University of Cape Town

Department of Physics
PHY2014F
Vibrations and Waves
Part 2
Coupled oscillators
Normal modes of continuous systems
The wave equation
Fourier analysis

… covering (more or less)

French Chapters 2, 5 & 6 Andy Buffler
Department of Physics
University of Cape Town
[email protected]
1
 Problem-solving and homework

Each week you will be given a take-home problem set to

complete and hand in for marks ...

In addition to this, you need to work through the following

problems in French, in you own time, at home. You will not be
asked to hand these in for marks. Get help from you friends, the
course tutor, lecturer, ... Do not take shortcuts.
Mastering these problems is a fundamental aspect of this course.
The problems associated with Part 2 are:
2-2, 2-3, 2-4, 2-5, 2-6, 5-2, 5-8, 5-9, 6-1, 6-2, 6-6, 6-7, 6-10,
6-11, 6-14

You might find these tougher: 5-4, 5-5, 5-6, 5-7

2
 French The superposition of periodic motions
page 20
Two superimposed vibrations of equal frequency

=x1 A1 cos(ω0t + φ1 ) combination can be written as

=x2 A2 cos(ω0t + φ2 ) =x A cos(ω0t + φ )

z1 = A1e j (ω0t +φ1 )

Using complex numbers: z= z1 + z2
z2 = A2 e j (ω0t +φ2 )

= {
∴ z e j (ω0t +φ1 ) A1 + A2 e j (φ2 −φ1 ) }
A2 Phase difference φ= φ2 − φ1
A
Then
β
A1 φ2 − φ1 A2 = A12 + A22 + 2 A1 A2 cos(φ2 − φ1 )
A sin β A2 sin (φ2 − φ1 )
and=
ω0t + φ1 3
 Superposed vibrations of slightly different frequency: Beats
If we add two sinusoids of slightly different frequency ω1 and ω2
… we observe “beats”…
 ω1 − ω2   ω1 + ω2 
cos ω1t + cos ω2t =
2cos  t  cos  t
 2   2 
French
page 22
cos ω1t cos ω t
2
x1
x2
t

x1+x2
t
2π  ω − ω2 
cos ω1t + cos ω2t Tbeat = cos  1 t
ω1 − ω2  2 
4
 Combination of two vibrations at right angles French
page 29
=x A1 cos(ω1t + φ1 )
???
=y A2 cos(ω2t + φ2 )

Consider case where frequencies are equal and let initial

phase difference be φ
Write x = A1 cos(ω0t ) =
and y A2 cos(ω0t + φ )

Case 1 : φ = 0 x = A1 cos(ω0t ) A2
y= x Rectilinear motion
y = A2 cos(ω0t ) A1

Case 2 : φ = π 2 x = A1 cos(ω0t )
y= A2 cos(ω0t + π 2) =
− A2 sin(ω0t )
x2 y 2 Elliptical path in
∴ 2+ 2 =1
A1 A2 clockwise direction 5
 Combination of two vibrations at right angles …2

Case 3 : φ = π x = A1 cos(ω0t ) A2
y= − x
y= A2 cos(ω0t + π ) =
− A2 cos(ω0t ) A1

Case 4 : φ = 3π 2 x = A1 cos(ω0t )
y= A2 cos(ω0t + 3π 2) = + A2 sin(ω0t )
x2 y 2
∴ 2+ 2 = −1 Elliptical path in
A1 A2 anticlockwise direction

Case 5 : φ = π 4 x = A1 cos(ω0t )
= y A2 cos(ω0t + π 4)

Harder to see … use a graphical approach …

6
Superposition of simple
harmonic vibrations at
right angles with an initial
phase difference of π 4

7
Superposition of two perpendicular simple
harmonic motions of the same frequency for
various initial phase differences. 8
Abbreviated construction for the superposition of
vibrations at right angles … see French page 34.

9
Perpendicular motions with different frequencies: Lissajous figures

See French page 35.

Lissajous figures for

ω2 = 2ω1 with various
initial phase differences.

φ= 0 π 4 π 2 3π 4 π 10
 ω2 : ω1
1:1
Lissajous figures
1:2

1:3

2:3

3:4

3:5

4:5

5:6
φ= 0 π 4 π 2 3π 4 π
11
 French Coupled oscillators
page 121

When we observe two weakly coupled identical oscillators

A and B, we see:

xA
t

xB
t

… these functions arise mathematically from the addition of two

SHMs of similar frequencies … so what are these two SHMs?

