I have the following question:

Let $\mathbb{R}_d$ denote $\mathbb{R}$ with the discrete topology. $\mathcal{B}(X)$ denotes the base or basis of the topology of $X$. Then show that $\mathcal{B}(\mathbb{R}_d) \times\mathcal{B}(\mathbb{R}_d) \neq \mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d).$

The $\mathcal{B}(\mathbb{R}_d)$ is the family of singletons and each singleton is both open and closed.

Another posting regarding the product of discrete topologies , Product Topology of Discrete Sets, seems to make sense but I can't reconcile that explanation with what I am asked to prove.

If we consider the set $S = \{ (x,y) : x = y \}$ , then $S$ is closed in $\mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d)$. But $S$ is not closed in $\mathcal{B}(\mathbb{R}_d) \times\mathcal{B}(\mathbb{R}_d)$. I do not understand why that is.

Any clarification would be appreciated.

Edit: The problem above is a modification of a problem from Folland's Real Analysis Text (Chapter 7, Exercise 29):

If $X$ is a set of cardinality > $\mathfrak{c}$ with the discrete topology, hen $\mathcal{B}_{X \times x}\neq \mathcal{B}_X \otimes \mathcal{B}_X$. In fact, $D = \{ (x,y) : x=y \}$ is closed but not in $\mathcal{B}_X \otimes \mathcal{B}_X$ (Use Exercise 5 in Section 1.2 and Proposition 1.23). If $D \in \mathcal{B}_X \otimes \mathcal{B}_X$ then $D \in \mathcal{M}$ where $\mathcal{M}$ is a sub-$\sigma$-algebra of $\mathcal{B}_X \otimes \mathcal{B}_X$ generated by a countable family of rectangles, hence $D \in \mathcal{N} \otimes \mathcal{N}$ where $\mathcal{N}$ is a countably generated sub-$\sigma$-algebra of $\mathcal{B}_X$. Then $\{x \} \in D_x \in \mathcal{N}$ for all $x$ but card$(\mathcal{N}) \leq \mathfrak{c}$. The same reasoning applies if $X$ is replaced by its one-point compactification.

I will re-read the chapter again. I am missing something fundamental.

  • 2
    $\begingroup$ "Is a basis ...". There could be many distinct bases... $\endgroup$ – Eric Towers Nov 21 '17 at 4:02
  • $\begingroup$ I see nothing at your link claiming that $S$ is not closed in the topology with basis $\mathcal{B}(\mathbb{R}_d) \times \mathcal{B}(\mathbb{R}_d)$. Where does the claim that $S$ is not closed originate? $\endgroup$ – Eric Towers Nov 21 '17 at 4:18
  • $\begingroup$ @Eric Towers, I was referencing, perhaps incorrectly, an exercise from Folland's Real Analysis Text. It's Exercise 29 on page 231 of the second edition. $\endgroup$ – testguy Nov 21 '17 at 4:26
  • $\begingroup$ @Eric TowersI can't discern the difference between $\mathcal{B}(\mathbb{R}_d) \times\mathcal{B}(\mathbb{R}_d)$ and $ \mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d)$. That's my problem. $\endgroup$ – testguy Nov 21 '17 at 4:31
  • 1
    $\begingroup$ It's not about the base but about the Borel sets of those spaces! $\endgroup$ – Henno Brandsma Nov 21 '17 at 22:38

We take for a basis of the discrete topology the collection of singleton subsets, $\mathcal{B}(\mathbb{R}_d) = \{ \{x\} : x \in \mathbb{R}\}$ and $\mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d) = \{ \{(x,y)\} : x \in \mathbb{R}, y \in \mathbb{R} \}$.

Let $x \in \mathbb{R}$. Then $\{x\} \in \mathcal{B}(\mathbb{R}_d)$ and $\{x\} \times \{x\} \in \mathcal{B}(\mathbb{R}_d) \times \mathcal{B}(\mathbb{R}_d)$.

But, although $\{(x,x)\} \in \mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d)$, $\{(x,x)\} \not \in \mathcal{B}(\mathbb{R}_d) \times \mathcal{B}(\mathbb{R}_d)$. (Why? $\{x\} \times \{x\} = (\{x\},\{x\}) \neq \{(x,x)\}$.)

