You are on page 1of 71

Mathematical Tripos IA: 0.

Introduction
Dynamics
0.1 Mathematical Preliminaries
John Stewart
R is the set of all real numbers. Rn is a n-
Centre for Mathematical Sciences
dimensional vector space, where n = 1, 2, . . .. If,
Wilberforce Road Cambridge CB3 0WA
in addition, we have a scalar or “dot” product, as
j.m.stewart@damtp.cam.ac.uk
defined in course C1/2 the norm or magnitude of
http://www.damtp.cam.ac.uk/user/john/c8
x ∈ Rn is √
Constructive comments or amendments should be |x| = x.x . (1)
set to the email address above. Printed material Rn with this norm is called Euclidean space En.
is available at the web site in “gzipped Postscript”
format to users from “.cam.ac.uk” sites. If U and V are vector spaces, U × V is the set
of ordered pairs (u, v), where u ∈ U , v ∈ V . U × V
I have only one book recommendation for the is also a vector space.
course as a whole.
Let (i, j) denote the standard basis for E2, i.e.,
• M. Lunn, A First Course in Mechanics, Oxford |i| = |j| = 1, i.j = 0. Then any vector x ∈ E2 can
(1991). be written

I will be recommending other sources for specific x = xi + yj or x = (x, y). (2)


topics.
There are four problem sheets, intended for Since we usually regard x as a “column vector” it
supervisions in weeks 3, 5, 7 and 9. Week 1 is would be more
 accurate (and more cluttered) to
next week; week 9 is next term. write x = xy or x = (x, y)T .

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 1 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 2

0.2 Trajectories ẋ(t1) is called the velocity of the trajectory


x(t) at t = t1.
A function x : R → E2, t → x(t) = x(t)i + y(t)j
(where y(t) is a function R → R) is a trajectory. Further
y d2x(t) d2 x d2 y
≡ ẍ(t) = i + j
t1 dt2 dt2 dt2

t2 is the acceleration of the trajectory x(t) at t.

Exercise 1. Verify that for example 1

x ẋ(t) = (u, v − gt), ẍ(t) = (0, −g).


Example 1.

x(t) = (ut, vt − 12 gt2),


Exercise 2. Define trajectories in E3 and their
where u, v and g are constants. velocities and accelerations. (A typical point in
E3 is x = xi + yj + zk.)
Functions R → E2 can be differentiated in an
obvious manner:
dx(t) dx dy
≡ ẋ(t) = i + j ≡ (ẋ, ẏ).
dt dt dt
Note that dj/dt = 0 etc. Here we regard the basis
as being fixed.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 3 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 4
0.3 The Gradient Operator  
∂ 2f ∂ ∂f
≡ = −ex sin y,
∂x∂y ∂x ∂y
 
Consider a function f : R2 → R, e.g., ∂ 2f ∂ ∂f
≡ = −ex sin y,
∂y∂x ∂y ∂x
f (x, y) = log x + sin y + ex cos y.  
∂ 2f ∂ ∂f
≡ = − sin y − ex cos y.
The partial derivative of f with respect to y ∂y 2 ∂y ∂y
is defined by the following prescription: regard x as It can be shown that for “smooth” functions f ,
fixed, so that f is a function of the single variable y, partial derivatives commute, i.e., ∂ 2f /∂y∂x =
and differentiate it ∂ 2f /∂x∂y.

∂f Each function f : R2 → R defines a vector


= cos y − ex sin y.
∂y written as ∇f , pronounced “grad f ”, via
 
We can do the same for x ∂f ∂f
∇f = , ,
∂x ∂y
∂f 1
= + ex cos y,
∂x x
and for our example
and we can repeat the operation
1 
  ∇f = + ex cos y, cos y − ex sin y .
∂ 2f ∂ ∂f 1 x
2
≡ =− + ex cos y,
∂x ∂x ∂x x2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 5 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 6

Exercise
 3. [Polar coordinates] Consider the functions 1. Introduction to Mechanics
r = x2 + y 2 and θ = tan−1 (y/x), and show
x y  1.1 Mechanics: experimental facts
∇r = , = (cos θ, sin θ),
r r The following approximately true experimental
 y x 1 (3)
∇θ = − 2 , 2 = (− sin θ, cos θ), facts are the basis for mechanics.
r r r

which will be needed later. (A) Our space is three-dimensional, homogeneous,


isotropic and Euclidean, and time is one-
Exercise 4. How would you define partial dimensional.
derivatives and the gradient operator for functions
f : Rn → R? (B) Galileo’s principle of relativity holds: there exist
special coordinate systems called inertial with
If for some vector f , there exists a scalar function the properties
U such that f = −∇U , then U is said to be a
potential for f . The minus sign is conventional. (1) all the laws of nature at all moments of time
are the same in all such coordinate systems,
(2) all coordinate systems in uniform rectilinear
motion with respect to an inertial one
are themselves inertial, as are those which
differ from an inertial one by a constant
translation, rotation or reflection.

(C) Newton’s principle holds: the initial state of a


mechanical system (the totality of positions and

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 7 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 8
velocities of its points at some initial moment in 1.2 Space and time
time) uniquely determines all of its motion.
It is a mistake to say our space is the vector
We now examine them in detail. space E3, for every vector space contains a privileged
vector, 0, which destroys the homogeneity. However
relative positions behave like vectors.

It doesn’t make much sense (until we agree an


origin of time) to say “the lecture started at 1005”
but the assertion “the duration of the lecture was 55
minutes” is readily verifiable. The position of King’s
College Chapel is not a vector, but the displacement
from there to the Cockroft Lecture Theatre is one.

In fact our space and time constitute a 4-


dimensional affine space AR4, whose points are
called events. For the record1 we give a precise
definition of an affine space, although you will never
be asked to use it.

1
For further details see e.g., P. Bamberg & S. Sternberg, A course in
mathematics for students of physics, vol 1, CUP 1990.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 9 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 10

Affine spaces From remark 3 above q = o ⊕ (u + v) and so


q  p = v. To see that this is independent of the
Let V be a vector space. The affine space AV choice of o consider another choice o = o⊕w. Then
consists of a set of points p, q, . . . and an operation p = o ⊕ (−w + u) and q = o ⊕ (−w + u + v) so
⊕ which assigns to each p ∈ AV and each v ∈ V that q  p = v as before.
another point in AV denoted p ⊕ v. The operation This construction is implicitly assumed in
satisfies the axioms: textbooks.

1. associativity: (p ⊕ u) ⊕ v = p ⊕ (u + v) for any


p ∈ AV , u, v ∈ V ,

2. identity: p ⊕ 0 = p for any p ∈ AV ,

3. transitivity: given p, q ∈ AV there is a v ∈ V


such that q = p ⊕ v,

4. faithfulness: if for all p ∈ AV the equality p⊕u =


p ⊕ v holds then u = v.

These imply that given o, p ∈ AV there is a


unique u ∈ V such that p = o⊕u and it is convenient
to write p  o = u. Thus if we choose a preferred,
fixed o, then any point p ∈ AV can be identified
with u ∈ V .

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 11 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 12
1.2 Space and time (continued) 1.3 Galilean Relativity

For Newtonian dynamics (but not special Suppose that we have chosen our frame to be
relativity) we assume that our AR4 is actually an inertial one in the sense of §1.1. Condition B(2)
AR×AE3. The first affine space represents time, the implies that the change
second physical space. We now choose a particular
o ∈ AR × AE3 and measure displacements relative (t, x) → (t, x) = (t, x + ut), (4)
to o so that our space becomes R × E3. Finally we
choose a standard basis (i, j, k) for E3, see exercise 2. where u ∈ E3 is a fixed relative velocity,
The choice of o and basis is a choice of reference defines a new inertial frame. Galileo’s principle
frame or frame. Because of the arbitrariness we (based on empirical evidence) requires that constant
must permit two types of transformation. The first translations
reflects the choice of o, the second the choice of
basis: (t, x) → (t, x) = (t + τ, x + ξ), (5)

1. Translations (t, x) → (t, x) = (t + τ, x + ξ), where τ ∈ R and ξ ∈ E3, and constant
where τ ∈ R, ξ ∈ E3, rotations/reflections

2. Rotations/reflections (t, x) → (t, x) = (t, x) → (t, x) = (t, U x), (6)
(t, U (t)x), where U is a unitary2 3-matrix which
can depend on time. (where U is a constant unitary matrix) generate
new inertial frames. These three transformations
can be combined, each has an inverse and includes
2
U is unitary if | det(U )| = 1, i.e., U represents a rotation or a the identity. They form the Galilean group of
reflection transformations from one inertial frame to another.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 13 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 14

1.4 Special Relativity∗3 A4 If the velocity of S relative to S is positive, then


the relative velocity of S with respect to S is
negative.
The discussion of the previous two sections
reflects the special status allocated to time in classical
A5 If the velocity of S relative to S is positive, and
mechanics. Let us, for simplicity, suppress two space
the velocity of a third inertial frame S relative to
dimensions and consider our spacetime to be an
S is positive, then the velocity of S relative to S
AR2, subject to axiom (B) of §1.1. We now make
is also positive.
the more general assumptions:
The motivation for the first two assumptions
A1 We are considering linear transformations has already been given. A3 is an isotropy
between inertial frames S and S, whose assumption based on empirical evidence. In
coefficients depend continuously on the relative many situations there is no preferred spatial
velocity of S with respect to S. direction. There are situations where this is
not the case—they involve gravitation or more
A2 The transformations form a group. This is often generally general relativity. The fourth assumption
called the relativity principle. An alternative asserts simply that if I see you moving away
version is that if a physical phenomenon holds in from me, then you see me moving away from
one inertial frame, it must hold in every other one, you. (Note that this is a tautology for Galilean
i.e., physics should be frame-independent. transformations, but might not hold for more
exotic transformation laws.) Similarly assumption
A3 There is no preferred direction; any physical A5 reflects common experience, and is a tautology
situation remains invariant under t → t, x → −x, for Galilean transformations.
provided of course, all velocities are reversed. Now, almost miraculously, there are only two
3
An asterisk denotes material which is non-examinable for this course. groups of transformations which satisfy these

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 15 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 16
assumptions/axioms. Obviously the Galilean 1.5 Newton’s principle
group is one of them. The other turns
out to be the Lorentz group in which We now examine Newton’s principle, (C) of
there is a finite upper limit to the relative section 1.1. For a system consisting of a single
speed between two inertial frames. This is particle it asserts that there is a function A such
the basis for special relativity, invented by that
Albert Einstein in 1905 from (almost) totally ẍ = A(t, x, ẋ). (7)
different premises. The justification for these
assertions is not available in most of the Specification of A is the definition of the system.
standard textbooks4 but is discussed in detail in However Galilean relativity places some
http://www.damtp.cam.ac.uk/user/john/srib/. constraints on the form of A. Recall that time
translations preserve inertial frames. Thus if
x = ϕ(t) is a solution of (7), then so also is
x = ϕ(t + s) for any s ∈ R. It follows that in an
inertial frame the right hand side of (7) does not
depend on time

ẍ = Φ(x, ẋ). (8)

Next consider a system of n particles. Fixed


spatial translations preserve inertial frames and so
if xi = ϕi(t) (i = 1, 2, . . . , n) are solutions of
4
The exception is W. Rindler, Essential Relativity, Springer,  1977. (8), then so are xi = ϕi(t) + r where r ∈ E3

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 17 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 18

is fixed. It follows that Φ can only depend on Example 2. Consider a system consisting of a
relative positions single particle. Then the acceleration vector
A cannot depend on t, x or ẋ and so must
ẍi = Φi({xj − xk }, ẋl). (9) be a constant vector, which in addition must
be invariant under rotations. Therefore the
acceleration vanishes; the particle moves with
uniform rectilinear motion. This is usually called
By a similar argument, since inertial frames are Newton’s first law.
preserved by uniform rectilinear motion, Φ can
only depend on relative velocities: Exercise 5. You may have encountered
Newton’s first law before. (If not then consult
ẍi = Φi({xj − xk }, {ẋl − ẋm}), (10) a conventional textbook.) Check for, and if
necessary, resolve any inconsistencies.
where i, j, k, l, m = 1, 2, . . . , n.
Example 3. Consider a system consisting of two
Finally let U be a constant 3 × 3 unitary matrix. particles, which are initially at rest. This system
Because constant reflections/rotations preserve must remain invariant under rotations about the
inertial frames we must require that if xi = ϕi(t) line which joins the two particles initially. Thus
is a solution of (10) then so is U ϕi(t), or the particles remain on this line.

Φi(U {xj − xk }, U {ẋl − ẋm}) = Exercise 6. Consider a system consisting of two


particles. Show that there exists an inertial
U Φi({xj − xk }, {ẋl − ẋm}). (11)
coordinate system in which the two particles
remain in a fixed plane.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 19 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 20
Example 4. [Gravitation] Experiments, first Show that if U = g.x then
conducted by Galileo, show that for a particle close
to the earth’s surface, the height z is governed by ẍ = −∇U.

z̈ = −g, (12)

where g ≈ 9.8m/s2 is a universal constant. If we


introduce the potential U = gz then clearly Example 5. [Non-localized gravitation] Experiments
show that Galileo’s equation (12) is invalid for
dU very large heights and has to be replaced by a
z̈ = − .
dz generalized Galileo’s equation

r02
z̈ = r̈ = −g , (13)
Exercise 7. (The same as example 4 but in r2
vectorial notation.)
where r = r0 + z.
x
1111111
0000000
0000000
1111111
r r0 1111111
0000000
0000000
1111111
g 0000000
1111111
z 0000000
1111111
0000000
1111111
0000000
1111111
The vector g has magnitude g and points vertically Note that if z is small, |z|/r0
1, (13) reduces
downwards, and x is a general “position vector”. to Galileo’s equation (12).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 21 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 22

This acceleration can also be expressed as (minus) 1.6 Mass and Force
the gradient of a potential, z̈ = −∇U where
Most textbooks assume that you already know,
gr 2 intuitively, what mass and force are. We look a
U = − 0.
r little more closely at these concepts.

Example 6. [Inertial mass, Newton’s second law]


Suppose we have a spring with one end fixed and
a particle attached to the other end, constrained
to move along the direction of the spring. If the
displacement of the particle is x, then experiments
show that the acceleration is

ẍ = −α2x,

where α is a constant, which too can be derived


from a potential U = 12 α2x2. Experiments
show that if the particle is replaced by two

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 23 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 24
identical ones, then for the same extension x, the Experiments show that all masses are positive.
acceleration is only half as large. It is established It follows from equation (14) that the product
experimentally that for any two bodies the ratio of of inertial mass and acceleration mẍ does not
the accelerations ẍ1/ẍ2 under the same extension depend on the body and is a characteristic of the
of a spring does not depend on the type of spring extension of the spring. This product is called the
but only on the bodies themselves. We therefore force of the spring acting on the body.
associate a property inertial mass with every
This is an example of a more general,
body. We define the ratio of the inertial
experimentally verified, property
masses to be the inverse of the ratio of their
accelerations
ẍ1 m2 force = inertial mass × acceleration, (15)
= . (14)
ẍ2 m1
usually called Newton’s second law. (We met
Newton’s first law in example 2.) As unit of force
we take the newton with abbreviation N . If one
Although we have defined this ratio for
litre of water is suspended on a vertical spring at
accelerations produced by springs, it is an
the surface of the earth, the spring acts with a
experimental fact that equation (14) holds for
force of 9.8 newtons.
any acceleration mechanism.
Newton’s second law in the form (15) is usually
Eq.(14) does not define the unit of inertial taken as an axiom of mechanics. Note however
mass. By convention we take the unit to be that in the experiments used to substantiate it,
a kilogramme, the mass of a litre ( = 10−3 m3) the mass of the particle is constant. Thus an
of water5 at a fixed temperature. Then the mass entirely equivalent formula is
m2 of a body can be determined by comparing
its acceleration with that of a litre of water. d  
5
It is assumed you are familiar with SI units for length and time.
force = inertial mass × velocity . (16)
dt
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 25 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 26

1.7 Gravitational Force


We call the product inertial mass × velocity the
momentum of the particle, usually denoted p. All bodies exert gravitational attraction, and
Thus (16) can be rewritten as we can quantify this by assigning every body
a gravitational mass M . Suppose we
d have two particles of inertial masses m1, m2,
force = momentum. (17) gravitational masses M1, M2 at positions x1 and
dt
x2 respectively. It is observed experimentally, and
codified as Newton’s law of gravitation that
particle 1 is accelerated towards the other, and the
acceleration is proportional to M2 and inversely
proportional to the square of their separation

x2 − x1
ẍ1 = GM2 , (18)
|x2 − x1|3

where G ≈ 6.67 × 10−11Nm2kg−2 is the


gravitational constant. (Note that this
formula is consistent with Newton’s principle
(11).) Thus the gravitational force exerted by
particle 2 on particle 1 is

x2 − x1
F1←2 = Gm1 M2 . (19)
|x2 − x1|3

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 27 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 28
If we denote by F2←1 the gravitational force Example 7. [Freely falling particle] Suppose that
exerted by particle 1 on particle 2 then it is m2 m1 so that the acceleration of particle 2
obvious that F2←1 = −(m2/M2)(M1/m1)F1←2. is negligible. Then setting x2 = 0, x1 = x and
r = |x|, equation (18) reduces to
It is a remarkable experimental fact (the Eötvös
experiment, to be discussed in a later example)
x
that the ratio of masses m/M is the same for ẍ = −Gm2 .
r3
all bodies, and by choosing units appropriately
we can arrange M = m. Thus there is no need
to distinguish between inertial and gravitational Suppose that at t = 0, |x| = R and ẋ = 0. How
masses despite their very different origins. In long will it take for the particles to meet?
future we just refer to mass. This is clearly a one-dimensional problem
One corollary of this result is that F2←1 =
−F1←2. This is an example of Newton’s third Gm2 dU
r̈ = − =− , (20)
law: action and reaction are equal and opposite. r2 dr
The third law appears to hold for all interaction
forces, not just gravity. where U = −Gm2 /r, see example 5. The easiest
way to integrate this is to note that
Exercise 8. We shall prove later that the
gravitational acceleration produced by an d  1 2 dU d
ṙ = ṙr̈ = − ṙ = − U (r(t)),
extended spherically symmetric body at points dt 2 dr dt
outside it is the same as that produced by a point
of equal mass situated at the centre of the body. where the chain rule has been used, so that 12 ṙ 2 =
Use this fact to deduce Galileo’s law (13) from Gm2/r + const. The initial conditions give ṙ 2 =
Newton’s law (18). 2Gm2(1/r − 1/R).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 29 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 30

Thus the time to the collision is 1.8 Newton’s 3rd Law & Closed
0 Systems
dr
T =−
R |ṙ|
R 1/2 Consider a system of n particles; the ith particle
Rr has momentum pi, and the kth particle exerts
= dr
0 2Gm2 (R − r) a force Fi←k on it. Such forces are internal
π/2 
R2 sin2 θ 1/2 to the system, while forces produced by outside
= 2R sin θ cos θ dθ influences are external. A closed system has
0 2Gm2R cos2 θ
no external forces.
The total momentum of
π  2R3 1/2 n
the system is P = i=1 pi.
= ,
4 Gm2
Recall that for a single particle subject to no forces
where the substitution r = R sin2 θ was used. the momentum is conserved, see (17).

Theorem 1. Consider a closed system of


particles where the internal forces obey Newton’s
third law. Then the total momentum is conserved.

Proof. Using the notation above, eq.(17)


generalizes to

dpi
= Fi←k .
dt
k=i

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 31 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 32
Thus 1.9 Electrical Forces
dP
= Fi←k .
dt i k=i While all bodies have non-negative mass, they
The double sum is a sum over pairs of particles. possess another property, electrical charge q which
The contribution from the pair (2, 5) is F2←5 + can have either sign or be zero. For time-
F5←2 = 0 by Newton’s third law. Thus the sum independent situations the Coulomb law of
vanishes and so P is constant. 2 electrostatics holds. In the notation of §1.7
the force exerted by particle 2 on particle 1 is
Of course the question then arises as to which
systems are closed. q1 q2 x 2 − x 1
F1←2 = − , (21)
4π0 |x2 − x1|3

where 0 is a new universal constant. Thus


electrostatics is effectively gravity in disguise. The
electric field produced at point 1 by particle 2
is
q2 x 2 − x 1
E1←2 = − , (22)
4π0 |x2 − x1|3
so that F1←2 = q1E1←2.
In time dependent problems there is another
vector, the magnetic field to contend with.
Suppose we have a “test particle”, ie., one whose
charge (or mass) is so small that it does not affect
the electrical fields E(t, x) and B(t, x). (This is
the same approximation as that made in example

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 33 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 34

7.) Then if the particle is at x(t) with velocity 1.10 Frictional Forces
ẋ(t), its acceleration is given by the Lorentz
force
The world is a complicated place and we often
mẍ(t) = FLorentz make gross but useful simplifications. It’s difficult
 
= q E(t, x(t)) + ẋ(t) × B(t, x(t)) . cycling against the wind, but rather than discuss
(23) the effect of all of the 1023 air molecules hitting
us, we model it by a simple phenomenological
We shall study simplified versions of this later in frictional force.
§3.4.

