Sie sind auf Seite 1von 6

Chapter 3

Uniform Convergence

Lecture 9
Sequences of functions are of great importance in many areas of pure and applied
mathematics, and their properties can often be studied in the context of metric
spaces, as in Examples 2.17.4 and .5.
Unless otherwise stated, the functions under consideration take the form
f : D R, where D R denotes a generic domain in the Euclidean line.
Usually, D is an interval.

Definition 3.1. A sequence (fn : n 1) of such functions converges pointwise


to f on D whenever the sequence of real numbers (fn (x)) converges to f (x) in R
for every x D.

Pointwise convergence need not behave well with respect to standard proper-
ties such as continuity.

Example 3.2. Let fn (x) = xn for all n 1 and x [0, 1]. Then limn fn (x)
is the function (
0 for 0 x < 1
f (x) =
1 for x = 1.
So f is discontinuous on [0, 1], even though fn is continuous for every n.

Definition 3.3. A sequence (fn : n 1) of functions converges uniformly to f


on D whenever

 > 0, N () N such that n N () |fn (x) f (x)| < 

for every x D.

The force of this definition is that the same N = N () must work for every
x D simultaneously. This requirement fails in Example 3.2. For take  = 1/2;

26
then for any N , however large, |xN 0| 1/2 for all 1 > x 21/N (that is, for
all x sufficiently close to 1). On the other hand,

(3.4) fn f uniformly on D fn f pointwise on D .

always holds.

Examples 3.5.

1. The sequence (xn ) converges to the constant function 0 on the subinterval


[0, 1). However, the argument above remains valid, and shows that this
convergence also fails to be uniform on [0, 1).

2. Let an a in R as n . Then the sequence (fn ) of constant functions


fn (x) = an converges both pointwise and uniformly to the constant function
f (x) = a on any domain D.

3. If fn f uniformly on the domain D, then fn f uniformly on any


subdomain D0 D.

It is now possible to link uniform convergence with the sup metric described
in Lemma 1.18.

Proposition 3.6. Let f and fn : D R be functions on the domain D; then


fn f uniformly on D iff supxD |fn (x) f (x)| exists for sufficiently large n,
and tends to 0 as n .

Proof. Assume that Mn = supxD |fn (x) f (x)| for sufficiently large n, and that
Mn 0. So for any  > 0, there exists an integer N such that n N implies
Mn < ; hence |fn (x) f (x)| <  for every x D, and fn f uniformly on D.
Conversely, assume that fn f uniformly on D. So for any  > 0, there
exists an integer N such that |fn (x) f (x)| < /2 for all n N and every x D.
The set {|fn (x) f (x)| : x D} is therefore bounded above by /2, and its
supremum Mn satisfies Mn /2 <  for all n N .
The sup metric arises on spaces of bounded or continuous functions, and
was tentatively related to uniform convergence in Examples 2.17.4. In that con-
text, Proposition 3.6 confirms that ordinary convergence in the function spaces
C[a, b] B[a, b] is the same as uniform convergence of the corresponding sequence
of functions on the domain [a, b].

Example 3.7. Let Asup C[0, 1] denote the subspace {h : h(0) 6= 0}, with
respect to the sup metric. For each h Asup , let  = |h(0)| > 0 and consider the
open ball B (h) C[0, 1]. Then f B (h) implies that supx[0,1] |f (x)h(x)| < ,
and therefore that |f (0) h(0)| <  = |h(0)|. So f (0) 6= 0, and f must also lie in
Asup . Hence Asup is open in C[0, 1].

27
Example 3.7 shows that the complement {h : h(0) = 0} of Asup is closed.
Such properties depend on the metric, as is apparent from Examples 2.17.5.
Example 3.8. Let A1 L1 [0, 1] denote the subspace {h : h(0) 6= 0}, with
respect to the L1 metric of Lemma 1.23. Assume h A1 has h(0) > 0, and for
any  > 0 define g L1 [0, 1] to be linear on [0, /h(0)], with
(
h(0) if x = 0
g(x) =
0 if x [/h(0), 1] .

Hence g 0 := h g has g 0 (0) = 0, so g 0


/ A1 . But
Z 1
d1 (g 0 , h) = g(t)dt = /2 <  ,
0

so g 0 lies in B (h). Thus h


/ A1 . A similar argument applies to the case h(0) < 0,
so A1 is empty; in particular, A1 cannot be open.

