4. Measurable Spaces

4.1 σ\sigma-Fields and Measurable Spaces

Definition 4.1 (σ\sigma-field): Let Ω\Omega be a nonempty set. A family of subsets F\sigmaF of Ω\Omega is called a σ\sigma-field on Ω\Omega if the following three items are satisfied:
  1. ΩF\Omega \in \sigmaF
  2. AF    AcFA \in \sigmaF \implies A^c \in \sigmaF
  3. if

    \qty{A_i: i \in \N}

    is a collection of sets s.t. AiFA_i \in \sigmaF for any iNi \in \N, then iNAiF\bigcup_{i \in \N} A_i \in \sigmaF
Example (Trivial σ\sigma-field): Let Ω\Omega \neq \varnothing be an arbitrary set. Let

\sigmaF = \qty{\varnothing, \Omega}

.
Then, F\sigmaF is a σ\sigma-field on Ω\Omega.
  1. Clearly, ΩF\Omega \in \sigmaF.
  2. Let AFA \in \sigmaF. Then, there are only two cases, eiter A=A = \varnothing or A=ΩA = \Omega. In each case, AcFA^c \in \sigmaF.
  3. Consider a countable collection

    \qty{A_i : i \in \N} \subset \sigmaF

    .
    This collection is composed only of the sets Ai=A_i = \varnothing or Ai=ΩA_i = \Omega, iNi \in \N. Thus iNAi={Ωif i s.t. Ai=Ωotherwise \bigcup_{i \in \N} A_i = \begin{cases} \Omega & \text{if } \exists i \text{ s.t. } A_i = \Omega \\ \varnothing & \text{otherwise} \end{cases} and iNAiF\bigcup_{i \in \N} A_i \subset \sigmaF.
Example (Power set): Let Ω\Omega \neq \varnothing be an arbitrary set. Let F\sigmaF be the family which consists of all possible subsets of Ω\Omega, i.e.

\sigmaF = \qty{A : A \subset \Omega}

.
Then, F\sigmaF is a σ\sigma-field on Ω\Omega.
  1. Since ΩΩ\Omega \subset \Omega, ΩF\Omega \in \sigmaF.
  2. Let AFA \in \sigmaF, then by definition Ac=ΩAΩA^c = \Omega \setminus A \subset \Omega and hence AcFA^c \in \sigmaF.
  3. Leet

    \qty{A_i : i \in \N} \subset \sigmaF

    .
    Then, by definition

    \bigcup_{i \in \N} A_i = \qty{\omega \in \Omega : \exists i \text{ s.t. } \omega \in A_i} \subset \Omega

    and iNAiF\bigcup_{i \in \N} A_i \subset \sigmaF.
Note: The trivial σ\sigma-field is the smallest possible σ\sigma-field on Ω\Omega. The power set also denoted with P(Ω)\powerset(\Omega) or 2Ω2^{\Omega} is the largest possible σ\sigma-field on Ω\Omega.
Example (Uncountable Ω\Omega, countable A,AcA, A^c): Let Ω\Omega be an uncountable set. We consider the family

\sigmaF = \qty{A: A \subset \Omega \text{ s.t. } A \text{ is countable or } A^c \text{ is countable}}

Then, F\sigmaF is a σ\sigma-field on Ω\Omega.
  1. We have that Ωc=\Omega^c = \varnothing. We know that #=0\#\varnothing = 0, in particular \varnothing is countable. Thus, ΩF\Omega \in \sigmaF.
  2. Let AFA \in \sigmaF. Thus, AA is countable or AcA^c is countable. Since (Ac)c=A(A^c)^c = A, AcFA^c \in \sigmaF.
  3. Let

    \qty{A_i : i \in \N} \subset \sigmaF

    .
    If there exists jNj \in \N s.t. AjA_j is uncountable, we have that

    \qty(\bigcup_{i \in \N} A_i)^c = \bigcap_{i \in \N} A_i^c = \bigcap_{i \in \N, \ i \neq j} A_i^c \cap A_j^c \subset A_j^c

