Part I, Chapter 1

Lebesgue spaces

The objective of the four chapters composing Part I is to recall (or gently
introduce) some elements of functional analysis that will be used throughout
the book: Lebesgue integration, weak derivatives, and Sobolev spaces. We
focus in this chapter on Lebesgue integration and Lebesgue spaces. Most of
the results are stated without proof, but we include various examples. We
refer the reader to Adams and Fournier [3], Bartle [16], Brezis [46], Demengel
and Demengel [83], Evans [95], Grisvard [106], Malý and Ziemer [132], Rudin
[161, Chap. 11], Rudin [162], Sobolev [172], Tartar [181], Yosida [193].
In this book d is the space dimension, and D denotes a nonempty subset
of Rd . Vectors in Rd , d ≥ 2, and vector-valued functions are denoted in
bold font. We abuse the notation by denoting position vectors in Rd in bold
font as well. Moreover k·kℓ2 (Rd ) denotes the Euclidean norm in Rd (we write
k·kℓ2 when the context is unambiguous), and a·b denotes the Euclidean inner
product between two vectors a, b ∈ Rd . For every pair of integers m ≤ n, we
use the notation {m:n} := {p ∈ N | m ≤ p ≤ n}.

1.1 Heuristic motivation

If one strictly restricts oneself to computational considerations, the Riemann
integral is the only notion of integration that is needed in numerical analysis,
since the objects that one manipulates in practice are piecewise smooth func-
tions (e.g., polynomials) defined on meshes. However the Riemann integral
becomes useless when one starts to investigate questions like passage to the
limit. For instance assume that one has an interval I := (a, b), a sequence of
finite partitions of this interval, say (Ih )h∈H , and a sequence of real-valued
functions (vh )h∈H defined on I such that vh is smooth on each subinterval
of Ih for all h ∈ H. Here H is a countable set with 0 as unique accumulation
point. In the context of finite elements, the index h refers to the size of the
mesh that is used to construct the function vh . Assume also that one can
2 Chapter 1. Lebesgue spaces

a priori prove that the sequence (vh )h∈H is Cauchy in the following sense:
Rb
for every ǫ > 0,
there is h(ǫ) such that a |vh1 (x) − vh2 (x)| dx ≤ ǫ for all
h1 , h2 ∈ 0, h(ǫ) ∩ H. One may then wonder whether vh converges to some
object with interesting properties when h → 0. The answer to this question
becomes very intricate if one restricts oneself to the Riemann integral, but
it becomes simple if one adopts Lebesgue’s point of view. Since the above
question arises constantly in this book, we now take some time to recall the
key ingredients of Lebesgue’s theory.

1.2 Lebesgue measure

To define the Lebesgue integral of a function defined on a subset D of Rd , one
needs to measure the volume of sets in Rd . For every bounded rectangular
parallelepiped R := [a1 , b1 ] × · · · × [ad , bd ], with ai ≤ bi for all i ∈ {1:d},
we define
Q the Lebesgue (outer) measure of R to be its volume, i.e., we set
|R| := i∈{1: d} (bi − ai ).

Definition 1.1 (Lebesgue’s outer measure). Let E be a set in Rd . Its

Lebesgue’s outer measure is defined as
nX [ o
|E|∗ := inf |Ri | | E ⊂ Ri , Ri ∈ R(Rd ) , (1.1)
i∈N i∈N

where R(Rd ) is the set of all the rectangular parallelepipeds in Rd .

We expect |E|∗ to be a reasonable estimate of the volume of E if E is a

reasonable set. The outer Lebesgue measure has the following properties: (i)
|∅|∗ = 0; (ii) If E ⊂ F , then S
|E|∗ ≤ |F |∗ ; P
(iii) If {Ei }i∈N is a countable collec-
tion of subsets of R , then | i∈N Ei |∗ ≤ i∈N |Ei |∗ (countable subadditivity
d

property; see [161, Thm. 11.8]).

Example 1.2 (Countable sets). The outer Lebesgue measure of a count-

S 1
able set A := k∈N {xk } is zero. Let indeed ǫ > 0. We have {xk } ⊂ R(xk , ǫ d ),
where R(z, r) is the cube of side r centered at z. Hence |xk |∗ ≤ ǫ, i.e.,
|xk |∗ = 0 since ǫ > 0 is arbitrary. Invoking subadditivity yields |A|∗ = 0. For
example this implies that outer measure of the set of the rational numbers is
zero, i.e., |Q|∗ = 0. ⊓
⊔

Definition 1.3 (Lebesgue’s measure of a set). A set E ⊂ Rd is said to

be Lebesgue-measurable if |S|∗ = |S ∩E|∗ +|S ∩E c |∗ for every subset S ⊂ Rd ,
where E c is the complement of E in Rd .

