 3D can be defined as a subsequent composition of the above three elementary
rotations, e.g., if a body is first rotated by ϕ1 about the eC1 -axis (into the M1 -frame), then by ϕ2
about the eC2 -axis (into the M2 -frame), and finally by ϕ3 about the eC3 -axis (into the M3 -frame),
then the complete rotation is characterized by
RM3 C = RM3 M2 RM2 M1 RM1 C (3.243)
with components
   
cos ϕ3 − sin ϕ3 0 cos ϕ2 0 sin ϕ2 1 0 0
[RM3 C ]C =  sin ϕ3 cos ϕ3 0  0 1 0  0 cos ϕ1 − sin ϕ1  . (3.244)
0 0 1 − sin ϕ2 0 cos ϕ2 0 sin ϕ1 cos ϕ1
Note that the order of the rotation tensors is reversed when compared to the order of applying
the rotations, because the action of a tensor R1 on a vector v defined as R1 v, so a subsequent
application of R2 yields R2 (R1 v) = (R2 R1 )v.
115
We note without proof here that a rotation about an axis with unit vector e by an angle ϕ is
generally described by a rotation tensor R whose Cartesian components are given by
[Rij ]C = ei ej + (δij − ei ej ) cos ϕ − ijk ek sin ϕ for i, j = 1, . . . , 3, (3.245)
where ijk denotes the permutation symbol, and δik is Kronecker’s delta. This reduces to the above
elementary rotations when choosing e = eCi and ϕ = ϕi for each of the cases i = 1, 2, 3.
Example 3.20. Passive and active rotations
As an example of passive and active rotations let us consider two frames of reference defined as
follows:
• frame C has a fixed, Cartesian basis {eC1 , eC2 , eC3 } = {e1 , e2 , e3 } and origin O,
• frame M is obtained from C by first rotating the basis vectors by π/4 about the eC3 -axis,
followed by a rotation by −π/2 about the eC2 -axis, with the same origin O.
The rotation from C to M is hence described by

cos −π 0 sin −π cos π4 − sin π4 −1
    
2 2 0 0√ 0√
MC
[R ]C =  0 1 0   sin π4 cos π4 0 = 1/√2 1/ √2 0  , (3.246)
− sin −π2 0 cos −π
2 0 0 1 1/ 2 −1/ 2 0
for which it is a simple exercise to verify that (RMC )T = (RMC )−1 = RCM : the inverse rotation is
found by first rotating by +π/2 about the eC2 -axis, then by −π/4 about the eC3 -axis:
√ √ 
cos −π − sin −π cos π2 0 − sin π2
   
4 4 0 0 1/√2 1/ √2
[RCM ]C =  sin −π 4 cos −π4 0  0 1 0  =  0 1/ 2 −1/ 2 . (3.247)
0 0 1 sin 2 0 cos π2
π
−1 0 0
The basis vectors of the M-frame are eM

i =R
MC eC whose components are the columns of RMC :
i
     
0 0 −1
1 1
[eM ]
1 C = √  1  , [e M
]
2 C = √  1  , [e M
]
3 C =  0 . (3.248)
2 1 2 −1 0
As an example vector, v = e2 has components [v1 ]C = [v3 ]C = 0 and [v2 ]C = 1, whereas in the
M-frame we obtain
 √ √    
3 0 1/√2 1/ √2 0 1
X
CM 1
[vi ]M = [Rij ][vj ]C ⇒ [v]M =  0 1/ 2 −1/ 2 1 = √ 1 . (3.249)
j=1 −1 0 0 0 2 0
As a final check, we can also revert the coordinate transform according to

−1
     
3 0√ 0√ 1 0
X
MC 1
[vi ]C = [Rij ][vj ]M ⇒ [v]C = 1/√2 1/ √2 0  √ 1 = 1 . (3.250)
j=1 1/ 2 −1/ 2 0 2 0 0
The two bases along with vector v and its M-components are illustrated below.
116
e2C e2C e2C

e
2
M
e2M
e1M e1M
p/4
v
p/4
p/4
e1C e3M p/4 e1C e3M e1C
p/2 p/2
e3C e3C e3C
3.4.2 Rotating frames of reference
Active rotations act on vectors and perform a rotation (e.g., from the
e2
C
C-basis to the M-basis). This rotation can be time-dependent, e.g., t
e1(t)
M
e3(t)
if the C-basis is time-invariant (inertial frame) while the M-basis is M
rotating as a function of time (non-inertial frame). When expressing

Pd t
a vector in the M-frame as r(t) = i=1 ri (t)eM i (t), it is important
to compute the time derivatives of the basis vectors eM i in order to e1C
compute ṙ(t). Differentiating the M-basis with respect to time (and R (t) MC
exploiting that the Cartesian frame is time-invariant) leads to

e3C
e2(t)
M
t
eM
i (t) = R
MC
(t) eCi ⇒ ėM
i (t) = Ṙ
MC
(t) eCi . (3.251)
From (3.239) it follows that (dropping the t-dependence for brevity)
T T T h T iT
RMC RMC = I ⇒ ṘMC RMC = −RMC ṘMC = − ṘMC RMC . (3.252)
This motivates the definition of the spin tensor

−ΩM ΩM
 
T 0 3 2
W MC = ṘMC RMC with [W MC ]C =  ΩM
3 0 −ΩM1
. (3.253)
−Ω2 M M
Ω1 0
The latter component form follows directly from the relation (3.252), which shows that the spin ten-
T
sor W MC = − W MC is skew-symmetric (having zero diagonal components and anti-symmetric
off-diagonal components) with only three independent components.
Notice that for any vector v

 M
Ω1
W MC v = ΩM × v with [ΩM ]C = ΩM
2
, (3.254)
M
Ω3
where ΩM is the angular velocity vector of frame M. Here and in the following, we use the
superscript M to indicate that the rotation is that of the M-frame.
117
From this derivation, it follows – starting with (3.251) – that

ėM
i = Ṙ
MC C
ei = ṘMC (RMC )−1 eM i = Ṙ
MC
(RMC )T eM i =W
MC M
ei = ΩM × eM
i . (3.255)
The same relation can be re-interpreted as follows. We have seen that, for a rigid body moving in
3D with an instantaneous angular velocity vector Ω(t), the velocity of any point on the body can
be calculated by the velocity transfer formula (3.10). Let us treat the reference frame M like a
moving body with basis vectors {eM M M
1 , e2 , e3 } which rotate with an instantaneous angular velocity
M
vector Ω (t). In this case, using a fixed origin, the velocity transfer formula gives
d M
e (t) = ėM M M
i (t) = Ω (t) × ei (t), (3.256)
dt i
which is equivalent to (3.255). This simplifies our derivations significantly. The absolute, inertial
time derivative of any vector y in the M-frame, when measured by an inertial observer in the
C-frame, is
3 3 3 3 3
C d X X X X X
ẏ = [yi ]M eM
i = [ẏi ]M eM
i + [yi ]M ėM
i = [ẏi ]M eM M
i +Ω × [yi ]M eM
i . (3.257)
dt
i=1 i=1 i=1 i=1 i=1
The above obviously deviates from the time derivative measured by an observer moving with frame
M, which is
3
X
ẙ M = [ẏi ]M eM
i , (3.258)
i=1
since for the moving observer the basis vectors eM

i appear to be time-invariant. Note the difference
in notation: a superscript M refers to a quantity associated with the M-frame (such as the angular
velocity ΩM , or the basis vectors eMi , or – here – the time derivative with respect to frame M),
while [·]M refers to the components of a vector measured in the M-frame.
The relation between the two time derivatives follows from the above two expressions, viz.
ẏ C = ẙ M + ΩM × y ⇒ ẏ C = ẙ M + ΩM × y (3.259)
|{z} |{z}
(absolute) time (relative) time
derivative in C derivative with
respect to M
This relation is quite important. It means that, if we are in a moving frame of reference M, we
do not need to measure our velocity with respect to some inertial basis C. All we need to know is
our velocity ẙ M with respect to the moving frame, the angular velocity ΩM with which our frame
is rotating, and our current position y to uniquely calculate our velocity ẏ C in the inertial frame.
(Note that y is a vector that can be expressed in either frame, C or M, which is why it has no
superscript in the above relation.)
In practice, Eq. (3.259) can only be evaluated if all vectors are expressed in the same frame.
Therefore, it makes sense to express the right-hand side in (3.259) exclusively in the M-frame
118
(including the vectors y and also ΩM ). To this end, we point out that any vector, such as an
angular velocity Ω, can be expressed with respect to both bases and the relations (3.236) and
(3.237) apply as well, e.g.,
3
X 3
X
Ω= [Ωi ]C eCi = [Ωi ]M eM
i and CM T
[Ωi ]C = [Rij ] [Ωj ]M (3.260)
i=1 i=1
with the rotation defined by (3.236) and (3.237).
Example 3.21. Rotations in 2D
As a special case, let us consider two reference frames related by a rotation in 2D, so we may be
interested in a rotating basis (see the schematic below)
g1 (t) = cos ϕ(t) eC1 + sin ϕ(t) eC2 ,

g2 (t) = − sin ϕ(t) eC1 + cos ϕ(t) eC2 , (3.261)
g3 (t) = eC3 ,
e2C j
where the time-varying angle ϕ(t) describes the rotation from the g2 g1
e3=g3 e1
C C
Cartesian to the co-rotating basis vectors in the plane (see the
schematic on the right). Notice that the (absolute) time derivatives M
measured in the C-frame are W(t)
ġ1C = −ϕ̇ sin ϕ eC1 + ϕ̇ cos ϕ eC2 = ϕ̇ g2 ,
(3.262)
ġ2C = −ϕ̇ cos ϕ eC1 − ϕ̇ sin ϕ eC2 = −ϕ̇ g1 ,
so that, writing ΩM (t) = ϕ̇(t), we conclude

ġ1C = ΩM g2 , ġ2C = −ΩM g1 , ġ3C = 0. (3.263)
The above relations may conveniently be abbreviated by introducing the angular velocity vector
ΩM (t) = ΩM (t)eC3 = ΩM (t)g3 (3.264)
such that indeed
ġiC = ΩM × gi for i = 1, 2, 3. (3.265)
This confirms relation (3.259), since the time derivative of a vector (expressed in the co-rotating
basis vectors)
3
X
v= [vi ]M gi (3.266)
i=1
becomes
3
X 3
X 3
X
v̇ C = ([v̇i ]M gi + [vi ]M ġi ) = v̊ + [vi ]M ΩM × gi = v̊ + ΩM × [vi ]M gi
i=1 i=1 i=1
(3.267)
= v̊ + Ω × v.
————
119
Example 3.22. Polar coordinates revisited
Recall that we had derived in Section 1.1.3 the velocity and acceleration components of a rotating
particle in polar coordinates (measured with respect to the rotating M-frame) as
r̈ − rϕ̇2
   
ṙ
[v]M = rϕ̇ , [a]M = 2ṙϕ̇ + rϕ̈ , (3.268)
0 0
and we made the important conclusion (as evident from the above vector components) that
d
[a]M 6= [v]M . (3.269)
dt
In other words, we cannot simply differentiate the vector components to obtain the acceleration
components:
d d
ṙ 6= r̈ − rϕ̇2 and (rϕ̇) 6= 2ṙϕ̇ + rϕ̈. (3.270)
dt dt
Equation (3.259) immediately explains why:
r̈ − rϕ̇2
       
r̈ 0 ṙ
v̇ = v̊ M + ΩM × v ⇒ [v̇]M = ṙϕ̇ + rϕ̈ +  0  × rϕ̇ = 2ṙϕ̇ + rϕ̈ . (3.271)
0 ϕ̇ 0 0
In other words, we may not only differentiate the components of [v]M in the rotating frame, but
we must also account for the fact that the basis vectors in polar coordinates rotate and hence have
a non-zero time derivative (as also discussed in Example 3.21 above). Using the correct relation
(3.259) hence yields the correct time derivative (here expressed in components [a]M measured with
respect to the rotating frame).
————
Now that we understand the kinematic relations between inertial and non-inertial frames, let us
revisit the kinetic balance laws and begin with linear momentum balance. For generality, we
consider the following two frames:
• frame C has a fixed, Cartesian basis {eC1 , eC2 , eC3 } = {e1 , e2 , e3 } and origin OC ,
• frame M has a rotating basis {eM M M M

1 , e2 , e3 } and a moving origin O .
Note that, unlike in our previous discussions, we here also consider distinct origins of the two
reference frames. This does not affect the relations between the two basis vectors (which are still
governed by the rotation relations derived above).
120
The position of any point P in space can be expressed with x2C

respect to both origins, the fixed Cartesian one and the
moving origin OM , viz.
rP (t) = rOM (t) + rOM P (t), (3.272) rP x1M

where rOM P is the position of point P relative to the mov- eC2 x2M rO P M
e3C e1C x1C

ing frame. Taking a time derivative on both sides of the O
C
equation with (3.259) gives rOM

e2M e1M
(t) × rOM P (t), (3.273) x3

C M
ṙPC (t) = ṙOM (t) + r̊O e3M O
M M
M P (t) + Ω
while the second time derivative becomes (dropping the t-

dependence for simplicity) x3M
◦◦ M dΩM
r̈PC = r̈OM + r OM P + × rOM P + 2ΩM × r̊O
M
MP + Ω
M
× (ΩM × rOM P ) . (3.274)
| dt {z } | {z } | {z }
Coriolis acceleration centripetal acceleration
Euler acceleration
We notice that – beyond the first two terms – three additional terms have appeared because of the
non-inertial basis. These three new acceleration terms are referred to as the Coriolis acceleration
aCoriolis , centripetal acceleration acentripetal and Euler acceleration aEuler , as shown above.
Having established a relation between the accelerations measured in both systems, let us return to
Newton’s axioms which, as we discussed, apply only in inertial reference frames. For a rigid body,
we may make the particular choice rP = rCM , so that linear momentum balance implies
X
C
M r̈CM = Fiext . (3.275)
i
If we denote the position, velocity and acceleration of the center of mass measured with respect to
the moving M-frame by, respectively,
M M M ◦◦ M
rCM = rOM CM , vCM = r̊OM CM , aM
CM = r OM CM , (3.276)
then linear momentum balance for a rigid body in a moving frame becomes
X
M aM
CM = Fiext + FCoriolis + FEuler + Fcentrifugal − M aOM (3.277)
i
Besides the externally applied forces, the right-hand side includes new additional forces, viz.
• the Coriolis force: FCoriolis = −2M ΩM × vCM

M
• the centrifugal force: Fcentrifugal = −M ΩM × (ΩM × rCM

M
)
dΩM M
• the Euler force: FEuler = −M × rCM
dt
• the (fictitious) force −M aOM due to the moving origin.
121
Note the changes in sign resulting in the centripetal acceleration becoming centrifugal force. The
three above forces are apparent forces (or fictitious forces) and not external, physical forces;
i.e., to an observer in the moving frame objects look like they are moving due to these forces
(because their frame of reference is moving), but those forces are not observed in an inertial frame.
It is helpful to keep in mind that the
• Coriolis force is present only if the body is moving with respect to the rotating frame, i.e., if
v M = r̊ M 6= 0. If a body is fixed onto a rotating frame, a co-rotating observer notices no
Coriolis force.
• Euler force appears only if the angular velocity of the reference frame is changing with time,
i.e., Ω̇M 6= 0. If a rotating system rotates with constant angular velocity, no Euler force is
experienced.
• centrifugal force always acts in a rotating system. Note that
Ω × (Ω × r) = (Ω · r)Ω − Ω2 r with Ω = |Ω|, (3.278)
which is helpful in particular in 2D rotations where Ω · r = 0, so the centrifugal force in 2D

simplifies to
Fcentrifugal,2D = M (ΩM )2 rCM

M
. (3.279)
In most practical cases we will be able to use the same fixed origin, so OC = OM and the final term
in (3.277) vanishes.
In summary, when formulating the balance of linear momentum in a moving frame M, then we
must not forget – in addition to the physical forces F ext – the three apparent/fictitious/inertial
forces given above.
Example 3.23. Rotation in 2D revisited
We consider a 2D rotation about the e3 -axis, so the motion is constrained to the x1 -x2 -plane and
the reference frame M is rotating with angular velocity
ΩM = Ωe3 , (3.280)
while the origin is fixed. Assume that we are an observer within the rotating system, so we measure
the 2D position and velocity of any particle or body in the rotating system as, respectively
2
X 2
X
r M (t) = yi (t)eM
i (t) and v M (t) = r̊ M (t) = ẏi (t)eM
i (t), (3.281)
i=1 i=1
and the acceleration measured in the M-frame follows analogously as

2
M ◦◦ M X
a (t) = r (t) = ÿi (t)eM
i (t) (3.282)
i=1
122
Note that we here labeled the components of the position as yi (t) instead of [ri ]M (t) for simplicity
and brevity.
Writing out the components of the individual terms in (3.277) in the moving frame for a particle
(or body) of mass m gives
       
ÿ1 −ẏ2 y1 −y2
m ÿ2  = [F ext ]M −2mΩM  ẏ1  + mΩ2 y2  −mΩ̇  y1 , (3.283)
0 0 0 0
| {z } | {z } | {z }
Coriolis force centrifugal force Euler force
where the components of F ext must be those with respect to the rotating basis M, as indicated.
Here, it is apparent that the centrifugal force is acting radially outward. All fictitious forces are
schematically shown below for a particle moving along the dashed trajectory in the rotating frame
(all forces are in the plane).
m Fcentrifugal
M
e3 r M
M
r M
e2
e1 FCoriolis FEuler
M
W (t)
M
————
Example 3.24. Particle on a taut string revisited

In Example 1.10 we derived that the force in a taut string (of length l
l) due to a particle of mass m attached to its end and rotating at a e3
M
e2M
constant angular velocity ω is given by
w e1M S
2
S = ma = −mlω er . (3.284)
What we derived back then can now easily be retrieved by our formulation of a rotating frame of
reference, in which linear momentum balance takes the form (3.277). We pick a moving M-frame
which rotates with the particle and string and whose origin is fixed at the center of rotation, as
shown. In this frame, the position of the particle is
M
rCM = leM
1 = const. ⇒ aM
CM = 0, aM
CM = 0. (3.285)
Consequently, FCoriolis = 0. Further, we have ΩM = ωeM 3 = const. so FEuler = 0. Altogether,
linear momentum in the moving frame (3.277) simplifies to
0 = −S − mΩM × (ΩM × rCM
M
) ⇔ S = −mΩM × (ΩM × rCM
M
) = −mω 2 leM
1 . (3.286)
We hence recover the solution (3.284)
————
123
Example 3.25. Satellite in geostationary orbit

A satellite in so-called geostationary or geosynchronous equatorial
msat
orbit (GEO) is surrounding the Earth with a radius of rotation
rGEO such that it remains geostationary (i.e., above exactly the F
same location on the Earth’s ground) if no external forces act on the
satellite except for gravity. What is rGEO ? rGEO
This is a particular case of the scenario of Example 3.24. Instead of

a taut string, gravity is attracting the satellite of mass msat towards mEarth
the Earth’s center, the magnitude of the gravitational force being
w
msat mEarth
F =G , (3.287)
r2
where G ≈ 6.674 · 10−11 m3 kg−1 s−2 is the gravitational constant, mEarth ≈ 5.974 · 1024 kg the Earth’s
mass, and r the radius of rotation of the satellite.
In geostationary orbit, the above gravitational force balances the centrifugal force in the absence
of any other forces acting on the satellite. Therefore, linear momentum balance (3.286) with
S = −F eM1 becomes
GmEarth 1/3

msat mEarth 2
G 2 = msat ω rGEO ⇔ rGEO = . (3.288)
rGEO ω2
For geostationarity, the satellite must rotate (above the equator) with the Earth’s angular velocity,
which is
2π 2π
ω= = ≈ 7.27 · 10−5 s−1 . (3.289)
1d 24 · 60 · 60s
This leads to
1/3
G mEarth
rGEO = ≈ 42.3 · 103 km, (3.290)
ω2
which is approximately 35.8 · 103 km above the Earth’s surface. The satellite moves at a constant
speed of vsat = ωrGEO ≈ 3.07 km/s.
————
Example 3.26. Particle in a rotating barrel

Consider a cylindrical barrel that is rotating with angular velocity
ω(t) in the x1 -x2 -plane, i.e., ω(t) = ω(t)e3 . Inside the barrel slides x
g mg m
a particle of mass m (assume that the barrel constrains the particle e 3
M
motion along the barrel’s axis, i.e., the particle fits right into the
e2M
barrel so it can only move along the barrel). If released from rest
O e1M Fext
at an initial distance x0 from the center, what is the motion of the
w(t)
particle and what forces are acting on the particle?
124
Here, it is natural to use a moving frame with basis {eM M M

1 , e2 , e3 } that is rotating with the barrel,
M M
so that Ω = ω (and we take O = O , so the origin is fixed and identical in both frames and
aOM = 0).
In the rotating frame, only a single degree of freedom x along the axis of the barrel is needed to
M = xeM . The kinematics
describe the position of the particle relative to the moving frame as13 rCM 1
of the particle motion in the rotating M-frame then follows as
M
rCM = xeM
1 ⇒ M
vCM = ẋeM
1 ⇒ aM M
CM = ẍe1 . (3.291)
Linear momentum balance in the non-inertial frame M is described by Eq. (3.277), viz.
maM
CM = Fext + FCoriolis + FEuler + Fcentrifugal
(3.292)
= Fext − 2mΩM × vCM
M
− mΩ̇M × rCM
M
− mΩM × (ΩM × rCM
M
).
Insertion of (3.291) along with ΩM (t) = ω(t) = ω(t)eM

3 yields
mẍeM M M M M 2 M
1 = Fext − 2mωe3 × ẋe1 − mω̇e3 × xe1 + mω xe1
(3.293)
= Fext − 2mω ẋeM M 2 M
2 − mω̇xe2 + mω xe1
Here, the resultant external force is the sum of gravity and the force exerted by the barrel onto
the particle (which can only act perpendicular to the direction of motion for frictionless sliding of
the particle), therefore we may write Fext = F2 eM M
2 + (F3 − mg)e3 with yet unknown F2 and F3
components. This allows us to decouple the above linear momentum balance into its components:
mẍ = mω 2 x ẍ − ω 2 x = 0,
0 = F2 − 2mω ẋ − mω̇x ⇒ m(2ω ẋ + ω̇x) = F2 , (3.294)
0 = F3 − mg mg = F3 .
The first of these equations is the equation of motion to be solved for x(t) with suitable initial
conditions. For example, for a particle initially at rest at x(0) = x0 at time t = 0 and if ω(t) =
ω = const., then the particle is accelerating outward exponentially according to
x0
x(t) = [exp(ωt) + exp(−ωt)] . (3.295)
2
The second and third equations in (3.294) provide the components of the reaction force from the
barrel onto the particle. For the example solution (3.295) with ω(t) = ω = const. those become
F2 (t) = 2mω ẋ(t) = mω 2 x0 [exp(ωt) − exp(−ωt)] and F3 = mg. (3.296)
Note that we could also have obtained the reaction force F2 from the balance of angular momentum
around the origin O in an inertial system. We have
MO,3 = ḢO with HO = IO ω = mx2 ω and MO,3 = F2 x, (3.297)

13
Even though for a particle we do not need to work with the center of mass (CM), we use this notation rCM here
for consistency with the above derivation.
125
so that we again obtain

d
mx2 ω = 2mxẋω + mx2 ω̇

F2 x = ⇒ F2 = 2mẋω + mxω̇. (3.298)
dt
As a variation, consider the barrel being inclined by an angle α against the horizontal plane, so
the particles moves upwards under an angle α in the barrel. In this case, we may want to introduce
a new frame R, which is obtained by rotating the M-frame with RRM about the eM 2 -axis by α,
such that the eR
1 -axis aligns with the barrel. Here, the same derivations as above applies with
ΩR = ω where, expressed in the R-frame, ΩR = ω sin α eR R

1 + ω cos α e2 . (3.299)
With the external force Fext = −mg(sin αeR R R R

1 + cos αe2 ) + F2 e2 + F3 e3 , we now arrive at
mẍeR R R 2 2 R
1 = Fext − 2mω ẋ cos αe2 − mω̇x cos αe2 + mω cos αxe1 , (3.300)
which admits analogous conclusions as the derivation above.

As a further variation, assume that the particle is attached to a x2
spring of stiffness k, which is attached to the origin and unstretched
m
in initial position x0 . Let us find the steady-state location of the par- x0
ticle under a constant angular velocity ω (where the apparent forces k
balance the spring force). e2
M
x
O e1 x1
M
We align the spring with the rotating x1 -axis as before, so that the
radial motion of the particle is again described in the moving frame
by w
M
rCM (t) = x(t)eM
1 . (3.301)
M = 0 and
When assuming a steady-state solution with respect to the moving frame (i.e., r̊CM
◦◦ M
r CM = 0) and a constant angular velocity (ω̇ = 0), the equations of linear momentum balance
in the rotating frame simplify significantly. Linear momentum balance in the x1 -direction in the
moving frame under these conditions gives
mẍ1 = F1 + mω 2 x = 0. (3.302)
Since the force component F1 is the resultant of the stretched spring, we have
kx0
F1 = −k(x − x0 ) ⇒ k(x − x0 ) = mω 2 x ⇔ x= . (3.303)
k − mω 2
p
Note that this solution only applies if ω 6= k/m. Under static conditions (ω = 0), we recover the
equilibrium solution x = x0 . When ω is increased from zero, the denominator decreases, so that
the spring is increasingly stretched due to the centrifugal force.
————
126
Example 3.27. Throwing a particle on a merry-go-round
In Example 1.9 we discussed how to compute the flight trajectory of a particle that is only subjected
to gravity. For example, if we throw a particle from the initial position r(0) = h0 e3 with an initial
velocity v(0) = v0 e1 in the presence of a constant gravitational acceleration g = −ge3 , then we
know the particle follows a parabola described by
gt2

g 2
r(t) = r(0) + v(0)t + t = h0 − e3 + v0 te1 . (3.304)
2 2
Now, assume that the same particle is thrown with exactly the same conditions, but we are sitting
on a rotating system that is rotating at a constant angular velocity Ω (such as on a merry-go-round),
and observe the flight of the particle as a rotating observer. As shown above, linear momentum
balance for the particle of mass m in the frame rotating with ΩM = Ωe3 = const. around a fixed
origin O in the moving frame reads
   ext     
ÿ1 F1 −ẏ2 y1
m ÿ2  = F2ext  − 2mΩ  ẏ1  + mΩ2 y2  , (3.305)
ÿ3 F3ext 0 0
where Fiext = [Fiext ]M are the external force components with respect to the rotating basis, and we
again denote by yi the ith component of the particle’s position as observed by the rotating observer
in the rotating M-frame with basis eM M
1 (t), e2 (t), e3 .
Note that the y3 -component in (3.305) is decoupled from the other two components (if the external
forces do not depend on position, which we assume here). If the particle is thrown, as above, under
the action of gravity as the only external force, then [F1ext ]M = [F2ext ]M = 0 and [F3 ]M = −mg, so
that the third component of (3.305) admits integration as before:
gt2 gt2
ÿ3 (t) = −g ⇒ y3 (t) = − + ẏ3 (0) t + y3 (0) = h0 − . (3.306)
2 2
The y1 - and y2 -components must be obtained by solving a system of ordinary differential equations
(ODEs), viz. the first two components of (3.305):
ÿ1 = 2Ωẏ2 + Ω2 y1 , ÿ2 = −2Ωẏ1 + Ω2 y2 . (3.307)
The solution of the system of ODEs is nontrivial in general. Therefore, let us again consider the
special case of a particle thrown with initial position r(0) = y(0) = h0 e3 with an initial velocity
v(0) = ẙ(0) = v0 e1 .
v0 m
g
h0 e3
r
M
M
e2
W (t)
M
e1 M
127
The initial conditions for the motion of the particle in the plane are expressed in the rotating frame
as

y1 (0) 0 ẏ1 (0) v
[y(0)]M = = and [ẙ(0)]M = = 0 . (3.308)
y2 (0) 0 ẏ2 (0)) 0
As we can easily verify by insertion, (3.307) admits the following solution:
y1 (t) = v0 t cos(Ωt), y2 (t) = −v0 t sin(Ωt). (3.309)
That is, the rotating observer does not see the particle flying along a straight line in the x1 -x2 -
plane and only losing altitude, as was the case for an outside, inertial observer (cf. (3.304)). Instead
the rotating observer watches the particle on a curved trajectory described by (3.309). Note that
inserting the initial conditions into (3.306) further yields y3 (t) = h0 − 12 gt2 .
We can verify that those two descriptions agree by converting the inertial observer’s trajectory
(3.304) into the rotating observer’s basis. To this end, we exploit that
eC1 = cos ϕ(t) eM M

1 (t) − sin ϕ(t) e2 (t), eC3 = eM
3 with ϕ(t) = Ωt, (3.310)
so that
gt2

r = h0 − eC3 + v0 teC1
2
gt2

eM M M

= h0 − 3 + v0 t cos(Ωt)e1 (t) − sin(Ωt)e2 (t)
2 (3.311)
gt2

= h0 − eM M M
3 + v0 t cos(Ωt) e1 (t) −v0 t sin(Ωt) e2 (t).
2 | {z } | {z }
| {z } y1 (t) y2 (t)
y3 (t)
The latter agrees with (3.309) and (3.306), so the solutions y(t) and r(t) are indeed identical –
only described in two different reference frames. Shown below are the two solutions as seen in the
rotating and inertial frames (plotted is the projection into the x1 -x2 - or y1 -y2 -plane as seen from
above).
y2 x2
g
y2 y1
e2 C
e
M
e2
M
e2 1 j
M
e1 M
y1 e1 r(t) x1
C
y(t)
trajectory observed by trajectory observed by

a rotating observer an inertial observer
128
Let us also verify the apparent forces. Obviously, the inertial observer sees the flight trajectory
(3.304) with
a(t) = −g, (3.312)
so the observer concludes that
F = ma ⇒ F = −mg. (3.313)
In other words, to the inertial observer the particle flies as if (correctly) only the gravitational force
is acting on the particle.
By contrast, the rotating observer witnesses the curved trajectory (3.311) and measures the (rela-
tive) particle velocity
ẙ(t) = ẏ1 (t)eM M M

1 + ẏ2 (t)e2 + ẏ3 (t)e3
= v0 cos ϕ(t)eM M
+ v0 t −ϕ̇ sin ϕ(t)eM M
− gteM

1 − sin ϕ(t)e2 1 − ϕ̇ cos ϕ(t)e2 3 (3.314)
M M M M M

= v0 cos ϕ(t)e1 − sin ϕ(t)e2 − Ω v0 t sin ϕ(t)e1 + cos ϕ(t)e2 − gte3 ,
where we used ϕ(t) = Ωt and Ω = ϕ̇. Taking one more time derivative in the rotating frame and
exploiting that Ω̇ = ϕ̈ = 0 yields
◦◦
y(t) = v0 Ω − sin ϕeM M
+ v0 Ω − sin ϕeM M
− Ω2 v0 t cos ϕeM M
− geM

1 − cos ϕe2 1 − cos ϕe2 1 − sin ϕe2 3
= −2Ωv0 sin ϕe1 + cos ϕe2 + Ω v0 t cos ϕe1 − sin ϕe2 − 2Ω v0 t cos ϕe1 − sin ϕe2 − geM
M M 2 M M 2 M M

3
= −2Ω v0 sin ϕeM M M M 2 M M M

1 + cos ϕe 2 + Ωv 0 t cos ϕe1 − sin ϕe 2 + Ω v 0 t cos ϕe 1 − sin ϕe 2 − ge3
= −2Ω × ẙ − Ω × (Ω × y) + g.
Therefore, the co-rotating observer “sees” that the particle is accelerated as if it obeyed the linear
momentum balance with fictitious forces included:
◦◦
my = mg − 2mΩ × ẙ − mΩ × (Ω × y) = F ext + FCoriolis + Fcentrifugal . (3.315)
The following schematic shows the apparent forces during the particle’s flight in the plane, i.e.,
omitting the y3 -component (note that the Euler force vanishes since Ω̇ = 0, but its hypothetical
direction is included only for reference).
y1 y(t) fug
al
cen
t r i
g
(Eu
ler
iolis
=0
)
Cor
net force
y~2
129
Example 3.28. Is the Earth an inertial frame?
Considering the Earth’s rotation about its own axis, one may ques- W
tion if the Earth can be treated as an inertial frame of reference. Let
us verify the influence of the apparent forces on particles moving on v
the Earth’s surface. The Earth’s radius and angular velocity are,
r
respectively, e3 M
e1 e2
M
M
2π 2π
R ≈ 6.4 · 106 m, Ω≈ = = 7.27 · 10−5 s−1 .
24h 24 · 3600s
Considering the individual apparent forces, we conclude that: R
• Euler force: since Ω = const., there is no Euler force: Ω̇M = 0 so that FEuler = 0.
• Coriolis force: for a particle of mass m moving with velocity v on the Earth’s surface:
|FCoriolis | = 2m|ΩM × v M | ≤ 2m|ΩM | |v M | ≈ 14.58 · 10−5 s−1 · mv. (3.316)
Therefore, in most applications the Coriolis force only matters when m is large (e.g., ocean
water) or if |v| is large (e.g., bullets, missiles). (It is for this reason that toilets flushing
clockwise vs. counter-clockwise on the northern vs. southern hemisphere is only a myth.)
• centrifugal force: for a particle of mass m on the Earth’s surface, we have
|Fcentrifugal | = m|ΩM × (ΩM × r M )| ≤ mΩ2 R ≈ 3.4 · 10−2 m/s2 · m. (3.317)
This force/acceleration, which is acting in radial direction (i.e., perpendicular to the Earth’s
axis of rotation into space) is essentially a correction to the gravitational acceleration. How-
ever, it is orders of magnitude smaller than gravity (9.81m/s2 ) and therefore usually negligible.
In conclusion, the Earth is close to an inertial frame for ordinary masses and velocities. However,
inertial forces may become important in case of fast traveling objects (see Example 3.29) or of large
masses moving on the Earth (see Example 3.30).
————
Example 3.29. Inertial forces on a fast traveling vehicle
Consider a vehicle (e.g., an airplane or a missile) traveling fast with velocity v M (measured relative
to the Earth’s surface) in straight Southern direction on the Northern hemisphere of the Earth, as
shown below schematically. What are the inertial forces acting on the vehicle?
The Earth is rotating with ΩM = Ωe3 , so that the motion of the vehicle (on the rotating Earth in
the co-rotating coordinate system shown below) results in inertial forces. The centrifugal force is
Fcentrifugal = −mΩM × (ΩM × r M ) = −mΩeM M M 2 M