Example 8. Linear friction models the force


as proportional to, but opposing the velocity. E.g.,
consider a particle of mass m falling under gravity,
with position z measured downwards

mz̈ = mg − k ż, (24)

or

z̈ = g − αż, (25)

where α = k/m.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 35 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 36
z be kmM/(M + m) times the relative speed of
the bullet with respect to the plate. Show that
the bullet passes through if u > ka, and if so
determine the displacement of the plate and the
g final speed of the bullet.
V0 =
α Let x denote the displacement of the plate.
Because of the form of the frictional force it is
t convenient to describe the position of the bullet by
its displacement y relative to the plate, so that
It is clear that ż = V0 ≡ g/α is a solution of (25). the absolute displacement is x + y. The equation
In fact all solutions ż(t) → V0 as t → ∞. V0 is of motion of the plate is M ẍ = kmM ẏ/(M + m)
the terminal velocity. or
kmẏ
ẍ = ,
M +m
Exercise 9. Suppose that ż(0) = U . Show that and by a similar argument the equation of motion
the solution of equation (25) is of the bullet is

ż(t) = V0 − (V0 − U )e−αt, kM ẏ


ẍ + ÿ = − .
M +m
and hence confirm the statement made above. We can deduce ÿ = −k ẏ so that ẏ + ky = const.
Initially x = y = 0, ẋ = 0, ẏ = u and so ẏ + ky =
Example 9. A slab of armour plate of mass M u. It is straightforward to obtain ẏ = u exp(−kt)
and length a lies on a smooth horizontal bench and y = (u/k)(1 − exp(−kt)). The bullet passes
and a bullet of mass m is fired at it with speed through iff ẏ > 0 at y = a ie., u > ka, the first
u. The resistive force on the bullet is known to result required. If the bullet reaches the far side

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 37 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 38

y = a at t = t1 then exp(−kt1) = 1 − ka/u. We Example 10. Quadratic friction is sometimes


now have ẍ = kmu exp(−kt)/(M + m) leading more realistic. Equation (24) is replaced by
to ẋ = mu(1 − exp(−kt))/(M + m) and
mz̈ = mg − K ż 2sign(ż), (26)

mut mu or, setting v = ż, β 2 = K/m, γ 2 = g and


x= − (1 − e−kt). assuming v = U > 0 at t = 0 and v  0
M + m (M + m)k
v̇ = γ 2 − β 2v 2. (27)
At t = t1 the displacement is
Note that there is a terminal velocity V0 = γ/β
and that the figure of example 8 is valid here.
mu  u  ka Thus
x= log − , v v 
(M + m)k u − ka u dv 1 1 1
t= = + dv
U γ −β v 2γ U γ − βv γ + βv
2 2 2
 
1 γ + βv γ − βU
and the speed of the bullet is = log ,
2βγ γ + βU γ − βv

and so
kM a  
ẋ + ẏ = u − .
M +m (V0 + U )e2βγt − (V0 − U )
v = V0 .
(V0 + U )e2βγt + (V0 − U )

As before ż = v → V0 as t → ∞ independent of
the initial conditions. See also example 24.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 39 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 40
Example 11. In other situations the Coulomb 2 Dimensional Analysis
or dry friction model may be the most
appropriate one. Here the force is We have three fundamental quantities, and with
each of them we can associate an abstract
F = −Ksign(ż). (28) dimension and possible units

• Mass: [M ]: kg, ounce, ton, stone, hundredweight,


carat,

Example 12. A variant on this idea is the • Length: [L]: m, cm, inch, mile, perch, verst,
concept of constraint forces or constraints.
Imagine a rigid straight wire along the x-axis, • Time: [T ]: second, day, month, year, leapyear,
y = z = 0, with a bead sliding along it. The bead millennium.
is constrained to stay at y = z = 0 because
any attempt to move away is opposed by the Now it makes sense to say “length1 = length2”,
(extremely strong) reaction force of the wire. for if it’s true in metres then it’s true in miles.
However a statement “length1 = time2” could
be a numerical coincidence only in one set of
units. Schematically we allow [L] = [L] but not
[L] = [T ]. Besides the fundamental quantities,
we can construct derived ones with their abstract
dimensions, e.g.,

[area] = [L2], [velocity] = [LT −1],


[density] = [mass/volume] = [M L−3],

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 41 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 42

and in an equation a = b + c, a, b and c must Example 13. [Pythagoras] We obtain Pythagoras’


have the same abstract dimension, a useful check theorem from “elementary” premises.
on the algebra.
A
6
Dimensioned quantities occur only in power laws ,
e.g., volume = length3, but never in other c b
functions; what does sin(2 inches) mean?
Only quantities which are dimensionless can occur θ
B C
in general functions, e.g., in sin α = sin(length × a
length/area) where [α] = [L][L]/[L2] = We can specify a right-angled triangle by giving
[L2]/[L2] = [1]. This idea, often called
its hypotenuse (longest side) a and smallest angle
dimensional analysis, although superficially θ. Its area A must therefore be a function of a
simple, has powerful consequences because one
and θ, A = A(a, θ). Notice that θ and A/a2 are
can obtain significant results from (apparently)
dimensionless and so A/a2 = f (θ), or A(a, θ) =
insignificant information. We study some
a2f (θ).
examples of dimensional analysis.
A

c θ b

θ
B C
a D
6
This key result is proved in the appendix to this chapter. Similarly the areas ABD and ADC are c2f (θ)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 43 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 44
and b2f (θ) and their sum is the original area, i.e., Next we construct all of the dimensionless
quantities that can be formed from these. There
c2f (θ) + b2f (θ) = a2f (θ), is essentially one, τ = g
−1ω −2 . Of course 2τ and
τ 2 are also dimensionless, but they are not really
different. As was stated above, in any physical
or
equation the dimensions must match. It follows
a2 = b2 + c2.
that they should be expressible in dimensionless
You should examine, very carefully, what has form, and so the physics of pendulums implies
been assumed here (Euclidean geometry leading that τ will satisfy a relation of the form f (τ ) = 0.
to similarity of triangles, addition of areas), and Now apart from exceptional cases this fixes τ as
what has not been assumed, an explicit formula a (positive) zero of f , i.e., τ = C −2 a constant.
for f (θ), which (trigonometric identity) implies Thus g
−1ω −2 = C −2 whence
the Pythagoras theorem.

g
ω=C .
Example 14. [Simple pendulum] As a more

physical example, consider a simple pendulum,


consisting of a point mass m connected to a fixed Those of you who have seen the simple pendulum
point by a massless thread of length
. The before can verify that (for a particular value of C)
pendulum performs oscillations in a vertical plane this is the correct formula.
(under the influence of gravity) of frequency ω.
What can be said using dimensional analysis? Exercise 10. There is of course a systematic
way to obtain the form of τ . Posit a relation
We first establish the dimensions of m,
, g and
ω τ = g α lβ ω γ ,

[m] = [M ], [
] = [L], [g] = [LT −2], [ω] = [T −1]. where α, β and γ are constants and determine

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 45 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 46

their values by requiring that τ be dimensionless.


Example 15. [Flow in a pipe] Consider a long
straight pipe of length
and circular cross section
of radius a. Suppose that a liquid with viscosity
coefficient η flows through the tube. ([η] =
[M L−1T −1].) We are interested in the flow Q
per unit time as a function of the pressure drop
P per unit length. Now

[P ] = [force][area]−1[L−1] = [M L−2T −2],

and [Q] = [L3T −1]. Note that the physics


suggests that
is irrelevant here. Thus
essentially the only dimensionless parameter is
θ = P a4/(Qη) and so by our usual argument

Ca4P
Q= ,
η

for some constant C. Thus if an oil refinery


doubles the radius of a pipe it can send 16
times as much oil down it. We all have two
large important pipes in our bodies, the left and
right coronary arteries. When fatty material is
deposited on their walls the tubes narrow. There

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 47 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 48
is a substantial safety factor built in, but the implies a dimensionless equation of the form
fourth power law means that once the danger
level is reached, quite small further deposits can f (τ1, τ2) = 0,
have catastrophic results.
and in general this is all that we can say. Subject
Example 16. [Surface waves] Consider surface to smoothness requirements though, the implicit
waves in an ocean of depth H These have function theorem implies τ1 = τ1(τ2). However
wavelength λ and travel with speed c. How without extra information we cannot proceed
does c depend on H, λ, g and ρ the density of further
water? Suppose though that the ocean is very deep, i.e.,
λ
H. Then one might expect the physics to be
independent of H, or in our notation f (τ1, τ2) =
λ g(τ2/τ1) = g(c2/(gλ)). Thus c2/(gλ) = A2, a
(constant) positive zero of g, and we deduce that
for deep waves.
H

c = A gλ.
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
111111111111111111111111111111
000000000000000000000000000000
The other extreme is a very shallow ocean,
We have already written down the dimensions of H
λ. In this case we expect c to be almost
all of the quantities involved and it is easy to λ-independent, or in dimensionless notation,
verify that there are essentially two independent f (τ1, τ2) = h(τ2). By a similar argument we
dimensionless quantities τ1 = λ/H and τ2 = can deduce 
c2/(gH). The physics of water waves presumably c = B gH,

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 49 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 50

where B is a constant. Both types of behaviour Taylor knew from theoretical reasoning that K ≈
are verified experimentally. 1 and so  2 1/5
t E
r≈ .
ρ
Example 17. [Atomic bombs] Our final example
relates to atomic bomb explosions. In the period The value of ρ is no secret, and so plotting r
1944–48 the US government regarded the amount as a function of t using the movie film, Taylor
of energy E released as top secret. They did deduced, correctly, the value of E.
however release movies of explosions, and from
these the Cambridge mathematician G.I. Taylor
deduced and published the result E ≈ 1021erg,
causing great embarrassment all round. How did
he do this?

Taylor modelled the explosion as the release of


energy E in a small volume, essentially a point, at
t = 0. There is a spherical blast wave of radius r.
He argued that r can depend only on t, E and ρ
the density of air. We know the dimensions of all
quantities, viz. [r] = [L], [t] = [T ], [ρ] = [M L−3]
and you should know [E] = [M L2T −2]. Positing
a relation r = Ktαρβ E γ where K, α, β and
γ are dimensionless constants, Taylor used the
technique of exercise 10 to obtain the exponents

α = 25 , β = − 15 , γ = 15 .

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 51 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 52
Appendix A: Composite dimensions Equating the two forms for a2/a1 we obtain a
are always power laws∗ functional equation for f

Earlier in this chapter it was claimed that all f (L2, M2, T2)
physical quantities have dimensions given by = f (L2/L1, M2/M1, T2/T1).
f (L1, M1, T1)
power laws e.g., [velocity] = LT −1. We now
prove this result. Suppose the quantity a has
To solve it we first take the partial derivative with
dimension given by
respect to L2 keeping all other quantities fixed.
[a] = f (L, M, T ),
1 ∂f  L2 M2 T2 
∂f
∂L (L2 , M2, T2)
= , , .
and that in some standard unit system it takes f (L1, M1, T1) L1 ∂L L1 M1 T1
the value a0. Suppose now we move to a new
system in which the units are decreased by factors Next set L2 = L1 = L etc.
L1, M1 and T1. Then the new value will be a1 =
a0f (L1, M1, T1). Suppose that a2 is similarly ∂f
∂L (L, M, T ) 1 ∂f α
defined. Then = (1, 1, 1) = ,
f (L, M, T ) L ∂L L
a2 f (L2, M2, T2)
= .
a1 f (L1, M1, T1) say. Now the solution of y  (x)/y(x) = α/x is
y = cxα, where c is a constant. Here we are
However we could regard system 1 as the standard, differentiating with respect to L holding M and
and system 2 having been obtained by decreasing T fixed, and so the “constant of integration” can
the units by factors L2/L1, M2/M1 and T2/T1, depend on M and T . Thus
i.e.
a2
= f (L2/L1, M2/M1, T2/T1).
a1 f (L, M, T ) = g(M, T )Lα.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 53 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 54

Substituting this back into the original functional 3 Simple Examples


equation, we obtain a new one of identical type

g(M2, T2) 3.1 Oscillations


= g(M2/M1, T2/T1).
g(M1, T1)
The extension of an elastic spring of unstretched
We solve this by the same technique finding length
0 and actual length
>
0 is (

0)/
0.
g(M, T ) = h(T )M β and a further application It is observed empirically, Hooke’s law, that the
of the technique gives h(T ) = cT γ , where c spring exerts a force or tension T given by
is a constant “constant of integration”. Thus
f (L, M, T ) = cLαM β T γ , a power law as claimed.  


0
T =λ , (29)

0

where λ is the modulus of elasticity of the


spring (see example 6). An elastic string behaves
similarly whenever the extension is positive.

Example 18. [Longitudinal oscillations] An unstretche


spring P O has one end P fixed at x = −
0 and
the other end O free at x = 0. A particle of mass
m is attached at O and is given a displacement
in the x-direction. What happens?
If the particle is at x(t) then the length of the
spring is
0 + x(t) and the tension is λx(t)/
0 .

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 55 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 56
If x > 0 the tension acts on the particle in the ẋ
negative x-direction and so
a
mẍ = −T = −λx/
0,
a x
or ω

ẍ + ω 2x = 0, ω 2 = λ/(m
0). (30)
Clearly x = (a/ω) sin θ(t), ẋ = a cos θ(t) satisfies
equation (31) for any choice of θ(t). Imposing
This is the equation of simple harmonic the condition ẋ = dx/dt gives θ̇ = ω, i.e., x =
motion (SHM). To integrate equation (30) (a/ω) sin(ωt + θ0). The motion is periodic with
multiply it by 2ẋ period τ = 2π/ω. Note that the physics requires
a non-negative length for the spring. Thus a/ω <
2ẋẍ + ω 2(2xẋ) = 0,
0 .

or
d 2
(ẋ + ω 2x2 ) = 0,
dt
which implies

ẋ2 + ω 2x2 = a2, const. (31)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 57 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 58

Example 19. [Transverse oscillations] A particle Example 20. [Large oscillations] A simple pendulum
of mass m is attached to the mid point P of consists of a rigid massless rod OP of length

a tightly stretched string AB of length 2b and with the end O fixed. P carries a mass m and the
tension T . It is displaced slightly through y system is released from rest at an angle α to the
perpendicular to the direction of the string. Find vertical. Find the period of the oscillations.
the period of the resulting oscillations. O
P θ
T T Q
y T

A B
b O P
 θ
Note that AP = b2 + y 2 = b(1 + (y/b)2)1/2 ≈ mg
b(1 + 12 (y/b)2) ≈ b if terms quadratic in y/b are
neglected. Thus the tension remains T and
Because we are not interested in the tension
y y
mÿ = −2T  ≈ −2T T we resolve vectors along the direction P Q
2
b +y 2 b perpendicular to OP . P has an instantaneous
velocity
θ̇ and acceleration
θ̈ along P Q and this
by 
the same argument. This is SHM of period is opposed by the component of gravity mg sin θ
2π mb/(2T ). along P Q. Thus mlθ̈ = −mg sin θ or

g
θ̈ + ω 2 sin θ = 0, ω2 = . (32)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 59 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 60
We try the usual trick, multiplication by 2θ̇, 3.2 Projectiles
finding
d 2 We look at a selection of problems involving
(θ̇ − 2ω 2 cos θ) = 0 motion under gravity.
dt
so that θ̇ 2 − 2ω 2 cos θ = C, a constant. When
Example 21. A ball is thrown from the origin
θ = α, θ̇ = 0 and so
O with speed V at an angle α to the horizontal.
Show that if V is sufficiently large, there will
θ̇ 2 = 2ω 2 (cos θ − cos α). (33) be two trajectories passing through the arbitrary
point P , (x = X, y √ = Y ), and that the times
Thus the time taken for θ to change from 0 to α taken satisfy t1t2 = 2 X 2 + Y 2/g.
is α Neglecting air resistance we have ẍ = 0, ẋ =
1 dθ
τ= √ √ , (34) V cos α and x = V t cos α. Similarly ÿ = −g,
ω 2 0 cos θ − cos α
ẏ = V sin α − gt and y = V t sin α − 12 gt2. Now
and the period is 4τ . The integral (34)
set x = X, y = Y . The answer appears not to
cannot be evaluated in terms of simple functions;
involve α and so we eliminate it:
the indefinite integral is a Jacobean elliptic
function. If α is small we can find τ 1
either by approximating the integral (34) or by X 2+(Y + gt2 )2 = V 2t2 cos2 α+V 2t2 sin2 α = V 2t2,
2
approximating the equation (33). In the latter
case we may set cos α ≈ 1 − 12 α2 and, since or
|θ|  α, cos θ ≈ 1− 12 θ 2, finding θ̇ 2 +ω 2θ 2 = ω 2α2 1 24
which is SHM with period 2π/ω. g t + (gY − V 2)t2 + X 2 + Y 2 = 0.
4

This is a quadratic for t2. Provided gY − V 2 <


0 and (gY − V 2)2 > g 2(X 2 + Y 2) (which will

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 61 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 62

certainly be the case if V is sufficiently large) The next example involves both the use of vectors
there will be two positive roots t12 and t22 with and linear friction.
t12t22 = 4(X 2 + Y 2)/g 2.
Example 23. A ball is thrown from the origin
Example 22. [Transition to vector equations] with initial velocity V in a crosswind of velocity
Recall the dynamics content of the previous U. Assuming that the air resistance produces a
example. We had force of −k times the relative velocity, determine
the orbit.
ẍ = 0, ẋ = V cos α, x = V t cos α, Let the ball have position x(t), and let the
gravitational force be mg and the frictional force
ÿ = −g, ẏ = V sin α − gt, y = V t sin α − 12 gt2. be F = −k(ẋ − U). Thus mẍ = mg − k(ẋ − U),
or
We can combine these two sets of equations, ẍ + β ẋ = g + βU, (35)
setting x = x(t)i + y(t)j, V = V (cos αi + sin αj) where β = k/m. The complementary function is
and g = −gj. Then easily seen to be x = c+de−βt where c and d are
constant vectors. We need a particular integral,
ẍ = g, ẋ = V + gt, x = Vt + 12 gt2. and so we guess x = at finding a = U + g/β.
The general solution of equation (35) will be the
Notice that we could start from the vector form of sum of these two solutions. We fix c and d by
the equation of motion and integrate it directly. requiring x(0) = 0 and ẋ(0) = V so that
This is a very useful technique, and it works for
1 1 1
any linear problem. x(t) = (V − g − U)(1 − e−βt) + (U + g)t.
β β β

It should be noted that this is planar motion.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 63 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 64
The next example extends example 10 log(1 + x) = x − 12 x2 + . . . and so 2ky ≈ (1 − 12 )
on quadratic friction and illustrates some or  
approximation techniques. V2 kV 2
y≈ 1− .
2g 2g
Example 24. A particle is fired vertically
upwards with speed V . Besides gravity there In order to find the time we start from v̇ =
is an acceleration due to air resistance opposing −g − kv 2 again. Separating variables directly
the motion with magnitude kv 2 where v is the (dv/(g + kv 2 ) = −dt) we find (recall v = V at
speed. Assuming air resistance is small, find the t = 0)
maximum height and the time taken to reach it.
   
What does “small” mean? Clearly kV 2/g is 1 −1 k 1 −1 k
√ tan v = −t+ √ tan V
dimensionless and we assume  ≡ kV 2/g
1. kg g kg g
We can proceed as in example 10 or as follows. Let
v = ẏ. On the upward path v̇ = −g − kv 2 . Now The highest
√ √point occurs when v = 0, t =
dv/dt = (dv/dy)(dy/dt) = v(dv/dy) = −g −kv 2 (tan−1
)/ kg. Now Gregory’s series is
and so tan−1 x = x − 13 x3 + . . . when |x| < 1, and this
dv 2
= −2g − 2kv 2. leads, after some simple algebra, to
dy
Separation of variables yields log(2kv 2 + 2g) = √    
  V kV 2
−2ky + C. Since v = V at y = 0 t≈√ 1− ≈ 1− .
kg 3 g 3g
 2 
kv + g
log = −2ky.
kV 2 + g
Example 25. [Example 24 continued] What is
The highest point occurs when v = 0 and 2ky = its speed U when it reaches the ground again, and
log((kV 2 + g)/g) = log(1 + ). Now if |x| < 1, why is it different from V ?

