content/set-theory/card-arithmetic/card-arithmetic.tex
1% Part: set-theory2% Chapter: card-arithmetic34\documentclass[../../../include/open-logic-chapter]{subfiles}56\begin{document}78\olchapter{sth}{card-arithmetic}{Cardinal Arithmetic}910\olimport{opps}11\olimport{simp}12\olimport{expotough}13\olimport{ch}14\olimport{fix}1516\OLEndChapterHook1718\end{document}
content/set-theory/card-arithmetic/opps.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{card-arithmetic}{opps}6\olsection{Defining the Basic Operations}78Since we do not need to keep track of order, cardinal arithmetic is9rather easier to define than ordinal arithmetic. We will define10addition, multiplication, and exponentiation simultaneously. 1112\begin{defn}13When $\cardfont{a}$ and $\cardfont{b}$ are cardinals:14\begin{align*}15 \cardfont{a} \cardplus \cardfont{b} &\defis 16 \card{\cardfont{a} \disjointsum \cardfont{b}}\\17 \cardfont{a} \cardtimes \cardfont{b} &\defis 18 \card{\cardfont{a} \times \cardfont{b}}\\19 \cardexpo{\cardfont{a}}{\cardfont{b}} &\defis 20 \card{\funfromto{\cardfont{b}}{\cardfont{a}}}21\end{align*}22where $\funfromto{X}{Y} = \Setabs{f}{f \text{ is a function } X \to Y}$.23(It is easy to show that $\funfromto{X}{Y}$ exists for any sets $X$24and $Y$; we leave this as an exercise.) 25\end{defn}2627\begin{prob}28Prove in $\Zminus$ that $\funfromto{X}{Y}$ exists for any sets $X$29and~$Y$. Working in $\ZF$, compute $\setrank{\funfromto{X}{Y}}$ from30$\setrank{X}$ and $\setrank{Y}$, in the manner of31\olref[sth][ord-arithmetic][using-addition]{rankcomputation}. 32\end{prob}3334It might help to explain this definition. Concerning addition: this35uses the notion of disjoint sum, $\disjointsum$, as defined in36\olref[ord-arithmetic][add]{defdissum}; and it is easy37to see that this definition gives the right verdict for finite cases.38Concerning multiplication: \olref[sfr][set][pai]{cardnmprod} tells us39that if $A$ has $n$ members and $B$ has $m$ members then $A \times B$40has $n \cdot m$ members, so our definition simply generalises the idea41to transfinite multiplication. Exponentiation is similar: we are42simply generalising the thought from the finite to the transfinite.43Indeed, in certain ways, transfinite cardinal arithmetic looks much44more like ``ordinary'' arithmetic than does transfinite ordinal45arithmetic:4647\begin{prop}\ollabel{cardplustimescommute}48$\cardplus$ and $\cardtimes$ are commutative and associative. 49\end{prop}5051\begin{proof}52For commutativity, by53\olref[cardinals][cardsasords]{lem:CardinalsBehaveRight} it54suffices to observe that $\cardeq{(\cardfont{a} \disjointsum55\cardfont{b})}{(\cardfont{b} \disjointsum \cardfont{a})}$ and56$\cardeq{(\cardfont{a} \times \cardfont{b})}{(\cardfont{b} \times57\cardfont{a})}$. We leave associativity as an exercise.58\end{proof}5960\begin{prob}61Prove that $\cardplus$ and $\cardtimes$ are associative.62\end{prob}6364\begin{prop}65$A$ is infinite iff $\card{A} \cardplus 1 = 1 \cardplus \card{A} = \card{A}$.66\end{prop}6768\begin{proof}69As in70\olref[cardinals][classing]{generalinfinitycharacter}, from71\olref[ord-arithmetic][using-addition]{ordinfinitycharacter} and72\olref[cardinals][cardsasords]{lem:CardinalsBehaveRight}. 73\end{proof}7475This explains why we need to use different symbols for ordinal versus76cardinal addition/multiplication: these are genuinely \emph{different}77operations. This next pair of results shows that ordinal versus78cardinal exponentiation are also different operations. (Recall that79\olref[z][infinity-again]{defnomega} entails that $2 = \{0,801\}$):8182\begin{lem}\ollabel{lem:SizePowerset2Exp}83$\card{\Pow{A}} = \cardexpo{2}{\card{A}}$, for any $A$.84\end{lem}8586\begin{proof}87For each subset $B \subseteq A$, let $\chi_B \in \funfromto{A}{2}$ be given by:88\begin{align*}89 \chi_{B}(x) &\defis90 \begin{cases}91 1 & \text{if }x\in B\\92 0 & \text{otherwise.}93 \end{cases}94\end{align*}95Now let $f(B) = \chi_B$; this defines !!a{bijection} $f \colon \Pow{A}96\to \funfromto{A}{2}$. So $\cardeq{\Pow{A}}{\funfromto{A}{2}}$. Hence97$\cardeq{\Pow{A}}{\funfromto{\card{A}}{2}}$, so that98$\card{\Pow{A}} = \card{\funfromto{\card{A}}{2}} =992^{\card{A}}$.100\end{proof}101102This snappy proof essentially subsumes the discussion of103\olref[sfr][siz][red-alt]{sec}. There, we showed how to ``reduce'' the104uncountability of $\Pow{\omega}$ to the uncountability of the set of105infinite binary strings, $\Bin^\omega$. In effect, $\Bin^{\omega}$ is106just $\funfromto{\omega}{2}$; and the preceding proof showed that the107reasoning we went through in \olref[sfr][siz][red-alt]{sec} will go108through using any set~$A$ in place of~$\omega$. The result also yields109a quick fact about cardinal exponentiation:110111\begin{cor}\ollabel{cantorcor}112$\cardfont{a} < \cardexpo{2}{\cardfont{a}}$ for any cardinal~$\cardfont{a}$.113\end{cor}114115\begin{proof}116From Cantor's Theorem (\olref[sfr][siz][car]{thm:cantor}) and117\olref{lem:SizePowerset2Exp}.118\end{proof}119\noindent120So $\omega < \cardexpo{2}{\omega}$. But note: this is a result about121\emph{cardinal} exponentiation. It should be contrasted with122\emph{ordinal} exponentiation, since in the latter case $\omega =123\ordexpo{2}{\omega}$ (see \olref[ord-arithmetic][expo]{sec}).124125Whilst we are on the topic of cardinal exponentiation, we can also be126a bit more precise about the ``way'' in which $\Real$ is127!!{nonenumerable}.128129\begin{thm}\ollabel{continuumis2aleph0}130$\card{\Real} = \cardexpo{2}{\omega}$131\end{thm}132133\begin{proof}[Proof skeleton]134There are plenty of ways to prove this. The most straightforward is to135argue that $\cardle{\Pow{\omega}}{\Real}$ and136$\cardle{\Real}{\Pow{\omega}}$, and then use Schr\"oder-Bernstein to137infer that $\cardeq{\Real}{\Pow{\omega}}$, and138\olref[card-arithmetic][opps]{lem:SizePowerset2Exp} to infer139that $\card{\Real} = \cardexpo{2}{\omega}$. We leave it as an140(illuminating) exercise to define injections $f \colon141\Pow{\omega} \to \Real$ and $g \colon \Real \to \Pow{\omega}$.142\end{proof}143144\begin{prob}145Complete the proof of146\olref[sth][card-arithmetic][opps]{continuumis2aleph0}, by147showing that $\cardle{\Pow{\omega}}{\Real}$ and148$\cardle{\Real}{\Pow{\omega}}$.149\end{prob}150151\end{document}
content/set-theory/card-arithmetic/simp.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}4\olfileid{sth}{card-arithmetic}{simp}56\olsection{Simplifying Addition and Multiplication}78It turns out that transfinite cardinal addition and multiplication is9\emph{extremely} easy. This follows from the fact that cardinals are10(certain) ordinals, and so well-ordered, and so can be manipulated in11a certain way. Showing this, though, is \emph{not} so easy. To start,12we need a tricksy definition:1314\begin{defn}15We define a \emph{canonical ordering}, $\canonord$, on pairs of16ordinals, by stipulating that $\tuple{\alpha_1, \alpha_2} \canonord17\tuple{\beta_1, \beta_2}$ iff either:18\begin{enumerate}19 \item $\max(\alpha_1, \alpha_2) < \max(\beta_1, \beta_2)$; or20 \item $\max(\alpha_1, \alpha_2) = \max(\beta_1, \beta_2)$ and21 $\alpha_1 < \beta_1$; or22 \item $\max(\alpha_1, \alpha_2) = \max(\beta_1, \beta_2)$ and23 $\alpha_1 = \beta_1$ and $\alpha_2 < \beta_2$24\end{enumerate}25\end{defn}2627\begin{lem}28$\tuple{\alpha \times \alpha, \canonord}$ is a well-order, for any29ordinal $\alpha$.30\end{lem}3132\begin{proof}33Evidently $\canonord$ is connected on $\alpha \times \alpha$. For34suppose that neither $\tuple{\alpha_1, \alpha_2}$ nor $\tuple{\beta_1,35\beta_2}$ is $\canonord$-less than the other. Then $\max(\alpha_1,36\alpha_2) = \max(\beta_1, \beta_2)$ and $\alpha_1 = \beta_1$ and37$\alpha_2 = \beta_2$, so that $\tuple{\alpha_1, \alpha_2} =38\tuple{\beta_1, \beta_2}$.3940To show well-ordering, let $X \subseteq \alpha\times\alpha$ be41non-empty. Since $\alpha$ is an ordinal, some $\delta$ is the least42member of $\Setabs{\max(\gamma_1, \gamma_2)}{\tuple{\gamma_1,43\gamma_2} \in X}$. Now discard all pairs from44$\Setabs{\tuple{\gamma_1,\gamma_2} \in X}{\max(\gamma_1, \gamma_2) =45\delta}$ except those with least first coordinate; from among these,46the pair with least second coordinate is the $\canonord$-least element47of $X$.48\end{proof}49\noindent50Now for a teensy, simple observation:5152\begin{prop}\ollabel{simplecardproduct}53If $\cardeq{\alpha}{\beta}$, then $\cardeq{\alpha \times \alpha}{\beta54\times \beta}$. 55\end{prop}5657\begin{proof}58Just let $f \colon \alpha \to \beta$ induce $\tuple{\gamma_1,59\gamma_2} \mapsto \tuple{f(\gamma_1), f(\gamma_2)}$.60\end{proof}61\noindent62And now we will put all this to work, in proving a crucial lemma:63\begin{lem}\ollabel{alphatimesalpha}64$\cardeq{\alpha}{\alpha \times \alpha}$, for any infinite ordinal65$\alpha$66\end{lem}6768\begin{proof}69For reductio, let $\alpha$ be the least infinite ordinal for which70this is false. \olref[sfr][siz][zigzag]{natsquaredenumerable} shows71that $\cardeq{\omega}{\omega\times\omega}$, so $\omega \in \alpha$.72Moreover, $\alpha$ is a cardinal: suppose otherwise, for reductio;73then $\card{\alpha} \in \alpha$, so that74$\cardeq{\card{\alpha}}{\card{\alpha} \times \card{\alpha}}$, by75hypothesis; and $\cardeq{\card{\alpha}}{\alpha}$ by definition; so76that $\cardeq{\alpha}{\alpha\times\alpha}$ by77\olref{simplecardproduct}. 7879Now, for each $\tuple{\gamma_1, \gamma_2} \in \alpha \times \alpha$,80consider the segment:81\begin{align*}82 \text{Seg}(\gamma_1, \gamma_2) &= \Setabs{\tuple{\delta_1, \delta_2} \in \alpha \times \alpha}{\tuple{\delta_1, \delta_2} \canonord \tuple{\gamma_1, \gamma_2}}83\end{align*}84Letting $\gamma = \max(\gamma_1, \gamma_2)$, note that $\tuple{\gamma_1, \gamma_2} \canonord \tuple{\gamma+1, \gamma + 1}$. So, when $\gamma$ is infinite, observe:85\begin{align*}86 \text{Seg}(\gamma_1, \gamma_2) & 87 \precsim ((\gamma \ordplus 1)\ordtimes (\gamma \ordplus 1))\\88 &\approx (\gamma \ordtimes \gamma)89 \text{, by \olref[ord-arithmetic][using-addition]{ordinfinitycharacter} and 90 \olref{simplecardproduct}}\\91 &\approx \gamma \text{, by the induction hypothesis}\\92 & \prec \alpha\text{, since $\alpha$ is a cardinal}93\end{align*}94So $\ordtype{\alpha\times \alpha, \canonord} \leq \alpha$, and hence95$\cardle{\alpha \times \alpha}{\alpha}$. Since of course96$\cardle{\alpha}{\alpha \times \alpha}$, the result follows by97Schr\"oder-Bernstein. 98\end{proof}99100Finally, we get to our simplifying result:101102\begin{thm}\ollabel{cardplustimesmax}103If $\cardfont{a}, \cardfont{b}$ are infinite cardinals, then:104\[105 \cardfont{a}106\cardtimes \cardfont{b} = \cardfont{a} \cardplus \cardfont{b} =107\text{max}(\cardfont{a}, \cardfont{b}).108\]109\end{thm}110111\begin{proof}112Without loss of generality, suppose $\cardfont{a} = \max(\cardfont{a},113\cardfont{b})$. Then invoking \olref{alphatimesalpha},114$\cardfont{a}\cardtimes\cardfont{a} = \cardfont{a} \leq \cardfont{a}115\cardplus \cardfont{b} \leq \cardfont{a} \cardplus \cardfont{a} \leq116\cardfont{a} \cardtimes \cardfont{a}$. \end{proof}\noindent Similarly,117if $\cardfont{a}$ is infinite, an $\cardfont{a}$-sized union of118$\leq\cardfont{a}$-sized sets has size $\leq\cardfont{a}$:119120\begin{prop}\ollabel{kappaunionkappasize}121Let $\cardfont{a}$ be an infinite cardinal. For each ordinal $\beta122\in \cardfont{a}$, let $X_\beta$ be a set with $\card{X_\beta} \leq123\cardfont{a}$. Then $\card{\bigcup_{\beta \in \cardfont{a}} X_\beta}124\leq \cardfont{a}$.125\end{prop}126127\begin{proof}128For each $\beta \in \cardfont{a}$, fix !!a{injection} $f_\beta \colon129X_\beta \to \cardfont{a}$.