You are on page 1of 75

Level 3 Classical Mechanics 2011

Durham University Physics


Contents
1 Generalized Description of Mechanical Systems 1
1.1 Validity of Newtons Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Dynamical Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Degrees of Freedom (DoF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4 Types of Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 Generalized coordinates q
k
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.6 Generalized Velocities q
k
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.7 Generalized forces T
k
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
HOMEWORK 1: LAGRANGIAN OF A SMALL BLOCK ON AN INCLINED PLANE 3
2 The Lagrangian 5
2.1 Kinetic Energy T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Generalized Equations of Motion (EoM) . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Conservative Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 The Lagrangian L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Hamiltons Principle of Least Action . . . . . . . . . . . . . . . . . . . . . . . . 6
EXAMPLE CLASS 1: APPLICATIONS OF THE LAGRANGIAN APPROACH 7
3 Linear Oscillators 9
3.1 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Lagrangian Near Static Equilibrium (1 DoF) . . . . . . . . . . . . . . . . . . . . 9
3.3 Simple Harmonic Oscillator (SHO) . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Damping Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.5 Damped Simple Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . 10
3.6 Overdamped SHO: Q <
1
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.7 Underdamped SHO: Q >
1
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.8 Critically Damped SHO: Q =
1
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
EXAMPLE CLASS 2: DAMPED, DRIVEN PENDULUM 12
4 Driven Oscillators 13
4.1 Oscillator Driven by an External Force . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 Dirac -Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3 Response to a q-Independent Impulsive Force . . . . . . . . . . . . . . . . . . 13
4.4 Time-Evolution Sequence of a SHO with an Impulsive Force . . . . . . . . . . 14
4.5 (Causal) Greens Function G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
i
4.6 General Driving Force F(t) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.7 Driving an Oscillator in Resonance . . . . . . . . . . . . . . . . . . . . . . . . . 15
5 Small Oscillations 16
5.1 Coupled Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.2 Small Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.3 Setting up the Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
HOMEWORK 2: DOUBLE PENDULUM 17
6 Normal Modes 20
6.1 Matrix Form of L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.2 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.3 Working Towards a Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.4 Normal Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7 Central Forces 22
7.1 Two Interacting Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7.2 Translational Symmetry and Conservation of Momentum . . . . . . . . . . . . 22
7.3 Ignorable Centre of Mass (CoM) Motion and Conservation of Momentum . 23
7.4 Rotational Invariance and Conservation of Angular Momentum . . . . . . . 23
7.5 Motion Takes Place in a Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
7.6 Equal Areas are Swept out in Equal Times (Keplers Second Law) . . . . . . . 24
7.7 Equivalent 1D Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
HOMEWORK 3: DIATOMIC MOLECULE 25
8 Gravitational Attraction 26
8.1 Solving 1D Systems by Quadrature . . . . . . . . . . . . . . . . . . . . . . . . . 26
8.2 Gravitational Interaction Potential . . . . . . . . . . . . . . . . . . . . . . . . . . 26
8.3 The u = 1/r Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
8.4 Elliptical Orbits (Keplers First Law) . . . . . . . . . . . . . . . . . . . . . . . . . . 27
8.5 Keplers Third Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
8.6 Keplers Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
9 Noethers Theorem and Hamiltonian Mechanics 30
9.1 Invariance of L under Continuous Transformations . . . . . . . . . . . . . . . . 30
9.2 Noethers Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
9.3 Legendre Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
9.4 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
9.5 Hamiltons Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
9.6 Hamiltons Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
HOMEWORK 4: CANONICAL TRANSFORMATION OF THE SHO 32
10 Canonical Transformations and Poisson Brackets 33
10.1 Canonical Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.2 Equivalence of Lagrangians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.3 The Generating Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
10.4 The Transformed Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
ii
10.5 Four Forms of Generating Function . . . . . . . . . . . . . . . . . . . . . . . . . 35
10.6 Poisson Brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
10.7 Poisson Bracket Formulation of Hamiltons Equations . . . . . . . . . . . . . . . 36
11 Rotating Reference Frames 37
11.1 Accelerating Reference Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
11.2 Rotated Reference Frames in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . 37
11.3 Finite Rotations in 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
11.4 Innitesimal Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
11.5 Velocity and Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
11.6 Inertial Forces in a Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . . . . 39
EXAMPLE CLASS 3: FOUCAULTS PENDULUM 40
12 Inertial Forces on Earth 41
12.1 Procedure for Determining a Local Coordinate Systems . . . . . . . . . . . . 41
12.2 Inertial Forces in the Local Coordinate System . . . . . . . . . . . . . . . . . . 41
12.3 Centrifugal Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
12.4 Effective Gravity on Earths Surface . . . . . . . . . . . . . . . . . . . . . . . . . 42
12.5 Deviation of a Plumb Bob . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
12.6 Coriolis Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
12.7 Loose Ends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
13 Rotational Inertia, Angular Momentum and Kinetic Energy 44
13.1 DoF of Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
13.2 Displacement of a Rigid Body . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
13.3 Moment of Inertia I about an Axis of Rotation . . . . . . . . . . . . . . . . . . 44
13.4 Angular Momentum J and Rotational Kinetic Energy . . . . . . . . . . . . . . 44
13.5 Denition of the Inertia Tensor

I . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
13.6 Generalized Description of the Rotational Kinetic Energy . . . . . . . . . . . . 46
13.7 Centre of Mass (CoM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
13.8 Separation off of the CoM Angular Momentum and Kinetic Energy . . . . . . 46
HOMEWORK 5: ROTATING BOOK 47
14 The Parallel and Principal Axis Theorems 48
14.1 Displaced Axis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
14.2 Parallel Axis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
14.3 Symmetric Nature of the Inertia Tensor . . . . . . . . . . . . . . . . . . . . . . . 49
14.4 Orthogonal Matrices and Orthogonal Diagonalization . . . . . . . . . . . . . 49
14.5 Principal Axis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
14.6 Deducing the Principal Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
15 Rigid Body Dynamics and Stability 52
15.1 Eulers Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
15.2 Torque-Free Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
iii
A Second-Order Ordinary Linear Differential Equations with Constant Coefcients I
A.1 Denition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2.1 Auxiliary Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2.2 General Solution (Roots not Repeated) . . . . . . . . . . . . . . . . . . I
A.3 Possible Complications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
A.3.1 Repeated Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
A.3.2 Additional Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
A.4 Representing Position in a Plane by a Single Complex Number . . . . . . . . II
B Taylor Series III
B.1 Taylors Theorem of the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
B.2 Taylor Series in One Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
B.3 Taylor Series in Two Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV
B.4 Taylor Series in Three Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV
C Determinants, Eigenvalues and Eigenvectors V
C.1 Symmetric Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V
C.2 Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V
C.3 The Conventional Eigenvalue Problem . . . . . . . . . . . . . . . . . . . . . . . V
C.4 Generalized Eigenvalue Problem . . . . . . . . . . . . . . . . . . . . . . . . . . VII
D Exam Questions and Solutions VIII
D.1 Coriolis Force (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII
D.1.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII
D.1.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII
D.2 Small Oscillations (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX
D.2.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX
D.2.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X
D.3 Stable and Unstable Rigid-Body Rotation (2007) . . . . . . . . . . . . . . . . . X
D.3.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X
D.3.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI
D.4 Diatomic Molecule (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XII
D.4.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XII
D.4.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
D.5 Normal Coordinates (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
D.5.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
D.5.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIV
D.6 Block in a Bowl (2009, Question Only) . . . . . . . . . . . . . . . . . . . . . . . . XV
D.7 Atom Falling through a Laser Standing Wave (2009, Question Only) . . . . . XV
D.8 Ice Sliding Between Glass Plates (2010, Question Only) . . . . . . . . . . . . . XVI
D.9 Double Pendulum (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVII
D.9.1 Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVII
D.9.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XVIII
iv
1 Generalized Description of Mechanical Systems
SEE ALSO HAND AND FINCH, 1.11.6
1.1 Validity of Newtons Laws
Newtons second law
F = p (= m v = m r for constant m) (1)
is valid in an inertial (non-accelerating) frame for objects moving at much less than
light speed, and with dimensions much larger than atoms or molecules.
1.2 Dynamical Variables
These are any set of variables that describe completely the conguration (positions
of parts) of a mechanical system.
They are usually positions and/or angles, e.g. cartesian, cylindrical, or spherical co-
ordinates for a point mass in free space.
They can change under the action of forces the motion of a system is determined
by specifying the dynamical variables as functions of time.
Newtons laws imply that these functions are found by solving second order differen-
tial equations hence the initial positions/angles and velocities/angular velocities
(or equivalent information) must be known.
1.3 Degrees of Freedom (DoF)
The motion of a point mass r(t) is described by 3 independent dynamical variables,
i.e., the point mass has N = 3 degrees of freedom (DoF).
A system of M point masses has N = 3M DoF, but the existence of j independent
constraints reduces this number to N = 3M j DoF.
M=3
j=3
M=3
j=2
(flexible joint)
M=2
j=1
N=5 DoF N=6 DoF N=7 DoF
E.g., point masses connected by rigid massless rods. N.b. Any rigid body more com-
plicated than 2 point masses connected by a rigid rod has 6 DoF (or fewer, if there
are additional constraints).
1
1.4 Types of Constraints
If constraints can be expressed in the form
f (r
1
, r
2
, . . . , r
M
, t) = 0 (e.g. constraints on M point masses), (2)
the constraints are called holonomic.
Such constraints may have an explicit time dependence: Rheonomic (running law)
constraints. If not [i.e., f (r
1
, r
2
, . . . , r
M
) = 0]: Scleronomic (rigid law) constraints.
Nonholonomic constraints involve:
velocities (rolling without slipping on a surface),
inequalities (particle sliding down the outside of a bowling ball),
most velocity dependent forces (friction).
1.5 Generalized coordinates q
k
The existence of j constraints means that the (cartesian) coordinates of e.g. M point
masses are no longer independent.
We require only as many coordinates q
k
as there are DoF (N = 3M j).
We can express the position r
i
of each part of a system as
r
1
= r
1
(q
1
, . . . , q
k
, . . . , q
N
, t),
.
.
.
r
i
= r
i
(q
1
, . . . , q
k
, . . . , q
N
, t),
.
.
.
r
M
= r
M
(q
1
, . . . , q
k
, . . . , q
N
, t),
(3)
if the constraints are holonomic. There may be an explicit time dependence partic-
ularly if the constraints are rheonomic.
1.6 Generalized Velocities q
k
The q
k
are the rst time derivatives of q
k
.
We regard q
k
and q
k
as independent variables (this is very important, and means for
example that q
k
/ q
k
= q
k
/q
k
= 0).
By the chain rule, one can derive the following identity, which can be a useful check:
v
i
= r
i
=

k
r
i
q
k
q
k
+
r
i
t

r
i
q
k
=
r
i
q
k
cancelling the dots,
(4)
2
1.7 Generalized forces T
k
If we freeze the motion of a mechanical system at a point in time, and displace
one DoF q
k
an innitesimal amount q
k
, the position of part r
i
of the system will
change by r
i
=
r
i
q
k
q
k
.
Due to nonconstraint forces, virtual work is done by this virtual displacement. If
each q
k
is innitesimally changed, the total virtual work W is the sum of the
virtual work by each generalized coordinate variation:
W =

i
F
i
r
i
=

i
F
i

k
r
i
q
k
q
k
=

k
_

i
F
i

r
i
q
k
_
. .
Generalized forces T
k
=
W
q
k
q
k
. (5)
Homework 1: Lagrangian of a Small Block on an Inclined Plane
X
Y
M
m

frictionless surfaces
Consider the situation of a small block (SB) of mass m sliding on a frictionless inclined plane (IP)
of height h, which itself has a mass M and rests on a at surface, without any friction between
this surface and the inclined plane. Consider a set of cartesian axes such that the Y axis points
up along the side of the initial location of the inclined plane, and the X axis points along the
underside of the inclined plane (we ignore the Z direction). Take (X
SB
, Y
SB
) and (X
IP
, Y
IP
) to be the
locations of the centre of mass of the SB, and the right-angled corner of the IP in this coordinate
system, respectively.
1. (Study up to section 1.4). Because the motion of the SB is constrained to remain on the upper
surface of the IP, the value for the dynamical variable Y
SB
can be determined exactly if the
values for the dynamical variables X
SB
, X
IP
are known. Derive an equation (a constraint
equation) describing this relationship.
2. (Study up to section 1.5). Consider an alternative dynamical variable d, the distance of the SB
centre of mass from the top of the IP. Determine expressions for X
SB
, Y
SB
in terms of X
IP
, d, h
and . (Study up to section 1.6). Taking the time-derivative, determine equivalent expressions
for

X
SB
,

Y
SB
.
3
3. (Study up to section 2.1). Trivially, the IP is constrained not to move in the Y direction, and so
Y
IP
and

Y
IP
can be neglected. (Study up to section 2.3). Determine an expression for the total
kinetic energy T of the SB and IP system in terms of

X
IP
and

d. Determine an expression for
the total potential energy V in terms of X
IP
and d.
4. (Study up to section 2.4). Write down the Lagrangian L using your expressions for T and
V. From the Lagrangian, derive Lagranges equations of motion for d(t) and X
IP
(t). Solve
the equations of motion, writing the solutions in terms of the initial positions and velocities
d(0), X
IP
(0),

d(0),

X
IP
(0).
5. From these solutions, derive an expression for the distance the IP has moved once the SB
has reached the bottom of the IP, and hence (assuming = /4 radians, that the small
block starts at the top of the inclined plane, and that both the small block and the inclined
plane are initially at rest) determine how large m must be relative to M for this distance to
be = h/2.
4
2 The Lagrangian
SEE ALSO HAND AND FINCH, 1.72.5
2.1 Kinetic Energy T
The most general form for a holonomic system is
T =
1
2

i
m
i
r
i
r
i
= T(q
1
, . . . , q
N
, q
1
, . . . , q
N
, t). (6)
T may have an explicit time dependence, particularly (although not necessarily)
for rheonomic constraints.
2.2 Generalized Equations of Motion (EoM)
From the chain rule
T
q
k
=

i
m
i
r
i

r
i
q
k
=

i
p
i

r
i
q
k
, (7)
T
q
k
=

i
m
i
r
i

r
i
q
k
=

i
p
i

r
i
q
k
[using Eq. (4)]. (8)
From the product rule, the total time derivative
d
dt
_
T
q
k
_
=

i
p
i
..
F
i

r
i
q
k
. .
T
k
+

i
p
i

r
i
q
k
. .
T
q
k
. (9)
Hence, the generalized equations of motion, below, follow:
T
k
=
d
dt
_
T
q
k
_

T
q
k
. (10)
2.3 Conservative Forces
If the nonconstraint forces are conservative, then by denition a potential energy
function V exists.
A nonconstraint force on the ith point mass of the system is derived from such a
potential energy function by F
i
=
i
V(r
1
, . . . , r
i
, . . . , r
M
). Correspondingly,
T
k
=

i
V
r
i
q
k
=
V
q
k
(follows from the chain rule). (11)
5
2.4 The Lagrangian L
L = T V (assumes V to be velocity independent).
Substituting T = L +V into the generalized equations of motion [Eq. (10)], we obtain
the Euler-Lagrange equations:
d
dt
_
L
q
k
_

L
q
k
= 0. (12)
These are the fundamental equations of Lagrangian mechanics.
2.5 Hamiltons Principle of Least Action
Alternatively, we can derive the Euler-Lagrange equations by minimizing
(strictly speaking extremizing) the action functional S. For a single generalized
coordinate q
S[q] =
_
t
f
t
i
L(q(t), q(t), t)dt. (13)
[The argument of a function is a number, and a number is returned the ar-
gument of a functional is a function, and a number is returned. (A functional
usually involves an integral)].
Consider variations of S due to variations of q(t) with the end points q(t
i
), q(t
f
)
xed.
S = S[q + q] S[q] =
_
L(q + q, q + q, t)dt
_
L(q, q, t)dt. (14)
Hamiltons principle asserts the physical path is that which minimizes (strictly
speaking extremizes) S, i.e., such that S = 0.
We Taylor expand the variation of L at a xed value of t:
L(q + q, q + q, t) = L(q, q, t) +
L
q
q +
L
q
q + higher-order terms in q, q.
(15)
Hence, for innitesimal variations, it follows that S =
_
_
L
q
q +
L
q
q
_
dt.
We Integrate the second term of S by parts
_
note q =
d(q)
dt
_
:
_
L
q
d(q)
dt
dt =
L
q
q

t
f
t
i
. .
vanishes no variations at end points

_
d
dt
_
L
q
_
qdt. (16)
6
Hence, as q is an arbitrary innitesimal variation, we derive the Euler-Lagrange
equations:
S =
_
_
L
q

d
dt
_
L
q
__
qdt implies S = 0
if and only if
d
dt
_
L
q
_

L
q
= 0.
(17)
Example Class 1: Applications of the Lagrangian Approach
1. Consider a point mass (of mass m) falling in the Earths gravitational eld (take the acceler-
ation = g to be constant, i.e., independent of the distance of the point mass from the Earths
surface, and ignore air resistance).
(a) What is the kinetic energy T of the point mass? What is the potential energy V? Hence,
determine the Lagrangian L = T V.
(b) Use the Euler-Lagrange equation to determine the equation of motion for the vertical
motion (z direction) of the point mass.
(c) Solve the equation of motion, and express the solution in terms of the point masss
initial position z(0) and velocity z(0).
2. Consider a point mass (of mass m) attached to a spring (spring constant k), constrained
somehow to move only in the x direction, where x = 0 is taken to be the position of the
point mass when it is at rest. This is just a one-dimensional simple harmonic oscillator
(undamped), with characteristic oscillation frequency =

k/m.
(a) What is the kinetic energy T of the point mass? What is the potential energy V? Hence,
determine the Lagrangian L = T V.
(b) Use the Euler-Lagrange equation to determine the equation of motion of the point
mass.
(c) The equation of motion is a linear, homogeneous, second-order differential equation.
The rst step towards solving this is to nd the roots of the auxiliary equation. Do this,
referring to Appendix A of these notes for guidance.
(d) Find the general solution of the equation of motion, in terms of the point masss initial
position x(0) and velocity x(0).
(e) Two independent real constants A and B can equally well be represented in terms of
two other real constants A
/
and
/
, through A = A
/
sin
/
, and B = A
/
cos
/
. Use this,
along with the trigonometric identity sin( + ) = sin cos + cos sin , to show
your solution can also be expressed in the form
x(t) = A
/
sin(t +
/
), (18)
and express A
/
and
/
in terms of the initial position and velocity.
7
ADDITIONAL EXERCISE
3. Consider the situation of a plane pendulum, consisting of a pendulum bob of mass m con-
nected by a rigid massless rod of length R to a pivot at a point in space O. Let us set the
origin of our coordinate system at O. An upright connects the pivot, rigidly, to the centre of
a disc which is driven by a motor to spin around the z (vertical) axis with a constant angular
velocity , rotating the plane pendulum with it.
x
z
y

