Let $(X,d)$ be a metric space and let $K\subset X$ be totally bounded. We are given some arbitrary $\epsilon >0$. Let $B\epsilon$ be a family of $\epsilon$ sized balls, such that any 2 balls from this family are disjoint. Assume $B\epsilon$ has infinite cardinality. Is it true to say that $K$ intersects a finite number of balls from the family $B\epsilon$?

  • $\begingroup$ Yes . By definition of totally boundedness. $\endgroup$ – user118494 Sep 8 '15 at 17:07
  • $\begingroup$ This is not the definition of totally boundedness. The definition is that a totally bounded set can be covered by finitely many $\epsilon$ balls. Here they do not even intersect. And clearly you do not know they cover all of K. $\endgroup$ – User666x Sep 8 '15 at 17:15
  • $\begingroup$ @user118494 if you have an explanation to your answer, please ellaborate $\endgroup$ – User666x Sep 8 '15 at 17:42
  • $\begingroup$ I may be wrong since I see an answer posted negating it with an example but this is what I thought : Let $K$ intersect infinitely many balls. And if they do not cover all of $K$ let $K_{0}$ be the subset of $K$ that that is equal to $K\cap \{B_{\epsilon}\}$ . Now subset of a totally bounded set is totally bounded so $K_{0}$ is totally bounded and hence only finitely many of the $B_{\epsilon} $'s are required to cover $K_{0}$ . Since the $B_{\epsilon}$'s do not intersect , the number of such balls intersecting $K$ should be finite . $\endgroup$ – user118494 Sep 8 '15 at 18:53
  • $\begingroup$ @user118494 The total boundedness of a set $K_0$ means that it can be covered by a finite number of $\epsilon$ sized balls, and not means that any cover of $K_0$ by $\epsilon$ sized balls has a finite subcover. $\endgroup$ – Alex Ravsky Sep 9 '15 at 0:04

NO.Let $ N$ be the positive integers Let $X$ be the set of real sequences $(x_n)_{n\in N}$ for which there is at most one $n>2$ with $x_n \ne 0$. Let $ d((x_n)_{n \in N},(y_n)_{n\in N})=\sum_{n \in N}|x_n-y_n|.$ (Of course at most $3$ terms in the summation are non-zero.) Let $$K=\{(x_n)_{n \in N} \in X : 0\le x_1 \le 1 \text{ and } x_n=0\text{ for } n>1\}.$$ $K$ is compact so it is totally bounded.Take real sequences $(A_j)_{j \in N}$ and $(B_j)_{j\in N}$ where $$0<B_{j+1}<A_j<B_j<1\text{ for }j\in N.$$ Now for each $j$ let $ D_j$ be the open disc of radius $1$ centered at $$C_j=(\sigma_{j,n})_{n\in N}$$ where $$\sigma_{j,1}=(A_j+B_j)/2 \text{ and } \sigma_{j,j}=1-(B_j-A_j)/2.$$ (All the other co-ordinates of $C_j$ are $0$.). Now I leave it to you to confirm that $\{D_j\}_{j\in N}$ is a pair-wise disjoint family and every D_j intersects K.

  • $\begingroup$ I do not need the centers to be disjoint, I need the balls $D_j$ for $j\in N$ to be pairwise disjoint $\endgroup$ – User666x Sep 9 '15 at 8:58
  • $\begingroup$ the balls are disjoint. i made a typo. $\endgroup$ – DanielWainfleet Sep 9 '15 at 9:17
  • $\begingroup$ What happens when j=1? Your 2 definition for sigma do not coincide in this case $\endgroup$ – User666x Sep 9 '15 at 16:02

It seems the following.

In general, the answer is negative. Let $X=\ell^1$ and $\{e_n\}$ be the standard orts of the space $\ell^1$. Put $K=\{e_1/n: n\in\Bbb N\}$, $\epsilon=1$ and the set of centers of the closed balls of the family $B\epsilon$ is $\{e_n+e_1/n: n\in\Bbb N \}$.

  • $\begingroup$ Interesting... You gave me food for thought that is for sure... $\endgroup$ – User666x Sep 8 '15 at 19:16
  • $\begingroup$ The same idea as mine but a lot easier to describe and to verify. $\endgroup$ – DanielWainfleet Sep 8 '15 at 19:20
  • $\begingroup$ @Alex Ravsky is it necessary that the balls be closed? $\endgroup$ – User666x Sep 8 '15 at 22:14
  • $\begingroup$ @AvivEshed In user254665's counterexample the balls are open. $\endgroup$ – Alex Ravsky Sep 9 '15 at 0:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.