Professional Documents
Culture Documents
Department of Mathematics
Fourier Theory
B.M.N. Clarke
Table of Contents
∂2u ∂2u
Laplace's equation, + = 0, for u(x, y) a function of two variables,
∂x 2 ∂y 2
∂u ∂2u
the heat equation, ∂t – κ = 0, for u(x, t) a function of two variables.
∂x 2
In the heat equation, x represents the position along the bar measured from some origin, t
represents time, u(x , t) the temperature at position x, time t. Fourier was initially
concerned with the heat equation. Incidentally, the same equation describes the
concentration of a dye diffusing in a liquid such as water. For this reason the equation is
sometimes called the diffusion equation.
In the wave equation, x, represents the position along an elastic string under tension,
measured from some origin, t represents time, u(x, t) the displacement of the string from
equilibrium at position x, time t.
In Laplace's equation, u(x, y) represents the steady temperature of a flat conducting plate
at the position (x, y) in the plane.
Since both the heat equation and the wave equation involve a single space variable x, we
sometime refer to them as the one dimensional heat equation and the one dimensional
wave equation respectively.
Laplace's equation involves two spatial variables and is therefore sometimes called the
two-dimensional laplace equation. Laplace's equation is connected to the theory of analytic
functions of a complex variable. If f(z) = u(x, y) + iv(x, y), the real and imaginary parts
u(x, y), v(x, y) satisfy the Cauchy-Riemann equations,
∂u ∂v ∂u ∂v
∂x = ∂y , ∂y = – ∂x ,
Then
∂2u ∂2v ∂2u
= =– 2
∂x 2
∂x∂y ∂y
or
∂2u ∂2u
+ = 0.
∂x 2 ∂y 2
Similarly,
∂2v ∂2v
+ = 0.
∂x 2 ∂y 2
Heat conduction and wave propagation usually occur in 3 space dimensions and are
described by the following versions of Laplace's equation, the heat equation and the wave
equation;
∂2u ∂2u ∂2u
+ + = 0,
∂x 2 ∂y 2 ∂z2
1. Introduction. 2
∂u ∂ u
2 ∂2u ∂2u
∂t – κ + + = 0,
∂x 2 ∂y 2 ∂z2
∂2u 2 ∂ u + ∂ u + ∂ u = 0.
2 2 2
– c
∂t2 ∂x 2 ∂y 2 ∂z2
2. Linear differential operators. 3
2. Linear differential operators.
All of the above mentioned partial differential equations can be written in the form
L[u] = F
where
2 ∂2u ∂2u ∂2u
L[u] ≡ ∇ u = + + in Laplace's equation,
∂x 2 ∂y 2 ∂z2
L [u ] is in each case, a linear partial differential operator. Linearity means that for any
two functions u1 , u2 , and any two constants c 1, c2,
Let u(x1 ,x2 ,...,xn) be a function of n variables x = (x1 ,x2 ,...,xn). Then the most general linear
partial differential operator is of the form
n n
∂2u n
∂u
L[u] = ∑ ∑ aij (x)∂x ∂x +
i j
∑ bi(x) ∂x i + c(x)u
i=1 j=1 i=1
The highest order partial derivative appearing is the order of the partial differential
operator. Henceforth we will consider only second order partial differential operators of the
form
n n
∂2u n
∂u
L[u] = ∑ ∑ aij (x)∂x ∂x +
i j
∑ bi(x) ∂x i + c(x)u.
i=1 j=1 i=1
The general linear second order partial differential equation is of the form
L[u] = F(x)
where F(x) is a given function. When F(x) ≡ 0, the equation L[u] = 0 is called homogeneous.
If F(x) ≠ 0, the equation L[u] = F(x) is called non-homogeneous.
Linearity of L is essential to the success of the Fourier method. There are usually
(infinitely) many solutions of a linear partial differential equation. The number of
2. Linear differential operators. 4
solutions may be restricted by imposing extra conditions. Often these extra conditions are
given as linear equations involving u and its derivatives on the boundary of some region Ω
⊂ Rn. These equations are written as boundary conditions
B[u(x)] = φ(x)
∂u 2
∂t –κ ∇ u = 0, x ∈ Ω, t > 0.
u(x, 0) = f(x), x ∈ Ω.
L[uj ] = Fj , j = 1, ... , k
B[uj ] = φ j , j = 1, ... , k,
then the linear combination u = c1u1 + .... + c kuk satisfies the linear partial differential
equation
u(x, t) = X(x)T(t).
X(x)T'(t) = κ X"(x)T(t).
T'(t) X"(x)
= .
κ T(t) X(x)
Since the variables x, t appear on separate sides of this equation, each side of this equation
can only be equal to a constant, say λ. Then
These are constant coefficient ordinary differential equations. λ is real but could be
positive, negative or zero. Assume for the moment that λ > 0. The solutions are
0 = X(0) = A + B
0 = X(l) = Ae l
√ λ + Be –l √
λ.
Eliminating A, B leads to the condition
e l
√ λ – e –l √
λ = 0
or
3. Separation of variables. 6
The boundary conditions u(0, t) = 0, u(l, t) = 0, t > 0, are satisfied if and only if X(0) = 0, X(l)
= 0. That is
0 = X(0) = A , 0 = X(l) = Bl
or
A = B = 0.
also satisfies the heat equation and the boundary conditions. For the initial condition to be
satisfied
∞
nπ x
u(x, 0) = f(x) = ∑ bn sin l ,
0 < x < l.
n=1
That is we must show that f(x) can be represented as a series expansion of sines. We are
mainly concerned wuth those functions f(x) which have this property.
Example 2. Suppose the conducting bar is insulated at each end, the temperature u(x,t)
satisfies the same heat equation and initial condition but different boundary conditions
Separation of variables u(x, t ) = X(x)T(t) in the heat equation leads to the same ordinary
differential equations
0 = X'(0), 0 = X'(l).
That is
0= λ
√ (A – B), 0= λ (Ae l√ λ
√ – Be – l √
λ ).
Eliminating A, B again leads to the condition, λ ≠ 0,
e l
√λ – e – l √
λ = 0
or
sinh (l√ λ)= 0.
There are no values of λ > 0 which satisfy this condition. So λ cannot be positive.
Suppose that λ < 0 and let λ = – µ2 for some real µ . Then √ λ = iµ and hence µ satisfies
i µ
e –e l – i µ l =0
or
sin (µl)= 0.
nπ
This has solutions µ nl = nπ , n = ±1, ±2,..., and hence µn = l , n = ±1, ±2,... ,
nπ 2
λ n = – µn 2 = – l , n = 1, 2,... ,
κ n 2 π2 t inπ x inπx
–
l2 l – l nπ x
Tn (t) = e , Xn (x) = e +e = 2 cos l , n = 1, 2,...
If λ = 0, T(t) = C, and X(x) = A + Bx. The boundary conditions u(0, t) = 0, u(l, t) = 0, t>
0, if and only if X'(0) = 0, X'(l) = 0. That is
0 = X'(0) = B , 0 = X'(l).
3. Separation of variables. 8
or
X(x) = A 0 , T(t) = C0
also satisfies the heat equation and the boundary conditions. For the initial condition to be
satisfied
∞
nπ x
u(x, 0) = f(x) = ∑ an cos l , 0 < x < l.
n=0
A function f(x), f: R→R is called periodic if f(x + P) = f(x) for all x ∈ R. P > 0 is called the
period of f.
Suppose f(x ) is periodic with period 2π , then an important question is whether f(x ) has a
Fourier series expansion of the form
a0 ∞
f(x) = 2 + ∑ (an cos nx + b n sin nx ), 0 < x < 2π .
n=1
a0
The constant term is taken as 2 as a matter of convenience.
The formulas
e ix = cos x + i sin x, e – ix = cos x – i sin x
an – ibn an + ib n a0
cn = 2 , c–n = 2 , n = 1, 2, ... , c0 = 2 .
Assuming for the moment that the 2π -periodic function f has a Fourier series expansion,
the Fourier coefficients cn will be determined using the following orthogonality property of
the complex exponentials e inx , n = 0, ±1, ±2, ... ,.
Lemma.
π 0; n≠m
∫e
inx – imx
e dx = .
– π 2π ; n=m
π π
∫e ∫e
inx – imx i (n–m )x
Proof. For n ≠ m, e dx = dx
– π – π
1 i (n–m )π
= i(n – m) [e – e –i (n–m )π ]
2
= (n – m) sin(n – m)π = 0.
π π
∫e ∫ 1 dx = 2 π.
inx – inx
If n = m, e dx =
– π – π
¯
4. Fourier Series. 10
∞
Since f(x) = ∑ cne inx ,
–∞
π π ∞
inx – imx
∫ ∫ ∑ cne
– imx
f(x) e dx = e dx
– π – π – ∞
∞ π
= ∑ cn ∫e
– i (n – m )x
dx = 2 π c m .
–∞ – π
Hence
π
∫ f(x) e
1 – imx
cm = 2π dx , m = 0, ±1, ±2, ... ,
– π
Since
an – ibn an + ib n a0
cn = 2 , c–n = 2 , n = 1, 2, ... , c0 = 2 ,
∫ f(x) dx,
1
a0 = 2c0 = π
– π
π π
∞ a0 ∞
∑ cne inx
, 2 + ∑ (an cos nx + b n sin nx ),
–∞ n=1
is called a Fourier series expansion of f(x). The first, the exponential form and the second
the trigonometric form.
Notice that if f(x) is an even periodic function, i.e. f(–x) = f(x) for all x, then
π 0 π
Thus the Fourier series expansion of an even function f(x) has only cosine terms
a0 ∞
2 + ∑ an cos nx
n=1
Furthermore, the coefficients a n
π 0 π
π 0 π
∫ f(x) sin nx dx ,
2
= π n = 1, 2, ... .
0
It is worth noting that for a periodic function f(x), the constant term in the Fourier series
π
a0
∫ f(x) dx
1
2 = 2π
– π
4. Fourier Series. 12
– 1; – π < x < 0
f(x) = .
1; 0 < x < π
1.5
0.5
-5 -3 -1 1 2 3 4 5 6
-1
-1.5
-6.28 6.28
This is called a squarewave function. It is obviously an odd function, hence its Fourier
cosine coefficients a n are all zero and
π π
0; n even
= .
4
; n odd
nπ
4
Therefore b 2n = 0, b2n– 1 = , n = 1, 2, ... . The Fourier series of f(x) is
(2n – 1) π
∞ ∞
∑ bn sin nx, = ∑ b2n– 1 sin (2n – 1)x
n=1 n=1
∞
4
= ∑ sin (2n – 1)x
n=1 (2n – 1) π
4 ∞ 1
= ∑
π n=1 2n –1
sin (2n – 1)x.
(b). Let f(x) be periodic with period 2 π, f(x) = |sin x|, – π < x < π,
– sin x; – π<x<0
f(x) = .
sin x; 0< x<π
4. Fourier Series. 13
1.5
1.25
0.75
0.25
-5 -3 -1 1 2 3 4 5 6
-0.5
-6.28 6.28
=
1 1 (1 – cos( n + 1) π ) – 1 (1 – cos( n – 1) π)
π n + 1 n– 1
0; n odd
= 2 – 2 ; n even
1
π n + 1 n – 1
0; n odd
= 4 .
–
(n2 – 1)π
; n even
4
Therefore a 2n = – , a2n– 1 = 0 and the Fourier series of f(x) is
((2n) 2 – 1)π
∞
a0 2 4 ∞ 1
2 + ∑ an cos nx =
π
– ∑
π n=1 4n2 – 1 cos 2nx.
n=1
5. Bessel's Inequality. 14
5. Bessel's inequality.
∞ π
∑ ∫ |f(x)|2 dx.
1
|cn|2 ≤ 2π
n=– ∞ – π
N 2 N N
f(x) –
∑ cne ∑ cne ∑ cme imx
inx inx
0 ≤ f(x) – = f(x) –
n=– N n=– N m=–N
N N N N
= |f(x)|2 – ∑ cne inx
f(x) – ∑ cm e – imx
f(x) + ∑ ∑ cn cm e i (n – m )x
n=– N m=–N n=– N m=–N
π N π N π
∫ ∑ ∫e ∑ ∫e
1 1 inx 1 – imx
0≤ 2π |f(x)|2 dx – cn 2 π f(x) dx – cm 2 π f(x) dx
– π n=– N – π m=–N – π
N N π
∑ ∑ ∫e
1 i (n – m )x
+ cn cm 2 π dx
n=– N m=–N – π
π N N N
∫ ∑ ∑ ∑
1
= 2π |f(x)|2 dx – cn cn – cm cm + cn cn
– π n=– N m=–N n=– N
π N
∫ ∑
1
= 2π |f(x)|2 dx – |cn|2 for any N.
– π n=– N
Bessel's inequality will later be shown to be actually an equality but for now it implies that
∞
the series ∑ |cn|2 converges where c n are the Fourier coefficients of the Riemann
n=– ∞
integrable function f.
an – ibn an + ib n a0
cn = 2 , c–n = 2 , n = 1, 2, ... , c0 = 2 ,
∞ a02 ∞ ∞ a + ib 2
an – ibn2
∑ |cn|2 = 4 + ∑ 2 + ∑ n 2 n
n=– ∞ n=1 n=1
a02 ∞
∑ [|an|2 + |b n|2]
1
= 4 + 2
n=1
∫ |f(x)|2 dx.
1
≤ 2π
– π
∞ ∞
∑ |an|2 , ∑ |b n|2 also converge, where an, bn are the Fourier cosine and
n=1 n=1
sine coefficients of f.
6. Convergence results for Fourier series. 16
6. Convergence results for Fourier series.
We will consider the question: For what functions f do the Fourier series
∞ a0 ∞
∑ cne inx , 2 + ∑ (an cos nx + b n sin nx ),
–∞ n=1
converge?
Because we are dealing with functions, the concept of convergence must be made precise.
Do we mean the numerical series converging at every x ∈ [– π , π]? Can convergence be
different at different points x, indeed can we have convergence at some points and not at
others and if so, which are the points of convergence? Can we have some sort of average
convergence on [– π , π ]?
f is called piecewise smooth if f and its derivative f' are piecewise continuous on [a, b].
N 1 π
= ∑
2 π – ∫π
e – iny
f(y) dy e inx
n=– N
π N
1
= ∫ 2π ∑ e in (x – y ) f(y) dy.
– π n=– N
N
∑
1
We define D N(x) = 2π e inx. Then
n=– N
π π
SNf(x) = ∫ D N(x – y)f(y) dy = ∫ D N(y)f(x – y) dy .
– π – π
[The last equality follows from a change of variable and the periodicity of the integrand
DN(x – y)f(y)]. The function DN(x) is called the Dirichlet kernel and
2N
∑
1 – iNx
DN(x ) = 2π e e inx
n=0
i (2 N + 1) x
= 2 π e – iNx
1 e –1
ix
e –1
6. Convergence results for Fourier series. 17
i (N + 1) x
– e – iN x
= 2 π
1 e
ix
e –1
1 1
e i (N + 2)x
– e – i (N + 2)x
=
1
2π
ix ix
e 2 – e– 2
=
1 sin (N + 12) x.
2π x
sin (2 )
Lemma.
0 π
∫ ∫
1
D N(x) dx = D N(x) dx = 2 .
– π 0
N N
∑ ∑
1 1 1 1
Proof. DN(x) = 2 π + 2 π e inx = 2 π + π cos nx
n=– N n=1
n≠0
N π
x sin nx
∑
1 1
= 2π + π n 0 2
= .
n=1
¯
f 1
lim SN(x) = 2 [f(x–) + f(x+ ) ]
N→ ∞
for every x. Hence lim SNf(x) = f(x) for each point x of continuity of f.
N→ ∞
π
Proof.
f
SN(x) –
1
2 [ f(x–) + f(x+ ) ]= ∫ 1
D N(y)f(x – y) dy – 2 [f(x–) + f(x+ ) ]
– π
π 0
=
0
∫ D N(y)[f(x – y) – (f(x–)]dy + ∫ D N(y)[f(x – y) – f(x+ )]dy
– π
(from lemma above)
6. Convergence results for Fourier series. 18
π 0
=
0
∫ D N(y)[f(x – y) – f(x–)]dy + ∫ D N(–y)[f(x + y) – f(x+ )](–dy)
π
π
= ∫
0
D N(y)[f(x – y) – f(x–) + f(x + y) – f(x+ )]dy
π
sin (N + 12) y [f(x – y) – f(x–) + f(x + y) – f(x+)]dy
∫
1
= 2π y
0 sin (2 )
π
y2 f(x – y) – f(x–) f(x + y) – f(x+ )
∫
1 1
= π sin (N + ) y 2 y y
+
y
dy
0
sin (2 )
For fixed x define
y2 f(x – y) – f(x–) f(x + y) – f(x+ )
g(y) =
y + ,
y y
sin (2)
which is an odd piecewise continuous function for [– π , π ] by the condition of piecewise
smoothness on f.
