Set Theory

Cardinal Arithmetic

Reading preferences

Optional display controls need JavaScript. All reading content and navigation work without it.

Source file content/set-theory/card-arithmetic/card-arithmetic.tex

Source file content/set-theory/card-arithmetic/opps.tex

Defining the Basic Operations

Since we do not need to keep track of order, cardinal arithmetic is rather easier to define than ordinal arithmetic. We will define addition, multiplication, and exponentiation simultaneously.

Definition one in this chapter

When a\cardfont{a}source and b\cardfont{b}source are cardinals:

ab=|ab|ab=|a×b|ab=|ba|\cardfont{a} \cardplus \cardfont{b} &\defis \card{\cardfont{a} \disjointsum \cardfont{b}}\\ \cardfont{a} \cardtimes \cardfont{b} &\defis \card{\cardfont{a} \times \cardfont{b}}\\ \cardexpo{\cardfont{a}}{\cardfont{b}} &\defis \card{\funfromto{\cardfont{b}}{\cardfont{a}}}source

where XY={f:f is a function XY}\funfromto{X}{Y} = \Setabs{f}{f \text{ is a function } X \to Y}source. (It is easy to show that XY\funfromto{X}{Y}source exists for any sets XXsource and YYsource; we leave this as an exercise.)

Exercise one in this chapter

Prove in Z\Zminussource that XY\funfromto{X}{Y}source exists for any sets XXsource and YYsource. Working in ZF\ZFsource, compute rank(XY)\setrank{\funfromto{X}{Y}}source from rank(X)\setrank{X}source and rank(Y)\setrank{Y}source, in the manner of lemma four in chapter “Ordinal Arithmetic”.

It might help to explain this definition. Concerning addition: this uses the notion of disjoint sum, \disjointsumsource, as defined in definition one in chapter “Ordinal Arithmetic”; and it is easy to see that this definition gives the right verdict for finite cases. Concerning multiplication: proposition in chapter “Sets” tells us that if AAsource has nnsource members and BBsource has mmsource members then A×BA \times Bsource has n·mn \cdot msource members, so our definition simply generalises the idea to transfinite multiplication. Exponentiation is similar: we are simply generalising the thought from the finite to the transfinite. Indeed, in certain ways, transfinite cardinal arithmetic looks much more like “ordinary” arithmetic than does transfinite ordinal arithmetic:

Proposition one in this chapter

\cardplussource and \cardtimessource are commutative and associative.

Proof

For commutativity, by the lemma on Cardinals Behave Right it suffices to observe that (ab)(ba)\cardeq{(\cardfont{a} \disjointsum \cardfont{b})}{(\cardfont{b} \disjointsum \cardfont{a})}source and (a×b)(b×a)\cardeq{(\cardfont{a} \times \cardfont{b})}{(\cardfont{b} \times \cardfont{a})}source. We leave associativity as an exercise.

Exercise two in this chapter

Prove that \cardplussource and \cardtimessource are associative.

Proposition two in this chapter

AAsource is infinite iff |A|1=1|A|=|A|\card{A} \cardplus 1 = 1 \cardplus \card{A} = \card{A}source.

Proof

As in theorem one in chapter “Cardinals”, from lemma five in chapter “Ordinal Arithmetic” and the lemma on Cardinals Behave Right.

This explains why we need to use different symbols for ordinal versus cardinal addition/multiplication: these are genuinely different operations. This next pair of results shows that ordinal versus cardinal exponentiation are also different operations. (Recall that definition of the natural numbers and omega in chapter “Steps towards Z” entails that 2={0,1}2 = \{0, 1\}source):

Lemma one in this chapter

|(A)|=2|A|\card{\Pow{A}} = \cardexpo{2}{\card{A}}source, for any AAsource.

Proof

For each subset BAB \subseteq Asource, let χBA2\chi_B \in \funfromto{A}{2}source be given by:

χB(x)={1if xB0otherwise.\chi_{B}(x) &\defis \begin{cases} 1 & \text{if }x\in B\\ 0 & \text{otherwise.} \end{cases}source

Now let f(B)=χBf(B) = \chi_Bsource; this defines a bijection f:(A)A2f \colon \Pow{A} \to \funfromto{A}{2}source. So (A)A2\cardeq{\Pow{A}}{\funfromto{A}{2}}source. Hence (A)|A|2\cardeq{\Pow{A}}{\funfromto{\card{A}}{2}}source, so that |(A)|=||A|2|=2|A|\card{\Pow{A}} = \card{\funfromto{\card{A}}{2}} = 2^{\card{A}}source.

