Chapter 4.

The dominated convergence theorem and applica-

tions
4.1 Fatous Lemma
This deals with non-negative functions only but we get away from monotone sequences.
Theorem 4.1.1 (Fatous Lemma). Let f
n
: R [0, ] be (nonnegative) Lebesgue measurable
functions. Then
liminf
n
_
R
f
n
d
_
R
liminf
n
f
n
d
Proof. Let g
n
(x) = inf
kn
f
k
(x) so that what we mean by liminf
n
f
n
is the function with
value at x R given by
_
liminf
n
f
n
_
(x) = liminf
n
f
n
(x) = lim
n
_
inf
kn
f
k
(x)
_
= lim
n
g
n
(x)
Notice that g
n
(x) = inf
kn
f
k
(x) inf
kn+1
f
k
(x) = g
n+1
(x) so that the sequence (g
n
(x))

n=1
is monotone increasing for each x and so the Monotone convergence theorem says that
lim
n
_
R
g
n
dx =
_
R
lim
n
g
n
d =
_
R
liminf
n
f
n
d
But also g
n
(x) f
k
(x) for each k n and so
_
R
g
n
d
_
R
f
k
d (k n)
or
_
R
g
n
d inf
kn
_
R
f
k
d
Hence
liminf
n
_
R
f
n
d = lim
n
_
inf
kn
_
R
f
k
d
_
lim
n
_
R
g
n
d =
_
R
liminf
n
f
n
d
Example 4.1.2. Fatous lemma is not true with equals.
For instance take, f
n
=
[n,2n]
and notice that
_
R
f
n
d = n as n but for each
x R, lim
n
f
n
(x) = 0. So
liminf
n
_
R
f
n
d = >
_
R
liminf
n
f
n
d =
_
R
0 d = 0
This also shows that the Monotone Convergence Theorem is not true without Monotone.
2 200910 Mathematics MA2224
4.2 Almost everywhere
Denition 4.2.1. We say that a property about real numbers x holds almost everywhere (with
respect to Lebesgue measure ) if the set of x where it fails to be true has measure 0.
Proposition 4.2.2. If f : R [, ] is integrable, then f(x) R holds almost everywhere
(or, equivalently, [f(x)[ < almost everywhere).
Proof. Let E = x : [f(x)[ = . What we want to do is show that (E) = 0.
We know
_
R
[f[ d < . So, for any n N, the simple function n
E
satises n
E
(x)
[f(x)[ always, and so has
_
R
n
E
d = n(E)
_
R
[f[ d < .
But this cant be true for all n N unless (E) = 0.
Proposition 4.2.3. If f : R [, ] is measurable, then f satises
_
R
[f[ d = 0
if and only if f(x) = 0 almost everywhere.
Proof. Suppose
_
R
[f[ d = 0 rst. Let E
n
= x R : [f(x)[ 1/n. Then
1
n

En
[f[
and so
_
R
1
n

En
d =
1
n
(E
n
)
_
R
[f[ d = 0.
Thus (E
n
) = 0 for each n. But E
1
E
2
and

_
n=1
E
n
= x R : f(x) ,= 0.
So
(x R : f(x) ,= 0) =
_

_
n=1
E
n
_
= lim
n
(E
n
) = 0.
Conversely, suppose now that (x R : f(x) ,= 0) = 0. We know [f[ is a non-negative
measurable function and so there is a monotone increasing sequence (f
n
)

n=1
of measurable
simple functions that converges pointwise to [f[. From 0 f
n
(x) [f(x)[ we can see that
x R : f
n
(x) ,= 0 x R : f(x) ,= 0 and so (x R : f
n
(x) ,= 0) (x R :
Lebesgue integral 3
f(x) ,= 0) = 0. Being a simple function f
n
has a largest value y
n
(which is nite) and so if we
put E
n
= x R : f
n
(x) ,= 0 we have
f
n
y
n

En

_
R
f
n
d
_
R
y
n

En
d = y
n
_
R

En
d = y
n
(E
n
) = 0.
From the Monotone Convergence Theorem
_
R
[f[ d =
_
R
_
lim
n
f
n
_
d = lim
n
_
R
f
n
d = lim
n
0 = 0.
(Note: we dont really need the Monotone Convergence Theorem here. See Tutorial sheet 9 for
more direct way, using just the denition of the integral of a positive measurable function.)
The above result is one way of saying that integration ignores what happens to the integrand
on any chosen set of measure 0. Here is a more useful and
Proposition 4.2.4. Let f : R [, ] be an integrable function and g : R [, ]
a Lebesgue measurable function with f(x) = g(x) almost everywhere. Then g must also be
integrable and
_
R
g d =
_
R
f d.
Proof. Let E = x R : f(x) ,= g(x) (think of E as standing for exceptional) and note that
f(x) = g(x) almost everywhere means (E) = 0.
Write f = (1
E
)f +
E
f. Note that both (1
E
)f and
E
f are integrable because they
are measurable and satisfy [(1
E
)f[ [f[ and [
E
f[ [f[. Also

_
R

E
f d

_
R
[
E
f[ d = 0
as
E
f = 0 almost everywhere. Similarly
_
R

E
g d = 0.
So
_
R
f d =
_
R
((1
E
)f +
E
f) d
=
_
R
(1
E
)f d +
_
R

E
f d
=
_
R
(1
E
)f d + 0
=
_
R
(1
E
)g d
The same calculation shows
_
R
(1
E
)[g[ d =
_
R
[f[ d < , so that (1
E
)g must be
integrable. Thus g = (1
E
)g +
E
g is also integrable and we have
_
R
g d =
_
R
(1

E
)g d +
_
R

E
g d =
_
R
f d + 0.
4 200910 Mathematics MA2224
Remark 4.2.5. It follows that we should be able to manage without allowing integrable functions
to have the values . The idea is that, if f is integrable, it must be almost everywhere nite.
If we change all the values where [f(x)[ = to 0 (say) we are only changing f on a set of xs
of measure zero. This is exactly changing f to the (1
E
)f in the above proof. The changed
function will be almost everywhere the same as the original f, but have nite values everywhere.
So from the point of view of the integral of f, this change is not signicant.
However, it can be awkward to have to do this all the time, and it is better to allow f(x) =
.
4.3 Dominated convergence theorem
Theorem 4.3.1 (Lebesgue dominated convergence theorem). Suppose f
n
: R [, ] are
(Lebesgue) measurable functions such that the pointwise limit f(x) = lim
n
f
n
(x) exists.
Assume there is an integrable g : R [0, ] with [f
n
(x)[ g(x) for each x R. Then f is
integrable as is f
n
for each n, and
lim
n
_
R
f
n
d =
_
R
lim
n
f
n
d =
_
R
f d
Proof. Since [f
n
(x)[ g(x) and g is integrable,
_
R
[f
n
[ d
_
R
g d < . So f
n
is integrable.
We know f is measurable (as a pointwise limit of measurable functions) and then, similarly,
[f(x)[ = lim
n
[f
n
(x)[ g(x) implies that f is integrable too.
The proof does not work properly if g(x) = for some x. We know that g(x) < almost
everywhere. So we can take E = x R : g(x) = and multiply g and each of the functions
f
n
and f by 1
E
to make sure all the functions have nite values. As we are changing them
all only on the set E of measure 0, this change does not affect the integrals or the conclusions.
We assume then all have nite values.
Let h
n
= g f
n
, so that h
n
0. By Fatous lemma
liminf
n
_
R
(g f
n
) d
_
R
liminf
n
(g f
n
) d =
_
R
(g f) d
and that gives
liminf
n
__
R
g d
_
R
f
n
d
_
=
_
R
g d limsup
n
_
R
f
n
d
_
R
g d
_
R
f d
or
limsup
n
_
R
f
n
d
_
R
f d (1)
Repeat this Fatous lemma argument with g + f
n
rather than g f
n
. We get
liminf
n
_
R
(g + f
n
) d
_
R
liminf
n
(g + f
n
) d =
_
R
(g + f) d
Lebesgue integral 5
and that gives
liminf
n
__
R
g d +
_
R
f
n
d
_
=
_
R
g d + liminf
n
_
R
f
n
d
_
R
g d +
_
R
f d
or
liminf
n
_
R
f
n
d
_
R
f d (2)
Combining (1) and (2) we get
_
R
f d liminf
n
_
R
f
n
d limsup
n
_
R
f
n
d
_
R
f d
which forces
_
R
f d = liminf
n
_
R
f
n
d = limsup
n
_
R
f
n
d
and that gives the result because if limsup
n
a
n
= liminf
n
a
n
(for a sequence (a
n
)