These two modes are known as normal modes … which are states
of the system in which all parts of the system oscillate with SHM
12
either in phase or in antiphase.
 Coupled oscillators

A B

xA
t

xB
t

13
 The double mass-spring oscillator

k xA xB k
m m

Individually we know that mxA = − kx A and mxB = − kxB

k
For both oscillators: ω0 =
m
Now add a weak coupling force:

k xA kc xB k
m m

For mass A: mxA =

− kx A + kc ( xB − x A )
k kc
x A = −ω x + Ω ( xB − x A )
or  2 2
where ω = ,Ω =
22
0 14
0 A m m
 The double mass-spring oscillator …2
For mass A: x A = −ω02 x A + Ω 2 ( xB − x A )


For mass B: xB = −ω02 xB − Ω 2 ( xB − x A )



… two coupled differential equations … how to solve ?

d2
Adding them: 2
( x A + x B ) =− ω 0 ( x A + xB )
2

dt
d2
Subtract B from A: 2
( x A − x B ) =−ω 2
0 ( x A − x B ) − 2 Ω 2
( x A − xB )
dt
q= x A + xB
1 … called “normal
Define two new variables: q= x A − xB
2 coordinates”

d 2 q1 d 2 q2
Then 2
= −ω 2
0 q1 and 2
=− (ω 2
0 + 2 Ω 2
)q2
dt dt 15
 The double mass-spring oscillator …3

The two equations are now decoupled …

d 2 q1
2
= − ω 2
s q1 ωs2 = ω02 s = “slow”
dt
Write
d 2 q2 ω 2f = ω02 + 2Ω 2
f = “fast”
2
= −ω 2
f q2
dt
= q1 C cos(ωs t + φ1 )
… which have the solutions:
= q2 D cos(ω f t + φ2 )

Since q=
1 x A + xB and q=
2 x A − xB

We can write =
xA 1
2 ( q1 + q2 ) and =
xB 1
2 ( q1 − q2 )
16
 The double mass-spring oscillator …4
C D
= xA cos(ωs t + φ1 ) + cos(ω f t + φ2 )
2 2
Then C D
= xB cos(ωs t + φ1 ) − cos(ω f t + φ2 )
2 2
So xA and xB have been expressed as the sum and difference of
two SHMs as expected from observation.
… C, D, φ1 and φ2 may be determined from the initial conditions.
… when xA = xB ,then q2 = 0 … there is no contribution from the
fast mode and the two masses move in phase … the coupling spring
does not change length and has no effect on the motion … ωs = ω0
… when xA = −xB ,then q1 = 0 … there is no contribution from the
slow mode … the coupling spring gives an extra force … each mass
experiences a force − ( k + 2kc ) x giving k + 2 kc
ωf =
2

m
= ω02 + 2Ω 2 17
 The double mass-spring oscillator …5
symmetric mode antisymmetric mode mixed mode

18
 The double mass-spring oscillator …6

We now have a system with two natural frequencies, and

experimentally find two resonances.
Amplitude

Frequency
19
 French
Pitch and bounce oscillator page 127

Two normal modes (by inspection):

d
k k Bouncing xA xB
x A = xB
m
Restoring force = −2kx
L d 2x 2k
∴m 2 = −2kx ωbounce =
2

dt m
Pitching
xA
x A = − xB
xB
θ
Centre of mass stationary
τ = − Iθ
6kd 2
6k d 2
kd ( 12 θ d ) = − 121 mL2θ → θ = 2 θ ωpitch =
2

mL m L2 20
 N=2

1π
=ω1 2=
ω0 sin ω0
2 ( 2 + 1)

2π
=ω2 2=
ω0 sin 3ω0
2 ( 2 + 1)

21
N=3

N=4

22
 French
page 136 N-coupled oscillators
Tension T
fixed l fixed

1 2 3 p−1 p p+1 N
Each bead has mass m
… consider transverse displacements that are small.
αp
α p−1
y

1 2 3 p−1 p p+1 N

−T sin α p−1 + T sin α p

Transverse force on pth particle: Fp =
y p − y p−1 y p+1 − y p
=
−T +T
l l
for small α 23
 N-coupled oscillators …2
d 2 yp y p − y p−1 y p+1 − y p
Fp =
m 2 = −T +T
dt l l

d 2 yp
∴ + 2ω02 y p − ω02 ( y p+1 − y p−1 ) =
0
dt 2
T
where ω = 2
0 , p = 1, 2 … N
ml
… a set of N coupled differential equations.