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 65 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 66

Here we have to be careful about the sign of the 3.3 Rockets


air resistance. On the downward path we have

dv 2
= −2g + 2kv 2, A rocket has mass m(t), velocity v(t) and
dy
is acted upon by a force F(t). It emits
exhaust gas at a rate |ṁ| (where ṁ < 0)
with first integral log(2g−2kv 2 ) = 2ky+C, where with (usually constant) velocity U relative to the
C is a constant. (The apparently gratuitous factor rocket. The details of this problem involve very
of −1 is to ensure that we take log of a positive complicated physics. Fortunately there are so
argument.) When v = 0 we know (see top of many microscopic particles in the gas that we can
the previous page) that 2ky = log(1 + ) and so invoke statistical physics, which tells us that we
C = log(2g/(1 + )). Finally when y = 0 we find can regard a packet of identical particles as a
  macroscopic particle with mass equal to the sum
kU 2 of the microscopic masses and velocity equal to
(1 + ) 1 − = 1,
g the average velocity.

or    
g  kV 2 Let us compare the system at times t and t + δt.
U= ≈V 1− .
k 1+ 2g At time t we have a rocket mass m(t), velocity
U < V because air resistance has reduced the v(t). At time t + δt the rocket has mass m + δm,
particle’s energy. velocity v + δv and in addition there are exhaust
gases of mass −δm with (approximate) velocity
v − U. Suppose for the moment that F = 0, i.e.,
we have a closed system. Then, Theorem 1, the
total momentum is conserved.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 67 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 68
t Example 26. A rocket of initial mass m0 burns
m, v fuel at a constant rate α and lifts off vertically
under gravity. Compute the speed as a function
t + δt of time.
m + δm, v + δv
−δm, v − U Here m(t) = m0 − αt. Resolve equation (36)
At time t it is mv and at time t + δt it is vertically, setting F = −mg, and dividing by
(m + δm)(v + δv) + (−δm)(v − U + O(δt)), see m(t)
figure. Thus dv αU
= −g + .
dt m0 − αt
mv + mδv + δmv − δmv + δmU − mv ≈ 0, Since v = 0 at t = 0 a simple integration gives
 
valid for small δt, where the approximation is αt
v = −gt − U log 1 − . (37)
ignoring terms of order (δt)2. If we include the m0
external force F we obtain
Two points should be noticed.
mδv + δmU ≈ Fδt.
• The condition for liftoff is (dv/dt) > 0 at t = 0,
valid for small δt. Now divide by δt and take the i.e., U α/m0 > g. If this condition is satisfied then
limit as δt → 0. There is a first order small term (dv/dt) > 0 for 0 < t < m0/α.
which vanishes and we are left with the (exact)
rocket equation • Let the “payload” or “final mass” (no more fuel)
be βm0. Then the final speed V0 can be computed
mv̇ = F − ṁU. (36) from (37) as V0 = −gt − U log β. If β
1,
V0/U can be arbitrarily large, although gravity
eventually slows the rocket.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 69 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 70

Another approach is to take m as the independent


The opposite to a rocket,(ṁ > 0) is an avalanche
variable.
or a condensing raindrop. This is a longer example
illustrating three different ways of solving the same
Example 27. Consider a rocket, as above, problem, which involves expansions in terms of a
subject to air resistance with a force kv opposing small parameter. Quite often it is hard to decide
the motion, and for simplicity ignore gravity. which is the optimal approach, and so one needs
The rocket equation is mv̇ = αU − kv. Now to keep a store of applicable methods.
(dv/dt) = (dv/dm)(dm/dt) = −α(dv/dm) so
that after separating the variables in the rocket Example 28. [Falling raindrop] A small raindrop
equation falling through a stationary cloud acquires
(k/α)dv k dm
= . moisture by condensation from the cloud. When
(k/α)v − U αm
the mass of the raindrop is m, the rate of increase
The integration of this differential equation is of the mass per unit time is km. The raindrop
straightforward. Using the condition v = 0 at starts from rest. Neglecting resistance due to
m = m0 we find motion, find the relation between the velocity v
 k/α and the distance fallen y, and prove that if k is
kv m small enough the velocity is given approximately
1− = .
by 
αU m0  
2y
v = 2gy 1 − 3 k
2 2
.
Finally m = βm0, v = V0 where V0 = (αU/k)(1− g
β k/α), and so air resistance imposes an upper limit
on V0, a terminal velocity .
First we establish the dynamics. Obviously we
can use the rocket equation (36), resolving in the
downward vertical direction. The force F is m(t)g

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 71 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 72
and ṁ = km. Because the cloud is at rest the that if we regard v as a function of y rather than
relative velocity is U = v. Thus mÿ = mg − kmẏ t we have
or, with the obvious change in notation, dv
v = g − kv.
dy
v̇ = g − kv. This is, admittedly, non-linear, but it is of
first order and is easily solved by separation of
variables. Using v = 0 at y = 0 we obtain
 
The naı̈ve, autopilot, approach is to spot that v g kv
y = − − 2 log 1 − ,
this is a linear equation with constant coefficients. k k g
Setting y = v = 0 at t = 0 we have
which is the required relation between v and y.
g  g g 
v = 1 − e−kt , y = t − 2 1 − e−kt . Next we assume that kv/g
1 and expand the
k k k
log term as a Taylor series, finding
While the integration is easy, and the results  
are correct, it is hard to obtain what was asked v g kv 1 k 2v 2 1 k 3v 3
y=− − 2 − − − − ...
for, because we next need to eliminate t between k k g 2 g2 3 g3
these two equations. The first stage is to note that
v v v 2 kv 3
t = −(1/k) log (1 − kv/g). The details are clearly =− + + + + ...
messy and are left as an exercise for masochists. k k 2g 3g 2
 
v2 2 kv
A little more thought might suggest that since = 1+ + ...
2g 3 g
the answer does not involve t there is little point
in introducing t as the independent variable. We v2  
= 1 + O(kv/g) .
have another choice for v̇ = (dv/dy)(dy/dt), so 2g

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 73 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 74

 
Thus v 2 = 2gy 1 + O(kv/g) which implies The zero order equation vv  = g integrates

v = 2gy 1 + O(kv/g) . Inserting this back into immediately to give v 2 = 2gy (for we required v =
the series, 0 at y√= 0) and we deduce v 2 = 2gy +O(kv/g) or
v = 2gy(1 + O(kv/g)). Substituting this back
 
v2 2 k into the equation
y= 1+ 2gy + O((kv/g)2 ) ,
2g 3g
dv   
v = g − k 2gy 1/2 + O (kv/g)2 ,
dy
and this is easily manipulated into
which integrates to give
  
2
v 2 = 2gy 1 − k
2y  
+ O((kv/g)2 ) . 2   
3 g v 2 = 2 gy − k 2gy 3/2 + O (kv/g)2
3
  
Well that was long, but at least it got the answer! 2 2y  
= 2gy 1 − k 2
+ O (kv/g) ,
The third method of solution relies on the fact 3 g
that, provided all functions are smooth, i.e., they
have Taylor series expansions in the appropriate the answer!
small parameter, then successive approximation
(as above) commutes with solving ordinary The purpose of this extended example is to
differential equations, even non-linear ones! emphasize that there is often more than one way
to solve a problem. To find which method is
We return to our differential equation optimal for a given problem requires some insight
and some experience.
dv  
v = g − kv = g 1 + O(kv/g) ,
dy

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 75 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 76
3.4 Electrodynamics Exercise 11. With the same notation show that

We recall the Lorentz force described by equation (x × y)˙ = ẋ × y + x × ẏ.


(23)
 
mẍ(t) = FLorentz = q E(t, x(t))+ẋ(t)×B(t, x(t)) .
(38) Example 29. Now suppose that E = 0, and
For a first study of this equation we make the choose axes so that B = Bk where k is a unit
significant simplification that both E and B are vector in the z-direction and B = |B|. Then the
constant both in time and space. We consider first Lorentz equation (38) can be written as
the case of single fields, and then a specialization
of the general case. The case B = 0, i.e., ẍ =
ẍ = ω ẋ × k, ω = qB/m, (39)
(q/m)E is easily despatched for this is constant
acceleration. The remaining cases make extensive
use of the following result. where it is easily seen that [ω] = [T −1] so that ω
is a frequency, the gyromagnetic frequency.
Lemma 1. If vectors x and y depend on time There are many ways to solve eq.(39). A simple-
minded approach is to write x = (x, y, z), and to
(x.y)˙ = ẋ.y + x.ẏ. resolve the equation in components

ẍ = ω ẏ, ÿ = −ω ẋ, z̈ = 0. (40)

Proof. We sketch the proof in two dimensions We obtain immediately z = z0 +w0 t and ẏ = a0 −
where x.y = x1y1 + x2y2. Then (x.y)˙ = ẋ1 y1 + ωx where z0, w0 and a0 are constants. Eliminating
ẋ2y2 + x1ẏ1 + x2ẏ2. 2 ẏ from the first equation we obtain ẍ+ω 2x = ωa0.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 77 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 78

It is straightforward to solve this SHM equation the equation with k gives k.ẍ = 0 or z̈ = 0 which
finding x = a0/ω + (V0/ω) cos(ωt − α) where V0 implies z = z0 + w0t as before. Dotting the
and α are constants of integration. We now have equation with ẋ gives ẋ.ẍ = 0. Using the lemma
ẏ = a0 − ωx = −V0 cos(ωt − α) and finally, we can deduce ẋ2 = V02 + w02 a constant. This
implies ẋ2 + ẏ 2 = V02 and so we set
V0 V0
x = x0+ cos(ωt−α), y = y0−
sin(ωt−α),
ω ω ẋ = −V0 sin θ(t), ẏ = −V0 cos θ(t),
(41)
where x0 = a0/ω and y0 are constants. where θ(t) is to be determined. Now dotting
A more sophisticated (and generalizable) way (39) with i gives ẍ = ω ẏ which implies θ̇ = ω
to solve the first two equations (40) is to and θ = ωt − α where α is a constant. Another
introduce the complex variable ζ = x + iy integration now gives (41).
finding ζ̈ + iω ζ̇ = 0. We can solve this as Suppose we shift the origin to (x0, y0, z0), and for
ζ̇ = [−iV0 exp(iα)] exp(−iωt), where the term the moment set w0 = 0. Then eq.(41) describes
in square brackets is the “constant of integration” motion in a circle in the (x, y)-plane with constant
and V0 and α are real constants. This is easily angular velocity −ω.
integrated to give
j
V0
ζ = ζ0 + e−i(ωt−α) ,
ω

where ζ0 is a complex constant. Setting ζ0 = i


x0 + iy0 we recover (41).
Another powerful approach is to treat eq.(39) as a
vector equation, setting x = xi+yj+zk. Dotting

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 79 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 80
If we allow z = w0t the circle becomes a helix and a suitable particular integral is ζ = (λ/ω)t.
with axis parallel to k, The constant C corresponds to a fixed change or
origin, and so we discard it. Setting ζ0 = R0eiα
V0 V0 (R0 = V0/ω) we have
x= cos(ωt − α)i − sin(ωt − α)j + w0tk,
ω ω
λ
which is a helix with axis in the 3-direction. x= t + R0 cos(ωt − α), y = −R0 sin(ωt − α),
ω

Example 30. Now we assume that both E and so that the helix is sheared into the x-direction.
B are nonzero, but we restrict attention to the, Notice that λ/ω = (qE/m)/(m/(qB)) = E/B
physically important, case E.B = 0. Making use depends only on the fields and is the same for
of the results from the previous example we set all particles with q = 0. The velocity vL =
x = (x, y, z), E = (0, E, 0) and B = (0, 0, B), (E/B, 0, 0) is the Larmor drift velocity
as well as ω = qB/m, and λ = qE/m. Now the
Lorentz equation (39) implies E×B
vL = .
|B|2
ẍ = ω ẏ, ÿ = λ − ω ẋ, z̈ = 0,
The easiest way to visualize the motion is to
and so z = z0 + w0t as before. In order to solve choose an inertial frame so that x(0) = ẋ(0) = 0.
the other two equations we set ζ = x + iy, finding then it is easy to verify that, setting R = E/(Bω),

ζ̈ = iλ − iω ζ̇. x = R(ωt − sin ωt, 1 − cos ωt, 0),

Here the complementary function is ζ = C + and


ζ0e−iωt, where C and ζ0 are complex constants, ẋ = Rω(1 − cos ωt, sin ωt, 0).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 81 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 82

Note that 4 Energy and Stability

(x − Rωt)2 + (y − R)2 = R2.


4.1 One degree of freedom

Consider a bicycle wheel radius R in the z = 0


plane rolling without slipping along the x-axis with A system is said to possess one degree of
angular velocity ω. Then the centre has velocity freedom if it can be described by a scalar
(Rω, 0, 0). The above equation implies that the equation of motion of the form
particle behaves like a point on the rim and so
traces out a cycloid. mẍ = F (x). (42)

The potential energy is


x
U (x) = − F (ξ) dξ, (43)

defined up to an arbitrary additive constant. Then


mẍ = −dU/dx which we have seen before. The
kinetic energy is T = 12 mẋ2 and the total
energy is E(x, ẋ) = T + U . The following result
is fundamental.

Theorem 2. For such a system E is a constant.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 83 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 84
Proof. 4.2 Conservative systems
dE d 1  Suppose we consider systems described by the
= mẋ2 + U (x)
dt dt 2 vector equation of motion
= mẋẍ + U (x)ẋ
mẍ = F(x). (44)
= ẋ(mẍ − F )
=0 by (42). The system is said to be conservative if ∃ U (x)
such that F = −∇U . Then the potential
2 energy is U (x) and the kinetic energy is
T = 12 mẋ.ẋ.
Note that the more general case F = F (x, ẋ)
is excluded. E.g., mẍ = −mg − k ẋ. Setting Theorem 3. For a conservative system the
U = mgx we have (d/dt)(T + U ) = −k ẋ2 < 0. total energy E = T + U is conserved.

Surprisingly the general case with more degrees


Proof.
of freedom has not yet been solved. There is
however an important special case where we can dE
make progress. = mẋ.ẍ + (∇U ).ẋ
dt
= ẋ.(mẍ − F) = 0.

Note that the potential and total energies are not


defined for non-conservative systems.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 85 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 86

P
4.3 When is a system conservative? In one dimension W = O F dx = −(U (P ) −
F U (O)) but how do we calculate (45) in higher
dimensions? Suppose the path is specified by
P x = x(s) where s = 0 at O and s = 1 at P .
x Then dx = (dx/ds) ds = x ds and
O
1
Suppose we have a constant force F whose point 
W = F(x(s)).x(s) ds, (46)
of application is displaced through x from O to 0
P . Then the work done is W = F.x. More
generally we want to consider varying forces and where the integrand is a scalar function of s and
curved paths. We split the path into a large the integral is computable by elementary methods.
number, N , of tiny segments δx(i) and let the
force, restricted to the ith segment be F(i).

F(i) P

δx(i)
O

Define the work function by

P
W = lim F(i).δx(i) → F.dx. (45)
N →∞ O
i

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 87 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 88
Example 31. [Path independence] In two dimensions Exercise 12. [Path non-independence] Consider
let O be (0, 0) and P be (1, 1) and consider the the same setup but with F = (y, −x). Show
three paths shown. (b is given by y = x2.) Wa = 0, Wb = − 13 and Wc = −1, so that the
integral depends on the path taken.
P
In the case of example 31, any path from O to
P will give the same answer. For fixed O, the
a b c
answer depends only on P and not on the path
taken. Thus setting P at x we can define
O M x

Suppose F = (y, x). U (x) = − F(ξ). dξ. (47)


0

For path a, x(s) = (s, s), x = (1, 1) and Wa =


1 1 Then if Q is at x + δx,
0
(s, s).(1, 1) ds = 0 2s ds = 1.

For path
 1 b, x(s) = (s, s2),
 1 x 2 = (1, 2s) and
x+δ x
2 δU = U (x + δx) − U (x) = − F(ξ). dξ
Wb = 0 (s , s).(1, 2s) ds = 0 3s ds = 1. x

For the path OM x = (s, 0), x = (1, 0) ≈ −F(x).δx


1
and Wci = 0 (0, s).(1, 0) ds = 0. For the ≈ −F1δx − F2δy,

path
1 M P x = (1, s), x = (0, 1) and Wcii =
0
(s, 1).(0, 1) ds = 1, and so Wc = 1. where we have set F = (F1, F2). It follows that
The work integral appears to be path-independent. F = (−∂U/∂x, −∂U/∂y) = −∇U and we have a
conservative system. (For example 31 U = −xy.)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 89 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 90

In exercise 12 none of this goes through and there A similar argument applies for the second (and
is no potential energy. indeed any further) dimension. 2
So what’s the difference? Most textbooks refer
you to other books at this point, but we will try
to be a little more complete. First a notational
point. In the next lemma and theorem δx, δy etc
will be small but variable quantities and we shall
frequently take limits as they tend to zero. A
quantity which has magnitude |(δx)m(δy)n| will
be denoted by O(m+n)

Lemma 2. [Taylor Series] Suppose f : R2 →


R is a smooth function. Then

∂f ∂f
f (x+δx, y+δy) = f (x, y)+ δx+ δy+O(2),
∂x ∂y
(48)
where the partial derivatives are computed at the
point (x, y).

Proof. Suppose first that δy = 0. Then (48) is


equivalent to the assertion

∂f f (x + δx, y) − f (x, y)
= lim .
∂x δx→0 δx
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 91 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 92
2 ⇒ 1: Regard A as fixed and B as a variable
Theorem 4. The following assertions are point. Define a function U via U (B) = U (A) -
B
equivalent: A
F.dx which is independent of the path chosen.
Then
1. F = −∇U for some U (x),
δU = U (x + δx) − U (x)
B x+δx
2. F.dx is path-independent,
A =− F.dx = −F(x).δx + O(2).
 x
3. C
F.dx = 0 for all closed curves C,
Choosing δx = (δx, 0, 0) gives δU = −F1δx
 which implies F1 = −∂U/∂x. A similar result
4. F.dx = 0 for all closed infinitesimal rectangles,
holds for the other components.
5. ∇ × F = 0. B
C2
Here
C1
 ∂F ∂F2 ∂F1 ∂F3 ∂F2 ∂F1 
3
∇×F = − , − , − . A
∂y ∂z ∂z ∂x ∂x ∂y
2 ⇒ 3: Suppose  B that A and B lie on a closed
curve C. Then A F.dx is the same whether we
Proof. use path C1 or C2. Thus
B B B   
1 ⇒ 2: A F.dx = − A ∇U.dx = − A dU =
− F.dx = 0 = F.dx.
U (A) − U (B) independent of the path chosen. C1 C2 C

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 93 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 94

3 ⇒ 2: Choose any two paths C1 and


 C2 from
A to B, and let C = C1 − C2. Then C1 F.dx =

C2
F.dx.
Consider a small rectangular path from (x, y) to
3 ⇒ 4: This step is obvious. (x + δx, y + δy) and back again, and integrate
F around it. The contribution from the vertical
4 ⇒ 3: sides is
δy
 
F2(x + δx, y + s) − F2(x, y + s) ds
0

δy
∂F2 ∂F2 ∂F2
= F2(x, y) + δx + s − F2(x, y) − s ds
0 ∂x ∂y ∂y
∂F2
= δxδy + O(3),
Regard a smooth closed curve C and its interior as ∂x
being composed of a large number of infinitesimal where the Taylor lemma 2 was used. Including the
rectangles Rn with sides
parallel
 to the coordinate contribution from the horizontal sides we have
axes. By condition 4, n Rn F.dx = 0. However  ∂F ∂F1
all the interior sides are traversedtwice in opposite 2
F.dx = − δxδy + O(3).
senses and so cancel out. Thus C F.dx = 0. ∂x ∂y

1 ⇒ 5: This step is an (almost) obvious identity. Thus if ∇ × F = 0 the integral vanishes. 2

5 ⇒ 4: Thus a system with many degrees of freedom is

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 95 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 96
conservative iff ∇ × F = 0. 4.4 Qualitative dynamics with one
degree of freedom

A system with one degree of freedom is necessarily


conservative, theorem 1, and the energy equation
1 2
2 mẋ + U (x) = E constant, can be rewritten as

ẋ2 = f (x) ≡ 2(E − U )/m. (49)

Differentiating with respect to time, and dividing


by 2ẋ gives
1
ẍ = f (x). (50)
2
This has only been derived for points where ẋ = 0.
In general zeros of f will be isolated and so (50)
holds, by continuity, at these points too.
We start with the obvious remark that the motion
is possible only for those points where f (x)  0,
and if f (x) > 0 two motions, right-moving (ẋ >
0) and left-moving (ẋ < 0) are possible.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 97 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 98

acceleration there and so the velocity becomes


f
positive.√ Thus for t  t0 √we have to integrate
ẋ = +c x, with solution x = 12 c(t − t0). We
can combine both cases as x2 = 14 c2(t − t0)2.
Although this behaviour is exact only for linear
f the general behaviour is qualitatively correct.
0 x The left-moving system reaches rest at x = 0
in a finite time (assuming f (0) > 0) and then
The figure shows a typical configuration. Motion becomes a right-mover.
is only possible for x  0 and the tangent to the
graph of f at x = 0 is shown. If for x > 0 If we reflect the figure in the y-axis this qualitative
the motion is right-moving it remains so, and behaviour is still correct provided we interchange
since f (x) > 0 it accelerates. On the other “left” and “right”. Further there is nothing special
hand if it is left-moving it approaches x = 0 about x = 0; any zero of f would do.
decelerating as it does so. In order to discuss what
happens as x → 0+ we approximate f locally by
its tangent f (x) ≈ c2x where c2 = f (0) >√ 0.
Thus ẋ2 = c2x. Initially ẋ < 0 and so ẋ = −c x.
It is easy to integrate this equation, finding

√ 1
x = c(t0 − t),
2

where t0 is a constant. The interpretation is that


as t → t0− the particle comes to rest at the
origin. Because of equation (50) it has positive

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 99 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 100
4.5 The stability of systems with one
Example 32. For motion in the earth’s degree of freedom
gravitational field, U = −mgr02/r, example 5,
We consider systems described by equations (49)
and so, with a slight change of notation,
and (50). The system is said to have an
equilibrium point at x = x0 if x(t) = x0,
2 mgr02  constant, is a possible motion of the system—in
ṙ 2 = f (r) = E+ .
m r other words if x does not change. Now suppose
that the system has an equilibrium point at x0 and
Suppose a rocket is fired vertically upwards with the system is given a “small perturbation”, i.e.,
speed V0 . If E < 0 then it is easy to see that the functional form of f (x) is changed slightly.
f has a single zero at a finite value rmax of r. If the system remains in a small neighbourhood
Thus the rocket will ascend to r = rmax and then of x = x0 then the equilibrium is said to be
return to earth. However if E > 0, f (r) > 0 for stable, otherwise it is unstable. (Note that the
all r, and so the particle can escape the earth’s perturbed system may have no equilibrium points
gravitational field. The critical case√ is where at all. Stability is the requirement that the motion
E = 0 which corresponds to V0 = 2gr0, the is confined to a small neighbourhood.)
so-called escape velocity. (V0 ≈ 11.2km s−1.)
Example 33. Empirical experience shows that
Of course we do not need all this abstract a simple pendulum consisting of a point mass
machinery to treat simple situations like example attached to a fixed point via a weightless rigid rod
32. Its real purpose is to reduce the study of has two equilibria. That in which the pendulum
stability of systems with one degree of freedom to hangs vertically downwards is stable, while that
a sequence of simple “A/S level” problems. where the point mass is exactly above the pivot
is unstable. (We shall see soon how to treat this
example quantitatively.)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 101 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 102

at x = x0 and so we have an equilibrium. This


Theorem 5. [The stability algorithm] The set proves the first sentence of the theorem.
of equilibria of systems governed by equations Suppose that x = x0 is an equilibrium point, and
(49) and (50) is the set of critical points consider a motion x(t) = x0 + (t) where  is
x0 of f (f (x0) = 0) with critical values initially small. Then, using a Taylor expansion,
zero, i.e., f (x0) = 0. Stable critical points
correspond to maxima, unstable ones to minima. 1
(The case of an inflection point needs special ẍ = ¨ = f (x0 + )
2
treatment.) Perturbed motion about a stable 1 1
critical point is = f (x0) + f (x0) + O(2)
 simple harmonic motion with 2 2
frequency ω = − 12 f (x0). 1 
= f (x0) + O(2),
2
Thus the investigation of stability reduces to the
determination of maxima and minima of functions where Taylor’s theorem was used in the second
f : R → R. line. If f (x0) > 0 then (t) grows exponentially
fast, and the solution passes outside any small
Proof. If x = x0 is an equilibrium point, then x(t) neighbourhood of x = x0. Conversely if f (x0) <
and all derivatives thereof must vanish at x = x0. 0 we have ¨(t) + ω 2(t) = O(2) (with ω defined
In particular we must have f (x0) = f (x0) = 0 in the statement of the theorem) which proves the
from equations (49) and (50). Suppose conversely second sentence. 2
that f (x0) = f (x0) = 0, so that ẋ and ẍ both
vanish when x = x0 . Differentiating (50) gives
d3x/dt3 = 12 f (x)ẋ, and so the third derivative
of x vanishes at x = x0 . By a simple induction
argument we see that all derivatives of x(t) vanish

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 103 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 104
It is instructive to examine the two cases using Example 34. [Simple stability] A bead P of
the qualitative methods of the previous section. mass m can slide on a fixed smooth wire shaped
as a circle of radius a, centre O. The circle lies
stable in a vertical plane. Determine the equilibria and
their stability. Let θ be the angle between OP
a b x and the downward vertical.
unstable

In the unstable case a minimum (with value 0) is


shown as a solid curve, together with two possible O
perturbations (dotted curves). In both cases the θ
P
motion moves out of a small neighbourhood of the
critical point. In the stable case one perturbation
is impossible for f (x) < 0. For the other f (x) > 0
only for x ∈ (a, b) and so we have bounded Clearly T = 12 ma2θ̇ 2, U = −mga cos θ + const.
motion. and we set E = T + U After some simple algebra

2 g
θ̇ 2 = f (θ) ≡ 2
E + 2 cos θ,
ma a

and we can deduce that f (θ) = −2(g/a) sin θ


and f (θ) = −2(g/a) cos θ. Equilibrium requires

f (θ) = 0 = f (θ).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 105 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 106

The second of these fixes θ = 0 or θ = π. The and absorbing all energy constants into E we have
first is boring—it merely fixes E. However E is
only defined up to the arbitrary constant which 2E αg
ẏ 2 = f (y) ≡ + 2gy − (y −
0)2.
occurs in U , and so its value is usually of no m
0
physical interest.
This is the equation of an upturned parabola.
Next note that f (π) > 0, and so θ = π is
There is a single maximum when
0 = α(y −
unstable.