O
m
(a) With this system it is more convenient to use spherical coordinates, i.e., r, the distance
of the pendulum bob from the pivot, , the angle of elevation of the pendulum (how
far the pendulum is from pointing straight up), and , the azimuthal angle describing
how far the whole apparatus is rotated about the z axis. Write down the expressions
for x, y, z, in terms of r, , , and hence determine expressions for x, y, z in terms of r,
r, ,

, ,

.
(b) Consider a point mass of mass m moving in 3 dimensions. Using the expressions you
have derived for x, y, z, rewrite the standard expression for the kinetic energy T of the
point mass in spherical polar form (i.e., in terms of r, r, ,

,

). Rewrite the expression
for the potential energy V = mgz in spherical polar form.
(c) Write down equations that express the two constraints (that the point mass is always a
xed distance R from the origin, and that the azimuthal angle describing the location
of the point mass changes in time at a constant rate ) in mathematical form. Use these
constraint equations to simplify the kinetic energy T and the potential energy V, and
hence determine the Lagrangian L for the spinning plane pendulum, where is the
only dynamical variable.
(d) Use the Euler-Lagrange equation to determine the equation of motion for . From the
equation of motion determine a condition for the existence of horizontal circular orbits,
i.e., motion of the pendulum bob such that the angle of elevation never changes. For
what range of angles are such orbits possible? (Do not consider = 0 or ).
8
3 Linear Oscillators
SEE ALSO HAND AND FINCH, 3.13.3
3.1 Equilibrium
A mechanical system that remains at rest is in equilibrium. This occurs at points in
conguration space where all generalized forces T
k
vanish.
For a conservative system, this corresponds to congurations where the potential
energy V(q
1
, . . . , q
N
) is stationary, i.e., its rst derivatives with respect to the q
k
vanish.
3.2 Lagrangian Near Static Equilibrium (1 DoF)
Carry out a 2nd order Taylor series expansion of some L(q, q) around q = q
S
[where q
S
is a stationary point, i.e., V/q[
q=q
S
= 0] and q = 0. Hence,
L L
approx
=L(q
S
, 0) + q
L
q

q
S
,0
+ q
L
q

q
S
,0
+
1
2
_
q
2

2
L
q
2

q
S
,0
+ 2q q

2
L
q q

q
S
,0
+ q
2

2
L
q
2

q
S
,0
_
= A + Bq + C q + Dq
2
+ Eq q + F q
2
.
(19)
Substitute the resulting L
approx
into the Euler-Lagrange equation (n.b., B = 0 by
the denition of equilibrium)
d
dt
(
L
approx
q
..
C + Eq + 2F q) =
L
approx
q
..
2Dq + E q . (20)
This yields q (D/F)q = 0, which is identical to the EoM for a particle of mass
m derived from
L =
m
2
_
q
2
+
D
F
q
2
_
. (21)
3.3 Simple Harmonic Oscillator (SHO)
If D/F < 0, Eq. (21) is the Lagrangian for a simple harmonic oscillator, which models
e.g. the oscillation of a mass on a spring with spring constant k = m(D/F).
Close to equilibrium one therefore observes simple harmonic motion with character-
istic frequency =

D/F. Hence, the motion is described as stable (if D/F > 0, it


is unstable).
9
q +
2
q = 0 is a linear, homogeneous differential equation, and can be solved to
give
q(t) = q(0) cos(t) +
q(0)

sin(t). (22)
3.4 Damping Force
A frictional force F
d
, acts to suppress motion, i.e., it must act in opposition to motion,
and vanish when the motion ceases.
The simplest such force is F
d
= q = (m/Q) q (expressed in convenient form for a
damped SHO), where Q is the dimensionless quality factor of the oscillator.
3.5 Damped Simple Harmonic Oscillator
We insert the SHOLagrangian L = (m/2)( q
2

2
q
2
) into the Euler-Lagrange equation
and incorporate the (non-conservative) damping force separately
d
dt
_
L
q
_

L
q
= F
d
. (23)
[This corresponds to taking Eq. (10), grouping the conservative forces (in the form of
a potential energy function V) with T to make L, and keeping the non-conservative
forces separate].
The result is a linear, homogeneous differential equation, q + (/Q) q +
2
q = 0.
3.6 Overdamped SHO: Q <
1
2
The solution to the equation of motion is given by
q(t) = Qe
t/2Q
_
q(0) cosh(
/
t) +
1

/
_

2Q
q(0) + q(0)
_
sinh(
/
t)
_
, (24)
where
/
= (/2Q)
_
1 4Q
2
, or, equivalently
q(t) =
e
t/2Q
2
_
e

/
t
_
q(0)
_
1 +
/
/
2Q
_
+
q(0)

/
_
. .
if = 0, rapid decay
+ e

/
t
_
q(0)
_
1
/
/
2Q
_

q(0)

/
_
. .
if = 0, slow decay
_
.
(25)
10
3.7 Underdamped SHO: Q >
1
2
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
t
q
0 5 10 15 20
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
t
q
Figure 1: Plots of the coordinate q for the damped SHO. Left (overdamped): Q = 1/8, q(0) =
q(0)(
/
+ /2Q), q(0) = 1 (solid line); Q = 1/8, q(0) = q(0)(
/
/2Q), q(0) = 1 (dashed line).
Right (underdamped): Q = 4, q(0) = 1 , q(0) = q(0)(/2Q) (solid line); q(0) = 0, q(0)/
/
= 1
(dashed line). is an arbitrary length unit, e.g. a metre, centimetre, etc..
3.8 Critically Damped SHO: Q =
1
2
In this special case the solution is given by
q(t) = e
t/2Q
_
q(0) +
_

2Q
q(0) + q(0)
_
t
. .
grows initially, before decay takes over
_
. (26)
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
t
q
Figure 2: Plot of the coordinate q for the critically damped SHO. q(0) = 1 , q(0)/ = q(0)/2Q
(solid line); q(0) = 0, q(0)/ = 1 (dashed line). is an arbitrary length unit.
11
Example Class 2: Damped, Driven Pendulum

Consider a plane pendulum with an iron bob of mass m suspended from a pivot by a massless
rigid (non-magnetizing) rod of length l in a time-dependent magnetic eld so that the pendulum
bob is driven, being alternatingly attracted to the left or right. We use the generalized coordinate
, the angle the pendulum makes with the vertical [for an xz cartesian coordinate system with its
origin at the pendulum pivot, x = l sin(q/l), z = l cos(q/l)]. In the small angle approximation,
the Lagrangian (without the magnetic eld) is that of a simple harmonic oscillator with frequency
=
_
g/l.
1. (Study up to section 3.5). There will inevitably be some friction, e.g. from the pivot. In-
corporate a model frictional generalized force F
d
= (ml
2
/Q)

into the Euler-Lagrange
equation for and determine the equation of motion. (Study up to section 3.7). Solve the
equation of motion for the underdamped SHO (Q > 1/2) in terms of (0) and

(0) [use the
shorthands = /2Q, and
/
=
_
4Q
2
1 to describe the oscillation frequency, here
and from now on].
2. (Study up to section 4.3). At a given point in time t
/
the underdamped pendulum has velocity

(t
/
) and position (t
/
), when the pendulum bob is struck sharply with a hammer. Mod-
elling the effect of this blow with an impulsive potential V
I
(t) = Kl(t t
/
), derive the
corresponding generalized force and determine the values of

and immediately after the
hammer is applied.
3. (Study up to section 4.5). If we dene G(t t
/
) = (t)ml/K, and consider the underdamped
pendulum to be at rest (i.e., = 0,

= 0) prior to being struck sharply by the hammer, what
are the values of G(t t
/
) when t t
/
0, and when t t
/
0?
4. (Study up to section 4.6). We now include the effect of the time-dependent magnetic eld
on the pendulum bob, modelled as a sinusoidal driving potential V
D
(t) = F
0
l sin(
0
t)
applied from t = 0. Using a Greens function technique, determine an integral expression
for (t) for the driven underdamped pendulum equivalent to Eq. (32) in the notes. (Study
up to section 4.7). Solve the integral [expanding the sine terms, i.e., using sin = e
i
/2i
+ complex conjugate] and show the evolution of (t) to be proportional to the solution
described by Eq. (35) and Eq. (36) in the notes.
See also the additional exercises to Homework 2.
12
4 Driven Oscillators
SEE ALSO HAND AND FINCH, 3.43.8
4.1 Oscillator Driven by an External Force
We consider an (undamped) oscillator, driven by a time-dependent external force
F(t) (no spatial dependence).
Whatever supplies the driving force is not considered to be part of the dynamical
system, in particular we do not consider any effect of forces the oscillator exerts on
the source of the external force.
The corresponding Lagrangian is given by
L =
m q
2
2
..
T

_
m
2
q
2
2
F(t)q
_
. .
V
, (27)
which yields a linear, inhomogeneous differential equation: q +
2
q = F(t)/m.
4.2 Dirac -Functions
It is instructive to rst consider consider an impulsive (instantaneously applied
and innitely intense) force, and for this we make use of -functions. For a func-
tion f (t), the effect of (t t
/
) is dened (necessarily under an integral) as
f (t
/
) =
_

dt(t t
/
) f (t). (28)
(t t
/
) is the zero-width limit of a function/distribution of t, sharply peaked at
t = t
/
, with unit area:
_
dt(t t
/
) = 1 (dimensionless). Consequently, if t has
the dimension of time, (t t
/
) must therefore have the dimension of inverse
time, i.e., frequency.
4.3 Response to a q-Independent Impulsive Force
Consider an (impulsive) force of the form F(t) = K(t t
/
). Integrating the
EoM for the driven oscillator (section 4.1) around t = t
/
then yields
_
t
/
+
t
/

dt( q +
2
q) =
_
t
/
+
t
/

dt
K
m
(t t
/
),
q(t
/
+ ) q(t
/
) +
2
_
t
/
+
t
/

dtq(t)
. .
0 as 0, unless q(t
/
)
=
K
m
.
(29)
13
Hence, there is an instantaneous change in the velocity: q(t
/
+
) = q(t
/

) + K/m,
where t
/

= lim
0
t
/
(this is true in general; one can dene an impulsive
force as having this effect).
Note that q(t
/
+
) = q(t
/

), i.e., there is no instantaneous change in position when


the impulsive force is applied.
4.4 Time-Evolution Sequence of a SHO with an Impulsive Force
1. t < t
/
system evolves as a SHO, to position q(t
/

), velocity q(t
/

).
2. t = t
/
instantaneous jump in velocity: q(t
/
) = q(t
/

), q(t
/
) = q(t
/

) + K/m.
3. t > t
/
system evolves as a SHO:
q(t) = q(t
/

) cos([t t
/
]) +
q(t
/

) + K/m

sin([t t
/
]). (30)
4.5 (Causal) Greens Function G
Consider G(t t
/
) = q(t)m/K [equal to G(0) when the impulsive force applied]. The
EoM of G is given by (see section 4.1)

G(t t
/
) +
2
G(t t
/
) = (t t
/
)
We consider the oscillator to be at rest prior to application of the impulsive force.
Hence,
G(t t
/
) =0, t t
/
0,
G(t t
/
) =
1

sin([t t
/
]), t t
/
0.
(31)
4.6 General Driving Force F(t)
By denition of the -function [Eq. (28)] the equation of motion for the driven oscilla-
tor (section 4.1) can be written as q +
2
q = (1/m)
_

dt
/
F(t
/
)(t t
/
).
We can then eliminate (t t
/
) by substituting in the EoM of the Greens function G
(section 4.5): q +
2
q = (1/m)
_

dt
/
F(t
/
)[

G(t t
/
) +
2
G(t t
/
)].
Comparing terms, and substituting in the solution to G(t t
/
) [Eq. (31)] it follows that
q(t) =
1
m
_
t

dt
/
F(t
/
)G(t t
/
) =
1
m
_
t

dt
/
F(t
/
) sin([t t
/
]). (32)
14
4.7 Driving an Oscillator in Resonance
Consider a driving force of the form F(t) = F
0
sin(
0
t), which is applied from
t = 0 (i.e., F(t) = 0 for t < 0). In general (when
0
,= ) one observes oscillatory
behaviour
q(t) =
F
0
m
_
t
0
dt
/
sin(
0
t
/
) sin([t t
/
])
=
F
0
m
_
sin(
0
t) + sin(t)
2(
0
+ )

sin(
0
t) sin(t)
2(
0
)
_
.
(33)
In resonance (when
0
= ), the value of q grows without limit
q(t) =
F
0
m
_
t
0
dt
/
sin(t
/
) sin([t t
/
]) =
F
0
m
_
sin(t)
2
. .
oscillates

t cos(t)
2
. .
grows
_
. (34)
If we consider an underdamped SHO subject to the same driving force, the equa-
tion of motion can be solved to give q(t) = q
s
(t) + q
t
(t), which has been conve-
niently divided into the steady state solution
q
s
(t) =
F
0
2m
/
__


2
+ (
0
+
/
)
2



2
+ (
0

/
)
2
_
cos(
0
t)
+
_

0
+
/

2
+ (
0
+
/
)
2


0

/

2
+ (
0

/
)
2
_
sin(
0
t)
_
,
(35)
and the transient solution
q
t
(t) =
F
0
e
t
2m
/
__

0
+
/

2
+ (
0
+
/
)
2
+

0

/

2
+ (
0

/
)
2
_
sin(
/
t)

_


2
+ (
0
+
/
)
2



2
+ (
0

/
)
2
_
cos(
/
t)
_
,
(36)
where
/
=
_
4Q
2
1, and = /2Q.
The transient solution oscillates with the intrinsic oscillator frequency
/
, but
dies away for large t; the steady-state solution oscillates with the driving fre-
quency
0
, and remains.
After the transient has died away, the dynamical behaviour is like that of a SHO,
but at the driving frequency. Note, however, that there is no singularity at reso-
nance the effect of resonance has been softened by the presence of damping.
15
5 Small Oscillations
SEE ALSO HAND AND FINCH, 9.19.2
5.1 Coupled Oscillators
The generic Lagrangian for two coupled oscillators (e.g. two objects of equal mass
m oscillating, with frequency , on identical springs, connected by a third identical
spring) is given by
L =
m
2
( q
2
1
+ q
2
2
)
m
2
2
[q
2
1
+ q
2
2
+ (q
1
q
2
)
2
] =
m
2
( q
2
1
+ q
2
2
) m
2
(q
2
1
+ q
2
2
q
1
q
2
) (37)
We now introduce the centre of mass q
C
= (q
1
+ q
2
)/2 and relative q
R
= q
2
q
1
coordinates. Substituting these into the Lagrangian transforms it to
L = m
_
q
2
C
+
q
2
R
4
_
m
2
_
q
2
C
+
3q
2
R
4
_
, (38)
which yields two separated harmonic oscillator EoMs: q
C
+
2
q
C
= 0, q
R
+3
2
q
R
= 0.
5.2 Small Oscillations
We consider motion around a stable equilibrium point of a mechanical system (small
oscillations). If this motion does not depart too far from a stable equilibrium point,
the mechanical system resembles a system of coupled oscillators.
In a system of N coupled oscillators there always exist N generalized coordinates
that separate, yielding N independent, homogeneous, linear differential equations.
Those generalized coordinates undergoing simple harmonic motion are called nor-
mal coordinates.
5.3 Setting up the Problem
Find an equilibrium conguration, i.e., values of the generalized coordinates where
all generalized forces T
k
= V/q
k
are = 0.
Taylor expand the Lagrangian L to second order in the generalized coordinates and
velocities, around values for the generalized coordinates q
k
given by the equilibrium
conguration, and the values of the generalized velocities q
k
set = 0.
For N DoF, the Taylor expansion must in principal be in 2N variables (all generalized
coordinates and velocities). In practice, expansions in fewer variables may be ade-
quate, e.g. if some terms are already in quadratic form.
16
Homework 2: Double Pendulum
1. Consider a double plane pendulum, consisting of a mass M attached to a pivot by a rigid
massless rod of length R, and a second mass m attached to a pivot on the rst mass, also by
a rigid massless rod of length R.

