%! TEX root = LST.tex % vim: tw=50 % 22/02/2024 09AM \begin{flashcard}[upward-lowenheim-skolem-thm] \begin{corollary}[Upward L\"owenheim-Skolem Theorem] \label{upward_ls} \cloze{Let $S$ be a first-order \gls{theory}. If $S$ has an infinite \gls{model}, then $S$ has an uncountable \gls{model}.} \end{corollary} \begin{proof} \cloze{We introduce an uncountable set of new constants $\{c_i \st i \in I\}$ to the language. We let \[ S' = S \cup \{\pnot c_i = c_j \st i, j \in I, i \neq j\} .\] Let $A$ be an infinite \gls{model} of $S$. Then $A$ is a \gls{model} of any finite subset of $S'$. By \nameref{fo_compactness}, $S'$ has a \gls{model}. A \gls{model} of $S'$ is a \gls{model} $B$ of $S$ together with an injection $I \to B$. So $B$ is uncountable.} \end{proof} \end{flashcard} \begin{remark*} For any set $X$, can take $I = \hartgamma(X)$ (from \nameref{hartogs}). The proof above shows that $S$ has a \gls{model} $B$ with an injection $I \to B$. So then there will be no injection $B \to X$. \end{remark*} \begin{flashcard}[downward-lowenheim-skolem-thm] \begin{corollary}[Downward L\"owenheim-Skolem Theorem] \label{downward_ls} \cloze{Let $S$ be a \gls{cons} first-order \gls{theory} in a countable \gls{folang} ($\pOmega, \pPi$ are countable). Then if $S$ has a \gls{model}, then $S$ has a countable \gls{model}.} \end{corollary} \begin{proof} \cloze{Since $S$ is \gls{cons} (by \nameref{fo_soundness}), the proof of \cref{fo_model_existence} builds a countable \gls{model} (since the \gls{folang} is countable).} \end{proof} \end{flashcard} \subsection{Peano Arithmetic} \glssymboldefn{pas}{$s$}{$s$} \glsnoundefn{PA}{PA}{N/A} We want to axiomatise $\NN$ as a first-order \gls{theory}. \glsref[folang]{Language}: \[ \Omega = \{0, s, +, \times\}, \qquad \pPi = \emptyset \] with \glspl{arity} $0, 1, 2, 2$. $s$ means ``successor'', and the others are clear. Axioms of Peano Arithmetic (PA): \begin{align*} &(\pforall x)(\pnot \pas x = 0) \\ &(\pforall x)(\pforall y)(\pas x = \pas y \pimpl x = y) \\ &(\pforall x)(x \times 0 = 0) \\ &(\pforall x)(\pforall y)(x \times (\pas y) = (x \times y) + x) \\ &(\pforall t_1) \cdots (\pforall t_n)[(\subst p[0 / x] \pand (\pforall x)(p \pimpl \subst p[\pas x / x])) \pimpl (\pforall x) p] \end{align*} \glsnoundefn{paind}{induction}{N/A} where the last \gls{sent} is for every \gls{form} $p$ with $\FV(p) = \{x, t_1, \ldots, t_n\}$. This is the axiom-scheme for induction. \begin{remark*} Let $p$ be the \gls{form} $x + (y + z) = (x + y) + z$. Then you can prove in \gls{PA} that $(\pforall x)(\pforall y)(\pforall z) p$ by \gls{paind} on $z$ with $x, y$ parameters. You \glsref[fproof]{prove}: \[ (\pforall x)(\pforall y)(\subst p[0 / z] \pand (\pforall z)(p \pimpl \subst p[\pas z / z])) \] \end{remark*} \begin{note*} $\NN_0 = \{0\} \cup \NN$ is a \gls{model} of \gls{PA}. We can also interpret $\NN$ as a \gls{model} of \gls{PA} by taking a bijection with $\NN_0$ (but this would be rather unnatural to do). By \nameref{upward_ls}, there are uncountable \glspl{model} of \gls{PA}. Didn't we lean $\NN_0$ is uniquely determined by its properties? Yes, but \emph{true} induction says: \[ (\forall A \subset \NN_0) ((0 \in A \wedge (\forall x)(x \in A \implies \pas x \in A)) \implies A = \NN_0) \] In first-order \gls{theory}, we cannot quantify over subsets of \glspl{struc}. The axiom scheme for induction captures only countably many subsets of $\NN_0$. \end{note*} \begin{flashcard}[definable-set-defn] \begin{definition*}[Definable set] \cloze{A subset $A$ of $\NN_0$ is \emph{definable} if there's a \gls{form} $p$ in \gls{folang} of \gls{PA} with \gls{freev} \gls{var} $x$ such that $p_{\NN_0} = A$, i.e. \[ \{a \in \NN_0 \st \text{$a$ satisfies $p$}\} = A .\]} \end{definition*} \end{flashcard} \begin{example*} Set of primes: use \[ p = (\pforall y)((\pexists z)(y \cdot z = x) \pimpl (y = \ub{1}_{=\pas 0} \por y = x) \] Powers of $2$: use \[ p = (\pforall y)(((y \mid x) \pand (\text{$y$ is a prime})) \pimpl y = \ub{2}_{= \pas \pas 0}) \] \end{example*} \vspace{-1em} A consequence of G\"odel's Incompleteness Theorem: there exists a \gls{sent} $p$ such that $p$ holds in $\NN_0$, but $\text{\gls{PA}} \not \ssynentails p$. \newpage \section{Set Theory} \label{sec5} We will describe set theory as just another example of first-order \gls{theory}. We want to understand what the ``universe of sets'' looks like. \subsubsection*{Zermelo-Frankel Set Theory (ZF)} \glsnoundefn{ZF}{ZF}{N/A} \glssymboldefn{zin}{$\in$}{$\in$} \glsref[folang]{Language}: $\pOmega = \emptyset$, $\pPi = \{\in\}$, $\in$ has \gls{arity} $2$. A \gls{struc} is a set $V$ together with $[\in]_V \subset V \times V$. An element of $V$ is called a ``set''. If $a, b \in V$ and $(a, b) \in [\zin]_V$, we say ``$a$ belongs to $b$'' or ``$a$ is an element of $b$''. $V$ will be the ``universe of sets'' (when $V$ is a \gls{model} of \gls{ZF}). \begin{center} \includegraphics[width=0.6\linewidth]{images/741ae34fd8af4c70.png} \end{center} There will be $2 + 4 + 3$ axioms of \gls{ZF}. \begin{enumerate}[(1)] \item \begin{flashcard}[extensionality-ax] \glsnoundefn{Ext}{(Ext)}{N/A} \prompt{\cloze{(1) }}\textbf{\cloze{Axiom of Extensionality} (Ext):} \cloze{``If two sets have the same members, then they are equal''. \[ (\pforall x)(\pforall y)((\pforall z)(z \zin x \iff z \zin y) \pimpl x = y) \]} \end{flashcard} \item \begin{flashcard}[separation-ax] \glsnoundefn{Sep}{(Sep)}{N/A} \prompt{\cloze{(2) }}\textbf{\cloze{Axiom of Separation} (Sep):} \cloze{``We can form subsets of a set.'' \[ (\pforall t_1) \cdots (\pforall t_n) [(\pforall x)(\pexists y)(\pforall z)(z \zin y \iff (z \zin x \pand p))] ,\] where $p$ is any \gls{form} with $\FV(p) = \{z, t_1, \ldots, t_n\}$. By \gls{Ext}, the set $y$ whose existence is asserted is unique. We denote it by $\{z \zin x \st p\}$. (Formally, we introduce an $(n + 1)$-ary operation symbol to the \gls{folang}; informally, this is an abbreviation).} \end{flashcard} \begin{example*} \vspace{1em} Given $t, x$, we can form $\{z \zin x \st t \zin z\}$. \end{example*} \end{enumerate}