It turns out that not all the sets of Rd are Lebesgue-measurable, but the class
of Lebesgue-measurable sets (in short, measurable sets) of Rd , say L(Rd ), is
sufficiently vast that we will only encounter measurable sets in this book. In
Part I. Elements of functional analysis 3

particular, (i) If E is measurable, then E c is also measurable; (ii) Open sets

of Rd and closed sets of Rd are measurable (so that all the usual geometric
objects, e.g., parallelepipeds or balls, are measurable); (iii) Countable unions
and intersections of measurable sets are measurable.
Henceforth, the map |·| : L(Rd ) → [0, ∞] such that |E| := |E|∗ for all
E ∈ L(Rd ) is called (d-dimensional) Lebesgue measure. Since the action of the
Lebesgue measure on measurable sets is simply the outer Lebesgue measure,
we infer that (i) |∅| = 0; (ii) If A, B ∈ L(Rd ) and A ⊂ B, then |A| ≤ |B|; (iii)
The countable subadditivity property holds true on countable collections of
measurable sets. By restricting our attention to measurable sets, the property
we have gained is that |A1 ∪ A2 | = |A1 | + |A2 | for disjoint measurable sets
(since (A1 ∪A2 )∩A1 = A1 and (A1 ∪A2 )∩Ac1 = A2 ). Moreover,S if {Ak }k∈N is a
countable
S family of
P measurable disjoint sets, the union k∈N Ak is measurable
and | k∈N Ak | = k∈N |Ak |; see [161, Thm. 11.10].
Example 1.4 (Null sets). Let A ⊂ Rd . If |A|∗ = 0, then A is measurable.
Let indeed S ⊂ Rd . Then |A ∩ S|∗ ≤ |A|∗ = 0, i.e., |A ∩ S|∗ = 0. Moreover
|S|∗ ≥ |S ∩Ac |∗ = |S ∩Ac |∗ + |S ∩A|∗ , and the subadditivity property implies
that |S|∗ ≤ |S ∩ A|∗ + |S ∩ Ac |∗ , whence the result. ⊓
⊔
Example 1.5 (Cantor set). To define the Cantor ternary set, one starts
with the interval [0, 1], then one deletes the open middle third from [0, 1],
leaving two line segments: [0, 13 ] ∪ [ 32 , 1]. Next the open middle third of each
of the two remaining segments is deleted, leaving four line segments: [0, 91 ] ∪
[ 29 , 13 ] ∪ [ 23 , 79 ] ∪ [ 98 , 1]. This process is continued ad infinitum. Setting C0 :=
[0, 1] and Cn := 31 Cn−1 ∪ ( 32 + 13 Cn−1 ), the Cantor ternary set is defined
by C∞ := {x ∈ [0, 1] | x ∈ Ck , ∀k ∈ N}. Then C∞ is measurable (as the
complement of a countable union of measurable sets), |C∞ | ≤ |Ck | for all
k ∈ N, so that |C∞ | = 0, but it can be shown that C∞ is not countable. ⊓ ⊔
Definition 1.6 (Equality a.e.). Let D ⊂ Rd be a measurable set, i.e., D ∈
L(Rd ). Let f : D → R and g : D → R be two functions. We say that f and g
are equal almost everywhere if |{x ∈ D | f (x) 6= g(x)}| = 0. Henceforth we
write f (x) = g(x) for a.e. x ∈ D, or f = g a.e. in D.
Definition 1.7 (ess sup, ess inf). Let D ⊂ Rd be a measurable set and let
f : D → R be a function. We define

ess sup f (x) := inf{M ∈ R | f (x) ≤ M for a.e. x ∈ D}, (1.2a)

x∈D
ess inf f (x) := sup{m ∈ R | f (x) ≥ m for a.e. x ∈ D}. (1.2b)
x∈D

Definition 1.8 (Measurable function). Let D ⊂ Rd be a measurable set.