3 × (Ωe3 × de1 ) = mΩ de1 , (3.318)
pointing radially away from the axis of rotation, as shown below.
130
e3 M
d e3 M
W W W
FCoriolis Fcentrifugal
e3
M
vM FCoriolis Fcentrifugal
v M
e1
M
e2 M
vM
+
FEuler= 0
Finally, the Coriolis force evaluates to

FCoriolis = −2mΩM × v M = −2mΩeM
3 ×v
M
= −2mΩ|eM M M
3 × v |e2 , (3.319)
perpendicular to both the axis of rotation and the relative velocity vector, as shown below. Finally,
since Ω = const., we know that FEuler = 0.
————
Example 3.30. The spinning of cyclones

A cyclone is a storm system whose spiral arrangement is
rapidly rotating and produces thunderstorms with heavy
rain and strong winds. (Depending on its location and
strength, a (tropical) cyclone is also known as hurricane,
typhoon, tropical or cyclonic storm.) The term “hurri-
vM
cane” usually refers to a tropical cyclone in the Atlantic
or northeastern Pacific Ocean. Any cyclone is character-
ized by a low-pressure center (the “eye”), which leads to
strong humid winds blowing towards the center.
This, in turn, leads to the migration of large masses of humid air towards the center of the cyclone
in a radial fashion, as demonstrated above and below (see the shown velocity v M ).
Let us follow the humid air masses dragged towards the center of the cyclone with some speed v M
(measured relative to the Earth’s surface which is rotating with the Earth at a constant angular
velocity ΩM = ΩeM 3 ). On the northern hemisphere, the relative velocity v
M along with the
Earth’s rotation ΩM results in a Coriolis force that deflects the air masses in a counter-clockwise
fashion (cf. the red arrows in the schematic below). This results in the shown counter-clockwise
rotation of hurricanes on the northern hemisphere (such as those forming over the Atlantic and
frequently moving towards the U.S. East Coast). By contrast, on the southern hemisphere (not
illustrated here) the Coriolis force results in the opposite (clockwise) rotation of cyclones (such as
those forming over the Pacific and threatening East Asia)14 .
14
As pointed out before, toilets flushing clockwise vs. counter-clockwise on the northern vs. southern hemisphere
is only a myth and not comparable, as the involved linear momentum mv is negligible.
131
e3M
northern hemisphere: southern hemisphere:

W
vM vM
FCoriolis FCoriolis
————
Example 3.31. Aboard a moving train

Consider a passenger on a high-sped train (moving with
a velocity v(t)), and assume the passenger is observing a
mechanical system; e.g., he is throwing a ball or watching e2M v
a mechanical pendulum on the moving train. In this case, e1
M
a convenient reference frame M is the one attached to the

OM
train. M is characterized by a moving coordinate origin
OM (which moves at the same speed v(t) as the train) but
its Cartesian basis does not rotate (ΩM = 0).
As a result, the balance of linear momentum (3.277) for the moving frame M reduces to
X
M aMCM = Fiext − M aOM . (3.320)
i
We realize that this equation is identical to that of an inertial, non-moving frame except for the
last term involving the acceleration of the moving coordinate origin. Further, notice that (3.320)
does not depend on the velocity vOM = v(t) of the origin, but it only depends on the acceleration
aOM = v̇(t) of the origin. Therefore, a train moving at constant speed v(t) = const. (so aOM = 0)
is in fact an inertial frame: the same laws of mechanics apply as in a non-moving frame. By
contrast, if the train’s acceleration is non-zero aOM 6= 0), it is a non-inertial frame and the last
term in (3.320) may become significant.
For example, throwing a particle of mass m on the moving train is governed by linear momentum
balance maM CM = m(g − aOM ), so the particle’s flight trajectory – as viewed from within the train
– is affected by the train’s acceleration. Specifically, aM
CM = g − aOM implies that the particle’s
apparent acceleration changes in the direction of the train’s movement. Recall that, of course, the
particle obeys the same physical laws on the train as on an inertial ground. The flight parabola
of the particle is exactly the same; it is just that the moving observer on the train sees a different
flight path when the train is accelerating or decelerating.
————
132
3.4.4 Balance of angular momentum, Euler equations
Having discussed linear momentum balance, let us proceed to derive the analogous relation for the
balance of angular momentum for a rigid body in a moving frame M in 3D. As discussed previously,
relation (3.259) for time derivatives in a moving frame applies to all vectors, so it specifically also
applies to the angular momentum vector:
ḢBC = H̊BM + ΩM × HB with HB = IB ω + M (rCM − rB ) × vB (3.321)
If we choose point B such that B = CM or vB = 0, then
HB = IB ω ⇒ ḢBC = (IB ω)◦M + ΩM × IB ω (3.322)
Note that ω (the angular velocity vector of the rigid body) and ΩM (the angular velocity vector of
the chosen rotating coordinate system M) do not have to be identical. We will see in the following
that it is convenient to choose ΩM = ω but this is not a requirement for those relations derived
here to hold (and sometimes we may indeed want to choose ΩM 6= ω, e.g., if multiple rotations are
involved).
The balance of angular momentum for a rigid body, formulated in the moving frame M, thus
follows as
MB = (IB ω)◦M + ΩM × IB ω if B = CM or vB = 0 (3.323)
If the M-frame rotates with the rigid body, i.e., if ΩM = ω, then
MB = (IB ω)◦M + ω × IB ω = IB ω̊ M + ω × IB ω, (3.324)
where we exploited that I˚BM = 0, since we compute the time derivative of the rigid body’s moment
of inertia tensor in its co-rotating reference frame (i.e., the coordinate system is rotating with the
body, so the moment of inertia tensor with respect to that reference frame is time-invariant).
In order to further simplify the last term in (3.324), it is wise to align the basis of M with the
principal directions of the rotating body B (which we denote by the principal frame M̂), so that the
moment of inertia tensor in this frame is diagonal:
ˆ   ˆ   ˆ
(I3 − Iˆ2 )ω3 ω2
 
I1 0 0 ω1 I1 ω1
[IB ]M̂ =  0 Iˆ2 0  ⇒ [ω × IB ω]M̂ = ω2  × Iˆ2 ω2  = (Iˆ1 − Iˆ3 )ω1 ω3  . (3.325)
0 0 Iˆ3 ω3 Iˆ3 ω3 (Iˆ2 − Iˆ1 )ω2 ω1
Therefore, if M̂ aligns with the principal axes of B, then in the M̂-frame


Iˆ1 ω̇1 + (Iˆ3 − Iˆ2 )ω3 ω2 = [MB,1 ]M̂ 
ωi = [ωi ]M̂


Iˆ2 ω̇2 + (Iˆ1 − Iˆ3 )ω1 ω3 = [MB,2 ]M̂ if ΩM = ω, and (3.326)
 ω̇i = [ω̊i ]M̂
Iˆ ω̇ + (Iˆ − Iˆ )ω ω = [M ]


3 3 2 1 2 1 B,3 M̂
Equations (3.326) are known as Euler’s equations of motion, valid in the so-called body frame
and for an arbitrarily-shaped body.
133
It is important to recall that in (3.326) all quantities are measured with respect to the
moving frame: [MB,i ]M̂ is the resultant torques around eM̂ i with respect to point B measured
ˆ
in the M̂-frame, Ii is the ith eigenvalue of IB in the M̂-frame, and ωi are the components [ω]M̂
of the angular velocity vector measured in the moving frame M̂. This particular frame M̂, which
aligns with the principal axes of body B and rotates with the same angular velocity (ΩM = ω) is
also referred to as the body frame.
Let us point out a common misconception. We introduced the following notation to compute time
derivatives with respect to moving and non-moving frames, which also applies to ω:
3
X 3
X 3
X
ωi (t)eM ω̇i eM M
ω̊ M = ω̇i eM

ω(t) = i (t) ⇒ ω̇ = i + ωi ėi , i . (3.327)
i=1 i=1 i=1
While ω̇ represents the absolute angular velocity (with respect to an inertial frame), ω̊ M is the
angular velocity measured with respect to the moving frame. The latter is the one which enters
the Euler equations (3.326), and its components are [ω̊i ]M̂ = ω̇i . It is important to realize that,
even though ΩM = ω (so the M-frame is rotating with the body), the time derivative ω̊ is not
necessarily zero. When we say “measured in the moving frame” or “measured with respect to the
moving frame”, we mean a quantity measured with respect to the moving coordinate basis. For
example, an object fixed on a rotating disk has a non-zero velocity v = ẋ with respect to an inertial
frame, but it has a vanishing velocity v̊ with respect to the corotating coordinate system M (an
observer sitting on the disk will not see any (relative) motion of the other points on the disk).
However, even though the observer is rotating with M, they will still feel a change in the angular
velocity. Therefore, even with respect to the corotating observer, the angular velocity of the disk
is changing and ω̊ 6= 0 unless ω = const.
Special case: axisymmetric bodies
In various problems, we will be dealing with rotating axisymmetric bodies. These present a special
case because the moment of inertia does not change, if the body is spinning about its axis of
rotational symmetry. Therefore, one may introduce a moving frame B which aligns with the
principal axes of the body but which does not spin with the body (so ΩB 6= ω). In this scenario,
we cannot use Euler’s equations directly but instead use (3.323), which here becomes
MO = (IO ω)◦B + ΩB × IO ω = IO ω̊ + ΩB × IO ω, (3.328)
where we exploited that both IO is time-invariant in the B-frame by axisymmetry (so I˚O
B = 0). In
components we have
 B  ˆ    ˆ B
I3 Ω2 ω3 − Iˆ2 ΩB

Ω1 I1 ω1 3 ω2
[ΩB × IO ω]B = ΩB 2
× Iˆ2  ω2  = Iˆ1 ΩB ω1 − Iˆ3 ΩB ω3  ,
3 1 (3.329)
ΩB
3 Iˆ3 ω3 Iˆ2 ΩB
1 ω 2 − ˆ1 ΩB ω1
I 2
so that the balance of angular momentum in the B-frame of the axisymmetric body becomes

Iˆ1 ω̇1 + Iˆ3 ΩB ω
2 3 − Iˆ Ω B
2 3 2 ω = [M ]
O,1 B 
ωi = [ωi ]B


ˆ ˆ B ˆ B
I2 ω̇2 + I1 Ω3 ω1 − I3 Ω1 ω3 = [MO,2 ]B for (3.330)
 ω̇i = [ω̊i ]B
Iˆ3 ω̇3 + Iˆ2 Ω1 ω2 − Iˆ1 Ω2 ω1 = [MO,3 ]B
B B 

134
These equations apply if the B-frame aligns with the principal axes of an axisymmetric body but
does not necessarily spin with the body about its axis of axisymmetry. If ΩB = ω, then B = M̂
and (3.330) reduces to Euler’s equations (3.326). This special case will become important when
discussing the dynamics of a spinning top in Section 3.5.
Example 3.32. Poorly aligned rotating shaft

A cylindrical mechanical shaft of length L and ra-
dius R is supposed to rotate with a constant an- a
a Bv B w
gular velocity ω about its longitudinal axis, sup- R
ported by two floating bearings. Unfortunately,
Av CM
the shaft and the bearings were poorly aligned,
A + B
so that the shaft’s symmetry axis is offset by an Bh
angle α from the axis of rotation, as shown on the a
right. What are the resultant forces acting on the AA L
h
two bearings at locations A and B?
Since the floating bearings allow for axial motion, they do not support any axial forces (along the
direction of the axis of rotation). Therefore, the forces A and B onto the bearings can only lie in
the plane perpendicular to the axis of rotation (as indicated by the shown horizontal and vertical
components). Since the center of mass of the cylindrical shaft can move only along the axis of
rotation, its acceleration is of the form a = aCM,k with aCM,k ||ω. Because there are no other
external forces, linear momentum balance tells that
A + B = M aCM,|| but A, B⊥aCM,|| ⇒ A = −B and a = aCM,|| = 0. (3.331)
That is, forces A and B are of the same magnitude and opposite orientation, so
Ah = −Bh , Av = −Bv , (3.332)
and the (linear) acceleration of the cylinder is zero. With this we have exhausted linear momentum
balance and turn to angular momentum balance to find the force magnitude(s).
Let us choose a rotating frame M that is aligned with e2M
the principal directions of the cylindrical shaft, specif-
ically with the eM
Av a Bv w
1 -axis aligned with the longitudinal
e3 e1 M
+
+
M
axis of the cylinder, as sketched. We here show a pro-
Ah R
jection of the cylinder in the plane of angle α (i.e., a Bh
L
top view of the above 3D drawing). a a
In this frame, the forces in the bearings are assumed to point in the shown directions; i.e., they
rotate with the body and the frame (so they always point in the shown direction relative to the tilt
axis). The moment of inertia tensor is diagonal and the components are those from Example 3.7,
viz.
M R2 M
Iˆ1 = Iˆ2 = Iˆ3 = L2 + 3R2 .

, (3.333)
2 12
Also expressed in the moving M-frame, the angular velocity vector ω is expressed as
ω = ω cos α eM M
1 − ω sin α e2 ⇒ ω1 = ω cos α, ω2 = −ω sin α, ω3 = 0. (3.334)
135
To compute the reaction forces in the bearings, we turn to Euler’s equations in the M-frame (which
is a body frame since it aligns with the principal axes and rotates with the body):
Iˆ1 ω̇1 + (Iˆ3 − Iˆ2 )ω3 ω2 = MB,1

Iˆ2 ω̇2 + (Iˆ1 − Iˆ3 )ω1 ω3 = MB,2 (3.335)
Iˆ3 ω̇3 + (Iˆ2 − Iˆ1 )ω2 ω1 = MB,3 ,
In the corotating frame ω1 , ω2 , ω3 from (3.334) are constant, so ω̇1 = ω̇2 = ω̇3 = 0. Further using
Iˆ2 = Iˆ3 from (3.333), Euler’s equations reduce to
MB,1 = 0, MB,2 = 0, MB,3 = (Iˆ2 − Iˆ1 )ω2 ω1 . (3.336)
The external torque components with respect to the center of mass follow from the forces onto the
bearings, shown above:
MB,1 = −Av a sin α + Bv a sin α, MB,2 = −Av a cos α + Bv a cos α, MB,3 = −Ah a + Bh a.
(3.337)
Combining (3.337) with (3.332) immediately shows that
Av = Bv = 0, MB,3 = 2Bh a, (3.338)
i.e., the forces onto the bearings act only in the plane of misalignment angle α.
Finally, inserting the external moment MB,3 from (3.338) into the third of Euler’s equations yields
M
MB,3 = 2Bh a = (Iˆ2 − Iˆ1 )ω2 ω1 = − L2 − 3R2 ω 2 sin α cos α.

(3.339)
12
and therefore (with 2 sin α cos α = sin 2α)
M M
L2 − 3R2 ω 2 sin α cos α L2 − 3R2 ω 2 sin 2α.

Bh = −Ah = − ⇒ |A| = |B| =
24a 48a
(3.340)
This shows that the misalignment of the shaft results in reaction forces in the bearings that grow
with misalignment angle α and quadratically with the angular velocity ω. Since significant forces
can arise in this fashion, proper alignment of the shaft is essential. Note that these reaction forces
A and B rotate with the shaft at angular velocity ω, as they were defined in the rotating M-frame.
But where do those reaction forces come from? Fcentrifugal
Simply put, the misalignment of the shaft leads to
pairs of centrifugal forces acting in opposite direc-
e2 M
tions. As an intuitive example, imagine the mis-

Av Bv w
e3 e1 M
+
+
aligned shaft as two particles, as shown on the right. M

The centrifugal forces onto the two particles are equal, Ah Bh
so no net force arises. Fcentrifugal L
However, the misalignment leads to a torque with respect to the center of mass, and that torque
must be balanced by the reaction forces. (Notice that no such torque would exist, if the shaft was
136
perfectly lined.) As discussed before, the extra terms in the Euler equations stem from inertial
forces. In fact, here the centrifugal forces are responsible.
————
Example 3.33. Pan mill (“Kollergang”)
A pan mill (Kollergang in German) is a circular wheel (radius R, mass M ) whose axis (length
a, assumed massless) makes it rotate around a fixed pole, as shown below. For simplicity, let us
assume that the wheel rolls without slipping and that it rotates with a constant angular velocity
Ω around the fixed hinge or pole. What is the reaction force N from the ground onto the wheel?
a
W
O R M
g w1
N
In a static problem, the answer would simply be N = M g. Here, however, dynamic effects con-
tribute. For convenience, let us choose a co-rotating basis {eM M M
1 , e2 , e3 } whose origin O
M is fixed
and coincides with the inertial frame’s origin O and whose coordinate axes align with the symmetry
axes, as shown below (with eM C
3 = e3 normal to the plane of rotation). We formulate all governing
equations in the moving frame M which rotates with the body (with Ω) around the fixed pole.
From Example 3.3 we know the instantaneous axis of
rotation is OA, so that we must have ω||rOA , and we
e3
M
know its component [ω3 ]M = [ω3 ]C = Ω (since eM3 = g
eC3 ). In the co-rotating frame, we thus have W
e2M
ω= ΩeM
3 + ω1 eM
1 . (3.341) Mg
OO
Angular velocity component ω1 is obtained from the
IAR C
e1M
kinematic constraint between the two rotations (ac-
counting for the proper signs by inspecting the rotation w1 F
A r
directions):
N
Ω R Ωa M
=− ⇒ ω = ΩeM
3 − e . (3.342)
ω1 a R 1
The kinematics of the problem is now sufficiently described by the known constant angular velocity
vector ω. To compute the normal force N , we need a kinetic balance equation. Linear momentum
137
balance introduces unnecessary reaction forces (e.g., the unknown reaction force O in the hinge in
the above sketch as well as the friction force Fr between the wheel and the ground), while angular
momentum balance around O can avoid the forces in the hinge. O is also fixed (vO = 0), which
avoids complicating terms in the angular momentum balance.
In the M-frame, the coordinate axes align with the symmetry axes of the cylinder, so we know,
using Steiner’s theorem, that the moment of inertia tensor of the wheel has components
   M R2 
0 0 0 2 0 0
[IO ]M = [IC ]M + M 0 a2 0  =  0 M 2 2
12 (L + 3R ) + M a
2 0 ,
0 0 a2 0 0 M 2 2
12 (L + 3R ) + M a
2
(3.343)
where we reused the components of IC computed in Example 3.7. Since IO is diagonal in the
chosen frame, we conclude that M is a principal frame of the rotating disk (i.e., the coordinate
axis coincide with the principal axes of the disk).
However, it is important to realize that the chosen M-frame (which rotates about the eC3 -axis with
a fixed eM C
3 = e3 ) does not rotate with the disk; specifically, we have
Ωa M
ΩM = ΩeM
3 whereas ω=− e + ΩeM
3 . (3.344)
R 1
Therefore, we unfortunately cannot use Euler’s equations but need to return to their original form
(3.323), which here becomes
MO = (IO ω)◦M + ΩM × IO ω = ΩM × IO ω, (3.345)
where we used that ω = const. and IO = const. in the chosen frame. Due to the diagonal form of
IO in the chosen M-frame and with [ω2 ]M = 0, this balance law of angular momentum reduces to,
completely expressed in the M-frame,
   ˆ
−Ω Ra
    
MO,1 0 IO,11 0 0 0
MO,2  =  0  ×  0 IÔ,22 0   0  = −IÔ,11 Ω2 Ra  (3.346)
MO,3 Ω 0 0 ˆ
IO,33 Ω 0
or
a
MO = Ωe3 × IO,11 (−Ω Ra )eM M
= −IO,11 Ω2 eM

1 + IO,33 Ωe3 . (3.347)
R 2
For the torques with respect to point O, we may neglect the reaction force O at the hinge, but we
must account for the torques due to the force N from the ground, gravity acting on the wheel, and
due to friction force Fr acting from the ground onto the wheel, which may be required to ensure
the rolling without slipping condition. From the above sketch we obtain
MO = −Fr ReM M M
1 + (M g − N )a e2 − Fr ae3 , (3.348)
Insertion into (3.347) lets us conclude that Fr = 0. Note that we assume Ω = const., which is why
no applied external torque is needed to keep the (conservative) system going; to conserve energy,
the wheel will keep rolling at constant angular velocity.
138
We hence obtain from (3.347) that
a MR 2
−IO,11 Ω2 = (M g − N )a ⇔ − Ω = M g − N. (3.349)
R 2
The reaction force from the ground follows as
M R Ω2
N = Mg + . (3.350)
2
Therefore, we conclude that the reaction force N is larger than that in the static case (N > M g).
In fact, the extra force given by the second term (which grows quadratically with the angular
velocity Ω) helps the milling process and is the rationale behind the pan mill design.
Finally, let us ask: where does the normal force
e3
M
M g
come from? As discussed before, it is inertial forces W
vM
that cause the extra terms in the Euler equations, which
here are responsible for the force N from the ground. As
e2
M
FCoriolis
the sketch illustrates, Coriolis forces must exist, because O
C
the wheel is rotating about its own axis, so that points e1M
on the wheel have a relative velocity v M with respect to
the M-frame’s rotation about the eM 3 -axis. We know Fr
the Coriolis force scales as FCoriolis ∼ −ΩM ×v M , which
gives the force directions shown in the sketch. vM N
Altogether, these Coriolis forces produce a positive net torque with respect to point O about the
eM
2 -axis, which must be balanced by force N from the ground.
————
Example 3.34. Pan mill revisited

As an alternative approach to the previous problem of the pan mill,
we may indeed use Euler’s equations but we need to choose a moving x3 R
frame different from M defined above. Recall that the Euler equations
can only be used if Ω = ω, so let us make this choice and denote the
new moving frame by R with
W
ΩR = ω. (3.351)
x2R
As a consequence, the coordinate system now rotates with the milling O a

wheel (i.e., eR M R
1 = e1 still points along the axis OA, but e2 and e3
R x
R 1
M
C
must rotate with the wheel, as shown on the right). In this case we mg R
can use Euler’s equations, but now we need to transform the angular
velocity vector ω (determined above in the M-frame) into the R- A
frame. This is accomplished by the coordinate transform N
[ω]R = [RRM ]T [ω]M , (3.352)
139
so that in the R-frame we have (with α measured positive counter-clockwise about eR

1 , as shown)
    a   a 
ω1 1 0 0 −RΩ −RΩ
[ω]R = ω2  = 0 cos α sin α   0  =  Ω sin α  . (3.353)
ω3 R 0 − sin α cos α Ω Ω cos α
Time differentiation in the rotating R-frame yields the angular acceleration vector (with α̇ = − Ra Ω)
as
       
ω̇1 ω1 0 0
ω̇2  = d ω2  =  Ωα̇ cos α  = −Ω2 a cos α . (3.354)
dt R
ω̇3 R ω3 R −Ωα̇ sin α Ω2 Ra sin α
Now, we may use Euler’s equations, when expressing all ingredients in the R-frame. [ω]R and [ω̇]R
are given above. The moment of inertia tensor is identical to (3.343) due to rotational symmetry,
i.e., [IO ]M = [IO ]R . We only need to compute the components of the resultant torque MO in the
R-frame, which is accomplished by a coordinate rotation, analogous to relation (3.352), giving
    
1 0 0 0 0
[MO ]R = 0 cos α sin α  (mg − N )a =  (mg − N )a cos α  . (3.355)
0 − sin α cos α 0 −(mg − N )a sin α
Overall, Euler’s equations (3.326) now take the form
(IÔ,33 − IÔ,22 )Ω2 cos α sin α = MO,1 = 0,

a a
−IÔ,22 Ω2 cos α − (IÔ,11 − IÔ,33 ) Ω2 cos α = MO,2 = (mg − N )a cos α,
R R
ˆ 2a ˆ ˆ a
IO,33 Ω sin α − (IO,22 − IO,11 ) Ω2 sin α = MO,3 = −(mg − N )a sin α.
R R
The first equation is satisfied trivially since IÔ,33 = IÔ,22 . Inserting IÔ,33 = IÔ,22 shows that the
second and third equations reduce to a single (duplicated) equation:
a a 1
IÔ,33 Ω2 − (IÔ,33 − IÔ,11 ) Ω2 = −(mg − N )a ⇔ IÔ,11 Ω2 = −(mg − N ), (3.356)
R R R
which is identical to (3.349). With IÔ,11 = 12 M R2 we again arrive at
M R 2 Ω2 1
= −(mg − N ) ⇔ N = mg + M RΩ2 . (3.357)
2 R 2
Hence, we obtain the same solution as in Example 3.33. As a summary, recall that we may choose
any (co-rotating) reference frame convenient to us. In Example 3.33 we chose the M-frame, in
which the resultant torque was simple to compute, but we could not use Euler’s equations. Here
by contrast, the frame was oriented conveniently so we could use Euler’s equations – at the cost
of having to calculate the torques less conveniently. If correctly done, both approaches lead to the
same solution, of course.
————
140
3.5 Application: Spinning Tops and Gyroscopes
3.5.1 General description
Spinning tops are interesting applications of the above kinetic relations derived for co-rotating
reference frames. A spinning top (Kreisel in German) is characterized by a free rotation in
3D about a stationary point O (i.e., although the body spins around multiple axis, point O is a
stationary, fixed point for all times). Let us consider a spinning top of mass M with center of
mass CM, whose tip is fixed at a point O (as shown below). The axis OC (through O and CM) is
a principal axis but not necessarily an axis of rotational symmetry.
Since the body is fixed at point O and therefore cannot
translate, the body’s motion can be described by a total of e3 =e3 A C
y
6−3 = 3 independent DOFs. For example, we may describe
its motion by three angles. Here we choose to use the so-
called Euler angles, which follow the so-called “3-1-3“ n
construction and characterize the motion of the body in
3D. For a spinning top on a flat ground (with normal eC3 ) CM + e2 A
we use the so-called precession angle ϕ (describing the

rotation around the eC3 -axis), the nutation angle ν (i.e.,
+
the tilt from the eC3 -axis), and the spin angle ψ (rotation O
about its own axis). All three angles are illustrated in the j
e1
A
schematic on the right.

In order to provide proper rotating frames to be used in the Euler equations later, let us define the
above three angles rigorously. We start with a Cartesian basis {eC1 , eC2 , eC3 } (with eC3 normal to the
ground) and describe each of the above angles by a new basis (shown in the figures below). Since
point O is stationary, all reference frames use the same origin O.
e3B
e3=e3
R B
e3A=e3C e3A=e3C e3A=e3C

y
n n
e2 B
e2
R
e2 B
CM + e2 A
CM + e2 A
CM + e2 A
+ + +
O O O e1 R
j e1A j e1=e1
B A j e1=e1
B
A A
precession nutation spin
1. First, we describe the precession of the body as a rotation from C to a new frame A with
basis {eA A A C C A
1 , e2 , e3 = e3 }, obtained by rotating the axes about e3 to bring e1 orthogonal to
C
the plane of rOC and e3 . The angle ϕ(t) of this rotation is the precession angle and results
in the body rotating with an angular velocity
ωϕ = ϕ̇eC3 = ϕ̇eA
3. (3.358)
141
2. Second, we describe the nutation of the body as a rotation from A to a new frame B with
basis {eB A B B A B
1 = e1 , e2 , e3 }, obtained by rotating the axes about e1 to align e3 with OC. The
angle ν(t) of this rotation is the nutation angle, which contributes the angular velocity
ων = ν̇eA B
1 = ν̇e1 . (3.359)
3. Third, we describe the spin of the body about its own axis as a rotation from B to a new
frame R with basis {eR R R B B R
1 , e2 , e3 = e3 }, obtained by rotating the axes about e3 to align e1
and eR2 with the principal axes. The angle ψ(t) of this rotation is the spin angle, which
describes a rotation with angular velocity
ωψ = ψ̇eR B
3 = ψ̇e3 . (3.360)
We have thus created three rotating reference frames A, B and R, which we may conveniently use
in the following. The total angular velocity of the rigid body is given by the superposition of all of
the above three rotations, viz.
ω = ωϕ + ων + ωψ = ϕ̇eA B R
3 + ν̇e1 + ψ̇e3 . (3.361)
To understand the kinetics, we could formulate linear momentum balance in any of the rotating
frames, e.g., in the B-frame (which is rotating but not spinning with the body). This leads to
dΩB
M aB B B
CM = Fext − 2M Ω × vCM − M
B
× rOC − M ΩB × (ΩB × rOC
B
). (3.362)
dt
For a freely spinning top, the only external forces (aside from gravity) can come through the fixed
origin O. Therefore, it will be more convenient to exploit angular momentum balance with respect
to O and formulate the associated Euler equations in one of the moving frames defined above,
e.g., in R with basis {eR R R
1 , e2 , e3 } and origin O. Since the R-axes are aligned with the principal
directions of the rotating body, its moment of inertia tensor [IO ]R is diagonal in this frame. Also
ΩR = ω. Therefore, the balance of angular momentum in the R-frame reduces to Euler’s equations
– with all vector and tensor components measured with respect to the R-frame:
Iˆ1 ω̇1 + (Iˆ3 − Iˆ2 )ω3 ω2 = MO,1 ,

Iˆ2 ω̇2 + (Iˆ1 − Iˆ3 )ω1 ω3 = MO,2 , (3.363)
Iˆ3 ω̇3 + (Iˆ2 − Iˆ1 )ω2 ω1 = MO,3 .
Note that the eigenvalues of the moment of inertia tensor are given by Steiner’s theorem as
Iˆ1 = [ICM,11 ]R + M |rO−CM |2 , Iˆ2 = [ICM,22 ]R + M |rO−CM |2 , Iˆ3 = [ICM,33 ]R , (3.364)
while the resultant torque with respect to O is

 
MO,1
MO = M ext + rOC × mg ⇒ MO,2  = [MO ]R . (3.365)
MO,3
142
Also measured in the R-frame, the angular velocity vector is obtained via coordinate transforma-
tions as
[ω]R = ϕ̇[eA B R
3 ]R + ν̇[e1 ]R + ψ̇[e3 ]R
       
sin ν sin ψ cos ψ 0 ϕ̇ sin ν sin ψ + ν̇ cos ψ (3.366)
= ϕ̇ sin ν cos ψ  + ν̇  sin ψ  + ψ̇ 0 = ϕ̇ sin ν cos ψ − ν̇ sin ψ  .
cos ν 0 1 ϕ̇ cos ν + ψ̇
Insertion of (3.364), (3.365) and (3.366) into Euler’s equations (3.363) yields a system of PDEs to
be solved for ϕ(t), ν(t) and ψ(t) with suitable initial conditions. Unfortunately, a general solution
for this problem is not feasible in closed form, which is why we proceed to study the most common
special cases that do admit an analytical treatment.
3.5.2 Axisymmetric bodies
One special case of interest in many applications is that of an axisymmetric body spinning about its
axis of rotational symmetry. In this case we can describe the rotation of the body conveniently in
the B-frame, since it is a principal frame of IO so [IO ]B is diagonal and also I˚O = 0 in the B-frame
(a spinning rotation of an axisymmetric body does not alter [IO ]B ). Furthermore, in the B-frame
we have
[ω]B = ϕ̇[eA B B
3 ]B + ν̇[e1 ]B + ψ̇[e3 ]B
         
0 1 0 ν̇ ω1 (3.367)
= ϕ̇ sin ν + ν̇ 0 + ψ̇ 0 =
       ϕ̇ sin ν  = ω2 

cos ν 0 1 ϕ̇ cos ν + ψ̇ ω3
and
     
ν̇ ν̈ ω̇1
d 
[ω̊]B = ϕ̇ sin ν  =  ϕ̈ sin ν + ϕ̇ν̇ cos ν  = ω̇2  .
 (3.368)
dt
ϕ̇ cos ν + ψ̇ ϕ̈ cos ν − ϕ̇ν̇ sin ν + ψ̈ ω̇3
Note that we now have ΩB 6= ω since

   B
ν̇ Ω1
[ΩB ]B = ϕ̇[eA ]
3 B + ν̇[eB
]
1 B =  ϕ̇ sin ν  = ΩB
2
. (3.369)
ϕ̇ cos ν B
Ω3
Therefore, we cannot use Euler’s equations directly but instead use (3.323), which here becomes
MO = (IO ω)◦B + ΩB × IO ω = IO ω̊ + ΩB × IO ω, (3.370)
as discussed in Section 3.4.4. Writing out angular momentum balance in the B-frame, cf. (3.330)
for an axisymmetric body with ωi = [ωi ]B and ω̇i = [ω̊i ]B becomes
Iˆ1 ω̇1 + Iˆ3 ΩB ˆ B

2 ω3 − I2 Ω3 ω2 = [MO,1 ]B ,
3 ω1 − I3 Ω1 ω3 = [MO,2 ]B , (3.371)
Iˆ3 ω̇3 + Iˆ2 ΩB ω2 − Iˆ1 ΩB ω1 = [MO,3 ]B .
1 2
143
Notice from (3.369) and (3.367) that ω1 = ΩB B B

1 and ω2 = Ω2 whereas ω3 = Ω3 + ψ̇. Axisymmetry
further implies that Iˆ1 = Iˆ2 . Therefore, the above angular momentum balance relations become
Iˆ1 (ω̇1 − ΩB B ˆ B B
3 Ω2 ) + I3 Ω2 (Ω3 + ψ̇) = MO,1 ,
Iˆ1 (ω̇2 + ΩB B ˆ B B
3 Ω1 ) − I3 Ω1 (Ω3 + ψ̇) = MO,2 ,
Iˆ3 ω̇3 = MO,3 .
Inserting the components ΩB

i from (3.369) and ω̇i from (3.368) finally yields
Iˆ1 (ν̈ − ϕ̇2 sin ν cos ν) + Iˆ3 ϕ̇ sin ν(ϕ̇ cos ν + ψ̇) = MO,1 ,
Iˆ1 (ϕ̈ sin ν + 2ϕ̇ν̇ cos ν) − Iˆ3 ν̇(ϕ̇ cos ν + ψ̇) = MO,2 , (3.372)

Iˆ3 ϕ̈ cos ν − ϕ̇ν̇ sin ν + ψ̈ = MO,3 .
In order to solve the system of ODEs in (3.372) for ϕ(t), ν(t) and ψ(t), we still need to apply
further assumptions, as the general case does not admit closed-form solutions. One such example
is the scenario of a steady precession, which occurs frequently (at least approximately) in practical
applications and which will be discussed next.
Example 3.35. Steady precession of an axisymmetric spinning top

Let us assume the special case of a steady precession, which implies
the spinning top’s motion has a solution of the form
j
ν = const., ϕ̇ = const., ψ̇ = const. (3.373)
y
Such a solution is typically found for axisymmetric spinning tops mov-
ing without any externally applied torques other than gravity. In this
case, applying (3.373) to (3.372) simplifies the equations dramatically
n=const.
since ν̇ = 0, ϕ̈ = ψ̈ = 0. Consequently, (3.372) reduces to
CM +

−Iˆ1 ϕ̇2 cos ν sin ν + Iˆ3 ϕ̇ sin ν ϕ̇ cos ν + ψ̇ = MO,1 ,
0 = MO,2 , (3.374) +O
0 = MO,3 .
This shows that steady precession is observed only if the external torques are parallel to the axis
of nutation, since the above implies that MO,2 = MO,3 = 0, so we must have
MO = MO,1 eB
1 (3.375)
for a steady precession. The steady precession may thus be understood as a steady-state solution,
in which the external torque due to gravity balances the gyroscopic angular acceleration about the
eB
1 -axis computed above, viz.

−Iˆ1 ϕ̇2 cos ν sin ν + Iˆ3 ϕ̇ sin ν ϕ̇ cos ν + ψ̇ = MO,1 (3.376)
144
As an example, consider gravity as the only force acting on the spinning top’s center of mass. This
results in
   
MO,1 mgrOC sin ν
MO = rOC × mg = mgrOC sin ν eB 1 ⇒ MO,2  =  0  (3.377)
MO,3 0
with rOC = |rOC |. Hence, (3.376) presents a relation between the three constant kinematic variables
ϕ̇, ψ̇ and ν:

Iˆ3 − Iˆ1 ϕ̇2 cos ν + Iˆ3 ϕ̇ψ̇ = mgrOC . (3.378)
A steady-state solution of the above form is most likely for a fast-spinning top (implying |ψ̇| |ϕ̇|).
In this case, (3.378) provides the precession rate for a given spin:
mgrOC
|ψ̇| |ϕ̇| ⇒ ϕ̇2 |ϕ̇ψ̇| ⇒ ϕ̇ ≈ (3.379)
Iˆ3 ψ̇
The limit |ψ̇| |ϕ̇| is known as the gyroscopic limit, since gyroscopes are characterized by having
one angular velocity component (ψ̇) significantly larger than all other components. In this limit,
one usually assumes that ψ̇ = const., so that – given a spin rate ψ̇ – the precession rate ϕ̇ follows
from (3.379).
We may find a similar relation more generally by considering the limiting case
|Iˆ3 ω3 | |Iˆ2 ω2 | ⇒ ΩB × IO ω ≈ ϕ̇eA ˆ R

3 × I3 ψ̇e3 . (3.380)
For the steady precession of an axisymmetric body (IO ω)◦B = 0, angular momentum balance be-
comes
MO = (IO ω)◦B + ΩB × IO ω = ΩB × IO ω ≈ ϕ̇eA ˆ R

3 × I3 ψ̇e3 , (3.381)
the orientations of the spin, precession, and torque must be related via
1 MO
ϕ̇eA R
3 × ψ̇e3 ≈ MO or, simply put, ϕ̇ × ψ̇ ≈ (3.382)
ˆ
I3 Iˆ3
where we abbreviated ϕ̇ = ϕ̇eA R

3 and ψ̇ = ψ̇e3 . Relation (3.382) is sometimes referred to as
the TSP-rule, linking the Torque axis, the Spin axis and the Precession axis of a fast-spinning
axisymmetric rigid body.
————
145
Example 3.36. Gyroscopes

A gyroscope is a spinning wheel (or disc) mounted
on a suspension such that its spin axis can freely as-
sume any orientation. For it to reorient freely, the
mount must be designed such that it cannot produce
any torques onto the spinning body (an example is
shown on the right). As a consequence, by angular mo- .
mentum conservation the spin axis must be unaffected y
by tilting or rotating the mount; i.e., the mount can be .
n
re-oriented in an arbitrary fashion – the spinning body (free to rotate)
will main a constant spin axis. It is for this reason that .
j (free to rotate)
gyroscopes are exploited, among others, in navigation
instruments since they can indicate a unique, absolute
direction independent of the orientation of the vehicle
onto which they are mounted.
Gyroscopes usually operate at the gyroscopic limit, i.e., they spin at very high speed ψ̇ compared
to any precession or nutation speed (and the spin is approximately constant).
For example, consider the flywheel shown on the right, which spins e3 A
.
with ψ̇ about its axis of rotational symmetry lying in the horizon- j
tal plane (ν = π/2). If the spinning wheel is initially held steady
with ψ̇ 6= 0 and then released, we know from (3.376) and the above .
y
derivation that H
V e3
B
mgrOC
ψ̇eR A
3 ⊥ ϕ̇e3 and ϕ̇ = . (3.383)
IO,33 ψ̇ mg
Hence, if spun counterclockwise (ψ̇ > 0 as shown), then the body will start to precede with ϕ̇ > 0
(as shown) in a counter-clockwise fashion. If spun clockwise (ψ̇ < 0), then the flywheel will precede
with ϕ̇ < 0 (as shown) in a clockwise fashion.
Interestingly, the flywheel is stable although the pin support at the center can only transmit (vertical
and in-plane) forces. These forces of magnitudes H and V , respectively, can be obtained from the
balance of linear momentum in the B-frame (which rotates with the body, so the motion of the center
of mass relative to the B-frame vanishes). For the case of a steady precession, linear momentum
balance gives
0 = Fext − M ΩB × (ΩB × rOC
B
) with ΩB = ϕ̇eA
e . (3.384)
For the particular case of a flywheel spinning around a horizontal axis of length rOC , the only
external forces are gravity and the support forces at O, so Fext = −mgeA B A
3 + He3 + V e3 (since
no other force components act in the e2 -direction, there is no support reaction in this direction
either). This leads to
−M (ϕ̇)2 rOC eB A B A
3 + mge3 = He3 + V e3 ⇒ H = −M (ϕ̇)2 rOC , V = mg. (3.385)
Therefore, V is a non-zero vertical force balancing the force due to gravity acting on the wheel. H
is a non-zero horizontal force balancing the centrifugal force on the wheel. (Note that it is because
146
of this latter horizontal force component that, when the wheel is supported by a string instead of
a pivot, the string cannot be perfectly vertical.)
We still want to answer the question: why does the flywheel
e3A
not fall down? Although forces are balanced, as discussed above,
.
there is no apparent torque balancing the action of gravity on the
j
wheel, yet it does not fall. Euler’s equations give the answer by the Fcfg
additional Ω × Iω-term which balances the torque due to gravity – FCoriolis
but what does it imply physically? v .
Recall that linear momentum balance (3.277) in the moving frame y
has additional components because of the frame’s rotation, the iner- e3B
tial/fictitious forces. We obtained angular momentum balance origi- v

nally from linear momentum balance and could, in principle, do the
same here. As a result, the balance equations of angular momentum mg FCoriolis
would have additional terms due to the torques produced by the in-
ertial forces, specifically the Coriolis and centrifugal forces (the Euler
force here is zero due to the constant angular velocities).
Fcfg
It is exactly these extra terms that make up those additional terms in the Euler equations. In this
particular example, we assumed |ψ̇| |ϕ̇|, so the dominant inertial forces are the Coriolis force due
to the spinning wheel in the preceding frame (FCoriolis = −2ΩB × v with v = ψ̇r and ΩB = ϕ̇) and
the centrifugal force due to the fast spinning (Fcfg = −ψ̇ 2 r), for any point at radius r on the wheel.
Note that each point on the flywheel experiences a different force (and integration over the wheel
is required to obtain the resultant forces). For a qualitative argument, we plotted above those two
force components for the top and bottom points on the wheel. While the centrifugal forces cancel,
the Coriolis force pair at the top and bottom of the wheel produces a torque which counteracts
gravity. Therefore, the torque due to the Coriolis forces acting onto the fast spun wheel balance
the torque due to gravity – and hence the wheel does not fall down.
e3B
HO e3
B
HO e3B
MO MO HO MO
TIME
We close by illustrating the reason for the precession of the flywheel. Recall that ḢO = MO , so
a change in angular momentum is related to a torque according to dHO = MO dt. Assume that
the flywheel is initially spinning about the eB
3 -axis as shown above on the left (as a top view from
above). The torque MO changes HO as shown. This in turn changes the orientation of MO , so
the torque keeps chasing the angular momentum vector. Consequently, the flywheel precedes.
147
Example 3.37. Intermediate axis theorem (tennis racket theorem)
As a final example, let us discuss the intermediate axis theorem (also known as the tennis
racket theorem). First seen in space by Russian cosmonaut Vladimir Dzhanibekov but theoret-
ically known for more than a century, this theorem describes the intriguing motion of a 3D rigid
body with three distinct principal moments of inertia. Instead of providing a rigorous derivation of
the underlying stability conditions here, we provide a qualitative explanation and leave the detailed
description to more advanced dynamics courses.
Consider a rigid-body in 3D without the action of any applied torques, so that Euler’s equations
become
Iˆ1 ω̇1 + (Iˆ3 − Iˆ2 )ω3 ω2 = 0, (3.386a)

Iˆ2 ω̇2 + (Iˆ1 − Iˆ3 )ω1 ω3 = 0, (3.386b)
Iˆ3 ω̇3 + (Iˆ2 − Iˆ1 )ω2 ω1 = 0. (3.386c)
Let us assume that Iˆ1 > Iˆ2 > Iˆ3 and consider a rotation about ω1 and small initial angular velocities
|ω2 | |ω1 |, |ω3 | |ω1 |. In this case (3.386a) implies that ω̇1 ∝ ω3 ω2 ≈ 0. Differentiating (3.386b)
with respect to time and inserting ω̇3 from (3.386c) yields
(Iˆ1 − Iˆ3 )(Iˆ2 − Iˆ1 ) 2

ω̈2 = ω1 ω2 (3.387)
Iˆ2 Iˆ3
which we rewrite as
(Iˆ1 − Iˆ3 )(Iˆ2 − Iˆ1 )
ω̈2 = Cω2 with C= < 0. (3.388)
Iˆ2 Iˆ3
From analysis we know that this ODE has a solution ω2 (t) which is of harmonic (sine/cosine) type.
Also, if the initial angular velocity ω2 is small, this oscillation is expected to be small.
The same argument holds true when considering a rotation with ω3 (we skip the analogous deriva-
tion).
By contrast, now consider a rotation about ω2 with initially |ω1 | |ω2 |, |ω3 | |ω2 |. In this case
(3.386b) implies that ω̇2 ≈ 0. Differentiating (3.386a) and inserting ω̇3 from (3.386c) yields
(Iˆ2 − Iˆ3 )(Iˆ1 − Iˆ2 ) 2

ω̈1 = ω2 ω1 (3.389)
Iˆ1 Iˆ3
which we rewrite as
(Iˆ2 − Iˆ3 )(Iˆ1 − Iˆ2 ) 2
ω̈2 = Dω2 with D= ω2 > 0. (3.390)
Iˆ1 Iˆ3
From analysis we know that this ODE has a solution ω1 (t) of exponential type. Therefore, initially
small values of ω1 can quickly grow exponentially, implying an instability.
In summary, this shows that, for a rigid body with three distinct principal moments of inertia, an
initial angular velocity around the principal axes corresponding to the largest or smallest moments
148
of inertia leads to a stable rotating motion (possibly superimposed by stable oscillations of bounded
amplitude). When rotating about the principal axis corresponding to the intermediate moment of
inertia, however, any small initial perturbation results in an unstable tumbling motion.
Typical examples of bodies with distinct principal moments of inertia are books but also tennis
rackets (therefore the eponymous theorem) and the famous T-handle in space.
A question for you: which of the three configurations is prone to show tumbling?
————
The following boxes summarize the key relations derived for rigid bodies in 3D (and special cases
in 2D). We here extend our concepts from particles to continuous bodies and also include the case
of non-inertial reference frames.
149
velocity and acceleration transfer formulae:
vB = vA + ω × rAB
aB = aA + ω̇ × rAB + ω × (ω × rAB )
center of mass and instantaneous center/axis of rotation (from a point P):
1 1 ω × vP
Z
rCM = rρ dV, rPΠ = eω × vP =
M B ω ω2
balance of linear momentum if M = const.:

X d X
Fiext = Ṗ = (M vCM ) ⇒ Fiext = M aCM
dt
i i
balance of angular momentum with respect to an arbitrary point B:
MB = ḢB + vB × P
angular momentum with respect to a point B on body B:
HB = IB ω + M (rCM − rB ) × vB
moment of inertia tensor (B ∈ B serves as coordinate origin):

 2
x2 + x23 −x1 x2 −x1 x3

Z
[IB ] =  −x1 x2 x21 + x23 −x2 x3  ρ dV
B −x1 x3 −x2 x3 x21 + x22
parallel axes theorem (Steiner’s theorem) with ∆x = rB − rCM :
(∆x2 )2 + (∆x3 )2
 
−∆x1 ∆x2 −∆x1 ∆x3
[IB ] = [ICM ] + M  −∆x1 ∆x2 (∆x1 )2 + (∆x3 )2 −∆x2 ∆x3 
−∆x1 ∆x3 −∆x2 ∆x3 (∆x1 )2 + (∆x2 )2
balance of angular momentum for B ∈ B if B = CM or vB = 0 and if I˙B = 0:

in 2D
MB = IB ω̇ ⇒ MB = IB ω̇ with IB = ICM + M (∆x)2
angular momentum transfer formula for arbitrary points A and B:
HB = HA + P × rAB
centroidal moments of inertia in 2D:

M L2 M R2
slender rod : ICM = , disk/cylinder : ICM =
12 2
150
kinetic energy of a rigid body for a point C ∈ B with C = CM or vC = 0:

1 1
T = M |vC |2 + ω · IC ω
2 2
work–energy balance:
XZ ri (t2 ) Z ϕ(t2 )
T (t2 ) − T (t1 ) = W12 W12 = Fiext · dri + M ext · dϕ
i ri (t1 ) ϕ(t1 )
time derivative relation between an inertial frame C and a non-inertial frame M:
ẏ C = ẙ M + ΩM × y
balance of linear momentum in a moving frame M:
M aM
CM = Fext + FCoriolis + FEuler + Fcentrifugal − M aOM
Coriolis, Euler and centrifugal forces:

FCoriolis = −2M ΩM × vCM
M
,
dΩM M
FEuler = −M × rCM ,
dt
Fcentrifugal = −M ΩM × (ΩM × rCM
M
)
balance of angular momentum in a moving frame M:
MB = (IB ω)◦M + ΩM × IB ω if B = CM or vB = 0
Euler equations in a rotating (principal) body frame M̂ (ΩM̂ = ω):


Iˆ1 ω̇1 + (Iˆ3 − Iˆ2 )ω3 ω2 = [MB,1 ]M̂ 
ωi = [ωi ]M̂


Iˆ2 ω̇2 + (Iˆ1 − Iˆ3 )ω1 ω3 = [MB,2 ]M̂ if ΩM = ω and
 ω̇i = [ω̊i ]M̂
Iˆ ω̇ + (Iˆ − Iˆ )ω ω = [M ]


3 3 2 1 2 1 B,3 M̂
angular momentum balance in a rotating (principal) frame B if I˚ = 0 (ΩB̂ 6= ω):


2 ω3 − I2 Ω3 ω2 = [MB,1 ]B̂ 
ωi = [ωi ]B̂


ˆ ˆ B ˆ B
I2 ω̇2 + I1 Ω3 ω1 − I3 Ω1 ω3 = [MB,2 ]B̂ where
 ω̇i = [ω̊i ]B̂
Iˆ3 ω̇3 + Iˆ2 ΩB ω2 − Iˆ1 ΩB ω1 = [MB,3 ]


1 2 B̂
TSP-rule for a fast-spinning top:

MO
ϕ̇ × ψ̇ ≈
Iˆ3
151
4 Vibrations
In several of our early examples we arrived at an equation of motion, viz. a partial differential
equation which we said had to be solved with appropriate initial conditions – without actually
solving those (see, e.g., Examples 1.16, 1.17 or 3.26). While the solution of such equations of
motion may be quite complicated in general, we can indeed derive analytical solutions for some
important classes of problems of engineering interest; one such example are mechanical vibrations.
Before discussing vibrations, however, let us first introduce the helpful concept known as Lagrange
equations.
4.1 Lagrange Equations
Let us consider a system of N particles whose positions are denoted by {r1 , . . . , rN } (we can later
generalize our derivations to rigid bodies, which may be interpreted as collections of infinitely many
rigidly connected particles, as discussed before). As mentioned for systems of N particles, we may
introduce any holonomic constraint as
ri = ri (q1 , . . . , qn , t) for i = 1, . . . , N, (4.1)
where (q1 , . . . , qn ) are so-called generalized degrees of freedom that are independent (and
n ≤ d × N in d dimensions). For a general system with k independent constraints and d × N
degrees of freedom, we thus identify n = d × N − k generalized degrees of freedom. We can think of
these generalized degrees of freedom as any set of (e.g., translational and/or rotational) independent
kinematic variables that uniquely define the state of a system. Their choice may not necessarily
be unique (and we should, as always, aim for a convenient choice), but the number of generalized
degrees of freedom is fixed as n = d × N − k for a system of particles.
Lagrange’s equations, which we derive in the following, are an oftentimes convenient shortcut to
deriving the equations of motion for those generalized DOFs. We note that there are multiple ways
to arrive at Lagrange’s equations (and various approaches are found in textbooks). Here, we use a
simple (but easy to follow) derivation as follows.
For a system of N particles, Newton’s second law may be formulated as

∂V
mi r̈i = Ficons. + Finon-cons. = − + Finon-cons. for i = 1, . . . , N, (4.2)
∂ri
where we accounted for both conservative forces Ficons. , deriving from a potential V = V (r1 , . . . , rN ),
and non-conservative forces Finon-cons. . Multiplying (4.2) by the respective particle velocity ṙi and
summing over all particles leads to
N
X ∂V non-cons.
0= ṙi · mi r̈i + − Fi , (4.3)
∂ri
i=1
which must obviously hold for a system of particles satisfying the equations of motion (4.2). Note
that we may reinterpret this latter equation by considering, for simplicity, a conservative system
152
whose total energy must remain constant over time, so

d X mi
T + V = const. ⇒ (T + V ) = 0 with T = |ṙi |2 . (4.4)
dt 2
i
Inserting the definition of the kinetic energy and applying the time derivative yields
N N N
d X X ∂V X ∂V
0= (T + V ) = mi ṙi · r̈i + · ṙi = ṙi · mi r̈i + , (4.5)
dt ∂ri ∂ri
i=1 i=1 i=1
so we recover (the conservative part of) (4.3) (which can be extended to non-conservative forces
as shown above). One could also multiply (4.2) by a virtual displacement δri instead of ṙi , which
arrives at an equivalent form known as the principle of virtual work. Here, we simply accept that
(4.3) holds.
Let us apply a few algebraic manipulations. First, we use the constraints (4.1) to write
N N n n X
N n
X ∂V X ∂V X dṙi X ∂V dri X ∂V
· ṙi = · q̇j = · q̇j = q̇j (4.6)
∂ri ∂ri dq̇j ∂ri dqj ∂qj
i=1 i=1 j=1 j=1 i=1 j=1
with the generalized velocity q̇j . Further, we use that

n n
d2 ri X ∂ri

X d ∂ri d ∂ri
mi r̈i · ṙi = mi 2 · q̇j = mi ṙi · − ṙi · q̇j
dt ∂qj dt ∂qj dt ∂qj
j=1 j=1
n n
d ∂ |ṙi |2 ∂ |ṙi |2

X d ∂ ṙi ∂ ṙi X
= mi ṙi · − ṙi · q̇j = mi − q̇j (4.7)
dt ∂ q̇j ∂qj dt ∂ q̇j 2 ∂qj 2
j=1 j=1
n
|ṙi |2

X d ∂Ti ∂Ti
= − q̇j with Ti = mi
dt ∂ q̇j ∂qj 2
j=1
being the kinetic energy of particle i. Next, we define the non-conservative generalized force
N
X ∂ri
Qnc
j = Finon-cons. · (4.8)
∂qj
i=1
conjugate to the generalized coordinate qj , such that

N n X
N n
X X ∂ri X
Finon-cons. · ṙi = Finon-cons. · q̇j = Qnc
j q̇j . (4.9)
∂qj
i=1 j=1 i=1 j=1
Altogether, we thus arrive at

N n
X "N #
X ∂V X d ∂Ti ∂Ti ∂V
non-cons. nc
0= ṙi · mi r̈i + − Fi = − + − Qj q̇j . (4.10)
∂ri dt ∂ q̇j ∂qj ∂qj
i=1 j=1 i=1
153
Since (4.10) must be satisfied for any generalized velocities and since the generalized coordinates
qi (and hence velocities q̇i ) are independent, we conclude that we must have
N N
!
d X ∂Ti ∂ X
+ V − Ti − Qncj =0 for j = 1, . . . , n. (4.11)
dt ∂ q̇j ∂qj
i=1 i=1
If we exploit that V = V (q1 , . . . , qn ) is independent of the generalized velocities q̇i so ∂V /∂ q̇j = 0

for all j = 1, . . . , n, we may define the so-called Lagrangian
N
X |ṙi |2
L=T −V with T = mi (4.12)
2
i=1
such that

d ∂L ∂L
− = Qnc
i for i = 1, . . . , n with L=T −V (4.13)
dt ∂ q̇i ∂qi
These are the so-called Lagrange equations, which are equivalent to the equations of motion for
the generalized degrees of freedom. If the system is conservative, then Qnc
i = 0 for each i = 1, . . . , n
and the Lagrangian L alone determines the equations of motion. Otherwise, we recall that
N
X ∂rj
Qnc
i = Fjnon-cons. · (4.14)
∂qi
j=1
The Lagrange equations provide a convenient alternative to LMB and AMB, since they allow us
to obtain the equations of motion of a mechanical system without the need to formulate balance
laws. All that is needed are an admissible choice of the generalized degrees of freedom as well as
the total potential and kinetic energies. Then, (4.13) yields the equations of motion directly by
differentiation. Examples will follow after the discussion of a special case – mechanical equilibrium.
4.2 Mechanical Equilibrium
In the special case of static problems, the Lagrange equations with T = 0 reduce to
∂V
= Qnc
i for i = 1, . . . , n, (4.15)
∂qi
which – when written out for a specific system – is equivalent to the vanishing sums of all forces
and torques, hence describing mechanical equilibrium.
Furthermore, in the special case of a static and conservative system (Qnc

i = 0) we have
∂V
=0 for i = 1, . . . , n. (4.16)
∂qi
154
This implies that an equilibrium of a conservative system is a stationary point of its potential
energy. The simplest example to picture this relation is a particle of mass m exposed to gravity
in a mountain landscape of varying altitude h(x), so V (x) = mgh(x). The only places where the
particle can come to rest (i.e., is in equilibrium) are mountain valleys and mountain tops, where
the slope vanishes, so ∇V = mg ∇h = 0. Of course, we understand that mountain valleys are
stable equilibria (after perturbation, the particle remains in the valley), whereas mountain tops
are unstable equilibria (once perturbed, the particle will roll into the next valley and not return).
Whether or not an equilibrium is stable can generally be verified by inserting a found solution
{q 1 , . . . , q n } into the Hessian matrix with components
∂2V
Hij = , (4.17)
∂qi ∂qj
to determine whether the found optimum is an energy minimum or maximum. A minimum implies
a stable equilibrium and corresponds to a positive-definite Hessian. By contrast, a maximum
is an unstable equilibrium and corresponds to a negative-definite Hessian. Saddle points are
metastable equilibria.
For the special case of a system with a single degree of freedom q and total potential energy V (q),
we thus conclude:
∂2V
stable equilibrium ⇔ energy minimum ⇔ >0
∂q 2
(4.18)
∂2V
unstable equilibrium ⇔ energy maximum ⇔ <0
∂q 2
Example 4.1. Vibrating mass–spring system
Consider a particle of mass m attached to a linear spring of stiffness k in 1D and a dashpot with
velocity-proportional viscosity d, so there is only one DOF q = x. The dashpot produces a drag
force Qnc = −dẋ = −dq̇, while the kinetic and potential energy for the system shown below read
1 1 1 1
T = mq̇ 2 , V = −mgq + kq 2 ⇒ L = T − V = mq̇ 2 + mgq − kq 2 , (4.19)
2 2 2 2
where we define q such that the spring is unstretched when q = 0.
The static equilibrium position q0 is obtained as
∂V mg k q0 g
0= = −mg + kq0 ⇒ q0 = ,
∂q q=q0 k d
and we can easily verify that this is a stable equilibrium since
x
∂ 2 V /∂q 2 = k > 0. m
The equation of motion (e.o.m.) is obtained as Lagrange’s equation, which here reduces to

d ∂L ∂L
−dq̇ = − = mq̈ − mg + kq ⇔ mq̈ + dq̇ + kq = mg. (4.20)
dt ∂ q̇ ∂q
155
This is a second-order, linear, inhomogeneous ordinary differential equation (ODE) to be solved

for q(t) with appropriate initial conditions. Note that if we introduce the degree of freedom as
x(t) = q(t) − q0 (so x = 0 is the equilibrium position), then the EoM becomes mẍ + dẋ + kx = 0.
————
Example 4.2. Equilibrium of a particle in a rotating arm

The system shown on the right consists of a particle of
mass m frictionlessly sliding in a massless arm of length
a with a spring of stiffness k1 and unstretched length g
b
l0,1 = a/4. The tip of the arm is connected to a vertical k2
spring of stiffness k2 and undeformed length l0,2 = b. k1
Considering gravity acting downwards as shown, what are m j
the equilibrium configurations of the system? y
a/4
We start by formulating the total potential energy as a func- x
tion of the angle ϕ and the elongation x of the spring within
the arm (these are our chosen two DOFs that uniquely de- a
scribe the system):
1 1
V = −mg (l0,1 + x) sin(ϕ) + k1 x2 + k2 (a sin ϕ)2 (4.21)
2 2
In order to find the equilibria of this static, conservative system, we compute
∂V mg
= −mg sin ϕ + k1 x = 0 ⇒ x= sin ϕ
∂x k1

∂V 2 mg 2
= −mg (l0,1 + x) cos ϕ + k2 a sin ϕ cos ϕ = −mg l0,1 + sin ϕ + k2 a sin ϕ cos ϕ = 0,
∂ϕ k1
whose solution is
π mgak1
ϕ=± ∨ sin ϕ = (4.22)
2 4 (k1 k2 a2 − m2 g 2 )
along with
mg m2 g 2 a
x0 = ± ∨ x0 = . (4.23)
c1 4 (k1 k2 a2 − m2 g 2 )
It is intuitive (and can be verified by second derivatives) that the ϕ = ± π2 solutions are unstable
equilibria, while the second solution in (4.22) presents a stable equilibrium angle.
————
Example 4.3. Swinging pendulum
Consider pendulum consisting of a particle of mass m attached to a hinged massless rod of length
l. What are the equilibria, and what is the equation of motion of the system?
156
Here, the kinetic and potential energy with the single DOF q = ϕ are
1 1
T = mv 2 = m(lq̇)2 , V = −mgl cos q. (4.24) g
2 2 l
j massless
This leads to the Lagrangian
1
m
L = T − V = m(lq̇)2 + mgl cos q. (4.25)
2
The equilibrium positions follow from
∂V
0= = mgl sin q0 ⇒ q0 = 0 ∨ q0 = π. (4.26)
∂q q=q0
Using second derivatives reveals whether or not these equilibria are stable:
∂2V
stable ⇔ energy minimum ⇔ >0
∂q 2
(4.27)
∂2V
unstable ⇔ energy maximum ⇔ <0
∂q 2
Here, we have
∂2V ∂2V ∂2V
= mgl cos q ⇒ = mgl > 0, = −mgl < 0 (4.28)
∂q 2 ∂q 2 q=0 ∂q 2 q=π
As expected, q0 = 0 is a stable equilibrium, while q0 = π is an unstable equilibrium.
The equation of motion is derived as Lagrange’s equation for the single DOF q, viz.

d ∂L ∂L
0= − = ml2 q̈ + mgl sin q ⇒ lq̈ + g sin q = 0. (4.29)
dt ∂ q̇ ∂q
This is a nonlinear ODE to be solved for the unknown function q(t). Since this is difficult in
general, one often turns to the special case of assuming small oscillations around a stable equilibrium
(e.g., imagine the pendulum is initially at rest and then perturbed by a small deflection to cause
oscillatory motion around the equilibrium). In such cases, we may assume q(t) = q0 + x(t) with
a stable equilibrium q0 and perturbation a |x(t)| 1. Using a Taylor series to approximate the
trigonometric term around q0 = 0 yields
x2 q0 =0
sin(q0 + x) = sin q0 + x cos q0 − sin q0 + O(x3 ) ⇒ sin(q) ≈ x + O(x3 ). (4.30)
2
d 2
Further using q̈(t) = dt2 [q0 + x(t)] = ẍ(t), the equation of motion reduces to a linear, homogeneous,
second-order ODE (independent of the mass m):
g
ẍ + x ≈ 0, (4.31)
l
which is to be solved for x(t) with appropriate initial conditions. We will discuss solutions to
equations of this type in the next section.
————
157
4.3 Single-Degree-of-Freedom Vibrations
4.3.1 Definitions and equation of motion
Generally speaking, vibrations for our purposes are small oscillations around a stable equilibrium.
We have seen that the equations of motion that govern such oscillations are usually differential
equations. We hence need to solve those differential equations with suitable initial conditions in
order to find the vibrational motion of a particle or body.
The prototypical equation of motion for a vibration with a single degree of freedom (see, e.g.,
Examples 4.1 and 4.3) has the form15
mẍ + cẋ + kx = F (4.32)
This is a second-order, linear ODE in time, which is a homogeneous ODE if F = 0 (this case is
termed free vibration) or otherwise inhomogeneous ODE if F 6= 0 (this case is referred to as
forced vibration). Solving the ODE for x(t) results in the solution in its so-called state space
representation (which refers to finding x as a function of time t).
For convenience, let us introduce the quantities

r
c k F (t)
δ= ≥ 0, ω0 = ≥ 0, f (t) = (4.33)
2m m m
to rewrite the second-order ODE (4.32) in its normalized form:
ẍ + 2δ ẋ + ω02 x = f (4.34)
Note that we may reduce any linear system of second-order to a first-order system as follows. We
introduce the new degree-of-freedom vector y (so-called phase space representation) such that

y1 = x y1
⇒ [y] = . (4.35)
y2 = ẋ y2
Consequently, the second-order ODE (4.34) becomes a first-order system of ODEs, viz.

0 1
ẏ = Ay + F with [A] = , (4.36)
−ω02 −2δ
which is to be solved for y(t) and usually visualized in the (y1 , y2 )-phase space.
The general solution of the equation of motion (4.34) is obtained by starting with an exponential
ansatz with unknowns A, λ ∈ C, whose insertion into the equation of motion (4.34) yields
x(t) = Aeλt λ2 + 2δλ + ω02 x(t) = f (t).

