FS Notes 1 PDF
FS Notes 1 PDF
M.T.Nair
Department of Mathematics, IIT Madras
Contents
1 Fourier Series 2
1.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
The consideration of Fourier Series can be traced back to the situation which Fourier encountered in
the beginning of last century while solving heat equation:
Consider a thin metallic wire of length `. Suppose an initial temperature is supplied to it, and sup-
pose the temperature at both the end points kept at 0. Then one would like to know the temperature
at each point of the string at a particular time.
Let us represent the string as an interval [0, `]. Let u(x, t) be the temperature at the point x ∈ [0, `]
at time t. It is known that u(·, ·) satisfies the partial equation
∂u ∂2u
= c2 2 , 0 < x < `, (1.1)
∂t ∂t
where c > 0 is the heat conductivity of the material. Since the temperature at both the end points
kept at 0, we have
u(0, t) = 0 = u(`, t), t > 0. (1.2)
Exercise 1.1. Equation (1.1) satisfying (1.2) and (1.3) cannot have more than one solution. ♦
In order to find u(x, t), we use a procedure called method of separation of variables. In this method,
we assume first that u(x, t) is of the form:
u(x, t) = φ(x)ψ(t).
Then we have
∂u ∂2u
= φ(x)ψ 0 (t), = φ00 (x)ψ(t).
∂t ∂t2
Hence, from (1.1),
φ(x)ψ 0 (t) = c2 φ00 (x)ψ(t).
Hence,
ψ 0 (t) φ00 (x)
= = K, const.
c2 ψ(t) φ(x)
Hence,
ψ 0 (t) = Kc2 ψ(t), φ00 (x) = Kφ(x). (1.4)
φ(x) = ax + b.
2
By (1.2), φ(0)ψ(t) = 0 = φ(`)ψ(t).
3
satisfies (1.1), (1.2) and (1.3).
What can we say if f is, in some sense, arbitrary? The consideration of the functions of the form
in (1.5) suggests the following query:
∞
X
If f is of the form f (x) = an sin(nπx/`), can we say that
n=1
∞
X 2 2
u(x, t) := an e−λn c t sin(nπx/`)
n=1
is a solution of (1.1) satisfying (1.2) and (1.3) with appropriate notion of convergence?
Assume further that, term by term integration of the above series is possible. Then, we have
Z ` ∞ Z `
mπx X nπx mπx
f (x) sin( )dx = an sin( ) sin( )dx.
0 ` n=1 0 ` `
Since Z `
nπx mπx
sin( ) sin( )dx = 0 for m 6= n,
0 ` `
and
` `
1 − cos 2( nπx
` )
Z Z
nπx `
sin2 ( )dx = dx = ,
0 ` 0 2 2
we obtain Z `
2 nπx
an = f (x) sin( )dx. (1.7)
` 0 `
Note that, if f has the form as in (1.6), then f is 2`-periodic, i.e.,
Assuming that the above series is convergent and can be differentiated term by term, we see that
u(·, ·) is a solution of (1.1) satisfying (1.2) and (1.3).
d2
Exercise 1.2. Show that each λn := nπ/` is an eigenvalue of the operator wit corresponding
dx2
eigenvector sin λn x. ♦
4
1.2 Fourier Series of 2π-Periodic functions
The above series is a special case of the Fourier series that we are going to introduce. Let us consider
a few definitions.
converges at a point x ∈ R, then it has to converge at x + 2π as well; and hence at x + 2nπ for all
integers n. This shows that we can restrict the discussion of convergence of a trigonometric series
to an interval of length 2π. Hence, we cannot expect to have a trigonometric series expansion for a
function f : R → R if it is not a 2π-periodic function.
We know that a convergent trigonometric series is 2π-periodic. What about the converse?
5
Suppose, for a moment, that we can write
∞
X
f (x) = c0 + (an cos nx + bn sin nx)
n=1
for all x ∈ R. Then what should be the coefficients c0 , an , bn ? To answer this question, let us further
assume that
For instance if the above series is uniformly convergent to f in [−π, π], then term by term integration
is possible. By Weierstrass test,we have the following result:
P∞ P∞
If n=0 (|an | + |bn |) converges, then c0 + n=1 (an cos nx + bn sin nx) is a dominated series
on R and hence it is uniformly convergent.
Thus, under the assumption that f is integrable on [−π, π] and the series can be integrated term by
term, we obtain Z π
1
c0 = f (x)dx,
2π −π
Z π Z π
1 1
an = f (x) cos nxdx, bn = f (x) sin nxdx.
