Skip to content
Topos Uranos

Summary

A deduction is a finite sequence and uses only a finite number of premises; the satisfaction of an infinite set of formulas, by contrast, requires a single valuation that gives the value 11 to infinitely many formulas at once, and nothing in its definition says that it suffices to examine finite parts. The compactness theorem asserts that it does suffice: a set of formulas is satisfiable if and only if each of its finite subsets is. This lesson proves it by a direct semantic argument, without going through deduction: the values of the variables v1,v2,…v_1, v_2, \ldots are fixed one by one, preserving at each step the invariant that every finite subset of the set is still satisfiable by a valuation that respects what has already been fixed; if neither of the two possible choices preserved it, the union of the two finite witnesses of that failure would contradict the previous invariant, and at the end each formula is satisfied by the coincidence lemma, because it has a finite number of variables. Since the variables are numbered, the construction does not need the axiom of choice. From compactness follow its equivalent forms (a set is unsatisfiable if and only if some finite part of it is; a formula is a consequence of a set if and only if it is a consequence of some finite part) and, with the completeness for finite sets of premises of the previous lesson, strong completeness: Γ⊨φ\Gamma \models \varphi implies Γ⊢φ\Gamma \vdash \varphi for every set Γ\Gamma, whence deduction and semantic consequence always coincide. The lesson ends with combinatorial applications: the coloring of infinite graphs from the colorings of their finite parts, the finite bounds hidden in infinite statements and the limits of what a set of formulas can express.

Learning objectives

  1. Define finitely satisfiable sets and prove the compactness theorem by fixing the values of the variables one after another, explaining the role of the invariant, of the two-witness argument and of the coincidence lemma, and why the construction does not need the axiom of choice.
  2. State, prove and apply the equivalent forms of compactness, the unsatisfiability form and the consequence form, and pass from each of them to the others.
  3. Derive strong completeness from compactness, from completeness for finite sets of premises and from monotonicity, and obtain from it the coincidence of deduction with semantic consequence and of consistency with satisfiability.
  4. Encode in sets of formulas combinatorial problems about infinite sets (colorings, orders, matchings) and solve them with compactness, distinguishing what the theorem ensures from what it does not.

What is used from the previous lessons

This lesson works almost entirely in semantics, and uses the following from it. A valuation is a function v ⁣:V→{0,1}v\colon V \to \{0, 1\} on the set VV of the variables v1,v2,v3,…v_1, v_2, v_3, \ldots, which extends in a unique way to all formulas; a valuation satisfies a set Γ\Gamma of formulas, v⊨Γv \models \Gamma, if it gives the value 11 to each of them, and Γ\Gamma is satisfiable if some valuation satisfies it and unsatisfiable otherwise. All this is in the lesson on the semantics of propositional logic, together with the central instrument of this lesson, the coincidence lemma: if two valuations agree on the variables of φ\varphi, which form a finite set V(φ)V(\varphi), they give φ\varphi the same value.

From the lesson on semantic consequence and equivalence, the following are used: the definition of Γ⊨φ\Gamma \models \varphi (every valuation that satisfies Γ\Gamma satisfies φ\varphi), its structural properties, in particular monotonicity (if Γ⊆Δ\Gamma \subseteq \Delta and Γ⊨φ\Gamma \models \varphi, then Δ⊨φ\Delta \models \varphi) and the fifth (an unsatisfiable set has every formula as a consequence), and the consequence and unsatisfiability theorem: Γ⊨φ\Gamma \models \varphi if and only if Γ∪{¬φ}\Gamma \cup \{\neg\varphi\} is unsatisfiable. From the lesson on the generalized De Morgan and distributive laws, the value lemma is used: ⋀i=1nφi\bigwedge_{i=1}^{n} \varphi_i has the value 11 if and only if all its terms have the value 11, and ⋁i=1nφi\bigvee_{i=1}^{n} \varphi_i, if and only if some term has the value 11. Finally, ⊤:=(v1→v1)\top := (v_1 \rightarrow v_1) is a tautology and ⊥:=¬⊤\bot := \neg\top, a contradiction.

On the side of deduction, the lesson on formal deductive systems proved monotonicity (Mon: if Γ⊆Δ\Gamma \subseteq \Delta and Γ⊢φ\Gamma \vdash \varphi, then Δ⊢φ\Delta \vdash \varphi) and finiteness (Fin: if Γ⊢φ\Gamma \vdash \varphi, some finite subset Γ0\Gamma_0 of Γ\Gamma is such that Γ0⊢φ\Gamma_0 \vdash \varphi), and the lesson on the four deduction techniques checked that both hold in the course system (the schemas A1, A2 and A3, modus ponens and the replacement of double negation, RDN), to which the symbol ⊢\vdash refers from now on. Finally, the lesson on soundness and completeness proved the soundness theorem (Γ⊢φ\Gamma \vdash \varphi implies Γ⊨φ\Gamma \models \varphi, for every Γ\Gamma) and completeness for finite sets of premises: if {γ1,…,γk}⊨φ\{\gamma_1, \ldots, \gamma_k\} \models \varphi, then {γ1,…,γk}⊢φ\{\gamma_1, \ldots, \gamma_k\} \vdash \varphi.