This Should Help Your Lazy Ass In Analisys B

Charles A. Mongus, world famous jazz composer and bassist

January 5, 2025

Contents
1 Normed Spaces 2
1.1 The space of linear applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Compactness and continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Finite dimensional spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1 Normed Spaces
Proposition 1.1

A subset C ⊂ X is a closed subset of (X, ||·||) whenever the limit of any convergent sequence
(xn )n ⊂ C belongs to C.

In other words, C is closed if and only if once a sequence of elements of C is converging, the limit
cannot escape C.

Proof

1 C closed and (xn )n ⊂ C =⇒ limn xn = x ∈ C.

We argue by contradiction that x ̸∈ C: then x must be in X \ C which must be open.
But if that is an open set then there must exist a ball of x for some r > 0 such that
B(x, r) ⊂ X \ C. But due to the definition of convergence there must exist a N such that
||xn − x|| < r ∀ n ⩾ N , but then that xn should be outside C which is a contradiction!
2 limn xn = x ∈ C =⇒ C closed and (xn )n ⊂ C.
We argue by contradiction that C is not closed, so X \ C must be not open. Not open
sets are such that there exist x ∈ X \ C such that for every r > 0 you can’t find a ball
that it is entirely in X \ C which means B(x, r) ∩ C = ̸ ∅. Then pick xn ∈ B(x, n1 ) ∩ C
so that every xn also belongs to C. Clearly this is a sequence of n such that for any n ⩾ 1
we have ||xn 9 − x|| < n1 and this means that limn xn = x. We picked x ∈ X \ B so we
have found a limit of a sequence of elements of C that doesn’t belong to C which is a
contradiction!

Lemma 1.1
Young’s inequality. Let p > 1 and let q < 1 be its conjugate exponent. Then, for any
nonnegative a, b ∈ R it holds
1 1
ab ⩽ ap + bq .
p b

Theorem 1.1
Holder Inequality. Let 1 ⩽ p ⩽ ∞ and p1 + 1q = 1. Assume that f ∈ Lp (S, µ) and g ∈ Lq (S, µ).
Then f · g ∈ L1 and
||f g||1 ⩽ ||f ||p ||g||q .

1
Proof
The proof is trivial if p = 1 and q = 1. Remember Young’s inequality
1 p 1 q
ab ⩽ a + b .
p q

Now let’s say a = |f (s)| and b = |g(s)|. Now our inequality becomes
1 1
|f (s)g(s)| ⩽ |f (s)|p + |g(s)|q for µ-a.e. on S.
p q
Now we integrate over S and we get
Z Z Z
1 p 1
|f (s)g(s)| ⩽ |f (s)| + |g(s)|q
S p S q S
1 p 1 q
=⇒ ||f g||1 ⩽ ||f ||p + ||g||q
p q

which means that ||f g||1 is finite and therefore f g ∈ L1 (S, µ). To end the proof let’s substitute
f with λf, ∀ λ > 0. We get
λp p 1 q
λ ||f g||1 ⩽
||f ||p + ||g||q ∀λ > 0
p q
λ λp p 1 q
=⇒ ||f g||1 ⩽ ||f ||p + ||g||q ∀λ > 0
λ pλ qλ
λp−1 p 1 q
=⇒ ||f g||1 ⩽ ||f ||p + ||g||q ∀ λ > 0.
p λq

q
1
Now choose λ = ||f ||p · ||g||qp . When we substitute this value for λ, we get:
 p p
1
λp ∥f ∥p · ∥g∥qq
∥f ∥pp = ∥f ∥pp .
p p
 p p
1
Expanding ∥f ∥p · ∥g∥qq , we get:
p
∥g∥pq

1 p

· ∥g∥qq = .
∥f ∥p ∥f ∥pp

Substituting this, we have:

∥g∥p
q
λp ∥g∥pq
∥p
∥f
∥f ∥pp =  ∥fp
∥
p = p
.
p p p
Substitute back into the inequality:

∥g∥pq 1
λ∥f g∥1 ⩽ + ∥g∥qq .
p q
∥g∥p ∥g∥qq
 
1 1 1 1
Since p
q
+ q = ∥g∥pq p + q and p + q = 1, we get:

λ∥f g∥1 ⩽ ∥g∥pq .

p
1 q
Dividing by λ = ∥f ∥p ∥g∥q :
∥f g∥1 ⩽ ∥f ∥p ∥g∥q .
This is Hölder’s inequality:
∥f g∥1 ⩽ ∥f ∥p ∥g∥q .