A function f : D → R is said to be measurable if {x ∈ D | f (x) > r} is
measurable for all r ∈ R; see also [161, Def. 11.13].
The meaning of the above definition is that one says that a function is mea-
surable if all its upper level sets are (Lebesgue) measurable.
4 Chapter 1. Lebesgue spaces

Lemma 1.9 (Characterization). Let D ⊂ Rd be a measurable set. Let

f : D → R. The function f is measurable iff any of the following statements
holds true:
(i) For all r ∈ R, the set {x ∈ D | f (x) > r} is measurable.
(ii) For all r ∈ R, the set {x ∈ D | f (x) ≥ r} is measurable.
(iii) For all r ∈ R, the set {x ∈ D | f (x) < r} is measurable.
(iv) For all r ∈ R, the set {x ∈ D | f (x) ≤ r} is measurable.
Proof. Item (i) is the
T definition of the measurability of f . The identity {x ∈
1
D | f (x) ≥ r} = n∈N {x ∈ D | f (x) > r − n+1 } proves that (i) implies
c
(ii). {x ∈ D | f (x) < r} = D
T ∩ {x ∈ D | f (x) ≥ r} proves that (ii) implies
1
(iii). {x ∈ D | f (x) ≤ r} = n∈N {x ∈ D | f (x) < r − n+1 } proves that (iii)
implies (iv), and {x ∈ D | f (x) > r} = D ∩ {x ∈ D | f (x) ≤ r}c proves that
(iv) implies the measurability of f . (See also [161, Thm. 11.15].) ⊓
⊔
Since every open set in R is a countable union of open intervals, the above
result shows that f is measurable iff {x ∈ D | f (x) ∈ U } is measurable for
every open set U of R.
Example 1.10 (Measurable functions). Functions that are piecewise
continuous and more generally all the functions that are integrable in the
Riemann sense are measurable.
Corollary 1.11 (Measurability and equality a.e.). Let D ⊂ Rd be a
measurable set. Let f : D → R be a measurable function. Let g : D → R be a
function. If f = g a.e. in D, then g is measurable.
Proof. See Exercise 1.2. ⊓
⊔
Example 1.12 (Measurability). Let D := (0, 1). Let f : D → R be
defined by f (x) := x. Let C∞ be the Cantor set (see Example 1.5). Let
g : D → R be defined by g(x) := −2x if x ∈ C∞ , and g(x) := x if x 6∈ C∞ .
The function f is measurable since it is continuous. Recalling that |C∞ | = 0,
g is also measurable by virtue of Corollary 1.11 since f = g a.e. in D. ⊓
⊔

1.3 Lebesgue integral

We say that g : D → R is a simple non-negative function if there exist m ∈ N,
a collection of disjoint measurable sets {Ak }k∈{1: m}Pin D, and a collection
of non-negative numbers {vk }k∈{1: m} such that g = k∈{1: m} vk 1Ak (where
1Ak (x) := 1 if x ∈ Ak and
R 1Ak (x) := P0 otherwise). The Lebesgue integral of
g over D is defined by D g(x) dx := k∈{1: m} vk |Ak |.

Theorem 1.13 (Simple functions). Let D ∈ L(Rd ). Let f : D → [0, ∞]

be a non-negative measurable function. Then there exist simple functions
{gk }k∈N s.t. 0 ≤ g1 ≤ g2 . . . ≤ f and limk→∞ gk (x) = f (x) for all x ∈ D.
Part I. Elements of functional analysis 5

Proof. See [162, Thm. 1.17]. ⊓

⊔

Definition 1.14 (Lebesgue integral). Let f be a non-negative measurable

function. The Lebesgue integral of f over D is defined in [0, ∞] as follows:
Z nZ o
f (x) dx := sup g(x) dx | g is simple non-negative and g ≤ f .
D D

This definition agrees with the Riemann integral of f if f is Riemann-

integrable.
R + Let fR be measurable but not necessarily non-negative. If either
D
f (x) dx or D f − (x) dx Ris finite, where Rf ± := max(±f,R0), the Lebesgue
integral of f is defined by D f (x) dx := D f + (x) dx − D f − (x) dx. We
say that f is (Lebesgue-)integrable
R R on D if both
R terms in the sum are fi-
nite, and we set D |f (x)| dx := D f + (x) dx + D f − (x) dx. Observe that by
R R
construction D f (x) dx ≤ D |f (x)| dx.
An important
R property of the Lebesgue integral is that if f is integrable
on D, then D |f (x)| dx = 0 if and only if f vanishes everywhere on D up
to a set of zero measure. This leads us to introduce a notion of equivalence
classes. Two functions are said to belong to the same class if they coincide
almost everywhere (henceforth, a.e.), i.e., everywhere but on a set of zero
Lebesgue measure. Elements of Lebesgue spaces are, strictly speaking, equiv-
alence classes, although we refer to them simply as functions that are defined
almost everywhere. For instance the function φ : (0, 1) → {0, 1} that is 1 on
the rational numbers and is zero otherwise is in the same equivalence class as
the zero function. Hence φ = 0 a.e. on (0, 1). Integrals are always understood
in the Lebesgue sense throughout
R this book. RWhenever the context is unam-
biguous, we simply write D f dx instead of D f (x) dx. We refer the reader
to [162, Chap. 1] for more elaborate notions on the measure theory.

Example 1.15 (Cantor set). Let f : [0, 1] → R be such that f (x) := 1 if

x is in C∞ (see Example 1.5) and f (x) := 0 otherwise. Then f is measurable
R1
(see Corollary 1.11) and 0 f (x) dx = 0. ⊓
⊔

Remark 1.16 (Literature). It is reported in Denjoy et al. [84, p. 15] that

Lebesgue explained his approach to integration as follows: “I have to pay a
certain sum, which I have collected in my pocket. I take the bills and coins
out of my pocket and give them to the creditor in the order I find them until
I have reached the total sum. This is the Riemann integral. But I can proceed
differently. After I have taken all the money out of my pocket, I order the
bills and coins according to identical values and then I pay the several heaps
one after the other to the creditor. This is my integral.” To get a clearer
connection with the integration process, one could say that Lebesgue went to
a grocery store every day in a month, bought items, and asked for credit until
R 30
the end of the month. His debt at the end of a 30-day month is 0 f (t) dt,
where f (t) is the amount of money he owes per day. What Lebesgue has
R 30
described above are two different ways to compute 0 f (t) dt. ⊓
⊔
6 Chapter 1. Lebesgue spaces

1.4 Lebesgue spaces

This section introduces the Lebesgue spaces and reviews their key properties.

1.4.1 Lebesgue space L1 (D)

Definition 1.17 (Space L1 ). Let D be an open set in Rd . L1 (D) is the
vector space composed of all the real-valued measurable functions that are
1
RLebesgue-integrable on D, and we equip L (D) with the norm kf kL1 (D) :=
D
|f | dx to make it a normed space.

Theorem 1.18 (Monotone convergence, Beppo Levi). Let D be an

open set in Rd . Let (fn )n∈N be a sequence of functions in L1R(D) such that
0 ≤ f0 ≤ f1 ≤ . . . ≤ fn ≤ fn+1 ≤ . . . a.e. on D and supn∈N D fn dx < ∞.
Then fn (x) converges to a finite limit for a.e. x in D. Denoting by f (x) the
limit in question, f is in L1 (D) and limn∈N kfn − f kL1 (D) = 0.

Proof. See [46, Thm. 4.1] or [162, Thm. 1.26]. ⊓

⊔

Theorem 1.19 (Lebesgue’s dominated convergence). Let (fn )n∈N be a

sequence of functions in L1 (D) such that:
(i) fn (x) → f (x) a.e. in D.
(ii) There is g ∈ L1 (D) such that |fn (x)| ≤ g(x) a.e. in D for all n ∈ N.
Then f ∈ L1 (D) and fn → f in L1 (D).

Proof. See [16, p. 123], [46, Thm. 4.2], [162, Thm. 1.34]. ⊓
⊔

Example 1.20. Let fn : D := (0, 1) −→ R, n ∈ N, with fn (x) := 1 if x < n1

and fn (x) := x otherwise. We have fn (x) → x a.e. in D and fn (x) ≤ g ≡ 1
a.e. in D. Hence fn → 1 in L1 (D). ⊓
⊔

Theorem 1.21 (Fischer–Riesz). L1 (D) equipped with the L1 -norm from

Definition 1.29 is a Banach space.

Proof. See [3, Thm. 2.16], [16, p. 142], [46, Thm. 4.8], [162, Thm. 3.11]. ⊓
⊔

Remark 1.22 (Lebesgue vs. Riemann). The two key results the notion
of Lebesgue integration gave us that were missing in the Riemann integration
are Lebesgue’s dominated convergence theorem and the fact that L1 (D) is
now complete, that is, it is a Banach space. This answers the question raised
in §1.1. ⊓
⊔

Theorem 1.23 (Pointwise convergence). Let (fn )n∈N be a sequence in

L1 (D) and asume that f ∈ L1 (D) is such that kfn − f kL1(D) → 0. Then there
exist a subsequence (fnk )k∈N and a function g ∈ L1 (D) such that fnk (x) →
f (x) a.e. in D and |fnk (x)| ≤ g(x) a.e. in D for all k ∈ N.

