5
$\begingroup$

Let $(X,d_X)$ and $(Y, d_Y)$ be compact metric spaces.

  1. Show that for a nonempty set $A\subseteq X,$ the function $f : X\to \mathbb{R}, f(x) = \inf \{d_X(x,a) : a\in A\}$ is continuous.
  2. Show that if a map $f : X\to Y$ is both continuous and bijective, then its inverse $f^{-1} : Y\to X$ is also continuous.
  3. Let $C$ be the Cantor set with metric from $(\mathbb{R},|\cdot|)$ and $P = \prod_{k=1}^\infty \{0,\frac{1}{2^k}\}$ with metric from $(\ell_1, \lVert \cdot \rVert_1).$ Show that there exists a continuous bijection $f : C\to P.$

For $1),$ let $x\in X$ and let $\epsilon > 0.$ Choose $\delta = \epsilon.$ Then if $y\in X$ is such that $d(x,y) < \epsilon, we $ have \begin{align}d(f(x), f(y)) &= \inf\{d(x,a) | a \in A \} - \inf\{d(y,a) | a\in A\}\\ &\leq \inf \{d(x,a) | a\in A\} - (\inf \{d(x,a) - d(x,y) | a \in A\})\\ &= \inf \{d(x,a) | a\in A\} - (\inf \{d(x,a) | a \in A\} - d(x,y)\}\\ &= d(x,y) < \epsilon.\end{align}

Is this part correct?

For $2),$ suppose $f:X\to Y$ is continuous and bijective. Let $y\in Y$ and $\epsilon > 0.$ We want to show that there exists $\delta > 0 $ so that if $d_Y(x,y) < \delta$ then $d_X(f^{-1}(x), f^{-1}(y)) < \epsilon$. Let $(y_n)$ be a sequence in $Y$ such that $y_n\to y.$ Then $(f^{-1}(y_n))\subseteq X$ and since $X$ is compact, it has a convergent subsequence, say $(f^{-1}(y_{n_k}))_k.$ Since $f$ is bijective, we can write $y_n = f(x_n)$ for each $n\in\mathbb{N},$ where $x_n \in X.$ Let $x = \lim_{k\to\infty} f^{-1}(y_{n_k}) = \lim_{k\to \infty} x_{n_k}.$ Since $f$ is continuous, $f(x) = \lim_{k\to\infty} f(x_{n_k}).$ But here I'm stuck as I'm not sure how to show $f^{-1}$ is continuous from here.

As for the third part, every element of the Cantor set can be uniquely written in the form $\sum_{k=1}^\infty \frac{a_k}{3^k}, a_k \in \{0,2\}\forall k.$ Then it suffices to show that the function $f : C\to P, f(\sum_{k=1}^\infty \frac{a_k}{3^k}) = \prod_{k=1}^\infty \{\frac{a_k/2}{2^k}\}$ is continuous. I think one could show it's sequentially continous, but I'm not really sure about the details.

$\endgroup$
1
  • $\begingroup$ For 3, consider choosing $\delta = min_k \frac{2^k}{3^k} \leq \epsilon$ $\endgroup$
    – colossal
    Feb 1 at 16:22
1
+50
$\begingroup$

I think (1) is correct, but I wouldn't bother with the infima: let $a,b\in A$. Then

$$|d(x,a)-d(y,b)|\leq |d(x,a)-d(a,y)|+|d(a,y)-d(y,b)|\leq d(x,y)+d(a,b)$$ where we used the triangle inequality. Take infimum on both sides as $a\in A$, while $b$ is (for now) fixed, so $|f(x)-d(y,b)|\leq d(x,y)+0=d(x,y)$. Now this is true for all $b\in A$, so take infimum over $b$ and conclude that $|f(x)-f(y)|\leq d(x,y)$, I think this is more neat.

For (2) this is a standard topological fact: If $f:X\to Y$ is a continuous bijection from a compact space to a Hausdorff space, then it is a homeomorphism. Indeed, all we need to do is show that $f^{-1}$ is continuous: if $E\subset X$ is a closed subset, we want to show that the inverse image of $E$ through $f^{-1}$ is closed in $Y$, which is the same as $f(E)$ is closed in $Y$. Now $X$ is compact and $E$ is a closed subset, so $E$ is also compact. Now $f$ is continuous, so $f(E)$ is a compact subset of $Y$. Now $Y$ is Hausdorff, so all compact sets are closed, thus $f(E)$ is closed in $Y$. Sometimes it helps viewing things in a more abstract frame; for example the metrics here are not essentially needed, so the extra information causes extra confusion.

