PC235 Winter 2013 Chapter 12.

Coupled Oscillators and Normal Modes Slide 1 of 49

Chapter 12.
Coupled Oscillators and Normal Modes

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 2 of 49
Outline

In chapter 6, we studied the oscillations of a single body subject to a Hookes law

force. Now, we wish to examine the oscillations of several bodies, each of which is
connected by springs to one or more of the others. An example of such a system is
a molecule, which we can model as a system of masses (the atoms) connected by
springs (the force between atoms, which takes a Hookes law form around the
atoms equilibrium separation.) If each of the masses were attached to a separate,
fixed spring, with no connections between the masses, then each would oscillate
independently. However, the forces that connect the masses to each other turn this
into a coupled system. We shall find that - unlike the case of a single oscillator which
exhibits a single natural oscillation frequency - two or more coupled oscillators have
a corresponding number of natural (or normal) frequencies, and that the general
motion is a combination of vibrations at all the different normal frequencies.
Table of Contents

Two Masses and Three Springs

Lagrangian Approach: The Double Pendulum

The General Case
Three Coupled Pendulums

Identical Springs and Equal Masses

Two Weakly Coupled Oscillators

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 3 of 49
Two Masses and Three Springs

Two Masses and Three Springs JRT 11.1

As a first example, consider the two carts shown in Fig. 1. The carts move without
friction on a horizontal track, between two fixed walls. Each is attached to its
adjacent wall by a spring (force constants k1 and k3 ), and to each other by a spring
with force constant k2 . Clearly, in the absence of spring 2, the carts would oscillate
independently of one another (assuming that they dont collide, of course.) However,
the presence of the coupling spring ensures that - to at least a small extent - the
motion of one cart affects the motion of the other.
In the following discussion, it is assumed that when the two carts are at their
equilibrium positions the three springs are neither stretched nor compressed.
However, depending on the distance between the walls, it could be that all three
springs are compressed or all three are stretched. It is a relatively simple task to
show that the results of this discussion are not affected by these possibilities.

Fig. 1:

Two carts attached by springs to fixed walls and to each other - see JRT Fig. 11.1

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 4 of 49
Two Masses and Three Springs

Two Masses and Three Springs cont

JRT 11.1

For this system, the equations of motion can be found using either Newtons second
law or Lagranges equations. The latter are easier to write down, but the former is a
bit more instructive, so that is the path we will take.
Suppose the carts have moved distances x1 and x2 , measured to the right, relative to
their equilibrium positions. Spring 1 is stretched by an amount x1 , so it exerts a force
k1 x1 to the left on cart 1. Spring 2 is more complicated since it is affected by the
position of both carts; it exerts a force k2 (x2 x1 ) to the right on cart 1. Of course,
these forces can be negative if the situation warrants. The net force on cart 1 is
k1 x1 + k2 (x2 x1 ) = (k1 + k2 )x1 + k2 x2 .
(12.1)
The net force on cart 2 is found in an identical fashion. By Newtons second law, the
two equations of motion are
(12.2)
m1 x1 = (k1 + k2 )x1 + k2 x2 , and m2 x2 = k2 x1 (k2 + k3 )x2 .

Before we try to solve these coupled equations, we note that they can be written in a
compact matrix form
= Kx.
Mx
(12.3)
Here, we have introduced the vector and matrices
#
"
#
"
#
"
k2
x1
m1 0
k1 + k2
, M=
, K=
.
(12.4)
x=
0 m2
x2
k2 + k3
k2

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 5 of 49
Two Masses and Three Springs

Two Masses and Three Springs cont

JRT 11.1

The mass matrix M for this simple case is a diagonal matrix, with the masses m1
and m2 down the diagonal. The spring-constant matrix K has nonzero off-diagonal
elements, reflecting that the elements of x are coupled to each other. Notice that the
matrix equation (12.3) has the form of Newtons second law, and indeed, if there was
just one cart then M, K, and x would reduce to scalars. Notice as well that both M
and K are symmetric, as will be true of all the corresponding matrices in this chapter.
To solve the equations of motion, we might reasonably guess that there could be
solutions in which both carts oscillate sinusoidally with the same angular frequency
; that is,
x1 (t ) = 1 cos(t 1 ), and x2 (t ) = 2 cos(t 2 ).
(12.5)
Furthermore, if such a solution exists, then there must also be a solution of the same
form but with the cosines replaced by sines (this is just a shift in the time axis):

y1 (t ) = 1 sin(t 1 ), and y2 (t ) = 2 sin(t 2 ).

(12.6)

And then, we can combine these two solutions into a single complex solution,

z1 (t ) = x1 (t ) + iy1 (t ) = 1 e i (t 1 ) = a1 e i t ,
where a1 = 1 e

i 1

(12.7)

, and likewise,

z2 (t ) = x2 (t ) + iy2 (t ) = 2 e i (t 2 ) = a2 e i t , where a2 = 2 e i 2 .

(12.8)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 6 of 49
Two Masses and Three Springs

Two Masses and Three Springs cont

JRT 11.1

Next, we combine the two complex solutions into a single vector solution of the form
"
# "
#
z1 (t )
a1
e i t = ae i t
(12.9)
z(t ) =
=
z2 (t )
a2
where the vector a is a constant, made up of two complex numbers,
#
"
# "
a1
1 e i 1
.
a=
=
2 e i 2
a2

(12.10)

In seeking solutions of the equation of motion, we must remember that when we find
such solutions, the actual motion x(t ) is equal to the real part of z(t ).
When we substitute eq. (12.9) into eq.(12.3), we obtain the equation
2 Mae i t = Kae i t ,

(12.11)

or, cancelling the common terms,

(K 2 M)a = 0.
This is known as a generalized eigenvalue equation.

