a or 2b is defined to be =. In case a=b, the interval aS 2b degenerates to a point and has length zero. Thus length is a non-negative real number. Since an interval J is a set of points, we see that its length is a set function of having value L(1) 20. AREA AND VOLUME ‘The idea of length is easily generalized to two, three and higher dimensional Euclidean spaces. For example, if we consider two dimensions, we can think of a rectangle as a generalized interval a mE) A-5 [Monotonicity]. If E:CEs, then m(E) = m(Bs). A6 [Translation invariance]. If each point of a set F is translated equal distances in the same direction on the real line, the measure of the translated set is the same as m(E). A-l. If E is an interval then m(F) = L(E), the length of the interval. Note that if requirement A-4 holds, then A-3 also holds. However, the converse need not be true. The two requirements A-3 and A-4 are sometimes together called the require- ment of countable additivity. We can show that requirement A-5 follows from the other requirements (see Problem 2.31} It is possible to show that for general point sets on the real line we cannot satisfy all of these requirements. Thus if we wish to keep all the requirements A-2 through A-7 we must conclude that not all sets have a measure. Similarly if we want all sets to have a measure, then we must sacrifice one or more of the requirements A-2 through A-7. EXTERIOR OR OUTER MEASURE OF A SET The exterior or outer measure of a set E, written m.(E), has the following properties which can be considered as axioms. Bl. m(E) is defined for each set B. B2. m(E) = 0. BB, me(E,UE2U+++) S meE) + mE) + for all sets E1,Es,... disjoint or not. B-4, Exterior measure is translation invariant. Note that the requirements A-3 and A-4 for ideal measure, i.e. the additivity require- ments, have been dropped and replaced by B-3.CHAP. 2} MEASURE THEORY 31 MEASURABLE SETS The exterior measure, although defined for all sets, does not satisfy in general the additivity requirements A-3 and A-4, as we have noted. Now we know that every set 7 can be written in the form {see Problem 1.10, page 12] T = (rngyucrnéy w ‘Thus if there is to be any chance at all of satisfying the additivity requirements, we ought to have e mT) = m&( TOE) + mAT OE) oy for all sets T, since 71 E and 7.0 H are disjoint. The result (5) will hold [if at all} only for a restrictive class of sets E. This leads to the following Definition 2.5. A set E is respect to an id to be measurable, or more precisely measurable with exterior measure, if for all sets T m(T) = mATOE) + m(TOB) (6) Tn such case m.(B) = m(B) is called the measure of E. We can show that if we restrict ourselves to measurable sets, then the above ideal require- ments for measure are satisfied. Since it is true that for all T [see Problem 2.8, page 36] mT) S mT OE) + mA TAB) the Definition 2.5 can be replaced by the following equivalent one. Definition 26. A set E is measurable if for all sets T mAT) = m(TOE) + mATNB) ” ‘This is often used in practice to show that a set is measurable. ‘The sets T in the above definitions are often called test sets since they are used to test measurability. LEBESGUE EXTERIOR OR OUTER MEASURE Thus far we have not actually demonstrated the existence of an exterior or outer measure although there are in fact several which satisfy the axioms given above. The one which is most famous and useful is attributed to Lebesgue who investigated it in the early part of the 20th century. Definition 2.7. ‘The Lebesgue exterior or outer measure of a set E is mE) = glib. L(0) for all open sets 0 B ite. the greatest lower bound of the lengths of all open sets O which contain E. ‘We can show that this does indeed satisfy the axioms for exterior measure [see Problem 2.4]. We can prove that if B is any interval I, then m.(I) = L(J). Also if S is any open set, then m.(S)=L(S). See Problems 2.2, 2.26 and 2.27. The following theorem is important. Theorem 2-1. If Eis any given set, then given ¢ > 0 there exists an open set 0 # such that m(H) < m(O)+« and m(0) = m(E).32 MEASURE THEORY [oHAP. 2 LEBESGUE MEASURE Definition 28. If a set E is measurable with respect to the Lebesgue exterior measure of Definition 2.7, then we say that E' is Lebesgue measurable and define the Lebesgue measure of E as m(E) = m.(E), In this book we shall, unless otherwise specified, be concerned only with Lebesgue exterior measure or Lebesgue measure of sets. Consequently we shall often refer only to exterior measure or measure without mentioning the name Lebesgue explicitly, In his investigations, Lebesgue did not actually use Definition 2.5 given above to define measurable sets. Instead he considered sets E in the bounded interval [a,] and first defined the interior or inner measure of a set Has m(E) = b—a — m (i) (8) He then called # measurable if the interior and exterior measures are equal, ie. if mB) = b—a— medi) (9) Now if we let 1 = [a,b], (9) can be written m(1) = meINE) + mlb) (19) which is a special case of (6) with T=. Thus the actual definition which Lebesque used is a special case of (6). Since Lebesgue started with sets contained in [a,b], ie. bounded sets, appropriate modifications had to be made for unbounded sets. Such modifications, however, are not needed if Definition 2.5 is used. Furthermore Definition 2.5 has the advantage that it can be used in more general theories of measure. Consequently we shall use Definition 2.5 in various proofs. It is of interest that if (10) is satisfied then (6) will also be satisfied [see Problem 2.18). It should be mentioned that a modification of the Lebesgue procedure is sometimes used. According to this the interior measure of a set H is defined as m(E) = Lub. L(C) for all closed sets CCE (a) i.e. the least upper bound of the lengths of all closed sets C contained in B. Then E is called measurable if m(E)= mE) where me(E) is given by Definition 2.7. Other pro- cedures are also possible [see Problems 2.42-2.44]. All of these procedures can in fact be shown equivalent. It should also be noted that for any set E, m(H) and m.(E) always exist and m{B) = m.(E) (12) This result is equivalent to (7) with T=. THEOREMS ON MEASURE Theorem 2-2. If E is m ible. then # is measurable and converselyCHAP. 2} MEASURE THEORY 83 Theorem 2-7. If E, and Ey are measurable, then their intersection E, 9 Es is measurable. ‘The result can be extended to any finite intersection of sets. Theorem 2-8. If Ey and E: are measurable and Ey C Es, then m(E3) = m(Bs). Theorem 2-9. If E; and E2 are measurable, E:C E., and E, has finite measure, then E:~E; is measurable and m(E2— Ey) = m(Bs) — m(E1). Theorem 2-10. If E, and E: are any measurable sets, then ‘m(E,UE:) = m(E1) + m(Bs) — m(Ei0 Es) This reduces to Theorem 2-6 if E; and Ey are disjoint. The result ean be generalized [see Problem 2.41). Theorem 2-11. If Ey,Es,... are measurable, then E = ,U, Bx is measurable. Theorem 2-12. If E1,E»,... are mutually disjoint measurable sets, then m(BiUBaU +++) = m(Es) + mB) + ++ Theorem 2-13. If E,E:,Es,... are measurable sets such that EiCK2CEsC-++, then B= 0, By is measurable. Theorem 2-14. If B:,B:,Es,... are measurable sets such that Ey CH:CByC--+, then E=,0, By is measurable and (5, Ex) = limmB,) Theorem 2-15. If Es,E:,Es,... are measurable sets such that Ey E> Hs> +++ and at least one of the F}, has finite measure, then E'= 9, By is measurable and (A, 5) = limmE.) Theorem 2-16. If Tis any interval, m(I) = L(I). Theorem 2-17. If 0 is any open set, m(0) = L(0). Theorem 2-18. If C is any closed set, m(C) = L(C). ALMOST EVERYWHERE A property which is true except for a set of measure zero is said to hold almost everywhere. BOREL SETS ‘The class of sets which can be obtained by taking countable unions or intersections of open or closed sets is called the class of Borel sets. We have the following Theorem 2-19. Any Borel set is measurable. VITALI'S COVERING THEOREM The following theorem due to Vitali is an important and interesting one which is reminiscent of the Heine-Borel theorem {see page 7).34 MEASURE THEORY (oar. 2 Theorem 2-20 [Vitali]. Suppose that each point of a bounded, measurable set H is covered by a class J of intervals having arbitrarily small length [ealled a Vitali covering]. ‘Then there exists a denumerable set of disjoint intervals /,, Is, such that 0) Jk covers E except for a set of measure zero. NON-MEASURABLE SETS It would seem from the above results that all sets in R are measurable. This however is not true, i.e. there are sets in R which are not measurable. For an example of a non- measurable set see Problem 2.21. Solved Problems LENGTHS OF SETS 21, Find the length of the set -f, 1 1 va = {ergy fh. te eee so that Lh) = 1-4, LU) = bm dy eee DU) = eae ‘Then since the Jy are mutually disjoint, 4L0 a] 3h = a-p+a-p+ Thus 2f On] = ame [e EXTERIOR OR OUTER MEASURE 2.2. Prove that if S is any open set, then m.(S) LS). Suppose that 0 is any open set containing S, ie. 03S. Then each of the component intervals of $ is contained in some component interval of O. Now S is the smallest open set which contains itself, Le. $3, Then by definition of exterior measure we have mS) = gi.b.L(0) where 0>8 = L(S) 23, Prove Theorem 2-1, page 81: If E is any given set, then given «> 0 there exists an open set ODE such that (a) me(E) < me(O) +e and (b) m(O) = me(B). (a) By definition of Lebesgue exterior measure, mE) = gb. 1{0) for all open sets ODE Since the greatest lower bound exists, it follows that m,(E) exists and by definition of greatest ower bound we will have for any ¢> 0 mE) < 10) +« However, since L(0) ~ m(O) by Problem 2.2, the required result follows. (8) This follows from Problem 2.25.CHAP. 2] MEASURE THEORY 35 24, 25. 26. Prove that the Lebesgue definition of exterior measure satisfies axiom B-3, page 80, for exterior measure, ‘We must show that for any sets B,,Es, . m (ye) = Zee ‘The inequality is trivial in ease one of the sets has infinite exterior measure, so that we can restrict ourselves to the case where all sets have finite exterior measure, Tn such case given e>0 there are open sets ODE, where k=1,2,... auch that [seo cea 1{0,) = malO,) < meEx) + 2k L(Oy) = me(Or) mAEy) Then ny *) = n(y os) = Brio) = Ymlby + B® = Fm) te ‘Thus since ¢ can be taken arbitrarily small, m(y me) = Emin (ys) = 3 ‘We can also show that the other axioms on page 30 are satisfied, Prove that if F is any countable set of real numbers, then m.(E) = 0. Let the points (real numbers) of the set be ay, dz,ds,.... ‘Then we ean enclose 4, 0,5, in open intervals of lengths less than or equal to «/2,/22,¢/98, ... respectively. Thus mB) 2 E+ Et Eto se Since « can be taken arbitrarily small, it follows that m,(E) = Prove that for each set 7 and any finite collection of mutually disjoint sets Hs, ..., B m.(to{o,m}) ‘We use mathematical induction, For m= the result is certainly true, Assuming it to be true for n—1 sets, we would have Emarngy ™ (ro ftGies}) = 53 marney “ Adding m(T7E,) to both sides yields mJT0E,) +m, (ro{iim}) = 3 murony ® This ean be written [see Problem 2.30] w(tofd.ecbom) +m (rf Now since His measurable, we have for each set T mT) = mkTOE,) + mT AB, w mATOEy) “6 son wetng T= rf Cie, oy becomes 6) Hence the result is true for n sets if it is true for m—1 sets. But since it is true for 1 set it follows that it is true for 2 sets, 3 sets and so on.36 MEASURE THEORY (omar. 2 MEASURABLE SETS AND THEOREMS ON MEASURE 27. Prove that if H is measurable, then the complement # is also measurable. Since E is measurable, we have for exch set T mT) = mfTOE) + mT By a In order that £ be measurable we must show that for each set 7 maT) = mi TOE) + meTAE\-} = mirnB) + mT0z) @ But (2) is identical with (1), ‘Thus # is measurable, 28 Prove that Definitions 2.5 and 2.6, page 31, are equivalent. For any sets T and we have ~ T= (roByu (rab) ® ‘Thus by Problem 2.4, mT) = m( TNE) + mits) @ Now according to Definition 2.6 a set E is measurable if mT) = m{TE) + m7 0B) (s) ‘Thus we ean conclude that « set E is measurable if for any set 7” m(T) = m(TAE) + ml TOE) w Which together with (2) yields (). Conversely if we adopt Definition 2.6, then in view of (2) we see that His measurable according to Definition 2.5. 29. Prove that E is measurable if m.(E) = 0. For any set T we have TNE CE, $0 that m(TOE) mB) = 0 Then since m,(TNE) we have mATNE) = 0 @ For any set T we have T> TF, so that using (1), mA) = m(TOB) = ml TE) + mi TrBy @) from which we see that B is measurable and_m(B) = m,(E) = 0. 2.10, Prove that if Z, and Hy are measurable, then EU E> is measurable. We would like to show that for any sat m{P) © mE 0B\UEY) + mA 7 (BYUE IM) Since B, and Ey are measurable, we have for any sat 7 me) = m(POE,) + miro) APOE) + miToB,) #,) + meliroky vB) maT OB) + m(ToB,B) + mirnk By) = miPOE) + miTOB OB) + ml 0 (EVE) mT 0 (B,UB,]) + ml 0 (EVE) Thus BU By is measurable,CHAP. 2) MEASURE THEORY 37 2. 212. 213. 21d. 2.16. Prove that if H, and E; are measurable, then E10 Es is measurable. By Theorem 112, page 4, we have BE, = (GB, Since 5, and By are measurable t follows from Problema 21 and 210 that B, 0B in meaattable, Prove that a countable union of measurable sets is measurable. ‘We can always choose the sets in a countable union to be mutually disjoint page 18], We must thus show that if B,,E,,... are mutually disjoint measurable sets, then B= 0, By is measurable Now since any finite union of measurable sets is measurable, we have for any set T win = m(en{8)) += (0>{ Bal) w(0o{a)) += (e>{B4F) = B maton + mrad Since m{f) is independent of n, we have on taking the limit as n= mit) = Z,mcrowy + marok) = mite) + m(toB) Thos B= O, By te measurable Prove that if H1,s,... are mutually disjoint measurable sets, then m (2 #) = Smee) This follows at once from Problem 212 by leting T=, the set of real numbers Prove that the measure of any countable set E is zero. By Problem 2.5, m,(E) = 0. ‘Thus by Problem 2.9, the measure of His given by m(E) m.tB) Prove that a countable union of countable sets has measure zero. Since a countable union of countable sets is a countable set, the measure of this set is zero by Problem 2.14 Given the measurable sets > H,>Es> +++ where m(E) is finite. Prove that (5.5) = lim m(Es) Let B= (0) By Then we have By-E — By) U (By—BQ)U © Since E,— Bs, By—Ey... are mutually disjoint and all the sets are measurable, we have by Problem 2.13. mm(B,—B) = m(B,— By) + (By E+ ®38 247, 218, MBASURE THEORY (CHAP. 2 Then since HyDE, Hy3 By, By2By ..., (2) can be written sm(dy) — m() = Bim [my Bs) + (By By) + °° + Ey 4B) = Bion [rm(y) ~ mC) + om(y) — (Bi) + + + (Bs) ~ mB) = Kim [m(By ~ m()] = ml) — Sim mle) ‘Thus since m(B) is fnite, me = m(8,m) = 3m wien Prove that the Cantor set is measurable and find its measure. In obtaining the Cantor set, the interval [0,1] is subdivided into 8 parts and the middle third is removed, ‘Then the remaining intervals are in turn divided into 8 parts and the middle thirds of each are removed. ‘The process is repeated indefinitely and what is left is the Cantor set. Now at the frst step 1 interval of length 1/9 is removed, At the second step 2 intervals of lengths 1/2 are removed. In general after n steps; 2? intervals of lengths 1/2» are removed. IW fellows that the total measure of the Cantor set K is $2 may = 1- 3% One of the interesting properties of the Cantor set is that although it is non-countable [see Problem 1.22, page 16] it has measure zero. Let I be the interval [a,b]. Prove that a necessary and sufficient condition for a set 2 in I to be measurable is that ml) = mE) + mei) Necessity. By definition a sot H is measurable if for all sets 7 mT) = m(TAE) + mi TB) we in this case THE = B ‘Then if in particular we let T= 1, the required reeult follows and TAH = B, Sufficiency. Let T be any set and suppose that G is a measurable set containing 7 such that m7) = m(G) [ace Problem 2.46). Since G is measurable, we have m{B) = m{BAG) + m(EOG) mB) = m(EnG) + mE) It follows that m() = ml) + m(B) = m(BNG) + m(8G) + m(FnG) + msn) = [m(EnG) + mEnG)) + Ime) + mE nd) = m(@) + mG) = mG) + mG) = mn ‘We thus see that m(ENG) + m(BOG) + mG) + mlENG) = mG) + mG) w ‘Now we know that mEOG) + mEG) = mG) = ma) @ Subtracting (2) from (1), m(ENG) + m(EnG) = mG) Now since TCG, we have ENTCENG and EnT CHG, 50 that m{BOT) + mEAT) = m(ENG) + m(EAG) = mG) = mT) ‘Then by Problem 2.8 it follows that H is measurable,CHAP. 2] MEASURE THEORY 39 Tt should be noted that the theorem proved here is proved under the assumption that for any set T there is a measurable set G such that m,(T) = m(G). This is true for Lebesgue exterior measure as defined in Definition 2.7, page 31 (see Problem 2.45]. An exterior measure for which this is true is often called a regular exterior measure. VITALI'S COVERING THEOREM 2.19. Prove Theorem 2-20 [Vitali’s covering theorem]. Let O be an open set of finite measure containing B. ‘Then we can assume that each interval of .f belongs to O. Suppose that hy is the least upper bound of the lengths of all these intervals of (Se Choose an interval J, € J whote length is greater than $h; ie. Lil,) > 4h. Now if m,(2—J;) =0, the required result is established. If not, let hz be the least upper bound of the lengths of all those intervals of .J which are disjoint from I,. Denote by I, an interval of J which is disjoint from Z, and whose length is greater than Jha, ie. [([,) > jhe If m,(H—1,U1,) = 0, the required result follows. If not we continue the process to find ‘an interval Js, ete, Clearly the process will either terminate, in which case the result is established, or it will not terminate, ‘We must thus investigate the ease where the process does not terminate. In such case we will have a sequence [1,1y,... of mutually disjoint intervals with the property that m() = Th) > dhe R= 1,2, ® Then since each 1,€0 s0 that 0, fc O, we have hy Problem 238 am(O) eo ™ (3,4) mt) = 3 Ly Thus the series in (2) converges and as a consequence of this and (1), lim Mth) = 0 and fim hy = 0 Consider now the set S = E— ) J which is not empty. If 25S, then it belongs to F and does not belong to Iy,ly,-.-11y. By the definition of Vitali covering there will be some interval TE J containing 2. Also /'is disjoint from Iy,...5lq. ‘The interval J must have at least one point in common with some interval I, where k > n. For if J were disjoint from Ins, Tura, --- for k = n+1, 042, ..., it would follow on noting that LU) = hy for k = nt4,m'+2, ... and lim hy = 0 that L() = 0 so that J would not bbe an interval, ed ‘We can assume that the first interval in the sequence (J,) with which J has common points is Ty. We see that in such case UD) S hese < 2Ellss) “6 Since # € I, we see from Fig. 21 that the lens distance from 2 to the midpoint of Ix. cannot Bs exceed LUN) + $L(In+1) oF FL(n+:1) using (8). a If therefore we consider an interval J, with —_—T the same midpoint as J, but five times as long, 7 wwe ace that 2 € Jey Fig.24 From thi follows that all points # belonging t» B ~ 5, fy alo belong tO) Je be B- Onc Om « Then by Problem 225, . . Es G a n) = 3 wy = 5, 3 uw40 MEASURE THEORY [cHAP. 2 But since 3 1(%:) converses, it follows that given «>0 there is a number my such that the last sum can Be made less than «/5 for n> my, Le, m (B= Oe) <¢ tor n> ny © Prove that the conclusion of Problem 2.19 is the same as the statement m(e—Or) =0 or that the mutually disjoint intervals 1, Js, ... cover E almost everywhere. ‘The result (5) of Problem 2.19 can be written equivalently as m(e-Gn) = 0 ‘Thus from Problem 29 we have . n(i -o n) NON-MEASURABLE SETS 2.21, Demonstrate the existence of a non-measurable set, We shall show that there is a non-measurable set in the interval [0,1], For convenience in arriving at this set we shall map this interval on to the circumference C of a circle so that to each point in [0,1] there will be one and only one point on C and conversely, Let x and y denote any two points on C. We define x and y to be equivalent points if the are joining them has rational length and in such ease write x ~ y. ‘We now consider subsets of C, denoted by Aq, having the property that two points x and y are in the same Ay if x ~ y. The sets A, are often called equivalence classes. Since the rational numbers are countable, it is then clear that each A, contains a countably infinite set of points, i.e. is countable, ‘The sets Ay are also mutually disjoint. For if the same point z belongs to different sets Ay, and Aq, then all points of Ag, and Aq, are the same so that A, is identical with A,,. Since the sets Ay are mutually disjoint, it follows that there is @ non-ountable number of such sets, To see this we simply have to observe that each set must contain a distinet irrational number and that the set of irrational numbers is non-countable. ‘We now construct a set S which consists of one and only one point 2, from each of the sets Aq so that any two distinet points of S are not equivalent. To do this we must use the axiom ‘of choice, ‘The set S is the required non-measurable set, To show that S is non-measurable let us first consider the set of rational numbers between © and 1, denoted by ry, Tq 75, .... For simplicity let r,=0, For any positive integer k let Sy be the set of points of C which we obtain by counterclockwise rotation of S through an are of length 7. It is clear that S=S, and all the S, are congruent, go that they are either all m urable or all non-measurable. The sets S_ do not have any points in common, i.e. they are mutually disjoint, To see this let us consider any two S,, for the sake of argument S, and S;. If S, and S; are not disjoint there is a point in common, say p. Since p€S, there is an clement y@S for which the are length from p to y is rq Also, since p €S; there is an element 2€S for which the are length from p to < is r;, Thus the are length from y to + is rational, ic, y~2. This however contra- dicts the fact that no two distinct points of S can be equivalent, and thus we must have y=~= s0 that we must have ry =r; which is impossible, This contradiction shows that the Sy are mutually disjoint, Now if we take any point 2 €C, then we have x € Ay for some a. Since rq € Aq it follows that the are length from 2 to z, equals 1, for some positive integer k so that 2 € S,.” Thus each point of C belongs to some S,.CHAP. 2} MEASURE THEORY 41 We thos ents that © = 5S, where th Sa tly nt, Then € te, are measurable we have m(S) MSy) US) and m(C) = m(S,) + m(Sy) + = m(S) + m(S) + +++ Now if m(S)>0 we would have m(C)==, and if m(S)=0 we would have m(C)=0. Since ‘m(C) = 1, we cannot have m(C) = of 0, This contradiction shows that S cannot be measurable as we were required to show. Supplementary Problems LENGTHS OF SETS 222, Find the lengths of each of the followi (@) (@:-2<2<0) () fe: 385 Ans. (8 (4 (9 sets, B}U(-4S2<-2) (@) (2: -~80 there exists an open set ODE such that_m(Q~E) 0. there exists a closed set CCE such that m(B—C) 0 there exist an open set ODE and a closed set CCH such that_m(O—C) <« If A iis any set, prove that there is a measurable set BDA such that m,(A) = m(B). See Prob- Jem 2.18, Suppose that all the numbers between 0 and 1 are expressed as non-terminating expansions in the scale of 10, i.e. as infinite decimals, Prove that the measure of the set of all such numbers in which one particular digit [say 5) is omitted is zero. Prove Theorem 2-19, page 38. Use Lebessue's definition of a measurable sot (see (5) or (9), page 22| to prove (a) Theorem 2-6, (0) Theorem 2-11, Show how to prove some of the theorems on pages 32 and 93 if we start with the definition that E is measurable if there is an open set ODE such that m,(O—E)« is measurable, The set of values « € E for which f(z) >. can be written (2B: (2) >} or briefly Elf(e) >). If f(a) is measurable we call it a measurable function, ‘THEOREMS ON MEASURABLE FUNCTIONS Theorem 3-1. The function f(z) is measurable on E if and only if for each real number x one of the following sets is measurable (@) Elf@) x], E[/(a) ] is measurable for each rational number «. Theorem 3-5. If f(z) is measurable on a set Ei and if BC Es, then f(z) is measurable on Es. Theorem 3-6. If f(z) is measurable on a countable class of disjoint sets Hs, H2,..., then it is measurable on their union 0, Ek. Theorem 3-7. If fx(:t) and f2(2) are measurable on B, then Elf:(2) > fe(2)] is measurable. Theorem 3-8. A constant function is measurable. Theorem 3-9. If f(x) is measurable on E, then for any constant ¢, f(z)+e and ef(2) are also measurable on B.44 MEASURABLE FUNCTIONS (CHAP. 8 Theorem 3-10. If f(x) is measurable on B, then (f(z2)]? is measurable on E. Theorem 3-11. If fx(z) and fo(z) are measurable on B, then fi(z) + fale), fale) — fale), fila) fale) and fs(2)/fo(z) where fa(x) +0 are measurable on E. Theorem 3-12. If f,(2) and fa(z) are measurable on E, then the maximum and minimum of fi(z) and fo(x), i.e, max {f:(x), fo(#)} and min {f:(z), f2()}, are measurable. Theorem 3-13. If f(x) is measurable on E, then (f(z)| is measurable on E. Theorem 3-14. If f(z) is continuous on B, it is measurable on E. Theorem 3-15. If f,(w) is continuous and w= fe(z) is measurable, then fi(f2(z)) is meas- urable. In other words a continuous function of a measurable function is measurable. However, a measurable function of a measurable funetion need not be measurable, Theorem 3-16. Let (fa(z)) be a sequence of functions measurable on E. Then F(z) = Lu.b. fa(2), called the upper boundary function, and G(x) = g.Lb. falz), called the lower boundary function, are also measurable on B. Theorem 3-17. Ii (f.(x)) is a monotonic sequence of functions measurable on E such that lim f.(2) = f(x), then f(z) is measurable on E. (Note: The sequence (Jn(2)) is said to be monotonic inereasing if fi(2) S fe(2) S ---, monotonic decreasing if f:(x) = fa(x) -, and monotonic if the sequence is mono- tonic increasing or monotonic decreasing.] Theorem 3-18. Let (fa()) be a sequence of functions measurable on B. Then lim f,() and lim /,(#) are measurable on B. ne Theorem 3-19. Let (fa(2)) be a sequence of functions measurable on E such that lim fa(z) = f(z). Then f(z) is measurable on E. The result is also valid in case lim f,(2) = f(2) almost everywhere. Theorem 3-20. If f,(x) is measurable on a set E and f,(ct) = f(z) almost everywhere on E, then f2(z) is measurable on E. BAIRE CLASSES From the above theorems we see that all continuous funetions are measurable and all limits of sequences of measurable functions are measurable. It follows that limits of sequences of continuous functions, which may be discontinuous, are measurable and that limits of these discontinuous functions are measurable, ete. Thus we obtain a hierarchy of measurable functions. ‘This has led Baire to give a classification of functions. A function is said to belong to the Baire class of order zero [or briefly Baire class 0) if it is continuous. A function which is the limit of a sequence of continuous functions (i.e. functions belonging to Baire class 0] but which is itself not continuous, belongs to the Baire class of order one [or briefly Baire class 1}. In general a function is said to belong to Baire class p if it is the limit of a sequence of functions of Baire class p—1 but does not itself belong to any of the Baire classes 0,1,...,p—1. It follows that every function which belongs to some Baire class is measurable.CHAP. 3} MEASURABLE FUNCTIONS 45 EGOROV'S THEOREM Theorem 3-21 [Egorov]. Let (fa(z)) be a sequence of measurable functions which converges toa finite limit f(t) almost everywhere on a set E of finite measure. Then given any number 6 >0, there exists a set F of measure greater than ‘m(E)— 8 on which fa(z) converges to f(z) uniformly. Solved Problems 3.1. Prove that f(x) is measurable on # if and only if for each real number « the set E(f(t) Sx] is measurable, If f(z) is measurable on E, then for each real number « the set E[/(a) > «] is measurable. Then by Theorem 22, page 82, ite complement with respect to E given by E(f(z) = «] is measurable Conversely if E{/(x) = x] is measurable, so also is E{f(x) > «], i.e. flz) is measurable on B. 32. Prove that f(z) is measurable on E if and only if for each real number « the set Eff(a) =x] is measurable. I (2) is measurable on FB, then for each real number « the set E{/(e) > «] is measurable, Then B[fe)> 2] fs mente for a= 1,2... ands Belper >e-4] = mye) =a x measurable Caveat Bt) omar in B[ © 2+] emer 98 and 50 L Sefer] = ayes is measurable, ic. f(z) is measurable on E. 3.3. Prove that f(x) is measurable on E if and only if for each real number « the set Eff(a) <«] is measurable. If f(z) is measurable on E, then E{f(e) > «] is measurable for each real number x, Then by Problem 3.2, E[f(z)= x] is measurable and so its complement with respect to E given by Elf(e) <} is measurable. Conversely if Biftz) <] is measurable, then its complement with respect to E given by E(/(e) = «| is measurable and so by Problem 32 f(z) is measurable on B. 3.4, Prove that if f(2) is measurable on E, then E{f(e)=¥] is measurable for each real number x. We have Rifle) =] = Blfle) = 4] 0 Bifle) 5 6) Since the sets on the right are measurable, their intersection is also measurable and so the result follows.36. 3. 38. 39, MEASURABLE FUNCTIONS [CHAP. a Prove that if f(z) is measurable on B, then (a) Bif(2) are measurable. =] and (6) Elf(z) = (o) We have Byte) =*) = faine> «| and since a countable intersection of measurable vets Is measurable the reguted rept flows () We have Blfiz) = — 0) Else) < — and the required result follows as in part (a). Give an example to show that even if E[f()=«) is measurable for each real number ‘s it may be true that /(z) is not measurable on Z. Compare Problem 3.4. Suppose that S is a set which is non-measurable (see Problem 2.21, page 40, for example] and let this set be made to correspond in a 1-1 manner with some subset of the set of points for which +> 0. Suppose further that the complement of S, i.e. 5, corresponds in a 1-1 manner with some subset of the set of points for which = Now let us suppose that a function f(z) is defined by these 1-1 correspondences. Then if « is any real number, the set H[/(z) =) contains not more than one point and is thus measurable, However, E[f(z) > 0] which is the same as S is non-measurable, i. f(2) is not measurable, Prove Theorem 3-4, page 43: A function f(z) is measurable on E if (a) Elf(a) > x) is measurable for each rational number « or (b) E[f(x) =x] is measurable for each rational number «. (a) Let « be the limit of a sequence of rational numbers (e,) where a; > mp > As > |, Bile) > a We have BUfle) > 5) = ‘Then since each of the sets on the right is measurable, their countable union is also measurable and so f(z) is measurable on E. (6) This follows as in part (a), since we can write Bifla) = a) ere Prove ‘Theorem 3. is measurable. , Page 43: If f(x) and fa(2) are measurable on E, then Elfs() > fe(2)] For every 2 © # we can always find a rational number r such that f,(2) > r> fale). Since Blila) > > fale)] = Elfale) > 7] 9 Eifel) fale] = VY Bthale) > 7 > fated) ® where the union on the right is taken over all rational numbers r such that f,(z) > r > fate). Since the set of rational numbers is countable, the right side of (2) is a countable union of measurable sets which is measurable, Prove that if /(z) is measurable on E, then for any real constant c, f(z) +e is meas- urable on E. Since f(z) is measurable on E, we have E(f(z) >x—e) is measurable. Thus E(flz)+e >) is measurable and so f(z)-+e is measurable onCHAP. 3} MEASURABLE FUNCTIONS aT 3.10. 3.1. B12. 313. Bd, 3.15. Prove that if f(x) is measurable on E, then for any real constant ¢, c f(z) is measurable on E. If ©>0, Eleflz) > x] = Blf(e) > w/e] which is measurable exo, Bese) >] Ite Elf(z) < w/e] which is measurable, | ef(ez) = 0 which is measurable, ‘Thus ¢ (2) is measurable on E for any real constant ¢. Prove that if f:(z) and fa(z) are measurable on F, then f:(z) + fa(z) is measurable on E. Method 1. ‘We have for each real number «, Eih@) + h@) >«] = FI w By Problem 3.9 and 2.10, «—f,(2) is measurable, Thus by Problem 38 the right side of (1) is measurable, so that f(z)‘ fa(z) is measurable on E. Method 2 We have for each real number «, Blfule) + fale) > a) =U BUG > v1 9 Blfala) > er] ® where the union on the right is taken over all rational numbers r and is thus a countable union. Since f,{(2) and fa(2) are measurable on B, so also are the sets on the right of (2) and a countable union of these sets. ‘Thus the left side of (2) is measurable and so f,(2) + fa(2) is measurable on E. Prove Theorem 3-10, page 44: If f(z) is measurable on E, then {f(c)}* is measurable on BE. We have B[V}? > 6] = Blfle) > Ve] U E[fle) <-Ve] and the required result follows since the union of measurable sets is measurable. Prove that if f:(#) and fo(z) are measurable on E, then f1(z) fe(x) is measurable on E. We have A@ fle = He) + flay? — (he) — LE] ‘Thus by Problems 3.10 and 8.11, f,(2) fa(2) is measurable on E. Prove Theorem 3-14, page 44: If f(x) is continuous on E, it is measurable on B. Consider the set E[f(x) =x). If 2» is any limit point of this set, then every neighborhood of zp contains points such that f(2) =x." By the continuity of f(2) at xq it follows that f(x) = « We thus see that the limit point zp belongs to the set E[/(x) =x). Then this set is closed since it contains all its limit points. It follows that Elf(z) =) is measurable since any closed set is measurable, ‘Thus the complement #[/(2)> x) is measurable and so f(r) is measurable on B. Let f(x) be measurable on E and let O be an open set, Prove that the set E/f(x) € 0] is measurable. Let 0= GIy where I, = (ay,b,) are the component open disjoint intervals (see Theorem 1-19, page 7]. It follows that48 3.16, 37. BAB, 3.19. MEASURABLE FUNCTIONS [cHAP. 3 BYeE0) = ORE) = G(s) > a] 9 BY) a) and Eif(z) x) is measurable, Now Elflfele) >) = Elfel2) € 0] wy where O = Hs hw>® is open since /; is continuous (see Problem 1.41]. Then since the right side of (1) is measurable, the required result follows, Prove Theorem 3-16, page 44: Let (f,(z)) be a sequence of functions measurable on BE. Then F(a) = Lub. (fe(2)) = Lub. (f1(2), falt), --.) and G(x) = gb. (fa(2)) = Lb. (f:(2), f2(a),...} are also measurable on E. We have E(P(x) 24) = 0 Blfale) ‘Then since a countable union of measurable sets is also measurable, the required result follows. A similar argument shows that G(2) = gilb. f,(#) is measurable on E [see Problem 3.45). If fa(z) and fo(2) are measurable on E, prove that (a) max {f,(2),fe(a)) and (b) min (f,(z), fo(@)), where max and min denote maximum and minimum, are also measurable on E. (a) We have max (fs(2), fa(2)) = Lah, f4(2), fel)? so that the required result follows from Problem 3.17. (®) We have min {fi(2), fal@)} = wb. 4(2), fale) so that the required result follows from Problem 3.17. Prove Theorem 3-17, page 44: If (f,(x)) is a monotonic sequence of functions meas- urable on E’ such that lim f.(2) = f(z), then f(e) is measurable on E. Constr the cae where for al #8, file) = fale) = f(z) = We hve fresh rel unter x elim nin =] = Denies) A similar proof ean be given for the case where f,(2) = fa(z) = fala) = +++ oF cases where the equality is omitted.CHAP. 3) MEASURABLE FUNCTIONS 49 8.20. Prove Theorem 3-18, page 44: Let (fa(t)) be a sequence of functions measurable on E. Then lim f.(2) and lim f(z) are measurable on E. call F(z) = Lub. (fy(2), fale, =.) P(x) = Lub. (fale) fale) Fy(z) = Lud. fale), fala --) and and so on, It follows that Fy(2) = Fa(z) = Fy(o Tim fale) = tim Fae) Now hy Problem 8.19, the right side is measurable and so the required result follows. A similar argument shows that lim f,(2) is measurable [see Problem 3.46] 3.21. Prove Theorem 3-19, page 44: Let (fa(z)) be a sequence of functions measurable on E such that lim fa(z) = f(z). ‘Then f(2) is measurable on 2. By Problem 3.20, Tim f,(2) and lim f,(2) are measurable on K. Tf lim f(z) = f(2), then ye) = Ti tote) = tim 100 and #0 is measurable on B. ™ 3.22, Prove that there are functions which are non-measurable. et [see Problem 2.21] and consider the characteristic function _ fi tees = 10 tes Let $ be a non-measurabl ‘Then x(z) is non-measurable. BAIRE CLASSES 3.23. Let fo(2) = gata for any real number 2. (a) To what Baire class does each f.(2) belong? (b) To what Bi class does lim f-(z) belong? (a) Since each fu(#) is continuous, it belongs to the Baire class of order zero, ie. Baire class 0. ea =I, (0) For [el<4, fim fy(a) = 1. Por f|>1, tim fale) = 0. For [ol = dim fale) = Thus if tim, fy(2) = Fe), “then . { ica F@) = 4 Bl=a o iel>a Since F(z) is the limit of a sequence of functions in Baire class zero but does not itself belong ‘to Baire class zero, it belongs to Baire class 1. 3.24, Prove that the sum of two functions of Baire class p is also of Baire class 2. The result is true if p=0, since the sum of two continuous funetions is also continuous, To prove that the result is true for p= 1, we observe that by definition if f(2) and g(2) belong. to Baire class 1 then there are sequences of functions (f(x)) and (g4(z)) such that lim f(z) = (2)50 MEASURABLE FUNCTIONS (CHAP. 3 and lim g,(z) = ge) where each f,(2) and g(x) belongs to Baire elass 0, Now since f(z) + gn(x) belongs to Baire class 0 it follows that f(z) +9(z) = lim {f,() + gq(2)) belongs to Baire class 1. ‘The same argument can be used to prove the result for any value of p. EGOROV'S THEOREM 3.25. Prove Theorem 8-21 [Egorov’s theorem], page 45: Let (fa(1)) be a sequence of measurable funetions which converges to a finite limit f(x) almost everywhere on a set Hof finite measure. ‘Then given any number §>0, there exists a set F of measure greater than m(E) — 8 on which fa(:t) converges to f(z) uniformly. Suppose that the set of all x € B for which fy(<) converges to fiz) is denoted by H. ‘Then we have mH) = m(E) or m(B—H) = 0 o f 1 je: 2H, ys) for ne Ie-fi<3 Let Hy; Now if we fix p it follows that Hy, C Hye C Hya Coo and #= Om, ‘Then by Theorem 2-14, page 22, we have fim mGH,,) = mitt) 0 that there must be a natural number j(p) for which mH — Hyp) < $e ® etl B= Btn wwe see that on F Va fl <2 forall nz sip) fe. fa(2) converges to f(z) uniformly on F. We now have only to show that m(F) > m(E)—8 or equivalently m(Z—F) <$. To do this note that . HP CU GH.) ‘so that by (1), (2) and Theorem 2-8, we have mE-F) = m—F) SS mH-Hyyy) < ie, mF) <8 as required,CHAP. 3} MEASURABLE FUNCTIONS 51 326, 321. 2.28, 3.29, 3.0, ast. 332, 3.33 aM, 335, 3.3. 3a, 3.38, 3.39, 3.4 3.45, 346. 37, Supplementary Problems Prove Theorem 3-3, page 43, Prove a theorem analogous to that of Theorem 3: inequality symbols < is replaced by 5. page 43, where either one or both of the Prove that a constant function is measurable, Investigate the measurability of the function _ [1 if is a rational namber in (0,1) AE) = V0. itz isan irrational number in (0,1) Prove that if f,(2) and fa(2) are measurable, then ¢, f,(z) + ¢2 (2), where e and ey are any constants, is also measurable. Prove that f(z) is measurable on F if and only if E(f(z)<] is measurable for each rational umber », Prove that any function defined on a set of measure zero is measurable. Prove that the sum and product of any finite number of measurable functions is also measurable. Is the result of Problem 8.83 true for an infinite number of measurable functions? Explain, Prove Theorem 3-5, page 43. Prove Theorem 3-6, page ¢3, Prove that if f,(z) and /a(z) are measurable on E, then f;(z)/fa(2) is measurable on E if fa(t) * 0. [Hint: First prove that 1/fa(2) is measurable) If f(z) is measurable, prove that any positive integral power of (2) is also measurable, Prove Theorem 8-12, page 44. Prove Theorem 5-13, page 44. Prove Theorem 219, page 44, for the case where lim fq(z) = fle) almost everywhere in E. Prove Theorem 3-20, nage 44 Prove that the function f(a) defined in the interval [0,6] by fa axe m). Prove that g(f(z) is measurable, such that 9(2,) = g(x) for52 3.48, 39, 350, 351 352, 3.50, MEASURABLE FUNCTIONS Prove the result of Problem 8.47 if g(2) is monotonic decreasing [cHAP. s ola) = o(e,) for > x) ‘Thus prove the result if g(z) is monotonic [ie, monotonie increasing or monotonic decreasing). Show that lim lim (cos p! rz)% is measurable, ‘To what Baire class does the function of Problem 3.49 belong? Let fa(2)= em" for 03251 and let F(z) = lim f,(z). To what Baire classes do the funetions (a) f,(2) and (6) F(z) belong? Let S be a non-measurable set. Suppose that for each pES we define a function if2eS fl =p fa {0 otherwise’ (@) Show that F(2) = Lub. [fp(e) where p € 5] (6) F(@) is non-measurable. Does the result of Problem 8.52 contradict Theorem 3-16, page 447 Prove the result of Problem 8.24 for the case where p = Prove that if j(2) is of Baire class p, then so also are (a) [f(z)l, (®) (f(z) Sole) = Prove that the product of two functions of Baire class » is also of Baire class p. Prove that if f,(2) and fo(2) are of Baire class p, then so also are (a) max {/;(2), fa(z)} and (@) min ((2), fl@)). (Hints max (F,(2), fale) HAG) +A) ~ HAG) ~ A] Use Problem 3.57 to work Problem 3.18, (02) + F202] + C2) — fa), min 2), F2)) = Let (fy (z)) be a sequence of functions of Baire class p which converges uniformly to f(z). Prove that f(z) is also of Baire class p. Is the result of Problem 8.59 true in cate the convergence is not uniform? Explain,Chapter 4 THE RIEMANN INTEGRAL The integral usually treated in introductory courses in the calculus is called the Riemann integral after the mathematician who contributed largely to its development. A discussion of this integral and its properties is given in Appendix A, pages 154-174, for the benefit of those who wish a review of it. ‘The Riemann integral has certain defects which can be remedied by use of the Lebesgue integral, as we shall see in this and later chapters, Although it is possible to define the Lebesgue integral in many ways, we shall adopt a procedure which parallels as closely as possible the definition given in Appendix A for the Riemann integral. ‘The main difference between the Riemann and Lebesgue integrals is that the former uses intervals and their lengths while the latter uses more general point sets and their measures. Thus it is not surprising that the Lebesgue integral is more general than the Riemann integral. DEFINITION OF LEBESGUE INTEGRAL FOR BOUNDED MEASURABLE FUNCTIONS Let f(z) be bounded and measurable on the interval [a,b]. Suppose that « and f are any two real numbers such that « < f(t) u and m(E{") = 0], it follows that the upper sum eannot increase by adding a point of subdivision and the required result is proved.CHAP. 4] THE LEBESGUE INTEGRAL FOR BOU 45. 4a. DED FUNCTIONS 59 Prove that an upper sum for any partition is not less than a lower sum corresponding to the same or any other partition. ‘The required result has already been proved when the partitions are the same [Problem 4.1] ‘To prove the result for different partitions, suppose that the upper and lower sums correspond- ing to one partition are given by Sp,e, while the upper and lower sums corresponding to the other partition are given by S3,65. Consider the partition obtained by superimposing the two above partitions [ie. the partition obtained by taking the union of all the subdivision points] and denote the corresponding upper and lower sums by Sj,8y. Since this new partition is a refinement of each of the given partitions, we have by Problem 4.4, S: 5 Sy 2 05 a But we also have by Problem 4.1, Sey ® ‘Thus from (1) and (2) we have SB © (or S; = s, on interchanging subseripts 2 and 3], i.e. an upper sum for any partition is not Tess than a lower sum for any other partition Using the notation of Problem 4.3, prove that [= J. Assume the contrary, ie, I< J. Now I is the sreatest lower bound of all upper sums. Thus there must be some partition or mode of subdivision giving an upper s . sum S to the left of the midpoint of the interval 1 << / [Fig. 42), 7 7 Also since J is the Ieast upper bound of all lower sums, there must be some partition giving a sum # to the right of the midpoint of interval 1J. Fig. 4-2 ‘This, however, is impossible because S cannot be less than #, Thus we have arrived at a contradiction on assuming I< J, and it follows that 1 = J. Note that although we have used a geometric argument the proof can be easily formulated analytically (see Problem 4.33), IfS is an upper sum corresponding to any partition and s is a lower sum correspond- ing to the same or different partition, prove that s JBe Since J is the greatest lower bound of all upper sums, we must have $=. Since J is the least upper bound of all lower sums, we must have J = s. Also from Problem 4.6 we have I = J, Thus S=1ZJEe, 12 Prove that f(x) is Lebesgue integrable if and only if for any ¢>0 there exists a partition with upper and lower sums S,s such that S—s 0 there is a partition such that S~e<, then by Problem 6.7, OS I-J 5 8-4 0 there exists a partition such that SJ~<«/2 [since J is the least upper bound of all lower sums). ‘Then S—s < (4d2)—-U-W2) =60 THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS (CHAP. 4 4.9. (a) Show directly from the definition that the function Ne) 1 isa rational number in (0, 1) (2) 0 x is.an irrational number in [0, 1] is Lebesgue integrable in [0,1] and (b) find the value of the Lebesgue integral of F(a) in (0, 1). (a) In the definition of the Lebesgue integral (see page 53] we choose subdivision points such that mS WSS St Se =P and consider sets By = ei mn S10) 1, > 0, tyr <1) By = me fe) 0 while =a is any number = 0, T= lubS = lab = 0 J = glbs = ghbw = 0 ‘Thus since T= J=0 we see that f(x) is Lebesgue integrable in [0, 1]. (®) Since the common value of the upper and lower integrals T and J is 0, it follows that the Lebesgue integral of /(2) in [0,1} is 0, Le. fii =o Note that the function is not Riemann integrable [see Problem A.2, page 159].CHAP. 4] THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS él 4.10. Prove that if f(x) is bounded and measurable on (a, b], then it is Lebesgue integrable on [a,b]. Using the notation on page 58, we have se = 3 nn vmey Now we can corsnyasrang to have «pon o mide of athvson wo hat wen < GE ‘Thus we can say that 3 may ) = 7253, 0 that from Problem 4.8, f(z) is Labessrue integrable. THE LEBESGUE INTEGRAL AS THE LIMIT OF A SUM 4,11. Prove that the definition of the Lebesgue integral as a limit of a sum [see page 55) follows from the definition given on page 53. For the case where the number of subdivision points is m, let S, and s, be the corresponding upper and lower sums, Then if J is the g.lb. of all upper sums and J'is the Lub. of all lower sums, ‘we must have asssrss, w Now 8, = 3 nme, 6 = 3 erminy so that os s- Bunn dmey ® ‘Then given «> 0 we ean choose n> my 80 that yy — v1 < mee 3, - mB) = « 7 eS ie. lim (S,~e,) = 0. Thus from (1) we see that rar f'pie lim Sy, ‘We can also show the converse {see Problem 4.29]. THEOREMS INVOLVING THE LEBESGUE INTEGRAL 4.12, Prove Theorem 4-5, page 55: If A=f(z)=B, then Am) = Jf fede = Bm(B) From Problems 4.2 and 4.7, AmB) 3821218 = Bm) ‘Since we are assuming that f(z) is Lebesgue integrable so that I= J, it follows that Ame) sf fede = Bm62 THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS (CHAP. 4 4.13. Prove Theorem 4-6, page 55: If E = E\UK2 where E; and 2 are disjoint, then S eae = S, fla) de + S. Hla) dx by using the Lebesgue integral as a limit of a sum. Choosing subdivision points y_ such that ee) the set H is divided into subsets H, while Z and E, are divided into disjoint subsets E{?, Eg? such that Ey = EL) uB® and my) = (HQ?) + mE?) Then SJ, serae = rim 3 vem) = tim 3 remo Sto + f toe For another method see Problem 4.14. 4.14. Prove Theorem 4-6, page 55 without using the Lebesgue integral as the limit of a sum [see Problem 4.13]. Suppose that for EE, « < f(z)

©, we have S vey +eae Nee) de + mB) 4.17. Prove Theorem 4-8, page 56: If f(x) and g(z) are bounded and measurable on E, then S; (feo) +oeae = f fade + Sf omae at B= 0, By whore the Hare given by (2) of Problem 436. Then SJ, +oenae Sf v0) Bf, tao Smaimen + Bf ae st fone |64 4.18, 4.19. 4.20, THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS [CHAP. 4 Thos we have Sverssioiie = fous + forse wo Sinan Sversoenes = 3S yor een w 3S m+ oe ae = Snmep + 3 J, ate) s+ Sota Thos we have Ser +onde = f tear > ff ateras ro) rom (1) and (2) we must have Si ttertomier = ff pores + f oar The result is easily extended to any finite number of functions [sce Problem 4.48) Prove Theorem 4-10, page 56: If f(z) =g(z) on E, then Sioa = Sf o@ax Since g(z) — fle) is non-negative on E, ie. g(z)— f(x) = 0, we have S012) ~ 10) azo e Sports = fn Prove Theorem 4-11, page 56: If f(a) is bounded and measurable on E, then |/(z)} is Lebesgue integrable on Z. Conversely if [f(z)| is bounded and measurable on E, then f(«) is Lebesgue integrable on E. If f(a) is bounded and measurable on E, then so also is |/(o)| [see Theorem 8-18, page 44), ‘Thus a) is Lebesgue integrable on B. Similarly if [(2)| is bounded and measurable on F, then 30 also is f(z). Thus f(x) is Lebeseue Integrable on E. Prove Theorem 4-12, page 56: | S, teax| = F \ieax under the conditions of ‘Theorem 4-11. ‘Method 1. ‘We have Wey] = fey = [ple] ‘Then by Problem 4.18 we have on integrating over the set E, ~Sieias = Sseoae s J peolae a [Jr = S wenaCHAP. 4] THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS 65 421, Method 2. If B+ = Bif(e)=0] and B J, fede = J, sere + S fear = f. \ey| ae ~S Weoiae and Sweoiae = Sf ieoiae + J eiae |S veoiae — f yee es| Si veride +f weniae = f yariae Bile) <0}, then ne [fsa 4 Prove Theorem 4-14, page 56: If f(x)=0 is bounded and measurable on FE and f feae = , then f(z)=0 almost everywhere on B. Method 1. ‘Since f(z) is bounded, there is a constant M such that 0 = f(x) MM. Consider the sets E, = (@: f@)=0), By = fa F< so = Jp M an “}. ay = fe fcsasZh Me fand in general Ey {e Rie) Since the sets are measurable and disjoint, mE) = (Ey) + m(B,) + + where B= Now by the mean-value theorem, M mE), men = J sora From the inequality on the left and the fact that the integral over K, cannot exceed the integral over B, we have mins BS finde = BS page = 0 m(EguEyu--+) = m(By) + m(By) +--+ = 0 ee ene eae af rial See al ere Method 2 Suppose that contained in E there is a set A of positive measure for which f(z) >0. ‘Then wwe can express A as the union of mutually disjoint measurable sets Ay = Ge: fla) > 1), Ay = es fl) > Hh, Ay = (es > D Since m(A) = (Ay) + m(da) + +++ and m(A) > 0, it follows that there is some set Ay such that, ‘m(A,) is a positive number, say p. Then by the mean-value theorem and the fact that the integral over A, cannot exceed the integral over B, we have may) _ p SJ, terae = J tear = MP = E> 0 which contradicts the hypothesis that the integral over E is zero, Thus the set A cannot be of positive measure and it follows that f(z) =0 almost everywhere.66 THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS (omar. 4 LEBESGUE'S THEOREM ON BOUNDED CONVERGENCE 4.22. Prove Lebesgue’s theorem on bounded convergence [Theorem 4-15, page 56]: Let (fa(z)) be a sequence of functions measurable on E’ such that lim fa(z) = f(a). ‘Then if [fala] SM, lim f, faa) dx S teyae From the fact that [/,(2)| SM and lim fy(e fla), it follows that || =M. Then fle) {is bounded and measurable and thus integrable. ‘The result to be proved is equivalent to proving that Jim f° le) — fede = 0 o or since \S. tHe) — facade | =f fe) — p60) ae st will follow if we ean prove that im f Wie) steiide = 0 «) Let E be represented as the union of the disjoint measurable sets By = @ I hl0? Explain, Prove that if /(2) and g(z) are bounded and measurable on E, then S,ise~otenae = ff perae ~ foe ae vent f feyie = f paide+ f torte-S paren Generalize the result of Problem 4.54 to three or more sets Ey, Ey, Ey, If (x(z) b=1,2,..., are bounded and measurable on Band cy, k constants, prove that f [ 3 ants] w= 3a Stecorae are any row it [S sinsterde| = 3Lf meas +f inate 51) nd) 08 de nd mele Prove Thre 48 see 1 fx(2) and fala) are Bounded and mensurable in (a, 6) and Steeped = 0 rove that f,(2) = fa(z) almost everywhere in (a,b). State and prove a theorem analogous to that of Problem 4.28 if the interval {o,6] is replaced by a set E, Beane fsinede, Ame 2CHAP. 4] THE LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS 1 + de es oe ee 461, Let (2) be bounded and integrable on B. Prove that given «> 0, there exists 8 >0 such that if Ais any measurable set contained in E then S Woaride < © whenever may p e ‘Then for each p, [/(z)}» is bounded and measurable and thus Lebesgue integrable. We define the Lebesgue integral of f(z) on E as Star = tim f Verde ® ‘The limit is either a non-negative number or is infinite. If the limit is a non-negative number, then the Lebesgue integral of f(x) exists and is equal to this number and we say that f(x) is Lebesgue integrable or briefly integrable on E. If the limit is infinite we say that f(z) is not Lebesgue integrable on E’ or that the integral does not exist, THE LEBESGUE INTEGRAL FOR ARBITRARY UNBOUNDED FUNCTIONS If f(@) SO we can define the Lebesgue integral of f(z) as Si tear = -§ yeas ) where the integral on the right is obtained as above since |f(2)] = 0. In the general case where f(z) may have arbitrary sign, we let tiny. [fl@) forall 2H such that f(z) 20 re = i 0 forall e€E such that f(z) <0 od _ 0 forall 2 €F such that f(z) =0 re {No for all « €H such that f(z) <0 © where it is to be noted that f* (e) and f~(z) are both non-negative, ‘Then it follows that f@) = f(2) - fF) (6) This leads us to define the integral of /(2) over Eas the difference of the integrals of non- negative functions, ie. Ste = ~r@a-f rae ® 2CHAP. 5) THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 3 The integral on the left of (7) will certainly exist when the two integrals on the right exist and in such case we will say that f(z) is Lebesgue integrable or briefly integrable on E. It should be mentioned that in case f(x) is bounded the definition given by (7) reduces to that given in Chapter 4. [See Problem 5.43.] THEOREMS INVOLVING LEBESGUE INTEGRALS OF UNBOUNDED FUNCTIONS, Most of the theorems of Chapter 4 involving Lebesgue integrals of bounded measurable functions can be extended to the case of unbounded measurable functions. In general, proofs of such extended theorems are achieved by first using a limiting procedure as in (2), page 72, for the case where the functions are non-negative and then using the definition (7), page 72, to obtain the case where the funetions are arbitrary in sign. In the following we list some important theorems many of which are related to those in Chapter 4. Unless otherwise stated we assume that all sets and functions are measurable and that the functions may have arbitrary sign and may be unbounded. Theorem 5-1. If f(t) bounded. , thon f(a) de exists if and only if f° [/(2)]p de is uniformly Theorem 5-2. If |f(x)| = g(x) where g(2) is integrable on E, then f(z) is also integrable on E and S; vealae = S$. oeeyae ‘The result is also true if |f(x)| = g(x) almost everywhere on EF. Theorem 5-3. A function f(x) is integrable on E if and only if |f(2)| is integrable on E and in such ease IS fleax| = S [Fear Because of this we say that f(z) is integrable on E if and only if it is absolutely integrable on B. Theorem 54. If [ f(z) dz exists, then f(2) is finite almost everywhere in E. Theorem 5-5. If H has measure zero, then S fadz = 0 Theorem 5-6. uf f(z) dz exists and if A is a measurable subset of #, then S f(z) dx also exists. In such case we have S; weolae = J eae Theorem 5-7, If E=E\UE: where Ey and E: are disjoint, then S fle)dz = S, fla)de + S. fleyde ‘The result is easily generalized to finite unions of sets.4 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS (CHAP. 5 Theorem 58. If E=E\UE,U--- where E,,B,,... are mutually disjoint, then if aft fle) dee exists . Stemi = S pede + S seyae + ‘The result is not necessarily true if f(z) is integrable on Es, Bs, mention of its integrability on E [see Problem 5.16). . without specific Theorem 54. S; fe sods = SF fade + § omar ‘The result is easily generalized to any finite number of functions. Se een este The result is easily generalized to any finite number of functions. Theorem 5-11. It j(z) is integrable on H and g(z) is bounded, then /(2) (2) is integrable on BE. Theorem 5-12. If f(x) = g(2) almost everywhere on E, then S fade = ff ofeyaz and the existence of one integral implies the existence of the other. Theorem 618. 1 42) 20 and ff fe)de=0, then fe) =0 amon everywhere on ‘The result is also true if f(z) =0 almost everywhere on E. Theorem 5-14. If j(x) is integrable on E, then given ¢>0 there exist a 8 >0 and a set ACE such that if m(A) <6 [Si teaz| a(z) and if lim ex(2) = (2), then us = S {& te} dz = SS uel) de (8) ‘We also have the following theorem which is often useful. Theorem 5-18. It = tu(z) is such that |so(z)| = | > aul) »| M for some constant M and if 0(2) is bounded and measurable on E, then S fo 3 Sum} a= % S ‘o(a) ua() de (9) FATOU’S THEOREM Theorem 5-19 [Fatou]. Let (fs(z)) be a sequence of non-negative measurable functions defined on E and suppose that lim fx(z) = f(a). Then tim f flayde 2 f) fede 0) Note that the theorem is also true if lim f,(z) = f(z) almost everywhere. If (2) is not integrable, the left hand side of (10) is infinite THE MONOTONE CONVERGENCE THEOREM Theorem 5-20. Let (fa(2)) be a sequence of non-negative monotonic increasing functions on a set B and suppose that the sequence converges to f(z). Then tim f fotzyde = ff fa)az (a1) provided that either side is finite. The theorem can also be stated as a theorem on series as follows Theorem 5-21. Let w(z)=0. Then S. {& w()}de = provided that either side converges. S une) dae (12) APPROXIMATION OF INTEGRABLE FUNCTIONS BY CONTINUOUS FUNCTIONS Although every continuous function in (a, b] is integrable, an integrable function need not be continuous. However, an integrable function can be approximated by continuous functions. A fundamental theorem in this connection is Theorem 5-22. Let f(x) be integrable in [a,b]. Then given any e> 0, there exists a con- tinuous function g(2) in [a,b] such that Se) - e142 <« (13)76 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS [onar. 5 An important theorem which is related to Theorem 5-22 is the following. Theorem 5-23. Let f() be integrable in [a,b]. ‘Then lim J +h) flzijar = 0 () THE LEBESGUE INTEGRAL ON UNBOUNDED SETS OR INTERVALS Up to now we have considered the Lebesgue integral on bounded sets. In order to extend the definition to unbounded sets such as the intervals (a, »), (~,b) or (—»,~), we need only adopt a suitable limiting case of bounded sets. Let us illustrate by defining the Lebesgue integrable on an unbounded set such as (a, «). Suppose first that (2) is non-negative, i.e. f(x) = 0, and that (2) is Lebesgue integrable in (@,5) for all finite values of b. ‘Then we define Sl Heya = tim Sted (15) and say that f(x) is Lebesgue integrable, or briefly integrable, on (a, «) if the limit in (15) exists, If f(a) has arbitrary sign, then we define Seas = Sr emae- fra 1) where /* (2), f- (2) are non-negative and say that f(z) is integrable on (a,~) if each of the integrals on the right of (16) exist in accordance with the definition (15). Obvious extensions can be made for the intervals (—x,b) and (—-, 2). In the last one, for example, we consider (a,b) and then let a-—=, 6 > =. If the set B is not an infinite interval but is nevertheless unbounded, we can define THEOREMS FOR LEBESGUE INTEGRALS ON UNBOUNDED SETS Most of the theorems which have been proved involving Lebesgue integrals on bounded sets are also valid for unbounded sets. For example, we have [compare ‘Theorem 5-3, page 73]. . Theorem 5-24. A function f(x) is integrable on E if and only if |f(x)] is integrable on E regardless of whether E is bounded or unbounded and in such case we have S, teas = Sf Hela (18) ‘Thus f(z) is integrable on £ if and only if f(z) is absolutely integrable on E. Similarly we can show that the other theorems proved for bounded sets, such as the Lebesgue dominated convergence theorem and Fatou’s theorem, are also valid in case the sets are unbounded.CHAP. 5] THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 7 COMPARISON WITH THE RIEMANN INTEGRAL FOR UNBOUNDED SETS Unlike the Lebesgue integral, the Riemann integral of f(z) on an unbounded set E can exist even though the Riemann integral of |/(z)| does not exist on E. For example, although exists as an (improper) Riemann integral, the integral does not exist [see Problems A.82, page 170, and A.33, page 171]. Thus the Riemann integral may exist when the Lebesgue integral does not exist. However, we have the following important theorem. sing @ Theorem 5-25, If \f(2)| is measurable and |f(z)| is Riemann integrable on B, then f(x) is both Riemann and Lebesgue integrable on £ and the two integrals are equal. Solved Problems DEFINITION OF THE LEBESGUE INTEGRAL FOR NON-NEGATIVE UNBOUNDED FUNCTIONS 5.1. Show that the function [/(x)]» defined on page 72 can be written as min (f(#),p) where min denotes minimum, We have fle) for all #@E such that f(z) =p urel, = {9 ean eee shit (>? ‘Then if fle) is less than or equal to p, [/(el]p = fle) = min (f(z), p) while & f(#) is greater than bs [flelly = p= min fle), Ph. In either case [f(2)], = min (f(z), P} 52. Show that lim (f(e)}»= f(a). ‘This follows at once from Problem 5.1 since min {/(s), =} = fle). 53. Show that [/(2)]p is bounded and measurable, and thus integrable for each p if (2) is measurable. From Problem 5.1 we see that [f(2)|p =p and is thus bounded for each p- 1 [flellp = min (J(e),P) where fl2) and p are measurable, it follows from Problem 3:8, page 48, that [/(2)]p is measurable, ‘The last result can also be shown directly from the definition of # measurable function [see Problem 5.20).8 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS (CHAP. 5 54, Prove that {” ‘¢ exists and find its value. , Va Define the funetion ; ; aE for Wes p or 2B 1iph Welle = 7 9 sor Wa>p or 2< tpt ‘Then s “te = lim f * Helly ds = mp paz +f] = nfo det, | (3 +6- = z =6 ‘Thos the integral exists and has the value 6, Note that the above Lebesgue integrals can be ‘evaluated as for Riemann integrals because of Theorem 4-17, page 57. Note also that the Riemann integral, (R) Se 2, exists as an improper integral defined as anal" i = in [be [] = an[o—$e*] = 6 ta tena thecone nth aba ie es Are he 18, The alee her nt ioe Sao ered ents el 5.5. Graph the functions f(z) =1/V% and [f(2)|, 0<2=8 of Problem 5.4 and thus illustrate geometrically the results of that problem. ‘The graphs of f(z) and [f(2)]p are indicated in Fig. 5-1 and 5-2 respectively. The function [larly is often called a truncated function since [f(2llp=p if flz) > p. v v= fe) The value tin J Verde = 6 represents the limit of the area shown shaded in Fig. 5-2 a8 p> =. DEFINITION OF THE LEBESGUE INTEGRAL FOR ARBITRARY FUNCTIONS 5.6. Show that the functions f*(e) and /~(z), page 72, can be defined by f*(z) = max {f(z),0}, f~ («) = max (0,—f(z))}. He) for all 2€B such that f(z) =O res { We have 0 forall «EH such that fle) <0CHAP. 5] THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 79 ‘Then if f(2) = 0, we have f+ (2) = f(x) = max (f(2),0) while if flz) <0, f+ (2) ‘max (f(z), 0}. In either case f* (2) = max (f(z), 0). By similar reasoning we show that f~(z) = max (0,~f(2)} 5.7. Prove that (a) f(x) = f*(x)— f-(a), (b) \f@)| = f*(@) +f @). (a) We have ft(2)= fle) if fle) 20 and fo(2)=0 if fle) ZO. Then f(x) =f*(2)—f-@) if f(z) =. Similarly we have f*(2) = 0 if flz)<0 and f-(@) Ma) =f (@)— F(a) if fz) <0. ‘Thus in all cases f(z) = f*(2)—f-(2). -fle) if fle)<0, Then @®) TE fle) ZO, then [f(a] = fle) = ft@+I—G). It fle) <0, then [f(2)| = —fle) = £5 @)+/-(@). Thus in all eases [fe] = f+ (2) + (-@). THEOREMS ON LEBESGUE INTEGRALS FOR UNBOUNDED FUNCTIONS 58. Prove Theorem 6-1, page 73: If f(e)=0, then S fle)de exists if and only if S [Me)]pdz is uniformly bounded. . orb fan of [1 in on tat (fel = (fall, Soo aw tat teres co lirrs oes “ Sf woee o is @ monotonic increasing sequence, It follows that if this sequence is uniformly bounded, i. if there is a constant M such that for all p f vera < 4 then im s Ul@llpds = J. fle) de convey it de et tn ane eye) a hee S veelae = f tarde =, we obtain S, Keds = S ote) de ® which follows since f” p(x) da exists and since the integral on the left of (2) exists an a result of Problem 58 (®) This follows at once from part (a) on replacing f(z) by |f(2)| which is non-negative. Prove Theorem 5-4, page 73,for the case where f(z) = 0. Let $= E(ftz) =~), ie, the set of all « for which f(2) is infinite. Then using the definition of [fe], we have Uelnde = Ff yeenpae = f rae = pmisy Taking the limit as p+ ©, we have ® Now itil) 0, wt yao cnteaition sc Bont fe) et, Th we Se eae wane irae ees Se ya cr Prove Theorem 5-6, page 73,for the case where f(x) = We have J Verran = f tarde “ Then taking the limit as p+ ©, tt follows that S tora = f nee ® ive. the integral on the left of (2) exists since the integral on the right of (2) exists, Prove Theorem 5-7, page 78, if f(z) We shall prove the theorem for any finite union of disjoint sets, ie. where E = E,UEU--- UE and ;,,,...,B, are mutually disjoint, Since the theorem is true for bounded functions [see Problem 4.14, page 62] we have S,terinte = f° pepae so +f Uteigde Then taking the limit as p> «, we have as required Si reree = f pede sn +f pnde since f(z) is supposed to be Lebesgue integrable on H and thus on the measurable subsets By, By. By Prove Theorem 5-7, page 78, if f(x) has arbitrary sign. ‘The result of Problem 5.12 is true for f* (a) and f~(2), so that we have Sree =f reise a f pene frm =f redetint f reacCHAP. 5 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 81 Bild, 5.15. eee aoe en ‘ tee pede f Free, we stn te Sete Siterde =f teddes +f porte Prove Theorem 5-8, page 74 for the case where (2) We have by Problem 4.15, page 62, J, Weolpae J, Weeds +f vets + Si, Worinde + +f Wepae w ‘Then by letting p> «, we have Sines =f porte fp P Letting > © in (1), we find , ° Si orde = ff soars farted ® Now for k=1,2,... we also have . . S werd = fees so that on summing over k, ™ " i ely asf Uepde tines iia ete + J Keyan + ia S,Werioae = J) teres “f Sea) de + > Letting p-> ©, we thus obtain Si tod = J fede +f fayde +--+ a From (2) and (8) it follows that Siri = ftom f pages Prove Theorem 5-8, page 74, for the case where f(z) has arbitrary sign. From Problem 5.14 we have Sem = JS, oer S reas Sea = S,roe+S reds Ten atectine and wine he fact at ff de f Pande ends, weottin the required result S toe = J, reef He) de + +82 5.16. BAT. THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS (CHAP. 5 Give an example to show that Theorem 5-8, page 74, need not be true if no mention is made of the integrability of f(z) on E. Let fle) = for n=1,2,. 4 =i Rati <** m7 ‘Then for each we have ger jae = cere eel ers = *Gti-ws)-*(afa- eA) 7 Now Si oiae = 3 f° yealae Spent - AS 12 f(a) dz does But since the last series diverges, it follows that f {/2)| de is infinite and so rot exist o Prove Theorem 5-9, page 74, for the case where f(z)=0 and g(x) Let (2) = fla) + of). ‘Then if we let [fa) for all EE such that f(2) = p Wee = py torall 2€E such that f(2)>p with similar definitions for (g(2llp and [h(2)]y we have [he@l]p = (Np + [ately = [aap ‘Thus we have Sorte = J yonyar+ Sf torte = erty ate Letting pe, we find J horae tite + f ade = f nnse «fret f see hich shoe that s te J, wersocnde = f pores + fore ‘The result is easily extended to any finite number of functions.Hap. 5] THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 83 5.18. Prove Theorem 5.19. 9, page 74, for the case where f(x) and g(x) have arbitrary sign. In case f(z) and g(z) have arbitrary sign and A(z) = f(z) + g(z), the following six cases arise: () fe) =0, gfx) = 0, Ma) ZO (fe) =0, gz) <0, hie) <0 @) fle) <0, ga) <0, hie) <0 (©) fle) <0, (2) =0, he) ZO 8) fle) 20, 92) <0, he) = (6) fle) <0, gz) =0, Me) <0 Let By, Eq, ...,B4 denote the respective sets in which each of these cases arises. It is then clear that B= By. Now if we can show that J, (hea) + o(ay]de = J, [iee)] de + for k=1,2,...,6 the required result will follow by addition and use of Problem 5.15. We have already established (t) for the case k= 1, ie. the set E in Problem 6.17. ‘The result can also be established in a similar manner for any of the other values of k. As an example let us illustrate the case where k= 4, In such ease we can write M2) = flr) + plz) as (fan) + [Aca] fon the set By we obtain as a consequence of Si nose = § pate +f -neae ote) de © ote) ‘Then since (2) = Problem 5.17, Ung Thorn 510 thi co ~f, ote =f torte f sore or Simoes = Sf parde+ f owrae le. Si, versace = f trees foes Prove Theorem 5-8, page 78: (a) f(2) is Lebesque integrable on E if and only if W(2)| is Lebesgue integrable on E and (6) |S fleyae| = f° Wayiee, (a) From Problem 5.18 we have Snorer = freee Sree S,moiee = freee f ree ‘Then if f(x) is Lebesgue integrable on E, s0 also is |f(z||; and conversely if |f(z)| is Lebesgue integrable on B, 30 also is fle). (®) rom part (a) we have Par] = |f prmae-f reas e I le ve = [free +|f ree = frees f pee =f weras84 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS (CHAR. 5 5.20. Prove Theorem 5-11, page 74. Since g(e) is bounded, we have |g(=)| = Mf so that We) ote) = Me] ‘Then we have S, Wie) 9(a)) aes uf Mee) ae and the required result follows since f(z), and thus |/(z)|, is integrable on B. 5.21. Prove Theorem 5-14, page 74: If f(x) is integrable on E, then given «> 0 there exist a8>0anda set ACE such that it maya |G reraz] 0 there exists natural number py sch that (V2) = (My a= p Uap = Since lim f(a) = fle), we have = tim alle = ele ‘Thus by Lebesgue's theorem of bounded convergence, inf Uatoheae = Step te a) Now since [f,l2)lp = falt) we have J, fede = Si, teteinde ® ‘Then taking lim of both sides of (2), we have anf tera = gm f topes ®86 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS [cHaP. 5 But since the limit as m-» = of the right hand side of (8) exists as in (1), we have da S itoae 2 Sf teny ate a) ‘Taking the limit as p= in (4), we thus have as required Hin J ninde & Sf serae 5.24, Prove Theorem 5-20, page 75. Suppose that tim ff fs) eis Site, ‘Then by Fatou's theorem we have Sf fede = im f° jor ae w so that the left side of (1) exist, Then since fs) f(a), we have Si tints = Sf teate e s0 that on taking the limit as m+ in (2) we have im f fuinvas sf fara “ anf norte = f nase Sinuny tt Jed tte ten since J) $f) we ave by Dabnew's demise convergence theorem, inf nirae = fn tterae = Sf ferae ‘Thus the required result is proved, APPROXIMATION OF INTEGRABLE FUNCTIONS BY CONTINUOUS FUNCTIONS 1 el 5.25. Let I: =(q:,b:] be an interval contained in I ,b]. Define f(x) = if aa Prove that given ¢> 0 there is a continuous function g(2) defined on J such that See) - omar < « Choose points P;:a;—6 and Q):5 +6 in the intervals (a,a;) and (b,,6) respectively where 0< 5 <2 as indicated in Fig. 5-3. Let Ay and B, denote the points where’ z= ay, fle)=1 and =D, fz) =1 respectively. Construct lines PyAy and B,Qy-CHAP. 5] THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 87 Let o(z) be the function whose graph is represented by AP,A,B,Q,B in Fig. 5-8, i ° a Lt (ays ay~ 8 ale) = 1 a 1-@-by/s by ° te It is clear that g(z) is continuous in [a,b]. Furthermore we have S02) 00) a0 S( Sr" : and the required result follows. 526. Let i= [abs], k=1,2,...,, be m disjoint intervals contained in I= a,b]. 1, r€Un . ' A Let f(e) = Prove that given ¢ > 0 there is a continuous function g(z) 0, e@Uh.” defined on I such that S ie)-owjar < « Tet fxla) 1 eel ‘Then by Problem 5.25 there is a continuous function g,(2) such that for > 0 0, céh” Sie) neolae 0 there is a continuous funetion g(x) such that Se) -otajae < « Since E is measurable, there an open set ODE such that if «, > 0, m(E) = m(0) < mE) + 6 wo Now by Theorem 1-19, page 7, the set O ean be expresged as a countable union of disjoint intervals 1, s0 that m0) = 3 min 2) From (1) and (2) we see that me) = Zmuy +, $m) < mi +a88 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS (CHAP. 5 ‘Since the series converges we can choose np sufficiently large so that eed MY <4 cc) and mB) = 3 mi) < mB) +a w Now consider the intervals Ty, Ta). ny and Tet Ingl@) = - 6 2e Uh ‘Then hy Problem 5.26 there is a continuous function g(z) such that Sige) sede <4 o Let By=BMIy b= 1,2,..., s0 that B= 0 By. Now if k=1,2,..4mp we have since I= EDUBy a Swer-tyeiae = fo ve sgolae + J. Wa) — fy) de ae poet hreas = Sf ae = mH) = my — my Hf k= rgb, mot2, cy then since fyg(x)=0 for 2 J voiiae J We) — fag 2) ‘Thus we have on using (8) and (4), Pro nyorde = BF wer-naeaide + 8,5 wer-yeorae % n SS [mi — mE] + | Smile) 2 ke * % = = [3 may-man] + [men ~ 3 men] + 3, mo % ae A = [3 muy—me] +S ment Smo La J} + Be wen = 3 may — me +, 3, mu +, 3, mn Satata = % Thus on using (5) we find, f'er-oeee = ff Ie) — Ingle da + fing) — ola de Choosing « ~ «/4, the required result follows.CHAP. 5 THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 39 528. Let Ey, k=1,2,.. 1, te Ey Defi = { fing O19 ne By that given «> 0, there is a continuous function 9(2) defined on I such that n, be m disjoint measurable sets contained in I= (a,b). and f(2) = D af,(a) where «, are constants. Prove Sie) -ocaax < « By Problem 5.27 given « > 0 we can find a continuous function g}(z) such that Per olaide 0 there is continuous function g(t) defined on I such that Sie) -oejaz < « Choose subdivision points em Sm SS te SB Suppose thatthe largest of the values yes is denoted by 8 and let By erm Sf < wd KEW ‘These sets are measurable since f(x) is measurable, (1 2em Tat ie) = 19) PERE and deine Nay = 3 ver hale) 20 that Ma) = wer for #€ By Now by Problem 5.28, given «, > 0, there is a continuous function g(z) such that [hey -seaide 0, there is a continuous function g(x) such that S Wle)-o@\dz < « o,f fale) = { reo Let _ fe Na) = ~ 0, fz) 0" (2) — fol) and SP rmae = Snorer ete ry = [ri fe 1 fia) > a? Define Hale), fale) Sm Fy) = { aoe ‘Then Fy(2) and Fy(2) are bounded and measurable in {0,0}. ‘Thus by Problem 5.29, given «, > 0, there are continvotis functions 9) and gs(2) such that SP wier-oieide 0 + i rs A Si veo-wde < 4 ries Also since g(2) is continuous in [e,d], it follows that for every «, > 0 there is a 5 >0 such that for all x in (a, 0], lole+ 8) — gel) < whenever [hl <8 ‘where we can suppose that |h| is less than the minimum of a—e and d~b. ‘Thus Se+m ~ oats = f° Fath) ~ fla) = [fet h) — gle + W) + [ole+) — ge} + (ole) — F02)] from which we have ine +b) — flay] so that for |Al <5, Now Ia + h) — ofa + b+ [ole + A) — o(e)| + Ifa) ~ ot) rn Seth ses wide f* meth ~ oeidn J" e+) ~ alae + fie) ~ eae So ee) = ata ae + fate + moto iae + fie = o(0\ ae = Swe) ota de + foie) — otaiaz +f We) - om) ae eee eta = ha Choosing «, = 0/3, we see that S Wet Halas < «tor neo fe. in, [e+ m— folds = 0 LEBESGUE INTEGRALS ON UNBOUNDED SETS 5.82, Prove that if f(x) is measurable on an unbounded set E, then \S. Aa)ar| =f ela ero By Problem 5.19 we have J f0)¢e| ‘Then taking the limit as a —e and b-> = the required result follows.92 ‘THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS [CHAP. 5 Supplementary Problems DEFINITION OF THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 5.38. 5a, 5.40, 5.44 5.45. 5.6, oO for a> 0, Determine [fell if (a) fla) =a, () 8 Drove nf" 4E st a» Laeger and vase, An 4 Prove that °° % does not exist. Inventions the existence of () f" 92, wy fis. 4 is Lebesgue integrable in (0,1) or not according as g<1 or not. Prove that flz) = Show that if f(z) is bounded, measurable and non-negative, then the definition (2) on page 72 for the Lebesgue integral reduces to that given in Chapter 4, Work Problem 5.3, page 77 by direct use of the definition of a measurable function, i.e. without relying on Problem 3.18, page 48. Prove that (a) [fy 5 [lalla tions (1) and (2) of Problem 5, 1 OS (lends =f erleee ‘Thus verity equa- Prove that if f(z) £0 then (a) fellp = fe) and (6) feripde = f° sede st (0) i integrable on. * Find (a) f*(2) and (6) f- (2) for the function f(x) = 3+ e —e-2, —% (a) Prove that if /(2) is bounded and measurable, then so also are f*(2) and f(z). (b) Use the result of part (a) to show that the definition of the Lebesgue integral given by (7), page 72, reduces to that given in Chapter 4 in case f(z) is bounded and measurable, Show that if f(z) is unbounded and non-negative, then the definition of the Lebesgue integral given by (7), page 72, reduces to that given by (2), page 72, Let B be the interval (-8,8) and f(s) = (el 1)/at, Use the definition to determine whether or not f- fla) de exists, If it exists find its value Give a geometric interpretation to the result of Problem 5.45. THEOREMS INVOLVING LEBESGUE INTEGRALS OF UNBOUNDED FUNCTIONS Bat. 551. 552. 553, Prove Theorem 5-4, page 73, if /(c) has arbitrary sign (compare Problem 6.10). Prove Theorem 5-5, page 72, if (a) f(z) = 0, (b) f(z) has arbitrary sign. Prove Theorem 5-6, page 78, if f(a) has arbitrary sign [compare Problem 6.11) Prove Theorem 5-8, page 74, for the cases (a) f(z) <0, gz) <0, (b) fz) = 0, gz) <0, Ma) = He) + 9a) = 0, (6) Hla) <0, oe) = 0, M2) = fle) + ole) ZO, (@) fla) < 0, le) 2 0, Mz) = Mle) + 9(2) <0. Generalize Theorem 5-9 to any finite number of functions. Prove Theorem 5-10, page 74. Prove Theorem 5-12, page 74, for the cases (a) f(z) = 0, o(2) = 0, (6) f(2), o(2) have arbitrary sign.CHAP. 5] THE LEBESGUE INTEGRAL FOR UNBOUNDED FUNCTIONS 93. Prove Theorem 5-18, page 74. 5.55. If ¢4,¢y are any constants and f,(x),/2(2) are integrable on £, prove that Si tenerteattaiide = of ierde + of floras 556, Generalize Problem 6.55 to any finite number of functions aan. 1, Wed ode = 0, roe at fo) =o) ane eer on 558, Prove Theorem 6-15, page 74. LEBESGUE'S DOMINATED CONVERGENCE THEOREM 4559. Discuss conditions under which Lebesgue's dominated convergence theorem reduces to Lebeseue’s theorem on bounded convergence. Uno = 2 5 8/00 O52 < Un! or Bin ced (a) Prove that lim f,(z)=0 in [0,1] but that the convergence is not uniform. pole, o, and monotonic decreasing it f(z) = f(z:) for 22>. If f(a) is either monotonic increasing or monotonic decreasing, it is called monotonic. 9596 DIFFERENTIATION AND INTEGRATION [onar. 6 The jump of f(z) at 2» is defined as f(z»+0) — fzo—0) where f(x0+0) = Jim f(x+h), {(eo-0) = lim flzo+h). If f(2) is monotonic increasing, the jump at any point is non- negative. If f(z) is monotonic decreasing, the jump is non-positive. If f(x) is continuous at 2», then the jump at z» is equal to zero. If f(z) is monotonic increasing, then —/(x) is monotonic decreasing and conversely. Because of this result many properties proved for monotonic increasing functions are also true for monotonic decreasing functions. SOME THEOREMS ON MONOTONIC FUNCTIONS Theorem 6-4. If f(z) is monotonic increasing in [a,b], then there can be at most a finite number of discontinuities with jump greater than some given «>0. A similar result is true for monotonic decreasing functions. Theorem 6-5. If f(2) is monotonic increasing [or monotonic decreasing] in [a,b], then the set of discontinuities of f(z) ean at most be denumerable. FUNCTIONS OF BOUNDED VARIATION Suppose that the interval (a, b] is divided into n parts by use of the points of subdivision a= a << +++ 0 there exists 8 >0 such that flax + ha) — f(@)| < « (a1) for every finite set of intervals (21, te + hx) such that98. DIFFERENTIATION AND INTEGRATION [cHAP. 6 Simi H@-1) for 2, > 2-1). Then if g(u) is absolutely continuous in [f(a), f(b)], g(f(2)) is absolutely continuous in (a, 5). In connection with Theorems 6-15 and 6-17 it is of interest to note that even a continuous function of bounded variation may not be absolutely continuous, ‘THEOREMS ON INDEFINITE LEBESGUE INTEGRALS. Theorem 6-21. If f(x) is (Lebesgue) integrable on (a, }], then Fe = Sf fwau is continuous and of bounded variation in [a,b] Theorem 6-22. If g(2) is integrable on [a,b] and S ouydu = 0 for all x in [a,}], then g(x) =0 almost everywhere in [a,b]. Theorem 6-23. If g(x) is continuous and monotonic increasing in [a,b], then g’(x) is integrable on [a,b] and Sv@ae s 90) ~o(0) or on replacing b by x, Swan = 92) ~ 910) Theorem 6-24. A function F(z) is an indefinite integral if and only if it is absolutely continuous.CHAP. 6] DIFFERENTIATION AND INTEGRATION 99 Theorem 6.25. If F(z) is absolutely continuous, then S P@a = Fe) - Fe or replacing b by 2, S “(udu = F(x) — F(a) Theorem 6-26. 1? F(z) exists everywhere in (a, 6] and is bounded, then Sf Pear = FO) ~ Fe or S Pmdu = F(x) — F(a) INTEGRATION BY PARTS Theorem 6-27. Let F(z) and G(z) be indefinite integrals of the integrable functions f(x) and (2) respectively. Then S Meaeae = Fal, — S teceyae = F()G(b) — Fla) Gla) — S f(z) G(x) dz A related theorem is the following. Theorem 6-28. If F(z) and Gz) are absolutely continuous, then LPecwmd = recall - fr eeeae CHANGE OF VARIABLES Theorem 6-29. If f(z) is integrable in [a,b] and z= ®(u) is an indefinite integral of a non- negative integrable function (uw), then Sieae = Lroomeman where « and f are such that (2) =a, (8) H@(w) ow) du100 DIFFERENTIATION AND INTEGRATION [cHAP. 6 Solved Problems MONOTONIC FUNCTIONS 6.1. Let f(x) be monotonic increasing in [a,b] and suppose that subdivision points are chosen such that @ © 20 that f(z) cannot be of bounded ¥: However f(z) is continuous [see Problem 1.40, page 19]. Prove Theorem 6-7, page 96: If f(z) is of bounded vari bounded in (a, b) Choose as subdivision points of [a, 6] the points a,#,b where a bounded variation, there exists a constant Mf such that [a,b], then it is ‘Then since fiz) is of Lite) — F(a) + 40) ~ fa] = m and so We) flay] = Me ‘Then Well = fle) Ha) + Hla] = fle) — Fal + [laa] = M + (Ha}l i.e, f(@) is bounded. Prove that the sum of two functions of bounded variation is also of bounded variation. Let f(z) and g(e) be the functions and A(z) = flz)+g(z). Then since f(@) and g(e) are of bounded variation, there exist constants M, and My such that for all possible partitions Ms w 3 Meo = sea) 5 Me 3, loted ~ 06%) ‘Then since ha) — Alay] = le) — fxn] + fale) — 9%] is wwe have Sra meaol = 3B vteo—stea + 3 loved ~oler dl a + Me so that h(z) = f(a) +9(2) is of bounded variation.102 68. 69. 6.10. DIFFERENTIATION AND INTEGRATION (CHAP. 6 Prove that the product of two functions of bounded variation is also of bounded variation. We have en) oCeH) — Hea) eqD] = [fend ole) ~ Hlen—1) Olen) + Hous) Gee) — Hee) old Slaten) [Med ~ Fem] + (Men! len) — 9(ex—0| Pie) ~ Hlex—o| + love) — ole] where P and Q are upper bounds of g(2) and f(z) which exist by Problem 6.6. Then summing from k=1 ton we have Si vewaen—ferdoteo| = PS wed fea + 0S tote) ~ ole = PM, + QMs since f(z) and g(x) are of bounded variation. Thus f(e) glz) is of bounded variation, Let f(x) = G(x) — H(z) where G(x) and H(z) are bounded and monotonic increasing in [a,b]. Prove that f(z) is of bounded variation in [a,b]. We have Wee) flex] = (Ge) lew) — (Ge, — Hees = [Glo1) ~ Cle, a)] ~ (Hla) ~ Hers) 1G) ~ Clay) + |Hle)— Herp] Gly) ~ Olay g) + Hla) — Hl) since Gla) = Glzy—2), Hey) = Hla). ‘Then Weed Foxaol 2 3 (Gla) Gla} + 3 Hey = G(b) ~ Gla) + Hie) — Hla) from which the required result follows. We can also prove the result if G(z) and H(q) are bounded and monotonic decreasing, Prove that if f(z) is of bounded variation in [a,b], then f(z) can be written as the difference of two monotonic increasing functions G(x) and H(z), ie. f(z) = G(a) — Hic). Let v= 3 ley fend) ” and let us denote the differences f(x,) — f(y) briefly by af. We can decompose the sum (1) into terms where Af >0 and those where Af <0, ‘Then we can write (1) as v= Po laf + Pe \af| = BB and so we have v= Pty e where P 2 of N ~ 2, af (3) Now : PON = Bart Bar = Bike feo) = 10)-f0) «) From (2) and (4) we find V = 2P+ flay— fo), V = 2N4 f(0)— Fea) 6)CHAP. 6) DIFFERENTIATION AND INTEGRATION 108 Let U, P2,2Y denote the least upper bounds of V,P,NV in [a,b] respectively. ‘Then from (5) we have UE = 2h + fle)— 70), VY = INI+ M0) — Ka) C ‘These results hold for the interval [a, ]. Now if we consider the interval a,] we have similarly Us = BPE + fla)~ fe), Ue = BNE+ Mle) — Ha) o ‘Then solving for f(s) in (7) we find Ha) = fla) + PE— Ne Since PE and ‘NJ are monotonic increasing functions of # we can write Ge) = fo) +75, HG) = NE so that as required fle) = G2) — Hla) where G(z) and H(z) are monotonic incres We call UE the total variation of f(z) in [a,2], PZ the positive variation of fiz) in [a,#] and (NG the negative variation of f(z) in (ay). DERIVATIVES OF MONOTONIC FUNCTIONS 6.1. Let f(2) be a monotonic increasing function in [a,b]. Prove that B = (x:D* fle) > Dsf(e)) has measure zero. Consider the sets Ey = fe: De fle) 0 there is an open set ODE,, such that m.(E,,) > m(0)—« Now since D, f(z) <1, it follows that corresponding to every point of B,, there is an interval [e,2 +A] in O whose length is arbitrarily small such that e+ M~Ie) o, or fleth)= fle) < rh By the Vitali covering theorem we ean choose a set of m disjoint intervals Ty = [psy + Mily B= A.-m, auch that if A= Oy mE, A) < ( Summing over theso intervals we find Zitrti fen) < rh = rm0) = veto «) Let us now consider the set B consisting of all points of A = Since m,(B,,—B) = m,(E,,— A), it follows from (1) that m(By,—B) < « ) 1g except for the end points,104 6.12. DIFFERENTIATION AND INTEGRATION (CHAP. 6 Denote by C the set of all points common to Hy, and B, ie. C=E,,AB. There will be a set of intervals [y,y +1) contained in B whose lengths are arbitrarily small, In view of the fact that D* fiz) > s, we have for this set of intervals MEDI > gor fy ~ fy) > at ” Now since this set of intervals is a Vitali covering of C we can choose « set of disjoint intervals T= pr Yp + ply B= Ay send, for whieh (2) holds, i.e. Illy +b) ~ Hy) > ay aa © and such that if @ =O) Jy , miC~6) < ©) Summing both sides of (6) from p= 1 to q we have &, lf +5) — fap! > 6 Bh 2) Now since f(2) is monotonie increasing and each interval J, ig contained in some interval I, we have 3 le tho ~ rep) ® Bin +) ~ flap) ‘Then from (2) and (8), (lon +h) — Hed] 834 < 3 ttt) su) eto) ) Now since $ 1, =m(G), we sce from (9) that rete mG) = ‘Then we have © = mB.) 2 m(B_OB) + miE,,0B) = m(O) + mE. —B) mic) + m{CNG) + mong) + « ™ mG) + miC~G) + « = Hie Solving the inequality for x, we find ° < hte Since « can be made arbitrarily small, it follows that «= m,(E,,) = 0 as required. Let /(x) be a monotonic increasing function in [a,b]. Prove that S = (x: D* f(z) has measure zero. Suppose the contrary, i.e. m,(S)=\>0. Then if K is any positive number, we ean find for each 2 € 8 a set of intervals [z,x +h] such that, Heth) — fla) fain > K Now by Vitali’s covering theorem there exists a disjoint set of intervals J, = (2.2, + kyl bate woh that A= Oy miS~A) < « woCHAP. 6 DIFFERENTIATION AND INTEGRATION 105 6.13. 6.14. 6.15. For this set of intervals we have Slat Ia) — fhe) he b Hate + hy) = Ht 5 a < fete ‘Then summing from K=1 ton we have ma) = Sn fla) and so on are also zero. It follows from these results that Dif = Df = Dtf = DF almost everywhere, so that the derivative exists [ie. is finite] almost everywhere. Prove that a monotonic decreasing function has a derivative almost everywhere. If f(2) is monotonie increasing, then —f(2) is monotonic decreasing, ‘Then since D f(z) exists almost everywhere, so also does D[-j(z)] =—D f(x) exist almost everywhere, Prove that a function of bounded variation has a derivative almost everywhere. ‘This follows at once from the fact that a function of bounded variation ean be written as the difference between two monotonic increasing functions which by Problem 6.13 have derivatives almost, everywhere. ABSOLUTE CONTINUITY 6.16, Prove Theorem 6-15, page 98: An absolutely continuous function is continuous. If fle) is absolutely continuous, then given «> 0 we can find § > 0 such that LB Meth) slog] < «whenever the] <8 By taking m=1 as a special case we have ife+ h) —fle)| < © whenever Al <8 0 that /(2) is continuous,106 6.17. 6.18. 6.19. DIFFERENTIATION AND INTEGRATION [cHAP. 6 If f(z) and g(z) are absolutely continuous, prove that f(z) +g(2) is absolutely continuous. Since /(z) and g(2) are absolutely continuous, given «> 0 there exists 6 >0 such that Bvien+ hy fey < § whenever 3 thi 0 there exists 8 >0 such that Beet ho fed! < «whenever Elke) 0. or (2) <0 on some set of measure grester than zero. we see that the integral of g(x) on any open interval is zero, Now any open set F can be expressed as a countable union of disjoint open intervals B,B,.... Sinee = Smee = 3S wma = 0 Similarly if g(x) <0 we arrive at a contradiction. ‘Thus g(z) = 0 almost everywhere. 625. Prove Theorem 6-23, page 98: If g(2) is continuous and monotonic increasing in {a,b}, then g'(x) is integrable and Sf ra ae = a0 ~ 910) Since g(2) is monotonie increasing we have by Problem 6.13, ole +h) — ole) mk = ge) exists almost everywhere ‘Then by Fatou’s theorem it follows that. tin f OED =H Gf" year Bot the left sie sequal to an Lf" setnae ~b fea} S eerte 3 foe an} fo snes + f°" woes} 16 wade — Ef nae} and 20 the required result follows,CHAP. 6] DIFFERENTIATION AND INTEGRATION 109 6.26. Prove Theorem 6-24, page 98, for the case where f(z) is bounded. If fle) is bounded, ie. [f(2)| = M, then Feth) * Letting +0 we find since the left side approaches F(z) almost everywhere, that Pe] =u ‘Then by the theorem of bounded convergence we have fem = froma Ef" Rw = Af roe {Moan = AST renee Ef ranges BS root — HAS nna = Ef rae Tras fin f FREE an = hf Rede ~ nf rae = Fe) — Fla) Then using (1) we have : Ped = FO-F@) = Shan stat fora, Sf wo rojan = 0 ‘Then by Theorem 5-18, page 74, we have F'(z) = f(2) almost everywhere, 6.27. Prove Theorem 6-24, page 98, for the case where f(z) is integrable but unbounded. It is sufficient to prove the result for f(z) = 0. since it will then be true for f* (2) and f~(2) and can be extended to the ease where f(z) has arbitrary sign as in Chapter 5 by using the fact that fle) = ft (2) — f-(). He), for fle) = We nave wwe = Y fose Then since {/(2)]p f(a) so that fla) ~ [f(el]p = 0, wwe have S100) — Hojpae = 0 Thos * (fw) ~ [Vd], du = 0 “sonde = ES" voy ie. EFe) = FW) = [lellp — slmost everywhere110 6.28. 6.29. DIFFERENTIATION AND INTEGRATION [cHaP. 6 Taking the limit as p-> =, we see that F(x) = f(z) almost everywhere so that on integrating ‘we have frrod © [ree = ro -K0 ” However by Problem 6.25, Fa)de = Fb) — Fla) @ Pres = ro r0 = free te S . “(e) fa) de = 0 ee for all « in [a, 8], Then since the integrand is non-negative it follows from Theorem 5-13, page 74, that F'(z) = f(z) almost everywhere. Prove that if F(z) is an integral, then it is absolutely continuous. — Fe = ffs Tren Fla thy) — Fea) = fsa free = fH au Thus Fla th) — Fe) = fino) aw Summing from k=1 ton, Brethren) = 3 f° pola Se a0 where E is the union of the intervals By = [ps2 + Iu), 1m, ‘Then using Theorem 5-15 we have for any given «> 0, SFetho— Ped < whenever mis) = 3th] < i.e, F(z) is absolutely continuous, Prove that a function F(x) is an integral if it is absolutely continuous. Since F(z) is absolutely continuous, it is continuous and of bounded variation by Problems 6.16 and 6.18, Thus there are continuous monotonic increasing functions G(z) and H(z) such that F(2) = Gla) — Hla) [see Problem 6.51] and we have almost everywhere F(a) = Ge) — Hie) 5, @(z) and H(z) are integrable and thus F"(2) is integrable, ic. S Prod ” exists, But the integral (1) is absolutely continuous (Problem 6.28] and thusCHAP. 6] DIFFERENTIATION AND INTEGRATION i ote) =f P%du ~ Frey ® is absolutely continuous, Then we have almost everywhere vo) = 2S Pdr = PFW = 0 ‘Thus from Problem 6.21, (2) (@) can be written wnt, and from (2) it follows that ¢=—Pia) so that Fe) = fr wdu + Fea i.e, F(a) is an intogral. G 6.30. Prove Theorem 6-25, page 99: If F(z) is absolutely continuous, then i ‘This follows at once from Equation (2) of Problem 6.29, ie. “(x)dx = F(b) ~ F(a) f Piw)du = Fla) — Flay SPe@d = [rma = FO) - Fe INTEGRATION BY PARTS 631. Prove Theorem 6-27, page 99: Let F(z) and G(x) be indefinite integrals of the integrable functions f(z) and g(x) respectively. Then ff Feeds = Fea) YS re aeae Since an integral is absolutely continuous [Problem 6.28], it follows that F(z) Glz) is absolutely: continuous [Problem 6.61]. ‘Then by Problem 6.19 Eire) ow) = P(e) Ge) + Fl) @e) = fiz) Gla) + Fe) (2) exists almost everywhere and, by Problem 6.30, f * 2 me cee = Fea s0 that Ste) Ge1 + Fe gay dz = Fe Gta) Thus SPPoae = Fepae|e~ Sf" MeG~ de2 DIFFERENTIATION AND INTEGRATION (CHAP. 6 Supplementary Problems MONOTONIC FUNCTIONS AND FUNCTIONS OF BOUNDED VARIATION 632. Prove a result analogous to that of Problem 6.1 if f(z) is monotonic decreasing in (a, 0]. 6.38, Prove Theorem 6-4, page 96, for the case where f(z) is monotonic decreasing in [a, 2]. 631, Prove Theorem 6-5, page 96, for the ease where f(z) is monotonic decreasing. 635. If f(2) and o(2) are monotonic increasing functions in [a,b], determine which of the following fune- Sons ae ap mootoni ineresing in Bl () #6) + 6) () fe)~ at, (6) fle), (2 Fava, ole) * 635. Work Problem 6.35 if (a) f(z) and g(2) are monotonic decreasing, (b) f(z) is monotonic increasing but g(2) is monotonic decreasing. 637. Prove that the difference of two functions of bounded variation is slso of bounded variation. 638 Prove that the quotient of two funetions of bounded variation is also of bounded variation provided ‘that there is no division by zero. 639. Work Problem 6.9, page 102, if G(e) and H(e) are bounded and monotonic decreasing in [at 640, Are the results of Problems 6.9 and 6.89 valid if the word bounded is not included? Explain. G41. Is the converse of Theorem 6-7 true? Explain, 642. If (2) and g(z) are of bounded variation, prove that (a) f(z)—g(z) and (B) flel/ole), gle) #0 in. (Hint: For (8) first show that 1g(2), ol) #0 is of bounded 6.43. Prove that (2) = 2!~ 823+ 2x*—5e—6 in of bounded variation in (a) the interval [1,2], (6) the interval [=8,3], (¢) any finite interval. 6.44. Prove that a polynomial is of bounded variation i any finite interval. feose, OS 2 fez) Show that if D* f(z) is continuous at =a, then the derivative of /(x) exists at =a. Is the result still true if one of the other derivates is used? Explain. Let f(z) be a monotonic increasing function in (a, 6]. Prove that the sets (a) (x: D, f(t) < D~ fle)}, () (2: D- f(z) < D> fla)} and (e) (2: D_ fl) 0 is absolutely continuous in [a,b If f(z) has a derivative (a) everywhere, (b) almost everywhere in (a, 6], is it absolutely continuous? Justify your statements, Compare Theorem 6-18. Prove Theorem 6-20. Consider the Weierstrass function defined by He) = 3 ar cos(ttes) where 00 isodd (a) Prove that f(z) is continuous everywhere. (8) Write the series for [f(z-+ A) ~ fla)\/h as A, +B, where A, represents the sum of the first m terms of the series and B, is the remainder after m terme and prove that . leh) fle)! Het N= Se) 2 Ip) — Ay soe aw (6) Use the result in (0) to show that if ob > 1489/2. then f'G) does not exist, thos showing that f(z) is continuous everywhere but has a derivative nowhere. (tlint: To prove the inequality for A, in (6, use the meun-value theorem on cos [Ms(=-+ h)] — cos [bt {To prove the inequality for By, fst let. Bz ~ py dy. where py is an integer and 1/2 = qq < 1/2 and prove that If k= G2 q,yo then 260/8 © KS 265. Then prove that for K =m, cos [Be(@-+ AY] = (1 +2, cos [bFr2] = (-1)%*# cos [6%] where Aad < + (Bal > Garb and estimate B, using the first term of corresponding series]14 DIFFERENTIATION AND INTEGRATION [cHAP. 6 ‘THEOREMS ON INDEFINITE INTEGRALS oer, 6.68. 6.69. 670, en, 672, em, em. 615. 616, en, 678, 679, 6.0, emir ee fee ou) do. prove that f(z) = g(z) at all points of (a,b Prove (a) Theorem 6.26, page 99, (b) Theorem 6-28, page 20. Prove that /'(2) is integrable in (a, 6) if f(z) is of bounded variation. If f'G) is integrable in [a,b] and (2) is of bounded variation, is it true that S"x\dx = f(b) — fla) Is there a result corresponding to that of Theorem 6-23 for monotonic decreasing functions? Explain. 2 asin(U/e), 0< 2 = Ve _ Let fla) = { " < Prove that (a) f'(2) existe everywhere in (0, Ve but @) /") is not integrable in [0, VHTz]. Discuss the relationship with Theorem 6-25, page 98 [iint: For part (0) use the fact Uhat f(a) is Lebesgue integrable if and only if [/(2) is Lebesgue integrable | If F(2)= J fw) du where f(z) is integrable on [2,0], prove that the total variation of F(z) in tt) if Yo Prove Theorem 6-29 on change of variables, page 99. [Hint: Use Problem 6.61(b).] State sufficient conditions under which we can write Srey eave = seal - Soe) rerae Let f'(2) exist everywhere in [a,b] and suppose that fz) = gle) almost everywhere in {a, 6) Prove that f(z) = g(2) everywhere in [a,] if g(z) is continuous almost everywhere in (a, Let f(z) be integrable in [a, 0] and g(z)=0 be monotonic increasing in [a,b]. Prove that there is a number 7 in [a, b] such the Sf Hera ae = 0-0 f pares ion of this is given in Problem 6.80. This is often called the second mean-value theorem. A geners If in Problem 6.78 g(z) > 0 is monotonic deereasing, prove that Sree ae ate 0 f "fends If in Problem 6.79 g(z) >0 is any monotonic function, prove that Si Heraeyde = ger) f"He\de + 90-0) {KeesChapter 7 L? SPACES ‘The space of all functions f(x) for which |/(z)|", p=1, is Lebesgue integrable on {a, d], ie. . Swords < 0 is denoted by L?[a, b], or briefly L” if the particular interval is not required. In such case we say that (2) belongs to L? or briefly f(a) EL’. If p=1 we denote L” by L. Although we shall use the interval [a,}] in this chapter, all results obtained can be restated using any measurable set E instead of (a, 5]. HILBERT SPACE ‘The space L? consisting of all functions f(z) for which SL veoras < 2 is called Hilbert space. Functions belonging to Hilbert space are often said to be square integrable. IMPORTANT INEQUALITIES 1. Schwarz's inequality IS. "120d = {vera} {Sf iawrrae} where f(x) € I, g(t) € 1, Equality holds if and only if /(2)/9(2) is constant almost everywhere. 