Appendix D

Measure theory and Lebesgue

integration
As far as the laws of mathematics refer to reality, they are not certain,
and as far as they are certain, they do not refer to reality.
Albert Einstein (1879-1955)
In this appendix, we will briey recall the main concepts and results about the measure theory and
Lebesgue integration. Then, we will introduce the L
p
spaces that prove especially useful in the analysis
of the solutions of partial dierential equations.
Contents
D.1 Measure theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
D.2 Riemann integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
D.3 Lebesgue integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
D.4 L
p
spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
D.1 Measure theory
Measure theory initially was proposed to provide an analysis of and generalize notions such as length,
area and volume (not strictly related to physical sizes) of subsets of Euclidean spaces. The approach to
measure and integration is axomatic, i.e. a measure is any function dened on subsets which satisfy a
cetain list of properties. In this respect, measure theory is a branch of real analysis which investigates,
among other concepts, measurable functions and integrals.
D.1.1 Denitions and properties
Denition D.1 A collection S of subsets of a set X is said to be a topology in X is S has the following
three properties:
(i) S and X S,
107
108 Chapter D. Measure theory and Lebesgue integration
(ii) if V
i
S for i = 1, . . . , n then, V
1
V
2
V
n
S,
(iii) if {V

} is an arbitrary collection of members of S (nite, countable or not), then

inS.
if S is a topology in X, then X is called a topological space and the members of S are called the open
sets in X.
Denition D.2 A collection F of subsets of a set X is said to be a -algebra in X if F has the following
three properties:
(i) X F,
(ii) if A F, then A
c
F, where A
c
is the complement of A relatie to X,
(iii) if A =

n=1
A
n
and if A
n
F for all n, then A F.
If F is a -algebra in X, then the pair (X, F) is called a measurable space . If X is a measurable space,
Y is a topological space and f is a mapping of X into Y , then f is said to be measurable provided that
f
1
(V ) is a measurable set in X for every open set V in Y .
Denition D.3 A measure is a function dened on a -algebra F over a set X and taking values in the
interval [0, [ such that the following properties are satised:
(i) the emptyset has measure zero, () = 0;
(ii) countable additivity: if (E
i
) is a countable sequence of pairwise disjoint sets in F, then

_
i=1
E
i
_
=

i=1
(E
i
) .
The triple (X, F, ) is then called a measure space and the members of F are called measurable sets.
For measure spaces that are also topological spaces, various compatibility conditions can be placed for
the measure and the topology.
Theorem D.1 Let f and g be real measurable functions on a measurable space X, let be a continuous
mapping of the plane into a topological space Y and dene h(x) = (u(x), v(x)) for x X. Then,
h : X Y is measurable.
Proposition D.1 If E is a measurable set in X and if

E
(x) =
_
1 if x E
0 if x / E
then,
E
is a measurable function called the characteristic function of the set E.
Denition D.4 Let (a
n
) be a sequence in [, ] and put
b
k
= sup{a
k
, a
k+1
, a
k+2
, . . . } k = 1, 2, 3, . . .
and = inf{b
1
, b
2
, b
3
, . . . }. We call the upper limit of (a
n
) and write
= limsup
n
a
n
.
D.2. Riemann integration 109
The lower limit is dened analogously by interchanging sup and inf in the previous denition. Moreover,
we have
liminf
n
a
n
= limsup
n
(a
n
) ,
and, if (a
n
) converges, then
limsup
n
a
n
= liminf
n
a
n
= lim
n
a
n
.
Supose (f
n
) is a sequence of real functions on a set X. Then sup
n
f
n
and limsup
n
f
n
are the functions
dened on X by:
(sup
n
f
n
)(x) = sup
n
(f
n
(x)) , (limsup
n
f
n
)(x) = limsup
n
(f
n
(x)) .
Moreover, if f(x) = lim
n
f
n
(x) the limit being assumed to exist at every x X, then f is called the
pointwise limit of the sequence (f
n
).
Theorem D.2 If f
n
: X R is measurable, for n = 1, 2, . . . and
g = sup
n1
f
n
, h = limsup
n
f
n
,
then g and h are both measurable.
Corollary D.1 The limit of very pointwise convergent sequence of complex measurable functions is mea-
surable. If f and g are measurable then so are max(f, g) and min(f, g). In particular, this is true of the
functions
f
+
= max(f, 0) and f

