Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 1.6.30
Fix an  ∈ (0, 1). We apply the result of Exercise 28 to guarantee the existence
of two intervals I1 , I2 such that m(−E ∩ I1 ) > (1 − )m(I1 ) and m(F ∩ I2 ) >
(1 − )m(I2 ). Without loss of generality suppose that m(I1 ) ≤ m(I2 ). This
means that we can translate I1 by some amount h so that it lies inside of I2 .
Let δ = m(I1 ), so that for 0 < |α| < δ we see that the interval

I1,|α| = (I1 + h + α) ∩ I2

must have m(I1,|α| ) = δ − α > (1 − )δ. As a result of this, we must have that
I1,|α| ∩ I2 6= ∅. Pick some x ∈ (I1,|α| ∩ I2 ) and observe that this means that
there is some u ∈ E and v ∈ F such that

x = v = −u + h + α

So this means that u + v = (h + α). So E + F must contain the interval

(h − δ, h + δ) and we are done.

1
Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 2.5.20
We want to show that if E ⊂ R2 is a Borel set, then for any y the slice E y is (a
parallel translate of) a Borel set in R. We begin by considering the following

C = {E ⊂ R2 | ∀y, E y is Borel}

We first prove the following

Claim. C is a σ-algebra.

Proof. It is clear that C is non-empty; the open unit disk is in C, for instance.
Next we will see that C is closed under taking complements. Let E ∈ C, then
we know that for any y the set E y is Borel and that (E c )y = (E y )c must also
be Borel because the Borel sets are a σ-algebra. So E c ∈ C. Next, suppose that
{Ek }∞
k=1 is a countable collection of sets in C. We see that

∞
!y ∞
[ [
Ek = Eky
k=1 k=1

Then we use the fact that each of the Eky is Borel, and so their union must be
as well because the Borel sets are closed under countable union. Thus, C is a
σ-algebra.

Now to complete the proof we need to prove the following

Proposition. If E ⊂ R2 is open then E ∈ C.

Proof. We begin with the proof in the case of E being an open cube. We note
here that we must have E = (a, b) × (a + h, b + h) for some h > 0. We then
take the slice E y which is the same as (a, b) + y, where y ∈ R. Geometrically,
we simply have the line segment (a, b) translated in the y-direction a distance y.
Clearly, this is and open interval and hence, Borel. We then prove the claim for
the family of closed cubes. The proof is the same, we note that the slices E y of
a closed cube is now the closed interval,
T which can be written as the countable
intersection of intervals of the form n (a − 1/n, b + 1/n), and therefore is Borel.
We then use the fact that every open set in R2 can be written as the countable
S∞
union of almost disjoint closed cubes to see that we can write E = j=1 Qk ,
where each of the Qk is a closed cube. We then use the fact that each of the
Qk ∈ C and that C is closed under countable unions to see that E ∈ C.

Finally, we recall that the Borel sets is the smallest σ-algebra containing the
open sets. We have shown that C contains the open sets. And therefore, it
contains the Borel sets. This completes the proof.

2
Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 3.5.6
For functions f : R → R we define
Z x+h
∗ 1
f+ (x) = sup |f (y)|dy
h>0 h x

We want to show that the set

∗
Eα+ = {x ∈ R | f+ (x) > α}

is measurable with measure

Z
1
m(Eα+ ) = |f (y)|dy
α +
Eα

Following the hint, we wish to apply Lemma 3.5 (Rising Sun Lemma) to the
function Z x
F (x) = |f (y)|dy − αx
0

We first observe that if x ∈ Eα+ then there is some h > 0 such that
Z
1
+xx+h |f (y)|dy > α
h
Z
+xx+h |f (y)|dy > hα

So on the right side we consider

hα = hα + xα − xα = (x + h)α − xα
R R x+h R x+h
and note that because |f (y)| is increasing have x |f (y)|dy = 0 |f (y)|dy−
Rx
0
|f (y)|dy. So the inequality becomes
Z x+h Z x
|f (y)|dy − |f (y)|dy > (x + h)α − xα
0 0
Z x+h Z x
|f (y)|dy − (x + h)α > |f (y)|dy − xα
0 0

Consequently, we must have

Eα+ = {x ∈ R | ∃h > 0, F (x + h) > F (x)}

If f is integrable then F is absolutely continuous by the absolute continuity of

the integral. We then apply Lemma 3.5 to write Eα+ as the disjoint union of
intervals so
[∞
Eα+ = (ak , bk )
k=1

3
Ben Carriel Math 6110 Midterm Exam October 16, 2012

such that F (ak ) = F (bk ). As a result, we see that

Z ak Z bk
|f (y)|dy − αak = |f (y)|dy − αbk
0 0
Z bk Z ak
αbk − αak = |f (y)|dy − |f (y)|dy
0 0
Z bk
α(bk − ak ) = |f (y)|dy
ak

Moreover, because all of the intervals are disjoint we have that

∞
X
m(Eα+ ) = (bk − ak )
k=1

We then observe that

Z ∞ Z
X bk ∞
X
|f (y)|dy = |f (y)|dy = α (bk − ak ) = αm(Eα+ )
+
Eα k=1 ak k=1

So that Z
1
m(Eα+ ) = |f (y)|dy
α +
Eα

And we are done.

4
Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 3.5.10
Let {rn }∞
n=0 be an enumeration of the rationals. Consider the function
X
f (x) = 2−n
rn <x

I claim that this function is increasing on R,continuous on R − Q and discon-

tinuous elsewhere (i.e. only on Q). First we verify that f is increasing. Pick
x, y ∈ R such that x 6= y and assume without loss of generality that x < y.
Because Q is dense in R we can find some rational rm such that x < rm < y.
So we can see that
X X
f (y) = 2−n > 2−m + 2−n = 2−m + f (x) > f (x)
rn <y rn <x

Now we will show that f is discontinuous on Q. To do this we examine the left

and right hand limits of f at a point x. So consider

f + (x) = inf f (t) and f − (x) = sup f (t)

t>x t<x

Then f is continuous at x iff f + (x) − f − (x) = 0. For any rational rm we have

X X
f + (rm ) − f − (rm ) = inf 2−n − sup 2−n
rp >rm rq <rm
rn <rp rn <rm
X X
−n
= inf 2 + f (rm ) − sup 2−n
rp >rm rq <rm
rm <rk ≤rp rn <rm
X
−n
= inf 2
rp >rm
rm <rk ≤rp

= 2−m

So f is discontinuous on Q. We will now show that f is continuous on R − Q.

We note that for x 6∈ Q we have that x 6= rn for all n. So the same computation
shows that f + (x) − f − (x) = 0 because the last line of the computation will
never have any terms in the sum because rp 6= x for all p.

5
Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 3.5.32
We need to verify that the following

Lemma. A function f : R → R is Lipschitz with constant M if and only if f

has the following two properties:
(i) f is absolutely continuous.
(ii) |f 0 (x)| ≤ M for almost every x.

Proof. In the forward diretion we suppose that f is Lipschitz. This means that

|f (x) − f (y)| ≤ M |x − y|

for some M and all x, y ∈ R. Pick an  > 0 and observe that δ = /M
immediately gives that
N
X N
X
|bj − aj | < δ implies |f (bj ) − f (aj )| < 
j=0 j=1

so f is absolutely continuous. We then use this fact to apply Theorem 3.11 and
get that f must be differentiable almost everywhere. If x is a point for which
f 0 exists we observe that for any h > 0 we must have that

f (x + h) − f (x)
| |≤M
h
We then take the limit as h → 0 to see that |f 0 (x)| ≤ M whenever f 0 exists (i.e.
almost everywhere).
In the reverse direction, we have that f is absolutely continuous and |f 0 (x)| ≤
M almost everywhere. We apply Theorem 3.11 again to see that for any x, y ∈ R
with x ≤ y we have Z y
f (x) − f (y) = f 0 (t)dt
x
Taking the absolute value gives
Z y  Z y Z y
f 0 (t)dt ≤ |f 0 (t)|dt ≤
 
|f (x) − f (y)| =  M dt = M |x − y|
x x x

So f satisfies a Lipschitz condition and we are done.

6
Ben Carriel Math 6110 Midterm Exam October 16, 2012

Problem 4.6.4
We are considering the space
∞
( )
X
2 2
` (Z) = (. . . , a−2 , a−1 , a0 , a1 , a2 , . . .) | ai ∈ C, |ak | < ∞
k=−∞

when endowed with the inner product and norm

∞ ∞
!1/2
X X
2
(a, b) = ak bk and kak = |ak |
k=−∞ k=−∞

We will verify that `2 is a Hilbert space. It is clear that `2 is a vector space

with pointwise addition and multiplication by scalars because if a, b ∈ `2 then
∞
X ∞
X ∞
X
|ak + bk |2 ≤ 4 |ak |2 + 4 |bk |2 < ∞
k=−∞ k=−∞ k=−∞

Note that we have used the inequality |ak +bk | ≤ 2 max{|ak |, |bk |}. Furthermore,
for any λ ∈ C
∞
X ∞
X ∞
X
|λak |2 = |λ|2 |ak |2 = |λ|2 |ak |2 < ∞
k=−∞ k=−∞ k=−∞

So `2 is indeed a vector space.

We then check the properties of the inner product that we gave `2 . It is clear
that for fixed b the map a 7→ (a, b) is linear, it follows from the linearity of the
sum because for x, y ∈ `2 we have
∞
X ∞
X ∞
X
(xk + yk )bk = x k bk + yk bk = (x, b) + (y, b)
k=−∞ k=−∞ k=−∞

The fact that (a, b) = (b, a) follows from the symmetry in the inner product. To
see that (a, a) ≥ 0 for every a ∈ `2 , observe that
∞
X ∞
X
(a, a) = ak ak = |ak |2 ≥ 0
k=−∞ k=−∞

Because the last term is a sum of non-negative values. Now we check that
kak = (a, a)1/2 , which is straightforward because
∞
!1/2 ∞
!1/2
X X
(a, a)1/2 = ak ak = |ak |2 = kak
k=−∞ k=−∞

The next thing we have to verify is that kak = 0 if and only if a = 0. This
2 1/2
P∞
P∞
is clear because k=−∞ |ak | = 0 implies that k=−∞ |ak |2 = 0, which

7
Ben Carriel Math 6110 Midterm Exam October 16, 2012

is only possible if each of the terms |ak |2 = 0. This immediately gives that
|ak | = 0, so that means ak = 0 for every k. So a = 0.
Now we will prove the Cauchy-Schwartz inequality in `2 . Choose a, b ∈ `2 . In
the case that a and b are linearly dependent we have equality. So if a and b are
linearly independent we define
(a, b)
c=a− b
(b, b)
We then observe that
 
(a, b) (a, b)
(c, b) = a− b, b = (a, b) − (b, b) = 0
(b, b) (b, b)
So this means that c is orthogonal to b. We then apply the Pythagorean theorem
to a = (a,b)
(b,b) b + c to see

 (a, b) 2 2 2
 
2
kak =    kbk2 + kck2 = (a, b) + kck ≥ (a, b)
(b, b)  kbk2 kbk2
So we multiply by kbk and take square roots to see that |(a, b)| ≤ kakkbk. The
triangle inequality follows immediately, the proof is exactly the same as the one
given on the bottom of page 158 in Stein and Shakarchi modulo a change of
variable, so we forgo it here.
We will now show that `2 is complete in the metric induced by the norm. Let
{ai }i∈Z be a Cauchy sequence in `2 . We can imagine that the elements of each
sequence ai are the rows of an infinite matrix with entries ai,j . We know that
the rows of this matrix all converge to unique limits in C. We also can get the
convergence of the columns by noting that for fixed k
|ai,k − aj,k |2 ≤ kai − aj k2 → 0
when we send min{i, j} → ∞ because {ai } is Cauchy. This means that each
column {ai,j }i∈Z converges to some unique limit bj ∈ C. We then set b =
{bj }j∈Z . It is clear from above that d(ai , b) → 0 as i → ∞. We now need to
show that b ∈ `2 . To show this, we need the following

Lemma. Every Cauchy sequence in `2 is bounded in the norm.

Proof. Let {xn }n∈Z be a Cauchy sequence in `2 . Then for any  > 0 we can
find N > 0 such that |n| > N implies kx|n| − xN k < . We then see that
kx|n| k ≤ kx|n| − xN k + kxN k < kxN k + 
We then note that if |n| < N that if
M = max{kx−N +1 k, . . . , kx0 k, kx1 k, . . . , kxN −1 k}
As a result we have that kxn k ≤ max{M, kxN k + } < ∞, so {xn } is bounded
in the norm.

8
Ben Carriel Math 6110 Midterm Exam October 16, 2012

With the lemma in hand we can proceed to see that for any fixed j we have
an M such that
X∞
|ak,j |2 ≤ M 2
k=−∞

So we can see that for any finite N we have that

X X
|bn |2 = lim |am,n |2 < M 2
m→∞
|n|<N |n|<N

Taking N → ∞ gives the result.

Now we show that `2 is separable. We construct the natural basis for `2
consisting of the vectors ek = (. . . , 0, 1, 0, . . .), which have a 1 in the k th position.
We need to show that finite linear combinations of the ek are dense in `2 . For
each element x ∈ `2 we consider the combination
X
SN (x) = xk ek
|k|≤N

It is clear that SN (x) and x agree for all indices |k| ≤ N . We then see that
X
kSN (x) − xk = |xk |
|k|>N

Because x ∈ `2 we have that k |xk |2 converges and so kP

P P
|xk | must converge
as well. We apply the definition of convergence to see that |k|>N |xk | → 0 as
N → ∞. Thus, finite linear combinations of the ek are dense in `2 .