Let $\mathcal L_Q$ be the logical system that includes first order logic together with the quantifier $Q$ which is defined as follows:

For an interpretation $\mathfrak I=(\mathfrak A, \beta)=((A,\mathfrak a), \beta)$,

$\mathfrak I\models Qx\varphi$ iff there exist uncountably many $a\in A$ so that $\mathfrak I \frac a x\models \varphi$.

My question is the following:

Let $\Phi$ be a set of formulas satisfied by a model of cardinality $\geq \aleph_1$, is $\Phi$ satisfiable by a model of cardinality $>\kappa$ for any cardinality $\kappa$?

If the answer to the above is negative, does the weaker version where $\kappa=\mathfrak c$ have a positive answer?

  • $\begingroup$ Have you tried anything so far? Some ideas to get started: 1. Does Lindström's theorem apply? 2. Can we use Henkin models or Herbrand models to prove this? $\endgroup$ – Johannes Kloos Oct 21 at 20:55
  • 1
    $\begingroup$ @JohannesKloos Neither of those ideas seems relevant. Lindstrom's theorem shows that there can't be a compact logic with the downwards Lowenheim-Skolem property, but $\mathcal{L}_Q$ lacks the compactness property and the OP is asking about the upwards Lowenheim-Skolem property. Henkin models have to do with second-order logic, I don't see how they'd be relevant to $\mathcal{L}_Q$. And Herbrand models are as small as possible - they're exactly the opposite of what the OP would be looking for. $\endgroup$ – Noah Schweber Oct 22 at 0:41

No, it doesn't.

Consider $\omega_1$, thought of as a linear order. This satisfies - in addition to the usual axioms of linear order - the axioms "The universe is uncountable" and "For each $x$, the set $\{y: y<x\}$ is countable." But every uncountable linear order with the countable predecessor property has cardinality $\aleph_1$.

What about $\aleph_2$?

Well, we can use a variant of the same trick! Consider the structure defined as follows:

  • We start with $\omega_2$ as a linear order.

  • Now we attach to each $x\in\omega_2$ a copy of $\omega_1$ as a linear order, together with a surjection from that copy to the initial segment $\{y: y<x\}$ of $\omega_2$. We now have a two-sorted structure, one of the sorts being the $\omega_2$-part and the other sort being the disjoint union of a bunch of $\omega_1$s, and some stuff connecting them appropriately.

Any $\mathcal{L}_Q$-elementarily equivalent structure must have, in place of the $\omega_2$-part, a linear order with the $\le\aleph_1$-predecessor property by the argument above for $\omega_1$. But this rules out models of cardinality $\aleph_3$ or above.

Similarly we can whip up a theory with models of size $\aleph_4$ but no models of size $\ge\aleph_5$, and so on. And in fact we don't need theories: all of this can be accomplished by individual $\mathcal{L}_Q$-sentences.

The first point where things become a bit slippery is $\aleph_\omega$; if I recall correctly we can continue well past this, but I don't immediately see the details.

| cite | improve this answer | |

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.