This snappy proof essentially subsumes the discussion of section “Reduction” in chapter “The Size of Sets”. There, we showed how to “reduce” the uncountability of (ω)\Pow{\omega}source to the uncountability of the set of infinite binary strings, Bω\Bin^\omegasource. In effect, Bω\Bin^{\omega}source is just ω2\funfromto{\omega}{2}source; and the preceding proof showed that the reasoning we went through in section “Reduction” in chapter “The Size of Sets” will go through using any set AAsource in place of ω\omegasource. The result also yields a quick fact about cardinal exponentiation:

Corollary one in this chapter

a<2a\cardfont{a} < \cardexpo{2}{\cardfont{a}}source for any cardinal a\cardfont{a}source.

Proof

From Cantor's Theorem (the theorem on cantor) and the lemma on Size Powersettwo Exp.

noindent So ω<2ω\omega < \cardexpo{2}{\omega}source. But note: this is a result about cardinal exponentiation. It should be contrasted with ordinal exponentiation, since in the latter case ω=2(ω)\omega = \ordexpo{2}{\omega}source (see section “Ordinal Exponentiation” in chapter “Ordinal Arithmetic”).

Whilst we are on the topic of cardinal exponentiation, we can also be a bit more precise about the “way” in which \Realsource is non-enumerable.

Theorem one in this chapter

||=2ω\card{\Real} = \cardexpo{2}{\omega}source

Proof

[Proof skeleton] There are plenty of ways to prove this. The most straightforward is to argue that (ω)\cardle{\Pow{\omega}}{\Real}source and (ω)\cardle{\Real}{\Pow{\omega}}source, and then use Schröder-Bernstein to infer that (ω)\cardeq{\Real}{\Pow{\omega}}source, and the lemma on Size Powersettwo Exp to infer that ||=2ω\card{\Real} = \cardexpo{2}{\omega}source. We leave it as an (illuminating) exercise to define injections f:(ω)f \colon \Pow{\omega} \to \Realsource and g:(ω)g \colon \Real \to \Pow{\omega}source.

Exercise three in this chapter

Complete the proof of theorem one in chapter “Cardinal Arithmetic”, by showing that (ω)\cardle{\Pow{\omega}}{\Real}source and (ω)\cardle{\Real}{\Pow{\omega}}source.

Source file content/set-theory/card-arithmetic/simp.tex

Simplifying Addition and Multiplication

It turns out that transfinite cardinal addition and multiplication is extremely easy. This follows from the fact that cardinals are (certain) ordinals, and so well-ordered, and so can be manipulated in a certain way. Showing this, though, is not so easy. To start, we need a tricksy definition:

Definition two in this chapter

We define a canonical ordering, \canonordsource, on pairs of ordinals, by stipulating that α1,α2β1,β2\tuple{\alpha_1, \alpha_2} \canonord \tuple{\beta_1, \beta_2}source iff either:

  1. max(α1,α2)<max(β1,β2)\max(\alpha_1, \alpha_2) < \max(\beta_1, \beta_2)source; or

  2. max(α1,α2)=max(β1,β2)\max(\alpha_1, \alpha_2) = \max(\beta_1, \beta_2)source and α1<β1\alpha_1 < \beta_1source; or

  3. max(α1,α2)=max(β1,β2)\max(\alpha_1, \alpha_2) = \max(\beta_1, \beta_2)source and α1=β1\alpha_1 = \beta_1source and α2<β2\alpha_2 < \beta_2source

Lemma two in this chapter

α×α,\tuple{\alpha \times \alpha, \canonord}source is a well-order, for any ordinal α\alphasource.

Proof

Evidently \canonordsource is connected on α×α\alpha \times \alphasource. For suppose that neither α1,α2\tuple{\alpha_1, \alpha_2}source nor β1,β2\tuple{\beta_1, \beta_2}source is \canonordsource-less than the other. Then max(α1,α2)=max(β1,β2)\max(\alpha_1, \alpha_2) = \max(\beta_1, \beta_2)source and α1=β1\alpha_1 = \beta_1source and α2=β2\alpha_2 = \beta_2source, so that α1,α2=β1,β2\tuple{\alpha_1, \alpha_2} = \tuple{\beta_1, \beta_2}source.

