7
$\begingroup$

Let $X \neq \emptyset$ be a set. We say that ${\cal F} \subseteq {\cal P}(X)$ is a down-set if ${\cal F}$ is closed under taking subsets. Whenever $a \in X$, we let ${\cal F}_a = \{ S \in F : a \in S\}$.

We say ${\cal G} \subseteq {\cal P}(X)$ is intersecting if $S \cap T$ is non-empty whenever $S, T \in {\cal G}$.

Consider the following statement:

(C): Whenever $X$ is a non-empty set, ${\cal F} \subseteq {\cal P}(X)$ is a down-set, and ${\cal G} \subseteq {\cal F}$ is an intersecting family, then there is $a \in X$, and an injection $\iota: {\cal G} \to {\cal F}_a$.

It is not known whether (C) is true for finite non-empty sets, and this open problem is called Chvatal's conjecture. I was thinking that for $X = \omega$ maybe a family with the finite intersection property could be used to show that (C) is false for infinite ground sets $X$, but I failed to come up with an argument.

Question. What is a counterexample to (C) in the infinite case?

$\endgroup$
3
  • 6
    $\begingroup$ Ordinarily in the definition of ideal one also requires that it is closed under unions. Is that what you intend here? $\endgroup$ Commented Oct 11, 2022 at 12:18
  • 1
    $\begingroup$ Doesn't a nonprincipal ultrafilter provide a counter-example? (Or have I misinterpreted the question?) $\endgroup$ Commented Oct 11, 2022 at 12:45
  • 2
    $\begingroup$ @BillBradley If F is dual to a nonprincipal ultrafilter, then every $F_a$ is enormous, so it will have those injections. $\endgroup$ Commented Oct 11, 2022 at 12:59

2 Answers 2

3
$\begingroup$

Let $\kappa$ be a cardinal of countable cofinality that is strictly larger than the continuum. I will construct a counterexample to (C) on $X=(\kappa\times \omega)\cup \omega$.

Since $\kappa$ has countable cofinality, there exists a countable increasing sequence of cardinals $(\kappa_i)_{i\in\omega}$ with limit $\kappa$. For each ordinal $\mu<\kappa$, let $n_{\mu}$ be the least $n\in\omega$ such that $\mu\in \kappa_n$. For each $\mu<\kappa$ let $A_{\mu}\subseteq X$ be the countable set $$ A_{\mu} = (\{\mu\}\times \omega)\cup \{2^c\cdot 3^d\;|\; c\leq n_{\mu} \leq d\}. $$ Let $\mathcal G = \{A_{\mu}\;|\;\mu<\kappa\}$. Let $\mathcal F$ be the down-closure of ${\mathcal G}$ in ${\mathcal P}(X)$. Necessarily $\mathcal F$ is a down-set in ${\mathcal P}(X)$ containing $\mathcal G$. I claim that

  • $\mathcal G$ is intersecting.
  • $\mathcal G$ has cardinality $\kappa$.
  • Each ${\mathcal F}_a$ has size $<\kappa$.

The last two items show that there is no injection from $\mathcal G$ to any ${\mathcal F}_a$.

To show that $\mathcal G$ is intersecting, note that if $\mu \leq \nu$, then $2^{n_{\mu}}3^{n_{\nu}}\in A_{\mu}\cap A_{\nu}$.

To see that $|{\mathcal G}| = \kappa$, it is enough to observe that the $A_{\mu}$'s are indexed by $\kappa$ and that they are distinct from each other since $(\mu,0)\in A_{\mu}-A_{\nu}$ when $\mu\neq \nu$.

It remains to show that each ${\mathcal F}_a$ has size $<\kappa$. Here there are cases depending on whether (i) $a \in \kappa\times \omega$, (ii) $a$ has the form $2^c\cdot 3^d$ for some $c\leq d$, or (iii) $a$ does not belong to any of the sets $A_{\mu}$.

In case (i) $a=(\lambda,k)\in \kappa\times \omega$. There is exactly one set $A_{\mu}$ containing $a$, namely $A_{\lambda}$. Since ${\mathcal F}$ is the order ideal generated by $\mathcal G$, the set ${\mathcal F}_a$ is ${\mathcal P}(A_{\lambda})_a$. Since $A_{\lambda}$ is countably infinite, ${\mathcal P}(A_{\lambda})_a$ has size continuum, which is $<\kappa$.

In case (ii) we have $a=2^c\cdot 3^d$ for some $c\leq d$. If $a\in A_{\mu}$ for some $\mu$, then $c\leq n_{\mu}\leq d$. There are at most finitely many $n_{\mu}$'s that can satisfy this, let $N$ be the largest such one. The elements of ${\mathcal F}_a$ are subsets of those $A_{\mu}$'s where $c\leq n_{\mu}\leq d$. Hence ${\mathcal F}_a\subseteq \bigcup_{\mu\leq \kappa_N} {\mathcal P}(A_{\mu})_a$. Since each $A_{\mu}$ is countably infinite ${\mathcal P}(A_{\mu})_a$ has size $2^{\aleph_0}$. Therefore $|{\mathcal F}_a|\leq 2^{\aleph_0}\cdot \kappa_{N} = \max(2^{\aleph_0},\kappa_N)<\kappa$.

In case (iii) ${\mathcal F}_a$ is empty, so has size $<\kappa$. \\\

$\endgroup$
1
  • $\begingroup$ Thanks, Keith, for this long and detailled answer - I don't have the time now to fully digest it but will be able to do it tonight and get back to you. $\endgroup$ Commented Oct 12, 2022 at 9:12
8
$\begingroup$

The usual definition of ideal also includes the requirement that $F$ is closed under unions.

But in this case, I claim that Chvatal's property holds for all infinite ideals.

First, if $a\in Y\in F$, then $\{a\}\in F$. And in this case, we may associate any $Z\in F$ that omits $a$ with the set $Z\cup\{a\}$, and this makes a bijection between $F-F_a$ and $F_a$.

Thus, if $F$ is infinite, we have split it in half, and these halves must therefore have the same size as $F$ itself. So $F_a$ is bijective with $F$. And so $G$ will inject into such an $F_a$ as desired.

So perhaps you don't intend to use ideals, but merely downsets?

$\endgroup$
1
  • 1
    $\begingroup$ First, thanks Joel for your answer! That's right, I wanted to know about down-sets (I mentioned down-sets in the original post already, but my use of the word "ideal" created confusion; I will change this). $\endgroup$ Commented Oct 12, 2022 at 9:09

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.