(12.12)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 7 of 49
Two Masses and Three Springs

Two Masses and Three Springs cont

JRT 11.1

If the matrix (K 2 M) has a nonzero determinant, then the only solution to eq.
(12.12) is the trivial solution a = 0, which corresponds to no motion at all. On the
other hand, if
det(K 2 M) = 0,

(12.13)

then there is a nontrivial solution of eq. (12.12) and hence a solution of the equations
of motion with our assumed sinusoidal form. This is the solution that we seek.
In the present case, since M and K are 2-by-2 matrices, there will be two (possibly
identical) solutions to eq. (12.13). This implies that there are two possible
frequencies at which the carts can oscillate sinusoidally.
These frequencies are called the normal frequencies of the system. They depend on
the values of the two masses and the three spring constants.
If we had N coupled masses, we would have an N -by-N eigenproblem to solve,
which would provide us with N normal frequencies. In PC235, we will restrict
ourselves to the case where N = 2 or 3. In many branches of physics (condensed
matter physics and optics in particular), we often have N so large that it is essentially
infinite! Thankfully, mathematicians have provided us with closed-form expressions
for the properties of symmetric infinite matrices.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 8 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses JRT 11.2

We will continue to examine the two carts of Fig. 1, but now we suppose that the
two masses are equal, m1 = m2 = m, and that the three springs constants are
also equal, k1 = k2 = k3 = k . In that case, the matrices defined in the previous
#
"
#
"
section become
2k k
m 0
and K =
.
(12.14)
M=
k 2k
0 m
The matrix of the eigenvalue equation is then
"
#
2k m2
k
(K 2 M) =
(12.15)
2
k
2k m
and its determinant is
det(K 2 M) = (2k m2 )2 k 2 = (k m2 )(3k m2 ).

(12.16)

The two normal frequencies are determined by the condition that this determinant be
r
r
zero, and are therefore
k
3k
1 =
and 2 =
.
(12.17)
m
m
These are the frequencies at which the two carts can oscillate in purely sinusoidal
motion. The first one, 1 , is precisely the frequency of a single mass m on a single
spring k . We will soon see why this is the case.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 9 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses - The First Normal Mode

JRT 11.2

The preceding equation tells us the two possible frequencies of the system, but we
have not yet described the corresponding motions. Recall that the actual motion is


given by the vector x(t ) = Re z(t ) , where the complex vector z(t ) = ae i t , and a is
#
"
the vector
a1
,
(12.18)
a=
a2
which must satisfy the eigenvalue equation
(K 2 M)a = 0.

(12.19)

Now that we know the possible normal frequencies, we must solve this equation for
the vector a for each normal frequency in turn. The sinusoidal motion with any one
of the normal frequencies is called a normal mode of the coupled oscillators.

If we choose equal to the first normal frequency, 1 = k /m, then the matrix
2
(K M) becomes
#
"
k k
2
(K 1 M) =
.
(12.20)
k
k

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 10 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses - The First Normal Mode

cont JRT 11.2
For this case, the eigenvalue equation, eq. (12.19), reads
#"
"
#
a1
1 1
= 0,
1
1
a2

(12.21)

which is equivalent to the two equations

a1 a2

a1 + a2

0.

(12.22)

These equations imply that a1 = a2 = Ae i . The complex vector z(t) is therefore

#
"
#
"
a1
A
(12.23)
z(t ) =
e i t =
e i (1 t ) ,
a2
A


and the corresponding actual motion is given by the vector x(t ) = Re z(t ) , or
# "
#
"
x1 (t )
A
=
cos(1 t ).
x(t ) =
(12.24)
A
x2 (t )
That is,

x1 (t ) = A cos(1 t ), and x2 (t ) = A cos(1 t ).

(12.25)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 11 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses - The First Normal Mode

cont JRT 11.2
We see that in the first normal mode the two carts oscillate in phase and with
the same amplitude A , as shown in Figs. 2-3. A striking feature of these figures is
that, because x1 (t ) = x2 (t ), the middle spring is at no time either stretched or
compressed. This spring is in fact irrelevant, and each cart oscillates simply as if it
were attached to the nearest wall by a single spring. This explains why the first
normal frequency is the same as for a single cart on a single spring.

Fig. 2:

Fig. 3:

The first normal mode for two equal-mass carts with three identical springs - see JRT Fig. 11.2

In the first normal mode, the two positions oscillate with equal amplitudes and in phase - see JRT Fig. 11.3

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 12 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses - The Second Normal

Mode JRT 11.2
The second normal frequency at which our system can oscillate is 2 =
which, when substituted into eq. (12.15), gives
"
#
k k
(K 22 M) =
.
k k
For this case, the eigenvalue equation, eq. (12.19), reads
#"
"
#
a1
1 1
= 0,
a2
1 1
i
which implies that a1 = a2 = Ae . That is,
x1 (t ) = A cos(2 t ), and x2 (t ) = A cos(2 t ).

3k /m,
(12.26)

(12.27)
(12.28)

In this second normal mode, the two carts oscillate with the same amplitude A ,
but exactly out of phase, as shown in Figs. 4-5.
Note that in the second normal mode, when cart 1 is displaced to the right, cart 2 is
displaced to the left by an equal distance. This means that when the outer two
springs are stretched, the middle spring is compressed by twice as much. Thus, for
example, when the left spring pulls cart 1 to the left, the middle spring is pushing cart
1, also to the left, with a force that is twice as large. This means that each cart
moves as if it were attached to a single spring with force constant 3k .