\footnote{How are these ``fixed''? See \olref[sth][choice][countablechoice]{sec}.} Define !!a{injection} $g \colon130\bigcup_{\beta \in \cardfont{a}} X_\beta \to \cardfont{a} \times131\cardfont{a}$ by $g(v) = \tuple{\beta, f_\beta(v)}$, where $v \in132X_\beta$ and $v \notin X_\gamma$ for any $\gamma \in \beta$. Now133$\bigcup_{\beta \in \cardfont{a}} X_\beta \preceq \cardfont{a} \times134\cardfont{a} \approx \cardfont{a}$ by \olref{cardplustimesmax}.135\end{proof}136137\end{document}
content/set-theory/card-arithmetic/expotough.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}4\olfileid{sth}{card-arithmetic}{expotough}56\olsection[Some Simplifications]{Some Simplification with Cardinal Exponentiation}78Whilst defining $\canonord$ was a little involved, the upshot is a9useful result concerning cardinal addition and multiplication,10\olref[simp]{cardplustimesmax}. Transfinite exponentiation, however,11cannot be simplified so straightforwardly. To explain why, we start12with a result which extends a familiar pattern from the finitary case13(though its proof is at a high level of abstraction):1415\begin{prop}\ollabel{simplecardexpo}16$\cardexpo{\cardfont{a}}{\cardfont{b} \cardplus \cardfont{c}} =17\cardexpo{\cardfont{a}}{\cardfont{b}} \cardtimes18\cardexpo{\cardfont{a}}{\cardfont{c}}$ and19$\cardexpo{(\cardexpo{\cardfont{a}}{\cardfont{b}})}{\cardfont{c}} =20\cardexpo{\cardfont{a}}{\cardfont{b} \cardtimes \cardfont{c}}$, for21any cardinals $\cardfont{a}, \cardfont{b}, \cardfont{c}$.22\end{prop}2324\begin{proof}25For the first claim, consider a function $f \colon26(\cardfont{b}\disjointsum\cardfont{c}) \to \cardfont{a}$. Now ``split27this'', by defining $f_\cardfont{b}(\beta) = f(\beta, 0)$ for each28$\beta \in \cardfont{b}$, and $f_\cardfont{c}(\gamma) = f(\gamma, 1)$29for each $\gamma \in \cardfont{c}$. The map $f \mapsto30(f_{\cardfont{b}} \times f_\cardfont{c})$ is !!a{bijection}31$\funfromto{\cardfont{b} \disjointsum \cardfont{c}}{\cardfont{a}} \to32(\funfromto{\cardfont{b}}{\cardfont{a}} \times33\funfromto{\cardfont{c}}{\cardfont{a}})$. 3435For the second claim, consider a function $f \colon \cardfont{c} \to36(\funfromto{\cardfont{b}}{\cardfont{a}})$; so for each $\gamma \in37\cardfont{c}$ we have some function $f(\gamma) \colon \cardfont{b} \to38\cardfont{a}$. Now define $f^*(\beta, \gamma) = (f(\gamma))(\beta)$39for each $\tuple{\beta, \gamma} \in \cardfont{b} \times \cardfont{c}$.40The map $f \mapsto f^*$ is !!a{bijection}41$\funfromto{\cardfont{c}}{(\funfromto{\cardfont{b}}{\cardfont{a}})}42\to \funfromto{\cardfont{b} \cardtimes \cardfont{c}}{\cardfont{a}}$. 43\end{proof}4445Now, what we would \emph{like} is an easy way to compute46$\cardexpo{\cardfont{a}}{\cardfont{b}}$ when we are dealing with47infinite cardinals. Here is a nice step in this direction:4849\begin{prop}\ollabel{cardexpo2reduct}50If $2 \leq \cardfont{a} \leq \cardfont{b}$ and $\cardfont{b}$ is51infinite, then $\cardexpo{\cardfont{a}}{\cardfont{b}} =52\cardexpo{2}{\cardfont{b}}$53\end{prop}5455\begin{proof}56\begin{align*}57 \cardexpo{2}{\cardfont{b}} &\leq 58 \cardexpo{\cardfont{a}}{\cardfont{b}}\text{, as $2 \leq \cardfont{a}$}\\59 &\leq \cardexpo{(2^\cardfont{a})}{\cardfont{b}}60 \text{, by \olref[opps]{lem:SizePowerset2Exp}}\\61 &= \cardexpo{2}{\cardfont{a} \cardtimes \cardfont{b}}62 \text{, by \olref{simplecardexpo}} \\63 &= \cardexpo{2}{\cardfont{b}}64 \text{, by \olref[simp]{cardplustimesmax}}65\end{align*}66\end{proof}6768We should not really expect to be able to simplify this any further,69since $\cardfont{b} < \cardexpo{2}{\cardfont{b}}$ by70\olref[card-arithmetic][opps]{lem:SizePowerset2Exp}.71However, this does not tell us what to say about72$\cardexpo{\cardfont{a}}{\cardfont{b}}$ when $\cardfont{b} <73\cardfont{a}$. Of course, if $\cardfont{b}$ is \emph{finite}, we know74what to do.7576\begin{prop}77If $\cardfont{a}$ is infinite and $n \in \omega$ then78$\cardexpo{\cardfont{a}}{n} = \cardfont{a}$79\end{prop}8081\begin{proof}82$\cardexpo{\cardfont{a}}{n} = \cardfont{a} \cardtimes \cardfont{a}83\cardtimes \ldots \cardtimes \cardfont{a} = \cardfont{a}$, by \olref[simp]{cardplustimesmax}.84\end{proof}85\noindent 86Additionally, in some other cases, we can control the size of87$\cardexpo{\cardfont{a}}{\cardfont{b}}$:8889\begin{prop}90If $2 \leq \cardfont{b} < \cardfont{a} \leq91\cardexpo{2}{\cardfont{b}}$ and $\cardfont{b}$ is infinite, then92$\cardexpo{\cardfont{a}}{\cardfont{b}} = \cardexpo{2}{\cardfont{b}}$93\end{prop}9495\begin{proof}96$\cardexpo{2}{\cardfont{b}}\leq \cardexpo{\cardfont{a}}{\cardfont{b}}97\leq \cardexpo{(\cardexpo{2}{\cardfont{b}})}{\cardfont{b}} =98\cardexpo{2}{\cardfont{b}\cardtimes\cardfont{b}} =99\cardexpo{2}{\cardfont{b}}$, reasoning as in \olref{cardexpo2reduct}.100\end{proof}101\noindent 102But, beyond this point, things become rather more subtle.103104\end{document}
content/set-theory/card-arithmetic/ch.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{card-arithmetic}{ch}67\olsection{The Continuum Hypothesis}89The previous result hints (correctly) that cardinal exponentiation10would be quite \emph{easy}, if infinite cardinals are guaranteed to11``play straightforwardly'' with powers of $2$, i.e., (by12\olref[opps]{lem:SizePowerset2Exp}) with taking powersets. But we13cannot assume that infinite cardinals \emph{do} play straightforwardly 14powersets. 1516To start unpacking this, we introduce some nice notation.1718\begin{defn}19Where $\cardsucc{\cardfont{a}}$ is the least cardinal strictly greater20than $\cardfont{a}$, we define two infinite sequences:21\begin{align*}22 \aleph_{0} &\defis \omega & 23 \beth_{0} &\defis \omega\\24 \aleph_{\alpha \ordplus 1} &\defis \cardsucc{(\aleph_{\alpha})} &25 \beth_{\alpha+1} &\defis \cardexpo{2}{\beth_{\alpha}}\\26 \aleph_{\alpha} &\defis \bigcup_{\beta< \alpha} \aleph_{\beta} &27 \beth_{\alpha} &\defis \bigcup_{\beta < \alpha}\beth_{\beta} & \text{when $\alpha$ is a limit ordinal}.28 \end{align*}29\end{defn}3031The definition of $\cardsucc{\cardfont{a}}$ is in order, since32\olref[cardinals][classing]{lem:NoLargestCardinal} tells us that, for each33cardinal $\cardfont{a}$, there is some cardinal greater than34$\cardfont{a}$, and Transfinite Induction guarantees that there is a35\emph{least} cardinal greater than $\cardfont{a}$. The rest of the36definition of $\cardfont{a}$ is provided by transfinite recursion. 3738Cantor introduced this ``$\aleph$'' notation; this is \emph{aleph},39the first letter in the Hebrew alphabet and the first letter in the40Hebrew word for ``infinite''. Peirce introduced the ``$\beth$''41notation; this is \emph{beth}, which is the second letter in the42Hebrew alphabet.\footnote{Peirce used this notation in a letter to43Cantor of December 1900. Unfortunately, Peirce also gave a bad44argument there that $\beth_\alpha$ does not exist for $\alpha \geq45\omega$.} Now, these notations provide us with infinite cardinals.4647\begin{prop}48$\aleph_\alpha$ and $\beth_\alpha$ are cardinals, for every49ordinal $\alpha$. 