π −π π −π
Definition 1.4. The Fourier series of a 2π-periodic function f is the trigonometric series
∞
a0 X
+ (an cos nx + bn sin nx) ,
2 n=1
Z π Z π
1 1
where an = f (x) cos nxdx and bn = f (x) sin nxdx and this fact is written as
π −π π −π
∞
a0 X
f (x) ∼ + (an cos nx + bn sin nx) .
2 n=1
6
If f is a trigonometric polynomial, then its Fourier series is itself.
Writing
1 inx 1 inx
cos nx = [e + e−inx ], sin nx = [e − e−inx ],
2 2i
we have
an inx bn
an cos nx + bn sin nx = [e + e−inx ] + [einx − e−inx ]
2 2i
an bn an bn
= + e inx
+ − e−inx .
2 2i 2 2i
Thus, writing
an bn an bn
cn := + , c−n := − ,
2 2i 2 2i
we have
∞
a0 X X
+ (an cos nx + bn sin nx) = cn einx .
2 n=1 n∈Z
inx
P
Now, suppose f (x) = n∈Z cn e with cn ∈ C, and this series can be integrated term by term. Then,
we have Z π Z π
X
f (x)e−imx dx = cn ei(n−m)x dx.
−π n∈Z −π
But, (
Z π
i(n−m) 0 if m = m,
e dx =
−π 2π if m 6= m.
Z π
Hence, ei(n−m) dx = 2πcm , i.e.,
−π
Z π
1
cn = f (x)e−inx dx, n ∈ Z.
2π −2π
The following theorem show that there is a large class of functions which can be represented by their
Fourier series (see Bhatia [1]). We shall come back to this theorem at a later stage.
In Theorem 1.5 we used the terminology piecewise differentiable as per the following definition.
7
Remark 1.7. It is known that there are continuous functions f defined on [−π, π] whose Fourier
series does not converge pointwise to f . Its proof relies on UBP (see [2]). We shall consider this at a
later occasion. ♦
Although each term and the partial sums of a Fourier series are infinitely differentiable, the sum
function need not be even continuous at certain points. This fact is illustrated by the following
example.
(
0, −π ≤ x ≤ 0,
Example 1.8. Let f (x) = By Dirichlet’s theorem (Theorem 1.5), the Fourier
1, 0 < x ≤ π.
series of f converges to f (x) for every x 6= 0, and at the point 0, the series converges to 1/2. Note
that (
Z π
1 1, n = 0,
an = cos nxdx =
π 0 0, n 6= 0,
and for n ∈ N,
2
π n
, n odd,
1 1 − cos nπ 1 1 − (−1) πn
Z
1
bn = sin nxdx = = =
π 0 π n π n
0, n even.
f (−x) = f (x) ∀ x ∈ X.
Then f (x) cos nx is an even function and f (x) sin nx is an odd function, so that
1 π 2 π
Z Z
an = f (x) cos nxdx = f (x) cos nxdx,
π −π π 0
1 π
Z
bn = f (x) sin nxdx.
π −π
8
• Suppose f is an odd function, i.e.,
f (−x) = −f (x) ∀ x ∈ X.
Then f (x) cos nx is an odd function and f (x) sin nx is an even function, so that
1 π
Z
an = f (x) cos nxdx = 0,
π −π
1 π 2 π
Z Z
bn = f (x) sin nxdx = f (x) sin nxdx.
π −π π 0
In particular,
∞ ∞
a0 X a0 X
f (0) = + an , f (π) = + (−1)n an ,
2 n=0
2 n=1
respectively.
In particular,
∞
X
f (π/2) = (−1)n b2n+1 .
n=0
with Z π
2
a0 = x dx = π
π 0
and for n = 1, 2, . . .,
π Z π
2 π
Z
2 sin nx sin nx
an = x cos nxdx = x − dx
π 0 π n 0 0 n
n
2 cos nx
h i π 2 (−1) − 1
= =
π n2 0 π n2
9
Thus,
−4
a2n = 0, a2n+1 = , n = 1, 2, . . .
π(2n + 1)2
so that
∞
π 4 X cos(2n + 1)x
|x| ∼ − , x ∈ [−π, π].
2 π n=0 (2n + 1)2
Example 1.10. Let f (x) = x, x ∈ [−π, π]. In this case, f is an odd function. Hence, the Fourier
series is
∞
X
bn sin nx, x ∈ [−π, π]
n=1
with
2 π
Z Z π
2 h cos nx iπ cos nx
bn = x sin nx dx = −x + dx
π 0 π n 0 0 n
2n cos nπ o (−1)n+1 2
= −π = .