m
M
(a) Consider a cartesian coordinate system with its origin located at the pivot to which the
mass M is attached. We consider Z to be the vertical axis, X the horizontal, and motion
to take place in the XZ plane; i.e., that motion is subject to the constraints y
M
= 0,
y
m
= 0, where y
M
, y
m
are the Y locations of the masses M, m. Determine equations
describing any other constraints the two masses are subject to, in terms of x
M
, x
m
, z
M
,
z
m
(the X and Z locations of the masses).
(b) Express the dynamical variables x
M
, x
m
, z
M
, z
m
in terms of the new dynamical vari-
ables and (which are the two angles shown in the diagram), and the corresponding
velocities x
M
, x
m
, z
M
, z
m
in terms of

,

, , . Use these expressions to determine
the Lagrangian L for this system in terms of

,

, , , assuming the case where M =
3m. [Hint: simplify your expression by using cos() cos( + ) + sin() sin( + ) =
cos()].
(c) (Study up to section 5.3). Consider the limit of small oscillations, i.e., such that , , (and
therefore + ) remain small during any dynamics. Hence, determine an approxima-
tion L
approx
to L which is in quadratic form. Write L
approx
in matrix form, and determine
the corresponding equations of motion for and (also in matrix form).
(d) (Study up to section 6.3). (You will need to determine such eigenvalues and eigen-
vectors associated with the matrix form of the approximate Lagrangian in or-
der to determine the normal coordinates. See Appendix C of these notes for
guidance on calculating these). (Study up to section 6.4). Determine the normal
coordinates of this system.
Note: This exercise involves solving a generalized eigenvalue problem, the eigenvectors of
which are orthogonal (and normalized) with regard to the metric . This means that, given
the 2 eigenvectors q
1
, q
2
, and the kinetic energy matrix , then q
T
j
q
k
= 0 for j ,= k (and = 1
for j = k if the eigenvectors are normalized).
ADDITIONAL EXERCISES
2. The linear ion trap is at present probably the most promising experimental conguration for
a quantum computer. Its normal modes are therefore of considerable interest. Consider a
classical model of a linear ion trap, consisting of 3 charged spheres of mass m and charge c
sliding on a frictionless rod of length 4l propped between two charged parallel plates, also
of charge c.
17
1 3 2
x
(a) (Study up to section 5.2). Treat the 3 charged spheres as point charges, and the two
charged plates as immovable point charges. Use the Coulomb potential V
C
= [c
2
/(4
0
)](1/r)
to model the interactions between the effective point charges (r is the distance between
the point charges). Hence, determine the total potential energy V(x
1
, x
2
, x
3
), where x
1
,
x
2
, x
3
are the distances of the leftmost, middle, and rightmost charged sphere from
the left-hand charged plate. To simplify matters, consider only interactions between
immediately neighbouring objects (plates or spheres).
(b) (Study up to section 5.3). Determine the equilibrium values of x
1
, x
2
, x
3
, hence showing
explicitly that the charges must be equally spaced at equilibrium (as one would expect).
(c) (Study up to section 5.3). Carry out a second-order Taylor series expansion of the Coulomb
potential V
c
about an arbitrary value r = a. Substitute this into V(x
1
, x
2
, x
3
) (with the
values of a determined by the equilibrium values of x
1
, x
2
, x
3
, as appropriate). (See
Appendix B of these notes for guidance on determining the appropriate Taylor
expansion).
(d) (Study up to section 6.1). Hence, determine the approximate second-order Lagrangian in
terms of new (shifted) generalized coordinates which all have equilibrium value = 0.
Write it down in matrix form.
(e) (Study up to section 6.3). (You will need to determine such eigenvalues and eigen-
vectors associated with the matrix form of the approximate Lagrangian in order
to determine the normal coordinates. See Appendix C of these notes for guid-
ance on calculating these). (Study up to section 6.4). Determine the (normalized)
normal coordinates of this system.
Check/hint: If r
1
, r
2
, r
3
are the normalized normal coordinates, the approximate Lagrangian
should take the form L = (m/2)( r
2
1
+ r
2
2
+ r
2
3
) (m/2)(
2
1
r
2
1
+
2
2
r
2
2
+
2
3
r
2
3
) k, where
2
1
,

2
1
,
2
3
are the eigenvalues associated with the normal coordinates, and k is an irrelevant
constant. (Finding a set of coordinates that separate in this way is the purpose of the whole
procedure).
3. Consider the motion of a length l rigid plane pendulum around its unstable equilibrium
point, i.e., with the (mass m) pendulum bob directly above the pivot.
(a) Derive the Lagrangian for this system in terms of a single appropriate generalized co-
ordinate (with its corresponding generalized velocity). Assume all the mass of the
pendulum to be concentrated in the pendulum bob (i.e., the moment of inertia = ml
2
).
(b) Carry out a second-order Taylor series expansion of this Lagrangian around the pen-
dulums unstable equilibrium point, and hence derive an approximate Lagrangian.
(c) Derive and solve the equation of motion produced by this approximate Lagrangian, in
terms of the initial values of the generalized coordinate and generalized velocity.
(d) Consider two possible initial conditions. In each case, the pendulum is initially at rest,
but the locations of the pendulum bobs are slightly different. Show that, according to
18
the approximate Lagrangian you have derived, for long times the difference in nal
locations of the pendulum bobs always diverges exponentially in time.
(e) How accurately does the previous result reect the complete dynamics of a plane pen-
dulum? Explain your answer.
4. Derive the general solution for the coordinate of a critically damped harmonic oscillator,
noting that the auxiliary equation in this case has only one root. Check that this is consistent
with the solution of the underdamped oscillator in the limit Q 1/2.
19
6 Normal Modes
SEE ALSO HAND AND FINCH, 9.29.4
6.1 Matrix Form of L
If a
k
are the equilibrium values of an initially used set of generalized coordinates q
/
k
,
we then dene a new set of generalized coordinates q
k
= q
/
k
a
k
.
These q
k
will have equilibrium values = 0, and L can then take the form
L = q
T
q
..
T
q
T
q
..
V
+c (a constant), (39)
where q is a column vector, and its transpose q
T
= (q
1
, . . . , q
N
) is a row vector.
The matrices and are symmetric (i.e., equal to their transposes, so that e.g.
kj
=

jk
), with matrix elements given by

jk
=
1
2

2
T
q
j
q
k

q
j
, q
k
=0
,
jk
=
1
2

2
V
q
j
q
k

q
j
,q
k
=0
. (40)
6.2 Equations of Motion
We note that partial differentiation of the Lagrangian L with respect to the general-
ized coordinates and velocities yields
L
q
j
= 2
N

k=1

jk
q
k
= 2( q)
j
,
L
q
j
= 2
N

k=1

jk
q
k
= 2( q)
j
, (41)
and that the Euler-Lagrange equations therefore yield a linear system of differential
equations: q + q = 0.
6.3 Working Towards a Solution
Inserting a trial solution q = q
0
e
i
0
t
into the EoM yields
(
2
0
+ )q
0
= 0. (42)
We also know (due to a theorem from linear algebra) that, unless q
0
= 0, this can
only be fullled if the determinant of
2
0
+ (also written as [
2
0
+ [) is = 0.
Calculating the determinant will yield an Nth order polynomial in
2
0
. Hence, there
are N generalized eigenvalues
2
j
(possible values of
2
0
), each with its correspond-
ing eigenvector q
j
[specic q
0
which, with its associated specic value of
2
0
, will
solve Eq. (42)].
20
6.4 Normal Coordinates
It is common to normalize the eigenvectors q
j
, i.e., to have q
T
j
= (b
j1
, b
j2
, . . . , b
jN
) such
that
_

N
k=1
b
2
jk
= 1. Using the eigenvector elements, we may dene a new set of
separated generalized coordinates r
j
:
r
j
=
N

k=1
b
jk
q
k
. (43)
If
2
j
> 0, the equilibrium conguration is stable, and the normal coordinate r
j
evolves as r
j
(t) = r
j
(0) cos(
j
t) + ( r
j
(0)/
j
) sin(
j
t). Such motion is called a normal
mode, with motion in general described by a superposition of normal modes.
If
2
j
< 0, the equilibrium is unstable, and if
2
j
= 0, the motion is also not oscillatory,
usually corresponding to e.g. rotation or translation of the centre of mass.
A normal mode is the motion of a normal coordinate.
21
7 Central Forces
SEE ALSO HAND AND FINCH, 4.34.4
7.1 Two Interacting Bodies
Consider two bodies (with position vectors r
1
and r
2
) interacting via a central force,
dened as a force depending only on the distance r between the two bodies, di-
rected along the imaginary line between them.
The potential energy V
12
from which such a force is derived is dependent only on r,
i.e., V
12
= V
12
(r
1
, r
2
) = V
12
([r
1
r
2
[) = V(r) for such a force.
If we consider the general case of 2 point masses interacting via a central force,
and both point masses subject to external forces, the Lagrangian is given by
L =
1
2
(m
1
r
2
1
+ m
2
r
2
2
) V
1
(r
1
) V
2
(r
2
)
. .
external potentials
V
12
([r
1
r
2
[).
. .
interaction potential
(44)
In the absence of external potentials (V
1
= V
2
= 0), this Lagrangian is translationally
invariant: L(r
1
, r
2
, r
1
, r
2
) = L(r
1
+ a, r
2
+ a, r
1
, r
2
).
7.2 Translational Symmetry and Conservation of Momentum
Consider an innitesimal translation a, and Taylor expand the Lagrangian L to
rst order in :
L(r
1
+ a, r
2
+ a, r
1
, r
2
) = L(r
1
, r
2
, r
1
, r
2
) + a (
1
+
2
)L +O(
2
). (45)
We neglect the O(
2
) term. Since a arbitrary, if L is to be translationally invari-
ant, then (
1
+
2
)L must = 0, i.e.,
L
x
1
+
L
x
2
= 0,
L
y
1
+
L
y
2
= 0,
L
z
1
+
L
z
2
= 0. (46)
From the Euler-Lagrange equations for x
1
and x
2
(and equivalently for y and z):
d
dt
_
L
x
1
_
+
d
dt
_
L
x
2
_
=
L
x
1
+
L
x
2
= 0. (47)
L/ x
1
= p
1x
denes the canonically conjugate momentum p
1x
to x
1
(here this is
equal to the mechanical momentum m x
1
, but this need not in general be so).
Hence,

P = p
1
+ p
2
= 0, and so the total momentum P is conserved.
22
7.3 Ignorable Centre of Mass (CoM) Motion and Conservation of Momentum
Conservation of the total momentum P implies that the CoM[coordinate R = (m
1
r
1
+
m
2
r
2
)/(m
1
+ m
2
) moves like a free particle.
We rewrite L In terms of R and the relative coordinate r = r
1
r
2
:
L =
1
2
M

R
2
+
1
2
r
2
V
12
(r), (48)
where M = m
1
+ m
2
(total mass) and = m
1
m
2
/(m
1
+ m
2
) (the reduced mass).
R does not appear in the Lagrangian L. Such coordinates are called ignorable. For
any ignorable generalized coordinate q, it follows from the Euler-Lagrange equation
that d(L/ q)/dt p
q
= 0, and hence that P is conserved in this case.
7.4 Rotational Invariance and Conservation of Angular Momentum
If we transform to the centre of mass frame (a frame moving with the CoM), R and

R are both dened = 0, and L = (1/2) r


2
V
12
(r). We note that V
12
(r) is dependent
on distance only, not direction, and hence that L is rotationally invariant.
In terms of spherical coordinates (x = r cos sin , y = r sin sin , z = r cos ), we
deduce that r
2
= x
2
+ y
2
+ z
2
= r
2
+ r
2

2
+ r
2

2
sin
2
. Hence,
L =
1
2
[ r
2
+ r
2

2
+ r
2

2
sin
2
] V
12
(r). (49)
is ignorable, as it does not appear in L. Hence
p

= r
2

sin
2
= constant J
z
..
can show to be the z component of J = r r
(50)
If J
z
is the only nonzero component of J, the total angular momentum constant.
However, the orientation of the coordinate system is arbitrary, and therefore J is al-
ways constant.
7.5 Motion Takes Place in a Plane
J = r r implies that p r and r must be perpendicular to J.
Hence, p and r must be in a plane perpendicular to J, and remain in that plane
if J is to remain constant.
If J is dened along the z direction, the angle of elevation = /2,

= 0, and
J = [J[ = J
z
= r
2

.
23
7.6 Equal Areas are Swept out in Equal Times (Keplers Second Law)
d
r
r+dr
dr
{
height h
Figure 3: The triangle swept out by changing the position vector r by the increment dr. The
incremental area of the triangle dA is deduced by the usual half-base-times-height.
If the position vector changes incrementally from r to r + dr, with the angle
swept through the angle increment d, the area increment dA is given by the
area of the triangle formed by r, dr, and r + dr.
By straightforward trigonometry dA = (1/2)[r + dr[[r[ sin(d). Noting that we
assume = /2 throughout, and that we must have only rst-order incremen-
tal terms, we deduce that dA = (1/2)r
2
d.
It follows that
dA
dt
=
1
2
r
2

=
J
2
, (51)
i.e., as J is conserved, that the rate at which area is swept out is constant, and
hence that equal areas are swept out in equal times.
Keplers second law is therefore a consequence of angular momentum con-
servation, itself a consequence of considering a central force, and has nothing
specically to do with gravity as such.
7.7 Equivalent 1D Problem
It is straightforward to determine the total energy E = T + V from L = T V as given
in Eq. (48). Hence, E = (1/2)[ r
2
+ r
2

2
+ r
2

2
sin
2
()] + V
12
(r), which, if we make use
of

= 0, = /2,

= J/r
2
, simplies to
E =
1
2
r
2
+
J
2
2r
2
+ V
12
(r).
. .
V
eff
(r)
(52)
The effective potential V
eff
(r) combines the interaction potential with the so-called
angular momentum barrier (responsible for the centrifugal force).
The total energy E is the same as that corresponding to an effective Lagrangian
L
eff
= (1/2) r
2
V
eff
, from which the Euler-Lagrange equation for r yields
r =
J
2
r
3

V
12
r
. (53)
24
Homework 3: Diatomic Molecule
1. A simple model for a diatomic molecule (such as H
2
) consists of two mass points connected
by a spring. We constrain the motion to a plane, considering two mass points, each of
mass m, connected by a spring of spring constant k and natural length d, sliding on a at,
frictionless surface.
(a) Determine an expression for the kinetic energy in terms of the centre of mass (x
C
, y
C
)
and relative (x
R
, y
R
) coordinates.
(b) Drawa diagramto showwhat the generalized coordinates R and correspond to phys-
ically, where x
R
= Rcos(), y
R
= Rsin(), and R 0. Also indicate the location of the
centre of mass. Hence, determine an expression for the potential energy in terms of R
only.
(c) Consider the centre of mass to be xed at the origin, and write the Lagrangian that
remains in terms of R and . Determine an expression for J

, the canonically conjugate


momentum to . Use the Euler-Lagrange equations to show that J

is conserved. Use
this result together with the Euler-Lagrange equation for R to determine an equation
of motion for R in closed form (i.e., independent of ). Use this equation to determine
the equilibrium extension of the spring if the angular momentum J

= 0.
(d) Is it in general possible to consider the motion of the two connected masses to take
place in a plane, even when this system is in free (three-dimensional) space? Explain
your answer.
ADDITIONAL EXERCISE
2. The purpose of this exercise is to derive Keplers equation, as outlined in Central Forces,
and Gravitational Attraction in the notes. Consider two bodies, of mass m
1
and m
2
, re-
spectively, in free space (i.e., there are no external potentials), interacting gravitationally
(potential energy V
G
(r) = Gm
1
m
2
/r).
(a) Determine the Lagrangian describing the motion of the relative coordinate r = r
1
r
2
.
Hence, determine an expression for the total energy E in the CoM frame.
(b) Dene an effective 1D Lagrangian L
eff
which yields the same value of E, but with T =
(1/2) r
2
(where r = [r[ and is the reduced mass). With the Euler-Lagrange equation,
derive from L
eff
the equation of motion for r.
(c) Express the energy in the centre of mass frame in terms of u 1/r and the azimuthal
angle , and solve the differential equation (of u, differentiated with respect to ) re-
sulting from the expression for dE/d, to nd a relationship between r and .
(d) Express this solution in terms of cartesian coordinates x and y rather than polar coordi-
nates r and , and hence show that bound states must take the form of ellipses, of the
general form (x + x
c
)
2
/a
2
+ y
2
/b
2
= 1, in the centre of mass frame.
(e) Use the identity cos
2
+ sin
2
= 1 to parametrize x and y in terms of , and hence
determine an equivalently parametrized form of r. Use this to solve the effective 1D
problem by quadrature and determine Keplers equation.
25
8 Gravitational Attraction
SEE ALSO HAND AND FINCH, 4.1, 4.54.6
8.1 Solving 1D Systems by Quadrature
If we consider a 1D system, for example of a point mass in an external potential, and
hence with Lagrangian L = (1/2)m q
2
V(q), then it has energy E = (1/2)m q
2
+V(q).
One can then deduce
q
dq
dt
=
_
2[E V(q)]
m
dt = dq
_

