content/set-theory/choice/choice.tex
1% Part: set-theory2% Chapter: choice34\documentclass[../../../include/open-logic-chapter]{subfiles}56\begin{document}78\olchapter{sth}{choice}{Choice}910\olimport{introduction}11\olimport{tarskiscott}12\olimport{hartogs}13\olimport{wellorderingproblem}14\olimport{countablechoice}15\olimport{justifications}16\olimport{banach}17\olimport{vitali}1819\OLEndChapterHook2021\end{document}
content/set-theory/choice/introduction.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{intro}67\olsection{Introduction}89In \crefrange{sth:cardinals::chap}{sth:card-arithmetic::chap}, we10developed a theory of cardinals by treating cardinals as ordinals.11That approach depends upon the Axiom of Well-Ordering. It turns out12that Well-Ordering is equivalent to another principle---the Axiom of13Choice---and there has been serious philosophical discussion of its14acceptability. Our question for this chapter are: How is the Axiom15used, and can it be justified?1617\end{document}
content/set-theory/choice/tarskiscott.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{tarskiscott}6\olsection{The Tarski--Scott Trick}78In \olref[cardinals][cardsasords]{defcardinalasordinal}, we9defined cardinals as ordinals. To do this, we assumed the Axiom of10Well-Ordering. We did this, for no other reason than that it is the11``industry standard''.1213Before we discuss any of the philosophical issues surrounding14Well-Ordering, then, it is important to be clear that we \emph{can}15depart from the industry standard, and develop a theory of cardinals16\emph{without} assuming Well-Ordering. We can still employ the17definitions of $\cardeq{A}{B}$, $\cardle{A}{B}$ and $\cardless{A}{B}$,18as they appeared in \olref[sfr][siz][]{chap}. We will just need a new19notion of \emph{cardinal}.2021A na\"ive thought would be to attempt to define $A$'s cardinality thus:22\[23 \Setabs{x}{\cardeq{A}{x}}.24\]25You might want to compare this with Frege's definition of $\# x Fx$,26sketched at the very end of \olref[cardinals][hp]{sec}. And, for27reasons we gestured at there, this definition fails. Any singleton set28is equinumerous with $\{\emptyset\}$. But new singleton sets are29formed at every successor stage of the hierarchy (just consider the30singleton of the previous stage). So $\Setabs{x}{\cardeq{A}{x}}$ does31not exist, since it cannot have a rank.3233To get around this problem, we use a trick due to Tarski and Scott:\footnote{A reminder: all formulas may have parameters (unless explicitly stated otherwise).}3435\begin{defn}[Tarski--Scott]36For any formula $\phi(x)$, let37$[ x : \phi(x)] $ be the set of all $x$, of least possible rank, such38that $\phi(x)$ (or $\emptyset$, if there are no $\phi$s).39\end{defn}4041We should check that this definition is legitimate. Working in $\ZF$,42\olref[spine][foundation]{zfentailsregularity} guarantees that43$\setrank{x}$ exists for every $x$. Now, if there are any entities44satisfying $\phi$, then we can let $\alpha$ be the least rank such45that $(\exists x\subseteq V_\alpha)\phi(x)$, i.e., $(\forall \beta46\in \alpha)(\forall x \subseteq V_\beta)\lnot \phi(x)$. We can then47define $[x : \phi(x)]$ by Separation as $\Setabs{x \in48V_{\alpha+1}}{\phi(x)}$. 4950Having justified the Tarski--Scott trick, we can now use it to define51a notion of cardinality:5253\begin{defn}54The \textsc{ts}-cardinality of $A$ is $\text{tsc}(A) = [x :55\cardeq{A}{x}]$.56\end{defn}5758The definition of a \textsc{ts}-cardinal does not use Well-Ordering.59But, even without that Axiom, we can show that60\emph{\textsc{ts}-cardinals} behave rather like \emph{cardinals} as61defined in \olref[cardinals][cardsasords]{defcardinalasordinal}.62For example, if we restate63\olref[cardinals][cardsasords]{lem:CardinalsBehaveRight} and64\olref[card-arithmetic][opps]{lem:SizePowerset2Exp} in terms of65\textsc{ts}-cardinals, the proofs go through just fine in $\ZF$,66without assuming Well-Ordering. 6768Whilst we are on the topic, it is worth noting that we can also69develop a theory of ordinals using the Tarski--Scott trick. Where70$\tuple{A, <}$ is a well-ordering, let $\text{tso}(A, <) = [\tuple{X,71R} : \ordeq{\tuple{A, <}}{\tuple{X, R}}]$. For more on this treatment72of cardinals and ordinals, see \citet[chs.~9--12]{Potter2004}.7374\end{document}
content/set-theory/choice/hartogs.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{hartogs}6\olsection{Comparability and Hartogs' Lemma}78That's the plus side. Here's the minus side. Without Choice, things9get \emph{messy}. To see why, here is a nice result due to10\cite{Hartogs1915}:1112\begin{lem}[\emph{in $\ZF$}]\ollabel{HartogsLemma}13For any set $A$, there is an ordinal $\alpha$ such that $\cardnless{\alpha}{A}$14\end{lem}1516\begin{proof}17If $B \subseteq A$ and $R \subseteq B^2$, then $\tuple{B, R} \subseteq18V_{\setrank{A}+4}$ by19\olref[ord-arithmetic][using-addition]{rankcomputation}. So, using20Separation, consider:21\[22 C = \Setabs{\tuple{B, R} \in V_{\setrank{A}+5}}{B\subseteq A 23 \text{ and $\tuple{B, R}$ is a well-ordering}}24\]25Using Replacement and26\olref[ordinals][ordtype]{thmOrdinalRepresentation}, form the set: 27\[28 \alpha = \Setabs{\ordtype{B, R}}{\tuple{B, R} \in C}.29\]30By \olref[ordinals][basic]{corordtransitiveord}, $\alpha$ is an31ordinal, since it is a transitive set of ordinals. After all, if32$\gamma \in \beta \in \alpha$, then $\beta = \ordtype{B, R}$ for some33$B \subseteq R$, whereupon $\gamma = \ordtype{B_b, R_b}$ for some $b34\in B$ by \olref[ordinals][iso]{wellordinitialsegment}, so that35$\gamma \in \alpha$. 3637For reductio, suppose there is !!a{injection} $f \colon \alpha \to A$.38Then, where:39\begin{align*}40 B &= \ran{f}\\41 R &= \Setabs{\tuple{f(\alpha), f(\beta)} \in A \times A}{\alpha \in \beta}.42\end{align*}43Clearly $\alpha = \ordtype{B, R}$ and $\tuple{B, R} \in C$. So $\alpha44\in \alpha$, which is a contradiction.45\end{proof}4647This entails a deep result:4849\begin{thm}[\emph{in $\ZF$}]50The following claims are equivalent:51\begin{enumerate}52 \item\ollabel{equivwo} The Axiom of Well-Ordering53 \item\ollabel{equivcompare} Either $\cardle{A}{B}$ or54 $\cardle{B}{A}$, for any sets $A$ and $B$55\end{enumerate}56\end{thm}5758\begin{proof}59\emph{\olref{equivwo} $\Rightarrow$ \olref{equivcompare}.} Fix $A$ and60$B$. Invoking \olref{equivwo}, there are well-orderings $\tuple{A, R}$61and $\tuple{B, S}$. Invoking62\olref[ordinals][ordtype]{thmOrdinalRepresentation}, let $f \colon63\alpha \to \tuple{A, R}$ and $g \colon \beta \to \tuple{B, S}$ be64isomorphisms. By \olref[sth][ordinals][basic]{ordinalsaresubsets}, either $\alpha \subseteq \beta$ or $\beta \subseteq \alpha$. If $\alpha \subseteq \beta$, then $\comp{f^{-1}}{g} \colon A \to B$ is !!a{injection}, and hence65$\cardle{A}{B}$; similarly, if $\beta \subseteq \alpha$ then $\cardle{B}{A}$.6667\emph{\olref{equivcompare} $\Rightarrow$ \olref{equivwo}.} Fix $A$; by68\olref{HartogsLemma} there is some ordinal $\beta$ such that69$\cardnless{\beta}{A}$. Invoking \olref{equivcompare}, we have70$\cardle{A}{\beta}$. So there is some !!{injection} $f \colon A \to71\beta$, and we can use this injection to well-order the elements of72$A$, by defining an order $\Setabs{\tuple{a, b} \in A \times A}{f(a)73\in f(b)}$.74\end{proof}75\noindent76As an immediate consequence: if Well-Ordering fails, then some sets77are \emph{literally incomparable} with regard to their size. So, if78Well-Ordering fails, then transfinite cardinal arithmetic will be79messy. For example, we will have to abandon the idea that if $A$ and80$B$ are infinite then $\cardeq{\cardeq{A \disjointsum B}{A \times81B}}{M}$, where $M$ is the larger of $A$ and $B$ (see82\olref[card-arithmetic][simp]{cardplustimesmax}). The problem is83simple: if we cannot \emph{compare} the size of $A$ and $B$, then it84is nonsensical to ask which is larger.8586\end{document}
content/set-theory/choice/wellorderingproblem.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{woproblem}6\olsection{The Well-Ordering Problem}78Evidently rather a lot hangs on whether we accept Well-Ordering. But9the discussion of this principle has tended to focus on an equivalent10principle, the Axiom of Choice. So we will now turn our attention to11that (and prove the equivalence). 1213In \citeyear{Cantor1883}, Cantor expressed his support for the Axiom14of Well-Ordering, calling it ``a law of thought which appears to me to15be fundamental, rich in its consequences, and particularly remarkable16for its general validity'' (cited in \citeauthor{Potter2004}17\citeyear[p.~243]{Potter2004}). But Cantor ultimately became convinced18that the ``Axiom'' was in need of proof. So did the mathematical19community. 2021The problem was ``solved'' by Zermelo in \citeyear{Zermelo1904}. To22explain his solution, we need some definitions. 2324\begin{defn}25A function $f$ is a \emph{choice function} iff $f(x) \in x$ for all $x \in \dom{f}$. We say that $f$ is a \emph{choice function for $A$} iff $f$ is a choice function with $\dom{f} = A \setminus \{\emptyset\}$.26\end{defn}2728Intuitively, for every (non-empty) set $x \in A$, a choice function29for $A$ \emph{chooses} a particular element, $f(x)$, from $x$. The30Axiom of Choice is then:3132\begin{axiom}[Choice]33 Every set has a choice function.34\end{axiom}3536Zermelo showed that Choice entails well-ordering, and vice versa:3738\begin{thm}[in $\ZF$]\ollabel{thmwochoice}39Well-Ordering and Choice are equivalent.40\end{thm}4142\begin{proof}43\emph{Left-to-right.} Let $A$ be a set of sets. Then $\bigcup A$44exists by the Axiom of Union, and so by Well-Ordering there is some45$<$ which well-orders $\bigcup A$. Now let $f(x) = \text{the $<$-least46member of }x$. This is a choice function for $A$.4748\emph{Right-to-left.} Fix $A$. By Choice, there is a choice function,49$f$, for $\Pow{A} \setminus \{\emptyset\}$. Using Transfinite50Recursion, define a function:51\begin{align*}52 g(0) &= f(A)\\53 g(\alpha) &= 54 \begin{cases}55 \text{stop!{}} &\text{if }A = \funimage{g}{\alpha}\\56 f(A \setminus \funimage{g}{\alpha}) & \text{otherwise}\\ 57 \end{cases}58\end{align*}59The indication to ``stop!'' is just a shorthand for what would60otherwise be a more long-winded definition. That is, when $A =61\funimage{g}{\alpha}$ for the first time, let $g(\delta) = A$ for all62$\delta \leq \alpha$. Now, in the first instance, we can only be sure that this defines a \emph{term} (see the remarks after \olref[sth][spine][recursion]{transrecursionschema}); but we will show that we indeed have a function.6364Since $f$ is a choice function, for each $\alpha$ (when defined) we have $g(\alpha) =65f(A \setminus \funimage{g}{\alpha}) \in A \setminus66\funimage{g}{\alpha}$; i.e., $g(\alpha) \notin \funimage{g}{\alpha}$.67So if $g(\alpha) = g(\beta)$ then $g(\beta) \notin68\funimage{g}{\alpha}$, i.e., $\beta \notin \alpha$, and similarly69$\alpha \notin \beta$. So $\alpha = \beta$, by Trichotomy. So $g$ is70!!{injective}.7172Next, observe that we do stop!{}, i.e.\ that there is some (least) ordinal $\alpha$ such that $A = g[\alpha]$. For suppose otherwise; then as $g$ is !!{injective} we would have $\cardless{\alpha}{\Pow{A} \setminus \{\emptyset\}}$ for73every ordinal $\alpha$, contradicting \olref[hartogs]{HartogsLemma}. Hence also $\ran{g} = A$.7475Assembling these facts, $g$ is !!a{bijection} from some ordinal to $A$. Now $g$ can be used to well-order $A$.76\end{proof}7778So Well-Ordering and Choice stand or fall together. But the question79remains: do they stand or fall?8081\end{document}
content/set-theory/choice/countablechoice.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{countablechoice}6\olsection{Countable Choice}78It is easy to prove, without any use of Choice/Well-Ordering, that:910\begin{lem}[in $\Zminus$]11Every finite set has a choice function. 12\end{lem}1314\begin{proof}15Let $a = \{b_1, \ldots, b_n\}$. Suppose for simplicity that each $b_i16\neq \emptyset$. So there are objects $c_1, \ldots, c_n$ such that17$c_1 \in b_1, \ldots, c_n \in b_n$. Now by18\olref[z][pairs]{prop:pairsconsequences}, the set $\{\langle b_1,19c_1\rangle , \ldots, \langle b_n, c_n\rangle\}$ exists; and this is a20choice function for~$a$.21\end{proof}2223But matters get murkier as soon as we consider infinite sets. For24example, consider this ``minimal'' extension to the above:2526\begin{defish}27\emph{Countable Choice.} Every \emph{countable} set has a choice function. 28\end{defish}2930This is a special case of Choice. And it transpires that this31principle was invoked fairly frequently, without an obvious awareness32of its use. Here are two nice examples.\footnote{Due to33\citet[\S9.4]{Potter2004} and Luca Incurvati.}3435\begin{ex}36Here is a natural thought: for any set $A$, either37$\cardle{\omega}{A}$, or $\cardeq{A}{n}$ for some $n \in \omega$. This38is one way to state the intuitive idea, that every set is either39finite or infinite. Cantor, and many other mathematicians, made this40claim without proving it. Cautious as we are, we proved this in41\olref[cardinals][classing]{generalinfinitycharacter}. But42in that proof we were working in $\ZFC$, since we were assuming that43any set $A$ can be well-ordered, and hence that $\card{A}$ is44guaranteed to exist. That is: we explicitly assumed Choice.4546In fact, \citet{Dedekind1888} offered his own proof of47this claim, as follows:4849\begin{thm}[in $\Zminus + \text{Countable Choice}$]50For any $A$, either $\cardle{\omega}{A}$ or $\cardeq{A}{n}$ for some51$n \in \omega$.52\end{thm}5354\begin{proof}55Suppose $\cardneq{A}{n}$ for all $n \in \omega$. Then in particular56for each $n < \omega$ there is subset $A_n \subseteq A$ with exactly57$\cardexpo{2}{n}$ elements. Using this sequence $A_0, A_1, A_2,58\ldots$, we define for each $n$:59\[60 B_n = A_n \setminus \bigcup_{i < n} A_i.61\]62Now note the following63\begin{align*}64 \card{\bigcup_{i < n}A_n} 65 &\leq \card{A_0} + \card{A_1} + \ldots + \card{A_{n-1}}\\66 &=1 + 2 + \ldots + 2^{n-1}\\67 & = 2^n - 1\\68 & < 2^n = \card{A_n}69\end{align*}70Hence each $B_n$ has at least one member, $c_n$. Moreover, the $B_n$s71are pairwise disjoint; so if $c_n = c_m$ then $n = m$. But every $c_n72\in A$. So the function $f(n) = c_n$ is an injection $\omega \to A$.73\end{proof}74\noindent 75Dedekind did not flag that he had used Countable Choice. But, did76\emph{you} spot its use? Look again. (Really: \emph{look again}.)7778The proof used Countable Choice twice. We used it once, to obtain79our sequence of sets $A_0$, $A_1$, $A_2$, \dots\@ We then used it80again to select our elements $c_n$ from each~$B_n$. Moreover, this use81of Choice is ineliminable. \citet[p.