15
$\begingroup$

My question is the following:

Let $M$ be a c.t.m. of $\mathsf{ZFC}$. Are there two reals $r_0,r_1 \in \mathbb{R}$ such that $r_i$ is Cohen over $M$ for $i=0,1$ and such that $\omega_1^M$ is countable in $M[r_0,r_1]$?

Thanks

$\endgroup$

3 Answers 3

14
$\begingroup$

The answer is yes. This is a consequence of the non-amalgamation phenomenon.

Let $M$ be a countable transitive model. Fix an $M$-generic real $z$ collapsing $\omega_1^M$, and let us define two reals $c$ and $d$, so that each is an $M$-generic Cohen real, but together they code $z$.

Enumerate the dense sets $D_0$, $D_1$, and so on for Cohen forcing in $M$, forcing with finite binary sequences under end-extension. Let $c_0$ be the shortest condition in $D_0$. Let $c_1$ be all zeros up to this length, then a $1$, then the first digit of $z$, then extended so as to be in $D_0$. Now extend $c_0$ with all $0$s, up to this length, then a $1$, then the next digit of $z$, and so forth.

enter image description here

Individually, they are $M$-generic Cohen reals, but they are not mutually generic for this forcing, since if we have them together we can observe the coding blocks and get $z$. So $M[c,d]$ collapses $\omega_1^M$, as requested.

I wrote a short article on this kind of argument, which is available at:

I had first heard this kind of construction from W. Hugh Woodin, when I was a graduate student, but I'm not sure of the exact provenance. The ideas in that paper led eventually to the following paper:

  • Habič, Miha E.; Hamkins, Joel David; Klausner, Lukas Daniel; Verner, Jonathan; Williams, Kameryn J., Set-theoretic blockchains, Arch. Math. Logic 58, No. 7-8, 965-997 (2019). ZBL1468.03063.
$\endgroup$
3
  • 1
    $\begingroup$ According to Guzicki's paper below PAMS, 1990), the idea of building "incompatible" cohen reals is due to Solovay, more specifically Solovay showed that for each subset a of ω there exist Cohen reals which are different exactly in a (whence those Cohen reals cannot be in an extension of M with the same ordinals, if a codes the order-type of the M-ordinals). Here is the link for Guzicki's paper: www-ams-org.ezproxy.ub.gu.se/journals/proc/1990-110-04/… $\endgroup$
    – Ali Enayat
    Commented Nov 1, 2023 at 20:26
  • $\begingroup$ Thanks, Ali. That is great. $\endgroup$ Commented Nov 1, 2023 at 20:31
  • $\begingroup$ The same idea as in my first comment can be used to large families (2 to the power of continum) of mutually incompative expansions of a countable model of ZFC of power continuum. This result is due to Matt Kaufmann (1984), and extends a result of Mostowski. Here is the link for Kaufmann's paper: matwbn.icm.edu.pl/ksiazki/fm/fm121/fm121121.pdf $\endgroup$
    – Ali Enayat
    Commented Nov 1, 2023 at 20:33
17
$\begingroup$

Here's a slightly different example. This example has the additional benefit that the analogous argument using measure shows that two random reals can also collapse cardinals.

By Solovay's characterization we know that the set of Cohen reals over a countable transitive model is comeager, and we also know that Baire category notions are preserved by translation (understood here as pointwise addition mod $2$).

Now let $M$ be a countable transitive model and let $z$ be a real that collapses $\omega_1^M$. The sets $\{x\in 2^\omega\mid x \text{ is Cohen over } M\}$ and $\{x\in 2^\omega\mid x\oplus z \text{ is Cohen over } M\}$ are both comeager, where $\oplus$ denotes pointwise addition mod $2$. Hence there must be some real $x$ such that both $x$ and $x\oplus z$ are Cohen over $M$ (just take intersection of the two sets). But any model containing both $x$ and $x\oplus z$ will also contain $z$, so $\omega_1^M$ is countable in $M[x, x\oplus z]$.

$\endgroup$
2
  • 1
    $\begingroup$ In fact, any $M[z]$-generic Cohen real $x$ will have both $x$ and $x\oplus z$ being $M$-generic Cohen reals, since the operation is an automorphism of the forcing in $M[z]$. $\endgroup$ Commented Oct 30, 2023 at 14:38
  • 1
    $\begingroup$ Indeed. And in fact in the argument above we could have picked any $z$ to begin with. For instance a $z$ that codes the ordertype of $Ord^M$. The same argument will supply an example of two generic extensions with no common extension with the same ordinals, which the blockchain argument in Joel's paper and answer above can also demonstrated. $\endgroup$ Commented Oct 30, 2023 at 19:52
9
$\begingroup$

Indeed one can prove a more general theorem, from which one can conclude the required question.

Theorem. Suppose $R$ is a real in $V.$ Then there are two reals $a$ and $b$ such that $a$ and $b$ are Cohen generic over $V$, the models $V, V[a], V[b]$ and $V[a, b]$ have the same cardinals and $R \in L[a, b]$.

For a proof see Killing the GCH everywhere with a single real , Theoem 4.1.

To see the theorem implies your requested result, let $V=L[G]$, be the generic extension of $L$ by collapsing some cardinal $\lambda$ into $\omega$ and let $R \in L[G]$ be a real coding such a collapse. Now apply the theorem.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .