Scribd
Upload
46 views3 pages

08 More Real Line

This document summarizes key definitions and results from a lecture on the real line. It introduces the definitions of isolated points, limit points, and perfect sets. It proves some lemmas, including that any nonempty perfect set is equivalent to the real numbers and any closed uncountable set contains a perfect subset. It then discusses the Cantor-Bendixson rank of closed sets and proves related theorems and corollaries.

Uploaded by

Marco Garcia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
46 views3 pages

08 More Real Line

This document summarizes key definitions and results from a lecture on the real line. It introduces the definitions of isolated points, limit points, and perfect sets. It proves some lemmas, including that any nonempty perfect set is equivalent to the real numbers and any closed uncountable set contains a perfect subset. It then discusses the Cantor-Bendixson rank of closed sets and proves related theorems and corollaries.

Uploaded by

Marco Garcia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Lecture 8: The Real Line, Part II

February 11, 2009

Definition 6.13. a ∈ X is isolated in X iff there is an open interval I for which


X ∩ I = {a}. Otherwise, a is a limit point.
Remark. Another way to state this is that a is isolated if it is not a limit point
of X.
Definition 6.14. X is a perfect set iff X is closed and has no isolated points.
Remark. This definition sounds nice and tidy, but there are some very strange
perfect sets. For example, the Cantor set is perfect, despite being nowhere
dense!
Our goal will be to prove the Cantor-Bendixson theorem, i.e. the perfect set
theorem for closed sets, that every closed uncountable set has a perfect subset.
Lemma 6.15. If P is a perfect set and I is an open interval on R such that
I ∩ P 6= ∅, then there exist disjoint closed intervals J0 , J1 ⊂ I such that int[J0 ] ∩
P 6= ∅ and int[J1 ] ∩ P 6= ∅. Moreover, we can pick J0 and J1 such that their
lengths are both less than any ² > 0.
Proof. Since P has no isolated points, there must be at least two points a0 , a1 ∈
I ∩ P . Then just pick J0 3 a0 and J1 3 a1 to be small enough. SDG

Lemma 6.16. If P is a nonempty perfect set, then P ∼ R.


Proof. We exhibit a one-to-one mapping G : 2ω → P .
Note that 2ω can be viewed as the set of all infinite paths in a full, infinite
binary tree with each edge labeled by 0 or 1. We label each node in the tree by
the sequence of labels on the path from the root to the node.
Now we associate an interval Is to each node s, with the properties that
• Is is closed,
• Is ∩ P 6= ∅,
• Is,b ⊂ Is ,
• Is,0 ∩ Is,1 = ∅, and
• |Is | < 1/(|s| + 1),
where |I| denotes the length of interval I and |s| denotes the length of sequence
s.
In particular, if hi denotes the empty sequence, let Ihi be the closure of I ∩ P
for some open interval I with length at most 1 whose intersection with P is
nonempty. Then, given a set Is satisfying the above properties, by Lemma 6.15
choose Is,0 and Is,1 to be disjoint closed subintervals of Is shorter than 1/(|s|+2)
whose intersection with P is nonempty.

25
Now, for all f ∈ 2ω , define
\
G(f ) = If (i) ,
i∈ω

where f (n) = {f (0), f (1), . . . , f (n)}. Actually, we are abusing notation a bit
here: what we mean is that G(f ) is the unique member of the given intersection;
we must show that this intersection does indeed result in a singleton set. This
follows from the fact that we have an infinite intersection of nested, closed in-
tervals of arbitrarily small length and that the real numbers are order-complete.
To see that G(f ) ∈ P , note that G(f ) is an intersection of decreasing inter-
vals, each of which has a nonempty intersection with P ; if we pick one point
from the intersection of each interval with P , they form a sequence with limit
G(f ), which is contained in P since P is closed.
Finally, suppose f, f 0 ∈ 2ω with f 6= f 0 . Let n ∈ ω be the smallest index
for which f (n) 6= f 0 (n). Then If (n) ∩ If 0 (n) = ∅ by construction, and therefore
G(f ) ∩ G(f 0 ) = ∅. This shows that G is injective. SDG

Theorem 6.17 (Cantor-Bendixson). If C ⊆ R is closed and uncountable, then


there exists some perfect, nonempty P ⊆ C.
Remark. In a sense, this is where set theory started. This proof is what mo-
tivated the development of transfinite ordinals, since it describes a recursive
process that is not completed after the first limit stage.

Proof. Let C ⊆ R be closed. Define the Cantor-Bendixson derivative

C 0 = { a ∈ C | a is a limit point of C }.

This operation maps closed sets to closed sets, since closed sets in R are those
which contain all their limit points, and the derivative is monotone and retains
all limit points. Then define

C0 = C
Cα+1 = (Cα )0
\
Cλ = Cβ (lim(λ)).
β<λ

Note that Cβ is closed for all β by induction.


Claim: Cγ = Cγ+1 for some γ. For if not, Cα 6= Cβ for any α 6= β, since C
is monotone. Then C(−) would be an injection Ord → P(C), which is absurd.
Note that Cγ is perfect, since it consists solely of limit points and is closed.
If Cγ 6= ∅, we are done.
We claim that Cγ cannot be ∅ since this would imply that C is countable.
Consider Cβ − Cβ+1 , which contains all the isolated points in Cβ . That is, if
a ∈ Cβ − Cβ+1 , there exists an open interval Ia 3 a such that Cβ ∩ Ia = {a}.
In particular, we may choose Ia to be an open interval with rational endpoints.

26
Note that each Ia is distinct; otherwise, at the earliest stage when Ia arose,
it would have contained more than one point. Therefore, we have an injection
from C into the set of intervals with rational endpoints, which is countable. SDG

Remark. The above proof shows that every closed set can be decomposed into
a perfect subset and a countable subset. (In fact, it turns out that every closed
set can be uniquely so decomposed.)
Definition 6.18. The smallest γ in the above proof for which Cγ = Cγ+1 is
called the Cantor-Bendixson rank of C, and the above proof shows that γ < ℵ1 .
Exercise: construct closed sets whose Cantor-Bendixson rank is strictly
greater than ω. In fact, it can be shown that for every γ < ℵ1 , there exists
a closed C ⊆ R with Cantor-Bendixson rank γ.
Corollary 6.19. For every C ⊆ R, if C is closed and uncountable then C ∼ R.
This follows from Lemma 6.16 and Theorem 6.17.
Remark. We might hope that every uncountable set has a perfect subset; this,
of course, would resolve the CH. However. . .
Theorem 6.20. There exists a set X with card(X) = 2ℵ0 = card(R − X) such
that for every perfect set P , P 6⊆ X and P 6⊆ R − X.
Proof. We use the AC to construct X. Let Pα , α < 2ℵ0 be an ordering of the
perfect sets (there are 2ℵ0 perfect sets; see Lemma 6.21). Also, let xα be an
ordering of R. Now define rγ to be the real number with next-to-least index in
the sequence xα which comes after all rβ , β < γ, and for which rγ ∈ Pγ . We
can keep picking such rγ since each Pα has cardinality 2ℵ0 and therefore cannot
be contained in any initial segment of the xα ’s. SDG

Lemma 6.21. There are 2ℵ0 perfect sets.


Proof. There are at least 2ℵ0 perfect sets, since there is an injection from P(N)
to the set of all perfect sets: for each set of naturals, identify each natural with
a small closed interal containing it, and take the union. There are at most 2ℵ0
perfect sets since there are 2ℵ0 closed sets (which, in turn, follows from the
fact that any closed set can be expressed as a countable intersection of rational
intervals). SDG

27

You might also like