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 13 of 49
Identical Springs and Equal Masses

Fig. 4:

Fig. 5:

The second normal mode for two equal-mass carts with three identical springs - see JRT Fig. 11.4

In the second normal mode, the two positions oscillate sinusoidally with equal amplitudes but exactly out of phase see JRT Fig. 11.5

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 14 of 49
Identical Springs and Equal Masses

Identical Springs and Equal Masses - The General Solution JRT

11.2

We have now found two normal-mode solutions, which we can rewrite as

#
#
"
"
1
1
cos(1 t 1 ) and x(t ) = A2
cos(2 t 2 ),
(12.29)
x(t ) = A1
1
1
where 1 and 2 are the normal frequencies. Both of these solutions satisfy the
equation of motion for any values of the four real constants A1 , A2 , 1 , and 2 .
Because the equation of motion is linear and homogeneous, the sum of these two
solutions is also a solution:
#
#
"
"
1
1
cos(1 t 1 ) + A2
cos(2 t 2 ).
(12.30)
x(t ) = A1
1
1

Because the equation of motion is really two second-order differential equations for
the two variables x1 (t ) and x2 (t ), its general solution has four constants of
integration. Thus, eq. (12.30), with its four arbitrary constants, is the general
solution.

When we allow A1 , A2 , 1 , and 2 to be completely arbitrary, the general solution can

be difficult to visualize. The
motion of each cart is a mixture of the two frequencies
1 and 2 . Since 2 = 31 , the motion never repeats itself except in the special
case that one of the constants A1 or A2 is zero. Figure 6 shows graphs of the two
positions in a typical non-normal mode.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 15 of 49
Identical Springs and Equal Masses

Fig. 6:

General solution for x1 (t ) and x2 (t ), for A1 = 1, A2 = 0.7, 1 = 0, 2 = /2 - see JRT Fig. 11.6

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 16 of 49
Identical Springs and Equal Masses

Identical Springs & Equal Masses - Normal Coordinates JRT 11.2

We have just seen that in any possible motion of the two-cart system, both of the
coordinates x1 (t ) and x2 (t ) vary with time. In the normal modes, the time
dependence is sinusoidal and coupled; one cart cannot move without the other.
It is also possible to introduce an alternative set of normal coordinates, which have
the convenient property that each can vary independently of the other. This
statement is true for any system of coupled oscillators, but is especially easy to see
in the present case of two equal masses joined by three identical springs. For this
case, the normal coordinates are
1
1
1 = (x1 + x2 ), and 2 = (x1 x2 ).
(12.31)
2
2
It is easy to show that for the first normal mode, the new variables are given by
1 (t ) = A cos(1 t ), and 2 (t ) = 0,
(12.32)
whereas for the second normal mode, we have
1 (t ) = 0, and 2 (t ) = A cos(2 t ).
(12.33)
In this sense, the new coordinates are independent ; oscillation of one does not
result in oscillation of the other. Of course, the general motion of the system is a
superposition of both modes, and thus 1 and 2 both oscillate, but 1 oscillates with
frequency 1 only, while 2 oscillates with frequency 2 only. In problems that are
much more complicated than the simple system of this section, the use of normal
coordinates represents a considerable simplification.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 17 of 49
Two Weakly Coupled Oscillators

Two Weakly Coupled Oscillators JRT 11.3

In the last section we discussed the oscillations of two equal masses joined by three
equal springs. For this system, the two normal modes were easy to understand and
to visualize, but the non-normal oscillations were somewhat complicated (Fig. 6). A
system where some of the non-normal oscillations are readily visualized is a pair of
oscillators which have the same natural frequency and which are weakly coupled.
For example, the system in Fig. 7 consists of two identical carts, attached to their
respective walls by identical springs (force constant k ) and to each other by a much
weaker spring (force constant k2 k ).

Fig. 7: Two weakly coupled carts. The middle spring which couples the two carts is much weaker than the outer two springs
- see JRT Fig. 11.7

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 18 of 49
Two Weakly Coupled Oscillators

Two Weakly Coupled Oscillators cont

JRT 11.3

We can solve for the normal modes of this system in exactly the same fashion as
before. The spring matrix K takes the form
"
#
k2
k + k2
K=
,
(12.34)
k + k2
k2
and thus

"

#
k + k2 m2
k2
.
(12.35)
2
k + k2 m
k2
The determinant is (k m2 )(k + 2k2 m2 ), and we conclude that the two normal
frequencies are
r
r
k
k + 2k2
and 2 =
.
(12.36)
1 =
m
m
The first frequency is the same as in the previous example, and for the same reason
(the carts move in unison, so the middle spring is irrelevant.) As for the second
mode, the carts oscillate exactly out of phase; this will be proven shortly. However, in
this mode, the strength of the middle spring becomes relevant since it does
experience tension and compression.
(K 2 M) =

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 19 of 49
Two Weakly Coupled Oscillators

Two Weakly Coupled Oscillators cont

JRT 11.3

In the present case, 2 is very close to 1 , since k2 k . To take advantage of this

closeness, we will define 0 to be the average of the two normal frequencies,
1 + 2
.
(12.37)
0 =
2
Since 1 and 2 are very close to each other, 0 is very close
to either, and for most
purposes, we can think of 0 as being the same as 1 = k /m. To show the
difference between 1 and 2 , we can write
1 = 0 and 2 = 0 + .
(12.38)
That is, the small number is half the difference between the two normal
frequencies.
The two normal modes of the weakly coupled carts can now be written as
#
#
"
"
1
1
z(t ) = C1
e i (0 )t and z(t ) = C2
e i (0 +)t ,
1
1
and, as before, their sum is also a solution,
#
#
"
"
1
1
i (0 )t
e
+ C2
e i (0 +)t .
z(t ) = C1
1
1

(12.39)

(12.40)

