My proof is very different from my reference, hence I am wondering is I got this right?

Apparently, $F$ is continuous, and the identity matrix is closed. Now we want to show that the preimage of continuous function on closed set is closed.

Let $D$ be a closed set, Consider a sequence $x_n \to x_0$ in which $x_n \in f^{-1}(D)$, and we will show that $x_0 \in f^{-1}(D)$.

Since $f$ is continuous, we have a convergent sequence $$\lim_{n\to \infty} f(x_n) = f(x_0) = y.$$

But we know $y$ is in the range, hence, $x_0$ is in the domain. So the preimage is also closed since it contains all the limit points.

Thank you.

  • 2
    $\begingroup$ Your last sentence needs to be rephrased. $y$ must be in $D$ (since $D$ is closed and each $f(x_n)\in D$); thus $x_0$ is in $f^{-1}(D)$. $\endgroup$ Jul 7, 2013 at 1:33
  • $\begingroup$ Thanks @DavidMitra - that indeed is what I intended. $\endgroup$ Jul 7, 2013 at 1:35
  • $\begingroup$ would you add the reference for your reference? $\endgroup$ Sep 11, 2018 at 23:09

3 Answers 3


Yes, it looks right. Alternatively, given a continuous map $f: X \to Y$, if $D \subseteq Y$ is closed, then $X \setminus f^{-1}(D) = f^{-1}(Y \setminus D)$ is open, so $f^{-1}(D)$ is closed.

  • 7
    $\begingroup$ However, why you can make the assertion that $X \setminus f^{-1}(D) = f^{-1}(Y \setminus D)$? $\endgroup$ Jul 7, 2013 at 2:01
  • 4
    $\begingroup$ @Jellyfish Just do an element chase to prove that they're equal. It has nothing to do with topology. $\endgroup$
    – Ink
    Jul 7, 2013 at 2:09
  • 4
    $\begingroup$ $X=f^{-1}(Y)$ requires $f$ to be a total function. $\endgroup$
    – Witiko
    Apr 1, 2018 at 7:42
  • 1
    $\begingroup$ Would you please add some reference on this? $\endgroup$ Sep 11, 2018 at 23:08
  • 6
    $\begingroup$ That doesn't seem necessary, as generally the definition of a function includes totality. $\endgroup$
    – Not Legato
    Mar 17, 2020 at 17:16

Yes, your proof is correct, but since you are using sequences this works on metric spaces, not on topological ones.

  • $\begingroup$ This answer interests me. If I want to work on topological ones, how should I do? Does the discussion on metric spaces not be included in the topological space? $\endgroup$ Aug 4, 2016 at 20:50
  • $\begingroup$ @sleevechen I think you probably need the topological space to be Hausdorff $\endgroup$ Mar 23, 2020 at 18:40

Let me add an analogous approach that works for topological spaces. We need to introduce an additional concept that is basically a generalisation of sequences for topological spaces.

Let $S$ be a directed set (reflexive, transitive and any pair of elements has a upper bound) and let $X$ be a topological space. Then a net is a function $x: S \rightarrow X$. We often use the notation $x_{\alpha}$ instead of $x(\alpha)$ for evaluation to mirror the notation used for sequences. And we say that a net converges to a point $x \in X$ when for all neighbourhoods of $x$ $N_{x}$ there exists an $\alpha_{0}$ such that

$$\left \{ x_{\alpha}:\alpha \geq \alpha_{0} \right \} \subset N_{x}.$$

We denote this fact by $x_{\alpha} \rightarrow x$.

Using this we can characterise closed set for general topological spaces in a "sequential manner". Let's compare the two characterisations that may be used:

  • In a first countable space (every metric space is first countable), $A$ is closed iff for every sequence $(x_{n}) \subset A:x_{n} \rightarrow x$ we have $x \in A$.
  • In a topological space (no additional separation assumptions are needed), $A$ is closed iff for every net $(x_{\alpha}) \subset A:x_{\alpha} \rightarrow x$ we have $x \in A$.

Note that we also have a theorem for nets that states that we can interchange a continuous function and the limiting operation. Thus allowing us to do exactly (notation wise) what 1LiterTears did in the initial post (consider the same setup, just replace sequence with net):

$$\text{lim}f(x_{\alpha}) = f(\text{lim}x_{\alpha})=f(x).$$

This of course allows us to also get the "same result" if we replace sequence by net in the initial statement.

In short, nets are a great way to move from first countable spaces to general topological spaces while still being able to recycle some of the "standard arguments" from analysis.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .