8 Coupled Oscillators and Normal Modes

Fall 2003

An undamped harmonic oscillator (a mass m and a Hooke's-law spring with force

constant k) has only one natural frequency of oscillation, o = k m. But when two
or more such oscillators interact, several natural frequencies are possible.
Let's consider a system of masses and Hooke's-law springs that has a stable equilibrium
position, such that each mass can vibrate around its equilibrium position. Are there
possible motions in which every mass moves with simple harmonic motion, all masses
with the same frequency? Such a motion, when it exists, is called a normal-mode motion.
We will now develop general methods for finding the possible normal modes of such a
system and their associated frequencies. We'll assume throughout that the spring forces
are linear functions of displacement. We'll illustrate the general method by use of
the following specific example.
Example
Let's consider the system shown below. The two masses move along a straight line. In
the equilibrium positions, the springs are neither stretched nor compressed, and the
coordinates x 1 and x 2 are the displacements of the particles from equilibrium.

If k ' = 0 (i.e., if the center spring is removed), we have two uncoupled harmonic
oscillators; each one can vibrate with angular frequency = k m, with arbitrary
amplitude and phase. When the central spring is included, there are two cases where the
masses can oscillate with the same frequency:
1) If x 1 = x 2 at each instant, then spring k ' is never stretched or compressed, and it can
be ignored. The two masses vibrate sinusoidally, in phase, with the same angular
frequency = k m, and with equal amplitudes.
2) If x 1 = x 2 at each instant, the midpoint of spring k ' is stationary, and the force it
exerts on each mass is like that of a spring with force constant 2 k '. The total force on
each mass is the same as for a spring with force constant k + 2 k '. In this case, the two
masses move sinusoidally with angular frequency = ( k + 2 k ' ) m , with equal
amplitudes but a half-cycle out of phase.
Thus this system has two normal modes, one with angular frequency =

k m , the

other with = ( k + 2 k ' ) m . Each mode has a characteristic vibration pattern (i.e., a
relation between the amplitudes and phases of the motions of the two masses).

8-2

8 Coupled Oscillators and Normal Modes

General Method
In our example, the symmetry of the problem allowed us to guess the normal modes, but
we need a more systematic and general approach. To develop this approach, we start
with the equations of motion, from F = ma. They are
kx1 + k ' ( x2 x1) = mx&&1 ,
kx2 k ' ( x2 x1 ) = mx&&2 .

(1)

(Be sure you understand the various + and signs in these equations.)
Now we guess that these equations have a solution in the form

x1 = a1 cos t + ,

x 2 = a 2 cos t + ,

(2)

where is not yet known and the amplitudes a1 and a2 may be related.
To test whether (or under what circumstances) these expressions really do satisfy Eqs.
(1), we carry out the derivatives and substitute back into Eqs. (1). After dividing out the
common factor cos (t + ) and re-arranging, we get
k a1 + k ' ( a2 a1 ) = m 2 a1 ,
k a2 k ' ( a2 a1 ) = m 2 a 2 ,
or
( k + k ' m 2 )a1 k ' a2 = 0,
k ' a1 + ( k + k ' m 2 ) a2 = 0.

(3)

That is, Eqs. (2) are a solution of the equations of motion, Eqs. (1), if (and only if)
Eqs. (3) are satisfied.
Eqs. (3) are a pair of simultaneous, homogeneous equations for the amplitudes a1 and
a2 . They always have the trivial solution a1 = a2 = 0. A fundamental theorem of linear
algebra states that non-trivial solutions for these equations exist if (and only if) the
determinant of the system is zero. Thus a necessary and sufficient condition for the
existence of non-trivial solutions is
k + k ' m 2
k '

k'
= 0.
k + k ' m 2

(For further discussion of this point, please refer to Section 9, pages 9-4 and 9-5.)
The values of k, k ', and m are fixed, so we conclude that the assumed solution is valid
only for certain special values of , the values that satisfy Eq. (4). This equation for
is called the secular equation. When we multiply out the determinant, we get

(4)

8 Coupled Oscillators and Normal Modes

( k + k ' m 2 ) 2 = k ' 2

8-3

k + k ' m 2 = k '.

and

(5)

Assuming 0, we get
=

k
m

and

k + 2k '
,
m

(6)

in agreement with our previous result.

Each value of in Eqs. (6) is a normal-mode frequency. To get the amplitude relations
for each normal mode, we substitute each value for back into Eqs. (3):
For 2 =

k
,
m

k ' a1 k ' a2 = 0,
k ' a1 + k ' a2 = 0.

I.e.,

a1 = a2 .

(7)

For 2 =

k + 2k '
,
m

k ' a1 k ' a2 = 0,
k ' a1 k ' a2 = 0.

I.e.,

a1 = a2 .

