Let $X, Y$ be topological spaces, $Y$ an ordered set in the order topology and $f_i:X\to Y, i=1,2$ be continuous. Show that $A=\{x:f_1(x)<f_2(x)\}$ is open.

(This differs from the related questions I found in that here $Y$ has only a topological structure; in the other questions (1, 2) I found $Y$ was the real line or a metric space)

I think I was able to come up with a proof (see below), but it feels like it is more complicated than it had to be. Any ideas for a shorter, tidier alternative?

We show that $A$ is open by showing that for every $x\in A$ there exists open $U$ with $x\in U\subseteq A$.

So let $x\in A$; then $f_1(x)\not=f_2(x)$, and since $Y$ is Hausdorff we may find disjoint open $V_i\ni f_i(x)$; and then since open intervals form a basis, we may find open intervals $I_i=(a_i,b_i)\ni f_i(x)$ such that $I_i\subseteq V_i$ and thus $I_1\cap I_2=\emptyset$.

Lemma. For all $t_i \in I_i$, we have $t_1 < t_2$.

Using this lemma then, put

$$ U = \bigcap_i f_i^{-1}(I_i) $$

Since $f_i$ is continuous and $I_i$ is open, $f_i^{-1}(I_i)$ is open and then so is $U$; also since $f_i(x)\in I_i$, we have that $x\in U$, and finally, to show that $U\subseteq A$, let $u \in U$; then $f_i(u)\in I_i$ and thus by the lemma, $f_1(u) < f_2(u)$ so that $u\in A$, q.e.d.

Proof of Lemma. Let $t_i \in I_i$, and assume by way of contradiction $t_2 \le t_1$. If $t_2=t_1$ then $I_1\cap I_2\not=\emptyset$, contradiction. So try $t_2 < t_1$. In that case we have two options.

$t_1 < b_2$. In that case, from $t_2 < t_1$ and $a_2 < t_2$ (since $t_2 \in I_2$) we deduce that $t_1 \in I_2$ again contradicting $I_1\cap I_2=\emptyset$.

So $b_2 \le t_1$. Then since $f_2(x)\in I_2$ we deduce that $f_2(x) < t_1$, and since $t_1\in I_1$, we have that $t_1 < b_1$ whence we deduce that $f_2(x) < b_1$; from $f_1(x)\in I_1$ we deduce that $a_1 < f_1(x)$ and then from $f_1(x) < f_2(x)$ we deduce that $a_1 < f_2(x)$, which now means that $f_2(x) \in I_1$, again contradicting $I_1 \cap I_2 = \emptyset$.


Here’s an approach that isn’re really essentially different but may be a little neater. Define

$$f:X\to Y\times Y:x\mapsto\langle f_1(x),f_2(x)\rangle\;.$$

If $V$ and $W$ are open sets in $Y$, $f^{-1}[V\times W]=f_1^{-1}[V]\cap f_2^{-1}[W]$ is open in $X$, so $f$ is continuous. Now let $U=\{\langle y_1,y_2\rangle\in Y\times Y:y_1<y_2\}$, the subset of $Y\times Y$ lying strictly above the diagonal; $U$ is open in $Y\times Y$, so $A=f^{-1}[U]$ is open in $X$.

To prove that $U$ is open, let $\langle y_1,y_2\rangle\in U$. If there is a $z\in(y_1,y_2)$, let $V_1=(\leftarrow,z)$ and $V_2=(z,\to)$; otherwise let $V_1=(\leftarrow,y_2)$ and $V_2=(y_1,\to)$. In either case $V_1$ and $V_2$ are disjoint open nbhds of $y_1$ and $y_2$, respectively, so $V_1\times V_2$ is an open nbhd of $\langle y_1,y_2\rangle$. Clearly $v_1<v_2$ whenever $v_1\in V_1$ and $v_2\in V_2$, so $V_1\times V_2\subseteq U$, and $U$ is open.

You could use the same idea to simplify slightly the proof of your lemma.

  • $\begingroup$ That is indeed much neater, thanks. BTW $(z,\to)=\{z':z'>z\}$ I presume, right? $\endgroup$ May 6 '20 at 15:28
  • $\begingroup$ @whatadisgrace: You’re welcome. Yes, you understood the notation correctly; I prefer the arrow notation for rays to the more common $\pm\infty$ notation. $\endgroup$ May 6 '20 at 15:31
  • $\begingroup$ @BrianM.Scott you mean $f^{-1}[V\times W]=f^{-1}_1(V)\cap f^{-1}_2(W)$, right? $\endgroup$
    – PtF
    May 6 '20 at 15:34
  • $\begingroup$ @PtF: I sure do; thanks! $\endgroup$ May 6 '20 at 15:35

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.