⇒ (4.37)
Depending on the choice of δ and f , we obtain different types of solutions. Several cases will be
discussed in the following, including free vibrations (f = 0) and forced vibrations (f 6= 0).
15
Even though the equation of motion may be much more complex in general (writing, e.g., g(ẍ, ẋ, x) = F (t) with
some nonlinear function g), it can always be linearized to the form (4.32) through Taylor expansions of g. Therefore,
(4.32) presents the general linearized equation of motion for small-amplitude oscillations.
158
4.3.2 Free vibrations
If F (t) = 0 and hence f (t) = 0, then we are concerned with a free vibration; i.e., the system
vibration is caused only by the initial conditions, there is no applied time-dependent forcing that
promotes vibrations. In this case, Eq. (4.37) simplifies to the so-called characteristic equation
q
λ2 + 2δλ + ω02 = 0 ⇒ λ1,2 = −δ ± δ 2 − ω02 . (4.38)
Depending on the values of δ and ω0 , this admits four types of solutions, which will be discussed
in the following in detail. It is convenient to introduce Lehr’s damping ratio
δ c
D= = , (4.39)
ω0 2mω0
which defines the four cases discussed below.
• Case 1: undamped vibration (D = 0)

If there is no damping, then δ = 0 and the above solutions turn into a pair of purely imaginary
values for λ, viz.
q √
δ=0 ⇒ λ1,2 = ± −ω02 = ±iω0 where i = −1, ω0 ∈ R. (4.40)
The solution is hence of the form (with complex coefficients C1 , C2 ∈ C)
x(t) = C1 exp(iω0 t) + C2 exp(−iω0 t). (4.41)
Using Euler’s identity, we transform the above into the alternative form
x(t) = C1 [cos(ω0 t) + i sin(ω0 t)] + C2 [cos(ω0 t) − i sin(ω0 t)]

(4.42)
= (C1 + C2 ) cos(ω0 t) + i(C1 − C2 ) sin(ω0 t).
Since the solution must be real-valued (i.e., we seek x(t) ∈ R), we must have
Im(C1 + C2 ) = 0, Re(C1 − C2 ) = 0 s.t. C1 + C2 , i(C1 − C2 ) ∈ R. (4.43)
This allows us to introduce two alternative (real-valued) coefficients
A1 = C 1 + C 2 ∈ R and A2 = i(C1 − C2 ) ∈ R, (4.44)
so that
x(t) = A1 cos(ω0 t) + A2 sin(ω0 t) (4.45)
with constants A1 , A2 ∈ R. Note that we may use the trigonometric identity A sin(ω0 t+ϕ0 ) =
A cos(ϕ0 ) sin(ω0 t) + A sin(ϕ0 ) cos(ω0 t) to introduce two alternative unknowns A, ϕ0 ∈ R and
define
A1
q
A1 = A sin(ϕ0 ), A2 = A cos(ϕ0 ) ⇒ A = A21 + A22 , tan ϕ0 = . (4.46)
A2
159
such that we now have the two alternative forms of the solution
x(t) = A1 cos(ω0 t) + A2 sin(ω0 t) = A sin(ω0 t + ϕ0 ) with Ai , A, ϕ0 ∈ R. (4.47)
ω0 is the eigenfrequency (also known as fundamental frequency or natural frequency)

of the system’s free vibration. A is the amplitude of the vibration. The time period of a
complete vibration cycle is
T = 2π/ω0 . (4.48)
The coefficients in the solution (4.47) are to be found from appropriate initial conditions; for
example, if x(0) = x0 and ẋ(0) = v0 , then A1 = x0 and A2 = v0 /ω0 . Note that, while A1 and
A2 (or A and ϕ) depend on the initial conditions, the eigenfrequency ω0 only depends on the
system characteristics and not on the initial conditions. Shown below is an example of a free,
undamped vibration.
2p
T=
w0
x 1.5
0.5
0
x(t)
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80
0 tt
Case 1: undamped vibration
• Case 2: overdamped vibration (D > 1)

In case of significant damping with δ > ω0 , two distinct, real eigenvalues λ1 , λ2 ∈ R exist:
q q
δ > ω0 ⇒ λ1 = −δ − δ 2 − ω02 < λ2 = −δ + δ 2 − ω02 < 0. (4.49)
Consequently, the solution with constants A1 , A2 ∈ R (again to be found from initial condi-
tions) becomes
q
x(t) = A1 eλ1 t + A2 eλ2 t with λ1,2 = −δ ± δ 2 − ω02 and Ai ∈ R. (4.50)
This solution decays exponentially with time (since λ1 , λ2 < 0) without displaying harmonic
vibrations. For example, if x(0) = x0 and ẋ(0) = v0 , then A1 , A2 are determined from
A1 + A2 = x0 and A1 λ1 + A2 λ2 = v0 . Shown below is an example of a free, overdamped
vibration for different cases of v0 .
160
0.12
v0 > 0
0.1
v0 = 0
0.08
v0 < 0
v0 < 0
x
0.06
x(t)
0.04
0.02
0 -0.02
tt
0 50 100 150
0
Case 2: overdamped vibration
• Case 3: critically damped vibration (D = 1)

In the special case of a critically damped system we have
δ = ω0 ⇒ λ1 = λ2 = −δ = −ω0 , (4.51)
so only one unique, real eigenvalue exists. The solution is therefore of the form
x(t) = A1 e−δt + A2 te−δt with A1 , A2 ∈ R. (4.52)
The constants are again to be found from initial conditions. Since the second law of ther-
modynamics requires δ > 0 in general, this solution decays exponentially with time without
displaying harmonic vibrations. Shown below is an example of a free, critically damped
vibration for different cases of the initial velocity v0 .
0.12
v0 > 0
0.1
v0 = 0
0.08
v0 < 0
v0 < 0
x
0.06
x(t)
0.04
0.02
0 -0.02
0 2 4 6 8 10 12 14 16 18 20
0 tt
Case 3: critically damped vibration
• Case 4: underdamped vibration (0 < D < 1)

In this final case of low damping, we obtain two complex eigenvalues:
q q
0 < δ < ω0 ⇒ λ1,2 = −δ ± −(ω0 − δ ) = −δ ± i ω02 − δ 2 = −δ ± iωd (4.53)
2 2
p
with ωd = ω02 − δ 2 ∈ R>0 . Expanding the exponentials in the solution and rewriting terms
as done for undamped vibrations (see Case 1 above) leads to
q
x(t) = e−δt [A1 cos(ωd t) + A2 sin(ωd t)] with ωd = ω02 − δ 2 A1 , A2 ∈ R. (4.54)
161
Coefficients A1 and A2 are again to be found from initial conditions.

In this case, we observe vibrations with the damped natural frequency ωd < ω0 so that
the period extends to
Td = 2π/ωd > T. (4.55)
The amplitude of underdamped vibrations decays exponentially, and that decay grows with δ.
Shown below is an example of a free, underdamped vibration along with the definition of Td .
2p 2p
Td = >T = w
wd 0
x 2.5
1.5
0.5
x(t)
0 0
-0.5
-1
-1.5
-2
tt
0 10 20 30 40 50 60 70 80 90 100
0
Case 4: underdamped vibration
Example 4.4. Vibration of a rolling cylinder

Consider a cylindrical body of mass m and radius R, which
+
is rolling without slipping on a curved ground having the
shape of a circular arch with radius r, as shown on the right. g j
What is the eigenfrequency of the cylinder when oscillating r
about its equilibrium position?
m
y
Let us use Lagrange’s equation to derive the equation of mo- R
tion. The kinetic and potential energy and the resultant La- A
grangian are (ψ being the rolling angle of the cylinder) +
1 1
T = IA ψ̇ 2 , V = mg(r−R)(1−cos ϕ) ⇒ L = T −V = IA ψ̇ 2 −mg(r−R)(1−cos ϕ) (4.56)
2 2
with IA = 23 mR2 (where we exploit that contact point A is the ICR) and the kinematic constraint
Rψ = (r − R)ϕ due to rolling without slipping. Consequently, the equation of motion for the single
DOF ϕ follows as
d ∂L ∂L (r − R)2 3
0= − = IA ϕ̈ + mg(r − R) sin ϕ = m(r − R)2 ϕ̈ + mg(r − R) sin ϕ, (4.57)
dt ∂ ϕ̇ ∂ϕ R2 2
which we re-arrange into
2g
ϕ̈ + sin ϕ = 0. (4.58)
3(r − R)
162
Obviously, ϕ = 0 is a stable equilibrium in this problem (and a minimizer of the potential energy).
Hence, we linearize the above equation of motion with ϕ = 0 + x about ϕ = 0 for |x| 1, which
leads to
2g
ẍ + x ≈ 0. (4.59)
3(r − R)
From comparing this equation of motion to the general form (4.34), we identify the eigenfrequency
of this system as
s
2g
ω0 = . (4.60)
3(r − R)
We further note that the system is undamped and without applied forcing, so we except an un-
damped free vibration with the above eigenfrequency.
————
Example 4.5. Vibration of a spring-mass-damper system

We consider the system shown on the right, consisting of a cylinder (mass M ,
radius r) connected to a linear spring (stiffness k) and dashpot (viscosity c)
in series and, via a massless inextensible rope, to a particle of mass m and
the ground. If the particle is given an initial downward speed v0 (and if the
viscosity c is very low), what is the resulting motion of the system?
k c
g
To find the equation of motion, we define the kinetic and potential energy as
j A
1 1
T = mẋ2 + IA ϕ̇2 , V =
k 2
y , (4.61) r y
+
2 2 2
so the Lagrangian follows as M
1 1 k
L = T − V = mẋ2 + IA ϕ̇2 − y 2 , (4.62) x
2 2 2 m
where we exploit that point A is the ICR, and IA = 32 M r2 . The kinematic constraints of the system
reveal that
x ẋ ẋ
x = 2y, x = 2rϕ ⇒ y= , ẏ = , rϕ̇ = . (4.63)
2 2 2
Insertion of those relations into the Lagrangian reduces the latter to depend only on a single DOF,
here we choose x:
2 k x2

1 2 13 2 ẋ 1 3 k
L = mẋ + Mr 2
− = m + M ẋ2 − x2 . (4.64)
2 22 4r 2 4 2 8 8
The non-conservative force in the dashpot is
∂y
Qnc
x = −cẏ , (4.65)
∂x
163
where the sign stems from the force acting against coordinate y, and the partial derivative from
the definition (4.14).
Lagrange’s equation yields

d ∂L ∂L 3 k ∂y c
0= − = Qnc
x ⇒ m+ M ẍ + x = −cẏ = − ẋ (4.66)
dt ∂ ẋ ∂x 8 4 ∂x 4
or, after some re-arrangement,

c k
ẍ + 3 ẋ + x = 0. (4.67)
4m + 2 M 4m + 23 M
| {z } | {z }
=2δ =ω 2
Comparing this equation of motion with the general form (4.34), we find
s
k c
ω0 = 3 , δ= (4.68)
4m + 2 M 8m + 3M
and further
r
δ 8m + 3M
D= =c . (4.69)
ω0 k
As discussed above, depending on the value of D, the vibration can be underdamped (D < 1),
overdamped (D > 1), or critically damped (D = 1).
For example, in case of

plow damping, the underdamped vibration manifests in a decaying oscillation
with frequency ωd = ω02 − δ 2 and the general solution (4.54):
x(t) = e−δt [A1 cos(ωd t) + A2 sin(ωd t)] . (4.70)
Applying the initial conditions x(0) = 0 and ẋ(0) = v0 leads to
x(0) = A1 = 0,
(4.71)
ẋ(0) = −δA1 + ωd A2 = v0 ⇒ A2 = v0 /ωd ,
so that the particle moves according to

v0 −δt
q
x(t) = e sin(ωd t) with ωd = ω02 − δ 2 ≤ ω0 . (4.72)
ωd
Notice that the vibration frequency in the damped case is lower than that of the undamped case.
As a result, the period of vibration T = 1/ω is longer in the damped case than in the undamped
case.
————
164
4.3.3 Forced vibrations
If the forcing f (t) is non-zero, then we deal with a so-called forced vibration. In this case, the
system is not free to vibrate but literally forced to vibrate through an applied force or torque that
varies periodically in time (consider, e.g., an engine that imparts a cyclic load onto its surroundings).
Here, the equation of motion is the inhomogeneous ODE
ẍ(t) + 2δ ẋ(t) + ω02 x(t) = f (t), (4.73)
whose solution is to be found as
x(t) = xhom (t) + xpart (t). (4.74)
The homogeneous solution xhom (t) is the same as the solution derived above for a free vibration.
Note that, except for the undamped case, xhom (t) always decays with time. The particular
solution xpart (t) depends on f (t) and survives in the presence of damping for all times. Therefore,
we may expect that for vibrations with δ > 0 the solution after long times is
x(t) ≈ xpart (t) as t → ∞. (4.75)
Shown below is an example of a solution x(t) = xhom (t) + xpart (t), indicated in red, consisting of
the homogeneous solution xhom (t) and the particular solution xpart (t) in response to a sinusoidal
loading, with xpart (t) indicated in blue. For nonzero damping, the homogeneous solution vanishes
exponentially over time, so that the general solution x(t) approaches the particular solution xpart (t).
3.5
2.5
x part(t)
x (t)
2
1.5
x 1
0.5
-0.5
-1
-1.5
t
0 2 4 6 8 12 14 16 18 20
In many real-world applications, the external forcing f (t) is time-harmonic with a known excitation
frequency. For example, consider vibrations caused to rotating machine or vehicle parts. Therefore
(and because it allows for analytical solutions), we will consider an applied time-dependent force16
f (t) with an excitation frequency Ω ∈ R, the most general form of which is
f (t) = fˆ cos(Ωt) + ĝ sin(Ωt) where fˆ, ĝ ∈ R (4.76)
are constant coefficients that define the amplitude of the excitation force. Note that, here and
in the following, we usually refer to f (t) as the excitation force, even though the concepts apply
equally to problems in which x is a rotational DOF, so that f (t) is in fact an applied excitation
torque.
16
Here and in the following, we simply speak of the excitation force f (t), even though it is, strictly speaking, force
per mass as f = F/m.
165
Brief mathematical interlude:
Before we present the solution to the above problem of a forced vibration, let us quickly introduce
(or review) a mathematical trick that will prove helpful here. For convenience (to avoid the use of
trigonometric relations), we exploit Euler’s identity
exp(iΩt) = cos(Ωt) + i sin(Ωt), (4.77)
so that we may write

h i
f (t) = fˆ cos(Ωt) + ĝ sin(Ωt) = Re (fˆ − iĝ) exp(iΩt) (4.78)
since, by (4.77),
h i
(fˆ − iĝ) exp(iΩt) = fˆ cos(Ωt) + ĝ sin(Ωt) + i −ĝ cos(Ωt) + fˆ sin(Ωt) . (4.79)
That is, we may rewrite (4.76) as

h i
f (t) = Re fˆ∗ exp(iΩt) with fˆ∗ ∈ C, (4.80)
whereby we replace the coefficients fˆ, ĝ ∈ R by a complex-valued unknown fˆ∗ ∈ C, whose real and
imaginary components define the amplitudes fˆ and ĝ. To differentiate the two, here and in the
following we use an asterisk (∗ ) to denote all quantities in the complex domain.
An important aspect to realize is that we are solving a linear ODE:

d2 d
ẍ(t) + 2δ ẋ(t) + ω02 x(t) = f (t) ⇔ L[x(t)] = f (t) with L= 2
+ 2δ + ω02 (4.81)
dt dt
being a linear differential operator. As a consequence, we may replace both F (t) and x(t) by
complex-valued functions F ∗ (t) and x∗ (t), so that, if x∗ (t) is a solution to
L[x∗ (t)] = f ∗ (t), (4.82)
then we also know (from taking real and imaginary parts of this latter equation) that
Re L[x∗ (t)] = L Re[x∗ (t)] = Re[f ∗ (t)].

(4.83)
In other words, when inserting f ∗ (t) and solving for x∗ (t), then we know that Re[x∗ (t)] is a solution
to the problem with excitation Re[f ∗ (t)], whereas – analogously – Im[x∗ (t)] is a solution to the
problem with excitation Im[f ∗ (t)]. Therefore, we will use the complex excitation
f ∗ (t) = fˆ∗ exp(iΩt) (4.84)
and solve for the complex solution x∗ (t), from which we conclude that the sought solution is
x(t) = Re[x∗ (t)]. (4.85)
In the following, we will follow this strategy to avoid the use of trigonometric functions.
————
166
Let us insert the general complex function f (t) = fˆ∗ exp(iΩt) into the governing ODE (4.73)
and solve for the resulting complex solution x∗ (t), from which we may conclude that the sought
particular solution for the forced-vibration problem is
xpart (t) = Re[x∗part (t)]. (4.86)
Insertion of F ∗ (t) = F̂ ∗ exp(iΩt) into the equation of motion (assuming the homogeneous solution
decays) yields
ẍ∗part (t) + 2δ ẋ∗part (t) + ω02 x∗part (t) = fˆ∗ exp(iΩt)t. (4.87)
Since this equation must hold for all times t, the solution must be of the form
x∗part (t) = x̂∗ exp(iΩt) with x̂∗ ∈ C. (4.88)
Insertion into (4.87) leads to
−Ω2 + 2iΩδ + ω02 x̂∗ exp(iΩt) = fˆ∗ exp(iΩt) ∀ t,

(4.89)
so that
fˆ∗ fˆ∗ ω02 − Ω2 fˆ∗ (−2Ωδ)

∗ ≡ x̂r + ix̂i .
x̂ = = + i
−Ω2 + 2iΩδ + ω02 (ω02 − Ω2 )2 + 4Ω2 δ 2 (ω02 − Ω2 )2 + 4Ω2 δ 2

Recall that, by Euler’s identity, (a + ib) = |a + ib| exp[i arctan(b/a)]. Therefore, we write

∗ ∗ ∗ x̂i
xpart (t) = x̂ exp(iΩt) = (x̂r + ix̂i ) exp(iΩt) = |x̂ | exp(iΩt) exp i arctan
x̂r (4.90)
∗
= |x̂ | exp [i(Ωt − ϕ)]
with the phase delay

x̂i 2Ωδ
ϕ = − arctan = arctan (4.91)
x̂r ω02 − Ω2
and the amplitude
|fˆ∗ |
q
|x̂∗ | = x̂2r + x̂2i = p 2 . (4.92)
(ω0 − Ω2 )2 + 4Ω2 δ 2
The phase delay indicates that the motion xpart (t) lags behind the applied forcing f (t). As can be
expected, the phase delay vanishes in case of zero damping (i.e., when δ = 0).
It is important to notice that both phase delay and amplitude depend on the excitation frequency
Ω. The amplitude peaks at resonance when Ω = ω0 , i.e., when the system is driven at the natural
frequency. At resonance (and also near resonance), the amplitude grows significantly but remains
bounded as long as δ 6= 0. If there is no damping (δ = 0), then the amplitude tends to infinity at
resonance. As Ω is increased from below through the resonance, the phase delay jumps from 0 to
π at resonance.
167
Introducing the frequency ratio η and re-using Lehr’s damping ratio D, defined as
Ω δ
η= , D= (4.93)
ω0 ω0
we may write the amplitude of (the particular solution of) a forced vibration as
|fˆ∗ | ω02 1
|x̂∗ | = V (D, η) with V (D, η) = p 2 =p (4.94)
ω02 2 2 2
(ω0 − Ω ) + 4Ω δ 2 (1 − η )2 + 4D2 η 2
2
where V is the so-called magnification factor (if V = 1, then the amplitude of the static response
with Ω = 0 is recovered). Using the same dimensionless ratios, the phase delay becomes

2Ωδ 2Dη
ϕ = arctan 2 = arctan (4.95)
ω0 − Ω2 1 − η2
Shown below is the amplification factor V (η; D), plotted as functions of the frequency ratio η
for various dimensionless damping values D. The maximum magnification for a given D-value is
achieved when
p p
η = 1 − 2D2 ⇔ Ω = ω0 1 − 2D2 . (4.96)
The dashed line indicates the curve of the maximum amplification V for the shown values of D.
3.5
D=0
3
D=0.15
2.5
D=0.2
2
D=0.3
V
1.5
D=0.5
1
0.5
D=1
0
-0.5
-1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
h
The next graphic below illustrates the phase shift ϕ(η; D), plotted as functions of the frequency
ratio η for various dimensionless damping values D. Note that for D = 0 the phase shift jumps at
the natural frequency ω0 (i.e., at η = 1) from 0 to π.
168
p D=0
D=0.15 D=0.2
D=0.3
in phase D=0.5
D=0.7
D=1
p/2
j
out of phase
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
h
Finally, recall that the solution required the recovery of only the real part of x∗part (t). Therefore,
e.g., if the applied excitation is of the form
f (t) = fˆ cos(Ωt) ⇔ fˆr = fˆ, fî = 0, (4.97)
then the particular solution is obtained as
fˆ
xpart (t) = Re x∗part (t)

⇒ xpart (t) = V (D, η) cos (Ωt − ϕ) (4.98)
ω02
By analogy, we may conclude that
fˆ
f (t) = fˆ sin(Ωt) ⇒ xpart (t) = V (D, η) sin (Ωt − ϕ) . (4.99)
ω02
Finally, recall that the complete solution of a forced vibration is
x(t) = xhom (t) + xpart (t) (4.100)
where, as discussed before, the homogeneous solution is that of a free vibration (which was derived
in Section 4.3.2), which decays over time if δ > 0.
We note that in the special case of resonance of an undamped system, i.e., if Ω = ω0 for the
case of δ = 0, the above general solution does not apply. As long as δ 6= 0, even if small, the above
relations apply. Yet, if δ = 0, we obtain the particular solution from solving the governing equation
(written in its complex form)
ẍ∗part (t) + ω02 x∗part (t) = f ∗ (t) = fˆ∗ exp(iω0 t), (4.101)
169
for which x∗part (t) = x̂∗ exp(iω0 t) is clearly not a solution (it renders the left-hand-side of (4.101)
zero). In this special case, the solution has the form
x∗part (t) = x̂∗ t exp(iω0 t). (4.102)
To verify this, let us insert (4.102) into (4.101). Note that
ẋ∗part = x̂∗ (1 + iω0 t) exp(iω0 t) ẍ∗part = x̂∗ 2iω0 − ω02 t exp(iω0 t),

and (4.103)
which yields upon insertion into (4.101)
fˆ∗ fˆ∗
x̂∗ 2iω0 − ω02 t + ω02 t exp(iω0 t) = fˆ∗ exp(iω0 t) x̂∗ =

⇒ = −i . (4.104)
2iω0 2ω0
Therefore, the special case of undamped resonance leads to, using −i = exp(−iπ/2),
fˆ∗ fˆ∗
x∗part (t) = −i t exp i(ω0 t − π2 ) ,

t exp(iω0 t) = (4.105)
2ω0 2ω0
which is out-of-phase by ϕ = π/2 (as in the damped cases above at resonance), and the magnifica-
tion is given by
|x̂∗ |ω02 ω0
V = = . (4.106)
|fˆ∗ | 2
As an example, if the system is excited by
f (t) = fˆ cos(ω0 t), (4.107)
then the undamped system response is
fˆ
x(t) = xhom (t) + t sin(ω0 t), (4.108)
2ω0
whose homogeneous contribution does not decay over time (in the absence of damping).
Importantly, the amplitude of the particular contribution in (4.108) grows linearly with time (of
course, only within the assumptions made to arrive at the linearized equation of motion). Therefore,
undamped forced vibrations quickly lead to excessive amplitudes at resonance (i.e., if the excitation
frequency coincides with the resonant frequency).
Why does resonance lead to excessive vibrations? Consider the simple example of pushing
a child on a swing-set. Without pexcitation it oscillates back and forth approximately at the natural
frequency of a pendulum, ω0 = g/l. In order to push the child higher, how would one optimally
excite the system? Of course, it’s intuitive to push the child at that exact same frequency –
e.g., whenever it reaches the front- or back-most point (but one would not push at some random
frequency while it is swinging in between). In order for the pushing to boost the child’s amplitude,
the forcing frequency should match the eigenfrequency. Admittedly, this is a simplistic analogy,
but it illustrates the concept quite effectively.
170
Example 4.6. Unbalanced rotation
Let us consider a motor of mass M sitting on a machine bed (approximated as a linear spring of
stiffness k and a linear dashpot of viscosity c). A particle of mass m is attached at a distance
d eccentrically to the massless rotor arm of the motor, which is rotating with a constant angular
velocity Ω, as schematically shown below. What is the system response?
In order to find the equation of motion of the system, we draw free-
body diagrams, as shown below on the right. Linear momentum M
balance for the large motor with DOF x pointing upwards reads d m Wt
x
M ẍ = −kx − cẋ + S sin ϕ, ϕ = Ωt, (4.109)
where S denotes the magnitude of the force in the rotor arm, and
ϕ = Ωt is the angle of the arm with the horizontal axis, as shown. To
relate the motion of the two masses, let us formulate the kinematic
relation k/2 c k/2
y = x + d sin(Ωt) ⇒ ÿ = ẍ − dΩ2 sin(Ωt), (4.110)
which indicates that the vertical motion y(t) of the particle is the sum
y
of the vertical motion x(t) of the large mass and the relative motion x M Sj m
due to the rotation of the particle. Linear momentum balance for the S
j=Wt
eccentric particle reveals
.
kx+cx
mÿ = −S sin(Ωt), (4.111)
which may be solved for S, inserting (4.110), to yield
S sin(Ωt) = −mẍ + mdΩ2 sin(Ωt). (4.112)
Finally, insertion into (4.109) leads to the equation of motion
(M + m)ẍ + cẋ + kx = mdΩ2 sin(Ωt) (4.113)
or, equivalently,
c k m
ẍ + ẋ + x= d Ω2 sin(Ωt) . (4.114)
M + m
| {z } M + m
| {z } M
| + m {z }
=2δ =ω02 =f (t)=fˆ sin(Ωt)
Comparing this equation of motion with the general form (4.34) yields
r √
k c δ c M +m Ω
ω0 = , δ= ⇒ D= = √ , η= . (4.115)
M +m 2(M + m) ω0 2 k ω0
The general solution is
x(t) = xhom (t) + xpart (t) (4.116)
171
with the homogeneous solution being that of a free damped vibration (the exact nature of the
solution depends on the value of D, as discussed in Section 4.3.2). The particular solution takes
the form (4.99), here
fˆ mdΩ2
xpart (t) = V (D, η) sin (Ωt − ϕ) = V (D, η) sin (Ωt − ϕ) . (4.117)
ω02 k
The maximum vibration amplitude is reached when

p
Ω = ω0 1 − 2D2 . (4.118)
————
4.4 Multi-DOF Vibrations
Systems of particles or bodies described by more than one generalized DOF may produce vibrations
referred to as multi-DOF vibrations. Consider a system described by n generalized degrees of
freedom. For convenience, we introduce vectors including all DOFs, including the vectors of
generalized DOFs: q(t) ∈ Rn ,

equilibrium positions: q0 ∈ Rn , (4.119)
n
deviations from equilibrium: x(t) = q(t) − q0 ∈ R .
As for the single-DOF case, vibrations for our purposes are understood as small oscillations about a
stable equilibrium q0 . To this end, we may linearize the (generally nonlinear) equations of motion
about a stable equilibrium and use the resulting linear equations of motion as the basis of our
analysis.
4.4.1 Equations of motion for multi-DOF vibrations
To arrive at the linearized equations of motion, we generally have two alternatives:
• We derive the (nonlinear) equations of motion (e.g., via the Lagrange equations, or from lin-
ear/angular momentum balance) and then linearize the nonlinear equations of motion about
the equilibrium.
• We expand the (nonlinear) kinetic and potential energies about the equilibrium to quadratic
order and then obtain the linearized equations of motion from the Lagrange equations.
We will briefly outline both approaches here for completeness.
Following the first approach, let us assume the (generally nonlinear) equations of motion can be
written as
F (q, q̇, q̈) = F ext (t), (4.120)
172
for which a stable equilibrium is identified to satisfy F (q0 , 0, 0) = 0. F ext (t) are small external
forces applied after equilibrium to excite vibrations. A linearization of the equations of motion
about the equilibrium via Taylor expansion leads to
∂F ∂F ∂F
F (q0 , 0, 0) + (q0 , 0, 0)x + (q0 , 0, 0)ẋ + (q0 , 0, 0)ẍ + h.o.t. = F ext (t) (4.121)
∂q ∂ q̇ ∂ q̈
or approximately (inserting F (q0 , 0, 0) = 0 by equilibrium and dropping all higher-order terms)
M ẍ + C ẋ + Kx = F ext (4.122)
with the three constant matrices17

∂F ∂F ∂F
M= (q0 , 0, 0), C= (q0 , 0, 0), K= (q0 , 0, 0), (4.123)
∂ q̈ ∂ q̇ ∂q
known as mass matrix, damping matrix and stiffness matrix, respectively. Note that these
definitions, in general, do not reveal anything about the structure or properties of the three matrices.
This will be different in the second approach demonstrated below, which accounts for the origin of
the equations of motion.
The second approach is a bit more elaborate as we start at the energetic level. Consider a kinetic
energy of the form T = T (q, q̇) as well as a potential energy V = V (q). For small deviations x
around a stable equilibrium q0 so that q(t) = q0 + x(t), we may Taylor expand the energies around
an equilibrium (q0 , q̇0 ) with q̇0 = 0. We expand up to quadratic order, so that differentiation in
the Lagrange equations leads to linear equations of motion. For the kinetic energy this leads to
∂T 1 ∂2T ∂T
T (q, q̇) = T (q0 , 0) + (q0 , 0) · x + x · (q0 , 0)x + (q0 , 0) · ẋ
∂q 2 ∂q∂q ∂ q̇
(4.124)
1 ∂2T 1 ∂2T
+ ẋ · (q0 , 0)ẋ + x · (q0 , 0)ẋ + h.o.t.
2 ∂ q̇∂ q̇ 2 ∂q∂ q̇
Note that for physical reasons we may expect that T (q, 0) ≡ 0 for vanishing velocities q̇ = 0, i.e.,
if no motion occurs. Therefore, we may set
∂T ∂2T
T (q0 , 0) = 0, (q0 , 0) = 0, (q0 , 0) = 0, (4.125)
∂q ∂q∂q
leading to
∂T 1 ∂2T 1 ∂2T
T (q, q̇) ≈ (q0 , 0) · ẋ + ẋ · (q0 , 0)ẋ + x · (q0 , 0)ẋ. (4.126)
∂ q̇ 2 ∂ q̇∂ q̇ 2 ∂q∂ q̇
Similarly, we consider conservative forces represented by a potential energy V = V (q). Taylor

expansion about equilibrium up to quadratic order gives
∂V 1 ∂2V
V ≈ V (q0 ) + (q0 ) · x + x · (q0 )x. (4.127)
∂q 2 ∂q∂q
17
The vector derivatives are to be evaluated component-wise; e.g., M = ∂F /q̈ is a short, symbolic notation for
the components Mij = ∂Fi /∂ q̈j . The other derivatives are defined analogously.
173
If non-conservative forces Qnc = Qnc (q, q̇) + F ext (t) act on the system, their approximation to
leading order is obtained from a Taylor expansion up to linear order (since Qnc enters the Lagrange
equations directly without differentiation, we truncate to linear order):
∂Qnc ∂Qnc
Qnc (q, q̇) ≈ Qnc (q0 , 0) + (q0 , 0)x + (q0 , 0)ẋ + F ext (t). (4.128)
∂q ∂ q̇
Note that we include time-dependent external forces F ext (t) here, which excite the system and do
not depend on q or q̇. Finally, we apply the Lagrange equations to the approximate forms (4.126),
(4.127) and (4.128). The first term in the Lagrange equations evaluates to
∂2T 1 ∂2T

d ∂L d ∂T d ∂T
= ≈ (q0 , 0) + (q0 , 0)ẋ + (q0 , 0)x
dt ∂ q̇ dt ∂ ẋ dt ∂ q̇ ∂ q̇∂ q̇ 2 ∂q∂ q̇
(4.129)
∂2T 1 ∂2T
= (q0 , 0)ẍ + (q0 , 0)ẋ,
∂ q̇∂ q̇ 2 ∂q∂ q̇
where we exploited that none of the derivatives depend on time, since they are evaluated at (q0 , 0).
Similarly, we obtain the second term in the Lagrange equations as
∂L ∂T ∂V 1 ∂2T ∂V ∂2V
= − ≈ (q0 , 0)ẋ − (q0 ) − (q0 )x.
∂q ∂x ∂x 2 ∂q∂ q̇ ∂q ∂q∂q
Altogether, the Lagrange equations

d ∂L ∂L
− = Qnc (4.130)
dt ∂ q̇ ∂q
now turn into
∂2T ∂V ∂2V
(q0 , 0)ẍ + (q0 ) + (q0 )x
∂ q̇∂ q̇ ∂q ∂q∂q
(4.131)
∂Qnc ∂Qnc
= Qnc (q0 , 0) + (q0 , 0)x + (q0 , 0)ẋ + F ext (t).
∂q ∂ q̇
Since the Lagrange equations at equilibrium dictate that
∂V
(q0 ) = Qnc (q0 , 0), (4.132)
∂q
the above further reduces to
∂2T ∂Qnc
2
∂Qnc

∂ V
(q0 , 0)ẍ − (q0 , 0)ẋ + (q0 ) − (q0 , 0) x = F ext (t). (4.133)
∂ q̇∂ q̇ ∂ q̇ ∂q∂q ∂q
Finally, we define the matrices
∂2T ∂Qnc ∂2V ∂Qnc
M= (q0 , 0), C=− (q0 , 0), K= (q0 ) − (q0 , 0), (4.134)
∂ q̇∂ q̇ ∂ q̇ ∂q∂q ∂q
so that the linearized equations of motion become
M ẍ + C ẋ + Kx = F ext (4.135)
174
which is identical to the equations of motion derived in (4.122) using the first approach.
No matter how we derive the linearized equations of motion, they involve the following quantities:
M ∈ Rn×n : mass matrix (pos.-definite, symmetric, constant)

n×n
C∈R : damping matrix (usually pos.-semi-definite, symmetric and constant)
n×n (4.136)
K∈R : stiffness matrix (usually pos.-semi-definite, symmetric and constant)
F ext (t) ∈ Rn : external force vector (time-dependent).
The matrix properties follow from their definitions (see, e.g., Eq. (4.134)). By usually in (4.136)
we mean all cases considered here and in the following, i.e., we will assume that both C and K
are symmetric, positive-semi-definite and constant. This is, e.g., the case if Qnc = Qnc (q̇) = −C q̇
with a constant, symmetric, positive-semi-definite C as for linear viscous drag. However, this may
not be the case in general (e.g., if complex non-conservative forces are involved, in case of contact
problems or non-smooth dynamics, etc.). Yet, we will not consider such cases in this course.
Note that based on the above matrices, we can also express the kinetic and potential energy of the
linear(ized) system with constant external forces F ext as18 , respectively,
1 1
T = ẋ · M ẋ, V = x · Kx − F ext · x (4.137)
2 2
The structure of C depends on the nature of the non-conservative forces involved. As a simple
example with damping, consider velocity-proportional damping as in the case of dashpots
between particles or linear viscous drag in a fluid. In this case, we may introduce a dissipation
potential, referred to as Rayleigh’s dissipation function D, such that

nc ∂D 1 d ∂L ∂L ∂D
Qi = − with D = ẋ · C ẋ ⇒ − + = 0. (4.138)
∂ ẋi 2 dt ∂ ẋi ∂xi ∂ ẋi
In many structural applications, structural damping is assumed as being velocity-proportional with

the approximation, known as Rayleigh damping,
C = αM + βK with α, β ≥ 0. (4.139)
This is a first-order approximation; constants α and β can be tuned to damp low- and high-
frequency vibrations, respectively. For example, if particles of equal mass m are submerged in
a fluid to experience linear viscous drag, then we consider drag forces Qnc i = Fdrag,i = −dẋi
d
with some damping coefficient d > 0, so that C = dI = m M = αM , since for identical particles
M = mI. Alternatively, if elastically coupled particles are interconnected by dashpots (one dashpot
in parallel to each spring connection, and the dashpot viscosity between particles i and j is denoted
by dij = βkij ), then we have C = βK. (4.139) presents a linera combination of these two cases.
18
These are the exact kinetic and potential energy for linear systems, while they are a approximations about an
equilibrium for nonlinear systems, following the Taylor expansions presented here.
175
Example 4.7. Elastic chain of particles
Consider a 1D chain of n particles of masses mi (i = 1, . . . , n) which are coupled through elastic

springs of stiffnesses ki,j (i, j = 1, . . . , n). In this case the equation of motion for each particle i
(using the scalar coordinates xi measured from equilibrium) can be obtained from a free-body
diagram (valid for all particles except i = 1 and i = n):
mi ẍi = −ki,i−1 (xi − xi−1 ) + ki,i+1 (xi+1 − xi ) + Fiext (t). (4.140)
xi-1 xi xi+1
mi-1 ki,i-1 mi ki+1,imi+1
Fi-1 Fi Fi+1
For the first and last particle (with one missing spring each) we have the shortened equations
m1 ẍ1 = k1,2 (x2 − x1 ) + F1ext (t), mn ẍn = −kn,n−1 (xn − xn−1 ) + Fnext (t). (4.141)
Alternatively and more conveniently for more complex systems, the equations of motion can also
be obtained from the Lagrange equations (4.13), using
n n
X mi X ki,i−1 d ∂L ∂L
T = ẋ2i , V = (xi − xi−1 ) 2
⇒ − = Fiext , (4.142)
2 2 dt ∂ ẋi ∂xi
i=1 i=2
which also yields the n equations given by (4.140) and (4.141).
Either way, the obtained equations of motion can be arranged into matrix form, viz.
 