~138]{Cohen1966} proved that the82result fails if we have no version of Choice. That is: it is83consistent with $\ZF$ that there are sets which are84\emph{incomparable} with~$\omega$.85\end{ex}8687\begin{ex} 88In \citeyear{Cantor1878}, Cantor stated that a countable union of89countable sets is countable. He did not present a proof, perhaps90indicating that he took the proof to be obvious. Now, cautious as we91are, we proved a more general version of this result in92\olref[card-arithmetic][simp]{kappaunionkappasize}. But our proof93explicitly assumed Choice. And even the proof of the less general94result requires Countable Choice.9596\begin{thm}[in $\Zminus + \text{Countable Choice}$]97If $A_n$ is countable for each $n \in \omega$, then $\bigcup_{n <98\omega} A_n$ is countable.99\end{thm}100101\begin{proof}102Without loss of generality, suppose that each $A_n \neq \emptyset$. So103for each $n \in \omega$ there is !!a{surjection} $f_n \colon \omega104\to A_n$. Define $f \colon \omega \times \omega \to \bigcup_{n <105\omega} A_n$ by $f(m, n) = f_n(m)$. The result follows because $\omega106\times \omega$ is countable107(\olref[sfr][siz][zigzag]{natsquaredenumerable}) and $f$ is108!!a{surjection}.109\end{proof}110\noindent 111Did you spot the use of the Countable Choice? It is used to choose our112sequence of functions $f_0$, $f_1$, $f_2$, \dots\footnote{A similar113use of Choice occurred in114\olref[card-arithmetic][simp]{kappaunionkappasize}, when we gave the115instruction ``For each $\beta \in \cardfont{a}$, fix !!a{injection}116$f_\beta$''.} And again, the result fails in the absence of any Choice117principle. Specifically, \citet{FefermanLevy1963} proved that it is118consistent with $\ZF$ that a countable union of countable sets has119cardinality~$\beth_1$. But here is a much funnier statement of the120point, from Russell:121\begin{quote}122 This is illustrated by the millionaire who bought a pair of socks123 whenever he bought a pair of boots, and never at any other time, and124 who had such a passion for buying both that at last he had125 $\aleph_0$ pairs of boots and $\aleph_0$ pairs of socks\dots\@ Among126 boots we can distinguish right and left, and therefore we can make a127 selection of one out of each pair, namely, we can choose all the128 right boots or all the left boots; but with socks no such principle129 of selection suggests itself, and we cannot be sure, unless we130 assume the multiplicative axiom [i.e., in effect Choice], that there131 is any class consisting of one sock out of each pair.132 \citep[p.~126]{Russell1919}133\end{quote}134In short, some form of Choice is needed to prove the following: If you135have countably many pairs of socks, then you have (only) countably136many socks. And in fact, without Countable Choice (or something137equivalent), a countable union of countable sets can fail to be138countable. 139\end{ex}140141The moral is that Countable Choice was used repeatedly, without much142awareness of its users. The philosophical question is: How could we143\emph{justify} Countable Choice? 144145An attempt at an intuitive justification might invoke an appeal to a146supertask. Suppose we make the first choice in $\nicefrac{1}{2}$ a147minute, our second choice in $\nicefrac{1}{4}$ a minute, \dots, our148$n$-th choice in $\nicefrac{1}{2^n}$ a minute, \dots\@ Then within $1$~minute, we will have made an $\omega$-sequence of choices, and defined149a choice function. 150151But what, really, could such a thought-experiment tell us? For a152start, it relies upon taking this idea of ``choosing'' rather153literally. For another, it seems to bind up mathematics in154metaphysical possibility. 155156More important: it is not going to give us any justification for157Choice \emph{tout court}, rather than \emph{mere} Countable Choice.158For if we need \emph{every} set to have a choice function, then we'll159need to be able to perform a ``supertask of arbitrary ordinal160length.'' Bluntly, that idea is laughable.161162\end{document}
