5
$\begingroup$

In David Marker's Model Theory: An Introduction, Exercise 2.5.20, he gives the ultrafilter/ultraproduct based proof of the Compactness theorem. I am concerned with a small detail of the proof.

enter image description here

Is there a small error in part (a)? Technically, a set $Y$ is in $D$ if and only if $Y$ extends $X_\phi$ for some $\mathcal{L}$-sentence $\phi$. So all $Y$ have some $X_\phi$ as a subset - where $X_\phi$ is the whole set of all finite theories containing $\phi$. By my reasoning, then, every $Y \in D$ must, for some $\phi$, contain all finite theories $\Delta$ that contain $\phi$.

So, choose $A$ and $B$ in $D$; then for some $\phi_1$ and $\phi_2$, \begin{equation} X_{\phi_1} \subset A \subset I \end{equation} \begin{equation} X_{\phi_2} \subset B \subset I \end{equation} So the intersection of $A$ and $B$ must contain $X_\psi$ for some $\psi$, but why is $A \cap B$ guaranteed to contain this, and what would $\psi$ be? I was thinking of $\psi = \phi_1 \land \phi_2$ because $A \cap B$ contains every finite theory that contains both $\phi_1$ and $\phi_2$, but just because a theory contains $\phi_1$ and $\phi_2$ doesn't mean it contains $\phi_1 \land \phi_2$, because there's nothing mentioned about taking logical completion and that doesn't make sense in this context. The proof on ProofWiki is slightly different than this one so I was wondering if this was a bug.

$\endgroup$
  • 1
    $\begingroup$ I think the picture is unrelated to the question. Do you want to correct it? $\endgroup$ – An old man in the sea. Jul 30 '17 at 19:48
  • $\begingroup$ LOL thanks. That's what happens when you save all your screenshots in a "misc" folder with random alphanumeric strings as titles. $\endgroup$ – Patrick N. Jul 30 '17 at 19:50
  • 1
    $\begingroup$ The proof I know is also a bit different, in that you consider $X_{\{\phi_1,..., \phi_n\}}$ with obvious notations, so that $X_{\{\phi_1,...,\phi_n\}}\cap X_{\{\psi_1,...,\psi_n\}} = X_{\{\phi_1,...,\phi_n, \psi_1,...,\psi_n\}}$, and thus it works fine $\endgroup$ – Max Jul 30 '17 at 20:04
  • $\begingroup$ Yeah that's what the proof on ProofWiki looks like that I mention at the end. $\endgroup$ – Patrick N. Jul 30 '17 at 20:29
  • 1
    $\begingroup$ Marker's book is not known for its lack of typos. I think you're right that the proof doesn't work as stated and should be amended as suggested by @Max, or by first adding to the theory each sentence which is a consequence of a finite subset. $\endgroup$ – Alex Kruckman Jul 30 '17 at 23:10
5
$\begingroup$

You're right. Interestingly, Marker has in his the errata that in the rest of the exercise you're supposed to work with the filter generated by $D$. So, as noted by Alex Kruckman, you just need to show that $D$ has the finite intersection property, which is not very difficult: in the case of the op, $A \cap B$ is not empty, since it contains $\{\phi_1, \phi_2\}$. As noted by Alex in his comments, this clearly generalizes for every finite intersection.

$\endgroup$
  • 1
    $\begingroup$ I don't think $D$ is a filter basis. The intersection of $X_{\phi_1}$ and $X_{\phi_2}$ does not contain anything of the form $X_\phi$ since every element of it have at least two elements. $\endgroup$ – Idéophage Jul 30 '17 at 23:39
  • $\begingroup$ @Idéophage - Yes, yes, I read the op's a bit too hastily; when he said "the intersection of $A$ and $B$ must contain...", I thought that was a conclusion! I'm going to edit the answer soon. $\endgroup$ – Nagase Jul 30 '17 at 23:41
  • 1
    $\begingroup$ Marker's errata is correct: $D$ is not a filter basis, but Marker doesn't claim it is. He says "work with the filter generated by $D$", meaning that we have to close $D$ under finite intersections. Any set with the finite intersection property (any intersection of finitely many elements is nonempty) generates a proper filter in this way. And if $Y_1,\dots,Y_n\in D$, witnessed by $X_{\varphi_i}\subseteq Y_i$ for all $1\leq i\leq n$, then $Y_1\cap \dots\cap Y_n$ is nonempty, since it contains $\{\varphi_1,\dots,\varphi_n\}$. $\endgroup$ – Alex Kruckman Jul 31 '17 at 0:02
  • $\begingroup$ @AlexKruckman - So another patch would be to first throw $Y_1 \cap \dots \cap Y_n$ into $D$ to get $D'$ and then generating a filter? $\endgroup$ – Nagase Jul 31 '17 at 0:09
  • 3
    $\begingroup$ Right. To form the filter generated by a family of sets, you close under finite intersections, then close upward. A family of sets is a filter base if you can omit the first step, and just close upward. $\endgroup$ – Alex Kruckman Jul 31 '17 at 0:17

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.