8.12: Integration and Differentiation
( \newcommand{\kernel}{\mathrm{null}\,}\)
I. We shall now link RN-derivatives (§11) to those of Chapter 7, §12.
Below, we use the notation of Definition 3 in Chapter 7, §10 and Definition 1 of Chapter 7, §12. (Review them!) In particular,
m:M∗→E∗
is Lebesgue measure in En (presupposed in such terms as "a.e.," etc.); s is an arbitrary set function. For convenience, we set
s′(¯p)=0
and
∫Xfdm=0,
unless defined otherwise; thus s′ and ∫Xf exist always.
We start with several lemmas that go back to Lebesgue.
With the notation of Definition 3 of Chapter 7, §10, the functions
¯Ds,D_s, and s′
are Lebesgue measurable on En for any set function
s:M′→E∗(M′⊇¯K).
- Proof
-
By definition,
¯Ds(¯p)=infrhr(¯p),
where
hr(¯p)=sup{sImI|I∈Kr¯p}
and
Kr¯p={I∈¯K|¯p∈I,dI<1r},r=1,2,….
As is easily seen (verify!),
En(hr>a)=⋃{I∈¯K|a<sImI,dI<1r},a∈E∗.
The right-side union is Lebesgue measurable by Problem 2 in Chapter 7, §10. Thus by Theorem 1 of §2, the function hr is measurable on En for r=1,2,… and so is
¯Ds=infrhr
by Lemma 1 of §2 and Definition 3 in Chapter 7, §10. Similarly for D_s.
Hence by Corollary 1 in §2, the set
A=En(D_s=¯Ds)
is measurable. As s′=¯Ds on A,s′ is measurable on A and also on −A (by convention, s′=0 on −A), hence on all of En.◻
With the same notation, let s:M′→E∗(M′⊇¯K) be a regular measure in En. Let A∈M∗ and B∈M′ with A⊆B, and a∈E1.
If
¯Ds>a on A,
then
a⋅mA≤sB.
- Proof
-
Fix ε>0. By regularity (Definition 4 in Chapter 7, §7), there is an open set G⊇B, with
sB+ε≥sG.
Now let
Kε={I∈¯K|I⊆G,sI≥(a−ε)mI}.
As ¯Ds>a, the definition of ¯Ds implies that Kε is a Vitali covering of A. (Verify!)
Thus Theorem 1 in Chapter 7, §10, yields a disjoint sequence {Ik}⊆Kε, with
m(A−⋃kIk)=0
and
mA≤m(A−⋃Ik)+m⋃Ik=0+m⋃Ik=∑kmIk.
As
⋃Ik⊆G and sB+ε≥sG
(by our choice of Kε and G, we obtain
sB+ε≥s⋃kIk=∑ksIk≥(a−ε)∑kmIk≥(a−ε)mA.
Thus
(a−ε)mA≤sB+ε.
Making ε→0, we obtain the result.◻
If
t=s±u,
with s,t,u:M′→E∗ and M′⊇¯K, and if u is differentiable at a point ¯p∈En, then
¯Dt=¯Ds±u′ and D_t=D_s±u′ at ¯p.
- Proof
-
The proof, from definitions, is left to the reader (Chapter 7, §12, Problem 7).
Any m-continuous measure s:M∗→E1 is strongly regular.
- Proof
-
By Corollary 3 of Chapter 7, §11, vs=s<∞ (s is finite!). Thus vs is certainly m-finite.
Hence by Theorem 2 in Chapter 7, §11, s is absolutely m-continuous. So given ε>0, there is δ>0 such that
(∀X∈M∗|mX<δ)sX<ε.
Now, let A∈M∗. By the strong regularity of Lebesgue measure m (Chapter 7, §8, Theorem 3(b)), there is an open set G⊇A and a closed F⊆A such that
m(A−F)<δ and m(G−A)<δ.
Thus by our choice of δ,
s(A−F)<ε and s(G−A)<ε,
as required.◻
Let s,sk(k=1,2,…) be finite m-continuous measures, with sk↗s or sk↘s on M∗.
If the sk are a.e. differentiable, then
¯Ds=D_s=limk→∞s′k a.e.
- Proof
-
Let first sk↗s. Set
tk=s−sk.
By Corollary 2 in Chapter 7, §11, all tk are m-continuous, hence strongly regular (Lemma 4). Also, tk↘0 (since sk↗s). Hence
tkI≥tk+1I≥0
for each cube I; and the definition of ¯Dtk implies that
¯Dtk≥¯Dtk+1≥D_tk+1≥0.