Proof. See [46, Thm. 4.9], [162, Thm. 3.12]. ⊓

⊔
Part I. Elements of functional analysis 7

Example 1.24 (Dirac mass). The assumption that there exists some
g ∈ L1 (D) s.t. |fn (x)| ≤ g(x) a.e. in D for all n ∈ N, is crucial to ap-
ply Lebesgue’s dominated convergence theorem. For instance consider the
1 n
sequence of functions in L1 (R) s.t. fn (x) := 0 if |x| R > n and fn (x) := 2
otherwise. We have fn (x) → 0 for a.e. x in R and R |fn (x)| dx = 1, but fn
does not converge in L1 (R). Reasoning by contradiction, let us assume that
fn → f in L1 (R). Theorem 1.23 implies that there is a subsequence (fnk )k∈N
s.t. fnk (x) → f (x) for a.e. x in R. For all x 6= 0, we have fnk (x) = 0 for all nk
1
such that nk > |x| . This implies that f (x) = 0 for a.e. x in R. This argument
R
shows thatR R |f (x)| dx = 0, but since we assumed that fn → f in L1 (R), we
also have R |f (x)| dx = 1, which is a contradiction. ⊓
⊔

Definition 1.25 (Space L1loc (D)). Let D be an open set in Rd . The elements
of the following space are called locally integrable functions:

L1loc (D) := {v measurable | ∀ compact K ⊂ D, v|K ∈ L1 (K)}. (1.3)

Definition 1.26 (Support). Let D be a measurable set in Rd . The support

in D of a function ϕ : D → R, henceforth denoted supp(ϕ), is defined to be
the closure in D of the subset {x ∈ D | ϕ(x) 6= 0}.
Definition 1.27 (Space C0∞ (D)). We denote by C0∞ (D) the space composed
of the functions from D to R that are C ∞ and whose support in D is compact.
The members of C0∞ (D) are called test functions.
Theorem 1.28 (Vanishing
R integral). Let D be an open set in Rd . Let
v ∈ Lloc (D). Then D vϕ dx = 0 for all ϕ ∈ C0∞ (D) iff v = 0 a.e. in D.
1

Proof. See [46, Cor. 4.24], [132, p. 6]. ⊓

⊔

1.4.2 Lebesgue spaces Lp (D) and L∞ (D)

Definition 1.29 (Lp spaces). Let D be an open set in Rd . For all p ∈ [1, ∞],
let Lp (D) := {f measurable | kf kLp(D) < ∞}, where
Z  p1
kf kLp(D) := |f |p dx , if p ∈ [1, ∞), (1.4a)
D
kf kL∞ (D) := ess sup |f (x)| := inf{M ∈ R | |f (x)| ≤ M a.e. x ∈ D}. (1.4b)
x∈D

We write Lp (D; Rq ), q ≥ 1, for the space composed of Rq -valued functions

whose components are all in Lp (D), and we use the Euclidean norm in Rq ,
kf kℓ2 (Rq ) , instead of |f |, to evaluate the norms in (1.4). When q = d, we
write Lp (D) := Lp (D; Rd ).
Lebesgue’s dominated convergence theorem extends to all the Lp spaces,
p ∈ [1, ∞), i.e., if the dominating function g is in Lp (D), the convergence of
fn to f occurs in Lp (D).
8 Chapter 1. Lebesgue spaces

Theorem 1.30 (Pointwise convergence). Let p ∈ [1, ∞]. Let (fn )n∈N be a
sequence in Lp (D) and let f ∈ Lp (D) such that kfn −f kLp(D) → 0. Then there
exist a subsequence (fnk )k∈N and a function g ∈ Lp (D) such that fnk (x) →
f (x) a.e. in D and |fnk (x)| ≤ g(x) a.e. in D for all k ∈ N.

Proof. See [46, Thm. 4.9], [162, Thm. 3.12]. ⊓

⊔

Theorem 1.31 (Fischer–Riesz). For all p ∈ [1, ∞], Lp (D) equipped with
the Lp -norm from Definition 1.29 is a Banach space.

