I am seeking help on intuition behind the idea of the proof of this theorem. Part of my confusion stems from the fact that I don't quite understand what it means for an open cover to "have a Lebesgue number".

The definition I have for Lebesgue number for an open cover $\mathbb{U}$ of $A\subset X$ where $(X,d)$ is an arbitrary metric space is: a fixed real number $\epsilon>0$ is a Lebesgue number for $\mathbb{U}$ if for every $x\in A$ $\exists$ set $U(x)$ in $\mathbb{U}$ such that $B_\epsilon(x)\subset U(x)$.

So an open cover having a Lebesgue # means there is a certain radius of open ball for each element of A, which is contained in an element of the open cover? But I don't understand the significance behind this, and how this is related to the theorem I stated above.


You have the correct definition and interpretation of a Lebesgue number, but seem to be struggling with the intuition. Personally, I tend to find it more instructive to see what goes wrong (rather than what goes right if we're following along with a theorem). As such, consider the following "non-example":

Let $(X, \mathcal{T})$ be the (not sequentially compact) interval $(0,1)$ with the standard topology. Let $\mathbb{U}$ be the open cover $\{U_n\}_{n=1}^\infty$ where $U_n = (1/(n+2) , ~1/n)$, e.g, $U_1 = (1/3, 1)$, $U_2 = (1/4, 1/2)$, et cetera. Note that this is indeed an open cover: given $x \in (0,1)$, there exists $n \in \mathbb{N}$ such that $1/(n+2) < x < 1/n$.

Since any $x$ will have a $U_n(x)$, we can stick a ball of radius $\epsilon(x)$ about $x$ inside $U_n(x)$, that is $x \in B_{\epsilon(x)} (x) \subset U_n(x)$. However, the issue is that the radius of the ball depends on the particular $x$ value. As $x \to 0$, we have that $\epsilon(x) \to 0$ as well, so there is no way to pick one $\epsilon$ that works for every $x \in X$.

In some way, having a Lebesgue number for an open cover is sort of like a function being uniformly continuous, instead of just continuous. For a function to be continuous, we can pick $\delta$ that depends upon $x$ and $\epsilon$, whereas uniform continuity requires that we choose $\delta$ depending only on $\epsilon$--the same $\delta$ has to work for every $x$.

  • $\begingroup$ Excellent answer i think. Thanks. $\endgroup$ – Frank Swanton Nov 8 '16 at 3:08
  • $\begingroup$ =) I remember being confused upon first introduction as well--the key for me was the analogy with uniform continuity, the idea that this one Lebesgue number has to work for every single $x$, i.e., we don't get to pick a "different Lebesgue number for each $x$" or something like that. $\endgroup$ – erfink Nov 8 '16 at 3:10

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.