This action might not be possible to undo. Are you sure you want to continue?

The Problem

Let u(x, t) denote the vertical displacement of a string from the x axis at position x and time t. The

string has length ℓ. Its left and right hand ends are held ﬁxed at height zero and we are told its initial

conﬁguration and speed. For notational convenience, choose a coordinate system so that the left hand end

of the string is at x = 0 and the right hand end of the string is at x = ℓ.

x

u

ℓ

We assume that the string is undergoing small amplitude transverse vibrations so that u(x, t) obeys the wave

equation

∂

2

u

∂ t

2

(x, t) = c

2 ∂

2

u

∂x

2

(x, t) for all 0 < x < ℓ and t > 0 (1)

The conditions that the left and right hand ends are held at height zero are encoded in the “boundary

conditions”

u(0, t) = 0 for all t > 0 (2)

u(ℓ, t) = 0 for all t > 0 (3)

As we have been told the position and speed of the string at time 0, there are given functions f(x) and g(x)

such that the “initial conditions”

u(x, 0) = f(x) for all 0 < x < ℓ (4)

u

t

(x, 0) = g(x) for all 0 < x < ℓ (5)

are satisﬁed. The problem is to determine u(x, t) for all x and t.

Outline of the Method of Separation of Variables

We are going to solve this problem in three steps.

Step 1 In the ﬁrst step, we ﬁnd all solutions of (1) that are of the special form u(x, t) = X(x)T(t) for

some function X(x) that depends on x but not t and some function T(t) that depends on t but

not x. This is where the name “separation of variables” comes from. It is of course too much to

expect that all solutions of (1) are of this form. But if we ﬁnd a bunch of solutions X

i

(x)T

i

(t)

of this form, then since (1) is a linear equation,

i

a

i

X

i

(x)T

i

(t) is also a solution for any choice

of the constants a

i

. (Check this yourself!) If we are lucky (and we shall be lucky), we will be

able to choose the constants a

i

so that the other conditions (2–5) are also satisﬁed.

Step 2 We impose the boundary conditions (2) and (3).

Step 3 We impose the initial conditions (4) and (5).

The First Step – Finding Factorized Solutions

The factorized function u(x, t) = X(x)T(t) is a solution to the wave equation (1) if and only if

X(x)T

′′

(t) = c

2

X

′′

(x)T(t) ⇐⇒

X

′′

(x)

X(x)

=

1

c

2

T

′′

(t)

T(t)

January 21, 2007 Solution of the Wave Equation by Separation of Variables 1

The left hand side is independent of t. So the right hand side, which is equal to the left hand side, must be

independent of t too. The right hand side is independent of x. So the left hand side must be independent of

x too. So both sides must be independent of both x and t. So both sides must be constant. Let’s call the

constant σ. So we have

X

′′

(x)

X(x)

= σ

1

c

2

T

′′

(t)

T(t)

= σ

⇐⇒ X

′′

(x) −σX(x) = 0 T

′′

(t) −c

2

σT(t) = 0

(6)

We now have two constant coeﬃcient ordinary diﬀerential equations, which we solve in the usual way. We

try X(x) = e

rx

and T(t) = e

st

for some constants r and s to be determined. These are solutions if and only

if

d

2

dx

2

e

rx

−σe

rx

= 0

d

2

dt

2

e

st

−c

2

σe

st

= 0

⇐⇒

_

r

2

−σ

_

e

rx

= 0

_

s

2

−c

2

σ)e

st

= 0

⇐⇒ r

2

−σ = 0 s

2

−c

2

σ = 0

⇐⇒ r = ±

√

σ s = ±c

√

σ

If σ = 0, we now have two independent solutions, namely e

√

σx

and e

−

√

σx

, for X(x) and two independent

solutions, namely e

c

√

σt

and e

−c

√

σt

, for T(t). If σ = 0, the general solution to (6) is

X(x) = d

1

e

√

σx

+ d

2

e

−

√

σx

T(t) = d

3

e

c

√

σt

+ d

4

e

−c

√

σt

for arbitrary constants d

1

, d

2

, d

3

and d

4

. If σ = 0, the equations (6) simplify to

X

′′

(x) = 0 T

′′

(t) = 0

and the general solution is

X(x) = d

1

+ d

2

x T(t) = d

3

+ d

4

t

for arbitrary constants d

1

, d

2

, d

3

and d

4

. We have now found a huge number of solutions to the wave

equation (1). Namely

u(x, t) =

_

d

1

e

√

σx

+ d

2

e

−

√

σx

__

d

3

e

c

√

σt

+ d

4

e

−c

√

σt

_

for arbitrary σ = 0 and arbitrary d

1

, d

2

, d

3

, d

4

u(x, t) =

_

d

1

+ d

2

x

__

d

3

+ d

4

t

_

for arbitrary d

1

, d

2

, d

3

, d

4

The Second Step – Imposition of the Boundary Conditions

If X

i

(x)T

i

(t), i = 1, 2, 3, · · · all solve the wave equation (1), then

i

a

i

X

i

(x)T

i

(t) is also a solution for

any choice of the constants a

i

. This solution satisﬁes the boundary condition (2) if and only if

i

a

i

X

i

(0)T

i

(t) = 0 for all t > 0

This will certainly be the case if X

i

(0) = 0 for all i. In fact, if the a

i

’s are nonzero and the T

i

(t)’s are

independent, then (2) is satisﬁed if and only if all of the X

i

(0)’s are zero. For us, it will be good enough to

simply restrict our attention to X

i

’s for which X

i

(0) = 0, so I am not even going to deﬁne what “independent”

means

(1)