content/set-theory/choice/justifications.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{justifications}6\olsection{Intrinsic Considerations about Choice}78The broader question, then, is whether Well-Ordering, or Choice, or9indeed the comparability of all sets as regards their size---it10doesn't matter which---can be justified. 1112Here is an attempted \emph{intrinsic} justification. Back in13\olref[z][story]{sec}, we introduced several principles14about the hierarchy. One of these is worth restating:15\begin{enumerate}16 \item[] \stagesacc. For any stage $S$, and for any sets which were17 formed \emph{before} stage $S$: a set is formed at stage $S$ whose18 members are exactly those sets. Nothing else is formed at19 stage~$S$. 20\end{enumerate}21In fact, many authors have suggested that the Axiom of Choice can be22justified via (something like) this principle. We will briefly provide23a gloss on that approach.2425We will start with a simple little result, which offers \emph{yet26another} equivalent for Choice:2728\begin{thm}[in $\ZF$]\ollabel{choiceset}29Choice is equivalent to the following principle. If the !!{element}s30of $A$ are disjoint and non-empty, then there is some $C$ such that $C31\cap x$ is a singleton for every $x \in A$. (We call such a $C$ a32{choice set} for $A$.)33\end{thm}3435The proof of this result is straightforward, and we leave it as an36exercise for the reader. 3738\begin{prob}39Prove \olref[sth][choice][justifications]{choiceset}. If you struggle,40you can find a proof in \cite[pp.~242--3]{Potter2004}.41\end{prob}4243The essential point is that a choice set for $A$ is just the range of44a choice function for $A$. So, to justify Choice, we can simply try to45justify its equivalent formulation, in terms of the existence of46choice sets. And we will now try to do exactly that. 4748Let $A$'s !!{element}s be disjoint and non-empty. By \stageshier{}49(see \olref[z][story]{sec}), $A$ is formed at some stage~$S$. Note50that all the !!{element}s of $\bigcup A$ are available before stage51$S$. Now, by \stagesacc{}, for \emph{any} sets which were formed52before~$S$, a set is formed whose members are exactly those sets.53Otherwise put: every \emph{possible} collections of earlier-available54sets will exist at~$S$. But it is certainly \emph{possible} to select55objects which could be formed into a choice set for~$A$; that is just56some very specific subset of $\bigcup A$. So: some such choice set57exists, as required.5859Well, that's a \emph{very} quick attempt to offer a justification of60Choice on intrinsic grounds. But, to pursue this idea further, you61should read Potter's (\citeyear[\S14.8]{Potter2004}) neat development62of it.6364\end{document}
content/set-theory/choice/banach.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{banach}6\olsection{The Banach--Tarski Paradox}78We might also attempt to justify Choice, as Boolos attempted to9justify Replacement, by appealing to \emph{extrinsic} considerations10(see \olref[replacement][extrinsic]{sec}). After all, adopting Choice11has many desirable consequences: the ability to compare every12cardinal; the ability to well-order every set; the ability to treat13cardinals as a particular kind of ordinal; etc. 1415Sometimes, however, it is claimed that Choice has \emph{undesirable}16consequences. Mostly, this is due to a result by17\cite{BanachTarski1924}. 1819\begin{thm}[Banach--Tarski Paradox (in $\ZFC$)]20Any ball can be decomposed into finitely many pieces, which can be21reassembled (by rotation and transportation) to form two copies of22that ball.23\end{thm}24\noindent 25At first glance, this is a bit amazing. Clearly the two balls have26\emph{twice} the volume of the original ball. But rigid27motions---rotation and transportation---do not change volume. So it28looks as if Banach--Tarski allows us to magick new matter into29existence.3031It gets worse.\footnote{See \citet[Theorem 3.12]{Wagon2016}.} Similar32reasoning shows that a pea can be cut into finitely many pieces, which33can then be reassembled (by rotation and transportation) to form an34entity the shape and size of Big Ben.3536None of this, however, holds in $\ZF$ on its own.\footnote{Though37Banach--Tarski can be proved with principles which are strictly weaker38than Choice; see \citet[303]{Wagon2016}.} So we face a decision:39reject Choice, or learn to live with the ``paradox''. 4041We're going to suggest that we should learn to live with the42``paradox''. Indeed, we don't think it's much of a paradox at all. In43particular, we don't see why it is any more or less paradoxical than44any of the following results:\footnote{\citet[276--7]{Potter2004},45\citet[16]{Weston2003}, \citet[31, 308--9]{Wagon2016}, make46similar points, using other examples.}47\begin{enumerate}48 \item There are as many points in the interval $(0,1)$ as in $\Real$. 49 \\\emph{Proof}: consider $\tan(\pi(r-\nicefrac{1}{2})))$.50 \item There are as many points in a line as in a square.51 \\See \olref[his][set][pathology]{sec} and \olref[his][set][cantorplane]{sec}.52 \item There are space-filling curves. 