The constants C1 and C2 are determined by the initial conditions - the positions and
velocities of the two carts at t = 0.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 20 of 49
Two Weakly Coupled Oscillators

Two Weakly Coupled Oscillators cont

JRT 11.3

To see some general features of the solution, we can factor out a term e i 0 t , leaving
)
#
#
"
( "
1
1
z(t ) = C1
e i t + C2
e i t e i 0 t .
(12.41)
1
1
The term in braces, { }, is a vector which depends on t . But since is very small,
this column varies very slowly compared to the second factor e i 0 t . Over a
reasonably short time interval, the first factor is essentially constant, and the solution
behaves like z(t ) = ae i 0 t , with a constant. That is, over any short time interval, the
two carts oscillate sinusoidally with angular frequency 0 . However, if we wait long
enough, the details of the motion will change.
Let us now examine the behaviour of eq. (12.41) for some simple values of the
constants C1 and C2 . First, if either of the constants is zero, then the solution reverts
to one of the normal modes. A more interesting case is that C1 and C2 are equal in
magnitude. For instance, let C1 = C2 = A /2. In this case,
"
#
"
#
A e i t + e i t
cos t
i 0 t
z(t ) =
e
=
A
e i 0 t .
(12.42)
i sin t
2 e i t e i t

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 21 of 49
Two Weakly Coupled Oscillators

Two Weakly Coupled Oscillators cont

JRT 11.3

Of course, the actual motion of the carts is given by the real part of z(t ), resulting in

x1 (t ) = A cos t cos 0 t , and x2 (t ) = A sin t sin 0 t .

(12.43)

This solution has a simple and elegant interpretation. At time t = 0, we have x1 = A ,

whereas x2 = x 1 = x 2 = 0. That is, the solution describes the motion when cart 1 is
pulled a distance A to the right and released at t = 0. Because is very small, there
is an appreciable interval (0 t 1/) during which the functions in eq. (12.43) that
involve t remain essentially unchanged. During this interval, we have

x1 (t ) A cos 0 t , and x2 (t ) 0.

(12.44)

That is, initially cart 1 oscillates with amplitude A and frequency 0 , while cart 2
remains stationary.
Eventually, of course, the factor t starts to become appreciable, and cart 2 starts to
oscillate, also at frequency 0 . The terms involving sin t grow, while the terms
involving cos t shrink toward zero. Eventually, when t = /2, the motion is

x1 (t ) 0, and x2 (t ) A sin 0 t .

(12.45)

The complete motion is shown in Fig. 8.

Those of you in photonics will return to the concept of weakly coupled oscillators
when you study such devices as directional couplers and gratings.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 22 of 49
Two Weakly Coupled Oscillators

Fig. 8:

The positions x1 (t ) and x2 (t ) of two weakly coupled oscillators - see JRT Fig. 11.8

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 23 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum JRT 11.4

The derivations in the previous sections were based on Newtons second law. For
such relatively simple systems, there is no real advantage to using the Lagrangian
approach. However, for more complicated systems, the Lagrangian approach is
considerably simpler.
Here, we will re-derive the equations for the system of two masses and three springs
using the Lagrangian. We shall then do the same for another simple system with two
degrees of freedom, the double pendulum.
Consider the two carts as in Fig. 1. The kinetic energy (assuming massless springs)
is
1
1
(12.46)
T = m1 x 12 + m2 x 22 .
2
2
To write down the potential energy, we must identify the extensions of the three
springs as x1 , x2 x1 , and x2 , from which it follows that the potential energy is

=
=

1
1
1
k1 x12 + k2 (x1 x2 )2 + k3 x22
2
2
2
1
1
(k1 + k2 )x12 k2 x1 x2 + (k2 + k3 )x22 .
2
2

(12.47)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 24 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum cont

JRT 11.4

These results immediately give us the Lagrangian L = T U , from which we find the
Lagrange equations of motion for the two variables x1 and x2 :

and

L
d L
=
dt x1
x1

or

m1 x1 = (k1 + k2 )x1 + k2 x2

(12.48)

d L
L
=
dt x2
x2

or

m2 x2 = k2 x1 (k2 + k3 )x2 .

(12.49)

These are precisely the two equation of motion that we found by the Newtonian
= Kx.
method, which we rewrote in the form Mx

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 25 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum cont

JRT 11.4

Now, consider a double pendulum, comprising a mass m1 suspended by a massless

rod of length L1 from a fixed pivot, and a second mass m2 suspended by a massless
rod of length L2 from m1 , as shown in Fig. 9. We choose as our generalized
coordinates the two angles 1 and 2 .
It is a straightforward matter to write down the Lagrangian L as a function of the two
generalized coordinates. When the angle 1 increases from 0, the mass m1 rises by
an amount L1 (1 cos 1 ) and gains a potential energy

U1 = m1 gL1 (1 cos 1 ).

(12.50)

Similarly, as 2 increases from 0, the second mass rises by L2 (1 cos 2 ) but, in

addition, its point of support (m1 ) has risen by L1 (1 cos 1 ). Thus,
i
h
(12.51)
U2 = m2 g L1 (1 cos 1 ) + L2 (1 cos 2 ) ;
that is, U1 depends only on 1 , while U2 depends on both 1 and 2 . The potential
energy is therefore

U (1 , 2 ) = (m1 + m2 )gL1 (1 cos 1 ) + m2 gL2 (1 cos 2 ).