_
m
2[E V(q)]
_
. (54)
The time t taken to move from q = 0 to q(t) is then
t =
_
m
2
_
q
0
dq
/

1
E V(q
/
)
. (55)
The time t(q) (elapsed time as a function of the coordinate) must then be inverted if
we wish to obtain the solution we are more used to, q(t) (this may have to be carried
out numerically).
A problem solved as an integral [i.e., put into integral form, as in Eq. (55)] is said to
be solved by quadrature.
8.2 Gravitational Interaction Potential
The potential energy between two gravitating bodies of masses m
1
and m
2
is V
12
(r) =
k/r, where k = Gm
1
m
2
and G is the gravitational constant. The EoM is then
r =
J
2
r
3

k
r
2
(56)
8.3 The u = 1/r Transformation
We choose to change the dependent variable from r to u = 1/r, and, from J =
r
2
(d/dt), determine that dt = [/(Ju
2
)]d and hence that
d
dt
=
Ju
2

d
d

dr
dt
=
1
u
2
du
dt
=
J

du
d
. (57)
Expressing the energy in the CoM frame [Eq. (52)] in terms of u and , we get
E =
J
2
2
_
_
du
d
_
2
+ u
2
_
ku. (58)
26
We differentiate E with respect to (note that E is constant, i.e., dE/d = 0) and
produce the differential equation
d
2
u
d
2
+ u =
k
J
2
. (59)
This can be solved to give (1/r) u = (k/J
2
) + Acos (A is an integra-
tion constant, and we have set a second integration constant, equivalent to an
arbitrary phase added to , to 0).
8.4 Elliptical Orbits (Keplers First Law)
If we dene p J
2
/k and pA, the solution to Eq. (59) can be written as p =
r + r cos . Expressed in cartesian coordinates, this becomes p =
_
x
2
+ y
2
+ x, or
equivalently
(1
2
)x
2
+ 2px + y
2
p
2
= 0. (60)
Substituting the solution for u [Eq. (59)] into E [as given in Eq. (58)] yields E = (k/2J
2
)(
2

1) = (k/2p)(
2
1). If 0 < 1, the energy E is negative, implying a bound state.
Restricting ourselves to 0 < 1, Eq. (60) describes an ellipse (Keplers rst law):
1
a
2
_
x +
p
1
2
_
2
+
y
2
b
2
= 1, (61)
where the semimajor axis a = p/(1
2
) and the semiminor axis b = p/

1
2
(with
these quantities dened, the energy takes the simple form E = k/2a).
8.5 Keplers Third Law
From Keplers second law dA/dt = J/(2), where A is the swept-out area [Eq. (51)].
Hence, the period of an elliptical orbit follows from the total area A = ab of that
orbit , i.e., = A2/J. Elementary algebraic manipulation yields
= 2

1
G(m
1
+ m
2
)
a
3/2
(62)
Keplers third law states that
2
a
3
with the same proportionality constant for all the
planets. As the mass of the sun dominates the masses of the individual planets, this
is approximately true.
27
8.6 Keplers Equation
From Eq. (61), we can use the identity cos
2
+ sin
2
= 1 to introduce the
parametrization
x
p
1
2
= a cos x = a(cos ),
y = b sin y = a
_
1
2
sin ,
(63)
from which it follows that r =
_
x
2
+ y
2
= a(1 cos ). Note that is a
different angle (the eccentric anomaly) to (the true anomaly), as illustrated in
Fig. 4.
a
r

Figure 4: Relationship of the eccentric anomaly to the true anomoly for an effective particle
placed a distance r and azimuthal angle from the CoM. The elliptical orbit, with semimajor axis
a, has been placed inside an imaginary circle of radius a.
We can solve the effective 1D problem [Eq. (53)] by quadrature. From Eq. (55) it
follows that
_
dt
/
=
_

2
_
dr
/
_
E + k/r
/
J
2
/(2r
/
2
)
. (64)
We are considering elliptical orbits, i.e., E = k/2a. We also use J
2
/(k) p
and p = a(1
2
) to obtain
_
dt
/
=
_

2k
_
r
/
dr
/
_
r
/
2
/(2a) + r
/
(a/2)(1
2
)
. (65)
28
Finally, we use r = a(1 cos ) and dr = a sin d to obtain
_
dt
/
=
_
a
3
k
_
d
/
(1 cos
/
). (66)
As written, the time integral measures the time elapsed between an initial and a
nal value of
/
(the eccentric anomaly). It is customary to choose t
/
= 0 when

/
=
/
= 0 (when the orbiting body is at its perihelion, i.e., closest to the sun
(which is almost at the position of the CoM as the mass of the sun generally
dominates the mass of the orbiting body).
Hence, if t() is the time elapsed from perihelion to when the eccentric anomaly
takes the value , we have Keplers equation:
t() =
_
a
3
k
( sin ). (67)
29
9 Noethers Theorem and Hamiltonian Mechanics
SEE ALSO HAND AND FINCH, 5.25.5
9.1 Invariance of L under Continuous Transformations
Consider a continuous family of transformations, i.e., transformations depend-
ing on one or more continuously variable parameters (e.g. translations along or
rotations about the z axis are single-variable families of transformations).
When these parameters (e.g. translational distance, angle) are = 0, this should
yield the identity transformation, i.e., no change.
Let s be a parameter characterizing a general transformation of coordinates.
If q
k
(t) is a solution of the original EoM for a Lagrangian L, let Q
k
(s, t) be a
solution for the Euler-Lagrange equations evaluated within a transformed set
of coordinates, where Q
k
(0, t) = q
k
(t).
The Lagrangian L is invariant (independent of s) under this transformation if
L
/
L(Q
1
(s, t), . . . , Q
N
(s, t),

Q
1
(s, t), . . . ,

Q
N
(s, t), t)
=L(q
1
, . . . , q
N
, q
1
, . . . , q
N
, t).
(68)
As the Lagrangian is independent of s, dL
/
/ds = 0, and therefore
dL
/
ds
=
N

k=1
_
L
/
Q
k
..
=
d
dt
_
L
/


Q
k
_
dQ
k
ds
+
L
/


Q
k
d

Q
k
ds
..
=
d
dt
_
dQ
k
ds
_
_
=
d
dt
_

k=1
L
/


Q
k
dQ
k
ds
_
= 0.
(69)
implies the existence of a constant of the motion I(q
1
, . . . , q
N
, q
1
, . . . , q
N
)

N
k=1
p
k
(dQ
k
/ds)[
s=0
, which is evaluated at s = 0 for convenience.
For multiple parameters s
1
, s
2
, . . ., the procedure can be repeated for each to nd
associated constants I
1
, I
2
, . . ..
This procedure effectively describes Noethers theorem.
9.2 Noethers Theorem
If the Lagrangian is invariant under a continuous symmetry transformation, there
are conserved quantities associated with that symmetry, one for each parameter
of the transformation. These can be found by differentiating each coordinate with
respect to the parameters of the transformation in the immediate neighbourhood of
the identity transformation, multiplying by the conjugate momentum, and summing
over the degrees of freedom.
30
9.3 Legendre Transformations
We assume a function A(x, y) of a passive variable x and an active variable y.
We then introduce a third variable z and dene B(x, y, z) yz A(x, y). Hence,
by application of the chain rule,
dB = zdy + ydz
A
x

y
dx
A
y

x
dy =
_
z
A
y

x
_
. .
vanishes, as we dene z = z(x, y)
A
y

x
dy + ydz
A
x

y
dx.
(70)
By this chosen denition of z, it follows that B has no explicit y dependence,
and that therefore y can be written in terms of x and z, i.e., y = y(x, z), B =
B(x, y(x, z), z) = B(x, z) = yz A(x, y). Furthermore, it follows from Eq. (70)
that
B
z

x
= y,
B
x

z
=
A
x

y
. (71)
9.4 The Hamiltonian
ALegendre transformation of a (time-independent, 1 DoF) Lagrangian L(q, q), where
the active variable is the velocity q, the passive variable is the coordinate q, and the
third variable is the canonically conjugate momentum p, yields the Hamiltonian H:
H(q, p) p q L(q, q), where p
L
q

constant q
. (72)
From dH [determined according to Eq. (70)] and the Euler-Lagrange equation, we
determine Hamiltons equations.
H
p

q
= q,
H
q

p
=
L
q

q
=
d
dt
L
q

q
= p, (73)
9.5 Hamiltons Principle
From Hamiltons principle (section 2.5), we know that variations in the action
S =
_
Ldt are given by S =
_
Ldt = 0. Using the relationship between L and
H given by the Legendre transformation [Eq. (72)] we calculate variations:
L = qp + p q
H
q
q +
H
p
p
. .
H
=
_
q
H
p
_
p
_
p +
H
q
_
q +
d
dt
(pq).
(74)
31
We note that d(pq)/dt vanishes on integration, as q = 0 at the endpoints of
the physical path, and that Hamiltons equations mean the coefcients of q, p
in Eq. (74) also vanish.
Hence, independent innitesimal variations in q, p from the physical path in
coordinate-momentum space (called phase space as opposed to the conguration
space of Lagrangian mechanics) do not change the action. This is summarized
by stating that dH = qdp pdq.
9.6 Hamiltons Equations of Motion
For N DoF, H
N
k=1
p
k
q
k
L (where L may be explicitly time-dependent). We there-
fore include a time differential in dH =
N
k=1
q
k
dp
k

N
k=1
p
k
dq
k
+ (H/t)dt. We can
also consider time as a passive variable in the Legendre transformation, which yields
H
t

q
1
,...,p
1
,...
=
L
t

q
1
,..., q
1
,...

dH
dt
=
H
t
=
L
t
(75)
If there is no explicit time dependence in L, H is a constant of the motion. If T is also
a quadratic form in the q
k
, H is the total energy E = T + V.
Finally, Hamiltons equations of motion in complete generality are:
q
k
=
H
p
k
, p
k
=
H
q
k
,
dH
dt
=
L
t
. (76)
Homework 4: Canonical Transformation of the SHO
Consider a one-dimensional simple harmonic oscillator (potential energy V = m
2
q
2
/2, mod-
elling, for example, a point mass of mass m on a spring of spring constant = m
2
.
1. (Study up to section 9.4). Determine the Hamiltonian H fromthe Lagrangian L for this system,
using a Legendre transformation. (Study up to section 9.6). What is the relationship between
H and the total energy E in this case? Is this expected? Explain your answer.
2. (Study up to section 10.3). Consider a generating function of the form F
1
(q, Q) = i[(mq
2
+
kQ
2
)/2

2kmqQ], where k is an arbitrary constant with the units of action (kgm


2
s
1
).
Determine expressions for p and P (the canonically conjugate momenta to q and Q, respec-
tively) from the implicit transformation equations.
3. (Study up to section 10.4). Use the expressions you have calculated for p and P in terms of q
and Q to determine expressions for p and q in terms of P and Q, and hence determine the
transformed Hamiltonian H
/
(P, Q).
4. (Study up to section 10.4). Use Hamiltons equations to determine the the equations of motion
for Q and P. From these, determine differential equations also for Q

and iP/k. Calculate


the general solution for Q as a function of time.
5. (Study up to section 10.6). Calculate the Poisson bracket Q, Q

.
32
10 Canonical Transformations and Poisson Brackets
SEE ALSO HAND AND FINCH, 6.16.3
10.1 Canonical Transformations
The Lagrangian formulation of mechanics makes coordinate transformations (com-
monly called point transformations), i.e., changing to new generalized coordinates
and their associated generalized velocities, fairly straightforward.
Coordinate transformations take place in conguration space. In phase space one
can consider new coordinates Q and momenta P which are functions of the original
coordinates q and their momenta p (and possibly the time t):
Q = Q(q, p, t), P = P(q, p, t). (77)
These are commonly called contact transformations.
A transformation is by denition canonical , if the structure of Hamiltons equations
for all dynamical systems is preserved by it. Hence, for a specic dynamical system,
a new Hamiltonian H
/
which is a function of Q and P is produced, and Hamiltons
equations for H
/
, P, and Q must describe the correct transformed dynamics.
10.2 Equivalence of Lagrangians
Consider two Lagrangians L and L
/
such that L
/
(q, q, t) = L(q, q, t) dF(q, t)/dt
(F is independent of q). Substituting L
/
into the Euler-Lagrange equation yields
d
dt
_
L
q
_

L
q
=
d
dt
_

q
_
dF
dt
__


q
_
dF
dt
_
=
d
dt
_

q
_
q
F
q
+
F
t
__


q
_
q
F
q
+
F
t
_
=
d
dt
_
F
q
_

_
q

2
F
q
2
+

2
F
qt
_
= 0,
(78)
i.e., the Euler-Lagrange equation derived from L. Hence, the descriptions of a
physical system provided by L and L
/
are equivalent.
We now consider instead a starting Lagrangian L(q, q) and a transformed La-
grangian L
/
(Q,

Q) related by a similar F which is a function of both q and Q,
and perhaps t, such that
L
/
(Q,

Q, t) = L(q, q, t)
dF(q, Q, t)
dt
(79)
( ,= 1 is associated with a change of units or scale transformation. We will always
consider = 1)
33
We next integrate Eq. (79) with respect to t (and with = 1):
_
t
2
t
1
L
/
dt =
_
t
2
t
1
Ldt + F(q(t
1
), Q(t
1
), t
1
) F(q(t
2
), Q(t
2
), t
2
) (80)
Recalling Hamiltons principle (section 2.5), if we take variations of Eq. (80)
[assuming that arbitrary variations q(t) imply arbitrary variations Q(t)] the
variation of
_
t
2
t
1
Ldt vanishes. If we further assume that F = 0 at the end points,
then we can say that the variation of
_
t
2
t
1
L
/
dt vanishes, and it follows that the
two descriptions are equivalent.
10.3 The Generating Function
F(q, Q, t) is called a generating function. From the chain rule, the time derivative for
the generating function is dF/dt = (F/q) q + (F/Q)

Q + F/t
By construction, q does not appear explicitly in L
/
, and so, from Eq. (79)
0 =
L
/
q
=
L
q

F
q
p =
F
q
. (81)
Similarly
L
/


Q
=
F
Q
P =
F
Q
. (82)
These are known as the implicit transformation equations.
With the implicit transformation equations we have two equations and two unknowns.
To nd an explicit form for the transformation, solve Eq. (81) to express Q = Q(q, p, t),
and then insert this relation into Eq. (82) to get P = P(p, q, t).
Extension to N > 1 DoF is in principle straightforward, although in practice the task
of determining 2N unknowns from 2N equations may be formidable.
10.4 The Transformed Hamiltonian
To nd the new Hamiltonian H
/
(Q, P), we rst return to the denition of H
/
in
terms of a Legendre transformation [Eq. (72)]. Hence,
H
/
(Q, P, t) P

Q L
/
=
F
Q

Q L +
F
q
q +
F
Q

Q +
F
t
= p q L +
F
t
=H(q, p, t) +
F(q, Q, t)
t
(83)
where we have made use of both Eq. (82) and Eq. (81). Finally, p and q appearing
in the RHS of Eq. (83) must be expressed in terms of P and Q.
34
If there is no explicit time dependence to F, then the new Hamiltonian H
/
is
simply the original Hamiltonian H, rewritten by inserting the inverse of the
transformation equations expressing P and Q in terms of p and q.
10.5 Four Forms of Generating Function
The generating function need not be written in terms of q and Q. Any combination
of one original and one new dynamical variable may be employed, in conjunction
with appropriately modied implicit transformation equations:
F
1
(q, Q, t), p =
F
1
q
, P =
F
1
Q
;
F
2
(q, P, t), p =
F
2
q
, Q =
F
2
P
;
F
3
(p, Q, t), q =
F
3
p
, P =
F
3
Q
;
F
4
(p, P, t), q =
F
4
p
, Q =
F
4
P
.
(84)
With some caveats, the different forms of generating function may be related via
Legendre transformations, or the implicit transformation equations can simply be
determined independently in a similar manner as in section 10.3.
10.6 Poisson Brackets
For N DoF, a Poisson bracket for two arbitrary functions F and G is dened as
F, G
N

k=1
_
F
q
k
G
p
k

F
p
k
G
q
k
_
=
F
q
G
p

F
p
G
q
if N = 1. (85)
Clearly, q, p = 1 (we consider 1DoF for brevity, but the result holds for N DoF). If
we consider a Poission bracket of q and p dened with respect to the canonically
transformed dynamical variables Q and P, then [using Eq. (84) and the chain rule]
q
Q
p
P

q
P
p
Q
=
q
Q

2
F
2
Pq
+

2
F
4
Pp
p
Q
=
Q
q
q
Q
+
Q
p
p
Q
=
Q
Q
= 1, (86)
i.e., the value of the Poisson bracket for q and p is independent of the representation.
Conversely, Q, P = 1 is a necessary and sufcient condition for a transformation to
be canonical.
Using the result of Eq. (86), we see that any Poisson bracket is independent of the
representation in which it is calculated:
35
F
Q
G
P

F
P
G
Q
=
_
F
q
G
p

F
p
G
q
__
q
Q
p
P

q
P
p
Q
_
=
F
q
G
p

F
p
G
q
, (87)
We therefore do not need to specify a particular set of coordinates and conjugate
momenta when discussing Poisson brackets.
10.7 Poisson Bracket Formulation of Hamiltons Equations
Hamiltons equations of motion can be reformulated in terms of Poisson brackets:
q = q, H, p = p, H. For an arbitrary function F(q, p, t),

F = F, H +
F
t
. (88)
36
11 Rotating Reference Frames
SEE ALSO HAND AND FINCH, 7.17.7
11.1 Accelerating Reference Frames
One can only specify the position r
S
of a point mass with respect to a chosen refer-
ence frame S.
One can equally choose another reference frame B to specify the position r
B
of the
same point mass, displaced by R from S. The positions of the point mass relative to
the two frames are then related by r
B
= r
S
R.
If S is an inertial frame, where Newtons laws hold, but

R ,= 0, then B is not an inertial
frame. Hence, in frame S, m r
S
= F, but in frame B,
m r
B
= m( r
S


R) = F
..
true force
m

R.
..
inertial force
(89)
11.2 Rotated Reference Frames in 2D
(a)
x
S

y
S

x
B

y
B

r

(b)
x
B

y
B

r
(c)
x
y
r
S

r
B


(a) A new reference frame B is dened with respect to an original reference frame S
by an anticlockwise rotation through angle . In frame S, the position vector r has
positive x and y components.
(b) In frame B, r has a positive x component and a negative y component.
(c) An observer in frame B, which has been rotated anticlockwise, does not perceive
that frames rotation, as his or her observations are with respect to that same frame.
Instead he or she observes that r appears to have rotated clockwise through an
angle , from r
S
to r
B
.
37
11.3 Finite Rotations in 3D
O
N
P
Q
n

N
P
Q
r
S

n
V

Figure 5: Left: Overall and Right: top views of a general clockwise rotation of a vector OP = r
S
through an angle about an axis dened by the unit vector n, to become OQ = r
B
. This is what
we would observe if the reference frame we were in rotated in an anticlockwise direction.
We wish to determine OQ = ON + NV + VQ. Firstly,
ON = (n r
S
)n (component of r
S
in n direction). (90)
Noting that OP = ON + NP NP = r
S
(n r
S
)n, and that NQ has the same
magnitude as NP, it follows that
NV =
NP
[NP[
[NQ[ cos() = [r
S
(n r
S
)n] cos() (91)
We note that r
S
n has direction perpendicular to r
S
= OP and n (as does VQ),
and magnitude [r
S
n[ = [NQ[. Hence,
VQ =
(r
S
n)
[r
S
n[
[NQ[ sin() = (r
S
n) sin(). (92)
Gathering the terms calculated above, we have
OQ
..
r
B
=
ON
..
(n r
S
)n +
NV
..
[r
S
(n r
S
)n] cos() +
VQ
..
(r
S
n) sin(), (93)
which is generally rearranged slightly to obtain the rotation formula:
r
B
= r
S
cos() + (n r
S
)n[1 cos()] + (r
S
n) sin(). (94)
38
11.4 Innitesimal Rotations
Letting r
S
= r, r
B
= r + dr ( dr = r
B
r
S
), = d, and taking the small-angle limit,
the rotation formula [Eq. (94)] transforms to r + dr = r + (r n)d.
Hence, (noting that for any two vectors U V = V U) the velocity of reference
frame B relative to reference frame S is given by:
dr
dt
=
_
n
d
dt
_
r =
..
angular velocity
r. (95)
Note that even if r is constant in reference frame B, it is time dependent if viewed
from reference frame S.
11.5 Velocity and Acceleration
If the vector r does have a time-dependence in reference frame S (e.g. represents
the position of a moving particle), then this must also be accounted for:
_
dr
dt
_
B
. .
velocity in B
=
_
dr
dt
_
S
. .
velocity in S
r.
. .
velocity of B relative to S
(96)
This equation is completely general for any vector (not just position vectors).
Having determined v
B
= v
S
r, we take the time-derivative (in frame B)
and make use of Eq. (96) (with v substituted for r) to determine the acceleration
in frame B:
_
dv
B
dt
_
B
=
_
dv
S
dt
_
B
..
_
dv
S
dt
_
S
v
S

_
d
dt
_
B
r v
B
=
_
dv
S
dt
_
S
(v
B
+ r)
_
d
dt
_
B
r v
B
.
(97)
11.6 Inertial Forces in a Rotating Frame
If considering motion of a point mass, then multiplying Eq. (97) by m and noting that
F = m(dv
S
/dt)
S
, we deduce that
m r = F 2m r
. .
coriolis force
m ( r)
. .
centrifugal force
m r.
. .
Euler force
(98)
39
Example Class 3: Foucaults Pendulum
1. A Foucault pendulum is a heavy weight suspended by a cable from a bearing that does
not restrict the motion (unlike e.g. a typical clock pendulum, the motion of which is con-
strained to a plane), thus making it sensitive to the earths angular velocity . The original
demonstration by Foucault in Paris was in 1851, and caused quite a sensation.
If the oscillation amplitude is small compared to the length of the pendulum l (small angle
approximation), we can neglect motion in the vertical (z) direction, and approximate the
relevant forces in the (x, y) plane tangential to the earths surface by F = m
2
0
(x + y ),
where
2
0
= g/l, at the earths surface.
(a) (Study up to section 11.6). Incorporating the force F, determine the relevant equations
of motion as perceived when standing on the (rotating with constant angular velocity
vector = ) Earth.
(b) (Study up to section 12.2). Derive the specic form taken by these equations in the ap-
propriate local coordinate system ( pointing east, pointing north,

k pointing out from
the centre of the earth).
(c) (Study up to section 12.3). In what follows we will neglect the centrifugal force terms.
This is acceptable if
0
is signicantly larger than [[. Why?
(d) (Study up to section 12.6). Determine the general solution of the motion of the pendulum
bob in the plane tangential to the earths surface (to the assumed level of approxima-
tion) in terms of the initial position and velocity. The mathematics will be easier if you
consider the time evolution of the complex number (t) = x(t) + iy(t) (see section
A.4 of the Appendix in these notes on linear differential equations). Make sure that you
consistently drop all terms of the same size or smaller than the neglected centrifugal
force terms (to do this you will need the approximate identity

1 + 1 + /2 for
small ).
(e) (Study up to section 12.7). The pendulum is allowed to start oscillating after being held
at an amplitude A due east from the origin. What is the position of the pendulum bob
after each complete oscillation? Determine how many such oscillations must elapse
before the pendulum bob (approximately) returns to its original position. How many
hours does this correspond to:
i. At the north pole;
ii. At the equator;
iii. In Durham ( latitude 54

47
/
north)?
ADDITIONAL EXERCISE
2. What angular velocity would be required for your effective weight at the equator to be half
that at the north pole? (Take the earths radius to be 6378 km). How many minutes would a
full rotation of the earth take?
40
12 Inertial Forces on Earth
SEE ALSO HAND AND FINCH, 7.77.10
12.1 Procedure for Determining a Local Coordinate Systems
x
z
y

z
y



Figure 6: Left: Overall view of a local coordinate system in the region of a point on the surface of
the spherical earth. The positive x direction is east, the positive y direction is north, the positive z
direction is out from the centre of the earth. Right: Cross-sectional view, showing the local y and z
axes only (the x axis points perpendicularly into the page). is the latitude, and the colatitude
(in spherical coordinates, the angle of elevation).
1. We model the Earth as a perfect, solid sphere, and require an appropriate coor-
dinate system to describe motion near a point on its surface. We rst consider a
cartesian coordinate system with the z axis running through the poles, and the x
and y axes emerging through the equator [in this frame = (0, 0, )].
2. We then rotate the coordinate system around its x axis so that the z axis emerges at
a chosen point on the Earths surface [now = (0, sin(), cos())].
3. Finally, we displace the origin of the coordinate system by R to point where the z
axis emerges from the Earths surface (where R = [R[ is the radius of the Earth).
12.2 Inertial Forces in the Local Coordinate System
Note that the derivation of centrifugal, coriolis, and Euler force terms assumes r to
be dened with respect to a point on the axis of rotation.
41
The nal displacement of the local coordinate system origin means the associated
position vector r
/
does not begin on the axis of rotation.
One must therefore substitute r = r
/
+ R into the inertial force terms, where r
/
is the
position vector associated with the local coordinate system.
12.3 Centrifugal Force m ( r)
The direction of the centrifugal force can be determined quite generally: r is
perpendicular to , r, and ( r) is perpendicular to both and r.
Hence, the centrifugal force m ( r) points directly away from the axis of
rotation, meaning it can counteract gravity.
12.4 Effective Gravity on Earths Surface
(a) (b) (c)
(a) Direction and magnitude of the force due to earths gravitation on the surface of a
spherical earth model.
(b) Direction of the centrifugal force on the surface of a spherical earth model (magni-
tude relative to the gravitational force greatly exaggerated for effect).
(c) Effect of adding the gravitational and centrifugal forces together. The net force for
a static object does not point exactly to the centre of the earth except at the poles
and the equator.
12.5 Deviation of a Plumb Bob
Setting r
/
= (0, 0, 0), one obtains m [ (r
/
+ R)] =
m
2
R(0, sin() cos(), sin
2
()). Incorporating the effect of gravity,
one obtains the EOM
m r
/
= m(0,
2
Rsin() cos(), g
2
Rsin
2
()). (99)
42
We now consider the deection angle
d
, dened through
tan(
d
) =

2
Rsin() cos()
g
2
Rsin
2
()


2
Rsin() cos()
g
(g
2
R). (100)
For = /4, 3/4,
d
is maximised at 1.7 milliradians. Buildings at colatitude
/4 are therefore tilted by this amount from the true vertical if aligned with a
plumb bob or spirit level.
12.6 Coriolis Force 2m r
In the local coordinate system dened for the surface of the Earth, the coriolis force
is given by
2m r
/
= 2m(sin() z cos() y, cos() x, sin() x) (101)
Note that it is proportional to velocities, and therefore only comes into play when
there is motion.
One often considers phenomena when motion is effectively constrained to be in
plane, when we ignore the z dimension. In this case the coriolis force x component
2mcos() y, and the coriolis force y component 2mcos() x.
12.7 Loose Ends
The Euler force is not zero on Earth, but is commonly taken to be negligible.
In a rotating reference frame which is also accelerating translationally with acceler-
ation

R, the general EOM is given by
m r = F 2m r m ( r) m r m

R. (102)
43
13 Rotational Inertia, Angular Momentum and Kinetic Energy
SEE ALSO HAND AND FINCH, 8.18.3
13.1 DoF of Rigid Bodies
Arigid body is a systemof mass points (or a continuum) constrained so that distances
between all points are constant (e.g. connected by massless rigid rods). A system of
2(3) connected point masses has 5(6) DoF (see section 1.3).
Connecting a further point mass to 3 original point masses fully species its position
relative to the rigid body, and contributes 3 coordinates 3 constraints = 0 DoF. Any
rigid body more complex than two connected point masses has 6 DoF.
13.2 Displacement of a Rigid Body
Translating one point of a rigid body by R without changing the rigid bodys orien-
tation involves 3 DoF. A second point on the rigid body can then be rotated to any
location, specied by azimuthal and elevation angles (i.e., 2 DoF), on the surface of
an imaginary sphere with its centre at the location of the rst point.
Having specied the position of the second point, the whole rigid body can be
rotated about the axis formed by drawing a line through the rst and second point
(1 DoF). The most general displacement of a rigid body is a translation plus a rotation
(Chasles theorem).
Alternatively, subsequent to translation (3 DoF), specify an axis of rotation (2 DoF)
and a nite rotation about that axis (1 DoF). The most general displacement of a
rigid body with one point xed is a rotation about some axis (Eulers theorem).
13.3 Moment of Inertia I about an Axis of Rotation
Consider N mass points, rotating with angular velocity anticlockwise about an axis
of rotation (assumed to pass through the origin), where the direction of the axis of
rotation is dened by the unit vector n (i.e., the angular velocity vector = n).
For one mass point I is the square of its distance from the axis of rotation multiplied
by its mass m. For N mass points, I is the sum of such terms. Hence,
I =
N

k=1
m
k
[r
2
k
(r
k
n)
2
] =

k=1
m
k
[(r
k
n)n r
k
[
2
. (103)
13.4 Angular Momentum J and Rotational Kinetic Energy
The angular momentum (about the origin) of a system of mass points is given by
J =
N
k=1
r
k
p
k
=
N
k=1
m
k
r
k
r
k
, (if the point about which we calculate J is
R ,= 0, then we replace r
k
, r
k
by r
/
k
, r
/
k
, where r
/
k
= r
k
R).
44
If rigid body constraints are imposed, any velocity r
k
is due to rotation only.
r
k
is the velocity of the kth mass point due to its anticlockwise rotation (in
contrast to the perceived velocity r of objects appearing to rotate clockwise
due to the anticlockwise rotation of a reference frame).
Using the vector identity a (b c) = (a c)b (a b)c and recalling = n,
J =
N

k=1
m
k
[r
k
( r
k
)] =
N

k=1
m
k
[r
2
k
(r
k
)r
k
]
J n =
N

k=1
m
k
[r
2
k
(r
k
n)
2
] = I.
(104)
The kinetic energy reduces to T =
1
2

N
k=1
m
k
r
2
k
=

2
2

N
k=1
m
k
[n r
k
[
2
. Noting
that [n r
k
[ = [(r
k
n)n r
k
[, the rotational kinetic energy takes a simple form:
T =

2
2
N

k=1
m
k
[(r
k
n)n r
k
[
2
=
I
2
2
. (105)
13.5 Denition of the Inertia Tensor

I
From J =
N
k=1
m
k
[r
2
k
(r
k
)r
k
], we infer that
J
x
=
N

k=1
m
k
(r
2
k
x
2
k
)
x
m
k
x
k
y
k

y
m
k
x
k
z
k

z
,
J
y
=
N

k=1
m
k
(r
2
k
y
2
k
)
y
m
k
y
k
x
k

x
m
k
y
k
z
k

z
,
J
z
=
N

k=1
m
k
(r
2
k
z
2
k
)
z
m
k
z
k
x
k

x
m
k
z
k
y
k

y
,
(106)
and use the linear transformation relating the angular velocity vector to the an-
gular momentum vector J to dene the inertia tensor

I:
_
_
J
x
J
y
J
z
_
_
. .
J
=
_
_
I
xx
I
xy
I
xz
I
yx
I
yy
I
yz
I
zx
I
zy
I
zz
_
_
. .

I
_
_

z
_
_
. .

with I

=
N

k
m
k
(r
2
k

r
k,
r
k,
), where

=
_
1 if =
0 if ,=
(107)
For continuous bodies I

=
_
volume
dxdydz(x, y, z)(r
2

), ( is a mass density).
45
13.6 Generalized Description of the Rotational Kinetic Energy
Using the vector identity a (b c) = b (a c) = c (a b), the rotational
kinetic energy can be phrased more generally in terms of the moment of inertia
tensor as
T =
1
2
N

k=1
m
k
r
2
k
=
1
2
N

k=1
m
k
r
k
( r
k
) =
1
2

_
N

k=1
m
k
(r
k
r
k
)
_
.
. .
J =

I
(108)
Hence T = (1/2) (

I) = (
2
/2)n
T

In, where n
T

In is the moment of inertia
I about the axis dened by n.
13.7 Centre of Mass (CoM)
The centre of mass position and velocity for a system of N mass points (mass of kth
point = m
k
) are dened by (total mass M =
N
k=1
m
k
):
R
C
=

N
k=1
m
k
r
k
M
,

R
C
=

N
k=1
m
k
r
k
M
. (109)
It immediately follows that the total momentum P =
N
k=1
m
k
r
k
= M

R
C
is equivalent
to the momentum of a single particle of mass M =
N
k=1
m
k
moving with the CoM.
13.8 Separation off of the CoM Angular Momentum and Kinetic Energy
The corresponding total angular momentum about some point (situated at the
origin of the coordinate system) is J =
N
k=1
m
k
(r
k
r
k
).
We let r
k
= r
/
k
+ R
C
. Hence,
J =
N

k=1
m
k
(r
/
k
+ R
C
) ( r
/
k
+

R
C
)
=
N

k=1
m
k
(r
/
k
r
/
k
+ R
C
r
/
k
+ r
/
k


R
C
+ R
C


R
C
)
=
N

k=1
m
k
(r
/
k
r
/
k
) + MR
C


N
k=1
m
k
r
/
k
M
+

N
k=1
m
k
r
/
k
M
. .
CoM velocity and position with respect to CoM, i.e., = 0
M

R
C
+ MR
C


R
C
.
(110)
J =
N
k=1
m
k
(r
/
k
r
/
k
) + MR
C


R
C
can be expressed as the angular momentum
about the CoM plus the angular momentum of the CoM about the origin.
46
Similarly, the CoM kinetic energy, and the kinetic energy for motion with re-
spect to the CoM can be separated:
T =
1
2
N

k=1
m
k
r
2
k
=
1
2
N

k=1
m
k
r
k
r
k
=
1
2
N

k=1
m
k
( r
/
k
+

R
C
) ( r
/
k
+

R
C
)
=
1
2
N

k=1
m
k
r
/2
k
+

R
C

k=1
m
k
r
/
k
. .
= 0
+
1
2
M

R
2
C
.
(111)
In general, if the only external forces acting on a rigid body act through the CoM
(e.g. gravity), the CoM motion can be considered separately.
Homework 5: Rotating Book
We consider the torque-free motion of a general right angled parallelepiped or cuboid (e.g., a
book), of uniform density.
1. (Study up to section 14.6). Derive the principal moments of inertia of a general cuboid of
height 2a, width 2b and depth 2c (all unequal), and having a uniform mass density .
2. (Study up to section 15.1). Noting that there are no external Torques, show that the system is
in equilibrium (meaning that never changes) when
2
=
3
= 0,
1
,= 0. Determine the
other two equilibrium solutions.
3. (Study up to section 15.1). Consider a near equilibrium solution, = (
1
,
2
,
3
), such that,
as
2
and
3
are considered small,
2

