Lecture #5: The Borel Sets of R

Statistics 451 (Fall 2013) September 13, 2013
Prof. Michael Kozdron
Lecture #5: The Borel Sets of R
We will now begin investigating the second of the two claims made at the end of Lecture #3,
namely that there exists a σ-algebra B1 of subsets of [0, 1] on which it is possible to define
a uniform probability.
Our goal for today will be to define the Borel sets of R. The actual construction of the
uniform probability will be deferred for several lectures.
Recall that a set E ⊆ R is said to be open if for every x ∈ E there exists some � > 0
(depending on x) such that the interval (x − �, x + �) is contained in E. Also recall that
intervals of the form (a, b) for −∞ < a < b < ∞ are open sets. A set F ⊆ R is said to be
closed if F c is open. Note that both R and ∅ are simultaneously both open and closed sets.
If we consider the collection O of all open sets of R, then it follows immediately that O is not
a σ-algebra of subsets of R. (That is, if A ∈ O so that A is an open set, then, by definition,
Ac is closed and so Ac ∈ / O.) However, we know that σ(O), the σ-algebra generated by O,
exists and satisfies O � σ(O) ⊆ 2R . This leads to the following definition.
Definition. The Borel σ-algebra of R, written B, is the σ-algebra generated by the open
sets. That is, if O denotes the collection of all open subsets of R, then B = σ(O).
Since B is a σ-algebra, we see that it necessarily contains all open sets, all closed sets,
all unions of open sets, all unions of closed sets, all intersections of closed sets, and all
intersections of open sets.
Exercise 5.1. The purpose of this exercise is to is to remind you of some facts about open
and closed sets. Suppose that {E1 , E2 , . . .} is an arbitrary collection of open subsets of R,
and suppose that {F1 , F2 , . . .} is an arbitrary collection of closed subsets of R. Prove that
�∞
(a) j=1 Ej is necessarily open,
�∞
(b) j=1 Ej need not be open,
�∞
(c) j=1 Fj is necessarily closed, and
�∞
(d) j=1 Fj need not be closed.
We often call a countable intersection of open sets a Gδ set (from the German Gebeit for
open and Durchschnitt for intersection) and a countable union of closed sets an Fσ set (from
the French fermé for closed and somme for union).
The following theorem characterizes open subsets of R and will occasionally be of use.
5–1
Theorem 5.2. If E ⊆ R is an open set, then there exist at most countably many disjoint
open intervals Ij , j = 1, 2, . . ., such that
∞
�
E= Ij .
j=1
Proof. The trick is to define an equivalence relation on E as follows. If a, b ∈ E, we say

that a and b are equivalent, written a ∼ b, if the entire open interval (a, b) is contained in
E. This equivalence relationship partitions E into a disjoint union of classes. We do not
know a priori that there are at most countably many such classes. Therefore, label these
classes Ij , j ∈ J, where J is an arbitrary index set. Note that Ij is, in fact, an interval for
the following reason. If aj , bj ∈ Ij , then aj ∼ bj so that the entire open interval (aj , bj ) is
contained in Ij . As aj , bj were arbitrary we see that Ij is, in fact, an interval. The next step
is to show that Ij is open. Let x ∈ Ij be arbitrary. Since x ∈ E and E is open, we know that
there exists an � > 0 such that (x − �, x + �) ⊆ E. However, we clearly have a ∼ x for every
a ∈ (x − �, x + �) which implies that this �-neighbourhood of x is contained in Ij . Thus, Ij
is open as required. The final set is to show that there are at most countably many Ij . This
follows from the fact that each Ij must contain at least one rational number. Since there are
countably many rationals, there can be at most countably many Ij .
It is worth noting that we have not yet proved that B �= 2R ; in other words, we have not yet
proved that there exist non-Borel sets. It turns out that the set H constructed in Lecture #4
is non-Borel, although we will not prove this at present. In fact, there is no simple procedure
to determine whether a given set A ⊆ R is Borel or not. However, one way to understand B
is that it is generated by intervals of the form (−∞, a] as the next theorem shows.
Theorem 5.3. The Borel σ-algebra B is generated by intervals of the form (−∞, a] where
a ∈ Q is a rational number.
Proof. Let O0 denote the collection of all open intervals. Since every open set in R is an
at most countable union of open intervals, we must have σ(O0 ) = B. Let D denote the
collection of all intervals of the form (−∞, a], a ∈ Q. Let (a, b) ∈ O0 for some b > a with a,
b ∈ Q. Let
1
an = a +
n
so that an ↓ a as n → ∞, and let
1
bn = b −
n
so that bn ↑ b as n → ∞. Thus,
∞
� ∞
�
(a, b) = (an , bn ] = {(−∞, bn ] ∩ (−∞, an ]c }
n=1 n=1
which implies that (a, b) ∈ σ(D). That is, O0 ⊆ σ(D) so that σ(O0 ) ⊆ σ(D). However,
every element of D is a closed set which implies that
σ(D) ⊆ B.
5–2
This gives the chain of containments
B = σ(O0 ) ⊆ σ(D) ⊆ B
and so σ(D) = B proving the theorem.
The Borel sets of [0, 1]

If we now consider the set [0, 1] ⊂ R as the sample space, then B1 , the Borel σ-algebra
of [0, 1], is the σ-algebra generated by the collection of open subsets of [0, 1]. As the next
exercise shows, we can equivalently think of B1 as the restriction of B to [0, 1]. Note that
B1 ⊂ B as collections of sets, but not as σ-algebras. That is, B1 is not a sub-σ-algebra of B.
The reason, of course, is that B is a σ-algebra of subsets of R whereas B1 is a σ-algebra of
subsets of [0, 1]; in order for one σ-algebra to be a sub-σ-algebra of another σ-algebra, it is
necessarily the case that the underlying sample spaces for both σ-algebras are the same.
Exercise 5.4. Suppose that Ω is the sample space and Ω� ⊆ Ω. Show that
(a) if F is a σ-algebra of subsets of Ω and F � = {A ∩ Ω� : A ∈ F}, then F � is a σ-algebra

of subsets of Ω� , and
(b) if C generates the σ-algebra F in Ω and C � = {A ∩ Ω� : A ∈ C}, then C � generates the

σ-algebra F � in Ω’.
5–3

Lecture #5: The Borel Sets of R

Uploaded by

Copyright:

Available Formats

Lecture #5: The Borel Sets of R

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture #5: The Borel Sets of R

Uploaded by

Copyright:

Available Formats

Statistics 451 (Fall 2013) September 13, 2013

Prof. Michael Kozdron

Lecture #5: The Borel Sets of R

Proof. The trick is to define an equivalence relation on E as follows. If a, b ∈ E, we say

and so σ(D) = B proving the theorem.

The Borel sets of [0, 1]

(a) if F is a σ-algebra of subsets of Ω and F � = {A ∩ Ω� : A ∈ F}, then F � is a σ-algebra

(b) if C generates the σ-algebra F in Ω and C � = {A ∩ Ω� : A ∈ C}, then C � generates the

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.