To show well-ordering, let Xα×αX \subseteq \alpha\times\alphasource be non-empty. Since α\alphasource is an ordinal, some δ\deltasource is the least member of {max(γ1,γ2):γ1,γ2X}\Setabs{\max(\gamma_1, \gamma_2)}{\tuple{\gamma_1, \gamma_2} \in X}source. Now discard all pairs from {γ1,γ2X:max(γ1,γ2)=δ}\Setabs{\tuple{\gamma_1,\gamma_2} \in X}{\max(\gamma_1, \gamma_2) = \delta}source except those with least first coordinate; from among these, the pair with least second coordinate is the \canonordsource-least element of XXsource.

noindent Now for a teensy, simple observation:

Proposition three in this chapter

If αβ\cardeq{\alpha}{\beta}source, then α×αβ×β\cardeq{\alpha \times \alpha}{\beta \times \beta}source.

Proof

Just let f:αβf \colon \alpha \to \betasource induce γ1,γ2f(γ1),f(γ2)\tuple{\gamma_1, \gamma_2} \mapsto \tuple{f(\gamma_1), f(\gamma_2)}source.

noindent And now we will put all this to work, in proving a crucial lemma:

Lemma three in this chapter

αα×α\cardeq{\alpha}{\alpha \times \alpha}source, for any infinite ordinal α\alphasource

Proof

For reductio, let α\alphasource be the least infinite ordinal for which this is false. proposition “Enumerability of pairs of natural numbers” in chapter “The Size of Sets” shows that ωω×ω\cardeq{\omega}{\omega\times\omega}source, so ωα\omega \in \alphasource. Moreover, α\alphasource is a cardinal: suppose otherwise, for reductio; then |α|α\card{\alpha} \in \alphasource, so that |α||α|×|α|\cardeq{\card{\alpha}}{\card{\alpha} \times \card{\alpha}}source, by hypothesis; and |α|α\cardeq{\card{\alpha}}{\alpha}source by definition; so that αα×α\cardeq{\alpha}{\alpha\times\alpha}source by proposition three in chapter “Cardinal Arithmetic”.

Now, for each γ1,γ2α×α\tuple{\gamma_1, \gamma_2} \in \alpha \times \alphasource, consider the segment:

Seg(γ1,γ2)={δ1,δ2α×α:δ1,δ2γ1,γ2}\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}}source

Letting γ=max(γ1,γ2)\gamma = \max(\gamma_1, \gamma_2)source, note that γ1,γ2γ+1,γ+1\tuple{\gamma_1, \gamma_2} \canonord \tuple{\gamma+1, \gamma + 1}source. So, when γ\gammasource is infinite, observe:

Seg(γ1,γ2)((γ+1)·(γ+1))(γ·γ), by lemma five in chapter Ordinal Arithmetic and proposition three in chapter Cardinal Arithmeticγ, by the induction hypothesisα, since α is a cardinal\text{Seg}(\gamma_1, \gamma_2) & \precsim ((\gamma \ordplus 1)\ordtimes (\gamma \ordplus 1))\\ &\approx (\gamma \ordtimes \gamma) \text{, by \olref[ord-arithmetic][using-addition]{ordinfinitycharacter} and \olref{simplecardproduct}}\\ &\approx \gamma \text{, by the induction hypothesis}\\ & \prec \alpha\text{, since $\alpha$ is a cardinal}source

So ord(α×α,)α\ordtype{\alpha\times \alpha, \canonord} \leq \alphasource, and hence α×αα\cardle{\alpha \times \alpha}{\alpha}source. Since of course αα×α\cardle{\alpha}{\alpha \times \alpha}source, the result follows by Schröder-Bernstein.

Finally, we get to our simplifying result:

Theorem two in this chapter

If a,b\cardfont{a}, \cardfont{b}source are infinite cardinals, then:

ab=ab=max(a,b).\cardfont{a} \cardtimes \cardfont{b} = \cardfont{a} \cardplus \cardfont{b} = \text{max}(\cardfont{a}, \cardfont{b}).source

Proof

Without loss of generality, suppose a=max(a,b)\cardfont{a} = \max(\cardfont{a}, \cardfont{b})source. Then invoking lemma three in chapter “Cardinal Arithmetic”, aa=aabaaaa\cardfont{a}\cardtimes\cardfont{a} = \cardfont{a} \leq \cardfont{a} \cardplus \cardfont{b} \leq \cardfont{a} \cardplus \cardfont{a} \leq \cardfont{a} \cardtimes \cardfont{a}source.

noindent Similarly, if a\cardfont{a}source is infinite, an a\cardfont{a}source-sized union of a\leq\cardfont{a}source-sized sets has size a\leq\cardfont{a}source:

Proposition four in this chapter

Let a\cardfont{a}source be an infinite cardinal. For each ordinal βa\beta \in \cardfont{a}source, let XβX_\betasource be a set with |Xβ|a\card{X_\beta} \leq \cardfont{a}source. Then |βaXβ|a\card{\bigcup_{\beta \in \cardfont{a}} X_\beta} \leq \cardfont{a}source.

