content/set-theory/ord-arithmetic/ord-arithmetic.tex
1% Part: set-theory2% Chapter: ord-arithmetic34\documentclass[../../../include/open-logic-chapter]{subfiles}56\begin{document}78\olchapter{sth}{ord-arithmetic}{Ordinal Arithmetic}910\olimport{introduction}11\olimport{addition}12\olimport{using-addition}13\olimport{multiplication}14\olimport{exponentiation}1516\OLEndChapterHook1718\end{document}
content/set-theory/ord-arithmetic/introduction.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}4\olfileid{sth}{ord-arithmetic}{intro}56\olsection{Introduction}78In \olref[ordinals][]{chap}, we developed a theory of ordinal numbers.9We saw in \olref[spine][]{chap} that we can think of the ordinals as a10spine around which the remainder of the hierarchy is constructed. But11that is not the only role for the ordinals. There is also the task of12performing ordinal arithmetic. 1314We already gestured at this, back in15\olref[ordinals][idea]{sec}, when we spoke of $\omega$,16$\omega+1$ and $\omega+\omega$. At the time, we spoke informally; the17time has come to spell it out properly. However, we should mention18that there is not much philosophy in this chapter; just technical19developments, coupled with a (mildly) interesting observation that we20can do the same thing in two different ways.2122\end{document}
content/set-theory/ord-arithmetic/addition.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{ord-arithmetic}{add}6\olsection{Ordinal Addition}78Suppose we want to add $\alpha$ and $\beta$. We can simply put a9{copy} of $\beta$ immediately after a copy of $\alpha$. (We need to10take \emph{copies}, since we know from11\olref[ordinals][basic]{ordinalsaresubsets} that either $\alpha12\subseteq \beta$ or $\beta \subseteq \alpha$.) The intuitive effect of13this is to run through an $\alpha$-sequence of steps, \emph{and then}14to run through a $\beta$-sequence. The resulting sequence will be15well-ordered; so by16\olref[ordinals][ordtype]{thmOrdinalRepresentation} it is isomorphic17to a (unique) ordinal. That ordinal can be regarded as the \emph{sum}18of $\alpha$ and $\beta$. 1920That is the intuitive idea behind ordinal addition. To define it21rigorously, we start with the idea of taking \emph{copies} of sets.22The idea here is to use arbitrary tags, $0$ and $1$, to keep track of23which object came from where:2425\begin{defn}\ollabel{defdissum}26The \emph{disjoint sum} of $A$ and $B$ is $A \disjointsum B = (A\times27\{0\}) \cup (B \times \{1\})$.28\end{defn}2930We next define an ordering on pairs of ordinals:3132\begin{defn}33For any ordinals $\alpha_1, \alpha_2, \beta_1, \beta_2$, say that:34\begin{align*}35 \tuple{\alpha_1, \alpha_2} \rlexless \tuple{\beta_1, \beta_2}\text{ iff }& 36 \text{either $\alpha_2 \in \beta_2$}\\37 & \text{or both $\alpha_2 = \beta_2$ and $\alpha_1 \in \beta_1$}38\end{align*} 39\end{defn}4041This is a \emph{reverse lexicographic} ordering, since you order by42the second element, then by the first. Now recall that we wanted to43define $\alpha \ordplus \beta$ as the order type of a copy of $\alpha$44followed by a copy of $\beta$. To achieve that, we say:4546\begin{defn}\ollabel{defordplus}47For any ordinals $\alpha$, $\beta$, their sum is $\alpha \ordplus48\beta = \ordtype{\alpha \disjointsum \beta, \rlexless}$.49\end{defn}50\noindent51Note that we slightly abused notation here; strictly we should write ``$\Setabs{\tuple{x,y}\in \alpha\disjointsum\beta}{x \rlexless y}$'' in place of ``$\rlexless$''. For brevity, though, we will continue to abuse notation in this way in what follows. 