. Similarly, u(x, t) =

i

a

i

X

i

(x)T

i

(t) satisﬁes the boundary condition (3) if and only if

i

a

i

X

i

(ℓ)T

i

(t) = 0 for all t > 0

(1)

It forbids, for example, T

1

= 7T

2

or T

1

= 3T

2

+ 5T

3

.

January 21, 2007 Solution of the Wave Equation by Separation of Variables 2

and this will certainly be the case if X

i

(ℓ) = 0 for all i. We are now going to go through the solutions that

we found in Step 1 and discard all of those that fail to satisfy X(0) = X(ℓ) = 0.

First, consider σ = 0 so that X(x) = d

1

+d

2

x. The condition X(0) = 0 is satisﬁed if and only if d

1

= 0.

The condition X(ℓ) = 0 is satisﬁed if and only if d

1

+ ℓd

2

= 0. So the conditions X(0) = X(ℓ) = 0 are both

satisﬁed only if d

1

= d

2

= 0, in which case X(x) is identically zero. There is nothing to be gained by keeping

an identically zero X(x), so we discard σ = 0 completely.

Next, consider σ = 0 so that d

1

e

√

σx

+ d

2

e

−

√

σx

. The condition X(0) = 0 is satisﬁed if and only if

d

1

+ d

2

= 0. So we require that d

2

= −d

1

. The condition X(ℓ) = 0 is satisﬁed if and only if

0 = d

1

e

√

σℓ

+ d

2

e

−

√

σℓ

= d

1

_

e

√

σℓ

−e

−

√

σℓ

_

If d

1

were zero, then X(x) would again be identically zero and hence useless. So instead, we discard any σ

that does not obey

e

√

σℓ

−e

−

√

σℓ

= 0 ⇐⇒ e

√

σℓ

= e

−

√

σℓ

⇐⇒ e

2

√

σℓ

= 1

In the last step, we multiplied both sides of e

√

σℓ

= e

−

√

σℓ

by e

√

σℓ

. One σ that obeys e

2

√

σℓ

= 1 is σ = 0.

But we are now considering only σ = 0. Fortunately, there are inﬁnitely many complex numbers

(2)

that

work. In fact e

2

√

σℓ

= 1 if and only if there is an integer k such that

2

√

σℓ = 2kπı ⇐⇒

√

σ = k

π

ℓ

ı ⇐⇒ σ = −k

2 π

2

ℓ

2

With

√

σ = k

π

ℓ

ı and d

2

= −d

1

,

X(x)T(t) = d

1

_

e

ı

kπ

ℓ

x

−e

−ı

kπ

ℓ

x

__

d

3

e

ı

ckπ

ℓ

t

+ d

4

e

−ı

ckπ

ℓ

t

_

= 2ıd

1

sin

_

kπ

ℓ

x

__

(d

3

+ d

4

) cos

_

ckπ

ℓ

t) + ı(d

3

−d

4

) sin

_

ckπ

ℓ

t

_¸

= sin

_

kπ

ℓ

x

__

α

k

cos

_

ckπ

ℓ

t) + β

k

sin

_

ckπ

ℓ

t

_¸

where α

k

= 2ıd

1

(d

3

+ d

4

) and β

k

= −2d

1

(d

3

−d

4

). Note that, to this point, d

1

, d

3

and d

4

are allowed to be

any complex numbers so that α

k

and β

k

are allowed to be any complex numbers.

The Third Step – Imposition of the Initial Conditions

We now know that

u(x, t) =

∞

k=1

sin

_

kπ

ℓ

x

__

α

k

cos

_

ckπ

ℓ

t) + β

k

sin

_

ckπ

ℓ

t

_¸

obeys the wave equation (1) and the boundary conditions (2) and (3), for any choice of the constants α

k

, β

k

.

It remains only to see if we can choose the α

k

’s and β

k

’s to satisfy

f(x) = u(x, 0) =

∞

k=1

α

k

sin

_

kπ

ℓ

x

_

(4

′

)

g(x) = u

t

(x, 0) =

∞

k=1

β

k

ckπ

ℓ

sin

_

kπ

ℓ

x

_

(5

′

)

But any (reasonably smooth) function, h(x), deﬁned on the interval 0 < x < ℓ, has a unique representation

(3)

h(x) =

∞

k=1

b

k

sin

kπx

ℓ

(7)

(2)

See, for example, the notes “Complex Numbers and Exponentials”.