(8)

In each case, the equations don't give us specific values for a1 and a2 , but they show
that they must be related in a very particular way.
Each of these possibilities is a normal mode, with a definite frequency and a relation
between the amplitudes, describing the pattern of the motion. The most general motion
of the system is a superposition of these normal modes, with arbitrary phases. Call the
modes Mode 1 and Mode 2, with
1 =

k
m

2 =

and

k + 2k '
.
m

(9)

Then for Mode 1, with a1 = a2 = a,

d
i
= a cos b t + g,

x1 = a cos 1 t + 1 ,
x2

(10)

and for Mode 2, with a1 = a2 = b,

x1 = b cos 2 t + 2 ,

(11)

x2 = b cos 2 t + 2 .
The most general motion of the system is described by equations with four arbitrary
constants a, b, 1 , and 2 , determined by the initial conditions:

d
i
d
i
= a cos b t + g b cos b t + g.

x1 = a cos 1 t + 1 + b cos 2 t + 2 ,
x2

(12)

The amplitude factor a and the phase angle 1 are associated with Mode 1, and b and
2 are associated with Mode 2.

8-4

8 Coupled Oscillators and Normal Modes

Normal Coordinates
Suppose we call the two sinusoidal variables q1 and q2 :

d
i
= b cos b t + g.

q1 = a cos 1 t + 1 ,
q2

x1 = q1 + q2 ,
x2 = q1 q2 .

Then

(13)

We can express the equations of motion, Eqs. (1), in terms of the variables q1 and q2 .
Substituting Eqs. (13) into Eqs. (1) and re-arranging, we get

b
g
b g
mb q&& q&& g = kq + b k + 2k 'gq .
m q&&1 + q&&2 = kq1 k + 2k ' q2 ,
1

(14)

Now note that by simply adding or subtracting these equations, we can obtain separate
equations for q1 and q2 :
mq&&1 = kq1,

(15)

mq&&2 = k + 2 k ' q2 .
These equations show that the coordinate q1 varies sinusoidally with angular frequency
1 = k m , and q2 with angular frequency 2 = ( k + 2 k ' ) m , as we should
expect. Thus if we had been lucky enough (or clever enough) to use q1 and q2 as
coordinates at the start, we would have obtained the normal modes immediately. Each
normal mode consists of a motion in which only one of the q's is different from zero.

The q's are called normal coordinates for this system, and the transformation from the
x's to the q's, given by Eqs. (13), is a normal-coordinate transformation. You can verify
that the inverse transformation, giving the q's in terms of the x's, is
q1 =
q2 =

1
2
1
2

b x + x g,
b x x g.
1

(16)

Energy
The total energy of the system can be expressed in terms of the normal coordinates. In
terms of the original coordinates x 1 and x 2 , the total energy is
E = T+V =

1
2

mx&12 +

1
2

mx&2 2 +

1
2

kx12 +

1
2

kx2 2 +

1
2

k ' x1 x2 .

(17)

Substituting Eqs. (13) into this expression and re-arranging, we get

E = m q&12 + q&2 2 + kq12 + k + 2 k ' q2 2

= mq&12 + kq12

mq& 2 2 + k + 2k ' q2 2 .

(18)

8 Coupled Oscillators and Normal Modes

8-5

We see that the total energy separates into terms containing only q1 and its derivative
and terms containing only q2 and its derivative. We invite you to verify that a similar
separation occurs with the Lagrangian function L = T V.
Limitations
Our entire analysis has made use of the fact that the restoring forces are linear functions
of the coordinates. When non-linear forces are present, in general there is no such thing
as normal-mode motion. Just as a single-mass oscillator with a non-linear restoring force
has a frequency that depends on the amplitude of the motion, so it is with more complex
systems when non-linear forces are present.
Matrix Formulation
The normal-coordinate transformation given by Eqs. (13) and (16) can be expressed
compactly using matrix language. Each set of coordinates is represented by a onecolumn matrix, and the linear transformation from one to the other is given by a square
matrix A. In the above example,
x=

FG x IJ ,
Hx K

q=

FG q IJ ,
Hq K
1

A =

FG1 1 IJ .
H1 1K

Then the transformation given by Eqs. (13) can be written simply as x = Aq.
Similarly, the inverse transformation can be written as q = A1 x, where
A 1 =

FG
H

1
2
1
2

1
2

1
2

IJ
K

FG
H

IJ
K

1 1 1
.
2 1 1

You can verify that AA1 = A1 A = I, where I is the identity matrix (or unit matrix):
I=

FG 1 0IJ .
H 0 1K

Matrix methods are a very powerful tool for analyzing normal modes of complex
systems. We'll discuss them in a separate chapter.
(If you need to review matrix multiplication or other aspects of matrix algebra, please
consult Edwards and Penney Differential Equations and Boundary Value Problems,
2nd ed., Section 5.1, pp. 284-290. This section also includes a brief discussion of
determinants.)

(19)

8-6

8 Coupled Oscillators and Normal Modes