Proof

For each βa\beta \in \cardfont{a}source, fix an injection fβ:Xβaf_\beta \colon X_\beta \to \cardfont{a}source.Footnote: How are these “fixed”? See section “Countable Choice” in chapter “Choice”. Define an injection g:βaXβa×ag \colon \bigcup_{\beta \in \cardfont{a}} X_\beta \to \cardfont{a} \times \cardfont{a}source by g(v)=β,fβ(v)g(v) = \tuple{\beta, f_\beta(v)}source, where vXβv \in X_\betasource and vXγv \notin X_\gammasource for any γβ\gamma \in \betasource. Now βaXβa×aa\bigcup_{\beta \in \cardfont{a}} X_\beta \preceq \cardfont{a} \times \cardfont{a} \approx \cardfont{a}source by theorem two in chapter “Cardinal Arithmetic”.

Source file content/set-theory/card-arithmetic/expotough.tex

Some Simplification with Cardinal Exponentiation

Whilst defining \canonordsource was a little involved, the upshot is a useful result concerning cardinal addition and multiplication, theorem two in chapter “Cardinal Arithmetic”. Transfinite exponentiation, however, cannot be simplified so straightforwardly. To explain why, we start with a result which extends a familiar pattern from the finitary case (though its proof is at a high level of abstraction):

Proposition five in this chapter

abc=abac\cardexpo{\cardfont{a}}{\cardfont{b} \cardplus \cardfont{c}} = \cardexpo{\cardfont{a}}{\cardfont{b}} \cardtimes \cardexpo{\cardfont{a}}{\cardfont{c}}source and (ab)c=abc\cardexpo{(\cardexpo{\cardfont{a}}{\cardfont{b}})}{\cardfont{c}} = \cardexpo{\cardfont{a}}{\cardfont{b} \cardtimes \cardfont{c}}source, for any cardinals a,b,c\cardfont{a}, \cardfont{b}, \cardfont{c}source.

Proof

For the first claim, consider a function f:(bc)af \colon (\cardfont{b}\disjointsum\cardfont{c}) \to \cardfont{a}source. Now “split this”, by defining fb(β)=f(β,0)f_\cardfont{b}(\beta) = f(\beta, 0)source for each βb\beta \in \cardfont{b}source, and fc(γ)=f(γ,1)f_\cardfont{c}(\gamma) = f(\gamma, 1)source for each γc\gamma \in \cardfont{c}source. The map f(fb×fc)f \mapsto (f_{\cardfont{b}} \times f_\cardfont{c})source is a bijection bca(ba×ca)\funfromto{\cardfont{b} \disjointsum \cardfont{c}}{\cardfont{a}} \to (\funfromto{\cardfont{b}}{\cardfont{a}} \times \funfromto{\cardfont{c}}{\cardfont{a}})source.

For the second claim, consider a function f:c(ba)f \colon \cardfont{c} \to (\funfromto{\cardfont{b}}{\cardfont{a}})source; so for each γc\gamma \in \cardfont{c}source we have some function f(γ):baf(\gamma) \colon \cardfont{b} \to \cardfont{a}source. Now define f*(β,γ)=(f(γ))(β)f^*(\beta, \gamma) = (f(\gamma))(\beta)source for each β,γb×c\tuple{\beta, \gamma} \in \cardfont{b} \times \cardfont{c}source. The map ff*f \mapsto f^*source is a bijection c(ba)bca\funfromto{\cardfont{c}}{(\funfromto{\cardfont{b}}{\cardfont{a}})} \to \funfromto{\cardfont{b} \cardtimes \cardfont{c}}{\cardfont{a}}source.