π n n
Thus the Fourier series is
∞
X (−1)n+1
2 sin nx.
n=1
n
In particular (using Dirichlet’s theorem), with x = π/2 we obtain the Madhava-Nīlakaṅtha series
∞ ∞
π X (−1)n+1 nπ X (−1)n
= sin = . ♦
4 n=1
n 2 n=0
2n + 1
(
−1, −π ≤ x < 0,
Example 1.11. Let f (x) = In this case, f is an odd function. Hence, the
1, 0 ≤ x ≤ π.
Fourier series is
∞
X
bn sin nx,
n=1
with Z π
2 2 2
bn = sin nx dx = (1 − cos nπ) = [1 − (−1)n ].
π 0 π π
Thus
∞
4 X sin(2n + 1)x
f (x) ∼ .
π n=0 2n + 1
10
Example 1.12. Let f (x) = x2 , x ∈ [−π, π]. Since f is an even function, its Fourier series is
∞
2 π 2
Z
a0 X
+ an cos nx, x ∈ [−π, π], an = x cos nx dx.
2 n=1
π 0
respectively. ♦
Suppose a function f is defined on [0, π]. By extending it to [−π, π] so that the extended function is
an odd function, we obtain Forier sine series of f , and by extending it to [−π, π] so that the extended
function is an even function, we obtain Fourier cosine series of f .
The odd extension and even extension of f , denoted by fodd and feven are defined by
(
f (x) if 0 ≤ x < π,
fodd (x) = ,
−f (−x) if − π ≤ x < 0,
(
f (x) if 0 ≤ x < π,
feven (x) =
f (−x) if − π ≤ x < 0,
respectively. Therefore,
∞
X
f (x) = fodd (x) ∼ bn sin nx, x ∈ [0, π]
n=1
and
∞
a0 X
f (x) = feven (x) ∼ + an cos nx, x ∈ [0, π]
2 n=1
with Z π Z π
2 2
an = f (x) cos nx dx, bn = f (x) sin nx dx.
π 0 π 0
Example 1.13. Let f (x) = x2 , x ∈ [0, π]. The even extension of f is itself. Its odd extension is:
(
x2 , if 0 ≤ x < π,
fodd (x) = 2
,
−x , if − π ≤ x < 0.
Hence,
∞
X
f (x) = fodd (x) ∼ bn sin nx, x ∈ [0, π],
n=1
11
with Z π h iπ Z π
2 2 2 2 cos nx cos nx
bn = x sin nx dx = −x + 2x dx .
π 0 π n 0 0 n
Note that h cos nx iπ cos nπ (−1)n+1
−x2 = −π 2 = π2 ,
n 0 n n
Z π π Z π
(−1)n − 1
h cos nx iπ
cos nx sin nx sin nx
2x dx = 2x − 2 dx = 2 =2 .
0 n n 0 0 n n2 0 n2
Thus,
n+1
(−1)n − 1 (−1)n+1 4 (−1)n − 1
2 2 (−1)
bn = π +2 = 2π + .
π n n2 n π n2
♦
Example 1.14. Let f (x) = x, x ∈ [0, π]. Its odd extension is itself, and
and
∞
π 4 X cos(2n + 1)x
x∼ − , x ∈ [0, π]. ♦
2 π n=0 (2n + 1)2
Example 1.15. Let us find the sine series expansion of the function
(
0, if 0 ≤ x < π/2,
f (x) =
1, if π/2 ≤ x < π.
The sine series of f is given by
∞
X
f (x) ∼ bn sin nx, x ∈ [0, π],
n=1
where π
2 cos nπ/2 − cos nπ
Z
2 2 h cos nx iπ
bn = sin nx dx = − = .
π π/2 π n π/2 π n
2
Note that b2n−1 = and
(2n − 1)π
(
2
2 − nπ if n odd,
b2n = [(−1)n − 1] =
2nπ 0 if n even.
Thus, for x ∈ [0, π], we have
π sin x sin 2x sin 3x sin 5x sin(4n − 3)x
f (x) ∼ − + + + ··· +
2 1 1 3 5 4n − 3
sin(4n − 2)x sin(4n − 1)x sin(4n + 1)x
− + + + ··· .
4n − 2 4n − 1 4n + 1
♦
12
1.5 Fourier Series of 2`-Periodic Functions
Suppose f is a T -periodic function. We may write T = 2`. Then we may consider the change of
variable t = πx/` so that the function
f (x) := f (`t/π),
where Z π Z `
1 `t 1 nπx
an = f cos ntdt = f (x) cos dx,
π −π π ` −` `
Z π Z `
1 `t 1 nπx
bn = f sin ntdt = f (x) sin dx.
