Lecture Notes 4: Fourier Series and PDE’s

1. Periodic Functions
A function f (x) defined on R is called a periodic function if there exists
a number T > 0 such that
f (x + T ) = f (x), ∀x ∈ R. (1.1)
The smallest number T for which the relation (1.1) holds is called the pe-
riod of f or fundamental period of f .

Lemma 1.1. If f is T - periodic continuous function, then

Z a+T Z T
f (x)dx = f (x)dx. (1.2)
a 0

Proof. Consider the function

Z x+T
F (x) = f (s)ds
x
It is clear that
F 0 (x) = f (x + T ) − f (x) = 0, ∀x ∈ R (f is T- periodic)
Thus F (x) is a constant function. Hence
Z a+T Z T
F (a) = f (x)dx = f (x)dx = F (0).
a 0


2. Functional Series
Let
f1 (x), f2 (x), ..., fn (x), ... (2.1)
be a sequence of functions defined on some interval I ⊂ R. We say that the
sequence (3.1) is convergent (or pointwise convergent) to a function f (x)
on I if for each fixed point x ∈ I the number sequence {fn (x)} converges
to the number f (x). If at least for one point x0 the sequence f (x0 ) is
divergent, then we say that the sequence of functions {fn (x)} is divergent
on I. .
1
2

A sequence of functions (3.1) is said to be uniformly convergent to a func-

tion f (x) on I if for each ε > 0 there exists a number Nε depending on ε
only, such that
|fn (x) − f (x)| ≤ ε
for all n ≥ Nε .

For a given sequence of functions (3.1) the series

∞
X
fn (x) (2.2)
n=1
is the following limit
lim SN (x),
N →∞
where
SN (x) = f1 (x) + f2 (x) + ... + fN (x)
is called the N -th partial sum of the series.
If the sequence of partial sums {SN (x)} converges to some function s(x)
on I, i.e.
X∞
fn (x) = s(x) (2.3)
n=1
, then we say that the series (3.8) is convergent on I to s(x). Otherwise
the series (3.8) is called divergent.
If the sequence {SN (x)} is uniformly convergent to s(x) then we say that
the series (3.8) is uniformly convergent.

Theorem 2.1. If the functions

f1 (x), f2 (x), ..., fn (x), ...
∞
P
are continuous on an interval [a, b] and the series fn (x) is uniformly
n=1
convergent on [a, b], then the sum of the series s(x) is a continuous function
on [a, b]. Moreover the series obtained term by term integration of this
series is also convergent and
∞ Z b
X Z b
fn (x)dx = s(x)dx.
n=1 a a
 3

Theorem 2.2. (Weierstrass Theorem) If the functions

f1 (x), f2 (x), ..., fn (x), ...
are continuous on an interval [a, b],
|fn (x)| ≤ an , ∀x ∈ [a, b], n = 1, 2, ...
∞
P ∞
P
and the series an is convergent then the series fn (x) is uniformly
n=1 n=1
convergent to some function that is continuous on [a, b].

3. Euler’s formulas and Fourier series

A series of the form
∞
a0 X
+ (an cos(nx) + bn sin(nx)) (3.1)
2 n=1

is called a trigonometric series.

Question: Which functions have trigonometric series expansion . If

f (x) has a trigonometric series expansion
∞
a0 X
f (x) = + (an cos(nx) + bn sin(nx)) (3.2)
2 n=1

how to compute a0 , a1 , ..., an , ..., b1 , b2 , ...?

Theorem 3.1. Suppose that f is 2π-periodic function and

∞
a0 X
f (x) = + (an cos(nx) + bn sin(nx)), (3.3)
2 n=1

where the series is converges uniformly on the real axis. Then

1 π
Z
an = f (x) cos(nx)dx, n = 0, 1, 2... (3.4)
π −π
1 π
Z
bn = f (x) sin(nx)dx, n = 1, 2... (3.5)
π −π
4