53 \\See \olref[his][set][pathology]{sec} and \olref[his][set][hilbertcurve]{sec}.54\end{enumerate}55None of these three results require Choice. Indeed, we now just regard56them as surprising, lovely, bits of mathematics. Maybe we should adopt57the same attitude to the Banach--Tarski Paradox.5859To be sure, a technical observation is required here; but it only60requires keeping a level head. Rigid motions preserve volume.61Consequently, the five\footnote{We stated the Paradox in terms of62``finitely many pieces''. In fact, \citet{Robinson1947} proved that63the decomposition can be achieved with \emph{five} pieces64(but no fewer). For a proof, see \citet[pp.~66--7]{Wagon2016}.} pieces65into which the ball is decomposed cannot all be \emph{measurable}.66Roughly put, then, it makes no sense to assign a volume to these67individual pieces. You should think of these as unpicturable,68``infinite scatterings'' of points. Now, maybe it is ``weird'' to69conceive of such ``infinitely scattered'' sets. But their existence70seems to fall out from the injunction, embodied in \stagesacc{}, that71you should form \emph{all possible} collections of earlier-available72sets. 7374If none of that convinces, here is a final (extrinsic) argument in75favour of embracing the Banach--Tarski Paradox. It immediately entails76the best math joke of all time:77\begin{enumerate}78 \item[] \emph{Question}. What's an anagram of ``Banach--Tarski''? 79 \item[] \emph{Answer}. ``Banach--Tarski Banach--Tarski''.80\end{enumerate}8182\end{document}
content/set-theory/choice/vitali.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{choice}{vitali}6\olsection{Appendix: Vitali's Paradox}78To get a real sense of whether the Banach-Tarski construction is9acceptable or not, we should examine its \emph{proof}. Unfortunately,10that would require much more algebra than we can present here.11However, we can offer some quick remarks which might shed some insight12on the proof of Banach-Tarski,\footnote{For a much fuller treatment,13see \cite{Weston2003} or \cite{Wagon2016}.} by focussing on the14following result:1516\begin{thm}[Vitali's Paradox (in $\ZFC$)]\ollabel{vitaliparadox}17Any circle can be decomposed into countably many pieces, which can be18reassembled (by rotation and transportation) to form two copies of19that circle.20\end{thm}2122Vitali's Paradox is much easier to prove than the Banach--Tarski Paradox. We have23called it ``Vitali's Paradox'', since it follows from Vitali's24\citeyear{Vitali1905} construction of an unmeasurable set. But the25set-theoretic aspects of the proof of Vitali's Paradox and the26Banach-Tarski Paradox are very similar. The essential difference27between the results is just that Banach-Tarski considers a28\emph{finite} decomposition, whereas Vitali's Paradox considers a29\emph{countably infinite} decomposition. As \citet{Weston2003}30puts it, Vitali's Paradox ``is certainly not nearly as striking as the31Banach--Tarski paradox, but it does illustrate that geometric32paradoxes can happen even in `simple' situations.'' 3334Vitali's Paradox concerns a two-dimensional figure, a circle. So we35will work on the plane, $\Real^2$. Let $\rotationsgroup$ be the set of36(clockwise) rotations of points around the origin by \emph{rational}37radian values between $[0,2\pi)$. Here are some algebraic facts about38$\rotationsgroup$ (if you don't understand the statement of the39result, the proof will make its meaning clear):4041\begin{lem}\ollabel{rotationsgroupabelian}42$\rotationsgroup$ forms an abelian {group} under composition of functions.43\end{lem}4445\begin{proof}46Writing $0_{\rotationsgroup}$ for the rotation by $0$ radians, this is47an identity element for $\rotationsgroup$, since48$\comp{0_{\rotationsgroup}}{\rho} = \comp{\rho}{0_{\rotationsgroup}} =49\rho$ for any $\rho \in \rotationsgroup$.5051Every element has an inverse. Where $\rho \in \rotationsgroup$ rotates52by $r$ radians, $\rho^{-1} \in \rotationsgroup$ rotates by $2\pi - r$53radians, so that $\rho \circ \rho^{-1} = 0_\rotationsgroup$.5455Composition is associative: $\comp{\rho}{(\comp{\sigma}{\tau})} =56\comp{(\comp{\rho}{\sigma})}{\tau}$ for any $\rho, \sigma, \tau \in57\rotationsgroup$5859Composition is commutative: $\comp{\rho}{\sigma} =60\comp{\sigma}{\rho}$ for any $\rho, \sigma \in \rotationsgroup$.61\end{proof}6263In fact, we can split our group $\rotationsgroup$64in half, and then use either half to recover the whole group:6566\begin{lem}\ollabel{disjointgroup}67There is a partition of $\rotationsgroup$ into two disjoint sets,68$\rotationsgroup_{1}$ and $\rotationsgroup_{2}$, both of which are a69basis for $\rotationsgroup$. 