Lecture 10
Uniformly convergent sequences of functions often interact well with analytic
properties such as continuity, and procedures such as integration.
By way of motivation, recall that Example 3.2 exhibits a non-uniformly con-
vergent sequence of continuous functions whose pointwise limit is discontinuous.
Example 3.9. On the interval [0, 1] R, let (gn : n 1) denote the sequence
of functions (
n if x (0, 1/n)
gn (x) =
0 otherwise .
Pointwise, gn (x) 0 as n , but the convergence is not uniform, because
supx[0,1] |gn (x) 0| = n does not tend to 0. Each gn is discontinuous at points
Rx 1= 0 (from the right) and 1/n; nevertheless, it is bounded and integrable, with
g (t)dt = 1 for all n. Thus
0 n
Z 1 Z 1
lim gn (t)dt = 1 6= 0 = lim gn (t)dt .
n 0 0 n

So the processes of integration and taking pointwise limits cannot necessarily be


interchanged, at least in situations where convergence is not uniform.
Neither of these anomalies can occur for a uniformly convergent sequence of
functions.
Theorem 3.10. If fn : [a, b] R is continuous for every n N , and fn f
uniformly on [a, b], then f is continuous on [a, b].

28
Proof. Since the convergence is uniform, for any  > 0 there exists an integer N
such that n N implies |fn (x) f (x)| < /3 for every x [a, b]. Moreover, fN
is continuous at each point x0 , so there exists > 0 for which |x x0 | < implies
|fN (x) fN (x0 )| < /3. For any such x, it follows that

|f (x) f (x0 )| |f (x) fN (x)| + |fN (x) fN (x0 )| + |fN (x0 ) f (x0 )| < 

from the triangle inequality. Thus f is continuous at x.


Care is needed to interpret certain steps of this proof at x = a or b.

Corollary 3.11. Suppose that fn : [a, b] R is continuous for every n N and


that the pointwise limit of the sequence (fn ) is discontinuous on [a, b]; then the
convergence cannot be uniform.

Examples 3.12.

1. In Lecture 9, the proof that the sequence (xn ) fails to converge uniformly
on [0, 1] involves the  and N of Definition 3.3. But Example 3.2 notes
that its pointwise limit is discontinuous at x = 1; so Corollary 3.10 shows
immediately that convergence cannot be uniform.

2. The function fn (x) = 1/(1 + nx) is continuous on [0, 1], for all n 1.
Moreover, (fn (x)) converges pointwise to the function
(
1 if x = 0
f (x) = ,
0 if x (0, 1]

which is discontinuous at x = 0. So Corollary 3.10 confirms that the con-


vergence is not uniform.

3. The function fn (x) = cosn x is continuous on [/2, /2], for all n 1.


Moreover, (fn (x)) converges pointwise to the function
(
1 if x = 0
f (x) = ,
0 otherwise

which is discontinuous at x = 0. So Corollary 3.10 confirms that the con-


vergence is not uniform.

Of course, Examples 3.12 may also be approached directly via Proposition


3.6. In each case, supxD |fn (x) f (x)| is easily computed to be 1 for all n, and
therefore 6 0; so the convergence of (fn ) is not uniform. It is instructive to
sketch the graphs of fn (x) for several values of n.

29
Theorem 3.13. If fn : [a, b] R is integrable on [a, b] for every n N , and
(fn ) converges uniformly on [a, b], then
Z b Z b
lim fn (t)dt = lim fn (t)dt .
n a a n

Proof. Let fn f uniformly. For any  > 0 there exists an integer N such that
n N implies supx[a,b] |fn (x) f (x)| < , by Proposition 3.6. Hence
b Z b Z b
Z


fn (t) f (t) dt
|fn (t) f (t)|dt dt = (b a)
a a a
Rb Rb 
for all n N . Thus a
fn (t)dt a
f (t)dt tends to 0 as n .
2
Example 3.14. Let fn (x) = xenx on [0, 1], for n 1. Since 0 < fn (x) < 1/nx
for each 0 < x 1, the pointwise limit of (fn ) is the zero function. To check for
uniform convergence, consider supx[0,1] |fn (x) 0|; this requires identifying the
turning points of fn (x) on [0, 1].
2 2
But dfn (x)/dx = enx + (2nx)xenx , which is 0 iff x = 1/(2n)1/2 ; fur-
thermore, the derivative is > 0 to the left of the turning point, and < 0 to the
right. Thus fn (1/(2n)1/2 ) = 1/(2ne)1/2 is the maximum value of fn (x). Hence
supx[0,1] |fn (x)| tends to limn0 (1/(2ne)1/2 ) = 0, and the convergence is uniform.
Then Theorem 3.13 confirms the well-known fact that
Z 1
2
lim tent dt = 0 .
n 0

30
Problems 3

Problem 28.
Interpret your solution to Problem 21 in terms of uniform convergence.

Problem 29.
Compute the integrals
Z 1 Z 2
4
et /n dt t2(sin nt)/n dt

lim and lim
n 0 n 1

justifying any procedures you may use.

31

Das könnte Ihnen auch gefallen