2. Holder's inequality |S reroeee] = {L"terras}” { Pioermacl™ where + and f(x) EL", g(x) € Le, Equality occurs if and only if {f(x)//\g(«)|* is constant almost everywhere. Note that if p=2, q=2 this inequality reduces to Schwarz’s inequality. = 4 p>l zie ain 15116 MEAN CONVERGENCE (CHAP.7 3 Minkowski’s inequality Sver+oe@rae = {fwerrar}” + {facnpas}” where p=1. Equality occurs if f(z)/g(2) is constant almost everywhere. Analogous inequalities exist where sums of real numbers are used in place of integrals [see Problems 7.26, 7.82, 7.45). ‘THE L’ SPACES AS METRIC SPACES If we define a distance between two functions f(z) and g(2) in L?, also called a norm in L’, as cab 1 py.o) = Wah = {Sf Me) ~aterrae} then we can show that L” is a metric space [see Problem 7.12]. For example the triangle inequality involving elements f(x), g(x) and h(z) in L’ can be written Wal) = =H + [atl and is a consequence of Minkowski’s inequality. ‘The functions are often called points of the space. Since two functions which are equal almost everywhere have the distance between them equal to zero, we shall assume that they represent the same point in the space. ‘When it is necessary to specify the space, we shall write ||f — g\lp rather than ||f — 9) THEOREMS INVOLVING FUNCTIONS IN L’ SPACES Theorem 7-1. If f(z)€L’, p=1, it also belongs to L, ie. L'CL. More generally if p>m=i, then L°CL Theorem 7-2. If g(x) €L? and |f(z)| =|g(z)|, then f(z) €L’. Theorem 7-3. If f(z) €L” and g(x) €L?, then f(z) g(x) € L’*. In particular if f(x) € L* and g(x) €L*, then f(x) g(x) €L. Theorem 7-4. Ut f(a) EL? and g(e) EL" where $+4=1, p>4, then fle) (2) €L, Theorem 7-5. If f(t) EL? and g(2) L’, then f(z)* g(a) EL’. MEAN CONVERGENCE Let (fa(z)) be a sequence of functions which belong to L”(a, b). If there exists a func! fx) EL? such that im f fala) — fl@Ppdx = 0 w we say that the sequence (fa(2)) converges in the mean or is mean convergent to f(x) in the space L*. In such case we often write (1) as Lim. fala) = fle) which is read “the limit in mean of fa(2) as n> = is f(z)".CHAP. 7] MEAN CONVERGENCE 17 Equivalently we can say that f,(2) approaches f(z) in the mean if for every «> 0 there exists a number no > 0 such that \lfatz) — f(@)\| < © whenever n > mo Theorem 74. . fa() exists, then it is unique [see Problem 7.18). CAUCHY SEQUENCES IN L’ SPACES ‘The sequence of functions (f,(z)) belonging to L’ is said to be a Cauchy sequence if tim [Yate ~ folnrde = 0 io or, in other words, if given « > 0, there exists a number ny > 0 such that \fm ~ fell S; Wale) ~falatdz < ¢ whenever m> no, n> no Note the analogy with Cauchy sequences of real numbers. We have the following Theorem 7-7. If (fa(x)) converges in mean to f(x) in L’, then (f«(2)) is a Cauchy sequence. COMPLETENESS OF 1’. THE RIESZ-FISCHER THEOREM ‘The space L” is said to be complete if every Cauchy sequence in the space converges in the mean to a function in the space. The following theorem, known as the Riesz-Fischer theorem, is of fundamental importance. Theorem 7-8 (Riest-Fischer]. Any L? space is complete. In other words if (fa(x)) is a sequence of functions belonging to L’ and jim Pirate) — fala) dx 0 ie. if (fa(2)) is a Cauchy sequence, then there is a function f(z) €L” to which fx(2) converges in the mean. This function is unique apart from a set of measure zero. CONVERGENCE IN MEASURE Let (fa(z)) be a sequence of measurable functions defined almost everywhere. Then (faz) is said to converge in measure to f(x) if 5 lim m(a : |fa(2)~f(@)| 2 0} = 0 for allo > 0. a ‘The following theorems hold. Theorem 7-9. If (fa(2)) converges almost everywhere to f(z), then it converges in measure to f(z). However, the converse is not valid but see Theorem 7-11. Theorem 7-10. If (f4(z)) converges in the mean to f(z), then it converges in measure to f(x). Theorem 7-11. If (fa(z)) converges in measure to f(2) on a set E, then there exists a sub- sequence (/,,(z)) which converges almost everywhere to f(z).118 MEAN CONVERGENCE (CHAP. 7 Theorem 7-12. If (fu(z)) converges in the mean to f(x) on a set Z, then there exists a sub- sequence (f,,()) which converges almost everywhere to f(z). It is of interest that if (f.(2)) converges to f(z) everywhere, it does not necessarily converge in the mean to f(z). Conversely not necessarily converge almost everywhere to f(z). See Problem 7.16. L? SPACES TA. Prove that fla) =1/V# (a) belongs to L[0,8] but (>) does not belong to 1*[0, 8}. 12 if /a(z)) converges in the mean to f(z), it does Solved Problems (®) We must investigate the existence of (a) The fact that f(2) belongs to L[0,8] follows from Problem 5.4, page 78. 2-2 Since 1/2 is unbounded in [0,8], we proceed as in Chapter 5. Define ‘Then [Ve for esp or x= Up (ale = | y for We>p or 2< ip ey lim," (W/z]p dz mle ft tim [plQ” + aie alk, tim [1 + In + Imp] which does not exit, Thus fla) = 15/2 does not belong to 19,8) It f(z) €L* and g(x) EL, prove that f(x)g(a) € L. We have 0 that ‘Then (see)| — late? = 0 a [rea — 21f(@| oz] + ot = 0 Wee ioe = Faye + lovely [nee 92) S MAP + we @ Wieoteldr = af Yertds + 4 f lovee “ Since f(x) €L? and g(2) € L2, the two integrals on the right of (8) exist so that the integral on the left of (8) exists, Note that the result is M2) gf) € Le special case of Theorem 7-8 page 116, where pCHAP. 7] MEAN CONVERGENCE 119 78. If f(t) ELA, prove that f(z) EL, ie. LCL. ‘This follows at once from Problem 7.2 by noting that g(2)=16 LA, Note, however, that the result is not valid if the interval (a,b) is infinite, INEQUALITIES 74. Prove Schware's inequality 15. 18. |S 1ero(2)¢e where f(x) €L?, g(x) € Lt. = {fverrarh” {f" cnyeae For all real numbers \ we have, using the fact that f(z) 12, g(x) EL? and f(a) ole) €L, Sse +ouyree 2 0 he, A+ BB+ ry where A= Syerta, B= Spode, = ® Now since A> 0, (1) can be written we Brek soo (x8) + ACee 4ayta 4 ae But this last inequality can be true for all real \ if and only if AC—B*=0, ie, B2S AC. Thus , ™ {f 12) 2) ex} 5 {f nee rae} {S'iawrree} Tees ies Prove that if f(z) €L#, g(x) € LA, SL eo aac = {Sr veorrael"{ f° toeypac} This follows at once from Problem 7.4 on replacing /(2) by [f(2)! and g(2) by lola). fH) 06) ae (a) If wand v are non-negative real numbers, then their arithmetic mean is defined as 4(u +) while their geometric mean is defined as the Vuv. Prove that the geometric mean is less than or equal to the arithmetic mean. (0) If we consider n; non-negative numbers all equal to u and mz non-negative numbers all equal to v, show that the geometric mean will be less than or equal to the arithmetic mean if and only if mu + ne m+ Me (c) Show that the inequality in (6) is equivalent to (um myivenitan wv? Sau + pv where a, are such that «+= 1,120 MBAN CONVERGENCE [cHaP. 7 (a) We must show that Vi = duty of utv—2w = 0 ie. Wa-var BO But since this last statement is clearly true, we can reverse the steps to obtain the required inequality. (b) The geometric mean of m positive numbers 2,5, ... 2%, i8 defined as Vawter En = Ceate tQ)!™ while the arithmetic mesn is defined aptapho ty, ‘Then the geometric mean of m, numbers equal to w and m, numbers equal to is (ur ompizontnn while their arithmetic mean is nye + ge so that the geometric mean is less than or equal to the arithmetic mean if and only if mater mm (er umyintnd = (©) The inequality in (b) ean be written suming gmeleni tm) Then letting « = 7 B= so that a+@=1 we obtain as required, wie? Sau + Bo ‘The validity of the inequality in (8) showing that the geometric mean is actually less than or equal to the arithmetic mean, is shown in Problem 7.7. 7.1. Prove that for any non-negative real numbers u,v and positive numbers a, such that «+f=1, uv? Sau + Bu thus demonstrating the truth of the inequality in Problem 7.6(b). If w=0 the inequ 1 ity is trivial so that we can assume w>0. Then letting = ux and , the required inequality can be written Ltpr—2—p = 0 ® Denoting the left side by F(z), it follows that the derivative is F'(z) = (L— 2°). Now F(z) =0 for 2=1, Thus since F(z) <0 for 0<2<1, it follows that F(x) =0 for 00 for 2>1, it follows that F(z) >0 for #>1, i.e, the inequality a) is true for z>1, Also equality holds for == 1. ‘Thus (1) is true for all z>0 and so the required inequality is proved.CHAP. 7] MEAN CONVERGENCE 121 78. Prove Holder's inequality |Srerorrae| = {f° nerrar}"”{ f° ereas}"™ where p>0, q>0, I/p+I/q=1 and f(a) EL’, g(x) EL’, In the inequality of Problem 7.7 let u= lle ys lol J 1 wd Fo o” P” q where F = fwd, = fierce w inate” Foie” pele, totzyle Then {ge} Gf © prt GG @ ‘Thus by integrating from a to b and using the definition (2) of F and G, we have from (2) * fine!” [iow Lyd 8 SHRP” (eg Y ae Leta a Since F and G are independent of z, we ean multiply both sides of (8) by F/G" to obtain ii Wel lola) de = {Slveerr ax} { f 7 corrae ‘The required inequality follows on noting that [f° reo a0 £ Gate) de 79. Prove Minkowski's inequality {Ste +oerrach” = {f° yeas} + ff" oceieaey” where p= 1. If p=1, the inequality is trivial. For p>, we have using Holder’ Svtoras = f° V+ ollf+olP-tae " Snirtorter + Siu V+ oP-ae {fo mee} {Sirearae}” ne : {fine} {S'vsoree eae Then avitng tats ste oy {fir+oirdeh” ” which ix assed tobe cert from sere 4 since otherwise the inequality is trivial], the required result follows. THEOREMS INVOLVING FUNCTIONS IN L” SPACES 7.10. Prove Theorem 7-2, page 116: If g(x) €L? and |f(2)|=|9(z)|, then f(2) €L?. If [f(2)| = Jo(e), then [fe))» = lola). Thos 50 that if the right side exists, Le. if (2) €L*, then the left side also exists, ie. flz) € L?.122 MEAN CONVERGENCE ouar.t TAL. Prove Theorem 7-3, page 116: If f(z)€L? and g(x) €L’, then f(x) g(x) EL". Since f(#) LP and gle) €L?, we have Sewer < =, S"wieras < i.e, the integrals |o(e)%, we have . Then by Schwarz’s inequality on replacing flz) by [fiz)|»/® and g(a) by if Wien? de} {Steve ae a) gz) €P", S Her atayoreae ‘Since the integrals on the right exist, the integral on the left exists, THE L’ SPACES AS METRIC SPACES 712. Prove that an L? space is a metric space. ‘The distance betwoen two elements (functions) of L? is defined as vin = {f ste) —atalr ae In order to prove that L” is a metric space, we must satisfy the requirements 1-4 on page 6, ie. 1. Dig) = 0 2 Dif.a) = Dif) 8 Dif.g) = 0 if and only if f= 0 4. Dif,9) 5 DU.) + Dib, 9) Now requirements 1 and 2 are obviously satisfied from the definition of D(f,g). Requirement 3 is satisfied if we consider the distance between f and g as zero, ic. if and only if the functions have the same values almost everywhere and we shall henceforth use this interpretation. Finally requirement 4 follows as a consequence of Minkowski’s inequality. MEAN CONVERGENCE, CAUCHY SEQUEN RIESZ-FISCHER THEOREM 713. Prove that if a sequence (f,()) converges in the mean to two different functions f(r) and g(z), then f(«) = g(z) almost everywhere. ‘Thus if Lim. f.(z) exists, it is unique. CES AND THE If Gu(2)) converges in mean to f(x), then sin fF Vale —Heras 0 w Similarly if (/,(z)) converges in mean to g(2), then tin J valey-ovmirde = 0 ® {Ser —acrrae}” {f" I=) — fala) rae)” + fr jac) oter ah” Thus letting n+ and using (1) and (2), we have Sf ve-sarae = 0 or flz) = oz) almost everywhere.CHAP. 7] MEAN CONVERGENCE 123 Td. 715. 7.16. Prove Theorem 7-7, page 117: If a sequence (f,(z)) converges in the mean to f(t) in the space L’, then it is a Cauchy sequence. By Minkowski’s inequality we have {JF tai —ncovasl”™ 5 V2) fara} + {fim — fn cjranl” or Wm Fall = hm — Sl + Wn 2i) But since (fm(z)) and (f4(2)) both converge in the mean to f(a), it follows that given «> 0 there exists my >D such that form > mq, > toy Um~M < $+ Weal < § Then Fn +h = 6 for m>ny n> M9 and the required result follows, If Lim. f.(x) = f(x), prove that tim fisiryrae = SL" airae or equivalently, lim |f| = |j/|- ‘We have from Minkowski’s inequality, Wall = Wa-A+Hll S Utes + Al or taking the limit as n=, Jim Iifall = HIM) ® Similarly Wal = Su Fall SUF Sull + all or taking the limit as n> ~, WAL = iow tial © ‘Then from (1) and (2) we have Bim Ufa = fil ‘The result is often interpreted by saying that the norm of the metrie space L? is continuous Let m, O Izy= mq we have q(t) 0]=0my, Le. lim f,(2)=0 for all zy in [0,1] () IE Lim, fae) then we would have tin {"intorae = 0124 711. MEAN CONVERGENCE (CHAP. 7 However, Si ateitde = fuer + fords = and s0 in fi iintevtae = = Lm. fale) # 0 (©) The result shows that a sequence /,(2)) may converge to f(z) at all points of an interval without converging in the mean to f(z) on the interval. ‘The example ean be extended to the case L? where p> 1_ [see Problem 7.51] Conversely we can show that a sequence (fq(2)) may converge in the mean to f(z) on an interval without converging to f(2) in the usual (pointwise) sense [see Problem 7.54). Prove the Riesz-Fischer theorem [Theorem 7-8, page 117]. Since the sequence of functions (f(z) in L? is a Cauchy sequence, ie. din S imt-faelrde = 0 it follows that to each natural number » there is a smallest natural number n, such that . . 1 2 2 Si itmted~ fue gs @ ‘Then S vei-f,crae =f (&) ae = Ley ® Combining this with (1), 1 1 F 2m) < Xo mie) < @ Consider now ByUByyUe = fe Uauer Fog! > gia» Vay ee foveal > sorte + } © It x doesnot belong to this set, it is clear that Vauss Soul © gam Vanes fond! © gorda ‘so that the series 6) convernes. But the measure of the set (6) is less than or equal to m(Bx) + miByys) $e GIN E GIN Ho = 8G which approaches zero as N~ ©, ‘This shows that the set of all x for which (5) does not converge is of measure zero or, in other ‘words, the series (6) converges almost everywhere.CHAP. 7] ‘MEAN CONVERGENCE 125 Now since the series (5) converges almost everywhere, so also does the series 3 Movs ~ fay) since an absolutely convergent series is convergent, Writing +S as he) it follows that Him fag = fay + 3 Waysr fo) exists, Denoting this limit by f(s) we thos see that Jim fale) = fle) almost everywhere We must now show that (f,(2)) does in fact converge in mean to f(z). We first prove that Jng(#) converges in mean to f(z). To do this we observe that by Fatov’s theorem, i Lingle) fale de = Sirage)— fleas ny But from the fact that (f4(2) is a Cauchy sequence we see that for every «> 0 there is a number K such that, Sine htelede <6 for k>K, o> K ‘Thus from (6), Soyo) feoirde Se for eo which shows that /,,(z) converges in mean to f(2) or equivalently a J ig@-tepae = 0 ‘To show that f,(z) converges in mean to f(z) we note that by Minkowski's inequality if Yate)— rey ae = {Stor tacrvae}™ + {J vaser-nenae} s0 that tin J inte—feyr ae = 0 i.e. fa(2) converges in mean to fle). ‘The limit function f(2) is unique apart from a set of measure zero or, in other words, if there are two limit functions f(z) and g(z) then f(z) = g(z) almost everywhere. CONVERGENCE IN MEASURE 718, Prove Theorem 7-10, page 117: If (fn(z)) converges in the mean to f(z), then it con- verges in measure to f(r). Let ¢>0 and B, = (2: |f,—f/ =o}. Then Ff ner—neiede =f orde = ovminy Since the left side approaches zero at >=, we sce that im m¢ in measure to (2. = ive. (fa(2)) converges126 719. MEAN CONVERGENCE (CHAP. 7 Prove Theorem 7-11, page 117: If (/a(2)) converges in measure to f(z) on a set B, then there is a subsequence fs,(7), fag(x), ..., mi a; >> ++ such that lim a =0 and let ny +z} ++ be a convergent series in which each term is positive, ae Choose the natural number m, such that for all x € B, met ay Bad <9 This number must exist because by hypothesis Jim me: afl Zoi) = 0 Similarly let ng > be @ natural number such that m2 Wag fl Boa) my> +++ > mj such that Mag H Bad < a mt This procedure defines the sequence (nm). ‘We now prove that lim fx,(2) = f(a). To do this suppose that bs i. Py &, fe: ly -N2od, @ ‘Then since Py>P;>P;++, it follows from Theorem 2-15, page 23, that m(Q) = Yim m(P) Co) wow sce mie) = So, 1 tvs that fy = 0 and trom () mQ) = 0 If we can now show that Bm fyle) = 1) for all s€E~Q, the required result will follow, ‘To do this suppose that 2)€H—Q. Then ‘there is some natural number jg such that y @ P,,, so that from the definition of P, % EE: yg NE ond for all k= jy It follows that Mnglte) — Heal] < om @ But since +0 as k->2, we see from (2) that fim fale) = fled) ve. for all se E~Q Him fyf@) = fe)CHAP. 7] MEAN CONVERGENCE 127 Supplementary Problems L? SPACES 120, ra. 22. 123. 1a, Prove that the function f(z) of Problem 7.1 (a) belongs to L?[0,8] if 13 p<3 but (b) docs not belong to L°[0, 8] if p= 8. (c) For what values of p does f(z) = Vz belong to L*(0,1]? Prove that if fiz)€L’, p21, then ef(e)€L? for any constant ¢. If fla) EL! and g(x) ELA, prove that (a) (f(z) + 9(2)] € LA, (®) [fl2)— ola] E12, If fy(2), fol), «+ fol) all belong to L* and cy, 0y)-+-)¢, are any constants, prove that eyfy(2) + cafe) +++ + egfglz) © 12. If a function belongs to L, does it also belong to £7 Explain. INEQUALITIES AND THEOREMS INVOLVING FUNCTIONS IN L” SPACES 135. 1.28. aa. 129. 730, 131, 132, 133. 138, 135. Prove that the equality holds in Schwarz’s inequality if and only if f(z)/a(z) is constant almost everywhere. Prove the following analog of Schwars's inequ: Jayby + + + aybyl (lay? + s+ + fag l*ibal® +--+ + byl) and show that the equality holds if and only if ay/by is constant. ity for real numbers a, by, fe = Obtain Schwarz’s inequality for integrals by using Problem 7.26 and the definition of an integral as a limit of a sum, Prove that if f(z) € a[f terrae] We] dz ‘Thus show that L°c L. Prove that f(z) = e~*/Vz€ L[0, »}. Prove Minkowski’s inoquality for p=2 by direct use of Schwars's inequality, Prove that the equality in Minkowski’s inequality hold everywhere. and only if f(z)/o(z) is constant almost TE ays byy be {lay + bP + oe + lat 60) = flay) ++ where the equality holds if and only if a,/6, is constant, Discuss the relationship with Minkowski’s inequality. 1+. are real numbers and p > 1, prove that + lag) + (Ube loa?) Prove that |j/-+4ll = [f+ llall_ and intorpret goometrially. Discuss the significance of lif|lp in ease p=. In this ease what is [lf—olly? Is L” a metric space? Explain,128 138, 131. 138. 738, 140, 7a 142, 7.43, 14, 146, MEAN CONVERGENCE, CAUCHY SEQUENCES AD 1a 748. 151. 152. 153, MEAN CONVERGENCE (cua. 7 Prove that if f(a) €L, p>1, then S verias = apne» Mee) ae| ‘Thus show that I°CL for p of Theorem 7-1, page 116. 4, generalizing the result of Problem 7.8 and proving the first part LCL if p>nZi. Prove the second part of Theorem 7-1, page 116, Discuss Holder’s inequality in ease p * cose 4,| e Yaeri | Prove that the equality in Holder's inequality holds if and only if f(z)/g(2) is constant almost everywhere. Prove Theorem 7-4, page 116, Prove Theorem 7-5, page 116, Prove that if f,(2)€L? and fl) €L?, then eyfy(z) + exfsla) € L? where e, and e, are any constants. Generalize the result of Problem 743 to n functions f,(2), ++ fal2)- State and prove an inequality for real numbers corresponding to Holder's inequality. Prove that the geometric mean of any set of non-negative numbers is less than or equal to their arithmetic mean, Is there a corresponding result involving integral ND THE RIESZ-FISCHER THEOREM If q(2)) converges in mean to f(z), prove that |fl| is bounded, Uf Lim. fa(2) = fle) and Lim. on(2) = 9(2) whore all functions are in L’, is it true that (e) ian, al2) + onCel] = f10)% o(2) (B) Liam, fC) one) =H) oe) () Him [fale + 992) = ie) * of]? Explain, Explain the relationship between Theorems 7-7 and 7-8, page 117. Prove that a sequence of functions in Hilbert space converges in the mean to a function in Hilbert space if and only if the sequence is a Cauchy sequence. Obtain a generalization of the remarks of Problem 7.18 for L? space, p> 1. Prove that if (fq(2)) converges in mean to f(2) in L” and g(2)€L*, then tin fp@oeree = fre oerde A. sequence (fa) in L? in anid to converge weakly to f(s) in L? if for every function ws) €L’, where ip +a = 1, . ° tim fo weer fuera = fle) see ade Prove that if (/_()) converges in the mean to f(x) in L’, then it also converges weakly to /(2).CHAP. 7) MEAN CONVERGENCE 129 734. 755. 156. Give an example to illustrate that if (f,(2)) converges in the mean to f(a), it does not necessarily converge almost everywhere to f(z). Under what conditions will $ exfx(e)€L? if u(x) EL?, = 1,2,..., and oy are constants? Justify your conclusions, "= itnve (©) Does fq(e) E14 for w= 1,2,...1 (6) Does Jim fale) €L22 (@) Does i cuss your results in connection with the Riess-Fiseher theorem. Let f(a) = for 05231 where n=1,2, (@) Is (fq(2)) & Cauchy sequence? fale) €L27 Dis CONVERGENCE IN MEASURE 151. 158. 159. 180, 761. Prove that if a sequence of functions (f,(z)) converges in measure to two different functions f(z) and g(z), then f(z) = o(z) almost everywhere. Can you aay that convergence in measure is, unique? Explain. Prove Fatou's theorem [see page 75] if the sequence (/,(z)) converges measure to f(z). Suppose that (fy (z)) converges in measure to (2) in Hilbert space. Prove that if |[f| is uniformly ounded, then (Jn (2)) converges weakly to (2) [see Problem 7.58]. Is the result of Problem 7.59 true for L” spaces in general? Explain. Prove Theorem 7-9, page 117.Chapter 8 DEFINITION OF A FOURIER SERIES Let f(2) be Lebesgue integrable in (—x,x). Suppose that f(x) is defined to be periodic with period 2 outside of this interval, ie. f(x 2ks) = f(a), k=1,2,.... The trigonomet- rie series . $ + Y (@rcosns + by sinnz) @ is called the Fourier series corresponding to f(a) if a = 1f f(z) cosnedz, by = 1s f(z) sin-ne dx @) We call a, and b, the Fourier coefficients corresponding to f(z). Since f(z) is periodic with period 2z, any other interval of length 2x can be used instead of (—x,z) such as for example (0,2z) or in general (c,¢-+2n) where ¢ is any constant. In such case the integration limits —x and = in (2) are changed to c and ¢+2r respectively, Extension to the case where f(x) has period 21 where 1 > 0, is easily made [see Problem 8.36]. THE RIEMANN-LEBESGUE THEOREM If f(a) is integrable in (a,b), then tim ['f(a)coserde = 0, tim f'f(a)sinaxds = 0 @ In particular the Fourier coefficients (2) approach zero as ><, CONVERGENCE OF FOURIER SERIES An important question which naturally arises is whether a Fourier series corresponding to an integrable function f(z) will converge and, if it does, whether it will converge to f(z). To answer this question we need to consider the partial sums of the series (1) given by $ + = (a, cosne +b, sin ne) «) Using the coefficients (2) we ean show [Problem 8.7] that Sue) = r" He 0) a 6) or [Problem 8.3] Sula) = S ett) + fle—9) se at Cy 130CHAP. 8] APPLICATIONS TO FOURIER SERIES 181 Subtracting S(x) from both sides of (6), we then find [Problem 8.9] that See + He—t ~ 254) OLN Su(2) = S(z) at a) ‘We thus arrive at the following important Theorem 8-1. The Fourier series converges to S() if and only if the integral in (7) approaches zero as M~> =, In such case lim Su(#) = S¢ Other important related theorems are the following. Theorem 8-2, The Fourier series converges to S(x) if and only if for some fixed number such that 0<8 =n, lim Erle +t) + fet) — 25(2)] ange qe dt = 0 Theorem 8-3. The Fourier series converges to S(z) if and only if for some fixed number 8 such that 0< 6 lim [fle +) + fle) — 28(2)] sin(M+a)t wand, t dt =0 SUFFICIENT CONDITIONS FOR CONVERGENCE OF FOURIER SERIES Various sufficient conditions exist under which a Fourier series will converge to S(2). ‘The following are two such conditions or tests as they are often called. Theorem 8-4 [Dini’s condition). The Fourier series converges to S(r) if for some fixed number 8 such that 0<8 5x, f fla +1) + fe—1) ~ 282) a4 exists 0 t Theorem 8-5 (Jordan's condition]. The Fourier series converges to sa) = Het tfe—oy if f(t) is of bounded variation in a neighborhood of t=s. If furthermore A(t) is continuous at t=, then the Fourier series converges to f(z). For other sufficient conditions see Problems 8.49, 8.50 and 8.55. INTEGRATION OF FOURIER SERIES Theorem 8.6. Let the Fourier series corresponding to the integrable funetion f(z) be given by ao 2 ‘Then if we integrate the series term by term from « to 8, the resulting series * fe) de. + ¥ (a cosne + by sinnz) (8) converges to ‘The remarkable thing about this theorem is that the series (8) need not be convergent. For a generalization of this theorem see Problem 8.60.182 APPLICATIONS TO FOURIER SERIES (cHAP. 8 FOURIER SERIES IN L? SPACES ‘Thus far we have been concerned with pointwise convergence of Fourier series. Two other important types of convergence can also be considered. The first involves mean con- vergence in L? spaces which is considered in this chapter and the second involves the concept of summability which is considered in Appendix B, page 175. It turns out that the theory of mean convergence of Fourier series in L? spaces is closely related to the theory of orthogonal (or orthonormal) series, since a Fourier series is a special case of an orthogonal series. For this reason we shall treat this more general theory. ORTHOGONAL FUNCTIONS ‘Two functions f:(2) and f(z) are said to be orthogonal in (a, 6) if Lneneae = 0 © A set of functions f,(2), fr(2), ... is said to be an orthogonal set in (a,b) if State tae = 0 men (10) and the functions are said to be mutually orthogonal or simply orthogonal in (a,b). Example: The functions {sin mz}, m = SF sme sn sete s+ are mutually orthogonal in (~r,7) since 0 men ORTHONORMAL FUNCTIONS If the set of functions f:(x),f2(x),... is orthogonal in (a,b), then if f,(z) EL? we will have . Sthpds = A n= 1.2... where A, > 0 is finite. Then if we let ¢,(z) = f,()/V/A,, we have JS [9,(@)Pde = 1 (1) S. sate) ote) ae = 0 men (12) The set of functions (4,(z)} is said to be orthogonal and normalized in (a, 6) and is referred to as an orthonormal set of functions in (a, 8) and the functions are said to be orthonormal in (a,b). Example: The functions = 1,2, ..., are orthonormal in (—z,7). See Problem 8.23. ORTHONORMAL SERIES A series ele) + eda) +o = Seal (18) where ¢,,¢,,... are constants and 4,(2),4,(z),... are orthonormal functions in (a,b) is called an orthonormal series. It = Steed «sycHap. 