(3)

See, for example, the notes “Fourier Series”.

January 21, 2007 Solution of the Wave Equation by Separation of Variables 3

as a linear combination of sin

kπx

ℓ

’s and we also know the formula

b

k

=

2

ℓ

_

ℓ

0

h(x) sin

kπx

ℓ

dx

for the coeﬃcients. We can make (7) match (4

′

) by choosing h(x) = f(x) and b

k

= α

k

. This tells us that

α

k

=

2

ℓ

_

ℓ

0

f(x) sin

kπx

ℓ

dx. Similarly, we can make (7) match (5

′

) by choosing h(x) = g(x) and b

k

= β

k

ckπ

ℓ

.

This tells us that

ckπ

ℓ

β

k

=

2

ℓ

_

ℓ

0

g(x) sin

kπx

ℓ

dx. So we have a solution:

u(x, t) =

∞

k=1

sin

_

kπ

ℓ

x

__

α

k

cos

_

ckπ

ℓ

t) + β

k

sin

_

ckπ

ℓ

t

_¸

(8)

with

α

k

=

2

ℓ

_

ℓ

0

f(x) sin

kπx

ℓ

dx β

k

=

2

ckπ

_

ℓ

0

g(x) sin

kπx

ℓ

dx

While the sum (8) can be very complicated, each term, called a “mode”, is quite simple. For each ﬁxed

t, the mode sin

_

kπ

ℓ

x

__

α

k

cos

_

kcπ

ℓ

t) + β

k

sin

_

kcπ

ℓ

t

_¸

is just a constant times sin

_

kπ

ℓ

x

_

. As x runs from 0

to ℓ, the argument of sin

_

kπ

ℓ

x

_

runs from 0 to kπ, which is k half–periods of sin. Here are graphs, at

ﬁxed t, of the ﬁrst three modes, called the fundamental tone, the ﬁrst harmonic and the second harmonic.

x

y

fundamental

ℓ

x

y

1

st

harmonic

ℓ

x

y

2

nd

harmonic

ℓ

For each ﬁxed x, the mode sin

_

kπ

ℓ

x

__

α

k

cos

_

kcπ

ℓ

t) +β

k

sin

_

kcπ

ℓ

t

_¸

is just a constant times cos

_

kcπ

ℓ

t) plus a

constant times sin

_

kcπ

ℓ

t). As t increases by one second, the argument,

kcπ

ℓ

t, of both cos

_

kcπ

ℓ

t) and sin

_

kcπ

ℓ

t)

increases by

kcπ

ℓ

, which is

kc

2ℓ

cycles (i.e. periods). So the fundamental oscillates at

c

2ℓ

cps, the ﬁrst harmonic

oscillates at 2

c

2ℓ

cps, the second harmonic oscillates at 3

c

2ℓ

cps and so on. If the string has density ρ and

tension T, then we have seen

(4)

that c =

_

T

ρ

. So to increase the frequency of oscillation of a string you

increase the tension and/or decrease the density and/or shorten the string.

Example 1 As a concrete example, suppose that

∂

2

u

∂ t

2

(x, t) = c

2 ∂

2

u

∂x

2

(x, t) for all 0 < x < 1 and t > 0

u(0, t) = u(1, t) = 0 for all t > 0

u(x, 0) = x(1 −x) for all 0 < x < 1

u

t

(x, 0) = 0 for all 0 < x < 1

This is a special case of equations (1–5) with ℓ = 1, f(x) = x(1 −x) and g(x) = 0. So, by (8),

u(x, y) =

∞

k=1

sin(kπx)

_

α

k

cos(ckπt) + β

k

sin(ckπt)

¸

with

α

k

= 2

_

1

0

x(1 − x) sin(kπx) dx β

k

= 2

_

1

0

0 sin(kπx) dx = 0

(4)

See, for example, the notes “Derivation of the Wave Equation”.

January 21, 2007 Solution of the Wave Equation by Separation of Variables 4

Using

(5)

_

1

0

xsin(kπx) dx =

_

1

0

−

1

π

d

dk

cos(kπx) dx = −

1

π

d

dk

_

1

0

cos(kπx) dx = −

1

π

d

dk

1

kπ

sin(kπx)

¸

¸

1

0

= −cos(kπ)

1

kπ

_

1

0

x

2

sin(kπx) dx =

_

1

0

−

1

π

2

d

2

dk

2

sin(kπx) dx = −

1

π

2

d

2

dk

2

_

1

0

sin(kπx) dx =

1

π

2

d

2

dk

2

1

kπ

cos(kπx)

¸

¸

1

0

= cos(kπ)

2−k

2

π

2

k

3

π

3

−

2

k

3

π

3

we have

α

k

= 2

_

1

0

x(1 −x) sin(kπx) dx = 2

_

−cos(kπ)

