Give an example of a metric space $(X, d)$ and a subset $E \subseteq X$ where $E$ contains an interior point that is not a limit point.

Hint: You won't be able to find an example of $E$ when $X = \mathbb{R}^2$ and $d$ the usual metric. Consider the discrete metric on $\mathbb{R}^2$.

From what I understand, the discrete metric $d(p, q)$ is equal to $1$ when $p \ne q$, and $0$ when $p = q$. In other words, the only "distance" you can have with the discrete metric is either $1$ or $0$, right?

I'm pretty stuck on this problem. Any point $p$ in some subset $E \subseteq \mathbb{R}^2$ can only have one neighborhood, $N_1 (p)$, since $N_0 (p)$ is a singleton, and singletons cannot be neighborhoods. However, it seems for $p$ to have a neighborhood contained in $E$ (i.e. to be an interior point), it needs to be $\ge 1$ "distance" from the boundary of $E$, but then this would also make it a limit point, since all neighborhoods of $p$ will have some point $q \in E$. How should I tackle this problem? Any tips/hints would be great. Thank you.

  • $\begingroup$ There's a trick. "Normally" if $z$ is an interior point of E so that $N(z)\subset X$ then there are $y \in N(z)\subset E$ and if $y \ne z$ then that mean $z$ must also be a limit point of $E$. The only way that can't happen is if there AREN'T any $y \in N(z)$ so that $y \ne Z$. SO that means $N(z) = \{z\}$ (!!!!!). But "normally" that is impossible because $\{n\}$ is normally not an open neighborhood. So ... that's the hint. Find metric space so that $N(z) = \{x \in X| d(x,z) < \epsilon\} = \{z\}$. The discreet space will do that, if $\epsilon < 1$. $\endgroup$ – fleablood Sep 14 '17 at 0:15
  • $\begingroup$ "and singletons cannot be neighborhoods." BINGO!!!!! That is the trick! For this to be true a singleton MUST be a neighborhood. In order for singletons to be neighborhoods you must use a different metric. In the discreet metric, singletons must certainly ARE neighborhoods. EVERY set is a neighboorhood. Hint: Let $N_{\frac 12}(0) = \{x\in X|d(x,z)< \frac 12\} = ?????$. That is a neighborhood by definition. $\endgroup$ – fleablood Sep 14 '17 at 0:20

You write "Any point $p$ in some subset $E\subseteq\mathbb{R}^2$ can only have one neighborhood, $N_1(p)$, since $N_0(p)$ is a singleton, and singletons cannot be neighborhoods." Actually, every point in $\mathbb{R}^2$ has many neighborhoods. In the topology given by the discrete metric, every subset of $\mathbb{R}^2$ is open. Therefore any subset of $\mathbb{R}^2$ will be a neighborhood of each of its points. In particular, the singleton $\{x\}$ is a neighborhood of $x$ because $\{x\}=B(x,1)$, where I am using $B(c,r):=\{y\in\mathbb{R}^2 \mid d(c,y) < r\}$ to denote the open ball centered at $c$ or radius $r>0$.

Let $x\in\mathbb{R}^2$ and set $A:=\{x\}$. Then $x$ is an interior point to $A$ because the open ball $B$ centered at $x$ of radius $1/2$ is equal to $\{x\}$, and hence $x\in B =\{x\} \subseteq A$. However $x$ is not a limit point of $A$ because $B$ is a neighborhood of $x$ which contains no points of $A$ other than $x$; i.e., in symbols, $B\cap (A\setminus\{x\})=\emptyset$.

  • $\begingroup$ I thought the discrete metric meant that distances can only be 0 and 1. So, how can we have a radius of 1/2? Doesn't that violate the definition of a discrete metric? Clearly there's something I'm not getting, and an explanation would be great. Thanks. $\endgroup$ – Max Sep 13 '17 at 23:49
  • $\begingroup$ @Max define ball in a metric space $\endgroup$ – user223391 Sep 13 '17 at 23:52
  • 1
    $\begingroup$ @Max Just because there is no point $y\in\mathbb{R}^2$ distinct from $x$ such that $d(x,y)<1$ doesn't mean we can't consider the open ball $B(x,1/2)$ (or $B(x,1/3)$ or $B(x,1)$ or $B(x,r)$ for $r>0$). It just means that $B(x,1/2)$ happens to be equal to $\{x\}$. In a metric space, we consider a set $A$ to be open if, for every point $a\in A$ there is $r>0$ such that $B(a,r)\subseteq A$. By taking any $0<r\le 1$, we get $B(a,r)=\{a\}$ (which is clearly a subset of $A$ if $a\in A$), and hence we have shown that every subset is open. $\endgroup$ – John Griffin Sep 13 '17 at 23:53
  • $\begingroup$ "I thought the discrete metric meant that distances can only be 0 and 1" Exactly! So if $N(z) = \{d(x,z) < 1/2\}$ then $N(z) = \{d(x,z) = 0\}=\{z\}$ because $0$ is the only acceptable distance less than $1/2$. A radius does not have to be a legitimate distance. A neighborhood with radius 1/2 is all points that have a distance of LESS than 1/2. And $0 \le \frac 12$, isn't it? That does NOT mean that there are any actual points with a distance of 1/2. $d(x,y) = \frac 12$ is impossible. But $d(x,y) =0 < \frac 12$ is NOT impossible. In fact it means $x$ MUST equal $y$. $\endgroup$ – fleablood Sep 14 '17 at 0:29
  • $\begingroup$ A radius of a set is a real number. It does not have to be a legitimate distance. Example a radius can be negative. If so, say $r = -3$ then the neighborhood is $\{x|d(x,y) < -3\} = \emptyset$. It's an empty neighborhood be it is well defined. Likewise let $X =\mathbb Q$. Let $r = \pi$. Then the neighborhood is $\{x\in \mathbb Q| d(x,y) < \pi\}$. That is perfectly okay, even though $d(x,y) = \pi$ is impossible. Likewise $\{x|d(x,y) < 1/2\}$ is well defined even if it is the discreet metric. $\{x|d(x,y)< 12\} = \{x\}$. $\endgroup$ – fleablood Sep 14 '17 at 0:45

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.