8) APPLICATIONS TO FOURIER SERIES 188 then the series (13) is called the orthonormal series corresponding to f(z). Because of the analogy with Fourier series, (13) is sometimes called a generalized Fourier series and (14) are the generalized Fourier coefficients. If f(a) and 4, (2) belong to L*, the coefficients (14) exist [see Problem 8.69]. PARSEVAL'S IDENTITY . Let Sie) = Be s(2) (15) be the partial sums of the series (13) with coefficients (14). Then by Problem 8.27, Ss@- horas = Sf yerpan - dea 1S.) — /@|h Now if it is true that Jim |S,(2) ~ fle) = 0 ile. if S,(x) converges in the mean to f(z), it follows that S vera = Se ro) which is called Parseval's identity. Conversely if Parseval’s identity holds, then S,(z) con- verges in the mean to f(z). The results can be summarized in the following Theorem 87. The generalized Fourier series corresponding to f(z) converges in the mean to f(x) if and only if Parseval’s identity is satisfied. BESSEL'S INEQUALITY Regardless of whether lim ||S, — fl] is or is not zero, it will certainly be true that sf apar an See Problem 8.28. This inequality is called Bessel’s inequality. The case of equality cor- responds to Parseval's identity. APPROXIMATIONS IN THE LEAST SQUARE SENSE If f(z) €L2, then we can think of the quantity 1 ° < aS, te - Zaeorae as) as the mean square error of f(z) from an approximating sum > a,¢,(z). ‘The following theorem is of interest. & Theorem 8-8. ‘The mean square error (18) is a minimum when the constants a, are the generalized Fourier coefficients, ie. when S Heeteyae «9 a = & Because of this theorem we say that > ¢,6,(#) approximates /(2) in the least square sense. Note that Parseval’s identity is satisfied if and only if the mean square error approaches zero as ©, See Problems 8.30 and 8.76.184 APPLICATIONS TO FOURIER SERIES [OHAP. 8 COMPLETENESS OF ORTHONORMAL SETS . If in approximating a function f(x) € L* by an orthonormal series ¥ ¢,¢,(z) we fail to include one or more of the functions ¢,(x), Parseval’s identity will not be satisfied. Because of this we adopt the following Definition. An orthonormal set (4,(2)} is said to be complete if for all functions f(z) € L* Parseval's identity is satisfied. We have the following Theorem 89. The set (#,(z)} is complete if and only if there is no function other than zero which is orthogonal to all the functions ¢,(2).. RIESZ-FISCHER THEOREM FOR GENERALIZED FOURIER SERIES Theorem 8-10 (Riesz-Fischer]. Given an orthonormal set (4,(2)} in (a,b) and a set of constants ¢, such that S ef converges, there exists a function f(z) € Lt such that the ¢, are Fourier coefficients corresponding to f(z), Parseval’s identity is satisfied and the series S ¢,¢,(2) converges in the mean to f(2). For Fourier series this theorem takes the following form Theorem 8-11. Let a,,b, be such that the series 8. Seve (20) converges. Then the trigonometric series ge se cos kx +b, sin ke) (1) is the Fourier series of some function f(x) € L*. Furthermore the partial sums of the series (21) converge in the mean to f(z) and Parseval’s identity (16) is satisfied. Solved Problems DEFINITION OF FOURIER SERIES 81. Show that if m and » are positive integers, (@) J cosme cosnzdr omen © cnene ti 0, men ) J, sine sine de = we mee oy S. sin mz cosnzdz = 0 men (ieee te [eos (m —n)x + eos (m+ n)z] de de Je 4 sin(m tmz) mom men | 1 [sin om: 2CHAP. 8) APPLICATIONS TO FOURIER SERIES 135 BS 6 Gh coamecosnede =f cote de 5 J + contmey ae = ery - 0 ten Sf sume sinnede = 1" fontanobe— nln tl sinimle P Lf sin (m+ nz = al 2 (eer site de = Ef” a -cosemside omy = tae . J sinmecoae de = 3 f tmnt anne cos (m+ nz _ cos (m—n)s mee ™ man =f sine cone ae = stimel" Lg 82. Ifthe series A + 3 (an cos nz + by sin nz) converges uniformly to f(x) in (—z,7), show 2.8.05 (0) as =f” fla) cosneds, (0) be =1 5 fle) sinnzde, that for n= (9 A = a2, (@) Multiplying 3 He) = A+ 3, (oy comms + by sine) 0 hy cone and intagrating from =r to, using Problem 8.1, we have Sse) conme ae = Af” conme de +S fof” came coonete + 5 f cance} = age if m=1,2,8,... Thus te = EFT sevconmede mat @136 APPLICATIONS TO FOURIER SERIES (CHAP. 8 (#) Multiplying (2) by sin mz and integra 18 from ~z to =, using Problem 8.1, we have Si fersinme de = in me de fon snmecanete +6, annem net} hee HLS” posnmete m=1.08,. (9 tnepain of) from ~e oye " narde = AS tot fog Soar « 6, annede} = a 2 as seen by formally putting m= 0 in equation (2) of part (a). Note that in all parts above, interchange of summation and integration is valid because the series is assumed to converge uniformly to f(z) in (x, x). Even when this assumption not warranted, the coefficients a, and b, as obtained above are called the Fourier coefficients corresponding to f(z), and the corresponding trigonometrie series with these values of , and 2, is then called the Fourier series corresponding to f(z). An important problem is to investigate conditions under which the series converges and if so whether it converges to f(z). <2 gnu = % Thor a= Ef me = 83. Find the Fourier series corresponding to the function (2) f(z) has period 2 outside of the interval (0, 27). Although it is of course not necessary to graph the function in order to determine the Fourier it is sometimes useful to do so, ‘The graph of f(z) showing the periodicity is given in 1, Since f(z) has period 2z, we can use (0,27) in Problem 8.2 instead of (—v, 2) 22, 0<2<2n, where ts ; e The Fourier coefficients corresponding to f(x) are nn by the following. ta = Af" he) oonede = 1 fetcoanede 2 o» (#832) — an(—s) + 2(- ay eee er be = Ef terainaeas = 16 "atannede Hoo) -on(-tr)«a(Sa)0 = 5CHAP. 8) APPLICATIONS TO FOURIER SERIES 137 ‘Then the required Fourier series is oe ; gt 2 a, cosnx + b, sin nx cy ee _ i. & (Leone funn) RIEMANN-LEBESGUE THEOREM 84, Prove the Riemann-Lebesgue theorem: If f(x) is integrable in (—x,=), then ©) lim fF" f(2) cosnede = 0 (a) lim (" f(c) sinnzde = (a) Let 2 =u to!n 90 that fi sesannete = =f" 4 (042) sana = =f 1008) since we assume f(z) has period 2s. " pasinnede = ~ f7 7(242)sinnede seeyrinneae =f" [ney (- + 5] sine de SI” nersinneae| = 3) §” [te1-1(242)] sinneae| = 3S |r@+8)- But by Problem 5.31, page 90, we have, since /(x) is integrable, tin” |r(242)- rola = 0 te anf tae seee = 0 ‘Thus and 2 de () This ean be proved in a manner similar to the method of part (a). See Problem 8.43. CONVERGENCE OF FOURIER SERIES eee sin (M+4)9 85. Prove that 2 + 2 cos 8 “2sindd We have cosna sin go = 4fsin (n+ d)e — sin (n—4)o] ‘Then summing from n=1 to M, speeeeeete eee = Hein §o—sin fo) + Hsin go—sin fo) + + + sin (+ $9 — sin (Je) = fain (M+ po — bain go 1,4 sin (M+ p)e ‘Thus ata = OO138 86. a7. APPLICATIONS TO FOURIER SERIES [cHAP. 8 1 (M+I ap 1 ql (? sin sno, _ Prove that (a) } spars be do = 5, w 2S sors a= (a) From Problem 8.5 we have on integrating from 0 to =, Mae = AL" [Le Swan] 2 sin we cose = Hee) ao (®) From Problem 8.5 we have on integrating from —r to 0, PS Soamptee = 2S" [EAS coon Heed. fl } Show that the partial sums of a Fourier series are given by 1 eg ay Sin (+ Hu Ste Dante au Su(z) = B+ Se (as cos na + by sin nz) Using the formulas for the Fourier coefficients, we have (EST 0 eons) cone + (ES to since) sine 2 PF” Heveosnt conn + sin nt sin na) dt a, cos nz + By sin nz = 2G 19 soem 2) a Also, af foe Then Sule) = (a, comms + by sin nz) = if. Koa 23 6" NO coan(t— 2) at = 1S" 10[h+ 3 coome-a)]e = ES 10 ROM ie ain Ha) using Problem 8.5, Letting t= 1, we then find Laces sin (r+ gu Sue) = gf et TE a Since the integrand has period 2x, we can replace the interval (—r—2,+— 2) by any other interval of length 2x, in particular (—z,). Thus we obtain the required result.CHAP. 8) APPLICATIONS TO FOURIER SERIES 139 88. 89. 8.10. il. Prove that 5 Sulz) = JS, [e+ t) + f(e—8)] se Dae From Problem 8.7, . sin Sul) = gf fone Me = sin at a sine A [ines otter f° pesnttbltia] _ Lif nen Ma Sie sina nna + De) a singe = Effnevotl nog + Ste OEM a] . sin at -éS eset Prove that Sula) — Sle) = aS [ile +0) + e— 0) — 28(2)] singe oh at From Problem 8.6(a), sey = 2 F251) RM Then using Problem 88, Sule) ~ Ste) Ue + f= 9) EP ae — ap sin + De Wee + 0+ feo — 280) Prove that convergence or divergence of a Fourier series corresponding to an integrable function f(z) at any particular point x depends only on the values of f(z) in a neighborhood of the point. From Problem 88 we have for some fixed number 8 such that 0<8 <=, sine + pt e » sn dt a Lf nero ren Oa Bay is integrable in (3, ), it follows from the Riemann-Lebesgue theorem Sule) = zs Wet + fee~ 0) 2 fet) +fle- aint that the last integral approaches zero as M+ =. ‘This shows that convergence or divergence of the Fourier series depends only on the integral over the interval (0,§), ie. on the values of f(z) in a neighborhood of the point 2. Now since Prove Theorem 8-1, page 181: The Fourier series corresponding to f(t) converges to S(x) if and only if Him f° le+8) + fla — 25(@)) OED me Jo in $¢ 0 ‘This follows at once from Problem 8.9 since in such case lim Sy(z) = Ste).140 APPLICATIONS TO FOURIER SERIES (CHAP. 8 812. Prove Theorem 8-2, page 181: The Fourier series corresponding to f(z) converges to S(z) if and only if for some fixed number 3 such that 0<85 x, im Stern +fe-t)— asa) See dt =0 If wo let Fie) = fle+) + fle) ~ 28(2), then by Problem 8.11 the Fourier series converses to S(a) if and only if din Sr a = ie, for some fixed number 6 such that 0< 8 <2, ae anges Nag (rg MEN] = 0 Now since the limit of the second integr result follows. sn [fr is zero by the Riemann-Lebesgue theorem, the required 813. Prove Theorem 8-3, page 131: The Fourier series corresponding to (x) converges to S(z) if and only if for some fixed number 8 such that 0<8 3, lim { [fle +8) + fet) — 25(2)] sin titae = 0 Using Problem 8.12 we see that the Fourier series converges to S(2) if and only if for some fixed number 8 such that a = 0 ® a ree integrable in the interval (0, ) [see Problem 8.45]. Thus from the Riemann-Lebesgue theorem, ~f]uarsiea = o ain Soa aint ‘Then from (1) and (2) we see that sin, Soro OB = 0 which proves the required result, SUFFICIENT CONDITIONS FOR CONVERGENCE OF FOURIER SERIES 814. Prove Theorem 8-4, page 131 and thus establish Dini’s condition for convergence of Fourier series: The Fourier series converges to S(2) if for some fixed number $ such that 0<3 0, we can choose 8; small enough so that (8,) < «/2B. Thee [So Mal < § ® Also by the Riemann-Labesgue theorem, in S vo tia = 0 se, we can choose Af large enough so that |Sive ye Di a <{ » Then from (2) and (4) it follows that if is sufcfently large, + sina + at, | Sno eal oa which proves the required result, 8.16. Prove the result of Problem 8.15 if y(t) is of bounded variation and lim y(t) = If y(t) is of bounded variation, then it can be expressed as the difference y1(0)—ya(t) of two monotonic increasing functions ya(@) and ya(@). Since lim, y(t) = 0, we have lim ys(@) = lim yalt) tnd theres eo of generat nearing hat tem ino ee, by Probie 815, Bn So a = tn foun = and so by taking the difference of these two limits we obtain as required - ep, tn fo EB ay oo142 APPLICATIONS TO FOURIER SERIES (CHAP. 8 817. Prove Theorem 8-5, page 131, and thus establish Jordan’s condition for convergence of Fourier series: The Fourier series converges to S(z) = 3{f(x+0) + f(e—0)] if i(t) is of bounded variation in a neighborhood of t If (0) is of bounded variation, so also is Fi) = fe + fle) 282) = fle+t) + fle) ~ fle +0) — fe-0) Furthermore it is clear that lim F(t) =0. ‘Thus by Problem 8.16, ain, fr MED which shows that the Fourier series converges to S(e) = a[/le +0) + fle~0)) 818. (a) Referring to Problem 8.3, page 136, prove that for 0 <2 <2z, we 4S a cosne — a) (0) To what value does the series converge for r=0 and 2=2r? Since f(e) = x* is of bounded variation, the series converges to Hifle +0) + fle —0)) (@) Since any point of the interval 0 << 2r is a point of continuity, the series converges to fla) = 22, ‘Thus for 0<2 <2, (®) The point x =0 is a point of discontinuity and we have from the graph of Fig. 8-1 or from direct analytical considerations the fact that at 2 = 0, Mle+0) + He—0)] = YO+4=) = 292 Thus at z=0 the series converges to 25%. (4 con nx 3 Similarly at the point x =2r the series converges to Hse +0) + fle—0)) = HOF 4x2) = Bet 1 819. tet From the fact that the series of Problem 8.18 converges to 2r? at z= 0, we have tet, SA at at = : rae # 1 Ae tate € INTEGRATION OF FOURIER SERIES 8.20. Prove Theorem 8-6, page 131. We shall prove the result for a= 0, @ =a. Consider S, [100-3] ® where ,, dy are the Fourier coofficients corresponding to f(z), ie. ote) oe = Af" neosted, by = 2S" hay sinkedeCHAP. 8) APPLICATIONS TO FOURIER SERIES 143, Since g(2) is continuous and of bounded variation [Theorem 6-21, page 08), it has a Fourier series which converges to g(z), 80 that oe) = $+3 (ey conker + dy sinkzy (2) where ee = EP oe coated, de = LP ole) sinks de ‘Now using integration by parts with g(z) given by (1), we have for k= 1,2,3, ce = Ef oercosteae = 1[orm it} 1 ks J, $0) sin be de 2) es fey cosh & = oe) sink de = [on S05H | + L ("Tey BS [0 ~ fomie ae Ef" rested = where we have used the fact that g(0) =0, as is evident from (1), and (2) =0 from the fact that g(z) has period 2. Substituting the values of , and dy in (2), oz) (3) Letting = a= “ so that a2) © From (1) and (©), = it, & msinks + bylt ~ cone) Grom = 4 8% ee But this is exactly what would be obtained if the Fourier series corresponding to f(z), ic. S +S toxcaskr + by sinke) is integrated term by term from 0 to x, and so the required result is proved. 11,1 * Show that petit a = 90 By Theorem 8-0 we can integrate with respect to # the series of Problem 8.3 to obtain 2 4 te Be (Sanne +S eosne) + 6 o whore ¢, is the constant of integration, ‘Then letting += 0 wo have, wsing Problem 8.19, Sta = 0 or a = eS144 APPLICATIONS TO FOURIER SERIES [cHAP. 8 ‘Thus (1) becomes sop ta = to Integrating both sides of (2) with respect to 2, wt _Detet | Bete Roos +5 = a Putting 2=0, a = (eee aeener SS aE ae °° eetses eed or aout 2g) ae = 0+ fea or, on carrying out the integration, ¢, = 4r4/00, so that as required, $1. a) ‘This last method can also be used to determine ¢, in (1) without using the result of Problem 8.19. 822. Prove that the series sing , sin2x nz * “Ind cannot be a Fourier series. + ah From Problem 8.20 we see that if the given series is a Fourier series, then 3 7# must con- verge. In the given series, however, b, = I/In(k-+ 1) and thus wk $e 2 § 1 _ aya 1 Ae > Aeiery ~ ina * zing * Fina + ‘hich diverges. Thus the given series cannot be a Fourier series, ORTHONORMAL FUNCTIONS AND SERIES 8.23, Prove that the functions ¢,(x) = n=1,2,3,... are orthonormal in (—x, ve ‘The required result follows at once since . _ act 0 it men Se enlereuter ate = ES sine sine de = { itn. by Problem 8.1(0). 824, Let (2,(z)), n= 1,2,8,..., bea set of orthonormal functions in (2,6). If 3 ¢, $,(2) converges uniformly to f(z), prove that * . = S reserCHAP. 8} APPLICATIONS TO FOURIER SERIES 145 It fe) $ ano then multiplying by #2), $2) mlz) = 3 en ml) oo(2) Integrating from a to b, we have J teremteyae = 3 eg Fouled gla) de ® where the integration of the series can be performed term by term since the series is supposed to be uniformly convergent in (a, ). Since the set of functions {¢,(x)} is orthonormal in (a,b), we have { if men ‘0 that (1) becomes as required, $2) eC) die The Hermite polynomials H(z), n= 0,1,2,8,..., are defined as polynomial solu- tions of degree n of the differential equation Hil - 20H, + 2nH, = 0 Prove that the set of functions (¢-*"/*H,(c)} is orthogonal in (~, ). ‘The differential equations satisfied by H(z) and H,(2) are respectively Ay — 2cH, + 2mH_ = 0 o (Hy — 22H, + 2nH, @ Multiplying the frst equation hy Hy the second by Hy and subtracting, we find HH ~ HyyHy, ~ 20(H Hq — HH) + (2m —2n)HyHy = 0 so that (m= 20H ally = Hy Ht! — Hyllf, ~ 2a(B Hl HH) or Cm—2n)glly = LZ Hy — H yb) ~ 22H yl — Hla) Multiplying by e~*", we can write this as Cm—2nle-P Hy, = Zle-# Edt — Myla) Then integrating from —= to =, m= 20) fo e-Hgliyde = 8g. HH ° Thos if men, Sf" centyae = 0 cy which is the same as saying that the set (e-*/2H,(2)} is orthogonal in (—#, ©).146 APPLICATIONS TO FOURIER SERIES (CHAP, 8 826, If uniformly to f(x), prove that cnH,(z), where H(z) are the Hermite polynomials of Problem 8.25, converges Senta ste) S. Peyrae We have He) = enkiy(e) 0 that by equation (s) of Problem 8.25, S eH fei Haleds = Se, 8 Hala) Hala) dz = ew * eH (a)}? de Co) J where the term by term integration is justified by the uniform convergence of the series. Solving for eq, we arrive at the required result, FOURIER SERIES IN L? SPACES 827. Let (¢,(z)} be an orthonormal set of functions € L? [a,b] and let S,(z) = Sz CGA) where ¢, = S Ke) $y(e) de, Prove that for any function (2) € L? [a,), (se) = Sse)-feyrae = fora — Sa We have, Issel = Sf" yey —suenpeae = f ae Fe) ~ 3 ensutel] =f {iter ~ 23 astern + 3 cnesed ac} ae = fuerte - 2 Saf rormiree + 3 3 oo Sater eterde “yore -2 38+ 34 Sverre - 3 where we have used the results = L He) ex) io _ JQ i*k Soe eras = i fan 8.28, From Problem 8.27, prove that Sa = S vera We have from the sesulé of Probie 827, Site - saareae =f" yerrae ~ 3 oCHAP. 8] APPLICATIONS TO FOURIER SERIES 147 8.30. Then since the left side is non-negative, it follows that 342 f vera ®) ‘Taking the limit as mr and noting that the right side of (2) is independent of n, we see that Sas Sine ade This inequality is called Bessel’s inequality, Prove that if lim |[Sa(z) — f(2)|_ = 0 where S,(x) and f(2) are as defined in Problem 8.27, then sa-f , Ifa) dx This follows at once from Problem 8.27, since if tim liSat@) ~ flail = 0 then lim [f° wera - 3] = 0 ie. S wera = Se The result ig called Parseval's identity. I Parseva's identity is satisfied for all functions Hla) € 12, we call the set of orthonormal functions {4y(2]) complete. ‘The mean square error of f(z) from an approximating sum «, $,(2) is given by aS, |W - Bence Prove that (a) the mean square error is a minimum when the constants «, are given by the generalized Fourier coefficients a = Steptoe and that (b) Parseval's identity is satisfied if and only if the mean square error approaches zero as n> ©, de (a) We have as in Problem 8.27, Se - = ifn Ineyae S verre S weereae S veres Bosna) de 3 em foe) ext ae 23 a J Med outa) de +148, a APPLICATIONS TO FOURIER SERIES [CHAP. 8 From this it is clear that a minimum is obtained when a = = f fedeterae For another method of obtaining the minimum, see Problem 876. This follows from part (a) and the result of Problem 8.29. RIESZ-FISCHER THEOREM FOR GENERALIZED FOURIER SERIES 831. Let (4,(z)} be an orthonormal set in (a,b) and let ¢, be a set of constants such that 2 converges. Prove that there exists a function f(x) Z* such that (a) the tial sums > ¢,4,(2) converge in the mean to f(x) and that (b) the o, are Fourier coefficients corresponding to f(z). @) ® Let 5 =, cess) dente the nh prt sum ofthe svi & esa). We have for m> Isn- Sil? = Sf 1SaGe)~ Siorrae de S. |S ane— Svc f 7 aa { 3. 3 cof) te} te " Mae Ms 5 Chef 6) aula de Now since 3 cf converges, it follows that lim 3 @ = 0 i lim, l|Sq— Sql = 0 ‘Thus by the Riesz-Fischer theorem there exists a function fle) E14 such that Sy(2) = Bytvex(2) converges in the mean to f(z) as n=. By part (a), since S,(z) converges in the mean to f(z) as n>, it follows from. Problem 7.52, page 128, that . Bin J Ser elerde = ff) ale) de o But Sila) gy(e) de = SL {Sectors} ae a Lf #r(e) #9(0) de = ey = Ste seb ae o ive, the o, are Fourier coefficients corresponding to f(s). ‘Thus (1) yieldsCHAP. 8] APPLICATIONS TO FOURIER SERIES 149 8.32. Prove that under the conditions specified in Problem 8.31 Parseval’s identity holds, i.e. S, reoras By Problem 8.81 the partial sums S,(z) converge in the mean to f(z) as n->©. ‘Thus by Problem 7.15, page 123, tin J" isertae = f"ieyeae ay bat Sswere = f° ae cx #02) Zeroee wc} de = 2 den f e@awe = 3a ‘Thus by (2), lin 34 = S weoveae or 34 = Sf wore «) Supplementary Problems DEFINITION OF FOURIER SERIES tm at foo = { "2ST snae fhe prl rb th Fer tf * sing _sin2z, sin3z_sindz | se toe oo 2 (ORE a He eee) 834, Show that the Fourier series for f(x) = cos pz, P40,21,22,... is given by © Sr, where the period is 2° and where 885. Extend the results of (a) Problem 8.1 and (b) Problem 8.2 to the interval (c, ¢+25) where ¢ is any real number, 856, Let f(x) be Lebesgue integrable in (—I,) or more generally in (c, ¢+2l) where 1>0 and ¢ is any real number. Suppose that /(2) is periodic with period 21 outside of this interval. Show that the Fourier series corresponding to f(e) is (on cor§E + by sin *s) where the Fourier coefficients are given by oe = EF He cos BE 8, = TS He) sin aeAPPLICATIONS TO FOURIER SERIES (CHAP. 8 431. Prove that if f(z) is an odd fonction in (1,0) [ie (2) = responding to fiz) 8 ot wae Stesin = where 6, = iS, fla) sin ™ da f(z), then the Fourier series cor- ‘The series is often called a Fourier sine series 838, Prove that if f(x) is an cven fumetion in (1,1) fie. fl-2) = fle], then the Fourier series corre- sponding to f(x) is _2ct mse on = Ef redone 839, Show that the Fourier series corresponding to f(x) = ene is given by ]o, -3<2<0 where the period is 6 3 5 § [Meas cost — Stone gee] 3 Find a Fourier sine series corresponding to fle) = cosz, 0<2 <=. 841, If the trigonometric series (1), page 150, converges almost everywhere to f(z) in (=x,=) and if the partial sums of the series are uniformly bounded, prove that (a) f(z) is integrable in (~z,) and (8) a, and b, are given by the formulas (2), page 190, (Hint: Use Lebesgue’s dominated convergence theorem ‘THE RIEMANN-LEBESGUE THEOREM AND CONVERGENCE OF FOURIER SERIES 842, Prove that Jim f-" S°"4de = 0 in two different ways, 844, In the proof of the Riemann-Lebesgue theorem in Problem 84, does have to be an integer? Explain. 2 ig intograble in (0,8) and thus complete the proof in 1 845. Prove that the function =fy755 ~ 7 Problem 8.18. 8.46, Prove that if /(x) is integrable in (—z,=) and f'(e) exists in (—x,x), then the Fourier series corre- sponding to fx) converges to fz) at each point # in (=7,=). 848, Prove that a Fourier series corresponding to /(e) will converge to f(z) if f(x) satisfies the condition fet) — fe)\ < Kur for |x) 0, then the Fourier series corresponding to f(z) is convergent and that it converges to Sie) if lim G(t) = 0, ‘This test is called de la Vallée Poussin’s teat. 856. Show that the test of Problem 8.55 is satisfied if either Dini’s or Jordan's tes satisfied. INTEGRATION OF FOURIER SERIES Jy. = ® a5 st. Prove that 4 837. Provethat f+ 3 1 +he 888, Prove that 3 Sater riy Averees and thus complete the proof in Problem 822. & sine ate, Prove that 3 se O0, Oif #=0 and -1if 2<0. Show that the functions {sgn (sin 2%r2)}, m=0,1,2,..., called Rademacher functions, are ortho- normal in (0, 1). 869, Prove that if /(z) and ¢q(z) belong to 12, then the coefficients (14), page 192, o 870. A sot of complex functions {¢n(2)} is called orthonormal in (a, 0) if ¢ if men moana = {0 fos where the bar over a function denotes complex conjugate. (a) Show that this definition reduces to the ordinary one in case the functions are real " =} where » is any integer is orthonormal in (—r,»)- (©) 1, 3. eget tn the Fourier series for f(s), prove that a nike, aS, faye de This is called the comples form of Fourier serie. 0) Sow th ot tae 871, Consider the set of functions 1,2, 2%2%,.... From these form a new set of functions jag tagt, a4 + age + age, .., where the nth function in the new set is obtained by multiplying ‘each of the first members of the first set by a constant and then adding. (a) Determine the con- stants ajay, ... 80 that the functions will be orthonormal in (~1,1) and (b) show that these polynomials satisfy Legendre’s differential equation [the actual Legendre polynomials have the property that P,(1) = 1]. See Problem 8.65. FOURIER SERIES IN I? SPACES AND THE RIESZ-FISCHER THEOREM 872, Let fx) © L* [-r,~} have the Fourier series $+ 3 co, cosne +, sinnz) Prove that Parseval’s identity is Co ee B+ Scqrey = WayeaeCHAP. 