    hence, using the fact that AjcA_j^c must be countable, we have

    \#\qty(\bigcup_{i \in \N} A_i)^c \leq \# A_j^c \leq \# \N

    and iNAiF\bigcup_{i \in \N} A_i \in \sigmaF. If for all iNi \in \N we have that AiA_i is countable we rely on TODO: reference Proposition (Union of countable sets) and conclude that iNAi\bigcup_{i \in \N} A_i must be countable as well and hence iNAiF\bigcup_{i \in \N} A_i \in \sigmaF.
Note: Notice for the aforementioned σ\sigma-field that since Ω\Omega is uncountable it is not true that F=P(Ω)\sigmaF = \powerset(\Omega). As a simple example consider Ω=[0,1)\Omega = [0,1), then both A=[0,0.5)A = [0,0.5) and [0.5,1)[0.5,1) are not countable.
Example (σ\sigma-field restricted on subset): Let Ω\Omega be a non empty set and F\sigmaF be a σ\sigma-field on Ω\Omega. Let Ω0Ω\Omega_0 \subset \Omega s.t. Ω0\Omega_0 \neq \varnothing. Then, the collection

\sigmaF \cap \Omega_0 = \qty{A \cap \Omega_0 : A \in \sigmaF}

is a σ\sigma-field on Ω0\Omega_0.
  1. Clearly, Ω0=ΩΩ0FΩ0\Omega_0 = \Omega \cap \Omega_0 \in \sigmaF \cap \Omega_0.
  2. If BFΩ0B \in \sigmaF \cap \Omega_0, then B=AΩ0B = A \cap \Omega_0 for some AFA \in \sigmaF and Ω0B=(Ω0A)(Ω0Ω0)=Ω0A=AcΩ0FΩ0\Omega_0 \setminus B = (\Omega_0 \setminus A) \cup (\Omega_0 \setminus \Omega_0) = \Omega_0 \setminus A = A^c \cap \Omega_0 \in \sigmaF \cap \Omega_0.
  3. Suppose that

    \qty{B_i : i \in \N} \subset \sigmaF \cap \Omega_0

    .
    Therefore, for any iNi \in \N, Bi=AiΩ0B_i = A_i \cap \Omega_0 for some AiFA_i \in \sigmaF. Then, since iNBi=(iNAi)Ω0\bigcup_{i \in \N} B_i = (\bigcup_{i\in\N} A_i) \cap \Omega_0 and iNAiF\bigcup_{i\in\N} A_i \in \sigmaF, it follows that iNBiFΩ0\bigcup_{i \in \N} B_i \in \sigmaF \cap \Omega_0.
Example (Infinte Ω\Omega, finite A,AcA, A^c): Let Ω\Omega be an infinite set. Define the family

\set{G} = \qty{A : A \subset \Omega \text{ s.t. } A \text{ is finite or } A^c \text{ is finite}}

Then, G\set{G} is not a σ\sigma-field on Ω\Omega. To see this, let

\qty{\omega_i : i \in \N}

be a countably infinite sequence of distinct points of Ω\Omega and define the set

A = \qty{\omega_{2i} : i \in \N}

.
We note that both AA and, since

\qty{\omega_{2i + 1} : i \in \N} \subset A^c

,
AcA^c are not finite. Let

A_i = \qty{\omega_{2i}}

s.t. A=iNAiA = \bigcup_{i \in \N} A_i. It is clear that AiGA_i \in \set{G} but A∉GA \not\in \set{G}, hence G\set{G} is not a σ\sigma-field on Ω\Omega.
Example (Left-open real intervals): Let Ω=R\Omega = \R and define the family

\set{R} = \qty{A = (a,b]: a,b \in \R} \cup \qty{\varnothing}

Then, R\set{R} is not a σ\sigma-field on Ω\Omega. To see this, let A=(a,b]RA = (a,b] \in \set{R} with AA \neq \varnothing, then Ac=(,a](b,)∉RA^c = (-\infty,a] \cup (b,\infty) \not \in \set{R}.
Definition 4.2 (Sub-σ\sigma-field): Let F\sigmaF be a σ\sigma-field on Ω\Omega. If AF\set{A} \subset \sigmaF and A\set{A} is a σ\sigma-field on Ω\Omega, then A\set{A} is referred to as a sub-σ\sigma-field on Ω\Omega.
Example (σ\sigma-field restricted on subset continued): We show Ω0F\Omega_0 \in \sigmaF iff FΩ0\sigmaF \cap \Omega_0 is a sub-σ\sigma-field of F\sigmaF. We prove the direction “Ω0F\Omega_0 \in \sigmaF    \impliesFΩ0\sigmaF \cap \Omega_0 is a sub-σ\sigma-field of F\sigmaF” first. We know that

