Universiteit Utrecht Mathematisch Instituut 3584 CD Utrecht

Measure and Integration: Solutions Final 2013-14

(1) Consider the measure space (R, B(R), λ), where B(R) is the Borel σ-algebra, and λ Lebesgue
measure. Determine the value of
Z
x x
lim (1 + )−n (1 − sin ) dλ(x).
n→∞ (0,n) n n
(2 pts)

Solution: Let un (x) = 1(0,n) (1 + nx )−n (1 − sin nx ). The positive sequence (1 + nx )−n n decreases

to e−x 1(0,∞) and the sequence 1 − sin nx n is bounded from below by 0 and from above by 2 and

converges to 1 as n → ∞. Thus, limn→∞ un (x) = 1(0,∞) e−x , and 0 ≤ un (x) ≤ 2(1 + x2 )−2 1(0,∞)
for n ≥ 2 and all x ∈ R. Since the function 2(1 + x2 )−2 1(0,∞) is measurable, non-negative and the
improper Riemann integrable on (0, ∞) exists, it follows that it is Lebesgue integrable on (0, ∞).
By Lebesgue Dominated Convergence Theorem (and taking the limit for n ≥ 2), we have
Z Z
x x
lim (1 + )−n (1 − sin ) dλ(x) = lim un (x)dλ(x)
n→∞ (0,n) n n n→∞
Z Z ∞
= 1(0,∞) e−x dλ(x) = e−x dx = 1.
0

(2) Let (X, F, µ) be a finite measure space. Assume f ∈ L2 (µ) satisfies 0 < ||f ||2 < ∞, and let
A = {x ∈ X : f (x) 6= 0}. Show that
( f dµ)2
R
µ(A) ≥ R 2 .
f dµ
(1.5 pts)

Solution: Since f = 0 on Ac , we have f dµ = f 1A dµ. Since µ is a finite measure and

R R

(1A )2 = 1A , then
||1A ||2 = (µ(A))1/2 < ∞.
Thus, 1A ∈ L2 (µ) and by Hölder’s inequality
Z
f dµ ≤ ||f ||2 ||1A ||2 = |f ||2 (µ(A))1/2 .

Squaring both sides and dividing by

Z
||f ||22 = f 2 dµ (> 0),

we get
( f dµ)2
R
µ(A) ≥ R 2 .
f dµ

(3) Let E = {(x, y) : y < x < 1, , 0 < y < 1}. We consider on E the restriction of the product Borel
σ-algebra, and the restriction of the product Lebesgue measure λ × λ. Let f : E → R be given
by f (x, y) = x−3/2 cos( πy
2x ).
(a) Show that f is λ × λ integrable on
R E. (0.5 pt)
(b) Define F : (0, 1) → R by F (y) = (y,1) x−3/2 cos( πy
2x ) dλ(x). Determine the value of
Z
F (y) dλ(y).

(2 pts)
1
2

Solution (a) : Notice that f is continuous, and hence measurable. Furthermore, |f (x, y)| ≤
x−3/2 . The function g(x, y) = x−3/2 is non-negative and measurable on E, hence by Tonelli’s
Theorem,
Z Z
|f (x, y)| d(λ × λ)(x, y) ≤ g(x, y) d(λ × λ)(x, y)
E E
Z 1 Z x
= x−3/2 dy dx
0 0
Z 1
= x−1/2 dx = 2.
0
Notice that the integrands are Riemann integrable, hence the Riemann integral equals the
Lebesgue integral. This shows that f is λ × λ integrable on E.

Solution (b) : By Fubini’s Theorem

Z Z Z Z
f (x, y) dλ(x) dλ(y) = f (x, y) dλ(y) dλ(x).

Notice that for each fixed 0 < x < 1, the function f (x, y) is Riemann-integrable in y on the
interval (0, x) and Z x
πy 2
x−3/2 cos( ) dy = x−1/2 ,
0 2x π
and the function π2 x−1/2 is Riemann-integrable in x on the interval (0, 1), and
Z 1
2 −1/2 4
x dx = .
0 π π
Thus,
Z Z Z Z 1Z x
πy 4
F (y) dλ(y) = f (x, y) dλ(x) dλ(y) = x−3/2 cos( ) dy dx = .
0 0 2x π

(4) Let 1 ≤ p < ∞, and suppose (X, A, µ) is a measure space. Let (fn )n ∈ Lp (µ) be a sequence
converging to f in Lp i.e. limn→∞ ||fn − f ||p = 0.
(a) Show that Z Z
|f | dµ ≤ lim inf |fn |p dµ.
p
n→∞
(1 pt)
(b) Show that lim np µ({|f | > n}) = 0. (1 pt)
n→∞

Solution (a): This is a simple consquence of the triangle inequality applied to the Lp -norm and
in fact the lim inf can be replaced by lim and the inequality by equality, namely
|||fn ||p − ||f ||p | ≤ ||fn − f ||p .
Taking limits, we get the desired result. (Remark: if we replace Lp -convergence by convergence
in measure, then the inequality is really needed).

Solution (b): Note that f ∈ Lp (µ) and hence by Corollary 10.13,

µ({|f |p = ∞}) = µ({|f | = ∞}) = 0.
Thus,
lim |f |p 1{|f |>n} = |f |p 1{|f |=∞} = 0 µ a.e.
n→∞
Since for each n, |f |p 1{|f |>n} ≤ |f |p and |f |p ∈ L1 (µ), we have by Lebesgue Dominated Conver-
gence Theorem, Z
lim |f |p 1{|f |>n} dµ = 0.
n→∞
Now, Z Z
np µ({|f | > n}) = np 1{|f |>n} dµ ≤ |f |p 1{|f |>n} dµ,
and from the above we get lim np µ({|f | > n}) = 0.
n→∞
 3

(5) Let (X, A, µ) be a finite measure space and fn , f ∈ M(A), n ≥ 1. Show that fn converges to f
|fn − f |
Z
in µ measure if and only if lim dµ = 0. (2 pts)
n→∞ 1 + |fn − f |
|fn − f |
Solution: First note that ≤ 1 for all n ≥ 1, and since µ(X) < ∞ we have 1 ∈ L1 (µ).
1 + |fn − f |
µ
Now assume that fn −
→ f , and let , δ > 0 , then there exists N such that
µ({x ∈ X : |fn (x) − f (x)| > δ}) < , for all n ≥ N.
Let A = {x ∈ X : |fn (x) − f (x)| > δ}, then for all n ≥ N
|fn − f | |fn − f | |fn − f |
Z Z Z Z Z
dµ = dµ + dµ ≤ 1 dµ + δ dµ.
1 + |fn − f | A 1 + |fn − f | Ac 1 + |fn − f | A Ac
Thus, for all n ≥ N
|fn − f |
Z
dµ ≤  + δµ(X).
1 + |fn − f |
|fn − f |
Z
Thus, lim dµ = 0.
n→∞ 1 + |fn − f |
|fn − f |
Z
Conversely, assume lim dµ = 0, and let  > 0. There exists N such that
n→∞ 1 + |fn − f |
|fn − f |
Z
dµ < 2 /(1 + ), for all n ≥ N.
1 + |fn − f |
Observe first that
|fn − f | 
|fn − f | >  ⇐⇒ > .
1 + |fn − f | 1+
Thus, by Markov Inequality, we have for all n ≥ N
|fn − f | |fn − f |
Z
 1+
µ({x ∈ X : |fn (x) − f (x)| > }) = µ({x ∈ X : > }) ≤ dµ < .
1 + |fn − f | 1+  1 + |fn − f |
µ
Thus, fn −
→ f.