content/set-theory/z/z.tex
1% Part: set-theory2% Chapter: z34\documentclass[../../../include/open-logic-chapter]{subfiles}56\begin{document}78\olchapter{sth}{z}{Steps towards $\Z$}910\olimport{story}11\olimport{separation}12\olimport{union}13\olimport{pairs}14\olimport{powerset}15\olimport{infinity-again}16\olimport{milestone}17\olimport{nat}18\olimport{arbintersections}1920\OLEndChapterHook2122\end{document}
content/set-theory/z/story.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{story}6\olsection{The Story in More Detail}7 8In \olref[story][approach]{sec}, we quoted Schoenfield's9description of the process of set-formation. We now want to write down10a few more principles, to make this story a bit more precise. Here11they are:12\begin{enumerate}13\item[] \stageshier. Every set is formed at some stage.14\item[] \stagesord. Stages are ordered: some come \emph{before}15others.\footnote{We will actually assume---tacitly---that the stages16are \emph{well-ordered}. What this amounts to is explained in17\olref[ordinals][]{chap}. This is a substantial assumption. In fact,18using a very clever technique due to \citet{Scott1974}, this19assumption can be \emph{avoided} and then \emph{derived}. (This will20also explain why we should think that there is an initial stage.) We cannot go into that here; for more, see \citet{ButtonLT1}.} 21\item[] \stagesacc. For any stage $S$, and for any sets which were22formed \emph{before} stage~$S$: a set is formed at stage~$S$ whose23members are exactly those sets. Nothing else is formed at stage~$S$.24\end{enumerate}25These are informal principles, but we will be able to use them to26vindicate several of the axioms of Zermelo's set theory. 2728(We should offer a word of caution. Although we will be presenting some29completely standard axioms, with completely standard names, the30italicized principles we have just presented have no particular names31in the literature. We simply monikers which we hope are helpful.)3233\end{document}
content/set-theory/z/separation.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{sep}6\olsection{Separation}78We start with a principle to replace Na\"{i}ve Comprehension:910\begin{axiom}[Scheme of Separation] For every formula $\phi(x)$, this is an11axiom: for any $A$, the set $\Setabs{x \in A}{\phi(x)}$ exists.12\end{axiom}1314Note that this is not a single axiom. It is a \emph{scheme} of axioms.15There are \emph{infinitely many} Separation axioms; one for every16formula $\phi(x)$. The scheme can equally well be (and normally is)17written down as follows:1819\begin{defish}20For any formula $\phi(x)$ which does not contain ``$S$'', this is an21axiom:22\[23 \forall A \exists S \forall x(x \in S \liff (\phi(x) \land x \in A)).24\]25\end{defish}2627In keeping with the convention noted at the start of28\olref[sth][][]{part}, the formulas~$\phi$ in the Separation axioms29may have parameters.\footnote{For an explanation of what this means,30see the discussion immediately after31\olref[sfr][infinite][induction]{natinductionschema}.}3233Separation is immediately justified by our cumulative-iterative34conception of sets we have been telling. To see why, let $A$ be a set.35So $A$ is formed by some stage~$S$ (by \stageshier). Since $A$ was36formed at stage~$S$, all of $A$'s members were formed before stage $S$37(by \stagesacc). Now in particular, consider all the sets which are38members of $A$ and which also satisfy $\phi$; clearly all of these39sets, too, were formed before stage~$S$. So they are formed into a set40$\Setabs{x \in A}{\phi(x)}$ at stage~$S$ too (by \stagesacc).4142Unlike Na\"ive Comprehension, this