5253The following result, together with54\olref[ordinals][ordtype]{thmOrdinalRepresentation}, confirms that our55definition is well-formed:5657\begin{lem}\ollabel{ordsumlessiswo} 58$\tuple{\alpha \disjointsum \beta, \rlexless}$ is a well-order, for59any ordinals $\alpha$ and $\beta$.60\end{lem}6162\begin{proof}63Obviously $\rlexless$ is connected on $\alpha \disjointsum \beta$. To64show it is well-founded, fix a non-empty $X \subseteq \alpha65\disjointsum \beta$. Let $Y$ be the subset of $X$ whose second coordinate is as small as possible, i.e.\ $Y = \Setabs{\tuple{\gamma, i} \in X}{(\forall \tuple{\delta, j} \in X)i \leq j}$. Now choose the element of $Y$ with smallest first coordinate. 66\end{proof}67\noindent 68So we have a nice, explicit definition of ordinal addition. Here is69an unsurprising fact (recall that $1 = \{0\}$, by70\olref[z][infinity-again]{defnomega}):71\begin{prop}72$\alpha \ordplus 1 = \ordsucc{\alpha}$, for any ordinal $\alpha$.73\end{prop}7475\begin{proof}76Consider the isomorphism $f$ from $\ordsucc{\alpha} = \alpha \cup77\{\alpha\}$ to $\alpha\disjointsum1 = (\alpha \times \{0\})78\disjointsum (\{0\} \times \{1\})$ given by $f(\gamma) =79\tuple{\gamma, 0}$ for $\gamma \in \alpha$, and $f(\alpha) = \tuple{0,801}$.81\end{proof}82\noindent83Moreover, it is easy to show that addition obeys certain recursive84conditions:8586\begin{lem}\ollabel{ordadditionrecursion}87For any ordinals $\alpha, \beta$, we have:88\begin{align*}89 \alpha\ordplus 0 &= \alpha\\90 \alpha \ordplus (\beta\ordplus 1) &= (\alpha \ordplus \beta) \ordplus 1\\91 \alpha \ordplus \beta &= \supstrict_{\delta < \beta}(\alpha \ordplus \delta) && \text{if $\beta $ is a limit ordinal}92\end{align*}93\end{lem}9495\begin{proof}96We check case-by-case; first:97\begin{align*}98 \alpha \ordplus 0 99 & = \ordtype{(\alpha \times \{0\}) \cup (0 \times \{1\}), \rlexless} \\100 &= \ordtype{(\alpha \times \{0\}) \cup \{0\}, \rlexless}\\101 &= \alpha\\102 \alpha \ordplus (\beta \ordplus 1) 103 %&= \ordtype{(\alpha\times \{0\}) \cup (\ordtype{\beta \ordplus 1}\times \{1\}), \rlexless} \\104 &= \ordtype{(\alpha\times \{0\}) \cup (\ordsucc{\beta}\times \{1\}), \rlexless} \\105 &= \ordtype{(\alpha\times \{0\}) \cup (\beta \times \{1\}), \rlexless} \ordplus 1\\106 &= (\alpha \ordplus \beta) \ordplus 1107\end{align*}108Now let $\beta \neq \emptyset$ be a limit. If $\delta < \beta$ then109also $\delta\ordplus 1 < \beta$, so $\alpha \ordplus \delta$ is a110proper initial segment of $\alpha \ordplus \beta$. So $\alpha111\ordplus \beta$ is a strict upper bound on $X = \Setabs{\alpha112\ordplus \delta}{\delta < \beta}$. Moreover, if $\alpha \leq \gamma <113\alpha \ordplus \beta$, then clearly $\gamma = \alpha \ordplus114\delta$ for some $\delta < \beta$. So $\alpha \ordplus \beta =115\supstrict_{\delta< \beta}(\alpha\ordplus \delta)$.116\end{proof}117118But here is a striking fact. To define ordinal addition, we could119\emph{instead} have simply used the Transfinite Recursion Theorem, and120laid down the recursion equations, exactly as given in121\olref{ordadditionrecursion} (though using ``$\ordsucc{\beta}$''122rather than ``$\beta \ordplus 1$'').123124There are, then, two different ways to define operations on the125ordinals. We can define them \emph{synthetically}, by explicitly126constructing a well-ordered set and considering its order type. Or we127can define them \emph{recursively}, just by laying down the recursion128equations. Done correctly, though, the outcome is identical. For129\olref[ordinals][ordtype]{thmOrdinalRepresentation} guarantees130that these recursion equations pin down \emph{unique} ordinals.131132In many ways, ordinal arithmetic behaves just like addition of the133natural numbers. For example, we can prove the following:134135\begin{lem}\ollabel{ordinaladditionisnice}136If $\alpha, \beta, \gamma$ are ordinals, then:137\begin{enumerate}138 \item\ollabel{ordaddition1} if $\beta < \gamma$, then $\alpha139 \ordplus \beta < \alpha \ordplus \gamma$140 \item\ollabel{ordaddition2} if $\alpha \ordplus \beta =141 \alpha\ordplus \gamma$, then $\beta = \gamma$142 \item\ollabel{ordaddition3} $\alpha \ordplus (\beta \ordplus143 \gamma) = (\alpha \ordplus \beta) \ordplus \gamma$, i.e.,144 addition is associative145 \item\ollabel{ordaddition4} If $\alpha \leq \beta$, then $\alpha146 \ordplus \gamma \leq \beta \ordplus \gamma$147\end{enumerate}148\end{lem}149150\begin{proof}151We prove \olref{ordaddition3}, leaving the rest as an exercise. The152proof is by Simple Transfinite Induction on $\gamma$, using153\olref{ordadditionrecursion}. When $\gamma = 0$:154\[155(\alpha \ordplus \beta) \ordplus 0 = \alpha \ordplus \beta = \alpha \ordplus (\beta \ordplus 0)156\]157When $\gamma = \delta\ordplus 1$, suppose for induction that $(\alpha158\ordplus \beta) \ordplus \delta = \alpha \ordplus (\beta \ordplus159\delta)$; now using \olref{ordadditionrecursion} three times:160\begin{align*}161 (\alpha \ordplus \beta) \ordplus (\delta \ordplus 1) & = ((\alpha \ordplus \beta) \ordplus \delta)\ordplus 1\\162 & = (\alpha \ordplus (\beta \ordplus \delta)) \ordplus 1\\163 & = \alpha \ordplus ((\beta \ordplus \delta)\ordplus 1)\\164 & = \alpha \ordplus (\beta \ordplus (\delta\ordplus 1))165\end{align*} 166When $\gamma$ is a limit ordinal, suppose for induction that if167$\delta \in \gamma$ then $(\alpha \ordplus \beta) \ordplus \delta =168\alpha \ordplus (\beta \ordplus \delta)$; now:169\begin{align*}170 (\alpha \ordplus \beta) \ordplus \gamma & = \supstrict_{\delta < \gamma}((\alpha \ordplus \beta) \ordplus \delta) \\171 &= \supstrict_{\delta < \gamma}(\alpha \ordplus (\beta \ordplus \delta))\\172 &= \alpha \ordplus \supstrict_{\delta < \gamma}(\beta \ordplus \delta)\\173 & = \alpha \ordplus (\beta \ordplus \gamma)174\end{align*}175\end{proof}176177\begin{prob}178Prove the remainder of179\olref[sth][ord-arithmetic][add]{ordinaladditionisnice}.180\end{prob}181182In these ways, ordinal addition should be very familiar. But, there is a crucial way in which ordinal addition is \emph{not} like addition on the natural numbers.183184\begin{prop}\ollabel{ordsumnotcommute}185Ordinal addition is {not} commutative; $1 \ordplus \omega = \omega <186\omega \ordplus 1$.187\end{prop}188189\begin{proof}190Note that $1 \ordplus \omega = \supstrict_{n < \omega} (1 \ordplus n)191= \omega \in \omega \cup \{\omega\} = \ordsucc{\omega} = \omega192\ordplus 1$.193\end{proof}194\noindent 195Whilst this may initially come as a surprise, \emph{it shouldn't}. On196the one hand, when you consider $1 \ordplus \omega$, you are thinking197about the order type you get by putting an extra element \emph{before}198all the natural numbers. Reasoning as we did with Hilbert's Hotel in199\olref[sfr][infinite][hilbert]{sec}, intuitively, this extra first200element shouldn't make any difference to the overall order type. On201the other hand, when you consider $\omega \ordplus 1$, you are202thinking about the order type you get by putting an extra element203\emph{after} all the natural numbers. And that's a radically different204beast!205206\end{document}
content/set-theory/ord-arithmetic/using-addition.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{ord-arithmetic}{using-addition}6\olsection{Using Ordinal Addition}78Using addition on the ordinals, we can explicitly calculate the ranks9of various sets, in the sense of \olref[spine][rank]{defnsetrank}:1011\begin{lem}\ollabel{rankcomputation}12If $\setrank{A} = \alpha$ and $\setrank{B} = \beta$, then:13\begin{enumerate}14 \item\ollabel{exrankpow} $\setrank{\Pow{A}} = \alpha\ordplus 1$15 \item\ollabel{exrankpair} $\setrank{\{A, B\}} = \max(\alpha,16 \beta) \ordplus 1$17 \item\ollabel{exrankcup} $\setrank{A \cup B} = \max(\alpha,18 \beta)$19 \item\ollabel{exranktuple} $\setrank{\tuple{A,B}} = \max(\alpha,20 \beta) \ordplus 2$21 \item\ollabel{exranktimes} $\setrank{A \times B} \leq \max(\alpha,22 \beta) \ordplus 2$23 \item\ollabel{exrankunion} $\setrank{\bigcup A} = \alpha$ when24 $\alpha$ is empty or a limit; $\setrank{\bigcup A} = \gamma$ when25 $\alpha = \gamma\ordplus 1$26\end{enumerate}27\end{lem}2829\begin{proof}30Throughout, we invoke \olref[spine][rank]{ranksupstrict}31repeatedly.3233\emph{\olref{exrankpow}.} If $x \subseteq A$ then $\setrank{x} \leq34\setrank{A}$. So $\setrank{\Pow{A}} \leq \alpha \ordplus 1$. Since $A35\in \Pow{A}$ in particular, $\setrank{\Pow{A}} = \alpha \ordplus 1$.3637\emph{\olref{exrankpair}.} By \olref[spine][rank]{ranksupstrict}3839\emph{\olref{exrankcup}.} By \olref[spine][rank]{ranksupstrict}.4041\emph{\olref{exranktuple}.} By \olref{exrankpair}, twice.4243\emph{\olref{exranktimes}.} Note that $A \times B \subseteq44\Pow{\Pow{A \cup B}}$, and invoke \olref{exranktuple}. 4546\emph{\olref{exrankunion}.} If $\alpha = \gamma\ordplus 1$, there is47some $c \in A$ with $\setrank{c} = \gamma$, and no !!{element} of $A$48has higher rank; so $\setrank{\bigcup A} = \gamma$. If $\alpha$ is a49limit ordinal, then $A$ has !!{element}s with rank arbitrarily close50to (but strictly less than) $\alpha$, so that $\bigcup A$ also has51!!{element}s with rank arbitrarily close to (but strictly less than)52$\alpha$, so that $\setrank{\bigcup A} = \alpha$.53\end{proof}54\noindent55We leave it as an exercise to show why \olref{exranktimes} involves an56\emph{in}equality.5758\begin{prob}59Produce sets $A$ and $B$ such that $\setrank{A \times B}=60\max(\setrank{A}, \setrank{B})$. Produce sets $A$ and $B$ such that61$\setrank{A \times B}\max(\setrank{A}, \setrank{B}) \ordplus 2$. Are62any other ranks possible?63\end{prob}6465We are also now in a position to show that several reasonable notions66of what it might mean to describe an ordinal as ``finite'' or ``infinite'' coincide:6768\begin{lem}\ollabel{ordinfinitycharacter}69For any ordinal $\alpha$, the following are equivalent:70\begin{enumerate}71 \item\ollabel{ord:notinomega} $\alpha\notin \omega$, i.e.,72 $\alpha$ is not a natural number73 \item\ollabel{ord:omegaplus} $\omega \leq \alpha$ 74 \item\ollabel{ord:oneplus} $1 \ordplus \alpha = \alpha$ 75 %\alpha \approx \alpha \ordplus 1$76 \item\ollabel{ord:plusone} $\alpha \approx \alpha\ordplus 1$,77 i.e., $\alpha$ and $\alpha\ordplus 1$ are equinumerous78 \item\ollabel{ord:infinite} $\alpha$ is Dedekind infinite 79 \end{enumerate}80\end{lem}81\noindent82So we have five provably equivalent ways to understand what it takes for an ordinal to be (in)finite.8384\begin{proof}85\emph{\olref{ord:notinomega} $\Rightarrow$ \olref{ord:omegaplus}.} By86Trichotomy. 8788\emph{\olref{ord:omegaplus} $\Rightarrow$ \olref{ord:oneplus}.} Fix89$\alpha \geq \omega$. By Transfinite Induction, there is some least90ordinal $\gamma$ (possibly $0$) such that there is a limit ordinal91$\beta$ with $\alpha = \beta \ordplus \gamma$. Now:92\[93 1 \ordplus \alpha = 94 1 \ordplus (\beta \ordplus \gamma) = 95 (1 \ordplus \beta) \ordplus \gamma = 96 \supstrict_{\delta < \beta} (1 \ordplus \delta) \ordplus \gamma = 97 \beta \ordplus \gamma = 98 \alpha.99\]100\emph{\olref{ord:oneplus} $\Rightarrow$ \olref{ord:plusone}.} There is101clearly !!a{bijection} $f \colon (\alpha \disjointsum 1) \to (1102\disjointsum \alpha)$. If $1 \ordplus \alpha = \alpha$, there is an103isomorphism $g \colon (1 \disjointsum \alpha) \to \alpha$. Now104consider $\comp{f}{g}$.105106\emph{\olref{ord:plusone} $\Rightarrow$ \olref{ord:infinite}.