Solutions Exam 2 — May 2, 2018

1. (a) (10 points) State the definition of a measurable function f : E → R.

A function f : E → R is measurable if and only if E is a measurable subset and for each

α ∈ R, f −1 ((−∞, α)) = {x ∈ E : f (x) < α} is a measurable set.

(b) (15 points) Let f : R → R be measurable and g : R → R be continuous. Prove that g ◦ f is

measurable.

As g is continuous Uα = g −1 ((−∞, α)) = Uα is an open set. A basic property of a measurable

function f is that for any open set U , f −1 (U ) is measurable, hence (g ◦ f )−1 ((−∞, α)) =
f −1 (Uα ) is measurable and so g ◦ f is also measurable. To see why this basic property holds,
we observe that for any open interval, I, f −1 (I) is measurable. Indeed, if I = (a, b) and
b < ∞, then f −1 (I) ∈ M by definition when a = −∞ and when a ∈ R
∞
[ 1
f −1 (I) = f −1 ((−∞, b))\ f −1 ((−∞, a − ))
n
n=1

and so the measurablity of f −1 (I) follows from the definition and the σ-algebra property of
M. When b = ∞ one has
∞
[
−1
f (I) = f −1 (I ∩ (−∞, n))
n=1

and each f −1 (I ∩ (−∞, n)) ∈ M by the previous observation and so f −1 (I) ∈ M by σ-algebra
property. As any open set, U , is the countable union of open intervals, i.e. U = ∪∞ n=1 In for
open intervals In . The fact that f −1 (U ) ∈ M follows from the σ-algebra property of M.
Math 416, Spring 2018 Solutions to Exam 2 — May 2, 2018 Page 2 of 6

2. Let f : R → [0, ∞] be a non-negative measurable function.

(a) (10 points) State the definition of the Lebesgue integral of f .

Given f , the Lebesgue integral of f is defined to by

Z Z 
f = sup φ : 0 ≤ φ ≤ f, and φ is a simple function .

Pn
Here φ is simple means that φ = i=1 ai χEi where ai ∈ R and Ei ∈ M and
Z n
Z X n
X
φ= ai χEi = ai m(Ei ).
i=1 i=1

(b) (10 points) Directly using the definition, show that

 1
x2
x 6= 0
f (x) =
∞ x = 0,

is not Lebesgue integrable.

R
Recall, f is Lebesgue integrable if and only if f < ∞. Consider the sequence of simple
functions φn = n2 χ(0, 1 ) . Clearly, 0 ≤ φn ≤ f and one has
n
Z
1
φn = n2 m((0, )) = n
n
R R
Hence, f ≥ n for all n ≥ 1 and so f = ∞. That is, f is not Lebesgue integrable.
Math 416, Spring 2018 Solutions to Exam 2 — May 2, 2018 Page 3 of 6

1
(c) (10 points) Show that f (x) = 1+x2
is Lebesgue integrable, i.e., is in L1 (R).

First observe that f is continuous and hence measurable. Let fn (x) = χ[−n,n] f (x). Clearly
each fn is measurable and the sequence (fn ) is monotone and limn→∞ fn = f pointwise.
Hence, by the monotone convergence theorem
Z Z Z n
1
f = lim fn = lim .
n→∞ n→∞ −n 1 + x2

1 d 1
As 1+x 2 is continuous and dx arctan(x) = 1+x2 one can use the fundamental theorem of
calculus to see that Z n
1
2
= arctan(n) − arctan(−n).
−n 1 + x
Hence, Z Z n
1
f = lim = lim (arctan(n) − arctan(−n)) = π < ∞
n→∞ −n 1 + x2 n→∞
and so f is Lebesgue integrable.
Math 416, Spring 2018 Solutions to Exam 2 — May 2, 2018 Page 4 of 6

3. Let (fn ) be a sequence of Lebesgue integrable functions (i.e., fn ∈ L1 (R)) that converge pointwise to
the zero function.
(a) (10 points) Show, by example, that it is possible for the fn to not converge in L1 (R).

Let fn (x) = χ[n,n+1] . For each x, asR long as n ≥ |x|

R + 1, one has fn (x) = 0. Hence,
limn→∞ fn = 0 pointwise. However, |fn − fn+2 | = fn + fn+2 = 2 and so the sequence
(fn ) is not Cauchy in L1 (R) and hence cannot converge to the zero function or to any other
function.

1
(b) (10 points) Suppose that, in addition, the fn satisfy |fn (x)| ≤ 1+x2
for a.e. x ∈ R. Show that the
fn converge in L1 (R) to the zero function.

1
In a previous problem we established that g(x) = 1+x 2 is Lebesgue integrable. Hence, by the
Lebesgue dominated convergence theorem one has
Z Z Z
lim |fn | = lim |fn | = 0 = 0.
n→∞ n→∞

That is, fn converges to 0 in L1 (R).

Math 416, Spring 2018 Solutions to Exam 2 — May 2, 2018 Page 5 of 6

4. (a) (15 points) Show that there is a continuous function g : [−π, π] → R with Fourier series given by
∞
X
2−k cos kx.
k=1

1
Rπ
What is the value of kgk22 = π −π g 2 (x) dx?

Notice that on [−π, π] one has,

sup |2−k cos kx| ≤ 2−k

[−π,π]

P∞ −k
and the series k=1 2 = 1 < ∞ is summable. Hence, by the Weierstrass M -test, the
functions
n
X
sn (x) = 2−k cos kx ∈ Tn
k=1

are continuous and satisfy sn → g uniformly for some continuous function g : [−π, π] → R.
We claim g has the desired Fourier series. To see this, we observe that for g ≥ k

1 π
Z
ak (sn ) = sn (x) cos(kx)dx = 2−k
π −π

Hence, for k ≥ 1

1 π 1 π 1 π
Z Z Z
ak (g) = g(x) cos(kx)dx = ( lim sn (x)) cos(kx)dx = lim sn (x) cos(kx)dx
π −π π −π n→∞ n→∞ π −π

where we are allowed to interchange limits and integration as the convergence is uniform.
Hence,
ak (g) = 2−k .
Similarly, one sees bk (g) = 0 for all k and a0 (g) = 0. That is, g has the desired Fourier series.
Finally, by using Parseval’s identity we can compute that
∞ ∞ ∞
a0 (g)2 X
X −k 2 X −k 1
kgk22 = + ak (g)2 + bk (g)2 = (2 ) = 4 = .
2 3
k=1 k=1 k=1
Math 416, Spring 2018 Solutions to Exam 2 — May 2, 2018 Page 6 of 6

(b) (10 points) Show that there is no continuous function f : [−π, π] → R with Fourier series
∞  
X 1 1
√ cos kx + sin kx .
k k
k=1

If there was such a f , then one would have

n  
X 1 1
sn (f ) = √ cos kx + sin kx
k k
k=1

And hence, by Bessel’s inequality and the orthogonality of the trigonometric functions,
n  
X 1 1
+ = ksn (f )k22 ≤ kf k22 < ∞.
k k2
k=1

Here the last inequality follows as f is continuous and hence bounded on [−π, π]. However,
n ∞
X 1 X1
lim = =∞
n→∞ k k
k=1 k=1

as the harmonic series diverges. This contradicts,

n n  
X 1 X 1 1
lim ≤ + ≤ kf k22 < ∞.
n→∞ k k k2
k=1 k=1

That is, there can be no such f .