Then
π
∫
1 1
f
SN(x) – 2 [ f(x–) + f(x+ ) ] =π
0
sin (N + 12) y g(y)dy
π π
y y
∫ sin (Ny){ cos (2) g(y) } dy + 0∫ cos (Ny){ sin(2) g(y) } dy
1 1
= π π
0
The last two terms are the Fourier sine and cosine coefficients BN , AN of
y y
1
2 cos ()
2
1
g(y), 2 sin 2 ()
g(y) respectively. By Bessel's inequality, BN , A N → 0 as N →
∞.
Therefore
SN(x) – 2 [f(x–) + f(x+ ) ]= BN + A N → 0, as N → ∞.
f 1
This result, that the partial sums of the Fourier series of a piecewise smooth 2 π−periodic
function converges pointwise to the mean of the left and right hand limits at x. If x is a
point of continuity of f, then the partial sums converge pointwise to f(x).
6. Convergence results for Fourier series. 19
Fourier series provide a useful method for summing certain numerical series.
Example. The Fourier series of the continuous periodic function f(x) =|sin x|, x ∈ [– π , π ],
is
2 4 ∞ 1
π
– ∑ 2
π n=1 4n – 1
cos 2nx
1.5
1.25
0.75
0.25
-5 -3 -1 1 2 3 4 5 6
-0.5
-6.28 6.28
Because x = 0 is a point of continuity of f, the Fourier series converges at x = 0 to f(0) = 0,
2 4 ∞ 1
0=
π
– ∑ 2
π n=1 4n – 1
or
∞
1
∑ 1
2 =2.
n=1 4n – 1
π
At x = 2 , the Fourier series converges to
π 2 4 ∞ 1
()
f 2 =1=
π
– ∑
π n=1 4n2 – 1 cos(n π)
2 4 ∞ (–1)n
= –
π
∑
π n=1 4n2 – 1
∞
(–1)n π – 2
∑ 2
4n – 1
= – 4 .
n=1
7. Differentiation and Integration of Fourier Series. 20
7. Differentiation and Integration of Fourier Series.
Fourier series can be differentiated term-by-term but the question is does the resulting
series converge and if so, to what does it converge? Similar concerns apply to the series
resulting from the term-by-term integration of a Fourier series.
∫ f'(x) e
1 – inx
c'n = 2π dx
– π
π π
–
∫ f(x) (–ine
1 – inx 1 – inx
= 2π f(x)e 2π )dx
–π – π
∫ f(x) e
1 – inx
= in 2π dx = incn.
– π
Theorem. Let f be 2 π -periodic, piecewise smooth, with piecewise smooth derivative f'.
Then the Fourier series of f' is
∞ ∞
∑ incne inx
= ∑ (nbn cos nx – nan sin nx )
–∞ n=1
and converge for each x where f'(x) exists. If f' is not continuous at x, then the series above
1
converge to 2 [f'(x–) + f'(x+ )].
∫
1
and let F(x) = f(t) dt. If c 0 = 2 a0 = 0, then for all x,
0
∞
cn inx A0 ∞ – bn an
F(x) = C0 + ∑ in e = 2 + ∑ n cos nx + n sin nx
n=– ∞ n=1
n≠0
7. Differentiation and Integration of Fourier Series. 21
A0 π
∫
1
where C 0 = 2 = 2π F(x) dx the average value of F on [– π , π ].
–π
If c 0 ≠ 0, then the series above converges to F(x) – c0x.
x
Proof. Since f(x) is piecewise continuous, F(x) =
0
∫ f(t) dt is continuous.
Therefore the Fourier series of F converges pointwise at each x ∈ [– π , π] to F(x). Since f(x) =
F'(x), by the two previous theorems,
an = nB n , bn = –nAn , cn = inC n ,
where an, bn, cn are the Fourier coefficients of f and An, B n, C n are the Fourier coefficients
of F. If n ≠ 0, this implies that
bn an cn
An = –
n , Bn = n , Cn = in .
A0 π
∫
1
The constant C0 = 2 = 2π F(x) dx is the constant term in the Fourier series of
–π
x
F(x) = ∫
0
f(t) dt.
If c 0 ≠ 0, f(x) – c0 has zero mean value on [–π , π ] and therefore its zeroth Fourier coefficient
is
π π
∫ ∫
1 1
2π (f(x) – c0)dx = 2π f(x) dx – c 0 = c0 – c 0 = 0.
–π –π
Applying the result just obtained to f(x) – c0 and its anti-derivative F(x) – c 0x completes
the theorem.
¯
Integrating and differentiating known Fourier series using the above results is a useful
way of obtaining new Fourier series.
0; – π<x<0
Example. f(x) =
1; 0<x<π
1.5
1.25
0.75
0.25
-5 -3 -1 1 2 3 4 5 6
-0.5
-6.28319 6.28319
is piecewise continuous with Fourier series
∞
1
∑ 2n – 1
1 2
f(x) = 2 + π sin (2n – 1)x
n=1
x 0; – π < x < 0
∫
1
(table #7). Since c 0 = 2 ≠ 0, F(x) = f(t) dt. = .
0 x; 0 < x < π
Therefore
∞
x A0 1
F(x) – 2 = 2 – π ∑
2
2 cos (2n – 1)x
n=1 (2n – 1)
A0 π ∞
π x π 1
∫ ∑
1 2
Since 2 = 2π F(x) dx = 4 , F(x) – 2 = 4 – π 2 cos (2n – 1)x.
–π n=1 (2n – 1)
x
–2; – π<x<0
x |x|
But F(x) – 2 = = 2
x
2; 0<x<π
Therefore
∞
|x| π 1
∑
2
2 = 4– π 2 cos (2n – 1)x
n=1 (2n – 1)
or
∞
π 1
∑
4
|x| = 2 – π 2 cos (2n – 1)x.
n=1 (2n – 1)
It is often convenient to represent a given function as a Fourier series which contains only
cosine terms or only sine terms, as in the initial examples.
Let f(x) be a piecewise smooth function given on the interval [0, π ] and extended to [–π , 0]
as an even function. That is, f(x) = f(–x), x ∈ [–π , 0]. Then the periodic extension to R , is
piecewise smooth and has Fourier series
a0 ∞
2 + ∑ an cos nx
n=1
where
π
a0 ∞ ∞
The Fourier series 2 + ∑ an cos nx, ∑ bn sin nx are known as half-range series for
n=1 n=1
f(x), x ∈ [0, π].
= |sin x|
8. Half-range Fourier series. 24
1.5
1.25
0.75
0.25
-2 -1 1 2 3
-0.5
-3.14159 3.14159
∞
1
∑ 2
2 4
f(x) = π –π cos 2nx (table #8).
n=1 2n – 1
9. General Intervals. 25
9. General Intervals.
The theory of Fourier series can be expanded to include functions which have arbitrary
period. Let f :R→R have period 2l > 0. That is f(x + 2l) = f(x), x ∈ R. Also define φ :R→R by
lx
φ(x) = f . Then
π
l(x + 2 π)
φ(x + 2 π) = f
lx lx
= f + 2 l = f = φ(x)
π π π
So φ has period 2π . Applying the earlier results to φ, results in the Fourier series
∞ a0 ∞
φ(x) = ∑ cne inx = 2 + ∑ (an cos nx + b n sin nx ),
–∞ n=1
where
π π π
l l
π y nπ y nπ y
∫l φ l sin l dy = l ∫ f(y) sin l dy,
1 1
bn = l
– –l
l inπ y l inπ y
1 π y – l 1 – l
c n = 2l
–
∫l φ l e
dy = 2l
–
∫l f(y) e dy.
Therefore
inπ x
∞
π x
∑
l
f(x) = φ l = cn e
–∞
∞
a0 nπ x nπ x
= 2 + ∑ an cos l + bn sin l
n=1
It follows from the above calculations that if f(x) is an even function on [–l, l], its Fourier
series contains no sine terms and is of the form
∞
a0 nπ x
2 + ∑ an cos l ,
n=1
with
l
nπξ
∫ f(ξ) cos
2
an = l
0
l d ξ
9. General Intervals. 26
Likewise if f(x) is an odd function on [–l, l], its Fourier series contains no cosine terms and
is of the form
∞
nπ x
∑ bn sin l ,
n=1
with
l
nπξ
∫ f(ξ) sin
2
bn = l
0
l d ξ.