0), i.e., y = y0 = (1 + α−1)
0, and so for an
However f (0) = −2g/a < 0 and so θ = 0 is appropriate (but boring) choice of E we have a
stable,
 and the frequency for small oscillations is stable equilibrium point. The frequency of small
ω = g/a. oscillations is ω where ω 2 = − 12 f (y0) = αg/
0.

Example 35. A massless spring of natural


length
0, modulus λ = αmg has the top
end fixed, while the bottom end carries a
mass m. Show that there is a single, stable,
equilibrium position and find the frequency of
small oscillations about it.

We measure y downwards from the unstretched


position. Clearly T = 12 mẏ 2 , and there is a
gravitational contribution to the potential energy
Ug = −mgy + const. Now the tension in the
spring is −λ(y −
0 )/
0 = −dUel/dy and so Uel =
2 λ(y −
0 ) /
0 + const. Setting E = T + Ug + Uel
1 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 107 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 108
modulus of elasticity of the string. The energy
Example 36. A light elastic loop of unstretched equation can be reduced to
length 4a passes over two small pegs at the same
level, distance 2a apart. A particle is attached 2E 2gy λ  2
ẏ 2 = f (y) ≡ 2
+ − 1 + y2 − 1 .
to the loop, and in equilibrium the loop forms an ma a ma
equilateral triangle. Show  that the frequency of
7g It is elementary to compute
vertical oscillations is ω = 4√ 3a
.
2g 2λ   y
Let the particle have mass m and let its depth f (y) = − 1 + y2 − 1  ,
a ma 1 + y2
below the pegs be ay where y is dimensionless.
 
a a  2λ 1 y2
f (y) = − 1− + .
ma (1 + y 2)1/2 (1 + y 2 )3/2
ay √
At √equilibrium we know that y = √3 and so
f ( √3) = 0, which√implies λ = 2mg/ 3. Then
f ( 3) = −7g/(2 3a) and so we have stable
m equilibrium with the frequency of small oscillations
as advertised.
The kinetic energy is T = 12 ma2ẏ 2 and
the gravitational potential energy is (ignoring
inessential constants) Ug = −mgay.  The
stretched length of the loop is 2a + 2 a + a y 2 .
2 2

Since the unstretched length is


0 =4a the elastic
energy is Uel = 12 λ(4a)−1 (2a + 2 a2 + a2y 2 −
4a)2 = 2 λa( 1 + y 2 − 1)2, where λ is the
1

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 109 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 110

or λs(x − xs) = 0 or x = xG = Λ−1 λsxs,


Example 37. [Planar stability] A particle P the x-coordinate of the centre of mass. A similar
moves in a plane under the influence of n fixed result holds for the y-direction and so the centroid
centres of attraction A1, A2,. . . , An in the plane. G is the unique equilibrium point.
The magnitude of the attractive force to As is
mλsrs, where λs is a positive constant and rs is
(x, y) = (xG + (t), yG + δ(t)). Then
Now write
the distance P As. If G is the centre of gravity of using λs(xG − xs) = 0 etc., it is easy to see
masses λ1, λ2, . . . , λn at A1, A2,. . . , An, prove that
that G is the unique position of stable equilibrium, ˙2 + δ̇ 2 = 2E /m − Λ(2 + δ 2).

and that the frequency ω
of small oscillations This is the equation for SHM with frequency Λ
about it satisfies ω 2 = Λ ≡ λs. in each direction.

Let the position of the particle be (x, y). Its


kinetic energy is 12 m(ẋ2 + ẏ 2) and, by analogy with
elasticity, its potential

energy is (up to an additive
constant) U = 2 m λs[(x − xs) + (y − ys)2]. If
1 2

E denotes the total energy, then simple algebra


gives

ẋ2 + ẏ 2 = f (x, y) ≡

2E/m − λs[(x − xs)2 + (y − ys)2].
s

For an equilibrium position, the condition of zero


acceleration in the x-direction implies ∂f /∂x = 0,

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 111 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 112
5 Impulses and collisions 5.2 Binary collisions in one dimension

We examine first the collision in one dimension,


5.1 Impulses where m2, u2 and v2 are the corresponding
quantities for ball 2. The force that ball 2 exerts
When two billiard balls collide at t = t0 they on ball 1 is F1←2 = I1←2δ(t − t0), and F2←1 is
almost instantaneously exchange momentum. We similarly defined. However by Newton’s third law
model this by saying that ball 1 suffers a force F2←1 = −F1←2 and so I2←1 = −I1←2. Clearly
F1 = I1δ(t − t0) or an impulse I1 at time t0.
If ball 1 has momentum m1u1 before and m1v1 m1v1 −m1u1 = I1←2, m2v2 −m2u2 = I2←1.
after the impulse then
Since we do not have, as yet, any further
m1v1 − m1u1 = I1. (51) information on the impulses we eliminate them
finding
Note that we do not try to analyze the detailed
physics of the actual collision. Equation (51) m1v1 + m2v2 = m1u1 + m2u2, (52)
is merely an empirical summary of the dynamics
which we henceforth treat as an axiom.
expressing conservation of momentum during the
course of the impulse. Note that this accords with
the discussion of closed systems in section 1.8.
In order to determine v1 and v2 in terms
of u1 and u2 we need a further (vector)
equation. Newton observed empirically that for 1-
dimensional collisions the relative velocities after

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 113 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 114

and before impact were linearly related. More and simple algebra on equations (54) and (55)
precisely shows
m2 m1
v2 − v1 = −e(u2 − u1), (53) v̂1 = e (û2 − û1), v̂2 = −e
(û2 − û1),
M M
(56)
where e is the coefficient of restitution. It giving v̂1 and v̂2, and hence v1 and v2.
depends on the physics of the collision, and Let us look at the energies involved. Initially
(usually) e  0. This provides the required second
relation. 1 1
Einit = m1(U + û1)2 + m2(U + û2)2
2 2
Actually ui and vi are not the most useful 1 1
variables. We define the total mass M = = M |U|2 + (m1|û1|2 + m2|û2|2)+
2 2
m1 +m2 and the centre of mass X = (m1x1 +
U.(m1û1 + m2û2),
m2x2)/M (where xi are the positions of the balls).
Now equation (52) asserts that M Ẋ is continuous
but the last term vanishes by virtue of (54).
at collisions, M V = M U, in a (hopefully) obvious
Using this property again we have
notation. Next set u1 = U + û1 etc., so that
1
Einit = M |U|2+
m1v̂1 + m2v̂2 = 0 = m1û1 + m2û2. (54) 2
1  
m1(m1 + m2)|û1|2 + m2(m1 + m2)|û2|2 −
2M
1
In this notation Newton’s law becomes |m1û1 + m2û2|2
2M
1 1 m1m2
v̂2 − v̂1 = −e(û2 − û1), (55) = M |U|2 + |û1 − û2|2,
2 2 M
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 115 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 116
after some algebra. A similar expression obtains calculated. We are making extensive use of (52)
for the final energy and (53). If v and w are the velocities of B and
C after the first impact then U = v + 7w = w − v
1 1 m1m2
Ef in = M |V|2 + |v̂1 − v̂2|2, giving
2 2 M
and using V = U and Newton’s law (55) we find v = − 34 U, w = 14 U.

1 m1m2 2
ΔE ≡ Ef in − Einit = (e − 1)|û1 − û2|2. If u and v are now the velocities of A and B after
2 M
(57) the second impact then − 34 U = 8u + v = u − v
We expect to lose energy in a collision, e.g., sound, giving
internal heating etc., and we deduce e2  1. The
case e = 1 ⇔ ΔE = 0 is said to be perfectly u = − 16 U, v= 7
12 U.
elastic.

Example 38. Three small spheres, A, B and C, If v and w are now the velocities of B and C after
whose masses are 8m, m and 7m are at rest in the third impact then
line, and AB = BC = a. The middle sphere B is
projected towards C with velocity U . Assuming
all of the spheres to be perfectly elastic, discuss
v + 7w = 7
12 U + 74 U, w−v = 7
12 U − 14 U,
the subsequent motion, and illustrate this by a
diagram showing the positions of the spheres at
any time after the start.
giving v = 0, w = 13 U , and there are no further
We first determine the velocities after the various collisions. The velocities in the four stages of
collisions; times and positions can then be motion are shown below.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 117 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 118

U 5.2 Newton’s law in two and three


A B C
dimensions
3 1
4U 4U
When two bodies collide there will be a line/plane
A B C of contact Σ with unit normal n.
1 7 1
6U 12 U 4U

A B C n
1 1
6U 3U

A B C
Suppose the initial positions to be (−a, 0, a). The Σ
time to the first collision is t = a/U after which
the positions are (−a, a, a).
The one-dimensional law (53) holds for the
The time to the second collision is t = = 2a/( 34 U ) velocity components parallel to n while the
8
3 a/U , and it is straightforward to calculate the components of v1 − u1 and v2 − u2 perpendicular
new positions to be (−a, −a, 53 a). to n (i.e., tangent to Σ) are continuous. Formally
You should now verify that the third collision
n.(v2 − v1) = −en.(u2 − u1),
takes place after a further time 8a/U at positions (58)
(− 73 a, 11 11
3 a, 3 a).
n × (v1 − u1) = 0 = n × (v2 − u2),

and there are of course “hatted” versions.


The only point in the previous section where we
really used the assumption of one dimension was

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 119 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 120
Newton’s law (53). In particular the derivations
of Einit and Ef in are unchanged. (You should
Example 39. [Bouncing ball] A ball is thrown
verify this.) Thus
into the air with initial speed V0 at an angle α0.
1 m1m2  
ΔE = |v1 − v2|2 − |u1 − u2|2 .
2 M

Now for any vector y,


V0 V0 V1
|y|2 = (n.y)2 + |n × y|2,
α0 α0 α1
A0 A1 A2
by Pythagoras’ theorem. Thus, using (58),
Let the ground be object 2 so that u2 = v2 = 0,
1 m1m2 2  2 and drop suffix 1 from the ball. At A1 the initial
ΔE = (e − 1) n.(u1 − u2) ,
2 M and final velocities of the ball are

with a similar interpretation to (57). u = (V0 cos α0, −V0 sin α0), v = (V1 cos α1, V1 sin α1).

Setting n = (0, 1) equations (58) imply

V1 sin α1 = eV0 sin α0, V1 cos α1 = V0 cos α0,

and so tan α1 = e tan α0. Now assume 0 < e < 1


and let the times and horizontal distances for each

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 121 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 122

bounce be t0, t1, . . . and d0, d1, . . .. By elementary 6 Rotation in two dimensions
mechanics

2 2 6.1 Angular velocity and momentum


t0 = V0 sin α0, d0 = (V0 sin α0)(V0 cos α0)
g g
Up to now we have been using inertial frames
and it is easy to see that t1 = et0, d1 = ed0 so with constant orthonormal bases, eg., (i, j) which
that tn = ent0 and dn = end0. Thus the ball we now write as (e1, e2). For many purposes
bounces an infinite number of times in a duration it is more convenient to use orthonormal bases
of t0/(1 − e) covering a distance d0/(1 − e). (Of which are in motion relative to our standard
course air resistance, linear friction, has been ones. Of course an equation like mẍ = F is
neglected!) basis-independent, but when we resolve it into
components with respect to these new bases we
may expect some unfamiliar terms to arise.
Of particular interest is the orthonormal basis
(er , eθ ) related to polar coordinates (r, θ).

eθ er

θ x

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 123 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 124
Note the basic relation
The general position vector is x(t) = (x, y) =
(r cos θ, r sin θ) in cartesians. This means
er = cos θe1 +sin θe2, eθ = − sin θe1 +cos θe2.
(59)
Now consider two frames, a fixed inertial one x = xe1 + ye2 = r cos θ e1 + r sin θ e2 = r er
based on (e1, e2) and a non-inertial one based
on (er , eθ ) which have a common origin. In using (59). Thus in polar coordinates x = (r, 0).
general θ = θ(t) and θ̇ is the angular velocity.
For velocity we have
Although e1 and e2 are fixed, eθ is changing in
time, precisely because θ is changing. Thus we
ẋ = ẋ e1 + ẏ e2 = (ẋ, ẏ) in cartesians,
have
= (r er )˙ = ṙ er + r θ̇ eθ (61)
ėθ = −θ̇ cos θe1 − θ̇ sin θe2 = (ṙ, r θ̇) in polars.
= −θ̇ er ,
For acceleration
and similarly ėr = θ̇ eθ , i.e.,
ẍ = ẍ e1 + ÿ e2 = (ẍ, ÿ) in cartesians,
ėr = θ̇ eθ , ėθ = −θ̇ er . (60) = (ṙ er + r θ̇ eθ )˙
= r̈ er + ṙ θ̇ eθ + ṙθ̇ eθ + r θ̈eθ − r θ̇ 2 er (62)
The concepts of position, velocity and acceleration
are fundamental to dynamics. What do the = (r̈ − r θ̇ 2, r θ̈ + 2ṙ θ̇) in polars,
vectors look like in the two frames? −1
= (r̈ − r θ̇ , r2 2
(r θ̇)˙).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 125 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 126

6.2 Central force problems


For the trajectory specified by x(t), θ̇ is also
called the angular velocity. Recall that for
For the moment we are working in three
a particle of mass m the linear momentum is
dimensions. A central force problem (CFP)
p = mẋ. The angular momentum (with
is one of the type
respect to the origin) of the trajectory is
a three-dimensional concept defined by
ẍ = −f (r)r −1x, (64)
h=x×p 
where r = x2 + y 2 + z 2 = |x|. These are both
= m(xẏ − y ẋ) e3 in cartesians,
(63) physically important and relatively simple.
= r er × m(ṙ er + r θ̇ eθ )
= mr 2θ̇ e3 in polars. Theorem 6. All CFPs are planar, i.e., two-
dimensional.

Proof. [Quite often one is required to show that


a trajectory x(t) lies in a plane. The technique
is to guess a constant vector k normal to the
plane, and to define ψ(t) = k.x(t). Using the
equation of motion for the trajectory one tries to
find an ordinary differential equation for ψ(t), and
to show ψ(t) ≡ 0. This means that the trajectory
lies in the plane perpendicular to k.]
Suppose that at some initial time t0, x(t0) = x0
and ẋ(t0) = v0. Define k = x0 × v0. If k = 0

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 127 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 128
then x0 and v0 are parallel. The system (64) and
the initial conditions are invariant under rotations Theorem 7. CFPs have conservative force
about the line joining the origin to x0 and so the fields.
motion remains on this line, which is definitely
planar. If k = 0 define ψ(t) = k.x(t). Then (64)
Proof. Because of the previous theorem we can
implies
work in a plane with cartesian coordinates (x, y).
ψ̈(t) = −f (r)r −1ψ(t),
Recall that r 2 = x2 +y 2 implies r dr = x dx+y dy.
and by construction ψ(t0 ) = 0 = ψ̇(t0). Now two Suppose the particle has mass m. Then from (64)
initial conditions of this type are sufficient to fix the work integral is
uniquely the solution of the ordinary differential
equation for ψ. By inspection ψ ≡ 0 satisfies the
equation and the initial conditions. We deduce F.dx = m f r −1 (x, y).(dx, dy)
that this is the solution. Thus the orbit lies in the
plane perpendicular to k, which we choose to be =m f r −1(x dx + y dy)
in the e3 direction.
If you are unhappy with this argument you =m f (r)r −1r dr by the remark above
may prefer the following one. The displayed
equation implies ψ̈(t0) = 0. Now differentiate =m f (r) dr = −U (r) say.
it, remembering that r depends on t. The right
hand side is the sum of terms each containing
ψ or ψ̇ as a factor. Thus the third derivative 2
of ψ vanishes at t = t0. By a simple induction
argument one can conclude that ψ(t) and all of
its derivatives vanish at t = t0. Thus ψ(t) has a
trivial Taylor series. 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 129 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 130

which is of the form (42). 2


Theorem 8. CFPs have one degree of freedom.

Proof. Using equation (62) we can resolve the


components of (64) with respect to our polar
basis

r̈ − r θ̇ 2 = −f (r), r −1(r 2θ̇)˙ = 0. (65)

The second equation implies that

r 2θ̇ = h, a constant. (66)

This equation, plus (63) imply that the angular


momentum h is constant (hence the notation used
above). It is instructive to verify this directly from
(63):

ḣ = m(x × ẋ)˙
= m(ẋ × ẋ + x × ẍ)
= mx × (−f r −1)x = 0 using (64).

Finally the first equation (65) and (66) imply

r̈ = −f (r) + h2r −3, (67)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 131 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 132
6.3 Ellipses and hyperbolae
In this section we review the equations of
ellipses and hyperbolae. These curves depend
on two real parameters a > 0 and e  0. In
cartesian coordinates the focus is (ae, 0), and
the directrix is the line x = a/e. In the figure
we assume e < 1.

y
P
D
F x
ae a a/e

The curves are the loci of points P such that


P F = e × P D.
Setting P = (x, y) we obtain (x − ae)2 + y 2 =
e2(a/e − x)2 or x2(1 − e2) + y 2 = a2(1 − e2).
Assuming e < 1 set b2 = a2(1 − e2). Then the
curve, an ellipse, has the cartesian equation

x2 y 2
+ = 1.
a2 b2
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 133 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 134

y
b

a x

Another useful parametric form for the ellipse


equation in cartesian coordinates is

(x, y) = (a cos ψ, b sin ψ). (68)

We shall also need the form of the ellipse in polar


coordinates. Note that the origin of polar
coordinates is chosen to be F , r = F P and
F P subtends an angle θ with the x-axis. Then
the defining equation implies

r = e(a/e − ae − r cos θ),

or
a(1 − e2)
r= (69)
1 + e cos θ
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 135 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 136
6.4 Kepler’s laws of planetary motion
Exercise 13. [Hyperbolae] Verify all of the
statements (and figures) for ellipses. Construct7 After many years of patient observation of the
an equivalent theory for hyperbolae, where e > 1. orbits of the planets, Kepler announced three
What about parabolae, where e = 1? laws. The motion of an individual planet around
a central Sun is a CFP, and these laws were
discovered before, and motivated, Newton’s three
laws and also his law of gravitation.
If we consider the motion of planets around the
Sun there is an important simplification. The
mass of the Sun is M
≈ 2 × 1030kg, while the
mass of the earth is m ≈ 6 × 1024kg. Since action
and reaction forces are equal and opposite, the
acceleration of the Sun due to the earth is about
one millionth of the acceleration of the earth due
to the Sun. Thus we may regard the Sun as fixed
at the (polar coordinate) origin, around which the
earth orbits.

7
The details can be found in appendix one of Lunn

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 137 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 138

gravitational field of the sun is a central force


Example 40. [Kepler’s second law] A straight problem.
line joining the Sun and a planet sweeps out equal
areas in equal times.

e
t + δt

r + δr

S δθ t
r

The area of the triangle is δA = 12 r(r +


δr) sin δθ = 12 r 2δθ to first order. Therefore

1 1
Ȧ = r 2θ̇ = h,
2 2

from (66). Thus Kepler’s second law is an


observational confirmation of the conservation of
angular momentum of solitary planets orbiting the
sun. In fact Newton used it as evidence that the

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 139 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 140
It is more instructive to return to equation (70)
We now want to apply the CFP concept and to look for a solution r(θ) rather than
to planetary motion. The central force is r(t), θ(t). In fact it pays to study u(θ) where
gravitational acceleration and so we set f (r) = u = 1/r. Note that (66) implies θ̇ = hu2 and so
GM/r 2 in equation (64), where M is the mass of
the Sun. Then (67) becomes ṙ = (u−1)˙ = −u−2(du/dθ)θ̇ = −u−2uhu2 = −hu,

GM h2 where u ≡ du/dθ. Further r̈ = −huθ̇ =


r̈ = − + 3. (70) −h2u2u. Now equation (70) reduces to
r2 r
GM
The standard way to handle this is to multiply u + u = , (73)
first by 2ṙ and then to integrate with respect to h2
time obtaining the energy equation whose general solution is

GM 1 h2 u(θ) = GM/h2 + β cos(θ − θ0), (74)


ṙ 2 = 2(E − V ) where V (r) = −
+ 2,
r 2r
(71) where β and θ0 are constants, and without loss
and E is a constant. Formally we can solve this of generality θ0 = 0. The energy equation (71)
by becomes
dr
t=  , (72)
2(E − V (r)) h2(u + u2) = 2E + 2GM u,
2
(75)
but not in terms of simple functions.
and inserting (74) in this gives, after a little
algebra,

β 2 = (2E + G2M 2/h2)/h2. (76)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 141 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 142

Now recall the energy equation (71) must have one positive and one negative zero, as
required.
 
