Measure Theory and Lebesgue Integration T. Muthukumar tmk@iitk.ac.in May 2, 2012iiContents Notations 1 Introduction ll 12 Riemann Integration and its Inadequacy 1.1.1 Limit and Integral: Interchange 1.1.2 Differentiation and Integration: Duality Motivating Lebesgue Integral and Measure 2 Lebesgue Measure on R” 24 2.2 2.3 24 25 26 27 28 Introduction Outer measure Measurable Sets Abstract Measure Theory Invariance of Lebesgue Measure Measurable Functions Littlewood’s Three Principles 2.7.1 First Prineiple 2.7.2 Third Principle 2.7.3 Second Principle Jordan Content or Measure 3 Lebesgue Integration 3.1 32 33 34 35 3.6 Simple Functions Bounded Function With Finite Measure Support Non-negative Functions al Integrable Functions L? Spaces Invariance of Lebesgue Integral woe oa 12 21 31 33 34 43 43 45 48 50 53 33 36 60 65 72 80CONTENTS: 4 Duality of Differentiation and Integration 4.1 Monotone Functions 4.2 Bounded Variation Functions 4.3 Derivative of an Integral 4.4 Absolute Continuity and FTC Appendices A Equivalence Relation B Cantor Set and Cantor Function © Metric Space D AM-GM Inequality Bibliography Index iv 83 84 89 97 101 109 ut 113 uy 119 123 125Notations Symbols 2 will denote the power set, the set of all subsets, of a set L(R") the class of all Lebesgue measurable subsets of IR” denotes the plane of complex numbers Q denotes the set of all rationals in R. Q" set of all vectors in R" with each coordinate being rational number B(R") the class of all Borel measurable subsets of R” B,(x) will denote the closed ball of radius r and centre at x Rg denotes the real line R" denotes the Buclidean space of dimension n B,(x) will denote the open ball of radius r and centre at x E* will denote the sot complement of Bc 8, $\ B Function Spaces R({a,6)) denotes the space of all Riemann integrable functions on the interval (a8) Lip(B) denotes the space of all Lipschitz functions on AC(E) denotes the space of all absolutely continuous functions on B BV (E) denotes the space of all bounded variation functions on ENOTATIONS vi O(X) the class of all real-valued continuous functions on X Co(X) denotes the space of all continuous functions vanishing at oo on X C.AX) denotes the spa x e of all compactly supported continuous functions on, (BE) denotes the space of all measurable p-int egrable functions on E M(B”) the class of all finite a.c. real valued Lebesgue measurable functions on R"Chapter 1 Introduction 1.1 Riemann Integration and its Inadequacy Let f = [a,b] + R be a bounded function. Let P be the partition of the interval [a, 6), a= to <2 <...<_—b. For i= 0,1,2,...,h, let M(P)= sup f(r) and m(P)=__ inf f(x) eles. ele ‘The upper Riemann sum of f with respect to the given partition P is, e U(P, f) =O MCP) (as — 2-2) a and the lower Riemann sum of f with respect to the given partition P is, k LP, f) = Som (PY (a = 21) We of upper sum and supremum of lower sum, over all partitions P of {a, 6], coincide and is denoted as the bounded function f is Riemann integrable on [a, 8] if the infimum, [ f(x) de := inf U(P, f) = sup L(P, f) P P If f =u + iv is a bounded complex-valued function on [a, }], then f is said to be Riemann integrable if its real and imaginary parts are Riemann 1CHAPTER 1. INTRODUCTION 2 [sou fo airs foes Exercise 1. Every Riemann integrable function’ is bounded integral and Let R({a, b)) denote the space of all Riemann integrable functions on [a, The space R((a, b]) forms a vector space over R (or C). It is closed under composition, if it makes sense. Theorem 1.1.1. If f is continuous on [a,b], then f € Ri(a, )) In fact even piecewise continuity is sufficient for Riemann integrability. ‘Theorem 1.1.2. If f is continuous except al finitely many points of [a,b] (piecewise continu- ous), then f € R([a,8)) But there are functions which has discontinuity at countably many points and are still in R((a, 6)) Example 1.1. Consider the function 1 ify <2 Bate) There by Riemann integrable we mean the upper sum and lower sum coincide and are finiteCHAPTER 1, INTRODUCTION 3 where H : R > R is defined as 1 ifx>0 H(x) = = @) {; ifz<0, The function f is discontinuous at all the points r, and can be shown to be in R((0, 1)), because it is bounded and monotone. Theorem 1.1.4. If f € R{la,b)) then f is continuous on a dense subset of (a,b) Example 1.3. An example of a function f : [0,1] + IR which is not Riemann, integrable is, 1 reQ f(x) = 0 £€(0,1]\@ A necessary and sufficient condition of Riemann integrability is given by Theorem 3.0.5. Thus, even to characterise the class of Riemann integrable functions, we need to have the notion of lenght (“measure”) (at least measure 2010) 1.1.1 Limit and Integral: Interchange Let us consider a sequence of functions {f,} C R([a,b]) and define f(a) = limg-sao fe(2), assuming that the limit exists for every x € (a,6]. Does f € Ri((a,b))? The answer is a “no”, as scon in example below. Example 1.4. Fix an enumeration (order) of the set of rationals in (0, 1]. Let the finite set 7, denote the first k elements of the set of rationals in [0,1] Define the sequence of functions fale) = {; ifrery 0 otherwise. Each fx € R{{0, 1)), since it has discontinuity at k (finite) number of points The point-wise limit of fx, f = limysoo fi, is _fi zeQ reo={} re (0,1]\Q which we have seen above is not Riemann integrable.CHAPTER 1. INTRODUCTION 4 Thus, the space R({a, b)) is not “complete” under point-wise limit. How- ever, R((a, 6)) is complete under uniform convergence, A related question is if the limit f © R({a, b)), is the Riemann integral of f the limit of the Riemann integrals of fy, Le., can we say [16 de = jin [never ‘The answer is a “no” again. Example 1.5. Consider the functions fela) = {; x € (0,1/k) 0 otherwise Then f(x) = lim, fx(z) = 0. Note that ff toe)ar= 1 Vk, but fy f(x) dr = 0. The interchange becomes possible under uniform convergence. Theorem 1.1.5. Let {fi} C Rila,b]) and f(z) > f(x) uniformly in ‘a, 6] Then f € R((a, b]) and But uniform convergence is too demanding in practice, The following more general result for interchanging limit and integral will be proved in this write-up. Theorem 1.1.6. Let {f.} © R(\a,6]) and f © Ri\a,b)). Also, let fale) > F(2) point-wise and fi, are uniformly bounded. Then ga fine fr The proof of above theorem is not elementary, thus in classical analysis we always prove the result for uniform convergence. Observe the hypothesis of integrability on f in the above theoremCHAPTER 1, INTRODUCTION 1.1.2 Differentiation and Integration: Duality An observation we make, once we have Riemann integration, is about the dual nature of differentiation and integration. Thus, one asks the following ‘two questions: 1. (Derivative of an integral) For which class of functions can we say Ef oa- sey Af toa re 2. (Integral of a derivative) For which class of functions can we say [ ferae= 10) — faye To answer the fizst question, for any f ¢ R{(a,6]), let us define the function F(2) [ f(tat. Exercise 2. Show that if f € R((a,6]) then F is continuous on (a, 6) The first question is answered by the following result of Riemann inte- gration Theorem 1.1.7. Let f € R({a,b]). If f is continuous at a point x € [a,b], then F is differentiable at x and F'(x) — f(x) What is the most general class of functions for which the above result holds true. The second question is answered by the famous Fundamental theorem of calculus (FTC). Theorem 1.1.8 (Fundamental Theorem of Calculus). If f is differentiable function on [a, 6] such that f" € R([a,b]), then * Ha) de F(b) ~ Fa) Note that the fundamental theorem of calculus fails under the following two circumstancesCHAPTER 1. INTRODUCTION 6 1, For a continuous function f on [a,6] which is nowhere differentiable on a, 6). Do such functions exist? 2, Derivative of f exists for all points in a, 6), but f” is not integrable. Do such functions exist? K. Weierstrass was the first to show in 1872 the existence of a everywhere continuous function which is nowhere differentiable. Prior to Weierstrass’ proof it was believed that every continuous function is differentiable except on a set of “isolated” points. This example of Weierstrass showed the existence of function for which FTC may not make any sense. The existence of a function f whose derivative exists everywhere but the derivative is not integrable, was shown by Vito Volterra, who was a student of Ulisse Dini, in 1881. His example was a clever modification of the function 5 {rae 2#0 o(z) * 0 2=0 which is differentiable. The derivative of g, g/(x) = 2xsin(1/r) — cos(1/), is discontinuous at 0. A natural question to ask was: Identify the class of functions for which FTC makes sense 1.2 Motivating Lebesgue Integral and Mea- sure Whatever the reasons are, we should be convinced now that it is worthwhile looking for a new type of integration which coincides for Riemann integrable functions and also includes “non-integrable” (Riemann) functions The Riemann integration was based on the simple fact that one can in- tegrate step functions (piecewise constant) and then approximate any given function with piecewise constant functions, by partitioning the domain of the function. Lebesgue came up with this idea of partitioning the range of the function A very good analogy to motivate Lebesgue integration is the following (cf. [Pug04]): Suppose A asks both B and C to give the total value for a bunch of coins with all denominations lying on a table. First B counts themCHAPTER 1, INTRODUCTION as he picks the coins and adds their denomination to come up with the total value. This is Riemann’s way of integration (partitioning the domain, if you consider the function to be coin mapped to its denomination). In his/her turn, C sorts the coin as per their denominations in to separate piles and counts the coins in each pile, multiply it with the denomination of the pile and sum them up for the total value. Both B and C will come up with the same value (assuming they counted right!). The way C counted is Lebesgue’s way of integration, We know that integration is related with the question of computing length/area/volume of a subset of an Euclidean space, depending on its di- mension, Now, if one wants to partition the range of a function, we need some way of “measuring” how much of the domain is sent to a particular region of the partition, This problem leads us to the theory of measures where we try to give a notion of “measure” to subsets of an Euclidean space.CHAPTER 1 INTRODUCTIONChapter 2 Lebesgue Measure on R” 2.1 Introduction In this chapter we shall develop the notion of Lebegue ‘measure’ in R" Definition 2.1.1. We say R is a cell (open) in RY af R is of the form (3,81) X (a2sb2) x =. (Qs by), He, {= (#1,-.-2tm) © R" | 25, € (a,b,) for all 1 0, |AR| = |R|, where AR = {(Aumi) | (a) € RhCHAPTER 2. LEBESGUE MEASURE ON R” 10 Exercise 4 (uniqueness of volume). Let C be the collection of all cells of R" and if v : C + (0, +00] is a well defined set-function on C such that v is invariant under translation and dilation. Show that v is same as the volume |-| up to a constant, ie., there exists a constant a > 0 such that (2) = a|R| for all cells R € C. In particular, if we additionally impose the condition that (0, 1") = 1 then v(R) = |R| for all REC. Erercise 5. Show that the volume satisfies monotonicity, ic., if RC Q, then IR| $|@| Definition 2.1.2. We say a collection of cells R, to be almost disjoint if the interiors of R, are pairwise disjoint Exercise 6. Ifa cell R = Ub, then |R| = 554 [Ril Exercise 7. If {R,}f ave cells in R" such that R C Uf,R,, then |R| < f NI 1R, such that R, are pairwise almost disjoint, Theorem 2.1.3. For every open subset © CR, there exists a unique count- able family of open intervals I, such that 9 = disjoint. 2h, where [,'s are pairwise Proof. Since {is open, for every x € 0, there is an open interval in © that contains . Let us pick the largest such open interval in @ that contains 2. How do we do this? Let, for each x € 9, aes oy = inf{(a,2) CO} and b, := sup{(x,6) CM} ibe Of course, a, and b, can take -Loo. Note that a, <# 2 Exercise 8. An open connected set 9 C R", n > 2 is the disjoint union of open cells iff (is itself an open cells, Exercise 9, Show that an open disc in IR? is not the disjoint union of open rectangles. However, relaxing our requirement to almost disjoint-ness will generalise Theorem 2.1.3 to higher dimensions R", n > 2. Theorem 2.1.4. For every open subset 2 CR”, there exists a countable family of almost disjoint closed cells R, such that 9 = UR R, Proof. ‘To begin we consider the grid of cells in R” of side length 1 and whose vertices have integer coordinates, The number cells in the grid is countable and they are almost disjoint. We ignore all those cells which are contained in 0°. Now we have two families of cells, those which are contained in Q, call the collection C, and those which intersect both 2 and 2°. We bisect the latter cells further in to 2* cells of side each 1/2. Again ignore those contained in 9° and add those contained in 2 to the collection C. Further bisecting the common cells in to cells of side length 1/4. Repeating this procedure, we have a countable collection C of almost disjoint cells in . By construction, UpecR C9. Let x € 9 then there is a cell of side length 1/2* {aficr bisecting k times) in C which contains x. Thus, UrecR = 9. a Again, as we did in one dimension, we hope to define the “volume” of an open subset 2 C R” as the sum of the volumes of the cells R obtained in above theorem. However, since the collection of cells is not unique, in contrast to one dimension, it is not clear if the sum of the volumes is independent of the choice of your family of cells We wish to extend the notion of volume to arbitrary subsets of an Eu- clidean space such that they coincide with the usual notion of volume for a cell, most importantly, preserving the properties of the volume. So, what are these properties of volume we wish to preserve? To state them, let's first regard the volume as a set function on the power set of R", mapping to a non-negative real number. Thus, we wish to construct a ‘measure’ 1, 4: 2" — [0,00] such that 1, If Ris any cell of RY, then u(R) = [R)CHAPTER 2. LEBESGUE MEASURE ON R™ 12 2. (Translation Invariance) For every # CR, (H+ 2) = u(E) for all x € RP. 3. (Monotonicity) If Bc F, then (2) < p(F) 4, (Countable Sub-additivity) If # = UP, B, then y(#) < 0%, u()) 5. (Countable Additivity) If B = U; disjoint then :() = D2, (Bi) 1; such that Ej are pairwise Exercise 10. Show that if p obeys finite additivity and is non-negative, then 1 is monotone. (Basically monotonicity is redundant from countable addi- tivity) 2.2 Outer measure To construct a ‘measure’ on the power set of R”, we use the simple approach of ‘covering’ an arbitrary subset of R” by cells and assigning a unique number using them. Definition 2.2.1. Let B CR", a subset of R". We say that a family of cells {Ri}er is a cover of B iff EC UserR;. If each of the cell R, in the cover ts an open (resp. closed) cell, then the cover is said to be open (resp. closed) cover of B. If the index set I is finite/countable/uncountable, then the cover is said to be a finite/countable/uncountable cover. ‘We shall not consider the case of uncountable cover in this text, because uncountable additivity makes no sense. Exercise 11. Every subset of IR” admits a countable covering! If we wish to associate a unique positive number to E € 22", satisfying monotonicity and (finite/countable) sub-additivity, then the association must satisfy HUE) < p(UserR,) (due to monotonicity) SHR) (due to finite/countable sub-additivity) ‘I AA SO IA| (measure same as volume). ‘aCHAPTER 2, LEBESGUE MEASURE ON 8” 13 ‘The case when the index set / is strictly finite corresponds to Riemann integration which we wish to generalise. Thus, we let [ to be a countable index set, henceforth. AA brief note on the case when index set J is strictly finite is given in § 2.8, Definition 2.2.2. For a subset E of R", we define its Lebesgue outer mea- sure’ p*(E) as, HE) = int, IRI, eter the infimum being taken over all possible countable coverings of B. The Lebesgue outer measure is a well-defined non-negative set function on the power set of R", 2°" Exercise 12. The outer measure is unchanging if we restrict ourselves to open covering or closed covering, ie., for every subset E CR”, H(E) = gcint, SIS where the infimum is taken over all possible countable closed or open cover- ings {Si} of E Before we see some examples for calculating outer measures of subsets of B®, lot us observe some immediate properties of outer measure following from definition. Lemma 2.2.3. The outer measure ji" has the following properties (a) For every subset E CR, 0 < p"(B) < +90 (6) (Translation Invariance) For every EC 8", "(E+ all x ER” wn(B) for (c) (Monotone) If F< F, then p*(B) < u°(F) (d) (Countable Sub-additivity) If E URE; then w(B) < So (Bi) TWhy wo call # “outer” and superscript with a + will be clear in the next seetionCHAPTER 2, LEBESGUE MEASURE ON 3 ua Proof. (a) The non-negativity of the outer measure is an obvious conse- quence of the non-negativity of | (b) The invariance under translation is obvious too, by noting that for each covering {R,} of EB or E +x, {R; + x} and {Rj — x} is a covering of E +x and E, respectively, and the volumes of the cell is invariant under translation. (c) Monotonicity is obvious, by noting the fact that, the family of covering of F is a sub-family of the coverings of E. Thus, the infimum over the family of cover for F is smaller than the sub-family. (a) If p*(B,) = +00, for some i, then the result is trivially true. Thus, we assume that je"(E,) < +00, for all i. By the definition of outer measure, for each ¢ > 0, there is a covering by cells {R}}®, for E; such that S vi 1 £ DIRS) + Since H = USE), the family {Ri}, is a covering for 2. Thus, MES IB SD (WE) + 5) = eB) +e Since choice of ¢ is arbitrary, we have the countable sub- ditivity of pt ao ‘We have seen the properties of outer measure. Let us now compute the outer measure for some subsets of R” Example 2.1. Outer measure of the empty set is zero, *(0) = 0. Every cell is a cover for the empty set. Thus, infimum over the volume of all cells is zero. Example 2.2. The outer measure for a singleton set {x} in RP is zero. The same argument as for empty set holds except that now the infimum is taken over all cells containing x. Thus, for each ¢ > 0, one can find a cell R, such that x € R, and |R,| < ¢. Therefore, *({r}) < for all ¢ > 0 and hence w({z}) = 0.CHAPTER 2, LEBESGUE MEASURE ON 8” 15, Example 2.3. The onter measure of any countable subset. # of IR” is zero. A countable set E = Usee{x}, where the union is countable. Thus, by countable sub-additivity, *(Z) < 0 and hence p"(#) = 0. Let’s highlight something interesting at this stage. Note that the set of all rationals, Q, in R is countable. Hence y:*(Q) = 0. Also Q is dense in R. Thus, we actually have a dense (‘scattered’) subset of R whose outer measure is zero (‘small’). Example 2.4. The situation is even worse. The converse of above example is not true, ie., we can have a uncountable set whose outer measure is zero. The outer measure of R"~! (lower dimensional), for n > 2, as a subset of R" is zero, ie., u*(R"!) = 0. Choose a cover {Rj} of R™- in R"? such that [Rina Then E; = R, x (3£,£) forms a cover for R"-! in R*. Thus, a wR) < Since the choice of ¢ > 0 could be as small as possible, we have j."(I""*) = 0, Example 2.5. Is there an uncountable subset of K. whose outer measure zero? Consider the Cantor set C’ (cf. Appendix B) which is uncountable. Let us compute the outer measure of C’. Recall that, for each i, C; is disjoint union of 2 closed intervals, each of whose length is 3-‘. Thus, y*(C,) = (2/3)', for all i. By construction, CC C, and hence due to monotonicity p*(C) < 4(Ci) = (2/3), for all i, But (2/3) 280. Thus, w*(C) = 0. Example 2.6. A similar argument as above shows that the outer measure of ‘the generalised Cantor set C (ef. Appendix B) is bounded above by w(C) Shim Parag... ay, We shall, in fact, show that equality holds here using “continuity from above” of outer measure (cf. Example 2.12) Example 2.7. If R is any cell of R*, then y*(R) = |R|, Since R is a cover by itself, we have from definition, /*(R) < |R|. It now remains to prove the reverse inequality. Let $'be a closed cell. Let (R,}}” be an arbitrary covering of S. Choose an arbitrary © > 0. For each ¢, we choose an open cell Q, such that R, CQ, and |Q,| < || +¢|Rj]. Note that {Q,} is an open covering ofCHAPTER 2. LEBESGUE MEASURE ON R” 16. S. Since $ is compact in IR” (closed and bounded), we can extract a finite sub-cover such that Sc UE ,Q;. Therefore, k [S| < S31Q) (ef. Exerc a k G+) SIRI a Since © can be chosen as small as possible, we have IA ‘ Is}< SRI < Sail ‘al Since {R,} was an arbitrary choice of cover for S, taking infimum, we get |S| < u*(S). Thus, for a closed cell we have shown that |S| = p“(S). Now, for the given cell R and ¢ > 0, one can always choose a closed cell $C R such that |R| < |S| +2. Thus, |R| < |S|+e=p'(S)+e 0 is arbitrary, |R) < y*(R) and hence |R| = p*(R), for any cell R. Example 2.8. The outer measure of R" is infinite, 4*(R") = too. For any M > 0, every cell R of volume M is a subset of R". Hence, by monotonicity of wt, °(R) < w*(R"). But p*(R) = |R| = M. Thus, p*(R") > M for all M > 0. Thus, u*(R") = +00. Brercise 13. Show that R is uncountable. Also show that the outer measure of the sct of inrationals in R is +00. Example 2.9. Does the outer measure of other basic subsets, such as balls (or spheres), polygons ete. coincide with their volume, which we know from geometry (calculus)? We shall postpone answering this, in a simple way, till we develop sufficient tools. However, we shall note the fact that for any non-empty open set {2 C R®, its outer measure is non-zero, ie., 4*(Q) > 0. This is because for every non-empty open set , one can always find a cell RCM such that |R| > 0. Thus, 2*(0) > [RI > 0. Exercise 14. If ECR” has a positive outer measure, “(E) > 0, does there always exist a cell RC E such that [R| > 0.CHAPTER 2, LEBESGUE MEASURE ON iv Theorem 2.2.4 (Outer Regularity). If EB CR" is a subset of R", then 10(B) = infose u*(M), where 9's are open sets containing E. Proof. By monotonicity, 4*(B) < y*(Q) for all open sets © containing E. Thus, "() < inf y*(Q). Conversely, for each ¢ > 0, we can choose a cover of cells {R,} of E such that Vial swe) +5 For cach Rj, choose an open cell Q; > R; such that IQ: <|Ril + pase Since each Q; is open and countable union of open sets is open, 2 = UR, is open. Therefore, by sub-additivity, HO) SIA Ss VIR +5 se) +e Thus, we have equality. a Recall that arbitrary union (intersection) of open (closed) sets is open (closed) and finite intersection (union) of open (closed) sets is open (closed). This motivates us to define the notion of Gs and F, subset of R* Definition 2.2.5. A subset E is said to be Gy? if it is a countable intersection of open sets in R". We say E is F,5 if it is a countable union of closed sets ink Corollary 2.2.6. For every subset EC RM there exists a Gs subset G of such that G > E ond p"(B) = u"(G). Proof. Using Theorem 2.2.4, we have that for every k € N, there is an open. set 4, > E such that meaning territory, and average or mean, respectively terminology comes from French word “fermé” and “somme” meaning closed and sum, respectivelyCHAPTER 2, LEBESGUE MEASURE ON 3 18 Let G:= Nf. Thus, Gis a Gs set. G is non-empty because B.C G and hence j"(E) < u"(G). For the reverse inequality, we note that GCM, for all k, and by monotonicity ah(G) 0, then p(B UF) = p(B) + u°(F) Proof. By the countable sub-additivity of yw", we have p"(BU FP) < p*(B) + 1°(F). Once we show the reverse inequality, we are done. For cach ¢ > 0, we can choose a covering {Rj} of EU F such that Dal s(BUF) +e a The family of cells {R,} can be categorised in to three groups: those inter- secting only E, those intersecting only F and those intersecting both B and F. Note that the third category, cells intersecting both B and F, should have diameter bigger than d(, F). Thus, by subdividing these cells to have diameter less than d(, F), we can have the family of open cover to consist of only those cells which either intersect with F or F. Let I, = {i : ROE #0} and I; = {i : R,OF # 9}. Due to our subdivision, we have I) In = 0. Thus, {R,} for i ¢ Jy is an open cover for # and {R,} for i € Ip is an open cover for F. Thus, W(E) +H (P) SOIR + DIR) S IRI Se (BUF) +5 i ih th = By the arbitrariness of ¢, we have the reverse inequality. oOCHAPTER 2, LEBESGUE MEASURE ON 19. Proposition 2.2.8. If a subset EC R" is a countable union of almost disjoint closed cells, i.e., B= UR1R, then (8) = SRI Proof. By countable sub-additivity, we already have yi*(E) < O%, |Ril. It only remains to prove the reverse inequality. For each ¢ > 0, let Q, CR, be a cell such that |R,| < |Q\| | ¢/2'. By construction, the eclls Q; are pairwise disjoint and d(Q,,Q;) > 0, for all i # j. Applying Proposition 2.2.7 finite number times, we have k #° (UELQ) = OQ] for each KEN. oa Since UL,Q; C E, by monotonicity, we have k c (BE) > et (UEIQ) = SO1@d = SO (VR - €/2") a a By letting k 00, we deduce VIR s H(A) +e Since ¢ can be made arbitrarily small, we have equality. oO A consequence above proposition and Theorem 2.1.4 is that for an open set 9, #(2) =O |R| irrespective of the choice of the almost disjoint closed cells whose union is 92. We now show that j* is not countably additive. In fact, it is not even. finitely additive. One should observe that finite additivity of j* is quite dif ferent from the result proved in Proposition 2.2.7. To show the non-additivity (finite) of p*, we need to find two disjoint sets, sum of whose outer measure is not the same as the outer measure of their union* This constr elated to the Banach-Tarski paradox which states that one ean partition the unit ball in R® in to a finite number of pieces which can be reassembled (after rotation and translation) to form two complete unit balls!CHAPTER 2. LEBESGUE MEASURE ON R” 20 Proposition 2.2.9. There exists a countable family {Nj}? of disjoint sub- sets of R® such that AV Proof. Consider the unit cube (0, 1)" in R”. We define an equivalence relation ~ (cf. Appendix A) on [0,1)" as, © ~ y whenever x —y € Q”. This equivalence relation will partition the cube [0, 1|* in to disjoint equivalence classes £4, [0,1)" = Ug. Now, lot N be the subset of [0,1)" which is formed by picking? exactly one element from each equivalence &,. Since Q" is countable, let {r;}° be the enumeration of all elements of Q". Let N, V +r; We first show that N,’s are all pairwise disjoint. Suppose ON; # @, then there exist 24,23 € N such that x4 +1) = 23 +7). Hence 2q— tq =1;—", € Q”. This implies that rq ~ xy which contradicts that fact that N contains exactly one representative from each equivalence class. Thus, N;'s are all disjoint. We now show that U®,.N; = R". It is obvious that URN; CRM. To show the reverse inclusion, we consider x € R". Then there isarg € Q” such that x € [0,1]" +7. Hence x ~ry € [0,1]. Thus x belongs to some equivalence class, i.e., there is a rq € N such that x—r ~ ra Therefore, x € Ng. Hence UPN; = R. Using the sub-additivity and translation-invariance of i", we have ) we (ue so =H) < De) =D) Thus, 1°(N) £0, ic, wr(N) > 0. Let J = {i| 7, € Q" (0, 1]"}. Note that J is countable. Now, consider the sub-collection {N;}jes and let F = UsesN,. Being a sub-collection of N's, N's are disjoint. By construction, cach NV; ¢ [0,2)", and hence F c (0,2)". By monotonicity and volume of cell, "(F) < 2". If countable additivity holds true for the sub-collection 1V,, then WP) = De) = Dew). se to 4*(F) is either zero or +00. But we have alteady shown that y2"(V) # 0 and hence y*(F) #0. Thus, p*(F°) = +00, which contradicts 4“(F) < 2". Thus, countable-additivity for N, cannot hold true. a "Possible due to Axiom of ChoiceCHAPTER 2, LEBESGUE MEASURE ON 21 ‘The clever construction of the set Vital. in the above proof is due to Giuseppe Corollary 2.2.10. There exists a finite family {N,}{ of disjoint subsets of RP such that ‘ Mi) AD eM) Proof. ‘The proof is ditto till proving the fact that p*(N) > 0. Now, it is always possible to choose a k € N such that ky"(N) > 2". Then, we pick exactly elements from the set J and for F to he the finite (k) union of {N,}f. Now, arguing as above assuming finite-additivity will contradict the fact that ku(N) > 2. ao The fact that the outer measure j.* is not countably (even finitely) addi- tivo® is a bad news and leaves our job of generalising the notion of volume, for all subsets of R", incomplete. 2.3. Measurable Sets The countable non-additivity of the outer measure, j:", has left our job in- complete. A simple way out of this would be to consider only those subsets of R” for which 4" is countably additive. ‘That is, we no longer work in the power set 2®", but relax ourselves to a subclass of 28" for which count- able additivity holds. For this sub-class of subsets, the notion of volume is generalised and hence we shall call this sub-class, as a collection of ‘measur- able’ sets and the outer measure j* restricted to this sub-class will be called ‘measure’ of the set. However, the problem is: how to identify this sub-class? Recall from ‘Theorem 2.2.4 that for any subsct EC IR” and for cach © > 0, there is an open sot > E, containing #, such that "(9) < p°(B) + or w"(Q)— p(B) <= Since 2 = BUM\ £ is a disjoint union, by demanding countable additivity, wwe expect to have 4"() — w"() > u"(@\ £)". Thus, we need to choose those subsets of R" for which *() — y"(E) > y*(O\ B). “However, tis possible to got a finitely additive set function on 2” which coincides with jo" on £(R") 7 The other inequality being true due to sub-additivityCHAPTER 2. LEBESGUE MEASURE ON R" Definition 2.3.1. We say a subset EF C R" is measurable (Lebesgue), if for any ¢ > 0 there exists an open set 2 > EB, containing B, such that w(Q\ B) < ¢. Further, we define the measure (Lebesgue), 4, of Eas H(B) = p(B). Let £(R") denote the class of all subsets of R” which are Lebesgue mea- surable. Thus, £(R") ¢ 2°”. The domain of outer measure j*, is 2°, whereas the domain for the Lebesgue measure, 4, is £(R"). The inclusion L£(R") 2%" is proper (cf, Exercise 21). By definition, the Lebesgue measure, j« will inherit all the properties of the outer measure 47". We shall see some examples of Lebesgue measurable subsets of R". Example 2.10. It is easy to see that every open set in R” belongs to £(R") Thus, 0,R", open cells etc, are all in £(R") Example 2.11. Every subset E of R" such that p*(E) = 0 is in £(R") By Theorem 2.24, for any ¢ > 0, there is an open set 2 D E such that 10) < w(E) + Since 2\ E C9, by monotonicity of pi", we have i(O\E) 0, there is an open set 1; > Ej such that - MO%\E) So HONE) S 5 Let 2 = U2), then F CQ and 2 is open. But @\ EC UR, (O,\ BE). By monotonicity and sub-additivity of *, H(Q\ B) < pt (UBM \ Bi) < Sw; \ By) Se. Thus, E is measurable. a Definition 2.3.3. We say E CR” is bounded if there is a cell R CR" of finite volume such that EC R.CHAPTER 2, LEBESGUE MEASURE ON R* 23. Exercise 16. If B is bounded then show that :*(B) < +00. Also, give an example of a set with *(E) < +00 but # is unbounded. Proposition 2.3.4. Compact subsets of R” are measurable Proof. Let F be a compact subset of R". Thus, u*(F) < +90. By Theo- rem 2.2.4, we have, for each ¢ > 0, an open subset > F such that w(O) < WP) +e If we show w"(@\ F) <¢, we are done, Observe that \ F is an open set (since F is closed) and hence, by Theorem 2.1.4, there exists almost disjoint closed cells R, such that Q\F=URR For a fixed k € N, consider the finite union of the closed eells A == UL Note that K is compact. Also KF = @ and thus, by Lemma C.0.20, a(F,K) >0. But KUF ¢. Thus, (2) > w(K UF) (Monotonicity) = w(K) +p°(F) (By Proposition 2.2.7) = WUE) +e (F) k SIRI + e°(F) (By Proposition 2.2.8) Hence, 7h, |Fi| < u*(O) — "(F) < ©. Since this is true for every k € N, by taking limit we have °°, |R,| <¢. Using Proposition 2.2.8 again, we have H(Q\ F) = SOIR se a Thus, F is measurable. a Corollary 2.3.5. Closed subsets of R" are measurable. Consequently, any F, set is measurable. Proof. Let T be a closed subset of R*. If P were bounded, we know it is compact and hence measurable from the above proposition. We need to check only for unbounded set I. Let F) =I B,(0), for i = 1,2,.... Note that each K, is compact and ’ = UR;i. Since each K, is measurable, using Theorem 2.3.2, we deduce that I is measurable. Any F, set is a countable union of closed sets and hence is measurable. 