= min(f, 0) ,
respectively called the positiveand negative parts of f.
We have f = f
+
f

and |f| = f
+
+f

.
Proposition D.2 If f = g h, g 0 and h 0 then f
+
g and f

h.
D.1.2 Completeness
A measurable set X is called a null set if (X) = 0. By extension, a subset of a null set is called a negligible
set. A negligible set need not be measurable, but every measurable negligible set is automatically a null
set. A measure is called complete if every negligible set is measurable.
A measure can be extended to a complete one by considering the -algebra of subsets Y which dier
by a negligible set from a measurable set X, that is, such that the symmetric dierence of X and Y is
contained in a null set. One denes (Y ) to equal (X).
D.1.3 Non-measureable sets
D.2 Riemann integration
Suppose that a function f is bounded on the interval [a, b], where a, b R and a < b and consider a
dissection : a = x
0
< x
1
< < x
n
= b of [a, b]. Then,
110 Chapter D. Measure theory and Lebesgue integration
Denition D.5 (Riemann sum) The lower Riemann sum of f(x) corresponding to the dissection is
dened as the following sum:
s(f, ) =
n

j=1
(x
j
x
j1
inf
x[x
j1
,x
j
]
f(x)
anf the upper Riemann sum of f(x) corresponding to the dissection is given by the sum:
S(f, ) =
n

j=1
(x
j
x
j1
sup
x[x
j1
,x
j
]
f(x) .
Theorem D.3 Suppose that a function f : R R is bounded on [a, b], where a < b, a, b R and that
and

are two dissections of [a, b] such that

. Then,
s(f,

) s(f, ) and S(f, ) S(f,

) .
If

is another dissection of [a, b] then,

s(f,

) S(f,

) .
Denition D.6 For all dissections of [a, b], the real number I

(f, a, b) = sup

s(f, ) is called the

lower integral of f(x) over [a, b] and the real number I
+
(f, a, b) = inf

S(f, ) is called the upper integral

of f(x) over [a, b].
Theorem D.4 Suppose that a function f is bounded on the interval [a, b], where a < b and a, b R.
Then
I

(f, a, b) I
+
(f, a, b) .
Denition D.7 Suppose that a function f is bounded on the interval [a, b], where a < b and a, b R
and suppose that I

(f, a, b) = I
+
(f, a, b). Then, the function f is said to be Riemann integrable over
[a, b] and we write
_
b
a
f(x) dx = I

(f, a, b) = I
+
(f, a, b) .
Lemma D.1 Suppose that a function f is bounded on the interval [a, b], where a < b and a, b R. The
following two statements are equivalent:
(i) f is Riemann integrable over [a, b]
(ii) given any > 0, there exists a dissection of [a, b] such that
S(f, ) s(f, ) < .
We have several additional properties of Riemann integrals.
Lemma D.2 Suppose that f, g are Riemann integrable over [a, b], where a < b and a, b R. Then
(i) f +g is Riemann integrable over [a, b] and
_
b
a
(f(x) +g(x)) dx =
_
b
a
f(x) dx +
_
b
a
g(x) dx,
(ii) for every c R, cf is Riemann integrable over [a, b] and
_
b
a
cf(x) dx = c
_
b
A
f(x) dx,
D.3. Lebesgue integration 111
(iii) if f(x) 0 for every x [a, b] then
_
b
a
f(x) dx 0,
(iv) if f(x) g(x) for every x [a, b] then
_
b
a
f(x) dx
_
b
a
g(x) dx.
Proposition D.3 Suppose that f is Riemann integrable over [a, b] where a, b R and a < b. Then,
for every real number c [a, b], f is Riemann integrable over [a, c] and Riemann integrable over [c, b].
Moreover, we can write:
_
b
A
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx.
Similarly, we have the following result.
Proposition D.4 Suppose that a, b, c R and that a < c < b. Suppose further that f is Riemann
integrable over [a, c] and is Riemann integrable over [c, b]. Then, f is Riemann integrable over [a, b] and
_
b
a
f(x)dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx.
Theorem D.5 Suppose that f is Riemann integrable over [a, b], where a, b R and a < b. Suppose that
f(x) = g(x) for every x [a, b], except possibly at x = x
0
Then g is Riemann integrable over [a, b] and
_
b
a
f(x) dx =
_
b
a
g(x) dx.
D.3 Lebesgue integration
Suppose we are considering integrating functions like (x), the characteristic function of the set S =
{x Q} R (i.e.
S
(x) = 1 if x S and
S
(x) = 0 if x / S). Or suppose we are considering real-valued
measurements x of a phenomenon and wondering what is the probability for x to be a rational number.
From the prababilistic theory, we know that if the measurements are distributed normally with a mean
of and a standard deviation of , then the probability is given by:
Pr[x Q] =
_
S
1

