Chapter 5

Wave equation. Waves

in strings. Sound
 Chapter 5

Introduction The method we developed in the previous chapter makes it

possible to study the vibrations of any real system. Let us
discuss for example a guitarÕs string. We can consider the
string as a chain of masses. The ÒmassesÓ of the chain
correspond to the atoms in the string. Because there are
zillions of atoms in a string, there is a correspondingly huge
number of normal modes. However, the modes we excite
when we play the guitar are a few modes for which the
wavelength is of the order of the size of the guitar, say 1 m.
This must be contrasted with the separation between atoms,
which is of the order of 1 •, or 10 -10 m. If you look at the
normal mode solutions with long wavelengths in systems
with very large N you notice that nearby masses tend to
have similar displacements. In the case of the string, with
the enormous discrepancy between atomic separations and
Òmusic-likeÓ wavelengths, it is apparent that a huge
number of atoms near a certain point x in the string are
undergoing virtually identical displacements. We can
thus make a ÒcontinuumÓ approximation. Rather than
labeling the atoms one by one, we assume that there is a
function ξ(x,t) that gives the displacement of the atoms near
position x at time t. Since the atoms are so closely spaced
compared with the wavelength of the Òmusic-likeÓ normal
modes, we can assume that the function ξ(x,t) is continuous
and differentiable. This leads to the famous wave equation.

00
11 111
000
The continuum approximation

11
00 111
000
1
0 111
000
Let us consider the many-mass chain depicted in Fig. 1

11
00
11
00
0000000000000000000
1111111111111111111
111
000
1
0 111
000
111
000
11
00 111
000 ξ(x,t)
Figure 1 The displacement pattern for the transverse
oscillations of a chain of masses for one of its normal
modes.

The displacement of the masses at position x is denoted by

ξ(x,t). Notice that in general there are masses only at certain
values of x (given by x = na). However, when we are consid-

phy 241 MenŽndez 76

 Chapter 5

ering a real guitar string, where the separation of atoms is so

small compared with the length of the string, the possible
values of x are so closely spaced that we can assume that x is
a continuous variable. We can thus differentiate the func-
tion ξ(x,t) not only with respect to t but also with respect to
x.

If the displacement of the particle is given by ξ(x,t), its

Mξ
velocity is the time derivative of this quantity, . Notice
Mt
that we use the partial derivative notation because ξ is a
function of x and t and we mean differentiation with respect
to t with x fixed. Similarly, the acceleration of the particle is
M 2ξ
given by 2 . We can thus rewrite Eq. (2) in Chapter 4 in
Mt
Mξ
terms of . We obtain
Mt
M 2 ξ(x,t) ξ(x + a,t) − ξ(x,t) ξ(x − a,t) − ξ(x,t)
m 2 = T +
Mt a a (1)

Since a is so small, the two terms on the right hand side are
virtually identical to the first derivatives of ξ(x,t). The first
term is the derivative at x, and the second term is minus the
derivative at x-a:
M 2 ξ(x,t) Mξ Mξ
m 2 = T (x, t) − (x − a, t) .
Mt Mx Mx (2)

If we divide by a on the two sides, the right hand side is by

definition (for very small a as in our case) the second deriva-
tive of ξ(x,t) evaluated at x-a:
M 2 ξ(x,t) M 2 ξ(x − a, t)
ρ 2 = T 2 ,
Mt Mx (3)

where ρ = m/a is the linear mass density. For a string with

total mass M and length L, ρ = M/L. In the limit a→0, this
becomes
M 2 ξ(x,t) T M 2 ξ(x, t)
=
Mt
2
ρ Mx
2
(4)

phy 241 MenŽndez 77

 Chapter 5

which can be written as

M ξ(x,t) 2 M ξ(x, t)
2 2

2 = c 2
Mt Mx (5)

This is the famous wave equation. Notice that we have

derived it solely on the basis of NewtonÕs laws combined
with some type of restoring force. However, while NewtonÕs
laws are universally valid, the wave equation is only valid in
the Òlong wavelength limitÓ, which includes only those
normal modes for which the wavelength λ is much longer
than the distance between particles. So we cannot obtain all
normal modes of a system from the wave equation, because
some of those normal modes do not satisfy the conditions
upon which the wave equation was derived. Still, the wave
equation is extraordinarily useful and important because in
many practical applications the normal modes of interest are
precisely the Òlong wavelengthÓ modes. The quantity c has
units of speed and is called the wave speed for reasons that
will become apparent later. For the transverse oscillations of
T
the linear chain, c = . Many other processes satisfy a wave
ρ
equation identical to Eq. (5). The expression for c changes
according to the physical process under consideration, but it
is always given by (the square root of) a quantity that
measures the strength of the restoring forces divided by a
quantity that measures the inertia of the system.