\sigmaF \cap \Omega_0 = \qty{A \cap \Omega_0 : A \in \sigmaF}

and from an exercise (TODO, make proposition) we know that if AFA \in \sigmaF and Ω0F\Omega_0 \in \sigmaF, then AΩ0FA \cap \Omega_0 \in \sigmaF, thus FΩ0F\sigmaF \cap \Omega_0 \subset \sigmaF. We already proved before that FΩ0\sigmaF \cap \Omega_0 is a σ\sigma-field. For the direction “FΩ0\sigmaF \cap \Omega_0 is a sub-σ\sigma-field of F\sigmaF    \impliesΩ0F\Omega_0 \in \sigmaF” we note that Ω0FΩ0\Omega_0 \in \sigmaF \cap \Omega_0 as ΩΩ0=Ω0\Omega \cap \Omega_0 = \Omega_0. Hence Ω0F\Omega_0 \in \sigmaF follows from FΩ0F\sigmaF \cap \Omega_0 \subset \sigmaF.
Definition 4.3 (Measurable space): Let Ω\Omega \neq \varnothing and F\sigmaF be a σ\sigma-field on Ω\Omega. The pair (Ω,F)(\Omega, \sigmaF) is referred to as a measurable space.
Note: If AFA \in \set{F}, then AA is said to be measurable.

4.2 σ\sigma-Fields generated by Families of Sets

The next result is of general importance as it shows that even though a family of subsets G\set{G} might not be a σ\sigma-field, one can always find a σ\sigma-field which is the smallest possible σ\sigma-field that contains G\set{G}.

Proposition 4.4 (σ\sigma-field generated by G\set{G}): Let Ω0\Omega \neq 0 and G\set{G} be a family of subsets of Ω\Omega. Then, there exists a σ\sigma-field σ(G)\sigma(\set{G}) which satisfies:
  1. Gσ(G)\set{G} \subset \sigma(\set{G})
  2. If GU\set{G} \subset \set{U} and U\set{U} is a σ\sigma-field, then σ(G)U\sigma(\set{G}) \subset \set{U}.

To prove proposition (TODO: ref) we rely on the following result:

Proposition 4.5 (Union of σ\sigma-fields is a σ\sigma-field): Let Ω\Omega \neq \varnothing be a set. Let Fi\sigmaF_i, iIi \in I be a collection of σ\sigma-fields on Ω\Omega over an arbitrary set II. Then F=iIFi\sigmaF = \bigcap_{i \in I} \sigmaF_i is a σ\sigma-field on Ω\Omega.

TODO: PROOF

The σ\sigma-field σ(G)\sigma(\set{G}) is referred to as the σ\sigma-field generated by G\set{G}.

Proposition 4.6 (Properties of σ\sigma-field generated by G\set{G}): Let σ(G)\sigma(\set{G}) be the σ\sigma-field generated by a family of subsets G\set{G} of Ω\Omega. Let A\set{A} be another family of subsets of Ω\Omega. Then,
  1. if A\set{A} is a σ\sigma-field s.t. GA\set{G} \subset \set{A} and Aσ(G)\set{A} \subset \sigma(\set{G}), then A=σ(G)\set{A} = \sigma(\set{G})
  2. AG    σ(A)σ(G)\set{A} \subset \set{G} \implies \sigma(\set{A}) \subset \sigma(\set{G})
  3. AGσ(A)    σ(A)=σ(G)\set{A} \subset \set{G} \subset \sigma(\set{A}) \implies \sigma(\set{A}) = \sigma(\set{G})
Example (

\set{G} = \qty{\varnothing}

): Let Ω\Omega \neq \varnothing and let

\set{G} = \qty{\varnothing}

.
Then

\sigma(\set{G}) = \qty{\varnothing, \Omega}

,
i.e. the trivial σ\sigma-field on Ω\Omega. By 1. of Proposition 4.6 it is enough to show that

\qty{\varnothing, \Omega} \subset \sigma(\set{G})

since

\qty{\varnothing, \Omega}

is a σ\sigma-field that contains G\set{G}. It is clear that

\qty{\varnothing, \Omega} \subset \sigma(\set{G})

is a σ\sigma-field, hence it must contain both, \varnothing and c=Ω\varnothing^c = \Omega.
Example (