2
2
exp
_

1
2
_
x

_
2
_
dx, (D.1)
and the Riemann integral is useless to evaluate this integral.
D.3.1 Sets of measure zero
The study of Lebesgue integral depends on the notion of zero measure sets in R.
Denition D.8 A set S R is said to have measure zero if, for every > 0 there exiss a countable
family F of intervals I such that
S
_
IF
I and

IF
(I) < ,
where, for every I F, (I) denotes the length of the interval I.
This denition states that the set S can be covered by a countable union of open intervals of arbitrarily
small total length.
Proposition D.5 Every countable set in R has measure zero. Furthermore, a outable union of sets of
measure zero in R has measure zero.
112 Chapter D. Measure theory and Lebesgue integration
D.3.2 Compact sets
Denition D.9 A set S R is said to be compact if and only if, for every family F of open intervals
I such that
S
_
IF
I
there exists a nite subfamily F
0
F such that
S
_
IF
0
I .
This denition means that every open covering of S can be achieved by a subcovering.
Theorem D.6 (Heine-Borel) Suppose that F R is bounded and closed. Then, F is compact.
D.3.3 Lebesgue integral
Denition D.10 A function is simple if its range is a nite set.
A simple function in R has always the following representation:
=
n

k=1
a
k

E
k
,
where a
k
are distinct values of and E
k
=
1
({a
k
}). Conversely, any expression of this form, where a
k
need not be distinct and E
k
not necessarily
1
({a
k
}) also denes a simple function. In the remainder,
we will consider that E
k
be measurable and that they partition the set X.
Denition D.11 Let (X, ) be a measure space. The Lebesgue integral over X of a R
+
-valued simple
function is dened as:
_
X
d =
_
X
n

k=1
a
k

E
k
d =
n

k=1
a
k
(E
k
) .
The quantity on the right represents tha sum of the areas below the graph of . If =

i
a
i

A
i
=

j
b
j

B
j
, where A
i
and B
j
partition X, then

i
a
i
(A
i
) =

i
a
i
(A
i
B
j
) =

i
b
j
(A
i
B
j
) =

j
b
j
(B
j
) .
The second equality follows since the value of is a
i
= b
j
on A
i
B
j
, so a
i
= b
j
whenever A
i
B
j
= .
The integral is thus well-dened. Similarly, for two simple functions then
_
X
d
_
X
d
(monotonicity of the integral).
Theorem D.7 For non-negative simple functions, the Lebesgue integral is linear.
Denition D.12 Let f : X [0, +] be measurable. Consider the set S
f
if all measurable functions
0 f. The integral of f over X is dened as:
_
X
f d = sup
S
f
_
X
d.
D.3. Lebesgue integration 113
The simple functions in S
f
are supposed to approximate f as close as possible. The integral of f is
obtained by computing the integrals of these approximations.
Theorem D.8 (Approximation theory) Let f : X [0, +] be measurable. Then, there exists
a sequence of non-negative functions {
n
} f, meaning
n
are increasing pointwise and converging
pointwise toward f. Moreover, if f is bounded, it is possible for the
n
to converge toward f uniformly.
Denition D.13 If f is not necessarily non-negative, we dene:
_
X
f d =
_
X
f
+
d
_
X
f

d,
provided that the two integrals are not both .
The functions f
+
and f

are measurable and represent the positive and negative part of f, respectively:
f
+
(x) = max(+f(x), 0) f