(12.52)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 26 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum cont

JRT 11.4

The velocity of m1 is just L1 1 in the tangential direction (remember, for circular

motion, v = r ), as shown in Fig. 9, so its kinetic energy is
1
(12.53)
T1 = m1 L12 21 .
2
The velocity of m2 is the vector sum of two velocities - the velocity L2 2 of m2 relative
to its support m1 plus the velocity L1 1 of its support. The angle between these two
velocities is (2 1 ), so the kinetic energy of m2 is
i 1
h
1
(12.54)
T2 = m2 L12 21 + 2L1 L2 1 2 cos(1 2 ) + m2 L22 22 ,
2
2
and the total KE is
1
1
T = (m1 + m2 )L12 21 + m2 L1 L2 1 2 cos(1 2 ) + m2 L22 22 .
(12.55)
2
2

Fig. 9: A double pendulum. The velocity of m2 is the

vector sum of the two velocities shown, separated by an
angle 2 1 - see JRT Fig. 11.9

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 27 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum cont

JRT 11.4

At this point, we could write down the Lagrangian T U and then find the two
Lagrange equations for 1 and 2 . However, the resulting equations are complicated
and not terribly illuminating (and they cant be solved analytically.) This situation is
reminiscent of the simple pendulum, for which the equation of motion
= g sin ) can be solved analytically only by resorting to special functions.
(L
As in that case, we can make a small-angle approximation for the double pendulum.
We assume that 1 and 2 are small, as are their derivatives 1 and 2 . Then, we
Taylor-expand the expressions for T and U and drop all terms that are of third power
or higher in these four small quantities. This gives
1
1
T = (m1 + m2 )L12 21 + m2 L1 L2 1 2 + m2 L22 22
(12.56)
2
2
and
1
1
(12.57)
U = (m1 + m2 )gL1 21 + m2 gL2 22 .
2
2
Now, the Lagrangian L = T U gives us the two Lagranges equations,
L
d L
1 + m 2 L1 L2
2 = (m1 + m2 )gL1 1
=
, or (m1 + m2 )L12
(12.58)
dt 1
1
and

d L
L
=
,
2
dt 2

1 + m2 L 2
or m2 L1 L2
2 2 = m2 gL2 2 .

(12.59)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 28 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum cont

JRT 11.4

These two equations for 1 and 2 can be rewritten as a single matrix equation
= K,
M

(12.60)

where we have introduced the following vector and matrices:

#
"
"
#
"
(m1 + m2 )gL1
1
(m1 + m2 )L12 m2 L1 L2
, K=
=
, M=
0
2
m 2 L1 L2
m2 L22

#
0
.
m2 gL2
(12.61)
The matrix equation is completely analogous to the case for the two carts on
springs. In the present case, the mass matrix M is not actually made up of masses,
but it still plays the role of inertia (note that the elements of M have the same units as
the moment of inertia I). Similarly, the spring-constant matrix K is not actually
made up of spring constants; its elements have units of energy.
Nevertheless, the problem is solved by the same methods. We seek to find solutions
- normal modes - in which the two coordinates 1 and 2 vary sinusoidally with the
same angular frequency . As before, any such solution (t ) can be written as the
real part of a complex solution z(t ) whose time dependence is just e i t . That is,
"
#


a1
(t ) = Re z(t ) where z(t ) = ae i t =
e i t .
(12.62)
a2

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 29 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum - Equal Lengths

and Masses JRT 11.4
To simplify the discussion, let us now restrict our attention to the case that our
double pendulum has equal masses, m1 = m2 = m, andpequal lengths, L1 = L2 = L .
The equations become much tidier if we recognize that g /L is the frequency of a
single pendulum of the same length L . If we call this frequency 0 , then we can
replace g with L 20 and the matrices M and K become
#
"
"
#
2 1
220 0
2
2
M = mL
and K = mL
.
(12.63)
1 1
0
20
The matrix (K 2 M) of the eigenvalue equation is therefore
"
#
2(20 2 )
2
.
(K 2 M) = mL 2
2
2
2

(0 )

(12.64)

The normal frequencies are determined by setting the determinant of the above
matrix to zero, which gives
2(20 2 )2 4 = 4 420 2 + 240 = 0,
(12.65)

2
2
with the two solutions = (2 2)0 . That is, the two normal frequencies are
q
q

(12.66)
1 = 2 20 0.770 and 2 = 2 + 20 1.850 .

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 30 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum - Equal Lengths

and Masses cont JRT 11.4
Knowing the two normal frequencies, we can now find the motion of the double
pendulum in the corresponding normal modes, by solving the equation
(K 2 M)a = 0 with = 1 and = 2 in turn. If we use 1 , we get
#
"

2
2

(K 21 M) = mL 2 20 ( 2 1)
.
(12.67)
2
1

Therefore, the equation (K 21 M)a = 0 implies that a2 = 2a1 , and if we write

i 1
a1 = A1 e , the two coordinates are
"
#
"
#
n
o
1
1 (t )
i 1 t

=
= Re ae
= A1
cos(1 t 1 ).
(12.68)
2 (t )
2
In this first normal mode, the two pendulums
oscillate exactly in phase, with the

amplitude of the lower pendulum 2 times that of the upper pendulum, as shown in
Fig. 10.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 31 of 49
Lagrangian Approach: The Double Pendulum

Fig. 10: The first normal mode for a double pendulum with
equal masses and equal lengths. The two angles 1 and 2
oscillate in phase, with the amplitude for 2 larger by a factor

of 2 - see JRT Fig. 11.10

Fig. 11:

The second normal mode for a double pendulum

with equal masses and equal lengths. The two angles 1 and
2 oscillate exactly out of phase, with the amplitude for 2

larger by a factor of 2 - see JRT Fig. 11.11

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 32 of 49
Lagrangian Approach: The Double Pendulum

Lagrangian Approach: The Double Pendulum - Equal Lengths

and Masses cont JRT 11.4
Turning to the second mode, we find that
(K

22 M)

= mL

20 (

2 + 1)

"

2
2

#
2
.
1

(12.69)