A function f(x) defined on an interval [0, l] can be extended to [–l, l] either as an even or an
odd function and the extended periodically with period 2l to the real line. The Fourier
series of such functions are half-range expansions on [0, l] and contain no sine terms in the
case of an even extension or no cosine terms for an odd extension.
10. Application to Laplace's equation. 27
10. Application to Laplace's equation.
∂2u ∂2u
+ = 0, 0 ≤ x ≤ a, 0 ≤ y ≤ b
∂x 2 ∂y 2
u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b.
u=f
y=b
u=0 u=0
x=a
u=0
Assuming a separation of variables solution of the form u(x, y) = X(x)Y(y) and substituting
into Laplaces equation,
X" Y + X Y" = 0
or
X" Y"
– X = Y = constant = λ
X(x) = A cos √
λx + B sin√
λx , Y (y) = C cosh√
λy + D sinh√
λy .
To satisfy the boundary conditions u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, requires that
nπ y
Y(0) = C = 0 and therefore Y (y) = D sinh a . The separation of variables solutions for
λ > 0 are
nπ x nπ y
u(x, y) = X(x)Y(y) = BD sin sinh a , n = 1, 2, ... ,
a
X(x) = A cosh√
µx + B sinh √
µx , Y (y) = C cos √
µy + D sin√
µy .
To satisfy the boundary conditions u(0, y) = u(a, y) = 0, 0 ≤ y ≤ b, requires that
X(0) = A = 0, X(a) = Ba = 0.
That is, A = B = 0 and the only separation of variables solution for λ = 0 is the trivial
solution.
nπ x nπ y
un(x, y) = Bn sin a sinh a , n = 1, 2, ... ,
where Bn are the constants BD for each n = 1, 2, ... . By superposition, a solution is also
given by
∞
u(x, y) = ∑ Bn u n(x, y)
n=1
∞
nπ x nπ y
= ∑ Bn sin a sinh a .
n=1
∞
nπ x nπ y
u(x, y) = ∑ Bn sin a sinh a .
n=1
Suppose all four sides of the rectangle have non-zero boundary conditions, then we can
break up the problem into four sub-problems each similar to the one we have just solved.
Each of these subproblems will have zero boundary conditions on three of the four sides
and can be solved as above. Then the solution to the boundary-value problem is the sum of
the four sub-problems by superposition.
∂2u 1 ∂u 1 ∂2u
∆u(r, θ) = + + = 0, r < a, 0 ≤ θ < 2 π,
∂r2 r ∂r r2 ∂θ2
u=
a
10. Application to Laplace's equation. 30
Using separation of variables, assume that
1 ' 1
R" Θ + R Θ + RΘ " = 0,
r r2
or separating variables,
r2R" + rR ' Θ"
= – = constant = λ2 (say),
R Θ
Consider the case λ = 0; then the solutions of the above equations are
Θ = A 0 + B 0θ,
R = C 0 + D0 log r
Θ = A λ cos λθ + B λ sin λθ
R = C λ rλ + Dλ r –λ
where the coefficients depend on λ . Again because solutions are periodic in θ with period
2π, we conclude that λ = n = 1,2,3,.., and because solutions are bounded for r ≤ a, Dn = 0, n
= 1,2,3,... By superposition we can combine these solutions to obtain
∞
u(r, θ) = ∑ Cnrn (A n cos nθ + B n sin nθ )
n=0
∞
u(r, θ) = ∑ rn (A n cos nθ + B n sin nθ )
n=0
At the boundary r = a,
∞
f( θ ) = u(a, θ ) = ∑ (A
an n cos nθ + B n sin nθ )
n=0
10. Application to Laplace's equation. 31
a0 ∞
= 2 + ∑ (a n cos nθ + b n sin nθ )
n=1
where
2π 2π
1 1
an =
π ∫f(t) cosnt dt , bn = π ∫f(t) sinnt dt
0 0
are the Fourier coefficients of the function f. Equating coefficients of cos n θ , sin nθ, in
these two expressions for f(θ), we obtain that
an bn
An = n , Bn =
a an
and hence
a0 ∞
r
n
u(r, θ ) = 2 + ∑ a (a n cos nθ + b n sin nθ )
n=1
We can get a closed form expression for u by substituting the formulae for the Fourier
coefficients as follows.
2π ∞ 2π
1 1 r n
f(t) dt + ∑ ∫f(t) ( cos nt cos nθ + sin nt sin nθ )dt
2π ∫0
u(r, θ ) =
π n = 1 a 0
2π ∞ 2π
1 1 rn
= ∫
2π 0
f(t) dt +
π
∑ a 0∫f(t) cos n(t – θ) dt
n=1
2π 2π ∞
1 1 r cos n(t– θ) dt
n
= ∫
2π 0
f(t) dt + ∫ f(t)
π0
∑ a
n=1
where the interchange of the order of summation and integration is allowed by the uniform
convergence of the series.
2π ∞
1 rn
Therefore u(r, θ ) = ∫ f(t) 1 + 2 ∑ cos n(t – θ) dt.
2π n=1
a
0
Using the exponential form, 2 cos x = eix + e –i x the term in the square brackets reduces to
two infinite geometric series whose sum is
∞
r cos n(t – θ) =
n a2 – r 2
1 +2 ∑
a a2 – 2arcos(t – θ) + r 2
n=1
(exercise). Therefore
2π
1 a2 – r 2
u(r, θ ) =
2π ∫ f(t)
a2 – 2ar cos(t – θ) + r 2
dt,
0
10. Application to Laplace's equation. 32
The functions sin nx , cos n x, einx, e – inx , n = 0, 1, 2, ... , are examples of orthogonal
functions on [– π, π ] Their orthogonality properties follow from the fact that they solutions
of linear second order ordinary differential equations.
– u"(x) = λ u(x),
for λ = n 2. The ordinary differential equation together with the boundary conditions is
called a boundary-value problem.
The set {e inx , n = 0, ±1, ±2, ... } form a basis for the vector space L2 [– π, π ] consisting of
π
functions f :[–π , π ] →R for which ∫ |f(x)|2 dx < ∞. A set of functions {φn(x); n = 1, 2, ... } is
– π
said to be a basis for L2 [– π, π ] if for any function f ∈ L2 [– π, π ] , there is a unique set of
scalars c n , n = 1, 2, ... such that
π N
2
lim ∫
N→ ∞ – π
f(x) –
∑ cnφn(x) dx = 0.
n=1
∞
Then we say that f has the expansion f = ∑ cnφn.
n=1
The interval [– π, π ] can be replaced by [a, b] and the theory of Fourier series generalised
to the expansion of arbitrary functions f ∈ L2[a, b] in terms of an orthonormal basis {φ n(x);
n = 1, 2,... } consisting of functions which are solutions of a boundary-value problem for
certain second order linear ordinary differential equations.
Let p (x ), q(x), w (x) be real continuous functions on the interval [a , b]. Let p(x) be
continuous and p(x) > 0, q(x) ≥ 0, w(x) ≥ 0 on [a, b]. Let α0 , α1 , β0, β1, be real constants.
α 0u(a) + α 1u'(a) = 0,
β 0u(b) + β 1u'(b) = 0.
The value of λ for which this boundary-value problem has non-trivial solutions are called
eigenvalues of the boundary-value problem and the corresponding solutions u(x) are called
eigenfunctions. It can be shown that the eigenvalues form a countable set { λn; n = 1, 2,... }
and the corresponding eigenfunctions { φn(x); n = 1, 2,... } are orthonormal with respect to
the weighted inner-product
b
1
L[f] Ê= w(x) [–(p(x)f')' + q(x)f ].
b b
= –p(x)f' g + ∫ (p(x)f' g' + q(x)f g )dx
a a
Also
b
<f ; L[g]>w =
a
∫ f(x) [(p(x)g' )' + q(x)g ] dx
b b
= –p(x)f g' + ∫ (p(x)f' g' + q(x)f g )dx.
a a
β1 β1
= –p(b) f'(b) – g'(b) – – f'(b) g'(b)
β0 β0
α1 α1
+ p(a) f'(a) – g'(a) – – f'(a) g'(a)
α0 α0
= 0.
α 0φ(a) + α 1φ'(a) = 0,
β 0φ(b) + β1φ'(b) = 0.
Then
λ ||φ||w2 = λ <φ ; φ>w = <λφ ; φ >w
= <L[ φ] ; φ>w = <φ ; L[ φ ]>w (above lemma)
Proof.
L[φ] = λφ L[ψ] = λψ
α φ(a) + α φ'(a) = 0
0 1 α ψ(a) + α ψ'(a) = 0
0 1
β φ(b) + β φ'(b) = 0
0 1
β ψ(b) + β ψ'(b) = 0
0 1
(λ – µ )<φ ; ψ>w = 0.
∞
β 0f(b) + β1f'(b) = 0, the series f(x) = ∑ cnφn (x) converges uniformly on [a, b]. That is
n=1
N
lim max f(x) –
N→ ∞ x ∈[a; b]
∑ cnφn(x) = 0.
n=1
12. Bessel Functions. 37
12. Bessel Functions.
∂u ∂u ∂ u + 1 ∂u + 1 ∂ u = 0,
2 2
u(a, θ, t) = 0,
0 ≤ θ < 2π , t > 0.
Assuming a solution of the form u(r, θ, t) = R(r) Θ(θ) T(t), substituting in the heat equation,
1 1
R Θ T' – κ R" ΘT + r R' ΘT + 2 RΘ"T = 0,
r
The equations for T, Θ are familiar but the equation for R is Bessel's equation and has
non-constant coefficients. We can write it in the Sturm-Liouville form,
ν2
– (rR' )' + r R = λ 2rR,
ν2
where p(r) = r ≥ 0, q(r) =
r > 0, w(r) = r ≥ 0.
A change of variable ρ = λ r in Bessel's equation results in
d2 S dS
ρ2 2 + ρ + ( ρ 2 – ν 2)S = 0,
dρ dρ
ρ
where S( ρ ) = R(r) = R . We denote a solution of this Bessel's equation as S( ρ ) = J ν(ρ )
λ
and then a solution of the original form of Bessel's equation is R (r) = S( λ r) = J ν(λ r). A
second solution of Bessel's equation is Y ν (λ r), a Bessel function of the second kind, and
therefore the general solution is
R(r) = AJ ν(λ r) + BJ –ν (λ r)
for constants A, B. The solution Jν (λ r) is called a Bessel's function of the first kind and a
series representation can be found
ν λ 2r2 λ 4r4
Jν (λ r) = ( λr) 1 – + + .... .
2(2 + 2ν) 2.4.(2 + 2 ν)(4 + 2ν)
For ν > 0, Jν(0) = 0, J 0 (0) = 1. Y ν(λ r), J –ν (λ r) are unbounded as r → ∞, hence the only
bounded solutions R(r) occur when B = 0.
The Bessel functions J ν(λ r) are oscillatory for r > 0. For ν > 0, let the zeros of Jν (ρ ) be
{ρ νm ; m = 0, 1, ... }, where ρν 0 = 0 for all ν > 0.
ρ νmr ∞
The Bessel functions Jν a , are orthogonal on the interval (0, a), with respect to
m=1
a
the weighted inner product <f ; g >w = ∫ r f(r) g(r) dr.
0
That is,
ρ νmr ρ νlr
a
∫
0
r J ν a Jν a dr = 0, if m ≠ l.
Returning to the heat equation on the disc 0 ≤ r < a, 0 ≤ θ < 2π , the solutions of the form
u(r, θ , t) = R(r) Θ(θ) T(t) are given by
By superposition, a solution of the heat equation on the disc satisfying the boundary
condition u(a, θ , t) = 0, is given by
∞ ∞ ρ nm κ t ρ nmr
2
u(r, θ , t) = ∑ ∑ exp –
J
a2 n a
(C nm cos(n θ) + Dnm sin(nθ )
).
n=o m=1
π
ρ nmr ρ nlr
a
0
∫ r J n J
a n a
dr = 0, if m ≠ l, and ∫ cos(nθ) sin (kθ ) dθ , if n ≠ k,
-π
;
ρ nmr ρ nmr ∞∞
imply the orthogonality of Jn cos(n θ) ; Jn a sin(nθ ) on the rectangle
a n=0; m=1
(r, θ) ∈ (0, a)×(– π , π ) with respect to the inner product
a π
π
ρ nmr ρ kl r
a
∫ ∫
0 –π
r Jn
a
sin(nθ ) Jk
a
sin(k θ) d θ dr
n≠ k
0;
0 ; n = k and m ≠ l
= a
ρ kl r2
π ∫ r Jk a dr; n = k and m = l
0
π
ρ nmr ρ kl r
a
∫ ∫
0 –π
r Jn
a
cos(n θ) Jk
a
sin(kθ ) d θ dr = 0 for all n, m, k and l.
∫ ∫
0 –π
r J k
a
cos(k θ) f(r, θ ) dθ dr = C kl π ∫ r J k
0 a
dr,
hence
π
ρ kl r
a
1
Ckl = a
ρ kl r2
∫ ∫
0 –π
r J k
a
cos(k θ) f(r, θ ) dθ dr.
π∫ r Jk dr
0 a
Similarly,
π
ρ kl r
a
1
Dkl = a
ρ kl r2
∫ ∫
0 –π
r J k
a
sin(kθ) f(r, θ) dθ dr,
π∫ r Jk dr
0 a
k = 0, 1, 2, ... , l = 1, 2, ... .
13. Fourier Transforms. 41
13. Fourier Transforms.
∞
A function f : R→R is said to be integrable on R if
–
∫∞ |f(x) |dx < ∞. We call the class of all
∞
∫∞ |f(x) |2dx
1
such functions L (R). Similarly, f : R→R is said to be square integrable on R if
–
f(x) = ,
0 ; otherwise
x– 3 ; x > 1
2
g(x) = .
0 ; otherwise
Then
∞ 1 2 1
–3 1
–
∫∞ |f(x) |dx = ∫
0
x dx = 3x 3| = 3 < ∞
0
and
∞ 1 4 1
–3 –3 1
–
∫∞ |f(x) |2dx =
0
∫x dx = – 3x |0 = ∞.
∞ ∞ 2 1
–3 ∞
–
∫∞ |g(x) |dx = ∫ x dx = 3x 3| 1 = ∞
1
and
∞ 1 4 1
–3 –3 ∞
–
∫∞ |g(x) |2dx = ∫x
0
dx = – 3x | 1 = 3 < ∞.
Given two function f ∈ L1 (R) and g ∈ L1 (R) , the product fg is not necessarily in L1(R).
Counter-examples are given by
x– 3 ; 0 < x < 1
2
f(x) = g(x) = .
0 ; otherwise
There is a product * for which f *g ∈ L1(R) whenever f ∈ L1(R) and g ∈ L 1(R). We define
the convolution product
13. Fourier Transforms. 42
∞
(f*g)(x) =
–
∫∞ f(x – y) g(y) dy.
Lemma. If f ∈ L1 (R) and g ∈ L1 (R), then f *g ∈ L1 (R).
∞ ∞ ∞
Proof. ∫
–∞
|(f*g)(x) |dx = ∫ ∫ f(x – y) g(y) dydx
– ∞ – ∞
∞ ∞
≤
–
∫∞ ∫∞ |f(x – y) g(y) | dydx
–
∞ ∞
≤
–
∫∞ |f(x – y) | ∫∞ |g(y) | dydx
–
∞ ∞
≤
–
∫∞ |f(x) | dx ∫∞ |g(y) | dx –
< ∞.
¯
∞ π ∞
–iξ x + π
=–
–
∫∞ f(x)e ξ dx = –
–
∫∞ f x – ξ e –iξxdx.
Therefore
∞ ∞
π e –iξxdx
2|f^(ξ)| = ∫∞ f(x)e –iξxdx – ∫ f x –
– –∞ ξ
∞
π
≤
–
∫∞ f(x) – f x – ξ |e –iξx| dx
13. Fourier Transforms. 43
∞
π
≤
–
∫∞ f(x) – f x – ξ dx → 0 as |ξ | → ∞.
