Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

November 8, 2024

Axiomatic Set Theory 8: Well Ordered Sets

Posted by Tom Leinster

Previously: Part 7. Next: Part 9.

By this point in the course, we’ve finished the delicate work of assembling all the customary set-theoretic apparatus from the axioms, and we’ve started proving major theorems. This week, we met well ordered sets and developed all the theory we’ll need. The main results were:

  • every family of well ordered sets has a least member — informally, “the well ordered sets are well ordered”;

  • the Hartogs theorem: for every set XX, there’s some well ordered set that doesn’t admit an injection into XX;

  • a very close relative of Zorn’s lemma that, nevertheless, doesn’t require the axiom of choice: for every ordered set XX and function φ\varphi assigning an upper bound to each chain in XX, there’s some chain CC such that φ(C)C\varphi(C) \in C.

I also included an optional chatty section on the use of transfinite recursion to strip the isolated points from any subset of \mathbb{R}. Am I right in understanding that this is what got Cantor started on set theory in the first place?

Diagram of an ordered set, showing a chain. I know, it's not *well* ordered

Posted at November 8, 2024 12:48 PM UTC

TrackBack URL for this Entry:   https://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/3578

6 Comments & 1 Trackback

Re: Axiomatic Set Theory 8: Well Ordered Sets

I’ve read someone jokingly say Cantor was studying convergence of Fourier series and did such a bad job he wound up inventing infinite ordinals. His focus on pointwise convergence of Fourier series looks a bit wrongheaded today: it’s good if you like really hard problems, but everything gets easier if you use L 1L^1 or L 2L^2 convergence.

From Nicholas Scoville’s Georg Cantor at the dawn of point-set topology:

Heine’s 1870 work showed that if a function is almost everywhere continuous and its trigonometric series converges uniformly, then the Fourier series is unique. As Dauben points out,

Requiring almost-everywhere continuity and uniform convergence, Heine’s theorem invited direct generalizations.

These generalizations would be taken up by Cantor. Important for our purposes is that Cantor developed a proof technique in his 1870 paper and modified it only slightly while weakening his hypotheses in the 1872 paper. More specifically, Cantor showed that, under certain hypotheses, the trigonometric representation of a function remains unique even when convergence or representation of the function is given up on certain infinite subsets of the open interval (0,2π).

Cantor defined the derived set PP' to be the set of limit points of a set PP \subseteq \mathbb{R}: in other words, PP minus its set of isolated points.

He then recursively defined the α\alphath derived set P (α)P^{(\alpha)} by

P (0)=P,P (α+1)=(P (α)) P^{(0)} = P, \quad P^{(\alpha + 1)} = (P^{(\alpha)})^'

and said PP \subseteq \mathbb{R} is a set of the α\alphath kind if P (α+1)P^{(\alpha + 1)} is empty.

It seems Cantor first showed this: if

lim N n=N Nc ne 2πnx=0 \lim_{N \to \infty} \sum_{n = -N}^N c_n e^{2\pi n x} = 0

pointwise, then c n=0c_n = 0 for all nn. This is a result about the uniqueness of Fourier series.

Then he got the same conclusion assuming the limit is zero except at a set of the first kind — i.e., except at a finite set of points.

Then he got the same conclusion assuming the limit is zero except at a set of the nnth kind for any nn \in \mathbb{N}.

But he didn’t want to stop there! So he defined ordinals, and for limit ordinals α\alpha he defined the α\alphath derived set by

P (α)= β<αP (β)P^{(\alpha)} = \bigcap_{\beta \lt \alpha} P^{(\beta)}

I want to know if Cantor succeeded in showing this: if

lim N n=N Nc ne 2πnx=0 \lim_{N \to \infty} \sum_{n = -N}^N c_n e^{2\pi n x} = 0

except at a set of the α\alphath kind for some countable ordinal α\alpha, does that imply c n=0c_n = 0 for all nn? That seems to be where his work was heading, but I don’t even know if it’s true!

Now I’m curious about how far Cantor got in this direction — he may have gotten distracted by studying ordinals and cardinals for their own sake.

According to S. M. Srivasta, Cantor did show that for every PP \subseteq \mathbb{R} there exists a countable ordinal α\alpha such that for all βα\beta \ge \alpha we have

P (β)=P (α) P^{(\beta)} = P^{(\alpha)}

So there’s no point in studying the α\alphath derived set of PP \subseteq \mathbb{R} for uncountable ordinals α\alpha.