(x) = max(f(x), 0) .
Let A be a measurable subset of X, we can dene:
_
A
d d =
_
X
f
A
d,
to introduce the Lebesgue integral on subsets of X.
Proposition D.6 A measurable function f : X [0, ] vanishes almost everywhere if and only if
_
X
f = 0.
D.3.4 Convergence results
Theorem D.9 (Monotone convergence) Let (X, ) be a measure space. Let f
n
be non-negative mea-
surable functions increasing pointwise toward f. Then,
_
X
f d =
_
X
_
lim
n
f
n
_
d = lim
n
_
X
f
n
d.
This theorem allows to prove linearity of the Lebesgue integral for non-simple functions. Given any
two non-negative measurable functions f, g we know (Theorem D.8) that there are non-negative simple
functions {
n
} f and {
n
} g. Then, {
n
+
n
} f + g and so, since the integral is linear for
simple functions:
_
f +g = lim
n
_

n
+
n
= lim
n
_

n
+
_

n
=
_
f +
_
g .
And, if f, g are not necessarily non-negative, then:
_
f +g =
_
(f
+
f

) + (g
+
g

) =
_
(f
+
+g
+
(f

+g

))
=
_
(f
+
+g
+
)
_
(f

+g

) =
_
f
+
+
_
g
+

__
(f

+
_
g

_
=
_
f +
_
g .
114 Chapter D. Measure theory and Lebesgue integration
Theorem D.10 (Beppo Levi) Let f
n
: X [0, ] be measurable. Then:
_

n=1
f
n
=

n=1
_
f
n
.
Theorem D.11 (Fatous lemma) Let f
n
: X [0, ] be measurable. Then,
_
liminf
n
f
n
liminf
n
_
f
n
.
Denition D.14 A function f : X R is called integrable if it is measurable and if
_
X
|f| < .
It follows that f is integrable if and only if f
+
and f

are both integrable. Moreover,

_
|f| < implies
that |f| < almost everywhere.
Theorem D.12 (Dominated convergence) Let (X, ) be a measure space. Let f
n
: X R be a
sequence of measurable functions converging pointwise toward f. Moreover, supose that there is an inte-
grable function g such that |f
n
| g for all n. Then f
n
and f are also integrable and:
lim
n
_
X
|f
n
f| d = 0 .
It is sucient to require that f
n
converge to f pointwise almost everywhere, or that |f
n
| is bounded above
by g almost everywhere. Using the triangle inequality, we conclude that
lim
n
_
X
f
n
d =
_
X
f d,
and this corresponds to the common application of this theorem.
D.3.5 Generalization
Theorem D.13 (Beppo Levi) Let (X, ) be a measure space and f
n
: X R be measurable functions
with
_
|f
n
| =
_
|f
n
| < . Then,

n=1
_
f
n
=
_

n=1
f
n
.
Proposition D.7 Let g : X [0, ] be measurable in the measure space (X, A, ). Let
(E) =
_
E
f d, E A.
Then, is a measure on (X, A) and for any measurable function f on X,
_
X
f d =
_
X
fg d,
and this result is usually written as: d = gd.
Lemma D.3 (Change of variables) Let X, Y be measure spaces and g : X Y , f : Y R. Then,
_
X
(f g) d =
_
Y
fd ,
where (B) = (g
1
(B)) is a measure dened for all measurable B Y .
D.3. Lebesgue integration 115
Applying these results together leads to:
Theorem D.14 (Dierential change of variables in R
n
) Let g : X Y be a dieomorphism of
open sets in R
n
. If A X is measurable and f : Y R is measurable then,
_
g(A)
f d =
_
A
(f g) d =
_
A
(f g) | det Dg| d.
And we have Lebesgue versions of results about the Riemann integral.
Theorem D.15 (First fundamental theorem of calculus) Let I R be an interval and f : I R
be integrable with Lebesgue measure in R. Then, the function
F(x) =
_
x
a
f(t) dt ,
is continuous. Moreover, if f is continuous at x, then F

(x) = f(x).
Theorem D.16 (Second fundamental theorem of calculus) Suppose f : [a, b] R is measurable
and bounded above and below. If f = g

for some g, then

_
b
a
f(x) dx = g(b) g(a) .
In this result, we shall not assume as strong hypotheses that f

is continuous or even that it is Lebesgue

integrable. If g exists, then it can also be computed as countable limit lim
n
n(g(x + 1/n) g(x)), thus
showing that g

is measurable.
Theorem D.17 (Continuous dependence on integral parameter) Let (X, ) be a measure space,
T be any metric space and f : X T R with f(, t) being measurable for each t T. Consider the
function
F(t) =
_
xX
f(x, t) .
Then, we have F continuous at t
0
T if the following conditions are met:
(i) for each x X, f(x, ) is continuous at t
0
I,
(ii) there is an integrable function g such that |f(x, t)| g(x) for all t T.
Theorem D.18 (Dierentiation under the integral sign) Using the same notations, with T being
an open real interval, we have
F

(t) =
d
dt
_
xX
f(x, t) =
_
xX

t
f(x, t) ,
if the following conditions are satised:
(i) for each x X,

t
f(x, t) exists,
(ii) there is an integrable function g such that

t
f(x, t)

g(x), for all t T.

This result can be generalized to T being any open set in R
n
, taking partial derivatives.
116 Chapter D. Measure theory and Lebesgue integration
D.4 L
p
spaces
The L
p
spaces, named after H. Lebesgue (1875-1941), are spaces of p-power integrable functions and form
an important class of examples of Banach spaces.
D.4.1 Denitions
Let be a domain in R
d
and let p [1, +[ be a positive real number. We denote L
p
(, d) the space
of all measurable functions u, dened on , for which
f
L

def
=
_

|f(x)|
p
d < .
The elements of L
p
are indeed equivalence classes of measurable functions satisfying the previous equality
(two functions being equivalent if they coincide almost everywhere in ). If p = , we denote by
L
p
(, d) the space of all measurable functions, dened on , such that:
f
L

def
= sup{/ {x, |f(x)| > } > 0} .
We need to introduce the notion of exponent conjugate. If p ]1, [, p

def
=
p
p 1
, if p = 1 then p

def
= +
and if p = +, p

def
= 1. The exponents p and p

are said to be (H older) conjugates of each other and,

under the convention that 1/= 0, we have
1
p
+
1
p

= 1 .
Theorem D.19 For any value p [1, [, the space L
p
(, d), endowed with the norm
L
p, is a
Banach space.
It is clear that if u L
p
() and c C, then cu L
p
(). Moreover, if u, v L
p
(), then since
|u(x) +v(x)|
p
(|u(x)| + |v(x)|)
p
2
p
(|u(x)|
p
+ |v(x)|
p
) ,
u +v L
p
(), so L
p
() is a vector space.
Proposition D.8 (Holder inequality) Let (, ) be a measure space, f be a function of L
p
(, d)
and g be a function of L
p

(, d). Then, the product fg is in L

1
(, d) and
_

|f(x)g(x)| d(x) f
L
pg
L
p
.
For 1 < p, p

< , the inequality becomes equality if and only if |f|

p
is proportional to |g|
p

almost
everywhere. When p = p

= 2, we retrieve the Cauchy-Schwarz inequality.

Proposition D.9 (Minkowskis inequality) If 1 p < , then for all u, v L
p
()
u +v
p
u
p
+v
p
.
Theorem D.20 (Jensens inequality) Let L
1
() be a non-negative function such that
_

(x)dx = 1.
Then, for every measurable function f such that f L
1
() and for every convex measurable function
: R R, we have:

__

f(x)(x) dx
_

(f(x))(x) dx.
D.4. L
p
spaces 117
In particular, for (x) = x
2
, we have:
__

f(x)(x) dx
_
2

(f(x))
2
(x) dx.
Theorem D.21 (Fubinis theorem) Suppose F L
1
(
1

2
). Then, for almost every x
1
,
F(x, ) L
1
(
2
) and for almost every y
2
, F(, y) L
1
(
1
). Furthermore, we have, by noting
d(x, y) = dx dy:
_