As {¯Dtk}↓, limk→∞¯Dtk exists (pointwise). Now set
Ar=En(limk→∞¯Dtk≥1r),r=1,2,….
By Lemma 1 (and Lemma 1 in §2), Ar∈M∗. Since
¯Dtk≥limi→∞¯Dti≥1r
on Ar, Lemma 2 yields
1rmAr≤tkAr.
As tk↘0, we have
1rmAr≤limk→∞tkAr=0.
Thus
mAr=0,r=1,2,….
Also, as is easily seen
En(limk→∞¯Dtk>0)=∞⋃r=1En(limk→∞¯Dtk≥1r)=∞⋃r=1Ar
and
m∞⋃r=1Ar=0.
Hence
limk→∞¯Dtk≤0a.e.
As
¯Dtk≥D_tk≥0
(see above), we get
limk→∞¯Dtk=0=limk→∞Dtk a.e. on En.
Now, as tk=s−sk and as the sk are differentiable, Lemma 3 yields
¯Dtk=¯Ds−s′k and D_tk=D_s−s′ka.e.
Thus
limk→∞(¯Ds−s′k)=0=lim(D_s−s′k),
i.e.,
¯Ds=limk→∞s′k=D_sa.e.
This settles the case sk↗s.
In the case sk↘s, one only has to set tk=sk−s and proceed as before. (Verify!)◻
Given A∈M∗,mA<∞, let
s=∫CAdm
on M∗. Then s is a.e. differentiable, and
s′=CA a.e. on En.
(CA= characteristic function of A.)
- Proof
-
First, let A be open and let ¯p∈A.
Then A contains some G¯p(δ) and hence also all cubes I∈¯K with dI<δ and ¯p∈I.
Thus for such I∈¯K,
sI=∫ICAdm=∫I(1)dm=mI;
i.e.,
sImI=1=CA(¯p),¯p∈A.
Hence by Definition 1 of Chapter 7, §12,
s′(¯p)=1=CA(¯p)
if ¯p∈A; i.e., s′=CA on A.
We calim that
¯Ds=s′=0a.e. on −A.
To prove it, note that
s=∫CAdm
is a finite (why?) m-continuous measure on M∗. By Lemma 4, s is strongly regular. Also, as sI≥0 for any I∈¯K, we certainly have
¯Ds≥D_s≥0.
(Why?) Now let
B=En(¯Ds>0)=∞⋃r=1Br,
where
Br=En(¯Ds≥1r),r=1,2,….
We have to show that m(B−A)=0.
Suppose
m(B−A)>0.
Then by (2), we must have m(Br−A)>0 for at least one Br; we fix this Br Also, by (3),
¯Ds≥1r on Br−A
(even on all of Br). Thus by Lemma 2,
0<1rm(Br−A)≤s(Br−A)=∫Br−ACAdm.
But this is impossible. Indeed, as CA=0 on −A (hence on Br−A),the integral in (4) cannot be >0. This refutes the assumption m(B−A)>0; so by (2),
m(En(¯Ds>0)−A)=0;
i.e.,
¯Ds=0=D_s a.e. on −A.
We see that
s′=0=CA a.e. on −A,
and
s′=1=CA on A,
proving the lemma for open sets A.
Now take any A∈M∗,mA<∞. As Lebesgue measure is regular (Chapter 7, §8, Theorem 3(b)), we find for each k∈N an open set Gk⊇A, with
m(Gk−A)<1k and Gk⊇Gk+1.
Let
sk=∫CGkdm.
Then sk↘s on M∗ (see Problem 5 (ii) in §6). Also, by what was shown above, the sk are differentiable, with s′k=CGk a.e.
Hence by Lemma 5,
¯Ds=D_s=limk→∞CGk=CA (a.e.).
The lemma is proved.◻
Let f:En→E∗(Er,Cr) be m-integrable, at least on each cube in En. Then the set function
s=∫fdm
is differentiable, with s′=f, a.e. on En.
Thus s′ is the RN-derivative of s with respect to Lebesgue measure m (Theorem 1 in §11).
- Proof
-
As En is a countable union of cubes (Lemma 2 in Chapter 7, §2), it suffices to show that s′=f a.e. on each open cube J, with s differentiable a.e. on J.
Thus fix such a J≠∅ and restrict s and m to
M0={X∈M∗|X⊆J}.
This does not affect s′ on J; for as J is open, any sequence of cubes
Ik→¯p∈J
terminates inside J anyway.
When so restricted,
s=∫f
is a generalized measure in J; for M0 is a σ-ring (verify!), and f is integrable on J. Also, m is strongly regular, and s is m-continuous.
