326 CHAPTER 14. THE LEBESGUE INTEGRAL

Definition 14.3.4 Let µ = µF be defined on B (R) as follows.

µ (E)≡ supn∈N

∫ n

−nXE∩[−n,n] (x)dF = lim

n→∞

∫ n

−nXE∩[−n,n] (x)dF

Does this definition make sense? I need to verify the limit exists because{∫ n

−nXE∩[−n,n] (x)dF

}∞

n=1

is an increasing sequence. If unbounded, then the limit is ∞ and if bounded, it is the reallimit of the sequence. Thus the definition does make sense.

Lemma 14.3.5{∫ n−n XE∩[−n,n] (x)dF

}∞

n=1 is increasing in n. Also, if E is a bounded set

contained in [−(n−1) ,n−1] , then∫ n+1−(n+1)XE∩[−(n+1),n+1] (x)dF =

∫ n−n XE∩[−n,n] (x)dF.

Proof:∫ n+1−(n+1)XE∩[−(n+1),n+1] (x)dF =

∫ −n

−(n+1)XE∩[−(n+1),n+1] (x)dF +

∫ n

−nXE∩[−(n+1),n+1] (x)dF

+∫ n+1

nXE∩[−(n+1),n+1] (x)dF ≥

∫ n+1

−(n+1)XE∩[−n,n] (x)dF

If E ⊆ [−(n−1) ,n−1] , then∫ n+1

−(n+1)XE∩[−(n+1),n+1] (x)dF =

∫ n+1

−(n+1)XE∩[−(n−1),n−1] (x)dF

=∫ −n

−(n+1)XE∩[−(n−1),n−1] (x)dF +

∫ n

−nXE∩[−(n−1),n−1] (x)dF

+∫ n+1

nXE∩[−(n−1),n−1] (x)dF

=∫ n

−nXE∩[−(n−1),n−1] (x)dF =

∫ n

−nXE∩[−n,n] (x)dF

because on [−(n+1) ,−n] and [n,n+1] all the sums in defining the integrals are 0. Thus,if E is bounded, the integrals giving the measure as a limit are eventually constant andµ (E)< ∞.

Next I need to verify that this is a measure. Let {Ei}∞

i=1 be disjoint sets in B (R). Then∪∞

i=1Ei ∈B (R) and so, from the monotone convergence theorem for generalized Riemannintegrals,

µ (∪∞i=1Ei) ≡ sup

n

∫ n

−nX∪∞

i=1Ei∩[−n,n] (x)dF = supn

limm→∞

∫ n

−nX∪m

i=1Ei∩[−n,n] (x)dF

= supn

limm→∞

∫ n

−n

m

∑i=1

XEi∩[−n,n]dF = supn

supm

∫ n

−n

m

∑i=1

XEi∩[−n,n]dF

326 CHAPTER 14. THE LEBESGUE INTEGRALDefinition 14.3.4 rer u =u p be defined on B(R) as follows.U(E) = sup Pen nyn| x) dF = tim | ZEA —n,n|neN- neoDoes this definition make sense? I need to verify the limit exists becauseco{ - ZL EA-na] (x) ar}is an increasing sequence. If unbounded, then the limit is co and if bounded, it is the reallimit of the sequence. Thus the definition does make sense.n=1Lemma 14.3.5 {[",, Zepinn (dP J is increasing inn. Also, if E is a bounded setcontained in [—(n—1),n—1], then ["" bet) Reo[— (nt) net] (X) dF = fo, Zea nnj (4) dF.Proof: ara bet) Ze \l-(n+1).n41] (x)dF =I, gay ZENO Dan (yar + [ Ze\l—(n+)nt1] (4) dFn+n+l+f Rep [—(n+1) nt] (x) dF =. Lenn] (x) dFIf FE C[—(n—1),n—1], thenn+ n+1ReA—(ntl).n x)aF = | LeQ—(n—1).n—1] (x) dFDic EN[-(n+1),n+1] () “ney PEN OD 1) @)= I, ay PEMD) x) dF + | ent (n—1),n—1] (4) dFn+l+ ZeEn|-(n-1),n—1] (x) dFn n= , LEA(-(n-1),n-1] (x) dF = , LEN [-na] (x) dFbecause on [—(n+1),—n] and [n,n + 1] all the sums in defining the integrals are 0. Thus,if E is bounded, the integrals giving the measure as a limit are eventually constant andU(E)<~Next I need to verify that this is a measure. Let {E£;};" , be disjoint sets in 4 (R). ThenUE; € @(R) and so, from the monotone convergence theorem for generalized Riemannintegrals,“n niu (Uj Ei) sup | Rose B:nt—nn| (x) dF = sup lim i Rom Bol-na] (x) dFn J— nn m-,eenm= sup lim i y Re p(—naldF = sup sup y A i- —n j=]n m—reo Nj=| m