Assignment 1

MTL874, Prof. Sivananthan Sampath

Ryan Azim Shaikh, 2022MT11923

March, 2025

Question 1. Let (X, M, µ) be a complete measure space and f, g : X → R if f = g pointwise µ-a.e. and f is
measurable, then g is measurable.

A1. Consider O ⊆ R open. Note,

g −1 (O) = {x ∈ X|g(x) ∈ O} = {x ∈ X|f (x) = g(x), f (x) ∈ O} ∪ {x ∈ X|f (x) ̸= g(x), g(x) ∈ O}

where the first set on the RHS is measurable as it is the intersection of two measurable sets, and the second set is a
subset of a zero measure set; thus, the completeness of (X, M, µ) completes the proof.

Question 2. Show that every Reimann integrable function defined on [a, b] is a measurable function on [a, b].

A2. Let f be Reimann integrable on [a, b]. So,

inf U (f, P ) = sup L(f, P )

P ∈P P ∈P

where, P is the set of all finite partitions on [a, b]. And for partition P = {a = x0 < x1 < · · · < xn = b},
n−1
X Z b n−1
X Z b
U (f, P ) = (xi+1 − xi ) sup f (x) = φf P and L(f, P ) = (xi+1 − xi ) inf f (x) = ψf P , where,
x∈(xi ,xi+1 ) a x∈(xi ,xi+1 ) a
i=0 i=0
φf P and ψf P are step functions defined by
n
X
φf P = χ(xi−1 ,xi ) sup f (x),
i=1 x∈(xi−1 ,xi )

n
X
ψf P = χ(xi−1 ,xi ) inf f (x).
x∈(xi−1 ,xi )
i=1

Step functions are Lebesque measurable since all intervals are Lebesque measurable. Also note by definition of φf P
and ψf P
ψf P ≤ f ≤ φf P .
Except maybe at finitely many points (A measure zero set). Thus, since f is Reimann integrable, we see that there
is a sequence of partitions {Pn } such that Pn+1 is a refinement of Pn for all n and
Z b Z b Z b
lim φf Pn = f (x)dx = lim ψf Pn .
n→∞ a a n→∞ a

1
This gives us sequences of step functions {φn } (decreasing) and {ψn } (increasing) such that ψn ≤ f ≤ φn a.e. on
(a, b) and
Z b Z b Z b
lim φn = f (x)dx = lim ψn .
n→∞ a a n→∞ a
Rb
So limn→∞ a
[φn − ψn ] = 0. Where, φn − ψn is nonnegative a.e. on [a, b]. Now for α > 0, let

Eα,n = {x ∈ [a, b] |φn (x) − ψn (x)| ≥ α}.

Since φn is decreasing and ψn is increasing {Eα,n }n∈N is a decreasing sequence of sets. By the Chebyschev inequality,
Z b
1
m(Eα,n ) ≤ [φn − ψn ].
α a

and so
lim m(Eα,n ) = 0.
n→∞

which implies
∞
!
\
m Eα,n = 0.
n=1

Note,
∞
[ ∞ \
[ ∞
{x ∈ [a, b] lim [φn (x) − ψn (x)] ̸= 0} = {x ∈ [a, b] lim |φn (x) − ψn (x)| ≥ 1/m} = E1/m,n .
n→∞ n→∞
m=1 m=1 n=1

Thus m({x ∈ [a, b] limn→∞ [φn (x) − ψn (x)] ̸= 0}) = 0. Since ψn ≤ f ≤ φn a.e. on [a, b], let An = {x ∈ [a, b] : ψn >
S∞
f or f > φn }. Then n=1 An is a measure zero set. Thus,
   
m {x ∈ [a, b] lim [φn (x) − f (x)] ̸= 0} = m {x ∈ [a, b] lim [f (x) − ψn (x)] ̸= 0} = 0.
n→∞ n→∞

Thus ψn → f and φn → f pointwise a.e. on [a, b], which shows that f is measurable on [a, b], since pointwise limits
of measurable functions are measurable.

1
k 3/2 x
Z
Question 3. Evaluate lim dx.
k→∞ 0 1 + k 2 x2
A3.
1 1
k 3/2 x k 3/2 2k 2 x
Z Z
lim dx = lim dx
k→∞ 0 1 + k 2 x2 k→∞ 2k 2 0 1 + k 2 x2
2
k Z
1 1
= lim
1/2
dt substituting t = k 2 x2
k→∞ 2k 0 1 + t
ln(1 + k 2 )
= lim
k→∞ 2k 1/2
= 0.

2
Question 4. Let f ∈ Lp (R), 1 ≤ p < ∞ and for each x ∈ R, define τx f (t) = f (t − x). Show that,

lim ||τx f − τx0 f ||p = 0.

x→x0

A4. We need to show Z

lim |f (t − x) − f (t − x0 )|p dt = 0.
x→x0 R

But since the lebesque measure is translation invariant, we have R |f |p = R |τx f |p for any x ∈ R. Without loss of
R R

generality, we may assume x0 = 0. Thus, we need to show

Z
lim |f (t − x) − f (t)|p dt = 0.
x→0 R

We show this result first for f ∈ Cc (R). If f is continuous limx→0 [f (t − x) − f (t)] = 0 on R. Further, since f is zero
outside a compact set, say K, we have f (x) is uniformly continuous on K, and since f = 0 everywhere outside, f is
uniformly continuous everywhere. Thus for ε > 0, there exists a δ > 0, such that |x| < δ implies |f (t−x)−f (t)| < ε1/p
everywhere. Since both f (t − x) and f (t) are zero outside the compact set M = K ∪ (K − δ) ∪ (K + δ), we have a
finite measure set µ(M ) ≤ 3µ(K), such that |f (t − x) − f (t)|p is zero outside M for all |x| < δ. Thus
Z
|f (t − x) − f (t)|p dt ≤ 3µ(K)ε.
R

Since K is fixed and ε can be made arbitrarily small by picking x from smaller neighbourhoods of zero we have the
required limit.
For the general case with f ∈ Lp (R), since Cc (R) = Lp (R) in Lp (R), for ε > 0 we have a function gε ∈ Cc (R) such
that ||f − gε ||p < ε. Since the norm || · ||p is translation invariant (||f ||p = ||τx f ||p ) we have

||τx f − τx0 f ||p ≤ ||τx gε − τx0 gε ||p + 2||f − gε ||p .

Each term on the right can be made arbitrarily small, ||f − gε ||p independently of x and ||τx gε − τx0 gε ||p as x → 0.
Thus, we have the required limit.

Question 5. Assume µ(X) < ∞. Suppose f ∈ Lp (µ) ∩ L∞ (µ) for some 1 ≤ p < ∞. Show that,

||f ||∞ = lim ||f ||q .

q→∞

A5. Let M = ||f ||∞ . Then by definition of least essential supremum |f | ≤ M a.e. on X and for any ε > 0,
Aε = {x ∈ X |f (x)| > M − ε} has non zero measure. Also χAε · (M − ε) ≤ |f | a.e. on X. Thus
Z 1/q Z 1/q Z 1/q
1/q q q q
(M − ε)µ(Aε ) = (M − ε) ≤ |f | ≤ M = M µ(X)1/q .
Aε X X

Now both µ(Aε )1/q and µ(X)1/q converge to 1, thus for q sufficiently large (say q > Q2ε ), µ(Aε )1/q > (M −2ε)/(M −ε)
and µ(X) < (M + 2ε)/M . Thus for q ≥ Q2ε
Z 1/q
q
(M − 2ε) < |f | < (M + 2ε).
X

3
Which gives the required limit as ε → 0.

Question 6. Prove or disprove that if f ∈ L1 (R) ∩ C(R), then lim f (x) = 0.

|x|→∞

A6. Define the triangle hat function h( a, b, c) for a < b < c, as follows

0
 if x ∈ (−∞, a] ∪ (c, ∞)
b−x
h(a,b,c) (x) = 1 − b−a if x ∈ (a, b]
1 − x−b

if x ∈ (b, c]

c−b

Note that h is nonnegative continuous and

Z
1
h(a,b,c) = (c − a).
R 2
P∞
Define the sequence of functions gn = h(n−1/2n ,n,n+1/2n ) and f as the pointwise limit of the series n=1 gn .
Consider a bounded interval IM = (−M, M ), then gn = 0 on IM for all n > M + 1. Thus, only finitely many gn
are non-zero anywhere in IM . Thus, f is continuous on IM . This is true for any bounded interval in R. Thus, f
is continuous on R, since for any neighborhood of any point in R at most finitely, many gn are possibly non-zero.
Further limx→∞ f (x) does not exist since f (x) = 1 for all x ∈ N and f (x) = 0 for all x ∈ N + 1/2.
If we show f ∈ L1 (R), we will be done. If we define An = (n − 1/2n , n + 1/2n ) and A = x ∈ R : f (x) = 0, by
countable additivity of domains of integration
Z X∞ Z Z X∞
f= f+ f= 1/2n = 1.
R n=1 An A n=1

This disproves the statement.

Question 7. Let f ∈ L2 (Rd ). Suppose that f ϕ dµ = 0 for all ϕ ∈ Cc (Rd ). Then show that f = 0 a.e.
R
Rd

A7. Assuming µ is a measure obtained using the Reisz representation theorem (Rd is an LCHS). We know
Cc (R ) = L2 (R). By Hölder’s inequality
d

Z Z Z 1/2 Z 1/2
f (f − ϕ) ≤ |f (f − ϕ)| ≤ f2 (f − ϕ)2

1/2
and by denseness of Cc (Rd ) in L2 (Rd ) we have (f − ϕ)2
R
can be made arbitrarily small. Thus
Z Z Z Z
2
f = f ϕ + f (f − ϕ) = f (f − ϕ) < ε

f 2 = 0, which gives f = 0 a.e. on Rd .

R
for arbitrarily small ε > 0. Thus

Question 8. State and prove Chebyshev’s inequality.

A8. The Chebyshev’s Inequality states that given a measure space (X, M, µ) and a measurable extended real-valued
function f on this space Z
1
µ({x ∈ X |f (x)| ≥ α}) ≤ · |f |.
α X

4
for any α > 0.
The case where f is not integrable over X is trivial since the right-hand side of the inequality becomes ∞. For the
other case, let A = {x ∈ X |f (x)| ≥ α}. A is measurable since f is measurable and [α, ∞] is borel. Now note
α · χA ≤ |f | on X thus Z Z
α · χA ≤ |f |.
X X
Which gives Z
α · µ(A) ≤ |f |,
X
which is the required inequality.

Question 9. Let f be a measurable function on (X, M, µ). Define the distribution function λf : (0, ∞) → [0, ∞] by
λf (α) = µ({x ∈ X| |f (x)| > α}).
Show that if f ∈ Lp (µ), 0 < p < ∞, then
Z Z ∞
|f |p dµ = p αp−1 λf (α) dα.
X 0

A9. Since f ∈ LP ⊆ L∞ , f is essentially bounded. Then λf (α) = 0 for all α >= ||f ||∞ = M (say). Thus
Z ∞ Z M
p−1
α λf (α) dα = αp−1 λf (α) dα.
0 0
p−1
Note that right hand side is Reimann Integrable since x is Reimann integrable and λf is monotone.
Now consider a partition of [0, M ], P = {0 = α0 < α1 < · · · < αn = M }. And define ϕP and ψP as
ϕP (x) = αi if αi < |f (x)| ≤ αi+1
ψP (x) = αi+1 if αi < |f (x)| ≤ αi+1
and
ϕP (x) = ψP (x) = 0 if f (x) = 0.
Thus we have
ϕP ≤ f ≤ ψP
a.e on X. Now
Z n
X
ϕpP = αip µ({x ∈ X|ϕP (x) = αi })
X i=0
n
X
= αip [µ({x ∈ X|ϕP (x) ≥ αi }) − µ({x ∈ X|ϕP (x) ≥ αi+1 })]
i=0
n
X
= (αip − αi−1
p
)µ({x ∈ X|ϕP (x) ≥ αi })
i=1
n
X
= (αip − αi−1
p
)µ({x ∈ X|f (x) > αi })
i=1
n
X
= (αip − αi−1
p
)λf (αi ). (1)
i=1

5
And similarly,
Z n
X
ψPp = αip µ({x ∈ X|ψP (x) = αi })
X i=0
n
X
= αip [µ({x ∈ X|ψP (x) ≥ αi }) − µ({x ∈ X|ψP (x) ≥ αi+1 })]
i=0
n
X
= (αip − αi−1
p
)µ({x ∈ X|ψP (x) ≥ αi })
i=1
n
X
= (αip − αi−1
p
)µ({x ∈ X|f (x) > αi−1 })
i=1
n
X
= (αip − αi−1
p
)λf (αi−1 ). (2)
i=1

For 0 < p ≤ 1 (xp is concave) we have

p−1
p
αip − αi−1
pαi−1 ≥ ≥ pαip−1
αi − αi−1
and λf (αi−1 ) ≥ λf (αi ). Which bounds 1 as follows
n
X n
X n
X
p(αi − αi−1 ) αip−1 λf (αi ) ≤ (αip − αi−1
p
)λf (αi ) ≤ p(αi − αi−1 ) p−1
αi−1 λf (αi−1 )
i=1 i=1 i=1

which gives
n
X Z n
X
p(αi − αi−1 ) αip−1 λf (αi ) ≤ ϕpP ≤ p(αi − αi−1 ) p−1
αi−1 λf (αi−1 )
i=1 X i=1

Consider a sequence of partitions Pm such that the width, maxni=1 (αi(m)

− αi−1(m) ) → 0 as m → ∞ and Pm+1
R M p−1
is a refinement of Pm . Both bounding limits are Reimann sums to p 0 α λf (α)dα and hence converge to
RM
p 0 αp−1 λf (α)dα. Also if Q is a refinement of P

ϕP ≤ ϕQ ≤ |f |

further ϕPm → f pointwise since for x ∈ [αi−1(m) , αi(m) ], ||f (x)| − ϕPm (x)| ≤ αi(m) − αi−1(m) → 0 as n → ∞. By
Lebesque Dominated Convergence and sandwich theorem, we have
Z Z M Z ∞
|f |p = p αp−1 λf (α) dα = p αp−1 λf (α) dα.
X 0 0

As required. Similarly for 1 < p < ∞ We have,

n
X n
X
p−1
p(αi − αi−1 ) αi−1 λf (αi−1 ) ≤ (αip − αi−1
p
)λf (αi−1 )
i=1 i=1

and
n
X n
X
(αip − αi−1
p
)λf (αi ) ≤ p(αi − αi−1 ) αip−1 λf (αi )
i=1 i=1

6
Or by 1 and 2,
n
X Z
p−1
p(αi − αi−1 ) αi−1 λf (αi−1 ) ≤ ψPp
i=1 X
Z n
X
ϕpP ≤ p(αi − αi−1 ) αip−1 λf (αi )
X i=1

Again, considering a similar monotonically refining sequence of partitions Pm with width going to zero, we see that

ψPm → |f | ← ϕPm

and thus, taking limit on the inequalities gives us

Z M Z Z M
p−1 p
p α λf α dα ≤ |f | ≤ p αp−1 λf α dα.
0 X 0

Which gives the required equality.

Question 10. Let (X, A) be a measurable space and the space M (X, A, C) be the collection of all complex measures
on (X, A). Show that M (X, A, C) is complete under the total variation norm.

A10. Consider a Cauchy sequence of complex measures in X, {µn }n∈N . Then for ε > 0, pick Nε ∈ N such that
n, m ≥ Nε implies
||µn − µm || = |µn − µm |(X) < ε.
For any A ∈ A, |µn −µm |(A) ≤ |µn −µm |(X) since µn −µm is a complex measure and thus |µn −µm | (total variation)
is a positive measure. Thus we have,

|µn (A) − µm (A)| ≤ |µn − µm |(A) ≤ |µn − µn |(X) < ε.

Thus, {µn (A)}n∈N is Cauchy in C (viz complete), which implies it is convergent, define

µ(A) = lim µn (A) ∀A ∈ A.

n→∞

If we show µ is a complex measure on (X, A), we will be done.

µ(∅) = lim µn (∅) = lim 0 = 0.

n→∞ n→∞

Now we only need countable additivity. Consider a countable disjoint collection {En }n∈N from A then
∞
[ ∞
[
µ( Ek ) = lim µn ( Ek ).
n→∞
k=1 k=1
∞
m
!!
X [
= lim µn (Ek ) + µn Ek
n→∞
k=1 k=m+1
∞
m
!
X [
= µ(Ek ) + lim µn Ek .
n→∞
k=1 k=m+1
∞ ∞ ∞
m
! !
[ X [ [
⇒ µ( Ek ) − µ(Ek ) = lim µn Ek = µ Ek (3)
n→∞
k=1 k=1 k=m+1 k=m+1

7
Now for any ε > 0 we have Nε ∈ N such that for l, n ≥ Nε ,

|µn (A) − µl (A)| < ε/2 ∀A ∈ A.

and since {µn (A)}n∈N converges in C we have some Mε,A such that for n ≥ Mε,A

|µn (A) − µ(A)| < ε/2.

So for some l ≥ max(Nε , Mε,A ) and n ≥ Nε ,

|µn (A) − µl (A)| < ε/2 ∀A ∈ A.

and
|µl (A) − µ(A)| < ε/2.
which gives
|µn (A) − µ(A)| < ε.
Note that the choice of n does not depend on A. So now for n ≥ Nε/2
∞ ∞
! !
[ [
µ Ek − µn Ek < ε/2
k=m+1 k=m+1
P∞ S∞
irrespective of m, also since k=1 µn (Ek ) = µn ( k=1 Ek ) ∈ C, we have that the tail of the series goes to 0, so for m
sufficiently large, (say m ≥ Lε,n )
∞
!
[
µn Ek < ε/2
k=m+1

Thus for m ≥ Lε,n

∞
!
[
µ Ek <ε
k=m+1

which by 3 completes the proof.