Normal mode solutions: y p = Ap sin ωt

Substitute to obtain N simultaneous equations
( −ω 2
+ 2ω02 ) Ap − ω02 ( Ap+1 + Ap−1 ) =
0
Ap+1 + Ap−1 −ω 2 + 2ω02
or =
Ap ω02 24
 N-coupled oscillators …3

From observation of physical systems we expect sinusoidal

“shape functions” of the form Ap = C sin pθ

Ap+1 + Ap−1 −ω 2 + 2ω02

Substitute into =
Ap ω02
And apply boundary conditions A0 = 0 and AN +1 = 0
nπ
… find that θ = n = 1, 2, 3, … N (modes)
N +1
There are N modes:
pnπ
= y pn A= sin ω t C sin sin ωnt
N +1
pn n n

nπ
and ω = 2ω sin
2 ( N + 1)
n 0
25
 N-coupled oscillators …4
nπ
ωn = 2ω0 sin
2 ( N + 1)
For small N:

ωn
2ω0

0 1 2 3 N+1 n
26
 N-coupled oscillators …5
ωn
In many systems of 2ω0
interest N is very
large… and we are
only interested in the
lowest frequency
modes.
linear region

0 n << N N+1 n
nπ nπ
For n << N : sin =
2 ( N + 1) 2 ( N + 1)
nπ  πω0 
=
then ωn 2=
ω0  n
2 ( N + 1)  N + 1 
i.e. ωn ∝ n for n << N
27
 N-coupled oscillators …6

N coupled oscillators have N normal modes and hence N resonances

response

ω
28
Continuous systems

29
 Continuous systems

Consider a string stretched between two rigid supports …

tension T
x=0 x=L
String has mass m and mass per unit length µ = m L
Suppose that the string is disturbed in some way:

The displacement y is a function of x and t : y ( x, t )

30
Normal modes of a stretched string
T θ + ∆θ
Consider the forces on a French
small length of string … page 162

θ T y

x x + ∆x x
Restrict to small amplitude disturbances … then θ is small and
∂y
cos θ = 1 sin θ= tan θ= θ=
∂x
The tension T is uniform throughout the string.
Net horizontal force is zero: T cos(θ + ∆θ ) − T cos θ =0
=
Vertical force: F T sin(θ + ∆θ ) − T sin θ
∂y ∂y
=
Then F T x +∆x − T 31
∂x ∂x
x
 Normal modes of a stretched string …2
∂y ∂y
= F T x +∆x − T
∂x ∂x
x

dg g ( x + ∆x) − g ( x)
Use =
dx ∆x
∂2 y
Then = F T 2 ∆x
∂x
∂2 y ∂2 y µ: mass per unit length
or ( µ∆x ) 2 = T 2 ∆x
∂t ∂x
Check: µ T has the
∂2 y  µ  ∂2 y dimensions 1 v 2
giving =  2
∂x 2
 T  ∂t Then v = T µ is
the speed at which a
∂2 y 1 ∂2 y
Write = 2 2 wave propagates along
∂x 2
v ∂t the string … see later
One dimensional wave equation 32
 Normal modes of a stretched string …3
Look the standing wave (normal mode) solutions …
Normal mode: all parts of the system move in SHM at the
same frequency …
Write: y ( x, t ) = f ( x) cos ωt

f ( x) is the “shape” function … substitute into wave equation

∂y 2 ( x, t ) d 2 f ( x)
= cos ωt
∂x 2
dx 2 d 2 f ( x) 1 2
2
cos ωt = − 2 ω f ( x) cos ωt
∂y 2 ( x, t ) dx v
∂t
= 2
f ( x ) ( −ω 2
cos ωt )
which must be true for all t