0CHAPTER 2. LEBESGUE MEASURE ON R” 24 Theorem 2.3.6. If E € £(IR”) then B° € £(R") Proof. Since F € £(R"), for each k € N, there exists an open set , > E such that 14(% \ B) < 1/k. Since Of is closed, it is in L(R"). Set F = UP, Note that F is a F, set and hence measurable. Since %; C E* for every k, we have FC E*. Also B°\ FC %\E, for all k € N. By monotonicity, W(E*\ F) < L/h, for all k © N. Therefore, p*(E*\ F) = 0 and hence is measurable. Now B° = (B°\\ F) UF is a union of two measurable sets and hence is measurable, a Corollary 2.3.7. If {E,}{° is a countable family m L(R"), then B= 02, BE; is in L(R"), ie., countable intersection of measurable sets is measurable. Consequently, any Gs set is measurable Proof. Note that B= 0%), = (URES. a Exercise 17. If B, F € £(R*) thon show that B\ F € LOR") Theorem 2.3.8 (Borel Regularity). For any subset E CRM, the following are equivalent: () BE LR’) (ii) For each € > 0, there is an open set 01> E such that p(Q\ B) 0, there is a closed set TC E such that p(B \T) 0, there is an open set > E such that wQ\B) Se. Since @\ E = 0 B*, it is measurable (intersection of measurable sets) ‘Thus, p°(0\ #) = W(0\ B) 0, there is an open set > B® such that H(O\ EY) SeCHAPTER 2, LEBESGUE MEASURE ON R” 25 Set [':= 0°. Then Pc EB. Note that H\T =0\ B*. Hence pH \T) 0 for alls # j. Therefore, using Proposition 2.2.7, we have for every KEN, H (VELA) = Sate Fc E, by monotonicity, we have ‘ HE) > a (UI = Doe = («4 - 5) Since the above inequality is true for every k and arbitrarily small c, we get M(B) > %, w(E,) and hence equality. Let E, be unbounded for some or all i. Consider a sequence of cells {R,}P such that Ry C Ryr, for all j = 1,2,..., and UR,R, = BY. Set Qi = Ry and Q, = Ry \ Ry. for all j > 2 Consider the measurable subsets i,j = E,Q,. Note the each E,, is pairwise disjoint and are each bounded. Observe that Ej ,E,5 and is a disjoint union. Therefore, H(B,) = D2, w( Bis). Also, E='U; Uy Bi, and is a disjoint union. Hence, Exercise 19. If E © F and y(E) < +00, then show that y(F \ E) = p(F) — (EB). Corollary 2.3.10. Let Ey, E>,... be measurable subsets of RB () (Continuity from below) If Ey © Fy ©... and B = us HE) = limgsoo w(x): (ii) (Continuity from above) If Ey > By 2 ..., 4(E,) < +00, for some i, and B=, K,, then p(B) = limy yoo pt( Bx). Proof. (i) Let Fi = Ey and F := EB; \ Ey-1, for i > 2. By construc tion, F, are measurable, disjoint and E = U®, F,. Hence, by countable additivity, Ej, then ME) = SO eR) = jim 2 Smn ") = im ps (Ui af) = Jim #( Fe). iarCHAPTER 2, LEBESGUE MEASURE ON a7 (ii) Without loss of generality, we assume that (21) < 0. Set Fy = Ey \ iss, for each i > 1. Note that By = BU (U%,) is a disjoint ‘union of measurable sets. Hence, ME) = wh) + SWF) = wl) + SWE) ~ plies) ba () + Jima SHC) — e(Biya)) w(E) + p(B) — Jim p(B) Fim w(Ex) = (EB) a Remark 2.3.11. Observe that for continuity from above, the assumption (E,) < +00 is very crucial. For instance, consider B, = (i,00) CR. Note that each 4(,) = +90 but p(B) = 0. Example 2.12. Recall that in Example 2.6 we proved an inequality regarding the generalised Cantor set C. We now have enough tools to show the equality. Note that each measurable Cis satisfy the hypothesis of continuity from above and hence C is measurable and w(C) = lim ayaa. ag In view of this example and Proposition B.0.18, we have generalised Cantor sot whose outer measure is positive Exercise 20 (Continuity from below for outer measure). Let Ei, H2,... be subsets’ of R” such that Ey C Ey ©... and E = Um)E;, then w"(E) = Timy- soo *(E) Proof. By monotonicity of outer measure, we get limy sa 4*(Ex) < 4¢(B). It only remains to prove the reverse inequality. Let Gy be a Gy set such that Gy 2 Ey, for all k, and w*(E,) = 4(G,). The {G,} may not be an increasing sequence. Hence, we set Fi, = oxG, and {F;,} is an increasing sequence of measurable sets. Also, Ek © Fy, since Ek CG; for all j > k "Not necessary measurableCHAPTER 2. LEBESGUE MEASURE ON 28 Moreover, since Ek © Fi, © Gy, we have 4i*(E) < u(Fe) < (Gx) and hence 1(Ex) = u(Fe) = (Ge). Since BC Uf Fe w(E) S MURA Fi) = Jim w(F,) = im p(B) a Recall that we showed the inner regularity of jz in Theorem 2.3.8. We can, in fact, better this for sets with finite measure, Corollary 2.3.12. If u(B) < +00 then there exists a compact set K CE such that p(E\ K) < Proof. We have, using (iii) of Theorem 2.3.8, that a closed set I’ < E such that 4(E\T) B\ Ky >... and E\T =08,(6\K,). Using, continuity from above and (12) < +00, we get © 2 w(E\T) = lim w(B\ Ki), Thus, for i large enough, we have u(E \ K,) Fy... and BE = M2,F,. Let Dy #(4) = L. By countable additivity, (A) < OX, (Bi) <0. By continuity from above, we have kt H(Z) = Jim (Fi) = Jim (UE, £,) S jim, Sue) = jm(e- YH). Thus, 1(E a The sot B in First Borel-Cantelli lemma is precisely the set of all x € R" such that x € E; for infinitely many 7. Let x € R” be such that 2 € By only for finitely many i. Arrange the indices i in increasing order, for which x © B, and let K be the maximum of the indices. Then, x ¢ F) for all j > K +1 and hence not in B. Conversely, if ¢ B, then there exists a j such that x ¢ Fi, for all k > j. Thus, either 2 € U8, or x ¢ H, for all i.CHAPTER 2, LEBESGUE MEASURE ON 29 Exercise 21. Show that the set NV constructed in Proposition 2.2.9 is not in £(R"). Thus, £(R") c 2°” is a strict inclusion. Exercise 22. Consider the non-measurable set N constructed in Proposi- tion 2.2.9, Show that every measurable subset E © N is of zero measure, Proof. Let EC N be a measurable set such that u(#) > 0, then for each 7, €QN(0,1], we set Ey -= B +r, and N,:= N +7,. Since N's are disjoint, E's are disjoint and are measurable. Since UX, ¢ [0,2], we have 22 MUR BE) = D> M(B) = Yo ulB) = 00. A contradiction due to the assumption that (HZ) > 0. Hence (Z)=0. Exercise 23. If B € £(R") such that p() > 0 then show that H has a non-measurable subset. Proof. We first show that every measurable set HC [0,1] such that (2) > 0 has a non-measurable subset. Consider the non-measurable subset NV of (0, 1} For each r, € Q, we set N, = N +r; and each of them are non-measurable. Also, we know that R = UR,N,. Let FC 0,1] be a measurable subset such that (EZ) > 0. Set B, = BON, Note that U2,B, = BN (URN) = EOR = E. If B, were measurable, for each i, then being a subset of N,, 4(E,) = 0. Thus, a contradiction, Thus, our assumption that E, are measurable is incorrect. Thus, £,’s are non-measurable subsets of E. Now, let E C R be any measurable subset such that 4{E) > 0. Note that E = User (EM[i,i + 1), where [i + 1) are disjoint measurable subsets of R. Hence, 0< ME) = OMEN e+) Thus, for some i, “(#1 [é,i+1)) > 0. For this i, set F = EN |i,i+1). Then F ~i c [0,1] which has positive measure and by earlier argument contains a non-measurable set M. Thus, M +7 F C E£ is non-measurable. aCHAPTER 2. LEBESGUE MEASURE ON R" 30 Exercise 24, Construct an example of a continuous function that maps a measurable (Lebesgue) set to a non-measurable set. Definition 2.3.14. We say a subset B C R® satisfies the Carathéodory criterion if w(S) = W(SNE) +e (SOE) forall subsets SCR” Equivalently, #(S) Note that since have w(SME)+°(S\ BE) for all subsets SCR” SAE)U(SME), by sub-additivity of w", we always (8) wh(SAB) + H(S\B)CHAPTER 2, LEBESGUE MEASURE ON 31 where the last inequality is due to monotonicity. Hence one way implication is proved. Conversely, let EC RR" satisfy the Carathéodory criterion. We need to show E € £(R"). To avoid working with oo, we assume K to be such that 1(E) < +00. We know, by outer regularity, that for each ¢ > 0 there is an open set 9 > E such that p*(M) < y*(E) +e. But, by Carathéodory criterion, we have H(O) = w(QN B) + (ON B) = WE) + (YB) Thus, B(O\ B) = (2) — WE) Se Hence, £ is measurable. It now only remains to prove for £ such that w(B) = +00. Let B= EM B,(0), for i = 1,2, Note that each i(E,) < too is bounded and E = Uf, E). Since using Theorem 2.3.2, we deduce that E is measurable. oa ach E, is measurable, 2.4 Abstract Measure Theory Let X be a non-empty set and 2* is the collection of all subsets of X. Definition 2.4.1. A set function p* : 2% + [0, co] és called @ outer measure on X if ) w= (i) FE CU then pt(E) < D2, u'(E) Exercise 25. Show that any outer measure is monotone?’ Definition 2.4.2. We say a subset EC X is measurable if 1(S) = (SOE) +ue(SME®) for all subsets Sc X. While classifying Lebesgue measurable sets, we noted some of their prop- erties. These properties can be summarised as follows *T is depende outer measure, (ji) is such that E = U%,A; then monotonicity does not follow from i on how one defines sub-additivity. If in the definition of definition. In that case monotonicity should be included in the definition of outer measureCHAPTER 2. LEBESGUE MEASURE ON R" 32 Definition 2.4.3. Let X be a non-emply set. We say a sub-collection MC 2 of subsets of X to be a o-algebra if (a) OEM. (b) If E 0 then p* is Borel, =W(E)+e(F), Definition 2.4.10. An outer measure is said to be Borel regular if u* is Borel and, for each EC X, there exists a Borel set F > E such that u*(B) = w(F). Definition 2.4.11. An outer measure is said to be a Radon measure #f ju" is Borel regular and y"(K) < +00, for each compact subset of X. Brercise 31. The Lebesgue measure on R” is a Radon measure, 2.5 Invariance of Lebesgue Measure Exercise 32. Show that if B ¢ £(R") (i) (Translation invariance) then E+ € £(R"), for every 7 € R”, and H(E) — p(E +2). (ii) (Refiection) then —B := {—x | x © E} is in £(R") and p(B) = w(-E). (iii) (Dilation) and A > 0, then AW := {Ax | 2 © B} is in £(R") and HOB) = Xu(E) Recall the definition of outer measure, which was infimum over all covers made up of cells. By a cell, we meant a rectangle in R” whose sides are parallel to the coordinate axes. Though it is not clear from the definition of Lebesgue outer measure, it turns out the Lebesgue measure is invariant if we Include rectangles whose sides are not parallel to the coordinate axes. The Lebesgue measure is invariant under orthogonal transformations.CHAPTER 2. LEBESGUE MEASURE ON R” 34 2.6 Measurable Functions Recall that our aim was to develop the Lebesgue notion of integration for functions on R". To do so, we need to classify those functions for which Lebesgue integration makes sense. We shall restrict our attention to real valued functions on R", Let IR == RU {—0, +90} denote the extended real line. Definition 2.6.1. We say a function f on R” is extended real valued if tt takes value on the extended real line K. Henceforth, we will confine ourselves to extended real valued functions unless stated otherwise. By a finite-valued function we will mean a function not taking too. Recall that we said Lebesgue integration is based on the idea of paxti- tioning the range. As a simple case, consider the characteristic function of a set BCR", afl feck XO V0 ite eB. [ower [=e me) But the last equality makes sense only when £ is measurable. ‘Thus, we expect to compute integrals of only those functions whose range when parti- tioned has the pre-image as a measurable subset of IR” We expect that Definition 2.6.2. We say a function f : BC R” + R is measurable (Lebesgue) if for alla € R, the set” f7'((-20,0)) = {f a} is measurable for every a € B. (iv) If f is measurable then show that —f is also measurable. (v) Let f be finite-valued. f is measurable iff {a < f < 8} is measurable for every a, 8 € RK Exercise 35. If f is measurable then (® f*, is measurable for all integers k > 1 (ii) f +A is measurable for a given constant A ¢ R. (iii) Af is measurable for a given constant \ € R. Exercise 36. If f,g are measurable and both finite-valued then both f +g and fg are measurable. Definition 2.6.4. A property is said to hold almost everywhere (a.¢.) if it holds except possibly on a set of measure zero. Consequently, we say a measurable function is finite a.c. if the sot on which it takes too is of measure zero. All the “finite-valued” statements above can be replaced with “finite a.c.”. Let M(R") denote the space of all finite a.c, measwable functions on R”. The class of functions M(R") forms a vector space over R, Note that M(R") excludes those measurable function which takes values on extended line on a non-zero measure setCHAPTER 2. LEBESGUE MEASURE ON R” 36 Definition 2.6.5. We say two functions f,g are equal ae, f= 9 ae., if the set {x| F(z) 4 9(2)}) is of measure zero. Define the equivalence relation f ~ g if f = g ac. on M(R"). Thus, we have the quotient space M(R")/ ~. However, as an abuse of nota- tion, it has become standard to identify the quotient space M(R")/ ~ with M(R"). Henceforth, by M(B") we refer to the quotient space. Thus, wh ever we say A C M(R*), we usually mean the inclusion of the quotient spaces A/ ~C M(R")/ ~. In other words, each equivalence class of M(R") has a representative from A Note that the support of a function f € M(R") is defined as the closure of the set E, B= {x| f(x) #0}, Thus, even though vq ~ 0 are in the same equivalence class and represent the same clement in M(R")/ ~, the support of xg is R whereas the support of zero function is empty set. Therefore, whenever we say support of a function J € M(B") has some property, we usually mean there is a representative in the equivalence class which satisfies the said properties Exercise 37 (Complete measure space). If f is measurable and f = g ac., then g is measurable Theorem 2.6.6. Let f be finite a.e. on R". The following are equivalent: (i) f is measurable. (ii) $-"(Q) is a measurable set, for every open set @ in R. (iti) f(D) is measurable for every closed set T in R. (iu) [>(B) is measurable for every Borel set B in R. Proof. Without loss of generality we shall assume that f is finite valued on ® () implies (ii) Let @ be an open subset of R. Then @ = U® (a;,b,) are intervals which are pairwise disjoint. Observe that LQ) = UES Mai, b,) = VELL > a} O{F < b}) (a1,6,) whereCHAPTER 2, LEBESGUE MEASURE ON R” 37 Since f is measurable, both {f > a,} and {f < 6} are measurable for all 4, Since countable union and intersection of measurable sets are measurable, we have f-(Q) is measurable. (Gi) implies (iii) f>2(P) = (J-(L9))*. Since T* is open, f-'(T) is mea- surable and complement of measurable sets are measurable (Gi) implies (iv) Consider the collection A= {FCR|f(F)€ LR} We note that the collection A forms a g-algebra, Firstly, 0 € A. Let F ¢ A FOP) = F(R) \ FCF) = (£7 (F))%. Also, f (USF) = UR fF). By (ii), every closed subset I’ € A and hence the Borel g-algebra of R is included in A iv) implies (i) Let f-!(B) be measurable for every Borel set B CR. In particular, (oo, a) is a Borel set of R, for any @ € R, and hence {f [0,1] be the function defined as a(v) = inf (fol@) = vb (262) where fc is the Cantor function (cf. Appendix B) Erereise 39. Show that g : [0,1] + C is bijective and increasing. Conse- quently, g is Borel Example 2.16 (Example of a Measurable set which is not Borel). Let V be the non-measurable (Lebesgue) subset of (0, 1] (constructed in Proposition 2.2.9) Let M := 9(N) is a subset of C. Since 4"(C) = 0, we have by monotonicity 4(M) = 0 and thus M is Lebesgue measurable (zero outer measure sets are measurable). If M were a Borel set then, by Borel measurability of g, N =g71(M) is also Borel, a contradiction. We provide another example using product of measures. Let V be the non-measurable (Lebesgue) subset of . Then N x {0} C R? has outer measure zero and hence is measurable subset of R?. If N x {0} were Borel set of R? then N should be a Borel set of R (since it is a section of V x {0} with y coordinate fixed). But NV is not Borel.CHAPTER 2. LEBESGUE MEASURE ON 38 What about composition of measurable functions. Let f : RY + R be measurable (Lebesgue) and g : R -+ R is also Lebesgue measurable. Then 90 f +1” + B need not be measurable because g~!(—o0,) need not be a Borel set. However, by relaxing the condition on g, we may expect the composition to be measurable. Proposition 2.6.7. If f is measurable, finite ae. on R" and g is Borel measurable on R then go f is (Lebesgue) measurable. In particular, go f is measurable for continuous function g. Consequenily, [~, ~, |f| and |f® for all p > 0 are all measurable. Also, for any two finite valued measurable functions f,9, max(f,g) and min(f, g) are measurable. Proof. Consider the set the interval (—oe, a) in R. By the Borel measurabil- ity of g, = g"(—00, a) is a Borel set of R. Using the measurability of f, (gf) *(-00, a) = f-'(O) is measurable. Prove the rest as an exercise. Example 2.17. The reverse composition f og, in general, is not (Lebesgue) measurable. Consider the Borel function g : [0,1] > C given in (2.6.1). Let N be the non-measurable subset of [0,1]. Set B := g(.V). Since BCC, E is Lebesgue measurable. Now, set f := xz, which is measurable on (0, 1) Observe that the composition f og = Xw, is not Lebesgue measurable since N is a non-measurable set. Exercise 40. We showed if [is measurable then |f| is measurable. The converse is not true. Given an example of a non-measurable function J such that |f| is measurable Proposition 2.6.8. If {f,} are a sequence of measurable functions then sup, fi(x), inf; fi(x), lim sup, fie) and liminf; y.. fi(x) are all measur- able. Consequently, f(x) = lim fi(2), if exists, is measurable. Proof. If f(x) := sup, fi(x) then {f > a} = Ui{f; > a} and is measurable If f(x) = inf, fi(x) then f(x) = —sup,(—fi(x)). Also, limsup,.. fi(2) = inf, (sup,>, f:) and lim inf, ee fi(x) = sup,(infis, fi) a ‘The space of measurable functions MM (R”) is closed under point-wise con- vergence. If C(JR") denotes the space of all continuous functions on R”, then wwe have already seen that C'(R") C M(R"). We know from classical analysis that C(R") is not closed under point-wise convergence and M(IR") can be thought of as the “completion” of G(R") under point-wise convergenceCHAPTER 2, LEBESGUE MEASURE ON 39 Recall that in Riemann integration, we approximated the graph of a given function by polygons, equivalently, we were approximating the given function by step functions. We shall now introduce a general class of functions which includes the step functions which corresponds to Lebesgue intergation Definition 2.6.9. A finite linear combination of characteristic functions is called a simple function, i.e., a function 6: E CR" + R is said to be a simple function if i is of the form : (2) = axe, for measurable subsets Ey CR” with 4(B,) < +00 and a; € R, for all i. A simple function @ is said to be a step function if Ej = R; are the (bounded) cells in BR By definition, a simple function is measurable and finite, hence in M(B") The class of a simple functions forms a vector subspace of M(R") as seen from the exercise below. Exercise 41. If @ and ¥ are simple functions on R” then ¢ + W and gv are simple too. Also, if ¢ is simple, A¢ is simple for all A € R. Note that the representation of the simple function , by our definition, is not unique. Definition 2.6.10. A simple function @ is said to have the canonical repre- sentation if * Yas, for disjoint measurable subsets E, CR" with p(E,) < +00 and a; £0, for all i, and a; # a; fori £5. Exercise 42. Every simple function can be decomposed in to its unique canon- ical representation. Proof. Let @ be a simple fmction, Then ¢ can take only finitely many distinct values, Let {bi,... be} be the distinct non-zero values attained by @. Define Ky = {x € BR” | 6(2) = b,}. By definition, £,’s are disjoint and we have the canonical representation of ¢. oCHAPTER 2. LEBESGUE MEASURE ON R" 40 Exercise 43. If ¢ is simple with canonical representation ¢ = 3, axe, then || is simple and |é| = 3°, lave, Theorem 2.6.11. For any finite a.e, measurable function f on R” such that J 20, there exists a sequence of simple functions {ox}? such that (i) oy = 0, for each k, (non-negative) (8) ox(2) < Ges1() (increasing sequence) and (284) limy 00 d(x) = f(x) for all x (point-wise convergence) Proof. Note that the domain of the given function may be of infinite measure We begin by assuming that f is bounded, | f(r)| 1, we partition the range [0, M) with intervals of length 1/2* and correspondingly define the set, a atl Eine {= €Rul ge SS) ah for all integers 0 1, we have a disjoint partition of the domain of f, Rar, in to {£,4};. Bix are all measurable due to the measurability of f. Hence, for each k, we define the simple function VFA), ih x(2) where J, is the set of all integers in (0,24). By definition @,’s are non- negative and dx(x) < f(x) for all 2 © Ray and for all k. In particular, [eel k 0 elsewhere. fia By construction, fx(x) < fey1(z), for all z € RY. Also, f(z) 2P f(x) converges point-wise, for all x € R". To see this fact, fix rs € R” and let € be the such that x ¢ Ry for all k < €. Thus, f(x) — 0 for all k= 1,2,...,€-1 Let m = f(z). Im < é then fe(x) = f(x) for all k > £ and hence the soquence converges point-wise. If m > é, choose the first integer i such that £44 > m > Eand we have x(x) = f(x) for all k > €+ i, and converges to Ste) Note that each fi, is measurable due to the measurability of f. Since the range of fi is [0,], it is enough to check the measurablility of fy for all a € (0,k]. The extreme cases, {fe < 0} = Rj, {fe < 0} = 0 and {fx < k} = IR” are measurable. For any a € (0,k), {fe < a} = REU{S 1} and, for k = 2,3,..., we define La ra f{rer seeps St eco} By construction, re) > Dive i ‘This is because at every stage k, ' J@)2 Dra) Clearly, the equality is true for {f = 0} and {f = +00}. Now, fix ar € RY such that 0 < f(x) < +00, then x ¢ B,,, for asubsequence ky, of k. Consequently, Letting ki, > 00, we have f(x) < Of; ¢Xe, (x). Hence, the equality holds. a In the next result, we relax the non-negativity requirement on f. Theorem 2.6.12. For any finite ae. measurable function f on R® there exists a sequence of simple functions {®,} such that (i) |Px(2)| < |Pxa(@)| and (i) Virmy 400 Pex) = f(@) for all x (point-wise convergence)CHAPTER 2, LEBESGUE MEASURE ON 43 In particular, |®4(2)] < |f(2)| for all « and k. Proof. Any function f can be decomposed in to non-negative functions as follows: f = f* — f>. Corresponding to each we have a sequence of dx and vp. satisfying properties of previous theorem and converges point-wise to f* and f-, respectively. By setting, B, = dy — Ye we immediately see that O(c) > f(x) for all x. Let Ey = {f < 0}, Bp = {f > 0} and Ey = {f =0}. Since both f* and f~ vanishes on Es, dx, Ux vanishes on Bs Thus, $, = 0 on Ky, Similarly, on By, &, = —vy <0 and on Ey, & = dy Thorefore, |x| = dj + Ve and hence is an increasing sequence. a In the above two theorems, we may allow extended real valued measurable function f, provided we allow the point-wise limit to take oo. The results above shows the density of simple functions in the space of finite valued functions in M(R") under the topology of point-wise convergence. 2.7 Littlewood’s Three Principles We have, thus far, developed the notion of measurable sets of R” and measur able functions on R”. J. E, Littlewood” simplified the connection of theory of measures with classical real analysis in the following three observations (i) Every measurable set is “nearly” a finite union of intervals. (ii) Every measurable function is “nearly” continuous. (iii) Every convergent sequence of measurable functions is “nearly” uni- formly convergent. Littlewood had no contribution in the proof of these principles. He sum- marised the connections of measure theory notions with classical analysis 2.7.1 First Principle Theorem 2.7.1 (First Principle). If E is a measurable subset of R” such that :(E) < +00 then, for every ¢ > 0, there exists a finite union of closed cells, say, such that (BAT) < €.'$ Tip his book on complex analysis titled Lectures on the Theory of Functions Sear = (BUP)\ (ENP)CHAPTER 2, LEBESGUE MEASURE ON 3” 44 Proof. For every = > 0, there exists a closed cover of cells {Ri} f° for E (B.C US;R,) such that DRI < HB) +5 Since (EZ) < +00, the series converges. ‘Thus, for the given > 0 there exists a k € N such that k 2 SOR — SOR ia LY le<§ th Set P= UE, Ri. Now, H(BAD) ME\T)+u(T\B) (by additivity) (USn Ri) +e (US)R: \Z) (by monotonicity) 1 (UR eri) + (UR) —n(E) (by additivity) Yo RI + OR = WP) (by sub-additivieyy Waa lA ao Note that in the one dimension case one can, in fact, find a finite union of open intervals satisfying above condition, At the end of last section, we saw that the sequence of simple functions were dense in M(R") under point-wise convergence. Using, Littlewood’s first principle, one can say that the space of step functions is dense in .M(IR") under a.e. point-wise convergence. Theorem 2.7.2. For any finite ae. measurable function [ on R there exists a sequence of step functions {G4} that converge to f(x) point-wise for ae reR" Proof. It is enough to show the claim for any characteristic function f = xx, where £ is a measurable subset of finite measure. By Littlewood’s fixst principle, for every integer k > 0, there exist Uf, finite union of closed cells such that w(BA UL, R,) < 1/2. It will be necessary to consider cellsCHAPTER 2, LEBESGUE MEASURE ON 45 that are disjoint to make the value of simple function coincide with f in the intersection. Thus, we extend sides of R, and form new collection of almost disjoint cells Q, such that Uf_,R, = U?;Q,. Further still, we can pick disjoint cells P, C Q, such that Fy = UEP, C UL) Ri and p((Uf)R,) \ Fe) < 1/2* Define wn on = xr, a Note that f(x) = de(x) for all x € US,P, and Ey := {2 € R"| f(z) # dx(z)} = BAF, with u(B) < 1/2). Set Gy := US, Bs, hence Gy > Gr > and u(Ge) < 1/2-?. Note that if 1 € Gy, then there exists a p > k such that f(x) 4 ¢p(z). Set G-=%,G;, then G ={f 4 ox} and w(G)=0. O The space of step functions on R” is dense in M(R") endowed with the point-wise a.e. topology. 2.7.2 Third Principle The time is now ripe to state and prove Littlewood’s third principle, a con- sequence of which is that uniform convergence is valid on “large” subset for a point-wise convergence sequence. Theorem 2.7.3. Suppose {f,} ts a sequence of measurable functions defined on a measurable set B with u(B) < +00 such that fi,(x) —> f(x) (point-wise) a.c. on. Then, for any given c,6 >0, there is a measurable subset Ff CB such that (2\ Ff) < 6 and an integer K € N (independent of x) such tha, for all x € F5, \fla) -f@|K Proof. We assume without loss of generality that fx(x) -* f(z) point-wise, for all x € E. Otherwise we shall restrict ourselves to the subset of where it holds and its complement in Z is of measure zero. For each ¢ > 0 and x € B, there exists a k © N (possibly depending on zz) such that lil) -f@)i<< Wek Since we want to get the region of uniform convergence, we accumulate all x € E for which the same k holds for a fixed ¢. For the fixed © > 0 and for cach & EN, we define the set BE {ee BIW) —S@l 0, there is ake €N such that (IE) ~ pl) = w(I2\ Ej) <8 Vk > kes If Bi, #0, set Ff = BE, and K = hy else set Ff to be the first non-empty sot E*, for m > ky and == m, We have, in particular, p(B \ Ff) < 6 and for all x € Fj, I@ —f@l 0, there is a closed subset FC E such that (E\ 15) <6 and an integer K €N (independent of x) such that, for all x eT, \felx) — f@)| f(x) (point-wise) a.e. on E. Then, for any given 6 > 0, there is a measurable subset Fs CE such that p(B \ Fs) <6 and fx > f uniformly on Fs Proof. From the theorem proved above, for a given 6 > 0 and k € N, there is a measurable subset Fi, C E such that 4(B\ Fi) < 6/2 and for all x € Fi, there is aN, € N [fil — f(z) Ne Sct Fy = OF. Thus, ME\ Fs) = MUR (E\ Fe) < SU ME\ Fi) <6 Now, it is easy to check that fe + f uniformly in Fs oOCHAPTER 2, LEBESGUE MEASURE ON 47 Example 2.18. The finite measure hypothesis on E is necessary in above theorems. Let fe = Xjn+) then fx(x) > 0 pointwise, Choose 6 = 1 and let F be any subsot of R such that w(F) < 1. Because u(F) < 1, Fen |k,k +1) 4 6, for all k, We claim that f, cannot converge uniformly to 0 on F*. For every k € N, there is a zy € F°M[k,k + 1) such that [fe(ox) — f(zx)| = Lfelaa)| The Egorov's theorem motivates the following notion of convergence in M(R") Definition 2.7.6. Let {fu}? and f be finite a.c. measurable functions on a measurable set ECR". We say f, converges almost uniformly™ to f on E, if for every 6 > 0, there exists measurable subset Fs CF such that u(F3) <€ and [y+ { uniformly on E\ Fs Exercise 46. Show that almost uniform convergence implies point-wise ae. convergence. Proof. Let fy + f almost uniformly converge. Then, by definition, for each k EN, there exists a measurable set F, C E such that u(F,) < 1/k and fie @ f uniformly on F\ Fy. Let F = 02, Fs. Thus, u(F) < (Fi) < 1/k for all & and hence j(F) = 0. For any x € E\ F, x € Uyai(E\ Fy) and hence x € B\ F;, for some k. Therefore, by the uniform convergence of fy f, we have f(r) —> f(x) point-wise. Thus, f,(x) -> f(x) for all 2 € E\ F and u(F) = 0, showing the point-wise a.c. convergence. a ‘The converse is not true, Example 2.18 gives an example of a point-wise a.e, converging sequence which do not converge almost uniformly. However, the converse is true for a finite measure set £. The Egorov’s theorem is pre- cisely the converse statement for finite measure set. Thus, on finite measure sot we have the following statement: Exercise 47. For finite measure set j1() < +00, a sequence of functions on E converges point-wise a.e. iff it converges almost uniformly. We shall end this section by giving a weaker notion of convergence on M(B’) Note that this notion of convergence is much weaker than demanding uniform conver- ence except on zero measure sets.CHAPTER 2, LEBESGUE MEASURE ON 3 48 Definition 2.7.7. Let {fi}? and f be finite a.e. measurable functions on a measurable set E CR". We say f, converges in measure to f on E, denoted as fy 4 f, if for everye > 0, sim ol) where = {x € B| |felx) — f(2)| > ep. Exercise 48. Almost uniform convergence implies convergence in measure. Proof. Since fy converges almost uniformly to f. For every ¢ > 0 and k € N, there exists a sot F, (independent of c) such that (4) < 1/k and there exists K €N, for all x € B \ Fz, such that IG@)-f@)|K. ‘Therefore Ef C F; for infinitely many j > K. Thus, for infinitely many j, (ES) < 1/k. In particular, choose jj, > k and u(E%,) < 1/k. a Example 2.19. Give an example to show that convergence in measure do not imply almost uniform convergence. However, the conver is truc upto a subsequence Theorem 2.7.8. Let {fi}? and f be finite a.e, measurable functions on a measurable set E CR" (not necessarily finite). If fx “ f then there is a subsequence {x,}{21 such that fr, converges almost uniformly to f 2.7.3 Second Principle Recall M(R") is decomposed in to equivalence classes under equality ae So, it would be a nice situation if for every measurable function there is a continuous function in its equivalence class. In other words, we wish to have for every measurable function f a continuous function g such that f = g ac. Unfortunately, this is not true. Exercise 49. Given an example of a measurable function f for which there is no continuous function g such that f = 9 a.c Littlewood’s second principle is “approximating” a measurable function on finite measure by a continuous function.CHAPTER 2, LEBESGUE MEASURE ON 49 Erercise 50. Let xp be a step function on R", where R is a cell with |R| < +00. Then, for ¢ > 0, there exists a subset E, CR” such that xp restricted to RY \ E, is continuous and ju(H.) <¢. Theorem 2.7.9 (Luzin). Let f be measurable finite a.e. on a measurable set E such that (BE) < +00. Then, for ¢ > 0, there exists a closed set Te C E such that p(E\T'.) <¢ and f |p, is a continuous function’® Proof. We know from Theorem 2.7.2 that step functions are dense in M(R") under point-wise a.c. convergence. Let {¢,}{2, be a sequence of step fune- tions that converge to f point-wise a.c. Fix © > 0. For each k € N, there exists a measurable subset Ey. © E such that (Ex) < (€/3)(1/2*) and by re- stricted to £\ Hy is continuous. By Egorov’s theorem, there is a measurable sot F, C E such that u(B\ F.) < ¢/3 and gy > f uniformly in F:, Note that dy restricted to Ge = F.