Srivasta shows that for every countable closed set, P (α)=P^{(\alpha)} = \emptyset for some countable ordinal α\alpha. Srivasta also says it “seems likely” that Cantor realized that if

lim N n=N Nc ne 2πnx=0 \lim_{N \to \infty} \sum_{n = -N}^N c_n e^{2\pi n x} = 0

except for xx in a countable closed set, then c n=0c_n = 0 for all nn.

That’s all I could quickly find about this, except: there exist sets of Lebesgue measure zero PP \subseteq \mathbb{R} such that

lim N n=N Nc ne 2πnx=0 \lim_{N \to \infty} \sum_{n = -N}^N c_n e^{2\pi n x} = 0

except on PP, yet we don’t have c n=0c_n = 0 for all nn.

Posted by: John Baez on November 8, 2024 11:46 PM | Permalink | Reply to this

Re: Axiomatic Set Theory 8: Well Ordered Sets

Cantor didn’t need to think about ordinals as the index for the derived sets, to end up with well-ordered sets. Tom gave the example of the set {1n}{0}\{\frac{1}{n}\}\cup\{0\} as being of the first kind, but perhaps more fruitfully, one can think of this as {nn+1}{1}\{\frac{n}{n+1}\}\cup\{1\}, which is order isomorphic to ω+1\omega+1. In light of Dedekind’s mentioning the monotone convergence theorem in his paper Continuity and irrational numbers (where he defines the reals using cuts), as a motivation, one might imagine that Cantor, in the same culture, might be thinking of increasing sequences, not decreasing sequences.

And so the first example of a set of the second kind that comes to my mind is in fact something order-isomorphic to ω 2+1\omega^2+1. One option is to stick in a scaled copy of the above increasing sequence converging to each individual term nn+1\frac{n}{n+1}, something like {nn+11(m+1)(n+1)}{1}\left\{\frac{n}{n+1} - \frac{1}{(m+1)(n+1)}\right\}\cup \{1\}.

Even if Cantor only considered sets of finite kind, to give examples of these sets he could have considered countable well-ordered subsets of the reals of the form ω na n++ωa 1+a 0\omega^{n}\cdot a_n + \cdots + \omega\cdot a_1 + a_0 for natural numbers a n,,a 0a_n,\ldots,a_0.

Posted by: David Roberts on November 9, 2024 11:14 PM | Permalink | Reply to this

Re: Axiomatic Set Theory 8: Well Ordered Sets

There is a book of Kechris and Louveau from the late 1980s, “Descriptive set theory and the structure of sets of uniqueness”, but I confess I have not looked into this. There are also some lecture notes by Kechris which include some discussion of what Cantor did or didn’t prove; see the Remark after the proof of Theorem 4.2.

Posted by: Yemon Choi on November 14, 2024 10:45 PM | Permalink | Reply to this

Re: Axiomatic Set Theory 8: Well Ordered Sets

Thanks, everyone, for all this information and the references.

Posted by: Tom Leinster on November 15, 2024 9:37 AM | Permalink | Reply to this

Re: Axiomatic Set Theory 8: Well Ordered Sets

I think there might be a typo near the start of the proof of Proposition 8.4.3 - \preccurlyeq rather than \leq. It’s invoking Hartogs, which only supplies the latter. And later, you invoke the property that any function WXW\to X can’t be injective, but how it was phrased was slightly confusing, because it seems like you were using the fact the function gg was order-preserving (esp. paired with the typo), when it’s not specific to gg.

Posted by: David Roberts on November 9, 2024 10:57 PM | Permalink | Reply to this

Re: Axiomatic Set Theory 8: Well Ordered Sets

Oh, yes. Thank you, David.

I’ll fix that, but for boring reasons to do with versions, the fix won’t show up until I post next week’s update.

Posted by: Tom Leinster on November 9, 2024 11:16 PM | Permalink | Reply to this
Read the post Axiomatic Set Theory 9: The Axiom of Choice
Weblog: The n-Category Café
Excerpt: The penultimate week of this axiomatic set theory course, based on Lawvere's Elementary Theory of the Category of Sets.
Tracked: November 15, 2024 2:27 PM

Post a New Comment