d2 f ω2
then 2
= − 2 f ( x)
dx v 33
 Normal modes of a stretched string …4

d2 f ω2
2
= − 2 f ( x)
dx v
d 2x
… which has the same form as the eq. of SHM: 2
= −ω 2
0x
dt
=
… has general solution: x A sin(ω0t + φ )
ω 
=
Thus we must have: f ( x) A sin  x + Φ 
v 
Apply boundary conditions: y = 0 at x = 0 and x = L
∴ f (0) = 0 and f ( L) = 0
ω 
x = 0,=f =0 : 0 A sin  0 + Φ  i.e. Φ =0
v 
ω  ω
x = L, f =0 : 0 = A sin  L  i.e. L = nπ n = 1,2,3,…
v  v 34
 Normal modes of a stretched string …5
nπ v
Write ωn = n = 1,2,3,…
L
 x nπ v   nπ x 
=
Therefore f ( x) A=n sin   An sin  
 v L   L 
shape function, or “eigenfunction”

x=0 x=L
f ( x) = A1 sin (π x L ) n=1 ω1 = π v L

f ( x) = A2 sin ( 2π x L ) n = 2 ω2 = 2π v L

f ( x) = A3 sin ( 3π x L ) n = 3 ω3 = 3π v L

f ( x) = A4 sin ( 4π x L ) n = 4 ω4 = 4π v L

f ( x) = A5 sin ( 5π x L ) n=5 ω5 = 5π v L
35
 Normal modes of a stretched string

n=1

n=2

n=3

n=4
36
37
 Normal modes of a stretched string …6

Full solution for our standing waves:

 nπ x  nπ v
y ( x, t ) = An sin   cos ωnt ωn =
 L  L
The mode frequencies are evenly spaced: ωn = nω1

ωn (recall the beaded string)

This continuum approach breaks
ω3 down as the wavelength approaches
ω2 atomic dimensions… also if there is
any “stiffness” in the spring which
ω1 adds an additional restoring force
which is more pronounced in the
high frequency modes.
0 1 2 3 n 38
 Normal modes of a stretched string …7

All motions of the system can be made up from the

superposition of normal modes
∞
 nπ x 
y ( x, t ) ∑
n =1
An sin 
 L 
 cos(ωnt + φn )

nπ v
with ωn =
L

Note that the phase angle is back since the modes may
not be in phase with each other.

39
 Whispering galleries

… best example is the inside

dome of St. Paul’s cathedral.

If you whisper just inside the

dome, then an observer close to
you can hear the whisper
coming from the opposite
direction … it has travelled
right round the inside of the
dome.

40
 French
Longitudinal vibrations of a rod page 170
x x + ∆x
section of massive rod

section is displaced and stretched

F1 F2 by an unbalanced force
x +ξ x + ξ + ∆x + ∆ξ

∆ξ ∆ξ
Average strain = Average stress = Y
∆x ∆x
Y : Young’s modulus
∂ (stress)
stress at x + ∆x = (stress at x) + ∆x
∂x

41
 Longitudinal vibrations of a rod …2

If the cross sectional area of the rod is α

∂ξ ∂ξ ∂ 2ξ
F1 = α Y and F2 =α Y + α Y 2 ∆x
∂x ∂x ∂x
∂ 2ξ
− F2 α Y =
F1= ∆x ma
∂x 2

∂ 2ξ ∂ 2ξ
∴α Y 2 ∆x= ρα∆x 2
∂x ∂t
∂ 2ξ ρ ∂ 2ξ
or =
∂x 2
Y ∂t 2

∂ 2ξ 1 ∂ 2ξ v=
Y
∴ 2 =
∂x v 2 ∂t 2 ρ
42
 Longitudinal vibrations of a rod …3
Look for solutions of the type: ξ ( x, t ) = f ( x) cos ωt
ω 
=
where f ( x ) A sin  x + Φ 
 v 
Apply boundary conditions: one end fixed and the other free

x = 0 : ξ (0, t ) = 0 i.e. Φ =0
∂ξ
x== L : F α= Y 0
∂x
 ωL 
… then cos  =0
 v 
ω
or L= ( n − 12 ) π n = 1,2,3,…
v
(=
n − 12 ) π v ( n − 12 ) π Y
=
The natural angular frequencies ωn
L L ρ 43
x=0 x=L
π Y
n=1 ω1 =
2L ρ
3π Y
n=2 ω2 =
2L ρ
5π Y
n=3 ω3 =
2L ρ
n=4