Consequently, $\mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d) \neq \mathcal{B}(\mathbb{R}_d) \times \mathcal{B}(\mathbb{R}_d)$

There's a basic "type mismatch" between elements of $\mathcal{B}(\mathbb{R}_d \times \mathbb{R}_d)$ and elements of $\mathcal{B}(\mathbb{R}_d) \times \mathcal{B}(\mathbb{R}_d)$. These can be tricky to notice, but once you do, they're hard to stop noticing...

| cite | improve this answer | |
  • $\begingroup$ Thank you very much Eric! It makes sense. $\endgroup$ – testguy Nov 21 '17 at 4:35
  • $\begingroup$ I don't know that I buy this reading of the question. it's perfectly normal in mathematics to overload notation based on the type of its arguments. I mean, using the usual definitions of natural numbers we have $1 = \{\{\}\}$ and $2 = \{\{\}, \{\{\}\}\}$, but nobody would argue that $$1 \times 2 = \{(\{\}, \{\}),(\{\}, \{\{\}\})\}$$ and is therefore not a number at all. Clearly anyone writing $\mathcal{B} \times \mathcal{B}$ intends to indicate the set $\{ b \times b' \; | \; b, b' \in \mathcal{B}\}$, not the literal Cartesian product of $\mathcal{B}$ with itself. $\endgroup$ – Daniel McLaury Nov 21 '17 at 5:00
  • $\begingroup$ @Daniel, the question is a modification of a question from Folland's Real Analysis, Exercise 29 in Chapter 7. I understand your point. I will post the question from Folland's text. $\endgroup$ – testguy Nov 21 '17 at 13:53
  • $\begingroup$ I posted the question from Folland's text. Thanks. I'll re-read the material again. $\endgroup$ – testguy Nov 21 '17 at 14:24

Prof. Brandsma pointed out that we are dealing with Borel sets not bases in the comments. Also, what I asked to prove is not true. My apologies.

If the space consisted of the set of ordinals less than or equal to the first uncountable ordinal $\omega_1$, with the order topology, then we can prove it. Exercise 7.28 on page 231 in Folland's Real Analysis states this. But $\mathbb{R}_d$ is not ordered.

There is a book by S. M. Srivastava on Borel sets: A Course on Borel Sets. This helped clear up my confusion as well as provide a solid proof.

Srivastava credits the proof to B.V Rao. Apparently Ulam first posed a question like this. B.V Rao provided the following proof\cite{sms}. First we identify the notation $\mathcal{B}(\mathbb{R}_d) = \mathscr{P}(\mathbb{R})$ and $\text{card}((\mathbb{R}_d))= \aleph_1 = \mathfrak{c}$ so as not to get confused. Also, $\text{card}(\omega_1) = \aleph_1.$ Under the Continuum Hypothesis, the following holds true.

Proposition: $\mathcal{P}(\omega_1) \otimes \mathcal{P}(\omega_1) = \mathcal{P}(\omega_1 \times \omega_1)$.

Proof. Let $A \subseteq \omega_1 \times \omega_1$. Write $A = B \cup C$, where $$B = A \bigcap \{(\alpha, \beta) \in \omega_1 \times \omega_1: \alpha \geq \beta\}$$ and $$C = A \bigcap \{(\alpha, \beta) \in \omega_1 \times \omega_1: \alpha \leq \beta\}$$ We shall show that $B$ is in the product $\sigma$-algebra. By symmetry, it will follow that $C$ is also in the product $\sigma$-algebra. The result then follows.

For each $\alpha < \omega_1$, $B_\alpha$ is countable, say $B_{\alpha} = \{\alpha_0,\alpha_1,\alpha_2, \dots \}$. Since the graph of every measurable function from a measurable space to a discrete measurable space whose cardinality is at most $\mathfrak{c}$ is measurable, $$G_n = \{ (\alpha, \alpha_n) : \alpha \in \omega_1 \}$$ is in the product of discrete $\sigma$-algebras. Now we note that $B = \bigcup_n G_n$.

If our space $X$ is Hausdorff and has cardinality greater than that of the continuum, then $\mathcal{B}(X \times X) \neq \mathcal{B}(X) \times \mathcal{B}(X).$ This is Exercise 7.29, page 231 in Folland. The proof is from Bogachev's Measure Theory Volume 2.

| cite | improve this answer | |

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.