} If107$\alpha \approx \alpha \ordplus 1$, there is !!a{bijection} $f \colon108(\alpha \disjointsum 1) \to \alpha$. Define $g(\gamma) = f(\gamma, 0)$109for each $\gamma < \alpha$; this !!{injection} witnesses that $\alpha$110is Dedekind infinite, since $f(0,1) \in \alpha \setminus \ran{g}$. 111112\emph{\olref{ord:infinite} $\Rightarrow$ \olref{ord:notinomega}.} This113is \olref[z][infinity-again]{naturalnumbersarentinfinite}.114\end{proof}115116\end{document}
content/set-theory/ord-arithmetic/multiplication.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{ord-arithmetic}{mult}67\olsection{Ordinal Multiplication}89We now turn to ordinal multiplication, and we approach this much like10ordinal addition. So, suppose we want to multiply $\alpha$ by~$\beta$.11To do this, you might imagine a rectangular grid, with width $\alpha$12and height $\beta$; the product of $\alpha$ and $\beta$ is now the13result of moving along each row, then moving through the next14row\ldots until you have moved through the entire grid. Otherwise put,15the product of $\alpha$ and $\beta$ arises by replacing \emph{each}16element in $\beta$ with a copy of $\alpha$. 1718To make this formal, we simply use the reverse lexicographic ordering19on the Cartesian product of $\alpha$ and $\beta$:2021\begin{defn}22For any ordinals $\alpha, \beta$, their product $\alpha \ordtimes \beta = \ordtype{\alpha \times \beta, \rlexless}$.23\end{defn}24\noindent25We must again confirm that this is a well-formed definition:2627\begin{lem}\ollabel{ordtimeslessiswo}28$\tuple{\alpha \times \beta, \rlexless}$ is a well-order, for any29ordinals $\alpha$ and $\beta$.30\end{lem}3132\begin{proof}33Exactly as for \olref[add]{ordsumlessiswo}.34\end{proof}35\noindent36And it is not hard to prove that multiplication behaves thus:3738\begin{lem}\ollabel{ordtimesrecursion}39For any ordinals $\alpha, \beta$:40\begin{align*}41 \alpha \ordtimes 0 &= 0\\42 \alpha \ordtimes (\beta \ordplus 1) &= 43 (\alpha \ordtimes \beta) \ordplus \alpha\\44 \alpha \ordtimes \beta &= 45 \supstrict_{\delta < \beta}(\alpha \ordtimes \delta) && 46 \text{when $\beta$ is a limit ordinal}.47\end{align*}48\end{lem}4950\begin{proof}51Left as an exercise.52\end{proof}5354Indeed, just as in the case of addition, we could have defined ordinal55multiplication via these recursion equations, rather than offering a56direct definition. Equally, as with addition, certain behaviour is57familiar:5859\begin{lem}\ollabel{ordinalmultiplicationisnice}60If $\alpha, \beta, \gamma$ are ordinals, then:61\begin{enumerate}62 \item\ollabel{ordtimes1} if $\alpha \neq 0$ and $\beta < \gamma$,63 then $\alpha \ordtimes \beta < \alpha \ordtimes \gamma$;64 \item\ollabel{ordtimes2} if $\alpha \neq 0$ and $\alpha \ordtimes65 \beta = \alpha\ordtimes\gamma$, then $\beta = \gamma$;66 \item\ollabel{ordtimes3} $\alpha \ordtimes (\beta \ordtimes67 \gamma) = (\alpha \ordtimes \beta) \ordtimes \gamma$;68 \item\ollabel{ordtimes4} If $\alpha \leq \beta$, then $\alpha69 \ordtimes \gamma \leq \beta \ordtimes \gamma$;70 \item\ollabel{ordtimes5} $\alpha \ordtimes (\beta \ordplus71 \gamma) = (\alpha \ordtimes \beta )\ordplus (\alpha\ordtimes72 \gamma)$.73\end{enumerate}74\end{lem}7576\begin{proof}77Left as an exercise.78\end{proof}7980You can prove (or look up) other results, to your heart's content.81But, given82\olref[ord-arithmetic][add]{ordsumnotcommute}, the83following should not come as a surprise:8485\begin{prop}86Ordinal multiplication is not commutative: $2 \ordtimes \omega =87\omega < \omega \ordtimes 