Chapter 4

Non-Relativistic Strings

A full appreciation for the subtleties of relativistic strings requires an un-

derstanding of the basic physics of non-relativistic strings. These strings
have mass, tension, and can vibrate both transversely and longitudinally. We
study the equations of motion for non-relativistic strings and develop the La-
grangian approach to their dynamics.

4.1 Equations of motion for transverse oscil-

lations
We will begin our study of strings with a look at the transverse fluctuations
of a stretched string. The direction along the string is called the longitudinal
direction, and the directions orthogonal to the string are called the transverse
directions. We consider, for notational simplicity, the case when there is
only one transverse direction – the generalization to additional transverse
directions is straightforward.
Working in the (x, y) plane, let the classical non-relativistic string have
its endpoints fixed at (0, 0), and (a, 0). In the static configuration, the string
is stretched along the x-axis between these two points. In a transverse oscil-
lation, the x coordinate of any point on the string does not change in time.
The transverse displacement of a point is given by its y-coordinate. The x
direction is longitudinal, and the y direction is transverse. To describe the
classical mechanics of a homogeneous string, we need two pieces of informa-
tion: the tension T0 and the mass per unit length µ0 . The total mass of the
string is then M = µ0 a.

87
88 CHAPTER 4. NON-RELATIVISTIC STRINGS

Let us look briefly at the units. Tension has units of force, so

[Energy]
[T0 ] = [Force] = . (4.1.1)
L
If you stretch a string an infinitesimal amount dx, its tension remains ap-
proximately constant through the stretching, and the change in energy equals
the work done T0 dx. The total mass of the string does not change. If we
were considering relativistic strings, however, a static string with more en-
ergy would have a larger rest mass. Using (4.1.1), noting that energy has
units of mass times velocity squared, and that µ0 has units of mass per unit
length, we have
M 2
[T0 ] = [v] = [µ0 ][v]2 . (4.1.2)
L
For a non-relativistic string, both T0 and µ0 are adjustable parameters, and
the velocity on the right-hand side above will turn out to be the velocity
of transverse waves. The above equation suggests that the string tension
T0 and the linear mass density µ0 in a relativistic string might be related
by T0 = µ0 c2 , since c is the canonical velocity in relativity. We will see in
Chapter 6 that this is indeed the correct relation for a relativistic string.

Figure 4.1: A short piece of a classical non-relativistic string vibrating transver-

sally. With different slopes at the two endpoints there is a net vertical force.

Returning to our classical non-relativistic string, let us figure out the

equation of motion. Consider a small portion of the string, extending from
x to x + dx, with y = 0, when the string is static. This piece is shown in
4.1. EQUATIONS OF MOTION FOR TRANSVERSE OSCILLATIONS89

transverse oscillation in Figure 4.1. At time t, the transverse displacement

of the string is y(t, x) at x and y(t, x + dx) at x + dx. We will assume that
the string oscillations are small, and by this we will mean that at all times
∂y
<< 1 , (4.1.3)
∂x
at any point on the string. This guarantees that the transverse displacement
of the string is small compared to the length of the string. The length of the
string changes little, and we can assume that the tension T0 is unchanged.
The slope of the string is a bit different at the points x and x + dx.
This change of slope means that the string tension changes direction and
the portion of string under consideration feels a net force. For transverse
oscillations we need only calculate the net vertical force – you may show in
Problem 4.1 that the horizontal net force is negligible. The vertical force at
(x + dx, y + dy) is accurately given by T0 times ∂y/∂x evaluated at x + dx
and is pointing up; similarly, the vertical force at (x, y) is T0 times ∂y/∂x
evaluated at x and is pointing down. Therefore the net vertical force dFv is
∂y