The equation (K 22 M)a = 0 implies that a2 = 2a1 , and if we write a1 = A2 e i 2 ,

the two coordinates are
"
#
"
#
o
n
1
1 (t )
i 2 t

(12.70)
= A2
cos(2 t 2 ).
=
= Re ae
2 (t )
2
In this second normal mode, the two pendulums
oscillate exactly out of phase, with

the amplitude of the lower pendulum 2 times that of the upper pendulum, as
shown in Fig. 11.
The general solution, of course, is a linear combination of these normal modes, with
relative amplitudes and phases that are determined by the initial conditions.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 33 of 49
Lagrangian Approach: The Double Pendulum

Example #1 - Cart and Pendulum JRT Prob. 11.19

Problem: A simple pendulum (mass M and length L is suspended from a cart (mass m) that can oscillate on the end of a
spring of force constant k , as shown in the figure.
(a) Assuming that the angle remains small, write down the systems Lagrangian and the
equations of motion for x and .
(b) Assuming that m = M = L = g = 1 and k = 2 (all in appropriate units), find the normal
frequencies, and for each normal frequency find and describe the motion of the corresponding
normal mode.

Fig. 12:

Geometry of the cart and pendulum problem

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 34 of 49
Lagrangian Approach: The Double Pendulum

Example #1 - Cart and Pendulum cont

JRT Prob. 11.19

Solution:
(a) Using the equilibrium position (x = = 0) as the reference point for potential energy, we have

U = Uspr + Upen =

1 2
1
1
kx + MgL (1 cos ) kx 2 + MgL 2 ,
2
2
2

(12.71)

where the small-angle approximation has been used. The kinetic energy has two components due to the two moving masses.
That of the cart is simply T = 21 mx 2 . As for the pendulum, the velocity of the bob is a bit trickier to describe - its equal to the
velocity of the bob relative to the pivot plus the velocity of the pivot relative to the origin. Therefore,
2
1
1 
1
1
T = mx 2 + M x + L = (m + M )x 2 + ML x + ML 2 2 .
(12.72)
2
2
2
2
All together,

L=T U =

The Lagrange equations read

1
1
1
1
(m + M )x 2 + ML x + ML 2 2 kx 2 MgL 2 .
2
2
2
2

(12.73)

d L
L

=
kx = (m + M )x + ML
x
dt x
L
d L

=
MgL = ML x + ML 2 .

dt

(12.74)

= Kq, where
(b) We can combine these equations in the form Mq
q=

"

, M=

"

m+M
ML

ML
ML 2

"

2
1

1
1

K=

"

k
0

0
MgL

"

2
0

0
1

(12.75)

(note that our choice of numerical values has made this problem mathematically equivalent to that of the double pendulum with
equal lengths and equal masses).

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 35 of 49
Lagrangian Approach: The Double Pendulum

Example #1 - Cart and Pendulum cont

JRT Prob. 11.19

Continuing, we have
K 2 M =

"

2 22
2

2
1 2

(12.76)

The requirement that the determinant of this matrix equals zero leads to the characteristic equation 4 42 + 2 = 0, with
resulting normal frequencies

1 =

2 = 0.77,

2 =

2+

2 = 1.85

(12.77)

(with proper units implied). As for the normal modes, as for the double pendulum problem, we have
1 a2 =

2a1 ,

2 a2 = 2a1 ,

In the first mode, and x are in phase, with (in radians) having
mode, they are out of phase with the same ratio of magnitudes.

(12.78)

2 times the magnitude of x (in meters). In the second

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 36 of 49
The General Case

The General Case JRT 11.5

We have now studied in detail the normal modes of two systems - a pair of carts
attached to three springs and a double pendulum. Now, we are ready to discuss the
general case of a system with n degrees of freedom that is oscillating about a point
of stable equilibrium. Since the system has n degrees of freedom, its configuration
can be specified by n generalized coordinates, q1 , , qn . We abbreviate this set as
q = (q1 , , qn ). For example, for the two carts that we studied earlier, q = (x1 , x2 )
and for the double pendulum, q = (1 , 2 ). Note carefully that q is not, in general, a
three-dimensional vector; it is a vector in the n-dimensional space of the generalized
coordinates.
We shall assume that the system is conservative, so that it has a potential energy

U (q1 , , qn ) = U (q)
and Lagrangian L = T U . The kinetic energy, as usual, is
1X
2
T=
m r ,
2
where the sum runs over all of the particles that make up the system.

(12.79)

(12.80)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 37 of 49
The General Case

The General Case cont

JRT 11.5

We must rewrite the KE in terms of the generalized coordinates, using the relation
between the Cartesian coordinates r and the generalized coordinates,
r = r (q1 , , qn ),

(12.81)

where we take for granted that this relation does not involve the time t explicitly (that
is, the coordinates are natural.) We saw back in our initial study of Hamiltonian
mechanics that if we differentiate eq. (12.81) with respect to t and substitute into the
kinetic energy, we find that
1X
=
T = T (q, q)
Ajk (q)q j q k
(12.82)
2 j ,k
where the coefficients Ajk (q) may depend only on the coordinates q. Under our
= T (q, q)
U (q),
present assumptions, the Lagrangian has the general form L(q, q)
is given by the previous equation and U (q) is an as-yet-unspecified
where T (q, q)
function of the coordinates q.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 38 of 49
The General Case

