Let $(X,d)$ be a metric space and define for $B\subset X$ bounded, i.e.

$$\operatorname{diam}(B)= \sup \{ d(x,y) \colon x,y\in B \} < \infty,$$

the measure

$$\beta(B) = \inf\{r > 0\colon\text{there exist finitely many balls of radius r which cover } B\},$$

or equivalently,

$$\beta(B)=\inf\big\lbrace r > 0|\exists N=N(r)\in{\bf N} \text{ and } x_1,\ldots x_N\in X\colon B\subset\bigcup_{k=1}^N B(x_k,r)\big\rbrace,$$

where $B(x,r)=\{y\in X\colon d(x,y)\}$ denotes the open ball of radius $r$ centered at $x\in X$. Let ${\bf K}(X)$ denote the collection of (non-empty) compact subsets in $X$.

I would like to prove

$$\beta(B)=d_H\big(B,{\bf K}(X)\big),$$ where $d_H$ is the Hausdorff distance.

I proved $d_H\big(B,{\bf K}(X)\big)\le\beta(B)$. Is there someone that knows how to prove the other inequality?


Let $d_0=d_H(B,K(X))$. So for $d>d_0$ we can find compact $K$ with $d_H(B,K)<d$. In particular, $B \subseteq \cup_{k\in K} B(k,d)$.

As $K$ is compact, for any $\epsilon>0$ we can find $k_1,\cdots,k_n\in K$ with $K\subseteq \cup_i B(k_i,\epsilon)$. For $b\in B$, we can find $k\in K$ with $d(b,k)<d$. Then we can find $i$ with $d(k,k_i)<\epsilon$, and so $d(b,k_i)<d+\epsilon$. Thus $\beta(B)\leq d+\epsilon$.

As $\epsilon>0$ and $d>d_0$ were arbitrary, we find that $\beta(B) \leq d_0$ as required.


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.