Proof. Really since

Z π Z π
cos(nx)dx = 0, sin(nx)dx = 0
−π −π

and due to uniform convergence of the series we can integrate (3.1) over
(−π, π) and get:
Z π
f (x)dx = a0 π,
−π
Let us multiply (3.1) by cos(mx) and integrate over (−π, π). Taking into
account Z π Z π
2
cos (mx)dx = π, sin2 (mx)dx = π
−π −π
we obtain Z π
1
an = f (x) cos(nx)dx, n = 0, 1, 2... (3.6)
π −π
Z π
1
bn = f (x) sin(nx)dx, n = 1, 2... (3.7)
π −π
P∞
Here we have used the fact
P∞ that for each m the series cos(mx) n=1 (an cos(nx)+
bn sin(nx)) and sin(mx) n=1 (an cos(nx) + bn sin(nx)) are uniformly con-
vergent. 
The series (3.1) where an and bn are defined by (3.6) and (3.7) is called the
Fourier series of the function f , the numbers an , bn are called the Fourier
coefficients of f .
Piecewise continuous function. A function f (x) is called piecewise
continuous on [a, b], if limx→b− f (x), limx→a+ f (x) exist f is continuous on
(a, b) except at finitely many of points in (a, b), where f has one-sided
limits.
Theorem 3.2. If 2π-periodic function f (x) and its derivative f 0 (x) are
piecewise continuous functions, then
∞
f (x+) + f (x−) a0 X
= + (an cos(nx) + bn sin(nx)) (3.8)
2 2 n=1

for each x ∈ R, where an and bn are defined by (3.4) and (3.5).

 5

Example 3.3. Consider the function Φ(x) given on [−π, π] by

(
π + x, x ∈ [−π, 0]
φ(x) =
π − x, x ∈ [0, π].
φ(x) is piecewise smooth.
π
1 π 2 π x2
Z Z 
2
a0 = φ(x)dx = (π − x)dx = πx − = 2π − π = π
π −π π 0 π 2 0

2 π
Z π
2 π
Z Z
an = (π − x) cos(nx)dx == 2 cos(nx)dx − x cos(nx)dx
π 0 0 π 0
Z π  0 Z π
2 1 2 π 2
=− x sin(nx) dx = − [x sin(nx)]0 + sin(nx)dx
π 0 n nπ nπ 0
π
(−1)n
  
2 1 2 1 1 2 1
= − cos(nx) = − cos(nπ) = −
nπ n 0 nπ n n nπ n2 n2
2
Thus an = 0 if n is and even number, and an = n2 , if n is and odd number.
∞
π 4X 1
φ(x) = + cos(2k − 1)x
2 π (2k − 1)2
k=1

Example 3.4. Using Fourier series show that

π2 1 1 1
= 1 + 2 + 2 + 2 + ...
8 3 5 7
3.1. Functions of any period p = 2l.
∞
a0 X   nπ   nπ 
f (x) = + an cos x + bn sin x (3.9)
2 n=1
l l

l
1 l
Z Z
1  nπ   nπ 
an = f (x) cos x dx, bn = f (x) sin x dx. (3.10)
l 0 l l 0 l
3.2. Even and Odd Functions. If a function f (x) is an even function ,
then
Z l
2  nπ 
bn = f (x) sin x dx = 0
l 0 l
6

and its Fourier series has the form

∞
a0 X  nπ 
f (x) = + an cos x , (3.11)
2 n=1
l
where
2 l
Z  nπ 
an = f (x) cos x dx, n = 0, 1, 2, ... (3.12)
l 0 l
If a function f (x) is an odd function , then
2 l
Z  nπ 
an = f (x) cos x dx = 0
l 0 l
and its Fourier series has the form
X∞  nπ 
f (x) = bn sin x x, (3.13)
n=1
l
where
2 l
Z  nπ 
bn = f (x) sin x dx, n = 1, 2, ... (3.14)
l 0 l
Let f (x) be defined on [0, l]. We define the even periodic extension fe of
f as follows
fe (x) = f (−x), if x ∈ [−l, 0], and fe (x) = fe (x + 2l), ∀x ∈ R.
An odd periodic extension f0 of f is defined as follows
f0 (x) = −f (−x), if x ∈ [−l, 0], and f0 (x) = f0 (x + 2l), ∀x ∈ R.
Example 3.5. Find Fourier series expansion for f (x) = 1−x2 , x ∈ [−1, 1].