 ẍ1   x1  F1ext
m1  ẍ2  k1,2 −k1,2
 x2 F2ext
. −k −k2,3 .   .
m2 k1,2 + k2,3

  1,2
 .
  . .
... . ... .   .

 mi

 ẍi + −ki−1,i ki−1,i + ki,i+1 −ki,i+1

 xi = Fiext
.
...  .
 ...  .
 
 .


mn . −kn,n−1 kn,n−1 . .
. . .
ẍn xn ext
Fn
This matrix equation is indeed of the form (4.135). Since the system is free of damping and non-
conservative forces, we observe that C = 0 in the above system of equations. The mass matrix and
external force vector are, respectively,
 ext 
F1
m1
 
F ext 
m2  2 
 .. 
 
 
 ...  ext
 . 
[M ] = 
  and [F ] =  F ext  ,
 (4.143)
mi 
 i 
 .. 
 
 ... 
 . 
mn
Fnext
176
while the stiffness matrix reads

k1,2 −k1,2
 
−k1,2 k1,2 + k2,3 −k2,3 
 
 ... 
[K] =  . (4.144)

 −ki−1,i ki−1,i + ki,i+1 −ki,i+1 

 ... 
−kn,n−1 kn,n−1
————
Example 4.8. Motion of a double-pendulum
Consider a double-pendulum consisting of two rods of masses m1 , m2 and
lengths l1 , l2 , as shown. When describing the system by the two angles α
and β as the generalized DOFs so q = {α, β}, the total kinetic energy is A
a m1,l1
1 1 2 1 2 2 1 2
T = IA α̇ + IC β̇ + m2 vC (4.145)
2 2 2 B
with g b +C
m2,l2
1 m1 l12 m2 l22
IA = , IC2 = . (4.146)
3 12
Note that the last term in the kinetic energy is required since the center of mass of body 2 is
moving. The required velocity of the CM of body 2 comes from the velocity transfer formula:
vC = vB + ω2 × rBC = ω1 × rAB + ω2 × rBC = (α̇e3 ) × rAB + (β̇e3 ) × rBC

(4.147)

l2 l2
= α̇l1 cos α + β̇ cos β e1 + α̇l1 sin α + β̇ sin β e2 ,
2 2
so that altogether with vC = |vC |
2
" 2 2 #
m1 l12 2 m2 l22 2 1

l2 l2
T = α̇ + β̇ + m2 α̇l1 cos α + β̇ cos β + α̇l1 sin α + β̇ sin β . (4.148)
6 24 2 2 2
The potential energy (dependent on the positions of the two centers of mass) becomes

cos α cos β m
1
cos β
V = −m1 gl1 − m2 g l1 cos α + l2 = −gl1 cos α + m2 − gm2 l2 . (4.149)
2 2 2 2
The Lagrange equations for this conservative system provide the (generally nonlinear) equations of
motion with L = T − V as
∂(T − V )

d ∂L ∂L d ∂T
− =0 ⇔ = (4.150)
dt ∂ α̇ ∂α dt ∂ α̇ ∂α
and
∂(T − V )

d ∂L ∂L d ∂T
− =0 ⇔ = . (4.151)
dt ∂ β̇ ∂β dt ∂ β̇ ∂β
177
The exact equations are left out here for conciseness but can be obtained in closed form by dif-
ferentatiating the above expressions of T and V . From the potential energy landscape one may
easily find that there is one stable equilibrium at α = β = 0 as well as unstable equilibria at
(α, β) = (0, ±π), at (α, β) = (±π, 0), and at (α, β) = (±π, ±π).
We may also obtain the linearized system matrices from the above potentials, viz.
∂2T ∂2T (m1 +3m2 )l12
! !
∂2T ,
∂ α̇2 ∂ α̇∂ β̇
l1 l2 m2
cos(α − β)
M= ⇒ [M ] = ∂ 2 T ∂ 2 T = l l m 3 2
l22 m2 (4.152)
∂ q̇∂ q̇ , 1 2 2
cos(α − β)
∂ α̇∂ β̇ ∂ β̇ 2 2 3
for the mass matrix, which for the stable equilibrium at α = β = 0 simplifies to
1 2(m1 + 3m2 )l12 3l1 l2 m2

[M ] = . (4.153)
6 3l1 l2 m2 2l22 m2
Similarly, we obtain the stiffness matrix as
∂2V ∂2V
!
∂2V ,

∂α2 ∂α∂β 1 l1 (m1 + 2m2 )g cos α 0
K= ⇒ [K] = ∂2V ∂2V = (4.154)
∂q∂q ∂α∂β , ∂β 2
2 0 l2 m2 g cos β
and for the stable equilibrium at α = β = 0

1 l1 (m1 + 2m2 )g 0
[K] = . (4.155)
2 0 l2 m2 g
The same matrices could also have been obtained by first deriving the equations of motion via
(4.150) and (4.151) and then linearizing about the equilibrium configuration. Either way we arrive
at the above matrices and we see that C = 0. Finally, note that both M and K are by definition
symmetric in this case and (when evaluated at equilibrium) positive-definite.
4.4.2 Free, undamped vibrations
When investigating vibrations of multi-DOF systems, let us first consider the case of free vibrations
(i.e., F ext (t) = 0) without damping (i.e., C = 0). In this case the governing equations become
M ẍ + Kx = 0. (4.156)
Analogous to the solution for single-DOF vibrations, we use the ansatz
x(t) = Re [x̂∗ exp(λt)] where x̂∗ ∈ Cn , λ ∈ C. (4.157)
Insertion of x∗ (t) = x̂∗ exp(λt) into (4.156) leads to
λ2 M x∗ (t) + Kx∗ (t) = 0 ∀t ⇒ (K + λ2 M )x̂∗ = 0. (4.158)
This is an eigenvalue problem, which may alternatively be written (M −1 K + λ2 I)x̂∗ = 0.

Hence, for non-trivial solutions (−λ2 ) must be an eigenvalue of the matrix M −1 K. λ is obtained
as a root of the characteristic equation
det K + λ2 M = 0

(4.159)
178
Since x is n-dimensional, we have n (not necessarily distinct) roots λ2i .
Recall that, by construction, M is positive-definite (and therefore guaranteed to be invertible),

while K is positive-semi-definite for those cases considered here. From linear algebra we know that
this ensures only real-valued eigenvalues (−λ2j ) ∈ R and, moreover, non-negative eigenvalues19 , i.e.,
q q
−λ2j ≥0 ⇒ λj = −(−λ2j ) = iωj with ωj = ± −λ2j ∈ R. (4.160)
Each of the ωj ∈ R (j = 1, . . . , n) is an eigenfrequency of the system, whereas the corresponding

eigenvector x̂∗j is known as its mode shape or eigenmode of the system. Note that, since
M −1 K is positive-semi-definite here, the eigenvalue problem also ensures that we have real-valued
eigenvectors x̂∗j = x̂j ∈ Rn (so we drop the asterisk).
Further, if ωj is a solution, then −ωj must also be a solution, so we have pairs of eigenfrequencies ±ωj
and consequently 2n solutions for ωj . We now need to consider two cases for each pair of eigen-
frequencies: either we have a pair of distinct non-zero eigenfrequencies ±ωj , or the eigenfrequency
vanishes, i.e., we have a duplicated eigenfrequency ωj = 0.
• pair of non-zero eigenfrequencies: in this case, the solution is analogous to the single-
DOF free, undamped vibrations discussed in Section 4.3.2. It can be written for each eigen-
frequency ωj (defining ωj as the positive root of the pair) as
xj (t) = Re [aj x̂j exp(iωj t) + bj x̂j exp(−iωj t)] with aj , bj ∈ C

(4.161)
= cj x̂j cos(ωj t + ϕj ) with cj , ϕj ∈ R.
• zero eigenfrequency: if λj = 0, then the duplicated ωj = 0 yields the solution
xj (t) = aj x̂j exp(0) + bj x̂j t exp(0) = (aj + bj t)x̂j with aj , bj ∈ R. (4.162)
This solution implies not vibrational motion but rigid-body motion.

To better understand this, let us insert λ = 0 into the equation of motion (4.158), leading to
Kxj = 0 so that 12 xj · Kxj = V (xj ) = 0. This implies that displacing the particles according
to xj (t) produces zero potential energy. Therefore, xj (t) represents rigid body motion, i.e.,
the particle system moves like a rigid body, leaving all particle distances unchanged.
We conclude that, since every rigid-body mode must result in a pair of zero eigenfrequency, every
system has as many pairs of zero eigenfrequencies as it has independent rigid-body
modes. For example, two elastically-coupled particles moving in 1D (as in Example 4.10 below)
have one rigid-body mode (viz., translation in 1D). If the two particles are moving in 2D, then the
system has three rigid-body modes (viz., translation in both directions and a rotation) and hence
three zero eigenfrequencies, etc.
Finally, we identify the complete solution by summing over all eigenfrequencies (since each of
them contributes a solution and the ODE is linear, the complete solution is a superposition of the
19
√
Here and in the following we use index j to avoid confusion with the imaginary i = −1.
179
eigenmodes). Assuming that there are n−l ≤ n rigid-body modes and that the eigenfrequencies are
sorted so that the first l eigenfrequencies are non-zero (and all eigenfrequencies are non-negative),
the general solution is
n
X l
X n
X
x(t) = xj (t) = cj cos(ωj t − ϕj )x̂j + (aj + bj t)x̂j (4.163)
j=1 j=1 j=l+1
The 2n unknown coefficients aj , bj , cj , ϕj ∈ R are to be determined from initial conditions, while

the eigenfrequencies and eigenmodes follow from the eigenvalue problem
det K − ωj2 M = 0 K − ωj2 M x̂j = 0

and (4.164)
We point out that the eigenmodes have a special orthogonal structure. Recall that the governing
ODE is linear, so that each term j in the above solution must satisfy the equation
−ωj2 M x̂j + K x̂j = 0 and hence also − ωk2 M x̂k + K x̂k = 0 (4.165)
for some 1 ≤ j, k ≤ n (for the rigid-body modes, the equations of motion are trivially satisfied,
as discussed above20 ). Pre-multiplying the first equation by x̂k and subtracting the second pre-
multiplied by x̂j yields (exploiting that M and K are symmetric here)
(ωk2 − ωj2 )x̂j · M x̂k = 0 ⇒ x̂j · M x̂k = 0 if ωj 6= ωk . (4.166)
Likewise, pre-multiplying the first equation by x̂k and adding the second pre-multiplied by x̂j (and
inserting (4.166)) leads to
−(ωj2 + ωk2 )x̂j · M x̂k + 2x̂j · K x̂k = 0 ⇒ x̂j · K x̂k = 0 if j 6= k. (4.167)
Example 4.9. Eigenfrequencies of a double-pendulum
Having established the linearized equations of motion and the associated mass and stiffness matrices
for a double-pendulum of two rods in Example 4.8, we may now compute the eigenfrequencies of
that system. The system is undamped, so we expect two imaginary eigenvalues. Further, since no
rigid body motion can occur (the upper beam is hinged, so the system cannot move through space
without changing the potential energy), we expect distinct non-zero eigenvalues.
We use the mass and stiffness matrices derived in Example 4.8,

1 2(m1 + 3m2 )l12 3l1 l2 m2

1 l1 (m1 + 2m2 )g 0
[M ] = , [K] = . (4.168)
6 3l1 l2 m2 2l22 m2 2 0 l2 m2 g
For the special case of equal rods, i.e., m1 = m2 = m and l1 = l2 = l, the two matrices reduce to
ml2 8 3

mgl 3 0
[M ] = , [K] = . (4.169)
6 3 2 2 0 1
20
The discussion here assumes distinct eigenvalues, which is not always the case. However, any set of eigenmodes
can be turned into an orthonormal basis, so the general procedure discussed here can still be applied after possible
orthogonalization.
180
Let us calculate the eigenvalues for this case, which leads to the characteristic equation
2 2 2 2

2
3mgl 2 8ml mgl 2 2ml 2 3ml
0 = det K − ω M = −ω −ω − −ω
2 6 2 6 6
2 4
2
(4.170)
m l g g
= 7ω 4 − 42 ω 2 + 27 2 ,
36 l l
√
whose roots are (noting that 6/ 7 < 3)
s r
2 6 g 6 g
ω = 3± √ ⇒ ω = 3± √ . (4.171)
7 l 7 l
We thus obtain two pairs of distinct eigenfrequencies:
s r s r
6 g p 6 g p
ω1 = ± 3 − √ ≈ ±0.856 g/l, ω2 = ± 3 + √ ≈ ±2.295 g/l. (4.172)
7 l 7 l
Dropping the negative eigenfrequencies for uniqueness, we conclude
p p
ω1 ≈ 0.856 g/l, ω2 ≈ 2.295 g/l. (4.173)
We can also compute the corresponding eigenmodes as the associated eigenvectors x̂j satisfying
(K − ωj2 M )x̂j = 0. Here, this leads to
1 √ 1 √
(1 + 2 7) (1 − 2 7)
[x̂1 ] = 9 , [x̂2 ] = 9 . (4.174)
1 1
For comparison, the eigenfrequency of a simple (single-rod) pendulum is obtained from its linearized
equations of motion as
r
ml2 l 3g 3g p
ϕ̈ + mg sin ϕ = 0 ⇒ ϕ̈ + ϕ ≈ 0 ⇒ ω0 = ≈ 1.225 g/l. (4.175)
3 2 2l 2l
In both cases, the eigenfrequencies are independent of the masses and only depend on the pendulum
length(s) and the gravitational acceleration.
————
Example 4.10. Elastic two-particle oscillator

Consider two particles of masses m1 and m2 , which are connected
x1 x2
by an elastic spring of stiffness k and slide frictionlessly on a planar
m1 k m2
ground in 1D. Let us find the eigenfrequencies and corresponding
eigenmodes of the system.
We denote the 1D positions of the two masses by x1 and x2 , and we write x = {x1 , x2 }. The
equations of motion in this case are a special form of those derived for the 1D elastic chain (see
Example 4.7), viz.

m1 0 ẍ1 k −k x1 0
+ = . (4.176)
0 m2 ẍ2 −k k x2 0
181
Insertion of x(t) = x̂ exp(iωt) yields

m1 0 k −k x̂1 0
−ω 2 + = . (4.177)
0 m2 −k k x̂2 0
The characteristic equation det(K − ω 2 M ) = 0 hence becomes

2
k −k m1 0 −ω m1 + k −k
0 = det − ω2 = det
−k k 0 m2 −k −ω 2 m2 + k (4.178)
2 2 2 2 2
= (−ω m1 + k)(−ω m2 + k) − k = −ω (m1 k + m2 k − ω m1 m2 ).
The eigenfrequencies ωj are obtained from the roots of this equation:

r
2 2 m1 + m2 m1 + m2
ω =0 ∨ ω = k ⇒ ω1,2 = 0, ω3,4 = ± k. (4.179)
m1 m2 m1 m2
As could have been expected, we have one rigid-body mode (identified by the repeated zero eigen-
frequency ω1,2 = 0). This becomes apparent when calculating the corresponding eigenmodes. To
this end, we insert the solutions for λ into (4.177) to identify x̂. For ω = 0 we thus obtain

m1 0 k −k x̂1 k −k x̂1 0 x̂1 1
0 + = = ⇒ = x̂ , (4.180)
0 m2 −k k x̂2 −k k x̂2 0 x̂2 1
which is defined up to an arbitrary coefficient x̂ ∈ R. This is obviously rigid body motion since
(x̂1 , x̂2 )T = x̂(1, 1)T implies that both masses translate by the same amount x̂. The rigid-body
solution in this case of a repeated eigenvalue ω = 0, with constants a, b ∈ R, reads

x1 (t) 1 1
[x(t)] = = [A exp(0) + B t exp(0)] x̂ = (a + bt) . (4.181)
x2 (t) 1 1
Analogously, we find the eigenmode associated with the non-zero eigenfrequency via
m1
−k m2 −k

m1 + m2 m1 0 k −k x̂1 x̂1 0
− k + = = ,
m1 m2 0 m2 −k k x̂2 −k −k m
m1
2
x̂ 2 0
which yields the eigenmode (again defined up to a constant x̂ ∈ R)

x̂1 m2
= x̂ . (4.182)
x̂2 −m1
In this case, the two particles move in opposite directions (which is clearly not rigid-body motion
since this eigenmode changes the spring length and hence causes changes to the potential energy).
The two eigenmodes are shown schematically below (recall that both mode shapes are only known
up to a constant). The non-zero eigenfrequencies are associated with simultaneous counter-motion
of the two particles.
x^1 x^2 x^1 x^2

w1,2=0 m1 m2 w3,4 m1 m2
182
The complete solution (4.163) thus becomes

r
x1 (t) m2 m1 + m2 1
[x(t)] = = x̂ cos k t + ϕ + (a + bt) , (4.183)
x2 (t) −m1 m1 m2 1
whose constants x̂, ϕ, a, b ∈ R are to be obtained from initial conditions (e.g., the two known initial
positions xi (0) and the two initial velocities ẋi (t) provide four equations for the four unknown
coefficients).
We point out that for arbitrary choices of x̂, ϕ, a, b, we realize that the total linear momentum P
of the two-particle system, evaluated as
P = m1 ẋ1 + m2 ẋ2 = [m1 x̂ m2 + m2 x̂(−m1 )] (−ω3 ) sin(ω3 t + ϕ) + (m1 + m2 )b

(4.184)
= (m1 + m2 )b = const.,
is conserved during vibrations. Specifically, the vibrational mode constitutes a vibration about the
center of mass – the location of the center of mass of the two-particle system remains unchanged
by the vibration and, of course, linear momentum is also conserved for the case of rigid-body
translation with constant velocity. The observation of a constant linear momentum of the system
makes indeed sense since, for a free motion without external forcing nor damping, linear momentum
is conserved (see Section 2.2.1). This, of course, also includes vibrational motion.
————
4.4.3 Damped and forced vibrations
Systems of particles undergoing forced vibrations (F ext (t) 6= 0) in the presence of damping (C 6= 0)
are governed, in the linear(ized) case, by
M ẍ + C ẋ + Kx = F ext (t). (4.185)
Like for single-DOF vibrations in Section 4.3.3, the solution is sought of the form
x(t) = xhom (t) + xpart (t), (4.186)
whose homogeneous part xhom (t) is the solution of a free vibration and, in the presence of any
damping, decays with time.
The homogeneous solution of a free vibration is found by considering
M ẍ + C ẋ + Kx = 0. (4.187)
Analogous to the solution for undamped vibrations, we start with the ansatz
xhom (t) = Re [x̂∗hom exp(λt)] where x̂∗hom ∈ Cn , λ ∈ C. (4.188)
Insertion into (4.187) leads to
λ2 M x∗hom (t) + λCx∗hom (t) + Kx∗hom (t) = 0 K + λC + λ2 M x̂∗hom = 0. (4.189)

⇒
183
The characteristic equation in this case reads
det K + λC + λ2 M = 0,

(4.190)
whose roots are the 2n (not necessarily distinct) eigenvalues λj . By construction, the case of non-
zero damping leads to eigenvalues that are not necessarily non-negative, so that complex-valued
eigenfrequencies ωj ∈ C are to be expected, leading to solutions that decay over time (analogous
to the single-particle discussion).
For a forced vibration, the particular solution depends on the forcing. Consider, e.g.,
h i
F ext (t) = Re F̂ ∗ exp(iΩt) where F̂ ∗ ∈ Cn , (4.191)
with an excitation frequency Ω ∈ R. This implies that the solution must be of the form
xpart (t) = Re x∗part (t) = Re x̂∗part exp(iΩt) x̂∗part ∈ Cn .

where (4.192)
Insertion of x∗part (t) = x̂∗part exp(iΩt) into (4.185) gives the sought mode shape as
−1 ∗
−Ω M + iΩC + K x̂∗part = F̂ ∗ x̂∗part = −Ω2 M + iΩC + K
2
⇒ F̂ . (4.193)
As before, the real part is to be taken as the final solution. Unfortunately, analytical solutions
of both (4.189) and (4.193) are hard to find, which is why we introduce the concept of modal
decomposition below.
Example 4.11. Tuned mass damping
A tuned mass damper is an engineering trick to avoid harmful vibrations at or near a resonant
frequency, which is found especially in civil engineering applications – from bridges to skyscrap-
ers. Rather than discussing a complicated engineering structure, let us study the simplest system
possible, which illustrates the concept of a tuned mass damper.
Consider the shown system of a large mass M attached to a linear spring of x1
stiffness k and a damper of viscosity c (this is our engineering structure). Limit- c
ing ourselves to 1D motion andp neglecting gravity, this simple one-DOF system M
has an eigenfrequency of ω0 = k/M , and free vibrations in the underdamped
k
case occur at the frequency F ext
r r
k k c
q
ωd = ω02 − δ 2 = − δ2 ≈ if δ = 1 (4.194)
M M 2M
for small damping. Now, assume the system is excited by an external force F ext (t), applied to the
large mass M at the eigenfrequency:
r !
ext k
F (t) = F̂ cos t . (4.195)
M
As discussed, this resonant excitation may lead to excessive vibration amplitudes that can harm
the structure (e.g., leading to the failure of bridges or tall buildings, or at least to making people
on/in those feel uncomfortable).
184
As a remedy, let us attach a small mass m through a second spring x1

of stiffness ke and a dashpot of viscosity ce , as shown on the right c ke x2
– effectively turning the above single-DOF system into a two-DOF M m
system. The vibrations of this two-DOF system are found from the k ext ce
equations of motion: F
ext
M ẍ1 c + ce −ce ẋ1 k + ke −ke x1 F (t)
+ + = . (4.196)
m ẍ2 −ce ce ẋ2 −ke ke x2 0
| {z } | {z } | {z } | {z }
=M =C =K =F ext (t)
Consider a harmonic excitation, so that

∗
ext ∗ F̂ ∗ x̂1
F (t) = F̂ exp(iΩt) = exp(iΩt) ⇒ xpart (t) = x̂ exp(iΩt) = exp(iΩt),
0 x̂∗2
where for simplicity we work with complex-valued functions (tacitly assuming that taking real parts
is required in the end). Insertion into the equations of motion yields
(−Ω2 M + iΩC + K)x̂∗ = F̂ ∗ ⇒ x̂∗ = (−Ω2 M + iΩC + K)−1 F̂ ∗ . (4.197)
The (lengthy) expression for the amplitude vector x̂∗ can in fact be evaluated
p analytically. For
example, the amplitude x̂∗1 of the big mass M for the special case of Ω = k/M (the critical
excitation frequency) evaluates to
v
Fu c2 γ 2 + (k − ξ)2
u
x̂∗1 q = t h i (4.198)
Ω= Mk
k γ 2 c2 γ 2 + k 2 + ξ 2 (c + 1)2 γ 2 + k 2 − 2γ 2 kξ
with the abbreviations

m ke ce c
ξ= , k= , c= , γ=√ . (4.199)
M k c kM
The idea is now to choose the added system parameters (m, ke and ce ) such as to minize the above
amplitude at the excitation frequency. Minimization of the amplitude with respect to, e.g., the
attached mass m (which is usually simplest to adjust in practice) shows that
ke c2
x̂∗1 q → min if m→M + e. (4.200)
Ω= Mk
k ke
2
Hence, choosing an extra mass m = M kke + kcee effectively reduces the vibration amplitude of the
large mass M to a minimum. Therefore, one speaks of the added (and carefully tuned) mass m as
a tuned mass damper.
p
Of course, minimizing the amplitude at Ω = k/M says little about what happens at other
frequencies,
p yet it helps avoid an excessive vibrational amplitude at the target frequency of Ω =
k/M . The plot below illustrates
p the amplitudes of the two masses M and m vs. the excitation
frequency Ω (normalized by k/M ).
185
|x|/(F/k)
4
x1 (t)
x2 (t)
3
Ω/ k M
0.0 0.5 1.0 1.5 2.0
As
p the plot indicates, the amplitude of x1 (t) becomes marginal at the target frequency of Ω =
k/M , while the amplitude of x2 (t) remains finite. Overall, we observe in the amplitude plot
two resonances of this two-DOF system, yet none of them is close to the target frequency. This
demonstrates the strategy of a tuned mass damper, i.e., an extra masses added to the system and
tuned such that the system vibration amplitudes of interest are reduced at a given target frequency.
————
4.4.4 Modal Decomposition
Acknowledging that the above treatment of damped and forced vibrations does not generally lead
to closed-form solutions (except for quite simple problems), we introduce the concept of modal
decomposition, which will prove helpful in this context. The key idea is to exploit that the n
eigenmodes x̂j can be turned into a basis of the n-dimensional space in which the sought solution
lies. This allows us to decompose the solution of any vibration as follows:
n
X
x(t) = x̂j yj (t) = Xy(t) (4.201)
j=1
where we introduced for convenience the constant matrix X and vector y(t) as
X = [x̂1 , . . . , x̂n ] ∈ Rn×n , y(t) = [y1 (t), . . . , yn (t)]T ∈ Rn . (4.202)
Here, x̂j are the (possibly normalized) eigenmodes of the free vibration, and y(t) ∈ R are some
unknown coefficients that are functions of time and to be determined. The strategy is thus to find
X a-priori from the free-vibration eigenvalue problem, and then to identify the unknown functions
yj (t).
Insertion of x(t) = Xy(t) into the equations of motion (4.185) and pre-multiplication by X T yields
X T M X ÿ + X T CX ẏ + X T KXy = X T F ext . (4.203)
186
Recall that in (4.166) and (4.167) we showed
x̂j · M x̂k = 0 and x̂j · K x̂k = 0 for j 6= k. (4.204)
We can further normalize the eigenvectors (knowing that x̂j · M x̂j > 0) according to
(
x̂j 1 if j = k,
x̂j ← p ⇒ x̂j · M x̂k = δjk = (4.205)
x̂j · M x̂j 0 else.
Using −ωk2 M x̂k + K x̂k = 0, the above normalization also leads to

(
ωj2 if j = k,
x̂j · K x̂k = ωk2 x̂j · M x̂k ⇒ x̂j · K x̂k = (4.206)
0 else.
With this normalization, the components of the terms X T M X and X T KX simplify to

(
ωj2 if j = k,
[X T M X]jk = x̂j · M x̂k = δjk and [X T KX]jk = x̂j · K x̂k = (4.207)
0 else.
This shows that both matrices X T M X and X T KX are in fact diagonal:

   2 
1 ω1
T
 1 
T
 ω22 
[X M X] =  and [X KX] =  . (4.208)
   
..  ..
 .   . 
1 2
ωn
If we further deal with structural damping, i.e., C = αM + βK with constants α, β ≥ 0, then

(4.203) decouples into n equations
ÿj + (α + βωj2 )ẏj + ωj2 yj = F ext · x̂j for j = 1, . . . , n (4.209)
which can be solved independently for each yj (t) (j = 1, . . . , n) with suitable initial conditions.
This is quite remarkable (and the key point of the method of modal decomposition): we started
with a system of n coupled differential equations and decoupled it into n independent ODEs which
can be solved one at a time. The only necessary step is the calculation of all eigenfrequencies and
associated eigenmodes (and possibly their normalization).
It is essential to observe that (4.209) bears the same structure as the equation of motion discussed for
single-DOF vibrations in Section 4.3. Therefore, all conclusions drawn there apply here as well for
each mode j individually – including the decomposition into homogeneous and particular solutions
for forced vibrations, the four cases of undamped, under-, over- and critically damped homogeneous
vibrations, and the case of forced vibrations as well as resonance. For F ext = F̂ sin(Ωt), each mode
is excited at the frequency Ω.
187
Example 4.12. Forced vibration of a damped mass-spring system
Consider two particles of equal masses m connected by an elastic spring of stiffness k. The first
particle is attached to the ground by another spring of stiffness k. The second particle is excited
harmonically by an external force F (t) of frequency Ω. The particles are subjected to structural
damping. What is the vibrational response of the system?
x1 x2
k k
m m F(t)
The system matrices M and K are obtained from the equations of motion, either via free-body
diagrams and linear momentum balance or by using the Lagrange equations. Either way one arrives
at (using structural damping for C)

m 2k −k
[M ] = , [K] = , C = αM + βK. (4.210)
m −k k
Two distinct eigenfrequencies are obtained from the characteristic equation as
√ √
r r r
1 1 k
ω1 = (3 − 5) ω0 , ω2 = (3 + 5) ω0 with ω0 = . (4.211)
2 2 m
Note that no rigid-body mode exists since the first particle is elastically attached to the ground, so
uniform translation of the two particles causes an increase in the potential energy and is therefore
not rigid-body motion. The eigenmodes corresponding to the above two eigenfrequencies are
1 √ 1 √
1 − 2 (1 + 5) 1 − 2 (1 − 5)
[x̂1 ] = q √ , [ x̂2 ] = √ . (4.212)
1 1
q
m m
2 (5 + 5) 2 (5 − 5)
Note that these have been normalized, as described above, such that x̂i · M x̂j = δij . Writing
X = (x̂1 , x̂2 ) yields
2
T 1 T ω1
[X M X] = , [X M X] = , (4.213)
1 ω22
while the structural damping assumption leads to
α + βω12

[X T CX] = . (4.214)
α + βω22
Thus, the system of equations of motion decouples into two ODEs by applying the modal decom-
position discussed above. Specifically, we write
x(t) = y1 (t)x̂1 + y2 (t)x̂2 , (4.215)
188
which decouples the equations of motion into
F (t)
ÿ1 + (α + βω12 )ẏ1 + ω12 y1 = F · x̂1 = q √ ,
m
2 (5 + 5)
(4.216)
F (t)
ÿ2 + (α + βω22 )ẏ2 + ω22 y2 = F · x̂2 = q √ ,
m
2 (5 − 5)
where we used that F (t) = [0, F (t)]T . For the case F (t) = F̂ ∗ exp(iΩt), we make the ansatz
y1 (t) = ŷ1∗ exp(iΩt), y2 (t) = ŷ2∗ exp(iΩt) with ŷ1∗ , ŷ2∗ ∈ C, (4.217)
whose insertion into (4.216) gives
F̂ ∗
ŷ1∗ = q √ ,
m
2 (5 + 5) −Ω2 + (α + βω12 )iΩ + ω12
(4.218)
F̂ ∗
ŷ2∗ =q √ .
m
2 (5 − 5) −Ω2 + (α + βω22 )iΩ + ω22
Therefore, we arrive at the complex-valued solution