Standing-wave solutions to the wave equation

Since the wave equation was derived as the continuum limit
of the equations of motion for a chain of atoms, it is
reasonable to expect that the solution to the wave equation
will be given by the continuum limit of the ÒexactÓ solution,
which is given by Eq. (13), Chapter 4. (For the transverse
oscillations, we write yn instead of xn in Eq. (4), Chapter 4).
In the continuum limit na → x, so that we can try a solution
of the form

ξ(x , t) = A sin kx cos (ωt + α ) (6)

If we substitute this into Eq. (5) we find that this expression

phy 241 MenŽndez 78

 Chapter 5

is indeed a solution to the wave equation provided that ω2

= c2k2 or ω = ck. Using ω = 2πν and λ = 2π/k, this can also
be written as

c = λν, (7)

which is easier to remember because the units of the

quantities involved are very obvious.

The result ω = ck was already derived as a long-wavelength

limit of the exact solution in Chapter 4. For the continuum
approximation to be valid λ >> a. This is equivalent to ka <<
4T ka 4T ka T
1. In this case, ω = sin • = k. So c =
am 2 am 2 ρ
T
, which is exactly the expression for the speed of the
ρ
wave derived from the wave equation for the case of
transverse oscillations in a string.

Change of notation in the discrete-to-continuum limit

While the transition from a discrete system of many masses
to a continuum of masses is conceptually straightforward,
some subtle changes in notation make it hard to understand
the physical meaning of certain expressions.

When we discuss a system of N masses connected by

springs, the displacement of mass n is given by a function
xn(t) (or yn(t) if we are discussing the transverse oscillations
of the system). Here the variable x gives the instantaneous
position of the particle, so that its velocity is given by dxn/dt
and its acceleration by d 2 x n /dt 2 . On the other hand, the
meaning of x when we make the continuum approximation
is completely different. Here x is not the instantaneous position
of the particle, which is given by ξ(x,t), but a label that
indicates the position of the particle along the string. Hence
x is the continuum equivalent of n. In words, ξ(x,t) is the
displacement, at time t, of a particle whose equilibrium
position is x. When we work in the continuum limit, the
expression dx/dt is not the velocity of any particle. In fact, x is
not a function of time at all, it is just a label, the same way n
is a label in the discrete case and we never use the derivative

phy 241 MenŽndez 79

 Chapter 5

dn/dt. The correct expression for the velocity of the particle

Mξ
in the continuum limit is v = ; the acceleration is given by
Mt
M 2ξ
a= 2 .
Mt

Standing waves in a guitar string

The standing waves in a guitar string with both ends fixed
are given by Eq. (6). The possible values of k are given in Eq.
(18), Chapter 4, with (N+1)a = L, where L is the length of
the string. Hence the expression for the possible
wavenumbers become
π 2π 3π
k1 = , k2= , k3= , ... (8)
L L L

This expression could have been derived directly from the

continuum solution Eq. (6). If both ends are fixed ξ(0,t) = 0
= ξ(L,t). This requires sin kL = 0 or kL = πi.

Using λ = 2π/k, the possible values of the wavelength λ are

given by
2L
λi =
i (9)

The first two modes are indicated in Fig. 3. Using ν = ω/2π,

T
ω = ck, and c = , the frequencies of the modes in the
ρ
string can be written as
i T
νi=
2L ρ (10)

The lowest possible frequency, given by ν 1 =1/2L(T/ ρ )1/2 is

called the fundamental frequency. The other possible
frequencies, called harmonics, turn out to be multiples of
the fundamental.

Completeness of the solutions

In the case of a discrete chain with N masses, we have N
different normal modes given by the N different possible
values of k. Each normal mode is a standing wave of the form

phy 241 MenŽndez 80

 Chapter 5

x n (t) = A sin kna cos (ωt + α). There are two adjustable
constants in this expression. A and α.. The most general
solution can be written as a sum of all possible normal
modes as in Eq. 16, Chapter 4. Such a solution contains 2N
adjustable constants, exactly the number needed to accom-
modate the 2N initial conditions (initial position and initial
velocity).

A continuum system has an infinite number of degrees of

freedom. This corresponds to an infinite number of normal
modes. In fact, the possible number of k values is no longer
limited to N. In Eq. (8) for the guitar string, for example,
any k value of the form k i = πi/L gives rise to a different
normal mode. This is very different from the discrete case,
where the solutions start to repeat themselves after the first
N values of k. In short, there are only N possible normal
modes (and N values of k) in the discrete chain because this
chain has N degrees of freedom, but there is a infinite
number of modes (and hence, infinite number of k values) in
the continuum chain because this chain has an infinite
number of degrees of freedom. The most general solution in
the continuum case can be written as

ξ(x , t) = A 1 sin k 1 x cos (ω 1 t + α 1 ) + A 2 sin k 2 x cos (ω 2 t + α 2 ) +

+ A 3 sin k 3 x cos (ω 3 t + α 3 ) + ...