8] APPLICATIONS TO FOURIER SER 158 818. If fle) EL? [2,2], use Problem 8.72 to prove that Do these results prove the Riemann-Lebesgue theorem of page 1907 Explain, 874, Verify Parseval’s identity corresponding to the function of Problem 8.18. 05, Use Parseval's identity corresponding to the fonction of Problem 8.24 to find the sum of the series $1 20 816, Work Problem 830 by using partial differentiation to determine the minimum value of the least square error. 8.17, Determine which if any of the following series are Fourier series corresponding to a function in 12 [=r] OB va O2m-1 9 Sea 878, Prove that if f(x) EL? and g(x)€L2 have the same Fourier coefficients, then f(z) = o(#) ‘almost everywhere. 879. Is it possible for a given trigonometric series to be a Fourier series corresponding to (a) a function in L? but not in L, (8) a function in L but not in 12? Explain, 880, Prove Theorem 8-9, page 134.Appendix A DEFINITION OF THE RIEMANN INTEGRAL Let f(2) be defined and bounded in [a,b]. Divide this interval into n subintervals by Points so, x1, @ Where @ = %o 0 there exists a partition with upper and lower sums S,s such that S—s <<. THE RIEMANN INTEGRAL AS A LIMIT OF A SUM ‘The Riemann integral can also be defined as a limit of a sum. To do this choose n points of subdivision @ = o 0 there exists a countable set of open intervals Ix, k= 1,2,..., such that EC Us; and such that the sum of the lengths of / is less than « ie. Y Lh) <« A fundamental theorem on the Riemann integral is the following. Theorem A-9. A necessary and sufficient condition that a bounded function f(2) be Riemann integrable in {a,b} is that the set of discontinuities of f(z) in [a,b] have measure zero. For a proof of this theorem see Problem 4.27, page 68 {compare Theorem 4-18, page 57]. THEOREMS ON RIEMANN INTEGRABLE FUNCTIONS Theorem A-10. If f:(x) and f2(z) are Riemann integrable in {a,b], then f(x) +f2(z) is Riemann integrable in {a,6) and Sfeax + She) + hear Theorem A-I1. If f(x) is Riemann integrable in [a,b] and c is any constant, then ¢ f(z) is Riemann integrable in {a,b] and St eftzyae = ef fade Theorem A-12. If f(z) and g(z) are Riemann integrable in [a,b], then f(x) g(z) is Riemann integrable in (a, 5). Arla) dx Theorem A-13, If f(z) is Riemann integrable in (a, 6], then Sirear = -Sima, Sseyae = 0 Theorem A-I4. If f(z) is bounded and Riemann integrable in [a,b] and ¢ is any point of (a,b), then r Spear = Sieyae + [rey ae Theorem A-I5. If f(2) and g(x) are Riemann integrable in {a,6] and f(x) = g(x), then Slievae = Theorem A-I6. If f(x) is Riemann integrable in [a,b] and has upper and lower bounds M and m in (a, b] respectively, then m(b—a) and if » is a number such that S f(z)dz = n(b-a) Theorem A-17 (Mean-value theorem}. If (x) is continuous in [a,b], then there exists a number ¢ € [a, 6] such that Lime = 6-076APPENDIX A] ‘THE RIEMANN INTEGRAL 167 Theorem A-18. It f(z) is Riemann integrable in [a,b] and |f(z)|=M_ for some constant M, then [Sted] = o-a) Theorem A-19. If f(x) is Riemann integrable in [a,b], then |f(z)] is Riemann integrable in [a,b] and . . [freee] = S° neuer DIFFERENTIATION AND INTEGRATION The indefinite Riemann integral is defined as Fa = Sf fan 6) where the variable of integration has been changed to to avoid confusion with the limit of integration «, The following theorems are important. Theorem A-20. If f(2) is bounded and Riemann integrable in (a, 6], then Fw) = S" fwau is continuous in {a, }). * Theorem A-21. If f(x) is Riemann integrable in [a,b] and F(x) = f' flu) du, then act Pe = Zr = £S twa = te) at each point of continuity of f(x). Theorem A.22 (Fundamental theorem of calculus). Let f(z) be Riemann integrable in [a,b] and suppose that there exists a function F(z) continuous on [a,b] such that F"(z) = f(z). ‘Then Srea = Sioa = Fb) — Fa) or Sra = Snwdu = Fee - Fe Theorem A-23 [Integration by parts]. Let f(«) and g(z) be Riemann integrable in {a,b} and suppose that F(x) and G(2) are such that F"(z) = f(e) and G'(z) = g(x) in [a,b]. ‘Then S. Feo@dz = FOOW) -~ Pec —~ f° Fe)G.) ae Payal, ~ S14) a0 Theorem A-24 [Change of variables}. Let f(«) be continuous in [a,b]. Let 2=g(u) have a continuous derivative in [a,8] where a= g(a), b= (9). Then Srerae =f owyawaw158 THE RIEMANN INTEGRAL [APPENDIX A THEOREMS ON SEQUENCES AND SERIES Theorem A-25. Let (fa()) be a sequence of continuous functions which converges uni- formly in [a,b] to a function f(x). ‘Then lim ("falc)ar = S evae = fF tim fuleyaz Theorem A-26. Let > u(x) be a uniformly convergent series of continuous functions in [a,b] and let s(z) be the sum of the series. Then Saaz = [ [Suwa = F [uae This is a reformulation of Theorem A-25 using series. Theorem A-27. Let S ws(2) converge to s(2) in [a,b]. Suppose that the derivatives u4(2) exist and are continuous in [a, b] and that 5 14(2) is uniformly convergent in [a,b]. Then s(x) = IMPROPER RIEMANN INTEGRALS If the interval (a, b] is infinite or if f(x) becomes infinite at one or more points of [a,b], then the Riemann integral of f(c) is often referred to as an improper integral. Such integrals are defined by appropriate limiting procedures and are said to exist if the corresponding limits exist. ‘The following are examples of such definition. Sl nevar = tim Syeyae ny S. fe)de = Nim Se de (7 So tovae If f(z) becomes unbounded when «= ¢, then we define Spear = tml (eae + [edz] @ im, {fe az ®) If the limit in (9) does not exist for «+, but does exist when « =«, we say that the corresponding value of the limit is the Cauchy principal value of f” f(z) dz. Also in such case we say that the Riemann integral of f(x) exists in the Cauchy principal value sense. A similar remark can be made for (8) for which the Cauchy principal value, if it exists, is obtained by taking a=—~b and letting b>».APPENDIX A] ‘THE RIEMANN INTEGRAL 159 Solved Problems DEFINITION OF THE RIEMANN INTEGRAL A.l. Interpret geometrically the case where f(x) is Riemann integrable in [a,b]. Refer to Fig. A-l which shows the curve C representing y= f(a). ‘The lower sum # corre- sponding to the partition shown represents the sum of areas of the shaded rectangles while the upper sum S represents the sum of areas of the larger rectangles. Clearly ¢=S, from the figure. Fig. Ad aera | oan ater enantiomer erat omen eee ae Peery A2. Prove that the function f(2) = not Riemann integrable. In any subinterval [24-1 24] of [a8] we have My=1, m,=0 since the rational and irrational numbers are dense in any subinterval. Then if S,e"are the upper and lower sums corresponding to any partition, we have 1, erational - eh 0, irrational *fined in the in {a, b] is s Suess, = Ban = b-0 a= Boman = Soran = 0 ‘Thus 1=b-a,J= sothat IJ and f(z) is not Riemann intograble. THEOREMS ON UPPER AND LOWER SUMS AND INTEGRALS AS. Prove Theorem A-1, page 154: If S,s are the upper and lower sums corresponding toa partition P and M,m are upper and lower bounds of f(z) in [a,b], then m(b-a) 3 8 = Ss M(b-a) From equations (1) and (2), page 154, we have s San, 6 = Ziman,160 THE RIEMANN INTEGRAL [APPENDIX A Then since om Sm, < My Mf we have on multiplying by ss, and summing over I from 1 to 3m = Soman = Suan = 3 wan nda cee s 2 wEan o moo) 22 5 8 = Mb~e) Ad. Prove Theorem A-2, page 155: If Si,s; are upper and lower sums corresponding to a new partition P; obtained by addition of points to the old partition P of Problem A.3, then S858 sea upper sums do not increase and lower sums do not decrease by addition of points. ‘The result will be proved if we can prove it when one point of subdivision is added to the given partition. To do this let the given subdivision points be a= Hy cmca cca = ‘Suppose that the additional point of subdi uso that xp << zp sion occurs in the interval (2,12) and is denoted by Now the contribution tothe upper aur corresponding t the subdivision points of the interval (eps * Myley— —) 1f we denote the least upper bounds of f(2) in (2p 4) and (u,24) by MS? and M5? respectively, then the contribution to the upper sum when the additional point of subdivision is taken into account MiP uz) + ME, —w) @ ‘The change in the original upper sum caused by the additional subdivision point is given by the difference between (1) and (2), ie. SY = Md 7 ayo) + Uy Me,— 09 Since this 1s negative or zero because MS" Mp, MS Mp, and > pn p>) folows that the upper sum eannot increase by adding’ point of subdivision andthe rea renal is proved Similarly we can prove that the lower sums cannot decrease by adding points of subdivision [sce Problem A.38) A5. Prove Theorem A-3, page 155: If S:,s: and Ss,s» are upper and lower sums corre- sponding to any partitions P; and Ps respectively, then Ss ie. any lower sum is never greater than any upper sum regardless of the partition used. ss SS: or & ‘The proof is identical with that given in Problem 4.5, page 69. AS. Prove Theorem A-4, page 155: An upper integral / is never less than a lower integral Jand we have S=1=J=s. ‘The proof is identical with that given in Problem 4.6, page 59.APPENDIX A] THE RIEMANN INTEGRAL 161 Aa, Prove Theorem A-5, page 155: A necessary and sufficient condition for a bounded function f(x) to be Riemann integrable in (a,b] is that given any «> 0, there exists a partition with upper and lower sums S,s such that S~s 0, then we ean choose 8 > 0 such that if 4-1 f& Sue ie . |S,revan - freee] < « am 5 8 s be the upper and lower sums corresponding to the given partition, Then since My where My,m, are the lab, and g.lb. of (2) in (2y—y,24), we have += Biman = s a ‘Also we have s fade = 6 «) 7 -s 5 -fijma = - «) Adding (1) and (s), forever “8-9 5S Kedar, or since S—e>0, |S wan - fred] = s—s Now since we can choose the largest values of Az, (ie. the norm] so small that S—2<, the required result follows, We can also prove that the definition of the Riemann integral given on page 154 follows from the definition as a limit of a sum [see Problem A.49]. Thus the two definitions are equivalent, THEOREMS INVOLVING THE RIEMANN INTEGRAL A9 Prove Theorem A. integrable in (a, b]. , Page 155: A continuous funetion f(2) in [a,b] is Riemann Since /(2) is continuous in the closed interval (a, 8, it is uniformly continuous. Let 2°) and 2° be any two points of an interval (x,.y,,). Then given «> 0, there exists § > 0 such that Wer) — fe] < whenever jx) ~ 2) < a Thus we can choose points of subdivision so that Mz - me < If the upper and lower sums corresponding to this partition are given by 5 = Sinan 2 = S mae162 A.10. All. Adz. THE RIEMANN INTEGRAL [APPENDIX A then we have * z Se = 3 Ohman < 3 52 ‘Thus S—s<« and it follows by Problem A.7 that f(x) is Riemann integrable. Prove Theorem A-7, page 155: A monotonic function f(z) in [a,b] is Riemann integrable in [a,b]. We shall assume that /(z) is monotonic increasing. The ease where f(2) is monotonic decreasing can be proved similarly [or by considering —f(2) instead of f2)] By definition we have for the given partition @ = xy 0, there exists § > 0 such that. Whe)—fla| < «whenever |e— a) <8 ”166 THE RIEMANN INTEGRAL [APPENDIX A ode — S"noee = fae = ES de 16 faa an = ES He — reo ae and it follows from (1) that if 0 <|Al <8, . ‘Thus Feet Ma FE) _ pay] — ta) vee au = RIS me ste) a aT Thos we EAS = sao on Pie) = feo ie, FP") = fle) at each point of continuity of f(e). A.L, (a) Prove Rolle’s theorem: Let f(z) be continuous in [a,b] and suppose that f(a) = f(b) = 0. Then if f(z) exists in (a,b), there is a point ¢ in (a,b) such that #0) = 0. (b) Prove the law of the mean or mean-value theorem for derivatives: Let f(z) be continuous in (a, 6] and suppose that /"(2) exists in (a,b). Then there is a point ¢ in (a, b) such that , re = Wate = bra (a) The theorem is trivial if f(z) =0 throughout [a,b]. Suppose then that (2) >0_at one point of (c,). Then there is point e in (a,2) at which f(z) attains a maximam value [see Theorem 1.25, page 8]. Now if (0) >0, ie. if fin HEN=L0 > 9 ai os HerN=H0 5 4 for [al <3 @ Thus if h>0, it follows from (2) that f{e+h) > fle) so that f(z) does not have a maximum at ¢ and we have a contradiction, ‘Thus /'(c) cannot be grester than zero, Similarly we can show that /'(c) cannot be less than zero and so we must have '(e) = 0. ‘The same result follows if we assume that f(z) <0 at some point of (a, ) (®) Consider the funetion Gla) = fle) ~ fle) — (4 ]te- a ® Then if 2) is continuous Sn (0,8), ao ia G2). Alao sf (2) exits in (a2), then G2) exists in (a,5). Furthermore wve have Gla) = GQ) = 0. ‘Thus G(s) satisfies Rolle’ thearem of part () tnd so there exists a point cin (o,®) such that G'() = 0. But this means that ow = ro [A=] = » ® and the required result follows.APPENDIX A} THE RIEMANN INTEGRAL 167 A.22 Prove Theorem A-22 [Fundamental theorem of calculus], page 157: Let f(x) be Riemann integrable in [a,b] and suppose that there exists a function F(2) continuous on [a,0] such that F’(x) = f(z). Then Sted = Fe) -F@ oS Heya Suppose that we form any partition a= 2) <2) <++- © and msx(z_~2,-2)70, it follows that Sted = ro- Fe oF Std = Fe) - Fe) asin (1/2), 240 A238. (a) Prove that the derivative of F(z) = { a peo isiven by Poe pesinaie,- cos(l/eh, 270 (b) Show that Ss. Pade = §& (a) If 20, we can use the usual rules of differential calculus to obtain Ztesnas = a{ont}(-4) + tesink = oe If ==, the derivative is given by fin, LOE EO sm sinh = = 1 = Him ksin () By part (a), Fe) exists at all values of 2 in (~2/,2/*). Thus by Problem A.22, 8 7" reyde = rane = 8 A.24, Prove Theorem A-28 [Integration by parts], page 157: Let f(x) and g(x) be Riemann integrable in [a,b] and suppose that F(«) and G(2) are such that F(z) = f(z): and G(@) = o(2) in [a,b]. Then S Fe oeae = F()G(b) — F(a) Gla) - Scere Fae. — SI fla) Gla) de Since Fe) and Gla have derivatives inf they ae continuous and thus Riemann integrable in 8 Then ance Fa) ana G0) ae al Reman integrate Inf flows that ON) tnd P") Ge) and thus tel sum. Fle) Ce) # a) Ga) ~ 2 (eC) negable a 5 Than toms Problem AE, oP f Eloaade = Fee) a = FOVGO) ~ Fea) Ga)168 ‘THE RIEMANN INTEGRAL [APPENDIX A or Swe + read = FOG) — FleGie Reinede = Moab) Pegi ~ He) Ge since F'(2) = f(z), Ge) = oz). ‘THEOREMS ON SEQUENCES AND SERIES A25, Prove Theorem A-26, page 158. Let > ua(zx) be a uniformly convergent series of continuous functions in [a,b] and let s(z) be the sum of the series. Then [Same = 3S uear eo ive If a function is continuous in [o, B, its integral exists [see Problem A.9], ‘Then sinco s(2),5,(¢) and r,(2) are continuous [see Problem 1.53, page 23], we have Sue = Sesoae + Sree = [fuer can be made arbitrarily small by choosing n large enough. ‘This however follows at once, since by | fserae — fone the uniform convergence of the series, given any >0 there exists my such that |r4(x)| for n> mo independent of 2 in a,b] and so Sinere| s Sree < f ‘This is equivalent to the statements “wade = fin Sacrae or in f'acraz = J Lim «| . . fn, {im A.26. Prove Theorem A-27, page 158: Let sae converge to s(z) in [a,b]. Suppose that the derivatives us(2) exist and are continuous in [a,b] and that Dy us(2) is uniformly convergent in [a,b]. Then va) = E3Sum = Let o(2) = 3 usa). By hypothesis this series converges uniformly in [0,2] so that by Problem ‘A.25 we can integrate term by term from a to z, where 2 € [a,b], to obtain Snow = Sf wow = § moa = Sum - So = we -20) Differentiating both sides of tT 9{t) dt = a(e)— fa) then shows that (2) = e'(z), which proves the theorem.APPENDIX A] THE RIBMANN INTEGRAL 169 AQT. A28, 29, Let a(x) = ne, n= 1,2,8,..., 0 (a) Determine whether lim {'s,(z)dz = ("tim s(z) de. (®) Explain the result in (a). (8 S'wtrde = faeewt te = tent = jacen Then tim [sleds = tim 4a =4 Also (2) = time(s) = lim nae-*# = 0 whether #=0 of 0<#¢1. Then te tatows tat tn fates tino) de, ithe tt ena be eten ends (#) Although the sequence #,(2) converges to 0 in [0,1] it does not converge uniformly to 0 [see Problem 1.52, page 23). "If the convergence were uniform the results would have been equal. The fact that it is not uniform allows for the possibility [but not a certainty] that the results are not equal. For a case where the sequence is not uniformly convergent but the integrals are nevertheless equal, see Problem A.59. Let (2) = $52" Prove that S 4. ten since $4; converse, the series for ee) i unttormly com vergent for all 2 and in particular for 0 ='z = z, by the Weierstrass M test (page 10}. Thus by Problem A.25 the series ean be integrated term by term and we have Sma = S(32)« if Ste +2ehy4.. at 5 We have |#inm2| Give an example of a sequence of uniformly bounded functions f,(2) such that im fala) de exists but * tim fa(2) de does not exist. Let @=0, b= 1 and consider the rational numbers in the interval [0,1]. Since the rational numbers in {0,1} are denumerable (Problem 1.16, page 13], we can write them in the form of a sequence 14, 75,75, ++ Lo for 2 = ry ty cea Define Info 0 otherwise ‘Then fuede = 0 and 80 exists. However, since /170 ‘THE RIEMANN INTEGRAL [APPENDIX A ; [1 ites rational EEO = Yo itz isirrational we see by Problem A.2 that Sf rm tines on IMPROPER RIEMANN INTEGRALS ° de jl A30, Prove that f, Gray exists and find its value We have by definition, S (e Sm ASI. Prove that f @ is -q exists as an improper Riemann integral provided that it is taken in the Cauchy principal value sense. ‘The integrand 1/(2—1)° is unbounded at z=1 so that the integral is improper. Cauchy principal value sense we define Soativs = lei an Sette Sete] 8g —1y-20 |S — 8g — 1-2 » [Beem — Sen In the hee] 20 + 8(-g-20 — 8-2 + So] + pa 2 — Faye + 510 | Note that if we define 2 ds f ,@-ia = Sey + Ste] where ¢ ¥« the integral does not exist. A382. Prove that f et ae exists, ‘The Riemann integral ff a0 exists, Also, on integrating by parts we haveAPPENDIX A] THE RIEMANN INTEGRAL wm ‘Then taking the limit ag b=, we see that mf te = fs Riemann integral. Now since if neg, ax diverges, ic. f° 8" de is not absolutely convergent. aly. a = {04}e-+gh) mit of the last series as n> © is infinite, the required result follows. Supplementary Problems DEFINITION OF THE RIEMANN INTEGRAL AA, Use the definition to show that f(z)=1 is Riemann integrable in (a, b) and find its value. 8 Ue the denon of he Rien integral to oan the vane off" fonde wer a THEOREMS ON RIEMANN INTEGRATION A236, Prove Theorem A-I1, page 156,172 Ast 38. 9, Aso, At, Aaa, Aas. At Aw. Aas. Ast. Aas, ‘THE RIEMANN INTEGRAL [APPENDIX A Prove that if f(z) is Riemann integrable in [a,b], then it is also Riemann integrable in any sub- interval (¢,¢] of [a, 6). Prove that the lower sums cannot decrease by adding points of sul of Problem A.4). ision [see remark at the end Prove Theorem A-13, page 156. Prove that /(2) is Riemann integrable in [a,] if flz) is bounded and has (a) one point of diseon- tinuity in [4,8], (6) any finite number of points of discontinuity in (a, 6). Prove that the Riemann integral of f(z) does not change in value if the values of f(z) in (a, ] are changed at a finite number of points. Prove Theorem A-I8, page 187, If f(z) is continuous in [a,b] and flz) $M, then if J fovae ‘M identically in (a, 6). Mea) prove that (2) Obtain the following generalization of Theorem A-16, page 156: If f(z) and o(2) are Riemann integrable in (a,6], g(2) = 0 and m= flz) = M, then ” jadole)de = Mf” oeyae J J Use Problem A.44 to prove that if f(z) is continuous in (a,6] in addition to the other assumptions, then there is a number such that S Herod = 19 fore is called the generalized mean-value theorem (see Theorem A-17, page 166). If f(z) is non-negative and continuous in [a,b] and if f(c) > 0 for some value ¢ in [a,b], prove that Pa Stora >o 1 /(2) is continuous in (a, 8] and S were = 0 prove that /(2 © in (2,8). Prove Theorem A-12, page 156. [Hint: First assume that f(z) and g(2) are both positive and have ‘upper and lower bounds M{”,mj!” and M{2,m) respectively in [o,-1, 4]. Then prove that if ‘Mg, my are upper and lower bounds of f(z) gz) in [zy-1, 24), then My — my = MG (ML —_m2>) + mQQMQ ~m{) and thus show that S—s<« If f(z) and g(2) are not both positive, add suitable constants s0 as to make them positive.] Prove that the definition of the Riemann integral given on page 154 follows from the definition as a limit of a sum.APPENDIX A] ‘THE RIEMANN INTEGRAL 173 DIFFERENTIATION AND INTEGRATION 50, Ast. ASS. ASE, « Onees . . a B SEEEE. co prams ref" tordy 05255, () Stow tt Fe) (ot Balin Ud) Doe "Fede = FO) FO? Let fie) is continuous in 0 Explain. 255, (e) Does Piz) eostiz, O< t= te °, 20 Consider the functions (a) f(z) Determine whether cos (1/2), OS 2 Sree Prove the second mean-value theorem: If f(z) and g(z) are Riemann integrable in {a, 0) and f(x) ia monotonic in (a, 8), then for some value 7 such that @ = y= Sl reraeree = Ho fade + 105 foe ae We, (b) fle) -{ exists and if it does find its value, tints tat Gte)= f"otode wie J te)aeide = fs) 6%) de and then ue integration oo pact J. Prove Theorem 24, page 187, Use the definition of the Riemann integral as a limit of a sum to show that tin (et + ate t THEOREMS ON SEQUENCES AND SERIES ASS. AS. Ast. Ass. As. 60, AB, Prove Theorem A-25, page 158. Prove at 3:5 * BT gee eset a ‘Suppose that the sequence of functions (fa(z)) converges to f(z) in (a, ]. State and prove a theorem giving suficient conditions under which the sequence of derivatives (/,(2)) converses to f'(2) Prove that F(z) = 3S!" jg continuous and has continuous derivatives of all orders for Let ae) = Bigg for O41. () Pv tat lot ler cmvenma, (0) Show ne . : cores and explain, Use Problem A.59 to give an example of a series which can be integrated term by term but which is not uniformly convergent. nz, 055 1m incest ain fineee =f" tm eines ut ne = { Determine whether and explain.174 ‘THE RIEMANN INTEGRAL [APPENDIX A IMPROPER RIEMANN INTEGRALS A62, Investigate the existence of the Riemann integrals co Si At 01 fete ant a their values if they exist. Ane. (a) 2V2, (b) does not exist 68, Evaluate the following integrals if they exist. (@) fH . O f" ofa. @ f° Ans. (a) 4, O) , (6) does not exist AG, Prove that if 0 = f(x) = g(2) and g(z) has an improper Riemann integral in [a,b] which exists, then so also does f(z) AG5. Use the result of Problem A.64 to show that (hae exists Ash Show tn the ites (6) f 5 est only nth Cashy pica! vl sense and find these values, Ans, (a) 0, (8) J InCONVERGENCE IN THE CESARO SENSE Let s = ti + t+ +++ +t, be the th partial sum [i.e. the sum of the first 1 terms} of the series ui + ue + If the sequence of partial sums s1, 2,80, ... converges to the limit J, then by definition the series converges to J or has sum 1. ‘Suppose however that the series does not converge. ‘Then consider a new sequence sits: Staton a Ee, Bite @ where the nth term is ‘the arithmetic mean of the first m partial sums. If this sequence converges to p, we say that the series u:+a2+-+- or Su is summable in the Cesaro sense, or is C-1 summable, to p. In case the series Dw does converge to I, then the series is also C-1 summable to 1 [see Problem B.3]. ‘The process can be repeated if necessary and we speak of C-2 summability, C-3 sum- mability, ete. CESARO SUMMABILITY OF FOURIER SERIES. FEJER'S THEOREM ‘The Cesaro summability of Fourier series was investigated by Fejer. Since the partial sums of a Fourier series are given by sin (n+ aa salt) = tS eta) + fe —wy) TM dw @) where n= 0,1,2,... [see Problem 8.8, page 139], we obtain the nth term of the arithmetic means [see Problem B.4), _ sola) + aly i sin? jnu te) = Bae Sy MEH) + Hew ANE ae) We can then show that for some function t(2), ta) — Ha) = gS e+) + fee) 24) a au @ and can prove the following theorems. Theorem B-1 (Fejer]. If f(x +0) and f(x—0) exist, then the Fourier series for f(2) con- verges in the Cesaro sense to 4[f(« +0) + f(#—0)], i. the Fourier series for A(2) is C-1 summable to 4{f(e +0) + f(e—0)]. Theorem B-2. The Fourier series for f() is C-1 summable to f(x) almost everywhere, 115176 SUMMABILITY OF FOURIER SERIES [APPENDIX B Solved Problems CONVERGENCE IN THE CESARO SENSE B.l. Prove that the series 1—1+1—1+ -++ does not converge. If 4, is the sum of the first m terms, we have s)=1, 82=0, #3 =1, 64 = sequence of partial sums is 1,0,1,0,... which does not converge. ‘Thus the B2. Prove that the series of Problem B.1 is C-1 summable to 4. From Problem B.1 we see that the sequence of arithmetic means of the partial sums is 140 14041 1+0+140 2 2 ie, Lhebbd If we let t, be the nth term of the last sequence, we see that ['s n= 1,35, b= 24,6, follows that the series 1141-14 or 1 7 CA summable to 4. ‘Then since lim t= 4, B3. (@) If limu=1, prove that lim“ +™@* ~~ + _7 and (6) discuss the signifi- se 7” cance of this in connection with Cesaro summability. (@) Let w= v,+L ‘Then we must show that [+ Iepaal + oe + Io o ‘we can choose p so that given any ¢>0, |v, p. Then Ieper t peal to + eel © gt Dee bez _ (=pwe Ct s 5 Put ® [After choosing p we can choose ny 80 that for n > ny > Py Intrettml < ¢ ® ‘Then using (2) and (8), (1) becomes aytetite| oe ivaeeeeee |e ‘thus proving the required result. (®) The result shows that if the sequence of partial sums of a series does converge to 1, ie. if the series converges to J, then the series is C-1 summable to 2.APPENDIX B] SUMMABILITY OF FOURIER SERIES mT CESARO SUMMABILITY OF FOURIER SERIES. FEJER’S THEOREM B4. Using Problem 8.8, page 139, for the partial sums #,(z) of a Fourier series, prove that oe sin’ de du sin? du Sele) + ole) +n bowl) LO ee ey + flew) tala) andy From Problem 8.8, we have sin (nt Du sin Je al) = 2 P Wetw + se-w) s0 that @ ole) + ay(2) + oe + tts) = . stay fin) [OMB ABE DT e+) + pew)? ee Now consider fy = sin gut sin gut o++ + sin (nda Maultplying by at Fusing = sintJu + sinfusin fe +--+ sini fusing = Ht —cosn) + flee cop du) 4 + 4608 (eB eae mg eeeeenee Thos = © and we obtains reed, . _ ttt tale) oS, Web + fee) SEA ae ® Bo. Py it (a) Prove that 1 ("sini gg = anz sin? yu ~ and (b) thus show that if ¢(x) is some function of x um = LS” a) ~ 24(ny] Sy toa) — a) = SRS” ie tay + Hee) — 2H SE a (@) If fl) =3, then t(2) =1 also. ‘Thus from (2) of Problem B.A we obtain as requse, 1 2 1 = oar , du wa (6) Multiplying both sides of (1) by t(2) and subtracting from equation (e) of Problem B.4, we obtain a int fe) = te) = fet + ey a = Af 200 I a sin? Jue = gf vert + new = ate) SE au178 SUMMABILITY OF FOURIER SERIES [APPENDIX B B6. Prove that a Fourier series is C-1 summable to ¢(z) if and only if for any number B such that 0<3<5, 0 where F(u) = f(x-+u) + f(e—u) — 2t(2). From Problem B.3(8) itis clear that a Fourier series is C-1 summable to ¢(2) it sin? J pore Pt ae =e ie. Jim te Se cote dre au + ale Feo: sd a} o But since Be Fo 0 Spe aul = 2” SEM tu wert for 0<8en rota a ‘Thus we see that the Fourier series is C-1 summable to t(2) if and only if tn Lf ro Etae = 0 BZ. Prove that a Fourier series is C-1 summable to ¢(x) if and only if 1 sin? nu wn fro eM = 0 We have 2 1 fmol - 1 since ante 7 Je is non-negative and continuous for 0 0, we can find 8, >0 so that |F(w| <« whenever |u| <8, or <5, since w is non-negative. ‘Then we have for fixed 8; <=, [2 f'n tie Ef ro anh FS ae " «) ® is measurable, The various properties of measurable functions are analogous to those developed in Chapter 8. ‘The following theorem is important. Theorem C-1. If f(x,y) is measurable on a set [(x,y): Sb,¢SySd), then f(z,y) is ‘a measurable function of. y for almost all x € [a,b] and f(z, y) is a measur- able function of « for almost all w € [c,d]. 