√ √
ŷ1∗ exp(iΩt) − 12 (1 + 5) ŷ2∗ exp(iΩt) − 12 (1 − 5)

∗
[x (t)] = [x̂1 ]y1 (t) + [x̂2 ]y2 (t) = q √ +q √
m 1 m 1
2 (5 + 5) 2 (5 − 5)
" 1 √
2 − 2 (1 + 5)
= √
(5 + 5) −Ω2 + (α + βω12 )iΩ + ω12 1

1 √ # ∗
2 − 2 (1 − 5) F̂ exp(iΩt)
+ √ .
2 2 1 m

2
(5 − 5) −Ω + (α + βω2 )iΩ + ω2
(4.219)
If the forcing is F (t) = F̂ cos(Ωt), then the forced vibration of the two particles is given by
x(t) = Re [x∗ (t)] with F̂ ∗ = F̂ , (4.220)
whose tedious analytical evaluation is omitted here for brevity. Instead, we present in the figure
below numerical results (for α = 0.1 and β = 0). Specifically, we plot the amplitudes x̂1 and
x̂2 as functions of the excitation frequency Ω. Clearly visible are the two resonance peaks near
ω1 ≈ 0.618ω0 and ω2 ≈ 1.618ω0 (and also the associated anti-resonances), at which the amplitudes
climax. Without damping, those peaks would tend to infinity; with damping, they remain finite.
Also apparent is the switch in the motion of the particles from in-phase (amplitudes of same sign) to
out-of-phase (amplitudes of opposite sign) below and above the first resonance, respectively. Here,
and in general, each resonant frequency is noticeable in the motion of each DOF in a multi-DOF
forced vibration.
189
^
xi/(F/k)
^
x2
^
x1
W/w0
190
Lagrange equations:
N
d ∂L ∂L X ∂rj
− = Qnc
i with L = T − V, Qnc
i = Fjnon-cons. ·
dt ∂ q̇i ∂qi ∂qi
j=1
equilibria of a conservative, static system with a single DOF q:
∂2V
stable equilibrium ⇔ energy minimum ⇔ >0
∂q 2
∂2V
unstable equilibrium ⇔ energy maximum ⇔ <0
∂q 2
general form of the equation of motion for a single DOF:

r
2 k c δ 2π
ẍ + 2δ ẋ + ω0 x = f (t) with ω0 = , δ= , D= , T =
m 2m ω0 ω0
general solution for undamped vibrations (D = 0):
x(t) = A1 cos(ω0 t) + A2 sin(ω0 t) = A sin(ω0 t + ϕ0 )
general solution for overdamped vibrations (D > 1):

√ √
− δ+ δ 2 −ω02 t − δ− δ 2 −ω02 t
x(t) = A1 e + A2 e
general solution for critically damped vibrations (D = 1):
x(t) = A1 e−δt + A2 te−δt
general solution for underdamped vibrations (0 < D < 1):

q
x(t) = e−δt [A1 cos(ωd t) + A2 sin(ωd t)] , ωd = ω02 − δ 2
general solution for forced vibrations:

fˆ
f (t) = fˆ cos(Ωt) ⇒ x(t) = xhom (t) + V cos (Ωt − ϕ)
ω02
with magnification and phase

1 2Dη Ω
V =p , ϕ = arctan where η=
(1 − η 2 )2 + 4D2 η 2 1 − η2 ω0
191
equations of motion for multiple-DOF vibrations:
M ẍ + C ẋ + Kx = F (t)
kinetic and potential energy for multiple-DOF linear(ized) vibrations:

1 1
T = ẋ · M ẋ, V = x · Kx
2 2
linearized system matrices around a stable equilibrium q0 :
∂2T ∂Qnc ∂2V ∂Qnc

M= (q0 , 0), C=− (q0 , 0), K= (q0 ) − (q0 , 0)
∂ q̇∂ q̇ ∂ q̇ ∂q∂q ∂q
structural damping:
C = αM + βK, α, β ≥ 0
eigenfrequencies ωj and eigenmodes x̂j are obtained from
det −ωj2 M + K = 0 −ωj2 M + K x̂j = 0

and
general solution of free vibrations for multiple-DOF systems:

n
X l
X n
X
x(t) = xj (t) = cj cos(ωj t + ϕj )x̂j + (aj + bj t)x̂j
j=1 j=1 j=l+1
192
5 Dynamics of Deformable Bodies
So far, we have limited our consideration to particles and rigid bodies. In reality, however, systems
are oftentimes not describable as systems of small particles nor as rigid bodies. In Mechanics
2 we saw that applied forces generally result in internal forces which, in turn, produce stresses
that lead to deformation if the body is not rigid. Unfortunately, although in Mechanics 2 the
foundation was laid to describe and predict the behavior of deformable bodies, that discussion was
limited to (quasi)statics. Therefore, we will here reformulate what we learned in Mechanics 2 in the
dynamic context, so that, e.g., accelerations (which we have seen to be interpretable as additional,
inertial forces) may also lead to deformation. As a simple example, consider a quickly rotating
turbine whose centrifugal forces may lead to internal tensile stresses and thus to an expansion of
the turbine blades. Or consider a heavy mass attached at the tip of a slender beam, so that an
excitation of the system may lead to vibrations that engage the beam in bending. Here, we will
discuss the various ways in which dynamics affects the mechanics of deformable bodies, and we
will restrict our study to linear elastic bodies (i.e., bodies whose stress–strain relation is linear).
5.1 Dynamics of Systems with Massless Deformable Bodies
Let us first consider the simple case of deformable bodies that are approximately massless (because
their mass is significantly smaller than that of other system components). We have seen previously
that massless bodies still obey static equilibrium. The deformation of such massless deformable
bodies can therefore be described as in Mechanics 2 and, especially for the analysis of vibrations,
can conveniently be reduced to simple models involving masses and springs. Let us review a few
examples:
EA EI F GJ Dq
F M
w
l Dl l l
• Bar: As a first example, consider a slender bar of length l, Young’s modulus E and cross-
sectional area A, which is deforming under the action of an axial force, as shown above on
the left. The relation between the change of length ∆l of the 1D bar and the applied force F
was found in Mechanics 2 as
σ Fl ∆l
∆l = εl = l= ⇔ F = EA . (5.1)
E EA l
If we reinterpret the bar as an elastic spring, then its effective spring stiffness keff (resulting
in the same relation between force and elongation) is
F EA
keff = = (5.2)
∆l l
193
• Beam: Next, consider a slender cantilever Euler-Bernoulli beam of length l, Young’s modulus
E and area moment I, which is loaded by a transverse force F applied at its tip. The relation
between the tip deflection w and the applied force F was found in Mechanics 2:
F l3 F 3EI
w= ⇒ keff = = 3 (5.3)
3EI w l
• Torsional rod: Finally, consider a cantilever rod of length l, shear modulus G and polar
moment J, which is loaded by an applied torsional torque M . The relation between the
relative twist angle ∆θ between the two end points and the applied torque M was derived in
Mechanics 2 as
Ml M GJ
∆θ = ⇒ keff = = (5.4)
GJ ∆θ l
If the above deformable bodies are assumed massless (compared to those masses attached to them),
then they can be replaced by elastic springs with the effective stiffness values derived above (since
for massless bodies the static equilibrium relations still apply). The same general procedure can be
applied to other deformable bodies and loading scenarios. As long as the relation between applied
force or torque and resulting displacement or twist is known and linear, it can be replaced by an
effective linear elastic spring.
Example 5.1. Flexural vibration of a particle on a beam

Consider an approximately massless beam (length l, Young’s EI
modulus E, area moment I) that is clamped at one end, while a m
heavy particle of mass m is attached to its free end, as shown. x
What is the flexural eigenfrequency of the system? l
Using the effective stiffness (5.3) for beam bending, we may write the equation of motion of the
particle as
3EI
mẍ = −F with F = keff x and keff = ⇒ mẍ + keff x = 0, (5.5)
l3
where we replaced the beam deforming in bending by a linear spring of effective m x

stiffness keff , as derived above. The free vibration of the system is thus governed
by the 1D equation of motion, which yields the eigenfrequency as
r keff
keff 2 keff 3EI
ẍ + x=0 ⇒ ω0 = ⇒ ω0 = . (5.6)
m m ml3
For example, for a circular cross-section of radius r, we have I = π4 r4 , so

r r
3EI 3Eπr4
ω0 = = . (5.7)
ml3 4ml3
194
Finally, recall that this solution based on the effective beam stiffness is only valid if the beam is
approximately massless.
————
Example 5.2. Vibration of a shaft with two gears

Consider an approximately massless shaft of length l and torsional
rigidity GJ (shear modulus G, polar area moment J), having gears
attached to its free ends (approximated as a disk of radius r1 and
mass m1 and a second disk of radius r2 and mass m2 ). GJ r2
r 1
j1
What are the eigenfrequencies of the system?
j2
Let us first consider torsional deformation of the gear system and
denote by ϕi the twist angle of disk i. We formulate the angular m1 l
momentum balance equations for both disks as m2
I1 ϕ̈1 = keff (ϕ2 − ϕ1 ), I2 ϕ̈2 = −keff (ϕ2 − ϕ1 ), (5.8)
with the two moments of inertia I1 = mr12 /2 and I2 = mr22 /2 with respect to the disks’ centers,
and the effective stiffness keff = GJ/l from (5.4). The above equations can be written in matrix
form as
j1 j2

I1 0

ϕ̈1

1 −1

ϕ1 0 keff
0 I2 ϕ̈2
+ keff
−1 1 ϕ2
=
0
. (5.9) I1 I2
We realize that this system of equations is equivalent to the equations of motion of the two-
particle oscillator in Example 4.10 (as schematically shown above). As a consequence, we can
re-use the eigenfrequencies derived there as Eq. (4.179), whose adaption here yields the torsional
eigenfrequencies
r s
I1 + I 2 m1 r12 + m2 r22 GJ
ω1tors. = 0, ω2tors. = keff = 2 . (5.10)
I1 I2 m1 r12 m2 r22 l
The zero eigenfrequency corresponds to rigid-body motion. The rigid-body mode here is a pure
rotation of the shaft without twisting, i.e., ϕ1 = ϕ2 (which is an eigenvector of the stiffness matrix).
Next, let us assume that the shaft can also undergo axial deformation (i.e., axial extension and
compression). In this case, the gear system can be reduced to two masses connected through a
linear spring with keff = EA/l, and the longitudinal eigenfrequencies are obtained analogously as
r r
m1 + m2 m1 + m2 EA
ω1long.
= 0, ω2long.
= keff = . (5.11)
m1 m2 m1 m2 l
Here, the rigid-body mode is pure axial translation of the system.
Finally, without the two bearings, the shaft could also undergo bending deformation, so the
flexural eigenfrequencies are
r r
flex. flex. m1 + m2 m1 + m2 3EI
ω1 = 0, ω2 = keff = , (5.12)
m1 m2 m1 m2 l3
195
where the rigid-body mode is translation perpendicular to the shaft’s axis (which is impossible,
however, in the presence of the two bearings).
We close by considering the example of a circular cross-section of the shaft, so that A = πr2 ,
I = π4 r4 and J = 2I = π2 r4 (and r l, r1 , r2 for a slender shaft). Also, for an isotropic material
G = E/2(1 + ν) with Young’s modulus E and Poisson’s ratio ν. In this case
v
u m r12 + m Eπr2
s u
m + m Eπr 2 3r 2 r2
flex. 1 2 tors. t 1 r2 2
ω2 = < ω 2 = 2
m1 m2 l 4l2 m1 m2 l 4(1 + ν)r12
s (5.13)
m1 + m2 Eπr 2
< ω2long. = ,
m1 m2 l
so the lowest non-zero eigenfrequency (the fundamental frequency) is the flexural mode – which,
however, can only engage without the two bearings. Otherwise, the torsional and longitudinal
eigenfrequencies are of similar order with the torsional expected lower since r < ri .
————
Example 5.3. Torsional vibration of a gear system

Let us consider a more complex system
consisting of two approximately mass less
shafts (each of length l and shear rigidity
GJ), which are connected through gears 2
and 3 that rotate without slipping. Further,
the two gears 1 and 4 are attached to the
two free ends of the shafts, as shown on the
right. All gears are approximated as disks
of radii ri and masses mi as shown.
What are the torsional eigenfrequencies of
the system for I1 = I4 and r2 = r3 ?
Since we are concerned with rotations of the
shafts, let us formulate the equations of mo-
tion in this case as the equations of angular
momentum balance.
We denote by ϕ1 , ϕ2 , ϕ3 and ϕ4 the (twist) rotation of each of the four gears (all with the same
sense). Angular momentum balance for each gear with Ii = 21 mi ri2 and keff = GJ/l reads
I1 ϕ̈1 = keff (ϕ2 − ϕ1 ), I3 ϕ̈3 = keff (ϕ4 − ϕ3 ) + FT r3 ,

(5.14)
I2 ϕ̈2 = −keff (ϕ2 − ϕ1 ) + FT r2 , I4 ϕ̈4 = −keff (ϕ4 − ϕ3 ),
where FT represents the unknown force transmitted between gears 2 and 3 (note that it appears
positive in both AMB equations, since ϕ2 and ϕ3 are both defined with the same sense); see the
schematic drawing of the two gears in question below.
196
From the gear kinematics we know that r2 ϕ2 = −r3 ϕ3 or ϕ3 = − rr32 ϕ2 ,

FT
so that the equations for ϕ2 and ϕ3 can be combined. Let us intro- r3 r2
duce generalized coordinates q1 = ϕ1 , q2 = ϕ2 and q3 = − rr32 ϕ4 such
that q1 = q2 = q3 corresponds to rigid-body motion. j3
j2
FT
This turns the above four equations into
I1 q̈1 = keff (q2 − q1 ),

I2 q̈2 = −keff (q2 − q1 ) + FT r2 ,
(5.15)
−I3 rr23 q̈2 = keff (− rr23 q3 + r2
r3 q2 ) + FT r3 ,
−I4 rr32 q̈3 = −keff (− rr23 q3 + rr23 q2 ).
Multiplying the third equation by r2 /r3 and subtracting the third from the second equation elimi-
nates the transmission force FT and results in the reduced equations of motion:
I1 q̈1 = keff (q2 − q1 ),

2 2 2
r2
I2 + I3 r3 q̈2 = −keff q2 − q1 − r3 q3 + rr32 q2 ,
r2
(5.16)
I4 q̈3 = −keff (q3 − q2 ).
In matrix form, the equations of motion of a free vibration with x = (q1 , q2 , q3 )T become
   
I1 1 −1
r2
 I2 + s2 I3  ẍ + keff −1 1 + s2 −s2  x = 0 with s= . (5.17)
r3
I4 −1 1
We notice that in the special case r2 = r3 (i.e., s = 1) the above stiffness matrix is analogous to
that of three masses coupled by elastic springs with the effective stiffness keff and moving in 1D
(see Example 4.7).
The resulting free, undamped vibration defines three eigenfrequencies (and associated eigenmodes),
which are obtained from solving det(K + λ2 M ) = 0 and using λ = iω. For example, for the special
case I1 = I4 and r2 = r3 , the three eigenfrequencies are obtained analytically as
v  
u
r u
keff u 1 2 GJ
ω1 = 0, ω2 = , ω3 = utkeff  I + with keff = . (5.18)

2 
I1 l
1
I2 + I3 rr32
As expected, there is one zero eigenfrequency due to the rigid-body mode x̂ = (1, 1, 1)T .
————
197
5.2 Dynamics of Deformable Bodies with Non-Negligible Mass
If the mass of a body is no longer negligible, then all those relations derived under static conditions
may no longer apply, including the balance equations as well as the force–displacement relations
characterizing the deformation behavior. Here, the balance equations of linear and angular momen-
tum must be re-formulated in the presence of inertial effects. Importantly, for deformable bodies,
we may no longer assume that the body moves as a rigid body, easily described by the motion of
its center of mass and the vectors of angular velocity and acceleration. Instead, each point on the
body can undergo a very different trajectory.
As an example, consider the vibrating beam from Example 5.1: while EI
its clamped end is not moving at all, its free end undergoes signifi- m
cant motion, and each point along the beam undergoes yet another x
motion. We need to account for this complication by considering
spatially varying fields describing the motion of a deformable body. l
As a consequence, formulating the kinetic relations can be done both at the global, macroscopic
level, i.e., formulating the balance equations of linear and angular momentum for the entire body
(as done in the previous sections), as well as at the local, microscopic level, i.e., at any point inside
the body (as usually done in continuum mechanics and as was done in Mechanics 2 in the context
of stresses in equilibrium).
The global balance of linear momentum of deformable bodies is identical to that of rigid
bodies. Specifically, exploiting that internal forces in sum perform no work (as shown for systems
of particles in Section 2), we still know that
X
Fiext = M aCM (5.19)
i
with the usual definitions of the total mass and the acceleration of the center of mass. To realize
this, recall that we had interpreted rigid bodies as collections of infinitely many rigidly connected
particles. If we now relax the latter constraint and picture a deformable body as a collection of
infinitely many particles that are not rigidly connected (but, e.g., elastically coupled by linear
spring connections between all pairs of particles), then we arrive at the description of a deformable
body, and the same linear momentum balance (5.19) was shown to apply to both rigid and non-
rigid collections of particles (in fact, we derived for the general case of systems of particles; see
Section 2.2.1).
It is important to note that the balance equation (5.19) also holds for subparts of the body (e.g.,
when the body is cut into pieces to reveal internal forces), in which case the internal forces become
external forces and must be considered in the balance equations (see Example 5.4 below). Angular
momentum balance, of course, follows analogously and was derived in previous sections.
Deformable bodies move differently from rigid bodies because each material point of the body can
now show an independent time-dependent motion. Unlike for rigid bodies where, e.g., the velocity
and acceleration transfer formulae uniquely linked the velocity and acceleration, respectively, of
each point on the body, here we may have both motion and deformation, and the latter leads to
variations across the body.
198
Having established that (5.19) also applies to an

arbitrary subset of a deformable (or rigid) body, syy(x)+syy,y(x)dy
we may apply it to an infinitesimal mass element
dm cut from a body Ω at a position x ∈ Ω, as
sxy(x)+sxy,y(x)dy
shown on the right in 2D (the 3D case follows syx(x)+syx,x(x)dx
dy
analogously). On each of the cut surfaces, we in- sxx(x) sxx(x)+sxx,x(x)dx
troduce the normal stress and shear stress com- syx(x) dx
ponents acting on that respective surface, and we
use a Taylor expansion to find the stresses on the sxy(x) dm
right and top edges of the small element of size syy(x)
dx × dy:
σ(x + dx) = σ(x) + ∇σ(x) · dx + h.o.t. (5.20)
For example, for the stress component σxx on the right surface this gives (writing x = (x, y))
σxx (x + dx, y) ≈ σxx (x) + σxx,x (x) dx. (5.21)
Note that we here use coordinates (x, y, z) for consistency with Mechanics 1 and 2. Here and in
the following we denote by (·),x a derivative of (·) with respect to x.
To apply (5.19), we multiply all stress components by respective length of the edge on which they
are acting (thus turning stresses into forces, or to be specific, forces per thickness in 2D). Further
considering a body force density f with components fx ,fy in 2D (such as, e.g., gravity with
f = ρg), we obtain
X
Fi,x = [σxx (x) + σxx,x (x) dx − σxx (x)] dy + [σxy (x) + σxy,y (x) dy − σxy (x)] dx + fx dx dy = dm ax ,
i
X
Fi,y = [σyy (x) + σyy,y (x) dx − σyy (x)] dx + [σyx (x) + σyx,x (x) dy − σyx (x)] dx + fy dx dy = dm ay ,
i
which, using dm = ρ dx dy (per unit thickness in 2D), simplifies to
σxx,x (x) dx dy + σxy,y (x) dy dx + fx dx dy = ρ dx dy ax σxx,x + σxy,y + fx = ρ ax

⇔ (5.22)
σyy,y (x) dx dx + σyx,x (x) dy dx + fy dx dy = ρ dx dy ay σyx,x + σyy,y + fy = ρ ay .
Altogether, we have arrived at the local balance of linear momentum of deformable bodies as
3
X dσij
+ fi = ρai for i = 1, 2, 3 (5.23)
dxj
j=1
where we replaced (x, y, z) by (x1 , x2 , x3 ) to allow for the summation. Recall that in Mechanics 2,
equilibrium on an infinitesimal mass element was shown to obey the equation div σ + f = 0. The
above is the dynamic version thereof. Inserting a = 0 automatically reduces (5.23) to the static
version. (5.23) can be expressed even shorter in symbolic form as
div σ(x, t) + f (x, t) = ρ a(x, t) (5.24)
199
Note that, as a novelty compared to Mechanics 2, all fields (including displacements, strains and
stresses) are now dependent on both position and time.
We can alternatively derive the local balance law also from the following argument. We may write
(5.19) for a body Ω with boundary ∂Ω as
Z Z Z
σn dS + f dV = ρa dV, (5.25)
∂Ω Ω Ω
where the first term represents all external forces applied to the boundary ∂Ω of the body through
tractions σn, the second term denotes external forces in the form of distributed body forces, and
the right-hand side stems from applying our definition of the center of mass and its acceleration,
cf. (3.3). Applying the divergence theorem to the first term and rewriting the equation gives
Z
(div σ + f − ρa) dV = 0. (5.26)
Ω
We know that this relation must hold not only for the whole body Ω but also for any sub-body cut
free from Ω. For the integral to vanish for any arbitrary subset of Ω, we must have
div σ + f − ρa = 0 ⇔ div σ + f = ρa, (5.27)
which is exactly the local balance of linear momentum stated in (5.24).
In this section we went contrary our usual order of introducing the dynamic governing equations –
we started by introducing the above kinetic relations linking accelerations and forces (or stresses).
Let us point out that the kinematics of deformable bodies are the same as those discussed before
for rigid bodies, enriched by those relations derived in Mechanics 2 for the deformation of bodies.
Specifically, we use a displacement field u(x, t) to define the current position y of a material
point originally at an undeformed position x at time t via
y(x, t) = x + u(x, t). (5.28)
For our purposes, we assume that u is continuously differentiable, so that we may work with the
strain tensor
1
ε(x, t) = [∇u(x, t) + u(x, t)∇] (5.29)
2
or, in component form,

1 ∂ui ∂uj
εij (x, t) = (x, t) + (x, t) . (5.30)
2 ∂xj ∂xi
The constitutive law governing the relation between stresses and strains (e.g., Hooke’s law for
linear elasticity) remains valid here – only that stresses and strains are now time-dependent.
200
Example 5.4. Rotating elastic beam (using global balance equations)
Consider an elastic beam (length l, mass m, cross-sectional area A), made of a

linear elastic material with Young’s modulus E. The beam is rotating about a l m
hinge in 2D, driven by a constant external torque M0 and starting from rest.M
0
x
Neglecting gravity, what are the inner reactions (axial force, transverse force, j
bending torque) due to the rotation of the system? B
Let us first describe the motion of the system by relating the applied torque M0 to the rotation.
To this end (and to avoid the reaction forces at the hinge), we use the global balance of angular
momentum about the hinge B, viz.
ml2 M0
IB ϕ̈ = M0 with IB = ⇒ ϕ̈ = . (5.31)
3 IB
By inserting the initial conditions ϕ(0) = 0 and ϕ̇(0) = 0, we find ϕ(t) by integration:
M0 2 M0 2
ϕ(t) = t + at + b and a=b=0 ⇒ ϕ(t) = t . (5.32)
2IB 2IB
To find the inner forces, we consider a free-body diagram of the free end of the
Q m(x)
rotating beam, cut at a distance 0 ≤ x ≤ l from the hinge, as shown. l CM
M C
We now apply the balance equations of linear and angular momentum to the
shown subpart of the body, accounting for the fact that the drawn inner forces N
B x
are now external forces to be considered in the balance equations.
For a cut at a distance x from the hinge, the cut end of the beam shown above has the mass
l−x
m(x) = m . (5.33)
l
Let us formulate linear momentum balance in polar coordinates, which in the radial direction
applied to the cut end of the beam gives the axial force via
l−x x+l 2 l 2 − x2 2 l2 − x2 M02 m 2
m(x)ar (x) = −N (x) ⇒ N (x) = m ϕ̇ (t) = m ϕ̇ (t) = t , (5.34)
l 2 2l l 2IB2
where we exploited that the radial (centripetal) acceleration during the circular motion is
l−x

2 x+l 2
ar = −rB-CM ϕ̇ = − x + ϕ̇2 = − ϕ̇ . (5.35)
2 2
Similarly, linear momentum balance in the tangential direction with

x+l
aϕ = rB-CM ϕ̈ = ϕ̈ (5.36)
2
gives
l−xx+l l 2 − x2 l 2 − x2 M 0 m
m(x)aϕ (x) = Q(x) ⇒ Q(x) = m ϕ̈ = m ϕ̈ = . (5.37)
l 2 2l l 2IB
201
Finally, balance of angular momentum about the center of mass of the cut sub-body yields
l−x M0 l−x
ICM ϕ̈ = −M (x) − Q(x) ⇒ M (x) = −ICM (x) − Q(x) (5.38)
2 IB 2
with the moment of inertia
1
ICM (x) = m(x) (l − x)2 . (5.39)
12
Altogether, this leads to
(l − x)3 (4l + 3x) (l − x)3 (4l + 3x) M0 m
M (x) = −m ϕ̈(t) = − . (5.40)
12l2 12l2 IB
We can easily verify that all inner reactions, N (x), Q(x) and M (x), vanish at the free end (at
x = l). They are shown as functions of the radial coordinate x in the figures below (note that we
inserted IB = ml2 /3). Under static conditions, all three reactions would vanish due to the absence
of any external loads – the dynamic effects are responsible for the inner forces and torque.
2 2
9M0 t 3M0 -M0
2ml
3 N(x) 2l Q(x) M(x)
+ + -
0 l/2 l 0 l/2 l 0 l/2 l
From these inner forces and torque one can also determine the deformation of the rotating beam as
a function of time by computing the axial elongation of the rod from N (x), and the deflection due
to bending from Q(x) and/or M (x). It is important to recall that we assume small strains (as is
typical, e.g, for linear elasticity). As a consequence, deformation is infinitesimal so that changes in
the body’s shape due to deformation are not considered when calculating distances in the kinematic
relations (e.g., the centripetal force is calculated based on the distance from the center of rotation
to the undeformed center of mass, etc.)
————
Example 5.5. Rotating elastic bar (using local balance equations)

Let us approximate a rotor blade by a slender bar of initial length l, mass
m and cross-sectional area A, made of a linear elastic material with Young’s
modulus E. The bar is assumed to rotate in 2D at a constant angular velocity x
Dl
Ω about its end point O. How does the length of the bar change during
rotation? What is the maximum stress inside the bar? l
Since Ω = const., the acceleration at a point x inside the bar (with the x-W
coordinate measured from the hinge, as shown) is given by the radial and tan-
gential components
ar (x) = −xϕ̇2 = ar (x) = −xΩ2 , at (x) = xϕ̈ = 0. (5.41)
202
Therefore, points on the bar are accelerated only in the radial direction, which results in an axial
extension of the bar (but not transverse forces or bending torques resulting in bending of the beam,
as in Example 5.4). Consequently, this is a 1D problem and the only non-zero stress component is
σxx = σxx (x).
The local balance of linear momentum, div σ + f = ρa here simplifies considerably: first, there
are no body forces (f = 0); second, there is only one non-zero stress component (viz. σxx ); third,
all deformation and motion depends only on the axial coordinate x. Therefore, linear momentum
balance at every point of the rod, introducing the mass density ρ = m/(Al), reduces to
dσxx
= ρar (x). (5.42)
dx
As an alternative derivation, we could also consider an infinitesimal segment of the bar of mass
dm = ρA dx, for which linear momentum balance reads
sxx(x)+sxx,x(x)dx
dm ar (x) = A [σxx (x) + σxx,x (x) dx] − Aσxx (x), (5.43)
dm A
so that dividing by A dx gives again x dx
dσxx
ρ ar (x) = (x). (5.44) sxx(x)
dx
Either way, inserting the acceleration (5.41) into the local linear momentum balance and assuming
linear elasticity with
dux
σxx = Eεxx and εxx = (5.45)
dx
leads to

d
2 d dux
−xρΩ = (Eεxx ) = E . (5.46)
dx dx dx
Assuming that Young’s modulus E is constant along the length of the rod gives an ODE for u(x),
whose solution is obtained by integrating twice:
d2 ux ρΩ2 3
E = −xρΩ2 ⇒ ux (x) = − x + c1 x + c2 . (5.47)
dx2 6E
Using the boundary conditions of vanishing displacements at the hinge (the bar is fixed here) and
zero tractions at the free end (i.e., σxx (l) = 0) shows that
dux ρΩ2 2 ρΩ2 2
ux (0) = c2 = 0, E (l) = − l + c1 = 0 ⇔ c1 = l , (5.48)
dx 2E 2E
so that we finally obtain
ρΩ2 2
ux (x) = (3l x − x3 ). (5.49)
6E
The total change of length ∆l of the rod is identical to the displacement at the tip:
ρΩ2 l3
∆l = ux (l) = . (5.50)
3E
203
The stress inside the bar follows as

dux ρΩ2 2 ρΩ2 l2
σxx (x) = E = (l − x2 ). ⇒ σxx,max = σxx (0) = . (5.51)
dx 2 2
As could have been expected, the maximum (tensile) stress is reached at the hinge. Both the
maximum stress and the bar’s elongation grow quadratically with angular velocity Ω.
As an alternative, the elongation could also have been obtained from the inner forces determined
in Example 5.4. From the inner axial force obtained there, see Eq. (5.34), we conclude
l−xx+l 2 N (x) l 2 − x2 2 l 2 − x2 2
N (x) = m ϕ̇ (t) ⇒ σxx (x) = =m Ω = ρΩ . (5.52)
l 2 A(x) 2Al 2
From the stress we obtain the axial strain and the resulting change of length as
Z l
σxx (x) l 2 − x2 2 ρΩ2 l3
εxx (x) = = ρΩ ⇒ ∆l = εxx (x) dx = . (5.53)
E 2E 0 3E
We conclude by noting that the above derivations were based on the assumption Ω = const. If the
angular velocity of the rod was not constant, then we would also need to account for a tangential
acceleration component at = xΩ̇, resulting in an inner transverse force Q(x). The latter, in turn,
by angular momentum balance produces an inner torque M (x). Overall, these inner reactions will
be analogous to those in Example 5.4 in additionally produce bending deformation of the rod.
————
5.3 Waves and Vibrations in Slender Rods
In Section 5.1 we discussed how deformable bodies that are approximately massless can support
vibrations by acting as effective linear springs. There, the mass of the system was concentrated into
particles while the deformable bodies (e.g., bars and beams) only provided the system with stiffness
against motion. Here, we consider the case were the deformable body is not approximately massless
but its mass matters and affects the vibrational behavior of the system. As an introductory example,
we consider slender rods in this section which may show either axial, longitudinal deformation
(i.e., stretching and compression), flexural deformation (i.e., bending), or torsional deformation.
The deformation behavior of slender rods was well described in Mechanics 2, however, there with
the assumption of static equilibrium and without considering a time-dependent system response.
5.3.1 Longitudinal wave motion and vibrations
Let us first consider a long and slender bar undergoing axial deformation in the form of linear
elastic extension and compression. The local balance of linear momentum, as already discussed in
Example 5.5, is obtained from an infinitesimal volume element of length dx, having an infinitesimal
mass dm = ρ(x)A(x) dx and a (possibly varying) cross-sectional area A(x):
dm a(x) = N (x + dx, t) − N (x, t), (5.54)
204
where N (x) is the axial force inside the bar. Here, we do not assume any particular motion of
the bar (other than that it is one-dimensional), so that we generally characterize the motion and
deformation of the bar by the 1D displacement field u(x, t), such that the undeformed position x
at time t = 0 and the deformed position y(x, t) of that same point at time t are linked via
y(x, t) = x + u(x, t). (5.55)
The acceleration of any material point x = const. thus follows as
d d2
a(x) = y(x, t) = 2 [x + u(x, t)] = ü(x, t). (5.56)
dt dt
Insertion into (5.54) along with dm = ρ(x)A(x) dx and taking the limit of dx → 0 yields
N (x + dx, t) − N (x, t) d
ρ(x)A(x)ü(x, t) = lim = N (x, t) = N,x (x, t). (5.57)
dx→0 dx dx
We further know that N (x, t) = A(x) σ(x, t) with the axial stress σ depending on position x and
time t, so that
ρ(x)A(x)ü(x, t) = [A(x)σ(x, t)],x . (5.58)
For a linear elastic bar with Young’s modulus E(x) we know σ(x, t) = E(x)ε(x, t) = E(x)u,x (x, t).
This turns the above linear momentum balance into
ρ(x)A(x)ü(x, t) = [E(x)A(x)u,x (x, t)],x . (5.59)
Finally, let us assume the special case of a homogeneous rod with constant modulus E(x) = E,
density ρ(x) = ρ and cross-sectional area A(x) = A. In this case we arrive at:
s
E
ρü(x, t) = Eu,xx (x, t) ⇒ ü(x, t) = c2 u,xx (x, t) with c= (5.60)
ρ
This is the one-dimensional wave equation, a linear, homogeneous partial differential equation
(PDE) of second order in both time and space. This equation is analogous to (5.47) in the example
of the rotating bar, except that here we make no assumption about the acceleration of the bar
(unlike in the case of the rotation example where the acceleration was given by the centripetal
acceleration). The above wave equation is sufficiently general to study 1D waves and vibrations in
long and slender bars.
Consider, e.g., the general form of a wave moving in the
x-direction with a speed v, for which the displacement is
assumed to be expressed as v
u(x, t) = f (x−vt) = f (ξ) with ξ = x−vt, (5.61)
where f is any smooth, differentiable function describing

the invariant shape of the wave, and we introduce the co-
ordinate ξ for convenience. Note that t
f(x)
∂ ∂ ∂ ∂ x
= and = −v . (5.62)
∂x ∂ξ ∂t ∂ξ
205
Inserting the general from (5.61) into (5.60) and using the relations (5.62) leads to
v 2 f,ξξ (ξ) = c2 f,ξξ (ξ) ⇒ v = ±c. (5.63)
That is, any longituginal wave in the 1D bar must move with the speed c, which is the longitudinal
wave speed in the bar21 . The solution is thus of the general form (with smooth, differentiable
functions fi (ξ))
u(x, t) = f1 (x − ct) + f2 (x + ct). (5.64)
The solution involves so-called characteristics moving in

u(x,t) opposite directions at the longitudinal wave speed c. The
c c
exact form of f1 and f2 depends on the initial/boundary
conditions. For example, if a bar is excited mechanically at
one of its ends by an applied force or displacement, then
a wave travels through the bar at speed c. If the bar is
x excited at an interior point, waves of speed c will travel in
both directions from the point of excitation.
Obviously, the specific solution depends on the boundary and initial conditions. When waves travel
in objects of finite size, e.g., in a rod of length l, then they interact with the ends of the rod. At
the end, waves are reflected, which – over time – results in a superposition of waves that may turn
into a standing wave (depending on the boundary conditions applied to the ends of the bar),
whose eigenfrequencies and mode shapes are analogous to those of vibrations discussed before. For
example, consider hitting a guitar string to produce music: initially (in the first fraction of a second)
waves propagate through the string, reflect at the ends and travel back. But what is ultimately
observed is the string vibrating in a standing wave – the ends being fixed and the string’s center
moving periodically at approximately fixed amplitude. Let us identify such standing waves (and
neglect the exact initial conditions that have produced the latter in a transient manner).
If the boundary conditions do not depend on time (e.g., when a rod is fixed at both ends so that
u(0, t) = u(l, t) = 0 for all times t), then the PDE and boundary conditions admit a separable
solution as follows. In order to solve the wave equation
ü(x, t) = c2 u,xx (x, t), (5.65)
we make the separable ansatz u(x, t) = û(x)q(t) with unknown functions û(x) and q(t). Insertion
into the wave equation yields
q̈(t) û,xx (x)
û(x)q̈(t) = c2 û,xx (x)q(t) ⇔ = c2 . (5.66)
q(t) û(x)
Note that the fraction on the left-hand side of the equation can only depend on t, while the fraction
on the right-hand side can only depend on x. This implies that both fractions are independent of
both x and t and must therefore be constant. If we call that constant −ω 2 , then
q̈(t) û,xx (x)
= −ω 2 = const. and c2 = −ω 2 = const. (5.67)
q(t) û(x)
21
p
As an example, consider steel with E ≈ 210 GPa, ρ ≈ 7870 kg/m3 so that c = E/ρ ≈ 5166 m/s.
206
Now we can independently solve each of these two ODEs, starting with
ω2
q̈(t) + ω 2 q(t) = 0 and û,xx (x) + û(x) = 0. (5.68)
c2
These second-order ODEs are of the same type as the equation of motion for a single-DOF free
vibration (discussed in Section 4.3), so the solution is of harmonic type. We thus write the general
solution as
u(x, t) = û(x)q(t) with

