%! TEX root = LC.tex % vim: tw=50 % 27/11/2024 11AM If $f(\ol{x})$ is $\mu y . g(\ol{x}, y) = 0$ and the graph of $g$ is definable by a $\sigone$-formula, then the graph of $f$ is definable by \[ \exists u . ((u)_y = 0 \wedge \forall i < y . ((u)_i \neq 0 \wedge \ub{\forall j \le y . \exists v (v = g(\ol{x}, j) \wedge v = (u)_j)}_{(*)})) \] by using \nameref{lemma:2.2.9} to code $g(\ol{x}, 0), g(\ol{x}, 1), \ldots, g(\ol{x}, f(\ol{x}))$. Again, this is equal to a $\sigone$-formula if the graph of $g$ is given by $\exists \ol{w} \varphi(\ol{x}, y, z, \ol{w})$ with $\varphi \in \delz$, then $(*)$ is equal in $\Nbb$ to \[ \exists s . \forall j \le y . \exists v, \ol{w} \le s . (v = (u)_j \wedge \varphi(\ol{x}, j, v, \ol{w})) . \qedhere \] \end{enumerate} \end{proof} \begin{fccoro}[] \label{coro:2.2.11} % Corollary 2.2.11 \begin{iffc} \lhs A subset $A \subseteq \Nbb^k$ is \gls{renum} \rhs there is a $\sigone$-formula $\psi(x_1, \ldots, x_k)$ such that, given $\ol{x} \in \Nbb^k$, we have $\ol{x} \in A$ if and only if $\Nbb \models \psi(x)$. \end{iffc} \end{fccoro} \begin{iffproof} \rightimpl If $A$ is \gls{renum}, then there is a \gls{rec} $f$ such that $A = \dom(f)$. Given $\ol{x} \in \Nbb^k$, we thus have $x \in A$ if and only if $\Nbb \models \exists v . v = f(\ol{x})$. But $\exists v . v = f(\ol{x})$ is equal to a $\sigone$-formula by \cref{thm:2.2.10}. \leftimpl Conversely, if $A$ is defined in $\Nbb$ by a $\sigone$-formula $\psi$, define $f(\ol{x}) = 0$ if $\Nbb \models \psi(\ol{x})$, and $f(\ol{x}) \uparrow$ otherwise. The graph of $f$ is given by $y = 0 \wedge \psi(\ol{x})$, which is $\sigone$, and so $f$ is \gls{rec} by \cref{thm:2.2.10}. But $A = \dom(f)$, so $A$ is \gls{renum}. \end{iffproof} Any model of \PAm{} includes a copy of $\Nbb$ inside of it: consider the \emph{standard natural numbers} \[ \ul{n} = \ub{SSS \ldots S}_{n}0 .\] In fact, $\Nbb$ embeds in any model \PAm{} as an initial segment: essentially because \[ \PAm \syn \forall x . (x \le \ul{k} \to x = \ul{0} \wedge x = \ul{1} \wedge \cdots \wedge x = \ul{k}) .\] In Example Sheet 4, you will see that $\Nbb$ is a $\delz$-elementary substructure of any model of \PAm: every $\delz$-sentence $\varphi(\ul{n})$ true in $\Nbb$ is also true in the model. \begin{fcdefn}[Representation of a total function] \glsadjdefn{reped}{represented}{function}% \glsadjdefn{reple}{representable}{function}% \glsnoundefn{repon}{representation}{representations}% \glsnoundefn{rep}{represent}{represents}% Let $f : \Nbb^k \to \Nbb$ be total and $T$ be any $L_{\PA}$-theory extending \PAm. We say that $f$ is \emph{represented in $T$} if there is an $L_{\PAm}$ formula $\theta(x_1, \ldots, x_k, y)$ such that, for all $\ol{n} \in \Nbb^k$: \begin{cenum}[(a)] \item $T \syn \exists! y . \theta(\ol{n}, y)$ \item If $k = f(\ol{n})$, then $T \syn \theta(\ol{n}, \ul{k})$ \end{cenum} \end{fcdefn} \begin{fclemma} \label{lemma:2.2.13} % Lemma 2.2.13 Every total recursive function $f : \Nbb^k \to \Nbb$ is $\sigone$-\gls{reped} in \PAm. \end{fclemma} \begin{proof} The graph of $f$ is given by a $\sigone$-formula by \cref{thm:2.2.10}, say $\exists \ol{z} . \varphi(\ol{x}, y, \ol{z})$ where $\varphi \in \delz$. Without loss of generality, we may assume that $\ol{z}$ is a single variable (for example, rewrite $\exists z . \exists \ol{w} < z . \varphi(\ol{x}, y, \ol{w})$). Let $\psi(\ol{x}, y, z)$ be the $\delz$-formula \[ \varphi(\ol{x}, y, z) \wedge \forall u, v \le y + z . (u + v < y + z \to \neg \varphi(\ol{x}, u, v)) .\] Then the $\sigone$-formula $\theta(\ol{x}, y) \defeq \exists z . \psi(\ol{x}, y, z)$ \glspl{rep} $f$ in \PAm. We show $\PAm \syn \theta(\ol{n}, k)$ first, where $k = f(\ol{n})$. Note that $k$ is the unique element of $\Nbb$ such that $\Nbb \models \exists z . \varphi(\ol{n}, k, z)$, as $f$ is a function. Take $l$ to be the first natural number such that $\Nbb \models \varphi(\ol{n}, k, l)$. Then $\Nbb \models \psi(\ol{n}, k, l)$ too, whence $\Nbb \models \exists z . \psi(\ol{n}, k, z)$. But any $\sigone$-sentence true in $\Nbb$ is true in any model of \PAm (c.f. Example Sheet 4), so $\PAm \syn \exists z . \psi(\ol{n}, k, z)$, i.e. $\PAm \syn \theta(\ol{n}, k)$. To see that $\PAm \syn \exists ! y . \theta(\ol{n}, y)$, let $l$ be the first number such that $\Nbb \models \varphi(\ol{n}, k, l)$, where $k = f(\ol{n})$. Suppose $a, b \in \mathcal{M} \models \PAm$, with $\mathcal{M} \models \psi(\ol{n}, a, b)$. We will show that $a = k$. Completeness settles the claim. Again, $\varphi(\ol{n}, k, l)$ is a $\delz$-sentence true in $\Nbb$, thus true in $\mathcal{M}$. Using the fact that $<$ is a linear ordering in $\mathcal{M}$, we have $a, b \le k + l \in \Nbb$, so $a, b \in \Nbb$ (as $\Nbb$ is an initial segment of $\mathcal{M}$). Now $\mathcal{M} \models \psi(\ol{n}, a, b) \in \delz$, hence $\Nbb \models \psi(\ol{x}, a, b)$ and thus $\Nbb \models \exists z . \varphi(\ol{n}, a, z)$. Thus $a = k$ as needed. \end{proof} \begin{fccoro} \label{coro:2.2.14} % Corollary 2.2.14 Every \gls{rec} set $A \subseteq \Nbb^k$ is $\sigone$-\gls{reple} in \PAm. \end{fccoro} \begin{proof} The characteristic function $\chi_A$ of $A$ is \gls{tr}, so $\chi_A(\ol{x}) = y$ is \gls{reped} by some $\sigone$-formula $\theta(\ol{x}, y)$ in \PAm. But then $\theta(\ol{x}, 1)$ \glspl{rep} $A$ in \PAm. \end{proof}