Convergence Relationships

Michael Van Opstall

Throughout, let X

µ
be a measure space and f n be a sequence of µ -measurable functions.

1 Definitions
 f almost everywhere if µ
x f x  f x  0.
Definition 1.1. f n n



Definition 1.2. f  f almost uniformly if for every ε 0, there exists a set A of measure less than ε such that f  f
n
n
uniformly off A.
Definition 1.3. f  f in L if  f  f   0.
n
p
n p

Definition 1.4. f  f in measure if for all ε 0, there exists N such that whenever n N, µ
x f x  f x 
ε ε.
n

n

2 Convergence with no Restrictions

Theorem 2.1. If f n  f a.u., then f  f a.e. n

Proof. Given ε 0, we can choose a set A with µ A ε such that f  f uniformly off A. Since ε is arbitrary, the set
of points where f x  f x has measure 0.

 n
n

Theorem 2.2. If f  f a.u., then f  f in measure.

n n

Proof. Given ε 0, we can choose a set A with µ A ε such that f  f uniformly off A, that is, there exists N such
that on X A,  f x  f x  ε for all x when n N. Then µ
x f x  f x  ε  µ A ε , so f  f in measure.

 n
n


n


 n

Theorem 2.3. If f  f in L , then f  f in measure.

n
p
n

Proof. Suppose f  f in measure. Then there exists ε 0 such that for all n, µ
x f x  f x  ε  ε . Let
An
x f x  f x  ε . Then for all n, we have
n

n




n

f  f  f x  f x  dx
n p  n


p
1 p

 f x  f x  dx An
n


p
1 p

1 p
p

ε
dx
An
p
1 p
 ε

  f  0.
so fn

1
3 Finite Measure Spaces
If we add the restriction that µ X , we gain the following results:


Theorem 3.1. Suppose µ X  . If f  f a.e., then f  f a.u.

n n

Proof. This is Egoroff’s theorem. The proof is not too difficult, but lengthy. See [1] for a sketch.
Theorem 3.2. Suppose µ X . If f  f a.e., then f  f in measure.
 n n

Proof. By 3.1, if f  f a.e., then f  f a.u. By 2.2, if f  f a.u., then f  f in measure.

n n n n

4 Sequences Dominated by an L p Function

or not µ X  ).  L (1 p ) such that  f  g, we add the following results (whether

If we require that there exists a function g p  n 

Theorem 4.1. Suppose
f  is dominated by an L function. If f  f a.e., then f  f a.u.

n
p
n n

Proof. Given ε 0, except for a set of finite measure, g x  ε since g  L , so the f converge uniformly except on
a set of finite measure, since  f x  g x  ε . But this set of finite measure is taken care of by Egoroff’s theorem



p
n
n



(3.1).
Theorem 4.2. Suppose
f  is dominated by an L function. If f  f a.e., then f  f in measure.
n
p
n n

Proof. By 4.1, if f  f a.e., then f  f a.u. By 2.2, if f  f a.u., then f  f in measure.

Theorem 4.3. Suppose
f  is dominated by an L function. If f  f a.e., then f  f in L .
n n n n
p p
n n n

Proof. This is a result of the dominated convergence theorem. For a proof of the DCT, see [1].


Theorem 4.4. Suppose f n is dominated by an L p function. If f n f a.u., then f  f in L . n
p

Proof. By 2.1, if f  f a.u. then f  f a.e. By 4.3, if f  f a.e., then f  f in L .

n n n n
p

Theorem 4.5. Suppose
f  is dominated by an L function. If f  f in measure, then f  f in L .

n
p
n n
p

Proof. See 4.3. The dominated convergence theorem also holds for convergence in measure: Every subsequence
of a sequence which converges in measure converges in measure. Every sequence which converges in measure has
a subsequence which converges a.e. Hence every subsequence of a sequence which converges in measure has a
subsequence which converges a.e., to which we can apply the DCT. Now the sequence of integrals is a sequence of
real numbers, so since every subsequence of it has a subsequence which converges, the sequence of integrals converges.

5 Counterexamples
Example 5.1. Let f n  χn
n
1 . Then
1. f n converges to 0 almost everywhere.
2. f n does not converge almost uniformly.
3. f n does not converge in measure.
4. f n does not converge in L p .
Proof.

1. Given x  Ê, whenever n x, f x 0.

n


2. For all n,  f x  0 1, for all x  n n

n
   1, and n n  1 has measure 1, so f n does not converge almost
uniformly.

2

  f x 
3. As above, µ x f n x


12  1 for all n.
4. The p-norm of each f n is 1, and hence cannot converge to 0, even though f n converges to 0 pointwise.

Example 5.2. Let f n  nχ0
1 n on the finite measure space 0 1. Then
1. f n converges to 0 almost everywhere.
2. f n converges to 0 almost uniformly.
3. f n converges to 0 in measure.
4. f n does not converge to 0 in L p .
Proof.

1. Given x  0 1 , whenever n

1x, f n x
 0.
2. Except for a set of measure 1n, f n x
 0.
3. Convergence almost uniformly implies convergence in measure.
4. The p-norm of each f n is 1, and hence cannot converge to 0, even though f n converges to 0 pointwise.


Example 5.3. Let f 2k
j  χ j 2k
j
1 2k  for all k and for j  0    2 k 1 for each k ([1]). Note that the measure
space 0 1 is finite and the functions are dominated by the characteristic function of 0 1 which is in L p .
1. f n does not converge almost everywhere.
2. f n does not converge almost uniformly.
3. f n converges in measure.
4. f n converges in L p .
Proof.

1. f n x
 1 for infinitely many values of n for every x and f n x
 0 for infinitely many values of n for every x.

2. If f n converged a.u., it would converge a.e.

 
3. For n
2k , the set x f n x

0 has measure less than 2 k
.
4. A sequence which converges in measure and is dominated by an L p function converges in L p .

References
[1] Royden, H. Real Analysis. Prentice Hall. 1988.

3
This paper is summarized in the following diagrams:
N O RESTRICTIONS :

Lp  m

ae au

W HEN µ X
 :

Lp m



 

ae  au

W HEN DOMINATED BY AN L p FUNCTION :

Lp m

 

 
ae  au