Let $\kappa$ be an uncountable regular cardinal and let $[\kappa]^{\aleph_0}$ be the set of all countably infinite subsets of $\kappa$. A (proper) ideal $I$ on $[\kappa]^{\aleph_0}$ is then a non-empty collection of subsets of $[\kappa]^{\aleph_0}$ such that

  1. $[\kappa]^{\aleph_0}\notin I$
  2. If $X\in I$ and $Y\subseteq X$ then $Y\in I$
  3. If $X,Y\in I$ then $X\cup Y\in I$.

My question is then what normality of $I$ means, as I've stumbled across two ostensibly different definitions. Recall that a set $X\subseteq [\kappa]^{\aleph_0}$ is $I$-positive if $X\notin I$.

Definition 1. $I$ is normal if whenever $X$ is $I$-positive and $f:X\to\kappa$ is a function satisfying that $f(\sigma)\in\sigma$ then there's an $I$-positive $Y\subseteq X$ on which $f$ is constant.

Definition 2. $I$ is normal if whenever $X$ is $I$-positive and $f:X\to\kappa$ is a function satisfying that $f(\sigma)<\sup(\sigma)$ for all $\sigma\in X$, then there's an $I$-positive $Y\subseteq X$ on which $f$ is constant.

So it boils down to what a regressive function in this context is. Are these definitions equivalent? For the sake of argument we can also assume that $I$ is fine, meaning that $\{\sigma\in[\kappa]^{\aleph_0}\mid\xi\notin\sigma\}\in I$ for every $\xi<\kappa$, and countably complete, meaning that it's closed under countable unions.

  • $\begingroup$ What is $\sigma(i)$ here? The $i$th member in some fixed enumeration, or...? $\endgroup$ – Asaf Karagila Mar 12 '18 at 11:06
  • $\begingroup$ Ah right. Yeah the $i$'th member in increasing order. $\endgroup$ – Dan Saattrup Nielsen Mar 12 '18 at 11:18
  • $\begingroup$ But that doesn't make sense, as that would only allow $\omega$-type subsets, and there are far more countable sets. $\endgroup$ – Asaf Karagila Mar 12 '18 at 11:22
  • $\begingroup$ Hm, you're right. I suppose we could fix some enumeration, but in that case the definition suddenly depends upon the choice of enumeration. Instead, I'll just remove that second definition. $\endgroup$ – Dan Saattrup Nielsen Mar 12 '18 at 11:27
  • $\begingroup$ It should be enough to prove that a positive set has limit order type. From there it should be fairly simple. $\endgroup$ – Asaf Karagila Mar 12 '18 at 13:14

Here's an answer. The argument for the 'only if' direction is due to Stamatis Dimopoulos.

Claim. Let $I$ be a fine ideal on $[\kappa]^{\aleph_0}$ satisfying the first normality definition. Then $I$ satisfies the second normality definition iff it's $\kappa$-complete.

Proof. $(\Rightarrow)$ Assume that $I$ satisfies the second normality defintion. Let $\gamma<\kappa$ and $\langle X_\xi\mid\xi<\gamma\rangle\in{^\gamma}I$ be a $\gamma$-sequence of elements of $I$. Note that the second definition implies that $I$ is closed under the following ostensibly stronger notion of diagonal union:

$$ \mathscr{D}_{\xi<\kappa}Z_\xi:=\{\sigma\in[\kappa]^{\aleph_0}\mid\sigma\in\bigcup_{\xi<\sup(\sigma)}Z_\xi\}$$

Define $Z_\xi:=X_\xi$ if $\xi<\gamma$ and $Z_\xi:=\emptyset$ for $\xi\in[\gamma,\kappa)$, so that $\mathscr{D}\vec Z\in I$. If now $\sigma\in X_\xi$ for some $\xi<\gamma$ then either $\sup(\sigma)<\gamma$ or $\sigma\in\mathscr{D}\vec Z$. Since

$$Y:=\{\sigma\in[\kappa]^{\aleph_0}\mid\sup(\sigma)<\gamma\}=\bigcap_{\xi<\gamma}\{\sigma\in[\kappa]^{\aleph_0}\mid\xi\notin\gamma\}\in I$$

by fineness, this means that $\bigcup\vec X\subseteq\mathscr{D}\vec Z\cup Y\in I$ and so $\bigcup\vec X\in I$, making $I$ $\kappa$-complete.

$(\Leftarrow)$ Assume that $I$ is $\kappa$-complete, let $X$ be $I$-positive and let $f:X\to\kappa$ satisfy $f(\sigma)<\sup(\sigma)$. Define $g:X\to\kappa$ as $g(\sigma)$ being the least $\xi\in\sigma$ such that $f(\sigma)<\xi$. Then $g(\sigma)\in\sigma$ for all $\sigma\in X$, so by definition 1 we get an $I$-positive $Y\subseteq X$ with $g[Y]=\{\xi_0\}$ for some $\xi_0<\kappa$. Define now for each $\eta<\xi_0$ the set $Y_\eta:=\{\sigma\in Y\mid f(\sigma)=\eta\}$, and note that $Y=\bigcup_{\eta<\xi_0}Y_\eta$. Then $\kappa$-completeness of $I$ implies that there's some $\eta_0<\xi_0$ such that $Y_{\eta_0}$ is $I$-positive. As every $\sigma\in Y_{\eta_0}$ satisfies that $f(\sigma)=\eta_0$ we get that $I$ satisfies definition 2. QED

  • $\begingroup$ Tell Stamatis that dancing on two weddings always ends up with the dancer falling down and breaking their hip. $\endgroup$ – Asaf Karagila Mar 14 '18 at 12:14

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.