Z 1
4
a0 = 2 (1 − x2 )dx = ,
0 3
Z 1 Z 1
an = 2 (1 − x2 ) cos(nπx)dx = −2 cos(nπx)dx+
0 0
Z  0 Z 1
2 1 2
2 x sin(nπx dx = − 2x sin(nπx)dx =
nπ nπ 0
Z 1  0
4 1 4 1 4
x − cos(nπx dx = − 2
x cos(nπx)| 0 = − 2 2
(−1)n .
nπ 0 nπ (nπ) nπ
 7

Thus we have
∞
2 4 X (−1)n+1
f (x) = + 2 cos(nπx).
3 π n=1 n2

4. Riemann -Lebesgue Lemma

Proposition 4.1. If f (x) is a continuous function , then
Z π
lim In = lim f (x) cos(nx)dx = 0,
n→∞ n→∞ −π
Z π
lim Jn = lim f (x) sin(nx)dx = 0. (4.1)
n→∞ n→∞ −π

Proof of 4.1. Since cos α = − cos(α + π) we have

Z π Z π h π i
In := f (x) cos(nx)dx = − f (x) cos (x + )n dx.
−π −π n
π
Making change of variables x + n = y and using the Lemma 1.1 we obtain
Z π Z π+ π2 Z π
π π
f (x) cos(nx)dx = − f (y − ) cos(ny)dy = − f (y − ) cos(ny)dy
−π −π+ π2 n −π n

Hence we have
Z π Z π
π
In + In = 2 f (x) cos(nx)dx = [f (x) − f (x − )] cos(nx)dx
−π −π n
Z π  Z π
  π
= 2 f (x) cos(nx)dx ≤ |f (x) − f (x − )|dx.
−π −π n

The function f is continuous on [−π, π] thus it is uniformly continuous on [−π, π]. Therefore the
integral in the right hand side of (R) tends to zero as n → ∞. So In → 0 as n → ∞. Similarly
we can show that Jn → 0 as n → ∞.

Problem. Let f (x) be 2π -periodic and f 0 (x) is continuous function. Show

that
   
1 1
an = o , bn = o .
n n
8

5. Bessel inequality and mean value approximation.

Theorem 5.1. (Bessel inequality) If f (x) is a piecewise continuous
function on (−π, π), then the following inequality caleed the Bessel inequal-
ity holds true
∞
a20 X 2 1 π 2
Z
2
+ (an + bn ) ≤ f (x)dx, (5.1)
2 n=1
π −π

where a0 , an and bn , n = 1, 2, ... are Fourier coefficients of f .

Proof. It is clear that
Z π Z π Z π
0 ≤ EN = f 2 (x)dx − 2 SN (x)f (x)dx + SN2 (x)dx. (5.2)
−π −π −π
By using Euler formulas we get
Z π
SN (x)f (x)dx
−π
Z π N Z π
a0 X
= f (x)dx + f (x) [ak cos(nx) + bk sin(nx)] dx
2 −π n=1 −π
 
1 2 2 2 2 2
= π a0 + a1 + ... + aN + b1 + ... + bN , (5.3)
2
Z π  2 
2 a0
[SN (x)] dx = π + a21 + ... + a2N + b21 + ... + b2N . (5.4)
−π 2
Substituting (5.3) and (5.4) into (5.2) we obtain:
Z π " N
#
2
a X
0 ≤ EN = f 2 (x)dx − π 0 + (a2n + b2n ) .
−π 2 n=1
or
N Z π
a20 X 2 1
+ (an + b2n ) ≤ f 2 (x)dx. (5.5)
2 n=1
π −π

We cam pass to the limit as N → ∞ and get (5.1). 

By using the Bessel inequality we can prove the following theorem
Theorem 5.2. Suppose that f is continuous 2π-periodic function and f 0
is piecewise continuous function. Then the Fourier series of f converges
absolutely and uniformly to the function f .
 9

Proof. Let us calculate the Fourier coefficients of f 0 :

1 π 0
Z
1
α0 = f (x)dx = (f (π) − f (−π)) = 0,
π −π π

1 π 0
Z Z π
1 x=π
αn = f (x) cos(nx)dx = cos(nx)f (x) +n f (x) sin(nx)dx = nbn ,

π −π π x=−π −π

1 π 0
Z
βn = f (x) sin(nx)dx
π −π
Z π
1 x=π
= sin(nx)f (x) −n f (x) cos(nx)dx = nan .

π x=−π −π

So we have
αn = nbn , n = 0, 1, 2, ..., βn = nan , n = 1, 2, ..., (5.6)
Employing the inequality
1
|ab| ≤ a2 + b2
4
we obtain from (5.6) the following inequlity
1 1 1
|an | + |bn | = |βn | + |αn | ≤ 2 + αn2 + βn2 .
n n 2n
Due to the Bessel inequality the series
∞
X
(αn2 + βn2 )
n=1

is convergent. Hence the series ∞

P
n=1 (|an | + |bn |) is also convergent. There-
fore the series
∞
a0 X  nπ nπ 
+ an cos x + bn sin x
2 n=1
l l
is uniformly convergent to a continuous function f .

where an and bn are Fourier coefficients of the function f
10

Example 5.3. Assume that the Fourier series of f (x) on [−π, π] converges
to f (x) and can be integrated term by term. Multiply
∞
a0 X  nπ nπ 
+ an cos x + bn sin x
2 n=1
l l
by f (x) and integrate the obtained relation from −π to π to derive the
identity

π ∞
a2 X 2
Z
1
f (x)dx = 0 +
2
(an + b2n ). (5.7)
π −π 2 n=1
This identity is called the Parseval identity.
6. Heat Equation. Method of Separation of Variables
We consider the problem
ut = a2 uxx , x ∈ (0, l), t > 0, (6.1)
u(x, 0) = f (x), x ∈ [0, l], (6.2)
u(0, t) = u(l, t) = 0, t ≥ 0, (6.3)
We assume that the solution of the problem has the form
u(x, t) = X(x)T (t),
where X(x) and T (t) are nonzero functions. Substituting into (6.1) we get
X(x)T 0 (t) = a2 X 00 (x)T (t).
Dividing both sides of the last equality by a2 X(x)T (t) we obtain
T 0 (t) X 00 (x)
= (6.4)
a2 T (t) X(x)
Since the left hand side of (6.4) depends only on t and the right hand
side depend only on x each side of this equality can only be equal to some
constant. Thus
T 0 (t) X 00 (x)
= = −λ, λ = constant
a2 T (t) X(x)
or
T 0 (t) = λa2 T (t) (6.5)
X 00 (x) + λX(x) = 0. (6.6)
 11

It follows from (3) that

X(0) = X(l) = 0. (6.7)
So we have to find the values of λ for which the equation (6.6) has nonzero
solution which satisfy (6.7). The values of l for which (6.6) has nonzero
solution satisfying (6.7) are called eigenvalues of the problem (6.6),(6.7)
When l = 0 (6.7), the corresponding solutions - eigenfunctions of (6.6),(6.7).
When l = 0 the equation has a general solution
X(x) = Ax + B
This function satisfies (6.7) just for A = B = 0. Thus λ = 0 is not an
eigenvalue.
If λ < 0 then general solution of (6.6) has the form
√ √
|λ|x − |λ|x
X(x) = C1 e + C1 e .
It is easy to see that this function satisfies (6.7) just when C1 = C2 = 0.
So (6.6),(6.7) has not negative eigenvalues.
If l > 0 then the general solution of (6.6) has the form
√ √
X(x) = C1 cos( λx) + C2 sin( λx).
Substituting into (6.7) we obtain
√
X(0) = C1 = 0, X(l) = C2 sin( λl) = 0
The second equality holds for ll = nπ, n = ±1, ±2, ... Thus the numbers
n2 π 2
λn = 2 , n = 1, 2, ...
l
are eigenvalues of the problem (6.6),(6.7), and the functions
 nπ 
Xn (x) = sin x , n = 1, 2, ...
l
are the corresponding eigenfunctions. It is easy to see that the general
solution of (6.5) for l = ln has the form
2
Tn (t) = Dn e−a λn t
, n = 1, 2, ...
2
Hence for each n = 1, 2, ... the function e−a λn t sin nπ


l x satisfies (6.1),(6.3).
Since the equation (6.6) is a linear equation for each N
N  nπ 
−a2 λn t
X
uN (x, t) = Dn e sin x , (6.8)
n=1
l
12

where Dn , n = 1, ..., N are arbitrary constants also satisfies (6.1),(6.3).

Next we try to satisfy the initial condition (6.2):
N
X  nπ 
uN (x, 0) = Dn sin x = f (x)
n=1
l
We see that the solution of the problem (6.1)-(6.3) has this form just when
the initial function is linear combination of functions
p p k2π2
sin λ1 x, ...., sin λn x, λk = 2 .
l
Let us consider the series
∞  nπ 
−a2 λn t
X
u(x, t) = Dn e sin x . (6.9)
n=1
l
∞
P
Let us note that if |Dn | < ∞ , then this series is uniformly convergent
n=1
on [−l, l] × [0, T ], ∀T > 0. The function u(x, t) defined by (6.9) satisfies
the boundary conditions (6.3) since each term of the series satisfies these
conditions. It follows from (6.9) that u(x, t) satisfies the initial condition
(6.2)
X∞  nπ 
f (x) = u(x, 0) = Dn sin x (6.10)
n=1
l
iff Z l
2  nπ 
Dn = fn = f (x) sin x dx.
l 0 l
Theorem 6.1. If f (x) is continuous on [0, L], f 0 (x) is piecewise continuous
on [0, L], f (0) = f (l) = 0 then the function
∞  nπ 
−a2 λn t
X
u(x, t) = fn e sin x (6.11)
n=1
l
satisfies (6.1)-(6.3).
To prove this theorem we need the following proposition
Proposition 6.2. Assume that the functions vn (x, t), n = 1, 2, ... are continuous QT = [a, b] ×
[t0 , T ] and
|vn (x, t)| ≤ an , ∀(x, t) ∈ QT , n = 1, 2, ...
 13