1

kπ

−cos(kπ)

2−k

2

π

2

k

3

π

3

+

2

k

3

π

3

¸

=

4

k

3

π

3

[1 −cos(kπ)]

=

_

8

k

3

π

3

for k odd

0 for k even

and

u(x, y) =

∞

k=1

k odd

8

k

3

π

3

sin(kπx) cos(ckπt)

Example 2 As a second concrete example, suppose that

∂

2

u

∂ t

2

(x, t) = c

2 ∂

2

u

∂x

2

(x, t) for all 0 < x < 1 and t > 0

u(0, t) = u(1, t) = 0 for all t > 0

u(x, 0) = sin(5πx) + 2 sin(7πx) for all 0 < x < 1

u

t

(x, 0) = 0 for all 0 < x < 1

This is again a special case of equations (1–5) with ℓ = 1. So, by (8),

u(x, y) =

∞

k=1

sin(kπx)

_

α

k

cos(ckπt) + β

k

sin(ckπt)

¸

This time it is very ineﬃcient to use the integral formulae to evaluate α

k

and β

k

. It is easier to observe

directly, just by matching coeﬃcients, that

sin(5πx) + 2 sin(7πx) = u(x, 0) =

∞

k=1

α

k

sin(kπx) ⇒α

k

=

_

1 if k = 5

2 if k = 7

0 if k = 5, 7

0 = u

t

(x, 0) =

∞

k=1

ckπβ

k

sin(kπx) ⇒β

k

= 0

So

u(x, y) = sin(5πx) cos(5cπt) + 2 sin(7πx) cos(7cπt)

(5)

Note that we cannot impose the condition that k is an integer until after evaluating the

d

dk

derivatives.

January 21, 2007 Solution of the Wave Equation by Separation of Variables 5

Using Fourier Series to Solve the Wave Equation

We can also use Fourier series to derive the solution (8) to the wave equation (1) with boundary

conditions (2,3) and initial conditions (4,5). The basic observation is that, for each ﬁxed t ≥ 0, the unknown

u(x, t) is a function of the one variable x and this function vanishes at x = 0 and x = ℓ. Thus, by the Fourier

series theorem and the odd periodic extension trick, u(x, t) has, for each ﬁxed t, a unique expansion

u(x, t) =

∞

k=1

b

k

(t) sin

_

kπx

ℓ

_

(9)

By using other periodic extensions, like the even periodic extension, we can get other expansions of u(x, t)

too. But the odd periodic expansion (9) is particularly useful because, with it, the boundary conditions (2,3)

are automatically satisﬁed. If we substitute x = 0 into the right hand side of (9) we necessarily get zero,

regardless of the value of b

k

(t), because every term contains a factor of sin(0) = 0. Similarly, if we substitute

x = ℓ into the right hand side of (9) we again necessarily get zero, for any b

k

(t), because sin(kπ) = 0 for

every integer k.

The solution u(x, t) is completely determined by the, as yet unknown, coeﬃcients b

k

(t). Furthermore

these coeﬃcients can be found by substituting u(x, t) =

∞

k=1

b

k

(t) sin

_

kπx

ℓ

_

into the three remaining re-

quirements (1), (4), (5) on u(x, t). First the wave equation (1):

0 =

∂

2

u

∂t

2

(x, t) −c

2 ∂

2

u

∂x

2

(x, t) =

∞

k=1

b

′′

k

(t) sin

_

kπx

ℓ

_

+

∞

k=1

k

2

π

2

c

2

ℓ

2

b

k

(t) sin

_

kπx

ℓ

_

=

∞

k=1

_

b

′′

k

(t) +

k

2

π

2

c

2

ℓ

2

b

k

(t)

_

sin

_

kπx

ℓ

_

This says that, for each ﬁxed t ≥ 0, the function 0, viewed as a function of x, has Fourier series expansion

∞

k=1

_

b

′′

k

(t) +

k

2

π

2

c

2

ℓ

2

b

k

(t)

_

sin

_

kπx

ℓ

_

. Applying (9) with h(x) being the zero function and with b

k

replaced

by

_

b

′′

k

(t) +

k

2

π

2

c

2

ℓ

2

b

k

(t)

_

then forces

b

′′

k

(t) +

k

2

π

2

c

2

ℓ

2

b

k

(t) =

2

ℓ

_

ℓ

0

0 sin

_

kπx

ℓ

_

dx = 0 for all k, t (1

′

)

Substituting into (4) and (5) gives

u(0, t) =

∞

k=1

b

k

(0) sin

_

kπx

ℓ

_

= f(x)

∂u

∂t

(0, t) =

∞

k=1

b

′

k

(0) sin

_

kπx

ℓ

_

= g(x)

By uniqueness of Fourier coeﬃcients, once again,

b

k

(0) =

2

ℓ

_

ℓ

0

f(x) sin

_

kπx

ℓ

_

dx (4

′

)

b

′

k

(0) =

2

ℓ

_

ℓ

0

g(x) sin

_

kπx

ℓ

_

dx (5

′

)

For each ﬁxed k, equations (1

′

), (4

′

) and (5

′

) constitute one second order constant coeﬃcient ordinary

diﬀerential equation and two initial conditions for the unknown function b

k

(t).