(11)

where this is now sum with an infinite number of terms.

The infinite number of normal modes in the continuum

chain correspond to an infinite number of initial conditions
(position and velocity) for the infinite number of infinitesi-
mal particles in this chain. Of course, we cannot list the
initial conditions for an infinite number of infinitely closed
particles. Instead, these conditions are given by continuous
functions themselves: ξ(x,0) = a(x) is the initial position
function, and ∂ξ/∂t(x,0) = b(x) is the initial velocity function.
The constants A and α in Eq. 11 are determined from the
requirement that ξ(x,0) be equal to the initial position

phy 241 MenŽndez 81

 Chapter 5

function a(x) and b(x) equal to the initial velocity function

∂ξ/∂t(x,0). The mathematical theory of Fourier analysis shows
that these conditions are sufficient to determine the
coefficients uniquely. However, we will not deal with this
topic in this course.

The above discussion shows that there is a complete analogy

between the discrete and continuum theory of a linear chain.
The problem of N masses is completely specified and has a
unique solution once the 2N initial conditions are deter-
mined. Similarly, the continuum problem is completely
specified and has a unique solution once the initial condi-
tions functions a(x) and b(x) are known. The mathematical
theory of Fourier analysis tells us that a solution can be
found for arbitrary functions a(x) and b(x). Physically,
however, we must be careful. If any of the functions a(x) and
b(x) change abruptly over very short distances, the theory of
Fourier analysis predicts that the coefficients in the expan-
sion of Eq. (11) will be different from zero for very large k
values. In other words, if our initial conditions functions
contain Òsharp edgesÓ, we will excite normal modes of very
short wavelength. However, the wave equation was shown to
be valid in continuum media for wavelengths that are long
compared with the separation between particles. Therefore,
we may be able to obtain an exact solution to the wave
equation, but the wave equation itself may not apply to our
problem if the normal mode we excite has a very short
wavelength.

phy 241 MenŽndez 82

 Chapter 5
000
111 111
000
111
000
000
L 111
000
111
000
000
111
111
000
000
111
111
1111111111111111
0000000000000000
000 000
111
111
000
111
000
1111111111111111
0000000000000000
111
000
111
000
111 000
111
000
111
000 000
111
111
000
λ= 2L
000
111
111
000
111
000
111
000
111
000
111 000
111
111
000
0000000000000000
000 000
111
1111111111111111
111
000
111
000
111
000
111
000
111
000 000
111
111
111
000 λ= L 000
111
000
111
Figure 3 Standing transverse waves on a string with
both ends fixed.

Longitudinal vibrations in a string

Let us consider the continuum limit of Eq. (1), Chapter 4,
for the longitudinal vibrations of a chain of masses. Using
the usual transformations, this equation can be written as
M 2ξ ξ(x + a) − ξ(x) ξ(x − a) − ξ(x)
m 2 = Ka +
Mt a a (12)

which, in analogy with the section on transverse waves

becomes
M 2 ξ Ka 2 M 2 ξ
2 =
Mt m M x2 (13)

The quantity m/a is the mass per unit length of the chain,
which we call ρ. On the other hand, the total spring constant
of a chain with N springs of individual constant K is K/N
(see homework problem). This means that the inverse spring
constant per unit length is a constant characteristic of the

phy 241 MenŽndez 83

 Chapter 5

material of which the spring is made. The inverse spring

constant per unit length is K -1 /a = 1/Ka. This can also be
written as 1/K LL, where L is the length of the chain and K L
the total spring constant of the chain. Substituting these
expressions, the wave equation for a linear chain becomes
M 2 ξ K LL M 2 ξ
2 =
Mt ρ M x2 (14)
K LL
so that the wave speed is c = .
ρ

Other boundary conditions

So far, the only boundary conditions we have considered for
the transverse or longitudinal standing waves in a chain are
the ones that require fixed ends, i.e., our chains are attached
to fixed walls. Of course, the chains could also be free at
these ends. The corresponding boundary conditions would
therefore be different. To understand the transition from
fixed end to free end, we could imagine that one of the
ÓwallsÓ is in fact a mass M. When the mass is very large, we
are in the limit of fixed ends. When the mass M is light, we
approach the limit where one of the ends is free. Let us
derive the normal modes for this problem. The system is

000
111
depicted in Fig. 4

111
000
111
000
111
000
M

Figure 4 A spring with a mass M attached to an end.

