I.

The Hölder Inequality

1 1
Hölder: kf gk1 ≤ kf kpkgkq for p + q = 1.

What does it give us?

Hölder: (Lp )∗ = Lq (Riesz Rep), also: relations between Lp spaces

I.1. How to prove Hölder inequality.

p q
(1) Prove Young’s Inequality: ab ≤ ap + bq .
(2) Then put A = kf kp, B = kgkq . Note: A, B 6= 0 or else trivial. Then
let a = |f (x)| |g(x)|
A , b = B and apply Young’s:

|f (x)g(x)| |f (x)|pap bq |g(x)|q

ab = AB≤ + pAp= + qAq
p q
Z Z Z
1
AB |f (x)g(x)|dµ ≤ pA1 p |f |p dµ + qB1 q |g|q dµ
R R
but Ap = |f |p dµ and B q = |g|q dµ, so this is
1 1 1
kf kp kgkq kf gk1 ≤ p + q =1
kf gk1 ≤ kf kpkgkq

I.1.1. How to prove Young’s inequality.

There are many ways.
1. Use Math 9A. [Lapidus]
Wlog, let a, b < ∞ (otherwise, trivial).
p
Define f (x) = xp + 1q − x on [0, ∞) and use the first derivative test:

f 0 (x) = xp−1 − 1, so f 0 (x) = 0 ⇐⇒ xp−1 = 1 ⇐⇒ x = 1.

So f attains its min on [0, ∞) at x = 1. (f 00 ≥ 0).
Note f (1) = 1p + 1q − 1 = 0 (conj exp!).

xp
So f (x) ≥ f (1) = 0 =⇒ p + 1q − x ≥ 0
xp 1
=⇒ p
+ q
≥x
2 Real Analysis Qual Seminar

Ansatz: x = ab1/(1−p). Then xp = ap bp/(1−p) = ap b−q :

ap b−q p
ab1/(1−p) ≤ p + 1
q 1−p = −q
 p
 p
 p 1 p
ap − 1−p 1 − 1−p 1− 1−p − 1−p
ab ≤ p b 1−p b + qb b =b
ap bq
ab ≤ p + q

2. Use Math 9B. [Cohn]

Consider the graph of t = sp−1:

Figure 1. The graph of t = sp−1

t
t = s p-1

(1)
(2)

a s

Since
1 1 1 1 q−1 1
p + q = 1 =⇒ p =1− q = q =⇒ p = qq − 1 =⇒ p − 1 = q−1 ,

this is also the graph ofis = tq−1 .

Ra p a p
Now (1) = 0 sp−1 = sp = ap ,
R b q−1 ib 0 q
tq
and (2) = 0 t = q = bq .
0
ap q
Thus the area of the entire shaded region is (1) + (2) = p + bq , which
is clearly always larger than the box of area ab:
I.1.2. A proof without Young’s inequality.
Use convexity [Rudin]:
ϕ ((1 − λ)x + λy) ≤ (1 − λ)ϕ(x) + λϕ(y).
Real Analysis Qual Seminar 3

Figure 2. The inherent inequality

t t
t = s p-1 t = s p-1

extra b extra
b

ab ab

a s a s

Since f ∈ Lp , g ∈ Lq , we have 0 < kf kp, kgkq < ∞, wlog.

Define F (x) = |fkf(x)| |g(x)|
kp and G(x) = kgkq so that

Z Z R
|f (x)|p |f |p dµ
F p dµ = dµ = R =1
kf kpp |f |p dµ
R
and Gq = 1 similarly.
Now define
 p  q
|f (x)| |g(x)|
s(x) = log , t(x) = log ,
kf kp kgkq

so that

F (x) = es(x)/p and G(x) = et(x)/q .

1
Since ex is a convex function, put λ = q
and get

s(x) t(x) s(x) t(x)

+ q 1 1
e p ≤ pe
p + qe
q

F (x)p G(x)q
=⇒ F (x)G(x) ≤ p + q .
4 Real Analysis Qual Seminar

Now integrate the left side and get

Z
kF Gk1 = |F G|dµ
Z
|f g|
= dµ
kf kpkgkq
Z
1
= |f g|dµ
kf kpkgkq
kf gk1
= ,
kf kpkgkq
and we integrate the right side to get
Z   Z Z
F (x)p G(x)q
+ dµ = 1p F p dµ + 1q Gq dµ
p q
= 1p + 1q
=1
Thus,
kf gk1
≤ 1 =⇒ kf gk1 ≤ kf kpkgkq .
kf kpkgkq
Advantage of this method? No need for Young’s!
Real Analysis Qual Seminar 5

I.1.3. Recap - 3 good ways to prove a functional inequality.

To prove a(x) ≤ b(x):
1. Use basic calculus on a difference function:
Define f (x) := a(x) − b(x).
Use calculus to show f (x) ≤ 0 (by computing f 0 , etc)

2. Use geometry.

3. Exploit another inequality. E.g., for any convex function ϕ(x),

ϕ ((1 − λ)x + λy) ≤ (1 − λ)ϕ(x) + λϕ(y).
Candidates for ϕ: ex , xp, . . ..

I.1.4. What did we not do yet? case p = 1, ∞.

|g(x)| 6ae kgk∞
|f (x)| · |g(x)| 6ae |f (x)| · kgk∞
|f (x)g(x)| 6ae |f (x)| · kgk∞
kf (x)g(x)k1 6ae |f (x)| · kgk∞
p = ∞ is exactly the same.

I.2. How to use the Hölder inequality. Assume (X, M, µ) is a measure

space with µX ≥ 1, f : X → R is measure, and Lp = Lp(X, µ).
1. For
R 1 p≤ p ≤ Rq < ∞, if |f (x)| ≥ 1, then kf kp ≤ kf kq . If µX = ∞, then
q
X |f | dµ = X |g| dµ = ∞, so let µX < ∞.
Then
|f |p ≤ |f |q
Z Z
p
|f | dµ ≤ |f |q dµ. (I.1)
X X
R
If X |fR |q dµ = ∞, it is trivial, so assume not.
Then X |f |q dµ < ∞ =⇒ f ∈ Lq , and (I.1) =⇒ f ∈ Lp.
Now p = q =⇒ kf kp = kf kq and we are done trivially, so let p < q. We
would like to use Hölder with g(x) = 1 and some conjugate exponents
6 Real Analysis Qual Seminar

α, β with α1 + β1 = 1.
Ansatz : Let α = pq and β = q
q−p , so
1 1 p q−p q
α + β = q + q = q = 1.

Now use Hölder with f = f p to get

kf pgk1 ≤ kf pkα kgkβ . (I.2)
Now remembering that g = 1, we have
Z
p p
kf gk1 = kf k1 = |f |p = kf kpp, and
X
Z p/q Z p/q
p p q/p q
kf kα = |f | = |f | = kf kpq, and (*)
X X
Z 1/β Z 1/β
kgkβ = 1β = 1 = (µX)1/β .
X X

So (I.2) becomes
q−p
kf kpp ≤ kf kpq · (µX) q
kf kp ≤ kf kq · (µX)(q−p)/pq
kf kp ≤ kf kq

q/p
2. For 1 ≤ p ≤ q < ∞, |f (x)| ≤ 1 ∀x ∈ X =⇒ kf kp ≥ kf kq .
p = q is trivial, so take p < q.
Then
p < q, |f | ≤ 1 =⇒ |f |p ≥ |f |q
Z Z
|f |p ≥ |f |q
Z 1/p Z 1/p
p q
|f | ≥ |f |

kf kp ≥ kf kq/p
q .
Real Analysis Qual Seminar 7

3. Show p ≤ q ≤ r and f ∈ Lp , f ∈ Lr =⇒ f ∈ Lq .
e
Let A = {|f | ≥ 1} and B = {|f | < 1} = A.

Z Z Z
f ∈ Lp =⇒ |f |p = |f |p + |f |p < ∞
ZX A B

|f |p < ∞
=⇒ (I.3)
ZB Z Z
f ∈ Lr =⇒ |f |r = |f |r + |f |r < ∞
ZX A B

=⇒ |f |r < ∞. (I.4)
A

R R
On A, |f |q ≤ |f |r =⇒ RA |f |q ≤ RA |f |r .
On B, |f |q ≤ |f |p =⇒ B |f |q ≤ B |f |p .
So

Z Z Z
|f |q = |f |q + |f |q
X
ZA ZB
≤ |f |r + |f |p
A B
<∞ by (I.3),(I.4)

q
shows that f ∈ L
R . R
Moral: to show X f (x) ≤ X g(x), try splitting X.
8 Real Analysis Qual Seminar

4. Show there is a bounded linear operator ϕ : Lq → (Lp )∗ given by

Z
ϕf (g) = ϕ(f )(g) = f gdµ, ∀f ∈ Lq , ∀g ∈ Lp
X
R
so that ϕ : g 7→ f gdµ is the functional “integration against f”.
kϕ(f )k
kϕk = sup
f ∈Lq ,f 6=0 kf kq
kϕ(f )(g)k
= sup sup def of kϕk
f ∈Lq ,f 6=0 g∈Lp ,g6=0 kf kq kgkp
R
| X f g dµ|
= sup sup def of ϕ(f )(g)
f ∈Lq ,f 6=0 g∈Lp ,g6=0 kf kq kgkp
Z Z
kf gk1
≤ sup sup | f g| ≤ |f g|
f ∈Lq ,f 6=0 g∈Lp ,g6=0 kf kq kgkp
kf kq kgkp
≤ sup sup Hölder
f ∈Lq ,f 6=0 g∈Lp ,g6=0 kf kq kgkp

so kϕk ≤ 1 and ϕ is bounded.

To see ϕ is linear, let f1 , f2, f ∈ Lq , g ∈ Lp, and α ∈ R: we show two
things in (Lp )∗ are equal by showing that they act the same way on
any g ∈ Lp .
Z
ϕ(f1 + f2 )(g) = (f1 + f2) g dµ
Z Z
= f1g dµ + f2g dµ
= ϕ(f1)(g) + ϕ(f2)(g) = (ϕ(f1) + ϕ(f2)) (g)
shows ϕ(f1 + f2) = ϕ(f1) + ϕ(f2), and
Z Z
ϕ(αf )(g) = αf g dµ = α f g dµ = αϕ(f )(g)

shows ϕ(αf ) = αϕ(f ).