Now, what we would like is an easy way to compute ab\cardexpo{\cardfont{a}}{\cardfont{b}}source when we are dealing with infinite cardinals. Here is a nice step in this direction:

Proposition six in this chapter

If 2ab2 \leq \cardfont{a} \leq \cardfont{b}source and b\cardfont{b}source is infinite, then ab=2b\cardexpo{\cardfont{a}}{\cardfont{b}} = \cardexpo{2}{\cardfont{b}}source

Proof

2bab, as 2a(2a)b, by the lemma on Size Powersettwo Exp=2ab, by proposition five in chapter Cardinal Arithmetic=2b, by theorem two in chapter Cardinal Arithmetic\cardexpo{2}{\cardfont{b}} &\leq \cardexpo{\cardfont{a}}{\cardfont{b}}\text{, as $2 \leq \cardfont{a}$}\\ &\leq \cardexpo{(2^\cardfont{a})}{\cardfont{b}} \text{, by \olref[opps]{lem:SizePowerset2Exp}}\\ &= \cardexpo{2}{\cardfont{a} \cardtimes \cardfont{b}} \text{, by \olref{simplecardexpo}} \\ &= \cardexpo{2}{\cardfont{b}} \text{, by \olref[simp]{cardplustimesmax}}source

We should not really expect to be able to simplify this any further, since b<2b\cardfont{b} < \cardexpo{2}{\cardfont{b}}source by the lemma on Size Powersettwo Exp. However, this does not tell us what to say about ab\cardexpo{\cardfont{a}}{\cardfont{b}}source when b<a\cardfont{b} < \cardfont{a}source. Of course, if b\cardfont{b}source is finite, we know what to do.

Proposition seven in this chapter

If a\cardfont{a}source is infinite and nωn \in \omegasource then an=a\cardexpo{\cardfont{a}}{n} = \cardfont{a}source

Proof

an=aaa=a\cardexpo{\cardfont{a}}{n} = \cardfont{a} \cardtimes \cardfont{a} \cardtimes \ldots \cardtimes \cardfont{a} = \cardfont{a}source, by theorem two in chapter “Cardinal Arithmetic”.

noindent Additionally, in some other cases, we can control the size of ab\cardexpo{\cardfont{a}}{\cardfont{b}}source:

Proposition eight in this chapter

If 2b<a2b2 \leq \cardfont{b} < \cardfont{a} \leq \cardexpo{2}{\cardfont{b}}source and b\cardfont{b}source is infinite, then ab=2b\cardexpo{\cardfont{a}}{\cardfont{b}} = \cardexpo{2}{\cardfont{b}}source

Proof

2bab(2b)b=2bb=2b\cardexpo{2}{\cardfont{b}}\leq \cardexpo{\cardfont{a}}{\cardfont{b}} \leq \cardexpo{(\cardexpo{2}{\cardfont{b}})}{\cardfont{b}} = \cardexpo{2}{\cardfont{b}\cardtimes\cardfont{b}} = \cardexpo{2}{\cardfont{b}}source, reasoning as in proposition six in chapter “Cardinal Arithmetic”.

noindent But, beyond this point, things become rather more subtle.

Source file content/set-theory/card-arithmetic/ch.tex

The Continuum Hypothesis

The previous result hints (correctly) that cardinal exponentiation would be quite easy, if infinite cardinals are guaranteed to “play straightforwardly” with powers of 22source, i.e., (by the lemma on Size Powersettwo Exp) with taking powersets. But we cannot assume that infinite cardinals do play straightforwardly powersets.

To start unpacking this, we introduce some nice notation.

Definition three in this chapter

Where a\cardsucc{\cardfont{a}}source is the least cardinal strictly greater than a\cardfont{a}source, we define two infinite sequences:

0=ω0=ωα+1=(α)α+1=2αα=β<αβα=β<αβwhen α is a limit ordinal.\aleph_{0} &\defis \omega & \beth_{0} &\defis \omega\\ \aleph_{\alpha \ordplus 1} &\defis \cardsucc{(\aleph_{\alpha})} & \beth_{\alpha+1} &\defis \cardexpo{2}{\beth_{\alpha}}\\ \aleph_{\alpha} &\defis \bigcup_{\beta< \alpha} \aleph_{\beta} & \beth_{\alpha} &\defis \bigcup_{\beta < \alpha}\beth_{\beta} & \text{when $\alpha$ is a limit ordinal}.source

The definition of a\cardsucc{\cardfont{a}}source is in order, since the lemma on No Largest Cardinal tells us that, for each cardinal a\cardfont{a}source, there is some cardinal greater than a\cardfont{a}source, and Transfinite Induction guarantees that there is a least cardinal greater than a\cardfont{a}source. The rest of the definition of a\cardfont{a}source is provided by transfinite recursion.

