2

I have an answer to this question which does not use the Hausdorff property, but I suspect there is an issue with my solution opposed to the question. I am aware that there are already solutions to this question (e.g. here ) but I am not quite able to understand the posted solution in full. I also understand that the result can be shown by utilising the the fact that the diagonal is closed and that projections onto components are continous

The question stated formally is:

Let $X, Y$ be topological spaces and have $f: X \mapsto Y$ a set map. Let us define $G(f) = \{(x,f(x)) \in X \times Y | x \in X\} $ as the graph of $X$. Then show that if f is continuous and Y hausdorff then $G(f)$ is closed.

Definition of closure I use is the following:

Closure point of a subset $A$ is any point $x$ with the property that any nbhd $V$ of $x$ has $V \cap A \neq \emptyset$.

My solution:

Let $x^* = (x,f(x))$ be an arbitrary point in $G(f)$. If we consider $V$ in the nbhd of $f(x)$, then continuity gives us that $f(U) \subset V$ where $U$ is in the nbhd of $x$.

Then denoting $U^* = U \times V$, we have $U^* \cap G(f) = (U \times V) \cap G(f) \subset (U \times f(U)) \cap G(f) \neq \emptyset $

Since $U^*$ is a nbhd of $x^*$ and the choice of $x^*$ was arbitrary, we can conclude that $G(f)$ is closed.

  • 4
    You seem to have just showed that for every point $x^$ in the graph, there is a neighborhood whose intersection with the graph is non-empty. But that is completely obvious because you took a point in the graph to begin with. What you need to do is take a point in the closure* of the graph, and show it actually belongs to the graph (using continuity of $f$ and Hausdorffness). Also your use of continuity is not phrased clearly; what comes first? $U$ or $V$? Universal or existential quantifer? – peek-a-boo May 29 '24 at 14:08
  • Is the graph of $id : X \to X$ always closed? See https://math.stackexchange.com/q/1920963 – Paul Frost May 29 '24 at 22:33
  • @peek-a-boo I was working under the assumption that the closure of a set is the set of all closure points. I attempted to take an arbitrary point in the graph of $f$ and show that it is always in the closure proving that the graph is closed (since every point is in the closure).

    Thanks for the point on continuity. I have rephrased the question.

    – JamesLevine May 30 '24 at 08:09
  • 1
    ok and once again, your argument doesn’t really prove much because you’re simply saying that every point of a set $A$ is a “closure point of $A$” (not standard terminology), i.e $A\subset\overline{A}$ (and this is true in full generality for any subset of any topological space). But to show that $A$ is closed, you need to show the reverse inclusion. – peek-a-boo May 30 '24 at 09:08
  • @CatharticEncephalopathy Seems like this is not a duplicate. This question is about continuous $\implies$ closed graph, but linked question is closed graph $\implies$ continuous. – Mike Earnest Sep 01 '24 at 15:15
  • In the paper mentioned the equivalent characterizations are mentioned @MikeEarnest. Read answer again – Clemens Bartholdy Sep 01 '24 at 15:36
  • 1
    I read the answer carefully. Their answer assumes “closed graph”, and derives that four things are equivalent. No where does that answer prove that a graph is closed assuming continuity. – Mike Earnest Sep 01 '24 at 15:40
  • 1
    https://math.stackexchange.com/questions/2488309/f-is-continuous-if-and-only-if-its-graph-is-closed-in-x-times-y?rq=1 @Mike Earnest this is another dupe target – Clemens Bartholdy Sep 01 '24 at 16:07

2 Answers2

2

You already seem to have gotten satisfactory answers to your concerns with your proof attemts, so instead here's an alternative solution if you're familiar with some of the basics of nets.

Let $(x,y)\in\overline{\operatorname{graph}(f)}$. Then there exists a net $\{(x_\alpha,f(x_\alpha))\}_{\alpha\in I}$ in $\operatorname{graph}(f)$ with limit $(x,y)$. As projections are continuous, we then have that $x_\alpha\to x$ and $f(x_\alpha)\to y$. Furthermore, as $f$ is continuous, $f(x_\alpha)\to f(x)$. As $Y$ is Hausdorff, nets in $Y$ have unique limits, and so $y=f(x)$. Consequently $(x,y)=(x,f(x))\in\operatorname{graph}(f)$, proving the claim.

Lorago
  • 12,173
1

Let $x \times y \in {\rm Cl} G (f)$ be arbitrary. We show that in fact $x \times y \in G (f)$, and since $x \times y$ was arbitrary, $G (f)$ must be closed.

Note that $x \times y \in G (f) \Longleftrightarrow y = f (x)$ by definition of "graph". Suppose for the sake of contradiction that $y \ne f (x) = z$. Since $Y$ is Hausdorff, there exist disjoint neighborhoods $U$ of $y$ and $V$ of $z$. Since $f$ is continuous, $f^{-1} (V)$ is a neighborhood of $x$. Since $f^{-1} (V) \times U \subseteq X \times Y$ is a neighborhood of $x \times y$ (by the definition of the product topology), it intersects $G (f)$ by hypothesis, say at a point $x' \times f (x')$. The first factor shows $x' \in f^{-1} (V) \implies f (x') \in V$, and the second factor shows $f (x') \in U$, which imply $f (x') \in U \cap V$. This contradicts the fact that $U$ and $V$ are disjoint.

Hausdorffness is required here. Consider for example $X = \{ 0, 1 \}$ in the discrete topology and $Y = \{ 0, 1 \}$ in the indiscrete topology. Define $f: X \to Y$ by $f (x) = 0$, which is continuous. The graph $G (f) = \{ 0 \times 0, 1 \times 0 \}$ is not closed in $X \times Y$. You can check this: every neighborhood of the point $0 \times 1 \notin G (f)$ contains the point $0 \times 0 \in G (f)$.

K. Jiang
  • 10,436
  • Thank you for the answer. It's all clear to me apart from one point - how do we know that $f^{-1}(V) \times U$ has a non empty intersection with $G(f)$? I don't quite see how this open product being a neighborhood of $x \times y$ necessarily gives us this non empty intersection. Is it because of the original assumption that $x \times y$ is in the closure of $G(f)$? – JamesLevine May 30 '24 at 10:11
  • 1
    @JamesLevine Exactly. Since $x \times y \in {\rm Cl} G (f)$ is assumed, then every neighborhood of $x \times y$, including $f^{-1} (V) \times U$, intersects $G (f)$. – K. Jiang May 30 '24 at 13:31