∞
P
where the sequence of positive numbers {an } is so that the series an is convergent. Then the
n=1
∞
P
series vn (x, t) is absolutely and uniformly convergent on QT . Moreover the function
n=1
∞
X
v(x, t) = vn (x, t)
n=1
is continuous on QT .
If    2 
 ∂vn   ∂ vn 
 ∂t (x, t) ≤ bn ,  ∂x2 (x, t) ≤ dn , ∀(x, t) ∈ QT , n = 1, 2, ...
   

and
∞
X ∞
X
bn < ∞, dn < ∞,
n=1 n=1
then the series
∞ ∞
X ∂vn X ∂ 2 vn
(x, t) and (x, t)
∂t ∂x2
n=1 n=1
uniformly converge to vt (x, t) and vxx (x, t) in QT . Moreover these functions are continuous in
QT .
Proof of Theorem 6.1. Since f is piecewise smooth the series
∞
X
|fn |
n=1
is convergent. Thus the Proposition 6.2 implies that the function u(x, t) is continuous on [0, l] ×
[0, ∞). Let us show that the series
∞ ∞
X ∂un X ∂ 2 un
(a) (x, t) and (b) (x, t) (6.12)
∂t ∂x2
n=1 n=1
are convergent uniformly for t ≥ t0 , x ∈ [0, l], where t0 is an arbitrary positive number. Con-
tinuity of f on [0, l] implies boundedness of the sequence {fn }. So there exists M > 0 so
that
|fn | ≤ M, n = 1, 2, ...
Thus for each t ≥ t0 and x ∈ [0, l] we have
2 2
   
 ∂un a2 n2 π 2 a2 n2 π 2
 = −fn a2 n π e− l2 t sin nπ x  ≤ M1 n2 e− l2 t0 ,
  
 (x, t) (6.13)
 ∂t   l2 l 
2
where M1 = M a2 πl2 .
 2
n2 π 2 − a2 n22 π2 t
   nπ 
 ∂ un a2 n2 π 2
 ≤ n2 M1 e− l2 t0 ,
 

 ∂x2 (x, t)  =  −fn e l sin x (6.14)
  l2 l  a2
It is easy to see that the series
∞
M1 a2 n2 π2
n2 2 e− l2 t0
X
a
n=1
is convergent. Therefore due to the Proposition 6.2 the function u(x, t) defined by (7.10) is a
solution of the problem (6.1)-(6.3).
14

∞ ∞
a2n b2n be convergent. Show that
P P
Problem 6.3. Let the series and
n=1 n=1

∞ ∞
!1/2 ∞
!1/2
X X X
|an bn | ≤ a2n b2n
n=1 n=1 n=1

6.1. Nonhomogeneous Equation.

ut = a2 uxx + h(x, t), x ∈ (0, l), t > 0, (6.15)
u(x, 0) = f (x), x ∈ [0, l], (6.16)
u(0, t) = u(l, t) = 0, t ≥ 0, (6.17)
We look for solution of the problem (6.15)-(6.17) in the form
X∞  nπ 
u(x, t) = un (t) sin x (6.18)
n=1
l
We expand h(x, t)
∞
X  nπ 
h(x, t) = hn (t) sin x
n=1
l
By using (6.18) we obtain from (6.15)
∞
X p
 0
un (t) + λn a2 un (t) − hn (t) sin( λn x) = 0.

n=1
This equality holds iff
u0n (t) + λn a2 un (t) = hn (t), n = 1, 2, ... (6.19)
Taking into accoun the initial condition (6.16) we obtain
X∞  nπ  X∞  nπ 
u(x, 0) = hn (0) sin x = f (x) = fn sin x.
n=1
l n=1
l
It follows then
un (0) = fn , n = 1, 2, ... (6.20)
The initial value problem (6.19),(6.21) has the solution
Z t
−λn a2 t 2
un (t) = e fn + e−λn a (t−s) hn (s)ds
0
 15

Inserting the expression for un (t) into (6.18) we get

∞  Z t   nπ 
−λn a2 t −λn a2 (t−s)
X
u(x, t) = fn e + e hn (s)ds sin x. (6.21)
n=1 0 l
n2 π 2
Let us recall that λn = l2 .