\set{G} = \qty{\qty{1}}

): Let

\Omega = \qty{1,2,3}

and define

\set{G} = \qty{\qty{1}}

.
Clearly, σ(G)\sigma(\set{G}) must contain

\qty{1}

,

\qty{1}^c = \qty{2,3}

,
Ω\Omega and \varnothing. Thus we claim

\sigma(\set{G}) = \qty{\varnothing, \qty{1}, \qty{2,3}, \Omega}

.
More generally, if AΩA \subset \Omega, then

\sigma(\qty{A}) = \qty{\varnothing, A, A^c, \Omega}

.
Note: We often omit the braces and use the notation

\sigma(A) = \sigma(\qty{A})

.
Example (

\set{G} = \qty{\qty{\omega}: \omega \in \Omega}

): Let Ω\Omega be an uncountable set and

\sigmaF = \qty{A: A \subset \Omega \text{ s.t. } A \text{ is countable or } A^c \text{ is countable}}

be the σ\sigma-field introduced in a previous example. We show that F=σ(G)\sigmaF = \sigma(\set{G}) where

\set{G} = \qty{\qty{\omega}: \omega \in \Omega}

.
Clearly GF\set{G} \subset \sigmaF since each set

\qty{\omega}

has cardinality one. Thus, it remains to show that Fσ(G)\sigmaF \subset \sigma(\set{G}). Let AFA \in \sigmaF. Then either AA or AcA^c is countable. Suppose that AA is countable, then

A = \qty{\omega_i : i \in \N}

for some collection of singletons ωiΩ\omega_i \in \Omega. Therefore,

A = \bigcup_{i \in \N}\qty{\omega_i}

and thus Aσ(G)A \in \sigma(\set{G}) since

\qty{\omega_i} \in \sigma(\set{G})

.
If AcA^c is countable, by the latter argument, Acσ(G)A^c \in \sigma(\set{G}) and hence A=(Ac)cσ(G)A = (A^c)^c \in \sigma(\set{G}).
Note: The set

\set{G} = \qty{\qty{\omega}: \omega \in \Omega}

is referred to as the point-partition on Ω\Omega.
Proposition 4.7 (σ\sigma-field generated by

\qty{G \cap \Omega_0 : G \in \set{G}}

): Let Ω\Omega be a non empty set and F\sigmaF be a σ\sigma-field on Ω\Omega. Let Ω0Ω\Omega_0 \subset \Omega s.t. Ω0\Omega_0 \neq \varnothing. Assume F=σ(G)\sigmaF = \sigma(\set{G}) for some family G\set{G} of subsets of Ω\Omega. Then

\sigmaF \cap \Omega_0 = \sigma(\qty{G \cap \Omega_0 : G \in \set{G}})

.

4.3 Borel Sets of Real Coordinate Spaces

Definition 4.8 (Borel σ\sigma-field on the real numbers): Let Ω=R\Omega = \R and

\set{R} = \qty{(-\infty,x] : x \in \R} \cup \qty{\varnothing}

be the family of left open intervals with the empty set adjoined. The σ\sigma-field B(R)=σ(R)\borelB(\R) = \sigma(\set{R}) is referred to as the Borel σ\sigma-field on R\R.
Definition 4.9 (Borel σ\sigma-field on real coordinate spaces): Let Ω=Rk\Omega = \R^k, kNk \in \N. Define

\set{R}_k = \qty{A : A = \bigtimes_{i=1}^k (a_i,b_i], \ a_i, b_i \in \R, \ i = 1,\ldots,k} \cup \qty{\varnothing}

i.e. the family of rectangles in Rk\R^k. Then, the σ\sigma-field B(Rk)=σ(Rk)\borelB(\R^k) = \sigma(\set{R}_k) is referred to as the Borel σ\sigma-field on Rk\R^k.
Definition 4.10 (Borel σ\sigma-field on real coordinate spaces subset): Let ERkE \subset \R^k s.t. EE \neq \varnothing. The Borel σ\sigma-field on EE is defined by

\borelB(E) = \borelB(\R^k) \cap E = \qty{A \cap E: A \in \borelB(\R^k)}

.

Upon Proposition 4.7, the following result is obtained.

Proposition 4.11 (Alternative formulation of B(E)\borelB(E)): Given any ERkE \subset \R^k s.t. EE \neq \varnothing, we have

\borelB(E) = \sigma(\qty{G \cap E : G \in \set{R}_k})

.