In order to derive the boundary condition at the point

where the mass M is attached, it is convenient to go back to
the discrete chain and then make the continuum approxima-
000
111
tion. Let us consider the chain depicted in Fig. 5
111
000
111
000 M

Figure 5 A discrete chain of masses and springs with a mass M

attached to an end.

phy 241 MenŽndez 84

 Chapter 5

M2ξ
The acceleration of the mass M must be equal to M t2
,
x= L

since the mass is attached to the chain. The last spring on

the chain excerts a force on the mass, so that its equation of
motion is

M 2ξ ξ(L) − ξ(L − a)
M = − K ξ(L) − ξ(L − a) = − Ka
M t2 x= L
a (15)

If we now make the usual continuum approximation for a

small, and use the relation KLL = Ka, we can write the above
boundary condition in terms of macroscopic quantities only.
We thus obtain:

M 2ξ Mξ
M = − K LL
M t2 x= L
Mx x= L (16)

This is the new boundary condition at L. At the other end, x

= 0, we still have the ÒoldÓ boundary condition

ξ(0, t) = 0 (17)

We now propose as a solution

ξ(x , t) = A sin kx cos ωt (18)

This is known to be a solution to the wave equation which

in addition satisfies automatically the boundary condition at
x = 0. When we substitute this solution into Eq. (16), we
obtain, after dropping common factors,

Mω 2 sin kL = K L L cos kL (19)

which can be rewritten, using ω = ck and c2 = KLL/ρ, as

Š m0 ‘“ 1
tan kL = ŒŒ “ , (20)
‹M ’ kL
where we have used m 0 = ρL, the total mass of the chain.
This is a trascendental equation for the wavenumber k. Its
solution must be computed numerically or graphically. The
solution is the intersection betwen the curves for tan kL and
the curve for (m0/M) 1/kL, as shown in Fig. 6.

phy 241 MenŽndez 85

 Chapter 5

1.0
0.8
tan kL
0.6
0.4
0.2 (m0/M )(1/k L)
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
π 2π 3π
kL
Figure 6 Graphical method for the determination of the
possible wavenumbers for the normal modes
of a continuous spring chain of total mass m0
with a mass M attached to one end. The figure
represents the case m0/M = 0.2

Let us now discuss the solutions to this equation. There is a

set of solutions which occur for wavenumbers close to the
condition kL = ¹i, with i = 1, 2, 3,....This is the condition
(Eq. 8) derived for a string with the two ends fixed. In fact,
for very large values of M, that is, for m0/M → 0, the condi-
tion k L = ¹i becomes almost exact. This makes sense,
because a very large mass M is equivalent to a fixed wall.
When the mass M is not infinite, the k values are slightly
larger than the predictions for a string with both ends fixed.
This means that the wavelengths of the modes is slightly
shorter.

Figure 6 shows a solution for small values of k that does not

belong to the series of solutions close to the condition kL =

phy 241 MenŽndez 86

 Chapter 5

¹i. Since the system has one extra degree of freedom relative
to the chain fixed at both ends (this degree of freedom
corresponds to the motion of mass M ), an additional
solution is to be expected. Let us study this solution. In the
limit m 0/M → 0, kL is very small, so that tan kL approaches
kL. We can thus rewrite Eq. (20) as
m0
‡ kL Ž = M
2
(21)

Using the relations k = ω/c, c 2 = K L L/ρ, and ρL = m 0 this

becomes
KL
ω2 =
M (22)

This is exactly the expression for the frequency of a simple

harmonic oscillator of mass M and spring constant K L .
Hence the meaning of the first solution in Fig. 6 is clear: in
the limit of negligible spring mass, it corresponds to the
solution we found in Chapter 2 for a mass attached to
spring. The exact solution in Figure 6 corresponds to the
case where the mass of the spring is not neglected, as we
implicitely did in Chapter 2. In one of the homework
problems you will derive a simple expression for the frequen-
cy of this mode for the case where m 0 is small but not
negligible with respect to M.

Let us now examine the opposite limit where M is much

smaller than m 0 . Physically, this corresponds to a chain
where one of the ends is free. In this case, the intersections
of the two curves occur for kL = ¹/2, 3¹/2, 5¹/2, etc., so that
the possible wavenumbers are given by

π 3π 5π
k1= , k2= , k3= , ....
2L 2L 2L (23)

The corresponding wavelengths λ = 2¹/k can thus be written

as
4L
λi = , i = 1, 2, 3, etc.
2i − 1 (24)

phy 241 MenŽndez 87

 Chapter 5

Notice that the Òfree endÓ solutions have maximum displace-

ment at x = L. (You can see this by using the wavenumbers
from Eq. (23) in Eq. (18)). Points at which the amplitude is
maximum are called a n t i n o d e s . The existence of an
antinode at x = L is consistent with Eq. (16) for M ap-
Mξ
proaching zero, which requires that the first derivative M x be
zero. In other words, there can be no force at the free end of
a chain.