\\U2,£x is continuous. Therefore, its uniform, limit f restricted to G, is continuous. Also, HEN Ge) = WUE Ba) + WB \ Fe) < (26)/3. By inner regularity, pick a closed set I, C G, such that u(G. \T.) < €/3. Now, obviously, f |p, is continuous and M(E\T.) = w(B\ Ge) + w(Ge\Te) 0, there exists a continuous function 9 on E such that n(x € B | f(z) # ofz)} <= Proof. Use Urysobn lemma or Tietze extension theorem to find a continuous function g on H which coincides with f on I’. Consider the set {2 ¢B| f(2) 4 o(z)}= B\Te The measure of the above set is less than © a Exercise 51. For any finite a.c. measurable function f on R” there exists a sequence of continuous functions {f,} that converge to f(x) point-wise for ae. 2 eR WF restricted to De points of Pe ‘continuous but f as a function on E may not be continuous onCHAPTER 2. LEBESGUE MEASURE ON R" 50 2.8 Jordan Content or Measure Woe end this chapter with few remarks on the notion of “Jordan content of a se ®, developed by Gius dan. This notion ppe Peano and Camille Jo is related to Riemann integration in the ne way as Lebesgue measure is related to Lebesgue integation. The Jordan content is the finite version of the Lebesgue measure. Definition 2.8.1. Let E be bounded subset of R". We say that a finite family of cells {R,}ier is a finite covering of E iff E © UerR,, where I is a finite inder set Let BC 2®" be the class of all bounded subsets of R". The reason for restricting ourselves to B is because every element of B will admit a finite covering. Definition 2.8.2. For a subset EB € B, we define its Jordan outer content JME) as, the infimum being taken over all possible finite coverings of E ‘The term “measure” is usually reserved for a countably additive set fune- the term “content”. Otherwise Jordan content could be viewed as finitely additive measure. Some texts refer to it as Jordan measure or Jordan-Peano measure. tion, hence we ws Lemma 2.8.3. The Jordan outer content J* has the following properties (a) For every subset E € B, 0 < J*(E) < +00. (6) (Translation Invariance) For every B € B, J*(E +2) = J*(B) for all reR” (c) (Monotone) If Ec F, then J*(B) < J*(F) (d) (Finite Sub-additivity) For a finite index set I, F* (Vier Bi) < OIE) ‘aCHAPTER 2, LEBESGUE MEASURE ON 8” OL Proof. ‘The proofs are similar to those of outer measure case a Exercise 52. Show that J*(R) = |R| for any cell R CR", Consequently, 40(R) = J*(R) for every cell R. Exercise 53. If E © R? denotes region below the graph of a bounded function f: RR. Show that the Jordan content of F is same as the Riemann upper sum of f, Example 2.20, Jordan outer content of the empty set is zero, 4*(9) = 0. Every cell is a cover for the empty set. Thus, infimum over the volume of all cells is zero Brample 2.21. The Jordan outer content for a singleton set {2} in R* is zero. The same argument as for empty set holds except that now the infimum is taken over all cells containing x. Thus, for each ¢ > 0, one can find a cell Re such that x € Re and |Re| 0 and hence p*({x}) = 0. Example 2.22. The Jordan outer content of a finite subset E of R” is zero. A finite set = Uzee{x}, where the union is finite. Thus, by finite sub- additivity, j2"(E) < 0 and hence y"(E) = 0. This is precisely where the difference lies between Lebesgue measure and Jordan content. Recall the Q had Lebesgue outer measure zero, But the Jor- dan outer content of Q contained in a bounded cell is positive. For instance, consider E := QM (0, 1]. We shall show that J*(E) = 1. Note that there is no finite cover of Z which is properly contained in (0,1), due to the density of Q in (0,1). Thus, any finite cover of £ also contains [0, 1). In fact, (0, 1] is itself a finite cover of £. Infimum over all the finite cover is bounded below by 1, due to monotonicity. Thus, J*(£) = 1 Exercise 54. Show that J*(B) = J*(B), where B denotes the closure of E. Proof. Firstly, the result is true when FE = Ris a cell, since J*(R) = u*(R) = 4°(R) = J*(R). By monotonicity of J*, J*(E) < J*(B). For converse argument, let_{R,} be any finite cover of E, ic., B.C UsR; then Bc UR In general, U;R; © U;R;. However, since the union is finite we have equality, OR. Therefore, {R;} is a finite cover of E. Thus, FE) < SIR = Sik URCHAPTER 2, LEBESGUE MEASURE ON 3” 52 ‘Taking infimum over all finite covers of B, we get J*(E) < J*(B). a Note that in the proof above the finite cover played a crucial role. A. similar result is not true Lebesgue outer measure. For instance, «*(Q 9 (0, 1]) = 0 and its closure is (0, 1] whose outer measure in one. More generally, for every k > 0, we have a set B CR" such that, *(E) = 0 but p*(OB) = k, where OF is the boundary of E. For instance, consider B= Q" (0, kY/")" Then, JE = [0,k!/"]" and yi*(QB) = k. Exercise 55. Show that the Jordan content of a sot is same as the outer measure of its closure, ie., w"(E) = J*(B) Proof. Note that it is enough to show that :"(E) = J*(E). Firstly, it follows from defintion that *(E) < J*(E). For the reverse inequality, we consider {Rj} to be an countable cover of B. Since F is closed and bounded, hence compact, there is a finite sub-cover {Qj} of E. Thus, FES Yl < VR ‘Taking infimum over all countable covers {R,}, we get IB) <'B) Hence the equality holds. a To classify the Jordan measurable subsets, we need to identify the class of subsets of B for which finite additivity holds. Note that Q/ [0, 1] cannot be Jordan measurable. Since J*(@/0, 1]) = 1. A similar argument also shows that J*(Q°1 0,1] = 1. If finite additivity were true then J*(Q (0,1)) + I(QN(0,1) =2 41 = J*((0, 1). Theorem 2.8.4. A bounded set E CR" is Jordan measurable iff Xe is Riemann integrable. Now do you see why the characteristic fmction on Q was not Riemann integrable? Precisely because Q was not Jordan measurable. Do you also see how Lebesgue measure fixes this inadequacy?Chapter 3 Lebesgue Integration In this chapter, we shall define the integral of a function on R", in a progres- sive way, with increasing order of complexity. Before we do so, we shall state some facts about Riemann integrability in measure theoretic language Theorem 3.0.5. Let f : [a,] + R be a bounded function. Then f Ra, 8)) iff f as continuous a.e. on [a,b] Basically, the result says that a function is Riemann integrable iff its set of discontinuities are of length (measure) zero 3.1 Simple Functions Recall from the discussion on simple functions in previous chapter that the representation of simple functions is not unique. Therefore, we defined the canonical representation of a simple function which is unique. We use this canonical representation to define the integral of a simple function. Definition 3.1.1. Let @ be a non-zero simple function on R" having the canonical form o(z) = axe, a with disjoint measurable subsets E; CR” with y(B,) < +00 and a, #0, for all i, and a, # a, for i # j. We define the Lebesgue integral of a simple 33CHAPTER 3. LEBESGUE INTEGRATION 54 function on R", denoted as where p is the Lebesgue measure on R". Henceforth, we shall denote [ ddp as f édx, for Lebesgue measure. Also, we define the integral of @ on ECR” Jeorae= [ oayre(e)ar Note that the integral of a simple function is always finite. Though, we chose to define integral using the canonical representation, it turns out that integral of a simple function is independent of its representation. Proposition 3. . For any representation of the simple function 6 = Dhiaxe,, we have . [ 62) dr = SP ain). ™ 1 Proof. Let ¢ = Si, aixg, be a representation of ¢ such that y's are pairwise disjoint which is not the canonical form, ie., a; are not necessarily distinct and can be zero for some i. Let {b;} be the distinct non-zero clements of {a1,...,ax}, where 1 0 then [4 > 0. Consequently, if ¢ < ¥, then edz [ var. Les - (iv) (‘Triangle Inequality) We know for a simple function ¢, |9| is also simple. ‘Thus, J ode ae. then fo= fu <=] \oldx Ian (v) Ie Example 3.1. An example of a Lebesgue integrable function which is not Riemann integral is the following: Consider the characteristic function xq ‘We have already seen in Example 1.3 that this is not Riemann integrable. But [role de = W(Q) =0 hsCHAPTER 3. LEBESGUE INTEGRATION 36 However, yg = 0a. and zero function is Riemann integrable. Thus, for vq which is Lebesgue integrable function there is a Riemann integrable function in its equivalence class. Is this always true? Do we always have a Riemann integrable function in the equivalence class of a Lebesgue integrable function, ‘The answer is a “no”. Find an example! Exercise 57. Show that the Riemann integral and Lebesgue integral coincide for step functions. 3.2 Bounded Function With Finite Measure Support Now that we have defined the notion of integral for a simple function, we intend to extend this notion to other measurable functions. At this juncture, the natural thing is to recall the fact proved in Theorem 2.6.12, which estab- lishes the existence of a sequence of simple functions ¢ converging point-wise toa given measurable finite ae. function f. Thus, the natural way of defining the integral of the funetion f would be f(x) dr = tim, f oul) de, lan im Jan This definition may not be well-defined. For instance, the limit on the RHS may depend on the choice of the sequence of simple functions ¢, Example 3.2. Let f = 0 be the zero function, By choosing d, = X(,1/%) which converges to f point-wise its integral is 1/k which also converging to 0. However, if we choose Yx = kx(o,1/x) Which converges point-wise to f, but f vx = 1 for all & and hen nverges to 1 But zero function is trivially a simple function with Lebesgue integral zero. Note that the situation is very similar to what happens in Riemann’s notion of integration. Therein we demand that the Riemann upper sum and Riemann lower sum coincide, for a function to be Riemann integrable. In Lehesgue’s situation too, we have that the integral of different sequences of simple functions converging to a function f may not coincide, ‘The follow- ing result singles out a case when the limits of integral of simple functions coincide for any choice Proposition 3.2.1. Let f be a measurable function finite a.e. on a set B of finite measure and let {4x} be a sequence of simple functions supported on ECHAPTER 3, LEBESGUE INTEGRATION oT and uniformly bounded by M such that x(x) > f(x) point-wise ac. on B. Then L := limgsoo fy bu de is finite. Further, L is independent of the choice of {dr}, ie, ff =0 ae. then L=0. Proof. Since ¢,(x) + f(a) point-wise ac. on E and p(B) < +00, by Egorov's theorem, for a given 5 > 0, there exists a measurable subset Fy C E such that 4(E'\ Fs) < 6/(4M) and @ > f uniformly on Fs, Sot y= fp de We shall show that {Jz} is a Cauchy sequence in Rand hence converges Consider t= tal Sf \ou(e)~ dm (2)] (rine inequality) |. ~ I, (0) ela + f Jena) ~ bata) é K MOTB) 2 > < 6 forall k,m>K (By monotonicity of 2) Thus, {J,} is Cauchy sequence and converges to some L. If f = 0, repeating the above argument on Jy Wil < f de(2)| (triangle inoquality) le = flee + [ fs é < on(x)| + > [ hostel +t B66 pal 7 +5 allk * < MET ty forallb> Ke < § forall k > K' (By monotonicity of p), we get L=0. a We know from Theorem 2.6.12 the existence of a sequence of simple fune- tions $, converging point-wise to a given measurable finite a.e. function f If, in addition, we assume f is bounded and supported on a finite measure set E, then the ¢, satisfy the hypotheses of above Proposition. This motivates us to give the following definitionCHAPTER 3. LEBESGUE INTEGRATION 38 Definition 3.2.2. Let f be bounded measurable function supported on a set E of finite measure. The integral of f is defined as i Fe) de = Jim i x(x) de where {64} are uniformly bounded simple functions supported on the support of f and converging point-wise to f. Moreover, for any measurable subset FCE, [teres | seoxrtnae |; E Exercise 58. Show that all the properties of integral listed in Exercise 56 is also valid for an integral of a bounded measurable function with support on finite measure. Exercise 59. A consequence of (v) property is that if f =O a.c. then f f= 0. The converse is true for non-negative functions. Let f be a bounded measurable function supported on finite measure set. If f > 0 and f f = 0 then f =0 ac. The way we defined our integral of a function, the interchange of limit and integral under point-wise convergence comes out as a gift Theorem 3.2.3 (Bounded Convergence Theorem'). Let fy be a sequence of measurable functions supported on a finite measure set E such that | fe(x)| < M for all & and x € E and fx(x) + f(x) point-wise a.e. on E. Then f is also bounded and supported on E a.e. and sin fln-si=0 da [in fr Proof. Since f is a point-wise a.c. limit of fy, [f(x)| 0, there exists a measurable subset Fy C E such that (E\ Fs) < 6/(4M) and fy > f In particular, This statement not elementary’ ‘same as the one in Theorem 1.1.6 where we mentioned the proof isCHAPTER 3, LEBESGUE INTEGRATION 59 uniformly on Fs. Also, choose K€ N, such that |fx(x) — f(x)| < 6/2u(B) for all k > K. Consider A [itncer— seo sf inter —seori+ f rece = 100) eves ri) +8 for ak > K HC Daa +3 for > < 6 forallk>K (By monotonicity of 2). ‘Therefore, limg soo fp |e — [| =0 and, by triangle inequality, an [= It is now time to address the problem of Riemann integration which does not allow us to interchange point-wise limit and integral. We first observe that Riemann integration is same as Lebesgue integration for Riemann in- tegrable functions and thus, by BCT, we have the interchange of limit and integral for Riemann integrable functions, when the limit is also Riemann integrable. a Theorem 3.2.4. If f € R([a,b]) then f is bounded measurable and ‘ [ Sle) de F(a) de, a a the LHS is in the sense of Riemann and RHS in the sense of Lebesgue. Proof. Since f € R{(a,6]), f is bounded, |f(x)| 0. Also, the support of f, being subset of [a, 6], is finite. We need to check that f is measurable. Since f is Riemann integrable there exists two sequences of step functions {¢,} and {q,} such that andCHAPTER 3. LEBESGUE INTEGRATION 60 Also, |x| 0, we must have Ya > 0. Thus, by Exercise 59, ¥— =O ac, and hence Wa. Hence f is measurable. Thus, . [det I ae [ f(x) de a The same statement is not true, in general, for improper Riemann inte- gration (cf. Exercise 63) 3.3 Non-negative Functions We have already noted that (cf. Example 3.2) for a general measurable fune- tion, defining its integral as the limit of the simple functions converging to it, may not be well-defined, However, we know from the proof of Theorem 2.6.11 that any non-negative function f has truncation fi, which are each bounded and supported on a sct of finite measure, increasing and converge point-wise to f, There could be many other choices of the sequences which satisfy sim- ilar condition. This motivates a definition of integrability for non-negative functions. Definition 3.3.1. Let f be a non-negative (f > 0) measurable function. The integral of f is defined as, F(x) de vo | g(x) dxCHAPTER 3, LEBESGUE INTEGRATION 61 where g is a bounded measurable function supported on a finite measure set. ‘As usual, [ Sa) ax le since [xp > 0 loo, if f >0 F(2)xe(r) dx Note that the supremum could be infinite and hence the integral could take infinite value Definition 3.3.2. We say a non-negative function f is Lebesgue integrable y S(x)dr < +00. hyo Exercise 60. Show the following properties of integral for non-negative mea- surable fmctions: (i) (Linearity) For any two measurable functions f,g and a, 6 € R, I (i) (Additivity) For any two disjoint subsets #, F C IR” with finite measure Jp ften [tees [sae (iii) (Monotonicity) If f < g, then [fers [ aan In particular, if g is integrable and 0 < f < g, then f is integrable. (iv) If f=gac. then f f= fg. (v) If f =O and f f =0 then f =0 ae. (of + oa)de=a ff fatera ode. Ian (vi) If f is integrable then J is finite ae. Do we have non-negative functions which are not Lebesgue integrable, ie,, for which the supremum is infinite?CHAPTER 3. LEBESGUE INTEGRATION, 62 Example 3.3. The function _ [ifle| for |] <1 $@ = {i for |z| > 1 is not integrable. Following the definition of the notion of integral of a function, the im- mediate question we have been asking is the interchange of point-wise limit and integral. Thus, for a sequence of non-negative functions { f.} converging pointwise to f is / fale / fe We have already seen in Example 3.2 that this is not true, in general. How- ever, the following result is always true. Lemma 3.3.3 (Fatou). Let {fe} be a sequence of non-negative measurable? functions converging point-wise a.e. to f, then / f Sliminf / Se Proof. Let 0 < g < f, where g is bounded with support on a set of finite measure E. Let gx(r) = min(g(z), f(x), then gy is measurable. Also, gy is bounded by the bound of g and supported on B, since g, < g and g, fr are non-negative for all k. We claim that gp converges to g point-wise ae in B. Fix x € B. Then either g(x) ~ f(x) or g(x) < f(z). Consider the case when g(x) < f(x). For any given e > 0 there is a K € N such that | fa(z) — f(z)| < © for all k > K. In particular, this is true for all © < F(e) — g(x). Thus, g(x) < f(x) —e < fe(x) for all k > K and hence gx(e) = 9(2) for all k > K. On the other hand if g(r) = f(z) then g(x) is either f(x) or fe(x) and will converge to f(x). Thus, ge(2) g(x) ac. and by the BCT f gx — J 9. Moreover, ge < fy and, by monotonicity, foe < f fe and therefore, font fo =tmint fo / f. Corollary 3.3.4. Let {fg} be a sequence of non-negative measurable func- tions converging point-wise a.e. to f and fx(x) < f(x), then [top [ te Proof. By monotonicity, f fi < ff and hence timsup f fa < [i srimint | f. k ‘The second inequality is due to Fatou’s lemma and hence we have fin [i Corollary 3.3.5 (Monotone Convergence Theorem). Let {fx} be an inereas- ing sequence of non-negative measurable functions converging point-wise a.¢. to f, ice., felt) < feri(), for all k, then firm [6 Proof. Since fi, is increasing sequence, we have fe(sx) < f(x) and hence we have our result by previous corollary. a aCHAPTER 3. LEBESGUE INTEGRATION 64 The MCT is not true for a decreasing sequence of functions. Consider FiXthox) On R. fy are non-negative and measurable functions on R. ‘The sequence f, converges point-wise to f = 0, since for each fixed x € R, Ju) = 0 for infinitely many &’s. However, {f,} are decreasing, fua(e) < ‘fi (x) for all x and k. Moreover, ff, — 00 and f f = 0. Thus, [or riips [te Corollary 3.3.6. Let {fu} be a sequence of non-negative measurable func- tions. Then oo oo I(& rn) te= > | ied a a Proof. Set g(t) = OW, fila) and glz) = T, felz)- gm axe measurable and gm(x) < gmsi(z) and g,, converges point-wise g. Thus, by MCT, [ot fon [Xie = fa- in fon tw > | 1 fr ‘Thus, “YE fae ao The highlight of Fatow's lemma and its corollary is that they all remain true for a measurable function, i.e., we do allow the integrals to take oo. We end this section by giving a different proof to the First Borel-Cantelli theorem proved in Theorem 2.3.13. Theorem 3.3.7 (First Borel-Cantelli Lemma). If {E,}{°_¢ £(R") be a countable collection of measurable subsets of R” suck that T°, 4(E,) < 00 Then B= 1%, US, By has measure zero, Proof. Define fi, = xm, and f = ¥ fx. Since fi, are non-negative, we have from the above corollary that, [t-Xuen < +00 7CHAPTER 3, LEBESGUE INTEGRATION 65 Thus, f € L'(R") and hence is finite a.e. Thus, the set F := {x © R” | f(2) = co} has measure zero, We claim that E = F. If x € F, then 3o, H(Ex) = oe implies that x © Hy, for infinitely many k (a fact observed before) and hence x € E. Conversely, if x € E, then © € Ky for infinitely many k and hence ¢ € F. Thus, (E) = 0. o 3.4 General Integrable Functions In this section, we try to extend our notion of integral to all other measurable functions. Recall that any function f can be decomposed in to f = f*— fr and both f*, f~ are non-negative. Note that if f is measurable, both f* and f> are measurable, We now give the definition of the integral of measurable functions. Definition 3.4.1. The Lebesgue integral of any measurable real-valued func- tion f on R” is defined as l. sede = sear | f(x) dr, Any measurable function f is said to be Lebesgue integrable if [ [F(@)l ar < +00. Any measurable function f is said to be locally Lebesgue integrable if [ lf@) lax < +50, K for all compact subsets K CR” Observe that if f is measurable then || is measurable. But the conver is not true (cf. Exercise 40). In view this, one may have a non-measurable function f which is Lebesgue integrable. To avoid this situation, we assume the measurability of f in the definition of Lebesgue integrability of f Exercise 62. Show that the definition of integral of f is independent of its decomposition f = fy — fz where f, > 0 for i= 1,2. Exercise 63, Consider f(z) = “2 on (0,00). Using contour integration one can show that / is Riemann integrable (improper) and is equal to x/2 However, f is not Lebesgue integrable since f™ f* = [°° f> = 00.CHAPTER 3. LEBESGUE INTEGRATION 66 Definition 3.4.2. A compler valued measurable function f = w+ iv on R™ is said to be integrable if [f(2)|dx = f (w(x) + v?(2))"? da < +00. ne he and the integral of f is given by fi-foofe Exercise 64, Show that a complex-valued function is integrable iff both its real and imaginary parts are integrable. Exercise 65. Show that all the properties of integral listed in Exercise 56 and Exercise 60 is also valid for a general integrable function. ‘The space of all real-valued measurable integrable functions on R" is denoted by L'(R"). Thus, L'(R") c M(B"). We will talk more on these spaces in the next section, We introduce this notation early in here only to use them in the statements of our results. We highlight here that the non-negativity hypothesis in all results proved in the previous section (Fatou’s lemma and its corollaries) can be replaced with a lower bound g € L}(IR"), since then we work with gx = J — g > 0. As usual we prove a result concerning the interchange of limit and integral, called the Dominated Convergence Theorem. Theorem 3.4.3 (Dominated Convergence Theorem). Let { fu} be a sequence of measurable Junctions converging point-wise ae. to f. If |fk()| < (x), for all k, such that g € L'(R") then f € L\(R") and [rm Proof. Sinee |fs| < 9, |f| <9 and since g € L'(R"), by monotonicity, J © LR"). Note that fia) < |fe(a)| < g(x). Hence, g ~ fe = 0 and converges point-wise a.c. to g ~ f. By Fatou’s lemma, fof siimint fo~ ny Therefore, [o- fis [ar tonine (- fr) = [omtinsw f 5CHAPTER 3, LEBESGUE INTEGRATION 67 Thus, limsup [ fe < J J, since fg is finite. Repeating above argument for ‘the non-negative function g + fr, we get f f < liminf f fr. Thus, ff = lim J fe a Exercise 66 (Generalised Dominated Convergence Theorem). Let {9,} C L(R”) be a sequence of integrable functions converging point-wise a.c. to 9 € LR"). Let {fs} be a sequence of measurable functions converging point-wise ae, to f and |fx(2)] < ge(z). Then f € L'(R") and further if im fo [0 ain [i= fs Brample 3.4, The Lebesgue dominated convergence theorem is something more than demanding uniform convergence. Consider f,(2) = 2* on (0, 1). ‘fk(x) — 0 point-wise on [0,1). However, the convergence is not uniform. But |fe| S$ ] and 1 [ Harm [ ahdr= fo.) lo ky converges to zero. Example 3.5. We have already seen using Ux in Example 3.2 that the bound by g in the hypothesis of DCT cannot be done away with. In fact, one can modify vz in that example to have functions whose integrals diverge. For Instance, choose f(z) = kUx = k®X(o,/e) Which point-wise converges to zero and f fi =k which diverges Example 3.6. The condition that g € L}(R") is also crucial. For instance, let fel) = 1/kxjog and |fe(x)| < 1 on R. Note that fy converge uniformly to zer0, J fe =1 do not converge to zero. Why? Because g = 1 is not in L(IR) Corollary 3.4.4. Let {fx} C L'(R") such that > [ier $00. Then S73 , fel) € LR") and / (= )) de =¥ fee de then,CHAPTER 3, LEBESGUE INTEGRATION 68 Proof. Let 9 = S721 |ful Since |fe| is a non-negative sequence, by a corol- lary to Fatou’s lemma, we have that [oe [ (Suni) t= finiaec te Thus, g € L1(R"). Now, consider Ste) “ Therefore DE, f(z) € L4(R"), since g € L'(R"). Consider the partial sum SV A@l =o. { Y fi). a Note that [Fn (2)| < 9(2) for all k and F,,(x) + f(x) ae.. Thus, by DCT, lim, f Fn = f f. By finite additivity of integals, we have [itp [Po weX [aE | Hence proved a F(x) Note that the BCT (cf. Theorem 3.2.3) had a stronger statement than DCT above, In fact, we can prove a similar statement for DCT. EBnercise 67. Let {fi} be a sequence of measurable functions converging point-wise ae. to f. If |fi(xr)| < g(x) such that g € L1(R") then dim fin-s=0 Proof. Note that | fe ~ f| <2g and by DCT lige / lJCHAPTER 3, LEBESGUE INTEGRATION 69 The above exercise could also be proved without using DCT and it is good enough proof to highlight here. Theorem 3.4.5. Let {J,} be a sequence of measurable functions converging point-wise ae, to f. If|fi(z)| < a(x) such that 9 € L'(R") then sim [n-s1=0 sa f= fs Proof. Note that |f(x)| < g(x), since f is point-wise limit of fy. Let By. = {a | € By(0) and g(x) < k}. Note that g is non-negative. Set ge(x) = 9()x1 (2) is measurable, integrable and non-negative. Also, 94(x) < gx+(2) and gi(2r) converges point-wise to g(x). By MCT, we have tia fn fo ‘Thus, for any given ¢ > 0, there exists a K€ N such that £ 2 esi vk > K. For the # obtained above, fi restricted to Ex is uniformly bounded by I. Since f(x) + f(x) point-wise a.e. on Ex, by BCT, there exists a K€ N In particular, Jp ii-n<§ bow [ims [it-se [ean [n-neefs £ of £40k >K pha Vee K ‘We have IA A Hence lim, J | fi — f| > 0. °CHAPTER 3. LEBESGUE INTEGRATION 70 The idea of the proof above actually suggests the following result, Proposition 3.4.6. Let f € L'(IR"). Then, for every given > 0, (i) There exists a ball BCR” of finite measure such that lfl 0 such that [ [fl <2 whenever p(B) < 6. ie Proof. Let g(x) =| f(z)| and hence g > 0. (i) Let B, = B,(0) denote the ball of radius k centred at origin. Set 9x(2) == 9(x)xa,(x) is measurable and non-negative. Also, g(x) < 9x+1(x) and g,(z2) converges point-wise to g(x). By MCT, we have lim = wm [an fo ‘Thus, for the given © > 0, there exists a A € N such that [|f|— Sl flvn, K. Hence, o> [U-xelt= fin veo (fi) Let By s= {x | |f(x)| K. For any E € L(R"), fe = fea + fm [o-m- fix < [(o-m)=hw2) Choose 6 < ¢/2K and p(E) < 5. Then, Jas [(o-on+ Kut) <5 + Ke lACHAPTER 3, LEBESGUE INTEGRATION 7 a The first part of the proposition above suggests that for intograble fune- tions, the “integral of the function” vanishes as we approach infinity. How- ever, this is not same as saying the function vanishes point-wise as |r| ap- proaches infinity. Example 3.7. Consider the real-valued function f on R ro={5 eZ 0 rem f=Oae. and fy. f =0, however lim, soo f(z) = +00 Can we have a continuous function in the above example? Example 3.8. Let f = SP bxje.e+1/x)- Note that Ss 1 Yo | pea) dr = Sg < te ia 1 ‘The integral of f is / =e a and is in L1(8"). But limsup f(x) = +00. In fact, this is true for a continuous function in L'(R) (extend the f contin- uously to R). However, for a uniformly continuous function in L*(R) we will hhave limjs|-so0 f(x) = 0. The absolute continuity property of the integral proved in the Proposition above is precisely the continuity of the integral Exercise 68. Let f € L'((a, )) and F(x) = | fat. Then F is continuous on [a,b]CHAPTER 3, LEBESGUE INTEGRATION 72 Proof. Let x € (a,b). Consider |F@) - F@)| = [ros [vow Since f € L'(la,b)), by absolute continuity, for any given ¢ > 0 there is a 6 > 0 such that for all y € B= {y € [a,8] | |x — y| < 4}, we have |F@) — Fi M} has outer measure zero. The infimum of all such M is said to be the essential supremum of f. The class of measurable essentially bounded function is denoted by L*(E).CHAPTER 3, LEBESGUE INTEGRATION 73 Exercise 70, Show that L?(E) forms a vector space over IR (or C) for 0.< PS. Proof. ‘The case p = oo is trivial. Consider the case 1

0. Let 0

0. Assume wlog that a < b (else we swap their roles). ‘Thus, a+b < 2b=2max(a,b) and therefore (a+b) < 2 < 2°(a? +). Using this we get (\FC@)L + Lo(@)|)” < QP F(a)? + 2lgla) Therefore, [iterator 0.CHAPTER 3. LEBESGUE INTEGRATION 74 Exercise 71. Show that for each scalar A € R (or C), llAflle = LAlllflp VF © L°(2). (This is the reason for having the exponent 1/p in the definition of norm) Note that the norm of zero fumetion, f = 0 is zero, but the converse is not true, Exercise 72. For each 0 < p < co, show that |f\|p = 0 iff f =0 a. Observe from the above exercise that the “length” we defined is short of being a “real length” (usually called semi-norm). In other words, we have non-zero vectors whose length is zero. To fix this issue, we inherit the equivalence relation of M(R") defined in Definition 2.6.5 to L*(R"). Thus, in the quotient space L?(E)/ ~ length of all non-zero vectors is non-zero, In practice we always work with the quotient space L”(B)/ ~ but write it as 1*(B). Hence the remark following Definition 2.6.5 holds true for L?(E) (as the quotient space) It now remains to show the triangle inequality of the nom, Proving tri angle inequality is a problem due to the presence of the exponent 1/p (which was introduced for dilation property). For instance, the triangle inequality is true without the exponent 1/p in the definition of norm, Baercise 73. Let E € £(R"). Show that for 0

0. Assume wlog that a fle + llally Exercise 74, Show that for 0

0. Thus, when p > 1 the maximum is 1 and we have the triangle inequality of the norm for p > 1, called the Minkowski inequality. To prove this, we would need the general form of Cauchy-Sehwarz inequality, called Hélder's inequality. For each 1 < p < co, we associate with it a conjugate exponent q such that 1/p + 1/q = 1. If p = 1 then we set @ = 50 and vice versa Theorem 3.5.4 (Hélder’s Inequality). Let F< £(R") and 1

0. Set fi = Gif € I7(#) and gi = gig € L4(E) with ||/illp = llgilla = 1. Recall the AM-GM inequality (ef. Theorem D.0.22), yyseet (35.2) P Using this we get ben < pine +e LF) iar Til = wif (x)|” + TG lo(a) « ir ae aor Now, integrating both sides w.r-t the Lebesgue measure, we get fis < Wlolole (SEplvm+ Totals) aire!" state Wfilolalle Hence fg € L1(E) oO lA Remark 3.5.5, Equality holds in (3.5.1) iff equality holds in (3.5.2) which happens, by Theorem D.0.22, iff HOt = WAP for ae. x € BE. Thus, vl Tally * [f(@)IP = Mg(2)|? where A= Le Exercise 75. Show that for 0 < p <1, f € LP(E) and g € L*(B) where the the conjugate exponent of p (now it is negative), [foils = Wfllollglle Theorem 3.5.6 (Minkowski Inequality). Let B € £(R") and 1

+ llolle The triangle inequality fails for 0 < p <1 due to the presence of the exponent 1/p in the definition of | fllp- Thus, for 0 < p <1, we also have the option of ignoring the 1/p exponent while defining |/f||p. Define the metric 4d, : LF(R*) x 1°(R") = [0, 00) on L?(R") such that dy(f, 9) = ll — ally for 1< p< cand dy(f,9) = lif — all? for 0< p <1 Exercise 77. Show that d, is a metric on L?(R") for p > 0. Definition 3.5.7. Let E € £(R"). We say a sequence {J,} converges to f in L*(E), p> 0, if dp( fi, f) 40 as k + 00. Exercise 78, Show that f, — f in L? then then f, > f in measure. The converse is not true, in general. Moreover, L? convergence does not imply almost uniform or point-wise a.e. However, they are true for a subsequence. Exercise 79. Uf fi, converges to f in LP(E) then there exists a subsequence {Jus} of {fi} such that fi,(x) > f(x) point-wise for ae. x € B. Theorem 3.5.8 (Riesz-Fischer). For p > 0, L?(R") is a complete metric space.CHAPTER 3. LEBESGUE INTEGRATION 78 In general, we ignore studying L? for 0 < p < 1 because its dual’ is trivial vector space. Thus, henceforth we restrict ourselves to 1 < p < 00 Proposition 3.5.9. Let E € £(R") be such that (B) < +00. If

0 there is a6 > 0 such that ju(Hs) > 6 where Ess {x € B| |f(2)| > |[\loo —e}. Swill be introduced in a course of functional analysisCHAPTER 3, LEBESGUE INTEGRATION 79 ‘Thus, FIle = f LF? = (Ulf lles ~ €)? 6 Bs Since 64” + 1 as p00, we got lim inf, +2 I|fllp > llfllee — ©. Since choice of ¢ is arbitrary, we have linty yas» = Ilf lee a EBrercise 81. Let {fg} be a sequence of functions in L® (IR). Show that [lf — f\loe + 0 iff there is a set E € £(R”) such that 4(H) =O and f, > f uniformly on £°. Proof. It is enough to show the result for f = 0. Let |[filleo — 0. Let Ex = {x ER" | |fel(x)| > |lfelloo}. Note that (Ex) = 0. Set B = UP Ee By sub-additivity of Lebesgue measure, (#) = 0. Fix ¢ > 0. Then there is a K € N such that for all k > K, || filloo < €. Choose any x € E*. Then \fe(z)| < \\felloo for all k. Thus, for all x € E* and k > K, |fi()| < e. Thus, fy 3 0 uniformly on F°. Conversely, let B € £(R") be such that :(E) = 0 and fy > 0 uniformly on E*. Fix c > 0. For any x € B°, there exists a K € N (independent of 2) such that |f,(2)| K. For each k > K, Ilalloe 85 suPzere|fel*)| = sup |fela)| < Hence || fx \|oo — 0. o We now prove some density results of L? spaces. Recall that in Theo- rem 2.6.11 and Theorem 2.6.12 we proved the density of simple function in M(B”) under point-wise a.c. convergence. We shall now prove the density of simple fumetions in L? spaces. We say a collection of functions A c L? is dense in L? if for every f € LP and ¢ > 0 there is a g € A such that If -allp <= Theorem 3.5.11. Let E € C(I"). The class of all simple functions are dense in LP(E) for 1