January 21, 2007 Solution of the Wave Equation by Separation of Variables 6

You already know how to solve constant coeﬃcient ordinary diﬀerential equations. The function b

k

(t) =

e

rt

satisﬁes the ordinary diﬀerential equation (1

′

) if and only if

r

2

+

k

2

π

2

c

2

ℓ

2

= 0

which in turn is true if and only if

r = ±i

kπc

ℓ

so that the general solution to (1

′

) is

b

k

(t) = C

k

e

i

kπc

ℓ

t

+ D

k

e

−i

kπc

ℓ

t

with C

k

and D

k

arbitrary constants. Using e

±i

kπc

ℓ

t

= cos

_

kπc

ℓ

t

_

± i sin

_

kπc

ℓ

t

_

we may rewrite the solution

as

b

k

(t) = α

k

cos

_

ckπ

ℓ

t) + β

k

sin

_

ckπ

ℓ

t

_

with α

k

= C

k

+ D

k

and β

k

= iC

k

− iD

k

again arbitrary constants. They are determined by the initial

conditions (4

′

) and (5

′

).

(4

′

) =⇒ b

k

(0) = α

k

=

2

ℓ

_

ℓ

0

f(x) sin

_

kπx

ℓ

_

dx

(5

′

) =⇒ b

′

k

(0) =

ckπ

ℓ

β

k

=

2

ℓ

_

ℓ

0

g(x) sin

_

kπx

ℓ

_

dx =⇒ β

k

=

2

ckπ

_

ℓ

0

g(x) sin

_

kπx

ℓ

_

dx

This gives us the solution (8) once again.

January 21, 2007 Solution of the Wave Equation by Separation of Variables 7

d3 . For us. 3. So we have X ′′ (x) X(x) =σ ′′ ⇐⇒ X (x) − σX(x) = 0 ′′ T (t) − c σT (t) = 0 ′′ 1 T (t) c2 T (t) 2 =σ (6) We now have two constant coeﬃcient ordinary diﬀerential equations. for example. Let’s call the constant σ. So the right hand side. d3 and d4 . Similarly. then (2) is satisﬁed if and only if all of the Xi (0)’s are zero. So both sides must be independent of both x and t. then i ai Xi (x)Ti (t) is also a solution for any choice of the constants ai . d2 . t) = i ai Xi (x)Ti (t) satisﬁes the boundary condition (3) if and only if ai Xi (ℓ)Ti (t) = 0 i (1) for all t > 0 It forbids. u(x. So both sides must be constant. · · · all solve the wave equation (1). if the ai ’s are nonzero and the Ti (t)’s are independent. 2007 2 . In fact. must be independent of t too. t) = d1 e √ σx T ′′ (t) = 0 + d2 e− √ σx d3 ec √ σt + d4 e−c √ σt for arbitrary σ = 0 and arbitrary d1 . 2. the general solution to (6) is X(x) = d1 e √ σx √ √ s = ±c σ + d2 e− √ σx T (t) = d3 ec √ σt + d4 e−c √ σt for arbitrary constants d1 . namely ec σt and e−c σt . If σ = 0. d3 and d4 . T1 = 7T2 or T1 = 3T2 + 5T3 . which is equal to the left hand side. for T (t). d4 u(x. it will be good enough to simply restrict our attention to Xi ’s for which Xi (0) = 0. d2 . d2 .The left hand side is independent of t. Namely u(x. Solution of the Wave Equation by Separation of Variables January 21. d4 for arbitrary d1 . so I am not even going to deﬁne what “independent” means(1) . namely e σx and e− σx . We try X(x) = erx and T (t) = est for some constants r and s to be determined. we now have two independent solutions. the equations (6) simplify to X ′′ (x) = 0 and the general solution is X(x) = d1 + d2 x T (t) = d3 + d4 t for arbitrary constants d1 . which we solve in the usual way. We have now found a huge number of solutions to the wave equation (1). So the left hand side must be independent of x too. for X(x) and two independent √ √ solutions. This solution satisﬁes the boundary condition (2) if and only if ai Xi (0)Ti (t) = 0 i for all t > 0 This will certainly be the case if Xi (0) = 0 for all i. The right hand side is independent of x. i = 1. These are solutions if and only if rx 2 st d2 st d2 − σerx = 0 dx2 e dt2 e − c σe = 0 ⇐⇒ ⇐⇒ ⇐⇒ r2 − σ erx = 0 r2 − σ = 0 s2 − c2 σ)est = 0 s 2 − c2 σ = 0 √ √ r=± σ If σ = 0. t) = d1 + d2 x d3 + d4 t The Second Step – Imposition of the Boundary Conditions If Xi (x)Ti (t). d2 . d3 . If σ = 0.

