Constructing measures

Based on our experience with metric spaces and Euclidean geometry, it is reasonable to think of the interval (a,b) as having length ba. In this section, using the tools we have developed, we will construct a measure on R that agrees with that suggestion.

What will the domain of the measure be? We know that measures are defined on σ-algebras, and we want all of the intervals (a,b) to belong to the σ-algebra. Every σ-algebra that contains all of the intervals (a,b) must contain the Borel σ-algebra Bor(R) on R. We can therefore rephrase our goal as follows: construct a measure λ on Bor(R) such that λ((a,b))=ba for all a<b. In this section we will construct such a measure. To do so we will introduce two new ideas: outer measures and Lebesgue measurable sets.

Outer measure

If a Borel set B is covered by a union of intervals Bn=1(an,bn) then we must have λ(B)n=1λ((an,bn))=n=1bnan by subadditivity. The idea behind the definition of outer measure is that we should assign a size to B by finding the most efficient way of covering it by a union of intervals. We define λ(B)=inf{n=1bnan:Bn=1(an,bn)} for all sets BR.

The function λ is defined on P(R). It assigns a value in [0,] to every subset of R. However, as we have seen already, it is not a measure on the σ-algebra P(R) because it is not countably additive. However, it does have some properties in common with measures.

Theorem

The map λ:P(R)[0,] has the following properties.

  1. λ()=0.
  2. λ(A)λ(B) whenever AB.
  3. λ(A1A2)λ(A1)+λ(A2)+λ(A3)+ for all A1,A2,A3,R.
Proof:

These properties are proved in the length section.

Quite generally, a mapping Ξ from P(X) to [0,] satisfying the three properties above is called an outer measure.

Lebesgue measurable sets

The central step in constructing a measure from an outer measure is to identify as large a σ-algebra as possible on which λ defines a measure. Identifying a criterion for membership in such a σ-algebra was one of Lebesgue's great insights. Suppose that we have such a σ-algebra to hand. It must contain the open intervals, and thus for any set L in the σ-algebra we also have IL and IL in the σ-algebra. Moreover, the measure of I must be equal to the sum of the measures of IL and IL. Lebesgue defined a set to be measurable if, in that sense, it splits every open interval appropriately.

We will not proceed directly with Lebesgue's definition. Instead we will work with a stranger but more applicable definition due to Carathéodory in which not only the intervals, but all subsets of R must be split properly for a set to be measurable.

Definition

A subset L of R is Lebesgue measurable if λ(S)=λ(SL)+λ(SLc) for every set SR. Write Leb(R) for the collection of all Lebesgue measurable subsets of R.

Since λ is an outer measure we have λ(S)λ(SL)+λ(SLc) for all S,LR. It therefore suffices, when aiming to check a set L is Lebesgue measurable, that λ(S)λ(SL)+λ(SLc) for all sets SR.

Theorem

The collection Leb(R) is a σ-algebra and the restriction of λ to Leb(R) is a measure.

Proof:

We know that λ()=0. The empty set is therefore Lebesgue measurable because S= and S=S for every SR. Thus Leb(R) is non-empty.

The expression defining Lebesgue measurability is symmetric in terms of L and Lc so it is immediate that LLeb(R) implies Lc is Lebesgue measurable.

Next we prove that if L,MR are both Lebesgue measurable then so too is LM. Fix SR. We have λ(S)=λ(SL)+λ(SLc)=λ(SLM)+λ(SLMc)+λ(SLcM)+λ(SLcMc)λ(S(LM))+λ(S(LM)c) where we have used Lebesgue measurability of L for the first equality and Lebesgue measurability of M twice for the second equality. The inequality follows from subadditivity of outer measures. Since SR was arbitrary we conclude that LM is Lebesgue measurable.

Now, if L,M and Lebesgue measurable and disjoint, we can calculate λ(LM)=λ((LM)L)+λ((LM)Lc)=λ(L)+λ(M) which proves λ is additive on the σ-algebra Leb(R).

As Leb(R) is now known to be closed under complements and finite unions it is also closed under finite intersection. To prove it is a σ-algebra it therefore suffices to take pairwise disjoint sets L1,L2, in Leb(R) and prove that their union M is also in Leb(R). Fix SR and calculate for every NN that λ(S(L1LN))=λ(S(L1LN)LN)+λ(S(L1LN)LNc)=λ(SLN)+λ(S(L1LN1)) and therefore λ(S(L1LN))=n=1Nλ(SLn) by induction. We have for every NN that λ(S)λ(S(L1LN)c)+λ(S(L1LN))λ(SMc)+n=1Nλ(SLn) and therefore λ(S)λ(SMc)+n=1λ(SLn)λ(SMc)+λ(Sn=1Ln)λ(S) from which we conclude all terms above are equal. In particular M belongs to Leb(R) and taking S=M gives λ(n=1Ln)=n=1λ(Ln) which is countable additivity.

We find ourselves in the following situation: there is a σ-algebra Leb(R) on which λ defines a measure. Membership in Leb(R) was defined by a bizarre property that made it possible to prove the above theorem. What sets actually have this bizarre property? It may be the case that Leb(R) is the trivial σ-algebra {,R}. To know that (R,Leb(R)) is an interesting measurable space, and to realize our ambition of defining a measure on a σ-algebra that contains the intervals and agrees with length, we conclude by checking that Leb(R) contains the Borel σ-algebra.

Theorem

The Lebesgue σ-algebra Leb(R) contains the Borel σ-algebra Bor(R).

Proof:

Since Bor(R) is generated by the open intervals (a,b) with a<b it suffices to prove that each open interval (a,b) belongs to Leb(R). Fix SR. We need to verify that λ(S)λ(SI)+λ(SIc) where I=(a,b).

First we check that λ(I)=ba. Certainly λ(I)<ba because I is an interval that covers I. For the converse fix ϵ>0. There is a sequence n(an,bn) of intervals with (a,b)n=1(an,bn) and n=1bnanλ(I)+ϵ as well. Now n(an,bn) is an open cover of [a+ϵ,bϵ] and we can therefore find r(1),,r(v) such that (ar(1),br(1)),,(ar(v),br(v)) is a finite subcover of [a+ϵ,bϵ]. It must then be the case that n=1vbr(n)ar(n)ba2ϵ whence λ(I)(n=1bnan)ϵba3ϵ and we conclude, since ϵ>0 was arbitrary, that λ(I)=ba.

If λ(S)= there is nothing to prove, so assume λ(S) is finite. Fix ϵ>0 and let nJn be a sequence of intervals with n=1λ(Jn)λ(S)+ϵ that covers S. (Here we have used that λ assignes the correct lengths to intervals.) Now λ(S)+ϵn=1λ(Jn)=n=1λ(JnI)+n=1λ(JnIc)λ(SI)+λ(SIc) for every ϵ>0. Thus I belongs to Leb(R).