2
 Theorem 1.2
For any 1 ⩽ p ⩽ ∞, Lp (S, µ) is a vector space and ||·||p is a norm.

Remark

For 1 < p < ∞ the triangular inequality

||f + g||p ⩽ ||f ||p + ||g||p ∀ f, g ∈ Lp (S, µ)

is known as Minkowski’s Inequality.

Proof

We already know that if f ∈ Lp (S, µ) then λf ∈ Lp (S, µ). Homogeneity and uniqueness are
also existent for ||·||p so in order to show that Lp (S, µ) is a vector space we only need to prove
that if f, g ∈ Lp (S, µ) then f + g ∈ Lp (S, µ) and ||·|| is a norm.
Fix f, g ∈ Lp (S, µ). We know that for any x, y ∈ R we get
p
1 1 1 p 1 p
x+ y ⩽ |x| + |y|
2 2 2 2
since this mapping r → rp is convex. This also means that
p
|x + y| ⩽ 2p−1 (|x|p + |y|p ) .

and this implies in particular that

p
|f (s) + g(s)| ⩽ 2p−1 (|f (s)|p + |g(s)|p ) for µ-a.e. on s ∈ S.

If we integrate over S we get:

Z Z Z 
p
|f (s) + g(s)| ⩽ 2p−1 |f (s)|p + |g(s)|p
S S S

which means  
p p p
||f + g||p ⩽ 2p−1 ||f ||p + ||g||p

which means that f + g ∈ Lp (S, µ).

We now must prove the Minkowski’s inequality. We know that
Z Z
p p p−1
||f + g||p = |f + g| dµ = |f + g| |f + g| dµ
S S

but since we know that |f + g| ⩽ |f | + |g| then

Z Z
p
||f + g||p ⩽ |f ||f + g|p−1 dµ + |g||f + g|p−1 dµ.
S S

Call ψ = |f + g|p−1 . It clearly belongs to Lq (S, µ) because

p
|ψ|q = |f + g|p−1 q = |f + g|


since q(p − 1) = p

so Z  q1 Z  q1
p
q p
||ψ||q = |ψ| = |f + g| = ||f + g||pq < ∞
S S

And this means that |ψ| ∈ L (S, µ) =⇒ ψ ∈ Lq (S, µ). We also know that |f | ∈ Lp (S, µ) so
q 1

we can apply Holder’s inequality with f and ψ so that

Z
p
|f ||f + g|p−1 dµ = ||f ψ||1 ⩽ ||f ||p ||ψ||q = ||f ||p ||f + g||pq
S

and Z
p
|g||f + g|p−1 dµ ⩽ ||g||p ||f + g||pq
S

3
 So that p p
p
||f + g||p ⩽ ||f ||p ||f + g||pq + ||g||p ||f + g||pq .
p
Dividing by ||f + g||pq ̸= 0 (otherwise the proof is trivial) we get

p− p
||f + g||p q
⩽ ||f ||p + ||g||p .

Proposition 1.2

Let (Ω, F, P) be a probability triplet. then the following holds:

L∞ (Ω, F, P) ⊂ Lp (Ω, F, P) ⊂ Lq (Ω, F, P) ⊂ L1 (Ω, F, P).

Of course, this result remains valid for every other measure space (S, S, µ) as long as µ(S) < ∞. In
the special case in which
S=N F = P(N)
P
and µ(A) is the counting measure µ(A) = k∈A δk (A), A ∈ N then knowing that sequences n 7→ f (n)
can be identified as functions over N of the type f : N → R we see that
∞
( )
X
L (S, µ) = L (S, µ) = ℓ (N) = x = (xn )n ; ||x||1 :=
1 1 1
|xn | < ∞ .
actual functions, not equivalence classes n=1

This means that ℓ1 (N) is a L1 space for some special choice of S and µ. Since we chose our measure
as the counting measure, we get
Z ∞
X ∞
X
|f (n)| dµ(n) = |f (n)| = |xn |.
N n=1 n=1

Cool!

Proposition 1.3

Let p ⩾ 1 be given. We define the set

∞
( )
X p
p
ℓ (N) = x = (xn )n∈N ⊂ R such that |xn | < ∞ .
n=1

Then, ℓp (N) is a vector space. Moreover, if

∞
! p1
X p
||x||p := |xn | ∀ x = (xn )n ∈ ℓp (N)
n=1
 
then ℓp (N), ||·||p is a normed space.

1.1 The space of linear applications

Proposition 1.4

Let (X, ||·||X ) and (Y, ||·||Y ) be two normed spaces and let L; X 7→ Y be a linear application.
The following are equivalent:
1 L is continuous on X;
2 L is continuous at x = 0;
3 there is a positive constant C > 0 such that
||L(x)||Y ⩽ C ||x||X ∀ x ∈ X.

4
 Proof

Of course 1 =⇒ 2 , but now let’s prove 3 . Consider the definition of continuity at 0 for the
function L: this means that for any ε > 0 there exists a constant δ > 0 such that

||x − 0X ||X ⩽ δ =⇒ ||L(x) − L(0Y )||Y < ε.

We know that L is linear, so L(0X ) = 0Y and therefore we get

||x||X ⩽ δ =⇒ ||L(x)||Y < ε.

Choose ε = 1 so that we get

||x||X ⩽ δ =⇒ ||L(x)||Y < 1.

δ
Let x ∈ X \ {0X } and set y = ||x||X x. Now we have that

δ δ
||y||X = x = ||x||
=δ
X
||x||X X ||x||
 X

and, due to the linearity of L,

 
δ δ δ 1
||L(y)||Y = L x = ||L(x)||Y = ||L(x)||Y < 1 =⇒ ||L(x)||Y < ||x||X .
||x||X Y
||x||X ||x||X δ
1
So, setting C = δ proves 3 . We still need to prove 1 =⇒ 3 : ’tis easy, since by linearity we
have
||L(x) − L(y)||X = ||L(x − y)||X ⩽ C ||x − y||Y ∀ x, y ∈ X
and this clearly implies the continuity of L at any x ∈ X (actually, the uniform continuity).

Definition 1.1
If (X, ||·||X ) and (Y, ||·||Y ) are two normed spaces, we denote by L(X, Y ) the space of continuous
linear applications from X to Y . If X = Y we simply denote L(X) = L(X, X).

Proposition 1.5

If (X, ||·||X ) and (Y, ||·||Y ) are two vector spaces and X is of finite dimension, any linear application
L : X 7→ Y is continuous.

Remark

If dim(X) = n and dim(Y ) = p, the space L(X, Y ) can be identified with the space Mn×p (R)
of matrices with n lines and p rows.

1.2 Compactness
Definition 1.2
Let (X, ||·||X ) be a normed space and let K ⊂ X. We say that K is compact if every sequence
(xn )n contains a subsequence which converges to some x ∈ K.

Of course if K is compact then it is closed.

Lemma 1.2
If K is a compact subset of a normed space (X, ||·||X ) then K is closed and there exists M > 0
such that supx∈K ||x|| ⩽ M which means that K is bounded.

5
 Proposition 1.6

Product of compact spaces. Let (X1 , ||·||1 ) and (X2 , ||·||2 ) be two compact normed spaces and
let X = X1 × X2 . Then (X, ||·||+ ) and (X, ||·||max ) are compact normed spaces.

Remember that
||x||+ = ||x1 ||1 + ||x2 ||2
and
||x||max = max(||x1 ||1 , ||x2 ||2 )

Proof

Let (ξn )n ⊂ X be a given sequence. This means, being in a product space, that there exist
two sequences (x1 )n ⊂ X1 and (yn )n ⊂ X2 such that ξn = (xn , yn ) for any n. Consider the
following subsequences:

• since (X1 , ||·||1 ) is compact, there exists a subsequence (xφ(n) )n ⊂ X1 with a limit x ∈ X1 ;
• take the analogous subsequence (yφ(n) )n ⊂ X2 . Since (X2 , ||·||2 ) is compact there is
another (sub-)subsequence (yϕ(φ(n)) )n of (yφ(n) )n that converges to y ∈ X2 ;
• now take the subsequence (xϕ(φ(n)) )n of (xφ(n) )n which is still convergent in X1 to x.

We can easily see that the subsequence

(ξϕ(φ(n)) )n = (xϕ(φ(n)) ), (yϕ(φ(n)) ) n

is converging in X to x = (x, y). So every subsequence ξn contains a subsequence that converges

to a point in X and this means that X is compact.

Of course, the above result readily extends to any finite product of compact normed spaces. On R
it is easy to describe a large class of compact sets:

Lemma 1.3
Let R be endowed with the absolute value, | · |. Any interval [a, b] ⊂ R is compact.

Proposition 1.7

Let (X, ||·||) be a normed space and let K be a compact subset of X. If A ⊂ K is a closed
subset then A is compact.

Corollary

Heine-Borel theorem. A subset K of RN (where RN is endowed with, say, the usual Euclidean
norm) is compact if and only if it is closed and bounded.

Proof

⇐= We know that a compact subset of RN is closed and bounded.

=⇒ Let K ⊂ RN be closed and bounded. Being bounded, there exists R > 0 such that

K ⊂ [−R, R]N .

Since K is closed, from the previous proposition it is sufficient to prove that [−R, R]N is
a compact subset of RN which is the same as checking that [−R, R] is a compact subset
of R, since every closed subset of a compact subset is compact.

This corollary can be reformulated as:

Every bounded sequence of RN has a convergent subsequence.

6
1.3 Compactness and continuous functions
Proposition 1.8

Let (X, ||·||X ) and (Y, ||·||Y ) be two normed spaces and let f : X → Y be continuous. If K ⊂ X
is a compact subset of X then f (K) is a compact subset of Y .

Proof

Let (yn )n be a sequence of f (K). It means that there is a sequence (xn )n ⊂ K
such that
yn = f (xn ) for any n. Since K is compact, then there exists a subsequence xφ(n) n of (xn )n
which converges to x ∈ K. This means that limn xφ(n) − x X = 0, but since f is continuous
then we know that


lim f xφ(n) − f (x) Y = 0 ⇐⇒ lim yφ(n) − f (x) Y = 0.
n n

Since y = f (x) ∈ f (K), we get that any sequence of f (K) has a subsequence which converges
to a limit in f (K), which means that f (K) is compact.

This has the following consequence:

Theorem 1.3
Let (X, ||·||) be a normed space and let K ⊂ X be compact. Let f : K → R be continuous.
Then, f assumes its maximum and minimum on K.

Proof

We know that f (K) is a compact subset of R and thanks to the Heine-Borel theorem we know
that any compact subset of R is closed and bounded (and viceversa...). So f is bounded and
the fact that it reaches its maximum and minimum inside K is a simple consequence of the fact
that f (K) is closed. Let
M = sup {f (x), x ∈ K} .
By definition there is a sequence (xn )n ⊂ K such that limn f (xn ) = M . This sequence
(f (xn ))n lies in f (K) which is compact and is closed. But this means that also its limit M lies
in f (K) (i.e.) there is x ∈ K such that f (x) = M . This shows that M is the maximum value of
f and by proceeding in an analogous manner we can show the same thing for the minimum.

Theorem 1.4
Heine Theorem. Let (X, ||·||X ) and (Y, ||·||Y ) be two normed spaces and let K ⊂ X be compact.
Assume that f : K → Y is continuous. Then f is uniformly continuous on K.