avoid Russell's Paradox. For we43cannot simply assert the existence of the set $\Setabs{x}{x \notin44x}$. Rather, \emph{given} some set~$A$, we can assert the existence of45the set $R_A = \Setabs{x \in A}{x \notin x}$. But all this proves is46that $R_A \notin R_A$ and $R_A \notin A$, none of which is very47worrying.4849However, Separation has an immediate and striking consequence:5051\begin{thm}\ollabel{thm:NoUniversalSet}52There is no \emph{universal} set, i.e., $\Setabs{x}{x = x}$ does not exist.53\end{thm}5455\begin{proof}56For reductio, suppose $V$ is a universal set. Then by Separation, $R =57\Setabs{x \in V}{x \notin x} = \Setabs{x}{x \notin x}$ exists,58contradicting Russell's Paradox.59\end{proof}6061The absence of a universal set---indeed, the open-endedness of the62hierarchy of sets---is one of the most fundamental ideas behind the63cumulative-iterative conception. So it is worth seeing that,64intuitively, we could reach it via a different route. A universal set65must be !!a{element} of itself. But, on our cumulative-iterative66conception, every set appears (for the first time) in the hierarchy at67the first stage immediately after all of its !!{element}s. But this68entails that \emph{no} set is self-membered. For any self-membered set69would have to first occur immediately after the stage at which it70first occurred, which is absurd. (We will see in71\olref[sth][spine][rank]{defnsetrank} how to make this explanation72more rigorous, by using the notion of the ``rank'' of a set. However,73we will need to have a few more axioms in place to do this.)7475Here are a few more consequences of Separation and Extensionality.7677\begin{prop}\ollabel{prop:emptyexists}78If any set exists, then $\emptyset$ exists.79\end{prop}8081\begin{proof}82If $A$ is a set, $\emptyset = \Setabs{x \in A}{x \neq x}$ exists by Separation.83\end{proof}8485\begin{prop}86$A \setminus B$ exists for any sets $A$ and $B$87\end{prop}8889\begin{proof}90$A \setminus B = \Setabs{x \in A}{x \notin B}$ exists by Separation.91\end{proof}9293It also turns out that (almost) arbitrary intersections exist:9495\begin{prop}\ollabel{prop:intersectionsexist}96If $A \neq \emptyset$, then $\bigcap A = \Setabs{x}{(\forall y \in A)x \in y}$ exists.97\end{prop}9899\begin{proof}100Let $A \neq \emptyset$, so there is some $c \in A$. Then $\bigcap A =101\Setabs{x}{(\forall y \in A)x \in y} = \Setabs{x \in c}{(\forall y \in102A)x \in y}$, which exists by Separation.103\end{proof}104105Note the condition that $A \neq \emptyset$, though; for $\bigcap106\emptyset$ would be the universal set, vacuously, contradicting107\olref{thm:NoUniversalSet}.108109\end{document}
content/set-theory/z/union.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{union}6\olsection{Union}78\olref[sth][z][sep]{prop:intersectionsexist} gave us intersections.9But if we want arbitrary unions to exist, we need to lay down another10axiom:1112\begin{axiom}[Union] For any set $A$, the set $\bigcup A =13\Setabs{x}{(\exists b \in A) x \in b}$ exists.14\[15 \forall A \exists U \forall x(x \in U \liff (\exists b \in A)x \in b)16\]17\end{axiom}1819This axiom is also justified by the cumulative-iterative conception.20Let $A$ be a set, so $A$ is formed at some stage $S$ (by \stageshier).21Every member of $A$ was formed \emph{before} $S$ (by \stagesacc); so,22reasoning similarly, every member of every member of $A$ was formed23before $S$. Thus all of \emph{those} sets are available before $S$, to24be formed into a set at $S$. And that set is just $\bigcup A$.2526\end{document}