50\end{prop}5152\begin{proof}53Both results hold by a simple transfinite induction. $\aleph_0 =54\beth_0 = \omega$ is a cardinal by55\olref[cardinals][classing]{omegaisacardinal}. Assuming $\aleph_\alpha$ and56$\beth_\alpha$ are both cardinals, $\aleph_{\alpha+1}$ and57$\beth_{\alpha+1}$ are explicitly defined as cardinals. And the union58of a set of cardinals is a cardinal, by59\olref[cardinals][classing]{unioncardinalscardinal}.60\end{proof}61\noindent62Moreover, every infinite cardinal is an $\aleph$:6364\begin{prop}65If $\cardfont{a}$ is an infinite cardinal, then $\cardfont{a} =66\aleph_\gamma$ for some unique $\gamma$.67\end{prop}6869\begin{proof}70By transfinite induction on cardinals. For induction, suppose that if71$\cardfont{b} < \cardfont{a}$ then $\cardfont{b} =72\aleph_{\gamma_\cardfont{b}}$. If $\cardfont{a} =73\cardsucc{\cardfont{b}}$ for some $\cardfont{b}$, then $\cardfont{a} =74\cardsucc{(\aleph_{\gamma_\cardfont{b}})}=75\aleph_{\gamma_\cardfont{b}+1}$. If $\cardfont{a}$ is not the76successor of any cardinal, then since cardinals are ordinals77$\cardfont{a} = \bigcup_{\cardfont{b} < \cardfont{a}} \cardfont{b} =78\bigcup_{\cardfont{b} < \cardfont{a}}{\aleph_{\gamma_\cardfont{b}}}$,79so $\cardfont{a} = \aleph_\gamma$ where $\gamma =80\bigcup_{\cardfont{b} < \cardfont{a}}\gamma_\cardfont{b}$. 81\end{proof}8283Since every infinite cardinal is an $\aleph$, this prompts us to ask:84is every infinite cardinal a~$\beth$? Certainly if that \emph{were}85the case, then the infinite cardinals would ``play straightforwardly''86with the operation of taking powersets. Indeed, we would have the87following:8889\begin{defish}90\emph{Generalized Continuum Hypothesis} (GCH). $\aleph_\alpha = \beth_\alpha$, for all $\alpha$. 91\end{defish}9293Moreover, if GCH held, then we could make some considerable94simplifications with cardinal exponentiation. In particular, we could95show that when $\cardfont{b} < \cardfont{a}$, the value of96$\cardexpo{\cardfont{a}}{\cardfont{b}}$ is trapped by97$\cardfont{a}\leq \cardexpo{\cardfont{a}}{\cardfont{b}} \leq98\cardsucc{\cardfont{a}}$. We could then go on to give precise99conditions which determine which of the two possibilities obtains100(i.e., whether $\cardfont{a} = \cardexpo{\cardfont{a}}{\cardfont{b}}$101or $\cardexpo{\cardfont{a}}{\cardfont{b}} =102\cardsucc{\cardfont{a}}$).\footnote{The condition is dictated by103\emph{cofinality}.}104105But GCH is a \emph{hypothesis}, not a \emph{theorem}. In fact,106\citet{Godel1938} proved that if $\ZFC$ is consistent, then so is107$\ZFC + \text{GCH}$. But it later turned out that we can equally add108$\lnot$GCH to $\ZFC$. Indeed, consider the simplest non-trivial109\emph{instance} of GCH, namely: 110111\begin{defish}112\emph{Continuum Hypothesis} (CH). $\aleph_1 = \beth_1$. 113\end{defish}114115\citet{Cohen1963} proved that if $\ZFC$ is consistent then so is $\ZFC116+ \lnot\text{CH}$. So the Continuum Hypothesis is independent from $\ZFC$.117118The Continuum Hypothesis is so-called, since ``the continuum'' is119another name for the real line, $\Real$.120\olref[opps]{continuumis2aleph0} tells us that $\card{\Real} =121\beth_1$. So the Continuum Hypothesis states that there is no cardinal122between the cardinality of the natural numbers, $\aleph_0 = \beth_0$,123and the cardinality of the continuum, $\beth_1$.124125Given the \emph{independence} of (G)CH from $\ZFC$, what should say126about their \emph{truth}? Well, there is \emph{much} to say. Indeed,127and much fertile recent work in set theory has been directed at128investigating these issues. But two very quick points are certainly worth129emphasising. 130131First: it does not \emph{immediately} follow from these formal132independence results that either GCH or CH is \emph{indeterminate} in133truth value. After all, maybe we just need to add more axioms, which134strike us as natural, and which will settle the question one way or135another. G\"odel himself suggested that this was the right response. 136137Second: the independence of CH from $\ZFC$ is certainly138\emph{striking}, but it is certainly not \emph{incredible} (in the139literal sense). The point is simply that, for all $\ZFC$ tells us,140moving from cardinals to their successors may involve a less blunt141tool than simply taking powersets.142 %The operation of taking powersets moves us from one stage of the143 %hierarchy to its successor stage (from $V_{\alpha}$ to144 %$V_{\alpha+1}$). 145146With those two observations made, if you want to know more, you will147now have to turn to the various philosophers and mathematicians with148horses in the race.\footnote{Though you might want to start by reading \citet[\S15.6]{Potter2004}.}149150\end{document}