ε
To prove continuity of f^, let ε > 0 be given and a > 0 chosen such that ∫ |f(x) |dx < 4 and
|x| > a
ηx ηx
≤ 2 ∫ |f(x) | |sin ( 2 ) |dx
|x| > a
+ 2 ∫ |f(x) | |sin ( 2 ) |dx
|x| < a
ηx
≤ 2 ∫ |f(x) | dx
|x| > a
+ 2 ∫ |f(x) | | 2
|x| < a
|dx
ε ε ε
≤ 2 + aδ ∫ |f(x) |dx ≤ 2 + 2 = ε.
|x| < a
(τ f)h
^(ξ) =
–
∫∞ (τh)(x)e –iξxdx
∞
=
–
∫∞ f(x – h)e –iξxdx
∞
=
–
∫∞ f(x )e –iξ(x + h)dx
∞
–iξ h
= e
–
∫∞ f(x )e –iξx dx
= e –iξ hf^(ξ).
∞
=
–
∫∞ f(x )e –iξ(x – c )dx
= f^(ξ – c) = (τ f ^ )(ξ).
c
1
(δλ f)(x) = λ – 2f(λ –1 x), x ∈ R.
(δλ f) ^(ξ) =
–
∫∞ (δλ f)(x)e –iξxdx
∞
1
∫∞ f(λ –1 x) e –iξxdx
–2
= λ
–
∞
1
–iξλy
∫∞ f(y) e
–2
= λ λdy
–
1
^ ^
= λ 2 f ( λξ) = (δλ f )(ξ)
–1
13. Fourier Transforms. 45
1 ∂
3. Differentiation. Let f and f' ∈ L (R) and denote by D the differential operator D = ∂x .
The Fourier transform of Df is
∞
(Df)^(ξ) = ∫ (Df)(x)e –iξxdx
∞ –
∞
∞
= f(x)e –iξx| – ∫∞ f(x) (De –iξx)dx
–∞
–
= (iξ)f^(ξ).
∞
=
–
∫∞ (–ix)f(x)e –iξxdx.
Therefore (–ixf )^ = ∂f ^. By induction it follows that ∂ kf^ = ((–ix)kf)^ , k = 1, 2, ... .
1 1
5. Convolution. Let f, g ∈ L (R). Then f*g ∈ L (R) and has Fourier transform
∞
(f*g) ^(ξ) =
–
∫∞ (f*g)(x)e –iξxdx
∞ ∞
= ∫ ∫ f(x – y) g(y) dy e –iξxdx
– ∞ – ∞
∞ ∞
=
–
∫∞ ∫∞ f(x) g(y) e –iξ(x + y)dx dy
–
∞ ∞
= ∫ f(x)e –iξxdx ∫ g(y)e –iξydy
– ∞ – ∞
= f^(ξ)g^(ξ).
x2
–2
Example 1. Let f(x) = e . Then f ∈ L1 (R) and has Fourier transform
∞ x2
– 2 –i ξx
f^(ξ) =
–
∫∞ e e dx
13. Fourier Transforms. 46
∞ x2
∂f ^ ∂ – 2 –i ξx
∂ξ
(ξ ) =
∂ξ –
∫∞ e e dx
∞ x2
–2
= ∫∞ e (–ix)e –iξxdx
–
∞ 2
∂ – 2 –iξx
x
=i ∫
–∞
∂x e e dx
x2 ∞ x2
–
–iξxe 2 ∞ –2 ∂ –iξx)dx
= ie | –∞ – i ∫∞ e –
∂x (e
∞ x2
– 2 –i ξx
=– ξ
–
∫∞ e e dx
= – ξf^(ξ).
Therefore f^ satisfies a first order ordinary differential equation with solution f ^(ξ) =
ξ2
–2
ce .
∞ x2
2π ,
√
–2
Since f^(0) = c =
–
∫∞ e dx =
2 ξ2
– x2 ^
√2π
–2
e = e .
¯
1 ; |x| ≤ a
Example 2. Let f(x) = ≡ χa(x) ∈ L1 (R) and has Fourier transform
0 ; otherwise
∞ a
–iξxdx e –iξxdx
(χa)^(ξ) = ∫ χa(x) e = ∫
∞ – –a
=–
1 –iξx a
iξ
e | –a = –
1
iξ
(e –iξa – e –iξa)
2 sin a ξ
= .
ξ
14. Inverse Fourier Transforms. 47
14. Inverse Fourier Transforms.
∫a f^(ξ) e iξxdξ
1
fa(x) = 2π
–
a ∞
–iξydy e iξxdξ
∫a ∫ f(y)e
1
= 2π
– – ∞
∞ a
iξ(x – y )dξ f(y) dy
∫ ∫ e
1
= 2π
– ∞ – a
∞ ia(x – y ) –
e e –ia(x – y )
∫
1
= 2π
–∞
i(x – y) f(y) dy
∞ ∞
=
–
∫∞ Da(x – y) f(y) dy = ∫∞ Da(y) f(x – y) dy –
= (Da*f )(x)
sin ax
where Da (x ) = .
πx
Now
∞
1 sin ay
∫∞ f(x – y)
1 1
fa(x) – 2 [ f(x–) + f(x+ ) ]=π y dy – 2 [f(x ) + f(x ) ]
– +
–
∞
1 sin ay
=
π 0
∫ [f(x – y) – f(x–)] y dy
0 ∞
1 sin ay sin ay π
+
π ∫
–∞
[f(x – y) – f(x+)] y dy since
0
∫ y dy = 2
∞
1 sin ay
=
π ∫ [f(x – y) – f(x–)]
0
y dy
∞
1 sin ay
+
π ∫ [f(x + y) – f(x+)]
0
y dy
14. Inverse Fourier Transforms. 48
In the second integral,
∞
1 sin ay
π 0∫
[ f(x + y) – f(x+ )]
y dy
K ∞
1 sin ay 1 sin ay
=
π 0
∫ [f(x + y) – f(x+ ) ] y dy + π ∫K [f(x + y) – f(x+)] y dy
If K ≥ 1,
∞ sin ay ∞
∫ f(x + y)
y dy ≤ K∫ |f(x + y) |dy
K
and
∞ ∞
sin ay sin ay
∫
K
f(x + ) y dy = f(x ) K∫
+
y dy.
∞ ∞ ∞
sin ay sin ay
Since ∫ |f(x) |dx , 0∫
0
y dy are both convergent integrals,
K
∫ y dy → 0,
∞
K
∫ |f(x + y) |dy → 0 as K → ∞.
For the integrals over [0, K],
K ∞ iay –
sin ay e e – iay
0
∫ [f(x + y) – f(x+ ) ] y dy = ∫
–∞
2i g(y) dy
1
= 2i [g^(–a) – g ^(a) ]
f(x + y) – f(x+ )
;0<y<K
y
where g(y) = .
0 ; otherwise
Since f is piecewise smooth, f'(x+ ) exists for all x ∈ R and lim g(y) = f'(x+ ). Therefore
y →0 +
1
g is bounded on [0, K] and hence g ∈ L (R). By the Riemann-Lebesgue lemma, g ^ exists, is
continuous, g ^(±a) → 0 as a → ∞ and therefore
K
sin ay
∫0 [f(x + y) – f(x+)]
y dy → 0 as a →∞
for K ≥ 1. A virtually identical argument works for the first integral.
Therefore
|f (x) – 12[f(x–) + f(x+) ]|→ 0 as a → ∞
a
or
a
∫ f^(ξ) e iξxdξ
1
lim fa(x) = lim 2π
a→∞ a→∞ –a
∞
Theorem. Let f ∈ L1 (R) and let f be piecewise smooth on R. Then for every x ∈ R,
∫∞ φ(ξ) e iξxdξ
1 1
For φ ∈L (R), we call φ˘(x) = 2 π the inverse Fourier transform of φ and
–
∞
∫∞ φ(ξ) e iξxdξ
–1 –1 1
F defined by (F φ)(x) = 2π in called the inverse Fourier transformation.
–
1
If φ is the Fourier transform of a piecewise smooth function f ∈ L (R), that is φ = f ^, then
1
we define f(x) = 2 [f(x–) + f(x+ ) ]at a point of discontinuity of f.
∞
∫∞ f^(ξ) e iξxdξ.
1
Then f(x) = φ˘(x) = 2 π That is F –1 Ff = f.
–
describes the vertical vibrations of an infinite stretched elastic string, where u(x, t) is the
vertical displacement of the string from its rest position at position x, time t. Let the initial
displacement and velocity be given as
∂u
u(x, 0) = f(x),
∂t (x, t) = g(x), – ∞ < x < ∞.
We take Fourier transforms of the wave equation and the initial conditions with respect to
the x variable and denote by u^ (ξ, t ) the Fourier transform F (u(x, t) ). The using the
derivative properties of the Fourier Transform,
∂2u^
– (ci ξ)2u^ = 0, – ∞ < ξ < ∞, t > 0,
∂t2
or
∂2u^
+ (cξ)2u^ = 0.
