8
$\begingroup$

Suppose $\kappa = \mathfrak{c} =2^{\aleph_0}$ is a real valued measurable cardinal with a witnessing measure $m:\mathcal{P}(\kappa) \to [0, 1]$ - So $m$ is a diffused (points have zero measure) $\kappa$-additive measure on $\mathcal{P}(\kappa)$. Let $\mu = \otimes_{n < \omega} m$ be the countble product measure - So $\mu$ is defined on the sigma algebra generated by the family $\{\Pi_{n < \omega} X_n : (\forall n)(X_n \subseteq \kappa)\}$. Let $\mathcal{N}$ be the sigma ideal of $\mu$-null subsets of $\kappa^{\omega}$.

What can be say about the cardinal invariants associated with the ideal $\mathcal{N}$? For example, can either one of $add(\mathcal{N}), non(\mathcal{N}), cov(\mathcal{N})$ be $\aleph_1$?

Definitions: $non(\mathcal{N})$ is the least size of a set not in $\mathcal{N}$, $add(\mathcal{N})$ is the least size of a family $F \subseteq \mathcal{N}$ whose union is not in $\mathcal{N}$ and $cov(\mathcal{N})$ is the least size of a family of members of $\mathcal{N}$ whose union is $\kappa^{\omega}$.

$\endgroup$
2
  • 2
    $\begingroup$ For some background, what can be said about cardinal characteristics of a product of two total measures? $\endgroup$ May 2, 2017 at 20:17
  • $\begingroup$ For finite products, all invariants equal $\kappa$ (by Fubini's theorem). $\endgroup$
    – Ashutosh
    Jul 14, 2017 at 6:12

1 Answer 1

4
$\begingroup$
  • $non(\mathcal{N})=\kappa$.
    If $S \subseteq \kappa^\omega$ is of size $<\kappa$, then all projections $\pi_i(S)\subseteq \kappa$ are also of size $<\kappa$, hence have measure $0$. But then $S\subseteq \prod_{i<\omega} \pi_i(S)$ also has measure $0$.
    (It seems to me that it would be enough that one of the projections is null.)

  • $cov(\mathcal{N})$ is bounded by the covering number for the (Borel) null ideal on $2^\omega$, which I will call $\nu$. [EDITED. I do not know if $cov(\mathcal{N})$ can be less than $2^{\aleph_0}$. The "standard" model for real valued measurable is obtained by adding $\kappa$ random reals to a model where $\kappa$ is measurable, so in that model we have $\nu=2^{\aleph_0}$, in which case my claim ``$cov(\mathcal{N})\le \nu$'' is irrelevant.]

    • Let $f:\kappa \to 2$ be such that $f^{-1}(\{i\})$ has measure $\frac12$ for $i=0,1$.
    • Define $F:\kappa^\omega\to 2^\omega$ by $F(x)(n) = f(x(n))$ for all $x\in \kappa^\omega$.
    • Then $F^{-1}(U) \subseteq \kappa^\omega$ has the same measure as $U$ for all open sets $U\subseteq 2^\omega$; hence $F^{-1}(N)$ is a null set whenever $N$ is.
    • Let $(N_j:j<\nu)$ be a family of null sets covering $2^\omega$.
    • Define $M_j = F^{-1}(N_j) \subseteq \kappa^\omega$. Then $(M_j:j<\nu)$ is a cover of $\kappa^\omega$ by null sets.
$\endgroup$
4
  • $\begingroup$ I now think that $cov(\mathcal N)=\nu=2^{\aleph_0}$ always holds if $\mathfrak c$ is rvm, but I don't have a proof yet. $\endgroup$
    – Goldstern
    Jul 14, 2017 at 5:17
  • $\begingroup$ You can do this by constructing an inverse measure preserving map $f: \kappa \to 2^{\omega}$ ($\kappa$ is any atomlessly rvm). This is also how we lift the measure on $\kappa$ to a total extension of Leb. measure. $\endgroup$
    – Ashutosh
    Jul 14, 2017 at 5:45
  • $\begingroup$ If there is an isomorphism between the given measure algebra (call it $(B_m,m)$) on $\kappa$ and the Lebesgue measure on $2^\omega$ (call it $(B_\lambda,\lambda)$, then of course the measure algebra $(B_\mu,\mu)$ will also be isomorphic. But I cannot show that $B_m$ has Maharam weight $\aleph_0$. ($m$ might not be an extension of Lebesgue measure.) $\endgroup$
    – Goldstern
    Jul 14, 2017 at 13:37
  • $\begingroup$ If $\kappa$ is rvm and $m$ is a witnessing Maharam homogeneous measure on $\kappa$, then forcing with the associated measure algebra is same as adding $\theta$ random reals for some $\theta \geq \kappa^{+}$. In particular, the measure algebra of $m$ cannot be isomorphic to the standard Lebesgue measure algebra. This is a theorem of Gitik and Shelah. $\endgroup$
    – Ashutosh
    Jul 14, 2017 at 16:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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