70\end{lem}7172\begin{proof}73Let $\rotationsgroup_{1}$ consist of the rotations by rational radian74values in $[0, \pi)$; let $\rotationsgroup_{2} = \rotationsgroup75\setminus \rotationsgroup_1$. By elementary algebra,76$\Setabs{\comp{\rho}{\rho}}{\rho \in \rotationsgroup_1} =77\rotationsgroup$. A similar result can be obtained for78$\rotationsgroup_2$.79\end{proof}8081We will use this fact about groups to establish \olref{vitaliparadox}.82Let $\onesphere$ be the unit circle, i.e., the set of points83exactly~$1$ unit away from the origin of the plane, i.e.,84$\Setabs{\tuple{r,s} \in \Real^2}{\sqrt{r^2+s^2}=1}$. We will split85$\onesphere$ into parts by considering the following relation on86$\onesphere$:87\[88 r \sim s \emph{ iff }(\exists \rho \in \rotationsgroup)\rho(r) = s.89\]90That is, the points of $\onesphere$ are linked by this relation iff you can get from one to the other by a rational-valued rotation about the origin. Unsurprisingly:9192\begin{lem}93$\sim$ is an equivalence relation.94\end{lem}9596\begin{proof}97Trivial, using \olref{rotationsgroupabelian}. 98\end{proof}99100We now invoke Choice to obtain a set, $C$, containing exactly one101member from each equivalence class of $\onesphere$ under $\sim$. That102is, we consider a choice function $f$ on the set of equivalence103classes,\footnote{Since $\rotationsgroup$ is !!{enumerable},104each !!{element} of $E$ is !!{enumerable}. Since $\onesphere$ is105!!{nonenumerable}, it follows from \olref{vitalicover} and106\olref[card-arithmetic][simp]{kappaunionkappasize} that $E$ is107!!{nonenumerable}. So this is a use of \emph{un}countable Choice.}108\[109 E = \Setabs{\equivrep{r}{\sim}}{r \in \onesphere},110\]111and let $C = \ran{f}$. For each rotation $\rho \in \rotationsgroup$,112the set $\funimage{\rho}{C}$ consists of the points obtained by113applying the rotation $\rho$ to each point in $C$. These next two114results show that these sets cover the circle completely and without115overlap:116117\begin{lem}\ollabel{vitalicover}118$\onesphere = \bigcup_{\rho \in \rotationsgroup} \funimage{\rho}{C}$.119\end{lem}120121\begin{proof}122Fix $s \in \onesphere$; there is some $r \in C$ such that $r \in123\equivrep{s}{\sim}$, i.e., $r \sim s$, i.e., $\rho(r) = s$ for some124$\rho \in \rotationsgroup$. 125\end{proof}126127\begin{lem}\ollabel{vitalinooverlap}128If $\rho_1 \neq \rho_2$ then $\funimage{\rho_1}{C} \cap \funimage{\rho_2}{C} = \emptyset$. 129\end{lem}130131\begin{proof}132Suppose $s \in \funimage{\rho_{1}}{C} \cap \funimage{\rho_{2}}{C}$. So133$s = \rho_{1}(r_{1}) = \rho_{2}(r_{2})$ for some $r_{1}, r_{2} \in C$.134Hence $\rho^{-1}_2(\rho_1(r_1)) = r_2$, and135$\comp{\rho_1}{\rho^{-1}_2} \in \rotationsgroup$, so $r_{1} \sim136r_{2}$. So $r_1 = r_2$, as $C$ selects exactly one member from each137equivalence class under $\sim$. So $s = \rho_1(r_1) = \rho_2(r_1)$,138and hence $\rho_1 = \rho_2$.139\end{proof}140141We now apply our earlier algebraic facts to our circle:142143\begin{lem}\ollabel{pseudobanachtarski}144There is a partition of $\onesphere$ into two disjoint sets, $D_{1}$145and $D_{2}$, such that $D_{1}$ can be partitioned into countably many146sets which can be rotated to form a copy of $\onesphere$ (and147similarly for $D_{2}$).148\end{lem}149150\begin{proof}151Using $\rotationsgroup_{1}$ and $\rotationsgroup_{2}$ from \olref{disjointgroup}, let:152\begin{align*}153 D_{1} &= \bigcup_{\rho \in \rotationsgroup_1} \funimage{\rho}{C} & 154 D_{2} &= \bigcup_{\rho \in \rotationsgroup_2} \funimage{\rho}{C}155\end{align*}156This is a partition of $\onesphere$, by \olref{vitalicover}, and $D_1$157and $D_2$ are disjoint by \olref{vitalinooverlap}. By construction,158$D_1$ can be partitioned into countably many sets,159$\funimage{\rho}{C}$ for each $\rho \in R_1$. And these can be rotated160to form a copy of $\onesphere$, since $\onesphere = \bigcup_{\rho \in161\rotationsgroup}\funimage{\rho}{C} = \bigcup_{\rho \in162\rotationsgroup_1}\funimage{(\comp{\rho}{\rho})}{C}$ by163\olref{disjointgroup} and \olref{vitalicover}. The same reasoning164applies to $D_2$. \end{proof}\noindent This immediately entails165Vitali's Paradox. For we can generate \emph{two} copies of166$\onesphere$ from $\onesphere$, just by splitting it up into countably167many pieces (the various $\funimage{\rho}{C}$'s) and then rigidly168moving them (simply rotate each piece of $D_1$, and first transport169and then rotate each piece of $D_2$).170171Let's recap the proof-strategy. We started with some algebraic facts172about the group of rotations on the plane. We used this group to173partition $\onesphere$ into equivalence classes. We then arrived at a174``paradox'', by using Choice to select elements from each class.175176We use exactly the same strategy to prove Banach--Tarski. The main177difference is that the algebraic facts used to prove Banach--Tarski178are significantly more complicated than those used to prove179Vitali's Paradox. But those algebraic facts have nothing to do with180Choice. We will summarise them quickly. 