Formal Proof Systems - Examples and Constructions
Concrete examples illustrate how different proof systems construct derivations and how automated reasoning systems implement these ideas.
Prove using natural deduction:
1. [p β q]ΒΉ (assumption)
2. [q β r]Β² (assumption)
3. [p]Β³ (assumption)
4. q (β-elim, 1, 3)
5. r (β-elim, 2, 4)
6. p β r (β-introΒ³, 3-5)
7. (q β r) β (p β r) (β-introΒ², 2-6)
8. (p β q) β ((q β r) β (p β r)) (β-introΒΉ, 1-7)
Each assumption is eventually discharged by an introduction rule, indicated by the superscript numbers. The proof tree structure makes the logical flow transparent.
Prove using Hilbert axioms:
- (Axiom 1)
- (Axiom 2)
- Instance of Axiom 2:
This is actually an axiom instance! Hilbert systems excel when the desired formula matches an axiom schema. For formulas that don't, proofs can be quite long and unintuitive.
Resolution is a refutation-based proof method. To prove , we show is unsatisfiable by deriving the empty clause (contradiction).
Given clauses (disjunctions of literals):
The resolvent is:
Resolution is refutation complete: if is unsatisfiable, repeated resolution derives the empty clause .
Prove :
Show is unsatisfiable:
1. p β¨ q (premise)
2. Β¬p β¨ r (premise)
3. Β¬q β¨ r (premise)
4. Β¬r (negated conclusion)
5. Β¬p (resolve 2, 4 on r)
6. Β¬q (resolve 3, 4 on r)
7. q (resolve 1, 5 on p)
8. β‘ (resolve 6, 7 on q - empty clause)
The empty clause proves unsatisfiability, hence the original consequence holds.
The tableau method (also called analytic tableaux) systematically searches for a model of a formula by building a tree of possible truth assignments. For , we test if is satisfiable:
- Decompose formulas according to their main connective
- : must have both and (single branch)
- : have or (branch splits into two)
- : equivalent to (splits)
- : instantiate with all relevant terms
- : introduce a fresh constant (Skolemization)
A branch closes if it contains both and . The tableau proves unsatisfiability if all branches close.
Modern automated theorem provers combine these methods:
- SAT solvers use sophisticated variants of resolution (DPLL, CDCL algorithms)
- SMT solvers extend SAT with decision procedures for theories (arithmetic, arrays, etc.)
- Interactive provers (Coq, Isabelle, Lean) use natural deduction or sequent calculus with tactics
The choice of proof system dramatically affects both human usability and computational efficiency.