3
0. Employing this approximation, show that

2
=
2

2
, in the process deriving an expression for
2
.
4. (Study up to section 15.2). What are the general solutions of the equation of motion

2
(t) =

2
(t) in terms of
2
(0) and

2
(0), for the two separate cases where
2
> 0, and where

2
< 0? (For convenience, use the parameter

in your answer, where

2
= [
2
[ is dened
positive).
5. (Study up to section 15.2). We say that the equilibrium solution is unstable if, when the initial
angular velocity vector (0) is very slightly different from the equilibrium solution and its
initial rst time-derivative (0) = 0, then the subsequent time-evolution of (t) causes it to
move rapidly away from the equilibrium solution. When is the equilibrium solution under
consideration unstable?
47
14 The Parallel and Principal Axis Theorems
SEE ALSO HAND AND FINCH, 8.2
14.1 Displaced Axis Theorem
In a similar fashion to the angular momentum and rotational kinetic energy, one
can insert the CoM position R
C
into the elements of the Inertia tensor I

:
I

=
N

k=1
m
k
(r
2
k

r
k,
r
k,
) =
N

k=1
m
k
[[r
/
k
+ R
C
[
2

(r
/
k,
+ R
C,
)(r
/
k,
+ R
C,
)]
=
N

k=1
m
k
[(r
/2
k
+ 2r
/
k
R
C
+ R
2
C
)

(r
/
k,
r
/
k,
+ r
/
k,
R
C,
+ R
C,
r
/
k,
+ R
C,
R
C,
)]
=
N

k=1
m
k
(r
/2
k

r
/
k,
r
/
k,
) + M(R
2
C

R
C,
R
C,
)
+ 2R
C

k=1
m
k
r
/
k

. .
=0
R
C,
N

k=1
m
k
r
/
k,
. .
=0
R
C,
N

k=1
m
k
r
/
k,
. .
=0
(112)
As before, we have let r
k
= r
/
k
+ R
C
, and used the fact that the CoM position
with respect to the CoM is = 0.
Hence, the inertia tensor of a rigid body, dened with respect to rotations about
the origin can be related to the inertia tensor of the same rigid body, dened
with respect to rotations about its CoM,through

I =

I
CoM
+ M

A, (113)
where

A can be represented by a matrix, the elements of which are determined
by the elements of the CoM position vector: A

= R
2
C

R
C,
R
C,
. This is
the displaced axis theorem.
14.2 Parallel Axis Theorem
Consider a moment of inertia I
COM
about an axis of rotation passing through the
CoM. This is equivalent to considering a diagonal element of some representation
of the inertia tensor.
From the displaced axis theorem, the moment of inertia I about a parallel axis is
I = I
CoM
+ MR
2
C
, (114)
where R
C
is the distance of parallel axis from the axis of rotation passing through the
CoM. This is the parallel axis theorem.
48
14.3 Symmetric Nature of the Inertia Tensor
The inertia tensor

I is dened for a particular rigid body, and for rotations of that
body about a particular point (frequently the centre of mass).
In any representation (e.g. for some set of cartesian axes, conveniently chosen with
the origin at the point of rotation) the inertia tensor is represented by a symmetric
matrix, i.e., I

= I

. We say

I is a symmetric tensor.
The matrix elements depend on the chosen representation (the tensor is dened
representation free). We wish to know if there is a diagonal matrix representation of
the inertia tensor.
14.4 Orthogonal Matrices and Orthogonal Diagonalization
The inverse

P
1
of a matrix

P is dened such that

P
1

P =

P

P
1
=

1, the
identity matrix. A matrix

P for which an inverse

P
1
exists is called invertible.
A real invertible square matrix

P such that its inverse

P
1
=

P
T
, its transpose, is
said to be an orthogonal matrix. (So-called because its rows form an orthonormal
set, and its columns form an orthonormal set).
A general 2 2 orthogonal matrix has a single free parameter , and describes
a 2-dimensional rotation when acting on a 2-dimensional column vector:

P =
_
cos sin
sin cos
_
. (115)
The inverse matrix describes a rotation back in the opposite direction.
Similarly, in 3 dimensions, one can consider separate rotations about each axis:

P
x
=
_
_
1 0 0
0 cos sin
0 sin cos
_
_
,

P
y
=
_
_
cos 0 sin
0 1 0
sin 0 cos
_
_
,

P
z
=
_
_
cos sin 0
sin cos 0
0 0 1
_
_
.
(116)
(

P
z

P
y

P
x
)
T

P
z

P
y

P
x
=

P
T
x

P
T
y

P
T
z

P
z

P
y

P
x
=

1, and the product

P
z

P
y

P
x
is a way of
formulating a general 3 3 orthogonal matrix, describing a general rotation in
3-dimensions.
A square matrix

A is called orthogonally diagonalizable if there is an invertible
matrix

P such that

P
1

A

P =

P
T

A

P =

D is diagonal;

P is said to orthogonally di-
agonalize

A. If

Ais an NN matrix, then the following are equivalent. (a)

Ais
orthogonally diagonalizable; (b)

A has an orthonormal set of N eigenvectors;
(c)

A is symmetric.
49
Consider the 3 3 case. Let us say that a 3 3 matrix

A has 3 orthonormal
eigenvectors p
1
= (p
11
, p
21
, p
31
), p
2
= (p
12
, p
22
, p
32
), p
3
= (p
13
, p
23
, p
33
) with
corresponding eigenvalues
1
,
2
,
3
.
Let

P be the matrix whose columns are these eigenvectors:

P =
_
_
p
11
p
12
p
13
p
21
p
22
p
23
p
31
p
32
p
33
_
_
(117)
This implies that

A

P =
_
_

1
p
11

2
p
12

3
p
13

1
p
21

2
p
22

3
p
23

1
p
31

2
p
32

3
p
33
_
_
=
_
_
p
11
p
12
p
13
p
21
p
22
p
23
p
31
p
32
p
33
_
_
. .

P
_
_

1
0 0
0
2
0
0 0
3
_
_
. .

D
.
(118)
Since the columns of

P are orthonormal,

P is an orthogonal matrix. Hence,

A

P =

P

D

P
1

A

P =

P
T

A

P =

D, i.e.,

A is orthogonally diagonalizable.
To help see that

A must be symmetric, i.e., that

A
T
=

A, we use

D =

P
1

A

P =

P
T

A

P

P

D

P
1
=

P

D

P
T
=

A. Hence,

A
T
= (

P

D

P
T
)
T
=

P

D
T

P
T
=

P

D

P
T
=

A. (119)
14.5 Principal Axis Theorem


I is a symmetric tensor, and is always represented by a 3 3 symmetric matrix. There-
fore,

I has three eigenvalues (which may be repeated). These are the principal
moments of inertia, called I
1
, I
2
, I
3
.


I must also have three orthonormal eigenvectors (in the case of repeated eigenval-
ues, there is some freedom in choosing what form the corresponding eigenvectors
should take). The directions the eigenvectors point in determine the principal axes,
labelled 1, 2, 3.
It is always possible to nd a rotated coordinate system such that the representa-
tion of

I is diagonal. The diagonal elements are the principal moments of inertia
(eigenvalues of

I), the coordinate system is set by the principal axes (eigenvectors
of

I), and a principal moment of inertia I
k
is the moment of inertia associated with
rotation about the principal axis k.
Rotating the coordinate system such that the representation of

I is diagonal is a
principal axis transformation, and that this is always possible is the principal axis the-
orem.
50
14.6 Deducing the Principal Axes
Figure 7: Principal axes of an asymmetric ellipsoid, and an asymmetric parallelepiped, for rotation
about the centre of mass. The principal moments of inertia have been chosen to be identical for
the two objects.
For highly symmetric objects, it is often possible to guess a coordinate system for
which all off-diagonal elements of

I are = 0.
One can then deduce the principal moments of inertia directly, without having to
carry out a principal axis transformation.
51
15 Rigid Body Dynamics and Stability
SEE ALSO HAND AND FINCH, 8.48.5
15.1 Eulers Equations of Motion
We consider N mass points subject to rigid-body constraints, and take the time
derivative of the angular momentum vector J =
N
k=1
m
k
(r
k
r
k
), to get

J =
N

k=1
m
k
( r
k
r
k
)
. .
= 0, as a a = 0 for any vector a
+
N

k=1
m
k
(r
k
r
k
) =
N

k=1
r
k
F
k
.
. .
total torque N
(120)
Hence, we can say, viewing the rigid body from outside (space axes/coordinate
system) (dJ/dt)
S
= N.
Recalling Eq. (96) [that, for any vector a, (da/dt)
B
= (da/dt)
S
a], viewing the
rigid body while travelling with its rotational motion (body axes/coordinate system)
we can say that (dJ/dt)
B
+ J = N.
We assume a coordinate system dened by the principal axes, i.e., the decomposi-
tion J = (I
1

1
, I
2

2
, I
3

3
), and thus derive Eulers equations of motion:
I
1

1

3
(I
2
I
3
) = N
1
,
I
2

2

1
(I
3
I
1
) = N
2
,
I
3

3

2
(I
1
I
2
) = N
3
.
(121)
15.2 Torque-Free Motion
If there are no external torques (e.g. a freely spinning object moving through the
Earths gravitational eld), then we can set N
1
= N
2
= N
3
= 0.
Note that the total angular momentum (about the CoM) and the total kinetic en-
ergy (associated with motion about the CoM) are both conserved.
J
2
= J
2
1
+ J
2
2
+ J
2
3
describes a sphere in angular momentum space, and T = J
2
1
/2I
1
+
J
2
2
/2I
2
+ J
2
3
/2I
3
describes an ellipsoid.
Where the sphere and ellipsoid intersect shows the possible trajectories for the an-
gular momentum components J
k
=
k
I
k
.
52
Figure 8: Left: Contours of constant T on a sphere described by J
2
= 1, for a system where
I
1
> I
2
> I
3
. Rotation about principal axis 2 is unstable, rotation about principal axes 1 and 3 is
stable. Right: Equivalent diagram for the case where I
1
= I
2
> I
3
. Principal axes 1 and 2 can be
rotated freely about principal axis 3, and there are no points on the J
2
= 1 sphere about which
the rotation is unstable.
53
A Second-Order Ordinary Linear Differential Equations with
Constant Coefcients
A.1 Denition
A second-order homogeneous ordinary linear differential equation with constant coefcients has
the form:
a
d
2
x
dt
2
+ b
dx
dt
+ cx = 0 (122)
This has been written to appear to be a differential equation of a position variable with respect to
time, as this is most obviously relevant for classical mechanics.
A.2 Solution
A.2.1 Auxiliary Equation
The recipe for solving such a linear differential equation begins by substituting e
t
for x. Thus
a
2
e
t
+ be
t
+ ce
t
= 0 (123)
which, cancelling e
t
everywhere, becomes
a
2
+ b + c = 0. (124)
This is the auxiliary equation. One can then solve the auxiliary equation by nding the roots from
the quadratic formula:

=
b

b
2
4ac
2a
. (125)
Thus
(x
+
)(x

) = 0, (126)
i.e., assuming the roots are not repeated, there are 2 distinct values of solving the auxiliary
equation, which may be complex or imaginary.
A.2.2 General Solution (Roots not Repeated)
The general solution is then a linear superposition of the two particular solutions e

+
t
and e

t
,
i.e.,
x(t) = A
+
e

+
t
+ A

t
(127)
where the A
+
, A

are arbitrary constants (because we are dealing with a second-order differential


equation, there must be two integration constants in its general solution), often determined by the
initial position:
x(0) = A
+
+ A

(128)
and the initial velocity, through
x(t) = A
+

+
e

+
t
+ A

t
x(0) = A
+

+
+ A

. (129)
The roots
+
and

may be imaginary if this is the case and x is a real quantity (such as po-
sition) then A
+
and A

must have (complex) values such that the overall imaginary contribution
is zero.
I
A.3 Possible Complications
A.3.1 Repeated Roots
If
+
=

= , then e
t
solves the differential equation, and a second linearly independent
solution is given by te
t
. The general solution is thus given by
x(t) = Ae
t
+ Bte
t
. (130)
A.3.2 Additional Constant
If the differential equation has the form
a
d
2
x
dt
2
+ b
dx
dt
+ cx + k = 0 (131)
then this can be rephrased as
a
d
2
x
dt
2
+ b
dx
dt
+ c
_
x +
k
c
_
= 0. (132)
One can then dene a new variable x = x + k/c, for which it follows
a
d
2
x
dt
2
+ b
d x
dt
+ c x = 0. (133)
One then nds the general solution for x(t), as shown above, and the general solution for x(t)
from it, by x(t) = x(t) k/c.
A.4 Representing Position in a Plane by a Single Complex Number
If we have a situation where, for example, we have coupled motion in the x, y plane, i.e., the x
and y variables cannot be separated:
d
2
x
dt
2
+ a
x
dx
dt
+ b
x
dy
dt
+ c
x
x + d
x
y + k
x
= 0, (134)
d
2
y
dt
2
+ a
y
dx
dt
+ b
y
dy
dt
+ c
y
x + d
y
y + k
y
= 0, (135)
then we can try to simplify matters by considering the complex variable = x + iy, and adding
the bottomequation multiplied by i to the top equation. We then consider the differential equation
given by
d
2

dt
2
+ =
d
2
x
dt
2
+ + i
_
d
2
y
dt
2
+
_
. (136)
In general, this will produce an equation containing terms proportional to both and its complex
conjugate

, as well as their rst derivatives, which in no way simplies anything! However, in


a number of cases of physical interest, the result takes the form
d
2

dt
2
+
d
dt
+ + = 0, (137)
where , , are (in general) complex constant coefcients, i.e., there is no coupling to the complex
conjugate variable

. If this is the case, then the general solution for (t) is determined using
exactly the same methodology as described on the rst two pages. Having determined the general
II
solution for (t), one derives the general solution for x(t) by taking the real part of (t), and the
general solution for y(t) by taking the imaginary part.
The very substantial simplication of the mathematics offered by such an approach means it is
at least worth considering. Otherwise it is necessary to use methods developed for linear systems
of equations; this will also be necessary if motion in the x, y, and z directions cannot be sepa-
rated. Extension to higher-order linear differential equations is straightforward. See essentially
any textbook on ordinary differential equations for further details, if interested.
B Taylor Series
B.1 Taylors Theorem of the Mean
We consider a function of N variables f (x
1
, x
2
, . . . , x
N
). This could, for example, be a Lagrangian
describing the motion of n objects, which would be considered a function of n generalized coor-
dinates and n generalized velocities (which are considered independent variables), making for a
total of N = 2n variables. Or it could be just the potential energy term of such a Lagrangian, with
N = n variables.
If, within a region (i.e., some volume of n-dimensional space) all possible partial derivatives
of f exist, Taylors theorem of the mean implies
f (x
01
+ a
1
, . . . , x
0N
+ a
N
) =f (x
01
, . . . , x
0N
)
+

k=1
1
k!
_
a
1

x
1
+ + a
N

x
N
_
k
f (x
1
, . . . , x
N
)

x
01
,...,x
0N
(138)
(Taylors theorem of the mean is in fact a more general result see essentially any moderately
advanced calculus text if interested). It is also assumed that the function f is well-dened over
the region of interest (in particular, there should be no innities), i.e., the above series converges
within this region. The vertical line at the right-hand-side of the derivative terms means that the
derivative is rst calculated, and then evaluated at x
1
= x
01
, x
2
= x
02
, etc..
B.2 Taylor Series in One Variable
If we restrict ourselves to f being a function of just one variable f (x), then
f (x
0
+ a) =f (x
0
) +

k=1
1
k!
a
k

k
x
k
f (x)

x
0
.
(139)
If we write the above in a form where a = x x
0
, then we get the usual form of a single-variable
Taylor series:
f (x) =f (x
0
) +

k=1
1
k!
_
_

k
x
k
f (x)

x
0
_
_
(x x
0
)
k
. (140)
Note that, for example, the function f (x) = x
1
cannot be expanded around x
0
= 0, as this point
is singular [ f (x) as x 0]. Any value of x
0
slightly different from zero is acceptable.
III
B.3 Taylor Series in Two Variables
It is not so easy to write down such a compact formulation for a Taylor series in two or more
variables, but the procedure is essentially the same. Bearing in mind that we are frequently only
interested in (at most) the rst- and second-order derivative terms, we write down
f (x
0
+ a, y
0
+ b) =f (x
0
, y
0
) +
_
a

x
+ b

y
_
f (x, y)

x
0
,y
0
+
1
2
_
a
2

2
x
2
+ 2ab

2
xy
+ b
2

2
y
2
_
f (x, y)

x
0
,y
0
+ .
(141)
Similarly to the case for a single-variable Taylor series, we set a = x x
0
and b = y y
0
, yielding
f (x, y) =f (x
0
, y
0
) +
__

x
f (x, y)

x
0
,y
0
_
(x x
0
) +
_

y
f (x, y)

x
0
,y
0
_
(y y
0
)
_
+
1
2
__

2
x
2
f (x, y)

x
0
,y
0
_
(x x
0
)
2
+ 2
_

2
xy
f (x, y)

x
0
,y
0
_
(x x
0
)(y y
0
)
+
_

2
y
2
f (x, y)

x
0
,y
0
_
(y y
0
)
2
_
+ .
(142)
B.4 Taylor Series in Three Variables
Proceeding in an exactly analogous fashion, we determine for a function of three variables, f (x, y, z),
that
f (x, y, z) =f (x
0
, y
0
, z
0
)
+
__

x
f (x, y, z)

x
0
,y
0
,z
0
_
(x x
0
) +
_

y
f (x, y, z)

x
0
,y
0
,z
0
_
(y y
0
)
+
_

z
f (x, y, z)

x
0
,y
0
,z
0
_
(z z
0
)
_
+
1
2
__

2
x
2
f (x, y, z)

x
0
,y
0
,z
0
_
(x x
0
)
2
+
_

2
y
2
f (x, y, z)

x
0
,y
0
,z
0
_
(y y
0
)
2
+
_

2
z
2
f (x, y, z)

x
0
,y
0
,z
0
_
(z z
0
)
2
+ 2
_

2
xy
f (x, y, z)

x
0
,y
0
,z
0
_
(x x
0
)(y y
0
)
+ 2
_

2
xz
f (x, y, z)

x
0
,y
0
,z
0
_
(x x
0
)(z z
0
)
+2
_

2
yz
f (x, y, z)

x
0
,y
0
,z
0
_
(y y
0
)(z z
0
)
_
+ .
(143)
This can be extended to functions of more than three variables, and to include higher-order
derivatives, as desired, in a reasonably straightforward fashion.
IV
C Determinants, Eigenvalues and Eigenvectors
C.1 Symmetric Matrices
In classical mechanics, we can usually restrict ourselves to considering real, symmetric matrices,
i.e., matrices with real entries that are equal to their own transpose. Formally, we say a matrix

