$X=\mathbb{N}_0$ and $Y= \left\lbrace \frac{1}{n} ~|~ n\in \mathbb{N} \right\rbrace \cup \left\lbrace 0\right\rbrace $ are subspaces of euclidean space $\mathbb{R}$. See if function $f: X \to Y $

$ f(n) = \begin{cases} \frac{1}{n} & n\in \mathbb{N}\\ 0 & n =0 \\ \end{cases} $

is a homeomorphism.

How would you do this textbook style, leaving little unsaid and being formal as possible by using only beginner theorems? I know how to do this but my proof always involve some handwaving. All my topology is handwaving, to be honest all my proofs are handwaving, I never mastered the art of a good explanation. For once I just want to see an obnoxious proof. Thanks

  • $\begingroup$ What do you mean by "an obnoxious proof"? $\endgroup$ – Cameron Buie Dec 16 '15 at 23:48
  • $\begingroup$ formal and explaining the obvious, maybe I used the wrong word. for example take Gaffney's first sentance, formalize it, write the definition down, prove the obvious, etc.. $\endgroup$ – nik Dec 17 '15 at 0:08
  • $\begingroup$ @nik What's obvious about all but a single point being open? What's obvious about a bijection on a countable set being discontinuous? $\endgroup$ – T.J. Gaffney Dec 17 '15 at 15:14
  • $\begingroup$ If the current answers do not contain enough detail, comment what is lacking. $\endgroup$ – Jake1234 Dec 21 '15 at 19:30
  • $\begingroup$ will comment in detail tomorrow $\endgroup$ – nik Dec 25 '15 at 12:26

Assume $f^{-1}: Y\to X$ is continuous. Clearly, it maps $0\in Y$ to $0\in X$. One definition of continuity in topology is that preimages of open sets are open.

Open sets in $X$ resp. $Y$ are such sets that with each $x$ also contain all points closer to $x$ then $\epsilon$ for some $\epsilon>0$. (This is co-called metric topology induced by the standard distance function on $\Bbb R$. It's the most obvious topology on $X,Y\subseteq\Bbb R$ one can think of, unless otherwise stated.)

If you accept this, and accept that the one-point set $\{0\}$ is open in $X=\Bbb N_0$, then $\{0\}$ should be open in $Y$. But it is not, because any neighborhood of $0$ in $Y$ contains infinitely many other points.

Of course, if you want to be very formal, one should probably say, in the assignment, "$X,Y$ are topological subspaces of $\Bbb R$ with the metric/Euclidean topology". If you endow $\Bbb R$ with a discrete topology---that is, all subsets are open---then $f$ is a homeomorphism, because it is a bijection and preimages of "open" (that means, any) sets under $f$, resp. $f^{-1}$ are "open" (that means, any) sets.


There cannot be an homeomorphism between these two spaces, because $Y$ has a limit point in $Y$ (this limit point is $0$) while $X$ has no limit point in $X$.


Let $f$ be a bijection between $X$ and $Y$. For simplicity, name $g=f^{-1}$.

Consider $a=g(0)=f^{-1}(0) \in \mathbb{N}$, and consider the set $\{a\}$ which is equal to $\mathbb{N} \cap (a-(1/2),a+(1/2))$.

$(a-1/2),a+(1/2))$ is an open interval, and those form a base of open sets in standard topology on $R$, so clearly the interval itself is an open set. By definition, intersection of an open interval and a subset is a open set in the subspace topology of that specific subset - so we get that $\{a\}$ is an open set in the subspace $\mathbb{N}$ of $\mathbb{R}$.

If $g$ is to be continuous, from the definition of continuity, we have that for each $x \in Y$, and each open neightborhood $G$ of $g(x)$, there must exist an open neightborhood $F$ of $x$, such that the image of this neightborhood with g must be a subset of $G$. So for $x=0$ there must exist an open set $U$ in the space $Y$ containing $0$, such that the image of this set with $g$ is a subset of $\{a\}$, as we have shown that is an open neightborhood of $a$.

But every open set contaning $0$ also contains another point:

Take an open set $A$ containing $0$, and because open intervals form a base of the standard topology on $\mathbb{R}$, we can find an open interval $(x,y)$ inside $A$ containing $0$ - clearly $y > 0$, and because for each real number we can find a larger integer, take $n> (y^{-1})$ - we have now found an $n \in \mathbb{N}$ such that $n^{-1} < y$, so every open set in $\mathbb{R}$ also contains a point $n^{-1} $ for some $n \in \mathbb{N}$, distinct from $0$, and because this point is in $Y$, we have also shown we can find such a point, distinct from $0$, in any open set of $Y$ subspace of $\mathbb{R}$ containing $0$.

Because $g$ is a bijection, the image of any open neightborhood of $0$ in $Y$ will contain at least two elements, and so it cannot possibly be a subset of $\{a\}$, therefore $g$ is not continuous, and so $f$ is not a homeomorphism, as $f^{-1} = g$.

Thus, no bijection between $X$ and $Y$ is a homeomorphism, and so a homeomorphism between $X$ and $Y$ does not exist. In particular, the function

$f(n) = \begin{cases} \frac{1}{n} & n\in \mathbb{N}\\ 0 & n =0 \\ \end{cases}$

is not a homeomorphism.


If you're endowing these with the subset topology, then every point of $X$ is open, since each point is the intersection of some open set in $\mathbb R$ and all of $X$. For $Y$ every point is an open set, except $0$. You'd expect continuity problems with this. For example $f(\{0\})$ maps an open set to a non-open set. So $f^{-1}$ can't be continuous, and $f$ can't be a homeomorphism. (All homeomorphisms are open maps.)

  1. Any subset of a metric space is also a metric space under the induced metric.

  2. If $f^{-1}$ was continuous, then $0=f^{-1}(0)=f^{-1}(\lim 1/n)=\lim f^{-1}(1/n)=\lim n$.

This is a contradiction.


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.