∂y

∂2y
dFv = T0
− T0

T0 2 dx . (4.1.4)
∂x x+dx ∂x x ∂x
The mass dm of this piece of string, originally stretched from x to x + dx,
is given by the mass density µ0 times dx. By Newton’s law, the net vertical
force equals mass times vertical acceleration. So we can simply write
∂2y ∂2y
T0 dx = (µ 0 dx) . (4.1.5)
∂x2 ∂t2
We cancel dx on each side and rearrange terms to get
∂ 2 y µ0 ∂ 2 y
− = 0. (4.1.6)
∂x2 T0 ∂t2
This is just a wave equation! Recall that for the wave equation
∂2y 1 ∂2y
− = 0, (4.1.7)
∂x2 v02 ∂t2
the parameter v0 is the velocity of the waves. Thus for the transverse waves
on our stretched string, the velocity v0 of the waves is

v0 = T0 /µ0 . (4.1.8)
The higher the tension, or the lighter the string, the faster the waves move.
90 CHAPTER 4. NON-RELATIVISTIC STRINGS

4.2 Boundary conditions and initial conditions

Since equation (4.1.6) is a partial differential equation involving space and
time derivatives, in order to obtain solutions we must in general apply both
boundary conditions and initial conditions. Boundary conditions (B.C.) con-
strain the solution at the boundary of the system, and initial conditions
constrain the solution at a fixed starting time. The most common types of
boundary conditions are Dirichlet and Neumann boundary conditions.
For our string, Dirichlet boundary conditions specify the positions of the
string endpoints. For example, if we attach each end of the string to a wall
(see Figure 4.2, left side), we are imposing the Dirichlet boundary conditions
y(t, x = 0) = y(t, x = a) = 0 , Dirichlet B.C. (4.2.1)

Figure 4.2: Left side: string with Dirichlet boundary conditions at the endpoints.
Right side: string with Neumann boundary conditions at the endpoints.

Alternatively, if we attach a massless loop to each end of the string and

the loops are allowed to slide along two frictionless poles, we are imposing
Neumann boundary conditions. For our string, Neumann boundary condi-
tions specify the values of the derivative ∂y/∂x at the endpoints. Since the
loops are massless and the poles are frictionless, the derivative ∂y/∂x must
vanish at the poles x = 0, a (see Figure 4.2, right side). If this were not the
case, then the slope of the string at a pole would be nonzero, and a com-
ponent of the string tension would accelerate the rings in the y-direction.
Since each ring is massless, their acceleration would be infinite. This is not
possible, so in effect, we are imposing the Neumann conditions
∂y ∂y
(t, x = 0) = (t, x = a) = 0 , Neumann B.C. (4.2.2)
∂x ∂x
4.3. FREQUENCIES OF TRANSVERSE OSCILLATION 91

Let’ s see how we can solve the wave equation for a particular set of initial
conditions. The general solution of equation (4.1.6) is of the form

y(t, x) = h+ (x − v0 t) + h− (x + v0 t) , (4.2.3)

where h+ and h− are arbitrary functions. This represents a superposition of

two waves, h+ moving to the right and h− moving to the left. Suppose the
initial values of y and ∂y/∂t are known at time t = 0. Using equation (4.2.3)
we see that this information yields the equations

y(0, x) = h+ (x) + h− (x) , (4.2.4)

∂y
(0, x) = −v0 h
+ (x) + v0 h
− (x) , (4.2.5)
∂t
where the left-hand sides are known functions, and primes denote derivatives
with respect to arguments. Using (4.2.4) we can solve for h− in terms of h+ .
Substituting into (4.2.5), we get a first-order ordinary differential equation
for h+ . Once we have solved for h+ , using appropriate boundary conditions,
we can use (4.2.4) again, this time to find the explicit form of h− . With h+
and h− known, the full solution of the equations of motion is given by (4.2.3).

4.3 Frequencies of transverse oscillation

Suppose we have a string where each point is oscillating in the y-direction
sinusoidally and in phase. This means that y(t, x) is of the form

y(t, x) = y(x) sin(ωt + φ) , (4.3.1)

where ω is the angular frequency of oscillation and φ is the constant common

is phase. Our aim is to find the allowed frequencies of oscillation. Substitut-
ing (4.3.1) into (4.1.6) we find, after cancelling the common time dependence

d2 y(x) µ0
2
+ ω2 y(x) = 0 . (4.3.2)
dx T0
This is an ordinary second-order differential equation for the profile y(x)
of the oscillations. The allowed frequencies are selected by this equation,
together with the boundary conditions. Since ω, µ0 and T0 are constants,
92 CHAPTER 4. NON-RELATIVISTIC STRINGS

the differential equation is solved in terms of trigonometric functions. With

Dirichlet boundary conditions (4.2.1) we have the nontrivial solutions
 nπx 
yn (x) = An sin , n = 1, 2, . . . , (4.3.3)
a
where An is an arbitrary constant. The value n = 0 is not included above
because it represents a motionless string. Plugging yn (x) into (4.3.2), we find
the allowed frequencies ωn :

