1

A subspace of a completely regular space is completely regular. A product of completely regular spaces is completely regular.

My attempt: Let $Y\subseteq X$. Let $\{y\}$ be a singleton set in $Y$. Since $X$ is $T_1$, $\{y\}$ is closed in $X$. By theorem 17.2, $\{y\}= Y\cap \{y\}$ is closed in $Y$. Hence $Y$ is $T_1$. Let $x\in Y$ and $B$ is closed in $Y$ such that $x\notin B$ (i.e. $\{x\} \cap B =\emptyset$). By theorem 17.2, $B=C\cap Y$; $C$ is closed in $X$. If $x\in C$, then $x\in Y\cap C=B$. So $x\notin C$. Since $X$ is completely regular, $\exists f: X\to [0,1]$ such that $f$ is continuous, $f(x)=1$ and $f(C)=\{0\}$. By theorem 18.2, $f|_{Y}:Y\to [0,1]$ is continuous. Since $B\subseteq C$, we have $f|_Y(B)=f(B)\subseteq f(C)=\{0\}$. Hence $\exists f|_Y:Y\to [0,1]$ such that $f|_Y$ is continuous, $f|_{Y}(x)=f(x)=1$ and $f|_Y (B)=\{0\}$. Is this proof better version of Munkres proof? Munkres used theorem 17.4, $(\overline{B})_Y =\overline{B}\cap Y$. Which is special case of theorem 17.2.

I think, there is a typo, $f_i(X_{\alpha_i}-U_{\alpha_i})$ instead of $f_i(X-U_{\alpha_i})$. Munkres’ claim that $\phi_i:X\to [0,1]$ defined by $\phi_i(x)=f_i(\pi_{\alpha_i}(x))$ vanishes outside $\pi_{\alpha_i}^{-1}(U_{\alpha_i})$, i.e. $\phi_i( \prod X_\alpha -\pi_{\alpha_i}^{-1}(U_{\alpha_i}))=\{0\}$. Question: (1) How to prove $\pi_{\alpha_i}(\prod X_\alpha- \pi_{\alpha_i}^{-1}(U_{\alpha_i}))=X_{\alpha_i} -\pi_{\alpha_i}( \pi_{\alpha_i}^{-1}(U_{\alpha_i}))$? I known $\supseteq$ inclusion holds in general. (2) Munkres’ claim that $f$ vanishes outside $\prod U_\alpha$, i.e. $f(\prod X_\alpha - \prod U_\alpha)=\{0\}$. By theorem 19.5, $(\prod X_\alpha) - (\prod U_\alpha) =\bigcup_{\alpha \in J} \pi_{\alpha}^{-1}(X_\alpha \setminus U_\alpha)$. So $\phi_1 (\prod X_\alpha - \prod U_\alpha) =\phi_1 (\bigcup_{\alpha \in J} \pi_{\alpha}^{-1}(X_\alpha \setminus U_\alpha))=$$\bigcup_{\alpha \in J}\phi_1 (\pi_\alpha^{-1} (X_\alpha \setminus U_\alpha))=\bigcup_{\alpha \in J}\phi_1 (\prod X_\alpha -\pi_{\alpha}^{-1}(U_\alpha))$. Note $\phi_1(\prod X_\alpha -\pi_{\alpha_1}^{-1}(U_{\alpha_1}))=\{0\}$. Now how to show $f(\prod X_\alpha -\prod U_\alpha)=\{0\}$? (3) Can this result holds in box topology? Here I showed arbitrary product of regular space is regular in box topology.

Edit: My claim in (1), $\pi_{\alpha_i}(\prod X_\alpha- \pi_{\alpha_i}^{-1}(U_{\alpha_i}))=X_{\alpha_i} -\pi_{\alpha_i}( \pi_{\alpha_i}^{-1}(U_{\alpha_i}))$, is incorrect. The correct way to reach desired result, $\phi_i( \prod X_\alpha -\pi_{\alpha_i}^{-1}(U_{\alpha_i}))=\{0\}$, is in answer by Mateo.

user264745
  • 4,143

1 Answers1

1

(0) Yes, there's a typo.

(1)

  • $\prod X_\alpha- \pi_{\alpha_i}^{-1}(U_{\alpha_i}) = \pi_{\alpha_i}^{-1}\left(X_{\alpha_i} - U_{\alpha_i} \right)$;
  • $f(f^{-1}(A))=A$ if $f$ is surjective. Then $$\pi_{\alpha_i}\left(\prod X_\alpha- \pi_{\alpha_i}^{-1}(U_{\alpha_i})\right) = \pi_{\alpha_i}(\pi_{\alpha_i}^{-1}\left(X_{\alpha_i} - U_{\alpha_i} \right)) =X_{\alpha_i} -U_{\alpha_i}.$$

(2)

  • Recall that $U_\alpha = X_\alpha$ unless $\alpha\notin \{\alpha_1,\ldots \alpha_n\}$.
  • Let $x\notin \prod U_\alpha$. Then $x_{\alpha_i}\notin U_{\alpha_i}$ for some $1\leq i\leq n$.
  • We get $\phi_i(x)=f_i(x_{\alpha_i}) = 0$. Therefore $f(x)=0$.

(3) https://mathoverflow.net/questions/209661/which-topological-properties-are-preserved-under-taking-box-products

I'm not familiar with box topology, but I'll try to prove it. The problem with box topology is that we can't define $f$ this way, because we now need to take into account all the components (in the product topology almost all components were trivial, i.e. whole space) and we can't calculate the infinite product (possibly uncountable) of numbers. I'll try with infimum function.

Let $X=\prod_{\alpha\in A} X_\alpha$, where $X_\alpha$ are completely regular. On $X$ we consider box topology. Let $b=(b_\alpha)\in X$ and a basis element $U=\prod U_\alpha$ such that $b\in U$. Let $f_\alpha\colon X_{\alpha}\to [0,1]$ be continuous such that $f_\alpha(b_\alpha)=1$ and $f_\alpha\equiv 0$ outside $U_\alpha$. Now define $$\phi_\alpha(x) = f_\alpha(\pi_\alpha(x)),\quad f(x)= \inf_{\alpha\in A} \phi_\alpha(x).$$

  • Of course $f(b)=\inf f_\alpha(b_\alpha)=1$.
  • If $x\notin U$ then $x_\alpha\notin U_\alpha$ for some $\alpha$. Then $$\phi_\alpha(x)=f_\alpha(\pi_\alpha(x))=f_\alpha(x_\alpha)=0,$$ so $f(x)=0$. Therefore $f\equiv 0$ outside $U$.

It suffices to show the continuity of $f$. First, observe that $\phi_\alpha$ is continuous as a composition of two continuous functions. We'll show that $f^{-1}(S)$ is open for any $S$ from the subbase of the Euclidean topology.

To do it we observe that for any set $P\subset \Bbb R$:

  • $\inf P<t \iff \exists_{p\in P}\,p<t$;
  • $\inf P>t \iff \exists_{s>t}\,\forall_{p\in P}\, p>s$.

Now we are ready to prove the continuity.