Longitudinal vibrations in solids

0000000000000000
1111111111111111
A solid object, such as a crystal, can be viewed as a collection

1111111111111111
0000000000000000
of parallel linear chains of atoms

1111111
0000000
1
0
1111111111111111
0000000000000000
0000000
1
0 1111111111111111
0000000000000000
11111110000000000000000
1111111111111111
1111111
0000000
1
0
1111111
0000000
1
0 1
0
1111111111111111
0000000000000000
1
0
1111111111111111
0000000000000000
11111110000000000000000
00000001111111111111111
1
0
11111110000000000000000
000000001
Figure 7
1111111111111111
A simple model of a solid object as a collection of
linear chains of masses and springs.

The solid will probably feature transverse springs that help

keep the chains together. Moreover, in real solids the atoms
interact not only with the nearest neighbors but also with
more distant ones. This implies more springs than the ones
we show in the figure. We have omitted these springs
because the drawing would become too messy. Even worse,
the arrangement of atoms in real solids do not always follow
a neat square lattice as indicated in the figure.

Within our simplified model, the effect of having parallel

chains of atoms will increase the effective spring constant.
For example, the effective spring constant of two equal
springs acting in parallel is twice the spring constant of a
single spring. If the cross-sectional area of the bar in Fig. 7 is
A and its length is L, its effective spring constant can be
written as
YA
KL=
L (25)

phy 241 MenŽndez 88

 Chapter 5

This is because the number of chains is proportional to the

cross-sectional area and the spring constant is inversely
proportional to the length, as discussed above. The constant
Y is a property of the barÕs material, not of its shape. It is
called the Young modulus or elastic modulus. Eq. (25) is
very convenient because it separates the properties of the
material for the shape of the object. For example, it makes
no sense to talk about the spring constant of iron. On the
other hand, the Young modulus of the material iron is well
defined. Experiments show it is given by Y = 2.06 ×10 1 1
Nm -2. Using this value, we can calculate the spring constant
K L of any iron bar of length L and cross-section A by using
Eq. 25. Inserting Eq. (25) into Eq. (14) we obtain
M ξ Y M ξ
2 2

=
Mt
2
ρV M x2 (26)

where ρ V = ρ/A is simply the standard volume density.

Notice that the symbol ρ is always used for density, but its
meaning may change from situation to situation. Some-
times, we mean linear density, sometimes areal density, and
sometimes, as in the case of the solid bar, volume density. For
simplicity, the subscripts that reveal the type of density are
often dropped, and we are left to guess what type of density
we are talking about. The best hint is given by the type of
problem: if we are talking about a cord, we will neglect its
thickness, so that the density will refer to mass per unit
length. If we are talking about a bar, it makes sense to talk
about the mass per unit volume, so that this will be usually
the meaning of ρ. Another safe way to determine the type of
density we need for our problem is to look at the units of
the different factors, keeping in mind that the coefficient of
the right hand side in the wave equation (see Eq. 26) has
units of speed squared.

The wave speed for the problem in Eq. (26) is given by

Y
c=
ρV (27)

phy 241 MenŽndez 89

 Chapter 5

Later it will become apparent that this quantity is simply

the speed of sound in a solid object. Replacing values for
typical materials, we find that this number is about ten
times higher that than the speed of sound in air, which we
will determine below.

Strain-stress
Eq. (25) suggest an experiment to determine the Young
modulus Y of a given material. Consider a bar of length L
and cross -sectional area A. Suppose we attach one end of
the bar to a wall and exert a tensile force F on the other end.
The bar will stretch, according to HookeÕs law, by ∆L = F/KL.
Using Eq. (25) this can be written as F/A = Y ∆L/L. The
quantity ∆L/L, which gives the relative elongation of the bar,
is called the strain ε. The quantity F/A, which gives the force
per unit cross sectional area, is called the stress S. We
therefore arrive at the strain-stress relation

S= Yε (28)

This is sometimes known as HookeÕs law for a continuum

medium. By determining the stress and the strain, one can
compute the Young modulus. Notice that hard materials are
those whose Young modulus is high, so that their deforma-
tion for a given stress is small.

Shear deformations
Fig. 8 (a) shows the type of deformation produced when we
measure the Young modulus. As indicated by Fig. 8 (b), it is
possible to produce another type of deformation in a solid
object by applying forces tangent to the surface. This type of
stress is called shear stress.

phy 241 MenŽndez 90

 Chapter 5

∆x

00000000000000
11111111111111
11111111111111
00000000000000
F 00000000000000
11111111111111 F
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
(a) Longitudinal stress

000000000000000
111111111111111
111111111111111∆x
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
F 000000000000000
111111111111111 F
111111111111111
000000000000000
111111111111111
000000000000000
(b) Shear stress
Figure 8 Longitudinal and shear stress-strain in a solid bar.