T0  nπ 
ωn = , n = 1, 2, · · · . (4.3.4)
µ0 a

These are the frequencies of oscillation for a Dirichlet string. The strings on
a violin are Dirichlet strings. To tune a violin to the right frequency one must
adjust the string tension. The higher the tension, the higher is the pitch, as
predicted by (4.3.4).
For the case of Neumann boundary conditions (4.2.2), we obtain the
spatial solutions
 nπx 
yn (x) = An cos n = 1, 2, . . . . (4.3.5)
a
This time the n = 0 case is a little less trivial: the string does not oscillate,
but it becomes rigidly translated to y(t, x) = A0 . The oscillation frequencies,
found by plugging (4.3.5) in (4.3.2), are the same as those in (4.3.4). There-
fore, the oscillation frequencies are the same in the Neumann and Dirichlet
problems. The Neumann case, however, admits one extra solution not in-
cluded in our oscillatory ansatz (4.3.1): the string can translate with constant
velocity. Indeed, y(t, x) = at + b, with a and b arbitrary constants, satisfies
both the boundary conditions and the original wave equation (4.1.7).

4.4 More general oscillating strings

Let us discuss briefly problems closely related to the ones considered thus
far. For example, we can take the mass density of the string to be a function
of position µ(x). The form (4.1.6) of the wave equation will not change, since
it is derived from local considerations: the examination of a little piece of
4.5. A BRIEF REVIEW OF LAGRANGIAN MECHANICS 93

string that can be chosen to be sufficiently small so that the mass density is
approximately constant. We therefore get:
∂ 2 y µ(x) ∂ 2 y
− = 0. (4.4.1)
∂x2 T0 ∂t2
For normal oscillations, we use the ansatz in (4.3.1) and find

d2 y µ(x) 2
+ ω y(x) = 0 . (4.4.2)
dx2 T0
This equation is no longer simple to solve, and it can only be studied in detail
once the function µ(x) is specified. In Problems 4.4 and 4.5 you will consider
some specific mass distributions, and you will explore a variational approach
that gives an upper bound for the lowest oscillation frequency.

So far we have only considered strings that are oscillating transversally.

Strings also admit longitudinal oscillations, although the relativistic string
does not. Imagine a string which at equilibrium lies along the x-axis and
consider the infinitesimal segment which at in equilibrium, extends from x to
x + dx. Suppose now that at time t the ends of this infinitesimal segment are
displaced from their equilibrium positions by distances η(t, x) and η(t, x+dx),
respectively. If these two quantities are not the same, the piece of string
is being compressed or stretched. An equation of motion can be obtained
for this system, much as we did for transverse motion. It is not possible,
however, to assume that the tension T is constant throughout the string.
For transverse oscillations the net force acting on a little piece of string arose
from the different angles at which the tension was applied on opposite ends
of the piece. If the string always lies along the x-axis then a net force can
act on a segment only if the tension is different on its two ends. Therefore
the waves of an oscillating string are accompanied by tension waves! It is an
instructive exercise (Problem 4.2) to work out the equations of motion for a
longitudinally-oscillating string.

4.5 A brief review of Lagrangian mechanics

The Lagrangian L of a system is defined by

L=T −V , (4.5.1)
94 CHAPTER 4. NON-RELATIVISTIC STRINGS

where T is the kinetic energy of the system and V is the potential energy
of the system. For a point particle of mass m moving along the x axis
under the influence of a time-independent potential V (x), the non-relativistic
Lagrangian takes the form

1 dx(t)
L(t) = m [ ẋ(t)]2 − V [x(t)] , ẋ(t) ≡ . (4.5.2)
2 dt
We must emphasize that the above Lagrangian is implicitly a function of
time, but it has no explicit time dependence. All the time dependence arises
from the time dependence of the position x(t). The action S is defined as

S= L(t)dt , (4.5.3)
P

where P is a path x(t) between an initial position xi at an initial time ti ,

and a final position xf at a final time tf > ti . One such path is shown in
Figure 4.3.