n=5

44
 Normal modes for different boundary conditions

Simply
supported

Clamped
one end

Free both
ends

Clamped
both ends

n=1 n=2 n=3 45

 French
l The elasticity of a gas page 176

dp
Bulk modulus: K = −V
A ρ, p dV

m 2
Kinetic theory of gases: Pressure = ρv
=
1
3p 2
vrms
rms
3 Al
2 Ek
If E k = 1
2 mv 2
rms then p=
3A l
Now move piston so as to compress the gas
… work done on gas: ∆W =− pA∆l =∆Ek
2 ∆Ek 2 ∆l 2 ∆l 5 ∆l
Then ∆p = − 2
∆Ek = ( − pA∆l ) − ( p ) =− p
3A l 3A l 3A l 3 l
∆p 5
giving K adiabatic = −V = p
∆V 3
K 1.667 p
and= v =
ρ ρ 46
 French Sound waves in pipes
page 174

A sound wave consists of a series of compressions and

rarefactions of the supporting medium (gas, liquid, solid)
In this wave individual molecules move longitudinally with
SHM. Thus a pressure maximum represents regions in which the
molecules have approached from both sides, receding from the
pressure minima.

wave propagation

47
Longitudinal wave on a spring

48
 Standing sound waves in pipes

t = 0:
Pressure p
p0
x
Flow
velocity u
0 x

t =T 2:
p
x

u
x49
 Standing sound waves in pipes …2
Consider a sound wave in a pipe. At the closed end the flow
velocity is zero (velocity node, pressure antinode).
At the open end the gas is in contact with the atmosphere,
i.e. p = p0 (pressure node and velocity antinode).

Open end Closed end

pressure node pressure antinode
p
p0

u
0

50
velocity antinode velocity node
 Standing sound waves in pipes …3

Pipe closed at Pipe open at Pipe open at

both ends both ends one end

nλ nv ( 2n − 1) λ ( 2n − 1) v
=
L = =L =
2 2f 4 4 f
f =
nv ( 2n − 1) v
2L f =
4L
nπ v ( 2n − 1) π v
ωn = ωn = 51
L 2L
 Sound

Audible sound is usually a longitudinal compression wave in air to

which the eardrum responds.

Velocity of sound (at NTP) ~ 330 m s-1

By considering the transport of energy by a compression wave,
can show that P = 2π 2 f 2 ρ Avs 2
m

… where A is cross sectional area of air column and sm is

maximum displacement of air particle in longitudinal wave

P
Then intensity = = 2π 2 f 2 ρ vsm2 unit: W m-2
A

52
 Sound …2

The human ear detects sound from ~10-12 W m-2 to ~1 W m-2

… use a logarithmic scale for I :
 I 
Intensity level or “loudness”: β = 10log10   decibels
 I0 
where I0 = “reference intensity” = 10-12 W m-2

53
 Musical sounds

Waveforms from real musical instruments are complex, and may

contain multiple harmonics, different phases, vibrato, ...

Pitch is the characteristic of a sound which allows sounds to be

ordered on a scale from high to low (!?). For a pure tone, pitch is
determined mainly by the frequency, although sound level may
also change the pitch. Pitch is a subjective sensation and is a
subject in “psychoacoustics”.

The basic unit in most musical scales is the octave. Notes judged
an octave apart have frequencies nearly (not exactly) in the ratio
2:1. Western music normally divides the octave into 12 intervals
called semitones ... which are given note names (A through G with
sharps and flats) and designated on musical scales.
54
 Musical sounds ...2

Timbre is used to denote “tone quality” or “tone colour” of a

sound and may be understood as that attribute of auditory
sensation whereby a listener can judge that two sounds are
dissimilar using any criteria other than pitch, loudness or
duration. Timbre depends primarily on the spectrum of the
stimulus, but also on the waveform, sound pressure and
temporal characteristics of the stimulus.

One subjective rating scale for timbre (von Bismarck, 1974)

dull sharp
compact scattered
full empty
colourful colourless
55
 Two dimensional systems French
page 181

Consider an elastic
membrane clamped Δy
at its edges … Δx
x
… the membrane has mass per unit area σ ,
and a surface tension S which gives a force SΔl
perpendicular to a length Δl in the surface …

SΔx
The forces on the shaded SΔy SΔy
portion are …

SΔx 56
 Two dimensional systems …2
If the membrane is displaced SΔy
from the z = 0 plane then a θ + ∆θ
cross section through the
shaded area shows:
θ z
SΔy
x x + ∆x x
… looks exactly like the case of the stretched string.
∂2 z
The transverse force on the element will be S ∆y 2 ∆x
∂x
And if we looked at a cross section perpendicular to this …
∂ 2
the transverse force will be S ∆x z ∆y
∂y 2 57
 Two dimensional systems …3

The mass of the element is σ∆x∆y .