ω ω (5.69)
û(x) = B1 cos x + B2 sin x and q(t) = A1 cos(ωt) + A2 sin(ωt)
c c
with coefficients A1 , A2 , B1 , B2 ∈ R which, along with ω, depend on the boundary and initial
conditions.
In general, we will find more than one eigenfrequency ω (as shown in the examples below). In
fact, we may expect an infinite set of eigenfrequencies {ω1 , ω2 , . . .}. Since the wave equation is a
linear PDE, the complete solution can be expressed as as sum
∞
X
u(x, t) = ûn (x)qn (t), (5.70)
n=1
in which each term in the (generally infinite) sum is of the form (5.69), viz.
ω ω
n n
ûn (x) = Bn,1 cos x + Bn,2 sin x and qn (t) = An,1 cos(ωn t) + An,2 sin(ωn t). (5.71)
c c
The complete solution thus becomes
∞
X
u(x, t) = ûn (x)qn (t)
n=1
∞ h ω ω i (5.72)
n n
X
= Bn,1 cos x + Bn,2 sin x [An,1 cos(ωn t) + An,2 sin(ωn t)]
c c
n=1
All coefficients An,i , Bn,i ∈ R as well as the frequencies ωn ≥ 0 are to be found from the boundary
and initial conditions. The lowest non-zero frequency ω1 is known as the natural frequency (or
fundamental frequency).
Example 5.6. Vibrating bar in the pinned-pinned configuration

u(x,t)
Consider a vibrating bar of length l that is fixed at both ends t1
(this is known as the pinned-pinned or fixed-fixed configua- t2
tion). The general solution to the problem is given by (5.69). t3 x
To find the coefficients as well as ω, let us use the boundary t4
conditions x u(x,t)
u(0, t) = u(l, t) = 0 ∀ t ⇒ û(0)q(t) = û(l)q(t) = 0 ∀ t.
l
207
Since the boundary conditions hold for all times t, we can directly enforce û(0) = û(l) = 0. By
applying the boundary conditions to û(x) from (5.69), we obtain a system of equations which is
written in matrix form as

0 û(0) 1 0 B1
= = ω ω . (5.73)
0 û(l) cos c l sin c l B2
The first equation requires that B1 = 0. The second one has either a trivial solution (B2 = 0) or
it requires that
ω ω
sin l =0 ⇔ l = nπ for any integer n. (5.74)
c c
We have hence indeed obtained an infinite set of eigenfrequencies of a longitudinal bar in its
pinned-pinned configuration, viz.
cnπ
ωn = with n = 1, 2, . . . , ∞. (5.75)
l
Note that the above procedure can be generalized to other boundary conditions, in which case the
non-trivial solution is found from requiring

1 0 cnπ
det ω ω =0 ⇒ ωn = . (5.76)
cos c l sin c l l
With the above infinite set of eigenfrequencies, the complete solution u(x, t) = û(x)q(t) becomes
∞ x
X ct ct
u(x, t) = sin nπ An,1 cos nπ + An,2 sin nπ with An,1 , An,2 ∈ R, (5.77)
l l l
n=1
whose coefficients An,1 and An,2 (with n = 1, . . . , ∞) depend on the initial conditions. The lowest
(non-zero) solution involves the natural frequency ω1 = π cl . Like for the vibrations discussed
previously, each eigenfrequency ωn = cnπ l has an associated mode shape
x
ûn (x) = sin nπ . (5.78)
l
A mode shape is often referred to also as a normal mode (or, especially in the context of acoustic
waves, as a harmonic). That is, the nth term in the solution (5.77) has the shape of ûn (x) whose
amplitude is time-dependent and given by the term in brackets in (5.77) (dependent on the initial
conditions). The mode shapes for the first four (non-zero) eigenfrequencies are shown below on the
left. For general boundary conditions, usually more than one of the modes are excited, so that the
superposition in (5.77) indeed becomes the solution, with several (or many) modes overlaid.
Each mode shape comes with so-called nodes, i.e., points at which ûn (x) = 0 (shown above as
black dots). For example, for the pinned-pinned configuration we know that û(0) = û(l) = 0, so
we must have at least two nodes at the ends of the bar. For higher-order modes, internal nodes
appear as well. The wave length λn of the nth mode is obtained from
ωn 2l
λn =c ⇒ λn = , (5.79)
2π n
and it describes the length of a complete period of each mode shape.
208
^
u ^,xl
u
w1 w4
w3
w3
w1
x/l x/l
w2
w4 w2
pinned-pinned eigenmodes
From the solution, we can also calculate the axial stress σ(x) inside the bar for each mode shape,
which for a homogeneous bar with E(x) = E leads to
σ(x, t) = Eε(x, t) = Eu,x (x, t) = E û,x (x)q(t) ∝ û,x . (5.80)
The shape of the (normalized) stress distribution û,x l is included in the above figure on the right.
We can identify stress nodes at which the stress vanishes. Due to the pinned-pinned configuration,
there are no stress nodes at the two ends (since there are reaction forces).
————
Example 5.7. Vibrating bar in the pinned-free configuration
As another example, consider the case of a pinned-free (or fixed-free) configuration. As a

practical example, this scenario describes the air column in a long and slender pipe (such as those
found in musical wind instruments like flutes and in organ pipes). When excited by the player, the
air column develops a standing wave as a superposition of eigenmodes, which is how the length of
the pipe is uniquely linked to its tone (the latter being determined by the frequency of vibration).
Here, the boundary value problem reads
ü(x, t) = c2 u,xx (x, t) with u(0, t) = 0, u,x (l, t) = 0 ∀ t, (5.81)
where the second condition was obtained from requiring a zero axial force N (l) = EAu,x (l) = 0 at
the free end at x = l. We exploit again that the boundary conditions are independent of time, so
that we must enforce û(0) = û,x (l) = 0. Writing the system of equations in matrix form gives

0 û(0) 1 0 B1
= = ω ω
ω ω
. (5.82)
0 û,x (l) − c sin c l c cos c l B2
As in the pinned-pinned configuration, we obtain B1 = 0, which is now complemented by the

characteristic equation
ω ω ω π
cos l =0 ⇔ l = (2n − 1) for any integer n. (5.83)
c c c 2
209
The infinite set of eigenfrequencies in the pinned-free configuration is therefore
c(2n − 1)π
ωn = , with n = 1, 2, . . . , ∞, (5.84)
2l
and the complete solution, with An,1 , An,2 ∈ R for all n = 1 . . . , n, becomes
∞
(2n − 1)π x (2n − 1)π ct (2n − 1)π ct
X
u(x, t) = sin An,1 cos + An,2 sin . (5.85)
2 l 2 l 2 l
n=1
Here, the natural frequency is ω1 = π2 cl . Each eigenfrequency ωn = c(2n − 1)π/2l has an

associated mode shape and wave length of, respectively,
(2n − 1)π x

4l
ûn (x) = sin and λn = . (5.86)
2 l 2n − 1
The mode shapes for the first four (non-zero) eigenfrequencies are shown below, along with the
stress profiles associated with those normal modes.
^
u ^,xl
u
w1 w4
w3 w3
w1
x/l x/l
w2
w4
w2
pinned-free eigenmodes
————
For any other configuration, e.g., the free-free configuration, the eigenfrequencies and mode shapes
can be derived analogously. We here restrict our discussion to the above two examples.
210
5.3.2 Torsional wave motion and vibrations
Torsional waves in long and slender rods can be analyzed in complete analogy to
the previous section. Now using the twist angle θ(x) to describe the deformation x
of the rod (specifically, the in-plane rotation around the central axis in the cross-
sectional plane), the governing equation of local angular momentum balance
here is l, r,G
Z
I0 (x)θ̈(x, t) = [G(x)J(x)θ,x (x, t)],x . where I0 (x) = ρr2 da
A
is the moment of inertia per unit thickness. For the special case of a homoge- q(x,t)
neous rod with constant properties, this reduces significantly. Especially, the
moment of inertia I0 and the polar area moment J are related by equation
(3.113) so that I0 = ρJ.
This simplies the above into
s
G
ρJ θ̈(x, t) = GJ θ,xx (x, t) ⇒ θ̈(x, t) = c2T θ,xx (x, t) with cT = (5.87)
ρ
Notice that this equation, the torsional wave equation, has the exact same form as Eq. (5.60),
which governed longitudinal waves in slender rods – the only differences being the replacement of
the displacement u(x) by the twist angle θ(x), and the replacement of the wave speed c by the
torsional wave speed cT , also known as the shear wave speed.
Recall that G = E/2(1 + ν), so the shear wave speed is usually smaller than the longitudinal bar
wave speed:
p p
cT = G/ρ < c = E/ρ. (5.88)
The analogy with longitudinal waves also means that there is no need to re-derive all those relations
obtained for standing waves and mode shapes: since the governing equation is of the identical form,
all solutions derived in Section 5.3.1 for longitudinal waves also apply to torsional
waves, if we make the replacements
c → cT , u(x) → θ(x). (5.89)
The general solution of a standing wave is therefore again of type (5.69), whose coefficients and
frequencies are to be found from the specific boundary conditions being applied.
211
5.3.3 Flexural wave motion and vibrations
l Bending (or flexural) deformation is the final type

r(x),EIy(x),A(x) we would like to consider. Unfortunately, this
y x case is more involved than the previous ones due
w(x,t) to the nature of the governing equations. From
q(x,t) Mechanics 2 we know that bending of a beam
z dm (whose axis is along the x = x1 -direction) in-
volves both a transverse force Q(x, t) acting in
A(x) M(x)
q +CM M(x)+M,x(x)dx the z = x3 -direction and a bending torque M (x, t)
w acting around the y = x2 -axis, as shown on the
Q(x) Q(x)+Q,x(x)dx left for a beam of length l and an infinitesimal
dx beam element of length dx and mass dm.
For a slender, linear elastic Euler-Bernoulli beam we assume that the deformation is captured by
the (small) deflection w(x, t) = u3 (x, t) in the z-direction, and the local change in angle of the beam
follows as θ(x, t) = w,x (x, t).
Let us consider an infinitesimal beam element of length dx and mass dm = ρ(x)A(x) dx, whose
transverse forces Q and torques M (approximated by a Taylor expansion for a small beam element,
as before) are shown above. Linear momentum balance in the z-direction yields:
dm ẅ(x, t) = Q(x, t) − Q(x, t) + Q,x (x, t) dx ⇒ ρ(x)A(x)ẅ(x, t) = Q,x (x, t). (5.90)
Similarly, angular momentum balance around the center of mass of the beam element yields
dx ( dx)2
dI(x)θ̈(x, t) = M (x, t) − M (x, t) + M,x (x, t) dx − 2Q(x, t) − Q,x (x, t) , (5.91)
2 2
where dI(x) = dm Iy (x) = ρ(x)Iy (x) dx is the moment of inertia and Iy is the area moment of
inertia (which also governs the bending stiffness). Dividing (5.91) by dx leads to
dx
ρ(x)Iy (x)θ̈(x, t) = M,x (x, t) − Q(x, t) − Q,x (x, t) . (5.92)
2
The last term on the right-hand side (multiplying dx with | dx| 1) is significantly smaller than
the others and is therefore neglected in the following. In order to insert (5.90), we differentiate
(5.92) with respect to x, so that insertion of (5.90) yields
[ρ(x)Iy (x)θ̈(x, t)],x = M,xx (x, t) − Q,x (x, t) = M,xx (x, t) − ρ(x)A(x)ẅ(x, t). (5.93)
Rotational inertia is commonly much less significant than translational inertia for beam bending
(deflections are more severe than axis rotations, so |Iy θ̈| |Aẅ|). Therefore, one may also drop
the term on the left-hand side of the above equation, so that
−M,xx (x, t) + ρ(x)A(x)ẅ(x, t) = 0. (5.94)
Finally, we recall the constitutive relation of elastic beam bending,
M = −EIy w,xx , (5.95)
212
which transforms (5.94) into
[E(x)Iy (x)w,xx (x, t)],xx + ρ(x)A(x)ẅ(x, t) = 0. (5.96)
For the special case of a homogeneous beam with constant Young’s modulus E(x) = E and mass
density ρ(x) = ρ, and constant cross-section A(x) = A such that also Iy (x) = Iy = const., the
above reduces to
ρA
w,xxxx (x, t) + ẅ(x, t) = 0 (5.97)
EIy
This is a homogeneous PDE of fourth order in space and second order in time, which is to be solved
for the deflection w(x, t) for specific initial and boundary conditions.
As for the previous cases of longitudinal and torsional waves, we may seek solutions in the form of
a standing wave. If the boundary conditions do not depend on time, then we may again make a
separable ansatz and insert it into (5.97):
ρA
w(x, t) = ŵ(x) q(t) ⇒ ŵ,xxxx (x)q(t) + ŵ(x)q̈(t) = 0. (5.98)
EIy
As discussed for longitudinal deformation, we observe that the following two fractions must be
constant, so we define
ŵ,xxxx (x) ρA q̈(t)

=− = k 4 = const., (5.99)
ŵ(x) EIy q(t)
which leads to two ODEs in space and time, respectively:

EIy
ŵ,xxxx (x) − k 4 ŵ(x) = 0 and q̈(t) + k 4 q(t) = 0. (5.100)
ρA
The time-dependent part, as before, yields to a harmonic solution
EIy
q(t) = A1 cos(ωt) + A2 sin(ωt) with ω2 = k4 . (5.101)
ρA
The mode shape ŵ(x) is more complex here due to the fourth-order ODE. The solution is obtained
from the ansatz
ŵ(x) = B exp(λt) ⇒ λ4 ŵ − k 4 ŵ = 0 ⇒ λ = ±k ∨ λ = ±ik. (5.102)
With these four possible solutions, we thus arrive at22
ŵ(x) = B1 cos(kx) + B2 sin(kx) + B3 cosh(kx) + B4 sinh(kx). (5.103)
22
We here exploit that, as discussed for single-DOF vibrations, solutions of the type exp(ikx) and exp(−ikx) can
be converted into solutions cos(kx) and sin(kx). Likewise, we here use that cosh(kx) = 12 [exp(kx) + exp(−kx)] and
sinh(kx) = 12 [exp(kx) − exp(−kx)] to transform the remaining two terms in the solution. Also, as before we may
replace A1 cos(ωt) + A2 sin(ωt) everywhere by A cos(ωt + ϕ), if convenient.
213
The general solution of the flexural vibration problem follows as
w(x, t) = ŵ(x) q(t) (5.104)
with
ŵ(x) = B1 cos(kx) + B2 sin(kx) + B3 cosh(kx) + B4 sinh(kx) and

EIy (5.105)
q(t) = A1 cos(ωt) + A2 sin(ωt), ω2 = k4
ρA
where coefficients A1 , A2 and B1 , B2 , B3 , B4 ∈ R are to be obtained from initial/boundary

conditions. We observe that the mode shape ŵ(x) and the eigenfrequency ω are linked through
k. To determine the eigenfrequencies and associated mode shapes, we need to supply the above
general solution with boundary conditions.
The complete solution follows as the superposition of all modes:
∞
X
w(x, t) = ŵn (x) qn (t)
n=1
∞
(5.106)
X
= [Bn,1 cos(kn x) + Bn,2 sin(kn x) + Bn,3 cosh(kn x) + Bn,4 sinh(kn x)]
n=1
s
2 EIy
× [An,1 cos (ωn t) + An,2 sin (ωn t)] , ωn = kn
ρA
Example 5.8. Vibration of a cantilever beam
As an example, we consider the free vibration of a cantilever beam in its clamped-free config-
uration. If the beam is clamped at x = 0, then both the deflection and the tilt angle must vanish
there for all times t:
w(0, t) = 0, w,x (0, t) = 0 ∀ t ⇒ ŵ(0) = 0, ŵ,x (0) = 0. (5.107)
A free end at x = l implies that both the bending torque and the transverse force vanish at that end.
Since we know from beam theory that M = −EIw,xx ∼ w,xx and Q = M,x = EIw,xxx ∼ w,xxx , we
require that
w,xx (l, t) = 0, w,xxx (l, t) = 0 ∀ t ⇒ ŵ,xx (l) = 0, ŵ,xxx (l) = 0. (5.108)
The four conditions (5.107) and (5.108) applied to ŵ(x) from (5.105) become in matrix form
    
0 1 0 1 0 B1
0  0 k 0 k  B2  .
  
 = (5.109)
2 2 2 2
0 −k cos(kl) −k sin(kl) k cosh(kl) k sinh(kl)  B3 
0 k 3 sin(kl) −k 3 cos(kl) k 3 sinh(kl) k 3 cosh(kl) B4
214
The non-trivial solution is obtained from the characteristic equation

 
1 0 1 0
 0 k 0 k  6
−k cos(kl) −k sin(kl) k cosh(kl) k sinh(kl)  = 2k [1 + cos(kl) cosh(kl)] .
0 = det  2 2 2 2

k 3 sin(kl) −k 3 cos(kl) k 3 sinh(kl) k 3 cosh(kl)

(5.110)
Even though not possible analytically, one can numerically find the roots of 1+cos(kl) cosh(kl) = 0,
which yields an infinite set of solutions:
kn l = 1.8751, 4.6941, 7.8548, 10.9955, . . . (5.111)
The eigenfrequencies are obtained via

s s r
2 EIy 2 EIy 2 EIy
ωn = kn = (kn l) 4
= (kn l) for n = 1, . . . , ∞ (5.112)
ρA ρAl ml3
and evaluate to
r r r r
EIy EIy EIy EIy
ωn = 1.87512 , 4.69412 , 7.85482 , 10.99552 , ... (5.113)
ml3 ml3 ml3 ml3
For each eigenfrequency ωn , we can find the corresponding mode shape by inserting kn into the
coefficient matrix in (5.109) and solving for the eigenvector. This can be performed numerically.
For example, for ω1 we find the coefficients (not normalized)
(B1,1 , B1,2 , B1,3 , B1,4 )T = (1, −0.734096, −1, 0.734096)T , (5.114)
while the mode shape associated with ω2 has the coefficients
(B2,1 , B2,2 , B2,3 , B2,4 )T = (1, −1.01847, −1, 1.01847)T , (5.115)
and the third mode shape has
(B3,1 , B3,2 , B3,3 , B3,4 )T = (1, −1, −1, 1)T . (5.116)
Inserting the Bi -coefficients as well as the corresponding solution for k from (5.111) into ŵ(x) from
(5.105) provides the sought mode shapes. Note that the above coefficient vectors are, in principle,
only defined up to a constant, since the complete solution defines the amplitude of each mode
through qn (t).
The first three modes of the cantilever beam (normalized such that ŵ(l) = 1) are plotted below.
As before, the complete solution is a superposition of all modes whose amplitudes depend on the
initial conditions.
215
^
w
w1
x/l
w2 w3
cantilever beam mode shapes
————
Other beam configurations, such as the free-free configuration (zero bending torques and forces at
both ends) or the pinned-pinned configuration (zero deflection and zero torques at both ends), can
be treated analogously. The respective boundary conditions are applied to the general solution, and
the resulting system of equations admits a non-trivial solution for an infinite set of eigenfrequencies.
There are various experimental demonstrations of beam mode shapes (see, e.g., this movie).
Finally, notice that – unlike in the cases of longitudinal and torsional waves – traveling waves of
the form w(x, t) = f1 (x − ct) + f2 (x + ct) are not solutions of equation (5.97), since it is of fourth
order in space and second in time. The flexural wave speed is therefore dependent on the frequency
of the wave (to be discussed elsewhere). Moreover, the wave speed in bending depends not only on
material properties (such as E, G and ρ), but it also depends on the geometry of the beam through
the factor EIy /ρA since
EIy EIy Iy
ẅ(x, t) + w,xxxx (x, t) = 0 and = c2 . (5.117)
ρA ρA A
216
effective stiffness of a rod in extension, bending, and torsion:

F EA F 3EI M GJ
keff = = , keff = = 3 , keff = =
∆l l w l ∆θ l
global balance of linear momentum for bodies and sub-bodies:

X
Fiext = M aCM
i
local balance of linear momentum:

3
X dσij
div σ + f = ρa or + fi = ρai for i = 1, 2, 3
dxj
j=1
longitudinal wave equation for stretching/compression of a homogeneous slender bar:

s
E
ü(x, t) = c2 u,xx (x, t) with c=
ρ
torsional wave equation for twisting of a homogeneous slender bar:

s
G
θ̈(x, t) = c2T θ,xx (x, t) with cT =
ρ
general solution for longitudinal vibrations (and torsion analogously):
u(x, t) = û(x)q(t) with

ω ω
û(x) = B1 cos x + B2 sin x and q(t) = A1 cos(ωt) + A2 sin(ωt)
c c
flexural wave equation for bending of a homogeneous slender bar:
ρA
w,xxxx (x, t) + ẅ(x, t) = 0
EIy
general solution for flexural vibrations:
w(x, t) = ŵ(x)q(t) with q(t) = A1 cos(ωt) + A2 sin(ωt)

EIy
and ŵ(x) = B1 cos(kx) + B2 sin(kx) + B3 cosh(kx) + B4 sinh(kx), ω2 = k4
ρA
complete solution for longitudinal and flexural vibrations, respectively:
∞
X X∞
u(x, t) = ûn (x)qn (t), w(x, t) = ŵn (x)qn (t)
n=1 n=1
217
Appendix
A What is a tensor?
In mechanics, we frequently need the concept of a tensor, which may be nebulous at first. We here
try to summarize what a tensor is, how to work with it, and what the difference between a tensor
and a matrix is. In this context, it is best to start with something we understand well: vectors.
A.1 What is a vector?
A vector is an object characterized by a magnitude and a direction.

If we define a reference frame C with a Cartesian basis {e1 , e2 , e3 } in
3D, then we may define a vector v as v
v2
3
X e2
v = v1 e 1 + v2 e 2 + v3 e 3 = vi e i , (A.1)
e3 e1
i=1 v3
v1
where {v1 , v2 , v3 } are the components of vector v in this basis.
For any given basis {e1 , e2 , e3 }, one can obtain these components, e.g., by projecting vector v onto
the base vectors, since vi = v · ei (for i = 1, 2, 3).
Although we often refer to a vector v by its components

 
v1
[v]C = v2  ,
 (A.2)
v3
these three components are insufficient to describe the vector. Without knowing the basis of C
p used here, the information {v1 , v2 , v3 } is useless. It may reveal the magnitude of the vector
being
as v12 + v22 + v32 but, without knowing the chosen basis, it does not imply a particular direction.
Therefore, we differentiate between the vector v (see (A.1)) and its components [v] (see (A.2)) –
the vector being generally understood and independent of a chosen basis, while its components
make sense only within a particular reference frame.
Note that, if we know vector v, we can compute its components in any Cartesian frame; e.g., in a
frame D with basis {e1 , e2 , e3 }, since [vi ]D = v · ei . Hence, v contains the complete information
needed to find its components in any given basis (it contains much richer information than its
components alone). If we know the bases of two different frames, say {e1 , e2 , e3 } and {e1 , e2 , e3 }, we
can also relate the vector components in these two frames via a rotation matrix (see Section 3.4.1).
A.2 What is a matrix?
Next, we quickly introduce a matrix as a two-dimensional array of numerical values. For a matrix
T having n rows and m columns, we write T ∈ Rn×m and represent matrix T (e.g., for n = m = 3)
218
as
 
T11 T12 T13
T = T21 T22 T23  , (A.3)
T31 T32 T33
where Tij denotes the component in the ith row and the jth column of the matrix. Therefore,
matrix T is nothing but an ordered collection of 3 × 3 = 9 numerical values.
A.3 What is a tensor?
In many applications we need linear mappings from vectors onto vectors; simply
rotation:
put, a linear mapping is a function (or a black box) which takes a vector as
u R
an input and returns a vector as its output. For example, imagine projecting e2
vectors v onto a certain direction n or into a plane. You provide a vector v, and v
the projection finds a new vector u (which is the projection of v onto the given e3 e1
direction n or into the given plane). Another example are active rotations:
for any given vector v, we rotate v about some given axis by a given angle, projection: v
resulting in a new vector u pointing in a different direction (see Section 3.4.1).
Such linear mappings map vectors onto vectors. n u
The simplest linear mapping we know is the identity I: it maps any vector v onto itself, i.e.,
Iv = v, (A.4)
which we read as “I applied to v yields v”. For more complex mappings (such as the projections
or rotations mentioned above), we introduce a linear mapping operator T – analogous to the above
identity – such that
T v = u, (A.5)
meaning “T applied to v yields u” (where u 6= v in general).
If we now introduce our C-frame with a specific Cartesian basis, we may write any linear mapping
from v to u in 3D as
    
T11 T12 T13 v1 u1 X 3
T21 T22 T23  v2  = u2  or Tij vj = ui for i = 1, 2, 3. (A.6)
T31 T32 T33 v3 u3 j=1
Because the mapping is linear, we know that we must be able to express it in the above form (the
new vector u must be some linear combination of the components of the original vector v). For
example, for the identity mapping I we know its components
        
1 0 0 u1 1 0 0 v1 v1
[Iij ] = 0 1 0 so that u2  = 0 1 0 v2  = v2  (A.7)
0 0 1 u3 0 0 1 v3 v3
P3
or, for short, ui = j=1 Iij vj = vi for i = 1, 2, 3.
219
It is important to realize that eqns. (A.5) and (A.6) aim to give the same message – yet they contain
different information. Analogous to our discussion of vectors above, (A.5) is a general statement:
vector v is mapped onto vector u, irrespective of whatever reference frame one wants to use. By
contrast, (A.6) only applies in a particular chosen reference frame C, in which (v1 , v2 , v3 ) = [v]C and
(u1 , u2 , u3 ) = [u]C are the components of vectors v and u, respectively. Analogously, we interpret
the matrix with components Tij in (A.6) as the components of a tensor T in the chosen C-frame,
so that we may write
 
T11 T12 T13
T21 T22 T23  = [T ]C . (A.8)
T31 T32 T33
If we wanted to use a different frame with a different basis D to calculate u, then the components of
vectors v and u would change, which implies that also the components of T would need to change.
Just like [v]C is valid only in the C-frame, [T ]C are the components of T only the C-frame, which
would need to change when going to a different frame. Most importantly, the mapping in (A.6)
only makes sense if the components of T , v and u are all with respect to the same basis.
To understand the intricacies of a tensor, let us go one step futher and introduce the following vector
operation. We denote by ⊗ a tensor product (also known as outer product or dyadic product), which
can operate between two vectors a and b and is defined by the relation
(a ⊗ b)v = (b · v)a (A.9)
In other words, a ⊗ b acting on vector v results in a new vector which points in the direction of a
and has the magnitude (b · v)|a|. This defines a linear mapping from vectors onto vectors: for a
given vector v, the object a ⊗ b turns it into a new vector (b · v)a (which is linear in v). Therefore,
a ⊗ b must be a tensor. As an example, consider n ⊗ n for some unit vector n. When applied to
any vector v, this leads to
(n ⊗ n)v = (n · v)n ( with |n| = 1), (A.10)
which is nothing but the projection of vector v onto the n-direction (this defines the projection
tensor T = n ⊗ n).
To understand what a ⊗ b means, let us write the right-hand side of (A.9), using b · v = b1 v1 +
b2 v2 + b3 v3 , in some Cartesian frame as
    
(b1 v1 + b2 v2 + b3 v3 )a1 a1 b1 a1 b2 a1 b3 v1
[(b · v)a] = (b1 v1 + b2 v2 + b3 v3 )a2  = a2 b1 a2 b2 a2 b3  v2  . (A.11)
(b1 v1 + b2 v2 + b3 v3 )a3 a3 b1 a3 b2 a3 b3 v3
But since (A.9) tells us (b · v)a = (a ⊗ b)v and hence
[(b · v)a] = [(a ⊗ b)v] = [a ⊗ b][v], (A.12)
we can identify the components of tensor a ⊗ b from (A.11) as
 
a1 b1 a1 b2 a1 b3
[a ⊗ b] = a2 b1 a2 b2 a2 b3  . (A.13)
a3 b1 a3 b2 a3 b3
220
Therefore, the components of [a ⊗ b] form a 3 × 3-matrix with ai bj being the component in the
ithe row and jth column. For example, we can evaluate the following outer products (using the
Cartesian frame with basis {e1 , e2 , e3 } so [e1 ] = (1, 0, 0), [e2 ] = (0, 1, 0), etc.), e.g.,
     
1 0 0 0 1 0 0 0 0
[e1 ⊗ e1 ] = 0 0 0 , [e1 ⊗ e2 ] = 0 0 0 , [e2 ⊗ e3 ] = 0 0 1 , etc. (A.14)
0 0 0 0 0 0 0 0 0
Note that this allows us to write for any matrix

     
T11 T12 T13 1 0 0 0 1 0 X 3
T21 T22 T23  = T11 0 0 0 + T12 0 0 0 + . . . = Tij [ei ⊗ ej ], (A.15)
T31 T32 T33 0 0 0 0 0 0 i,j=1
which motivates the definition of a tensor T , using (A.9), as
3
X
T = Tij ei ⊗ ej (A.16)
i,j=1
We can also verify the action of this tensor T onto a vector v = 3k=1 vk ek as
P
 
X3 3
X X3 3
X
Tv =  Tij ei ⊗ ej  vk ek = Tij vk (ei ⊗ ej )ek = Tij vk ei (ej · ek ). (A.17)
i,j=1 k=1 i,j,k=1 i,j,k=1
Realizing that for orthonormal basis vectors we must have