Figure 4.3: A trajectory P representing possible one-dimensional motion of a

particle in the time interval [ti , tf ].

The action is a functional. Whereas a function of a single variable takes

one number – the argument – as input and gives another number as output,
a functional takes a function as the input, and gives a number as output.
Since a function is usually defined by an infinitely-many points, we can think
of a functional as a function of infinitely-many variables. In our present
4.5. A BRIEF REVIEW OF LAGRANGIAN MECHANICS 95

application, the input for the action functional is the function x(t) which
determines the path P. We can emphasize the argument of S by using the
notation S[x]. Here [x] represents the full function x(t). It is potentially
confusing to write S[x(t)], since it suggests that S is ultimately a function of
t, which it is not.
More explicitly, for any path x(t), the action is given by
 tf  
1 2
S[x] = m [ẋ(t)] − V [x(t)] dt . (4.5.4)
ti 2
It is very important to emphasize that the action S can be calculated for any
path x(t) and not only for paths that represent physically-realized motion.
It is because S can be calculated for all paths that it is a very powerful tool
to find the paths that can be physically realized.

Figure 4.4: A path x(t) and its variation x(t)+δx(t). This variation δx(t) vanishes
at t = ti and at t = tf .

Hamilton’s principle states that the path P which a system actually takes
is one for which the action S is stationary. More precisely, if this path P
is varied infinitesimally, the action does not change to first order in the
variation. In terms of the function x(t) specifying the path, the perturbed
path takes the form x(t) + δx(t), as shown in Figure 4.4. For any time t,
the variation δx(t) is the vertical distance between the original path and the
varied path. As in the figure, we consider variations where the initial and
final positions xi = x(ti ) and xf = x(tf ) are unchanged:
δx(ti ) = δx(tf ) = 0 . (4.5.5)
96 CHAPTER 4. NON-RELATIVISTIC STRINGS

We now calculate the action S[x + δx] for the perturbed path x(t) + δx(t):
 tf 
2 
m d
S[x + δx] = (x(t) + δx(t)) − V [x(t) + δx(t)] dt (4.5.6)
ti 2 dt
 tf  
d

= S[x] + mẋ(t) δx(t) − V [x(t)]δx(t) dt + O((δx)2 ) ,
ti dt

where in the last equation we have expanded V in a Taylor series about x(t).
The terms of order (δx)2 and higher are unnecessary to determine whether
the action is stationary. We have thus left them undetermined and only
indicate them by O((δx)2 ). We can write the new action as S + δS, where
δS is linear in δx. From the equation above we see that δS is given by
 tf  
d

δS = mẋ(t) δx(t) − V [x(t)]δx(t) dt . (4.5.7)
ti dt
To find the equations of motion, the variation δS must be rewritten in the
form δS ∼ δx{. . .}. In particular, no derivatives must be acting on δx.
This can be achieved using integration by parts:
 tf  
d

δS = [mẋ(t)δx(t)] − mẍ(t)δx(t) − V [x(t)]δx(t) dt
ti dt
 tf
= mẋ(tf )δx(tf ) − mẋ(ti )δx(ti ) + δx(t) {−mẍ(t) − V
[x(t)]} dt .
ti
(4.5.8)

Making use of (4.5.5), the variation reduces to

 tf
δS = δx(t) {−mẍ(t) − V
[x(t)]} dt . (4.5.9)
ti

The action is stationary if δS vanishes for every variation δx(t). For this to
happen, the factor multiplying δx(t) in the integrand must vanish:

mẍ(t) = −V
[x(t)] . (4.5.10)

This is Newton’s second law of motion for a particle in a potential V (x). We

have recovered the expected equation of motion by requiring that the action
be stationary.
4.6. THE NON-RELATIVISTIC STRING LAGRANGIAN 97

Suppose we have determined the path that the particle takes while going
from xi to xf . As we have seen, the action is then stationary under variations
that vanish at the initial and final times. Is the action also stationary under
variations that change the initial position at ti or the final position at tf ?
In general, the answer is no. This can be seen from equation (4.5.8). The
integral term vanishes by assumption, but if δx(tf ) = 0, the first term in the
right-hand side would not vanish unless mẋ(tf ), the final momentum of the
particle, happens to vanish. The situation is analogous for δx(ti ) = 0.

