proof of Vitali convergence theorem∗

stevecheng†
2013-03-21 23:45:35

Theorem. Let f1 , f2 , . . . be Lp -integrable functions on a measure space (X, µ),

for 1 ≤ p < ∞. The following conditions are necessary and sufficient for fn to
be a Cauchy sequence in the Lp (X, µ) norm:

(i) the sequence fn is Cauchy in measure;

(ii) the functions {|fn |p } are uniformly integrable; and

(iii) for each  > 0, there is a set A of finite measure, with kfn 1(X \ A)k < 
for all n.

Proof. We abbreviate |fn − fm | by fmn .

Necessity of (i). Fix t > 0, and let Emn = {fmn ≥ t}. Then
1 1
µ(Emn )1/p = kt 1(Emn )k ≤ kfmn k → 0 , as m, n → ∞.
t t

Necessity of (ii). Select N such that kfn − fN k <  when n ≥ N . The family
{|f1 |p , . . . , |fN −1 |p , |fN |p } is uniformly integrable because it consists of
only finitely many integrable functions.
So for every  > 0, there is δ > 0 such that µ(E) < δ implies kfn 1(E)k < 
for n ≤ N . On the other hand, for n > N ,

kfn 1(E)k ≤ k(fn − fN )1(E)k + kfN 1(E)k < 2

for the same sets E, and thus the entire infinite sequence {|fn |p } is uni-
formly integrable too.
Necessity of (iii). Select N such that kfn − fN k <  for all n ≥ N . Let ϕ be a
simple function approximating fN in Lp norm up to . Then kfn −ϕk < 2
∗ hProofOfVitaliConvergenceTheoremi created: h2013-03-21i by: hstevechengi version:

h39909i Privacy setting: h1i hProofi h28A20i

† This text is available under the Creative Commons Attribution/Share-Alike License 3.0.

You can reuse this document or portions thereof only if you do so under terms that are
compatible with the CC-BY-SA license.

1
 for all n ≥ N . Let AN = {ϕ 6= 0} be the support of ϕ, which must have
finite measure. It follows that

kfn 1(X \ AN )k = kfn − fn 1(AN )k ≤ kfn − ϕk + kϕ − fn 1(AN )k

= kfn − ϕk + k(ϕ − fn )1(AN )k
< 2 + 2 .

For each n < N , we can similarly construct sets An of finite measure, such
that kfn 1(X \ An )k < 4. If we set A = A1 ∪ · · · ∪ AN −1 ∪ AN , a finite
union, then A has finite measure, and clearly kfn 1(X \ A)k < 4 for any
n.
Sufficiency. We show fmn to be small for large m, n by a multi-step estimate:

kfmn k ≤ kfmn 1(A \ Emn )k + kfmn 1(Emn )k + kfmn 1(X \ A)k .

Use condition (iii) to choose A of finite measure such that kfn 1(X \A)k < 
for every n. Then kfmn 1(X \ A)k < 2.
Let t = /µ(A)1/p > 0, and Emn = {fmn ≥ t}. By condition (ii) choose
δ > 0 so that kfn 1(E)k <  whenever µ(E) < δ. By condition (i), take
N such that if m, n ≥ N , then µ(Emn ) < δ; it follows immediately that
kfmn 1(Emn )k < 2.
Finally, kfmn 1(A \ Emn )k ≤ tµ(A)1/p = , since fmn < t on the comple-
ment of Emn . Hence kfmn k < 5 for m, n ≥ N .

Remark. In the statement of the theorem, instead of dealing with Cauchy

sequences, we can directly speak of convergence of fn to f in Lp and in measure.
This variation of the theorem is easily proved, for:

• a sequence converges in Lp if and only if it is Cauchy in Lp ;

• a sequence that converges in measure is automatically Cauchy in measure;
• a simple adaptation of the argument shows that fn → f in Lp implies
fn → f in measure; and

• the limit in measure is unique.