(
1, if j = k,
ej · ek = (A.18)
0, else,
(A.17) simplifies correctly to

3
X 3
X
Tv = Tij vk ei (ej · ek ) = Tij vj ei , (A.19)
i,j,k=1 i,j=1
so that T v is indeed a vector, whose components are given by [T v] = Tij vj .
A.4 Brief summary
While a vector is defined as

 
3
X v1
v= vi ei with components [v] = v2  , (A.20)
i=1 v3
221
a tensor is defined as
 
3
X T11 T12 T13
T = Tij ei ⊗ ej with components [T ] = T21 T22 T23  . (A.21)
i,j=1 T31 T32 T33
Both v and T are independent of the chosen basis, while their components vi = [v]C and Tij = [T ]C
do depend on the chosen basis C. If we want to switch between different bases, then we may use
transformation rules to transform the components of either a vector or a tensor from one basis to
another. (For the transformation of vector components, see the discussin of passive rorations in
Section 3.4.1; for transformations of tensor components, we refer to the discussion in Mechanics 2
around Mohr’s Circle).
Technically speaking, a tensor T provides a mapping from vectors v onto vectors T v. An example
from mechanics is the stress tensor σ: for any given surface normal n, it provides the stress (or
traction) vector on that surface as t = σn. Another example is the (mass) moment of inertia
tensor IB : for any angular velocity vector ω, it provides the resulting angular momentum vector
HB = IB ω (assuming that B is the center of mass or a stationary point).
Finally, what we called a “tensor ” here is, strictly speaking, a second-order tensor, as the definition
in (A.21) involves two basis vectors (ei ⊗ ej ). A vector, as defined in (A.20), is also referred to as a
first-order tensor (a single basis vector is required), while a scalar can be viewed as a zeroth-order
tensor ; and one can also defined higher-order tensors, e.g., a third-order tensor as a ⊗ b ⊗ c or the
fourth-order stiffness tensor C (which is needed in Continuum Mechanics).
222
Index
acceleration, 4 critically damped vibration, 161

acceleration transfer formula, 72 curvature, 12
active rotations, 113, 114
amplitude, 160, 167 damping matrix, 173, 175
angular acceleration, 10 deformable, 3, 193
angular momentum, 30, 80, 81 degrees of freedom, 4
angular momentum balance, 30, 52 deviatoric moment, 88
angular momentum transfer formula, 97 displacement, 200
angular velocity, 10, 66, 117 DOF, 4
apparent forces, 122 drag, 5
area moment of inertia, 85, 87
eccentric collision, 105
axial moment of inertia, 86
eigenfrequency, 160, 179, 207
binormal vector, 13 eigenmode, 179
body, 3, 64 eigenvalue problem, 178
body force, 199 elastic, 8
body frame, 133, 134 elementary rotations, 114
equation of motion, 33
center of mass, 45, 64 equilibrium, 1, 154, 155
center of percussion, 109, 110 Euler acceleration, 121
centrifugal force, 121 Euler angles, 141
centripetal, 11 Euler force, 121
centripetal acceleration, 13, 121 Euler spiral, 27
centrode, 68 Euler’s equations, 133
centroidal moment of inertia, 84 excitation frequency, 165
characteristic equation, 159, 178 external force vector, 175
characteristics, 206 external forces, 48
clothoid, 27
fictitious forces, 122
coefficient of restitution, 39, 57, 106
flexural, 194
collision, 37, 56, 105
forced vibration, 158, 165, 184
component form, 4
free vibration, 158, 159, 183
conservation of energy, 23, 101
fundamental frequency, 160, 196, 207
conservation of linear momentum, 80
conservative, 101, 154 generalized degrees of freedom, 152
conservative force, 23, 50 generalized force, 153
conservative system, 22, 23, 50 generalized velocity, 153
constitutive law, 200 global balance of linear momentum, 198
constrained, 7 gyroscope, 146
constraint force, 7, 46 gyroscopic limit, 145
continuous fields, 65
Coriolis acceleration, 121 harmonic, 208
Coriolis force, 121 Hessian, 155
Cornu spiral, 27 holonomic constraint, 45, 152
223
Dynamics Prof. Dr. D.M. Kochmann
November 8, 2021 ETH Zürich
homogeneous ODE, 158 Newton’s first axiom, 15

Hooke’s law, 200 Newton’s second axiom, 16
Newton’s third axiom, 16
ICR, 67 nodes, 208
impact, 37 non-conservative force, 24
impulsive force, 38 non-conservative system, 21
inertial frame, 4, 15 non-inertial frame, 8, 11, 111
inertial reference frame, 111 normal mode, 208
inhomogeneous ODE, 158 nutation, 141, 142
instantaneous axis of rotation, 71
instantaneous center of rotation, 67 oblique collision, 105
intermediate axis theorem, 148 orthonormal basis, 3, 4
internal forces, 48 osculating circle, 12, 13
overdamped vibration, 160
kinematic constraint, 7, 45
kinematics, 1, 3, 15 parallel axis theorem, 89
kinetic energy, 20, 101 particle, 3
kinetic friction coefficient, 17, 18 passive rotations, 113, 114
kinetics, 1, 3, 15 path length, 11
period, 160
Lagrange equations, 154 period of compression, 38, 56
Lagrangian, 154 period of restitution, 38, 56
Lehr’s damping ratio, 159 phase delay, 167
linear elastic, 193 phase space representation, 158
linear elasticity, 200 planetary gear, 94
linear momentum, 15, 79 point, 65
linear momentum balance, 16, 49, 80, 121 polar coordinates, 8
local balance of linear momentum, 199 polar moment of area, 87
longitudinal, 195 position, 3, 4
longitudinal wave speed, 206 potential energy, 22
precession, 141
magnification factor, 168 principal axes, 88
mass accretion, 59 principal directions, 88
mass density, 64 principal normal vector, 12
mass loss, 60
mass matrix, 173, 175 radius of curvature, 12
metastable, 155 radius of gyration, 87
modal decomposition, 186 Rayleigh damping, 175
mode shape, 179, 208 Rayleigh’s dissipation function, 175
moment of inertia, 31 reaction force, 7
moment of inertia of a disk, 84 resonance, 167, 189
moment of inertia of a slender rod, 84 resultant force, 15
moment of inertia tensor, 81, 85 resultant torque, 30, 52
multi-DOF vibrations, 172 rigid, 3, 7, 45, 50
rigid body, 64
natural frequency, 160, 167, 207 rigid system, 45, 51
224
rigid-body motion, 179 structural damping, 187

rocket equation, 60
rotation in 2D, 119 tennis racket theorem, 148
rotation matrix, 112 thrust, 61
rotational kinetic energy, 101 time-invariant, 4
torque, 30
second-order tensor, 85 torsional, 195
shear, 199 torsional wave equation, 211
shear wave speed, 211 torsional wave speed, 211
slider crank, 68 total mass, 64
sliding friction, 18 translational kinetic energy, 101
space curve, 11 TSP-rule, 145
special orthogonal group, 114 tuned mass damper, 184
speed, 4
undamped vibration, 159
spin, 141, 142
underdamped vibration, 161
spin tensor, 117
unstable, 155
spinning top, 111, 135, 141
stable, 155 velocity, 4
standing wave, 206, 213 velocity transfer formula, 66, 72
state space representation, 158 vibration, 158
static, 154
steady precession, 144 wave, 205
Steiner’s theorem, 89 wave equation, 205
stiffness matrix, 173, 175 wave length, 208
strain, 85, 200 work, 20
stress, 85, 199 work–energy balance, 20, 50, 101
225
Glossary
acceleration “Beschleunigung”: the acceleration is the rate of change of the velocity of any particle
or point. ‘Acceleration’ generally refers to the acceleration vector a, but it is also sometimes
used for its magnitude a = |a|. 4
amplitude “Amplitude”: the peak-to-peak distance (i.e., the difference between peak (highest
value) and trough (lowest value) of a harmonically varying quantity, e.g., the DOF of a
vibration. 160, 167
angular acceleration “Winkelbeschleunigung”: the angular acceleration ω̇ = ϕ̈ (written in 2D)

is the rate of change of the angular velocity per time. 10
angular momentum “Drehimpuls”: the angular momentum vector is the rotational equivalent
of the linear momentum vector. 30, 80, 81
angular momentum balance “Drehimpulsbilanzgleichung”: the balance of angular momentum

states that any angular momentum applied to a system through external torques is converted,
without losses, into a change of the angular momentum of the system. 30, 52
angular velocity “Winkelgeschwindigkeit”: the angular velocity ω = ϕ̇ (written in 2D) is the rate
of angle change per time. 10, 66, 117
apparent forces “Trägheitskräfte”: forces observed by a moving observer (e.g., on a rotating

system) due to the motion of the reference frame. 122, 229
area moment of inertia “Flächenträgheitsmoment”: a geometrical property of the cross-sectional

area of a body, which characterizes how its shape is distributed with regard to an arbitrary
axis (needed in the analysis of bending). 85, 87
bending “Biegung”: flexural deformation of a slender body (especially of beams or plates). 226,
229
binormal vector “Binormalvektor ”: unit vector perpendicular to both the velocity vector and
the principal normal vector. 13
body “Körper ”: a body is an object of finite size and having a specific shape (unlike a particle
which is an infinitesimally small point mass). 3, 64
body force “Volumenkraft”: a distributed force that acts throughout the volume of a body. 199
center of mass “Massenmittelpunkt, Massenschwerpunkt”: the average position of all particles

of a system, weighted according to their masses; for a rigid body with uniform density, the
center of mass is the centroid. 64
center of percussion “Stosszentrum, Aufschlagmittelpunkt”: the center of percussion of a rigid

body attached to a hinge is that point for which a perpendicular impact onto the body will
produce no reactive forces at the pivot/hinge (note that this is our use of the definition, while
there are others out there). 109, 110
226
centrifugal force “Zentrifugalkraft”: apparent/fictitious force of a body due to a rotating refer-

ence frame, pushing the body outward from the axis of rotation. 121
centripetal “zentripetal ”: pointing radially away from the center (of a rotation). 11
centripetal acceleration “Zentripetalbeschleunigung”: acceleration of a body due to a rotating

reference frame, accelerating the body inward towards the axis of rotation. 13, 121
centrode “Polbahn”: locus of all ICRs of a moving body. 68
characteristic equation “charakteristische Gleichung”: equation obtained from equating the

characteristic polynomial of a matrix to zero. 159
characteristics “Charakteristik ”: for a PDE such as the wave equation, a characteristic is a a

line x = x(t) along which the field u(x, t) = const.. 206
coefficient of restitution “Stossziffer, Stosszahl ”: the ratio of impulsive force magnitudes during
restitution and compression periods, or the negatvie ratio of the normal relative velocity
components of impacting particles after and before the collision. 39, 106
component form “Komponentenform, Komponentenschreibweise”: we may express any vector

(or tensor) v in a given coordinate frame C through its components [v]C measured in that
frame. 4
conservation of energy “Energieerhaltung”: conservative systems maintain a constant total en-

ergy (sum of kinetic and potential energy) over time. 23, 101
conservation of linear momentum “Impulserhaltung”: the total linear momentum of a particle

or body remains constant if no external forces are acting. 80
conservative “konservativ ”: a force is conservative if it derives from a potential; a system is

conservative if all forces are conservative. 101, 154
conservative force “konservative/eingeprägte Kraft”: a force that derives from an energy poten-
tial and which along any closed path does not perform any total work. 23, 50
conservative system “konservatives System”: a system that maintains a constant total energy
(i.e., the sum of kinetic and potential energy remains constant). 22, 23, 50
constitutive law “Konstitutivgesetz ”: the material-dependent relation between stresses and strains.
200
constrained “eingeschränkt, beschränkt”: if a particle or body cannot move freely but is forced
to move in a particular fashion, its motion is constrained . 7
constraint force see reaction force. 7, 46
Coriolis acceleration “Coriolisbeschleunigung”: acceleration of a body due to a rotating refer-

ence frame, stemming from a relative velocity with respect to the moving frame. 121
227
Coriolis force “Corioliskraft”: apparent/fictitious force of a body due to a rotating reference

frame, stemming from a relative velocity with respect to the moving frame. 121
damping matrix “Dämpfungsmatrix ”: the matrix multiplying ẋ in the linearized equations of

motion, associated with viscous damping. 175
deformable “deformierbar ”: the ability of a body to change its shape (the opposite of rigid). 193
degrees of freedom “Freiheitsgrade”: a body or particle has d degrees of freedom (DOFs), if its
configuration in space can uniquely be described by d independent state variables (e.g., the
position vector x = {x1 , . . . , xd } of a particle moving through space, or the position vector
x = {x1 , . . . , xd } and its current rotation θ. 4
deviatoric moment “deviatorisches Flächenträgheitsmoment”: an off-diagonal component of the

area moment of inertia tensor. 88
displacement “Verschiebung”: the vector pointing from the undeformed position of a point to its
deformed position. 200
eccentric collision “exzentrischer Stoss”: a collision whose impulsive forces’ line of action does
not go through the centers of mass of the participating bodies. 105
eigenfrequency “Eigenfrequenz ”: a frequency at which a body or system tends to oscillate with-

out any applied forcing or damping. 160, 179, 207, 229, 231
eigenmode see mode shape . 179
eigenvalue problem “Eigenwertproblem”: a problem of the type Kx = λx to be solved for

eigenvectors (“Eigenvektoren”) x and eigenvalues “Eigenwertwerte” λ. 178
equation of motion “Bewegungsgleichung”: a differential equation in time which describes the

mechanical behavior of a system in terms of its time-dependent motion. Provided initial
conditions, the equation of motion is solved to determine the motion of the system. 33
equilibrium “Gleichgewicht, Gleichgewichtslage”: a configuration of a system in which the resul-

tant forces and torques vanish. 1, 154, 155
Euler acceleration “Eulerbeschleunigung (Azimutalbeschleunigung)”: acceleration of a body due

to a rotating reference frame, stemming from a rate of change of the angular velocity of the
moving frame. 121
Euler angles “Eulersche Winkel ”: introduced by Leonhard Euler, these three angles describe the
orientation of a rigid body with respect to a fixed 3D reference frame (used, e.g., for spinning
tops whose motion is described uniquely by the three angles). 141
Euler force “Eulerkraft”: apparent/fictitious force of a body due to a rotating reference frame,
stemming from a rate of change of the angular velocity of the moving frame. 121
228
Euler’s equations “Eulersche (Kreisel-)Gleichungen”: a vectorial first-order system of ODEs

describing the rotation of a rigid body in a rotating reference frame M̂ (which rotates with
the body and whose basis aligns with the body’s principal axes of inertia. 133
excitation frequency “Anregerfrequenz, Erregungsfrequenz ”: the frequency at which a forced

vibration is induced through an externally applied force or torque. 165
external forces “äussere Kräfte”: forces applied externally to a system (i.e., not forces between
particles or bodies in systems, or forces within deformable bodies). 48
fictitious forces “Trägheitskräfte”: see apparent forces. 122
flexural “Biege-”: having to do with bending (e.g., flexural deformation implies bending defor-
mation). 194
forced vibration “erzwungene Schwingung”: a vibration witho an applied periodic forcing (ap-
plied force, displacement, torque, etc.). 158, 165, 184
free vibration “freie Schwingung”: a vibration without any applied forcing. 158, 159, 183
fundamental frequency (lowest) eigenfrequency of a system. 160, 207
generalized degrees of freedom “generalisierte Koordinaten”: independent DOFs describing

the motion of a system uniquely (considering any imposed constraints). 152
generalized force “generalisierte Kraft”: a force that is conjugate to a generalized DOF. 153
generalized velocity “generalisierte Geschwindigkeit”: time derivative of a generalized degree of

freedom. 153
gyroscope “Gyroskop, Kreiselinstrument, Kreiselstabilisator ”: a fast spinning axisymmetric wheel,

whose axis of rotation is free to assume any orientation in space. 146
gyroscopic limit “gyroskopisches System”: the limit |ψ̇| |ϕ̇, i.e., the special case of a fast-
spinning top which behaves like gyrosope. 145
harmonic see mode shape. 208
Hessian “Hesse-Matrix ”: the Hessian matrix (or simply Hessian) is a square matrix containing
all second-order partial derivatives of a scalar function. 155
holonomic constraint “holonome Nebenbedingung”: a holonomic constraint is a kinematic con-

straint which can be defined as f (x1 , . . . , xn , t) = 0, i.e., which depends only on the degrees
of freedom and time (not their derivatives). 45, 152
Hooke’s law “Hooke’sches Gesetz ”: linear relation between stresses and strains (e.g., σ = Eε in
1D, σ = Cε in higher dimensions), cf. linear elasticity. 200, 230
229
impulsive force “Kraftstoss”: the change of momentum due to a collision; in other words, a force
(integrated over time) which acts on an object for a very short time interval during a collision.
38
inertial frame “Inertialsystem”: an inertial frame is an absolute frame of reference whose basis
and origin are independent of time. Only in inertial frames do Newton’s axioms hold. 4, 15,
232
inertial reference frame “Inertialsystem”: a frame of reference that is fixed and not moving (so
that Newton’s axioms apply). 111
instantaneous axis of rotation “Momentanachse”: the axis around which each point on a rigid
body undergoes a pure rotation at a given instance of time. 71
instantaneous center of rotation “Momentanpol ”: the point with respect to which each point
on a rigid body undergoes a pure rotation at a given instance of time in 2D. 67
intermediate axis theorem “Dschanibekow-Effekt”: a theorem describing the motion and tum-
bling of a rigid body with three distinct principal moments of inertia. 148, 235
internal forces “innere Kräfte”: forces not applied externally to a system; e.g., forces between
particles or bodies in systems of particles/rigid bodies, or forces within deformable bodies.
48
kinematic constraint “kinematische Nebenbedingung”: a kinematic constraint is a mathematical

relation that restricts the degrees of freedom of a dynamic system (e.g., a rigid ground imposes
the kinematic constraint that a particle or body cannot penetrate the ground, or two particles
connected by a rigid bar cannot change their distance. 7, 45
kinematics “Kinematik ”: the concepts of describing the relation between the geometric state of
a system (i.e., its motion and possibly deformation), on the one hand, and the applied forces
and torques that cause the motion (and deformation), on the other hand. 1, 3
kinetic energy “kinetische Energie”: the energy stored in a particle of body at a given time due
to its motion. 20, 101
kinetic friction coefficient “kinetischer (oder dynamischer) Reibkoeffizient”: the kinetic friction
coefficient is analogous to the static friction coefficient and characterizes the ratio between
friction and normal force magnitudes (µ = |R|/kN |) during dynamic frictional sliding. 17,
18, 234
kinetics “Kinetik ”: the concepts of describing the geometric state of a system, i.e., its motion
(and deformation, in general). 1, 3
linear elastic “linear elastisch”: a material behavior that is characterized by a linear relation
between the (infinitesimal) stress and strain components. 193
linear elasticity “lineare Elastizitt”: material behavior characterized by a linear relation between
stresses and strains (cf. Hooke’s law). 200, 229
230
linear momentum “(linearer) Impuls”: linear momentum of a body or particle is the product
of its mass m and the velocity of its center of mass, i.e., P = mvCM . Please do not confuse
“momentum” with the German ”Moment” (= “torque” in our terminology). 15, 79
linear momentum balance “Impulsbilanzgleichung”: the balance of linear momentum states

that any linear momentum applied to a system through external forces is converted, without
losses, into a change of the linear momentum of the system. 16, 49, 80
longitudinal “Längs-”: running lengthwise/along the axis; having to do with axial deformation.
195
longitudinal wave speed “Longitudinalwellengeschwindigkeit”: speed at which longitudinal (ax-

ial) waves propagate in a 1D linear elastic rod. 206
mass accretion “Massenzunahme”: mass accretion describes the continuous acquisition of ad-
ditional mass of a body over time (e.g., a droplet growing while absorbing more fluid from
damp air, or a snow ball rolling down a snowy ground). 59
mass density “Massendichte”: mass per volume of a continuous body. 64
mass loss “Massenverlust”: mass loss describes the continuous reduction of a body’s mass over
time (e.g., a rocket losing mass in the form of exhaust gas). 60
mass matrix “Massenmatrix ”: the matrix multiplying ẍ in the linearized equations of motion,
associated with kinetics/inertia. 175
metastable “metastabil ”: metastable equilibria corresponds to saddle points in the potential en-
ergy landscape. 155
mode shape “Eigenmode, Normalmode”: the eigenvector associated with an eigenfrequency of a

linearized dynamical system. 179, 208, 228, 229, 232
moment of inertia “Massenträgheitsmoment”: the moment of inertia characterizes a particle’s

or body’s inertia against rotational acceleration (analogous to the mass characterizing inertia
against linear acceleration. 31
moment of inertia tensor “Massenträgheitsmomententensor ”: this tensor characterizes a rigid

body’s moment of inertia around all possible axes of rotation. 81, 85
natural frequency (lowest) eigenfrequency of a system. 160, 167, 207
nodes “Knoten”: in the context of standing waves, nodes are points of constant zero motion (i.e.,
points at which the mode shape vanishes). 208
non-conservative force “nicht-konservative Kraft”: a force whose work, when applied to a mov-
ing particle, depends on the particle’s path and cannot be derived from a position-dependent
potential. 24
231
non-conservative system “Arbeitssatz ”: a system that reduces its total energy (i.e., the sum of
kinetic and potential energy) over time due to the action of non-conservative forces such as
friction or viscous drag. 21
non-inertial frame “Nicht-Inertialsystem”: a non-inertial frame is a moving frame of reference

whose basis and/or origin depend on time; see also inertial frame. 8, 11, 111
normal mode see mode shape. 208
nutation “Neigung, Nutation”: the tilt of a body away from the normal to the ground; varying
nutation results in a rocking or swaying motion of the axis of rotation of the body. 141, 142
orthonormal basis “orthonormale Basis”: a set of vectors {e1 , . . . , en } is orthonormal, if each

vector is of unit length (i.e., |ei | = 1) and if each pair of vectors is orthogonal (i.e., ei ·ej = δij ,
or ei ⊥ej for i 6= j). 3, 4
osculating circle “Schmiegekreis”: the circle which best approximates a space curve at a given
point (goes through the point and has the same curvature). 13
parallel axis theorem “Steinerscher Satz, Satz von Steiner ”: relates moment of inertia tensors
with respect to two different reference points to each other. 89, 235
particle “Massenpunkt, Punktmasse”: a body whose extensions are negligibly small compared
to its path through space can be assumed to be a particle; particles only have translational
degrees of freedom and no rotational degrees of freedom. 3
path length “Pfadlänge”: the distance travelled by a particle since some initial reference time.
11
period “Periode”: the time of a complete vibration/oscillation cycle. 160
period of compression “Kompressionsphase”: the (initial) period of time [t− , t0 ) during a colli-
sion, within which the two colliding objects move towards each other in the normal direction.
38, 56
period of restitution “Restitutionsphase”: the (later) period of time [t0 , t+ ] during a collision,
within which the two colliding objects move away from each other in the normal direction.
38, 56
phase delay “Phasenverschiebung”: the time delay between an applied forcing and the resultant
response of the system. 167
planetary gear “Planetengetriebe”: also known as epicyclic gear, a gear system consisting of one
or more outer planet gears revolving around a central sun gear . 94
polar coordinates “Polarkoordinaten”: a 2D coordinate system in which a point on a plane is

determined by a distance r from an origin and an angle ϕ (or θ) from a reference direction. 8
232
polar moment of area “polares Flächenträgheitsmoment”: a geometrical property of the cross-

sectional area of a body, which characterizes how its shape is distributed with regard to its
central axis (needed in the analysis of torsion). 87
position “Position, Lage”: a position in space is generally defined with respect to a basis and an
origin, e.g., r = x1 e1 + x2 e2 + x3 e3 in a Cartesian frame of reference. 3, 4
potential energy “potentielle Energie”: the position-dependent energy from which a (sum of)
conservative force(s) derives as − dV / dr. 22
precession “Präzession”: the change in the orientation of the rotational axis of a rotating body
(such as a spinning top). 141
principal axes “Hauptachsen”: the coordinate axes within which the components a second-order
tensor (e.g., the moment of inertia, stress or strain tensor) are diagonal (i.e., all off-diagonal
components vanish). The principal axes are defined by the eigenvectors of the tensor (the
principal values, i.e., the corresponding eigenvalues are the tensor’s diagonal entries in that
frame of reference). 88
principal normal vector “Hauptnormalenvektor ”: unit vector perpendicular to the velocity vec-
tor and pointing towards the center of curvature. 12
reaction force “Reaktionskraft”: a force that arises in response to a kinematic constraint; e.g., if
the trajectory of a particle is guided by a rigid ground, slope or slide, then the ground exerts
a reaction force onto the particle which is normal to the ground. Similarly, a particle attached
to a rigid stick or string experiences a reaction force along the direction of the stick or string.
7, 227
resonance “Resonanz ”: excitation of a system at an eigenfrequency, causing a large-amplitude

system response. 167, 189
resultant force “Resultierende”: the sum of all forces acting on a particle or body (i.e., the net
force). 15
resultant torque “resultierendes Moment”: the sum of all torques acting on a particle or body
with respect to a given point (i.e., the net torque). 30, 52
rigid “starr, unverformbar ”: a rigid body or link is not deformable, i.e., it maintains its shape and
size under arbitrary deformation (e.g., a rigid link is a connecting bar/beam that is infinitely
stiff and hence cannot extend or shrink. 3, 7, 45, 50
rigid body “Starrörper ”: a rigid body is not deformable (or infinitely stiff). 64
rigid system “starres System”: in a rigid system of particles, the relative distances between each
pair of particles remains constant at all times. 45, 51
rigid-body motion “Starrkörperbewegung”: motion of a system (or body) without changing the
relative distances between any of its particles (or material points), hence producing no change
in potential energy. 179
233
second-order tensor “Tensor zweiter Stufe”: the result of a dyadic/outer product of two vectors
(or a linear mapping of vectors onto vectors), defined by a matrix of components and and
associated basis. 85
shear “Schub”: loosely speaking, deformation associated with angle changes; characterized by the
off-diagonal components of the stress and strain tensors. 199
shear wave speed “Scherwellengeschwindigkeit”: speed at which shear waves propagate in a

linear elastic medium. 211, 235
slider crank “Schubkurbel(getriebe)”: arrangement of hinged mechanical parts designed to convert

straight-line motion into circular motion. 68
sliding friction “Gleitreibung”: sliding friction refers to the process of a two bodies sliding past
in which, whose contact is charaterized by both normal forces and friction forces (the latter
opposing the motion). The ratio between friction and normal forces is the kinetic friction
coefficient µ. 18
space curve “Raumkurve”: the trajectory of a particle through space. 11
special orthogonal group “Inertialsystem”: group of all second-order tensors satisfying RRT =
I (to be mathematically exact, this is the group of all distance-preserving isometries, where
the group operation is given by composing transformations). 114
speed “Schnelligkeit, Geschwindigkeit”: the speed v refers to the magnitude of the velocity vector
v, i.e., v = |v|. 4
spin “Spin”: rotation of a body about one of its principal axes, leaving the principal axis stationary
(often but not necessarily of rotational symmetry). 141, 142
spin tensor “Spintensor ”: a second-order tensor mapping any vector onto the cross product of
the angular velocity vector and that vector. 117
spinning top “Kreisel ”: a (often but not necessarily rotational symmetric) body that is spun
quickly around a fixed point (inertia causes it to remain precisely balanced on its tip in the
presence of gravity). 111, 135, 141
stable “stabil ”: loosely speaking, an equilibrium is stable if a small perturbation of the system
does not deviate far from the equilibrium; mathematically speaking, a stable equilibrium of
a conservative system is a potential energy minimum. 155
standing wave “stehende Welle, Stehwelle”: also known as a stationary wave, a standing wave
is a wave which oscillates in time but whose peak amplitude profile does not move in space.
206, 213
state space representation “Zustandsraumdarstellung”: description of a physical system as a

first-order ODE with state variables described as functions of time. 158
static “statisch”: a problem is (quasi)static if inertial effects are negligible, so the kinetic energy
can be neglected. 154
234
Steiner’s theorem “Steinerscher Satz, Satz von Steiner ”: see parallel axis theorem. 89
stiffness matrix “Steifigkeitsmatrix ”: the matrix multiplying x in the linearized equations of

motion, associated with linear spring connections. 175
strain “Dehnung”: deformation per original length. 85, 200
stress “Spannung”: force per original area. 85
tennis racket theorem see intermediate axis theorem. 148
thrust “Schubkraft, Vorschub”: the thrust ṁvm is the propulsive force of a rocket, produced by
gas leaving the rocket at a mass stream ṁ and relative velocity vm . 61
time-invariant “zeitlich invariant”: not changing with time. 4
torque “Moment, Drehmoment”: the product of force times lever arm with respect a given point B
(note that in American English the term ‘moment’ is usually preferred over ‘torque’). 30
torsion “Torsion”: the twisting of a deformable body (e.g., a shaft) due to applied torques. 235
torsional “Torsions-”: having to do with torsion. 195
torsional wave speed see shear wave speed). 211
unstable “instabil ”: loosely speaking, an equilibrium is unstable if a small perturbation of the

system leads to a response that deviates far from the equilibrium; mathematically speaking,
a stable equilibrium of a conservative system is a potential energy maximum. 155
velocity “Geschwindigkeit”: the velocity is the rate of change of the position of any particle or
point. ‘Velocity’ generally refers to the velocity vector v, whereas speed (“Schnelligkeit”)
refers to the magnitude v = |v| of the velocity vector. 4
velocity transfer formula expression relating the velocities of two points on the same rigid body
by the distance vector and the angular velocity vector. 66
vibration “Vibration, Schwingung”: a small-amplitude oscillation around a stable equilibrium.

158
wave length “Wellenlänge”: the spatial period of a periodic wave (i.e., the spatial distance over
which the wave’s shape repeats). 208
work “Arbeit”: amount of energy transferred to a system by moving it through the action of
applied forces. 20
work–energy balance “Arbeitssatz ”: the change in kinetic energy between two states equals the
work performed on the particle or body by external forces. 20, 101
235

Dynamics LectureNotes

Uploaded by

Copyright:

Available Formats

Dynamics LectureNotes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamics LectureNotes

Uploaded by

Copyright:

Available Formats

Dynamics

(Fall Semester 2021)

Prof. Dr. Dennis M. Kochmann

Mechanics & Materials Lab

Copyright c 2021 by Dennis M. Kochmann

2 Dynamics of Systems of Particles 45

3 Dynamics of Rigid Bodies 64

3.5.1 General description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

5 Dynamics of Deformable Bodies 193

A What is a tensor? 218

1.1.1 Kinematics in a Cartesian reference frame

We describe the positions of particles in space within a fixed Carte- x2

Example 1.1. Finding a particle trajectory from a time-dependent acceleration

so that, with v(t0 ) being the initial velocity at time t0 ,

As an example, consider gravity g accelerating a particle in the negative x2 -

Example 1.2. Finding a particle trajectory from a velocity- or position-dependent

Example 1.3. Particle on a circular trajectory

1.1.2 Constrained motion

Example 1.4. Unconstrained and constrained motion

1.1.3 Kinematics in polar coordinates

r(t) = r(t) cos ϕ(t)e1 + r(t) sin ϕ(t)e2 . (1.22)

Similarly, the acceleration follows as

a = r̈ cos ϕ − 2ṙϕ̇ sin ϕ − rϕ̇2 cos ϕ − rϕ̈ sin ϕ e1

or, in component form with respect to the Cartesian frame C,

x = xr er + xϕ eϕ with xr = x · er , and xϕ = x · eϕ . (1.29)

For example, the velocity vector v can be written as v = vr er + vϕ eϕ , vr vt

We here exploit that the co-rotating basis is orthonormal, i.e., er · eϕ = 0 and er · er = eϕ · eϕ = 1,

The velocity vector in the rotating polar frame R hence becomes

v = ṙer + rϕ̇eϕ (1.31)

er · a = ar = r̈ − rϕ̇2 and eϕ · a = aϕ = 2ṙϕ̇ + rϕ̈, (1.32)

a = r̈ − rϕ̇2 er + (2ṙϕ̇ + rϕ̈) eϕ

ar 6= v̇r and aϕ 6= v̇ϕ , (1.35)

Example 1.5. Particle rotating on a taut, inextensible string

1.1.4 Kinematics of space curves

r = r(s) with s = s(t). (1.38) v

and noticing that

This allows us to write the velocity vector in general as

a = s̈et + ṡėt . (1.43)

To find ėt , let us first realize that, since |et | = 1,

et · et = 1 ⇒ 2et · ėt = 0 ⇒ ėt ⊥ et , (1.44)

From geometry, we conclude that (approximately)

Insertion of (1.46) into (1.45) results in

To arrive at a 3D formulation, one may define the (unit) binormal vector

Example 1.6. Circular motion

Next, applying (1.51) with ρ = R = const. yields

We note that for a general 2D motion

1.1.5 A brief summary of particle kinematics

• Cartesian coordinates in d dimensions:

• polar coordinates in 2D:

• space curve description:

1.2.1 Newton’s axioms and balance of linear momentum

I. Newton’s first axiom:

II. Newton’s second axiom:

We will refer to this relation as linear momentum balance (abbreviated as LMB). It is

III. Newton’s third axiom:

Action = Reaction (1.65)

Example 1.7. Sliding block on a frictionless surface

r(t) = x1 (t)e1 , v(t) = ẋ1 (t)e1 , a(t) = ẍ1 (t)e1 . (1.67)

The constraint x2 (t) = 0 entails a reaction force N = N e2 to keep

ma = mẍ1 e1 = (N − mg)e2 . (1.68)