22.01.2014 Views

Metatheory - University of Cambridge

Metatheory - University of Cambridge

Metatheory - University of Cambridge

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.

6. Completeness 59<br />

if Γ are infinite, SimpleSearch will not terminate. Indeed, we will not even<br />

get past Step 1 <strong>of</strong> SimpleSearch, since we will never finish writing all <strong>of</strong> Γ and<br />

¬C as assumptions!<br />

A similar problem arises for the more abstract pro<strong>of</strong> <strong>of</strong> Lemma 6.2 <strong>of</strong>fered<br />

in §6.3. In the course <strong>of</strong> that pro<strong>of</strong>, I assumed that only finitely many<br />

atomic sentences occur as subsentences in the sentences among ∆, namely,<br />

A 1 , A 2 , . . . , A m , and I used this assumption to select my jointly polarising sentences,<br />

Ω m . But if ∆ are infinite, it might be that infinitely many atomic<br />

sentences occur as subsentences among ∆, and my pro<strong>of</strong> will break down.<br />

Fortunately, this last problem can be repaired. In making our repair, the<br />

crucial insight is that every TFL-pro<strong>of</strong> is only ever finitely long. This insight<br />

will allow us to prove Lemma 6.2 in full generality.<br />

Pro<strong>of</strong> <strong>of</strong> Lemma 6.2, in the fully general case. Suppose ∆ ⊬ ⊥; let A 1 , A 2 , . . .,<br />

be all the atomic sentences that occur as subsentences in any <strong>of</strong> the sentences<br />

among ∆. We define:<br />

Base clause. If ∆, A 1 ⊬ ⊥, then let Ω 1 be A 1 . Otherwise, let Ω 1 be ¬A 1<br />

Recursion clause. For each number n > 1:<br />

• If ∆, Ω n , A n+1 ⊬ ⊥, then let Ω n+1 be all <strong>of</strong> Ω n and A n+1<br />

• Otherwise, let Ω n+1 be all <strong>of</strong> Ω n and ¬A n+1<br />

Just as before, we can prove by induction that:<br />

Claim 1. ∆, Ω n ⊬ ⊥, for all n.<br />

Now let Ω be all and only the sentences that are to be found among some (any)<br />

Ω n . I claim:<br />

Claim 2. ∆, Ω ⊬ ⊥<br />

To establish this claim, suppose that ∆, Ω ⊢ ⊥. Since every TFL-pro<strong>of</strong> is<br />

only finitely long, there must be finitely many sentences, ±A i1 , ±A i2 , . . . , ±A ik ,<br />

among Ω such that:<br />

∆, ±A i1 , ±A i2 , . . . , ±A ik ⊢ ⊥<br />

But then all <strong>of</strong> these sentences, ±A i1 , ±A i2 , . . . , ±A ik are among some Ω n .<br />

(Specifically, they are in Ω n , when n is the largest number among i 1 , . . . , i k .)<br />

So ∆, Ω n ⊢ ⊥, which contradicts Claim 1.<br />

It follows from Claim 2 that Ω are jointly polarising. The remainder <strong>of</strong> the<br />

pro<strong>of</strong> now proceeds exactly as before.<br />

■<br />

In short, pursuing our more abstract pro<strong>of</strong>-strategy allows us to <strong>of</strong>fer a ‘completely<br />

general’ pro<strong>of</strong> <strong>of</strong> the Completeness Theorem. What is more, we can<br />

obtain an important result as an immediate consequence.<br />

Theorem 6.12. Compactness Theorem. Let ∆ be any sentences. Suppose<br />

that, whenever you take only finitely many sentences <strong>of</strong> ∆, those finitely many<br />

sentences are jointly consistent. Then ∆ themselves are jointly consistent.<br />

Pro<strong>of</strong>. We prove the contrapositive. Suppose that ∆ are jointly inconsistent.<br />

By the (completely general) Completeness Theorem, ∆ ⊢ ⊥. Since every TFLpro<strong>of</strong><br />

is only finitely long, there are finitely many sentences, ∆ 0 , among ∆ such<br />

that ∆ 0 ⊢ ⊥. By the Soundness Theorem, ∆ 0 are jointly inconsistent. ■

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

Saved successfully!

Ooh no, something went wrong!