4.6 The non-relativistic string Lagrangian

Let’s return now to our string with constant mass density µ0 , constant tension
T0 , and ends located at x = 0 and x = a. The kinetic energy is simply the
sum of the kinetic energies of all the infinitesimal segments that comprise the
string. So it can be written as
 a  2
1 ∂y
T = (µ0 dx) . (4.6.1)
0 2 ∂t
The potential energy arises from the work which must be done to stretch
the segments. Consider an infinitesimal portion of string which extends from
(x, 0) to (x + dx, 0) when the string is in equilibrium. If the string element is
momentarily stretched from (x, y) to (x + dx, y + dy), as in Figure 4.1, then
the change in length ∆l of the infinitesimal segment is given by

∆l = (dx)2 + (dy)2 − dx
  ∂y 2   2
1 ∂y
= dx 1+ − 1
dx , (4.6.2)
∂x 2 ∂x
where we have used the small oscillation approximation (4.1.3) to discard
higher order terms in the expansion of the square root. Since the work done
in stretching each infinitesimal unit is T0 ∆l, we have
 a  2
1 ∂y
V = T0 dx . (4.6.3)
0 2 ∂x
The Lagrangian for the string is given by T − V :
 a  a
1  ∂y 2 1  ∂y 2

L(t) = µ0 − T0 dx ≡ L dx , (4.6.4)
0 2 ∂t 2 ∂x 0
98 CHAPTER 4. NON-RELATIVISTIC STRINGS

where L is referred to as the Lagrangian density:

 ∂y ∂y  1  ∂y 2 1  ∂y 2
L , = µ0 − T0 . (4.6.5)
∂t ∂x 2 ∂t 2 ∂x

The action for our string is therefore

    
tf tf a
1  ∂y 2 1  ∂y 2
S= L(t)dt = dt dx µ0 − T0 . (4.6.6)
ti ti 0 2 ∂t 2 ∂x

In this action the “path” is the function y(t, x). To find the equations of
motion, we must examine the variation of the action as we vary: y(t, x) →
y(t, x) + δy(t, x). Performing the variation as before, we get
 tf  a  
∂y ∂(δy) ∂y ∂(δy)
δS = dt dx µ0 − T0 . (4.6.7)
ti 0 ∂t ∂t ∂x ∂x

We must have no derivatives acting on the variations, so we rewrite each of

the two terms above as a total derivative minus a term in which the derivative
does not act on the variation:
 tf  a 
∂ ∂y ∂2y
δS = dt dx µ0 δy − µ0 2 δy
ti 0 ∂t ∂t ∂t

∂ ∂y ∂2y

−T0 δy + T0 2 δy . (4.6.8)
∂x ∂x ∂x

The total derivatives can be integrated. The total time derivative on the first
line reduces to evaluations at tf and ti , while the total space derivative on
the second line gives evaluations at the string endpoints:
 a  t=tf  tf  x=a
∂y ∂y
δS = µ0 δy dx − T0 δy dt
0 ∂t t=ti ti ∂x x=0
 tf  a 
∂2y ∂2y
− dt dx µ0 2 − T0 2 δy . (4.6.9)
ti 0 ∂t ∂ x

Our final expression for δS contains three terms. Each one must vanish inde-
pendently. The third term, for example, is determined by the motion of the
string for x ∈ (0, a) and t ∈ (ti , tf ). The boundary conditions do not restrict
δy(t, x) here, so we set to zero the coefficient of δy, and recover our original
4.6. THE NON-RELATIVISTIC STRING LAGRANGIAN 99

equation (4.1.6). The first term in (4.6.9) is determined by the configuration

of the string at times ti and tf . If we specify these configurations, we are
in effect setting δy(ti , x) and δy(tf , x) to zero. This causes the first term
to vanish. We encountered an analogous situation in our study of the free
particle.
The second term in (4.6.9) is new: it concerns the motion of the string
endpoints y(t, 0) and y(t, a). We can make this term vanish by specifying
either Dirichlet or Neumann boundary conditions. Suppose we impose the
Dirichlet boundary conditions (4.2.1). Then the positions of our endpoints
are fixed throughout time, so we require that the variation δy vanishes for
x = 0 and x = a. This will cause the second term to vanish. If, on the
other hand, we impose the Neumann boundary conditions (4.2.2), then we
are setting

