4.

1 Different Notions of Convergence

4.1.1 Mean Square Convergence

Let {fm }∞m=1 be sequence of functions defined on [a, b]. Let f be defined on [a, b]. We say
that the sequence {fm }∞
m=1 converges in the mean square sense to f on [a, b] if
Z b
lim |f (x) − fm (x)|2 dx = 0
m→∞ a

4.1.2 Pointwise Convergence

Let {fm }∞ m=1 be sequence of functions defined on [a, b] and let f be defined on [a, b].
We say that {fm }∞ m=1 converges pointwise to f on [a, b] if for each x ∈ [a, b] we have
limm→∞ fm (x) = f (x). That is, for each x ∈ [a, b] and ε > 0 there is a natural number
N (ε, x) such that

|fn (x) − f (x)| < ε for all n ≥ N (ε, x)

4.1.3 Uniform Convergence

Let {fm }∞
m=1 be sequence of functions defined on [a, b] and let f be defined on [a, b]. We
say that {fm }∞
m=1 converges uniformly to f on [a, b] if for each ε > 0 there is a natural
number N (ε) such that

|fn (x) − f (x)| < ε for all n ≥ N (ε), and for all x ∈ [a, b]

There is one more interesting fact about the uniform convergence. If {fm }∞ m=1 is a se-
quence of continuous functions which converge uniformly to a function to f on [a, b], then
f is continuous.

4.1.4 Example 1

Let un = xn on [0, 1). Clearly, the sequence {un }∞

n=1 converges pointwise to 0, that is, for
fixed x ∈ [0, 1) we have lim un = 0. But it does not converge uniformly to 0 as we shall
n→∞
 Lecture Notes on Fourier Series

show that for given ε there does not exist a natural number N independent of x such that
|un − 0| < ε. Suppose that the series converges uniformly, then for a given ε with

|un − 0| < ε, (4.1)

we seek for a natural number N (ε) such that relation (4.1) holds for n > N . Note that
relation (4.1) holds true if
ln ε
xn < ε ⇐⇒ n >
ln x
It should be evident now that for given x and ε one can define
 
ln ε
N := , where [ ] gives integer rounded towards infinity
ln x

It once again confirms pointwise convergence. However if x is not fixed then ln ε/ ln x

grows without bounds for x ∈ [0, 1). Hence it is not possible to find N which depends only
on ε and therefore the sequence un does not converge uniformly to 0.

4.1.5 Example 2

xn
Let un = on [0, 1). This sequence converges uniformly and of course pointwise to 0.
n
For given ε > 0 take n > N := 1ε then noting 1ε > 1ε we have |un − 0| < xn /n < 1/n < ε
   

for all n > N Hence the sequence un converges uniformly.

Now we discuss these three types of convergence for the Fourier series of a function.

• Let f be a piecewise continuous function on [−π, π] then the Fourier series of f

convergence to f in the mean square sense. That is
Z π  ha m
X i 2
0
lim f (x) − + (ak cos kx + bk sin kx)  dx = 0
 
m→∞ −π 2
k=1

• Let f be a piecewise continuous function on [−π, π] and the appropriate one sided
derivatives of f at each point in [−π, π] exists then for each x ∈ [−π, π] the Fourier
series of f converges pointwise to the value (f (x−) + f (x+))/2.
• If f is continuous on [−π, π], f (−π) = f (π), and f 0 is piecewise continuous on
[−π, π], then the Fourier series of f converges uniformly (and also absolutely) to f
on [−π, π].

2
 Lecture Notes on Fourier Series

4.2 Best Trigonometric Polynomial Approximation

An interesting property of the partial sums of a Fourier series is that among all trigonomet-
ric polynomials of degree N , the partial sum of Fourier Series yield the best approximation
of f in the mean square sense. This result has been summarized in the following lemma.

4.2.1 Lemma

Let f be piecewise continuous function on [−π, π] and let the mean square error is defined
by the following function

Z π  hc N
X i 2
0
E(c0 , . . . , cN , d1 , . . . , dN ) = f − + (ck cos kx + dk sin kx)  dx
 
−π 2
k=1

then E(a0 , . . . , aN , b1 , . . . , bN ) ≤ E(c0 , . . . , cN , d1 , . . . , dN ) for any real numbers c0 , c1 , . . . , cN

and d1 , d2 , . . . , dN . Note that ak and bk are the Fourier coefficients of f .

4.3 Example Problems

4.3.1 Problem 1

Let the function f (x) be defined as

(
−π, −π < x < 0;
f (x) =
x, 0 < x < π.

Find the sum of the Fourier series for all point in [−π, π].
Solution: At x = 0, the Fourier series will converge to
f (0+) + f (0−) 0 + (−π) π
= =−
2 2 2
Again, x = ±π are another points of discontinuity and the value of the series at these point
will be
f (π−) + f ((−π)+) π + (−π)
= = 0;
2 2
At all other points the series will converge to functional value f (x).

3
 Lecture Notes on Fourier Series

4.3.2 Problem 2

Let the Fourier series of the function f (x) = x + x2 , −π < x < π be given by
∞
π2 X 4 2
h i
2
x+x ∼ + (−1)n 2 cos nx − sin nx
3 n n
n=1

Find the sum of the Fourier series for all point in [−π, π]. Applying the result on conver-
gence of the Fourier series find the value of
1 1 1 1 1 1
1+ 2
+ 2 + 2 + ... and 1− 2
+ 2 − 2 + ...
2 3 4 2 3 4

Solution: Clearly the required series may be obtained by substituting x = ±π and x = 0.

At the points of discontinuity x = ±π the series converges to

f (π−) + f ((−π)+) (π + π 2 ) + (−π + π 2 )

= = π2;
2 2
Substituting x = ±π into the series we get
∞ ∞
π2 X 4 X 1 π2
+ (−1)(2n) 2 = π 2 =⇒ =
3 n n2 6
n=1 n=1

At the point x = 0 is a point of continuity and therefore the series will converge to 0.
Substituting x = 0 into the series we obtain
∞ ∞
π2 X 4 X 1 π2
+ (−1)(n) 2 = 0 =⇒ (−1)(1+n) 2 = .
3 n n 12
n=1 n=1