The General Case cont

JRT 11.5

As a final assumption, we presume that the system is making small oscillations

about a stable equilibrium configuration. By redefining the coordinates if necessary,
we can arrange that the equilibrium position is q = 0. Then, since we are interested
only in small oscillations, we have only to concern ourselves with small values of the
coordinates q, and we can use Taylor expansions of T and U about the equilibrium
point q = 0. For U , this gives
X U
1 X 2 U
U (q) = U (0) +
qj +
qj qk +
(12.83)

q
2 j ,k qj qk
j
j
where all derivatives are evaluated at q = 0. Luckily, this equation can be simplified.
Since U (0) is a constant, we can simply redefine the zero-level of energy so that this
term equals zero. Next, since q = 0 is an equilibrium point, all of the first derivatives
U /qj are zero.
As a further simplification, we shall rename the second derivatives as
2 U /qj qk = Kjk , where Kjk = Kkj by inspection. And finally, since the oscillations
are small, we shall neglect all terms higher than second order in the small quantities
This reduces U to
q or q.
1X
U = U (q) =
Kjk qj qk .
(12.84)
2 j ,k

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 39 of 49
The General Case

The General Case cont

JRT 11.5

The kinetic energy is a bit simpler to rearrange. Every term in eq. (12.82) contains a
factor q j q k which is already second order in small quantities. Therefore, we can
ignore everything but the constant term in the expansion of Ajk (q). If we call this
constant term Ajk (0) = Mjk , this reduces the KE to
1X
=
T = T (q)
Mjk q j q k ,
(12.85)
2 j ,k
and the Lagrangian to

U (q).
= T (q)
(12.86)
L(q, q)
Now we can easily write down the equations of motion. Since there are n
generalized coordinates, there are n corresponding Lagrange equations,
L
d L
[i = 1, , n] .
=
(12.87)
dt qi
q i
To write these equations explicitly, we need to differentiate our expressions for T and
U . Differentiating sums can be a bit tricky. For example, for a system with n = 2, the
equation for U reads

1X
1
U=
Kjk qj qk =
K11 q12 + K12 q1 q2 + K21 q2 q1 + K22 q22
2 j ,k
2
(12.88)

1
=
K11 q12 + 2K12 q1 q2 + K22 q22 .
2

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 40 of 49
The General Case

The General Case cont

JRT 11.5

In this form we can easily differentiate with respect to either q1 or q2 . For example,
U
(12.89)
= K11 q1 + K12 q2
q 1
with a corresponding expression for U /q2 . For systems with more than 2 degrees
of freedom, we have a similar expression,
X
U
=
Kij qj [i = 1, , n] .
(12.90)
q i
j
Since differentiation of the kinetic energy works in exactly the same way, we can
write down the n Lagrange equations
X
X
j =
Mij q
Kij qj [i = 1, , n] .
(12.91)
j

These n equations can be grouped into a single matrix equation

= Kq
Mq

(12.92)

where M and K are the n-by-n mass and spring constant matrices with elements
Mij and Kij respectively.
Note that we only need to find the elements of the matrices M and K to solve the
problem at hand; it is not necessary to write down the Lagrangian or the Lagrange
equations.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 41 of 49
The General Case

The General Case cont

JRT 11.5

The preceding matrix equation is just the n-dimensional equivalent of the

two-dimensional equations that we derived earlier for the two carts and for the
double pendulum. It is solved in exactly the same way; the eigenvalues determine
the normal frequencies, and the eigenvectors determine the normal modes. Of
course, finding these values for an n-by-n matrix is often no easy task.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 42 of 49
Three Coupled Pendulums

Three Coupled Pendulums JRT 11.6

Consider three identical pendulums coupled by two identical springs as shown in
Fig. 13. We use the generalized coordinates 1 , 2 , and 3 , with the system at
equilibrium when all three angles are zero. Our first task is to write down the
Lagrangian for small displacements.
From the results of the previous sections the KE of the three pendulums is seen to
be
1
T = mL 2 ( 21 + 22 + 23 ).
(12.93)
2
The gravitational potential energy of each pendulum has the form
mgL (1 cos ) 21 mgL 2 . Thus, the total gravitational potential energy is
1
Ugrav = mgL (21 + 22 + 23 ).
(12.94)
2

Fig. 13:

Three identical pendulums of lengths L and

masses m are coupled by two identical springs with spring
constants k - see JRT Fig. 11.13

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 43 of 49
Three Coupled Pendulums

Three Coupled Pendulums cont

JRT 11.6

We also need to account for the potential energy due to the springs. This is a
function of how much each is stretched or compressed. For arbitrary values of the
angles , this is rather messy. However, for small angles, the only appreciable
stretching and compression comes from the horizontal displacement of the masses,
each of which moves a distance of approximately L (to the right, for positive .)
Thus, for example, the left spring is stretched by about L (2 1 ). The total spring
potential energy is
i
1 2h
(12.95)
Uspr =
kL (2 1 )2 + (3 2 )2
2

1 2 2
=
kL 1 + 222 + 23 21 2 22 3 .
2
At this point, the textbook introduces a new notation system called natural
units in which certain parameters (in this case, mass m and length L ) are
given the dimensionless value of unity. While there are good reasons for this
approach, I disagree with its use at this level, as it obscures some of the
physics. Thus, the remainder of this section of notes will differ slightly from
the way it is presented in the textbook.