GM 1 h2 Now set p = h2/(GM ) and e = pβ =
ṙ 2 = 2 E + − 2 . 
r 2r 1 + 2Eh2/(GM )2 so that (74) becomes pu =
1 + e cos θ or
The right hand side is negative for small r and for
p
large r it approaches E. We need real solutions, r= . (77)
i.e., a positive right hand side for some range of 1 + e cos θ
r. This requires that the quadratic expression
Comparing this with equation (69) we see that
we have an ellipse if e < 1 ⇔ E < 0. (The
Q = 2Er 2 + 2GM r − h2 case E > 0 leads to unbound hyperbolic orbits.)
Thus we have verified Kepler’s first law: the
has two positive zeros if E < 0 and one if E > 0. path of each planet is an ellipse with the
The condition for real zeros is sun at its focus. Newton used the observed
form (77) to compute the f (u) of (64) and thus
(GM )2 + 2Eh2 > 0, to verify that the gravitational force obeyed an
inverse square law.
which guarantees that β 2 defined by (76) is We next investigate Kepler’s third law:
positive, i.e., β is real. Notice, from the expression the square of each planet’s period is
for Q above, that if E < 0 the sum of the zeros proportional to the cube of the semi-
of Q is positive as is the product. Thus if the major axis of the elliptic orbit. The semi-
√ a = p/(1 − e )
2
roots are real they are both positive, as required. axes are, from
√ (69) and (77)
Conversely if E > 0 the sum and product of and b ≡ a 1 − e = p/ 1 − e .
2 2 We can
the zeros are both negative. For real zeros we parametrize the equation of the ellipse by (68),

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 143 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 144
i.e., x = a cos ψ, y = b sin ψ The area is 6.5 Miscellaneous examples
a We now look at some consequences of planetary
A=4 y dx motion. First we ask the question why does
0
π/2 our world have three spatial dimensions? Let
= 4ab sin2 ψ dψ the number be N . Clearly N > 2, for N = 2
0 places severe topological restrictions on our body
1 functions, but why not N = 11, N = 26? Note
= 4ab(π/2) = πab
2 that we exist. This means that there must exist
 planetary orbits which are close to circular and are
= πa2 1 − e2
stable. (This kind of argument is an anthropic
= πa3/2p1/2. principle.) We first investigate what gravity
looks like in N dimensions, and then we find
Now from example 40, Ȧ = 12 h. Thus the period which values of N permit stable planetary orbits.
is
Example 41. [Gravity in N dimensions] If U
A 2π is the gravitational potential energy, then g =
T = = 2πa3/2p1/2h−1 = √ a3/2, (78)
Ȧ GM −∇U . What scalar equation does U satisfy?
The simplest is ∇2U ≡ ∇.∇U = 0. In 2
which verifies the third law, for M is the same dimensions ∇U = (∂U/∂x, ∂U/∂y), ∇2U =
for all planets. Newton used this to deduce that ∂ 2U/∂x2 + ∂ 2U/∂y 2, and in N dimensions
gravity was an acceleration, i.e. the force acting  
on a planet was proportional to the mass of the ∂2 ∂2
+ . . . + U = 0. (79)
planet. ∂x12 ∂xN 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 145 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 146

We look for a solution U (r) where r = (x12 + Example 42. [Circular orbits in a central force]
. . . + xN 2)1/2. Now Recall equation (67)

∂U
=
dU ∂r
= U (r)
x1 r̈ = −f (r) + h2r −3,
∂x1 dr ∂x1 r

∂ 2U x 2
U x12 which has the first integral
= U  2 + − U 3 .
1
∂x1 2 r r r r
ṙ = H(r) ≡ 2E − 2
2
f (s) ds − h2/r 2. (81)
Thus

xi2 U xi2 We now use the stability algorithm, theorem 5 of


∇2U = U  +N − U §4.5: a circular orbit at r = a requires H(a) =
r2 r r3
i i 0 = H (a) and will be stable if H (a) < 0.
= U  + (N − 1)U /r. We can, by appropriate choice of E always set
H(a) = 0. Notice
Solving ∇2U = 0 gives
H (r) = −2f (r)+2h2/r 3, H (r) = −2f (r)−6h2/r 4.
U  = C/r N −1, U = D/r N −2, (80)
Thus H (a) = 0 fixes h2/a3 = f (a) and stability
requires H (a) = −2f (a) − 6h2/a4 < 0 or
where we are assuming N > 2, and C > 0,
D < 0 are constants. (The signs are determined
af (a) + 3f (a) > 0. (82)
by requiring gravity to be attractive.) Note that
for N = 3 we obtain the “correct” results.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 147 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 148
Assuming that U  is finite at r = 0,
We can now resolve the question as to the number
of spatial dimensions. Recall from equation (80) r
2 
that the gravitational force is proportional to r U (r) = 4πG ρ(s)s2 ds.
1/r N −1 with a positive proportionality constant. 0
Thus we substitute f (r) = 1/r N −1 in the stability
criterion (82) finding N < 4 for stability. It is no In particular, at r = R, R2U (R) = GM . Outside
accident that we live in three spatial dimensions. the sphere (83) implies r 2U (r) = C, and by
continuity C = GM . Thus outside the body the
We insert here the proof of a result used in §1.6. gravitational field is −U (r) = −GM/r 2. This
is as if all of the mass was concentrated at the
Example 43. [Gravitational field of a sphere] centre. (This result was used in example 7.)
In vacuum the gravitational potential U satisfies
∇2U = 0 and in three dimensions, assuming Example 44. [Startrek] A spaceship S has shut
spherical symmetry, i.e., U = U (r), U  + down its engine and is coasting near a strange
2r −1U  ≡ r −2(r 2U ) = 0. Thus in vacuum looking object O. The commander notices

r 2U (r) = C, C a constant. (83) • his angular momentum with respect to the object
is constant,
Now suppose we have a sphere mass M , radius R
made of material with mass density ρ(r). Inside • the orbit is a part of a circle, radius a, into the
the sphere Laplace’s equation ∇2U = 0 has to object.
be generalized to Poisson’s equation ∇2U =
4πGρ, and for spherical symmetry
What can he deduce about the object?

(r 2U ) = 4πGρr 2.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 149 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 150

distances of the particle from the centre are 2a


S
1
0
0
1 and 2a/3, and find the angle turned between a
maximum and the next minimum.
1111111
0000000
0000000
1111111

O
0000000
1111111
0000000
1111111
0000000
1111111
0000000
1111111
θ P From the initial conditions we have (setting√ the
0000000
1111111
0000000
1111111
0000000
1111111
0000000
1111111
0000000
1111111
mass m = 1) angular momentum h = a ak.
Using the displayed equation from example 44, we
The commander deduces he has a central force have u + u = 4/a − 3u, or
problem r 2θ̇ = h constant, governed by equation
(67), r̈ = −f (r) + h2/r 3. Recall also that if 4
u(θ) = 1/r then r̈ = −h2u2u and so u + 4u = ,
a
f (1/u)
u + u = . for which the solution is
h2u2
1 
Clearly r = OS = OP cos θ = 2a cos θ and so u= 1 + α cos(2θ) + β sin(2θ) ,
a
u = 1/(2a cos θ). Thus u + u = 8a2u3 and
f = 8h2a2r −5. It was a strange object!
where α and β are constants. (The
complementary function is obvious and the
Example 45. A particle is acted upon by an particular integral is easily guessed.) Initially we
attractive central force, and satisfies (67) with √
 set θ =√0. Initially r θ̇ = ak = u−1hu2 =
f (r) = k 4(a/r) − 3(a/r) .
2 3
Initially the hu = a aka−1 √(1 + α) and so√α = 0. Further
particle is distance a from the centre with both at θ = 0, ṙ = ak = −hu = ak(−2β) and so
the radial component of the velocity as well as β = − 12 . Clearly u achieves a minimum value of
that
√ in a perpendicular direction both equal to 1/(2a) at 2θ = π/2 and a maximum of 3/(2a)
ak. Show that the maximum and minimum at 2θ = 3π/2. Thus r achieves a maximum of

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 151 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 152
2a followed by a minimum of 2a/3 after turning energy equation directly! The kinetic energy
through a further Δθ = π/2. is 12 (ṙ 2 + r 2θ̇ 2) = 12 (ṙ 2 + h2/r 2) and is equal
  2
to 12 2
λ/3 initially. The elastic energy is
Example 46. A particle of unit mass moves on a
smooth horizontal table and is attached to a fixed
1
2 λ(r −
)2/
and is initially zero. Writing down
point of the table by an elastic string of natural the (conserved) energy and equating it to its initial
length
and modulus of elasticity λ. The particle value gives (after trivial algebra)
is held with the string just taut, and projected
 at ṙ 2 = 43
λ − λ(r −
)2/
− 43
3λ/r 2
right angles to the string with velocity 2
λ/3.
Show that the greatest length of the string in the or, setting r =
x, where x is dimensionless,
resulting motion is 2
. ẋ2 = (λ/
)f (x) where
This is clearly a CFP, and so angular momentum 4 1
is conserved. We obviously use polar coordinates f (x) = 1 − 2 − (x − 1)2.
3 x
with centre at the fixed point, and θ = 0 along
the initial direction of the
string. Then it is easy
to compute h =
× 2
λ/3 from the initial Oh dear, they want us to examine a quartic
conditions. One approach is to start from r̈ = equation! (And we have to be careful because
−f (r) + h2/r 3, where (from Hooke’s law) f (r) = if f (x) < 0, or if x < 1, then the physics
λ(r −
)/
. However when the string achieves used becomes invalid!) However, from the initial
its greatest length, ṙ = 0; we are interested in ṙ conditions we know that ṙ = 0 initially, and so
rather than r̈. x = 1 is a root. Guessing the answer, we can
verify that x = 2 is a root. It is easy to compute
Obviously we can multiply the above equation
by ṙ and integrate, obtaining the energy 8 8
equation. Alternatively we can write down the f (x) = 2(1 − x) + , f (x) = −2 − .
3x3 x4
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 153 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 154

Because f  < 0 we can deduce that f (x) has,


at most, one zero. It is easy to verify f (1) > 0
and f (2) < 0. Thus f (x) increases from zero
at x = 1. It reaches a positive maximum in Finally in this chapter we consider an example
x ∈ (1, 2) and returns to a zero value at x = 2. of unbound orbits produced by a repulsive
The assertions made in the question have been electrostatic force.
verified.

Example 47. [Rutherford scattering] A particle


of unit mass, charge q initial speed v is
approaching a stationary mass of charge Q with
impact parameter b.

Q
b
q V

We choose Qq > 0 so that the charges repel. By


what angle Θ is the orbit bent?

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 155 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 156
and so we have a hyperbola, see equation (77)

1 1
= (1 + e cos θ), (85)
Θ/2 r p

Q θ=0 where p = h2/k = 2Eb2/k. Note that θ = 0


corresponds to the minimum value of r. As
Θ/2 r → ∞, θ → ±θ0 = cos−1 (−1/e) where θ0 =
2 π + 2 Θ. Thus sin( 2 Θ) = − cos θ0 = 1/e and
1 1 1
−1
finally Θ = 2 sin (1/e).
We can also attempt this problem from first
principles, rather than rely on established results,
This is an electrostatics problem with an inverse but now we configure the geometry differently,
square central force, so that we can use Keplerian and we allow the particle to have mass m.
theory of §6.4 provided we replace GM by
k ≡ −Qq/(4π0), see §1.9 (and recall that we
are treating a particle of unit mass). Notice
that initially r is large, so that the electrostatic
potential energy
√ is negligible. Thus E = 12 v 2 and Θ rθ
h = bv = b 2E. Since E > 0 the eccentricity of Q
the orbit is
Because this is a central force problem we
  obviously have r 2θ̇ = h/m = bv. The initial
2Eh2  2Eb 2 horizontal component of the momentum is mv,
e= 1+ 2 = 1+ > 1, (84) and the initial energy is 12 mv 2. At late times
k k
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 157 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 158

the electrostatic potential energy is again zero, 1


The first approach established
√ sin 2 Θ = 1/e which
and so the final energy, all kinetic, is 12 mv 2 . would imply tan 2 Θ = 1/ e2 − 1. Using the
1
It follows that the final speed is again v, and displayed formula (84) for e it is easy to verify the
so the final horizontal momentum component consistency of the two approaches.
is mv cos Θ. The horizontal component of
the repulsive force is F = −kr −2 cos θ where Of course both of these approaches will also work
k = −Qq/(4π0) as above. Thus the reduction when the force is attractive, provided the orbit is
in the horizontal unbound, i.e., a hyperbola.
 t=∞ component of the momentum
should be t=−∞ F dt.
In order to perform this integral we change the
variable from t to θ using the expression for θ̇
given above. Thus the two expressions for the
change in horizontal momentum give
π−Θ   
k cos θ r2
mv(cos Θ − 1) = − − dθ
θ=0 r2 bv

or
k
mv(1 − cos Θ) = − sin Θ,
bv
which implies

Qq
tan 12 Θ = .
4π0mv 2b

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 159 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 160
7 Rotation in three dimensions 2

7.1 Introduction

For many purposes we can treat a frame centred


on the Sun, at rest with respect to the fixed
stars as inertial. However the centre of the earth
rotates with respect to the sun, and the surface
of the earth rotates relative to the centre. Frames
attached to the earth, although convenient, are
not inertial. How then are we to proceed?
We shall discuss two frames S and S, with a
common origin. S will always be inertial, and S
is in general rotating with respect to S. By this
we mean that if v is any vector then |v|2 = v.v
is the same in both frames, but v̂ = v/|v| is not.
Rotations preserve the lengths of vectors, but not
their directions.

Lemma 3. Rotations preserve angles between


two vectors u and v, i.e., u.v is the same in both
frames.

Proof.

|u + v|2 = |u|2 + |v|2 + 2u.v.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 161 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 162

where U is a 3 × 3 matrix. If we choose as


Thus rotations preserve shapes but not
a basis of S, e1 = (1, 0, 0)T , e2 = (0, 1, 0)T ,
orientations. However as we soon discover, rates
e3 = (0, 0, 1)T , then it is easy to see that the
of change of vectors in the two frames will be 
different. columns of U are the {ei} ,

Let {e1, e2, e3} be an orthonormal basis, fixed in    


U = e1|e2|e3 .
S, and
let u(t) be an arbitrary vector, so that
u = i ui(t)ei . The we define the rate of
change of u in S as Since we know that the columns of U are
  orthonormal, it follows that U is an orthogonal
du dui matrix. Although this has been deduced in a
= ei , (86)
dt S dt particular basis, it is of course a basis-independent
i
property. Let eiβ be the β-component with
i.e., we demand respect to the given basis of ei. In other words
eiβ = δiβ . Let ejα be the α-component of ej with
  respect to the same basis.
dei
= 0. (87)
dt S Then the equation ej = U ej means



Now let {ei} be an orthonormal basis fixed in S ,  ejα = Uαβ ejβ .
i.e. (dei/dt)S = 0 for each i = 1, 2, 3. Each ek β

can of course be expressed as a linear combination


of the {ei}. We express this as Indeed substituting ejβ = δjβ we have ejα = Uαj ,
confirming the displayed equation above.

ei = U ei , i = 1, 2, 3, To obtain another useful form for the components

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 163 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 164
of U note that which implies λ∗λ = 1. 2

ei . ej = eiαejα Theorem 9. [Euler] Let U be an orthogonal 3×

α
3 matrix which is proper, i.e., U = I. Then 1
= eiαUαβ ejβ is an eigenvalue of U of multiplicity 1, i.e., there
α β exists a vector e, unique up to multiplication, such
that
= δiαUαβ δjβ U e = e. (88)
α β

= Uij .
Proof. The eigenvalues are the roots of the
We now obtain some useful results about
secular equation
orthogonal matrices.

Lemma 4. If U is a real orthogonal matrix, then det(U − λI) = 0. (89)


its eigenvalues are of the form λ = exp(±iθ)
where θ is real, and are either real or occur in Because U is orthogonal U U T = I. This means
complex conjugate pairs. that U T (U − I) = I − U T , and we may deduce
det(U T ) det(U − I) = det(I − U T ) = det((I −
U )T ).
Proof. Because U is real, if e is an eigenvector
with eigenvalue λ then e∗ is an eigenvector with Now det(AT ) = det(A) for any square matrix A
eigenvalue λ∗. Thus and det(U ) = 1 for rotations. Thus

λ∗λe∗.e = (λ∗e∗)T (λe) det(U − I) = det(I − U ) = det(−(U − I))


= (U e∗)T (U e) = (e∗)T U T U e = e∗.e, = (−1)3 det(U − I).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 165 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 166

We therefore deduce that det(U − I) = 0. 7.2 Kinematical theorems


Comparison with the secular equation (89) shows8
that 1 is indeed an eigenvalue of U . Now
let the eigenvalues be λ1, λ2, λ3 = 1. Since Theorem 10. [Euler] The general rotation of
det(U ) = λ1λ2λ3 = 1 we deduce that λ1 λ2 = 1. an orthonormal frame with fixed origin leaves, at
Thus the eigenvalues are (1, 1, 1) (which is ruled each instant, points in one direction, the axis of
out since U is proper) or (eiθ , e−iθ , 1) where rotation at rest.
0 < θ < 2π. Thus λ3 = 1 has multiplicity
1. 2 Proof. This is an immediate consequence of
theorem 9. Since U e = e, where e is the unique
Corollary. Suppose we choose coordinates so eigenvector (up to a factor) with eigenvalue 1,
that the eigenvector e lies in the z-direction. points in the direction of e are not moved. Note
Then elementary algebra shows that however that U and hence e depend on time.
Thus the axis can, and usually does, change with
⎛ ⎞
cos θ sin θ 0 time. 2
U = ⎝− sin θ cos θ 0⎠
0 0 1 Theorem 11. [Angular velocity] If {ei} and

{ei} are orthonormal bases fixed in frames S
and S respectively, then there exists an angular
velocity vector ω such that
Hopefully some of these results are familiar, but
it is convenient to collect them all in one place.   
We now return to dynamics. dei 
= ω × ei , i = 1, 2, 3. (90)
dt S
8
There is nothing special about 3 here; any odd dimension would do.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 167 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 168

Proof. We have ei = U ei which implies ei = Then for any vector a,

U T ei. Thus ⎛ ⎞
ω2a3 − ω3a2
U̇ U T a = ⎝ω3a1 − ω1a3⎠ = ω × a,
      
dei dU dei ω1a2 − ω2a1
= ei + U .
dt S dt S dt S
where ω = (ω1, ω2, ω3)T . In particular
equation(90) holds. 2
The last term vanishes by definition. The

components of U are Uij = ei.ej and are the 
Corollary. Note that if ei is parallel to ω
same in either frame by lemma 3. We write 
(dU/dt)S = (dU/dt)S = U̇ . Thus then (dei/dt)S = 0. Thus the direction ω is
instantaneously constant. It is the instantaneous
axis of rotation implied by Euler’s theorem.
  
dei 
= U̇ ei = U̇ U T ei.
dt S

Now U U T = I implies U̇ U T + U U̇ T = 0 or
U̇ U T = −(U̇ U T )T so that U̇ U T is antisymmetric,
and we can write

⎞ ⎛
0 −ω3 ω2
U̇ U T = ⎝ ω3 0 −ω1⎠ .
−ω2 ω1 0

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 169 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 170

In the angular velocity theorem we implicitly


Theorem 12. [Rotating axes] For any vector asserted that ω was a vector without proving
a(t),     it. The next lemma shows that angular velocities
da da
= + ω × a. (91) add, which is what we require.
dt S dt S
Lemma 5. If frame S rotates with angular

  velocity ω  with respect to S, it rotates with
Proof. Set a = aiei. Then
angular velocity ω + ω  with respect to S.
   da    
da i   dei
= ei + ai . Proof. Recall that for an arbitrary vector a(t),
dt S dt S dt S
   
da da
 
Since ai = a.ei, lemma 3 implies ((dai)/dt)S =

= + ω  × a.
  dt S dt S
((dai)/dt)S . We note also that ((dei)/dt)S = 0
by definition. Thus Then
        
da d    da da
= aiei + aiω × ei = +ω×a
dt S dt S dt S dt S
     
da da
= +ω× aiei = + ω × a + ω × a
dt S dt S
   
da da
= + ω × a. = + (ω + ω ) × a.
dt S dt S

2 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 171 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 172
Theorem 13. [Acceleration] For an arbitrary Now (dω/dt)S = (dω/dt)S + ω × ω = (dω/dt)S
vector a(t), is unambiguous and we write it as ω̇. Expanding
the last term using (91)
     
d2 a d2 a da
= + 2ω × +      
dt2 dt2 S dt S (92) d2 a d2 a da
S = + ω̇ × a + ω × +
ω × (ω × a) + ω̇ × a. dt2 S dt2 S dt S
 
da
ω× + ω × (ω × a),
dt S
Proof. We have
 2    which is equation (92). 2
d a d da
=
dt2 S dt dt S S Note that in almost all applications ω̇ = 0 and
   
d da da so we usually discard this term, often without
= +ω× using (91), explicitly saying so.
dt dt S S dt S
     
d da da The term 2ω × (da/dt)S is the Coriolis
= 
+ω×a +ω× acceleration and ω×(ω×a) is the centripetal
dt dt S S dt S
      acceleration.
d da dω da
= + ×a + ω × +
dt dt S S dt S dt S
 
da
ω×
dt S

where (91) has been used again and again.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 173 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 174

7.3 Applications to dynamics Example 48. [Fairground wall] This is a large


circular drum of radius a which can rotate about
In an inertial frame S a particle of mass m at x(t) its vertical axis. Visitors pay to stand inside
subject to a force F satisfies Newton’s second law against the wall, the drum rotates, and when it
reaches full speed the floor slides down, leaving
 2  the visitors stuck to the wall.
d x
m = F.
dt2 S Let S be an inertial frame attached to the
fairground, and let S be attached to the drum,

Thus, from (92) with the vertical axes e3 and e3 coinciding. In

S the visitor is at rest, x = ae1, (dx/dt)S = 0,
 2    
d x dx (d2x/dt2)S = 0. Then, setting ω = ωe3, (93)
m +2mω× +mω×(ω×x) = F.
dt2 S dt S gives
(93)  2 
The extra terms reflect the non-inertial nature of d x   
m = mωe3×(ωe3×ae1) = −mge3+R+F,
S. Physicists love to confuse by “simplifying”, dt2 S
and rewrite (93) as
where R is the horizontal reaction force and F is
 2    the vertical frictional force. Resolving horizontally
d x dx 
m = F−2mω× −mω×(ω×x). we obtain −maω 2e1 = R, so the reaction is
dt2 S dt S
(94) directed inwards, and resolving vertically 0 =
They call the extra terms on the right hand −mge3 + F, so that the frictional force is
side the Coriolis force and centrifugal force upwards. Empirical friction laws suggest that
respectively. (Note that these are −mass × slipping will not occur if |F|  μ|R|, where
acceleration!) μ is a dimensionless friction coefficient, i.e., if
mg  μmaω 2. Thus the visitor “sticks” if aω 2/g

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 175 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 176

is sufficiently large. Setting Ω = Ωe3, g = (0, 0, −g) and x =
(r, 0, z), we find F = m(Ω2 r, 0, −g). This
Example 49. [Newton bucket experiment] A bucket is antiparallel to the normal n to the surface,
of water is rotated with vertical angular velocity as shown. An element t = (dr, 0, dz) of the
Ω about its axis. The shape of the surface adjusts surface,i.e., perpendicular to n is shown. This
so that the normal to the surface is antiparallel must be perpendicular to F and so Ω2r dr −
to the net force acting on the water. Using polar g dz = 0, or dz/dr = Ω2r/g so that the surface
coordinates (r, θ, z) determine the shape of the is a parabola z = z0 + 12 Ω2r 2/g, which is what
surface. Newton observed.
As usual S is inertial, fixed in space, and S is
fixed in the bucket. Example 50. [Eötvös experiment] Eötvös constructed
simple pendulums with different materials for the
Ω bobs.