Cantor introduced this “\alephsource” notation; this is aleph, the first letter in the Hebrew alphabet and the first letter in the Hebrew word for “infinite”. Peirce introduced the “\bethsource” notation; this is beth, which is the second letter in the Hebrew alphabet.Footnote: Peirce used this notation in a letter to Cantor of December 1900. Unfortunately, Peirce also gave a bad argument there that α\beth_\alphasource does not exist for αω\alpha \geq \omegasource. Now, these notations provide us with infinite cardinals.

Proposition nine in this chapter

α\aleph_\alphasource and α\beth_\alphasource are cardinals, for every ordinal α\alphasource.

Proof

Both results hold by a simple transfinite induction. 0=0=ω\aleph_0 = \beth_0 = \omegasource is a cardinal by corollary two in chapter “Cardinals”. Assuming α\aleph_\alphasource and α\beth_\alphasource are both cardinals, α+1\aleph_{\alpha+1}source and α+1\beth_{\alpha+1}source are explicitly defined as cardinals. And the union of a set of cardinals is a cardinal, by proposition three in chapter “Cardinals”.

noindent Moreover, every infinite cardinal is an \alephsource:

Proposition ten in this chapter

If a\cardfont{a}source is an infinite cardinal, then a=γ\cardfont{a} = \aleph_\gammasource for some unique γ\gammasource.

Proof

By transfinite induction on cardinals. For induction, suppose that if b<a\cardfont{b} < \cardfont{a}source then b=γb\cardfont{b} = \aleph_{\gamma_\cardfont{b}}source. If a=b\cardfont{a} = \cardsucc{\cardfont{b}}source for some b\cardfont{b}source, then a=(γb)=γb+1\cardfont{a} = \cardsucc{(\aleph_{\gamma_\cardfont{b}})}= \aleph_{\gamma_\cardfont{b}+1}source. If a\cardfont{a}source is not the successor of any cardinal, then since cardinals are ordinals a=b<ab=b<aγb\cardfont{a} = \bigcup_{\cardfont{b} < \cardfont{a}} \cardfont{b} = \bigcup_{\cardfont{b} < \cardfont{a}}{\aleph_{\gamma_\cardfont{b}}}source, so a=γ\cardfont{a} = \aleph_\gammasource where γ=b<aγb\gamma = \bigcup_{\cardfont{b} < \cardfont{a}}\gamma_\cardfont{b}source.

Since every infinite cardinal is an \alephsource, this prompts us to ask: is every infinite cardinal a \bethsource? Certainly if that were the case, then the infinite cardinals would “play straightforwardly” with the operation of taking powersets. Indeed, we would have the following:

Definition four in this chapter

Generalized Continuum Hypothesis (GCH). α=α\aleph_\alpha = \beth_\alphasource, for all α\alphasource.

Moreover, if GCH held, then we could make some considerable simplifications with cardinal exponentiation. In particular, we could show that when b<a\cardfont{b} < \cardfont{a}source, the value of ab\cardexpo{\cardfont{a}}{\cardfont{b}}source is trapped by aaba\cardfont{a}\leq \cardexpo{\cardfont{a}}{\cardfont{b}} \leq \cardsucc{\cardfont{a}}source. We could then go on to give precise conditions which determine which of the two possibilities obtains (i.e., whether a=ab\cardfont{a} = \cardexpo{\cardfont{a}}{\cardfont{b}}source or ab=a\cardexpo{\cardfont{a}}{\cardfont{b}} = \cardsucc{\cardfont{a}}source).Footnote: The condition is dictated by cofinality.

But GCH is a hypothesis, not a theorem. In fact, Kurt Gödel (1938) proved that if ZFC\ZFCsource is consistent, then so is ZFC+GCH\ZFC + \text{GCH}source. But it later turned out that we can equally add ¬\lnotsourceGCH to ZFC\ZFCsource. Indeed, consider the simplest non-trivial instance of GCH, namely:

Definition five in this chapter

Continuum Hypothesis (CH). 1=1\aleph_1 = \beth_1source.

Paul J. Cohen (1963) proved that if ZFC\ZFCsource is consistent then so is ZFC+¬CH\ZFC + \lnot\text{CH}source. So the Continuum Hypothesis is independent from ZFC\ZFCsource.