Problem 6.4. Show that the problem (6.1)-(6.3) has a unique solution.
Hint. Assume that v(x, t) is another solution of this problem, i.e.
vt = a2 vxx , x ∈ (0, l), t > 0,
v(x, 0) = f (x), x ∈ [0, l],
v(0, t) = v(l, t) = 0, t ≥ 0,
Then show that the function
w(x, t) = u(x, t) − v(x, t)
which is a solution of the problem
wt = a2 wxx , x ∈ (0, l), t > 0,
w(x, 0) = 0, x ∈ [0, l],
w(0, t) = w(l, t) = 0, t ≥ 0
satisfies Z l Z l
1d 2 2
[w(x, t)] dx + a [wx (x, t)]2 dx = 0.
2 dt 0 0

6.2. Nonhomogeneous boundary conditions. Let us consider the prob-

lem
ut = a2 uxx , x ∈ (0, l), t > 0, (6.22)
u(x, 0) = f (x), x ∈ [0, l], (6.23)
u(0, t) = A, u(l, t) = B, t ≥ 0, (6.24)
where A, B are given constants. The solution of the problem u(x, t) is
a summ of two functions v(x, t and W (x), where W is a solution of the
problem (
W 00 (x) = 0, x ∈ (0, l),
(6.25)
W (0) = A, W (l) = B
16

and v is a solution of the problem


2
vt = a vxx , x ∈ (0, l), t > 0,

v(x, 0) = f (x) − W (x), x ∈ [0, l], (6.26)

u(0, t) = u(l, t) = 0, t ≥ 0,

It is clear that
1
W (x) = A + (B − A)x
l
is a solution of the problem (6.25) Hence the solution of the proble (6.27)-
(6.29) is the function
∞  nπ 
X
−n2 a2 t 1
u(x, t) = qn e sin x + A + (B − A)x,
n=1
l l

where
Z l 
2 1  nπ 
qn = f (x) − A − (B − A)x sin x .
l 0 l l
Next we consider the following problem

ut = a2 uxx , x ∈ (0, l), t > 0, (6.27)

u(x, 0) = f (x), x ∈ [0, l], (6.28)

u(0, t) = φ(t), u(l, t) = ψ(t), t ≥ 0, (6.29)

we look the solution of the problem(6.27)-(6.29) in the form
∞
X p 
u(x, t) = un (t) sin λn x (6.30)
n=1

For ut we have

∞
X p 
ut (x, t) = u0n (t) sin λn x (6.31)
n=1

∞
X p 
uxx (x, t) = gn (t) sin λn x , (6.32)
n=1

where
Z l
2 p 
gn (t) = uxx (x, t) sin λn x dx.
l 0
 17

Integrating by parts we obtain

Z l
2h p il 2p p 
gn (t) = ux sin λn x − λn ux (x, t) cos ln x dx
l 0 l 0
 p
2 p l 2
Z l p 
=− λn u(x, t) cos λn x − λn u(x, t) sin λn x dx
l 0 l 0
2p
= λn [u(0, t) − u(l, t) cos(nπ)] − λn un (t).
l
Employing the boundary conditions (6.29) we obtain
2p
gn (t) = λn [φ(t) − ψ(t) cos(nπ)] − λn un (t)
l
Thus (6.32) implies
∞  √ √ 
X 2 λn 2 λn n
p 
uxx (x, t) = φ(t) − (−1) ψ(t) − ln un (t) sin λn x
l l
n=1
By using the last relation and (6.31) in(6.27) we obtain
∞   2√ √
2a2 λn

2 2a λn
X p 
0 n
un (t) − a φ(t) − (−1) ψ(t) − aλn un (t) sin λn x = 0.
l l
n=1
Therefore the Fourier coefficients un (t) satisfy
 √ √ 
0 2 2 λn 2 λn n
un (t) = a φ(t) − (−1) ψ(t) − λn un (t) , n = 1, 2, ... (6.33)
l l
The function u will satisfy the initial condition (6.28) iff
un (0) = fn , n = 1, 2, ... (6.34)
We solve the initial value problem (6.33),(6.34) and get
√ Z
−a2 λn t 2a2 λn t −a2 λn (t−s)
un (t) = fn e − e [(−1)n ψ(s) − φ(s)] ds, n = 1, 2, ... (6.35)
l 0
So the solution of the problem (6.27)-(6.29) has the form (6.30), where un (t), n = 1, 2, ... are
defined by (6.35)