So we require that d2 = −d1 . ℓ X(x)T (t) = d1 eı = sin kπ ℓ x √ √ √ √ = 2ıd1 sin kπ ℓ x kπ ℓ x − e−ı kπ ℓ x d3 eı ckπ ℓ t) ckπ ℓ t + d4 e−ı ckπ ℓ t (d3 + d4 ) cos + αk cos ckπ ℓ t) + ı(d3 βk sin ckπ t ℓ − d4 ) sin ckπ ℓ t where αk = 2ıd1 (d3 + d4 ) and βk = −2d1 (d3 − d4 ). t) = ∞ k=1 sin kπ ℓ x αk cos ckπ ℓ t) + βk sin ckπ ℓ t obeys the wave equation (1) and the boundary conditions (2) and (3). The condition X(0) = 0 is satisﬁed if and only if d1 = 0. 0) = ∞ k=1 ∞ k=1 αk sin kπ ℓ x (4′ ) (5′ ) βk ckπ sin ℓ kπ ℓ x But any (reasonably smooth) function. for example. deﬁned on the interval 0 < x < ℓ. It remains only to see if we can choose the αk ’s and βk ’s to satisfy f (x) = u(x. has a unique representation(3) h(x) = (2) (3) ∞ k=1 bk sin kπx ℓ (7) See. There is nothing to be gained by keeping an identically zero X(x). so we discard σ = 0 completely. Note that. Fortunately. we discard any σ that does not obey √ √ √ √ √ e σℓ − e− σℓ = 0 ⇐⇒ e σℓ = e− σℓ ⇐⇒ e2 σℓ = 1 In the last step. The condition X(0) = 0 is satisﬁed if and only if d1 + d2 = 0. h(x). See. in which case X(x) is identically zero. Solution of the Wave Equation by Separation of Variables January 21. So instead. In fact e2 σℓ = 1 if and only if there is an integer k such that √ √ 2 2 σℓ = 2kπı ⇐⇒ σ = k π ı ⇐⇒ σ = −k 2 π2 ℓ ℓ √ With σ = k π ı and d2 = −d1 . 2007 3 . One σ that obeys e2 σℓ = 1 is σ = 0. So the conditions X(0) = X(ℓ) = 0 are both satisﬁed only if d1 = d2 = 0. √ √ Next. βk . consider σ = 0 so that X(x) = d1 + d2 x. The condition X(ℓ) = 0 is satisﬁed if and only if d1 + ℓd2 = 0. consider σ = 0 so that d1 e σx + d2 e− σx . But we are now considering only σ = 0. to this point.and this will certainly be the case if Xi (ℓ) = 0 for all i. the notes “Complex Numbers and Exponentials”. the notes “Fourier Series”. d1 . for example. for any choice of the constants αk . First. 0) = g(x) = ut (x. then X(x) would again be identically zero and hence useless. we multiplied both sides of e σℓ = e− σℓ by e σℓ . The condition X(ℓ) = 0 is satisﬁed if and only if 0 = d1 e √ σℓ + d2 e− √ σℓ = d1 e √ σℓ − e− √ σℓ If d1 were zero. The Third Step – Imposition of the Initial Conditions We now know that u(x. d3 and d4 are allowed to be any complex numbers so that αk and βk are allowed to be any complex numbers. there are inﬁnitely many complex numbers(2) that √ work. We are now going to go through the solutions that we found in Step 1 and discard all of those that fail to satisfy X(0) = X(ℓ) = 0.

