MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $A,B\subseteq\omega$. We write $A\subseteq^*B$ if $A\setminus B$ is finite, if additionally $B\setminus A$ is infinite then we write $A\subsetneq^*B$, otherwise we write $A=^*B$.

We say that a $\cal A\subseteq P(\omega)$ is almost disjoint if for every two distinct $A,B\in\cal A$ we have $A\cap B=^*\varnothing$. We say that $\cal A$ is maximal almost disjoint, or MAD, if there is no $\cal B$ strictly containing $\cal A$ which is almost disjoint.

At the other end of the spectrum we say that $\cal A\subseteq P(\omega)$ is a tower if $\cal A$ is well-ordered by $\subsetneq^*$.

Finally, we define $\mathcal B=\{B_\alpha\mid\alpha<\kappa\}$ to be insane if it is MAD, and there exists a tower $\mathcal A=\{A_\alpha\mid\alpha<\kappa\}$ with the following property: $$\beta<\alpha\implies B_\beta\subseteq^*A_\alpha\\ \beta\geq\alpha\implies B_\beta\cap A_\alpha=^*\varnothing.$$ In that case we say that $\cal A$ is an associated tower for $\cal B$.

Note, for example, that if $\cal B$ is insane and $\cal A$ is an associated tower then $A_{\alpha+1}\setminus A_\alpha=^*B_\alpha$.


  1. Is the existence of insane families consistent with $\sf ZFC$?
  2. If the answer is yes to the previous question, is there an insane family in $L$?
  3. If the answer is yes to the previous question, can this notion be extended to every regular cardinal $\kappa$? (replacing "finite" by ${<}\kappa$ in the definition of $\subseteq^*$ and so on.)
share|cite|improve this question
I would be glad to hear constructive remarks, in additional to the less-constructive downvotes! – Asaf Karagila Sep 9 '13 at 17:13
+1: seems like you are having too much fun with your math, how non serious is that, tsk tsk... :-) – Suvrit Sep 9 '13 at 17:32
@survit: Well, MAD families is a common term (which I can't, in good conscience, claim as my own) and insane families are just... madder than usual, because that tower thingie is not at all obviously definable from every mad family. :-) – Asaf Karagila Sep 9 '13 at 17:37
Someone downvoted this? Weird. +1 from me, anyways. – Noah Schweber Sep 9 '13 at 18:51
If only you could have found a way to replace "L" with "the membrane" ... – Yemon Choi Sep 9 '13 at 23:13
up vote 11 down vote accepted

Your requirements are inconsistent; there is no insane family.

Suppose towards contradiction that we have an insane family $\mathcal{B}=\{B_\alpha\mid\alpha\lt\kappa\}$, witnessed by tower $\langle A_\alpha\mid\alpha\lt\kappa\rangle$. For finite $k$, let $b_k$ be any element in $[(A_\omega-A_k)\cap B_k]-\bigcup_{j\lt k}B_j$. There are such elements, since $B_k$ is almost disjoint from $A_k$ and from the earlier $B_j$ for $j\lt k$, and $B_k$ is almost contained in $A_\omega$. Note that the $b_k$ are distinct, and so $B=\{b_k\mid k\lt\omega\}$ is infinite. By maximality, $B$ must have infinite intersection with some $B_\beta$. Note that $B$ has exactly one element from each $B_k$ for $k\lt\omega$. So it must be that $\beta\geq\omega$. But in this case, since $B\subset A_\omega$, we have infinitely many elements in $B_\beta\cap A_\omega$, which violates the second insanity clause.

So there is no insane family.

share|cite|improve this answer
Oh. I suspected as much. Drats. This means that I am going to have so much more work... Oh well. Thanks. I suppose this will be the same argument if we require $\leq$ and $>$ instead... – Asaf Karagila Sep 9 '13 at 20:33
I find the issue related to the connection between maximal antichains and difference antichains in a complete Boolean algebra. See and a few other places. – Joel David Hamkins Sep 9 '13 at 20:37
Hmmm, yeah. Well, I guess this means that tomorrow there's a lot of work cut out for me. Thank you very much! – Asaf Karagila Sep 9 '13 at 21:34
In the case of $P(\kappa)$, if you use an ideal $I$ for which $P(\kappa)/I$ is a complete Boolean algebra, then you can realize any maximal antichain as the difference antichain of a tower in the Boolean algebra. But you've got to work modulo $I$ instead of modulo finite. – Joel David Hamkins Sep 9 '13 at 21:53
This question reminds me of your previous question about the countable completeness of $\mathcal{P}(\omega)$/fin. If an insane family exists, then $A_{\omega}$ (for example) is a least upper bound for the countable chain $\langle A_n: n < \omega \rangle$. For if $A_{\omega}' \subsetneq_* A_{\omega}$ was a strictly smaller upper bound, the difference set $X=A_{\omega} \setminus A_{\omega}'$ would be almost disjoint from all the $B_{\beta}$, contradicting maximality. But countable chains never have least upper bounds in $\mathcal{P}(\omega)$/fin, so it must be that there is no insane family. – Garrett Ervin Sep 10 '13 at 4:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.