The Continuum Hypothesis is so-called, since “the continuum” is another name for the real line, \Realsource. theorem one in chapter “Cardinal Arithmetic” tells us that ||=1\card{\Real} = \beth_1source. So the Continuum Hypothesis states that there is no cardinal between the cardinality of the natural numbers, 0=0\aleph_0 = \beth_0source, and the cardinality of the continuum, 1\beth_1source.

Given the independence of (G)CH from ZFC\ZFCsource, what should say about their truth? Well, there is much to say. Indeed, and much fertile recent work in set theory has been directed at investigating these issues. But two very quick points are certainly worth emphasising.

First: it does not immediately follow from these formal independence results that either GCH or CH is indeterminate in truth value. After all, maybe we just need to add more axioms, which strike us as natural, and which will settle the question one way or another. Gödel himself suggested that this was the right response.

Second: the independence of CH from ZFC\ZFCsource is certainly striking, but it is certainly not incredible (in the literal sense). The point is simply that, for all ZFC\ZFCsource tells us, moving from cardinals to their successors may involve a less blunt tool than simply taking powersets.

With those two observations made, if you want to know more, you will now have to turn to the various philosophers and mathematicians with horses in the race.Footnote: Though you might want to start by reading Michael Potter (2004), §15.6.

Source file content/set-theory/card-arithmetic/fix.tex

\alephsource-Fixed Points

In chapter “Stages and Ranks”, we suggested that Replacement stands in need of justification, because it forces the hierarchy to be rather tall. Having done some cardinal arithmetic, we can give a little illustration of the height of the hierarchy.

Evidently 0<00 < \aleph_0source, and 1<11 < \aleph_1source, and 2<22 < \aleph_2sourceldots and, indeed, the difference in size only gets bigger with every step. So it is tempting to conjecture that κ<κ\kappa< \aleph_\kappasource for every ordinal κ\kappasource.

But this conjecture is false, given ZFC\ZFCsource. In fact, we can prove that there are \alephsource-fixed-points, i.e., cardinals κ\kappasource such that κ=κ\kappa=\aleph_\kappasource.

Proposition eleven in this chapter

There is an \alephsource-fixed-point.

Proof

Using recursion, define:

κ0=0κn+1=κnκ=n<ωκn\kappa_0 &= 0\\ \kappa_{n+1} &= \aleph_{\kappa_n}\\ \kappa&= \bigcup_{n < \omega}\kappa_nsource

Now κ\kappasource is a cardinal by proposition three in chapter “Cardinals”. But now:

κ=n<ωκn+1=n<ωκn=α<κα=κ\kappa= \bigcup_{n < \omega} \kappa_{n+1} = \bigcup_{n < \omega}\aleph_{\kappa_n} = \bigcup_{\alpha < \kappa}\aleph_\alpha = \aleph_\kappasource

Boolos once wrote an article about exactly the \alephsource-fixed-point we just constructed. After noting the existence of κ\kappasource, at the start of his article, he said:

[κ\kappasource is] a pretty big number, by the lights of those with no previous exposure to set theory, so big, it seems to me, that it calls into question the truth of any theory, one of whose assertions is the claim that there are at least κ\kappasource objects. (George Boolos, 2000, p. 257)

And he ultimately concluded his paper by asking:

[do] we suspect that, however it may have been at the beginning of the story, by the time we have come thus far the wheels are spinning and we are no longer listening to a description of anything that is the case? (George Boolos, 2000, p. 268)

If we have, indeed, outrun “anything that is the case”, then we must point the finger of blame directly at Replacement. For it is this axiom which allows our proof to work. In which case, one assumes, Boolos would need to revisit the claim he made, a few decades earlier, that Replacement has “no undesirable” consequences (see section “Extrinsic Considerations about Replacement” in chapter “Replacement”).

But is the existence of κ\kappasource so bad? It might help, here, to consider Russell's Tristram Shandy paradox. Tristram Shandy documents his life in his diary, but it takes him a year to record a single day. With every passing year, Tristram falls further and further behind: after one year, he has recorded only one day, and has lived 364 days unrecorded days; after two years, he has only recorded two days, and has lived 728 unrecorded days; after three years, he has only recorded three days, and lived 1092 unrecorded days dotsFootnote: Forgetting about leap years. Still, if Tristram is immortal, Tristram will manage to record every day, for he will record the nnsourceth day on the nnsourceth year of his life. And so, “at the end of time”, Tristram will have a complete diary.