Proof. See [3, Thm. 2.16], [16, p. 142], [46, Thm. 4.8], [162, Thm. 3.11]. ⊓
⊔

Among all the Lebesgue spaces, L2 (D) plays a particular role owing to the
following important consequence of the Fischer–Riesz theorem.
Theorem 1.32 (L2 space).R L2 (D) is a Hilbert space when equipped with the
Rinner product (f, g)L2 (D) := D f g dx for real-valued functions, or (f, g)L2 (D) :=
D
f g dx for complex-valued functions.

Remark 1.33 (Continuous embedding on bounded sets). Assume

that D is bounded. For all p, q ∈ [1, ∞] with p ≤ q, Hölder’s inequality
implies that
1 1
kf kLp(D) ≤ |D| p − q kf kLq (D) , ∀f ∈ Lq (D), (1.5)

meaning that Lq (D) ֒→ Lp (D) (this notation means that Lq (D) is con-
tinuously embedded into Lp (D)). One can show that limp→∞ kf kLp(D) =
kf kL∞ (D) for all f ∈ L∞ (D). Moreover, if f ∈ Lp (D) for all p ∈ [1, ∞)
and if there is c, uniform w.r.t. p, s.t. kf kLp(D) ≤ c, then f ∈ L∞ (D) and
kf kL∞ (D) ≤ c; see [3, Thm. 2.14]. ⊓
⊔

Theorem 1.34 (Density of C0∞ (D)). Let D be an open set in Rd . Then

C0∞ (D) is dense in Lp (D) for all p ∈ [1, ∞).

Proof. See [162, Thm. 3.14].

Remark 1.35 (The case of L∞ (D)). C0∞ (D) is not dense in L∞ (D). If D
is bounded, the completion of C ∞ (D) in L∞ (D) is C 0 (D), and the completion
of C0∞ (D) is {v ∈ C 0 (D) | v|∂D = 0}. ⊓
⊔

1.4.3 Duality
Lemma 1.36 (Conjugate, Hölder’s inequality). Let p ∈ [1, ∞] be a real
number. The real number p′ ∈ [1, ∞] such that 1p + p1′ = 1, with the convention
that p′ := 1 if p = ∞ and p′ := ∞ if p = 1, is called conjugate of p. Let
′
f ∈ Lp (D) and g ∈ Lp (D). Then f g ∈ L1 (D) and
Z
|f g| dx ≤ kf kLp(D) kgkLp′ (D) . (1.6)
D
Part I. Elements of functional analysis 9

Proof. See [3, Thm. 2.4], [16, p. 404], [46, Thm. 4.6], [162, Thm. 3.8]. ⊓
⊔

For p = p′ = 2, R Hölder’s inequality becomes the Cauchy–Schwarz in-

equality in L2 (D): D |f g| dx ≤ kf kL2 (D) kgkL2(D) for all f, g ∈R L2 (D). This
inequality is useful to bound |(f, g)L2 (D) | since |(f, g)L2 (D) | ≤ D |f g| dx.

Theorem 1.37 (Riesz–Fréchet). Let p ∈ [1, ∞). The dual space of Lp (D)
′
can be identified with Lp (D).

Proof. See [3, pp. 45-49], [46, Thm. 4.11&4.14], [162, Thm. 6.16]. ⊓
⊔

Remark 1.38 (L∞ (D)). Theorem 1.37 fails for p = ∞. Indeed the dual of
L∞ (D) strictly contains L1 (D) (see [46, p. 102]). ⊓
⊔

Corollary 1.39 (Interpolation inequality). Let p, q ∈ [1, ∞] with p ≤ q

and let θ ∈ [0, 1]. Let r ∈ [1, ∞] be such that 1r := pθ + 1−θ
q . Then we have

kf kLr (D) ≤ kf kθLp(D) kf k1−θ

Lq (D) , ∀f ∈ Lp (D) ∩ Lq (D). (1.7)

Recall from §A.2 that for two Banach spaces V and W , L(V ; W ) is composed
of the linear operators that map V boundedly to W , and that the norm
k·kL(V ;W ) is defined in (A.2).
Theorem 1.40 (Riesz–Thorin). Let p0 , p1 , q0 , q1 be four real numbers such
that 1 ≤ p0 ≤ p1 ≤ ∞, 1 ≤ q0 ≤ q1 ≤ ∞. Let T : Lp0 (D) + Lp1 (D) −→
Lq0 (D)+Lq1 (D) be a linear operator that maps Lp0 (D) and Lp1 (D) boundedly
to Lq0 (D) and Lq1 (D), respectively. Then the operator T maps Lpθ (D) bound-
edly to Lqθ (D) for all θ ∈ (0, 1), where pθ and qθ are defined by p1θ := 1−θ θ
p0 + p1 ,
1 := 1−θ θ θ 1−θ
q0 + q1 . Moreover kT kL(L θ ;L θ ) ≤ kT kL(Lp0 ;Lq0 ) kT kL(Lp1 ;Lq1 ) .
p q
qθ