Ω
n
T
t Fc
Fg
F

Because of friction, the water is at rest relative to He argued that the bob is in equilibrium under
the bucket. Using (94) the net force as seen in S three forces:
is
F = mg − m(Ω.x)Ω + m|Ω|2x. • Fg , the gravitational force proportional to M the

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 177 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 178

gravitational mass of the bob,

• Fc, the centrifugal force proportional to m the Example 51. A bead of mass m slides on a rigid
inertial mass, smooth circular loop of radius a which is forced
to rotate with uniform angular speed ω about a
• T, the tension in the string, required to balance fixed vertical diameter. Determine the points of
the other two forces. equilibrium, and their stability.


Thus the pendulum will be deflected slightly e3
from the vertical (except at the equator) and,
resolving perpendicular to T, we see that the small
deflection depends on the dimensionless ratio 
|Fc|/|Fg |. He observed that whatever material
111111
000000
000000
111111
000000
111111
000000
111111
e1
000000
111111
000000
111111

was used for the bob, the deflection remained the θ


000000
111111
000000
111111
000000
111111
0
1
000000
111111
0
1
000000
111111
0
1
m
same, and hence he deduced that m/M was a
universal constant, and equal to 1 by a suitable
choice of units. This justifies the assertion made   
We choose axes e1, e2 and e3 fixed in the loop
in §1.6. 
with e3 vertical and the loop lying in the 13-plane.
We may set
All of these examples have involved only
centripetal accelerations. We now turn to
examples where the Coriolis acceleration plays a x = a(sin θ, 0, − cos θ),
rôle. There are easier relevant examples in the ẋ ≡ (dx/dt)S = aθ̇(cos θ, 0, sin θ),
book by Lunn, and you may wish to study those
first, before looking at the next two examples. ẍ ≡ (d2x/dt2)S = aθ̈(cos θ, 0, sin θ) − xθ̇ 2.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 179 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 180
We now want to use equation (93) in the form corresponds to one degree of freedom. It can
be integrated (in the usual way) to give
 2   
d x dx
m +2mω × + mω × (ω × x) 1
dt2 S dt S θ̇ 2 = const. − ω 2( cos 2θ − 2ν cos θ).
2
= F = mg + R,
It is now straightforward to verify that if ν > 1
where R is the reaction of the wire, and is then θ = 0 is stable, while θ = π is unstable, but
perpendicular to it since the wire is smooth. If, if ν < 1 then θ = θ0 is stable and both θ = 0 and
as here, we are not interested in R we can “dot” θ = π are unstable.
the above equation with ẋ which is parallel to the
wire. Incidentally this also removes the Coriolis
term. This give

a2θ̈ − a2ω 2 sin θ cos θ = −ag sin θ.

Equilibrium requires θ̇ = θ̈ = 0 and so either


sin θ = 0 or cos θ = ν ≡ g/(aω 2 ). Therefore θ =
0 and θ = π are always equilibria. If 0 < ν < 1
there is a third equilibrium point at θ = θ0 =
cos−1 ν.
The stability algorithm of theorem 5 does apply
here, even though energy is not conserved
(because work is done maintaining the constancy
of ω). However the displayed equation

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 181 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 182

equation (93) in the form


Example 52. A smooth plane revolves with  2   
uniform angular speed ω about a fixed vertical d x dx
m +2mω × + mω × (ω × x)
axis which intersects the plane at O. The normal dt2 S dt S
to the plane is at a fixed angle θ to the vertical. = F = mg + R,
A particle of mass m is placed at rest at O and
is released. Show that in the subsequent motion, where R is the reaction of the plane. Here
the particle’s speed v relative to the plane, the x = (x, y, 0), (dx/dt)S = (ẋ, ẏ, 0), (d2x/dt2)S =
distance r of the particle from the axis and the (ẍ, ÿ, 0), and R = (0, 0, R).
depth h of the particle below O are related by
ω z
v 2 = r 2ω 2 + 2gh. x
θ
O
Further ω = (ω sin θ, 0, ω cos θ) and g =
It is exceedingly difficult to describe the motion
(−g sin θ, 0, −g cos θ). Resolving in the x- and
of the plane and particle with respect to a fixed
y-directions gives
inertial frame S and so we won’t try it! We use a
non-inertial frame S attached to the plane, with
 ẍ − 2ω ẏ cos θ − ω 2x cos2 θ = −g sin θ,
e3 normal to the plane, and we choose cartesian

coordinates (x, y, z) adapted to the {ei} with ÿ + 2ω ẋ cos θ − ω 2y = 0.
origin at O. The rotation axis is fixed in S and
it is a consequence of the rotating axes theorem (We don’t need to resolve in the z-direction,
that it is also fixed in S. We can therefore choose because we are not interested in the reaction
it to lie in the xz-plane. We now want to use force.)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 183 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 184
7.4 Motion on the surface of the
These equations imply
earth
ẋẍ + ẏ ÿ − ω 2(xẋ cos2 θ + y ẏ) = −g ẋ sin θ,
ω

with first integral O x


R
ẋ + ẏ − ω (x cos θ + y ) = −2gx sin θ + C,
2 2 2 2 2 2
O
λ

where C is a constant. But initially x = y = ẋ =


ẏ = 0 and so C = 0. Now identify v 2 = ẋ2 + ẏ 2 ,
r 2 = x2 cos2 θ + y 2 and h = −x sin θ to obtain
the required result. We would like to use a frame with origin at O, a
point fixed on the surface of the earth. However
in our development we have always assumed that
frames S and S have a common origin. Thus we
must replace x in equation (93) by R + x where
R is fixed in S. For a particle subject only to
gravity (93) becomes
   
d2 x dx
+2ω × +ω ×(ω ×(R+x)) = g.
dt2 S dt S
(95)
In almost all applications |x|
|R| and so we
drop the x-contribution to the centripetal term.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 185 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 186

We move the R-contribution to the right hand set of axes, centred at O on the surface of the
side, defining earth as follows: O1 is east and horizontal, O2
is north and horizontal while O3 is vertical, (see
g = g − ω × (ω × R) (96) figure).

as the effective gravity. Since |R| ≈ 6.4 ×


106m, it is easy to verify that the relative
difference between g and g is tiny and we usually
ignore the difference. With these approximations
we have
 2   
d x dx
+ 2ω × = g, (97)
dt2 S dt S

and this allows us to study the Coriolis terms in a


particularly clean way. A convenient form of this
equation, suitable for physicists, is
   
d2 x dx
= g + C, C = −2ω × .
dt2 S dt S
(98)

A basic question is in which direction is a particle,


moving relative to the surface of the earth,
accelerated? We choose a standard right handed

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 187 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 188
ω 2(north)
2 3
U

O

λ
O θ
1(east)

We set O
 
dx horiz. compt. of C
= (U cos θ, U sin θ, V ),
dt S
ω = (0, ω cos λ, ω sin λ), If V = 0, so that the particle is
moving horizontally, then we have C =
so that 2ωU sin λ(sin θ, − cos θ, .). (We have not
computed the 3-component because it will
C = 2ω(U sin θ sin λ − V cos λ, −U cos θ sin λ, be dwarfed by gravity.) We deduce: a
particle with a horizontal velocity suffers
U cos θ cos λ).
a horizontal deflection to the right in the
northern hemisphere and to the left in the
We can disentangle two important cases here. southern one.
If U = 0, so that the particle is moving
vertically, then we have C = 2ωV cos λ(−1, 0, 0).
We deduce: a rising(falling) particle is
deflected to the west(east).

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 189 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 190

Example 53. [Projectiles in S] The projectile


has initial position x0, velocity V0 at t = 0. As an example consider dropping a particle from
Because we are ignoring O(ω 2 ) terms (see the rest down a lift shaft. We have, to this order of
comment just after equation (95)) we can use the approximation,
technique of “successive approximations” to solve
(98). First we write it as 1 1
z = − gt2, x = gt3 ω cos λ,
 2    2 3
d x dx
= g − 2ω × + O(ω 2 ) (99)
dt2 S dt S a deflection to the east.
= g + O(ω). (100)

Integrating (100) is easy:


 
dx
= V0 + gt + O(ω).
dt S

We now insert this into the right hand side of


(99):
 2 
d x
= g − 2ω × V0 − 2ω × gt + O(ω 2 ),
dt2 S

which integrates to give


1 1
x = x0+V0t+ (g−2ω×V0)t2− ω×gt3+O(ω 2).
2 3
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 191 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 192
approximation is good enough, and, setting
Example 54. [Foucault pendulum] This is a n2 = g/
we have, correct to first order,
simple pendulum consisting of a bob of mass
m attached by a string of length
to a fixed ẍ−2ω sin λẏ+n2x = 0, ÿ+2ω sin λẋ+n2y = 0.
point. We set the origin O at (0, 0,
) so that
in equilibrium (ignoring rotation) the bob is at
x = (0, 0, 0). In the motion we assume x/
and
y/
are both small. It follows that z, ż and z̈ are Setting ζ = x + iy we combine these as
second order small and we ignore them. Let T be
the tension in the string, and write (dx/dt)S = ẋ ζ̈ + 2iω sin λζ̇ + n2ζ = 0,
etc. Then (97) becomes
with solution
T
ẍ + 2ω × ẋ = g + ,
m
ζeiωt sin λ = Aeiμt + Be−iμt,
with components
where μ2 = n2 + ω 2 sin2 λ, and A and B are
ẍ − 2ω sin λẏ = −(T /m)(x/
), complex constants. Now choose new axes Ox,
Oy  which are rotating around Oz with angular
ÿ + 2ω sin λẋ = −(T /m)(y/
), velocity −ω sin λ. Then
−2ω cos λẋ = −g + T /m.
x + iy  = (x + iy)eiωt sin λ = Aeiμt + Be−iμt.
Thus T /m = g plus first order quantities.
However in the other equations T /m is multiplied For simplicity we consider the case where A and
by a small quantity, e.g., x/
. Thus this B are real and set C = A + B, D = A − B so

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 193 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 194

that
Example 55. [Introduction to meteorology] Consider
x = C cos μt, y  = D sin μt, the large scale horizontal wind circulation. Masses
of air tend to move from regions of high pressure
Clearly x and y  trace out an ellipse with to regions of low pressure. In the vertical direction
frequency μ, and since μ = n + O(ω 2) this is air close to the earth is warmer than higher
exactly what one would expect if the earth were air, and the resulting pressure gradient roughly
not rotating. However these coordinate axes are balances gravitational forces. In the horizontal
rotating with angular velocity −ω sin λ relative direction there are persistent long-range motions
to the surface of the earth, and so the ellipse of air masses, usually called winds. Here the
rotates with this angular velocity in the clockwise pressure gradients are comparable to the Coriolis
direction. Most science museums have a Foucault forces, and one can see both effects, particularly
pendulum. in cyclones, regions where the pressure achieves
a minimum.

no rotation
low pressure

with rotation
high pressure

If the earth did not rotate wind would flow from


high pressure to low pressure as shown. In

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 195 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 196
the northern hemisphere Coriolis forces deflect 8 Systems of particles
it to the right as shown in the figure. If the
deflection causes the wind to move nearly parallel
to the isobars (contours of equal pressure) as 8.1 Centre of Mass
shown, then the Coriolis force becomes almost
anti-parallel to the pressure gradient, and can Up to now we have concentrated on systems which
match it. One can then obtain geostrophic can be formulated in terms of a few or even a single
winds which follow the isobars (shown red in point particle. We now want to extend this to
the figure). Once friction is taken into account systems of many particles, labelled by A, B, . . . =
these winds are slowed and directed towards the 1, 2, . . . , N . (A continuum approach where sums
cyclone (green arrows). Thus in cyclonic weather are replaced by integrals is also possible.) The
the weather maps show direct evidence of Coriolis Ath particle has mass mA,
position xA, and is
effects. acted on by forces FA + B=A FA←B . The
internal forces are required to satisfy Newton’s
third law, FB←A = −FA←B . If, in addition, they
are directed along xB − xA,
low
FA←B = KAB (xA − xB ), (101)

they are said to be torque-free. (Some authors


assume, implicitly, that all internal forces are
torque-free.)
We now generalize two definitions originally

given
in §5.2. The total mass is M = A mA, and
the centre of mass or centroid is at X(t)

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 197 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 198

where M X = A mAxA. In this section all time


derivatives take place within an inertial frame S Theorem 15. Let TCoM denote the kinetic
and so we replace (d/dt)S by ˙. energy, as measured in the centre of mass frame.
Then the total kinetic energy is given by
Theorem 14. The centroid behaves like a
particle of

mass M subject to the net external 1


force F = A FA, i.e., T = M |Ẋ|2 + TCoM . (103)
2
M Ẍ = F. (102)

Proof. Write xA = X + x̂A, where x̂A is the


position of the particle in the CoM frame, and set
Proof.
ẋA = uA, Ẋ = U and x̂˙ A = ûA. Then
 
M Ẍ = mAẍA = FA + FA←B
A A
1 1
B=A T = mAuA.uA = mA(U + ûA).(U + ûA)
2 2
A A
=F+ FA←B .
1 1
A B=A = M U.U + U. mAûA + mAûA.ûA
2 2
A A
Consider the contribution to the double sum
1
from a pair of particles, e.g., 2 and 7. We = M |Ẋ|2 + TCoM .
have F2←7 + F7←2 = 0. Thus the double sum 2
vanishes. 2
(The middle term in the middle line vanishes by
Corollary. If F = 0, then the centre of mass the obvious generalization of equation (54).) 2
frame (CoM) defined by X = 0 is also inertial.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 199 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 200
where μ = m1m2/M is the reduced mass.
Example 56. [Reduced mass] Things are simplest
in the CoM frame when N = 2. Suppose we have The reduced mass also plays a rôle in the equation
two stars moving under their mutual gravity, and of motion, which can be written as
set r = x2 − x1. Then Newton’s law of gravity
implies Gm1m2
μr̈ = − r.
r3
m1ẍ1 = Gm1m2r −3r, m2ẍ2 = −Gm1m2r −3r,

and M Ẍ = 0 since there are no external forces.


Also

r̈ = ẍ2 − ẍ1 = −G(m1 + m2)r −3r = −GM r −3r,

so that the separation r evolves like the position


of a planet about a sun of mass M , see chapter
6.
In the CoM frame m1x̂1+m2 x̂2 = 0 and x̂2−x̂1 =
r imply x̂1 = −(m2/M )r and x̂2 = (m1/M )r so
that
 2  2
1 m2 1 m1
TCoM = m1 ṙ.ṙ + m2 ṙ.ṙ
2 M 2 M
1
= μ|ṙ|2
2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 201 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 202

8.2 Angular momentum


Now let O be an arbitrary point, not necessarily
the origin. The total angular momentum
We defined angular momentum with respect to
about O is
the origin O of a single particle by (63),

HO = (xA − xO ) × mA(ẋA − ẋO ), (106)
hO = x × mẋ.
A

Note that if mẍ = F then ḣO = x × F. and the total torque or net couple about O
is
F GO = (xA − xO ) × FA. (107)
x y A

O Theorem 16. Suppose the internal forces are


torque-free, i.e., (101) holds. Then
The couple or moment or torque of F about
O is
gO = x × F. (104) ḢO = GO + M (xO − X) × ẍO . (108)
Note that gO doesn’t depend on the choice of x.
In particular if either O = CoM or ẍO = 0 then
For if y − x is parallel to F then gO = y × F as
well. Thus we have
ḢO = GO , (109)
ḣ0 = g0. (105)
the obvious generalization of (105).
We now generalize this to systems.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 203 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 204
Proof. The deduction of (109) from (108) is Example 57. A bead of mass m1 can slide on a
trivial. From (106) smooth straight horizontal wire, and a particle of
mass m2 is attached to the bead by a light string
of length
. The system moves in the vertical
ḢO = (xA − xO ) × mA(ẍA − ẍO )
plane through the wire, and the string remains
A
taut. The horizontal displacement of the centre
= (xA − xO ) × FA + (xA − xO ) × FA←B of gravity, G, at time t is denoted by x and the
A A B=A angle the string makes with the downward vertical

− mAxA × ẍO + mAxO × ẍO . by θ. Prove that during the motion ẋ and
A A
1
(m1 + m2 sin2 θ)θ̇ 2 − (m1 + m2)(g/l) cos θ
The first term on the right hand side is GO . The 2
third and fourth are −M X × ẍ0 and M xO × ẍO
remain constant. Find the period of small
respectively. The contribution of particles 2 and
oscillations about the vertical.
7 to the double sum is
m1
(x2 − xO ) × F2←7 + (x7 − xO ) × F7←2
θ
= (x2 − xO − x7 + xO ) × F2←7 by Newton 3 G
= (x2 − x7) × K27(x2 − x7) by (101) m2
= 0.
The coordinates of G are X = (x, −(m2/M )
cos θ),
2 where M = m1 + m2. Because the external forces
(gravity and reaction at m1) are vertical, theorem

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 205 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 206

14 implies that the horizontal component of Ẋ is is simple harmonic motion with frequency ω where
constant, i.e., ẋ is constant. The potential energy ω 2 = (M/m1)(g/l).
of the system is obviously U = −m2g
cos θ plus
a constant. To compute the kinetic energy we We now look at two examples involving angular
use theorem 13. The kinetic energy of the momentum.
centre of mass is clearly 12 M |Ẋ|2 = 12 M (ẋ2 +
((m2/M )
θ̇ sin θ)2). Now particle 1 has relative Example 58. A horizontal wheel of radius r with
coordinates x̂1 = (m2/M )
(sin θ, − cos θ) and so buckets on its circumference revolves about a
particle 1 contributes 12 m1(m2/M )2
2θ̇ 2 towards frictionless vertical axis. When the wheel has
TCoM and there is a similar contribution from angular velocity ω, the angular momentum of
particle 2. Now theorem 15 implies that the total the wheel and buckets is Iω. Water falls into the
kinetic energy is buckets at a uniform rate of mass m per unit time.
Treating the buckets as small compared with the
1 1 wheel, find the angular velocity ω(t) of the wheel
T = M ẋ2 + (m2/M )(m1 + m2 sin2 θ)
2θ̇ 2
2 2 after time t if Ω be its initial value. Show that
after time t the wheel has turned through an angle
Conservation of energy implies
IΩ  mr 2t 
1 log 1 + .
(m1 + m2 sin2 θ)
2θ̇ 2 − M g
cos θ mr 2 I
2

is constant, which is equivalent to the required


result. The motion of and forces on the water and buckets
form a complicated problem, but a moment’s
For small values of θ we have correct to second thought reveals that all of the forces in this
order, 12 m1
2θ̇ 2 −M g
(1− 12 θ 2) = constant, which problem are vertical. Now let O be the centre

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 207 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 208
of the wheel. From equation (107) we can deduce
that the total torque of the forces has a vanishing Example 59. A particle of mass m is free to
component in the vertical direction. Now theorem move in a thin smooth uniform straight tube of
16, equation (109), implies that the vertical mass 3m and length a. The tube can turn freely
component of the total angular momentum is in a horizontal plane about one end O which is
constant. Initially this is IΩ. At time t the fixed. Initially the tube has angular velocity Ω
wheel and buckets contribute Iω(t). However the and the particle is at rest relative to the tube at
buckets now contain a mass mt of water all at a its midpoint. Find the velocity with which the
distance r from O. This water has a velocity of particle leaves the tube.
magnitude rω(t) directed circumferentially, and [A result, proved in chapter 9, is that when the
so it contributes mtr 2ω(t) to the total angular tube has angular velocity ω its angular momentum
momentum. Thus we have is ma2ω and its kinetic energy is 12 ma2ω 2.]