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 44 of 49
Three Coupled Pendulums

Three Coupled Pendulums cont

JRT 11.6

At this point, we could write down the Lagrangian and then the Lagrange equations.
But weve already done this. We already know that the result will be the familiar
matrix equation
= K
M
(12.96)

where in this case, is the 3-element vector containing the three angles 1 , 2 , and
3 , and the matrices are

2
kL 2
0

1 0 0
mgL + kL

mgL + 2kL 2
kL 2
kL 2
M = mL 2 0 1 0 , K =
. (12.97)

0
kL 2
mgL + kL 2
0 0 1
The normal modes of our
have the familiar form
n system
o


(t ) = Re z(t ) = Re ae i t , where a and are determined by the eigenvalue
equation (K 2 M)a = 0. Written out in full (and dividing through by L 2 ,) the matrix
is
mg

2
k
0
L + k m

mg

2
2

.
k
+
2
k

k
(K M) =
(12.98)

mg
2
0
k
+ k m
L

The determinant is (somewhat) easily evaluated as





g
g
g
3k
k
det(K 2 M) =
2
+
2
+
2 .
L
L
m
L
m

(12.99)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 45 of 49
Three Coupled Pendulums

Three Coupled Pendulums cont

JRT 11.6

Thus, the three normal frequencies are

r
r
g
g
k
, 2 =
+ ,
1 =
L
L
m

3 =

g
3k
+
.
L
m

(12.100)

For the first mode, a1 = a2 = a3 = Ae i . That is,

1 (t ) = 2 (t ) = 3 (t ) = A cos(1 t ).

(12.101)

In this case, the three pendulums oscillate in unison (with equal amplitudes and
phases,) as shown in Fig. 14(a). In this mode, the springs are neither stretched nor
compressed, and their presence is irrelevant;
each pendulum oscillates just like a
p
single pendulum with frequency 1 = g /L .

Similarly, the second mode gives us a1 = a3 = Ae i , a2 = 0. In this case, the outer

two pendulums oscillate exactly out of phase, while the middle one sits at rest, as
shown in Fig. 14(b). Finally, the third mode gives us a1 = 21 a2 = a3 = Ae i . In this
mode, the outer two pendulums oscillate in unison while the middle one oscillates
with twice the amplitude and exactly out of phase, as shown in Fig. 14(c).

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 46 of 49
Three Coupled Pendulums

Fig. 14:

The three normal modes for three coupled pendulums (equal masses, equal lengths, equal spring constants - see
JRT Fig. 11.14

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 47 of 49
Three Coupled Pendulums

Example #2 - Linear Triatomic Molecule

Problem: A simplistic model of a linear triatomic molecule such as carbon dioxide is shown in the figure below. For this
molecule, two identical atoms of mass m are connected by two identical springs (spring constant k ) to a single atom of mass
M . We assume that the system is confined to move in one dimension (that is, the three atoms are always colinear). The
equilibrium state of the molecule occurs when both springs are at their natural length.
(a) Write down the kinetic energy, the potential energy, and the Lagrangian for this system, using
the displacement of each atom from its equilibrium position as the three generalized
coordinates.
(b) Write down the Lagrange equations for the three generalized coordinates, and then express
these equations as a single matrix equation.
(c) Solve for the normal frequencies of the system, as a function of m, M , and k .
(d) Find and describe the motion in each of the three normal modes.

k
M

m
x2

x1
Fig. 15:

Geometry of the linear triatomic molecule

x3

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 48 of 49
Three Coupled Pendulums

Example #2 - Linear Triatomic Molecule cont

Solution: We will introduce the shorthand notation = M /m.
(a) The total KE is
1
1
1
1
mx 12 + M x 22 + mx 32 = m(x 12 + x 22 + x 32 ).
2
2
2
2
The total PE is

1
1
1
k (x2 x1 )2 + k (x3 x2 )2 = k (x12 + 2x22 + x33 2x1 x2 2x2 x3 ).
2
2
2

(12.102)

(12.103)

Therefore, the Lagrangian is

L=

1
1
m(x 12 + x 22 + x 32 ) k (x12 + 2x22 + x33 2x1 x2 2x2 x3 ).
2
2

(12.104)

(b) The Lagrange equations read

d L
L
=
dt x1
x1

mx1 = kx1 + kx2

(12.105)

L
d L
=
x2
dt x2

mx2 = kx1 2kx2 + kx3

(12.106)

d L
L
=
x3
dt x3

mx3 = kx2 kx3 .

(12.107)

= Kx, where
This can be expressed as the matrix equation Mx

M = 0

0
m
0

0
0 ,

and K = k

k
2k
k

0
k .

(12.108)

PC235 Winter 2013 Chapter 12. Coupled Oscillators and Normal Modes Slide 49 of 49
Three Coupled Pendulums

Example #2 - Linear Triatomic Molecule cont

(c) To find the normal frequencies, we must set det(K 2 M) = 0, where

k
k m2

K 2 M =
2k m2
k

0
k

0
k
k m 2

This determinant is a bit easier to deal with if we divide through by m and recall that k /m = 20 :

2
2
20
0

 0

1 
2
2
2
2
2

.
0
20
0
K M =

2
2
2
0
0
0


 
The determinant is 2 (2 20 ) 2 2+
20 , and therefore the normal frequencies are

r
r
r
k
2+
k
2k
0 =
1 = 0, 2 = 0 =
, 3 =
+
.
m

m
M

(12.109)

(12.110)

(12.111)

(d) For the case of 1 , the zero frequency just means that all three masses have identical motion; that is, the molecule is
drifting through space. This type of motion is possible because there are no forces anchoring the molecule to a particular
position in space (in contrast, the other systems we have studied, such as two carts with three springs and the double
pendulum, always have at least one fixed reference point...motion away from these points produces a restoring force.) Since
neither spring expands or contracts inthis mode, there is zero oscillation frequency.
For the case of 2 , substituting = k /m into (K 2 M)a = 0 results in the solution a2 = 0, a3 = a1 . In this mode, the
center mass M is stationary, while theqtwo outer masses oscillate with equal amplitudes and out of phase.

k
2k
2
For the case of 3 , substituting =
m + M into (K M)a = 0 results in the solution a1 = a3 = (/2)a2 . In this mode,
the two outer masses move with equal amplitude and in phase, while the middle mass moves out of phase with the outer
masses, and with an amplitude that is /2 as large.