MTH404: Measure Theory

Homework: These will be periodically updated. Odd Semester, 2025/26

1. Show that [0, 1] is not Lebesgue null.

Solution
By induction on n, show that for every a < b and a sequence hJk : 1 ≤ k ≤ ni of open
n
X
intervals that cover [a, b], we have length(Jk ) > b − a.
k=1

2. Show that the ternary Cantor set is Lebesgue null.

Solution
Let εT> 0. Choose n ≥ 1 such that (2/3)n < ε/2. Recall that the Cantor set
C = n≥1 Cn where Cn is a union of 2n closed intervals each of length 3−n . Let these
intervals be {Ik : 1 ≤ k ≤ 2n }. Let hJk : k ≥ 1i be a sequence of open intervals
defined as follows. For each k ≤ 2n , let Jk ⊇ Ik be an open intervalXsuch that
−n k n
length(Jk ) < 2 + ε/4 . If k > 2 , define Jk = ∅. Now check that length(Jn ) < ε
S n≥1
and C ⊆ Cn ⊆ n≥1 Jn .

3. Show that the ternary Cantor set is nowhere dense in R.

Solution
Let (a, b) be any open interval. We will find a subinterval (c, d) ⊆ (a, b) such that
C ∩ (c, d) = ∅. We can assume that 0 ≤ a < b ≤ 1 otherwise this is clear (if a < 0,
then (c, d) = (a, 0) works and if b > 1, then (c, d) = (1, b) works). Choose n ≥ 1 large
enough so that 3−n < (b − a)/5. Note that if we divide [0, 1] into 3n consecutive
closed subintervals each of length 3−n , then there must be 3 of these consecutive
subintervals say J1 , J2 , J3 which are all subsets of (a, b). Now observe that at least
one of the open intervals Interior(Jk ) for k ∈ {1, 2, 3} must be disjoint from Cn
(Why?). So we can take (c, d) to be this interval.

4. Let (X, d) be a metric space and D ⊆ X. Show that the following are equivalent.
(a) D is nowhere dense in X.
(b) cl(D) (closure of D) is nowhere dense in X.
(c) X \ D contains an open dense subset of X.
Solution
(a) =⇒ (c): Assume D is nowhereSdense in X. Let F be the family of all open sets
that are disjoint from D. Let V = F. Then V is open since the union of any family
of open sets is open. Since D is nowhere dense, every nonempty open set A has a
nonempty open subset B disjoint from D. But this means that B ∈ F and so B ⊆ V .
 So V meets every nonempty open set hence V is dense in X. As V ⊆ X \ D, V is as
required.
(c) =⇒ (b): Let U be an open dense subset of X contained in X \ D. Then X \ U is
closed as complements of open sets are closed. Since D ⊆ X \ U , cl(D) ⊆ X \ U
because closure of any set is the smallest closed set containing it. To show that cl(D)
is nowhere dense, it suffices to show that X \ U is nowhere dense since
cl(D) ⊆ X \ U . Let A be any nonempty open set. Put B = A ∩ U . Then B is
nonempty (as U is dense in X) open and B ∩ (X \ U ) = ∅. It follows that X \ U is
nowhere dense and we are done.
(b) =⇒ (a):: Clear since subset of nowhere dense sets are nowhere dense.

5. Let A be the set of all x ∈ [0, 1] whose decimal expansion contains only two digits
1, 2. Show that A is uncountable and nowhere dense in R.
Solution
That A is uncountable follows via a standard diagonalization argument. To see that
it is nowhere dense, it suffices to show that every open interval contains an open
subinterval disjoint from A. Let J be an open interval. By shrinking J, we can
assume that J = (a, b) where a < b have terminating decimal representations. Choose
c < d such that a < c < d < b and c, d have terminating decimal expansions that end
with 5 and 6 respectively. It is clear that (c, d) ∩ A = ∅.

6. Let f : R → R. Define

osc(f, x) = lim sup({|f (a) − f (b)| : a, b ∈ (x − h, x + h)})

h→0

(a) Show that f is continuous at x iff osc(f, x) = 0.

(b) Show that for each a > 0, the set {x ∈ R : osc(f, x) < a} is open in R.
(c) A subset E ⊆ R is a Gδ (G-delta) set if it is the intersection of some
countable family of open sets. Show that {x ∈ R : f is continuous at x} is a Gδ set.
(d) Show that Q is not a Gδ set.
(e) Show that there is no function f : R → R that is continuous at every rational
and discontinuous at every irrational.
Solution
For each h > 0, define the oscillation of f over (x − h, x + h) by

osc(f, x, h) = sup({|f (a) − f (b)| : a, b ∈ (x − h, x + h)})

and observe that 0 < h1 < h2 =⇒ 0 ≤ osc(f, x) ≤ osc(f, x, h1 ) ≤ osc(f, x, h2 ).

(a) First assume that f is continuous at x. We will show that osc(f, x) ≤ ε for
every ε > 0. It will follows that osc(f, x) = 0. Let ε > 0. Since f is continuous at x,
 we can choose h > 0 such that for every y ∈ (x − h, x + h), |f (x) − f (y)| < ε/2. Now
for every a, b ∈ (x − h, x + h),

|f (a) − f (b)| ≤ |f (a) − f (x)| + |f (b) − f (x)| < ε/2 + ε/2 = ε

It follows that osc(f, x, h) ≤ ε. Hence also osc(f, x) ≤ ε.

Next assume that osc(f, x) = lim osc(f, x, h) = 0. We will show that f is continuous
h→0
at x.
Let ε > 0. Choose δ > 0 such that for every h ≤ δ, osc(f, x, h) < ε. It follows that for
every a, b ∈ (x − δ, x + δ), |f (a) − f (b)| < ε. in particular, for every y ∈ (x − δ, x + δ),
|f (x) − f (y)| < ε.
It follows that f is continuous at x.
(b) Put U = {x ∈ R : osc(f, x) < a}. Let x ∈ U . We will show that x is an
interior point of U . Since osc(f, x) < a, we can fix h? > 0 such that osc(f, x, h? ) < a.
We claim that (x − h? , x + h? ) ⊆ U . To see this, suppose y ∈ (x − h? , x + h? ). Choose
h > 0 such that (y − h, y + h) ⊆ (x − h? , x + h? ). It follows that

osc(f, y) ≤ osc(f, y, h) ≤ osc(f, x, h? ) < a.

Hence y ∈ U and therefore (x − h? , x + h? ) ⊆ U .

(c) Put Df = {x ∈ R : f is continuous at x}. Then by part (a),
\
Df = {x ∈ R : osc(f, x) < 1/n}
n≥1

Since {x ∈ R : osc(f, x) < 1/n} is open in R for each n ≥ 1, it follows that Df is the
intersection of a countable family of open sets. Hence Df is a Gδ set.

\Suppose not and let {Un : n ≥ 1} be a countable family of open sets such that
(d)
Q= Un . Then each Un is open and dense in R (as Q ⊆ Un ). But this implies that
n≥1
\
Un must be uncountable (see the solution to (23)). So Q is not a Gδ set.
n≥1

(e) Follows from (c)+(d).

7. Call X ⊆ R strongly null iff for every sequence hεn : n ≥ 1i of positive

S reals, there
exists a sequence hJn : n ≥ 1i of open intervals such that X ⊆ n≥1 Jn and
length(Jn ) < εn for every n ≥ 1.
(a) Show that the family of all strongly null sets is a σ-ideal on R.
(b) Show that every strongly null set is also Lebesgue null.
(c) Show that the ternary Cantor set is not strongly null.
(d) Is there an uncountable strongly null set?
 8. Let A be a family of subsets of X such that ∅, X ∈ A. Assume f : A → [0, ∞] and
f (∅) = 0. Define m : P(X) → [0, ∞] by
( )
X [
m(A) = inf f (An ) : hAn : n ≥ 1i is a sequence of members of A and A ⊆ An
n≥1 n≥1

Show that m is an outer measure on X.

Solution
m(∅) = 0 because f (∅) = 0. The proof that m is monotone and countable subadditive
is identical to the proof that Lebesgue outer measure is monotone and countable
subadditive.

9. Let F be a σ-algebra on X. Show that F is either finite or uncountable.

Solution
Call A ∈ F an atom iff A 6= ∅ and no proper nonempty subset of A belongs to F. We
have two cases.
Case 1. For some nonempty B ∈ F, there is no atom A ∈ F such that A ⊆ B. In this
case, show that there is an infinite strictly descending sequence
B = B1 ⊇ B2 ⊇ B3 · · · of sets in F. Put Ck = Bk \ Bk+1 . ThenSE = {Bk : k ≥ 1} is an
infinite family of pairwise disjoint sets in F. Now note that |{ C : C ⊆ E}| = |P(N)|.
Case 2. For every B ∈ F, there exists an atom A ∈ F such that A ⊆ B. Let A be the
set of all atoms in F. Show A consists of pairwise disjoint subsets of X. Now check
that If |A| = n, then |F| = 2n and if A is infinite, then |F| ≥ |P(N)|.

10. Suppose X is a nonempty set and A is a collection of subsets of X. Show that there
is a smallest (under inclusion) σ-algebra on X that contains every set in A.
Solution
Let Σ = {F : F is a σ-algebra on X and A ⊆ F}. Then Σ 6= ∅ since F = P(X) ∈ Σ.
Let \
E= F
F ∈Σ

Check that E is as required.

11. Let (X, F, m) be a measure space. Show the following.

(a) Assume A, B ∈ F and A ⊆ B. Then m(A) ≤ m(B).
(b) Assume A, B ∈ F, B ⊆ A and m(B) < ∞. Then m(A \ B) = m(A) − m(B).
[
(c) Suppose An ∈ F and An ⊆ An+1 for every n ≥ 1. Put A = An . Then
n≥1
m(A) = lim m(An ).
n→∞
 \
(d) Suppose An ∈ F and An ⊇ An+1 for every n ≥ 1. Put A = An . Assume
n≥1
m(A1 ) < ∞. Then m(A) = lim m(An ).
n→∞
(e) Suppose An ∈ F. Define
A = {y ∈ X : y ∈ An for infinitely many n ≥ 1}.
X
Show that A ∈ F. If m(An ) < ∞, then m(A) = 0.
n≥1

Solution
(a) m(B) = m(A t (B \ A)) = m(A) + m(B \ A) ≥ m(A).
(b) m(A) = m(B t (A \ B)) = m(B) + m(A \ B). Since m(B) < ∞, we can
subtract it from both sides to get m(A \ B) = m(A) − m(B).
G
(c) Note that A = A1 t (Ak+1 \ Ak ). As m is countably additive, we get
k≥1

∞ n
!
X X
m(A) = m(A1 ) + m(Ak+1 \ Ak ) = lim m(A1 ) + m(Ak+1 \ Ak )
n
k=1 k=1

n
G
Also, An+1 = A1 t (Ak+1 \ Ak ). So
k≥1
n
X
m(An+1 ) = m(A1 ) + m(Ak+1 \ Ak )
k=1

It follows that m(A) = limn m(An+1 ) = limn m(An ).

S
(d) Put Bn = A1 \ An and B = n Bn = A1 \ A. Then Bn ’s are increasing so we
can apply part (c) to get m(B) = limn m(Bn ). Next, apply part (b) to get,
lim m(Bn ) = lim m(A1 \ An ) = m(A1 ) − lim m(An )
n n n

Now use m(A1 ) − m(A) = m(A1 \ A) = m(B).

(e) First check that \ [
A = lim sup An = Ak
n
n≥1 k≥n
\
It follows that A ∈ F. Put Bn = Ak . Then A ⊆ Bn for every n. Hence
k≥n
X
m(A) ≤ m(Bn ) ≤ m(An )
k≥n
X
which goes to 0 as n → ∞ because m(Ak ) < ∞. Hence m(A) = 0.
k≥1
12. Let E ⊆ Rn be Lebesgue measurable. Show the following.
(a) µ(E) = inf{µ(U ) : E ⊆ U and U is open}.
(b) µ(E) = sup{µ(K) : K ⊆ E and K is compact}.
(c) For every ε > 0, there are C ⊆ E ⊆ U such that C is closed, U is open and
µ(U \ E) < ε.
(d) There exists a Gδ -set G such that E ⊆ G and µ(G \ E) = 0.
(e) There exists an Fσ -set F such that F ⊆ E and µ(E \ F ) = 0.
Solution Done in lectures.

13. Show that there is no Lebesgue measurable E ⊆ R such that for every open interval
J, µ(J ∩ E) = 0.5µ(J).
Solution The existence of such a set contradicts Lebesgue density theorem.

14. Show that there exists a Lebesgue measurable E ⊆ R such that for every open
interval J, 0 < µ(J ∩ E) < µ(J).
Solution Done in Tutorial.

15. Let U ⊆ R be open. Show that the characteristic function of U is a Baire class one
function.
Solution If U = R, this is obvious. So assume U 6= R and put C = R \ U . Note that
C is a nonempty closed subset of R. For each n ≥ 1, define hn : R → R by
hn (x) = 1 − min(1, nd(x, C)) where d(x, C) = inf{|x − y| : y ∈ C}. Check that each
hn is continuous and hn ’s pointwise converge to 1U .

16. Show that 1Q ∈ FBaire2 (R) \ FBaire1 (R) where 1Q : R → R is the characteristic
function of the rationals.
Solution Let {rn : n ≥ 1} enumerate Q. Let hn to be the characteristic function of
{rk : 1 ≤ k ≤ n}. Then by previous problem, hn ∈ FBaire1 (R). Hence
1Q = limn hn ∈ FBaire2 (R). To see that 1Q ∈/ FBaire1 (R) use the fact (proved in class)
that a Baire class one function is continuous on a comeager set and 1Q is everywhere
discontinuous.

17. Let f : Rn → R. Show that f is Lebesgue measurable iff there exists a Borel function
g : Rn → R such that f and g agree almost everywhere (this means
µ({x ∈ Rn : f (x) 6= g(x)}) = 0).
Solution First assume that f is Lebesgue measurable. By Lusin’s theorem, for each
m ≥ 1, we can find U T⊆ Rn open such that µ(Um ) < 2−m and f  (Rn \ Um ) is
continuous. Put G = m≥1 Um and note that µ(G) = 0. Define g : R → R as follows.
If x ∈ G, g(x) = 0 and if x ∈ R \ G, then g(x) = f (x). Check that g is as required.
Next assume g : R → R is Borel and W = {x ∈ Rn : f (x) 6= g(x)} has measure zero.
For each Borel B ⊆ R, note that g −1 [B] \ W ⊆ f − [B] ⊆ g −1 [B] ∪ W . It follows that
 f −1 [B] = (g −1 [B] \ W ) ∪ T ) for some T ⊆ W . Since µ(W ) = 0, T is also Lebesgue
measurable. Also g −1 [B] \ W is Lebesgue measurable because g −1 [B] is Borel and W
is Lebesgue measurable. Hence f −1 [B] is Lebesgue measurable and we are done.

18. Suppose f : R → R is Lebesgue measurable and ε > 0. Show that there is a

continuous functions g : R → R such that µ({x : f (x) 6= g(x)}) < ε.
Solution By Lusin’s theorem, we can find an open set U ⊆ R such that µ(U ) < ε
and f  (R \ U ) is continuous. Since R \ U is a closed set, we can extend f  (R \ U )
to a continuous g : R → R. It is clear that g is as required.

19. Show that there is a Lebesgue measurable function f : R → R such that for every
continuous g : R → R, we have µ({x : f (x) 6= g(x)}) > 0.
Solution Take the characteristic function of the set in problem 14.

20. Let f : R → R satisfy f (x + y) = f (x) + f (y) for every x, y ∈ R. Show that if f is

Lebesgue measurable, then f (x) = ax for svery x ∈ R where a = f (1).
Solution By Lusin’s theorem, we can find a compact K ⊆ R such that µ(K) > 0 and
f  K is continuous. By Steinhaus theorem, K − K = {x − y : x, y ∈ E} contains an
open interval around 0. Fix h > 0 such that (−h, h) ⊆ K − K.
We claim that f is continuous at 0. Let hn be any sequence converging to 0. We will
show limn f (hn ) = f (0) = 0. We can assume that each hn ∈ (−h, h). Fix xn , yn ∈ K
such that xn − yn = hn . Since K is compact, WLOG, we can assume for some
x, y ∈ K, limn xn = x and limn yn = y. Now x − y = limn (xn − yn ) = limn hn = 0. So
x = y. By continuity of f  K we get limn f (xn ) = f (x) and limn f (yn ) = f (y). So
limn f (hn ) = limn f (xn − yn ) = limn f (xn ) − limn f (yn ) = f (x) − f (y) = 0. So f is
continuous at 0.
We next claim that f is everywhere continuous. To see this suppose limn xn = x.
Then limn (xn − x) = 0. By continuity of f at 0, we get limn f (xn − x) = 0. Now use
f (xn − x) = f (xn ) − f (x) to conclude limn f (xn ) = f (x).
Finally show that for every rational r ∈ Q, f (r) = rf (1) = ra. Then invoke
continuity of f to conclude f (x) = ax for every x ∈ R.

21. Let (X, F) be a measurable space. Suppose f, g : X → R are F-measurable,

h : R → R is Borel and a, b ∈ R. Then |f |, af + bg, f g and h ◦ f are F-measurable. If
0∈/ range(f ), then 1/f is F-measurable.
Solution
First prove the following. If f1 , f2 , · · · , fn : X → R are F-measurable, then the
function F : X → Rn is defined by

F (x) = (f1 (x), f2 (x), · · · , fn (x))

is (F, Borel(Rn ))-measurable.

22. Let (X, F) be a measurable space. Assume fk : X → R is F-measurable for each
k ≥ 1.
(a) Assume that for each x ∈ X, {fk (x) : k ≥ 1} is bounded from above in R.
Show that lim sup fk is F-measurable.
k

(a) Assume that for each x ∈ X, {fk (x) : k ≥ 1} is bounded from below in R.
Show that lim inf fk is F-measurable.
k
Solution
First show that f : X → R is F-measurable iff for every a ∈ R, f −1 [(∞, a)] ∈ F.
(a) Put f = lim sup fk . Note that
k

f (x) < a ⇐⇒ (∃m ≥ 1)(∃N ≥ 1)(∀k ≥ N )(fk (x) < a − 1/m)

So [ [ \
f −1 [(−∞, a)] = fk−1 [(−∞, a − 1/m)] ∈ F
m≥1 N ≥1 k≥N

for every a ∈ R.
(b) Use the fact that lim inf fk = − lim sup −fk .
k k