n=1
), it
implies that lim
n
a
n
exists and lim
n
a
n
= limsup
n
a
n
= liminf
n
a
n
.
Remark 4.3.2. The example following Fatous lemma also shows that the assumption about the
existence of the dominating function g cant be dispensed with.
4.4 Applications of the dominated convergence theorem
Theorem 4.4.1 (Continuity of integrals). Assume f : R R R is such that x f
[t]
(x) =
f(x, t) is measurable for each t R and t f(x, t) is continuous for each x R. Assume also
that there is an integrable g : R R with [f(x, t)[ g(x) for each x, t R. Then the function
f
[t]
is integrable for each t and the function F : R R dened by
F(t) =
_
R
f
[t]
d =
_
R
f(x, t) d(x)
is continuous.
Proof. Since f
[t]
is measurable and [f
[t]
[ g we have
_
R
[f
[t]
[ d
_
R
g d < and so f
[t]
is
integrable (for each t R). This F(t) makes sense.
To show that F is continuous at t
0
R it is enough to show that for each sequence (t
n
)

n=1
with lim
n
t
n
= t
0
we have lim
n
F(t
n
) = F(t
0
).
But that follows from the dominated convergence theorem applied to f
n
(t) = f(x, t
n
), since
we have
lim
n
f
n
(t) = lim
n
f(x, t
n
) = f(x, t
0
)
by continuity of t f(x, t). We also have [f
n
(t)[ = [f(x, t
n
)[ g(x) for each n and each
x R.
6 200910 Mathematics MA2224
Example 4.4.2. Show that
F(t) =
_
[0,)
e
x
cos(t) d(x)
is continuous.
Proof. The idea is to apply the theorem with dominating function g(x) given by
g(x) =
[0,)
(x)e
x
=
_
e
x
for x 0
0 for x < 0
We need to knowthat
_
R
g d < (and that g is measurable and that x
[0,)
(x)e
x
cos(t)
is measurable for each t but we do know that these are measurable because e
x
is continuous
and
[0,)
is measurable).
By the Monotone Convergence Theorem,
_
R
g d = lim
n
_
R

[n,n]
g d = lim
n
_
R

[0,n]
e
x
d(x) = lim
n
_
n
0
e
x
d(x)
You can work this out easily using ordinary Riemann integral ideas and the limit is 1. So
_
R
g d < .
Now the theorem applies because
[
[0,)
(x)e
x
cos(t)[ g(x)
for each (x, t) R
2
(and certainly t
[0,)
(x)e
x
cos(t) is continuous for each x).
Theorem 4.4.3 (Differentiating under the integral sign). Assume f : R R R is such that
x f
[t]
(x) = f(x, t) is measurable for each t R, that f
[t
0
]
(x) = f(x, t
0
) is integrable for
some t
0
R and
f(x,t)
t
exists for each (x, t). Assume also that there is an integrable g : R R
with [
f
t
[
(x,t)
[ g(x) for each x, t R. Then the function x f(x, t) is integrable for each t
and the function F : R R dened by
F(t) =
_
R
f
t
d =
_
R
f(x, t) d(x)
is differentiable with derivative
F

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

t
f(x, t) d(x).
Proof. Applying the Mean Value theorem to the function t f(x, t), for each t ,= t
0
we have
to have some c between t
0
and t so that
f(x, t) f(x, t
0
) =
f
t
[
(x,c)
(t t
0
)
Lebesgue integral 7
It follows that
[f(x, t) f(x, t
0
)[ g(x)[t t
0
[
and so
[f(x, t)[ [f(x, t
0
)[ + g(x)[t t
0
[.
Thus
_
R
[f(x, t)[ d(x)
_
R
([f(x, t
0
)[ + g(x)) d(x) =
_
R
[f(x, t
0
)[ d(x) +
_
R
g d < ,
which establishes that the function x f(x, t) is integrable for each t.
To prove the formula for F