Proof. See [181, Thm. 21.2], Bergh and Löfström [18, Chap. 1]. ⊓
⊔

Remark 1.41 (Interpolation). Corollary 1.39 and Theorem 1.40 are re-
lated to the interpolation theory between Banach spaces (see §A.5). For in-
stance Lp (D) can be defined for all p ∈ (1, ∞), up to equivalent norm, by
interpolating between L1 (D) and L∞ (D), i.e., Lp (D) = [L1 (D), L∞ (D)] 1′ ,p ;
p
see Tartar [181, p. 111]. ⊓
⊔

1.4.4 Multivariate functions

The following results on multivariate functions are useful in many situations.

Theorem 1.42 (Tonelli). Let f R: D1 ×D2 → R be a measurable function

such that the function D1 ∋ x1 7→ D2 |f (x1 , x2 )| dx2 is finite a.e. in D1 and
is in L1 (D1 ). Then f ∈ L1 (D1 × D2 ).

Proof. See [46, Thm. 4.4]. ⊓

⊔
10 Chapter 1. Lebesgue spaces

Theorem 1.43 (Fubini). Let f ∈ L1 (D1 ×D2 ). Then the function D2 ∋

1
Rx2 7→ f (x1 , x2 ) is in L (D 1
2 ) for a.e. x1 ∈ D1 , and the function D1 ∋ x1 7→
f (x 1 , x 2 ) dx 2 is in L (D 1 ). Similarly the function D1 ∋ x1 7→ f (x1 , x2 )
D2 R
is in L1 (D1 ) for a.e. x2 ∈ D2 , and the function D2 ∋ x2 7→ D1 f (x1 , x2 ) dx1
is in L1 (D2 ). Moreover we have
Z Z  Z Z 
f (x1 , x2 ) dx2 dx1 = f (x1 , x2 ) dx1 dx2 , (1.8)
D1 D2 D2 D1
R
and both quantities are equal to D1 ×D2 f (x1 , x2 ) dx1 dx2 , where dx1 dx2 is
the product measure on the Cartesian product D1 ×D2 .

Exercises
Exercise 1.1 (Measurability). Let W be a non-measurable subset of
D := (0, 1). Let f : W → R be defined by f (x) := 1 if x ∈ D\W and
f (x) := 0 if x ∈ W . (i) Is f measurable? (ii) Assume that there is a mea-
surable subset V ⊂ W s.t. |V | > 0. Compute supx∈D f (x), ess supx∈D f (x),
inf x∈D f (x), ess inf x∈D f (x). (iii) Is f a member of L∞ (D)? (iv) Assume now
that W has zero measure (hence W is measurable). Compute inf x∈D f (x)
and ess inf x∈D f (x).

Exercise 1.2 (Measurability and equality a.e.). Prove Corollary 1.11.

(Hint : consider the sets Ar := {x ∈ D | f (x) > r} and Br := {x ∈
D | g(x) > r} for all r ∈ R, and show that Br = (Ar ∩ (Ar \Br )c ) ∪ (Br \Ar ).)

Exercise 1.3 (Lebesgue’s theorem). Let D := (−1, 1). Let (fn )n∈N be a
sequence of functions in L1 (D) and let g ∈ L1 (D). Assume that fn → f a.e.
in D. Propose a counterexample to show that the assumption “|fn | ≤ g a.e.
for all n ∈ N” cannot be replaced by “fn ≤ g a.e. for all n ∈ N” in Lebesgue’s
dominated convergence theorem.

Exercise 1.4 (Compact support). Let D := (0, 1) and f (x) := 1 for all
x ∈ D. What is the support of f in D? Is the support compact?