7. Wave Equation
In this section we study the wave equation. The first problem is the initial
boundary value problem:
utt = c2 uxx , x ∈ (0, l), t > 0, (7.1)
u(x, 0) = f (x), ut (x, 0) = g(x), x ∈ [0, l], (7.2)
u(0, t) = u(l, t) = 0, t ≥ 0, (7.3)
We assume that the solution of the problem has the form
u(x, t) = X(x)T (t),
18

where X(x) and T (t) are nonzero functions. Substituting into (7.1) we get
X(x)T 00 (t) = a2 X 00 (x)T (t).
Dividing both sides of the last equality by c2 X(x)T (t) we obtain
T 00 (t) X 00 (x)
= (7.4)
c2 T (t) X(x)
Since the left hand side of (7.4) depends only on t and the right hand
side depend only on x each side of this equality can only be equal to some
constant. Thus
T 00 (t) X 00 (x)
= = −λ, λ = constant
c2 T (t) X(x)
or
T 00 (t) = λc2 T (t) (7.5)
X 00 (x) + λX(x) = 0. (7.6)
It follows from (7.3) that
X(0) = X(l) = 0. (7.7)
So we have to solve the eigenvalue problem (7.6),(7.7). We have seen that
the numbers
n2 π 2
λn = 2 , n = 1, 2, ...
l
are eigenvalues of the problem (7.6),(7.7), and the functions
 nπ 
Xn (x) = sin x , n = 1, 2, ...
l
are the corresponding eigenfunctions. It is easy to see that the general
solution of (7.5) for l = λn has the form
p p
Tn (t) = An cos(c λn t) + Bn sin(c λn t), n = 1, 2, ...
It is easy to see that for each N the function
N h
X p p i  nπ 
uN (x, t) = An cos(c λn t) + Bn sin(c λn t) sin x , (7.8)
n=1
l
where An , Bn , n = 1, ..., N are arbitrary constants satisfies (7.1),(7.3). But
this function may satisfy the initial condition (7.2) only when f is a linear
 19

combination of finitely many eigenfunctions. To satisfy the initial condition

(7.2) we consider the series
X∞ h p p i  nπ 
u(x, t) = An cos(c λn t) + Bn sin(c λn t) sin x . (7.9)
n=1
l
∞
P ∞
P
Let us note that if |An | < ∞ and |Bn | < ∞ , then this series
n=1 n=1
is uniformly convergent on [−l, l] × [0, T ], ∀T > 0. The function u(x, t)
defined by (7.9) satisfies the boundary conditions (7.3) since each term
of the series satisfies these conditions. It follows from (7.9) that u(x, t)
satisfies the initial conditions (7.2)
∞
X  nπ 
f (x) = u(x, 0) = An sin x
n=1
l
∞
X cnπ  nπ 
g(x) = ut (x, 0) = Bn sin x
n=1
l l
iff
Z l
2  nπ 
An = fn = f (x) sin x dx,
l 0 l
and
Z l
l 2  nπ 
Bn = gn = g(x) sin x dx,
cnπ cnπ 0 l
Similar to the corresponding theorem for the heat equation we can prove
the following
Theorem 7.1. If f (x), g(x), g 0 (x), f 0 (x), f 00 (x) are continuous g 00 (x), f 000 (x)
are piecewise continuous on [0, L],
f (0) = f (l) = 0, f 00 (0) = f 00 (l) = 0, g(0) = g(l) = 0
then the function
∞   p 
X  p  gn  nπ 
u(x, t) = fn cos c λn t + √ sin c λn t sin x (7.10)
n=1
c λ n l

is a solution of (7.1)-(7.3).
20

7.1. Nonhomogeneous Equation.

utt = c2 uxx + h(x, t), x ∈ (0, l), t > 0, (7.11)
u(x, 0) = f (x), ut (x, 0) = g(x), x ∈ [0, l], (7.12)
u(0, t) = u(l, t) = 0, t ≥ 0, (7.13)
We look for solution of the problem (7.11)-(7.13) in the form
∞
X  nπ 
u(x, t) = un (t) sin x (7.14)
l
n=1

