Uniform & Pointwise Convergence (II)
Uniform & Pointwise Convergence (II)
Consider sequences and series whose terms depend on a variable, i.e., those whose
terms are real valued functions defined on an interval as domain. The sequences
and series are denoted by {fn} and ∑fn respectively.
Point-wise Convergence
Definition. Let {fn}, n = 1, 2, 3,…be a sequence of functions, defined on an interval
I, a ≤ x ≤ b. If there exits a real valued function f with domain I such that
f(x) = lim {fn(x)}, ∀ x ∈I
n →∞
Then the function f is called the limit or the point-wise limit of the sequence {fn}
on [a, b], and the sequence {fn} is said to be point-wise convergent to f on [a, b].
Similarly, if the series ∑fn converges for every point x∈I, and we define
∞
f(x) = ∑f
n =0
n ( x) , ∀ x ∈ [a, b]
the function f is called the sum or the point-wise sum of the series ∑fn on [a, b].
Definition. If a sequence of functions {fn} defined on [a, b], converges poinwise
to f , then to each ∈ > 0 and to each x ∈ [a, b], there corresponds an integer N such
that
|fn(x) − f(x) | < ∈, ∀ n ≥ N (1.1)
Remark:
1. The limit of differentials may not equal to the differential of the limit.
sin nx
Consider the sequence {fn}, where fn(x) = , (x real).
n
It has the limit
f(x) = lim fn(x) = 0
n →∞
1
But
f n′ ( x ) = n cos nx
so that
f n′ (0) = n →∞ as n→∞
Thus at x = 0, the sequence { f 'n ( x )} diverges whereas the limit function f ′(x) = 0,
2. Each term of the series may be continuous but the sum f may not.
Consider the series
∞
x2
∑ f n , where fn(x) = (1 + x 2 ) n
(x real)
n =0
1
∴ ∫ f (x ) dx = 0
0
Again,
1 1
n
∫ f n (x ) dx = ∫ nx(1 − x ) dx =
2 n
0 0
2n + 2
2
so that
⎧⎪ 1 ⎫⎪ 1
lim ⎨∫ f n ( x ) dx ⎬ =
n →∞ ⎪ ⎪⎭ 2
⎩0
Thus,
⎧⎪ 1 ⎫⎪ 1 1
lim ⎨∫ f n dx ⎬ ≠ ∫ f dx = ∫ ⎡ lim {f n }⎤ dx
n →∞ ⎪ ⎪⎭ 0 ⎢⎣n →∞ ⎥⎦
⎩0 0
Uniform Convergence
3
Theorem (Cauchy’s Criterion for Uniform Convergence). The sequence of
functions {fn} defined on [a, b] converges uniformly on [a, b] if and only if for
every ε > 0 and for all x ∈ [a, b], there exists an integer N such that
|fn+p(x) − fn(x) | < ε, ∀ n ≥ N, p ≥ 1 …(1)
Proof. Let the sequence {fn} uniformly converge on [a, b] to the limit function f, so
that for a given ε > 0, and for all x ∈ [a, b], there exist integers n1, n2 such that
| fn(x) − f(x)| < ε/2, ∀ n ≥ n1
and
|fn+p(x) − f(x)| < ε/2, ∀ n ≥n2, p ≥ 1
Let N = max (n1, n2).
∴ |fn+p(x) − fn(x)| ≤ |fn+p(x) − f(x)| + |fn(x) − f(x)|
< ε/2 + ε/2 = ε, ∀ n ≥ N, p ≥ 1
Conversely. Let the given condition hold so by Cauchy’s general principle of
convergence, {fn} converges for each x ∈ [a, b] to a limit, say f and so the sequence
converges pointwise to f.
For a given ε > 0, let us choose an integer N such that (1) holds. Fix n, and
let p→∞ in (1). Since fn+p→f as p → ∞, we get
|f(x) − fn(x)| < ε ∀ n ≥ N, all x ∈[a, b]
which proves that fn(x) → f(x) uniformly on [a, b].
Remark. Other form of this theorem is :
The sequence of functions {fn} defined on [a, b] converges uniformly on [a, b] if
and only if for every ε > 0 and for all x ∈ [a, b], there exists an integer N such that
|fn(x) − fm(x)| < ε, ∀ n, m ≥ N
Theorem 2. A series of functions ∑fn defined on [a, b] converges uniformly on
[a, b] if and only if for every ε > 0 and for all x∈[a, b], there exists an integer N
such that
|fn+1(x) + fn+2(x) +…+ fn+p(x)| < ε, ∀ n ≥ N, p ≥ 1 …(2)
Proof. Taking the sequence {Sn} of partial sums of functions ∑fn , defined by
4
n
Sn(x) = ∑ f i (x)
i =1
5
For 0 < x ≤ k < 1, we have
|fn(x) − f(x)| = xn < ε
if
n
⎛1⎞ 1
⎜ ⎟ >
⎝x⎠ ε
or if
n > log (1/ε)/log(1/x)
This number, log (1/ε)/log (1/x) increases with x, its maximum value being
log (1/ε)/log(1/k) in ]0, k], k > 0.
