4
$\begingroup$

It is well known that if $\phi$ is a $\Delta_{1}$-formula and $x_{1},..,x_{n}$ in $V$ and $V[G]$ is a forcing extension, then $V\models\phi(x_{1},...,x_{n})$ if and only if $V[G]\models\phi(x_{1},...,x_{n})$. In other words, we say that $\Delta_{1}$ formulas are absolute for forcing extensions. More generally, if $\phi$ is $\Sigma_{1}$, and $V\models\phi(x_{1},...,x_{n})$, then $V[G]\models\phi(x_{1},...,x_{n})$. I am wondering about what kind of absoluteness theorems hold between $V^{B}$ and $V$. Perhaps, there are absoluteness theorems where the formulas in $V^{B}$ are slightly different than the formulas in $V$. Let me be more specific about what I mean.

Let $B$ be a complete Boolean algebra and let $V^{B}$ be the corresponding $B$-valued set theoretic universe. Suppose $\dot{X}\in V^{B}$. Let $\Gamma^{*}(\dot{X})=\{\dot{x}\in V^{B}:\|\dot{x}\in\dot{X}\|=1\}$. Let $\simeq$ be the equivalence relation on $\Gamma^{*}(\dot{X})$ where $\dot{x}\simeq\dot{y}$ iff $\|\dot{x}=\dot{y}\|=1$. Let $\Gamma(\dot{X})=\Gamma^{*}(\dot{X})/\simeq$.

Suppose that $n_{1},...,n_{r}$ are natural numbers and $V^{B}\models\dot{R}_{1}\subseteq\dot{X}^{n_{1}},...,\dot{R}_{k}\subseteq\dot{X}^{n_{r}}$. Let $X=\Gamma(\dot{X})$. If $1\leq i\leq r$, then let $R_{i}$ be the $n_{i}$-ary relation on $X$ where $([x_{1}],...,[x_{r}])\in R_{i}$ if and only if $V^{B}\models (x_{1},...,x_{n_{i}})\in\dot{R}_{i}$.

Then for which n-th order formulas $\Phi$ do we have $V^{B}\models((\dot{X},\dot{R}_{1},...,\dot{R}_{r})\models\Phi(\dot{a}_{1},...,\dot{a}_{r}))$ implies $(X,R_{1},...,R_{n})\models\Phi$ whenever $V^{B}\models\dot{a}_{1},...,\dot{a}_{r}\in\dot{X}$?

If $\Phi$ is a first order formula, then the answer to this question is fairly clear:

The structure $(X,R_{1},...,R_{r})$ is a Boolean product of the structures $(\dot{X},\dot{R}_{1},...,\dot{R}_{r})/U$ where $U$ ranges over the ultrafilters on $B$. However, Horn formulas are preserved under most Boolean products [see 1] including the Boolean products that one obtains from forcing. In particular, if $\Phi$ is a Horn formula, then $V^{B}\models((\dot{X},\dot{R}_{1},...,\dot{R}_{r})\models\Phi(\dot{a}_{1},...,\dot{a}_{r}))$ implies $(X,R_{1},...,R_{n})\models\Phi([\dot{a}_{1}],...,[\dot{a}_{n}])$ whenever $V^{B}\models\dot{a}_{1},...,\dot{a}_{r}\in\dot{X}$.

Furthermore, the Feferman-Vaught theorem also holds for most Boolean products [see 1], so the truth value of any formula in the structure $(X,R_{1},...,R_{n})$ is determined by the Boolean truth value of formulas in $(\dot{X},\dot{R}_{1},...,\dot{R}_{n})$.

However, it is not clear which higher order formulas are preserved by going between $V$ and $V^{B}$. I however must mention that some higher order sentences are preserved when one goes from $V^{B}$ to $V$. For example, it is a well known fact about iterated forcing that if $V^{B}\models``\dot{C}\,\textrm{is a complete Boolean algebra}"$, then $V\models``\Gamma(\dot{C})\,\textrm{is a complete Boolean algebra}"$. Perhaps this result can be generalize to a large class of first order formulas that contains the second order formula stating that the object is a complete Boolean algebra.

I would also appreciate it if someone supplied references for such absoluteness results. I am currently just interested in such absoluteness results between $V^{B}$ and $V$ and I am not as much interested right now in the corresponding absoluteness between $V^{B}/U$ and $V$ where $U$ is an ultrafilter.

1. Sheaf constructions and their elementary properties. Stanley Burris and Heinrich Werner. Trans. Amer. Math. Soc. 248 (1979), 269-309.

$\endgroup$
4
  • 1
    $\begingroup$ I suspect you may be interested in this question. In my answer there I give some references on generic absoluteness through the projective hierarchy. Joel's gives examples of a different nature. $\endgroup$ Mar 19, 2015 at 4:15
  • $\begingroup$ I deleted this question earlier, but then I undeleted the question since now that I think of it a little more, it appears to be an interesting question, but I am in the process of editing this question to make it better. $\endgroup$ Apr 1, 2015 at 19:26
  • $\begingroup$ I have edited this question. Now it should be much more answerable and clear what I am looking for. $\endgroup$ Apr 7, 2015 at 22:32
  • $\begingroup$ Some first remark, $\check x$ is a name for an object in $V[G]$ which is the canonical name for the object $x$ from $V$, and moreover $\check x^G=x$. So you can just write $V\models\phi(x)\iff V[G]\models\phi(x)$. If you want to put emphasis on generic objects, then $\phi(\dot x^G)$ is preferable. $\endgroup$
    – Asaf Karagila
    Apr 7, 2015 at 22:32

0

Your Answer

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

Browse other questions tagged or ask your own question.