Proof
Suppose that f is not uniformly continuous. This means that there exists ε0 > 0 such that

∀ δ > 0, ∃x, y ∈ K, with ||x − y||X < δ and ||f (x) − f (y)||Y ⩾ ε0 .
1
Now chossing δ = n, n ∈ N this allows to build two sequences (xn )n and (yn )n such that
1
||xn − yn ||X < and ||f (xn ) − f (yn )||Y ⩾ ε0 ∀ n ∈ N.
n


Since K is compact we can extract a subsequence of xn that we call xφ(n) n and that converges


to some x0 ∈ K. It follows that yφ(n) n also converges to x0 (why?). Since f is continuous



we get that limn→∞ xφ(n) = f (x0 ) = limn→∞ f yφ(n) in Y , i.e.



lim f xφ(n) − f yφ(n) Y = 0
n→∞



but this contradicts the fact that f xφ(n) − f yφ(n) Y
⩾ ε0 for each n ∈ N+ .

7
 So, in this case taking two sequences that get closer and closer does not correspond to the fact also
their functions get closer and closer... and this is not possible.

1.4 Finite dimensional spaces

Proposition 1.9

Let (X, ||·||) be a finite dimensional normed vector space with dim(X) = d and let {e1 , . . . , ed }
be a basis for X. Then, there are positive constants C0 , C1 > 0 such that
d
X d
X d
X
C0 |xi | ⩽ xi ei ⩽ C1 |xi | ∀ (x1 , . . . , xd ) ∈ Rd .
i=1 i=1 i=1

This proposition asserts that if dim(X) = d then any norm ||·|| is related to the ||·||1 norm of Rd .
This translates in the following:

Proposition 1.10

If X is a finite dimensional vector space, all norms over X are equivalent.

So there is no weird norm, but everything is comparable to the simple ||·||1 norm. This proposition
also allows us to identify in a continuous way a finite dimensional space (X, ||·||) and the space Rd where
d is the dimension of X. Indeed, introducing a basis {e1 , . . . , ed } of X, the mapping
Φ : X → Rd
Pd
which associates Φ(x) = (x1 , . . . , xd ) to some x = i=1 xi ei ∈ X, is a bijection from X to Rd which
is continuous whose inverse is also continuous. This results in the following:

Corollary

If (X, ||·||) is a finite dimensional vector space and K ⊂ X is closed and bounded then K is
compact.

Again, this is very specific to finite dimensional spaces and, as we shall see, this actually characterizes
finite dimensional spaces. Indeed, in infinite dimensional normed spaces, the closed unit ball cannot be
compact. This shows that, in infinite dimensional spaces, the compact subsets do not coincide with
closed and bounded subsets!! We first state the following technical lemma:

Lemma 1.4
Riesz Lemma. Let (X, ||·||) be a normed vector space and let Y be a closed subspace of X (i.e.
Y is closed in X and Y is a linear subspace of X). If Y ̸= X then for any ε ∈ (0, 1) there exists
x ∈ X with ||x|| = 1 such that
inf ||x − y|| ⩾ 1 − ε.
y∈Y

Remark

̸ X is a closed subspace then for any ε ∈ (0, 1) there is some unit

This lemma asserts that if Y =
vector x ∈ X such that dist(x, Y ) ⩾ 1 − ε.

Proof

Let z ∈ X \ Y . Since Y is closed and z ∈

/ Y , one has
α = dist(z, Y ) = inf ||z − y|| > 0.
y∈Y
α
Pick ε ∈ (0, 1). There exists y ∈ Y such that ||y − z|| ⩽ 1−ε (otherwise we would get
α
dist(z, Y ) ⩾ 1−ε > α!). Notice that y ̸= z so that r : + ||y − z|| > 0. Set
1
x := (z − y).
r

8
Clearly, ||x|| = 1. Let y ∈ Y be given. One can write
1
||x − y|| = ||z − y − ry||
r
and since Y is a linear subspace y + ry ∈ Y so that ||z − y − ry|| ⩾ α. Therefore ||x − y|| ⩾
α
r ⩾ 1 − ε by assumption on r = ||z − y||. Since this is true for any y ∈ Y , this proves the
result.