Hence (by the linearity of the integral), ϕ is linear.
Real Analysis Qual Seminar 9

II. The Dual of Lp

R
Proposition II.1. Show that ϕ : Lq → (Lp)∗ by ϕ : f 7→ f gdµ is an
isometry.
Proof. So we must show kϕ(f )k = kf k, ∀f ∈ Lq . Let 1 < p, q < ∞. Then

kϕ(f )(g)k
kϕ(f )k = sup def of kϕk
g∈Lp ,g6=0 kgkp
R
| X f g dµ|
= sup def of ϕ(f )(g)
g∈Lp ,g6=0 kgkp
Z Z
kf gk1
≤ sup | f g| ≤ |f g|
g∈Lp ,g6=0 kgkp
≤ sup kf kq Hölder
g∈Lp ,g6=0

Hence kϕ(f )k ≤ kf k. For kϕ(f )k ≥ kf k, use the fact that kϕ(f )k is defined
as a supremum: kϕ(f )k is the smallest number such that
kϕ(f )(g)k ≤ kϕ(f )k · kgk holds for all g (6= 0).
In other words, if we can find a g for which kϕ(fkgk)(g)k
≥ kf k, then
n o
kϕ(f )k = supg∈Lp ,g6=0 kϕ(fkgk
)(g)k
≥ kf k.
Ansatz : let g = |f |q/p sgn f .
Then |g|p = |f |q = f g.1 Thus, f ∈ Lq =⇒ g ∈ Lp. Now
Z Z
|g| = |f |q
p

Z 1/p Z 1/p
|g|p = |f |q
Z 1/q Z 1/p Z 1/q Z 1/p
|f |q |g|p = |f |q |f |q
Z 1/q+1/p Z 1
q q
kf kq kgkp = |f | = |f | = kf kqq

11 + 1
= 1 =⇒ q
+ q
= q =⇒ q
+ 1 = q, so f g = f · |f |q/p sgn f = |f | · |f |q/p = |f |1+q/p = |f |q .
p q p q p
10 Real Analysis Qual Seminar
R R
Thus, ϕ(f )(g) = f g dµ = |f |q = kf kqq = kf kq kgkp
=⇒ |ϕ(f )(g)|
kgkp ≥ kf kq .

Now suppose p = 1, q = ∞.
We have
R
| f gdµ| kgk1 · kf k∞
kϕ(f )k = sup ≤ sup = kf k∞
kgk1 kgk1
R R

as before. Now it remains to find a g ∈ L1 for which f g dµ ≥ |g|dµ kf k∞ .
We have f ∈ L∞ , so note kf k∞ < ∞. Then fix ε > 0 and define
B = {f ≥ kf k∞ − ε},
and let A be any measurable subset of B such that 0 < µA < ∞.2 Define
gε := χA sgn f.3
R
Then |gε |dµ = µA and
Z Z
f gε dµ = |f |dµ ≥ (kf k∞ − ε) µA
A
R
|f gε|dµ
≥ kf k∞ − ε
µA
R 
|f gε |dµ
sup R ≥ kf k∞ − ε.
|gε |dµ
nR o
|f g|dµ
Since this is true for any ε, let ε → 0 and obtain sup R
|g|dµ
≥ kf k∞.

Now suppose p = ∞,R q = 1.

| f gdµ|
Again, kϕ(f )k = sup kgk∞ ≤ kf k1, as before. Now find g ∈ L∞ such that
Z Z 
f g dµ ≥ kgk∞ |f |dµ .

Let α > 0 and define g = α sgn f be a constant function. Then

Z Z Z
f g dµ = α |f |dµ = kgk∞ |f |dµ.


2Note that if f is a constant function, µB could be ∞! µB > 0 by def of kf k .
∞
3Include the sgn f so that f g = |f | instead of just f , in the following derivation.
ε
Real Analysis Qual Seminar 11

II.1. The Riesz Representation Theorem.

Theorem II.2. Let F be a bounded
R linear functional on Lp, 1 ≤ p < ∞.
Then ∃g ∈ Lq such that F (f ) = f g dµ, ∀f ∈ Lp , and kF k = kgkq .
There are two common proofs for this theorem. One uses step functions and
absolute continuity of functions; the other uses simple functions and absolute
continuity of measures. Both follow a similar strategy:
R R
(1) Show F (χA ) = g χA dµ = A g dµ.
• use absolute continuity of Φ : [0, 1] → R by gF (s) = F (χs ); or
• use absolute continuity of νE = F (χE ).

(2) Extend to a dense subspace of Lp

• use step functions for gF (s)
• use simple functions for νE
(3) Establish kF k = kgkq
• extend F to bounded
R measurable functions, use Royden & Hölder
• define G(f ) = f gdµ on Lp and use density, continuity

(4) Extend to Lp
• approx by step functions
• use G
(5) Show uniqueness of g.

Method I:
• uses step functions– only applies for Lp ([a, b], λ)
• requires reference to 3 thms of Royden
• nice use of DCT, boundedness
• absolute continuity of a function is a bit more concrete
Method II:
• uses simple functions, so applies to Lp(X)
• requires reference to 1 thm of Royden
• uses Radon-Nikodym Theorem
• smooth use of general topology
• works for any σ-finite µ
12 Real Analysis Qual Seminar

Important note: must add “µ is σ-finite” in order to do the case p = 1!

Method I (Royden)
Proof.
Rs
1. For s ∈ [0, 1], let χs := χ[0,s] . Then F (χs ) = 0 g dλ is some real
number, so define Φ : [0, 1] → R by
Z s
Φ(s) = F (χs ) = g dλ.
0
Claim: Φ is absolutely continuous.

Fix ε > 0 and let {(ai, bi)}ni=1 be any finite collection of disjoint
subintervals of [0, 1] such that
X
(bi − ai ) < δ.
P
Then |Φ(bi) − Φ(ai)| = F (f ) for
n
X
f= (χbi − χai ) sgn (Φ(bi) − Φ(ai)) .4
i=1
R PR p P R bi P
Since |f |p = χ = 1 dλ = (bi − ai ) < δ,5
(ai ,bi ) ai
X
|Φ(bi) − Φ(ai)| = F (f ) ≤ kF k · kf kp < kF kδ 1/p.
Thus, total variation of Φ over any finite collection of disjoint intervals
is less than ε, as long as the total length of these intervals is less than
εp
δ= kF kp ,

which shows that Φ is absolutely continuous.

6
Then Φ has an antiderivative, by some theorem:
Z s
Φ(s) = g.
0
4Note that |Φ(b ) − Φ(a )| = |F (χ ) − F (χb )| = (F (χ ) − F (χ )) sgn (Φ(χ ) − Φ(χ )) .
i i ai i ai bi bi ai
5In the first equality, use (χ − χ ) = χ and | sgn h| = 1 ae.
bi ai (ai ,bi )
6Royden, Lemma 5.14 on page 110.
Real Analysis Qual Seminar 13
R1
Thus F (χs ) = 0 g χs .

2. Since every step function ϕ on [0, 1] is a linear combination

X
ϕ= ci χsi (λ-ae),
we get
Z 1
F (f ) = gϕ
0
R
by the linearity of F, .

3. Now extend F to the bounded measurable functions f on [0, 1].

Let f be such a function, and find a bounded sequence {ϕn} ⊆ Step
which converges λ-ae to f . This is possible by Royden, Prop 3.22.
Then {|f − ϕn |p } is uniformly bounded (∃M such that |f − ϕn |p ≤
M p , ∀n, x) and tends to 0, λ-ae.
This bound allows us to use the DCT and get
Z
n→∞
lim |f − ϕn |p = 0 =⇒ kf − ϕnkp −−−−→ 0.
n→∞

Then the boundedness of F and the fact that

|F (f ) − F (ϕn)| = |F (f − ϕn)| ≤ kF k · kf − ϕnk
together imply that F (f ) =Rlimn→∞ F (ϕ
R n).
Also, |gϕn | ≤ |g| · Mϕ =⇒ f g = lim gϕn by the DCT again.7
Putting this together,
Z Z
F (f ) = lim F (ϕn) = lim gϕn = f g, ∀f (bdd,measurable).

Then by Proposition 5.12 on page 131 of Royden,

Z 
 
 f g dλ = |F (f )| ≤ kF k · kf kpN · kf kp
 

and we have g ∈ Lq , kgkq ≤ N = kF k,.

Then by Prop II.1, kF k = kgkq .
7M is the uniform bound on the sequence {ϕ }.
ϕ n
14 Real Analysis Qual Seminar

4. Extend to f ∈ Lp .
∀ε, ∃ψ ∈ Step such that kf − ψkp < ε. R
Then ψ ∈ Step =⇒ ψ bounded =⇒ F (f ) = ψg, by (3).
Hence,
 Z   Z Z 
   
F (f ) − f g  = F (f ) − F (ψ) + ψf − f g 
   
Z Z 
 
≤ |F (f ) − F (ψ)| +  f g − ψf 
Z 
 
= |F (f − ψ)| +  (g − ψ)f 

≤ kF k · kf − ψkp + kf − ψkp · kgkq

< (kF k + kgkq ) ε
R
Since ε is arbitrary, this shows F (f ) = f g dλ.

5. If g1, g2 determine the same F in this way, then

Z Z Z
f g1 dλ − f g2 dλ = f (g1 − g2 ) dλ

gives the zero functional, and

kg1 − g2 kq = 0 =⇒ g1 = g.
ae 2

Method II (Royden)
Again, given a bounded
R linear functional F on Lp , we must find a g ∈ Lq
such that F (f ) = f g dµ.

Proof. First, consider a finite measure space (X, A, µ).

Then f bounded =⇒ f ∈ Lp (µ).

1. Define ν on A by
νE = F (χE ).
Real Analysis Qual Seminar 15

Then ν is a signed measure:

is the disjoint union of {En } ⊆ A, let αn = sgn F (χEn ) and define
If E P
f= αn χEn . Then F is bounded, so a lemma gives
X∞ X∞
|νEn| = F (f ) < ∞ and νEn = F (χE ) = νE.
n=1 n=1
So ν is a signed measure. Also,
µ = 0 =⇒ F (χE ) = 0,
so ν R µ. Then Radon-Nikodym applies and ∃g measure such that
νE = E g dµ.
Note: F bounded
R =⇒ F (χX ) = νX < ∞, so ν is finite.
Then νX = X g dµ < ∞ =⇒ g ∈ L1(µ).

R
2. Let ϕ be a simple function. Then by linearity of F and , we have
Z
F (ϕ) = ϕg dµ.

Now |F (ϕ)| ≤ kF k · kϕkp =⇒ g ∈ Lq by some Lemma on page 283.

R
3. (&4.) Define G(f ) = f g dµ for f ∈ Lp, so (G − F ) is a BLT on Lp
which vanishes on S.
(G − F ) bounded ≡ (G − F ) continuous,
so (G − F ) = 0 on Lp . R
Hence, ∀f ∈ Lp , F (f ) = f g dµ and kF k = kGk = kgkq .

