I ask this question only to have a clarification about the consequences of Halmos’ definition of Infinite Cartesian Product. Let $\{X_i\}_{i\in A}$ be an indexed family of subsets of a given set $X$. As you know $\displaystyle\prod_{\substack{i\in A}}X_i$ is defined to be the set of all functions with domain $A$ and codomain $\displaystyle\bigcup_{\substack{i\in A}}X_i$ such that $f(i)\in X_i ,\forall i\in A$. Now, let $\pi_i$ be the generic projection map and let $E\subseteq X_i$: how can $\pi_i^{-1}(E)$ be equal to $\displaystyle\prod_{\substack{j\in A}}E_j$ with $E_j=E$ if $j=i$ and $E_j=X_j$ otherwise? I mean, the codomain is relevant in this definition...

  • $\begingroup$ I assume the $f$ is actually meant to be $\pi_i$. $\endgroup$ – Theo Bendit Sep 18 '18 at 10:12
  • $\begingroup$ The result seems natural enough to me. What problem are you having with it? $\endgroup$ – Theo Bendit Sep 18 '18 at 10:16
  • $\begingroup$ As the domain of f is A, $f^{-1}(E)$ is a subset of A. $\endgroup$ – William Elliot Sep 18 '18 at 10:20
  • $\begingroup$ What is the final $f$ in $f^{-1}[E]$? $\endgroup$ – Henno Brandsma Sep 18 '18 at 10:51
  • $\begingroup$ It was $\pi_i$ indeed, thanks for noticing! $\endgroup$ – Daniele Leo Sep 19 '18 at 5:32

I agree with the definition of the Cartesian product, that's completely standard (a set of functions $f$ from the index set $I$ such that $f(i) \in X_i$ for all $i \in I$). The codomain is "relevant" in the sense that it allows us to justify this as a set (from the axioms): a function is a subset of a product $A \times B$ (defined as sets of ordered pairs, e.g. with Kuratowski's definition $(a,b) := \{\{a\},\{a,b\}\}$ etc.) where $A$ and $B$ are already given to be sets. In this case we want $A = I$ (a set) and as $B$ we choose $\bigcup_i X_i$ (a set by the union axiom), because we already "know" that all values of $f$'s in the Cartesian product we want to define are going to be in this union. So this way we can garantuee from the pairs/union/comprehension etc. axioms that $\prod_i X_i$ is going to be a well-defined set. But "really" it's just some collection of ordered pairs fulfilling some conditions. I give these conditions at the end in more formal terms.

The projection $\pi_j: \prod_{i \in I} X_i \to X_j$, for any fixed $j \in I$ is defined as $\pi_j(f) = f(j)$, which indeed is in $X_j$ by definition.

If $E$ is a subset of $X_j$, then consider what it means to be some $f$ in $\pi_j^{-1}[E]$: $f \in \prod_{i \in I} X_i$ so it is a function defined on $I$ with $f(i) \in X_i$ for all $i$ and $\pi_j(f) \in E$ forces the only extra restriction that in fact $f(j) \in E$. But these are also exactly the conditions for a function on $I$ to be in $\prod_{i \in I} E_i$ where all $E_i = X_i$ except $E_j=E$. They're the same conditions on $f$ as a set of ordered pairs. Hence that we can say

$$\pi_j^{-1}[E] = \prod_{i \in I} E_i \text{ where all } E_i \text{ (except } E_j = E \text{) equal } X_i$$

So to recap a set $f$ is in $\prod_{i \in I} X_i$ iff

  • $\forall x \in f: \exists i \in I: \exists p: x= (i,p)$.
  • $\forall i \in I: \exists p: (i,p) \in f$.
  • $\forall i \in I: \forall p,p': ((i,p) \in f \land (i,p') \in f) \implies p=p'$. (the last equality is equality of sets, so really $p=p'$ should be $\forall x: x \in p \leftrightarrow x \in p'$)
  • $\forall i \in I: \exists x \in X_i: (i,x) \in f$.
  • $\begingroup$ Functions in $\displaystyle\prod_{\substack{i\in A}}X_i$ have $\displaystyle\bigcup_{\substack{i\in A}}X_i$ as codomain; functions in $\displaystyle\prod_{\substack{i\in A}}E_i$ have $\displaystyle\bigcup_{\substack{i\in A}}E_i$ as codomain. How this can be formally contained in the first one? It’s just fussiness, I got all of the definitions, the concepts behind, just that missing point $\endgroup$ – Daniele Leo Sep 19 '18 at 5:48
  • $\begingroup$ @DanieleLee A function is a set of pairs; its domain and codomain can be inferred from the set. $\endgroup$ – Henno Brandsma Sep 19 '18 at 11:41
  • $\begingroup$ I got it! Thanks Henno! :) $\endgroup$ – Daniele Leo Sep 22 '18 at 10:43
  • $\begingroup$ @DanieleLee glad to be able to help! $\endgroup$ – Henno Brandsma Sep 22 '18 at 13:02

It's ${\bf\pi_i}^{-1}(E)$ that is $\prod_jE_j$ with $E_j=X_j$ if $j\ne i$ and $E_i=E$.


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.