0. By Theorem 2.6.11 we have an increasing sequence of non-negative simple functions {¢,} that converge point-wise a.c, to f and dy < f for all k. Thus [eu(e) — fa) < 2\F(@)PCHAPTER 3. LEBESGUE INTEGRATION 80 and by DCT we have lim dx — f18 = am, f |oe—P > 0. im, Be, For an arbitrary f € L?(E), we use the decomposition f = f* — f~ and triangle inequality to prove the result, a Theorem 3.5.12. Let E € L(R"). The space of all compactly supported continuous functions on E, denoted as C,(E) is dense in [°(E) for

0 there is an open (bounded) set such that > F and j(M\ F) <¢/2. Also, by inner regularity, there is a compact set KC F such that p(F\K) <¢/2. By Urysohn lemma there is a continnous function g : E —> R such that g = 0 on E\O, 9 = lon K and 0 <9 <1on®\K. Note that 9 € C,(B). Therefore, lice a= file —aP = [ee 9 < (0 #1) |» bv a Example 3.10. The class of all simple functions is not dense in L*(E). The space C,(E) is not dense in £%(E), but is dense Co(B) with uniform norm, the space of all continuous function vanishing at infinity. Theorem 3.5.13. For 1 < p < 00, L7(R") is separable but L™(R") is not separable. 3.6 Invariance of Lebesgue Integral Recall that in section 2.5, we noted the invariance properties of the Lebesgue measure, In this section, we shall note the invariance properties of Lebesgue integral.CHAPTER 3, LEBESGUE INTEGRATION al Definition 3.6.1. For any function f on R” we define its translation by a vector y € R", denoted r,f, as tyf(z) = fe — 9). Similarly, one can define notion similar to reflection SF) = f(-2) Also, dilation by > 0, is f(Ax) Exercise 82. Show that if f € L'(R") (i) (Translation invariance) then 7,f € LMR"), for every y € R", and ff=Srf (ii) (Reflection) then f € L'(R") and f f = ff. (iii) (Dilation) and A > 0, then f(Ar) € L1(R") and f f =" f (Ax).CHAPTER 3. LEBESGUE INTEGRATION 82Chapter 4 Duality of Differentiation and Integration The aim of this chapter is to identify the general class functions (within the framework of concepts developed in previous chapters) for which following is true: 1. (Derivative of an integral) ar =f a= 16) 2. (Integral of a derivative) : [ roar=s0)- 10) ‘We shall attempt to answer these questions in one-dimensional case to keep our attempt simple. In fact answering both these questions have far reaching consequences not highlighted in this chapter. Let f : [a, 5] > R be Lebegue integrable and define Fe = [sow x € (a,b) Answering first question is equivalent to saying F"(x) = f(x). Note that for a non-negative f, F is a monotonically increasing function. This observation motivates the study of monotone functions in the next section. 83CHAPTER 4. DUALITY OF DIFFERENTIATION AND INTEGRATION 84 4.1 Monotone Functions Recall that a fimction is said to be monotone if it preserves a given order A function f is said to be monotonically increasing if f(x) < f(y) whenever x 0 and for every x € KE, there exist a ball B= V such that x € B and y(B) 0, there is a finite disjoint sub-collection {B,} C V such that 1 (B\UB) 2, we pick a ball By such that By (UE}B,) = 0 and (By) > r2/2", where re = sup{u(B) | BOUETB, = 0}. Bey, Note that ry is finite for all &, since rz < u(@) < +00. If the set over which the supremum is taken is an empty collection for some k, ie., there is no BEV such that BNUE}B, =, then we already have EF C USB; and we are done. Otherwise, we have a countable disjoint collection of closed ballsCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 85 By such that (Bx) > 14/2". Also UgBk © M and hence, by monotonicity and additivity of 1, we got $00 > p() > e(UE Be) = > w(Be). ‘Thus, for any given ¢ > 0, there is K € N such that Y uey K. Suppose, for every i > K, B.0 Bi = 0, then (Bz) K. Let [be the smallest i > K (or first instance) when B, mects By Hence (Bz) < ry < 2"(By). We claim that B, C 5B). Let y € Be Bi Then, |x — y| < 2r,, where r, is the radius of B,. Also, |y—2)| Df (x) and D~ f(x) > D_f(x). If D*f(x) = D,J (x) # £00 then we say the right-hand derivative of f exists at x. Simi- larly, if D~ f(x) = D_ f(x) # £00 then we say the left-hand derivative of f exists at x. We say f is differentiable at x if D* f(x) = Dy f(x) = D~ f(x) = D_f(x) # £00 and f'(x) = D* f(x). Observe that for a increasing function the Dini derivatives are all non-negative. Exercise 83. Show that, for a given f : R R, if g(a —J(—2) for all 2 ER then Dt g(x) = D f(x) and D_g(x) = Dy f(z) Proof. Consider hy Degle) = tinsyy HEA ob) - =fl-a~h)+ fez) - F-, = limsup ik =D f(-2) Similarly, D-o(x) = limint a2) = oe ~A) = Tnint chee) + Ce +h) Lp. pa)CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 87 Theorem 4.1.4. If f : [a,b] IR is a increasing function then f is differ- entiable a.e. in |a,6). Moreover, the derivative f’ € L¥({a,6]) and [rs 70) - Proof. To show that f is differentiable a.e. in [a,b}, it is enough to show that D*f(x) < oo for ac. in [a,8) and D*f(x) < D_f(x) for ac. in [a,b]. Because, by applying this result to 9(y) = —f(-y), we would get D- f(-y) < Dy S(-y) for ac. —y € [a,0). Hence, D* f(x) < D_f(x) < D f(x) < D- f(x) and all are equal since D, f(z) < D* f(x). Thus, it is sufficient to show that the set E = {x (a,b) | D* f(x) > D_f(2)} has outer measure zero. In fact, a similar argument will prove the result for every other combination of Dini derivatives. Let p,q € Q such that p > q and define Enq = {2 € [a,b] | D*f(z) > p> q> D_f(2)} Note that B = UpyoBnq. We will show that (Eq) = 0 which will imply that *() =0. To begin we assume a non-empty Ey has j"(Epq) £0, for a fixed p,q € @ such that p > g, and arrive at a contradiction. We construct a Vitali cover of Eg. For any given © > 0, by outer regularity, there is an open set 2D E,,, such that y(Q) < w*(E,4) + €. For each « € E,,4, since Q is open, there is an interval [x — h, x] © such that (x) ~ f(z —h) < gh. The collection of all such intervals, for cach 2 € Epq, forms a Vitali cover of Byq. Therefore, by Vitali covering lomma, we have finite disjoint sub- collection {J,}7" from the Vitali cover such that # (Epa \URLL) 0 such that [y,y +k] ¢ J; and Sly +k) — f(y) > vk. By Vitali covering lemma, there is finite disjoint collection of intervals {J,}4 cach contained in J, for some i such that W(A\U GS) <= Set B= AN(Uf,J5) and A= BU(A\Uf4J5). Hence, (A) < "(B) + Therefore, we have ‘ Fu) +k) — Fy) > PIO = PD HG) = palUyJ;) > pue(B) > plu(A) ~e) > plu’ (Bya) ~ 22) mh | Now, for each fixed i, we sum over all j such that J, C J, to got the inequality DY lw +s) = Aas) $ fe) = fle —h) due to the increasing nature of f and disjointness of Jj. This implies that plt*(Epq) — 22) < q(u*(Eyq) + €). Since e is arbitrary, we have py"(Epq) < qe'(Ep,q) which will contradict q < p unless :*(Ep¢) = 0. Consequently, 4°(E) = 0 and f is differentiable a.c. in {a,6]. Hence f"(z) is defined a.c. in (a, 8). Set ye(x) = k (f(z + 1/k) ~ F(x) such that for all x > b, f(x) = f(b)!, Note that g(x) > f'(x) ae. in [a,b] Thus, f’ is measurable, due to the measurability of g, which follows from the measurability of f, a consequence of being a increasing function. Also, since f is increasing g, are non-negative and hence f" is non-negative. Using , for any ¢ > f(b), for all «> 6, and obtain ¢ ~ f(a) bout f(U) ~ f(a) is the best bound one can obtainCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 89 Fatou’s lemma, we have fr timint oe = limnint («fires 1/k) — tnuint (: / a) of “1) ese o+/h mint (/ fe) -*f reo) (F constant for a > 8) = limint (1-4 v) J(8) ~ f(a) (Fis increasing) IA [1@) IA a Note that the above result also holds true for decreasing functions. Also, observe that for any two increasing fimctions their sum and difference are also differentiable a.c., but the difference is not necessarily increasing or do- creasing, We wish to classify this class of functions which is the difference of two increasing functions. 4.2 Bounded Variation Functions The problem of finding area under a graph lead to the notion of integration, An equally important problem is to find the length of curves. Let 7 denote a continous curve in a metric space (X,d). Let the continuous function +: [a,8) + X be the parametrisation of the curve ~y with parametrised variable t € [a,8]. Let P be the partition of the interval (a, 8], a= to <1 < Shab Definition 4.2.1. The length of a curve +: [a,6] » X on a metric space (Xd) ts defined as : La) -ap {Soa ted}, ‘here the supremum is taken over all finite number of partitions P of [a, 6} If L(y) < +90 then the curve + is said to be rectifiableCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 90 ‘The length of the curve is defined as the supremum over the sum of length of all finite number of “line segments” approximating 7. If X is the usual Euclidean space with standard metric then the length of the curve has the form : uP {= I(t)? - exten} ‘at A interesting questions one can ask at this juncture is: under what conditions on the function 7 is the curve + rectifiable? ‘The length of a curve definition motivates the class of bounded variation functions. Definition 4.2.2. Let f :|a,0] > R(C) be any real or complex valued func- tion? Let P be a partition of the interval [a,b], a= 20 <2 <... Say = 5. We define the total variation of f on [a,b], denoted as V(f;(a,0)), as ‘ Vs [a,8)) = sup {= Mle) — Sa oi} We say f is of bounded variation if V(f;[a,6]) < +00 and the class of all bounded variation function is denoted as BV ((a,6)) LO) Comparing the definition of bounded variation with curves in C, we ex- pect that any curve + is rectifiable iff +: (a, 6] + C is a function of bounded variation. Example 4.1. Every constant function on (a, 6] belongs to BV ([a, 6) and its total variation, V(f;[a,6)) = 0, is zezo. Lemma 4.2.3. For any function f, V(f;(a,6)) tion on [a, b} 0 iff f is a constant func- Example 4.2. Any increasing function f on [a,b] has the total variation V(f;(a,6}) = f(b) — f(a). Consequently, if f is a bounded increasing fune- tion, then f € BV((q,6]). For any partition P = {a = 29 <... < ty = 6}, we have : : SMif@) = fled = YU) fle) = Fes) ~ Hla) + fle) ~ fer) bo | flows) ~ Flea) | 0) ~ flea) Fb) #0) jous as required for a curve“)CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION OL Thus, V(f;[a,b)) = f(b) — f(a). Similarly, if f is decreasing on {a,8] then Ff & BV((a,6)) and V(f; [a,5)) = fla) — f(b). The above example is very important and we will later see that every function of bounded variation can be decomposed in to increasing bounded. functions, a result due to Jordan. Example 4.3. The Cantor function fe on [0,1] is increasing and hence is BY ((0,1]). We already know fe is uniformly continuous (cf. Appendix B). Esample 44. Any differentiable function f = [a,8] + R such that f' is bounded (say by C) is in BV((a,8]). Using mean value theorem, we know that Ye,y € [a,0] and 2 € [x,y] Since the derivative is bounded, we get | f(x) — f(y)| < lx — yl. Thus, for any partition P = {a = 2 <... < a4 = b}, we have ‘ ‘ Sl fle) — Flea)] < CY as — aul = (ba) ‘a The above example is a particular case of the class of Lipschitz. functions on [a 6) Definition 4.2.4. A function f : [a,b] + R(C) is said to be Lipschitz on (a,8] sf there exists a Lipschitz constant C > 0 such that [f(@) ~ f(y) < Cle — yl ¥x,y € (a, 6 The space of all Lipschitz functions is denoted as Lip((a, )) Exercise 84. Any Lipschitz function is uniformly continuous, Lip([a,8]) C C((a, 8) Exercise 85. Every Lipschitz function is of bounded variation and V(filab]) SOF) — fea)| > n Choose x9 =a. Let 2; be an irrational between a and 6, Choose x2 to be an rational between x; and 6. Proceeding this way till x,42 = 6, we will have a partition P whose successive points, excluding a and 6, alternate between rational and irrational. Therefore, ne Vofsla,b)) = SOG) = flea)| Wv dl = [Ple2) — f(r) +--+ If 2os1) = Send [1-04 Jo-1+...=n. Thus, V(f;[a, )) = 20. a We have already seen that the Cantor function fe, which is (uniformly) continuous, is of bounded variation. But we do have functions which are continuous and not of bounded variation Erercise 88. Show that the following continuous function on (0, 1} xsin(I/z) 2 #0 0 r=0 J() do not belong to BV((0, 1]). More generally, ale) = a eee is in BV((0, 1}) iff a > 8.CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 93 Proof. For each k € N, note that (i) -° 5 (stan) =e If we choose the points in our partition alternating between 1/kn and then, for each m € N, aa V(Fila.b)) > Y)J0 m= Gp int ape Hence, V(fila,b)) 2 For each k € N, note that Jae atr( If we choose the points in our partition alternating between 1/(kx)"/" and echo then, for each me N, Tern fay Vesilae) > fal ey Ben + 7/2) (1 Hence, 7 ~ 1 Vile) > > Tap The series on the right converges iff a/@ > 1. Thus, g € BV([0,1)) iff a>s. Lemma 4.2.5. Let f : a,0] + R be a given function. Let P denote the parti- tion P = {a,21,...,%»—1,0} of the interval [a,b] and P! = {a, yi, ,Ym—1,0} be a refinement of P, i.c., PC P!. Then Vile) — Fal < VF) = Fo) FCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 4 Proof. We first. prove the result, by adding one point to the partition P and then invoke induction. Let y € P'. If y= 2, for some i, then the partition P remains unchanged. If y # 2, for all i then y © (24-1, x4) for some ke {0,1,...,n}. Consider, Sifle) — Fay = SO) ~ f@a)| + | l@e) — fe) + a + ie) = sey) a » = Sie) - feat YO vd - Fed) + [F(ex) ~ Fu) + FW) ~ Flena)| < Sled — flea) + Fee) - FO! + IFW) = Flamadl + D> [fle = fl@-s)| rst = YM) = f2)|. (by relabelling) Similarly, adding each point of P’ into the extended partition of P, we have our result, ao Eaercise 89. Show that BV{((a,8]) forms a vector space over R. Also, if fg € BV((a,6)) then (fg are in BV((a,6)) Gi) f/g € BV((a,0)) if 1/9 is bounded on [a,b] Theorem 4.2.6. Let f : {a,0] -» R be a function and let c € (a,b). If f belongs to both BV ({a,¢}) and BV((c, 6) then f € BV([a,6)) and VF lab) = VUS fa, €)) + VU le, 8).CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 95 Proof. Let P be any partition of [a6] and P’ = PU {c}, relabelled in in- creasing order. P’ being a refinement of P, arguing similar to the proof of Lemma 4.2.5, we get Vie) = fed) < Vie) -s@)I Sled = fad) + I) = fleas) ViFslae) + VUF lt) Hence, V(f; [a,6)) < V(f;[a,¢]) + V(f;[c, 4). On the other hand, let P; and Py be a partition of [a, ¢] and [e,8], respectively. Then P = P; U P, gives a partition of (a, 6]. Therefore, SV @)-fedl+ Ved - fea m Fr A IA Sie) - se) F < V(fila,0)) The above inequality is true for any arbitrary partition P, and P; of [a,¢) and [c,B], respectively, Thus, Vifilad) + VU (68) < VF (a, 8) and we have equality as desired a Exercise 90. Show that if f € BV((a,6]) then f € BV((c,d]) for all sub- intervals [¢, d] ¢ [a,b] Lot f : > R(C) be any real or complex valued function. Let BVige() denote the class of all f: RR such that f € BV ((a, b)) for all [a,6] cB. Definition 4.2.7. We define the total variation of f on IR, denoted as V(F:B), as VR) = aap VF; {a,6)), where the supremum is taken over all closed intervals [a,b] CR. We say f is of bounded variation on R if V(f;R) < +00 and denote the class as BV(R) Note that BV (IR) © BVigc(R) and the inclusion is strictCHAPTER 4. DUALITY OF DIFFERENTIATION AND INTEGRATION 96 Example 4.5. The function f 1 aj t#l sa) {5 r=l1 belongs to BV(0,1) but do not belong to BV ({0, 1)). Definition 4.2.8. For f © BV((a,b}), we define its variation function, {re [az] Ye € (0,8) VO= 19 ne Lemma 4.2.9. The variation function Vj(x) corresponding to a function J € BV{(a,8)) is an increasing function. Proof. Let x,y € [a8] be such that x Vj(x) and equality holds when f is constant on [zy] a ‘Theorem 4.2.10 (Jordan Decomposition). Let