For (3) you have a correct guess: Indeed, every element of the Cantor set $C$ can be written uniquely as $\sum_{k=1}^\infty\frac{a_k}{3^k}$, where $a_k$ is either $0$ or $2$. But it is important to note that the converse is also true: Any sum $\sum_{k=1}^\infty\frac{a_k}{3^k}$ where $a_k\in\{0,2\}$ is an element of the cantor set. Now define $f:C\to P$ as $$f\big(\sum_{k=1}^\infty\frac{a_k}{3^k}\big)=(\frac{a_k}{2^{k+1}})_{k=1}^\infty$$

  • $f$ is well-defined: since $a_k$ is either $0$ or $2$, we have that $a_k/2^{k+1}$ is either $0$ or $1/2^k$, so each $a_k/2^{k+1}$ does indeed belong to $\{0,\frac{1}{2^k}\}$, so $f$ does indeed take values in $\prod_{k=1}^\infty\{0,\frac{1}{2^k}\}$.

  • $f$ is one to one: this is obvious.

  • $f$ is surjective: indeed, let $x=(x_k)_{k=1}^\infty\in P$, so each $x_k$ is either $0$ or $1/2^k$. Set $a_k=2^{k+1}\cdot x_k$. Then $a_k$ is either $0$ or $2$, so if $t=\sum_{k=1}^\infty\frac{a_k}{3^k}\in C$ then $f(t)=x$.

  • $f$ is continuous: Let $\varepsilon>0$ and fix a point $x\in C$, say $x=\sum_{k=1}^\infty\frac{a_k}{3^k}$, where $a_k\in\{0,2\}$. Let $y\in C$ be another element, say $y=\sum_{k=1}^\infty\frac{b_k}{3^k}$ where $b_k\in\{0,2\}$. Note that $$d_P(f(x),f(y))=\|\{\frac{a_k}{2^{k+1}}\}_{k=1}^\infty-\{\frac{b_k}{2^{k+1}}\}_{k=1}^\infty\|_{\ell^1}=\|\{\frac{a_k-b_k}{2^{k+1}}\}_{k=1}^\infty\|_{\ell^1}=\sum_{k=1}^\infty\frac{|a_k-b_k|}{2^{k+1}} $$

Claim: Let $m\geq1$ be an integer. Then if $|x-y|<\frac{1}{3^m}$, then $a_k=b_k$ for all $k=1,\dots,m$.

Proof of the claim: I will only describe the idea here. Note that the construction of the Cantor set is done in steps: in each step we divide the segments in 3 pieces and throw out the middle part. In the expression $x=\sum_{k=0}^\infty\frac{a_k}{3^k}$ with $a_k=0$ or $2$, we have the following interpretation: in the first step of the construction, we divide $[0,1]$ in three segments and we only keep $[0,1/3]$ and $[2/3,1]$. If $a_1=0$, then $x$ lies in $[0,1/3]$. If $a_1=2$, then $x$ lies in $[2/3,1]$. So if $|x-y|<1/3$ where $y\in C,y=\sum_{k=1}^\infty\frac{b_k}{3^k}$ with $b_k=0$ or $2$, then $b_1=a_1$: otherwise we would have $x\in[0,1/3]$ and $y\in[2/3,1]$ or $x\in [2/3,1]$ and $y\in[0,1/3]$, which contradicts the fact that $|x-y|<1/3$. Let's assume that $a_1=0$, so $x\in[0,1/3]$. Now again, in the second step of the construction, we cut each of $[0,1/3]$ and $[2/3,1]$ in three pieces each and throw out the middle third of each. Let's look at $a_2$. If $a_2=0$, then $x$ lies in the left segment of $[0,1/3]$ that is left after the second step, i.e. $x\in[0,1/3^2]$. If $a_2=2$, then $x$ lies in the right segment of $[0,1/3]$ that is left after the second step, i.e. $x\in[2/3^2,1/3]$. Now if $|x-y|<1/3^2$ then we also have $|x-y|<1/3$, so $y$ lies in $[0,1/3]$ as we explained for the first step, i.e. $b_1=0$. Now if $b_2\neq a_2$, we have that $x,y$ are in different thirds of $[0,1/3]$ after the application of the second step of the construction, so $|x-y|\geq 1/3^2$, a contradiction. I believe that you get the idea.

Now let $m$ be so large that $\sum_{k={m+1}}^\infty\frac{1}{2^k}<\varepsilon$. Taking $\delta=\frac{1}{3^m}$; from the claim we have that if $y\in C$ satisfies $|x-y|<\delta$, then $a_k=b_k$ for all $k=1,\dots,m$, so $$d_P(f(x),f(y))=\sum_{k=m+1}^\infty\frac{|a_k-b_k|}{2^{k+1}}\leq\sum_{k=m+1}^\infty\frac{2}{2^{k+1}}<\varepsilon$$ and continuity follows. Observe that $\delta$ is independent of $x$, so $f$ is uniformly continuous.

Is this good news? No, this is expected, since $f$ is continuous on a compact space, thus it must be uniformly continuous. Also, by (2) this gives us a homeomorphism of the cantor set and $P$.

$\endgroup$

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.