∂t2
Then
u^ (ξ, t) = A( ξ) cos(c ξt) + B( ξ) sin(c ξt).
∂u^
When t = 0, u^ (ξ, t) = A( ξ) = f^(ξ), ∂t (ξ, t) = g^(ξ) = cξB(ξ). Therefore
g^(ξ) 2 sin(cξt)
u^ (ξ, t) = f^(ξ) cos(cξ t) + 2c
ξ
e i(ct) ξ + e – i(ct) ξ 1
= f^(ξ) + 2c (χ *g)^(ξ).
2 ct
∫∞
1 1
u(x, t) = 2 [f(x + ct ) + f(x – ct )] + 2c χ ct(x – ct) g(y) dy
–
∫
1 1
= 2 [f(x + ct ) + f(x – ct )] + 2c g(y) dy
|x – ct| < ct
x + ct
∫
1 1
= 2 [f(x + ct ) + f(x – ct )] + 2c g(y) dy.
x – ct
Consider Laplace's equation in two variables on the upper half-plane y > 0. Then
∂2u ∂2u
+ = 0, – ∞ < x < ∞, y > 0.
∂x 2 ∂y 2
be given for a function f ∈ L1 (R). Then taking the Fourier transform in the variable x,
∂2u^
(iξ)2u^ (ξ, y) + (ξ, y) = 0, – ∞ < x < ∞, y > 0.
∂y 2
or
∂2u^
(ξ, y) – ξ 2u^ (ξ, y) = 0,
∂y 2
which has solutions
We need two conditions to determne the functions A(ξ), B(ξ). In addition to u(x, 0) = f(x), –
∂u
∞ < x < ∞, let ∂y (x, 0) = g(x), – ∞ < x < ∞ for some function g ∈ L1 (R). We will find g
such that the solution u(x, y) is bounded for y > 0, in fact such that u(x, y) → 0 as y → ∞.
Taking transforms,
∂u^
u^ (ξ, 0) = f ^(ξ), ∂y (ξ, 0) = g (ξ),
^
or
A( ξ) + B(ξ) = f^(ξ), ξA( ξ) – ξB( ξ) = g^(ξ).
Solving for A(ξ), B(ξ),
1 g ^(ξ) g ^(ξ)
, B(ξ) = 2 f^(ξ) –
1
A( ξ) = 2 f^(ξ) + ξ
,
ξ
and
g ^(ξ) ξy 1 ^ g ^(ξ) – ξy
u^ (ξ, y) = 2 f^(ξ) +
1
ξ
ξ 2 ξ
e + f ( ) – e .
g ^(ξ)
For ξ > 0, u^ (ξ, y) → 0 as y → 0 if and only if f ^(ξ) + = 0, and for ξ < 0, u^ (ξ, y) → 0 as
ξ
g ^(ξ)
y → 0 if and only if f^(ξ) – = 0. Therefore
ξ
–ξf^(ξ) ; ξ > 0
g (ξ ) =
^
ξf^(ξ) ; ξ < 0
= – |ξ| f ^(ξ).
Hence
15. Applications to Differential Equations. 52
f^(ξ)e – ξy ; ξ > 0
u^ (ξ, y) =
f^(ξ)e ξy ; ξ < 0
= f^(ξ)e – ξ|y|.
y ^
Since e – ξ|y| = , by the convolution theorem,
2
π (x + y )
2
∞
y y
u(x, y) =
*f = ∫∞ f(s) ds.
π (x + y ) π ((x – s) 2 + y 2)
2 2
–
for f ∈ L1 (R), f continuous and bounded, can be solved using Fourier transforms. Taking
transforms in the variable x,
∂u^
∂t (ξ, t) – κ (i ξ) u (ξ, t) = 0,
2 ^
or
∂u^
∂t (ξ, t) + κ ξ u (ξ, t) = 0.
2 ^
2 ξ2
– x2 ^
2π
√
–2
Now e = e and using the dilation property of Fourier transforms
1 (λ x) 2^ (λ ξ) 2
–1
λ – 2e – 2 = 1
2π
√
– 2
λ e
2
(λ ξ) 2 (λ x) 2^
–1
– 2 1 e – 2 .
e =
λ√
2π
λ2
Let 2 = κ t or λ = √
2κt , then
15. Applications to Differential Equations. 53
1 x 2 ^
1 1 – 2
√
2
e – κξ t = e 2 κt
2π √
√
2κt
1 x 2 ^
1 – 2
= e
√ 2 κt .
√
4πκt
1 x
1 – 2
2
u(x, t) = f* e √ 2 κt (x)
√
4πκt
∞ (x – y )2
1 –
4 κ t f(y)
= ∫∞ e dy.
√ 4πκt –
x2
–
4κt
e
The function h(x, t) = , is called the heat kernel. We can then write
√ 4πκt
∞
u(x, t) =
–
∫∞ h(x – y, t) f(y) dy.
16. Plancherels' and Parsevals' Identities. 54
16. Plancherels' and Parsevals' Identities.
1 2
By the convolution theorem, for f, g ∈ L (R)∩L (R),
(f*g)^ = f^ g^.
Therefore
f *g = (f^ g^)˘
or
∞ ∞
1
∫
–∞
f(x – y) g(y) dy =
2 π ∫
–∞
e i ξx f ^(ξ) g ^(ξ) d ξ.
Set x = 0,
∞ ∞
1
∫
–∞
f(– y) g(y) dy =
2π – ∫∞
f^(ξ) g ^(ξ) d ξ
Replacing g(x) by g(–x) , the Fourier transform g^(ξ) is replaced by g^(ξ) , hence
∞ ∞
1
∫
–∞
f(– y) g(–y) dy =
2π – ∫∞
f^(ξ) g^(ξ) dξ
or
∞ ∞
1
∫
–∞
f( y) g(y) dy =
2 π ∫
–∞
f^(ξ) g^(ξ) dξ .
Examples.
1; |x| < a
1. Let f(x) = χ a(x) = .
0; otherwise
2 sin a ξ
Then f^(ξ) = , and by Parsevals' identity,
ξ
∞ ∞
1 sin a ξ2
–∞
∫ |χa(x) | dx = 2π – ∫∞ ξ dξ
2
∞
2 sin a ξ dξ = 4 πa
2
–
∫∞
ξ
or
∞
sin a ξ dξ = π a.
2
–
∫∞ ξ
16. Plancherels' and Parsevals' Identities. 55
2a
2. Let f(x) = e – a|x|, f^(ξ) = and by Parsevals' identity
a2 + ξ2
∞ ∞
1 2a 2
–
∫(
∞
e – a|x| ) 2dx =
2π ∫ 2 2 dξ
–∞ a + ξ
∞
1 2a2 1
a
=
π –
∫∞ (a2 + ξ 2)2
dξ
or
∞
1 π
–
∫∞ (a2 + ξ 2)2
d ξ =
2a3
.
17. Band Limited Functions and Shannon's Sampling Theorem. 56
17. Band Limited Functions and Shannon's Sampling Theorem.
The Fourier transform variable has the role of frequency and f ^(ξ) is referred to as the
frequency representation of f(x).
If f ^(ξ) = 0 for |ξ | > ξc > 0, then f(x) is called a band-limited function and ξ c is called the
cut-off frequency. Many functions from science and technology, are band-limited. For
example, human hearing is assumed to be limited to frequencies below about 20 kHz.
Therefore the acoustic signals recorded on compact discs are limited to a bandwidth of 22
kHz.
^
Consider the functions φ (ξ) given by
n
0; |ξ | > ξ c
φ (ξ ) =
^ inπξ
– ξ
2ξ e ; |ξ| ≤ ξ
n 1 c
√
c
c
^
The inverse Fourier transforms of φ n (ξ) are given by
∞
1 i xξ ^
φn(x) =
2π –
∫∞ e φ n (ξ ) d ξ
ξc inπξ
– ξ
i xξ 1
2π φ n(x) = ∫e dξ
c
e
– ξ
c
√ 2 ξc
ξc nπ
i ξ x –
1 ξc
= ∫ e dξ
√2ξc – ξ c
i ξx – nξπ ξ
1 e
c
c
e i (ξcx – n π) – e – i (ξcx – n π)
= √2ξc
2i(ξcx – nπ )
sin (ξcx – nπ )
=
√ 2 ξc
ξcx – nπ
17. Band Limited Functions and Shannon's Sampling Theorem. 57
=
√ 2π sin ξc(x – nL)
L ξc(x – nL)
where L =
π
ξc
.