First, suppose f is M0-simple on J, say,
f=q∑i=1aiCAi,
say, with 0<ai<∞,Ai∈M∗, and
J=q⋃i=1Ai (disjoint).
Then
s=∫f=q∑i=1ai∫CAi.
Hence by Lemma 6 above and by Theorem 1 in Chapter 7, §12, s is differentiable a.e. (as each ∫CAi is), and
s′=q∑i=1ai(∫CAi)′=q∑i=1aiCAi=f (a.e.),
as required.
The general case reduces (via components and the formula f=f+−f−) to the case f≥0, with f measurable (even integrable) on J.
By Problem 6 in §2, then, we have fk↗f for some simple maps fk≥0. Let
sk=∫fk on M0,k=1,2,….
Then all sk and s=∫f are finite measures and sk↗s, by Theorem 4 in §6. Also, by what was shown above, each sk is differentiable a.e. on J, with s′k=fk (a.e.). Thus as in Lemma 5,
¯Ds=D_s=s′=limk→∞s′k=limfk=f (a.e.) on J,
with s′=f≠±∞ (a.e.), as f is integrable on J. Thus all is proved.◻
II. So far we have considered Lebesgue (¯K) differentiation. However, our results easily extend to Ω-differentiation (Definition 2 in Chapter 7, §12).
The proof is even simpler. Thus in Lemma 1, the union in formula (1) is countable (as ¯K is replaced by the countable set family Ω); hence it is μ-measurable. In Lemma 2, the use of the Vitali theorem is replaced by Theorem 3 in Chapter 7, §12. Otherwise, one only has to replace Lebesgue measure m by μ on M. Once the lemmas are established (reread the proofs!), we obtain the following.
Let S,ρ,Ω, and μ:M→E∗ be as in Definition 2 of Chapter 7, §12. Let f:S→E∗(Er,Cr) be mu-integrable on each A∈M with μA<∞.
Then the set function
s=∫fdμ
is Ω-differentiable, with s′=f, (a.e.) on S.
- Proof
-
Recall that S is a countable union of sets Uin∈Ω with 0<μUin<∞. As μ∗ is G-regular, each Uin lies in an open set Jin∈M with
μJin<μUin+εin<∞.
Also, f is μ-measurable (even integrable) on Jin. Dropping a null set, assume that f is M-measurable on J=Jin.
From here, proceed exactly as in Theorem 1, replacing m by μ.◻
Both theorems combined yield the following result.
If s:M′→E∗(Er,Cr) is an m-continuous and m-finite generalized measure in En, then s is ¯K-differentiable a.e. on En, and ds=s′dm (see Definition 3 in §10) in any A∈M∗(mA<∞).
Similarly for Ω-differentiation.
- Proof
-
Given A∈M∗(mA<∞), there is an open set J⊇A such that
mJ<mA+ε<∞.
As before, restrict s and m to
M0={X∈M∗|X⊆J}.
Then by assumption, s is finite and m-continuous on M0 (a σ-ring); so by Theorem 1 in §11,
s=∫fdm
on M0 for some m-integrable map f on J.
Hence by our present Theorem 1, s is differentiable, with s′=f a.e. on J and so
s=∫f=∫s′ on M0.
This implies ds=s′dm in A.
For Ω-differentiation, use Theorem 2.◻
Let s be as in Corollary 1. Subject to Note 1 in §10, if f is s-integrable on A∈M∗(mA<∞), then fs′ is m-integrable on A and
∫Afds=∫Afs′dm.
Similarly for Ω-derivatives, with m replaced by μ.
- Proof
-
By Corollary 1, ds=s′dm in A. Thus Theorem 6 of §10 yields the result.◻
Note 1. In particular, Corollary 2 applies to m-continuous signed LS measures s=sα in E1 (see end of §11). If A=[a,b], then sα is surely finite on sα-measurable subsets of A; so Corollaries 1 and 2 show that
∫Afdsα=∫Afs′αdm=∫Afα′dm,
since s′α=α′. (See Problem 9 in Chapter 7, §12.)
Note 2. Moreover, s=sα (see Note 1) is absolutely m-continuous iff α is absolutely continuous in the stronger sense (Problem 2 in Chapter 4, §8).
Indeed, assuming the latter, fix ε>0 and choose δ as in Definition 3 of Chapter 7, §11. Then if mX<δ, we have
X⊆⋃Ik (disjoint)
for some intervals Ik=(ak,bk], with
δ>∑mIk=∑(bk−ak).
Hence
|sX|≤∑|sIk|<ε.
(Why?) Similarly for the converse.