We expand h(x, t)
∞
X  nπ 
h(x, t) = hn (t) sin x
l
n=1
By using (7.14) we obtain from (7.11)
∞
X p
 0
un (t) + λn a2 un (t) − hn (t) sin( λn x) = 0.

n=1

This equality hold iff

u00n (t) + λn c2 un (t) = hn (t), n = 1, 2, ... (7.15)
Taking into account the initial condition (7.12) we obtain
∞
X  nπ  ∞
X  nπ 
u(x, 0) = An (0) sin x = f (x) = fn sin x.
l l
n=1 n=1
∞ ∞
X cnπ  nπ  X  nπ 
ut (x, 0) = Bn (0) sin x = g(x) = gn sin x.
l l l
n=1 n=1
It follows then
l
An (0) = fn , Bn (0) = gn , n = 1, 2, ... (7.16)
cnπ
The initial value problem (7.15),(7.16) has the solution
 p  lgn  p  Z t h cnπ i
un (t) = fn cos c λn t + √ sin c λn t + sin (t − s) hn (s)ds
c ln 0 l
Inserting the expression for un (t) into (7.14) we get
∞   p 
X  p  gn  nπ 
u(x, t) = fn cos c λn t + √ sin c λn t sin x
c λn l
n=1
∞ Z t
X h p i  nπ 
+ sin c λn (t − s) hn (s)ds sin x (7.17)
0 l
n=1
n2 π 2
Remember that λn = l2
.

Problem 7.2. Show that the problem (7.1)-(7.3) has a unique solution.
 21

Hint. Assume that v is another solution of the problem. Then observe

that w(x, t) satisfies
wtt = c2 wxx , x ∈ (0, l), t > 0, (7.18)
w(x, 0) = 0, wt (x, 0) = 0, x ∈ [0, l], (7.19)
w(0, t) = w(l, t) = 0, t ≥ 0. (7.20)
Multiply the equation by wt , integrate the obtained relation with respect
to x over the interval (0, l) and obtain the equality
d l
Z
[wt (x, t)]2 + [wx (x, t)]2 dx = 0.


dt 0
7.2. The Cauchy problem for the wave equation. Now we consider
the initial value problem for the wave equation, i.e. we would like to find
solution of the equation
utt = c2 uxx , t > 0, x ∈ (−∞, ∞), (7.21)
under the initial conditions
u(x, 0) = f (x); ut (x, 0) = g(x), x ∈ (−∞, ∞), (7.22)
where f and g are given numbers and c > 0 is a given number. To solve
the problem we make the following change of variables
ξ = x − ct, η = x + ct.
By using the chain rule we find
ut = uξ ξt + uη ηt = −auξ + cuη = c(uη − uξ ),
utt = c(uηξ ξt + uηη ηt − uξξ ξt − uξη ηt ) = c2 (uξξ − 2uξη + uηη ), (7.23)
ux = uξ ξx + uη ηx = uξ + uη ,
uxx = uξξ ξx + uξη ηx + uηξ ξx + uηη ηx = uξξ + 2uξη + uηη (7.24)
Bu using (7.23) and (7.24) in (7.21) we find
uξη = 0. (7.25)
It is clear that for each differentiable functions u1 , u2 the function
u(ξ, η) = u1 (ξ) + u2 (η),
isa solution of (7.25) . Hence the function
u(x, t) = u1 (x − ct) + u2 (x + ct) (7.26)
22

is a solution of (7.21). Bu using the initial conditions (7.22) we get

u(x, 0) = u1 (x) + u2 (x) = f (x), (7.27)
ut (x, 0) = −cu01 (x) + cu02 (x) = g(x).
Integrating the last equality we obtain
1 x
Z
u2 (x) − u1 (x) = g(s)ds + C. (7.28)
c x0
Solving the system of equations (7.27), (7.28) we obtain
1 x
Z
1 C
u1 (x) = f (x) − g(s)ds − ,
2 2c x0 2
Z x
1 1 C
u2 (x) = f (x) + g(s)ds + .
2 2c x0 2
Iserting the values of u1 , u2 from the last two equalities into (7.26) we get
Z x−ct
1 x+ct
Z
1 1 1
u(x, t) = ϕ(x − ct) − g(s)ds + f (x + ct) + g(s)ds
2 2a x0 2 2c x0
or Z x+ct
f (x − ct) + f (x + ct) 1
u(x, t) = + g(s)ds, (7.29)
2 2a x−ct
The last equality is called the D’Alembert formula .