content/set-theory/z/pairs.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{pairs}6\olsection{Pairs}78The next axiom to consider is the following:910\begin{axiom}[Pairs]11For any sets $a, b$, the set $\{a, b\}$ exists.12\[13 \forall a \forall b \exists P \forall x (x \in P \liff (x = a \lor x = b))14\]15\end{axiom}1617Here is how to justify this axiom, using the iterative conception. Suppose $a$ is available at stage $S$, and $b$ is available at stage $T$. Let $M$ be whichever of stages $S$ and $T$ comes later. Then since $a$ and $b$ are both available at stage $M$, the set $\{a,b\}$ is a possible collection available at any stage after $M$ (whichever is the greater).1819But hold on!{} Why assume that there \emph{are} any stages after $M$? If there are none, then our justification will fail. So, to justify Pairs, we will have to add another principle to the story we told in \olref[sth][z][story]{sec}, namely:20\begin{enumerate}21 \item[] \stagessucc. There is no last stage.22\end{enumerate}23Is this principle justified? Nothing in Shoenfield's story stated24\emph{explicitly} that there is no last stage. Still, even if it is25(strictly speaking) an extra addition to our story, it fits well with26the basic idea that sets are formed in stages. We will simply accept27it in what follows. And so, we will accept the Axiom of Pairs too.2829Armed with this new Axiom, we can prove the existence of plenty more sets. For example:3031\begin{prop}\ollabel{prop:pairsconsequences}32For any sets $a$ and $b$, the following sets exist:33 \begin{enumerate}34 \item\ollabel{singleton} $\{a\}$35 \item\ollabel{binunion} $a \cup b$36 \item\ollabel{tuples} $\tuple{a, b}$37 \end{enumerate}38\end{prop}3940\begin{proof}41\olref{singleton}. By Pairs, $\{a, a\}$ exists, which is $\{a\}$ by42Extensionality.4344\olref{binunion}. By Pairs, $\{a, b\}$ exists. Now $a \cup b = \bigcup45\{a, b\}$ exists by Union.4647\olref{tuples}. By \olref{singleton}, $\{a\}$ exists. By Pairs, $\{a,48b\}$ exists. Now $\{\{a\}, \{a, b\}\} = \tuple{a, b}$ exists, by Pairs49again.50\end{proof}5152\begin{prob}53Show that, for any sets $a, b, c$, the set $\{a, b, c\}$ exists.54\end{prob}5556\begin{prob}57Show that, for any sets $a_1, \ldots, a_n$, the set $\{a_1, \ldots,58a_n\}$ exists.59\end{prob}6061%\begin{proof} By two applications of Pairs, $\{\{a_1, a_2\}, \{a_1,62% a_3\}\}$ exists. By Union and Extensionality, $\bigcup \{\{a_1,63% a_2\}, \{a_1, a_3\}\} = \{a_1, a_2, a_3\}$ exists. Repeat this64% trick as often as necessary. \end{proof}6566\end{document}
content/set-theory/z/powerset.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{power}6\olsection{Powersets}78We will proceed with another axiom:910\begin{axiom}[Powersets]11For any set $A$, the set $\Pow{A} = \Setabs{x}{x \subseteq A}$ exists.12\[13 \forall A \exists P \forall x(x \in P \liff (\forall z \in x)z \in A)14\]15\end{axiom}1617Our justification for this is pretty straightforward. Suppose $A$ is18formed at stage $S$. Then all of $A$'s members were available before19$S$ (by \stagesacc). So, reasoning as in our justification for20Separation, every subset of $A$ is formed by stage $S$. So they are21all available, to be formed into a single set, at any stage after $S$.22And we know that there is some such stage, since $S$ is not the last23stage (by \stagessucc). So $\Pow{A}$ exists.2425Here is a nice consequence of Powersets:2627\begin{prop}\label{thm:Products}28Given any sets $A, B$, their Cartesian product $A \times B$ exists.29\end{prop}3031\begin{proof}32The set $\Pow{\Pow{A \cup B}}$ exists by Powersets and33\olref[sth][z][pairs]{prop:pairsconsequences}. So by Separation, this34set exists:35\[36 C = \Setabs{z \in \Pow{\Pow{A \cup B}}}{(\exists x \in A)(\exists y \in B) z = \tuple{x, y}}.37\]38Now, for any $x \in A$ and $y \in B$, the set $\tuple{x, y}$ exists by39\olref[sth][z][pairs]{prop:pairsconsequences}. Moreover, since $x, y40\in A \cup B$, we have that $\{x\}, \{x, y\} \in \Pow{A \cup B}$, and41$\tuple{x,y} \in \Pow{\Pow{A \cup B}}$. So $A \times B = C$.42\end{proof}4344In this proof, Powerset interacts with Separation. And that is no45surprise. Without Separation, Powersets wouldn't be a very46\emph{powerful} principle. After all, Separation tells us which47subsets of a set exist, and hence determines just how ``fat'' each48Powerset is.4950\begin{prob}51Show that, for any sets $A, B$: (i) the set of all relations with52domain $A$ and range $B$ exists; and (ii) the set of all functions53from $A$ to $B$ exists.54\end{prob}5556\begin{prob}57Let $A$ be a set, and let $\sim$ be an equivalence relation on $A$.58Prove that the set of equivalence classes under $\sim$ on $A$, i.e.,59$\equivclass{A}{\sim}$, exists.60\end{prob}6162\end{document}
content/set-theory/z/infinity-again.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{infinity-again} 6\olsection{Infinity}78We already have enough axioms to ensure that there are infinitely many9sets (if there are any). For suppose some set exists, and so10$\emptyset$ exists (by \olref[sth][z][sep]{prop:emptyexists}). Now for11any set~$x$, the set $x \cup \{x\}$ exists by12\olref[sth][z][pairs]{prop:pairsconsequences}. So, applying this a few13times, we will get sets as follows:14\begin{enumerate}15 \item[0.] $\emptyset$ %& $0$ & stage $0$\\16 \item[1.] $ \{\emptyset\}$ %& $1$ & stage $1$\\17 \item[2.] $\{\emptyset, \{\emptyset\}\}$ %& $2$ & stage $2$\\18 \item[3.] $\{\emptyset, \{\emptyset\}, \{\emptyset, \{\emptyset\}\}\}$ %&$3$ & stage $3$\\19 \item[4.] $\{\emptyset, \{\emptyset\}, \{\emptyset, \{\emptyset\}\}, \{\emptyset, \{\emptyset\}, \{\emptyset, \{\emptyset\}\}\}\}$% & $4$ & stage $4$ 20\end{enumerate}21and we can check that each of these sets is distinct. 2223We have started the numbering from $0$, for a few reasons. But one of24them is this. It is not that hard to check that the set we have25labelled ``$n$'' has exactly $n$ members, and (intuitively) is formed26at the $n$th stage. 2728But. This gives us \emph{infinitely many} sets, but it does not29guarantee that there is an \emph{infinite set}, i.e., a set with30infinitely many members. And this really matters: unless we can find a31(Dedekind) infinite set, we cannot construct a Dedekind algebra. But32we want a Dedekind algebra, so that we can treat it as the set of33natural numbers. (Compare34\olref[sfr][infinite][dedekindsproof]{sec}.)3536Importantly, the axioms we have laid down so far do \emph{not}37guarantee the existence of any infinite set. So we have to lay down a38new axiom:3940\begin{axiom}[Infinity]41There is a set, $I$, such that $\emptyset \in I$ and $x \cup \{x\} \in I$ whenever $x \in I$.42\begin{align*}43 \exists I( & (\exists o \in I)\forall x\ x \notin o \land {}\\44 & (\forall x \in I)(\exists s \in I)\forall z(z \in s \liff (z \in x \lor z = x)))45\end{align*}46\end{axiom}4748It is easy to see that the set $I$ given to us by the Axiom of49Infinity is Dedekind infinite. Its distinguished element is50$\emptyset$, and the injection on $I$ is given by $s(x) = x\cup51\{x\}$. Now,52\olref[sfr][infinite][dedekind]{thm:DedekindInfiniteAlgebra} showed53how to extract a Dedekind Algebra from a Dedekind infinite set; and we54will treat this as our set of natural numbers. More precisely:55\begin{defn}\ollabel{defnomega}56Let $I$ be any set given to us by the Axiom of Infinity. Let $s$57be the function $s(x) = x \cup \{x\}$. Let $\omega =58\closureofunder{s}{\emptyset}$. We call the members of $\omega$ the59\emph{natural numbers}, and say that $n$ is the result of $n$-many60applications of $s$ to $\emptyset$.61\end{defn}6263You can now look back and check that the set labelled ``$n$'', a few64paragraphs earlier, will be treated \emph{as} the number $n$. 