π −π π ` −` `
In particular,
Z `
2 nπx
f even =⇒ bn = 0 and an = f (x) cos dx,
` 0 `
Z `
2 nπx
f odd =⇒ an = 0 and bn = f (x) sin dx.
` 0 `
and Z 1
bn = (1 − |x|) sin nπx dx = 0.
−1
Now,
Z 1 1
sin nπx
cos nπx dx = = 0,
0 n 0
Z 1 1 Z 1
(−1)n − 1
h cos nπx i1
sin nπx sin nπx
x cos nπx dx = x − dx = = .
0 n 0 0 n n 0 n
Hence, (
Z 1
2 0, neven,
an = 2 (1 − |x|) cos nπx dx = [1 − (−1)n ] =
0 n 4/n, n odd.
Thus,
∞
X 4
f (x) ∼ cos nπx. ♦
n=0
2n +1
13
1.6 Fourier Series on Arbitrary Intervals
Suppose a function f is defined in an interval [a, b]. We can obtain Fourier expansion of it on [a, b] as
follows:
a+b
Method 1: Let us consider a change of variable as y = x − 2 . Let
a+b
ϕ(y) := f (x) = f (y + ), where −`≤y ≤`
2
with ` = (b − a)/2. We can extend ϕ as a 2`-periodic function and obtain its Fourier series as
∞
a0 X nπ nπ
ϕ(y) ∼ + an cos y + bn sin y
2 n=1
` `
where Z ` Z `
1 nπy 1 nπy
an = ϕ(y) cos dy, bn = ϕ(y) sin dy.
` −` ` ` −` `
Thus,
∞
a0 X nπ nπ
f (x) ∼ + an cos y + bn sin y
2 n=1
` `
where Z ` Z `
1 nπy 1 nπy
an = f (x) cos dx, bn = f (x) sin dx
` −` ` ` −` `
a+b
with ` = (b − a)/2 and y = x − 2 .
(a) For y ∈ [−`, 0], define f˜e (y) = ϕf (−y). Thus f˜e on [−`, `] is an even function. In this case,
∞
a0 X nπ
ϕ(y) ∼ + an cos y
2 n=1
`
(b) For y ∈ [−`, 0], define f˜o (y) = −ϕ(−y). Thus f˜o on [−`, `] is an odd function. In this case,
∞
X nπ
ϕ(y) ∼ bn sin y
n=1
`
where Z `
1 nπy
bn = ϕ(y) sin dy.
` −` `
From the series of ϕ we can recover the corresponding series of f on [a, b] by writing y = x − a.
14
1.7 Exercises
1. Find the Fourier series of the 2π- period function f such that:
(
1, −π 2 ≤x< 2
π
(a) f (x) = π 3π
0, 2 < x < 2 .
(
−π π
x, 2 ≤x< 2
(b) f (x) =
π − x, π2 < x < 3π 2 .
(
1 + 2x
π , −π ≤ x ≤ 0
(c) f (x) = 2x
1− π , 0 ≤ x ≤ π.
x2
(d) f (x) = 4 , −π ≤ x ≤ π.
2. Using the Fourier series in Exercise 1, find the sum of the following series:
1 1 1 1 1 1
(a) 1 − + − + . . ., (b) 1 + + + + . . ..
3 5 7 4 9 16
1 1 1 1 1 1
(c) 1 − + − + . . ., (d) 1 + 2 + 2 + 2 + . . ..
4 9 16 3 5 7
(
π
sin x, 0 ≤ x ≤ 4
3. If f (x) = , then show that
cos x, π4 ≤ x < π2
8 π sin x sin 3x sin 10x
f (x) ∼ cos + + + ... .
π 4 1.3 5.7 9.11
15
8. Assuming that the Fourier series of f converges uniformly on [−π, π), show that
π ∞
a20 X 2
Z
1
[f (x)]2 dx = + (an + b2n ).
π −π 2 n=1
10. Write down the Fourier series of f (x) = x for x ∈ [1, 2) so that it converges to 1/2 at x = 1.
References
[2] Nair, M. T. (2002) (Fourth Print: 2014): Functional Analysis: A First Course, New Delhi:
Printice-Hall of India.
[3] Nair, M. T. (2014): Calculus of One Variable, Ane Books, Pvt Ltd.
[4] M.T. Nair, Measure and Integration, Notes for the MSc. Course, Jan-may, 2014.
[5] R. Radha and S. Thangavelu (2013): Fourier Analysis, NPTEL Notes, IIT Madras.
16