Let N be an integer ≥ log (1/ε)/log(1/k).
∴ |fn(x) − f(x)| < ε, ∀ n ≥ N, 0 < x < 1
Again at x = 0,
|fn(x) − f(x)| = 0 < ε, ∀ n≥1
Thus for any ε > 0, ∃ N such that for all x∈[0, k], k < 1
|fn(x) − f(x)| < ε, ∀n≥N
Therefore, the sequence {fn} is uniformly convergent in [0, k], k < 1.
However, the number log (1/ε)/log (1/x)→∞ as x→1 so that it is not
possible to find an integer N such that |fn(x) − f(x)| < ε, for all n ≥ N and all x in
[0, 1]. Hence the sequence is not uniformly convergent on any interval containing
1 and in particular on [0, 1].
Example . Show that the sequence {fn}, where
1
fn(x) =
x+n
is uniformly convergent in any interval [0, b], b > 0.
Solution. The limit function is
f(x) = lim fn(x) = 0 ∀ x ∈ [0, b]
n→
6
1
|fn(x) − f(x)| = <ε
x+n
if n > (1/ε) − x, which decreases with x, the maximum value being 1/ε.
Let N be an integer ≥ 1/ε, so that for ε > 0, there exists N such that
|fn(x) − f(x)| < ε, ∀n≥N
Hence the sequence is uniformly convergent in any interval [0, b], b > 0.
2
Example . The series ∑fn, whose sum to n terms, Sn(x) = nxe −nx , is pointwise and
not uniformly convergent on any interval [0, k], k > 0.
Solution. The pointwise sum S(x) = lim Sn(x) = 0, for all x ≥ 0. Thus the series
n →∞
Let N0 be an integer greater than N and e2ε2, then for x = 1/ N 0 and n = N0, (*)
gives
√N0/e < ε ⇒ N0 < e2ε2
so we arrive at a contradiction. Hence the series is not uniformly convergent on
[0, k].
Note . The interval of uniform convergence is always to be a closed interval, that is
it must include the end points. But the interval for pointwise or absolute
convergence can be of any type.
Theorem 3. Let {fn} be a sequence of functions, such that
lim fn(x) = f(x), x ∈ [a, b]
n →∞
and let
Mn = Sup |fn(x) − f(x)|
x∈[ a ,b ]
7
Proof. Let fn → f uniformly on [a, b], so that for a given ε>0, there exists an integer
N such that
|fn(x) − f(x)| < ε, ∀ n ≥ N, ∀ x ∈ [a, b]
⇒ Mn = Sup |fn(x) − f(x)| ≤ ε, ∀n≥N
x∈[ a ,b ]
⇒ Mn → 0, as n→ ∞
Conversely. Let Mn → 0, as n → ∞, so that for any ε > 0, ∃ an integer N such that
Mn < ε, ∀n≥N
⇒ Sup |fn(x) − f(x)| < ε, ∀n≥N
x∈[ a ,b ]
8
f(x) = lim fn(x) = 0, ∀x
n →∞
x
Mn = Sup | f n ( x ) − f ( x ) | = Sup
x∈I x∈I 1 + nx 2
1
= →0 as n → ∞
2 n
Hence {fn} converges uniformly on I.
⎡ x 1 1 ⎤
⎢⎣1 + nx 2 attains the maximum value 2 n at x = n , i.e. at the origin.⎥
⎦
Example . Show that the sequence {fn}, where
2
fn(x) = nxe −nx , x ≥ 0
is not uniformly convergent on [0, k], k > 0
Solution. f(x) = lim fn(x) = 0, ∀x≥0
n →∞
2 n 1
Also nxe −nx attains maximum value at x =
2e 2n
Now
Mn = Sup |fn(x) − f(x)|
x∈[ 0,k ]
2 n
= Sup nxe −nx = → ∞ as n→∞
x∈[ 0,k ] 2e
9
n −1
or x = 0 or .
n
d2y n−
A so 2
= − ve when x =
dx n
n −1
⎛ 1⎞ ⎛ n −1⎞ 1
∴ Mn = max y = ⎜1 + ⎟ ⎜1 − ⎟ → × 0 = 0 as n→∞.
⎝ n⎠ ⎝ n ⎠ e
Hence the sequence is uniformly convergent on [0, 1] by Mn-test.
Example. Show that 0 is a point of non-uniform convergence of the sequence {fn},
where fn(x) = 1−(1 − x2)n.