f : [4,6] > R be a real valued function. Then the following are equivalent: (i) f © BV((a,6)) (ii) There exist two increasing functions fy, fr : (a,0] -¥ R such that f = h-hk Proof. (ii) implying (i) is obvious, because any inereasing function is in the vector space BV ((a, b]). Conversely, let us prove (i) implies (fi). For a given f € BV((a, 8)) we know that V;, the variation function, is increasing in [a, 8 Set fi = V; and fo = V; — f. It only remains to show that fs is increasing let 2, y € [a,)] be such that x < y. Consider, Sly) ~ fale) = Vylo) ~ Flu) — Vyla) + #0) Vy(u) — Vp(@) — (f(y) — Fl) V(F[2,u]) — F@) - £@)) V(f 2, u)) — f(y) ~ f(@)] = 0. Thus, fy is increasing and f = fy — fo a vCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 97 Exercise 91. Show that in the above theorem one can, in fact, have strictly increasing functions fa, fo Theorem 4.2.11 (Lebesgue Differentiation Theorem). If f € BV ((a,6)) then f is differentiable a.e. in [a,b] and the derivative f’ € L'(\a,0)). Fur- ther, : fr Proof. We know the result for a increasing function and, by Jordan decom- position, any f € BV((a,6)) will satisfy the required result. Prove it as an exercise, oa 0 on a set B of non-zero measure, Since £ has non-zero Lebesgue measure, there exist a closed set PC E such that (I) > 0. Set 2 = [a, 6] \P. Since F is identically zero, we have , o=Fe=[ soa [y+ fy [ s--[r<0 Let = UR, (a;,0,) is a disjoint union of open intervals. Then, oe fr-¥fs Therefore, for some k € N, we have 04 [r= [4 [y= Pe Fe ‘Thus, either F(b,) # 0 or F(a,) # 0 which contradicts the hypothesis on F. a Thus, Theorem 4.3.3. Let f € L¥(|a,)) and ce R. Set F(e)=e+ [ F(t)at. Then ¢ = F(a) and F" = f ae. on (a,b)CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 99 Proof. ‘The fact that ¢ = F(a) is obvious, By Lemma 4.3.1, we have F € BV/(a,6)). By Lebesgue differentiation theorem, F is differentiable a.c. It is required to show that F’ = f a.e, We prove by cases. First let us assume f is bounded, ie., || flo < 00. Extend F as F(x) = F(b), for all x > b. Set re f(t) dt. Note that g(r) 3 F'(r) ae. in [a,b]. Since f is bounded, g,’s are all uniformly bounded and supported inside [a, 8]. Using BCT, for any d € [a, 8), we have fr = in [‘s = im (« [resi -« ["#00) = an (ef Re -+ f'n) = (ef ro-2[°" ee) Consider, for any ¢ € [a,b], g(x) = k (F(x + 1/k) — F(x) Foyt [" rover < pe [ |F(e) — F(2)| dx ‘ < sup Fle) — Ft@)) oy By continuity of F (cf, Exercise 68), F(e +h) + F(e) ash 0. Hence, we have Therefore, Thus, for all d € [a, 6), we have [w-n-oCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 100 and by Lemma 4.3.2, we have F’— f =0 ae. on [a,), ie, FY = fae. on (a, It now remains to prove the result for a unbounded function. Without loss of generality, we assume f is non-negative. Then F is increasing on [a,b] and, by Theorem 4.1.4, we have + [ F(x) dx < F(b) — F(a) Let fe be the truncation of f at k level, ic., _[7@) se) fel) acc. Consequently, F’(x) > f(z) ae. Thus 5 5 [ Pi(a)de > [ f(2)dz = F(b) ~ F(a) and we have equality above, since other inequality holds as noted above. Therefore, [er @-seyae-o and for F” — f > 0, integral zero implies that F'(z) = f(x) ae. on fab}. OCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 101 Recall the definition of derivative of F, him PE +A) = FQ) mo h Fe) Thus, we may reformulate the demand F"(x) = f(z) a.e. on (a, 6] as mo lint [Foes for ac. x € [a,b] Note that the integral (along with the fraction) on the LHS is the “average” or “mean” of f over [x,+h] and the equality says that the limit of averages of f around a interval I of x converges to the value of f at x, as the measure of interval I tends to zero. The theorem above validates this result for all f € L' in the one dimension case. This reformulation, in terms of averages, helps in stating the problem in higher dimensions. ‘Thus, in higher dimension, we ask the question: For all f € L'(R"), do we have 1 n Pa amy J 1 ae= fe) fora. 2 eR! where B is any ball in RY containing x 4.4 Absolute Continuity and FTC In this section we wish to identify the class of functions f for which : J fede 16) ~ fo) This is the second question we hoped to answer in the beginning of the chap- tor, Note that if f ¢ BV/(|a,})) then, by Lebesgue differentiation theorem, f is differentiable and the derivative is Lebesgue integral. So the question reducing to asking: For any f € BV(la,6)), do we have [ * Pa) de Unfortunately, the answer is a “No”, as seen in the example below. F(b) = Fla)?CHAPTER 4. DUALITY OF DIFFERENTIATION AND INTEGRATION 102 Example 4.6, Recall that the Cantor function fer € BV([0,1)), since for is increasing. Outside of the Cantor set C, fe is constant and hence f?, = 0 a.e. on [0,1]. Therefore, but fo(1) ~ fo(0) This motivates us to look for a sub-class of bounded variation functions for which fundamental theorem of calculus (FTC) is true. Definition 4.4.1. A function f :|a,b) > R is said to be absolutely contin- uous on (a, 6] if for every © > 0 there exist ad > 0 such that STi) - Se)| 0 be given. By Proposition 3.4.6, we have 6 > 0 such that [ If 0 be such that for any disjoint collection (Gnite or countable) of subintervals {(2,,y,)} of [a, 8] with Siw-al<6 we have Sif) - fey) <1 Let M denote the smallest integer such that (b— a)/8 R is both absolutely continuous and singular, then f is constant. Proof. It is enough to show that f(a) = f(c) for all ¢ € (a, 6]. Fix ac € (a,b). Due to the singular nature of f, we have a measurable set HC (a,c) such that (EZ) =e—aand f'(z) =0 on B. Due to the absolute continuity of f, for every ¢ > 0 there exist a 6 > 0 such that c DViiw - fel <5 for any disjoint collection (finite or countable) of subintervals {(2,,y,)} of [a,b] with Siw-al<é We now construct a Vitali cover for E. For each x € E, f is differentiable and derivative is zero. We choose the interval [2, x +h] C [a,) such that € sh Yea) By Vitali covering lemma, we can find a finite disjoint collection of closed intervals {J; = [x., 15 + hiJ}f such that w(E\ URAL) <6. Note that [a,c] is an interval and {2,,r; + hj}. for all i = 1, disjoint collection of intervals in (a, c]. Thus, relabelling as {9 = haya, t2 1 hay... 2k, te + hee = te41} and setting ho = 0, we have that eh) -—F@)| < k Dies — (sb dl < 6 =o Hence, by absolute continuity, we have Sean) — fet hd <5 ‘=0CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 105 Now, consider k : (FC) ~ Fl@)| = | Flea) ~ Flee hi) + SO Fle +h) ~ Fm) Since the choice of ¢ is arbitrary, we have |f(c) — f(a)| = 0 and f(c) F(a) a Theorem 4.4.5. If f : [a,b] +B then the following are equivalent: () f € AC((a,b)) (ii) f is differentiable a.e.,f/ € L'{{a,8]) and se) se) = [Fa ee tad Proof. We first prove (ii) implies (i). Let © > 0 be given. By Proposi- tion 3.4.6, since f’ € L'({a,6]), there exists a 6 > 0 such that [ I l, |f(ys) — f(a2)| < ¢. Consider Duo - se = Df" roa < Df vice Conversely, let f € AC{(a, )) then f € BV([a,8)). Thus, by Lebesgue differentiation theorem, f is differentiable a.c. and f" € L'((a,)). Define Fx) = [ fatCHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 106 By Example 4.8, F € AC{(a,8)) and hence g = f — F € AC((a,6)). By ‘Theorem 4.3.3, we have F=f’ ac. on (a,b). Thus, g = 0 a.c., hence g is singular. Therefore, g is constant, g = ¢ and se)=cxRty aes [roa res = [roa a The implication (ii) implies (3) is a kind of converse to Lebesgue differ- entiation theorem (with additional hypothesis). In fact, the exact converse of Lebesgue differentiation theorem, i.e., ’ exists and J" € L* implies that J € AC © BY, is tine, but requires more observation, viz. Banach-Zaretsky theorem. Corollary 4.4.6. If f € BV(|a,)) then f can be decomposed as f = fat Fe where fi € AC{(a,6)) and f, is singular on [a,b]). Moreover, f and f, are unique up to additive constants and f fi(é)dt = fal) Proof. Since f € BV(\a,6)), by Lebesgue differentiation theorem, f' exists and f’ € L¥((a,6)). Define fem f i"(2) de By (b), fa € AC((a,0)). By, the derivative of an integral we have, f= f! a.e. on (a,b). Define fy := f— fa. Thus, f! =O... and hence f, is singular. Let f = g-+h where g € AC((a,b]) and his singular. Then, ¢+h = futfs Hence 9 ~ f. = f. — h. LHS is in AC((a, 6) and RHS is singular. Therefore they must be equal to some constant c. g = fz +c and h = f,—c. Exercise 96 (Integration by parts). Show that if f,g € AC((a,6)) then , [ feoteraes [ peate)ae = 191) - flea)CHAPTER 4, DUALITY OF DIFFERENTIATION AND INTEGRATION 107 Theorem 4.4.7. If f € AC{(a,6]) then VCs (a, 8) i'l Theorem 4.4.8. Let f € AC{(a,6)). Then w(f(E)) = 0 for all E © [a,b] such that :(E) = 0. Proof. Let E © (a,6) be such that jx(Z) = 0. Note that we are excluding the end-points because (f(a), f(b)} is measure zero subset of f(). Since J € AC((a,6)), for every given © > 0, there exists a 4 > 0 such that for every sub-collection of disjoint intervals {(xi,y,)} C [a,b] with O,{y — ai) < 4, we have 3, [f(y.) — f(,)| < ¢. By outer regularity of E, there is an open set 2D E such that p(M) < 6. Wlog, we may assume 2 C (a,b), because otherwise we consider the intersection of with (a,8). But © = Ui(ni,4i), a disjoint countable union of open intervals and Yu- HO) <6. Consider, WF) S WED) = WFC) = w(US(e.¥) Seem) Let cj, d; € (x;,y:) be points such that f(c;) and f(d;) is the minimum and maximum, respectively, of f on (x;,y)). Note that |d, — c| < |yy — | and hence 33, |d; — cj] < 6. ‘Then, Le Ee.) = Oia) - fel The Cantor set C is the intersection of all the nested G's, C= oC Lemma B.0.12. C is compact. Proof. C is countable intersection of closed sets and hence is closed. CC (0, 1] and hence is bounded. Thus, C’ is compact. ao ‘The Cantor set Cis non-empty, because the end-points of the closed intervals in C,, for each i= 0,1,2,..., belong to C. In fact, the Cantor set cannot contain any interval of positive length. Lemma B.0.13. For any x,y € C, there isaz ¢C such that <2 C. o We show in example 2.5, that C’ has length zero. Since C’ is non-empty, how ‘big’ is C? The number of end-points sitting in C are countable. But C has points other than the end-points of the closed intervals C; for all i. For instance, 1/4 (not an end-point) will never belong to the the intervals being removed at every step i, hence is in C. There are more! 3/4 and 1/13 are all in C which are not end-points of removed intervals. It is easy to observe these by considering the ternary expansion characterisation of C. Consider the ternary expansion of every x € [0,1], 4 where a; =0,1 or 2 ‘The decomposition of x in ternary form is not unique'. For instance, 1/3 = O.1s = 0.022222. ..3, 2/3 = 0.2; = 0.1222...3 and 1 = 0.222...5. At the C; stage, while removing the open interval (1/3, 2/3), we are removing all num- bers whose first digit in temary expansion (in all possible representations) is 1. Thus, C; has all those numbers in [0,1] whose first digit in ternary expansion is not 1. Carrying forward this argument, we see that for each i, Cy contains all those numbers in [0,1] with digits upto ith place, in ternary expansion, is not 1. Thus, for any x € C, oe Lemma B.0.14. (is uncountable. Is O.aqyagas...3 where a; = 0,2. Proof. Use Cantor's diagonal argument to show that the set of all sequences containing 0 and 2 is uncountable. ao This is true for any positional system. For instance, 1 = 0.99999. .. in decimal systemAPPENDIX B, CANTOR SET AND CANTOR FUNCTION 115 Cantor Function We shall now define the Cantor function fe: C ++ [0,1] as, folt) = fe (=) =e Since a, = 0 or 2, the function replaces all 2 occurrences with 1 in the ternary expansion and we interpret the resulting number in binary system. Note, however, that the Cantor function fe is not injective. For instance, one of the representation of 1/3 is 0.0222...; and 2/3 is 0.2. Under fo y are mapped to 0.0111...2 and 0.13, respectively, which are different representations of the same point. Since fc is same on the end-points of the ved interval, we can extend fe to [0,1] by making it constant along the -d intervals. Theorem B.0.15. The Cantor function fo + [0,1] — [0,1] és uniformly continuous. Generalised Cantor Set We generalise the idea behind the construction of Cantor sets to build Cantor- like subsets of [0,1]. Choose a sequence {a,} such that ay € (0,1/2) for all k. In the first step we remove the open interval (a, 1 — a3) from [0, 1] to get Cy. Hence C; = 0,a)] U[1— a, I]. Let Ch:=(0,a;] and C? = [1 — a, 1]. Hence, C; = C} UG}. Note that Cj are sets of length a, carved out from the end-points of Cy. We repeat step one for each of the end-points of C} of length aya. Therefore, we get four sets Cp = [0,ara2] CF = [ay — ara2, ay), C$:=[l-a,l—a+aqa;] C$:= [1 — a2, 1] Define Cy = Ui ,C}. Each C} is of length ajax. Note that aay < a Repeating the procedure successively for each term in the sequence {az}, we get a sequence of sets Ck C [0,1] whose length is 2a,az...a,. TheAPPENDIX B, CANTOR SET AND CANTOR FUNCTION 116 “generalised” Cantor set C’ is the intersection of all the nested Cy’s, C = NFqCk and each C, = U}",C}. Note that by choosing the constant sequence a, = 1/3 for all k gives the Cantor set defined in the beginning of this Appendix. Similar arguments show that the generalised Cantor set C’ is compact. Moreover, C is non-empty, becaus intervals in Ck, for each k = 0,1,2,..., belong to C. the d-points of the closed Lemma B.0.16. For any x,y © C, there is az ¢ C such thatx [0,1 Thus, fe being uniform limit of continuous function is continuous and is the called the generalised Cantor function.Appendix C Metric Space The distance d between two sets is defined as d(H, F’) = inf Lemma C.0.20. If F is closed and K is compact such that PK = 0 Then d(F, K) > 0. Proof. Let x € K. Since FK = 0, ¢ F. Since F is closed and x € F*, here is a d, > 0 such that Bys,(r)F = 0. Thus, d(x, F) > 36,. Note or each x € K, the open balls Bys,() will form an open cover for , KC Uc Bos,(2) and the open cover does not intersect F. Since K is compact there is a finite sub-cover Bys,(z,) of K, for i = 1,2,..., We will show that d(F,K) > 6 > 0. Let x € K andy € F. Thus, for some i, 7 € Byg,(1%). Therefore, d(x, ,) < 26;. But, by our choice, we have d(x, y) > 38. Hence, 36; < (as, y) < das, x) + d(a,y) < 28 + d(z, 9), Therefore, d(x, y) > 6. Set 6 = min{d,,...,4,}. Then (x,y) >6>0. OAPPENDIX C, METRIC SPACE 118Appendix D AM-GM Inequality One of the most basic inequality in analysis is the AM-GM inequality. The arithmetic mean of a sequence of n real numbers, 2 = (2t1,:t2,..., tn), is defined as ie A(x) aoe The geometric mean of a sequence of non-negative real numbers « is given by, n\n Ge) ( ) (D.0.1) ia numbers is the Geometrically, th length of the si the n-dimensional cuboid whose sides are given by the n. geometric mean of n non-negative al to the area of mbers. fa n-dimensional cube whose area is e: Theorem D.0.21. For any non-negative sequence of n real numbers x (11,2,.--a4), the arithmetic mean always bounds the geometric mean, i.e., G(x) < A(x) The equality holds iff 11, =. =... = tq Proof (Cauchy (1821)). The result is trivially true for n = 1, where equality 119)APPENDIX D, AN-GM INEQUALITY 120 en = 2. Observe that (1 —m) > 0 2x2 Any Iv (xv)? Equality holds in the first inequality iff 2, = 12. Now, let us assume the result holds for n =m. We shall show that it also holds for n = 2m. Let 1, 2, --- sm; Vis Yas---sYm be the non-negative (i) = ("oy") (i) -)" 255 1 In ew) 2m real numbers. Then aa Moreover, equality holds iff 2) = 22 =... = 2m =i = 2 = ++. = Ym. Thus, we have proved the result when n = 1,2 and all power of 2. Suppose n is an arbitrary integer and let & be an positive integer such that n < 2* = N. Let Lyn 2 Da ai and sot ue =... = ay = 8. Therefore, Tp < (3 Honeo, [P21 $s" andAPPENDIX D, AN-GM INEQUALITY 121 A more general AM-GM inequality is given below. In practice, the p's in the result below will play the role of weights or probabilities. ‘Theorem D.0.22. Let 1,22 Ships b Then 1% 2 0 and pr,p2, Th? < Soe. ia iat with equality iff t) = 22 =... = tqAPPENDIX D, AM-GM INEQUALITY 122Bibliography (Pug04] Charles Chapman Pugh, Real mathematical analysis, Undergraduate ‘Texts in Mathematics, Springer-Verlag, Indian Reprint, 2004. 123BIBLIOGRAPHYIndex inequality am-gm, 119 generalised am-gm, 121