content/set-theory/card-arithmetic/fix.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{card-arithmetic}{fix}6\olsection{$\aleph$-Fixed Points}78In \olref[spine][]{chap}, we suggested that Replacement stands in need9of justification, because it forces the hierarchy to be rather tall.10Having done some cardinal arithmetic, we can give a little11illustration of the height of the hierarchy. 1213Evidently $0 < \aleph_0$, and $1 < \aleph_1$, and $2 < \aleph_2$\ldots14and, indeed, the difference in size only gets \emph{bigger} with every15step. So it is tempting to conjecture that $\kappa< \aleph_\kappa$16for every ordinal $\kappa$. 1718But this conjecture is \emph{false}, given $\ZFC$. In fact, we can19prove that there are \emph{$\aleph$-fixed-points}, i.e.,20cardinals $\kappa$ such that $\kappa=\aleph_\kappa$. 2122\begin{prop}\ollabel{alephfixed}23There is an $\aleph$-fixed-point.24\end{prop}2526\begin{proof}27Using recursion, define:28\begin{align*}29 \kappa_0 &= 0\\30 \kappa_{n+1} &= \aleph_{\kappa_n}\\31 \kappa&= \bigcup_{n < \omega}\kappa_n32\end{align*}33Now $\kappa$ is a cardinal by34\olref[cardinals][classing]{unioncardinalscardinal}. But now:35\[36 \kappa= \bigcup_{n < \omega} \kappa_{n+1} = 37 \bigcup_{n < \omega}\aleph_{\kappa_n} = 38 \bigcup_{\alpha < \kappa}\aleph_\alpha = \aleph_\kappa39\]40% By construction, $\kappa$ is the least cardinal greater than each41% $\kappa_n$. So $\aleph_\kappa$ is the least cardinal greater than42% each $\aleph_{\kappa_n}$, and hence greater than each43% $\kappa_{n+1}$. But equally $\kappa$ is the least cardinal greater44% than each $\kappa_{n+1} = \aleph_{\kappa_n}$. So $\kappa=45% \aleph_\kappa$.46\end{proof}4748Boolos once wrote an article about exactly the $\aleph$-fixed-point we49just constructed. After noting the existence of $\kappa$, at the start50of his article, he said:51\begin{quote}52 [$\kappa$ is] a \emph{pretty big} number, by the lights of those53 with no previous exposure to set theory, so big, it seems to me,54 that it calls into question the truth of any theory, one of whose55 assertions is the claim that there are at least $\kappa$ objects.56 \citep[p.~257]{Boolos2000}57\end{quote}58And he ultimately concluded his paper by asking:59\begin{quote}60 [do] we suspect that, however it may have been at the61 beginning of the story, by the time we have come thus far62 the wheels are spinning and we are no longer listening to63 a description of anything that is the case?64 \citep[p.~268]{Boolos2000}65\end{quote}66If we have, indeed, outrun ``anything that is the case'', then we must67point the finger of blame directly at Replacement. For it is this68axiom which allows our proof to work. In which case, one assumes,69Boolos would need to revisit the claim he made, a few decades earlier,70that Replacement has ``no undesirable'' consequences (see71\olref[replacement][extrinsic]{sec}).7273But is the existence of $\kappa$ so bad? It might help, here, to74consider Russell's \emph{Tristram Shandy paradox}. Tristram Shandy75documents his life in his diary, but it takes him a year to record a76single day. With every passing year, Tristram falls further and77further behind: after one year, he has recorded only one day, and has78lived 364 days unrecorded days; after two years, he has only recorded79two days, and has lived 728 unrecorded days; after three years, he has80only recorded three days, and lived 1092 unrecorded81days \dots\footnote{Forgetting about leap years.