(I + mr 2t)ω = IΩ or ω= .
I + mr 2t
 r
The angle turned is obviously ω dt which gives
the result displayed above. θ
O
We describe the position of the particle by polar
coordinates (r, θ). There are internal forces
between the tube and the particle and an external
reaction force at O, but there are no external
torques about O. Thus we have conservation of
angular momentum about O, theorem 16 , and of

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 209 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 210

course conservation of total kinetic energy. The 9 Rigid bodies


first implies
9.1 Introduction
ma2θ̇ + mr 2θ̇ = ma2Ω + m(a/2)2Ω = 54 ma2Ω,
Collections of particles are all very well, but we are
and the second actually interested in macroscopic rigid bodies. A
1 2 2 1 1 1 rigid body is a collection of particles labelled by
ma θ̇ + m(ṙ 2 + r 2θ̇ 2) = ma2Ω2 + m(a/2)2Ω2 A, B, . . . = 1, 2, . . . , N such that
2 2 2 2
5
= ma2Ω2. 1. |xA(t) − xB (t)| = const., ∀A, B,
8

It is simple algebra 2. the internal forces are torque-free; condition (101)


 to conclude that at r = a, holds.
θ̇ = 58 Ω and ṙ = 15/32aΩ, so that the ejection
velocity is
 15 5  We can specify its position by giving the situation
aΩ, aΩ . of three points (not on a straight line) at each
32 8
time t. Clearly x1(t) requires 3 numbers, but
x2(t) requires only 2 (for |x2 − x1| is fixed) and
x3(t) needs only one.
According to Newton’s principle, §1.1, we shall
also need 6 parameters to describe the velocity.
An evident choice is the velocity of one point and
in addition the angular velocity about it. In total
six parameters are needed.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 211 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 212
9.2 Linear momentum 9.3 Angular velocity

We learned in chapter 8, theorem 14, that the We need here to revert to the notation and
CoM behaves like a particle

of mass M subject to conventions of chapter 7. Let S be an inertial


the net external force F = A FA, frame fixed in space, and let S’ be a (possibly
non-inertial) frame fixed in the body.
M Ẍ = F, (110) If A and B are two points fixed in the body, then

xA − xB is fixed in S. It follows from theorem
where X = M −1 A mA xA is the position of the 119 of chapter 7 that there exists a unique angular
centroid. velocity vector ω, independent of A and B such
that

ẋA − ẋB = ω × (xA − xB ), (111)

where, as throughout this chapter, we write


(da/dt)S as ȧ.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 213 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 214

9.4 Angular momentum


If at time t there is a line of points instantaneously
at rest then this is the (instantaneous) axis of
Let O be an arbitrary point of the body. We
rotation.
define the angular momentum about O by
the usual formula, see (106),
Theorem 17. An axis of rotation exists iff
ω.Ẋ = 0, and it is parallel to ω.
HO = (xA − xO ) × mA(ẋA − ẋO ), (112)
A
Proof. The axis would have to satisfy
  It follows from theorem 16 that
0 = ẋ = ẋ + ω × (x − X) ,
ḢO = GO + M (xO − X) × ẍO . (113)
where we have set xA = x and xB = X in (111).

By dotting this equation with ω we see that it is where GO = A (xA − xO ) × FA is the net
inconsistent unless ω.ẋ = 0. If this condition is torque of the external forces about O. Note the
satisfied then the solution is simplification if

x = X + |ω|−2ω × ẋ + λω, 1. xO = X, i.e., O is the centre of mass,

where λ is arbitrary, and this defines a line parallel 2. ẍO = 0, in particular if O is at rest.
to ω. 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 215 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 216
Example 60. [Moment of inertia] A bicycle Example 61. [Continuation of example 60] Suppose
wheel, centre O, radius a, is mounted on an now that someone applies a torque G = Ge2 at
axle which is initially in the e1 direction. The rim O.
is lined with lead, so that essentially all of the
mass is in the rim. The wheel spins with angular 3
speed ω. Compute ω and HCoM .
G 2
Obviously ω = ωe1. For any point A on the rim,
xA − X lies in the 23-plane. Clearly ẋ = 0. Now H + Gδt
(111) implies ẋA = ω × (xA − X) and then the δϕ
definition (106) implies O 1
H
  Now (113) implies Ḣ = G. In time δt, H →
HCoM = mA(xA − X) × ω × (xA − X) H + Gδt. Note that H.G = 0 initially so that
A H = |H| is unchanged to first order, but H
(114)
is rotated through a small angle δϕ = Gδt/H
= mA(xA − X).(xA − X)ω (115) in the 12-plane. Thus the wheel acquires an
A angular velocity component dϕ/dt = G/H in the
= M a2ω. (116) 3-direction.

Here H and ω are proportional, and the constant Clearly the relationship between ω, H and G is
of proportionality, M a2, is the moment of non-trivial. We now explore it further, looking
inertia about the axle. first at energy, which is a scalar concept.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 217 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 218

9.5 Energy in the CoM frame where TCoM is the kinetic energy in the CoM
frame.
Recall that for a single particle with velocity ẋ,
linear momentum p, the kinetic energy is T =
2 ẋ.p. We might conjecture that 2 ω.H has a
1 1

similar interpretation. Suppose we look in the


CoM frame. Equation (111) implies that for each
particle A

ẋA − ẋ = ω × (xA − X). (117)

Setting O to be the centre of mass (112) asserts



HCoM = mA(xA − X) × (ẋA − ẋ),
A

and so9
1 1
ω.HCoM = mA(ẋA − ẋ).ω × (xA − X)
2 2
A
1
= mA|ẋA − ẋ|2
2
A

= TCoM ,
9
We use the cyclic property of the scalar triple product in the first
equation and (117) in the second.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 219 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 220
9.6 Inertia tensor and moments of
Now recall theorem 15. The total kinetic energy
is inertia
1
T = M |ẋ|2 + TCoM ,
2 In this section let O be any preferred point of the
which can be written as body (and not necessarily the centre of mass).
Equation (111) implies that for each particle A
1 1
T = M |ẋ|2 + ω.HCoM . (118)
2 2 ẋA − ẋO = ω × (xA − xO ),
The first term on the right is called the
while (112) asserts
translational kinetic energy and the second
is called the rotational kinetic energy.
Another way to write this is HO = mA(xA − xO ) × (ẋA − ẋO ).
A

1 1
T = Ẋ.P + ω.HCoM , (119) Combining these we have
2 2
 
where P = M ẋ is the linear momentum of the HO = mA(xA − xO ) × ω × (xA − xO ) .
centre of mass. A
(120)
We know that P and ẋ are proportional. We Clearly we can write this as HO = IO (ω) where IO
have seen from example 61 that the relation is a linear map R3 → R3, the inertia operator.
between ω and H is likely to be considerably more This is the linear relationship alluded to at the end
complicated, but we might reasonably expect of the last section. A linear map from one vector
linearity. space (here “angular velocity space”) to another
(here “angular momentum space”) is called a

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 221 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 222

tensor relationship, and will be explored in detail Converting to suffix notation we have
in the Methods course in term 4. However we

know from the course C1/2 that we can interpret HOi = mAimnrAmnkpωk rAp
I as a 3 × 3 matrix Iik . Invoking summation A
convention (in this section only) we write (120)
as = nimnkpωk mArAmrAp
A

= (δik δmp − δipδmk )ωk mArAmrAp.
HOi = IOik ωk , (121) A

Comparing this with (121) we have


where the IOik are the components of the
inertia tensor at O. IOik = (δik δmp − δipδmk ) mArAmrAp,
A

or
We now want to obtain an explicit formula for  
the inertia tensor at an arbitrary point O. It IOik = mA (rAmrAm)δik − rAirAk , (123)
is very convenient to introduce the abbreviation A

rA ≡ xA − xO . Then (120) shortens to


where rAi = xAi − xOi.


HO = mArA × (ω × rA). (122)
A

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 223 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 224
Consider first the evaluation of these quantities at
If we wish to regard the rigid body as continuous
the centroid. (The general point O will be treated
then we need to change the sum to an integral.
shortly.) If we are lucky the body will possess
We set mA = ρ(x) dV = ρ d3x, where ρ(x) is the
a symmetry axis through the centroid. Because
density, and dV = d3x is a volume element. Then
the body is invariant under rotations about this
(123) becomes
axis, the principal directions (an intrinsic property
of the body) must also share this invariance.
 
IOik = ρ(x) (rmrm)δik −rirk d3x, (124) Thus one principal axis at the centroid must be
Body along the symmetry axis, and the other two are
perpendicular to it. Further the principal moments
where ri = xi − xOi. about the other two directions must be equal, so
Notice that IO is symmetric, i.e., IOki = IOik . that we only need two integrals to compute the
This means that we can choose coordinate axes eigenvalues directly.
along the eigenvectors of IOik, so that IOik =
The next lemma shows how to compute principal
diag(I1, I2, I3), where the Ik are the eigenvalues moments about the centroid in this special case.
of IOik . It is obvious from the definition that
these axes are fixed in the body; they are called
the principal directions at O and the Ik are
called the (principal) moments of inertia at
O.
Masochists will note, with relish, that the
computation of the principal moments requires
6 integrals (to form IOik) followed by an
eigenvalue/eigenvector determination. However,
with some finesse, we can often avoid this.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 225 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 226

Lemma 6. Let e1 be a principal direction


through the CoM, and let RA be the perpendicular Theorem 18. [Parallel axes] For any point O,
distance of particle A from e1. Then  
HO = HCoM + M (xO − X) × ω × (xO − X) ,
I1 = mARA2. (125) (126)
A and

IOik = ICoM ik + M [(xOm − Xm)(xOm − Xm)δik


e1 
− (xOi − Xi)(xOk − Xk ) . (127)
CoM RA
A

Proof. Recall equation (122)


Proof. One way to verify the result is to set
i = k = 1 in (123). However it is constructive to
proceed as follows. Suppose the body was rotating HO = mArA × (ω × rA),
with angular velocity ω = ωe1. Clearly |ṙA| = A

RAω and so the kinetic


energy relative to the where rA = xA − xO = (xA − X) − (xO − X) =
centroid is TCoM = 12 A mARA2ω 2 . But from
RA − RO say. Then
(118), TCoM = 12 HCoM .ω. However because e1
is an eigenvector of ICom ik , HCoM = I1ω and so
TCoM = 12 I1ω 2 which proves the result. 2 HO = mARA × (ω × RA) − mARA × (ω × RO )
A A

If we know the inertia tensor at the centroid we − mARO × (ω × RA) + mARO × (ω × R
can compute it at an arbitrary point O. A A

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 227 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 228

The middle two terms vanish since A mARA = the plane is x3 = 0. Then equation (123) implies
0 and the first and the last give (126). Equation
⎛ ⎞
(127) follows on writing this out in suffix . . 0
notation. 2 IOik = ⎝ . . 0⎠
0 0 .

Another useful result is so that O3 is a principal axis, and the other two
lie in the plane.
Theorem 19. [Perpendicular axes] Consider a Choosing coordinates adapted to the principal
lamina, a 2-dimensional object with principal axes, lemma 6 implies
directions e1 and e2 in the plane, and e3
perpendicular to the plane. Then the principal
I1 = mA(xA22 + xA3 2),
moments at the centroid satisfy
A

I2 = mA(xA32 + xA1 2),
I3 = I1 + I2 . (128) A

I3 = mA(xA12 + xA2 2),
A

and since xA3 ≡ 0 the result follows. 2


Proof. Notice that the laminar is invariant under
reflections in its normal. The principal axes must
share this invariance and so can be taken in the
stated directions. An algebraic version of this
argument follows if we choose coordinates so that

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 229 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 230

9.7 Computation of moments of Example 63. [Continuation of example 62] What


inertia is the moment of inertia about an axis through E
perpendicular to the rod?

Example 62. Consider a uniform rod of mass Suppose the rod has angular velocity ω = ωe3.
M , length 2a, centre O, one end E. Then HCoM = 13 M a2ω. Now use the parallel
axes theorem, replacing O by E.
e3  
HE = HCoM + M (xE − X) × ω × (xE − X)
e1 = HCoM + M |xE − X|2ω
O 1
= M a2ω + M a2ω
3
E = 43 M a2ω
By symmetry the principal axes are e1 along the
where we used the result ω.(xE − X) = 0. This
rod and any two perpendicular directions. The
implies IE3 = 43 M a2.
density is ρ = M/(2a). Using the continuum
form a This is a typical use of the parallel axes theorem.
M 2 1 Note however that here the geometry is so simple
IO3 = x dx = M a2.
−a 2a 3 that we could compute
2a
M 2
Exercise 14. Show that IO1 = 0. What does IE3 = x dx = 43 M a2
0 2a
the perpendicular axes theorem say here?
directly.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 231 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 232
and IO1 = IO2 = 14 M a2.
Example 64. Consider a circular ring of mass
M , radius a lying in the 12-plane. Example 66. Next consider a uniform cylinder
of mass M , radius a and height h. Choosing
e3 the 3-axis along the axis of symmetry, obviously
e2 a principal axis, we have, taking O to be the
centroid,
O e1
M
IO3 = 2 (x2 + y 2) dV
πa h
h/2 a 2π
Clearly IO3 = M a2. By symmetry IO1 = IO2, M
and so the perpendicular axes theorem implies = 2 r 2dr rdθ dz,
πa h z=−h/2 r=0 θ=0
that IO1 = IO2 = 12 M a2.
but there is no need to evaluate this integral!
Example 65. The same setup, but now a We can collapse the 3-axis, turning this into the
uniform disc of radius a, mass M , density integral of the previous example, with the same
ρ = M/(πa2 ). Then if O is the centroid, result, I03 = 12 M a2. Clearly IO2 = IO1 where

M M
IO3 = (x2 + y 2) dx dy I01 = (x2 + z 2) dV
πa2 πa2h
a 2π
M M h/2 a 2π
= r 2 dr rdθ = 2 (r 2 cos2 θ + z 2)dr rdθ dz
πa2 r=0 θ=0 πa h
a z=−h/2 r=0 θ=0
M 1 M
= 2 2π r 3 dr = M a2 = (3a2 + h2),
πa 0 2 12
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 233 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 234

by conventional methods.
Example 67. Next consider a uniform sphere,
centre O, of mass M , radius a, density ρ =
3M/(4πa3). Any diameter is a principal axis, the
principal moments are all equal and, e.g.,

IO3 = ρ (x2 + y 2) dV.

Here it is easiest to decompose the sphere


√ into
discs z = const., thickness dz, radius a2 − z 2
giving

a  2π √a2−z 2 
2
IO3 = ρ r dr rdθ dz
z=−a θ=0 r=0
a
1 2
= 2πρ (a − z 2)2 dz
z=−a 4
= 25 M a2.

Exercise 15. What is the relation between the


height h and radius a of a uniform cylinder for
it to behave dynamically like a sphere, i.e., for

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 235 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 236
equality among the principal moments of inertia Example 68. Finally consider the problem of
at its centroid? determining the inertia tensor at the centroid O
of a uniform cube of mass M , side 2a.
We choose coordinates so that O is (0, 0, 0) and
the vertices are (±a, ±a, ±a). By symmetry Oz
is a principal axis and
a a a
M
IO3 = 3 (x2 + y 2 ) dx dy dz
8a −a −a −a
= 23 M a2.

By symmetry IO1 = IO2 = IO3 and so


⎛ ⎞
1 0 0
IO = 23 M a2 ⎝0 1 0⎠
0 0 1

Note that for a sphere


⎞⎛
1 0 0
IO = 25 M a2 ⎝0 1 0⎠
0 0 1

so that a cube behaves dynamically, pretty much


like a sphere.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 237 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 238

9.8 Miscellaneous examples O is at rest, that

4 2
3 M a θ̈ = −M ga sin θ,
Example 69. [Yet another pendulum] A uniform
rod, mass M , length 2a swings in a vertical plane or
about one end which is fixed. Determine the θ̈ + (3g/4a) sin θ = 0,
frequency of small oscillations.
corresponding to a simple pendulum with ω 2 =
(3g/4a). We can regard this as being a simple
O pendulum of equivalent length
= 4a/3.

θ C Example 70. [Example 69 continued] Now a


disc of mass 12M , radius 13 a can be clamped so
111
000
000
111D
000
111 that its centre D is on the rod at a distance x
from O. Show that the length
of the equivalent
Mg simple pendulum satisfies 23 a 
 2a.
The disc has moment of inertia 12 (12M )(a/3)2
about a horizontal axis through D, example 65,
The component of HO perpendicular to the plane and moment 23 M a2 + 12M x2 about a horizontal
is 43 M a2θ̇, see example 63. The only external axis through O, parallel axes theorem. Thus the
forces are gravity and the reaction at O. The total moment about O is M (12x2 + 2a2). The
reaction has zero couple at O and we can regard net couple is −(12M gx + M ga) sin θ and so
the gravitational force as acting through the
centroid C. Thus the net couple is −M ga sin θ. It 12x2 + 2a2

= ,
follows from (113), specialized to the case where 12x + a
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 239 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 240
which has a minimum of 2a/3 at x = a/3 and horizontal axis through O is 13 M a2 + 13 M a2 +M a2,
achieves its greatest value at both x = 0 and where the contributions come from OA, OB and
x = 2a. AB respectively. (Examples 63 and 64 apply
here.)
Example 71. A uniform wire of length 3a is
The main computational problem here would seem
bent to form a sector of a circle, with the two
be to determine the position of the centroid, and
radii OA, OB and the arc AB each of length
hence the gravitational torque. It is actually
a. The point O is fixed and the wire swings in
simpler to compute the gravitational potential
a vertical plane. Show that the period of small
energy directly. Further we need only compute
oscillations is
it in the rest, equilibrium, position. For when
 the system is deflected through an an angle θ,
5a the potential energy picks up a factor cos θ. At
2π .
3g(cos 12 + 2 sin 12 ) rest the potential energy of OA (and OB) is
−M g(a/2) cos 12 . The potential energy of AB is

−M ga −θ0 0 cos θ dθ = −2M ga sin 12 . Thus the
O total potential energy, at equilibrium, is
θ0  
1 1
−M ga cos + 2 sin .
2 2
A B
Therefore the total energy, in general position, is
Simple geometry shows that the angle AOB is 1    
(in radians). Thus θ0 = 12 . Let the total mass 1 5 1 1
E= M a θ̇ −M ga cos +2 sin
2 2
cos θ.
be 3M . Then the moment of inertia about a 2 3 2 2

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 241 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 242

Comparison with the standard theory for the Example 72. A uniform rod AB of length 2a
simple pendulum now produces the required slides with its ends fixed on a smooth circular wire
answer. loop whose centre is O. If b denotes the distance
of the centre C of AB from O, and θ is the angle
that OC makes with the vertical, prove that
6bg
θ̇ 2 = (cos θ − cos α),
a2 + 3b2
where α is the maximum value of θ. Find the
frequency of small oscillations when α
1.

O B
θ
C

A
Example 62 shows that IC = 13 M a2 and the
parallel axes theorem shows that IO = 13 M a2 +
M b2. Thus the total energy is
 
1 1
M a2 + b2 θ̇ 2 − M gb cos θ,
2 3

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 243 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 244
which has to be a constant, and this implies
the stated form for θ̇ 2 = f (θ). Now the Example 73. A lamina of mass M is at rest on
stability algorithm, theorem 5, implies that at a smooth horizontal table. A horizontal impulsive
equilibrium f (θ) ∝ sin θ has to vanish, and the force of magnitude F is applied at a point A
stable equilibrium is obviously at θ = 0. Further distant a from the centroid G, and the direction of
the frequency of small oscillations is given by the force is perpendicular to GA. What happens?
ω 2 = − 12 f (0) or

bg F
ω2 = .
a2 + 3b2
G A

Let IG be the moment of inertia about a vertical


axis through G. After the impact G will have
a linear velocity v in the direction of F and the
lamina will have a vertical angular velocity ω about
G.
By (110) M v = F and so v = F/M . We shall
assume that the internal forces are torque-free.
The vertical axis through G is a principal axis
and so HG = IGω. By (113) IGω = aF and so
ω = aF/IG.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 245 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 246

Further we can find F directly by taking moments


about G using (113) finding
Example 74. [Rolling Disc] A uniform disc of
mass M and radius a rolls without slipping down a
1 1
line of greatest slope of a plane inclined at an angle aF = M a2θ̈ = M aẍ,
α to the horizontal. Determine its acceleration 2 2
and the reaction forces.
and so
ẍ = 23 g sin α,
1 2 a constant value, but smaller than that of a
2 M a θ̈
particle sliding down the plane. Also,
G
M ẍ F F = 13 M g sin α, N = M g cos α.
N
D
Mg Notice that F/N = 13 tan α and so the friction
α
coefficient must exceed this value if no slipping is
to occur.
The centre of gravity moves with acceleration ẍ
parallel to the plane. Because there is no slipping
the angle turned is given by θ = x/a. The mass
accelerations and forces are shown in the figure.

