Vitali coverings on the real line

Jordan Bell
March 10, 2016

For x∈ℝ and r>0 write

B⁢(x,r)={y∈ℝ:|y-x|<r}.

Let λ be Lebesgue measure on the Borel σ-algebra of ℝ and let λ* be Lebesgue outer measure on ℝ.

A Vitali covering of a set E⊂ℝ is a collection 𝒱 of closed intervals such that for ϵ>0 and for x∈E there is some I∈𝒱 with x∈I and 0<λ⁢(I)<ϵ.

The following is the Vitali covering theorem.11 1 Klaus Bichteler, Integration – A Functional Approach, p. 161, Lemma 10.5; John J. Benedetto and Wojciech Czaja, Integration and Modern Analysis, p. 179, Theorem 4.3.1; Russell A. Gordon, The Integrals of Lebesgue, Denjoy, Perron, and Henstock, p. 52, Lemma 4.6.

Theorem 1 (Vitali covering theorem).

Let U be an open set in R with λ⁢(U)<∞, let E⊂U, and let V be a Vitali covering of E each interval of which is contained in U. Then for any ϵ>0, there are disjoint I1,…,In∈V such that

λ*⁢(E∖⋃j=1nIj)<ϵ.
Proof.

Suppose that I1,…,In∈𝒱 are pairwise disjoint. If E⊂⋃j=1nIj then I1,…,In satisfy the claim, and otherwise, let

Un=U∖⋃j=1nIj,

and there exists some x∈E∩Un. As x∈Un and Un is open, there is some η>0 such that B⁢(x,η)⊂Un and then as 𝒱 is a Vitali covering of E there is some I∈𝒱 with x∈I⊂B⁢(x,η)⊂Un. Thus δn>0 for

δn=sup⁡{λ⁢(I):I∈𝒱,I⊂Un},

and there is some In+1∈𝒱 with In+1⊂Un and λ⁢(In+1)>δn2.

For j≥1 write Ij=[xj-rj,xj+rj] and let Jj=[xj-5⁢rj,xj+5⁢rj], namely Jj is concentric with Ij and λ⁢(Jj)=5⁢λ⁢(Ij). Then, as the intervals I1,I2,… are pairwise disjoint Borel sets each contained in U,

∑j=1∞λ⁢(Jj)=5⁢∑j=1∞λ⁢(Ij)=5⁢λ⁢(⋃j=1∞Ij)≤5⁢λ⁢(U)<∞

and it follows from ∑j=1∞λ⁢(Jj)<∞ that ∑j=M∞λ⁢(Jj)→0 as M→∞, which with

λ⁢(⋃j=M∞Jj)≤∑j=M∞λ⁢(Jj)

yields λ⁢(⋃j=M∞Jj)→0 as M→∞.

Let M≥1. If x∈E∖⋃j=1∞Ij then x∈E∖⋃j=1MIj and so x∈UM, and as UM is open there is some η>0 with B⁢(x,η)⊂UM. But x∈E and 𝒱 is a Vitali covering of E, so there is some I∈𝒱 with x∈I and I⊂B⁢(x,η)⊂UM. Now, λ⁢(Ij+1)>δj2 and ∑j=1∞λ⁢(Ij)<∞ together imply δn→0 as n→∞, so there is some n for which δn<λ⁢(I). By the definition of δn as a supremum, this means that I⊄Un and so it makes sense to define N to be a minimal positive integer such that I⊄UN. M<N: if M≥N then I⊂UM⊂UN, contradicting I⊄UN. (We shall merely use that M≤N.) The fact that I⊄UN and I⊂UN-1 means that I∩IN≠∅ and also, by the definition of δN-1, λ⁢(I)≤δN-1<2⁢λ⁢(IN). Write I=[y-r,y+r]. I∩IN≠∅ tells us y-r≤xN+rN and y+r≥xN-rN, and λ⁢(I)<2⁢λ⁢(IN) tells us 2⁢r<4⁢rN, hence

y+r≤xN+rN+2⁢r≤xN+5⁢rN,y-r≥xN-rN-2⁢r≥xN-5⁢rN,

showing that

x∈I=[y-r,y+r]⊂JN⊂⋃j=M∞Jj.

This is true for each x∈E∖⋃j=1∞Ij, which means that

E∖⋃j=1∞Ij⊂⋃j=M∞Jj.

Because λ⁢(⋃j=M∞Jj)→0 as M→∞, this yields

λ*⁢(E∖⋃j=1∞Ij)=0.

But E∖⋃j=1nIj is an increasing sequence of sets tending to E∖⋃j=1∞Ij, therefore

λ*⁢(E∖⋃j=1nIj)→λ*⁢(E∖⋃j=1∞Ij)=0,n→∞,

so there is some n such that λ*⁢(E∖⋃j=1nIj)<ϵ and then I1,…,In satisfy the claim. ∎