∂y ∂y
(t, x = 0) = (t, x = a) = 0 , Neumann B.C. (4.6.10)
∂x ∂x

This will also cause the second term to vanish. Dirichlet boundary conditions
can be written in a form where the similarity to Neumann boundary condi-
tions is more apparent. If the string endpoints are fixed, the time derivatives
of the endpoint coordinates must vanish

∂y ∂y
(t, x = 0) = (t, x = a) = 0 , Dirichlet B.C. (4.6.11)
∂t ∂t

The similarity with (4.6.10) is quite striking. The only change is that spatial
derivatives were turned into time derivatives. If we write Dirichlet boundary
conditions in this form, we must still specify the values of the coordinates of
the fixed endpoints.

In order to appreciate further the physical import of boundary condi-

tions, we consider the momentum py carried by the string. There is no other
component to the momentum, because we have assumed that the motion
is restricted to the y-direction. This momentum is simply the sum of the
momenta of each infinitesimal segment along the string:
 a
∂y
py = µ0 dx . (4.6.12)
0 ∂t
100 CHAPTER 4. NON-RELATIVISTIC STRINGS

Let us see if this momentum is conserved:

 a  a  x=a
dpy (t) ∂2y ∂2y ∂y
= µ0 2 dx = T0 2 dx = T0 , (4.6.13)
dt 0 ∂t 0 ∂x ∂x x=0
where we used the wave equation (4.1.6). We see that momentum is con-
served for Neumann boundary conditions (4.6.10), but for Dirichlet boundary
conditions momentum is not conserved! Does this make sense? Certainly.
When the endpoints of a string are attached to a wall, the wall is constantly
exerting a force on the endpoints of the string in order that they remain
fixed. For example, in the lowest normal mode of a Dirichlet string the net
momentum constantly oscillates between the +y- and −y-directions.
Why is this important for string theory? For many years string theorists
did not take the possibility of Dirichlet boundary conditions all that seriously.
It seemed unphysical that the string momentum could fail to be conserved.
Moreover, to what could the endpoints of open strings be attached to? The
answer is that they are attached to D-branes – a new kind of dynamical ex-
tended object. If a string is attached to a D-brane then momentum can be
conserved – the momentum lost by the string is absorbed by the D-brane.
A detailed analysis of the spatial boundary term is crucial to recognize the
possibility of D-branes in string theory.

We conclude with a more general derivation of the equation of motion for

the string. For this we write the action as
 tf  a  ∂y ∂y 
S= dt dx L , , (4.6.14)
ti 0 ∂t ∂x
where we are using equation (4.6.5). We also define the quantities
∂L ∂L
Pt ≡ , Px ≡ , (4.6.15)
∂ ẏ ∂y

with y
= ∂y/∂x. These are simply the derivatives of L with respect to its
first and second arguments, respectively. Explicitly, they are
∂y ∂y
P t = µ0 , P x = −T0 . (4.6.16)
∂t ∂x
When we vary the motion by δy, the variation of the action is given by
 tf  a  tf  a
∂L ∂L

δS = dt dx δ ẏ +
δy = dt dx P δ ẏ +P δy . (4.6.17)
t x
ti 0 ∂ ẏ ∂y ti 0
4.6. THE NON-RELATIVISTIC STRING LAGRANGIAN 101