For shear stress, we define stress and strain in a similar way

as in the case of longitudinal deformation, S = F/A, and ε =
∆x/L, where A is the cross sectional area, L the length of the
bar, and ∆ x the lateral deformation. However, there is no
reason why the restoring force for a longitudinal deforma-
tion has to be equal to the restoring force for shear deforma-
tion, so that in general we will have

S = Gε (29)

with G ≠ Y. The quantity G is called the shear modulus of

the material. When a deformation as in Fig. 8 is induced in
a bar, transverse waves are setup. They satisfy a wave equation
identical to the one we derived for longitudinal waves, but
now the speed will be given by

phy 241 MenŽndez 91

 Chapter 5

G
c=
ρV (30)

Thus the speed for transverse waves is different from the

speed for longitudinal wave. As we will see in the next
chapter, this has important implications for applied science.
An important example is seismology.

When shear stress is applied to a liquid or a gas, there is no

restoring force, because the liquid or the gas donÕt oppose
being deformed as in Fig. 8 (b). In other words, for liquids
and gases G = 0. Thus transverse waves are not possible in
these media.

Vibrations in a column of air: sound

0000000000000
1111111111111
1111111111111
0000000000000
1111111111111
0000000000000
1111111111111
0000000000000 F = pA
1111111111111
0000000000000
AIR
1111111111111
0000000000000
1111111111111
0000000000000
Figure 9 A column of air enclosed in a tube is similar to
a spring in that it resists being compressed.

A column of air enclosed in a tube, as in Fig. 9, acts like a

spring. Therefore, we can assume that the longitudinal air
displacements satisfy an equation similar to Eq. (14). Our
task is to calculate the spring constant K L of the column of
air. A fundamental difference between a gas and a standard
spring is that the spring ÒconstantÓ of the gas is actually not
a constant: as the gas is compressed, it becomes harder and
harder to compress it. This is what makes air such a good
shock absorber. If we apply a force F to a spring of constant
K L, the spring stretches by an amount ∆L = F/KL, where ∆L
could in principle be a large displacement. However, if K L is
not a constant, this expression makes no sense, because KL
will be changing as the spring is compressed. This is exactly

phy 241 MenŽndez 92

 Chapter 5

what happens with a column of air. The right expression in

this case is

dF = − K L (L) dL (31)

For air, the force is given by F = pA, where A is the area of

the piston that encloses the air and p is the pressure. Thus
dF = Adp. Since pressure changes as we compress the gas, it
is convenient to write this expression as
Š dp ‘““
dF = A ŒŒ dV
‹ dV ’0
Š dp ‘““
= A ŒŒ A dL
‹ dV ’ 0
2 Š dp ‘“
= A ŒŒ “ dL
‹ dV ’ 0 (32)

where the subscript Ò0Ó means that the derivatives are

computed at equilibrium, that is, when the pressure inside is
equal to the pressure outside, which usually is the atmo-
spheric pressure. Comparing Eq. (31) with Eq. (32), we
conclude that the spring ÒconstantÓ of the air column is
dp ‘“
given by K L = − A 2 ŠŒŒ “ . Hence the speed in the wave
‹ dV ’ 0
equation is

A 2 L 0 ‡ dV Ž0 AL 0 ‡ dV Ž0
dp dp
K LL0
c =
2
=− =− =
ρ0 ρ0 ρ0 /A
V 0 ‡ dV Ž0
dp

=−
ρ V,0 (33)

where all subscripts Ò0Ó refer to equilibrium. Here, as usual,

ρV refers to the volume density (mass per unit volume). We
must now compute the derivative dp/dV. For an ideal gas, we
will show in later chapters that p = NkT/V, is the number of
molecules in the gas, k the so-called Boltzmann constant,
and T the temperature in Kelvin. If we try to compute the
volume derivative of this expression, we must know how the
temperature depends on the volume. The first who thought
about this problem was Newton himself. He reasoned that

phy 241 MenŽndez 93

 Chapter 5

compressions and rarefactions in a gas occur so rapidly when

a wave is excited, (thousands of times per second for
standard sound waves) that the temperature remains
essentially constant. With this assumption, dp/dV = -NkT/V2.
= -p/V. Using this expression in Eq. (33), we obtain c2 =
p0/ρ0. For air at atmospheric pressure, p0 = 1.01 ×105 N m-2
and ρ0 = 1.29 kg/m3. This yields c = 280 m/s., only 15% off
the known speed of sound in air, c = 332 m/s. Not too bad.
On the other hand, what could be wrong in our derivation
to give a 15% error? The answer is in our assumption that
the temperature remains constant.