5. If g1, g2 determine the same F in this way, then

Z Z Z
f g1 dµ − f g2 dµ = f (g1 − g2 ) dµ

gives the zero functional, and

kg1 − g2 kq = 0 =⇒ g1 = g.
ae 2

Now consider µ σ-finite. Choose {Xn } such that

[∞
X= Xn , Xn ⊆ Xn+1, µXn < ∞, ∀n.
n=1
16 Real Analysis Qual Seminar

Then the previous

R case gives a gn for peach Xn such that gn vanishes outside
Xn and F (f ) = f gn dµ for all f ∈ L that vanish outside Xn .
By construction, the uniqueness of gn on Xn gives


gn+1  = gn ,
X n

so for x ∈ X, define
g(x) := gn (x), where x ∈ Xn .
Since gn differs from gm on a set of at most measure 0 (on any Xi where both
are defined), discrepancies may be safely ignored and g is well-defined.
Moreover, |gn | increases pointwise to |g|. By MCT,
Z Z
|g| dµ = lim |gn |q dµ ≤ kF kq ,
q

so g ∈ Lq .
For general f ∈ Lp , define
(
f on Xn
fn = fn .
0 on X
pw Lp
Then fn −−−→ f and fn −−−→ f .
Then |fn g| ≤ |f g| ∈ L1, so DCT gives
Z Z
DCT
f g dµ = lim fn g dµ
Z
DCT
= lim fn gn dµ
= lim F (fn )
= F (f ).


Side note: if λ is Lebesgue measure and ν is the point mass at 0 (on

dν
(R, B(R))), then what would dµ be (if ν  µ)?
Real Analysis Qual Seminar 17

II.2. (L1)∗ = L∞ , but (L∞)∗ 6= L1.

Let X be [0, 1] so that we can safely consider C(X) as a subspace of L ∞ (X).

Define ζ : C(X) → R by ζ(f ) = f (0), so ζ ∈ (C(X))∗ .
By HBT, ∃ϕ ∈ (L∞ )∗ such that ϕ(f ) = f (0) ∀f ∈ C(X). To see that ϕ
cannot be given by integration against a function in L1, consider fn ⊆ C(X)
defined by fn(x) = max{1 − nx, 0}.

Figure 3. The functions fn .

1
fn

0 1
n 1

Then ϕ(fn) = fn (0) = 1 ∀n.

1
But fn (x) →
R 0 ∀x > 0, so fn g → 0 ∀g ∈ L .
If ϕ(fn) = fn g, then we would have
1 = ϕ(f ) = ϕ(lim fn )
= lim ϕ(fn) ϕ ∈ (L∞ )∗ =⇒ ϕ continuous
Z
= lim fng hypothesis
Z
= lim fng DCT
Z
= 0
=0

Slightly fancier version: use ζ(f ) = f (p) and fn (x) = max{0, 1 − n|x − p|}.
18 Real Analysis Qual Seminar

III. The Minkowski Inequality

(X, M, µ), 1 ≤ p ≤ ∞. Then kf + gkp ≤ kf kp + kgkp .

What does it give us? Lp is a normed vector space (4-ineq), Lp is Banach.

III.1. How to prove Minkowski inequality.

III.1.1. Use Hölder inequality.

case i) 1 < p < ∞.
1 1
For f, g ∈ Lp , define q by p + q = 1, so that p + q = pq:
1 1 p−1
q = 1− p = p
p
=⇒ q = p−1
 
p2 −p p p2 p
=⇒ p + q = p−1
+ p−1
= p−1
=p p−1
= pq

Now

q
p/(p−1)
|f + g|p−1 = |f + g|p−1 = |f + g|p .
Since Lp is a vector space, we have f +g ∈ Lp and hence |f +g|p−1 ∈ Lq .
We need to set up for Hölder:

|f + g|p = |f + g|p−1 · |f + g|
≤ |f + g|p−1 (|f | + |g|)
≤ |f + g|p−1|f | + |f + g|p−1 |g|
Z Z Z
|f + g|p dµ ≤ |f + g|p−1 |f | dµ + |f + g|p−1 |g| dµ

Now Hölder gives

Z
|f | · |f + g|p−1 dµ ≤ kf kpk|f + g|p−1kq and

Z
|g| · |f + g|p−1 dµ ≤ kgkpk|f + g|p−1 kq .
Real Analysis Qual Seminar 19

Thus
Z
|f + g|p dµ ≤ (kf kp + kgkp) k|f + g|p−1kq
Z 1/q
= (kf kp + kgkp ) |f + g|(p−1)q dµ .
R
If |f + g|p dµ = 0, then Minkowski is trivial, so assume not.
Then we may divide by
Z 1/q Z 1/q


p−1 q p
|f + g| dµ = |f + g| dµ

to get
Z 1−1/q
|f + g|p dµ ≤ kf kp + kgkp .

case ii) p = 1.
Then

|f + g| ≤ |f | + |g| 4-ineq
Z Z Z
|f + g|dµ ≤ |f |dµ + |g|dµ integration
kf + gk1 ≤ kf k1 + kgk1

case iii) p = ∞.
Then define the null sets

N1 := {|f (x)| > kf k∞}, N2 := {|g(x)| > kgk∞ }.

Then f, g ∈ L∞ =⇒ µN1 = µN2 = 0, so µ(N1 ∪ N2) = 0 also. On the

complement of N1 ∪ N2 we have

|f (x) + g(x)| ≤ |f (x)| + |g(x)|.

Then taking suprema gives

kf + gk∞ ≤ kf k∞ + kgk∞ .
20 Real Analysis Qual Seminar

III.1.2. Use convexity.

Let α = kf kp and β = kgkp . Note: α, β 6= 0 or else trivial. Then define
f0 := α1 |f |, g0 := β1 |g|
so that these functions satisfy
|f | = αf0 , |g| = βg0 , kf0kp = kg0kp = 1.
Note that this implies
kf0kpp = kg0kpp = 1 (III.1)
Set
α α+β α β
λ := α+β
so 1 − λ = α+β
− α+β
= α+β
.
Then we have
|f (x) + g(x)|p ≤ (|f (x)| + |g(x)|)p 4-ineq
= (αf0 (x) + βg0(x))p def of f0 , g0
 p
p α β
= (α + β) α+β f0(x) + α+β g0 (x)

= (α + β)p (λf0 (x) + (1 − λ)g0(x))p def of λ

= (α + β)p (λf0 (x)p + (1 − λ)g0(x)p) cvxty of tp
Z Z
|f (x) + g(x)|p dµ ≤ (α + β)p (λf0(x)p + (1 − λ)g0 (x)p) dµ integrating


kf + gkpp ≤ (α + β)p λkf0kpp + (1 − λ)kg0kpp linearity
≤ (α + β)p (λ + (1 − λ)) by (III.1)
= (α + β)p
kf + gkp ≤ (α + β) pth roots
Real Analysis Qual Seminar 21

III.2. The Riesz-Fischer Theorem: Lp(X, M, µ) is complete.

Road map:
Split into the two cases 1 ≤ p < ∞, p = ∞. For each case:
(1) Invoke the Banach characterization lemma.
(2) Define
(P
fn (x) behaves
f (x) =
0 else
P .
Use g(x) = k |fk (x)| for 1 ≤ p < ∞, use Nk = {x .. |fk (x)| > kfk k∞ }
for p = ∞.
P Lp
(3) Use Minkowski to show f ∈ Lp and nk=1 fk −−−→ f.

case i) 1 ≤ p < ∞.
1. By the lemma, it suffices to show that every series which converges
absolutely (in R) also converges in Lp , p ∈ [1, ∞).

P
2. Let ∞ k=1 kf
∞ p
k kp < ∞ for some {fk }k=1 ⊆ L .
P n→∞
NTS: kf − nk=1 fk kp −−−−→ 0 for some f ∈ Lp, since this is what
Pn Lp
f
k=1 k −− −→ f means.
Define
∞
X
g(x) = |fk (x)|
k=1

so that g ≥ 0 (g may take the value ∞). Note:

n
!p
X
|fk | ≥0 (III.2)
k=1

and since positive exponents preserve order, we also have

n
!p n+1
!p
X X
|fk | ≤ |fk | . (III.3)
k=1 k=1
22 Real Analysis Qual Seminar

Then we have
Z  p !1/p
X n 
 
kgkp = lim  |fk | dµ
n→∞  
k=1
Z n
!p !1/p
X
= lim |fk | dµ by (III.2)
n→∞
k=1
Z n
!p !1/p
X
= lim |fk | dµ by MCT,(III.3)
n→∞
k=1

Xn

= lim |fk | def of k · kp
n→∞
k=1 p
n
X
≤ lim kfk kp Minkowski
n→∞
k=1
∞
X
= kfk kp ,
k=1
which is finite, by hypothesis. Thus g ∈ Lp , so |g| 6ae ∞.
Hence we may define
(P
∞
k=1 fk (x) |g(x)| < ∞
f (x) =
0 |g(x)| = ∞
so that f is measurable and
|f |p ≤ g p =⇒ f ∈ Lp .
P P p
Since limn→∞ |f (x) − nk=1 fk (x)| = 0 and |f (x) − ∞k=1 fk (x)| ≤
P n→∞
g p are both true ae, the DCT gives kf − nk=1 fk kp −−−−→ 0.
Real Analysis Qual Seminar 23

case ii) p = ∞.
1. By the lemma, it suffices to show that every series which converges
absolutely (in R) also converges in L∞ .

P
2. Let ∞ k=1 kf
∞
k k∞ < ∞ for some {fk }k=1 ⊆ L .
∞
P n→∞
NTS: kf − nk=1 fk k∞ −−−−→ 0 for some f ∈ L∞ .
For each k, define
.
Nk := {x .. |fk (x)| > kfk k∞ },
so that µNk = 0, ∀k =⇒ µ (∪k Nk ) = 0. Then if x ∈
/ ∪ k Nk ,
X X X
|fk (x)| ≤ kfk k∞ =⇒ fk (x) < ∞,
k k k
by what we know of R. Now we may define
(P
k fk (x) x ∈
/ ∪ k Nk
f (x) =
0 x ∈ ∪ k Nk
so that f is µ-measurable and bounded, i.e., f ∈ L∞ .

3. Since µ(∪k Nk ) = 0,

Xn X ∞ ∞
X

f − fk ≤ fk ≤ kfk k∞ by Mink

k=1 ∞ k=n+1 ∞ k=n+1
Then taking limits,

Xn X∞

lim f − fk ≤ lim kfk k∞ = 0.
n→∞ n→∞
k=1 ∞ k=n+1
Pn n→∞
Thus, kf − k=1 fk k∞ −−−−→ 0.
24 Real Analysis Qual Seminar

Lemma III.1. (Banach Characterization Lemma).

Suppose that (X, k · k) is a normed vector space. Then
X is Banach ⇐⇒ every absolutely convergent series in X is convergent.
Proof.
(⇒) Suppose every Cauchy
P sequence converges.
Let {xk } be such that kxk k < ∞ so {xk } is absolutely convergent.
Then let
n
X X∞
sn := xk , ands := lim sn = xk .
n→∞
k=1 k=1
NTS: {sn } is Cauchy. Wlog, let n < m.