∂2 z ∂2 z ∂2 z
σ∆x∆y 2
Thus S ∆y 2 ∆x + S ∆x 2 ∆y =
∂x ∂y ∂t
∂2 z ∂2 z  σ  ∂2 z
or + 2 =  2
∂x ∂y
2
 S  ∂t

… a two dimensional wave equation

S
… with the wave velocity v =
σ

58
 Two dimensional systems …4

Look for normal mode solutions of the form:

z ( x, y, t ) = f ( x) g ( y ) cos ωt
∂z 2 d 2 f
= 2 g ( y ) cos ωt d2 f d 2g
∂x 2
dx g ( y ) cos ωt + 2 f ( x) cos ωt =
2
∂z 2 d 2 g dx dy
= 2 f ( x) cos ωt ω2
∂y 2
dy − f ( x) g ( y ) cos ωt
2
v
∂z 2
f ( x) g ( y ) ( −ω 2 cos ωt )
∂t 2 1 d 2 f 1 d 2g ω2
i.e. 2
+ 2
=
− 2
f dx g dy v
In a similar fashion to the 1D case, find …
 n1π x   n2π y 
f ( x) = An1 sin  g ( y ) = Bn2 sin 
 and  L 
 x 
L  y  59
 Two dimensional systems …5

 n1π x   n2π y 
…then z ( x, y, t ) = Cn1 n2 sin   sin   cos ωn1 ,n2 t

 Lx   Ly 
where the normal mode frequencies are
2
 n1π v   n2π v 
2

ωn1 ,n2
=   +  
 Lx   Ly 

e.g. for a membrane having sides 1.05L and 0.95L

π v  n1 
2 2
 n2 
=
then ωn1 ,n2  +
  
L  1.05   0.95 

60
 Normal modes of a rectangular membrane

up
1,1
down

2,1 2,2

3,1 3,2

61
 Normal modes of a circular membrane

1,0 2,0 3,0

1,1 2,1 2,2 62

Modes of vibration of a 38 cm cymbal. The first 6 modes
resemble those of a flat plate ... but after that the resonances
tend to be combinations of two or more modes. 63
Normal modes of a circular drum

64
Chladni plates

65
Soap films

66
Holographic interferograms of the top and bottom plates
of a violin at several resonances.
67
Holographic interferograms of a classical guitar top
plate at several resonances.
68
Holographic interferograms showing the vibrations of a 0.3 mm
thick trombone driven acoustically at 240 and 630 Hz.
69
Time-average hologram interferograms of inextensional
modes in a C5 handbell
70
 Normal modes of a square membrane
2
n2
v area per point =  v 
4 4  2L 
2L
v one point per
3 3 normal mode
2L
v
2 2 2 2
2L  n1v   n2 v 
f 4,3 =
f n1 ,n2   + 
v  2 L   2 L 
1 1
2L
0 0 v v v v v
0 1 2 3 4 5
2L 2L 2L 2L 2L
0 1 2 3 4 5 n1
Normal modes having the same frequency are said to be degenerate
71
 Normal modes of a square membrane … for large n1 and n2

n2
df
area = 14 (2π f )df

2
 v 
f area per mode =  
 2 L 

n1
Number of modes with 2
 2L 
1
frequencies between f and (f + df) = 4 (2π f ) df  
 v 
2π L f df
2
=
v2 72
 French Three dimensional systems
page 188

Consider some quantity Ψ which depends on x , y , z and t ,

e.g. the density of air in a room.