t) = c2 ∂2u ∂x2 (x. t) = with ℓ ℓ ∞ k=1 sin kπ ℓ x αk cos ckπ ℓ t) + βk sin ckπ ℓ t (8) αk = 2 ℓ 0 f (x) sin kπx dx ℓ βk = 2 ckπ 0 g(x) sin kπx dx ℓ While the sum (8) can be very complicated. t) = u(1. So we have a solution: ℓ ℓ ℓ u(x. called a “mode”. As t increases by one second.e. the notes “Derivation of the Wave Equation”. Here are graphs. Solution of the Wave Equation by Separation of Variables January 21. Similarly. 0) = x(1 − x) ut (x. each term.as a linear combination of sin kπx ’s and we also know the formula ℓ ℓ bk = 2 ℓ 0 h(x) sin kπx dx ℓ for the coeﬃcients. we can make (7) match (5′ ) by choosing h(x) = g(x) and bk = βk ckπ . by (8). t) = 0 u(x. the mode sin kπ x αk cos kcπ t) + βk sin kcπ t is just a constant times sin kπ x . which is k half–periods of sin. at ℓ ﬁxed t. Example 1 As a concrete example. 0) = 0 This is a special case of equations (1–5) with ℓ = 1. the mode sin kπ x αk cos kcπ t) + βk sin kcπ t is just a constant times cos kcπ t) plus a ℓ ℓ ℓ ℓ constant times sin kcπ t). 2007 4 . for example. suppose that ∂2 u ∂ t2 (x. As x runs from 0 ℓ ℓ ℓ ℓ to ℓ. t) for all 0 < x < 1 and t > 0 for all t > 0 for all 0 < x < 1 for all 0 < x < 1 u(0. For each ﬁxed t. ℓ ℓ ℓ ℓ This tells us that ckπ βk = 2 0 g(x) sin kπx dx. If the string has density ρ and T tension T . kcπ t. the argument. the ﬁrst harmonic and the second harmonic. So. y) = with αk = 2 0 (4) ∞ k=1 sin(kπx) αk cos(ckπt) + βk sin(ckπt) 1 1 x(1 − x) sin(kπx) dx βk = 2 0 0 sin(kπx) dx = 0 See. is quite simple. called the fundamental tone. y y y ℓ x fundamental 1st harmonic ℓ x 2nd harmonic ℓ x For each ﬁxed x. f (x) = x(1 − x) and g(x) = 0. We can make (7) match (4′ ) by choosing h(x) = f (x) and bk = αk . This tells us that ℓ αk = 2 0 f (x) sin kπx dx. of the ﬁrst three modes. u(x. then we have seen(4) that c = ρ . which is kc cycles (i. of both cos kcπ t) and sin kcπ t) ℓ ℓ ℓ ℓ c increases by kcπ . the ﬁrst harmonic ℓ 2ℓ c c oscillates at 2 2ℓ cps. periods). the argument of sin kπ x runs from 0 to kπ. So the fundamental oscillates at 2ℓ cps. So to increase the frequency of oscillation of a string you increase the tension and/or decrease the density and/or shorten the string. the second harmonic oscillates at 3 2ℓ cps and so on.

0) = ∞ k=1 αk sin(kπx) ⇒ αk = 1 2 0 if k = 5 if k = 7 if k = 5. just by matching coeﬃcients. y) = ∞ k=1 sin(kπx) αk cos(ckπt) + βk sin(ckπt) This time it is very ineﬃcient to use the integral formulae to evaluate αk and βk . u(x. It is easier to observe directly. t) = u(1. y) = 8 k3 π 3 sin(kπx) cos(ckπt) Example 2 As a second concrete example. that sin(5πx) + 2 sin(7πx) = u(x. 0) = sin(5πx) + 2 sin(7πx) ut (x. 0) = So ∞ k=1 ckπβk sin(kπx) ⇒ βk = 0 u(x. January 21. y) = sin(5πx) cos(5cπt) + 2 sin(7πx) cos(7cπt) (5) Note that we cannot impose the condition that k is an integer until after evaluating the d dk derivatives. t) for all 0 < x < 1 and t > 0 for all t > 0 for all 0 < x < 1 for all 0 < x < 1 u(0. by (8). So. 2007 Solution of the Wave Equation by Separation of Variables 5 . 7 0 = ut (x. t) = c2 ∂2u ∂x2 (x.Using(5) 1 1 1 1 1 d − π dk cos(kπx) dx = − π d dk 0 1 1 cos(kπx) dx = − π d kπ sin(kπx) dk x sin(kπx) dx = 0 1 0 0 1 1 0 1 = − cos(kπ) kπ x2 sin(kπx) dx = 0 1 d 1 d − π2 dk2 sin(kπx) dx = − π2 dk2 2 2 1 sin(kπx) dx = 0 1 d2 1 π 2 dk2 kπ cos(kπx) 1 0 = we have 1 2 2 π cos(kπ) 2−kπ3 k3 − 2 k3 π 3 αk = 2 0 π 1 x(1 − x) sin(kπx) dx = 2 − cos(kπ) kπ − cos(kπ) 2−kπ3 + k3 2 2 2 k3 π 3 = 4 k3 π 3 [1 − cos(kπ)] = and 8 k3 π 3 for k odd for k even ∞ k=1 k odd 0 u(x. 0) = 0 This is again a special case of equations (1–5) with ℓ = 1. t) = 0 u(x. suppose that ∂2 u ∂ t2 (x.