  • Let $S=(-\infty,t)$. Then $$f^{-1}(S) = \{x\in X\,|\,\phi_\alpha(x)<t\text{ for some }\alpha\} = \bigcup_\alpha \phi_\alpha^{-1}(S),$$ which is an open set.
  • Let $S=(t,\infty)$. Then $$f^{-1}(S) = \{x\in X\,|\,\inf\phi_\alpha(x)>t\} = \bigcup_{s>t} \{x\in X\,|\,\phi_\alpha(x)>s\text{ for all }\alpha\}=\ldots$$ $$\ldots = \bigcup_{s>t} \{x\in X\,|\,\pi_\alpha(x)\in f_\alpha^{-1}((s,\infty))\text{ for all }\alpha\} = \bigcup_{s>t} \prod_\alpha f_\alpha^{-1}((s,\infty)) $$ which is an open set as a sum of basis sets.
user264745
  • 4,143
Mateo
  • 4,956
  • Thank you so much for the answer. How did I forgot first bulletin of (1)! Your proof of (2) is slick. It should be $f_i(x_{\alpha_i})=0$ instead of $f_i(x_i)=0$. Another(very similar) way to show it, $\prod X_\alpha -\prod U_\alpha=\bigcup_{\alpha \in J} \pi_\alpha^{-1}(X_\alpha -U_\alpha)= \bigcup_{i \in J_n} \pi_\alpha^{-1}(X_{\alpha_i} -U_{\alpha_i})$, since $X_\alpha =U_\alpha$, $\forall \alpha \in J-{\alpha_1,..,\alpha_n}$. Let $x\in \prod X_\alpha -\prod U_\alpha$. Then $x\in \pi_\alpha^{-1}(X_{\alpha_i} -U_{\alpha_i})= \prod X_\alpha - \pi_{\alpha_i}^{-1}(U_{\alpha_i})$, – user264745 Jun 06 '22 at 02:56
  • for some $i\in J_n$. By (1) or since $\phi_i$ vanishes in $\prod X_\alpha - \pi_{\alpha_i}^{-1}(U_{\alpha_i})$, we have $\phi_i(x)=0$. Thus $f(x)=0$. Our desired result. – user264745 Jun 06 '22 at 02:56
  • I read that mathoverflow post. It is written, completely regular is preserved under box topology. But proof is not given. I guess proof is just generalisation of product topology, i.e. define $f:X\to [0,1]$ as $f(x)= \prod_{i\in J} \phi_i(x)$. How to show $\prod_{i\in J} \phi_i(x)$ is continuous? – user264745 Jun 06 '22 at 10:49
  • I added a part about box topology. Please check if everything is correct, as I've never considered box topology before. – Mateo Jun 06 '22 at 17:55
  • Everything looks nice. Since $0\leq \phi_\alpha (x)\leq 1$, $\forall \alpha \in A$, we have $f(x)=\inf_{\alpha \in A} \phi_\alpha (x)\in [0,1]$. I don’t know about $f^{-1}(S) = {x\in X,|,\phi_\alpha(x)<t\text{ for some }\alpha} = \bigcup_\alpha \phi_\alpha^{-1}(S)$ and $\bigcup_{s>t} {x\in X,|,\pi_\alpha(x)\in f_\alpha^{-1}(s,\infty)\text{ for all }\alpha} = \bigcup_{s>t} \prod_\alpha f_\alpha^{-1}(s,\infty)$ identity. Can you use different definition of continuity? – user264745 Jun 06 '22 at 18:28
  • Please be more specific. The problem is with the equalities (why left-hand side is equal to the right-hand side) or why to show continuity I used the preimages of semilines? – Mateo Jun 06 '22 at 21:10
  • Yes, why LHS=RHS. I’m sorry for being vague. – user264745 Jun 06 '22 at 21:11
  • 1
    I added some explanation. The last equality you ask for goes directly from the definition of the product and projections: ${x|\pi_\alpha(x)\in A_\alpha\text{ for all }\alpha} = \prod A_\alpha$. – Mateo Jun 06 '22 at 22:38
  • You have 220 iq. I proved all those equalities, which I was feeling uncomfortable with. I wanna ask, how did you thought to write $f^{-1}((-\infty,t))=\bigcup_\alpha \phi_\alpha^{-1}(S)$? because I lack intuition. First you showed $f:X\to \Bbb{R}$ is continuous. $f(X)\subseteq [0,1]$. By theorem 18.2(e), $f:X\to [0,1]$ is continuous. Our desired result. In last sentence you said “sum of basis sets”. I would have said, arbitrary union of basis sets, hence it’s open. Thank you once again! – user264745 Jun 07 '22 at 09:32
  • 1
    Thank you. I just needed to calculate this preimage and interpret the infimum in the definition of f. It's nice to have in mind the equivalent descriptions of the inequalities present in my last edit: $\inf P<t \iff\cdots$ and $\inf P>t \iff\cdots$. – Mateo Jun 07 '22 at 10:55