A
is symmetric if

A =

A
T
. (144)
In the case of a general 2 2 matrix

B

B
T
=
_
a b
c d
_
T
=
_
a c
b d
_
. (145)
Hence, if

B is to equal

B
T
, it follows that b must equal c, i.e., the most general 2 2 symmetric
matrix has the form

B =
_
a b
b d
_
. (146)
C.2 Determinants
The determinant of the matrix

B is given by
det(

B) =

a b
c d

= ad bc. (147)
There is considerable freedom in the choice of steps to take when calculating the determinant of
a 3 3 matrix

C. The most conventional formulation is
det(

C) =

c
11
c
12
c
13
c
21
c
22
c
23
c
31
c
32
c
33

= c
11

c
22
c
23
c
32
c
33

c
12

c
21
c
23
c
31
c
33

+ c
13

c
21
c
22
c
31
c
32

. (148)
We thus have a sum of 3 terms involving determinants of 2 2 matrices, which we already know
how to calculate.
Similarly, the determinant of an n n matrix

D is given by a sum of n terms involving deter-
minants of (n 1) (n 1) matrices. These (n 1) (n 1) matrices are conventionally formed
by removing the 1st row and the kth column from the matrix

D. In this case, the determinant is
formulated by
det(

D) =

d
11
d
12
d
1n
d
21
d
22
d
2n
.
.
.
.
.
.
.
.
.
.
.
.
d
n1
d
n2
d
nn

=
n

k=1
(1)
1+k
d
1k

d
2,k1
d
2,k+1

.
.
.
.
.
.
.
.
.
.
.
.
d
n,k1
d
n,k+1

. (149)
C.3 The Conventional Eigenvalue Problem
If we take a matrix

A, a vector v, and a scalar , such that

Av = v, (150)
V
we say that v is an eigenvector of the matrix

A, with eigenvalue . It obviously follows that

Av v = 0 (

A

1)v = 0, (151)
where

1 is the identity matrix [i.e., the matrix with ones down the main (top-left-to-bottom-right)
diagonal and zeros everywhere else]. It so happens that this can only be possible if the determi-
nant of the matrix (

A

1) is zero, discounting the trivial solution v = 0.


The procedure for nding the eigenvalues associated with the matrix

A then begins by calcu-
lating the determinant of (

A

1), and setting it equal to zero. If



A is a real, symmetric n n
matrix, this yields an nth-degree polynomial in which always has n real roots (although roots
may be repeated, or degenerate).
We will work this through for a real 2 2 symmetric matrix

B =
_
a b
b d
_
. (152)
To determine the eigenvalues of

B, we calculate

a b
b d

= (a )(d ) b
2
=
2
(a + d) + (ad b
2
) = 0. (153)
The values of that solve this equation can in this case be generally determined fromthe quadratic
formula. There are two solutions,
+
and

, given by

=
(a + d)
_
(a d)
2
+ b
2
2
. (154)
The eigenvector v
+
associated with the eigenvalue
+
can then be determined from
_
a
+
b
b d
+
__
v
1+
v
2+
_
=
_
0
0
_
(155)
which implies
(
+
a)v
1+
=bv
2+
(156)
(
+
d)v
2+
=bv
1+
. (157)
It is common to choose v
1+
= 1, and then normalize the result. Hence, from Eq. (157)
v
+
=
1
_
1 + (
+
d)
2
/b
2
_
1
(
+
d)/b
_
, (158)
and equivalently, for the determining of v

. Because the polynomial (a


+
)(d
+
) b
2
= 0
xes a relationship between the matrix components a, b, d, and the eigenvalue
+
, Eq. (156) and
Eq. (157) are not independent. This is why we have ignored Eq. (156) when determining v
+
.
When considering 3- and higher-dimensional vectors, some more involved algebraic manip-
ulation is usually required to produce isolated expressions for the eigenvector components. For
an n n matrix, we will end up with n equations like Eq. (156) and Eq. (157), and we can always
ignore one of them as being not independent.
If, for example two roots in the characteristic polynomial are identical (repeated roots), then we
can determine two independent (i.e., orthogonal) eigenvectors with the same eigenvalue. As any
linear combination of the two will also be an eigenvector with the same eigenvalue, there is some
freedom in choosing what form the two orthogonal eigenvectors should take. See essentially any
linear algebra text for more details and proofs of the above results.
VI
C.4 Generalized Eigenvalue Problem
It can happen (for example when determining normal modes) that we have a situation which is
expressed more like
(

A

B)v = 0. (159)
where

B is a matrix which is not necessarily proportional to the identity (although it will still be
symmetric, in the case of determining normal modes). It is still the case that the determinant of
(

A

B) must be zero for this equation to be nontrivially solvable.
This is a generalized eigenvalue problem, and the eigenvalues and eigenvectors are deter-
mined in much the same way, i.e., setting the determinant = 0, nding the roots of the resulting
polynomial in order to determine the eigenvalues, then using the calculated values of the eigen-
values to determine the relationships between the components of their corresponding eigenvec-
tors, and nally (usually) normalizing the eigenvectors.
VII
D Exam Questions and Solutions
D.1 Coriolis Force (2006)
D.1.1 Question
The general expression for the coriolis force in a frame of reference rotating with angular velocity
, acting on an object of mass m moving with velocity r is 2m r
(a) With the aid of a diagram or diagrams, show that in a local coordinate system oriented
with respect to a point on the surface of the earth (x-axis points east, y-axis points north,
z-axis points out from the centre of the earth), the angular velocity vector takes the form
(0, sin(), cos()) ( is the magnitude of the angular velocity and the colatitude). [4
marks]
(b) Within such a local coordinate system, derive the equations of motion in the x and y direc-
tions for a shell of mass m red horizontally from large gun. Neglect both the centrifugal
force, and vertical motion (motion in the z direction). [6 marks]
(c) Where (t) = x(t) + iy(t), show that the solution to the combined equation of motion is [6
marks]
(t) = (0) +
i

(0)
2cos()
[exp(i2cos()t) 1]. (160)
(d) Recalling that the centrifugal force has magnitude m
2
Rsin() (R is the radius of the earth),
explain briey the justication for neglecting the centrifugal force, while retaining the cori-
olis force. Express (t) in a form which is consistent with this approximation. [4 marks].
D.1.2 Solution
(a) See Lecture 12, specically 12.1 and 12.2.
(b) m r = 2m r, where
r =



k
0 sin() cos()
x y z

. (161)
Hence, x = 2[sin() z cos() y], which is = 2cos() y if z = 0, and y = 2cos() x.
(c) (There are a number of equally correct ways to answer this question. A typical method is as
follows).
Hence, x + i y =

= i2cos()

. If we replace with the trial solution e
t
, this yields

2
= i2cos() [ + i2cos()] = 0. (162)
Thus, = 0 or i2cos(), and the general solution is given by
(t) = e
i2cos()t
+ , (163)
where and are two integration constants. It immediately follows that (0) = + , and
from

(t) = i2cos()e
i2cos()t
(164)
VIII
that

(0) = i2cos(). Hence
=
i

(0)
2cos()
, = (0)
i

(0)
2cos()
, (165)
from which the solution follows.
(d) One neglects the centrifugal force (
2
) as being small compared to the coriolis force ( )
if can be considered small on the relevant timescales. Thus, to be consistent, we must
neglect all terms
2
,
3
, . . .. Expanding the exponential in the solution,
(t) =(0) +
i

(0)
2cos()
[1 i2cos()t 2
2
cos
2
()t
2
+ (higher order terms in ) 1]
=(0) +

(0)[t icos()t
2
].
(166)
D.2 Small Oscillations (2007)
D.2.1 Question
Consider a small block of mass m (considered to be a mass point), sliding without friction on
the inside of a parabolic bowl of curvature c and mass M. The bowl slides without friction on a
horizontal surface. Consider a coordinate system such that the origin is at the initial position of
the bottom of the bowl. We consider side-to-side and vertical (x
S
and y
S
) motions of the small
block, and side-to-side (x
B
) motion of the bowl only.
(a) Using the generalized coordinates x
B
, x
R
= x
S
x
B
and the constraint y
S
= cx
2
R
/2, show
that the Lagrangian can be written
L =
m
2
( x
2
R
+ 2 x
R
x
B
+ c
2
x
2
R
x
2
R
) +
m + M
2
x
2
B

mgc
2
x
2
R
, (167)
where g is the gravitational acceleration. [4 marks]
(b) We will neglect the quartic term c
2
x
2
R
x
2
R
. For what kind of motion is this approximation
justied? [2 marks]
(c) Showthat the centre of mass coordinate (for side-to-side motion) x
C
= x
B
+ [m/(m+ M)]x
R
.
[2 marks]
(d) Rewrite the approximate Lagrangian L (i.e., neglecting the quartic term) in terms of x
R
, x
C
,
x
R
, x
C
, and in separated form (i.e., without cross-terms combining the different coordinates
or their velocities). From this, determine the equations of motion for x
R
and x
C
using the
Euler-Lagrange equations:
d
dt
_
L
q
_

L
q
= 0 (168)
(q = x
R
or x
C
in this case). From the solutions to these equations, determine the position of
the bowl x
B
(t) when both the bowl and the small block are initially at rest, in terms of x
C
(0)
and x
R
(0). [10 marks]
(e) Your equation of motion for the side-to-side centre of mass coordinate should show that its
velocity never changes. Is this expected? Give a physical explanation for your answer. [2
marks]
IX
D.2.2 Solution
(a)
L = T V =
m
2
( x
2
S
+ y
2
S
) +
M
2
x
2
B
mgy
S
. (169)
Note y
S
= cx
2
R
/2 y
S
= cx
R
x
R
, and x
R
= x
S
x
B
x
S
= x
R
+ x
B
x
S
= x
R
+ x
B
.
Hence,
L =
m
2
[( x
R
+ x
B
)
2
+ c
2
x
2
R
x
2
R
] +
M
2
x
2
B

mgc
2
x
2
R
(170)
which can be rearranged to give the desired result.
(b) This is a small oscillations approximation, hence is justied when motion does not stray
too far from an equilibrium point.
(c) By denition,
x
C
=
Mx
B
+ mx
S
M + m
. (171)
Substituting in x
S
= x
R
+ x
B
then gives the desired result.
(d) Neglecting the quartic term and using x
B
= x
C
m x
R
/(M + m),
L =
m
2
x
2
R
+ m x
R
_
x
C

m x
R
m + M
_
+
(m + M)
2
_
x
C

m x
R
m + M
_
2

mgc
2
x
2
R
=
mM
m + M
x
2
R
2
+ (m + M)
x
2
C
2

mgc
2
x
2
R
.
(172)
The Euler-Lagrange equations then yield x
C
= 0,, x
R
= [(m + M)/M]gcx
R
. Therefore
x
C
(t) = x
C
(0) + x
C
(0)t, and
x
R
(t) = x
R
(0) cos
_
_
m + M
m
gct
_
+ x
R
(0)
_
m
(m + M)gc
sin
_
_
m + M
m
gct
_
. (173)
If initially at rest, x
C
(0) =

X
R
(0) = 0, and hence
x
B
(t) = x
C
(t)
m
m + M
x
r
(t) = x
C
(0)
m
m + M
cos
_
_
m + M
m
gct
_
x
R
(0) (174)
(e) This is expected because there are no external forces acting in the x direction on the small
block and bowl system as a whole.
D.3 Stable and Unstable Rigid-Body Rotation (2007)
D.3.1 Question
Consider the torque-free motion of an asymmetric top rotating about its centre of mass, for exam-
ple a (non-square) book spinning while falling freely in Earths gravitational eld.
(a) Eulers equations of motion for rigid bodies are in this case given by
I
1

1

3
(I
2
I
3
) =0, (175)
I
2

2

1
(I
3
I
1
) =0, (176)
I
3

3

2
(I
1
I
2
) =0, (177)
X
where I
1
, I
2
, I
3
, are the principal moments of inertia and
1
,
2
,
3
are the components of
the angular velocity vector . How is the coordinate system in which these quantities are
dened related to our outside view (watching the book spinning while it is falling to the
Earths surface)? [2 marks].
(b) Show that the system is in equilibrium (meaning that never changes) when
2
=
3
= 0,

1
,= 0. Determine the other two equilibrium solutions. [4 marks]
(c) Consider a near equilibrium solution, = (
1
,
2
,
3
), such that, as
2
and
3
are considered
small,
2

3
0. Employing this approximation, show that

2
=
2

2
, where

2
=
2
1
(I
1
I
2
)(I
3
I
1
)
I
3
I
2
. (178)
Note that
2
< 0 means that is pure imaginary. [4 marks]
(d) Solve the above equation of motion for
2
(t) in terms of
2
(0) and

2
(0) for the two separate
cases where
2
> 0, and where
2
< 0. [6 marks]
(e) We say that the equilibrium solution is unstable if, when the initial angular velocity vec-
tor (0) is very slightly different from the equilibrium solution and its initial rst time-
derivative (0) = 0, then the subsequent time-evolution of (t) causes it to move rapidly
away from the equilibrium solution. For which of the following relative magnitudes of the
principal moments of inertia
I
1
> I
2
> I
3
, I
1
> I
3
> I
2
, I
2
> I
1
> I
3
, I
2
> I
3
> I
1
, I
3
> I
1
> I
2
, I
3
> I
2
> I
1
,
(179)
is the equilibrium solution under consideration unstable, and for which is it stable? [4
marks]
D.3.2 Solution
(a) The coordinate system rotates with the book, and has as its origin the (moving) centre of
mass of the book.
(b) Substituting = (
1
, 0, 0) into Eulers equations of motion, we see that I
1

1
= I
2

2
=
I
3

3
= 0, hence never changes. The other equilibriumsolutions are = (0,
2
, 0), (0, 0,
3
).
(c) Taking the time derivative of the second Euler equation (for I
2

2
), we get I
2

2
= (

1
+

3

1
)(I
3
I
1
). However, looking at the rst Euler equation, we see that I
1

1
=
2

3
(I
2

I
3
) = 0
1
= 0. Hence I
2

2
=

1
(I
3
I
1
). Finally, substitute in the third Euler equation
(for I
3

3
) to get

2
=
2
1
(I
1
I
2
)(I
3
I
1
)
I
3
I
2

2

2
=
2
1
(I
1
I
2
)(I
3
I
1
)
I
3
I
2
.
(d) We dene

2
= [
2
[ as a convenient shorthand. For
2
< 0, the differential equation is
that for a harmonic oscillator

2
=

2
, which is solved by
2
(t) = Acos(

t) +Bsin(

t),
where
2
(0) = A and

2
(0) =

B. Hence,

2
(t) = (0) cos(

t) + (

(0)/

) sin(

t).
XI
For
2
> 0 (when the differential equation becomes that for an inverted oscillator,

2
=

2
), then
2
(t) = Ae

t
+ Be

t
, where
2
(0) = A + B and

2
(0) =

(AB), from which
A = (
2
(0) +

2
(0)/

)/2, B = (
2
(0)

2
(0)/

)/2. Hence

2
(t) =

2
(0) +

2
(0)/

2
e

t
+

2
(0)