X
m

ksn − sm k = xk

k=n+1
Xm
≤ kxk k 4-ineq
k=n+1
∞
X
m→∞
−−−−−→ kxk k
k=n+1
n→∞
−−−−→ 0
n→∞
Hence, {sn } Cauchy implies that sn −−−−→ s = limn→∞ sn ∈ X.
(⇐) Suppose that every abs. convergent series is convergent.
Let {xn} be a Cauchy sequence.
NTS: xn → x ∈ X.
Since {xn} is Cauchy, we can find a subsequence {xnk } which satisfies
1
kxnk − xnk+1 k < 2k .
Define
v 1 = x n1 , and vk = xnk+1 − xnk .
Then we have a telescoping sum:
N
X
vk = xn1 + (xn2 − xn1 ) + . . . + (xnN − xnN −1 ) = xnN ,
k=1
Real Analysis Qual Seminar 25

so ∞ ∞
X X
1
kvk k < 2k
=1
k=1 k=1
P∞
shows Pk=1 kvk k converges.
Hence, ∞k=1 vk converges by hypothesis to some v ∈ X. Then
∞
X N
X
vk = v = lim vk = lim xnN
N →∞ N →∞
k=1 k=1
N →∞
shows xnN −−−−−→ v. Now
kv − xnk ≤ kv − xnk k + kxnk − xnk
n→∞
shows that xn −−−−→ v also.

26 Real Analysis Qual Seminar

IV. Hilbert Space Review

Most material in this talk is from Reed & Simon.

Definition IV.1. A complex vector space is called an inner product space

(IPS) when
(i) hx, xi ≥ 0, and hx, xi = 0 iff x = 0,
(ii) hx, y + zi = hx, yi + hx, zi,
(iii) hx, αyi = αhx, yi,
(iv) hx, yi = hy, xi.
An inner
p product space is a Hilbert space iff it is complete under the norm
kxk = hx, xi.

Definition IV.2. Two vectors x 6= y are orthogonal iff hx, yi = 0.

A collection {xi} is an orthonormal set iff
hxi, xii = 1 and hxi , xj i = 0 ∀i 6= j.

Proposition IV.3. (Pythagorean Theorem)

Let {xn}Nn=1 be an orthogonal set in an IPS. Then
2
X N XN

xn = kxnk2

n=1 n=1
P P P P
Proof. k xn k2 = h xn , xn i = N n,m=1 hxn , xm i .
ThenP see that all the terms
PN with n2 6= m are 0 because of orthogonality, leaving
N
only n=1 hxn, xni = n=1 kxnk . 

Proposition IV.4. (Bessel’s Inequality)

If {xα }α∈A is an orthonormal set in an IPS, then for any x,
X
|hx, xαi|2 ≤ kxk2.
α∈A
Real Analysis Qual Seminar 27
P
Proof. It suffices to show that α∈F |hx, xα i|2 ≤ kxk2 for any finite F ⊆ A:
2
X

0 ≤ x − hx, xαixα

α∈F
* +
X X
= x− hx, xα ixα , x − hx, xα ixα
α∈F α∈F
* + 2
X X
2
= kxk − 2 Re x, hx, xα ixα + hx, xαixα

α∈F α∈F
2
X X
2
= kxk − 2 Re hx, xαi hx, xαi + hx, xα ixα

α∈F α∈F
X X
= kxk2 − 2 |hx, xα i|2 + |hx, xαi|2 (IV.1)
α∈F α∈F
X
= kxk2 − |hx, xα i|2
α∈F

Where (IV.1) comes by the Pythagorean Thm. 

.
Note that this theorem indicates {α .. hx, xαi 6= 0} is countable.

Proposition IV.5. (Schwartz Inequality)

If x and y are vectors in an IPS, then
kxk·kyk ≥ |hx, yi| .
y
Proof. The case y − 0 is trivial, so suppose y 6= 0. The vector kyk by itself
forms an orthonormal set, so applying Bessel’s inequality to any x gives
 2
2  y 
kxk ≥ hx, kyk i
|hx, yi|2
=
kyk2
kxk2kyk2 ≥ |hx, yi|2


28 Real Analysis Qual Seminar
p
Proposition IV.6. kxk = hx, xi really is a norm.
Proof. The first two properties of norm are clearly satisfied:
kxk = 0 ⇐⇒ x = 0, kxk ≥ 0,
kαxk = |α|·kxk.
To see the triangle inequality,
kx + yk2 = hx + y, x + yi = hx + y, xi + hx + y, yi
= hx, xi + hx, yi + hy, xi + hy, yi linearity
= kxk2 + 2 Rehx, yi + kyk2 z+z
2
= Re z
≤ kxk2 + 2 |hx, yi| + kyk2 Re z ≤ |z|
≤ kxk2 + 2kxk·kyk + kyk2 |hx, yi| ≤ kxk·kyk
= (kxk + kyk)2


Proposition IV.7. (Parallelogram Identity)

kx + yk2 + kx − yk2 = 2 kxk2 + kyk2 .
Proof. Add the two formulae
kx + yk2 = kxk2 + 2 Rehx, yi + kyk2
kx − yk2 = kxk2 − 2 Rehx, yi + kyk2.


Example. ( )
∞
.. X
`2 := {xn}∞
n=1 . |xn |2 < ∞
n=1
with the inner product
D E ∞
X
{xn }∞ ∞
n=1 , {yn }n=1 := xn y n .
n=1
Real Analysis Qual Seminar 29

Example.  
Z
.
L2 := f : X → C .. |f |2 dµ < ∞
X
with the inner product Z
hf, gi := f g dµ.
X

Example.  
Z
.
L2H := f : X → H .. kf (x)k2H dµ <∞
X
with the inner product
Z
hf, gi := hf (x), g(x)iH dµ.
X

IV.1. Bases.
Definition IV.8. An orthonormal basis of a Hilbert space H is a maximal
orthonormal set S (i.e., no other orthonormal set contains S as a proper
subset).

Theorem IV.9. Every Hilbert space has an orthonormal basis.

Proof. Let C be the collection of orthonormal subsets of H. Order C by
inclusion:
S1 ≺ S 2 ⇐⇒ S 1 ⊆ S2 .
Then (C, ≺) is a poset.
It is also nonempty since {x/kxk} is an orthonormal set, ∀x ∈ H.
Now let {Sα }α∈A be any linearly ordered subset of C.
Then ∪α∈A Sα is an orthonormal set which contains each Sα and is thus an
upper bound for {Sα }α∈A .
Since every linearly ordered subset of C has an upper bound, apply Zorn’s
Lemma and conclude that C has a maximal element.
This maximal element is an orthonormal set not properly contained in any
other orthonormal set. 
30 Real Analysis Qual Seminar

Theorem IV.10. (Orthogonal Decomposition and Parseval’s Rela-

tion)
Let S = {xα }α∈A be an orthonormal basis for a Hilbert space H. Then
∀y ∈ H:
X X
2
y= hxα , yixα , and kyk = |hxα , yi|2 .
α∈A α∈A

Proof. Proving Bessel’s inequality, we saw that

X
|hxα , yi|2 ≤ kyk2 ,
α∈A

and that there are at most countably many nonzero summands.

Collect these α’s for which hxα , yi 6= 0 to obtain a sequence {αj }∞
j=1 .
PN
As a positive-term series, j=1 |hxαj , yi|2 is monotone increasing.
It is also bounded above by kyk2 .
Thus, it converges to a finite limit as N → ∞. Define
n
X
yn := hxαj , yixαj .
j=1

We want to show lim yn = y. For n > m,

2
X n n
X
2
kyn − ym k = hxαj , yixαj = |hxαj , yi|2 ,

j=m+1 j=m+1

by the Pythagorean Thm. Letting n, m → ∞ shows {yn } is Cauchy.

Since H is Hilbert, it is complete and {yn } must converge to some y 0 ∈ H.
Let xα be any element of S. If ∃` α = α` , then by the continuity of norms:
* n
+
X
hy − y 0 , xα` i = lim y − hxαj , yixαj , xα` = hy, xα` i − hy, xα` i = 0
n→∞
j=1

and if not,
* n
+
X
hy − y 0 , xα i = lim y− hxαj , yixαj , xα =0
n→∞
j=1
Real Analysis Qual Seminar 31

because
* n
+
X
hy − y 0 , xα i = lim y− hxαj , yixαj , xα
n→∞
j=1
n
X

= hy, xα i − lim hxαj , yixαj , xα
n→∞
j=1
∞
X

=0− hxαj , yixαj , xα hy, xα i = 0 for α 6= α`
j=1
∞
X

= hy, xαj i xαj , xα
j=1
X∞
= hy, xαj i · 0 hxαj , xα i = 0 for α 6= α`
j=1
=0
So y − y 0 is orthogonal to every xα in S. Since S is a orthonormal basis, this
means we must have y − y 0 = 0. Thus
n
X
y = lim hxαj , yixαj ,
n→∞
j=1

and we have shown the first part. Finally,

2
X n

0 = lim y − hxαj , yixαj
n→∞
j=1
 * n + n 2 
X
 X
2
= lim kyk − 2 Re y, hxαj , yixαj + hxαj , yixαj 
n→∞
j=1 j=1
n n
!
X
 X
= lim kyk2 − 2 Re hxα , yi y, xα + hxα , yixα 2
j j j j
n→∞
j=1 j=1
n n
!
X 
 2 X  
= lim 2
kyk − 2  x αj , y  + hxαj , yi2 xαj 2
n→∞
j=1 j=1
32 Real Analysis Qual Seminar

n
!
X
2
= lim kyk − |hxαj , yi|2
n→∞
j=1
X
= kyk2 − |hxα , yi|2
α∈A
gives Parseval’s Relation:
X
kyk2 = |hxα , yi|2 .
α∈A


Definition IV.11. The coefficients hxα , yi are the Fourier coefficients of y

with respect to the basis {xα }.

IV.2. The Riesz Representation Theorem Again.

Definition IV.12. Let M be a closed subspace of H. Then M is a Hilbert
space under the inner product it inherits as a subspace of H. Define the
orthogonal complement of M to be
.
M := {x ∈ H .. hx, yi = 0 ∀y ∈ M}.

Theorem IV.13. (Projection Theorem)

If M is a closed subspace of H, then H = M ⊕ M⊥. That is, ∀x ∈ H, x can
be uniquely expressed as x = y + z, where y ∈ M, z ∈ M⊥. Moreover, y, z
are the unique elements of M and M⊥ whose distance to x is minimal.

If y ∈ H, then ϕy (x) = hx, yi defines a functional on H.

By the linearity of inner prod, it is a linear functional.
By the Schwartz inequality8,
kϕy k = supkxk≤1 kϕy (x)k = supkxk≤1 |hx, yi| ≤ supkxk≤1 kxk·kyk ≤ kyk
Shows that this functional is bounded/ continuous.

8Recall, the Schwartz ineq is just the Hölder ineq when p = 2.