6566We will discuss this significance of this stipulation in67\olref[sth][z][nat]{sec}. For now, it enables us to prove an68intuitive result:6970\begin{prop}\ollabel{naturalnumbersarentinfinite}71No natural number is Dedekind infinite.72\end{prop}7374\begin{proof}75The proof is by induction, i.e.,76\olref[sfr][infinite][induction]{thm:dedinfiniteinduction}. Clearly $077= \emptyset$ is not Dedekind infinite. For the induction step, we will78establish the contrapositive: if (absurdly) $s(n)$ is Dedekind79infinite, then $n$ is Dedekind infinite. 8081So suppose that $s(n)$ is Dedekind infinite, i.e., there is some82!!{injection} $f$ with $\ran{f}\subsetneq \dom{f} = s(n) = n \cup83\{n\}$. There are two cases to consider. 8485\emph{Case 1: $n \notin \ran{f}$.} So $\ran{f} \subseteq n$, and $f(n) \in n$. Let $g = \funrestrictionto{f}{n}$; now $\ran{g} =86\ran{f} \setminus \{f(n)\} \subsetneq n = \dom{g}$. Hence $n$ is87Dedekind infinite. 8889\emph{Case 2: $n \in \ran{f}$.} Fix $m \in \dom{f} \setminus \ran{f}$, and define a function $h$ with domain $s(n) = n \cup \{n\}$:90\[91h(x) = 92 \begin{cases}93 f(x) & \text{if }f(x) \neq n\\94 m & \text{if }f(x)=n95 \end{cases}96\]97So $h$ and $f$ agree everywhere, except that $h(f^{-1}(n)) = m98\neq n = f(f^{-1}(n))$. Since $f$ is !!a{injection}, $n \notin99\ran{h}$; and $\ran{h} \subsetneq \dom{h} = s(n)$. Now $n$ is100Dedekind infinite, using the argument of Case 1.101\end{proof}102103The question remains, though, of how we might \emph{justify} the Axiom104of Infinity. The short answer is that we will need to add another105principle to the story we have been telling. That principle is as106follows:107\begin{enumerate}108 \item[] \stagesinf. There is an infinite stage. That is, there is a stage which (a) is not the first stage, and which (b) has some stages before it, but which (c) has no immediate predecessor.109\end{enumerate}110The Axiom of Infinity follows straightforwardly from this principle.111We know that natural number $n$ is formed at stage $n$. So the set112$\omega$ is formed at the first infinite stage. And $\omega$ itself113witnesses the Axiom of Infinity. 114115This, however, simply pushes us back to the question of how we might116justify \stagesinf. As with \stagessucc, it was not an explicit part117of the story we told about the cumulative-iterative hierarchy. But118more than that: nothing in the very idea of an iterative hierarchy, in119which sets are formed stage by stage, forces us to think that the120process involves an \emph{infinite} stage. It seems perfectly coherent121to think that the stages are ordered like the natural numbers. 122123This, however, gives rise to an obvious problem. In124\olref[sfr][infinite][dedekindsproof]{sec}, we considered125Dedekind's ``proof'' that there is a Dedekind infinite set (of126thoughts). This may not have struck you as very satisfying. But if127\stagesinf{} is not ``forced upon us'' by the iterative conception of128set (or by ``the laws of thought''), then we are still left without an129intrinsic justification for the claim that there is a Dedekind130infinite set.131132There is much more to say here, of course. But hopefully you are now133at a point to start thinking about what it might \emph{take} to134justify an axiom (or principle). In what follows we will simply take135\stagesinf{} for granted.136137\end{document}