(5) on u(x. regardless of the value of bk (t). by the Fourier series theorem and the odd periodic extension trick.Using Fourier Series to Solve the Wave Equation We can also use Fourier series to derive the solution (8) to the wave equation (1) with boundary conditions (2. t) − c2 ∂ u (x. the boundary conditions (2.3) are automatically satisﬁed. t) = ∞ bk (t) sin kπx into the three remaining rek=1 ℓ quirements (1). for each ﬁxed t. the function 0. t) has. because sin(kπ) = 0 for every integer k. t) = ∞ k=1 bk (t) sin kπx ℓ (9) By using other periodic extensions. if we substitute x = ℓ into the right hand side of (9) we again necessarily get zero. t) too. equations (1′ ). (4′ ) and (5′ ) constitute one second order constant coeﬃcient ordinary diﬀerential equation and two initial conditions for the unknown function bk (t). (4).3) and initial conditions (4. Similarly. as yet unknown. because every term contains a factor of sin(0) = 0. t). a unique expansion u(x. t) = ∂x2 = 2 ∞ k=1 ∞ k=1 b′′ (t) sin k ′′ bk (t) + kπx ℓ + ∞ k=1 k 2 π 2 c2 ℓ2 bk (t) sin kπx ℓ k 2 π 2 c2 ℓ2 bk (t) sin kπx ℓ This says that. the unknown u(x. for any bk (t). coeﬃcients bk (t). for each ﬁxed t ≥ 0. January 21. like the even periodic extension. t) = ∂u ∂t (0.5). If we substitute x = 0 into the right hand side of (9) we necessarily get zero. ℓ bk (0) = b′ (0) = k 2 ℓ 2 ℓ f (x) sin 0 ℓ kπx ℓ kπx ℓ dx dx (4′ ) (5′ ) g(x) sin 0 For each ﬁxed k. we can get other expansions of u(x. u(x. First the wave equation (1): 0= ∂2u ∂t2 (x. has Fourier series expansion ∞ kπx k 2 π 2 c2 ′′ . with it. t) is completely determined by the. But the odd periodic expansion (9) is particularly useful because. 2007 Solution of the Wave Equation by Separation of Variables 6 . The solution u(x. for each ﬁxed t ≥ 0. viewed as a function of x. t (1′ ) Substituting into (4) and (5) gives u(0. Furthermore these coeﬃcients can be found by substituting u(x. once again. Thus. Applying (9) with h(x) being the zero function and with bk replaced k=1 bk (t) + ℓ2 bk (t) sin ℓ by b′′ (t) + k k 2 π 2 c2 ℓ2 bk (t) then forces b′′ (t) + k k 2 π 2 c2 ℓ2 bk (t) ℓ = 2 ℓ 0 sin 0 kπx ℓ dx = 0 for all k. t) is a function of the one variable x and this function vanishes at x = 0 and x = ℓ. The basic observation is that. t) ∞ k=1 ∞ k=1 bk (0) sin b′ (0) sin k kπx ℓ = f (x) = g(x) = kπx ℓ By uniqueness of Fourier coeﬃcients.

They are determined by the initial conditions (4′ ) and (5′ ).e rt You already know how to solve constant coeﬃcient ordinary diﬀerential equations. The function bk (t) = satisﬁes the ordinary diﬀerential equation (1′ ) if and only if r2 + k 2 π 2 c2 ℓ2 =0 which in turn is true if and only if r = ±i kπc ℓ so that the general solution to (1′ ) is bk (t) = Ck ei with Ck and Dk arbitrary constants. Using e±i ℓ as bk (t) = αk cos kπc kπc ℓ t + Dk e−i kπc ℓ t kπc ℓ t t = cos ± i sin ckπ ℓ t kπc ℓ t we may rewrite the solution ckπ ℓ t) + βk sin with αk = Ck + Dk and βk = iCk − iDk again arbitrary constants. ℓ (4′ ) =⇒ bk (0) = αk (5′ ) =⇒ b′ (0) = k ckπ ℓ βk = = 2 ℓ 2 ℓ f (x) sin 0 ℓ kπx ℓ kπx ℓ dx ℓ g(x) sin 0 dx =⇒ βk = 2 ckπ g(x) sin 0 kπx ℓ dx This gives us the solution (8) once again. 2007 Solution of the Wave Equation by Separation of Variables 7 . January 21.

Sign up to vote on this title

UsefulNot useful- Fourth Semester B.Tech. Degree Examination, December 2009
- m 344 Lecture 16
- Maths-4 (140001)
- Answers Hw 11
- Review of Chapter 6 Fourier Series
- Isc Sample Paper 2010 Mathematics 1
- 12th Question Paper
- Advanced Higher Maths Exam 2012
- 9650Guess paper 17-01-12
- Answers-Instructions-week-4.pdf
- Lab 7
- MA166 Exam #4(Example Problems)
- Tarea Aplicaciones de La Transformada
- 337 Week 0910weeklying
- Chap 7
- Pre-Calculus sine,cosine
- EE451.Ch2.F13
- [Herbert Kreyszig, Erwin Kreyszig] Student Solutio(BookZZ.org)
- HW 14 Solutions
- Example Solutions
- sol_9_fall_04
- Mathcad - CAPE - 2002 - Math Unit 2 - Paper 01
- MATHS PPT
- 9709_s07_qp_3
- Respuestas
- Solution 2005
- MEC4418 - Assignment 1 - Solution
- 9709_s07_qp_3
- CH_2
- Additional Revision Questions 2
- Solution of the Wave Equation by Separation of Variables