Exercise 1.5 (Pointwise limit of measurable functions). Let D be a

measurable set in Rd . Let fn : D → R for all n ∈ N be real-valued mea-
surable functions. (i) Show that lim supn∈N fn and lim inf n∈N fn are both
measurable. (Hint : recall that lim supn∈N fn (x) = inf n∈N supk≥n fk (x) and
lim inf n∈N fn (x) = supn∈N inf k≥n fk (x) for all x ∈ D). (ii) Let f : D → R.
Assume that fn (x) → f (x) for every x ∈ D. Show that f is measurable. (iii)
Let f : D → R. Assume that fn (x) → f (x) for a.e. x ∈ D. Show that f is
measurable.
Part I. Elements of functional analysis 11

Solution to exercises
Exercise 1.1 (Measurability). (i) Since the set {x ∈ D | f (x) < 1} = W
is not measurable, Lemma 1.9 implies that f is not measurable.
(ii) We have

sup f (x) = 1,
x∈D
ess sup f (x) = 1,
x∈D
inf f (x) = 0,
x∈D
ess inf f (x) = 0.
x∈D

(iii) Although ess supx∈D |f (x)| = 1 ≤ ∞, the function f is not a member of

L∞ (D) since it is not measurable.
(iv) Since we now assume that |W | = 0, i.e., f = 1 a.e., we have

inf f (x) = 0,
x∈D
ess inf f (x) = 1.
x∈D

Exercise 1.2 (Measurability and equality a.e.). Following the hint, let
Ar := {x ∈ D | f (x) > r} and Br := {x ∈ D | g(x) > r} for all r ∈ R.
By assumption the set Ar is measurable. We observe that Ar \Br ⊂ {x ∈
D | f (x) 6= g(x)} and Br \Ar ⊂ {x ∈ D | f (x) 6= g(x)}. Hence |Ar \Br |∗ = 0
and |Br \Ar |∗ = 0. This means that Ar \Br and Br \Ar are measurable (see
Example 1.4). After observing that Br ∩ Ar = Ar ∩ (Ar \Br )c , and Br ∩ Acr =
Br \Ar , we finally have Br = (Ar ∩ (Ar \Br )c ) ∪ (Br \Ar ). This shows that Br
is measurable since the sets Ar , (Ar \Br )c , and Br \Ar are measurable. Hence
g is measurable owing to Lemma 1.9.

Exercise 1.3 (Lebesgue’s theorem). The sequence {fn }n≥1 such that
fn (x) := −2n if |x| ≤ n1 and fn (x) := 0 otherwise is such that fn → 0
a.e. in D and fn ≤ 0 ∈ L1 (D), but fn does not converge to 0 in L1 (D) since
kfn kL1 (D) = 1.

Exercise 1.4 (Compact support). We have {x ∈ D | f (x) 6= 0} = D.

The slight subtlety here is that the closure of D in D is D itself. Hence the
support of f in D is D. Note that D is not compact since it is not a closed
set in R (the limit point of the sequence { n1 }n≥1 does not belong to D).

Exercise 1.5 (Pointwise limit of measurable functions). (i) Using the

hint, we have for all x ∈ D,
12 Chapter 1. Lebesgue spaces

lim sup fn (x) ≤ c ⇐⇒ inf sup fk (x) ≤ c

n∈N n∈N k≥n

1
⇐⇒ ∀j ≥ 1, ∃n ≥ 0, sup fk (x) ≤ c +
k≥n j
1
⇐⇒ ∀j ≥ 1, ∃n ≥ 0, ∀k ≥ n, fk (x) ≤ c +
j
\ [ \n 1o
⇐⇒ x ∈ y ∈ D | f (y) ≤ c + .
j
j≥1 n≥0 k≥n

This proves that

n o \ [ \n 1o
x ∈ D | lim sup fn (x) ≤ c = y ∈ D | f (y) ≤ c + .
n∈N j
j≥1 n≥0 k≥n

Hence the function lim supn∈N fn is measurable. The proof that lim inf n∈N fn
is a measurable function is similar.
(ii) Saying that fn (x) → f (x) for every x ∈ D means that lim supn∈N fn (x) =
f (x) = lim inf n∈N fn (x) for every x ∈ D. We conclude from (i) that f is
measurable.
(iii) Let S ⊂ D be such that S := {x ∈ D | f (x) = lim supn∈N fn (x) =
lim inf n∈N fn (x)}. By assumption we have |S c | = 0. Moreover we have

S = {x ∈ D | lim sup fn (x) = lim inf fn (x) = f (x)}

n∈N n∈N

⊂ {x ∈ D | f (x) = lim sup fn (x)}.

Hence {x ∈ D | f (x) 6= lim sup fn (x)} ⊂ S c . This means that the function f
and lim sup fn coincide almost everywhere. Corollary 1.11 implies that f is
measurable since lim sup fn is measurable.