Real Analysis MAA 6616

Lecture 11
Egorov and Luzin Theorems
Egorov’s Theorem
Egorov’s Theorem states that if a sequence of measurable functions converges pointwise a.e. on
a set of finite measure to a function that is a.e. finite, then it converges uniformly except on a
subset with arbitrarily small measure.
Start with an example. Consider the sequence of piecewise linear functions {fn } defined on
[0, 2] as follows:
▶ For n even:

1 1 1 1
fn (0) = fn ( ) = fn (1 − ) = fn (1 + ) = fn (2 − ) = fn (2) = 0;
n 2n 2n n
1 1 1
fn ( ) = fn (2 − ) = 1 and f (1) = 1 −
2n 2n n

and fn is the linear function connecting any two consecutive points so that f (x) = 2nx for
x ∈ [0, 1/2n];
▶ For n odd:

1 1 1 1
fn (0) = fn ( ) = fn (1 − ) = fn (1 + ) = fn (2 − ) = fn (2) = 0;
n 2n 2n n
1 1 1
fn ( ) = fn (2 − ) = −1 and f (1) = −1 +
2n 2n n

and fn is the linear function connecting any two consecutive points so that f (x) = −2nx
for x ∈ [0, 1/2n].
The sequence {fn } converges pointwise on [0, 1) ∪ (1, 2] to the function f = 0 and the
sequence diverges for x = 1.
Indeed f2j (1) = 1 − (1/(2j)) converges to 1 while f2j+1 (1) = −1 + (1/(2j + 1)) converges to
−1. If x = 0 or x = 2, then fn (x) = 0 and the sequence converges to 0. If x ̸= 0, 1, 2, let
1
δ = min(|x| , |1 − x| , |2 − x|). Let N ∈ N such that N ≥ . Then for every n ≥ N we have
δ
fn (x) = 0 and fn (x) converges to 0.
1
Now for any ϵ > 0 (no matter how small), let N ≥ , then {fn } converges uniformly to f = 0
4ϵ
ϵ ϵ ϵ ϵ
on the set [ , 1 − ] ∪ [1 + , 2 − ] with measure 2 − ϵ.
4 4 4 4
Lemma (1)
Let {fn } be a sequence of measurable functions on a measurable set E ⊂ Rq with finite
measure. Assume that {fn } converges pointwise a.e. on E to a function f such that f is finite a.e.
on E. Then for every ϵ > 0 and η > 0, there exists a measurable set A ⊂ E and an integer
N ∈ N such that m(E\A) < η and |fn (x) − f (x)| < ϵ for all x ∈ A and for all n > N.

Proof.
Let Z1 be the subset of E where f is not finite and let Z2 be the subset of E where {fn } fails to converge to f . Let Z = Z1 ∪ Z2 .
Then m(Z) = 0 by hypothesis. For every j ∈ N, let Aj = {x ∈ E\Z : |f (x) − fk (x)| < ϵ for all k ≥ j}. The set Aj is
\
measurable since it can be expressed as a countable intersection of measurable sets: Aj = {|f − fk | < ϵ} ∩ (E\Z).
k≥j
S
Note that Aj ⊂ Aj+1 and j≥1 Aj = E\Z (since fn −→ f on E\Z). It follows from the continuity of the Lebesgue measure
that lim m(An ) = m(E\Z) = m(E). Therefore for the given η > 0 there exists N ∈ N such that m(E\AN ) < η and for
n→∞
every x ∈ A = AN and for every n ≥ N we have |fn (x) − f (x)| < ϵ.
Egorov’s Theorem

Theorem (1)
Let {fn } be a sequence of measurable functions on a measurable set E ⊂ Rq with finite
measure. Assume that {fn } converge pointwise a.e. on E to a function f such that f is finite a.e.
on E. Then for every η > 0 there exists a closed set A ⊂ E such that m(E\A) < η and {fn }
converges uniformly to f on A.

