Real Analysis Midterm Solutions

Problem 1: (20 points)

(i) Let E be a subset of R that has finite outer measure. Prove that there is a Gδ subset
A of R such that
E ⊆ A and m∗ (E) = m∗ (A).

(ii) Prove that if m∗ (A ∼ E) = 0, then E is measurable.

Solution:

T are open sets {Ok } containing E such that

(i) Since E has finite outer measure, there
m∗ (Ok ) − m∗ (E) < 1/k, therefore A ≡ k Ok is Gδ and contains A such that

m∗ (A) − m∗ (E) = 0.

(ii) If m∗ (A ∼ E) = 0, then A ∼ E is measurable and therefore E = A ∼ (A ∼ E) is

measurable.


Problem 2: (20 points)

(i) State the Lebesgue Dominated Convergence Theorem and the Vitali Convergence The-
orem.

(ii) Deduce the Lebesgue Dominated Convergence Theorem from the Vitali Convergence
Theorem.

Solution:

(i) Lebesgue Dominated Convergence: Let {fn } be a sequence extended real-valued

measureable functions over X ⊆ R such that for some f , {fn } → f pointwise a.e. in
X. If there exists a nonnegtaive and integrable g such that |fn | ≤ g for all n, then f
is integrable and Z Z
lim fn = f.
n→∞ X X

1
 Vitali Convegence: Let {fn } be a sequence extended real-valued measureable func-
tions over X ⊆ R such that for some f , {fn } → f pointwise a.e. in X. If {fn } are
uniformly integrable, i.e. there exists a δ > 0 such that if m(A) < δ then
Z
|fn | < 
A

for all n, and tight, i.e. there exists a measurable set E ⊂ X for which
Z
|fn | < 
X∼E

for all n, then f is integrable over X and

Z Z
lim fn = f.
n→∞ X X

(ii) Suppose {fn } → {f } pointwise a.e. and suppose g dominates {fn } and is integrable. It
is not hard to see that {fn } are uniformly integrable and tight since by the itegrability
of g we may always find δ so that if m(A) < δ then
Z Z
|fn | ≤ g<
A A

and we can also find a set E of finite measure such that

Z Z
|fn | ≤ g < .
X∼E X∼E

Therefore we may apply Vitali Convergence.


Problem 3: (20 points)

(i) Let f : (0, 1] → R be continuous. Suppose the following finite limit exists:
Z 1
lim f (x) dx.
n→∞ 1/n

Is f Lebesgue integrable over (0, 1]?

(ii) If f is nonnegative, is f Lebesgue integrable over (0, 1]?

Solution:

2
 (i) This is not true, consider the functions f (x) = sin(2π/x)/x2 . Since f is continuous we
may change variables in the Riemann integral to find that
sin 2π
Z 1
Z n
x
lim dx = lim sin(2πx) dx = 0,
n→∞ 1
n
x2 n→∞ 1

However, f is not Lebesgue intergrable, since

Z 1 Z n
|f (x)| dx = | sin(2πx)|dx = 2n → ∞.
1
n
1

(ii) If f is nonnegative, then fn = f χ[1/n,1] is nonnegative and {fn } → f in a monotone

increasing fashion. Therefore by monotone convergence,
Z 1 Z 1
lim f (x) dx = f dx < ∞,
n→∞ 1
0
n

and so f is integrable.

Problem 4: (20 points) A real valued function on R is said to be Lipschitz provided there
is a constant c > 0 such that
|f (u) − f (v)| ≤ c|u − v| for all u, v.
(i) Show that a Lipschitz functions takes sets of measure zero to sets of measure zero.
(ii) Show that a Lipschitz function on a closed, bounded interval is of bounded variation.
Solution:
(i) Since f is Lipschitz, it maps intervals to intervals, and so for any open interval I
`(f (I)) = sup |f (x) − f (y)| ≤ c · sup |x − y| = `(I).
x,y∈I x,y∈I

Furthermore if {Ii }∞ ∞
i=1 is an open cover of a measureable set A, then {f (Ii )/2i+1 }i=1 is
an open cover of f (A), where f (I) is an  thickening of I given by
[
f (I) = (x − , x + ).
x∈f (I)

It follows that for any  > 0 we can choose the open covering so that
∞ ∞ 
X X 
m(f (A)) ≤ `(f (Ii )/2 ) ≤ c ·
i+1 `(Ii ) + i < c · (m(A) + 2) .
i=1 i=1
2

Taking  → 0 we conclude
m(f (A)) ≤ c · m(A).
Therefore m(f (A)) = 0 whenever m(A) = 0.

3
 (ii) Let {xi }N
i=0 be the end points of a finite partition PI of the interval I = [a, b]. If f is
Lipschitz, then
N
X N
X
|f (xi ) − f (xi−1 )| ≤ c · |xi − xi−1 | = c · (b − a) < ∞
i=1 i=1

Since this bound is independent of the partition, we conclude that f is bounded vari-
ation.


Problem 5: (20 points)

(i) Define what it means for a sequence {fn : R → R} of integrable functions to be uni-
formly integrable.

(ii) Show that a sequence of nonnegative, integrable functions {fn : R → R} is uniformly

integrable and tight if and only if the following property holds:

for each  > 0,there is a constant C such that for all n

Z
|fn | < .
{x∈R | fn (x)>c}

Solution:

(i) A sequence {fn } is uniformly integrable if for every  > 0 there is a δ > 0 so that if A
is measurable and satisfies m(A) < δ, then
Z
|fn | ≤ 
A

uniformly in n.

(ii) Suppose {fn } is uniformly integrable. Let  > 0 and choose δ > 0 as in the uniformly
integrability condition. The claim is that there is a c0 such that for all n

m({R | fn ≥ c0 }) < δ.

Using the claim it easily follows by uniform integrability that

Z
fn < ,
{R|fn >c0 }

uniformly in n. To see the claim, suppose not, that is suppose for every c > 0 there
exists an n0 such that
m({R | fn0 ≥ c}) > δ.

4
If this is indeed the case, then upon choosing c > 2/δ we may find a set An0 ⊆ {R |
fn0 ≥ c} such that
δ
≤ m(An0 ) < δ.
2
By Chebyshev and uniform integrability of {fn }, we conclude that
Z
cδ
< ≤ c · m(An0 ) ≤ fn < ,
2 An

which is a contradiction.
R
For the converse, choose c so that {R|fn >c}
fn < /2. Then for any measurable set
A ⊆ R we may write
Z Z Z
fn ≤ fn + fn
A {x∈R|fn >c} {x∈A|fn ≤c}

≤ /2 + c · m(A)

uniformly in n. Choosing m(A) < /2c, we conclude

Z
fn < ,
A

uniformly in n.