∂ 2
Ψ ∂ 2
Ψ ∂ 2
Ψ 1 ∂ 2
Ψ
In three dimensions: + + =
∂x 2 ∂y 2 ∂z 2 v 2 ∂t 2

1 ∂ 2
Ψ
which can be written: ∇ Ψ= 2 2
2

v ∂t
The solutions for a rectangular enclosure:
2
 n1π v   n2π v   n3π v 
2 2

ωn ,n ,n =   +   + 

 x   y   z 
L L L
1 2 3

πv 2 2 2
…and for a cube: ωn1 ,n=
2 ,n3
n1 + n2 + n3
L 73
 Three dimensional systems …2
How many modes are there with frequencies in the range
f and (f + df) … ?
Set up an imaginary cubic lattice with spacing v 2 L
n2 … and consider positive frequencies only.
df
Volume of shell = 1
8 (4π f 2 )df
f 3
 v 
n1 Volume per mode =  
 2 L 

n3
Number of modes with 3
 2L 
frequencies between f and (f + df) = (4π f 2 )df 
1
8 
 v 
4π L f df
3 2
= 74
v3
 Three dimensional systems …3
Number of modes with 4π V f 2 df
frequencies between f and (f + df) =
v3
… holds for any volume V … provided its dimensions are much
greater than the wavelengths involved.
… need to multiply by a factor of 2 when dealing with
electromagnetic radiation (2 polarization states) …
“Ultraviolet catastrophe” for blackbody radiation …
Equipartition theorem: in thermal equilibrium each mode has an
average energy 12 k BT in each of its two energy stores
Hence, energy density of radiation in a cavity:
4π V f 2 df 1 experiment
µ df = 2 3
2 ( 2 kT ) µ !?
c
8π f 2
or µ = 3
kT
c f
75
 Planck was able to show, effectively by
assuming that energy was emitted an
absorbed in quanta of energy hf , that the
average energy of a cavity mode was not kT
but hf
e hf kT − 1
where Planck’s constant h = 6.67 10-34 J K-1
8π f 2 df hf
Then µ df = Planck’s law
c3 e hf kT
−1

energy no. of average

density modes in energy per
range f mode
to f +df
76
… which agrees extremely well with experiment.
77
 French Introduction to Fourier methods
page 189

We return to our claim that any physically observed shape

function of a stretched string can be made up from normal
mode shape functions.
f ( x)

x ∞
 nπ x 
i.e. f ( x) = ∑ Bn sin  
n =1  L 
… a surprising claim … ? … first find Bn
 n1π x 
… multiply both sides by sin  
 L 
and integrate over the range x = 0 to x = L
L
n π
 1 x
L
 n1π x  ∞  nπ x 
∫0 f ( x)sin  L  dx = ∫0 sin  L  ∑ n =1
Bn sin 
 L 
dx 78
 Fourier methods …2
 n1π x   n1π x  ∞  nπ x 
L L

∫ f ( x)sin   dx = ∫ sin  ∑ n 
B sin dx
0  L  0  L  n=1  L 
If the functions are well behaved, then we can re-order things:
 n1π x  ∞
 n1π x   nπ x 
L L

∫ f ( x)sin   dx = ∑ Bn ∫ sin   sin   dx

0  L  n =1 0  L   L 
[n1 is a particular integer and n can have any value between 1 and ∞ .]

Integral on rhs:

 n1π x   nπ x  1   ( n1 − n ) π x  ( n1 + n ) π x 
L L

∫ sin=
  sin 
 L   L 
 dx ∫ cos 
2  
 − cos  dx
0 0
L  L 

79
 Fourier methods …3

Both (n1 + n) and (n1 − n) are

integers, so the functions
 ( n1 ± n ) π x 
cos  
 L 
on the interval x = 0 to L
must look like
… from which it is evident that
L
 ( n1 ± n ) π x 
∫0 cos  L  dx = 0
Except for the special case when n1 and n are equal … then …
 ( n1 − n ) π x  L
 ( n1 − n ) π x 
cos   = 1 and ∫ cos 
 L
 dx = L

 L  0
80
 Fourier methods …4

Thus all the terms in the summation are zero, except for the
single case when n1 = n i.e.
 nπx  1   ( n1 − n ) π x  ( n1 + n ) π x 
L L

∫
0
f ( x)sin  1  dx
 L 
Bn1 ∫
0
cos 
2   L
 − cos
 L
 dx

L
= Bn1
2
 n1π x 
L
2
i.e. Bn1 = ∫ f ( x)sin   dx
L0  L 
We have found the value of the coefficient for some
particular value of n1 … the same recipe must work for any
value, so we can write:
 nπ x 
L
2
Bn = ∫ f ( x)sin   dx
L0  L 
81
 Fourier methods …5

The important property we have used is that the functions

 n1π x   nπ x 
sin   and sin  L 
 L   
are “orthogonal over the interval x = 0 to x = L.”