Solution. Here
⎧0 when x = 0
f(x) = lim f n ( x ) = ⎨
⎩1 when 0 < 1 | x |< 2
n →∞
neighborhood ]0, k[ of 0 where k is a number such that 0 < k < 2 . There exists
therefore a positive integer m such that
∑ xe
n =0
− nx
in the closed interval [0, 1].
n −1
x (−11 / e nx )
Solution. Here fn(x) = ∑ xe −nx =
n =1 1 − 1/ ex
xe x ⎛ 1 ⎞
= x ⎜1 + nx ⎟
e −1⎝ e ⎠
10
⎧0 where x = 0
⎪ x
Now f(x) = lim f n ( x ) = ⎨ xe
⎪ e x − 1 when 0 < x ≤ 1
n →∞
⎩
We consider 0 < x ≤ 1. We have
Mn = sup {|fn(x) − f(x) : x ∈ [0, 1]}
⎧ xe x ⎫
= sup ⎨ : x ∈ [0,1]⎬
⎩ (e − 1)e
x nx
⎭
1 / n.e1 / n ⎛ 1 ⎞
≥ ⎜ Taking x = ∈ [0,1] ⎟
(e1 / n − 1)e ⎝ n ⎠
1 / n e1 / xn ⎡ 0⎤
Now lim 1 / n ⎢⎣Form
n →∞ (e − 1) 0 ⎥⎦
1 / ne1 / n (−1 / n 2 ) + (−1 / n 2 )e1 / n
= lim
n →∞ e.e1 / n − (−1 / n 2 )
(1 / n + 1) (0 + 1) 1
= lim = = .
n →∞ e e e
Thus Mn does not tend zero as n→∞.
Hence the sequence is non-uniformly convergent by Mn-test.
Here 0 is a point of non-uniform convergence.
Example . The sequence {fn}, where
nx
fn(x) =
1+ n 2x2
is not uniformly convergent on any interval containing zero.
Solution. Here
lim fn(x) = 0, ∀x
n →∞
nx 1 1 1
Now 2 2
attains the maximum value at x = ; tending to 0 as
1+ n x 2 n n
n→∞. Let us take an interval [a, b] containing 0.
Thus
11
Mn = Sup |fn(x) − f(x)|
x∈[ a ,b ]
nx
= Sup
x∈[ a ,b ] 1+ n2x2
1
= , which does not tend to zero as n→∞.
2
Hence the sequence {fn} is not uniformly convergent in any interval
containing the origin.
12
du n ( x )
Now un(x) is maximum or minimum when =0
dx
or (n + x2)2 − 4x2 (n + x2) = 0
3x4 + 2nx2 − n2 = 0
n n
or x2 = i.e. x = .
3 3
d 2 u n (x) n
If will be seen that is −ve when x = .
dx 2 3
n
3 3 3
Hence Max un(x) = 2
= = Mn .
⎛ n⎞ 16n 3 / 2
⎜n + ⎟
⎝ 3⎠
Therefore |un(x)| ≤ Mn.
But ∑ Mn is convergent.
Hence the given series is uniformly convergent for all values of x by Weierstrass’s
M test.
(ii) Here un(x) is Maximum or minimum when
n(1 + nx2) − 2n2x2 = 0 or x = ± 1/√(n).
1
It can be easily shown that x = makes un(x) a maximum.
n
1/ n 1
Hence Max un(x) = = = M n . But ∑ Mn is convergent.
n (1 + 1) 2.n 3 / 2
Hence the given series is uniformly convergent for all values of x by
Weierstrass’s M-test.
∞
x
Example :- Consider ∑ n (1 + nx 2 ) , x ∈R.
n =1
We assume that x is +ve, for if x is negative, we can change signs of all the terms.
We have
13
x
fn(x) =
n (1 + nx 2 )
and f n′ ( x ) = 0
1
implies nx2 = 1. Thus maximum value of fn(x) is
2n 3 / 2
1
Hence fn(x) ≤
2n 3 / 2
∞
1 x
Since ∑ n3 / 2 is convergent, Weierstrass’ M-Test implies that ∑ n (1 + nx 2 ) is
n =1
x
f n′ ( x ) =
(n + x 2 ) 2
(n + x 2 ) 2 − 2 x (n + x 2 )2 x
and so fn(x) =
(n + x 2 ) 4
Thus f n′ ( x ) = 0 gives
x4 + x2 + 2nx2 − 4nx2 − 4x4 = 0
− 3x4 − 2nx2 + n2 = 0
or 3x4 + 2nx2 − n2+ = 0
n n
or x2 = or x =
3 3
3 3 1
Also f n′′ ( x ) is −ve. Hence maximum value of fn(x) is 2
. Since ∑ is
16n n2
convergent, it follows by Weierstrass’s M-Test that the given series is uniformly
convergent.
14
x
Example . The series ∑ n p + x 2n q converges uniformly over any finite interval [a,
⎡ 1 1⎤
converges uniformly on ⎢− , ⎥ .
⎣ 2 2⎦
15