} Still, if Tristram82is \emph{immortal}, Tristram will manage to record every day, for he83will record the $n$th day on the $n$th year of his life. And so, ``at84the end of time'', Tristram will have a complete diary. 8586Now: why is this so different from the thought that $\alpha$ is87smaller than $\aleph_\alpha$---and indeed, increasingly, desperately88smaller---up until $\kappa$, at which point, we catch up, and $\kappa89= \aleph_\kappa$?9091Setting that aside, and assuming we accept $\ZFC$, let's close with a92little more fun concerning fixed-point constructions. The next three93results establish, intuitively, that there is a (non-trivial) point at94which the hierarchy is as wide as it is tall:9596\begin{prop}\ollabel{bethfixed}97There is a $\beth$-fixed-point, i.e., a $\kappa$ such that $\kappa=98\beth_\kappa$.99\end{prop}100101\begin{proof}102As in \olref{alephfixed}, using ``$\beth$'' in place of ``$\aleph$''. 103\end{proof}104105\begin{prop}\ollabel{stagesize}106$\card{V_{\omega+\alpha}} = \beth_{\alpha}$. If $\omega \ordtimes107\omega \leq \alpha$, then $\card{V_\alpha} = \beth_\alpha$.108\end{prop}109110\begin{proof}111The first claim holds by a simple transfinite induction. The second112claim follows, since if $\omega \ordtimes \omega \leq \alpha$ then113$\omega + \alpha = \alpha$. To establish this, we use facts about114ordinal arithmetic from \olref[ord-arithmetic][]{chap}. First note115that $\omega \ordtimes \omega = \omega \ordtimes (1 \ordplus \omega) =116(\omega \ordtimes 1) \ordplus (\omega\ordtimes\omega) = \omega117\ordplus (\omega \ordtimes \omega)$. Now if $\omega \ordtimes \omega118\leq \alpha$, i.e., $\alpha = (\omega\ordtimes\omega) \ordplus \beta$119for some $\beta$, then $\omega \ordplus \alpha = \omega \ordplus120((\omega \ordtimes \omega) \ordplus \beta) = (\omega \ordplus (\omega121\ordtimes \omega)) \ordplus \beta = (\omega \ordtimes \omega) \ordplus122\beta = \alpha$. 123\end{proof}124125\begin{cor}126There is a $\kappa$ such that $\card{V_\kappa} = \kappa$.127\end{cor}128129\begin{proof}130Let $\kappa$ be a $\beth$-fixed point, as given by \olref{bethfixed}.131Clearly $\omega \ordtimes \omega < \kappa$. So $\card{V_\kappa} =132\beth_\kappa= \kappa$ by \olref{stagesize}.133\end{proof}134135There are as many stages beneath $V_\kappa$ as there are !!{element}s136of $V_\kappa$. Intuitively, then, $V_\kappa$ is as wide as it is tall.137This is very Tristram-Shandy-esque: we move from one stage to the next138by taking \emph{powersets}, thereby making our hierarchy \emph{much}139bigger with each step. But, ``in the end'', i.e., at stage $\kappa$,140the hierarchy's width catches up with its height. 141142One might ask: \emph{How often does the hierarchy's width match its143height?} The answer is: \emph{As often as there are ordinals.} But144this needs a little explanation. 145146We define a term $\tau$ as follows. For any $A$, let:147\begin{align*}148 \tau_0(A) & \defis \card{A}\\149 \tau_{n+1}(A) & \defis \beth_{\tau_n(A)}\\150 \tau(A) & \defis \bigcup_{n < \omega}\tau_n(A)151\intertext{As in \olref{bethfixed}, $\tau(A)$ is a152$\beth$-fixed point for any $A$, and trivially $\card{A} < \tau(A)$.153So now consider this recursive definition:}154 W_0 &\defis 0\\155 W_{\alpha + 1} & \defis \tau(W_\alpha)\\156 W_\alpha & \defis \bigcup_{\beta < \alpha} W_\beta157 \text{, when $\alpha$ is a limit}158\end{align*}159The construction is defined for all ordinals. Intuitively, then,160$W$ is ``!!a{injection}'' from the ordinals to $\beth$-fixed points.161And, exactly as before, 162$V_{W_\alpha}$ is as wide as it is tall, for any $\alpha$.163164\end{document}