2
(0)/

2
e

t
,
with a possible alternative given by

2
(t) = (0) cosh(

t) + (

(0)/

) sinh(

t). (180)
(e) The dynamics are unstable when the linearized dynamics are inverted oscillator like, i.e.,
when
2
> 0. This occurs when I
2
> I
1
> I
3
, or when I
3
> I
1
> I
2
(1 mark), i.e., when I
1
is
intermediate in value between I
2
and I
3
. Otherwise the dynamics are stable.
D.4 Diatomic Molecule (2008)
D.4.1 Question
A simple model for a diatomic molecule (such as H
2
) consists of two mass points connected by
a spring. We constrain the motion to a plane, considering two mass points, each of mass m,
connected by a spring of spring constant k and natural length d, sliding on a at, frictionless
surface. The cartesian coordinates of the two mass points on the plane are x
1
, y
1
, and x
2
, y
2
,
respectively.
(a) Using the centre of mass x
C
= (x
1
+ x
2
)/2, y
C
= (y
1
+ y
2
)/2, and relative x
R
= x
2
x
1
,
y
R
= y
2
y
1
, coordinates, show that the kinetic energy can be written
T = m( x
2
C
+ y
2
C
) +
m
4
( x
2
R
+ y
2
R
).
[2 marks]
(b) Drawa diagramto showwhat the generalized coordinates R and correspond to physically,
where x
R
= Rcos(), y
R
= Rsin(), and R 0. Also indicate the location of the centre of
mass. Hence, explain how the potential energy of the system is given by V = k(R d)
2
/2.
[4 marks]
(c) Consider the centre of mass to be xed at the origin, and write the Lagrangian that remains
in terms of R and . Show that the total angular momentum is given by J = mR
2

/2, and
use the Euler-Lagrange equation:
d
dt
_
L
q
_

L
q
= 0
(for q = ), to show that J is conserved. Use this result together with the Euler-Lagrange
equation (for q = R) to determine an equation of motion for R in closed form (i.e., inde-
pendent of ). Use this equation to determine the equilibrium extension of the spring if the
angular momentum J = 0. [10 marks]
(d) It is in general possible to consider the motion of the two connected masses to take place in
a plane, even when this system is in free (three-dimensional) space. Explain why this is so.
[4 marks]
XII
D.4.2 Solution
(a) x
C
= ( x
1
+ x
2
)/2, y
C
= ( y
1
+ y
2
)/2, x
R
= x
2
x
1
, y
R
= y
2
y
1
, inserting all this into
T = (m/2)( x
2
1
+ y
2
1
) + (m/2)( x
2
2
+ y
2
2
) yields the answer.
(b) Diagram clearly showing R to be the distance between the two mass points, to be the
angle of a line, drawn between the two mass points, from the horizontal, and the centre of
mass to be halfway along this line. The potential energy is proportional to the square of the
extension of the spring from its natural length d, i.e., (R d)
2
. Multiplying this by half the
spring constant then gives the potential energy.
(c) x
R
= Rcos() x
R
=

Rcos() R

sin(), y
R
= Rsin() y
R
=

Rsin() + R

cos()
[1 mark]. Hence, x
2
R
+ y
2
R
=

R
2
+ R
2

2
, and L = T V = (m/4)(

R
2
+ R
2

2
) (k/2)(R
d)
2
.
By denition, for this system, with the centre of mass at the origin, J = m(x
1
y
1
y
1
x
1
+
x
2
y
2
y
2
x
2
). According to the above diagram, x
1
= x
2
x
1
= x
2
, y
1
= y
2

y
1
= y
2
, and hence J = 2m(x
2
y
2
y
2
x
2
) = (m/2)Rcos()[

Rsin() + R

cos()]
Rsin()[

Rcos() R

sin()] = (m/2)R
2

.
From the Euler-Lagrange equation for ,
d
dt
_
m
2
R
2

_
= 0,
from which it follows that J = (m/2)R
2

does not change with time, i.e., is conserved.
From the Euler-Lagrange equation for R,

R R

2
+ (2/m)k(R d) = 0. We can then sub-
stitute in J to get

R J
2
/[R
3
(m/2)
2
] + (2/m)k(R d) = 0. If we then set

R = 0 and J = 0,
we get R d = 0, so the equilibrium extension of the spring is d.
(d) In three dimensions, as in two, there will be no angular dependence to T or V, and hence
L, that is, the system may be arbitrarily rotated without there being any effect on the La-
grangian; formally, the Lagrangian is rotationally invariant (or any other suitable explana-
tion leading to the conclusion that angular momentum is conserved). Because of this, the
angular momentum vector will be conserved. A consequence of this is that the system dy-
namics take place in a plane, because moving out-of-plane would involve a change in
angular momentum.
D.5 Normal Coordinates (2008)
D.5.1 Question
Consider two identical small beads (mass m), sliding without friction on a straight horizontal rod
propped between two walls. The left-hand bead (distance from equilibrium position x
1
) is con-
nected to the left-hand wall by a spring of spring constant k
1
, and the right-hand bead (distance
from equilibrium position x
2
) is connected to the right-hand wall by a spring of spring constant
k
2
. Finally, the two beads are connected to each other by a spring of spring constant k
12
.
(a) Draw a diagram of this system, and show that the Lagrangian can be written in matrix form
as L = x
T
x x
T
x, where x is a column vector, its transpose x
T
= (x
1
, x
2
) is a row vector,
and the and matrices are given by
=
m
2
_
1 0
0 1
_
, =
m
2
_

2
1

2
12

2
12

2
2
_
.
XIII
(To simplify the mathematics, the frequencies
1
,
2
,
12
, dened through m
2
1
= k
1
+ k
12
,
m
2
2
= k
2
+ k
12
, m
2
12
= k
12
, have been introduced). [6 marks]
(b) (i) Using the Euler-Lagrange equation
d
dt
_
L
q
_

L
q
= 0,
determine the equation of motion for x in matrix form [2 marks].
(ii) Explain how setting the determinant of (2/m)(
2
0
+ ) equal to zero and solving
for
2
0
yields the normal mode frequencies, and hence show that they are given by
(

0
)
2
=
_
(
2
1
+
2
2
)
_
(
2
1

2
2
)
2
+ 4
4
12
_
/2 [6 marks].
(iii) For the particular case where k
1
= 2k
12
and k
2
= k
12
/2, determine the normal coordi-
nates. [4 marks]
(c) Briey explain the signicance of normal coordinates, for example when considering a sys-
tem of coupled oscillators. [2 marks]
D.5.2 Solution
(a) Diagram clearly indicating the springs, the walls, and the beads.
V = (k
1
/2)x
2
1
+ (k
2
/2)x
2
2
+ (k
12
/2)(x
1
x
2
)
2
[1 mark] = [(k
1
+k
12
)/2]x
2
1
+ [(k
2
+k
12
)/2]x
2
2

k
12
x
1
x
2
[1 mark] Hence, L = T V = (m/2)[ x
2
1
+ x
2
2
] (m/2)[
2
1
x
2
1
+
2
2
x
2
2
2
2
12
x
1
x
2
].
One can then see that T = (m/2)[ x
2
1
+ x
2
2
] = x
T
x [1 mark] and V =
2
2
x
2
2
2
2
12
x
1
x
2
] =
x
T
x [1 mark] from which the answer follows.
(b) (i) x + x = 0.
(ii) If one inserts a trial solution x = x
0
e
i
0
t
into the equation of motion, then (
2
0
+
)x
0
= 0 results [2 marks]. This is only nontrivially true if the determinant of the
matrix acting on x
0
is = 0. (Multiplying through by the constant factor 2/m makes no
difference to this argument).

2
1

2
0

2
12

2
12

2
2

2
0

=
4
0
(
2
1
+
2
2
)
2
0
+ (
2
1

2
2

4
12
) = 0,
the roots of which can be found using the quadratic formula:
(

0
)
2
=
(
2
1
+
2
2
)
_
(
2
1
+
2
2
)
2
4(
2
1

2
2

4
12
)
2
.
The answer appears as given following some elementary manipulation.
(iii) In this case
2
1
= 3
2
12
,
2
2
= 3
2
12
/2, and hence

0
2
= [(9/2)
2
12
(5/2)
2
12
]/2, i.e.,

+
0
2
= (7/2)
2
12
,

0
2
=
2
12
. Substituting these values into the matrix (2/m)(
2
0
+
),

2
12
_
1/2 1
1 2
__
a
+
b
+
_
= 0 a
+
= 2b
+

_
a
+
b
+
_
=
1

5
_
2
1
_

2
12
_
2 1
1 1/2
__
a

_
= 0 b

= 2a


_
a
+
b
+
_
=
1

5
_
1
2
_
XIV
(normalization not required) hence, the normal coordinates are r
+
= (x
2
2x
1
)/

5,
r

= (x
1
+ 2x
2
)/

5.
(c) In a system of coupled oscillators (for example), the normal coordinates are a set of coordi-
nates that separate, such that the time-evolution of each normal coordinate can be consid-
ered separately from all the other normal coordinates.
D.6 Block in a Bowl (2009, Question Only)
Consider a small block of mass m sliding without friction on the surface of an immovable radially
symmetric parabolic bowl of curvature c. Consider rst a cartesian coordinate system such that
the origin is at the bottom of the bowl.
(a) Using the generalized coordinates r, , dened implicitly through x = r cos and y = r sin ,
and the constraint z = cr
2
/2, show that the Lagrangian can be written
L =
m
2
( r
2
+ r
2

2
+ c
2
r
2
r
2
)
mgcr
2
2
,
where g is the gravitational acceleration. [5 marks]
(b) Show that the component of the angular momentum vector oriented in the vertical direction
is given by J
z
= mr
2

, and, using the Euler-Lagrange equations
d
dt
_
L
q
_

L
q
= 0
(q can be r or for the above Lagrangian L), show that J
z
is a conserved quantity. Are the
other components of the angular momentum vector conserved quantities? Explain your
answer. [6 marks]
(c) Using L and the Euler-Lagrange equations, determine the equation of motion for r. From
this equation of motion, show that (if moving with the correct velocity) the blocks motion
can describe a circular orbit around the bowl at a xed height above the ground. In the
process, show that the magnitude of the blocks velocity must be directly proportional to
the blocks radial distance from the origin. [6 marks]
(d) Determine an expression for the total energy of this system, and use this to explain how the
system can be considered to be effectively one-dimensional, and hence solved by quadra-
ture. [3 marks]
D.7 Atom Falling through a Laser Standing Wave (2009, Question Only)
Consider a classical model of an atom falling freely through an off-resonant laser standing-wave.
Do this by considering the vertical motion of a classical point particle of mass m under the inu-
ence of forces, due to gravity and also derived from a sinusoidal potential U cos(kz).
(a) Show that the Lagrangian can be written
L =
m z
2
2
mgz U cos(kz),
XV
where g is the gravitational acceleration. Now, determine p, the canonically conjugate mo-
mentum to z, and use the Legendre transformation H = p z L to determine that the Hamil-
tonian is
H =
p
2
2m
+ mgz + U cos(kz).
[5 marks]
(b) With respect to the dynamical variables z and p, the Poisson bracket for this system may be
dened
F, G =
F
z
G
p

F
p
G
z
,
where F and G are two arbitrary functions of z and p. Determine z, p, explaining how
this can conrm that z and p are canonically conjugate, as well as z, H and p, H. Hence,
solve the coupled Hamiltons equations of motion dened by
z = z, H, p = p, H,
for z(t) and p(t) in terms of the initial momentum and position, for the particular case where
U = 0. [7 marks]
(c) It is often convenient to work with a Hamiltonian that is spatially periodic, and so we wish
to transform the term mgz away. To do this, we employ the generating function F
2
(z, P, t) =
(P mgt)z, together with the implicit transformation equations
p =
F
2
z
, Z =
F
2
P
,

H = H +
F
2
t
,
where Z is the new coordinate, P the new canonically conjugate momentum, and

H the
new Hamiltonian subsequent to the transformation. Now, determine expressions for p, z,
and hence

H in terms of P and Z. Using these results, determine Hamiltons equations of
motion for Z and P. [7 marks]
(d) Do you expect Z, P to be different to z, p? Explain your answer. [1 mark]
D.8 Ice Sliding Between Glass Plates (2010, Question Only)
The general expression for the Coriolis force in a frame of reference rotating with (constant) an-
gular velocity , acting on an object of mass m moving with velocity r is 2m r.
(a) With the aid of a diagram or diagrams, show that in a local coordinate system oriented
with respect to a point on the surface of the earth (x-axis points east, y-axis points north,
z-axis points out from the centre of the earth), the angular velocity vector takes the form
(0, sin(), cos()) ( is the magnitude of the angular velocity and the colatitude). [4
marks]
(b) Within such a local coordinate system, consider the motion of a block of ice of mass m which
is allowed to fall to the earth with local gravitational acceleration g, by sliding frictionlessly
between two glass plates oriented exactly along the east-west axis, preventing any north-
south motion of the block.
(i) Write down a constraint equation describing this restriction of motion, and state what
this implies about the blocks velocity. [2 marks]
XVI
(ii) Derive equations of motion for the block, where you may neglect the centrifugal force
(the equations will therefore necessarily be approximate). [6 marks]
(iii) Where (t) = x(t) + iz(t), show that the exact solution to the combined approximate
equation of motion is
(t) = (0) +
i(1 e
i2 t
)
2
[

(0) + g/2 ] (g/2 )t,
where = sin . (Hint rst solve the differential equation for =

+ g/2 ). [6
marks]
(c) Truncate the exact solution to the approximate equation of motion for such that it is only
to rst order in , and comment on the validity of taking this step. [2 marks]
D.9 Double Pendulum (2010)
D.9.1 Question
Consider a double plane pendulum, consisting of a mass M attached to a pivot by a rigid massless
rod of length R, and a second mass m attached to a pivot on the rst mass, also by a rigid massless
rod of length R.

m
M
(a) Consider a cartesian coordinate system with its origin located at the pivot to which the mass
M is attached. We consider Z to be the vertical axis, X the horizontal, and motion to take
place in the XZ plane; i.e., that motion is subject to the constraints y
M
= 0, y
m
= 0, where
y
M
, y
m
are the Y locations of the masses M, m. Determine equations describing any other
constraints the two masses are subject to, in terms of x
M
, x
m
, z
M
, z
m
(the X and Z locations
of the masses). [4 marks]
(b) (i) Express the dynamical variables x
M
, x
m
, z
M
, z
m
in terms of the new dynamical vari-
ables and (which are the two angles shown in the diagram), and the corresponding
velocities x
M
, x
m
, z
M
, z
m
in terms of

,

, , . [5 marks]
(ii) Consider the case where M = 2m. Show that the Lagrangian for this system can then
be written, in terms of the generalized coordinates , , and velocities

,

, as
L =
mR
2
2
[3

2
+ (

+

)
2
+ 2

(

+

) cos()] + mgR[cos( + ) + 3 cos()]
[where we take cos() cos( + ) + sin() sin( + ) = cos()]. [6 marks]
(c) (i) Consider the limit of small oscillations, i.e., such that , , (and therefore + ) remain
small during any dynamics. Hence, determine an approximation L
approx
to L which is
in quadratic form. [3 marks]
XVII
(ii) Write L
approx
in matrix form, and determine the corresponding equations of motion for
and (also in matrix form), where the Euler-Lagrange equation for a generalized
coordinate q is
d
dt
_
L
q
_

L
q
= 0.
[2 marks]
D.9.2 Solution
(a) The two constraint equations are: x
2
M
+ z
2
M
= R
2
, and (x
m
x
M
)
2
+ (z
m
z
M
)
2
= R
2
.
(b) (i) x
M
= Rsin(), z
M
= Rcos(). Similarly, x
m
x
M
= Rsin( + ), z
m
z
M
=
Rcos( + ), and so x
m
= R[sin( + ) + sin()],
z
m
= R[cos( +) +cos()]. The generalized velocities are therefore x
M
= R

cos(),
z
M
= R

sin(), x
m
= R[(

+

) cos( + ) +

cos()], z
m
= R[(

+

) sin( + ) +

sin()].
(ii) First consider the kinetic energy T = m( x
2
M
+ z
2
M
) + m( x
2
m
+ z
2
m
)/2. Substituting in
expressions for the velocities,
T =mR
2

2
+
m
2
R
2

2
+ (

+

)
2
+ 2

(

+

)[cos() cos( + ) + sin() sin( + )]
=
mR
2
2
[3

2
+ (

+

)
2
+ 2

(

+

) cos()]
where we have used cos
2
(a) + sin
2
(a) 1 and cos(a) cos(b) + sin(a) sin(b) cos(a
b) cos(b a). Next, consider the potential energy V = 2mgz
M
+mgz
m
= MgR[cos( +
) + 3 cos()]. Taking L = T V then yields the answer.
(c) (i) Using cos() 1
2
/2 (for small ), an approximate Lagrangian, keeping only terms
of up to quadratic order is given by (discounting irrelevant constant terms)
L
approx
=
MR
2
2
[3

2
+ (

+

)
2
+ 2

(

+

)]
MgR
2
[( + )
2
+ 3
2
]
=
MR
2
2
[6

2
+ 4

2
]
MgR
2
[4
2
+ 2 +
2
]
or anything algebraically equivalent.
(ii) Hence, in matrix form,
L
approx
= (

,

)
_
mR
2
2
_
6 2
2 1
__ _

_
(, )
_
mgR
2
_
4 1
1 1
__ _

_
and the equations of motion, in matrix form, are
__
6 2
2 1
__ _

_
+
_
g
R
_
4 1
1 1
__ _

_
= 0.
XVIII

You might also like