MathLabs
TheoremProved

Löwenheim–Skolem Theorem (downward)

Statement

Let L\mathcal{L} be a countable language and M\mathcal{M} an infinite L\mathcal{L}-structure. Then M\mathcal{M} has a countable elementary substructure N\mathcal{N}, i.e. there is N\mathcal{N} with N⪯M\mathcal{N} \preceq \mathcal{M} and NN countably infinite.

Why is it true?

This shows first-order logic cannot pin down cardinality: any theory with an infinite model (e.g. the axioms of a field, of set theory, ...) already has a countable model, no matter how "big" the original model was built to be. Combined with the upward version (any infinite model has elementary extensions of every larger cardinality), this is the source of Skolem's Paradox — a countable structure can satisfy the same first-order sentences as an uncountable one, including sentences that "assert" uncountability from the inside.

Proof sketch

We build NN as the union of a countable increasing chain of countable subsets of MM, using the Tarski–Vaught test: a subset N⊆MN \subseteq M (as a substructure) is elementary iff for every L\mathcal{L}-formula ψ(x,yˉ)\psi(x, \bar{y}) and every tuple aˉ\bar{a} from NN, if there is some b∈Mb \in M with M⊨ψ(b,aˉ)\mathcal{M} \models \psi(b, \bar{a}), then there is already such a witness b∈Nb \in N.

Since L\mathcal{L} is countable, there are only countably many formulas ψ(x,yˉ)\psi(x,\bar y). Start with any countably infinite X0⊆MX_0 \subseteq M (possible since MM is infinite). Given a countable XnX_n, for each formula ψ(x,yˉ)\psi(x,\bar y) and each tuple aˉ\bar a from XnX_n (still only countably many pairs, since XnX_n is countable and L\mathcal{L} is countable), if ∃b∈M M⊨ψ(b,aˉ)\exists b \in M\, \mathcal{M} \models \psi(b,\bar a), choose one such witness bb (using the Axiom of Choice) and add it to form Xn+1X_{n+1}; this adds only countably many new elements, so Xn+1X_{n+1} stays countable.

Let N=⋃n<ωXnN = \bigcup_{n<\omega} X_n; a countable union of countable sets, so NN is countable (and infinite, since X0⊆NX_0 \subseteq N). We check the Tarski–Vaught test for NN: given ψ(x,yˉ)\psi(x,\bar y) and aˉ\bar a from NN, since aˉ\bar a is a finite tuple it lies entirely in some single XnX_n (the chain is increasing); if a witness b∈Mb \in M exists for ψ(b,aˉ)\psi(b, \bar a), then by construction of Xn+1X_{n+1}, a witness was already chosen and placed into Xn+1⊆NX_{n+1} \subseteq N.

By the Tarski–Vaught test, NN (with the induced L\mathcal{L}-structure N\mathcal{N}) satisfies N⪯M\mathcal{N} \preceq \mathcal{M}. Elementary substructures satisfy exactly the same sentences as the ambient structure, so N\mathcal{N} is a countably infinite model witnessing the theorem.

Topics that use this theorem

Step-by-step proofs

No step-by-step proof yet for this theorem.

References

  1. Katrin Tent, Martin Ziegler (2012). A Course in Model Theory
  2. Lou van den Dries (1998). Tame Topology and O-minimal Structures
  3. Jonathan Pila, Alex J. Wilkie (2006). The rational points of a definable set