Real Analysis Qual Seminar 33

Theorem IV.14. (Riesz Representation Theorem for Hilbert Spaces)

If ϕ ∈ H∗ , then ∃!y ∈ H such that ϕ(x) = hx, yi ∀x ∈ H. Also, kϕk = kyk.
Proof. If ϕ is the zero functional, then y = 0 and we’re done.
Otherwise, consider the nullspace
.
M := {x ∈ H .. ϕ(x) = 0}.
M is a proper closed subspace of H and M⊥ 6= {0} by the Projection Thm.
Thus we can find z ∈ M⊥ with kzk = 1 and define
u := ϕ(x)z − ϕ(z)x.
Then  
ϕ(u) = ϕ ϕ(x)z − ϕ(z)x = ϕ(x)ϕ(z) − ϕ(z)ϕ(x) = 0
shows that u ∈ M and hence that u ⊥ z. Thus,
0 = hz, ui = hz, ϕ(x)z − ϕ(z)xi
= hz, ϕ(x)zi − hz, ϕ(z)xi linearity
= ϕ(x)kzk2 − ϕ(z)hz, xi hz, zi = kzk2
= ϕ(x) − hx, ϕ(z)zi kzk = 1
Thus, ϕ(x) = hx, yi where y = ϕ(z)z.
As for uniqueness, if hx, yi = hx, y 0 i for all x, take x = y − y 0 and get
ky − y 0 k2 = hy − y 0 , y − y 0 i = hy − y 0 , yi − hy − y 0 , y 0 i = 0 =⇒ y = y0.

This shows that y 7→ ϕy is a conjugate linear isometry of H onto H∗.

Definition IV.15. Isomorphisms of Hilbert spaces are those transformations

U : H1 → H2 which preserve the inner product:
hU x, U yiH2 = hx, yiH1 ∀x, y ∈ H1 .
Such operators are called unitary.
For U : H → H, unitary operators are also characterized by U ∗ = U −1, where
T ∗ is the Hilbert space adjoint of T ∈ L(H) and is defined by
hT ∗ x, yi = hx, T yi.
34 Real Analysis Qual Seminar

V. A Practical Guide to Integral Problems

This talk covers the relation between Riemann and Lebesgue integration,
when you can differentiate under an integral, and other practical applications
of Lebesgue theory to standard integral problems.

V.1. Some related theorems.

Theorem V.1. Let f : [a, ∞) → R be locally R-integrable. Then

Z ∞ Z ∞ Z
f ∈ L1 [a, ∞) ⇐⇒ R |f | dx < ∞ and R f dx = L f dµ.
a a [a,∞)

Proof. (⇒) f ∈ L1 =⇒ f + , f − ∈ L1 .
Define fn = f +χ[a,a+n) so that fn ≤ fn+1, and fn → f + and fn ∈ L1.
Now for A := [a, ∞),
Z ∞ Z a+n
R + R
f dx = lim f +dx def of improper int
a n→∞
Za
L
= lim f +dx R = L on bounded
n→∞ [a,a+n)
Z
= lim L fn dµ def fn
n→∞ A
Z
= L lim fn dµ MCT
n→∞
ZA
= L f + dµ
A
R
R ∞ −
R
L −
Similarly, a f dx = Af dµ, so
Z Z Z ∞ Z ∞
L L + −

R + −

R
f dµ = f −f dµ = f −f dx = f dx.
A A a a

(⇐) Define fn (x) = |f |χ[a,a+n] so that fn % |f | and fn is R-integrable.

(support is compact)
Real Analysis Qual Seminar 35
R
R [a+n]
R
R [a,a+n]
R
L
Then a fn dx exists and a fn dx = [a,a+n] fn dµ, so

Z Z
L L
|f | dµ = lim fn dµ MCT
A n→∞ A
Z a+n
R
= lim fn dx R = L on bounded
n→∞ a
Z a+n
R
= lim |f |dx def of fn
n→∞ a
Z ∞
= R |f |dx def of impr int
a
<∞ hypothesis

shows that f ∈ L1 . 

R
Theorem V.2. Define F (t) = X f (x, t) dµ(x) for f : X × [a, b] → C.
(1) What is sufficient for F to be continuous? lim F (t) = F (t0), ∀t0 .
t→t0 Z
∂f
(2) What is sufficient for F to be differential? F 0 (t) = ∂t (x, t) dµ(x).
X

Royden:
(1) (i) ft (x) = f (x, t) is a measurable function of x for each fixed t.
(ii) ∀t, |f (x, t)| ≤ g(x) ∈ L1(X).
(iii) limt→t0 f (x, t) = f (x, t0) for each x (i.e., f (x, t) is continuous in t
for each x).
The proof follows by applying DCT to f (x, tn), where tn → t0 .
(2) (i) ∂f∂t
exists on X × [a, b],
∂f
(ii) ∂t is bounded on X × [a, b],
(iii) f is bounded on X × [a, b],
(iv) For each fixed t, f is a measurable function of x.
36 Real Analysis Qual Seminar

(3) Alternatively:
∂f
(i) ∂t exists on X × [a, b],
 
 ∂f 
(ii)  ∂t (x, t) ≤ g(x) ∈ L1(X) on X × [a, b].

advantages: f, ∂f
∂t need not be bounded
disadvantages: need f ∈ L1 .

Proof. (of the second version).

Pick any sequence {tn } ⊆ [a, b] with tn → t0 . Then define

f (x, tn) − f (x, t0)

hn (x) := .
tn − t 0

Then ∂f
∂t
(x, t0) = limn→∞ hn (x), so ∂f
∂t
(x, t0) is measurable as a limit of mea-
surable functions. It follows that ∂f
∂t (x, t) is measurable. By the mean value
theorem, there is a t between tn and t0 for which

f (x, tn) − f (x, t0) = (tn − t0 ) ∂f

∂t (x, t).

Then
 
 ∂f 
|hn (x)| ≤ sup  ∂t (x, t) ≤ g(x),
t∈[a,b]

since taking the supremum can only make it larger.

Invoke the dominated convergence theorem again and get
Z Z
F (tn ) − F (t0 ) ∂f
F 0 (t0) = lim = lim hn (x) dµ(x) = ∂t (x, t)dµ(x).
tn − t 0

Finally, exploit the compactness of [a, b] and [Rudin 4.2]:

lim g(x) = g(t) ⇐⇒ lim g(xn) = g(t), ∀{xn } ⊆ X, xn → t.

x→t n→∞


Real Analysis Qual Seminar 37

V.2. Solutions to the Nasty Integrals.

R
1. f : X → [0, ∞] is measurable and X f dµ = c where 0 < c < ∞. Let
α ∈ R be a constant. Show that

Z   α  
∞ 0 < α < 1
f (x)
lim n log 1 + dµ = c α=1
n→∞ X n 

0 α>1
Proof.

case i) α = 1.  n
f (x)
By basic calculus, 1 + n increases to ef (x) for each x, so
 n
f (x) n→∞
gn (x) = log 1 + −−−−→ f (x) ∈ L1 (increasing).
n
R R
Then limn→∞ X gn (x) = X f (x) = c by MCT.

case ii) α > 1. R

Note that f (x) ≥ 0, X f dµ = c ≤ ∞ show f is finite µ-ae, i.e.:
  
∃M < ∞ and ∃E ∈ M s.t. µE = 0 and f Ẽ(x) ≤ M,
(so E is where f is bounded). Since we can always find N such
f (x)
that n ≥ N =⇒ M n < 1, n ≤ Mn , µ-au. We’re concerned
with n → ∞, so this means f (x) n < 1 for our purposes. Hence,
α > 1 =⇒ α − 1 > 0 implies
 α−1
f (x)
0≤ n < 1. (V.1)

We need
  α 
f (x)
n log 1 + ≤ αf (x),
n
so define



t α
G(t) = n log 1 + n − αt.
38 Real Analysis Qual Seminar

Then G(0) = 0. Also,

 α−1 
0 nt
G (t) = nα +tα − 1 α diff
 α−1 
nt
≤ nα − 1 α drop the tα


t α−1
= n −1 α simp
 α−1
f (x)
<0 t = f (x), n < 1,
so G is decreasing and G(t) < 0 for t > 0, i.e.,
  α 
f (x)
n log 1 + ≤ αf (x).
n
Set  α  
f (x)
gn (x) = n log 1 + .
n
Since this is bounded by αf and f ∈ L1 by hypothesis, DCT gives
Z Z
lim gn dµ = lim gn dµ.
n→∞ X X n→∞
Now split the leading n and match the denominator:
  α    α nα
f
1−α α
gn (x) = n n log 1 + n =n 1−α
log 1 + nf α ,
so that   nα
fα
lim 1 + nα = ef 6ae eM
n→∞
shows   nα
fα
log 1 + nα
lim = 0.
n→∞ nα−1
case iii) α < 1.
 
fα
f > 0 =⇒ log 1 + nα
> 0,
R
and ifRwe define RA := {f > 0}, then µA > 0 because X f dµ > 0.
Thus A f dµ = X f dµ, and
 
fα
log 1 + nα > 0 on A.
Real Analysis Qual Seminar 39

But then   nα
fα
log 1 + nα n→∞
−−−−→ ∞
nα−1
because α < 1 =⇒ α − 1 < 0.
Thus, lim gn = ∞, so by Fatou’s Lemma,
Z Z Z
lim gn dµ ≥ lim gn = ∞ =⇒ lim gn dµ = ∞.
n→∞


40 Real Analysis Qual Seminar
Z ∞
e−xt
2. Define F (t) = dx, for t > 0.
0 1 + x2
a) Show that F is well-defined as an improper Riemann integral and
as a Lebesgue integral.
−xt
e
Riemann: 1+x 2 is continuous ∀t, so it is R-integrable on any bounded

interval (a, b). So only remains to show the convergence of

Z a
e−xt
lim 1+x2 dx.
a→∞ 0
Since the integrand is nonnegative,9
Z a Z b
−xt
e e−xt
b≥a =⇒ 1+x2
dx ≤ 1+x2
dx.
0 0
Thus, it suffices to consider a dominating function g(x):
e−xt 1 1
1+x2
≤ 1+x2
≤ x2
∀t > 0.
Since
Z a
1
 a a→∞
x2 dx = − x1 1 = − a1 + 1 −−−−→ 1
1
and the integrand is bounded by 1, ∀t > 0, we have
Z a
e−xt
1+x2 dx ≤ 2 ∀a,
0
and thus it converges as a → ∞.
R  e−xt 
Lebesgue: R+  1+x2  dµ exists because we can bound the integrand
as above.

b) Show F 00 (t) exists on (0, ∞).

e −xt ∂ϕ −xt
We have ϕ(t) = 1+x 1
2 ∈ L , so just need ∂t
(t) = − xe
1+x2
∈ L1 in
order to use the theorem and get
Z ∞ −xt
0 xe
F (t) = − 2
dx.
0 1 + x
9f (x) ≥ 0 =⇒ g(x) = R x f (t) dt is increasing.
0
Real Analysis Qual Seminar 41

Fix t > 0 and pick ε > 0 such that t − ε > 0.

