ProofComplete

Predicate Logic - Key Proof

Proof

Theorem (Compactness): A set Γ\Gamma of first-order sentences has a model if and only if every finite subset of Γ\Gamma has a model.

We prove compactness using Gödel's completeness theorem.

Direction (\Rightarrow): If Γ\Gamma has a model M\mathcal{M}, then M\mathcal{M} is also a model of every subset of Γ\Gamma, including all finite subsets. This direction is trivial.

Direction (\Leftarrow): Assume every finite subset of Γ\Gamma has a model. We must show Γ\Gamma has a model.

Step 1: Prove Γ\Gamma is consistent

Suppose for contradiction that Γ\Gamma is inconsistent, meaning Γ\Gamma \vdash \bot (proves a contradiction). By the definition of formal proof, any derivation of \bot uses only finitely many axioms. Therefore, there exists a finite subset Γ0Γ\Gamma_0 \subseteq \Gamma such that Γ0\Gamma_0 \vdash \bot.

By soundness of first-order logic, if Γ0\Gamma_0 \vdash \bot, then Γ0\Gamma_0 \models \bot, which means Γ0\Gamma_0 has no model. This contradicts our assumption that every finite subset of Γ\Gamma has a model.

Therefore, Γ\Gamma must be consistent.

Step 2: Apply completeness to obtain a model

By Gödel's completeness theorem, every consistent set of sentences has a model. Since we've shown Γ\Gamma is consistent, it follows that Γ\Gamma has a model.

Conclusion: We've shown that if every finite subset of Γ\Gamma has a model, then Γ\Gamma itself has a model, completing the proof.


Alternative Proof via Ultraproducts (sketch):

For readers familiar with ultraproducts, here's a model-theoretic proof that doesn't rely on completeness:

  1. Let F\mathcal{F} be the set of all finite subsets of Γ\Gamma
  2. For each ΔF\Delta \in \mathcal{F}, let MΔ\mathcal{M}_\Delta be a model of Δ\Delta (exists by assumption)
  3. Let U\mathcal{U} be an ultrafilter on F\mathcal{F} containing all sets of the form {ΔF:ϕΔ}\{\Delta \in \mathcal{F} : \phi \in \Delta\} for ϕΓ\phi \in \Gamma
  4. Form the ultraproduct M=ΔFMΔ/U\mathcal{M} = \prod_{\Delta \in \mathcal{F}} \mathcal{M}_\Delta / \mathcal{U}
  5. By Łoś's theorem, for any sentence ϕ\phi: Mϕ    {ΔF:MΔϕ}U\mathcal{M} \models \phi \iff \{\Delta \in \mathcal{F} : \mathcal{M}_\Delta \models \phi\} \in \mathcal{U}
  6. For any ϕΓ\phi \in \Gamma, the set {ΔF:ϕΔ}{ΔF:MΔϕ}\{\Delta \in \mathcal{F} : \phi \in \Delta\} \subseteq \{\Delta \in \mathcal{F} : \mathcal{M}_\Delta \models \phi\} is in U\mathcal{U}
  7. Therefore Mϕ\mathcal{M} \models \phi for all ϕΓ\phi \in \Gamma, so M\mathcal{M} is a model of Γ\Gamma

Historical Note: The compactness theorem was first proved by Gödel in 1930 as a consequence of his completeness theorem. The ultraproduct proof came later with the development of model theory in the 1950s. Compactness is now recognized as one of the most fundamental properties of first-order logic, with applications spanning all areas of mathematics.

Remark

The compactness theorem fails for many extensions of first-order logic, including second-order logic and infinitary logics like Lω1,ω\mathcal{L}_{\omega_1, \omega}. This failure is often seen as a limitation of these richer logics, but it also gives them greater expressive power—they can characterize properties that first-order logic cannot, precisely because they lack compactness.