Assignment 1

Abhinav Pradeep
March 11, 2024

1 Question 1
Let C := {{x}, x ∈ R}. Show that every set in σC is either countable or has a countable complement.

Define a collection of sets D to contain all sets on R such that the sets are either countable or have a
countable complement.

Show that D is σ-algebra that contains C. Therefore, by definition, σC ⊆ D.

To show that D is a σ-algebra:

1. D is closed under complement: The complement of any countable set in R must be uncountable
as R is itself uncountable. Moreover, D by definition contains only uncountable subsets of R such
that their complement is countable. Combining these two facts, it is clear that D is closed under
complement.

2. ∅ ∈ D: D was defined to contain all countable subsets of R. As ∅ is defined to be countable, ∅ ∈ D.

3. D is closed under countable union: Countable unions of countable sets are countable. Therefore
immediately all countable sets in D are closed under countable unions. Now consider any sequence An
of uncountable sets in D:
[
An
n

Clearly, the countable unions of uncountable sets will be uncountable. Now consider that D was
defined to contain any uncountable set whose complement was countable. Hence, it would suffice to
show that the complement of the above is countable:
!c
[
An
n

By De Morgans law,
\
Acn
n

By definition of uncountable sets An ∈ D, Acnis countable. Countable intersections of countable sets

are countable. Therefore the above is countable. Hence countable unions of uncountable sets in D will

1
lie in D. Now consider countable unions of some mixed sequence Bn of countable and uncountable
sets in D. This can be split as below by the commutativity and associativity of unions:

[ [ [
Bn = Cn ∪ Un
n Ic Iu

Where the sequence Cn is the subsequence of countable sets and Un is the subsequence of uncountable
sets and Ic and Iu are the corresponding index sets of the subsequences such that I = Ic ∪ Iu where I
is
S the countable index set that the union is being taken over (implicitly,Sas it was just denoted
S with
n ). As was preciously shown above, these unions lie in D and suppose C
Ic n = C and U
Iu n = U.
Hence,
[
Bn = C ∪ U
n

The union of a countable and uncountable set will be uncountable. Now consider the complement
!c
[
Bn = (C ∪ U )c
n

By De’Morgans,
!c
[
Bn = Cc ∩ U c
n

Now as U in the above lies in D, the complement will be countable. The complement of a count-
able subset of R is always uncountable. However, the intersect of a countable and uncountable set
c c c
must be countable. This is because the intersect
c c c c
S by defeition has that C ∩ U ⊆ U and therefore
|C ∩ U | ≤ |U | and U was countable. Hence n Bn is an uncountable subset of R with countable
complement and therefore by definition lies in D. Hence D is closed under complement.

Hence, D is a σ-algebra. Moreover, as D contains all countable sets in R, C ⊂ D and hence σC ⊆ D.

Show that all sets in D are contained in σC. Therefore, D ⊆ σC.

Consider any set S ∈ D. By the definition of D, S is either countable or has a countable complement.

Consider the case that S is countable. σC is the smallest σ-algebra that contains C. By definition
σC is closed under countable union. Any countable set in R can be formed from countable unions of
point-sets. Hence, S ∈ σC.

Consider if S has a countable complement. As was previously established, σC contains all countable
sets in R. Hence S c ∈ σC. Moreover, σC is closed under complement. Hence, S ∈ σC.

Hence, (D ⊆ σC) ∧ (σC ⊆ D) =⇒ D = σC.

2 Question 2
Let (E, E) be a measurable space. Let D ⊆ E (not necessarily in E). Define D as the collection of the
sets of the form A ∩ D where A ∈ E. Show that D is a σ-algebra on D. The measurable space (D, D)
is called the trace of (E, E) on D.

Show that the set D is a σ-algebra on D.

D = {A ∩ D | A ∈ E}
As the intersect is used in the definition, all sets in D are in D. Now consider that:

1. D is closed under complement: Consider for some A ∩ D ∈ D, its complement with respect to D
will be:

Ac ∩ D
A ∈ E =⇒ Ac ∈ E as E is a σ-algebra. Hence, by the definition of D, Ac ∩ D ∈ D.

2. D contains ∅: Consider that ∅ ∈ E. Hence, as ∅ ∩ D = ∅, ∅ ∈ D.

3. D is closed under countable union: Consider some sequence of sets Bn in D

Bn := An ∩ D

[ [
Bn = (An ∩ D)
n n

By the distributive law

[ [
Bn = (An ) ∩ D
n n
S S
An ∈ E ∀n =⇒ n (An ) ∈ E. Hence n (Bn ) ∈ D. Hence D is closed under countable union.
3 Question 3
Prove that Definition 1.10 of a d-system is equivalent to Definition 1.6 of a σ-algebra, replacing (in
the latter definition) E with D and Item 3. with:
S
3’. If A1 , A2 , · · · ∈ D are disjoint, then also An ∈ D.
n
New definition of σ-algebra =⇒ it is a d-system.

1. Is the same condition as for d-systems.

2. σ-algebras are defined to be closed under complement with respect to the universe. Hence A, B ∈ D,
then Ac , B c ∈ D. Consider some B ⊆ A. Want to show that A \ B ∈ D where:

A \ B := A ∩ B c
Consider the complement w.r.t the universe:

(A \ B)c := (A ∩ B c )c
by De’Morgans,

(A \ B)c := Ac ∪ B
As B ⊆ A, Ac and B are disjoint. By 3’ the union of disjoint sets lies in D. Hence:

(A \ B)c ∈ D
As σ-algebras are closed under complement w.r.t the universe,

(A \ B) ∈ D
3. Consider some increasing (An ⊆ An+1 ) sequence of sets An ∈ the σ-algebra. Deconstruct this into
sequence Bn as:

Bn = An \ A(n−1)
With B1 = A1 . As was previously shown, this set Bn ∈ the σ-algebra as An−1 ⊆ An implies closure
under relative complement and therefore, An \ An−1 ∈ the σ-algebra. Morover the sequence is disjoint.
Consider first the simplest case:

B1 ∩ B2 = A1 ∩ (A2 \ A1 )
Clearly B1 ∩ B2 = ∅

For full generality consider ∀ϵ ∈ N, ϵ ≥ 1:

Bn ∩ Bn+ϵ = (An \ A(n−1) ) ∩ (A(n+ϵ) \ A(n+ϵ−1) )

Pictorially:

Clearly as An ⊆ An+ϵ−1 (as ϵ ≥ 1), An ∩ (An+ϵ \ An+ϵ−1 ) = ∅. Hence as An−1 ⊆ An ,

(An \ A(n−1) ) ∩ (A(n+ϵ) \ A(n+ϵ−1) ) = ∅

Bn ∩ Bn+ϵ = ∅
S
S sequence such that each Bn ∈ the σ-algebra. Hence, n Bn ∈ the σ-algebra.
Hence, Bn is disjoint
This also implies n An ∈ the σ-algebra as Bn only breaks down the increasing sequence An into a
disjoint sequence of rings as illustrated before. The argumet to show this will be similar to that of the
above. Consider the simplest case:

B1 ∪ B2 = A1 ∪ A2 \ A1
Clearly,

B1 ∪ B2 = A2
As we removed A1 and added back A1 with the union. In general,
[ [
Bn = An \ An−1
n n

Clearly the removing and adding back process will keep happening such that,
[ [
Bn = An
n n
To show this more rigorously, consider that:
[ [
Bn ⊆ An
n n
S
Consider any e ∈ n Bn .SHence for some n = k, e ∈ Bk . By definition, Bk := Ak \ Ak−1 . Hence,
e ∈ Ak and therefore e ∈ n An .
[ [
An ⊆ Bn
n n
S
Consider some e ∈ n An . As An is increasing, there must intuitively exist some minimum n = k such
that e is in all An≥k . More rigorously, consider the set indices I := {n | e ∈ An }. Countable union
implies the index set is countable, i.e. has bijective mapping to the positive integers. By well ordering
of integers this them means that this set I has the smallest element k. Hence, Bk := Ak \ Ak−1 will
contain k. No set BS n>k will contain it by definition of B and the fact that An is increasing. Nonethe-
less, as e ∈ Bk , e ∈ n Bn
S S S
Hence, n Bn = n An and therefore n An ∈ the σ-algebra.

Definition of d-system =⇒ it is a σ-algebra under the new definition.

1. Same condition as for σ-algebras.

2. The d-system is defined to contain the universe, and sets in the d-system are subsets of the universe.
Hence, the second law of d-systems will satisfy closure w.r.t. the universe that σ-algebra demands.

3. Consider some disjoint sequence An ∈ the d-system. As was previously shown, Ac1 ∈ the d-system.
As A1 and A2 are disjoint, A2 ⊆ Ac1 . This can be then applied using condition 2. of d-systems. That
is Ac1 \ A2 ∈ D. Rewriting as:

Ac1 \ A2 = Ac1 ∩ Ac2

Note that as Ac1 \ A2 lies within D, the complement is an element of D. Hence,

(Ac1 \ A2 )c = (Ac1 ∩ Ac2 )c

by De’Morgans

(Ac1 \ A2 )c = A1 ∪ A2
Hence A1 ∪ A2 ∈ D. This argument can be repeated arbitrarily for any sequence of An where n is
finite. To show in the general case,
n
[
Bn := Ai
i=1

Clearly, Bn ⊆ Bn+1 ∀n. Moreover by S the proof before ∀n, Bn ∈ D. As d-systems are closed under
countable union of increasing sets, n Bn =: B ∈Sthe d-system. Moreover, as Bn is an increasing
sequence of sets, limn Bn = B. Hence as limn Bn = ni=1 Ai ∈ the d-system.
4 Question 4
Let f : E × F → G be (E ⊗ F)/G-measurable, where E, F and G are σ-algebras on E, F and G
respectively. Show that the function y 7→ f (x0 , y) is F/G-measurable for any fixed x0 ∈ E. Such a
function is called a section of f .

Consider the function h : F → E × F defined by y 7→ (x0 , y). Want to show that h is F/E ⊗ F-
measurable.

Consider that,
(
B if x0 ∈ A
h−1 (A × B) =
∅ if x0 ̸∈ A
Hence it is clear that for any A × B such that A ∈ E and B ∈ F, h−1 (A × B) ∈ F, as B ∈ F and
∅ ∈ F. The set (E ⊗ F)0 = {A × B | A ∈ E, B ∈ F} is what generates E ⊗ F. By Theorem 1.20,
if h−1 (S) ∈ F for any S ∈ (E ⊗ F)0 , then it holds that h−1 (S) ∈ F for any S ∈ E ⊗ F as E ⊗ F is
generated by (E ⊗F)0 . As h−1 (S) ∈ F for any S ∈ (E ⊗F)0 has been shown, h is F/E ⊗F-measurable.

Consider the function g : F → G defined by the map y 7→ f (x0 , y). Clearly, g = f ◦ h. As g is a

composition of measurable functions, it is a measurable function itself by Theorem 1.21.

5 Question 5
Let (E, E, µ) be a measure space and f an E-measurable (numerical) function. Prove that if f is
integrable, then f must be real-valued µ-almost everywhere. Hint: show this first for f ∈ E+ .

Consider some f + ∈ E+ .

f + is said to be integrable if µf + < ∞. For a proof by contradiction, consider if f + is integrable and

it is not real valued µ-almost everywhere. For f + to not be real valued µ-almost everywhere, then
f + → ∞ over some non-zero measured set S ∈ E.

Consider specifically the integral of f + on this set S. This is because,

µf + = µ f + 1S + µ f + 1S c

Hence,

µ f + 1S → ∞ =⇒ µf + → ∞

The integral of f + on S is defined as:

µ f + 1S = lim µfn+

Where fn+ are a sequence of increasing simple functions that point wise converge to f + 1S .

Define:

fn+ = n1S
Clearly,

lim fn+ = f + 1S
n↑∞

As f + was defined to be a positive function and not be real valued over S. Now consider that:

µfn+ = nµ(S)
And the integral of f + 1S is defined to be:

µ f + 1S = lim µfn+


n↑∞

µ f + 1S = lim nµ(S)


n↑∞

Clearly, as µ(S) ̸= 0,

lim nµ(S) = ∞
n↑∞

µ f + 1S = ∞

Hence, µf + → ∞ and therefore f + is not integrable. The above argument can be applied to f − ∈ E−
by applying a sign change. Moreover, for any f ∈ E, it can be decomposed as:

f = f+ − f−
Where f + := max{f, 0} and f − = max{−f, 0}

By the linearity of integration,

µf = µf + − µf −
Clearly, if either one of µf + → ∞ or µf − → ∞ then µf → ∞ and f is not integrable. If both
approach ∞ the integral is not defined and it is moot to speak of integrability. Hence, f must be real
valued µ-almost everywhere for it to be integrable.