Observe: x ≤ eεx for large enough x. Thus we pick M large enough
that x ≥ M =⇒ x ≤ eεx , and split the integral:

Z ∞ Z M Z ∞
−xe−xt −xe−xt −xe−xt
dx = dx + dx.
0 1 + x2 0 1 + x2 M 1 + x2

Then we have

Z M
 −xt  Z M
 xe   −xt 
  dx ≤ xe  dx,
 1 + x2 
0 0

which as a continuous function over a compact space is clearly

finite, and hence the integrand is in L1(0, M ). Also,

Z ∞
 
−xt  Z ∞  εx −xt 
−xe e e 
  dx ≤  
 1 + x2   1 + x2  dx by choice of M
M M
Z ∞  (ε−t)x 
e 
=  
 1 + x2  dx
M
Z ∞ (ε−t)x
e
= dx positive integrand
0 1 + x2

e(ε−t)x
So ε − t < 0 =⇒ ∈ L1 by (a).
 −tx 
1+x2
−xe−tx x2 e−tx
Thus 1+x2 ∈ L . Now if ∂t −xe
1 ∂
1+x2 = 1+x2 ∈ L1 , we’ll have

Z ∞
00 x2 e−tx
F (t) = dx.
0 1 + x2

To show this, find N such that x ≥ N =⇒ x2 ≤ eεx , and proceed

as before.
42 Real Analysis Qual Seminar

c) (Extra credit) Show F (t) satisfies F 00 (t)+F (t) = 1t . Compute F (t).

R ∞ x2 e−tx
We have F 00 (t) = 0 1+x2
dx from (b), so
Z ∞
00 x2 e−tx + e−tx
F (t) + F (t) = 2
dx
1 + x
Z0 ∞
(1 + x2)e−tx
= 2
dx
1 + x
Z0 ∞
= e−tx dx
 0 1 −tx ∞
= −te 0
= 0 − (− 1t ) = 1t .
Now solve the differential equation F 00 (t) + F (t) = 1t .

3. Let I be an open interval of R and suppose f : R → R such that

x 7→ ext f (x) is integrable for each fixed t ∈ I. Define F : I → R by
Z
F (t) = ext f (x) dx.
R
R
Show that F is differentiable with derivative F 0 (t) = R xe
xt
f (x) dx at
each t ∈ I.

Note: xetx f (x) may not be in L1! We would like to compute F 0 (t)
by Z  tn x 
0 e − e t0 x
F (t0) = lim f (x) dx, (V.2)
n→∞ tn − t 0
where {tn } is a sequence in I with tn → t0 .
To use DCT, we need to find g ∈ L1 such that
 tx 
 e n − e t0 x 
 
 tn − t0  ≤ g(x) ∀n.
Choose t0 ∈ I such that tn < t0 , ∀n. This is possible, since otherwise
there would be a subsequence of {tn } converging to sup{t ∈ I}.
<
. I is open and tn → t0 ∈ I.
By MVT,
 tx 
e n − et0 x  ≤ |xesx | · |tn − t0 | for some s ∈ [tn , t0].
Real Analysis Qual Seminar 43

Since s < t0 ,
 tx   tx  
 e n − et0 x   t0 x   e n − e t0 x   t0 x 

   
 tn − t0  ≤ xe  =⇒  tn − t0 f (x) ≤ xe f (x) , (V.3)

and we have a bound which no longer depends on n.

 0 
To see that g(x) = xet x f (x) is integrable, split the integral: pick
ε > 0 such that t0 + ε ∈ I and choose M be such that

x≥M =⇒ x ≤ eεx .

Now
Z M  Z M 
 t0 x   t0 x 
xe f (x) dx ≤ M e f (x) dx < ∞ t0 ∈ I, and
Z0 ∞   Z ∞ 0 
 t0 x   (t0 +ε)x 
xe f (x) dx ≤ e f (x) dx < ∞ t0 + ε ∈ I.
M M
R∞ 0  R0
Thus we have 0 xet x f (x) dx < ∞. For −∞ g(x) dx, pick ε > 0 such
that t0 − ε ∈ I and let M be such that

x≥M =⇒ x ≤ eεx ,
R∞
and proceed as for 0 g(x) dx.
Together, this gives g ∈ L1 . By (V.3), we can use the DCT in (V.2) to
obtain the result.

4. [2003] Let f be a bounded measurable function on [0, ∞). Show that

Z ∞
f (x)e−xt
F (t) = √ dx, t > 0
0 x

is continuously differentiable on (0, ∞).

Let us denote the integrand by ϕ(x, t) := f (x)e−xtx−1/2.

We would like to find F 0 (t) by choosing any sequence {tn } such that
44 Real Analysis Qual Seminar

tn → t0 and computing
Z ∞ 
0 f (x)e−xtn f (x)e−xt0
F (t0) = lim √ − √ dx
n→∞ 0 tn x t0 x
Z ∞
e−xtn − e−xt0 −1/2
= lim f (x) x dx
n→∞ 0 tn − t 0

Since f is bounded, |f (x)| ≤ M . Then

   −xt 
 e −xtn
− e −xt0   e n
− e −xt0 
f (x) x −1/2
≤ M  x −1/2
 tn − t 0   tn − t 0 .

Since t ranges over (0, ∞) and tn → t0 < ∞, we can certainly pick a

strict lower bound τ of {tn }, i.e. a number τ such that inf{tn } > τ .
By MVT,
 −t x   
e n − e−t0 x  ≤ xe−sx  · |tn − t0 | for some s ∈ [tn, t0 ].

Since s > τ ,
 −t x 
 e n − e−t0 x −1/2  −sx −1/2
  ≤ xe x  ≤ e−τ x x1/2 ∈ L1 .
 tn − t 0 x 

To verify the integrability of the dominating function, note that

u = x1/2 dv = e−x dx
2du = x−1/2dx v = −e−x

gives
Z h i∞ Z
e−x x1/2 = −e−x x1/2 +2 −e−x x−1/2dx
0
Z ∞
2 √
= (0 − 0) + 4 e−u du put u = x
0
√
= 4 · 2π
√
= 2 π.
Real Analysis Qual Seminar 45
R∞ √ R∞ pπ
Further, 0 e−x x−1/2dx = 2 π =⇒ 0 e−τ x x−1/2dx = 2 τ. Thus

Z ∞
0 e−xtn − e−xt0 −1/2
F (t0 ) = lim f (x) x dx
n→∞ 0 tn − t 0
Z ∞
e−xtn − e−xt0 −1/2
= lim f (x) x dx by DCT
0 n→∞ tn − t 0
Z ∞  −xt 
e − e−xt0
= f (x) lim x−1/2 dx
t→t 0 t − t0
Z0 ∞
∂ −xt

−1/2
= f (x) ∂t e x dx
0
Z ∞


= f (x) −xe−xt x−1/2 dx
0
Z ∞
=− f (x)e−xtx1/2 dx.
0

Let us denote this function

Z ∞
G(t) = − f (x)e−xtx1/2 dx.
0

Since we are required to show that F is continuously differential, we

must show that G is continuous.
2
Notice that another u-substitution with u = (tx)1/2, t
u du = dx
allows us to rewrite
Z ∞
2
G(t) = − 2t f (x)xe−x dx.
0

Thus, all we need to do is show the integral to be finite. Since f is

bounded by M , it suffices to show
Z ∞
2
xe−x dx < ∞.
0
46 Real Analysis Qual Seminar

Next, putting u = x2, 21 du = x dx,

Z ∞ Z ∞
−x2
xe dx = 2 1
e−u du
0
1
 0 −u ∞
= 2 −e 0
= 12 (0 − (−1))
1
= 2

which is as finite as it gets.

Z ∞
sin(x2t)
5. [2000] Show F (t) = 2
dx is continuous on R.
−∞ 1 + x

First, note that | sin x| ≤ 1 gives

 
 sin(x2t) 
  ≤ 1 ∈ L1 ,
 1 + x2  1 + x2

where the final inclusion is clear from

Z ∞
dx

2
= [arctan x]∞−∞ =
π
2 − − π2 = π.
−∞ 1 + x

Pick any {tn } ∈ R with tn → t0. Then

Z ∞
sin(x2tn )
lim F (tn ) = lim dx def of F
n→∞ n→∞ −∞ 1 + x2
Z ∞
sin(x2tn )
= lim 2
dx DCT
−∞ n→∞ 1 + x
Z ∞
sin(x2 limn→∞ tn )
= dx contin of sin x, mult by x2
−∞ 1 + x2
Z ∞
sin(x2t0 )
= 2
dx tn → t0
−∞ 1 + x
= F (t0).
Since this is true for all sequences tn → t0 , we have limt→t0 F (t) = F (t0 ).
Real Analysis Qual Seminar 47
R1
6. [1998] f ∈ C[0, 1] is such that 0 xnf (x) dx = 0 for n = 0, 1, 2, . . ..
Show that f ≡ 0.

By the Stone-Weierstraß Theorem, there is a sequence of polynomials

unif
{Pk (x)} such that Pk (x) −−−→ f (x). Then
Z Z
k→∞
Pk (x)f (x) dx −−−−→ [f (x)]2 dx by uniformity.

But since any polynomial may be written

m
X
P (x) = ai x i = a 0 + a 1 x + a 2 x 2 + · · · + a m x m ,
i=0
R1
the linearity of the integral and the hypothesis 0 xn f (x) dx = 0 give
Z 1 Z 1 Z 1 Z 1
P (x)f (x) dx = a0 dx + a1 x dx + · · · + am xm dx
0 0 0 0
= a0 · 0 + a 1 · 0 + a 2 · 0 + · · · + a m · 0
=0

Z
So Pk (x)f (x)dx = 0 ∀k
Z
f 2dx = 0

f2 ≡ 0 f2 ≥ 0

f ≡0

7. Compute the limits

R∞
−n

a) lim 0 1 + nx sin x
n dx
n→∞

Note that



x −n n→∞
sin x  ≤ 1 and 1+ −−−−→ e−x .
n n
48 Real Analysis Qual Seminar


−2
 x −n 
Then  1 + n  ≤ 1 + x2 ∀n ≥ 2, so the DCT gives
Z ∞ Z ∞
x −n x  x −n x
lim 1+ sin dx = lim 1 + sin dx
n→∞ 0 n n 0 n→∞ n n
Z ∞
= e−x sin(0) dx
0

= 0.

R∞
b) lim n(1 + n2x2 )−1 dx
n→∞ a

We do a u-substitution with u = nx, du = n dx:

Z ∞ Z ∞
2 2 −1 du
lim n(1 + n x ) dx = lim
n→∞ a n→∞ na 1 + u2

= lim [arctan u]∞

na
n→∞

 π
 a=0
π

2
= lim 2 − arctan na = 0 a>0.
n→∞ 

π a<0

8. a) Find the smallest constant c such that log(1 + et ) < c + t for

0 < t < ∞.
First, observe that
1+et
1 + e t < ec et =⇒ et
< ec .
Note that
t
1+et
lim 1+e
et = 2 and lim t = 1.
t→0 t→∞ e
1+et
Since et is monotonic, it is evidently monotonically decreasing:
d 1+et


dt e t = d 1
dt e t + 1 = dtd e−t = −e−t .
Thus, let e2 = 2 or c = log 2.
Real Analysis Qual Seminar 49
R1
b) Does lim n1 0 log(1 + enf (x) ) dx exist for every real f ∈ L1[0, 1], if
n→∞
f > 0?
1+enf (x)
From part (a), we get 1 < enf (x)
< 2, which gives
enf (x) < 1 + enf (x) < 2e nf (x)
 
nf (x)
nf (x) < log 1 + e < nf (x) + c (c = log 2)
Z 1 Z 1   Z 1
nf (x)
n f (x) dx < log 1 + e dx < n f (x) dx + c
0 0 0
(since integration is a pos linear functional and f ∈ L1.)
Z 1 Z 1   Z 1
1 nf (x)
f (x) dx < n log 1 + e dx < f (x) dx + nc
0 0 0

Then taking the limit as n → ∞, we get

Z 1 Z 1   Z 1
1 nf (x)
f (x) dx ≤ lim n log 1 + e dx ≤ f (x) dx.
0 n→∞ 0 0
Since f is integrable by hypothesis, the Sandwich Theorem gives
Z 1   Z 1
lim n1 log 1 + enf (x) dx = f (x) dx.
n→∞ 0 0
50 Real Analysis Qual Seminar

VI. The Hahn-Banach Theorem and applications

[Folland] It is not obvious that there are any nonzero bounded functionals
on an arbitrary normed vector space. That such functionals exist in great
abundance is one of the fundamental theorems of functional analysis.

[Reed & Simon] In dealing with Banach spaces, one often needs to construct
linear functionals with certain properties. This is usually done in two steps:
first one defines the linear functional on a subspace of the Banach space where
it is easy to verify the desire properties; second, one appeals to (or proves) a
general theorem which says that any such functional can be extended to the
whole space while retaining the desired properties. One of the basic tools the
second step is the following theorem,

Theorem VI.1. Let X be a vector space and p : X → R such that

(i) p(αx) = αp(x), ∀α ≥ 0, and
(ii) p(x + y) ≤ p(x) + p(y), ∀x, y ∈ X.
If S is a subspace of X and there is a linear functional
f : S → R such that f (s) ≤ p(s), ∀s ∈ S, then f may be extended to
F : X → R with F (x) ≤ p(x), ∀x ∈ X, with F (s) = f (s) ∀s ∈ S.
Proof. The idea of the proof is to first show that if x ∈ X but x ∈/ S, then
we can extend f to a functional having all the right properties on the space
spanned by x and S. We then use a Zorn’s Lemma / Hausdorff Maximality
argument to show that this process can be continued to extend f to the whole
space X.

(Sketch)
1. Consider the family
.
G := {g : D → R .. g is linear; g(x) ≤ p(x), ∀x ∈ D; g(s) = f (s), ∀s ∈ S},
where D is any subspace of X which contains S. So G is roughly the
collection of “all linear extensions of f which are bounded by p”.
Now G is a poset under


g1 ≺ g 2 ⇐⇒ Dom(g1) ⊆ Dom(g2) and g2  = g1 .
Dom(g1 )
Real Analysis Qual Seminar 51

2. Use Hausdorff maximality Principle (or Zorn) to get a maximal lin-

early ordered subset {gα } ⊆ G which contains f . Define F on the union
of the domains of the {gα } by F (x) = gα (x) for x ∈ Dom(gα ).

3. Show that this makes F into a well-defined linear functional which

extends f , and that F is maximal in that F ≺ G =⇒ F = G.

4. Show F is defined on all of X using the fact that F is maximal. Do

this by showing that a linear functional defined on a proper subspace
has a proper extension. (Hence F must be defined on all of X or it
wouldn’t be maximal.)


Proposition VI.2. (Hausdorff Maximality Principle)

(A, ≺) is a poset =⇒ ∃B ⊆ A such that B is a maximal linearly ordered
subset. I.e., if C is linearly ordered, then B ≺ C ≺ A =⇒ C = B or C = A.

Of course, the HBT is also readily extendable to the complex case:

Theorem VI.3. Let X be a complex vector space and p a real-valued func-
tion defined on X satisfying
p(αx + βy) ≤ |α|p(x) + |β|p(y) ∀x, y ∈ X, ∀α, β ∈ C with |α| + |β| = 1.
If S is a subspace of X and there is a complex linear functional
f : S → R such that |f (s)| ≤ p(s), ∀s ∈ S, then f may be extended to
F : X → R with |F (x)| ≤ p(x), ∀x ∈ X, with F (s) = f (s) ∀s ∈ S.
52 Real Analysis Qual Seminar

VI.1. Principle Applications of the HBT.

Most often, p(x) is taken to be the norm of the Banach space in question.
∗
1. M is a closed  subspace of X and x ∈ X\M =⇒ ∃f ∈ X such that
f (x) 6= 0, f M = 0. In fact, if δ = inf y∈M kx − yk, f can be taken to
satisfy kf k = 1 and f (x) = δ.

Define f on M + Cx by f (y + λx) = λδ for y ∈ M, λ ∈ C. Then

f (x) = f (0 + ·x) = 1 · δ = δ
but for m ∈ M ,
f (m) = f (m + 0) = 0 · δ = 0.
For λ 6= 0, we have
|f (y + λx)| = |λ|δ ≤ |λ| · kλ−1y + xk = ky + λxk
because δ = inf ky + xk ≤ kλ−1y + xk (putting in λ−1 for y). Using
p(x) = kxk, apply the HBT to extend f from M + Cx to all of X.

2. If x 6= 0, x ∈ X, then ∃f ∈ X ∗ such that kf k = 1 and f (x) = kxk.

M = {0} is trivially a closed subspace, so apply (1) with δ = kxk.

3. The bounded linear functionals on X separate points.

If x 6= y, then (2) shows ∃f ∈ X ∗ such that f (x − y) 6= 0. I.e.,

f (x) 6= f (y). This result indicates that X ∗ is BIG.

4. If x ∈ X, define x̂ : X ∗ → C by x̂(f ) = f (x), so x̂ ∈ X ∗∗ . Then

ϕ : x 7→ x̂ is a linear isometry from X into X ∗∗.

x̂(αf + βg) = (αf + βg)(x) = αf (x) + βg(x) = x̂(αf ) + x̂(βg),

so x̂ is linear. This verifies that x̂ ∈ X ∗∗.
For ϕ(x) = x̂, ϕ(ax + by) is defined by
ax +ˆ by(f ) = f (ax + by) = af (x) + bf (y) = ax̂(f ) + bŷ(f ),
Real Analysis Qual Seminar 53

so ϕ(ax+by) = ax + ˆ by = ax̂+bŷ = aϕ(x)+bϕ(y) shows that ϕ : x 7→ x̂

is linear. Finally,
|x̂(f )| = |f (x)| ≤ kf k · kxk
shows that
|x̂(f )| |f (x)| kf k · kxk
kx̂k = sup = sup ≤ sup = kxk.
kf k≤1 kf k kf k≤1 kf k kf k≤1 kf k

To get the reverse inequality, note that (2) provides a function f 0 for
which |x̂(f0)| = |f0 (x)| = kxk and kf k = 1. Then
kx̂k = sup |x̂(f )| ≥ |x̂(f0)| = kxk.
kf k=1

We saw this for Hilbert spaces, but this example is applicable to

general Banach spaces and requires none of the Hilbert space machinery
(orthonormal basis, projection theorem, etc.) as the HBT takes care
of a lot.

VI.2. Corollaries to the HBT.

1. If X is a normed linear space, Y a subspace of X, and f ∈ Y ∗ , then

there exists F ∈ X ∗ extending f and satisfying kF kX ∗ = kf kY ∗ .
Proof. Apply HBT with p(x) = kf kY ∗ kxkX ∗ . 

2. Let X be a Banach space. If X ∗ is separable, then X is separable.

Proof. Let {fn } be a dense set in X ∗ . Choose xn ∈ X, kxnk = 1 so
that
|fn (xn)| ≥ kfn k/2.
Let D be the set of all finite linear combinations of the kxnk with
rational coefficients. Since D is countable, we just need to show that
D is dense in X.
If D is not dense in X, then there is a y ∈ X\D and a linear functional
f ∈ X ∗ such that f (y) 6= 0 but f (x) = 0 for all x ∈ D, by application
(1).
54 Real Analysis Qual Seminar

Let {fnk } be a subsequence of {fn} which converges to f . Then

kf − fnk k ≥ |(f − fnk )(xnk )|
= |(fnk )(xnk )|
≥ kfnk k/2
which implies kfnk k → 0 as k → ∞. Thus f = 0. .<
Therefore D is dense and X is separable. 
The example of `1 and `∞ shows that the converse of this theorem
doesn’t hold. In fact, this corollary offers a proof that `1 is not the
dual of `∞ , provided you can show `∞ is not separable.
Real Analysis Qual Seminar 55

VII. The Baire Theorem and Consequences

Definition VII.1. D is dense in X iff D̄ = X, equivalently, iff ∀U open in
X, U ∩ D 6= ∅.

∼
Definition VII.2. E is nowhere dense iff Ē is dense in X.
Definition VII.3. X is meager 10 iff X is a countable union of nowhere dense
sets.
Theorem VII.4. Let (X, d) be a complete metric space. (Note: can substi-
tute LCH for complete, using f.i.p. def of compactness). Then
T∞
a) If {On }∞
n=1 are open dense subsets of X, then n=1 On is dense in X.

b) No nonempty open subset of X is a countable union of nowhere dense

sets. In particular, X is not.
Proof. The idea of the proof S is straightforward: Suppose that X is a com-
plete metric space and X = ∞ n=1 An with each An nowhere dense. We will
construct a Cauchy sequence {xm } which stays away from each An so that
its limit point x (which is inSX by completeness) is in no An , thereby contra-
dicting the statement X = ∞ n=1 An .

Since A1 is nowhere dense, we can find x1 ∈

/ Ā1 and an open ball B1 about
x1 such that B1 ∩ A1 = ∅, with the radius of B1 smaller than 1.

Since A2 is nowhere dense, we can find x2 ∈ B1 \Ā2. Let B2 be an open ball

about x2 such that B2 ∩ A2 = ∅, with the radius of B2 smaller than 21 .

Proceeding inductively, we obtain a sequence {xn} where xn ∈ Bn−1\A¯n,

B¯n ⊆ Bn−1, and Bn ∩ An = ∅.

This sequence is Cauchy because n, m ≥ N implies that for xn, xm ∈ BN ,

N →∞
ρ(xn , xm) ≤ 21−N + 21−N = 22−N −−−−−→ 0.
Let x = limn→∞ xn . Since xn ∈ BN for n ≥ N , we have
x ∈ B¯N ⊆ BN −1.
S
/ AN −1 for any N , which contradicts X = ∞
Thus x ∈ n=1 An . 
10Formerly called “of first category”.
56 Real Analysis Qual Seminar

VII.1. Applications of Baire.

1. Q is NOT a Gδ .

T∞
Proof. Suppose it were. Then Q = n=1 On , where the On are open.
T
Note: Q ⊆ On =⇒ Q ⊆ On , ∀n, so Q dense in R =⇒ On dense
in R, ∀n.

Let {qk }∞
k=1 be an enumeration of Q.
Consider the singleton set (not the sequence!) {qn }.

Then {qn } closed =⇒ {qn }∼ open. Also, {qn }∼ is dense in R. (It

contains all but 1 of the rationals, so, e.g. the sequence {qn − m1 }∞
m=1 ⊆
∼
{qn} and has qn as a limit point.)

Then On ∩ {qn }∼ is open and dense. To see dense, note that

On ∩ {qn }∼ = Q ∼ {qn }.
Then
\ \ \ 
∼ ∼
(On ∩ {qn} ) = ( On ) ∩ {qn } = Q ∩ (R ∼ Q) = ∅,
<
So ∅ is dense in R, by Baire’s Theorem. . 
S
However, Q is an Fσ : Q = ∞ n=1 {qn }.

2. A meager set with Lebesgue measure 1.

We start by constructing a nowhere dense set with positive measure.
Let Cα be the Cantor set formed by removing an open interval of
α
length 3n+1 from each of the remaining 2n pieces, where 0 < α < 1.
T
Then Cα = ∞ n
n=0 Cα is closed as an intersection of closed sets and
Cα = Cα . To show Cα is nowhere dense, it suffices to show that Cα
contains no nonempty open set.11 If O 6= ∅ is open, then it contains
11If a closed set F contains no open set, then it is nowhere dense: we use contrapositive. Suppose (F ) ∼
is not dense. Then ∃x ∈ F which is not a limit point of Fe. If no sequence in Fe converges to x, every point
of Fe must be at least ε > 0 away from x. Thus F contains Bε (x). The conditions are actually equivalent.
Real Analysis Qual Seminar 57

Figure 4. Construction of the Cantor set with measure α.

_
3
C0

C1

C2

C3

C4

an interval A of positive length ε > 0. Each of the 2n intervals of Cαn

(nth step of construction) are of length ` < 2−n, so for large enough
N , we have ` < 2−N < ε. Thus A is longer than any component of
CαN and hence cannot be contained in CαN or Cα . So Cα is nowhere dense.

Now we use the measure lemma12 to compute the measure of Cα :

∞
X ∞
X
 
α2n α 2 n α 1
µCα = 1 − 3n+1 =1− 3 3 =1− 3 1−2/3 = 1 − α.
n=0 n=0

Thus, every singleton {Cα } is a trivially a countable union of nowhere

dense sets which have positive Lebesgue measure.

12The Measure Lemmae.

Proposition VII.5. Let (X, A, µ) be a measure space.

a) If {Ak } is an increasing sequence (Ak ⊆ Ak+1 ) of sets of A, then µ(∪Ak ) = lim µAk .

b) If {Ak } is a decreasing sequence (Ak+1 ⊆ Ak ) of sets of A, and µAn < ∞ for some n, then µ(∩Ak ) =
lim µAk .
(µAn < ∞ is necessary, else let Ak = (k, ∞).)

Proposition VII.6. Let (X, A) be a measurable space and let µ be a finitely additive function µ : A → [0, ∞]
such that µ(∅) = 0. Then µ is a measure if either
a) lim µAk = µ(∪Ak ) for each increasing sequence {Ak } ⊆ A; or
T
b) lim µAk = 0 holds for each decreasing sequence {Ak } ⊆ A with Ak = ∅.
58 Real Analysis Qual Seminar

However, to complete
S∞ this in the manner suggested by Royden, we
now consider P = k=1 C1/k .
∞
!
[
µP = µ C1/k = lim (1 − k1 ) = 1.
k→∞
k=1
So P is a union of a countable infinite collection of nowhere dense sets,
and P has Lebesgue measure 1.

VII.2. Consequences of Baire.

Theorem VII.7 (Open Mapping Theorem). X, Y are Banach spaces, T ∈

L(X, Y ). Then T surjective =⇒ T open.
Proof. Nasty. (technical and longish) 

Corollary VII.8. If X, Y are Banach and T ∈ L(X, Y ) is bijective, then T

is an isomorphism.
Proof.
T is an isomorphism ⇐⇒ T −1 ∈ L(X, Y )
⇐⇒ T −1 is continuous
⇐⇒ T is open
But this last condition is exactly the result of the OMT;
T bijective =⇒ T surjective =⇒ T open.


Definition VII.9. The graph of T is the set

.
Γ(T ) = {(x, y) ∈ X × Y .. y = T x}.

Definition VII.10. T ∈ L(X, Y ) is closed iff Γ(T ) is a closed subspace of

X ×Y.
Real Analysis Qual Seminar 59

Theorem VII.11 (Closed Graph Theorem). If X, Y are Banach spaces, and

the linear map T : X → Y is closed, then T is bounded.
Proof. Let π1 , π2 be the projections of
.
Γ(T ) = {(x, y) ∈ X × Y .. y = T x}
to X and Y respectively:
π1 (x, T x) = x and π2 (x, T x) = T x.
Obviously, π1 ∈ L(Γ(T ), X) and π2 ∈ L(Γ(T ), Y ).
Since X, Y are complete, so is X × Y . Hence T closed implies Γ(T ) is a
closed subspace (by definition), and a closed subspace of a complete space is
complete, so Γ(T ) is also complete.

Now π1 : Γ(T ) → X is a bijection, hence an isomorphism by the corollary

to OMT, i.e., π1−1 is bounded.
But then T = π2◦π1−1 is bounded. 

Figure 5. T as a composition of projections

Y

-1 1
1

CGT restated: Let X, Y be Banach spaces, and T : X → Y be linear.

(1) Then T is bounded ⇐⇒ Γ(T ) is closed.
(2) Then T is continuous ⇐⇒ (xn → x, T xn → y =⇒ y = T x).

Note. For X, Y be Banach spaces, and S : X → Y unbounded ,

(a) Γ(S) is not complete,
(b) T : X → Γ(S) is closed but not bounded, and
(c) T −1 : Γ(S) → X is bounded and surjective but not open.
60 Real Analysis Qual Seminar

Theorem VII.12 (Uniform Boundedness Principle, aka Banach-Steinhaus

Theoreom). Suppose X, Y are normed vector spaces and and A ⊆ L(x, y).
a) If supT ∈A kT xk < ∞ for all x in some nonmeager set D ⊆ X, then
supT ∈A kT k < ∞.
b) If X is Banach and supT ∈A kT xk < ∞ ∀x ∈ X, then supT ∈A kT k <
∞.
Proof of (a). Let
.. \ .
En = {x ∈ X . sup kT xk ≤ n} = {x ∈ X .. kT xk ≤ n}.
T ∈A
T ∈A

Thus the En are closed as intersections of preimages of closed sets under con-
tinuous maps. Since supT ∈A kT xk ≤ N, ∀x ∈ D by hypothesis, EN contains
a nontrivial closed ball
B(x0, r), r > 0.
But then
kxk < R =⇒ (x − x0) ∈ EN
=⇒ kT xk ≤ kT (x − x0)k + kT x0k ≤ 2N,
so kxk ≤ r =⇒ kT xk ≤ 2N ∀T ∈ A, which implies that
2N
kT k ≤ r < ∞.

Proof of (b). X is a nonempty Banach space, so X is nonmeager by Baire.
(Baire’s Theorem says that every complete metric space is nonmeager.) Then
just apply (a). 
Rephrase of the UBP:
Either ∃M < ∞ such that kT k ≤ M, ∀T ∈ A,
or else supT ∈A kT xk = ∞, ∀x in some dense Gδ ⊆ X.

Geometrically, either there is a ball B ⊆ Y (with radius M , center 0) such

that every T ∈ A maps the unit ball of X into B, or there exists an x ∈ X
(in fact, a whole dense Gδ of them) such that no ball in Y contains T x, for
all T ∈ A simultaneously.
Real Analysis Qual Seminar 61

VII.3. Related Problems.

1. Find a Banach space X, a normed linear space Y , and a continuous

linear bijection f : X → Y such that f −1 is not continuous. (Note: Y
better not be Banach!)



Let X := L2 [0, 1], k · k2 and Y := L2[0, 1], k · k1 , and define f :
X → Y by f (x) = x.
f is clearly bijective and linear. To see f is continuous, it suffices to
show f is continuous at 0:

Let ϕ ∈ X and fix ε > 0. Since X, Y are finite measure spaces,

p < q =⇒ kϕkp ≤ kϕkq .
Thus, kϕk1 ≤ kϕk2 shows kϕk2 ≤ ε =⇒ kf (ϕ)k1 = kϕk1 ≤ ε.
To see f −1 is not continuous, consider {ϕn} ⊆ Y defined by
1
ϕn (x) = p .
x + 1/n
R1
Now ϕn(x) ≤ ϕn+1(x) ≤ ϕ(x) = √1x , ∀n, x. Since 0 √1x dx = 2 =⇒
ϕ ∈ L1 , the MCT gives
Z 1
dx
lim kϕk1 = lim p
0 x + 1/n
Z 1
= (lim ϕn)dx
0
Z 1
dx
= √
0 x
= 2 = kϕk1.
However,
Z 1 1/2
dx n→∞
kϕnk2 = = (log[1 + n])1/2 −−−−→ ∞
0 x + 1/n
shows that lim f −1(ϕn) does not converge.
R1 R1
Indeed, 0 ϕ2 dx = 0 dx x
= ∞. Hence, lim f −1(ϕn) 6= f −1(ϕ) shows
that f −1 is not continuous.
62 Real Analysis Qual Seminar

2. Let V be a Banach space.

a) If V is infinite-dimensional, construct an unbounded linear operator
f :V →V.
b) If V is finite-dimensional, show that every linear operator on it is
bounded.
Since V is a vector space, it has some basis {eλ }λ∈Λ . Then if {cλ }λ∈Λ
is any
Psequence (or net) in R, ∃!f : V → V such that for every element
v = vλ eλ of V , we have
X
f (v) = cλ v λ e λ .
(Just define f (eλ ) = cλ eλ .)
a) For V infinite-dimensional, {cλ } unbounded implies f unbounded.
b) For V finite-dimensional, kf k ≤ k · max{cλ } for some k. (Basically,
Λ finite =⇒ max{cλ } exists.)
See any book on Quantum Dynamics or Functional Analysis for
the examples of the position & momentum operators (which are un-
bounded), e.g., Reed & Simon.