0 if n1 ≠ n
 n1π x   nπ x 
L

i.e. ∫
0
sin   sin 
 L   L 
 dx = L
if n1 = n
2

Read French pages 195-6

82
 Fourier methods …6

The most general case (where there can be nodal or

antinodal boundary conditions at x = 0 and x = L) is

A0 ∞   nπ x   nπ x  
f ( x) = + ∑  An cos   + Bn sin  
2 n=1   L   L 

 nπ x 
L
2
An = ∫ f ( x) cos   dx
L0  L 
where
 nπ x 
L
2
Bn = ∫ f ( x)sin   dx
L0  L 

83
 Fourier methods …7
One of the most commonly encountered uses of Fourier methods is
the representation of periodic functions of time in terms of sine
and cosine functions … T
f (t )
2π
Put Ω =
T
This is the lowest frequency in f (t ) … clearly there are t
higher frequencies … by the same method as before, write …
A0 ∞  2π nt   2π nt 
f (t ) = + ∑ An cos   + Bn sin  
2 n=1  T   T 
A0 ∞
= + ∑ An cos nΩt + Bn sin nΩt
2 n=1
T T
2 2
where An ∫ f (t ) cos ( nΩt ) dt
= and Bn ∫ f (t )sin ( nΩt ) dt
T0 T0
84
 Waveforms of ...

a flute

a clarinet

an oboe

a saxophone

85
Fast Fourier transform experiments, 10 March 2008

86
Fast Fourier transform experiments, 10 March 2008

87
Fast Fourier transform experiments, 10 March 2008

88
Fast Fourier transform experiments, 10 March 2008

89
Fast Fourier transform experiments, 10 March 2008

90
 Odd functions

An odd periodic function f (−t ) =− f (t )

where −T 2 < t < T 2

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

t t t

…can be expressed as a sum of sine functions

∞
2π
=f (t ) ∑B
n =0
n sin nΩt Ω=
T
T
2
=Bn ∫ f (t )sin ( nΩt ) dt
T0 91
 Even functions

An even periodic function f (−t ) =+ f (t )

where −T 2 < t < T 2

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

t t t

…can be expressed as a sum of cosine functions

A0 ∞ 2π
f (t ) = + ∑ An cos nΩt Ω=
2 n=1 T
T
2
=An ∫ f (t ) cos ( nΩt ) dt
T0 92
 Fourier methods … Example
Find Fourier coefficients for the case:
f (t )
1
t
-1
∞
This is an odd function: f (−t ) =− f (t ) ∴ f=
(t ) ∑B
n =1
n sin nΩt
T
2
Bn ∫ f (t )sin ( nΩt ) dt
T0
T 2 T
2 2
= ∫ (1)sin ( nΩt ) dt + ∫ (−1)sin ( nΩt ) dt
T 0 TT2
2  1  2  1 
 cos ( nΩt )  0 + (−1)  −  cos ( nΩt ) T 2
T 2 T
= (1)  −
T  nΩ  T  nΩ  93
 Fourier methods … Example cont.
2π
Ω=
T
1 1
− [ cos nπ − cos 0] +
∴ Bn = [cos 2nπ − cos nπ ]
nπ nπ

2nπ cos
For even n: cos= = nπ 1 ∴ Beven n =
0

For odd n: cos nπ = −1 and cos 2nπ = +1

1 1 4
∴ Bodd n =− (−1 − 1) + (1 + 1) =
nπ nπ nπ

4 1 1 
∴=
f (t ) sin Ωt + sin 3Ωt + sin 5Ωt + ...
π 3 5 
94
 Fourier sums … Example 3

2 terms −T 2 20 terms
T 2

4 terms 50 terms

8 terms 200 terms

95
 Fourier sums … Example 1

2 terms 8 terms
0 T

3 terms 20 terms

4 terms 50 terms

96
 Fourier sums … Example 2

2 terms 20 terms
0 T

4 terms 50 terms

8 terms 200 terms

97
 Time domain Frequency spectrum

Fourier t
f
transforms f1
t
f
7 f1

t
f

t
f

t
f
f1 3 f1 5 f1 7 f1

t
f 98
f1 3 f1 5 f1 7 f1
99
100