We will show in later chapters that the energy of a gas is

directly proportional to its temperature. When the pressure
does work on the gas, its energy increases. If the temperature
is to remain constant, the work done by the pressure must
be compensated by an energy loss of exactly the same
magnitude, usually in the form of heat. However (this is the
crucial argument) it takes time for heat to dissipate. For
very rapid oscillations this time is not available. Hence a
reasonable assumption is not that the temperature remains
constant but that the heat transfer is zero. Under such an
assumption, we will show in the thermodynamics part of this
course that the temperature is proportional to V γ-1, where γ
= 1.4 for air. Hence the correct relationship between
pressure and volume is pV γ = constant. With this expres-
sion, dp/dV = -γp/V and
γp 0
c2 =
ρ V,0 (34)

This gives c = 332 m/s, which agrees exactly with the

experimental value.

Standing waves in tubes

The study of standing waves in tubes filled with air is
important to understand how musical instruments work.
Depending on whether the tubes are open or closed at the
ends, the boundary conditions change and different types of
standing waves are setup. From the boundary conditions, we

phy 241 MenŽndez 94

 Chapter 5

can determine the possible values of the wavenumber k (or,

equivalently, the wavelength λ). Using ω=ck, (or c=λν) with c
= 332 /s, we readily obtain the frequencies of vibration.

The simplest case is that of closed tubes. Here the boundary

conditions are quite obvious: since the tube is closed, there
must be nodes at the two ends, that is, the air displacement
is zero at these points. This is schematically illustrated in
Fig. 10. Notice that the figure suggests that the air displace-
ment is perpendicular to the length of the tube. This is not
true! The actual air displacement is longitudinal. Since we
cannot indicate in the figure a longitudinal wave, we have
drawn a transverse one. The figure illustrates the amplitude

00000000000000
11111111111111
of the oscillation, but is not a picture of the actual displace-

11111111111111
00000000000000
ments, which occur along the axis of the tube.

11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
Figure 10 The first two standing waves in a closed tube filled
with air.

From Fig. 10 it is easy to conclude that the possible wave-

lengths in a closed tube are given by λ = 2L/i, where i i s
1,2,3, etc. Hence the frequency of the wave is
Š c ‘““
ν i = i ŒŒ (35)
‹ 2L ’

phy 241 MenŽndez 95

 Chapter 5

Let us now consider the case where one of the ends is open.
This is illustrated in Fig. 11. The boundary condition for the
closed end remains the same: there is a node at this point.
At the open end, the situation is more complicated. We can
understand the boundary condition at this end by analogy
with the case of a spring with one end fixed and a mass M
attached to the other end. If there was no air outside the
tube, this would correspond to the case M = 0. Therefore,
we would obtain antinodes at the open end of the tube. Of
course, there must be air outside, so that this condition is
not fulfilled exactly. On the other hand, while the air inside
the tube will ÒresistÓ any compression, the air outside the
tube will offer very little resistance to compression. So we
could ÒmodelÓ an open tube by a spring with large spring
constant (air inside) attached to a spring of very low spring
constant (air outside). The solutions to this problem are
very similar to the solutions for the spring with a very small
mass M at one end, because they essentially reduce to the
condition that no force is applied to the system at the open
end. Hence there are antinodes very close to the open end
of the tube. For the solution of homework problems, we
usually assume that these antinodes are exactly at x = L.
Hence we can use Eqs. (23) and (24) to determine the
wavenumbers and wavelengths for the possible standing
waves. This leads to frequencies given by
c
ν i = (2i − 1) = ν 1 , 3ν 1 , 5ν 1 , ...
4L (36)

where the fundamental frequency is given by ν1 = c/4L. Notice

that only the odd harmonics of the fundamental are present.
From Fig. 6, We notice that the exact solution features
wavenumbers slightly smaller than the wavenumbers ob-
tained from the condition that an antinode exists exactly at
x = L. This means that in real tubes the antinode is slightly
outside the tube.

phy 241 MenŽndez 96

 00000000000000000
11111111111111111 Chapter 5
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
Figure 11 The lowest two vibrational modes in tubes
open at one end.

The case of a tube open at both ends is straightforward from

the discussion in the previous paragraphs and is the subject
of one of the homework problems.

