27.04.2015 Views

Computability and Logic

Computability and Logic

Computability and Logic

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

14.3. OTHER PROOF PROCEDURES AND HILBERT’S THESIS 181<br />

to which may be added the step<br />

Ɣ ∪{s = t}⇒{B(s)}∪<br />

which follows by (R8a), <strong>and</strong> again we have no counterexample.<br />

14.16 Corollary. Any sequent derivable using (R0)–(R9) is in fact derivable using only<br />

(R0)–(R8).<br />

Proof: Suppose there were a counterexample, that is, a derivation using (R0)–(R9)<br />

the last step Ɣ ⇒ of which was not derivable using just (R0)–(R8). Among all such<br />

derivations, consider a derivation that is as short as possible for a counterexample.<br />

Ɣ ⇒ is not of the form {A} ⇒{A}, since any sequent of that form can be derived<br />

in one step by (R0). So in the sequent Ɣ ⇒ is inferred by from one or more<br />

premisses appearing as earlier steps. Since the derivation down to any earlier step is<br />

too short to be a counterexample, for each premiss there is a derivation of it using<br />

just (R0)–(R8). If there is only one premiss, let 0 be such a derivation of it. If there<br />

are more than one premiss, let 0 be the result of concatenating such a derivation<br />

for each premiss, writing one after the other. In either case, 0 is a derivation using<br />

only (R0)–(R8) that includes any <strong>and</strong> all premisses among its steps. Let ′ be the<br />

derivation that results on adding Ɣ ⇒ as one last step, inferred by the same rule as<br />

in . If that rule was one of (R0)–(R8), we have a derivation of Ɣ ⇒ using only<br />

(R0)–(R8). If the rule was (R9a), then is of the form {A}∪ ′ , where we have a<br />

derivation of Ɣ ∪{∼A}⇒ ′ using only (R0)–(R8). In that case, the inversion lemma<br />

tells us we have a derivation of Ɣ ⇒ , that is, of Ɣ ⇒{A}∪ ′ , using only (R0)–<br />

(R8). Likewise if the rule was (R9b). So in any case, we have a derivation of Ɣ ⇒ <br />

using only (R0)–(R8), <strong>and</strong> our original supposition that we had a counterexample has<br />

led to a contradiction, completing the proof.<br />

14.17 Corollary. The proof procedure consisting of rules (R0)–(R8) is sound <strong>and</strong><br />

complete.<br />

Proof: Soundness is immediate from the soundness theorem for (R0)–(R9), since<br />

taking away rules cannot make a sound system unsound. Completeness follows from<br />

completeness for (R0)–(R9) together with the preceding corollary.<br />

(R10)<br />

Instead of dropping (R9), one might consider adding the following.<br />

Ɣ ⇒{(A → B)}∪<br />

Ɣ ⇒{A}∪<br />

.<br />

Ɣ ⇒{B}∪<br />

14.18 Lemma (Cut elimination theorem). Using (R0)–(R9), if there are derivations of<br />

Ɣ ⇒{(A → B)}∪ <strong>and</strong> of Ɣ ⇒{A}∪, then there is a derivation of Ɣ ⇒{B}∪.<br />

14.19 Corollary. Any sequent derivable using (R0)–(R10) is in fact derivable using<br />

only (R0)–(R9).<br />

14.20 Corollary. The proof procedure consisting of rules (R0)–(R10) is sound <strong>and</strong><br />

complete.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!