Processing math: 100%

Friday, August 29, 2014

Partial Converse to the Uniform Limit Theorem (3.26.10)

James Munkres Topology, chapter 3.26, exercise 10:

MathJax TeX Test Page (a) Let fn:X be a sequence of continuous functions, such that fn(x)f(x) for each xX. If f is continuous, and if the sequence (fn) is monotone increasing, and if X is compact, show that the convergence is uniform.

(b) Give examples to show that the converse fails to hold if either of the requirements of X being compact or fn being monotone increasing is deleted.

Proof: (a) For each n, let Kn={xX | |fn(x)f(x)|ε}. It is routine to show Kn is closed, by showing its complement is open: Suppose |fn(x)f(x)|<ε. Then f(x)(fn(x)ε,fn(x)+ε)=V, so let U be a neighborhood of f(x) contained in V, and we see f1(U)X is open such that f(U)V so that U is a neighborhood of x also contained in the complement of Kn. Now, since (fn) is monotone increasing, and hence f(x)fn(x) by convergence, we see Kn+1Kn for all n. If we assume each Kn is nonempty, then finite intersections Ki=Kmax i are also nonempty, so since X is compact, there exists xX such that xKn for all n, i.e. |fn(x)f(x)|ε for all n. However, this is impossible since fn(x)f(x) for all x. Hence KN=ø for some N, hence Kn=ø for all nN, that is, |fn(x)f(x)|<ε for all x, for sufficiently large n.

(b) Suppose X need not be compact; then the sequence of functions (fn) defined by fn(x)={0:xnnx:x>n is seen to be monotone increasing with fn(x)0 for all x, yet does not converge uniformly to the zero function.

Suppose fn need not be monotone increasing; then let [0,1]=[an,bn] be any countably infinite partition of [0,1] into disjoint closed intervals where anbn for all n. Then when cn=an+bn2, define a sequence of functions (fn) by fn(x)={0:x[an,bn]xancnan:x[an,cn]xbncnbn:x[cn,bn] It is clear by the pasting lemma that each fn is a continuous function [0,1][0,1]. As well, it is clear fn(x)0 for at most one value of n, so that fn(x)0 for each x. However, since fn(cn)=1 for each n, we do not have uniform convergence to the zero function. 

No comments:

Post a Comment