1.2: Preparation
( \newcommand{\kernel}{\mathrm{null}\,}\)
We begin with two useful convergence criteria for improper integrals that do not involve a parameter. Consistent with the definition on p. 152, we say that f is locally integrable on an interval I if it is integrable on every finite closed subinterval of I.
[theorem:2] Suppose g is locally integrable on [a,b) and denote
G(r)=∫rag(x)dx,a≤r<b.
Then the improper integral ∫bag(x)dx converges if and only if, for each ϵ>0, there is an r0∈[a,b) such that
|G(r)−G(r1)|<ϵ,r0≤r,r1<b.
For necessity, suppose ∫bag(x)dx=L. By definition, this means that for each ϵ>0 there is an r0∈[a,b) such that
|G(r)−L|<ϵ2\quad and\quad|G(r1)−L|<ϵ2,r0≤r,r1<b.
Therefore
|G(r)−G(r1)|=|(G(r)−L)−(G(r1)−L)|≤|G(r)−L|+|G(r1)−L|<ϵ,r0≤r,r1<b.
For sufficiency, [eq:9] implies that
|G(r)|=|G(r1)+(G(r)−G(r1))|<|G(r1)|+|G(r)−G(r1)|≤|G(r1)|+ϵ,
r0≤r≤r1<b. Since G is also bounded on the compact set [a,r0] (Theorem 5.2.11, p. 313), G is bounded on [a,b). Therefore the monotonic functions
¯G(r)=sup{G(r1)|r≤r1<b}\quad and\quadG_(r)=inf{G(r1)|r≤r1<b}
are well defined on [a,b), and
limr→b−¯G(r)=¯L\quad and\quadlimr→b−G_(r)=L_
both exist and are finite (Theorem 2.1.11, p. 47). From [eq:9],
|G(r)−G(r1)|=|(G(r)−G(r0))−(G(r1)−G(r0))|≤|G(r)−G(r0)|+|G(r1)−G(r0)|<2ϵ,
so
¯G(r)−G_(r)≤2ϵ,r0≤r,r1<b.
Since ϵ is an arbitrary positive number, this implies that
limr→b−(¯G(r)−G_(r))=0,
so ¯L=L_. Let L=¯L=L_. Since
G_(r)≤G(r)≤¯G(r),
it follows that limr→b−G(r)=L.
We leave the proof of the following theorem to you (Exercise [exer:2]).
[theorem:3] Suppose g is locally integrable on (a,b] and denote
G(r)=∫brg(x)dx,a≤r<b.
Then the improper integral ∫bag(x)dx converges if and only if, for each ϵ>0, there is an r0∈(a,b] such that
|G(r)−G(r1)|<ϵ,a<r,r1≤r0.
To see why we associate Theorems [theorem:2] and [theorem:3] with Cauchy, compare them with Theorem 4.3.5 (p. 204)