< ^ ^
φ n ; φm > =
–
∫∞
^ ^
φ n (ξ) φm(ξ) d ξ
ξc inπξ – imπξ
– ξ
1 ξc
∫ dξ
c
= e e
2 ξc – ξc
ξc i(m – n) πξ
1 ξc
=
2 ξc ∫ e dξ
– ξ c
0; m ≠ n
= .
1; m = n
^
So the functions {φ (ξ) ; n = 0, ±1, ±2, ... }, form an orthogonal set in L2 (– ξc, ξc). Since f^(ξ)
n
is band limited to |ξ | ≤ ξc, it has a Fourier series
∞ ^
f^(ξ) = ∑ cn φn (ξ)
n=– ∞
inπξ
∞ – ξ
1
∑ dξ
c
= cn e
n=– ∞
√ 2 ξc
ξc inπξ
ξc
cn = < > ^ ^
f ; φn =
–
∫
ξ c
f^(ξ)
√2ξc
1
e dξ .
By Plancherel's theorem
cn = < > ^ ^
f ; φn = 2π <f ; φ > n
so
∞ ∞
^ ^
f^(ξ) = ∑ cn φ n (ξ) = ∑ 2π <f ; φ >φ (ξ).
n n
n=– ∞ n=– ∞
>√
∞ 2π sin ξc(x – nL)
= ∑ < f ; φn L ξc(x – nL)
.
n=– ∞
17. Band Limited Functions and Shannon's Sampling Theorem. 58
π
The samples at intervals of length L = are
ξc
>√
∞ 2π sin ξc(kL – nL)
f(kL) = ∑ < f ; φn L ξc(kL – nL)
n=– ∞
>√
∞ 2π sin(k – n)π
= ∑ < f ; φn L (k – n)π
n=– ∞
<f ; φ >√
L 2π
= .
k
So
√
L
< f ; φk> = f(kL) , k = 0, ±1, ±2, ... .
2π
Finally therefore
>√
∞ 2π sin ξc(x – nL)
f(x) = ∑ < f ; φn L ξc(x – nL)
n=– ∞
The relationship ωL = 2π , where ω is the frequency of sampling in cycles per unit length,
shows that from Shannon's theorem, to reconstruct a band-limited function, it suffices to
2π
sample at a frequency ω = L = 2 ξc, twice the cut-off frequency.
18. HeisenbergÕs Inequality. 59
18. Heisenberg’s Inequality.
2 2
Let f ∈ L (R), xf ∈ L (R). Then the quantity
∫ (x – a) |f(x)| dx
∞
2 2
∆ f ≡
– ∞
a ∞
∫
|f(x) | dx 2
– ∞
is called the dispersion about the point x = a of f. The reasoning behind the definition is
that if f(x) is concentrated near x = a , then ∆af is smaller than when f is not close to zero
far from x = a.
1 ; |x| ≤ b
χb(x) =
0 ; otherwise
Notice that χb is concentrated near x = 0 for small b. The dispersion about the origin is
∫ x dx b
b
2
2
=
∆0 χb = 3
– b
b
∫ dx – b
^
This indicates that χ b is spread out away from x = 0.
The following result shows that there is a type of inverse relationship between the
dispersion of a function and that of its Fourier transform.
(∆ f )(∆ f^ ) =
a ^
α ≥4
1 – ∞
∞
– ∞
∞
∫ |f(x)| dx ∫ |f (ξ)| dξ 2 2
– ∞ – ∞
2
and equality holds if and only if f(x) = c e – kx for constants c ∈ R and k > 0.
∞ ∞
Proof. We firstly prove the result for a = α = 0.
–
∫∞ x2 |f(x) |2 dx and
–
∫∞ ξ2|f^(ξ)|2dξ are
both assumed finite since otherwise the result is trivial.
Let f *(x) ≡ ((i ξf^(ξ)) ˘ or (f*) ^(ξ) = iξf^(ξ). Then f* ∈ L2(R), and
∞ ∞ ∞ ∞
∫ x2 |f(x) |2 dx ∫ ξ 2|f^(ξ)|2dξ = ∫ x2 |f(x) |2 dx ∫ |(iξ)f^(ξ)| 2dξ
– ∞ – ∞ – ∞ – ∞
∞ ∞
= ∫ x2 |f(x) |2 dx ∫ |(f*) ^(ξ)| 2dξ
– ∞ – ∞
∞ ∞
= 2π ∫ x2 |f(x) |2 dx ∫ |f*(x) |2 dx (by Parseval’s identity).
– ∞ – ∞
Since
∞ ∞
2 2
∫∞ = ∫ x Re
1
x ( f*(x) f(x) + f*(x) f(x) ) dx ( f*(x) f(x) ) dx
–
2
– ∞
∞ ∞
2 2
= Re
–
∫∞ x f (x) f*(x) dx ≤ ∫ x f (x) f*(x) dx
– ∞
∞ ∞
≤ ∫ x2 |f(x) |2 dx ∫ |f*(x) |2 dx
– ∞ – ∞
(by the Cauchy-Schwartz inequality).
∞
2
∫∞
1
≥ 2π x ( f*(x) f(x) + f*(x) f(x) ) dx
–
2
18. HeisenbergÕs Inequality. 61
∞ ∞ ∞
π 2 π 1
= 2 ∫ |f(x) |2 dx = 2 ∫ |f(x) |2 dx 2π ∫∞ |f^(ξ)| 2dξ
– ∞ – ∞ –
∞ ∞
∫ |f(x) | dx ∫∞ |f^(ξ)| 2dξ .
1 2
= 4 – ∞ –
To complete the proof, we assume that f is continuous and piecewise smooth, This
assumption can be removed since functions in L 1 (R) are the uniform limit of such
functions.
f*(x) = f'(x)
As for the case of equality in Heisenberg’s inequality, this holds if and only if f(x) f*(x) is
real and f*(x) = K x f(x) for some complex constant K. That is,
Therefore K is real.
Kx2
– 2
f(x) = c e , c any real constant,
18. HeisenbergÕs Inequality. 62
Kx2
–
and f(x) = c e 2 ∈ L2(R) if and only if K > 0. Therefore equality holds in Heisenberg’s
– kx2
inequality only if f(x) = c e for constants c ∈ R and k > 0.
Kx2
– 2
Conversely, let f(x) = e for constant K > 0. Then
∞ ∞ ∞ ∞
2
∫ x |f(x) | dx ∫ ξ |f^(ξ)| dξ = 2π ∫ x |f(x) | dx ∫ |f*(x) | dx
2 2 2 2 2 2
– ∞ – ∞ – ∞ – ∞
∞ ∞
= 2π ∫ x2 |f(x) |2 dx ∫ |K x f(x) |2 dx
– ∞ – ∞
∞
2
= 2π K 2 ∫ x2 |f(x) |2 dx
– ∞
∞
2 – Kx dx
2 2
= 2π K2 ∫x e
– ∞
∞
x
= 2π K2 ∫ –2K
– ∞
(–2Kxe – Kx 2
) dx 2
∞
x
= 2π K 2 ∫ –2K
– ∞
(–2Kxe – Kx 2
) dx 2
∞
x d
= 2π K2 ∫ –2K
– ∞ dx
(e – Kx 2
) dx 2
∞
1 – Kx 2 2
= 2π K2 ∫ 2K e dx (integration by parts)
– ∞
∞
π – Kx 2 2
=2 ∫ e dx
– ∞
∞
π –t2 2 π2
= 2K ∫ e dt = 2K .
– ∞
Whereas,
∞ ∞ ∞ ∞
2 1
∫ |f(x) | dx ∫ |f ^(ξ)| dξ = 4 ∫ |f(x) | dx 2π ∫ |f(x) | dx
1 2 2 2
4 – ∞ – ∞ – ∞ – ∞
∞
π 2
= 2 ∫ |f(x) |2 dx
– ∞
18. HeisenbergÕs Inequality. 63
∞
π – Kx 2 2
=2 ∫ e dx
– ∞
The case of a ≠ 0, α ≠ 0, follows by observing that F(x) = e –iα x f(x + a) satisfies the same
hypotheses as f(x) and ∆a f = ∆ 0F and ∆α f^= ∆0 F^ for any a ≠ 0, α ≠ 0.