Math 557 Nov 17
The Diagonal Lemma
In the following, we consider the diagonal function \(d\), defined by
\[ d(n)= \begin{cases} \ulcorner \forall y\, (y = \underline{n} \to \sigma(y)) \urcorner & \text{if} \; n = \ulcorner \sigma(v_0) \urcorner \; \text{for an $L$-formula} \; \sigma(v_0)\\ 0 & \text{otherwise}. \end{cases} \]
Proof. If \(d\) is represented in \(T\) by the formula \(\delta(x,y)\), then define
\[ \psi(v_0):= \forall y \,(\delta(v_0,y) \to \theta(y)), \]
and set \(n:= \ulcorner \psi(v_0) \urcorner\).
Note: \(n\) is actually a number in which the variable \(v_0\) does not occur; however, the Gödel number of \(v_0\) does enter into the calculation of \(n\).
For \(G\), we now choose the sentence
\[ G:= \forall y \, (y = \underline{n} \to \psi(y)). \]
Then \(G\) has Gödel number \(d(\ulcorner \psi(v_0) \urcorner)\). If \(\delta(x,y)\) is a \(\Sigma_1\)-formula and \(\theta(v)\) is a \(\Pi_1\)-formula, then \(\psi\) and \(G\) are (equivalent to) \(\Pi_1\)-formulas. Thus it remains to show that \(T \vdash G \leftrightarrow \theta(\underline{\ulcorner G \urcorner})\):
It is clear that
\[ \begin{aligned} &T \vdash G \leftrightarrow \psi(\underline{n}), && \text{i.e., by the definition of } \psi\\ \text{(1)} \quad &T \vdash G \leftrightarrow \forall y (\delta(\underline{n},y) \to \theta(y)). \end{aligned} \]
But the fact that \(d\) is represented in \(T\) by \(\delta(x,y)\) means
\[ d(n)= \ulcorner G \urcorner \Rightarrow T \vdash \delta(\underline{n},\underline{\ulcorner G \urcorner}) \quad \text{ and } \quad T \vdash \exists! y \; \delta(\underline{n},y). \]
Thus in particular
\[ \begin{aligned} \text{(2)} \quad &T \vdash \forall y \; ( \delta(\underline{n},y) \leftrightarrow y = \underline{\ulcorner G \urcorner}) && \text{and together with (1) it follows}\\ &T \vdash G \leftrightarrow \forall y \; ( y = \underline{\ulcorner G \urcorner} \to \theta(y)), && \text{hence}\\ &T \vdash G \leftrightarrow \theta(\underline{\ulcorner G \urcorner}). \end{aligned} \]