2
F(x, y)d(x, y) =
_

1
_

2
F(x, y)dydx =
_

2
_

1
F(x, y)dxdy .
D.4.2 Properties of L
p
spaces
Corollary D.2 Let p, q ]1, [ be two real numbers such that 1/P + 1/q 1. Then, the function
L
p
L
q
L
r
, (f, g) fg is a bilinear continuous map if 1/r = 1/p + 1/q.
Corollary D.3 If is a space of nite measure, then L
p
(, d) L
q
(, d) if p q.
Lemma D.4 Given (, d) a measure space. Let f be a measurable function and p [1, [. Then, if
sup
g
L
p
1
_

|f(x)g(x)| d(x) < +,

then, f is in L
p
() and we have
f
L
p = sup
g
L
p

f(x)g(x) d(x)

.
The next result provides a reverse form of H older and Minkwski inequalities for the case 0 < p < 1. This
result is used to show the uniform convexity of certain L
p
spaces.
Theorem D.22 Let 0 < p < 1 so that the conjugate writes p

= p((p 1) < 0. Suppose f L

p
() and
0 < g
_

|g(x)|
p
d(x) < .
Then,
_

|f(x)g(x)| d(x)
__

|f(x)|
p
d(x)
_
1/p
__

|g(x)|
p

d(x)
_
1/p

.
and we have also
|u| + |v|
p
u
p
+v
p
.
Denition D.15 A function f, measurable on is said to be essentialy bounded on if there exists
a constant C such that f(x) C almost everywhere in . The greatest lower bound of C is called the
essential supremum of |f| on and is denoted by ess sup
x
|f(x)|.
We denote by L

() the space of all functions f that are essentially bounded on . It is then easy to
verify that the functional

dened by
f

= ess sup
x
|f(x)| ,
is a norm on L

().
118 Chapter D. Measure theory and Lebesgue integration
D.4.3 Density of continuous functions in L
p
spaces
At rst, we shall introduce new functional spaces, the L
p
loc
spaces.
Denition D.16 We denote L
p
loc
() the space of all p-power (1 p ) locally integrable functions
on . It is the set of functions f such that, for any compact subset K of
f L
p
(K, d) .
The Holder inequality implies that q p L
q
(K, d) L
p
(K, d) thus yielding the following result.
Proposition D.10 If p < p

then, L
p

loc
() is included in the space L
p
loc
.
Denition D.17 Let f be a function of L
1
loc
(). The support of f, denoted Supp f is the remainder of
the largest open subset U such that f
|U
0.
In other words, a function f L
1
loc
() is said to have nite support if f(x) = 0 for all but nitely many
x.
Proposition D.11 Let f be a function of L
1
loc
(). The support of f is the set of points x for which
every open neighborhood N(x) of x has positive measure:
Suppf = {x , /x some open N
x
, (N
x
) > 0 .
We introduce the following fundamental result about the density of compact support functions.
Theorem D.23 Given p [1, [, the space C
c
() of all continuous functions with compact support in
is dense in the space L
p
(, d).
Finally, we have a useful imbedding result for L
p
spaces over domains with nite volume.
Theorem D.24 Suppose vol =
_

1dx < and 1 p, q . If f L

q
(), then f L
p
() and
f
p
(vol)
(1/p)(1/q)
f
q
.
Hence, L
q
() L
p
(). If f L

(), then
lim
p
f
p
= f

.
Finally, if f L
p
() for 1 p < and if there is a constant C such that for all such p
f
p
C ,
then f L

() and
f

C .
Corollary D.4 L
p
() L
1
loc
() for 1 p and any domain .
D.4. L
p
spaces 119
D.4.4 Completeness of L
p
spaces
We give a sequence of important results about L
p
spaces.
Proposition D.12 (i) L
p
() is a Banach space if 1 p .
(ii) If 1 p , a Cauchy sequence in L
p
() has a subsequence converging pointwise almost every-
where on .
(iii) L
2
() is a Hilbert space with respect to the inner product
f, g =
_

f(x) g(x) dx.