One way to obtain F and N is by resolving forces


parallel to and perpendicular to the plane:

M g sin α − F = M ẍ, M g cos α − N = 0.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 247 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 248
that GC is vertical and downwards when the
Example 75. [Rolling cylinder] A uniform solid cylinder is in its lowest position, and let ϕ be the
circular cylinder, of mass M and radius a, rolls angle between GC and the downward vertical in
without slipping on the inside of a fixed hollow a general position. To establish the no-slipping
circular cylinder of radius a + b. The axes of the condition notice that the distance between C
cylinders are horizontal and parallel. Investigate and D is both a(θ + ϕ) and (a + b)θ, whence
the motion, and find the least coefficient of friction aϕ = bθ. The forces on the cylinder are gravity
if the motion starts from rest with the point of and reaction forces N and F as in the previous
contact at an angular distance α < 12 π from the example. The cylinder has a clockwise angular
lowest point of the fixed cylinder. acceleration ϕ̈ about G, and the centre of mass
has a tangential acceleration bθ̈ and a radial
O acceleration bθ̇ 2 towards O.
θ Resolving forces in the tangential and radial
directions we obtain

F −M g sin θ = M bθ̈, N −M g cos θ = M bθ̇ 2,


G
C and taking moments about G gives
ϕ
θ
1 1
D aF = − M a2ϕ̈ = − M abθ̈,
2 2

so that
Denote by θ the angle OG makes with the vertical. 2g
Let C be a fixed point on the small cylinder, such θ̈ + sin θ = 0,
3b
University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 249 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 250

with first integral


Example 76. [Backspin] A sphere of mass M ,
4g
θ̇ 2 = (cos θ − cos α). radius a, is launched with initial speed V and zero
3b angular velocity on a rough horizontal surface
with friction coefficient μ. Show that slipping
It is now easy to show that
stops after time t = 2V /(7μg), and that the final
5
kinetic energy is 14 M V 2.
F = 13 M g sin θ, N = 13 M g(7 cos θ − 4 cos α).

Now θ̇
F sin θ
= Mg
N 7 cos θ − 4 cos α
F R
is an increasing function of θ and so achieves its
greatest value, 13 tan α at θ = α.
There is another way to view this problem. There is no vertical motion and so the reaction of
The constraint forces act at point D which is the surface R = M g. Thus while slipping occurs
instantaneously at rest. Therefore they do no there is a frictional force F = μR = M μg. If x
work and energy is conserved. The translational is the horizontal displacement, Newton’s second
kinetic energy is 12 M (bθ̇)2 and the rotational law implies M ẍ = −F , and an easy integral gives
kinetic energy is 12 ( 12 M a2)ϕ̇2, so that the total ẋ = V − μgt. Using example 66, and equating
kinetic energy is 34 M b2θ̇ 2. The potential energy the rate of change of angular momentum to the
is U = −M gb cos θ, and we easily obtain the frictional torque, 25 M a2θ̈ = aF , and integration
equation for θ̇ 2 given above. gives aθ̇ = 52 μgt. Clearly ẋ is decreasing while
aθ̇ is increasing and they become equal at time
t = 2V /(7μg). Thereafter slipping stops, F → 0

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 251 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 252
and we get pure rolling with ẋ = aθ̇ = 57 V . Thus 10. Phase plane techniques
the final kinetic energy is

1 12 5 10.1 Introduction
T = M ẋ2 + M a2θ̇ 2 = M V 2.
2 25 14
In chapter 4 we looked at systems with one degree
of freedom, described by
Example 77. [Backspin continued] As above,
but the sphere has an initial angular velocity −Ω. ẋ2 = f (x), (129)
Show that the ball returns, rolling, to its initial
position if 5V < 2aΩ < 12V and determine its and we showed that equilibrium at x = a requires
final speed. f (a) = f (a) = 0. Further for small perturbations
x = a + (t),
Again we have ẍ = −μg, aθ̈ = 52 μg while slipping
occurs and so
¨ − 12 f (a) = O(2). (130)
ẋ = V −μgt, x = V t− 12 μgt2, aθ̇ = −aΩ+ 52 μgt.
This equation has the first integral
So if the ball is slipping it reverses velocity at
t1 = V /μg) and returns to x = 0 at time t3 = ˙2 − 12 f (a)2 = ΔE + O(3), (131)
2V /(μg). If it is to return rolling, then the
transition from slipping to rolling (when ẋ = aθ̇) where ΔE is a constant. If the equilibrium is
at t2 = 27 (V + aΩ)/(μg) must satisfy t1 < t2 < t3 stable then f (a) < 0. The solution curves of
which implies the inequality above. The final (131) are small ellipses around x = a, ẋ = 0
speed is of course the transition speed ẋ(t2) = in the xẋ-plane. This is a centre in standard
−(2aΩ − 5V )/7. ordinary differential equation (ODE) notation.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 253 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 254


10.2 Plane autonomous systems

Equation (129) in the form ẋ = f (x), ẍ =
1 
2 f (x) is a special case of a plane autonomous
system (PAS)
a x
ẋ = F (x, y), ẏ = G(x, y). (132)

A fixed point is a point (x0, y0 ) such that


If the equilibrium is unstable then f (a) > 0, and F (x0, y0) = 0 = G(x0 , y0).
the solution curves are hyperbolae.
Theorem 20. A solution curve of a PAS can
ẋ start, end or self-intersect only at a fixed point.

Proof. Equation (132) implies

x dy ẏ G(x, y)
= = .
dx ẋ F (x, y)

Thus if at most one of F or G vanishes at a point,


there is a well-defined tangent direction (which
In ODE notation this is a saddle. may be horizontal or vertical) and the curve can
be continued. This fails at a fixed point. 2

It is an immediate corollary that a solution curve


of (129) can self-intersect only at a saddle. A

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 255 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 256
solution curve through a saddle point is called a
separatrix. The set of separatrixes separates
classes of solution curves with different global E
3
behaviour. 2.5
2
1.5
1
0.5
0
-0.5
-1
2
1.5
1
0.5
-2 -1 0
0
-0.5 θ̇
1 2 -1
3 -1.5
θ 4 5 -2
Example 78. [The simple pendulum] With a
standard notation we have

E = 12 m
2θ̇ 2 + U (θ),

where U (θ) = −mg


cos θ. For simplicity we
choose units to set m = g =
= 1. We can However along the solution curves E = const.
interpret this equation as saying E = E(θ, θ̇) and Thus it is more informative to draw the contour
drawing this a surface above the θ θ̇-plane, the lines E = const. of the surface, for these are the
phase plane. solution curves.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 257 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 258

2
θ̇ 2

1.5

1 1

θ̇ 0.5

0 0

-0.5
−1
-1

-1.5
−2
-2
-6 -4 -2 0 2 4 6
θ
−3
−6 −4 −2 0 2 4 6
θ
Curves with −1 < E/(mg
) < 1 lie inside the
Notice that the points θ = 2nπ, θ̇ = 0 correspond separatrixes and are bound, i.e. ∃ θ1, θ2 such that
to stable equilibrium with E = −mg
, and appear θ1 < θ < θ2, and the motion librates, i.e., θ̇
as centres on the phase plane. However the points changes sign regularly. Curves with E > mg

θ = (2n + 1)π, θ̇ = 0 corresponding to E = mg


lie outside the separatrixes and are unbound.
are unstable equilibria and appear as saddles in (They correspond to a rotating pendulum.) Thus
the phase plane. the separatrix curves separate solution curves
corresponding to radically different behaviour.
The curves E = mg
define the separatrixes.
Note that one can even describe a damped
These curves are defined by
pendulum in the phase plane diagram of an
undamped one. This will be a curve on which
1 2 2
2 m
θ̇ − mg
cos θ = mg
,
E decreases and so it crosses the contour lines,
ending up at a stable centre.
or 
g
θ̇ = ±2 cos( 12 θ).

Example 79. [Anharmonic oscillator] We consider

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 259 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 260
a one-dimensional system with energy given by E = E0 = 14 ω 4/A2. This separatrix
√ intersects the
ẋ-axis at x = 0, ẋ = ±ω 2/( 2A)
√ intersects
and
E = 12 ẋ2 + U (x), the x-axis again at x = x1 = 2 − 1(ω/A),
ẋ = 0.
where
Here is the phase portrait.

1 2 4
U (x) = 12 ω 2x2 + 4A x x < 0,
− 14 A2x4 x > 0, 1
Phase portait for anharmonic oscillator

0.8

where A > 0. (The case A = 0 is the harmonic 0.6

oscillator.) 0.4

0.2

xdot
0
0.5
−0.2
0.4
−0.4
0.3 E0
−0.6
0.2
−0.8

0.1
−1
−1 −0.5 0 0.5 1 1.5
x
0
x1 0 x0
-0.1

-0.2
-1 -0.5 0 0.5 1 1.5
Notice that orbits outside the separatrix
√ are
unbounded and as x → ∞, ẋ ∼ ±Ax2/ 2.
We have stable equilibrium with a centre at x =
0 = ẋ. There is an unstable equilibrium at x = Note once more that stable equilibria correspond
x0 = ω/A, ẋ = 0, which corresponds to an energy to centres in the phase portrait, while unstable

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 261 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 262

equilibria correspond to saddles which may be (f (x, y), g(x, y)) has an isolated zero, then it is
joined by separatrixes. SS. We can construct a neighbourhood of the zero
in which the curves f (x, y) = 0 and g(x, y) = 0
In order to discuss the general case we need to
intersect transversally, and this is obviously SS.
introduce the idea of structural stability. A
precise definition involves serious mathematics,
and so we give a heuristic version, together with 5. If a smooth function f (x, y) has an isolated
examples. Basically if a system has a feature stationary point then it is SS. (Use item 4 applied
which remains invariant under small continuous to ∇f .)
perturbations of the system, then the feature is
structurally stable (SS). 6. A zero eigenvalue of a matrix is not SS. Consider
 
1. If a smooth function f (x) has an isolated simple 1 1
a=
zero then this property is SS. There is an interval 1 1+
such that the value of f changes sign at the
two endpoints, and this is preserved by small with eigenvalues
perturbations.

1
2. The property that f has an isolated double root λ=1− ± 1 + 14 2.
2
is not SS. Consider e.g., f (x) = x2. Adding a
small constant to f produces 0 or 2 real zeros.

3. The property that f has an isolated stationary Now we are interested in the stationary points of
point is SS. Simply apply item 1 to f (x). the surface E = E(x, ẋ). These are of course SS
by item 5. Further their type is determined by
4. If a smooth function F : R2 → R2, F (x, y) = the signs of the eigenvalues of the hessian matrix

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 263 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 264
of E. For a SS system they must be nonzero 10.3 Limit cycles
by item 6. If they both have the same sign we
have a maximum or a minimum, and because In the case of general systems, a closed periodic
of the nature of the kinetic energy term, only a curve is a limit cycle.
minimum can occur. If they have oppposite signs
then we have a saddle. Thus the analysis we have Example 80. Consider the system defined by
given applies to any SS system with one degree
of freedom.
ṙ = cr(r − 1), θ̇ = 1,
In particular we can analyze any smooth system
with one degree of freedom, e.g. the system with where c is a constant. Clearly r(t) = 1, θ(t) = t
the potential energy as shown. defines a limit cycle. If c > 0 then it is unstable
for ṙ > 0 if r > 1 and ṙ < 0 if r < 1. By a similar
U argument it is stable if c < 0.
D
B Another, almost trivial, example are the simple
pendulum orbits which are bound. They are
C neutrally stable.
x
A
Motion will be possible only for small |x| when
x < 0. There will be centres at A and C, and
saddles at B and D and the D-separatrix encloses
the B-one. Sketching the phase portrait is left as
an exercise.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 265 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 266

10.4 The not-so-simple pendulum The equation of motion then becomes

Suppose S is an inertial frame with coordinates


θ̈ + k θ̇ + (1 + aω 2 cos ωt) sin θ = 0. (133)
x, y and z. Consider a second frame S with
coordinates x = x, y  = y and z  = z + h(t).
because of gravity a particle in S feels an The extra a-term can change completely the
acceleration z̈ = −g. However a particle in S dynamics. Note that because of the t-dependence
feels z̈  = −g − ḧ. E.g., if S is freely falling, this is not a system with one degree of freedom,
h = vt − 12 gt2, then particles in S experience no even though it is one-dimensional. Nor is it
vertical acceleration. In this frame there is no conservative. The k-term takes energy out of
effective gravity. the pendulum. The a-term can add or subtract
energy, sometimes both in the same motion.
We now want to consider a simple pendulum of
length
whose pivot is oscillating in a vertical The conventional picture suggests that after fixing
direction z = A cos(ΩT ) where, for the moment, the parameters k, a and ω, and specifying the
T is the time coordinate. Thus the effective initial data θ = θ0 and θ̇ = ω0, then the solution
gravity is g + AΩ2 cos(ΩT ). Allowing for friction, is determined.
the equation of motion is
Consider first the case where k = 0 and θ is small
d2 θ dθ 1   so that we can replace sin θ by θ. Choosing a new
2
+K + g + AΩ2 cos(ΩT ) sin θ = 0. dimensionless time coordinate τ = ωt we have
dT dT

For further study it is sensible to use dimensionless d2 θ


+ (α + β cos τ )θ = 0,
variables dτ 2
T Ω A K where α = ω −2 and β = a. This is Mathieu’s
t= , ω= , a= , k= .

/g g/

g/
equation which has been extensively studied. In

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 267 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 268
particular it is known that for certain values of α this case stable!
and β the exact solution θ(t) ≡ 0 is unstable.

Figure 1: a = 0.2, omega = 10, theta0 = pi, omega0 = 0.01


0.02

0.015

0.01

0.005

thetadot
0

The shaded region corresponds to values of α and −0.005

β for which the exact solution θ(t) ≡ 0 is unstable.


Note in particular there is a region of instability −0.01

near α = 14 , β
1, which corresponds to small −0.015
a and ω being twice the natural frequency of the
pendulum. Then θ = 0 is unstable; an initially −0.02
3.13 3.135 3.14 3.145 3.15 3.155
theta
small disturbance will lead to a swinging motion
of steadily increasing amplitude.
Let us now return to the general case. Figure 1
shows the behaviour of a pendulum starting from Figure 2 corresponds to the same data, except that
θ = π, θ̇ = 0.01, with a = 0.2 and ω = 10 for θ̇ = −0.01 initially, and we are using a different
0  t  200π. The upright position θ = π is, in value for ω.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 269 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 270

which never self-intersects. What is being shown


is the projection onto a t = const. plane.
Figure 2: a = 0.2, omega = 7, theta = pi, omega = −0.01
0 0
3

Figure 3: a = 0.1, omega = 2, theta = 0.01, omega = 0


0 0
2
1.5

1 1
thetadot

0 0.5
thetadot

−1 0

−2 −0.5

−3 −1
−2 −1 0 1 2 3 4
theta

−1.5
−1.5 −1 −0.5 0 0.5 1 1.5
Now the pendulum drops to θ = 0 and oscillates theta

wildly before dissipation causes the solution to


settle at θ = 0. Both of these pictures are Figures 3 and 4 illustrate realistic limit cycles. In
completely different to the phase portrait of the figure 3 the bob is released from rest at θ = 0.01,
simple pendulum discussed earlier. The orbit but in this case θ = 0 is unstable equilibrium. The
appears to self-intersect but this is just an illusion. pendulum oscillates periodically about its lowest
In fact the orbit is the curve (t, θ(t), θ̇(t)) in R3 position but never comes to rest. Energy is being

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 271 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 272
pumped in at the pivot at a rate which balances
the frictional dissipation. The pendulum has one more secret to reveal.
When we solve a system like

Figure 4: a = 0.45, omega = 5, theta = pi, omega = 1.3


2
0 0
θ̈ + k θ̇ + sin θ = 0, θ(0) = θ0, θ̇(0) = ω0,

1.5
we know that the solution θ(t, θ0, ω0) is a
1 continuous function not only of t but also of
the initial data θ0 and ω0. This continuity is
0.5
essential for solving differential equations on a
thetadot

0
computer because floating point numbers (R\Z)
can only be stored approximately. If we store say
−0.5 8 significant figures we expect to get a reasonable
approximation to the exact solution.
−1

−1.5 However certain non-linear systems are chaotic,


i.e.,
−2
2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4 4.2
theta

1. the solution curves are very sensitive to the initial


In figure 4 the bob is released from θ = π with conditions,
θ̇ = 1.3 and moves to an elaborate but stable
limit cycle.
2. they never settle into an equilibrium or repeating
pattern.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 273 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 274

digit accuracy the evolution of such a pendulum


would be essentially random and unrepeatable. Of
Figure 5: a = 0.5, omega = 1.1, theta0 = pi +0.00001 x n, omega0 = 0
2.5
course this is only a toy model, but prediction is a
serious problem for weather forecasters, for their
2
equations are genuinely chaotic
1.5

How can one tell whether a system is chaotic?


1
One necessary condition is that the equations be
0.5 non-linear. A second such condition is that the
thetadot

0 phase space must have more than two dimensions.


A PAS cannot exhibit chaos. Suppose that in a
−0.5
PAS, a phase path starts at (x0, y0) and cannot
−1
leave a bounded area in phase space. Then the
−1.5 Poincaré–Bendixson theorem asserts that
−2
the phase orbit must eventually either

−2.5
−10 −5 0 5
theta
10 15 20 25 1. terminate at an equilibrium point,

2. return to (x0 , y0) giving a closed path,


Figure 5 shows a pendulum with a = 0.5, ω = 1.1,
ω0 = 0 and θ0 = π + n × 10−5 where n = 1, 2, 3, 4 3. approach a limit cycle.
corresponding to the black, blue, green and red
curves respectively. Clearly the orbits stay close
None of these possibilities is chaotic.
only for a short time. If one continues beyond
t = 200π the orbits wiggle and loop with no These conditions are however not sufficient, and
obvious pattern. Thus if my computer had only 4 there is a very simple example, extending chapter

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 275 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 276
9 to demonstrate this. Consider a rigid body, Exercise 16. Show that the kinetic energy T
centroid O, subject to no external forces. Then and the norm (squared) of the angular momentum
by theorem 16 the rate of change of the angular
momentum H0 with respect to an inertial frame
S is zero,   1
T = (Aω12 + Bω22 + Cω32),
dH0 2
= 0.
dt S |H|2 =A2ω12 + B 2ω22 + C 2ω32,
By theorem 12, the rotating axes theorem, as
seen in a frame S fixed in the body (with suffix are both constants of the motion.
S suppressed)
You can study chaos further in the Part II and
Ḣ0 + ω × H0 = 0. III courses on dynamical systems, and the lecture
schedules give useful references. The not-so-
Choose S so that the inertia tensor is diagonal, simple pendulum is based on some exercises in
IO = diag(A, B, C), and set ω = (ω1, ω2, ω3)T .
Then H0 = (Aω1, Bω2, Cω3) and • D. Acheson, From calculus to chaos: an
introduction to dynamics, Oxford, 1997,
Aω̇1 =(B − C)ω2ω3,
B ω̇2 =(C − A)ω3ω1, which also cites some more popular references.
C ω̇3 =(A − B)ω1ω2.

These are clearly 3-dimensional and nonlinear, but


they cannot exhibit chaos.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 277 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 278

Hamiltonian dynamics For each q1 the corresponding generalized


momentum is
Most of this course has been pretty similar to
A-level school courses. In reality dynamics is ∂L
pi = .
more sophisticated than this. Suppose we have ∂ q̇i
a system which can be described by coordinates
and angles. We unite them in the concept of Finally we form the hamiltonian function
generalized coordinates traditionally denoted
q1, q2, . . . , qn Suppose we can form the kinetic
n

energy and potential energy H(t, q, p) = piq˙i − L,


i=1

T =T (t, q1, q2, . . . , qn, q̇1, q̇2, . . . , q̇n)


where the generalized velocities are eliminated in
U =U (t, q1, q2, . . . , qn, q̇1, q̇2, . . . , q̇n). terms of the generalized momenta. Then the
equations of motion of the system, Hamilton’s
We then form the lagrangian function equations take the beautiful form

L(t, q, q̇) = T − U. dqi ∂H dpi ∂H


= , =− .
dt ∂pi dt ∂qi
There is a powerful theory of dynamics based on
the lagrangian function, but I will not describe In almost all situations H = T + U , the total
that here, because there is a brilliant stimulating energy, so that for any system if you can write
account in the Feynman lectures in physics, down the total energy, eliminate the velocities
volume 2, chapter 19, which will also be covered in terms of the momenta, then you have the
in the Part IB methods course. Instead I want to equations of motion of the system.
look at the hamiltonian extension.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 279 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 280
The first set of Hamilton’s equations dqi/dt =
Example 81. [Hamiltonian derivation of CFP equatio ∂H/∂pi are uninteresting (in this very simple
We study a system for which the kinetic and example). They assert ṙ = pr /m and θ̇ =
potential energies are given by pθ /(mr 2). However the second set has real
content.
1
T = m(ṙ 2 + r 2θ̇ 2), U = −GM m/r. Firstly dpr /dt = −∂H/∂qr asserts mr̈ = mr θ̇ 2 −
2
GM m/r 2. This is true, and reminds us that this
We set q1 = qr = r, q2 = qθ = θ, so that the formalism knows all about curvilinear coordinate
lagrangian is systems.

1 Secondly dpθ /dt = 0 since H is independent of


L = m(q̇r2 + qr 2q̇θ2) + GM m/qr . qθ = θ.
2
More generally we note that if H does not
Then pr = ∂L/∂ q̇r = mq̇r = mṙ, the depend on qi then pi is a constant. This is a
conventional momentum. However pθ = simple instance of Noether’s theorem: every
∂L/∂ q̇θ = mqr 2q̇θ = mr 2θ̇, the angular symmetry implies a conserved quantity.
momentum. (This is why p is a generalized
momentum.)
Here is another example. Systems with one degree
Next we form H = T + U as a function of the q’s of freedom do not involve the time coordinate
and p’s, explicitly, and so they are invariant under time
translations. This is a symmetry and the
pr 2 pθ 2 GM m conserved quantity is energy. Further examples
H= + − . occur in the Dynamics and General Relativity
2m 2mqr 2 qr
courses at Part II and beyond.

University of Cambridge: Mathematical Tripos IA: Dynamics: 


c J.M. Stewart 2002 281 University of Cambridge: Mathematical Tripos IA: Dynamics: 
c J.M. Stewart 2002 282

You might also like