Proof.
Let m ∈ N. It follows from Lemma 1 that there exists a measurable set Am ⊂ E and N(m) ∈ N such that
η 1
m(E\Am ) < and |fn − f | < for all n ≥ N(m).
2m+1 m
∞
\
Let à = Am . The set à is measurable and
m=1

 
∞ ∞
[ X η η
m(E\Ã) = m  (E\Am ) ≤ = .
m=1 m=1
2m+1 2

1 < ϵ. It follows from the

Now we prove that {fn } converges uniformly to f on Ã. Let ϵ > 0 and let m ∈ N such that m
1
definition of à and of Am that for every n ≥ N(m) and for every x ∈ à ⊂ Am we have |fn − f | < < ϵ. This proves the
m
uniform convergence on Ã
Next, since à is measurable we can find a closed set A ⊂ à such that m(Ã\A) < η/2. The sequence is uniformly convergent
on A and   η η
m(E\A) = m (E\Ã) ∪ (Ã\A) ≤ m(E\Ã) + m(Ã\A) ≤ + =η
2 2
Luzin’s Theorem states that a measurable function on E is "nearly continuous" in the following
sense: For any ϵ > 0, there is a subset whose measure is within ϵ to that of E and on which the
function is continuous. We start with the case of simple functions.

Proposition (1)
Let ϕ be a simple and measurable function defined on a set E. Then for every ϵ > 0 there exists
a closed F ⊂ E such that m(E\F) < ϵ and ϕ continuous on F.

Proof. n
X
We use the canonical representation of simple function to express ϕ as ϕ = aj χE for some collection of n disjoint
j
j=1
measurable Ssets E1 , · · · , En in E. For ϵ > 0, there exist closed sets F1 , · · · , Fn with Fj ⊂ Ej and m(Ej \Fj ) < ϵ/n.
Then F = nj=1 Fj is a closed subset in E and m(E\F) < ϵ. Note that the restriction of ϕ to each Fj is continuous since it is
constant on each Fj . It remains to verify that ϕ is continuous on their disjoint union F.
For N ∈ N, let BN = BN (0) be the ball in Rq with center 0 and radius N. Let FjN = Fj ∩ BN and F N = nj=1 FjN . To
S
N S N
prove the continuity of ϕ on F it is enough to prove continuity on F for an arbitrary N (since F = N F ). To prove
continuity on F N , it is enough to verify that the closed sets FjN ’s are separated. That is, there exists δ0 > 0 such that for j ̸= k,
and for every x ∈ FjN and y ∈ FkN we have |x − y| ≥ δ0 .
For j ̸= k, let δj,k = inf{|x − y| : x ∈ FjN , y ∈ FkN }. We claim that δj,k > 0. Indeed, if δj,k = 0, then there would be
sequences {xp }p ⊂ FjN and {yp }p ⊂ FkN such that lim xp − yp = 0. Since FjN and FkN are compact, then we can
p→∞
extract convergent subsequences {xpm }m ⊂ FjN and {ypm }m ⊂ FkN that converge to the same limit z ∈ FjN ∩ FkN (since
δj,k = 0) and this is a contradiction since Fj and Fk are disjoint. Hence δj,k > 0 and δ0 = min{δj,k : j ̸= k} is strictly
positive. Therefore the FjN ’s are separated and ϕ is continuous on F N .
Luzin’s Theorem

Theorem (2)
Let f : E ⊂ Rq −→ R be measurable and finite a.e. on E. Then for every ϵ > 0, there exists a
closed F ⊂ E such that m(E\F) < ϵ and f continuous on F.

Proof.
Since f is measurable, by the Simple Approximation Theorem, there exists a sequence {ϕn } of simple functions on E which
converges pointwise on E to f . According to Egorov’s Theorem, given any ϵ > 0, there exists a closed set C ⊂ E with
m(E\C) < ϵ/2 and {ϕn } converges uniformly on C. For each n ∈ N, there exists a closed set Fn ⊂ E with
T∞
m(E\Fn ) < (ϵ/2n+1 ) such that the simple function ϕn is continuous on Fn . Consider the closed set F = C ∩


S∞
n=1 Fn .
We have E\F = (E\C) ∪ n=1 (E\Fn ) so that

∞ ∞
X ϵ X ϵ
m(E\F) ≤ m(E\C) + m (E\Fn ) ≤ + = ϵ.
n=1
2 n=1
2n+1

Since each function ϕn is continuous on F ⊂ Fn and {ϕn } converges uniformly on F, then the limit f is also continuous on
F.