Importance of the boundary conditions

The frequencies of oscillations of a system depend on the
boundary conditions, as can be seen by comparing the
possible frequencies for closed and open-end tubes. This
statement is valid for all vibrating systems and is not
restricted to tubes. Hence the possible frequencies of sound
waves inside a room will change if we open a door. However,
we typically donÕt notice much of a change in the sounds
heard inside a room when a door is opened. We do notice
changes, on the other hand, when we sing while taking a
shower. Those of us who are lousy singers notice a definite
improvement in the quality of the sounds we emit. The
expalanations for these effects is the following: The
separation between a certain mode and the next mode is
proportional to 1/L. When the size of the vibrating system is
much larger than the wavelengths of the modes, the modes
are very closely spaced, so that the allowed frequencies do

phy 241 MenŽndez 97

 Chapter 5

not change much if we change the boundary conditions. This

makes sense, because in a very large system the boundaries
should eventually become unimportant. On the other hand,
the frequencies will change in a noticeable way if the
dimensions of the system are of the order of the
wavelengths. For a sound wave which has a frequency of 300
Hz, the wavelegth is of the order of 1 m. This is much
smaller than the size of a typical living-room but is of the
order of the distance between the wall and the shower
curtain.
When we discuss sound in open places, we can imagine that
we are dealing with volumes of air enclosed in very large
volumes. Thus the modes are extremely close to each other.
In other words, all frequencies are possible outdoors.

phy 241 MenŽndez 98

 Chapter 5

4. (Crawford 2.21) Find the mode configurations

Problems and frequencies for transverse oscillations of a
beaded string of 5 beads with one end fixed and one
free. Plot the five corresponding points on the
1. Consider a chain of 100 equal masses connected
dispersion relation ω(k).
by equal springs and attached to rigid walls at both
ends.
5. In the problem of a mass M attached to a spring
a) Using the procedure developed in Chapter 4, find
of mass m0 and constant K L, find the frequency and
the frequency of the first and last normal mode of
plot ξ(x,t) as a function of x for the lowest frequency
the system. In both cases, compute the ratio x59/x60
mode. Take m 0 /M = 0.1 and select the other
of the displacements of mass 59 and mass 60.
parameters arbitrarily. Compare the graph of ξ(x,t)
b) Repeat a) by using not the exact solution but the
with the deformation of the string when it is
continuum approximation developed in this
stretched statically.
chapter. Compare with a) and discuss.

6. A way to estimate a correction to the frequency

2. At time t = 0, a string of length L = 1 m is
of oscillation of a mass M attached to a spring due
deformed with the shape of a square pulse between a
to the mass m 0 of the spring is as follows: In Eq.

000
111 000
111
= 0.4 m and b = 0.6 m. The height of the pulse is h
(20), expand the tangent to terms of order (kL)3. In

111
000 111
000
= 0.05 m.
the last term of the expansion, substitute (kL) 2 =

111
000 111
000
h m0/M. Finally, show that the frequency of the lowest

111
000 111
000
mode of the system becomes

111
000 111
000
KL
ω2 = .

111
000 111
000
M+
1
a b 3 m0
Discuss the correctness of the approach used to
derive this expression.
The initial velocity of the string is zero at all points.
a) Verify (graph when needed) that the initial 7. (Alonso 28.13) A string of length 2 m and mass 4
conditions are satisfied by Eq. (11) if one chooses × 10 -3 kg is held horizontally, with one end fixed.
αm = 0 for all m, The other end passes over a pulley and supports a
A m = 0, mass of 2 kg. Calculate the velocity of transverse
for m even, and waves in the string.
2h
Am= cos mk 1 a − cos mk 1 b 8. (Alonso 28.15) A steel (Y = 2.0 × 1011 N m-2 )
mπ
wire having a length of 2 m and a radius of 5 × 10-4
for m odd. Eq. (11) is an infinite series. Of course, m hangs from the ceiling. a) If a body having a mass
you cannot add all terms in an infinite series. of 100 kg is hung from the free end, calculate the
However, you should get a good approximate elongation of the wire. b) Also determine the
answer if you take the first 20 terms or so in the displacement and the downward pull at a point in
expansion. the middle of the wire. c) Determine the velocity of
b) Graph the function ξ(x,t) at several times t > 0. longitudinal and transverse waves along the wire
Invent your own values for the speed c. when the mass is attached.

3. Show that the effective spring constant of a chain 9 . (Alonso 34.16) How is the fundamental
of N equal springs of spring constant K is K/N. frequency of a string changed if a) its tension is
Show that if you put the N springs in parallel, the doubled, b) its mass per unit length is doubled, c) its
effective spring constant becomes NK. radius is doubled, d) its length is doubled? Repeat

phy 241 MenŽndez 99

 Chapter 5

the problem if the quantities listed are halved.

10. Derive an expression for the frequencies of

oscillations in a tube with both ends open.

11. (Alonso 34.17) A tube whose length is 0.6 m is

a) open at both ends, and b) closed at one end and
open at the other. Find its fundamental frequency
and the first overtone. Plot the amplitude
distribution along the tube corresponding to the
fundamental frequency and the first overtone for
each case.

12. Compare two pipes of identical length, one of

them closed at both ends and the other one open at
one end. Find the length of the pipes so that the
fundamental frequencies differ by 1 Hz. Would you
be able to tell the difference between the sound
waves produced by the two tubes? What if the
length L is reduced by a factor of 100?

phy 241 MenŽndez 100