content/set-theory/z/milestone.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{milestone}6\olsection{$\Z^-$: a Milestone}78We will revisit \stagesinf{} in the next section. However, with the9Axiom of Infinity, we have reached an important milestone. We now have10all the axioms required for the theory $\Zminus$. In detail:1112\begin{defn}13The theory $\Zminus$ has these axioms: Extensionality, Union, Pairs, Powersets, Infinity, and all instances of the Separation scheme.14\end{defn}1516The name stands for \emph{Zermelo} set theory (\emph{minus} something17which we will come to later). Zermelo deserves the honour, since he18essentially formulated this theory in his19\citeyear{Zermelo1908Untersuchungen}.\footnote{For interesting20comments on the history and technicalities, see \citet[Appendix21A]{Potter2004}.}2223This theory is powerful enough to allow us to do an enormous amount of24mathematics. In particular, you \emph{should} look back through25\olref[sfr][][]{part}, and convince yourself that everything we did,26na\"ively, could be done more formally within~$\Zminus$. (Once you27have done that for a bit, you might want to skip ahead and read28\olref[sth][z][arbintersections]{sec}.) So, henceforth,29and without any further comment, we will take ourselves to be working30in $\Zminus$ (at least).3132\end{document}
content/set-theory/z/nat.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}4 5\olfileid{sth}{z}{nat}6\olsection{Selecting our Natural Numbers}78%Let us take it for granted that the informal principle \stagesinf{}9%can be justified. It is also, though, worth commenting on the10%particular axiom we extracted from that informal principle. 1112In \olref[infinity-again]{defnomega}, we explicitly defined13the expression ``natural numbers''. How should you understand this14stipulation? It is not a metaphysical claim, but just a decision to15\emph{treat} certain sets as the natural numbers. We touched upon reasons for thinking this in16\olref[sfr][rel][ref]{sec}, \olref[sfr][arith][ref]{sec} and17\olref[sfr][infinite][dedekindsproof]{sec}. But we can make these18reasons even more pointed.1920Our Axiom of Infinity follows \citet{VonNeumann1925}. But here is21another axiom, which we could have adopted instead:2223\begin{defish}24\emph{Zermelo's \citeyear{Zermelo1908Untersuchungen} Axiom of25Infinity.} There is a set $A$ such that $\emptyset \in A$ and26$(\forall x \in A)\{x\} \in A$. 27\end{defish}2829Had we used Zermelo's axiom, instead of our (von Neumann-inspired)30Axiom of Infinity, we would equally well have been given a Dedekind31infinite set, and so a Dedekind algebra. On Zermelo's approach, the32distinguished element of our algebra would again have been $\emptyset$33(our surrogate for $0$), but the injection would have been given by34the map $x \mapsto \{x\}$, rather than $x \mapsto x \cup \{x\}$. The35simplest upshot of this is that Zermelo treats $2$ as36$\{\{\emptyset\}\}$, whereas we (with von Neumann) treat $2$ as37$\{\emptyset, \{\emptyset\}\}$. 3839Why choose one axiom of Infinity rather than the other? The main40practical reason is that von Neumann's approach ``scales up'' to41handle transfinite numbers rather well. We will explore this from42\olref[ordinals][]{chap} onwards. However, from the simple43perspective of \emph{doing arithmetic}, both approaches would do44equally well. So if someone tells you that the natural numbers45\emph{are} sets, the obvious question is: \emph{Which sets are they?