(t) consider any sequence (t

n
)

n=1
so that lim
n
t
n
= t but
t
n
,= t for each t. We claim that
lim
n
F(t
n
) F(t)
t
n
t
=
_
R

t
f(x, t) d(x). (3)
We have
F(t
n
) F(t)
t
n
t
=
_
R
f(x, t
n
) f(x, t)
t
n
t
d(x) =
_
R
f
n
(x) d(x)
where
f
n
(x) =
f(x, t
n
) f(x, t)
t
n
t
.
Notice that, for each x we know
lim
n
f
n
(x) =
f
t
[
(x,t)
and so (3) will follow from the dominated convergence theorem once we show that [f
n
(x)[
g(x) for each x.
That follows from the Mean Value theorem again because there is c between t and t
n
(with c
depending on x) so that
f
n
(x) =
f(x, t
n
) f(x, t)
t
n
t
=
f
t
[
(x,c)
.
So [f
n
(x)[ g(x) for each x.
4.5 Whats missing?
There are quite a few topics that are very useful and that we have not covered at all. Some of the
things we have covered are simplied from the way they are often stated and used.
An example in the latter category is that the Monotone Convergence Theorem and the Dom-
inated Convergence Theorem are true if we only assume the hypotheses are valid almost every-
where. The Monotone Convergence Theorem is still true if we assume that the sequence (f
n
)

n=1
of measurable functions satises f
n
0 almost everywhere and f
n
f
n+1
almost everywhere
8 200910 Mathematics MA2224
(for each n). Then the pointwise limit f(x) = lim
n
f
n
(x) may exist only almost everywhere.
Something similar for the Dominated Convergence Theorem.
We stuck to integrals of functions f(x) dened for x R (or for x X L which
is more or less the same because we can extend them to be zero on R X) and we used only
Lebesgue measure on the Lebesgue -algebra L. What we need abstractly is just a mea-
sure space (X, , ) and -measurable integrands f : X [, ]. By denition f is -
measurable if
f
1
([, a]) = x X : f(x) a (a R),
and it follows from that condition that f
1
(B) for all Borel subsets B R. We can
then talk about simple functions on X (f : X R with nite range f(X) = y
!
, y
2
, . . . , y
n
),
their standard form (f =

n
j=1
y
j

F
j
where F
j
= f
1
(y
j
)), integrals of non-negative -
measurable simple functions (
_
X
f d =

n
j=1
y
j
(F
j
)), integrals of non-negative measur-
able f : X [0, ], (generalised) Monotone convergence theorem, -integrable -measurable
f : X [, ], (generalised) Dominated Convergence Theorem.
To make this applicable, we would need some more examples of measures, other that just
Lebesgue measure : L [0, ]. We did touch on some examples of measures that dont
correspond so obviously to total length as does. For instance if f : R [0, ] is a non-
negative (Lebesgue) measurable function, there is an associated measure
f
: L [0, ] given
by
f
(E) =
_
E
f d. (We discussed
f
when f was simple but the fact that
f
is a measure
even when f is not simple follows easily from the Monotone Convergence Theorem.) If we
choose f so that
_
R
f d = 1, then we get a probability measure from
f
(and f is called a
probability density function). The standard normal distribution is the name given to
f
when
f(x) = (1/

2)e
x
2
/2
. Thats just one example.
The Radon Nikodym theorem gives a way to recognise measures : L [0, ] that are of
the form =
f
for some non-negative (Lebesgue) measurable function f. The key thing is that
(E) = 0 (E) = 0. The full version of the Radon Nikodym theorem applies not just to
measures on (R, L) with respect to , but to many more general measure spaces.
And then we could have explained area measure on R
2
, volume measure on R
3
, n-dimensional
volume measure on R
n
, Fubinis theorem relating integrals on product spaces to iterated inte-
grals, the change of variables formula for integrals on R
n
, and quite a few more topics.
One place where measure theory comes up is in dening so-called Lebesgue spaces, which
are Banach spaces dened using (Lebesgue) integration. For example L
1
(R) is the space of inte-
grable functions f : R R with norm |f|
1
=
_
R
[f[ d. Or to be more precise it is the space of
almost everywhere equivalence classes of such functions. (That is so that |f|
1
= 0 only for the
zero element of the space, a property that norms should have.) To make L
1
(R) complete we need
the Lebesgue integral. Hilbert spaces like L
2
(R) come into Fourier analysis, for instance. By
denition L
2
(R) is the space of almost everywhere equivalence classes of measurable f : R R
that satisfy
_
R
[f[
2
d < with norm given by |f|
2
=
__
R
[f[
2
d
_
1/2
.
In short then, there is quite a range of things that the Lebesgue theory is used for (probabil-
ity theory, Fourier analysis, differential equations and partial differential equations, functional
analysis, stochastic processes, . . . ). My aim was to lay the basis for studying these topics later.
R. Timoney March 27, 2010