2$88\end{prop}8990\begin{proof}91$2 \ordtimes \omega = \supstrict_{n < \omega}(2\ordtimes n) = \omega \in \supstrict_{n < \omega}(\omega \ordplus n) = \omega \ordplus \omega = \omega \ordtimes 2$.92\end{proof}93\noindent 94Again, the intuitive rationale is quite straightforward. To compute $295\ordtimes \omega$, you replace each natural number with two entities.96You would get the same order type if you simply inserted all the97``half'' numbers into the natural numbers, i.e., you considered the98natural ordering on $\Setabs{\nicefrac{n}{2}}{n \in \omega}$. And, put99like that, the order type is plainly the same as that of $\omega$100itself. But, to compute $\omega \ordtimes 2$, you place down two101copies of $\omega$, one after the other. 102103\begin{prob}104Prove105\olref[sth][ord-arithmetic][mult]{ordtimeslessiswo},106\olref[sth][ord-arithmetic][mult]{ordtimesrecursion},107and108\olref[sth][ord-arithmetic][mult]{ordinalmultiplicationisnice}109\end{prob}110111\end{document}
content/set-theory/ord-arithmetic/exponentiation.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{ord-arithmetic}{expo}6\olsection{Ordinal Exponentiation}78We now move to ordinal exponentiation. Sadly, there is no \emph{nice}9synthetic definition for ordinal exponentiation.1011Sure, there \emph{are} explicit synthetic definitions. Here is one.12Let $\text{finfun}(\alpha,\beta)$ be the set of all functions $f13\colon \alpha \to \beta$ such that $\Setabs{\gamma \in14\alpha}{f(\gamma) \neq 0}$ is equinumerous with some natural number.15Define a well-ordering on $\text{finfun}(\alpha,\beta)$ by $f16\sqsubset g$ iff $f \neq g$ and $f(\gamma_0) < g(\gamma_0)$, where17$\gamma_0 = \text{max}\Setabs{\gamma \in \alpha}{f(\gamma) \neq18g(\gamma)}$. Then we can define $\ordexpo{\alpha}{\beta}$ as19$\ordtype{\text{finfun}(\alpha, \beta), \sqsubset}$. Potter employs20this explicit definition, and then immediately explains:21\begin{quote}22 The choice of this ordering is determined purely by our desire to23 obtain a definition of ordinal exponentiation which obeys the24 appropriate recursive condition\ldots, and it is much harder to25 picture than either the ordered sum or the ordered product.26 \citep[p.~199]{Potter2004}27\end{quote}28Quite. We explained addition as ``a copy of $\alpha$ followed by a29copy of $\beta$'', and multiplication as ``a $\beta$-sequence of30copies of $\alpha$''. But we have nothing pithy to say about31$\text{finfun}(\alpha, \gamma)$. So instead, we'll offer the32definition of ordinal exponentiation just \emph{by} transfinite33recursion, i.e.:3435\begin{defn}\ollabel{ordexporecursion}36\begin{align*}37 \ordexpo{\alpha}{0} &= 1\\38 \ordexpo{\alpha}{\beta\ordplus 1} &=\ordexpo{\alpha}{\beta} \ordtimes \alpha\\39 \ordexpo{\alpha}{\beta} &= \bigcup_{\delta < \beta}\ordexpo{\alpha}{\delta}& & \text{when $\beta$ is a limit ordinal}40\end{align*}41\end{defn}4243If we were working \emph{as} set theorists, we might want to explore44some of the properties of ordinal exponentiation. But we have nothing45much more to add, except to note the unsurprising fact that ordinal46exponentiation does not commute. Thus $\ordexpo{2}{\omega} =47\bigcup_{\delta < \omega}\ordexpo{2}{\delta} = \omega$, whereas48$\ordexpo{\omega}{2} = \omega \ordtimes \omega$. But then, we should49not \emph{expect} exponentiation to commute, since it does not commute50with natural numbers: $\ordexpo{2}{3} = 8 < 9 = \ordexpo{3}{2}$. 5152\begin{prob}53Using Transfinite Induction, prove that, if we define54$\ordexpo{\alpha}{\beta} = \ordtype{\text{finfun}(\alpha, \beta),55\sqsubset}$, we obtain the recursion equations of56\olref[sth][ord-arithmetic][expo]{ordexporecursion}.57\end{prob}5859\end{document}