and Holder inequality for L
2
() is simply the Schwarz inequality:
|f, g| f
L
2 g
L
2 .
(iv) The space L
1
(, d) L

(, d) L
p
(, d) is dense in the space L
p
(, d).
Corollary D.5 The space C
0
() is dense in L
p
() if 1 p < .
The space L
p
() is separable if 1 p < .
Remark D.1 C(), considered as a closed subspace of L

() is not dense in that space. The same

consideration applies to C
0
() and C

0
(); this leads to conclude that L

() is not separable.
Proposition D.13 C

0
() is dense in L
p
() if 1 p < .
D.4.5 The uniform convexity of L
p
spaces
Denition D.18 The norm on any normed space X is called uniformly convex if for every number
satisfying 0 < 2, there exists a number () > 0 such that if x, y X satisfy x
X
= y
X
= 1 and
x y
X
then
(x +y)/2
X
1 () .
For 1 < p < , the space L
p
() is uniformly convex, its norm
p
satisfying the condition above.
Clarkson showed this result via a set of inequalities for L
p
() that generalizes the parallelogram law in
L
2
().
Lemma D.5 (i) If 1 p < and a 0, b 0, then
(a +b)
p
2
p1
(a
p
+b
p
) .
(ii) If 0 < s < 1, the function f(x) = (1 s
x
)/x is a decreasing function of x > 0.
(iii) If 1 < p 2 and 0 t 1, then

1 +t
2

1 t
2

_
1
2
+
1
2
t
p
_
1/(p1)
,
where p

= p/(p 1) is the exponent conjugate to p.

120 Chapter D. Measure theory and Lebesgue integration
(iv) Let z, w C. If 1 < p 2, then

z +w
2

z w
2

_
1
2
|z|
p
+
1
2
|w|
p
_
1/(p1)
,
and if 2 p < , then

z +w
2

p
+

z w
2

1
2
|z|
p
+
1
2
|w|
p
,
Theorem D.25 (Clarksons inequalites) Let f, g L
p
(). For 1 < p < let p

= p/(p 1). If
2 p < , then
_
_
_
_
f +g
2
_
_
_
_
p
p
+
_
_
_
_
f g
2
_
_
_
_
p
p

1
2
f
p
p
+
1
2
g
p
p
,
_
_
_
_
f +g
2
_
_
_
_
p

p
+
_
_
_
_
f g
2
_
_
_
_
p

_
1
2
f
p
p
+
1
2
g
p
p
_
p

1
.
If 1 < p 2, then
_
_
_
_
f +g
2
_
_
_
_
p

p
+
_
_
_
_
f g
2
_
_
_
_
p

_
1
2
f
p
p
+
1
2
g
p
p
_
p

1
,
_
_
_
_
f +g
2
_
_
_
_
p
p
+
_
_
_
_
f g
2
_
_
_
_
p
p

1
2
f
p
p
+
1
2
g
p
p
.
Corollary D.6 If 1 < p < , the space L
p
() is uniformly convex.
D.4.6 The dual of L
p
()
The main result of this section is the following theorem that establishes that a linear functional can be
represented on L
p
spaces when p R.
Denition D.19 Let 1 p and let p

denote the exponent conjugate to p. For each element

g L
p

() we can dene a linear functional L

g
on L
p
() via
L
g
(f) =
_

f(x)vg(x)d(x) , f L
p
() .
By H olders inequality, |L
g
(f)| f
p
g
p
, so that L
g
(L
p
())

and
L
g

(L
p
())
g
p
.
Indeed, the equality must holds in the previous expression. Hence, the operator map g to L
g
is an
isometric isomorphism of L
p

() onto a subspace of (L
p
())

.
Theorem D.26 (Riesz representation theorem for L
p
) Let 1 < p < and let L (L
p
())

.
Then, there exists g L
p

() such that for all finL

p
()
L(f) =
_

f(x)g(x)d(x) .
Moreover, g
p
= L
(L
p
())
. Thus, (l
p
())

= L
p

().
Let L (L
1
())

. Then, there exists g L

() such that for all f L

1
()
L(f) =
_

f(x)g(x) d(x) ,
and f

= L
(L
1
())
. Thus, (L
1
())

= L

().