181182To prove Banach--Tarski, we start by establishing an analogue of183\olref{disjointgroup}: any \emph{free group} can be split into four184pieces, which intuitively we can ``move around'' to recover two copies185of the whole group.\footnote{The fact that we can use \emph{four}186pieces is due to \cite{Robinson1947}. For a recent proof, see187\citet[Theorem 5.2]{Wagon2016}. We follow \citet[p.~3]{Weston2003}188in describing this as ``moving'' the pieces of the group.} We then189show that we can use two particular rotations around the origin of190$\Real^3$ to generate a free group of rotations, $F$.\footnote{See191\citet[Theorem 2.1]{Wagon2016}.} (No Choice yet.) We now regard points192on the surface of the sphere as ``similar'' iff one can be obtained193from the other by a rotation in~$F$. We then \emph{use Choice} to194select exactly one point from each equivalence class of ``similar''195points. Applying our division of $F$ to the surface of the sphere, as196in \olref{pseudobanachtarski}, we split that surface into four pieces,197which we can ``move around'' to obtain two copies of the surface of198the sphere. And this establishes \citep{Hausdorff1914}:199200\begin{thm}[Hausdorff's Paradox (in $\ZFC$)] 201The surface of any sphere can be decomposed into finitely many pieces,202which can be reassembled (by rotation and transportation) to form two203disjoint copies of that sphere.204\end{thm}205206A couple of further algebraic tricks are needed to obtain the full207Banach-Tarski Theorem (which concerns not just the sphere's surface,208but its interior too). Frankly, however, this is just icing on the209algebraic cake. Hence Weston writes:210\begin{quote} 211 [\ldots] the result on free groups is the \emph{key step} in the212 proof of the Banach-Tarski paradox. From this point of view, the213 Banach-Tarski paradox is not a statement about $\Real^3$ so much as214 it is a statement about the complexity of the group [of translations215 and rotations in $\Real^3$]. \cite[p.~16]{Weston2003}216\end{quote}217That is: whether we can offer a \emph{finite} decomposition (as in218Banach--Tarski) or a \emph{countably infinite} decomposition (as in219Vitali's Paradox) comes down to certain group-theoretic facts about220working in two-dimension or three-dimensions.221222Admittedly, this last observation slightly spoils the joke at the end223of \olref[banach]{sec}. Since it is two dimensional,224``Banach-Tarski'' must be divided into a countable \emph{infinity} of225pieces, if one wants to rearrange those pieces to form ``Banach-Tarski226Banach-Tarski''. To repair the joke, one must write in three227dimensions. We leave this as an exercise for the reader.228229One final comment. In \olref[banach]{sec}, we mentioned that the230``pieces'' of the sphere one obtains cannot be \emph{measurable}, but231must be unpicturable ``infinite scatterings''. The same is true of our232use of Choice in obtaining \olref{pseudobanachtarski}. And this is all233worth explaining.234235Again, we must sketch some background (but this is \emph{just} a236sketch; you may want to consult a textbook entry on \emph{measure}).237To define a measure for a set $X$ is to assign a value $\mu(E) \in238\Real$ for each $E$ in some ``$\sigma$-algebra'' on $X$. Details here239are not essential, except that the function $\mu$ must obey the240principle of countable additivity: the measure of a countable union of241disjoint sets is the sum of their individual measures, i.e.,242$\mu(\bigcup_{n < \omega} X_n) = \sum_{n < \omega}\mu(X_n)$ whenever243the $X_n$s are disjoint. To say that a set is ``unmeasurable'' is to244say that no measure can be suitably assigned. Now, using our245$\rotationsgroup$ from before:246247\begin{cor}[Vitali]248Let $\mu$ be a measure such that $\mu(\onesphere) = 1$, and such that249$\mu(X) = \mu(Y)$ if $X$ and $Y$ are congruent. Then250$\funimage{\rho}{C}$ is unmeasurable for all $\rho \in251\rotationsgroup$. 252\end{cor}253254\begin{proof}255For reductio, suppose otherwise. So let $\mu(\funimage{\sigma}{C}) =256r$ for some $\sigma \in \rotationsgroup$ and some $r \in \Real$. For257any $\rho \in C$, $\funimage{\rho}{C}$ and $\funimage{\sigma}{C}$ are258congruent, and hence $\mu(\funimage{\rho}{C}) = r$ for any $\rho \in259C$. By \olref{vitalicover} and \olref{vitalinooverlap}, $\onesphere =260\bigcup_{\rho \in \rotationsgroup}\funimage{\rho}{C}$ is a countable261union of pairwise disjoint sets. So countable additivity dictates that262$\mu(\onesphere) = 1$ is the sum of the measures of each263$\funimage{\rho}{C}$, i.e.,264\[265 1 = \mu(\onesphere) = \sum_{\rho \in \rotationsgroup}\mu(\funimage{\rho}{C}) = \sum_{\rho \in \rotationsgroup}r266\]267But if $r = 0$ then $\sum_{\rho \in \rotationsgroup}r = 0$, and if $r268> 0$ then $\sum_{\rho \in \rotationsgroup}r = \infty$. 269\end{proof}270271\end{document}