Using the standard manipulations we find

 tf  a  ∂P t ∂P x 

∂ t ∂
δS = dt dx (P δy) + (P δy) − δy
x
+ . (4.6.18)
ti 0 ∂t ∂x ∂t ∂x

This gives the equation of motion

∂P t ∂P x
+ = 0, (4.6.19)
∂t ∂x
Using (4.6.16), one can see that this is the wave equation (4.1.6).
Note that P t , as given in (4.6.16), coincides with the momentum density
used before in equation (4.6.12). This is not a coincidence. In Lagrangian
mechanics, the derivative of the Lagrangian with respect to a velocity is
the conjugate momentum. For the string, ẏ plays the role of a velocity, and
therefore P t , a derivative of the Lagrangian density with respect to a velocity,
is a momentum density.
In addition, note that for free string endpoints, the vanishing of the varia-
tion δS requires that P x = 0. As we can see from (4.6.16), this is a Neumann
boundary condition. Furthermore, P t vanishes at the string endpoints when
impose a Dirichlet boundary condition ((4.6.11)). A more detailed analysis
of these ideas will be given in Chapter 8, where P t and P x will be shown to
have an interesting two-dimensional interpretation.
102 CHAPTER 4. NON-RELATIVISTIC STRINGS

Problems
Problem 4.1. Consistency of small transverse oscillations.

Reconsider the analysis of transverse oscillations in section 4.1. Calculate

the horizontal force dFh on the little piece of string shown in Figure 4.1. Show
that for small oscillations this force is much smaller than the vertical force
dFv responsible for the transverse oscillations.

Problem 4.2. Longitudinal waves on strings.

Consider a string with uniform mass density µ0 stretched between x = 0

and x = a. Let the equilibrium tension be T0 . Longitudinal waves are only
possible if the tension of the string varies as it stretches or compresses. Given
a piece of this string with length L and tension T0 , under a small stretching
∆L, the tension changes by ∆T where
1 1 ∆L
=
τ0 L ∆T
Find the equation governing the (small) longitudinal oscillations of this string.
Give the velocity of the waves.

Problem 4.3. Evolving an initial string configuration


A string with tension T0 , mass density µ0 , and wave velocity v0 = T0 /µ0 ,
is stretched from (x, y) = (0, 0) to (x, y) = (a, 0). The string endpoints are
fixed, and the string can vibrate in the y direction.

(a) Show that the above Dirichlet boundary conditions imply that in the
notation of equation (4.2.3)

h+ (u) = −h− (−u) , and, h+ (u) = h+ (u + 2a) . (1)

Now consider an initial value problem for this string. At t = 0 the transverse
displacement is identically zero, and the velocity is
∂y x x
(0, x) = v0 1− , x ∈ (0, a) . (2)
∂t a a
(b) Calculate h+ (u) for u ∈ (−a, a). Does this define h+ (u) for all x?
Problems for Chapter 4 103

(c) Calculate y(t, x) for x and v0 t in the domain

D = {(x, v0 t)
0 ≤ x ± v0 t < a}

Show D in a plane with axes x and v0 t.

(d) At t = 0 the midpoint x = a/2 has the largest velocity of all points in
the string. Show that the velocity of the midpoint reaches the value
of zero at time t0 = a/(2v0 ) and that y(t0 , a/2) = a/12. This is the
maximum vertical displacement of the string.

Problem 4.4. A configuration with two joined strings.

A string with tension T0 is stretched from x = 0 to x = 2a. The part

of the string x ∈ (0, a) has mass density µ1 , and the part of the string
x ∈ (a, 2a) has mass density µ2 . Consider the differential equation (4.4.2)
that determines the normal oscillations. What boundary conditions should
dy
be imposed on y(x) and dx (x) at x = a? Write the conditions that determine
the oscillation frequencies. Calculate the lowest frequency of oscillation of
this string when µ1 = µ0 and µ2 = 2µ0 .

Problem 4.5. Variational problem for strings.

Consider a string stretched from x = 0 to x = a, with a tension T0

and a position-dependent mass density µ(x). Equation (4.4.2) determines
the transverse oscillation frequencies ωi and associated profiles ψi (x) for this
string.

(a) Set up a variational procedure that gives an upper bound on the lowest
frequency of oscillation ω0 . (This can be done as in quantum mechanics,
where the ground state energy E0 of a system with Hamiltonian H
satisfies E0 ≤ ψ,Hψ
ψ,ψ
). A useful first step may be to define an inner
product · , · such that ψi , ψj vanishes when ωi = ωj . Explain why
your variational procedure works.

(b) Consider the case µ(x) = µ0 xa . Use your variational principle to find
a simple bound on the lowest oscillation frequency. Compare with the
answer ω02 = (18.956) µT0 a0 2 obtained by a direct numerical solution of
the eigenvalue problem.