} 4647This precise question was made famous by \citet{Benacerraf1965}. But48it is worth emphasising that it is just the most famous example of a49phenomenon that we have encountered many times already. The basic50point is this. Set theory gives us a way to \emph{simulate} a bunch of51``intuitive'' kinds of entities: the reals, rationals, integers, and52naturals, yes; but also ordered pairs, functions, and relations.53However, set theory never provides us with a \emph{unique} choice of54simulation. There are \emph{always} alternatives55which---straightforwardly---would have served us just as well. 5657\end{document}
content/set-theory/z/arbintersections.tex
1\documentclass[../../../include/open-logic-section]{subfiles}23\begin{document}45\olfileid{sth}{z}{arbintersections}6\olsection{Appendix: Closure, Comprehension, and Intersection}78In \olref[sth][z][milestone]{sec}, we suggested that you should look9back through the na\"ive work of \olref[sfr][][]{part} and check that10it can be carried out in~$\Zminus$. If you followed that advice,11\emph{one} point might have tripped you up: the use of12\emph{intersection} in Dedekind's treatment of \emph{closures}. 1314Recall from \olref[sfr][infinite][dedekind]{Closure} that15\begin{align*}16 \closureofunder{f}{o} & = \bigcap\Setabs{X}{o \in X \text{ and $X$ is17$f$-closed}}.18\intertext{The general shape of this is a definition of the form:}19 C & = \bigcap\Setabs{X}{\phi(X)}.20\end{align*}21But this should ring alarm bells: since Na\"ive Comprehension fails,22there is no guarantee that $\Setabs{X}{\phi(X)}$ exists. It looks23dangerously, then, like such definitions are \emph{cheating}. 2425Fortunately, they are not cheating; or rather, if they \emph{are}26cheating as they stand, then we can engage in some honest toil to27render them kosher. That honest toil was foreshadowed in28\olref[sth][z][sep]{prop:intersectionsexist}, when we explained why29$\bigcap A$ exists for any $A \neq \emptyset$. But we will spell it out30explicitly.3132Given Extensionality, if we attempt to define $C$ as33$\bigcap\Setabs{X}{\phi(X)}$, all we are really asking is for an34object $C$ which obeys the following:35\begin{equation}\ollabel{bicondelimarbintersection}36 \forall x(x \in C \liff \forall X(\phi(X) \lif x \in X))\tag{*}37\end{equation}38Now, suppose there is \emph{some} set, $S$, such that $\phi(S)$. Then39to deliver \olref{bicondelimarbintersection}, we can simply define $C$40using \emph{Separation}, as follows:41\[42 C = \Setabs{x \in S}{\forall X(\phi(X) \lif x \in X)}.43\]44We leave it as an exercise to check that this definition yields45\olref{bicondelimarbintersection}, as desired. 46% fix $x$. First suppose that $x \in C$, as (re)defined; then both $x47% \in S$ and $\forall X(\phi(X) \lif x \in X)$; but since48% $\phi(S)$, this reduces to the condition that $\forall X(\phi(X)49% \lif x \in X)$. Conversely, suppose $\forall X(\phi(X) \lif50% x \in X)$; then, since $\phi(S)$, we also have that $x \in S$; so51% $x \in c$, as (re)defined. 52And this general strategy will allow us to circumvent any apparent use53of Na\"ive Comprehension in defining intersections. In the particular54case which got us started on this line of thought, namely that of55$\closureofunder{f}{o}$, here is how that would work. We began the56proof of \olref[sfr][infinite][dedekind]{closureproperties} by noting57that $o \in \ran{f}\cup\{o\}$ and that $\ran{f} \cup \{o\}$ is58$f$-closed. So, we can define what we want thus:59\[60 \closureofunder{f}{o} = \Setabs{x \in \ran{f} \cup \{o\}}{(\forall X \ni o)(X \text{ is $f$-closed} \lif x \in X)}.61\]6263\end{document}