Now: why is this so different from the thought that α\alphasource is smaller than α\aleph_\alphasource---and indeed, increasingly, desperately smaller---up until κ\kappasource, at which point, we catch up, and κ=κ\kappa = \aleph_\kappasource?

Setting that aside, and assuming we accept ZFC\ZFCsource, let's close with a little more fun concerning fixed-point constructions. The next three results establish, intuitively, that there is a (non-trivial) point at which the hierarchy is as wide as it is tall:

Proposition twelve in this chapter

There is a \bethsource-fixed-point, i.e., a κ\kappasource such that κ=κ\kappa= \beth_\kappasource.

Proof

As in proposition eleven in chapter “Cardinal Arithmetic”, using “\bethsource” in place of “\alephsource”.

Proposition thirteen in this chapter

|Vω+α|=α\card{V_{\omega+\alpha}} = \beth_{\alpha}source. If ω·ωα\omega \ordtimes \omega \leq \alphasource, then |Vα|=α\card{V_\alpha} = \beth_\alphasource.

Proof

The first claim holds by a simple transfinite induction. The second claim follows, since if ω·ωα\omega \ordtimes \omega \leq \alphasource then ω+α=α\omega + \alpha = \alphasource. To establish this, we use facts about ordinal arithmetic from chapter “Ordinal Arithmetic”. First note that ω·ω=ω·(1+ω)=(ω·1)+(ω·ω)=ω+(ω·ω)\omega \ordtimes \omega = \omega \ordtimes (1 \ordplus \omega) = (\omega \ordtimes 1) \ordplus (\omega\ordtimes\omega) = \omega \ordplus (\omega \ordtimes \omega)source. Now if ω·ωα\omega \ordtimes \omega \leq \alphasource, i.e., α=(ω·ω)+β\alpha = (\omega\ordtimes\omega) \ordplus \betasource for some β\betasource, then ω+α=ω+((ω·ω)+β)=(ω+(ω·ω))+β=(ω·ω)+β=α\omega \ordplus \alpha = \omega \ordplus ((\omega \ordtimes \omega) \ordplus \beta) = (\omega \ordplus (\omega \ordtimes \omega)) \ordplus \beta = (\omega \ordtimes \omega) \ordplus \beta = \alphasource.

Corollary two in this chapter

There is a κ\kappasource such that |Vκ|=κ\card{V_\kappa} = \kappasource.

Proof

Let κ\kappasource be a \bethsource-fixed point, as given by proposition twelve in chapter “Cardinal Arithmetic”. Clearly ω·ω<κ\omega \ordtimes \omega < \kappasource. So |Vκ|=κ=κ\card{V_\kappa} = \beth_\kappa= \kappasource by proposition thirteen in chapter “Cardinal Arithmetic”.

There are as many stages beneath VκV_\kappasource as there are elements of VκV_\kappasource. Intuitively, then, VκV_\kappasource is as wide as it is tall. This is very Tristram-Shandy-esque: we move from one stage to the next by taking powersets, thereby making our hierarchy much bigger with each step. But, “in the end”, i.e., at stage κ\kappasource, the hierarchy's width catches up with its height.

One might ask: How often does the hierarchy's width match its height? The answer is: As often as there are ordinals. But this needs a little explanation.

We define a term τ\tausource as follows. For any AAsource, let:

τ0(A)=|A|τn+1(A)=τn(A)τ(A)=n<ωτn(A)As in proposition twelve in chapter Cardinal Arithmetic, τ(A) is a -fixed point for any A, and trivially |A|<τ(A). So now consider this recursive definition:W0=0Wα+1=τ(Wα)Wα=β<αWβ, when α is a limit\tau_0(A) & \defis \card{A}\\ \tau_{n+1}(A) & \defis \beth_{\tau_n(A)}\\ \tau(A) & \defis \bigcup_{n < \omega}\tau_n(A) \intertext{As in \olref{bethfixed}, $\tau(A)$ is a $\beth$-fixed point for any $A$, and trivially $\card{A} < \tau(A)$. So now consider this recursive definition:} W_0 &\defis 0\\ W_{\alpha + 1} & \defis \tau(W_\alpha)\\ W_\alpha & \defis \bigcup_{\beta < \alpha} W_\beta \text{, when $\alpha$ is a limit}source

The construction is defined for all ordinals. Intuitively, then, WWsource is “an injection” from the ordinals to \bethsource-fixed points. And, exactly as before, VWαV_{W_\alpha}source is as wide as it is tall, for any α\alphasource.