180APPENDIX C] DOUBLE LEBESGUE INTEGRALS AND FUBINI'S THEOREM 181 THE LEBESGUE INTEGRAL IN THE PLANE Given a rectangle P = (x,y): aS25b,eS by using subdivision points @ = m 2. Thus for example if n= 3, ie. three dimensional Euclidean space, we begin with the definition of the measure of a rectangular parallelepiped as its volume [the product of the lengths of three mutually perpendicular sides]. This leads to the notion of a triple integral. For 1>8 the ideas can be developed in an entirely analogous manner.Index of Special Symbols ‘The following list shows special symbols and notations used in this book together with the number of the page on which they first appear. Cases where a symbol has more than ‘one meaning will be clear from the context. AB, BS H(z) Tol SARS Lyle) Symbols seta, 1 elements or members of sets, 1 Fourier coefficients, 130 aleph null, cardinal number of a denumerable set, 5 aleph one or e, cardinal number of the set of real numbers R, 5 closed set, 30 open covering, 7 generalized Fourier coeficients, 12 norm of a partition, 154 positive numbers, 6, 7 function or mapping, 4 empty or null set, 1 Hermite polynomials, 145 interval, 29 1b, of upper sums for Lebesgue integrals, 58 for Riemann integrals, 154) intervals, 9 class of intervals, 34 Lub, of lower sums for Lebesgue intgerals, 58 [or Riemann integrals, 154 open subcovering, 7 Cantor set, 6 limit of fle) as 2 a, 7 Laguerte polynomials, 152 upper bound of f(a), 154 lower bound of f(x), 154 set of natural numbers or positive integers 1, 2,3, . open set, 29 partition, 154 rectangle, 180 Legendre polynomis 152 set of rational numbers, 2 set of real numbers, 2 remainder of a series after m terms, 23 upper sum for Lebesgue integral, 58 for Riemann integral, 154] 188184 Su Sx Sy), 0402) Siz) r ty oF f(a) u z INDEX OF SPECIAL SYMBOLS AND NOTATIONS lower sum for Lebesgue integral, 58 for Riemann integral, 154] upper sums for different partitions, 58, 160 lower sums for different partitions, 58, 160 partial sums of a series, 9 partial sums of Fourier series, 130, 175 limit of partial sums of Fourier series, 181 at eet, 31 arithmetic mean of partial sums, 175 universal set or universe, 1 set of integers 0, £1, #2, ....2 Notations {a,b,¢,...) set consisting of elements 8,6, ..251 (e:ete.} set of all elements x such that ete, 1 ES a belongs to [or is an element of] the set S, 1 @€S a does not belong to [or is not an element of] S, 1 4,b€S qand b belong to [or are elements of] S, 1 ACB Ai ADB A contains B,1 A=B Acquls B,1 AB Ais not equal toB,1 < less than, 2 ‘= _ less than or equal to, 2 srrester than, 2 sxceater than oF equal to, 2 (2,0) open interval a<2Y function or mapping from set X to set Y, 4 HA) set of all elements f(x) where x € A, 4 -"(B) inverse image of B under f, 4 J inverse function, 5INDEX OF SPECIAL SYMBOLS AND NOTATIONS 185 A~B A is equivalent to B, 5 (ey...) ordered n-tuplet, 6 alz,y) distance between 2 and y, 6 lal absolute value of a, 6 Re Huclidean space of m dimensions, 6 S' derived set of S,7 S closure of 8,7 lim f(z) limit of f(2) as x approaches a, 7 lim, f(z) right hand limit of f(z) as 2 approaches a, 8 lim fle) left hand limit of f(2) as 2 approaches a, 8 ay) Sequence ay, da, day +. -5 8 Tim aq oF lima, limit of sequence (a,) a8 m+», 8 i,iima, or limit superior, greatest limit or upper limit of (a), 9 lim sup a, a, or limit inferior, least limit or lower limit of (a,), 9 im in? a, (fy) oF (fa(2)) sequence of functions, fay or Za, series ay tat, 8 (1) length of an interval 1, 28 m(B) measure of a set F, 30 mB) exterior or outer measure of a set H, 80 ‘m(B) interior or inner measure of a sot E, 82 Ej(2) >} set of values x © H for which f(z)>, 43 max {f(2), fa(e)) maximum of f(x) and fa(e), 4 rnin {f/,(2),fa(2)) minimum of /,(z) and fala), 44 x(z) characteristic funetion, 49 J sede upper Lebesgue integral [except in Appendix A where 7 ‘isthe upper Riemann integral], 5¢ fice eee eee eae §e.te tags tga fe) oma nes Aen: & (R) f/f) de Riemann integral of f(z) from ato b, 54 Steed <= existence ofthe integral of fe) from a to 8, 54 ff fedde Levene trl) onthe eB 8 “Fe wiley 20), 66 E- BYfla) <0}, 65 f fle) for all 2€ B such that fl2) = p p forall @€B such that fle) > p’ Up196 INDEX OF SPECIAL SYMBOLS AND NOTATIONS 0 forall 2€B such that fla) <0” 0 for all #€B such that fle) =0 fe) for all #€E such that (2) <0" aa tA for all zB such that fle) = 0 re J Foo dn indefnive integra, 95 Hla), fle 0) lim, fleet M, lim laa +), 06 URN) or UE total variation of f(z) on [a, 5], 96 Die) = Fe) = Fpl) derivative of f(z), 97 D* fia), D~ f(z), derivates of f(z), 97 Dy fle), D_fle) P% and Nj positive and negative variations of f(z) on [ La, bor L? LP space, 115 1A Hilbert space, 115 D(f,9) oF |\f—g|| distance hetween functions f and g in L? space, 116 2), 108 limit in mean of f,(z) as n> =, 116 orthonormal set of functions, 132 orthonormal series or generalized Fourier series, 132 conjugate of 9q(2), 152 signum 2, 162 arety CA, 0-2, ete, Cosaro summability, 175 Sf tenaedy vpper Lebesgue integral of fle), 181 Sf tewaedy rower Lebesgue integral of fe), 181 erent SUS reac] ata teaAbsolute continuity, 97, 98, 105-107, 110, 111 ‘continuity and, 98, 105, 106 definition of, 97, 98 indefinite Lebesgue integrals and, 98, 99 ‘theorems on, 98, 105-107 Absolute convergence, of integrals, 73, 76, 171 of series, 10 relation of to convergence, 10 Absolute value, 6 Absolutely continuous functions (tee Absolute continuity] Absolutely integrable functions, 78, 76 ‘Accumulation points, 7,15 [vee also Limit points} Aaditivity of measure, 80 Aggregate, 1 [sce alzo Sets] ‘Aleph null and aleph one, 5 Algebra of sets, 4 Algebraic numbers, 26 Almost everywhore, 83 ‘Approximations in the least square sense, 183, 134, 147 Approximations of integrable functions by continuous funetions, 75, 76 proofs for, 86-91 Arithmetic mean, relation of to geometric mean, 119, 120 Associative law for unions and interseetions, 3 Baire classes, 44, 49, 50 Bornstein, Schroeder-, theorem, 5 Bessel’s inequality, 132, 147 proof of, 146, 147 Binary expansion, 15, 26 Bolzano-Welerstrass theorem, 7, 16, 21 Borel, Heine-, theorem, 7 Borel'sets, 83 Boundary function, upper and lower, 44 Boundary of a set, 7 Boundary points of a set, 6, 15 ‘Bounded convergence theorem of Lebesgue, for sequences, 56, 66 for series, 56 Bounded functions, 8 Lebesgue integral for, 68-71 Bounded sequences, 8 (see also Sequences] ‘uniformly, 56 Bounded sets, 2 ‘which are infinite, 7 Bounded variation, functions of [see ‘Functions of bounded variation) Rounds, greatest lower, 2, 11, 21 least upper, 2, 11, 21 ower, 2, 11 ‘upper, 2,11 INDEX 187 Cantor set, 5, 6 as a perfect set, 27 ‘countability of, 15 expressed as a ternary expansion, 15, 26 measure of, 38 Cardinal numbers, 5 ‘of denumerable set, 5 of empty set, 5 of finite set, 5 ‘operations with, 5, 26 Cartesian products, 4, 12 relationship of to functions, 4 Cauchy prineipal value of Riemann integral, 188 ‘Cauchy sequences, 9, 22 in L? spaces, 117, 122-125 Cesare summability, 175-179 Change of variables, for Lebeague integrals, 99 for Riemann integrals, 157 Class, 1 eee also Sets] Baire, 44, 49, 50 Closed intervals, 2 half, 2 Closed rectangle, 180 Closed sets, 7, 15, 16 complement of, 7, 16 Tengths of, 30 ‘unions and intersections of, 7, 17 Closed sphere, 6 Closure of a set, 7 Collection, 1 [see aleo Sets Commutative law for unions and Intersections of sets, 3 Compact sets, 7, 16 Complement, of a closed set, 7, 16 of an open set, 7, 16 of a set, 3 Completeness axiom, 2 Completeness, of L® spaces, 117, 122-125 of orthonormal sets, 124, 147 of sets, 9, 22 Component intervals, 7, 17, 24 rectangles, 180 Composition ‘function, 25 Continuity, 8, 18-20 [see also Continuous functions) absolute [see Absolute continuity) defined in terms of open sets, 8, 19 uniform, 8, 19, 20 Continuous functions, 8 [see aleo Continuity] approximation of integrable functions by, 5, 76, 86-91 Lebesgue integral of, 67, 168 relationship of with measurable functions, 44, 48 Riemann integral of, 51, 68, 156188 Continous functions (cont) theorems on, 8 uniformly convergent sequences and feries of, 9, 23 which have no derivatives, 106, 113 CCantinam hypothesis, 5 Continoum, real, 5 Conwergence, absolute [see Absolute convergence bounited, 66, 66 dominated, 1, 75, 84, 85 in meanure, 117, 18 rmean [ree Mean convergence) monotone, 75, 86 oF Cauchy sequences, 9,22, 117, 122-125 of Fourier nevis [ree Convergence of Fourier series] of infinite series, 10 [sce also Series! of integrals see Integrals] Of sequences, 8 [see also Sequences] ‘niform [see Uniform convergence) weak, 128,129 Convergence of Fourier series, 130, 13, 137-140 [ae aleo Fourier series] cesaro, 115-170 de In Vallée Poussin’s condition for, 161 Dini's condition for, 131, 140, 141 Dirichlet conditions for, 151 Bejer's theorem for, 179, 178, 178 in neighborhood of & point, 188 Jordan's eondtion for, 181, 14 pointwise, 137-140 Fuficient conditions for, 181 Cosine series, Fourier, 150 Countability, 8, 18-18 {eee aloo Countatie sets] of Cantor ae, 18 of countable sets, 5,19 of points on a line, 15 of rational nombers, 5, 13,14 of real numbers, , 14,15 Countable additivity of measure, 80 Countable sets, 5 (sce alae Countabiity] countably of, 5, 18 measure of, 32 Covering of « set, Heine-Borel theorem and, 1.22 I's theorem and, 35, 34, 39, 40 Covering of interval, 7 Decomposition theorem, Jordan's, 96, 102, 103 Decreasing or increasing monotonic functions [see Monotonic functions] Definite integrals, Lebesgue, 54 [eee also Lebesgue integrals) Riemann, 164 (see aleo Riemann integrals] de Ia Vallée Poussin’s test for convergence of Fourier series, 151 Deleted delta neighborhoods, 6 Delta (8) neighborhoods, 6 De Morgan's laws, 4 proofs of, 11, 12 Denumerable additivity of measure, 30 Denumerable sets, 5 cardinal numbers of, 5 Derivates of a function, 97 relationship of to derivatives, 97 Derivatives, 97 continuous funetions having no, 106, 113 functions of bounded variation and, 96, 97, 105 mean-value theorem for, 168 ‘of monotonic funetions, 97, 108-105 of series, 158 relationship of to derivates, 97 relationship of with integration, 95-114 right and left hand, 97 Derived set, 7, 15, Difference of sets, 3 Differentiation and integration, for Lebesgue integrals, 95-114 for Riemann integrals, 95, 157, 165-168 Dini’s condition for convergence of Fourier series, 131 proof of, 140, 141 Dirichlet conditions for convergence of Fourier series, 151 Discontinuities, functions of bounded variation and, 96 ‘monotonic functions and, 96, 100 of functions, 19 [see also Continuity] Riemann integral and, 57, 68, 156 Discontinuous functions, 19 (see also Continuity] Disjoint sets, 3 Distance, in Buclidean spaces, 6 in L” spaces, 116 Distributive laws for sets, 3 proof of, 12 Divergence, of series, 10 of sequences, 8 Domain, of a fonction, 4, 13 Dominated convergence theorem of Lebesgue, for infinite series, 75 for sequences, 74 proof of, 84, 85 Double Lebesgue integrals, 180-182 Duality, principle of, 4 Dummy symbol, 95 Egorov's theorem, 45 proof of, 50 Elements of a set, 1 Empty set, 1 cardinal number of, 5 length of, 29 Equality of sets, 1 Equivalent eets, 5 Error, mean aquare, 133, 147, 148 Euclidean spaces, 6, 20 distance in, 6 ‘generalizations of length in, 20 Even function, 150 Expansion, binary, 15, 26 series [sce Series] ternary, 15Exterior measure, 20, 24, 35 [ece also Lebesgue exterior measure] axioms for, 50 in the plane, 180 of a countable set of real numbers, 35 regular, 39 Exterior of a set, 7 Exterior point of a set, 6, 15 Fatow’s theorem, 75, 108, 125 roof of, 85 Fojer’s theorem, 175 proof of, 178, 179 Finite additivity of measure, 30 Finite sets, 5 cardinal number of, 5 Fischer, Riest-, theorem, 117, 122-125, 194, 148, 149 Fourier coefficients, 190, 126 generalized, 133 Fourier series, 130-158 complex form of, 152 convergence of [see Convergence of Fourier series} definition of, 130, 134-197 fonctions of bounded variation and, 131, 141, 142 generalized, 183, in L? or Hilbert spaces, 192, 140-148 integration of, 131, 142-144 ‘mean convergence of, 182, 188, 148, 149 partial sums of, 180, 181,'183, 138, 189 Riesz-Fischer theorem for [tee Riesz-Fiseher theorem] trigonometric series which are not, 144 Fourier sine and cosine series, 150 Fubini's theorem, 181 Fubini-Tonelli-Hobson theorem, 182 Functions, 4, 12, 18 absolutely continuous [see Absolute continuity] boundary, 44 bounded, 8, 53-71 ‘composition, 25 continuous [see Continuous functions] derivates of, 97 derivatives of [see Derivatives] discontinuous, 19 [see also Continuity] domains of, 4, 13 in L” spaces, 115, 116, 121, 122 integrable [sve Integrable functions) Jumps of, 96 limits of [see Limits of functions) lower and upper boundary, 44 ‘measurable [sce Measurable fonctions] monotonic [sce Monotonic functions) non-measurable, 49) odd and even, 150 of bounded variation [see Functions of bounded variation] fone to one, 4, 5, 13 onto, 4, 5, 18 INDEX 189 Funetions (cont.) product, 25 ranges of, 4, 18, sequences and series of, 9 square integrable, 115 step, 5t truncated, 78 unbounded [sce Unbounded functions) Functions of bounded variation, 96, 97, 100-108, 107, 158 absolute continuity and, 98, 106 definition of, 96 derivatives of, 96, 97, 105 discontinuities of, 96 Fourier series for, 131, 141, 142 integrals of, 162 monotonic functions and, 96, 100, 10% theorems on, 96, 97, 100-103 Fundamental theorem of calculus, 95, 157, 167 Generalized Fourier coefficients, 138 Generalized Fourier series, 133 ‘mean convergence of, 148, 149 Parseval’s identity for, 149 Riesz-Fischer theorem for ‘see Ries2-Fischer theorem! Geometric mean, relation of to arithmetic mean, 119, 120 Greatest limit of a sequence {see Limit superior] Greatest lower bound (g.lb.|, 2,11, 21 Half open or closed intervals, 2 Heine-Borel theorem, 7, 22 roof of, 22 relationship of to Weierstrass-Belzano theorem, 7 Hermite polynomials, 145, 146 Hilbert or L? spaces, 115 Fourier series in, 182, 146-148 Hobson, 182 Holder's inequality, 116 proof of, 121 Image, 4 Improper Riemann integral, 77, 78, 158, 170,171 Inclusion symbels, 4 Increasing or decreasing monotonic functions |ece Monotonie functions} Indefinite integrals, Lebesgue [sce Indefinite Lebessrue integrals] Riemann, 95, 157 Indefinite Lebesgue integrals, 95 [sce also Lebesgue integrals] derivatives of, 95-114 theorems on, 98, 99, 107-111 Inequality, Bessel’s, 188, 146, 147 Holder's, 115, 121 Minkowski's, 116, 121 Schwarz’s, 115, 119 triangle, 6 Inferior limit, 9, 21190 Infimum [inf], 2 Infinite sequences [sce Sequences} Infinite series, 9, 10 [see aleo Series) Infinite sets, 5 [see aleo Sets] Inner or interior measure, 32 Integers, 2 Integrable functions, approximation of by continuous funetions, 75, 76, 86-01 Lebesgue, 54, 59, 72, 78, 76 [see also Lebesgue integrals Riemann [eve Riemann integrable functions and Riemann integrals) Integrals, definite [sce Definite integrals) double, 180-182 indefinite [see Indefinite integrals) iterated, 181 Lebesgue [see Lebesirue integrals] multiple, 182 Integration [seo Integrals] of Fourier series, 191, 142.144 Integration by parts, for Lebesgue integrals, 99, 111 for Riemann integrals, 151, 167, 168 Interior of a set, 6 Interior or inner measure, 82 Interior point of a set, 6, 15, 16 Intermediate-value theorem, 8 Intersections, associative and commutative laws for, 2 of closed sets, 7,17 of open sets, 7, 16 of sets, 2 Intervals, 2 closed, 2 ‘component, 7, 17, 84 lengths of, 29, measures of, 38 nested, 9, 21 open covering of, 7 Inverse functions, 5 Inverse image, 4 Irrational numbers, 2 Iterated integrals, 181 Jordan's condition for convergence of Fourier series, 131 proof of, 142 Jordan's decomposition theorem, 96 proof of, 102, 103 Jump of a function, 96 Laguerre polynomials, 152. Law of the mean, 166 Least limit of a sequence [see Limit inferior] “Least square sense, approximations in the, 188, 134, 147 Least upper bound (L.u.b], 2, 11, 21 Least upper bound axiom, 2 Lebesgue definite integrals, 54 [see also Lebesgue integrals} Lebesgue exterior of outer measure, 31, 32 Lebesgue indefinite integrals, 95 {eee aleo Indefinite Lebesgue integrals) Lebesgue integrable functions, 54, 72, 78, 76 Inve also Lebesgue integrals} necessary and sufficient condition for Tunetions to be, 59 Lebesaue integrals, as limits of sums, 55, 61, 62 bounded convergence theorem for, 56, 66 change of variables in, 98 defined on a bounded measurable set, 55 definite, 64 definition of for bounded measurable functions, 58, 4, 57-61 definition of for unbounded measurable fonctions, 72, 73 derivatives and, 95-114 dominated convergence theorem for, TH, 15, 84, 85 double, 180-182 for bounded funetions, 53-71 for non-negative unbounded funetions, 2,77, 78 for unbounded funetions, 12-94 geometric interpretation of, 54, 78 indefinite [sce Indefinite Lebesgue integrals} integration by parts for, 99, 111 in the plane, 181, 182 lower and upper, 54, 181 lower and upper sums for, 53, 57-60 mean-value theorems for, 85, 61, 114, 141 ‘multiple, 182 notation for, 54 ‘on unbounded sets, 76, 77, 91 relationship with Riemann integral, BY, 67, 68, 17 theorems involving, 55, 56, 61-66, 73, 74, 76, 79-84 Lebesgue integration [see Lebesgue integrals] Lebesgue measurable functions |ece Measurable functions} Lebesgue measure [ace also Measure) definition of, 32 in the plane, 180 zero, 82, 33, 37, 156 Lebesgue, Riemann-, theorem, 190, 137, 189, 140 Lebesgue’s bounded convergence theorem, 56 roof of, 66 Lebesgue’s dominated convergence theorém, 1, 75 proof of, 84, 85 Loft and right hand limits, 8, 18, 19 Legendre polynomials, 152 Length, of an interval, 29 of a union of disjoint intervals, 29 of closed sets, 30 of empty set, 29 of open sets, 29 of sets, 4 relationship of with measure, 80, 81, 38 2er0, 29 Limit inferior, 9, 21Limit in mean [Li.m.], 116, 117 [see also Mean convergence] ‘uniqueness of, 117 Limit of sum definition, for Lebesgue integrals, 55, 61, 62 for Riemann integrals, 155, 161 Limit points, of a set, 7, 15, 16 Weicrstrats-Bolzano theorem and, 7, 16, 21 Limit superior, 9, 21 Limits of functions, 7, 18-20 right and left hand, 8, 18, 19 theorems involving, 7 uniqueness of, 7 Limits of sequences, 8 [sce also Sequences ‘and Convergence) uniqueness of, 8, 20 Lower boundary function, 44 Lower bounds, 2, 11 greatest, 2, 11 Lower Lebesgue integral, 54 in the plane, 181 Lower limit of a sequence [see Limit inferior] Lower Riemann integral, 154 theorems on, 155, 159-161 Lower sums for Lebesiue integrals, 53, 57-60, theorems on, 57-60 Lower sums for Riemann integrals, 154 theorems on, 154, 155, 159-161 L? spaces, as metric spaces, 116, 122 completeness of, 117 distance in, 116 Mappings, 4 Mean, arithmetic and geometric, 119, 120 Mean, law of the, 168 Mean convergence, 115-129 definition of, 116, 117 of Fourier series, 12, 183 of generalized Fourier series, 148, 149 relationship of to convergence in measure, 7 Riesz-Fischer theorem and [see Riess-Fischer theorem! theorems involving, 117, 118, 122-125 Mean square error, 133, 147 ‘condition for minimum, 147, 148 Parseval’s identity and, 147, 148 Mean-value theorem for derivatives, 166 ‘Mean-value theorem for Lebesgue integrals, 5 proof of, 61 second, 114, 141 ‘Mean-value theorem for Riemann integral, 156 generalized, 172 proof of, 104 second, 173 ‘Measurability, of Riemann integrable functions, 67, 68 of sets [ace Measurable acts] Measurable functions, 43-52 Baire classes as, 44, 49, 50 vounded, 53, 54, 57-61 191 Measurable functions (cont.) definition of, 43 in the plane, 180 limit superior and limit inferior as, 44 non, 49 relationship of with continuous functions, 44, 48 theorems on, 43, 44, 45-49 unbounded, 72, 73 Measurable sets, 31, 36-89 [ace aleo Measure] definition of, 31 Lebesgue integral defined on, 55 Measure, 29-42 additivity of, 80 convergence in, 117, 118 definition of, $1 exterior or outer [ace Exterior measure] {deal requirements for, 80 interior or inner, 32 Lebesgue [tee Lebesgue measure] Lebesgue inner, 32 Lebesgue integrals and Isee Lebesgue integrals] monotonicity of, 80 of Cantor set, 38, of countable sets, 37 of intersection of sets, 38, $7 of intervals, 33 of union of sets, 82, 38, 36, 37 plane, 180 relationship of with length, 80, 81, 33 theorems on, 82, 33 translation Invariance of, 30 20r0, 22, 83, 31, 156 Members of a set, 1 Metric spaces, 6 LP spaces as, 116, 122 Middle thirds, of Cantor set, 6 ‘Minkowski's inequality, 116, 121 ‘Mode of subdivision, partition or net, 58, 57-60, 154, 165 norm of, 165 refinement of, 155 Monotone convergence theorem, 75 proof off 86 Monotonic functions, 44, 95, 96 derivatives of, 97, 103-105 discontinuities and, 96, 100 fanctions of bounded variation and, 96, 100, 101 increasing or decreasing, 44, 95, 96, 100, 155 Riemann integrals of, 155 theorems on, 96, 100 Monotonic sequences, 8 functions [see Monotonie functions) Monotonicity of measure, 30 Multiple Lebesgue integrals, 182 ‘Mutually orthogonal functions, 132 Natural numbers, 2 Negative integers, 2 Negative variation, 103 ‘Neighborhoods, 6192, Nested intervals, 9, 21 Net, partition or mode of subdivi ‘58, 57-60, 154, 155 norm of, 155, refinement of, 155 Non-countable or non-denumerable sets, 5 Non-empty set, 10 ‘Non-measurable functions, 40 Non-measurable sets, 34, 40, 41 Non-negative property for distance in metric spaces, 6 Non-terminating series, 15 [eee alao Series] Norm, in L" spaces, 116 of a partition, 154 Normalized functions, 192 Nombers, cardinal, 5 Odd function, 150 ‘One to one correspondence, 4, 5 One to one function, 4, 5, 18 ‘onto and, 5 ‘Onto functions, 4, 13 ‘one to one and, 5 Open covering of a set, 7 and Heine-Borel theorem, 7, 22 Open intervals, 2 half, 2 ‘open sets as countable unions of, 7, 17 Open sets, 6, 15, 16 ‘as countable unions of open intervals, 7, 17 complements of, 7, 16 continuity defined in terms of, 8, 19 in the plane, 180 lengths of, 29 unions and intersections of, 7, 16 Open sphere, 6 Open sulicovering of a set, 7 Operations, with cardinal numbers, 5, 26 ‘with sets, 3, 11, 12 Ordered m-tuplet, 6 airs, 4 triplets, 4 Orthoxonat funetions, 132 Orthogonal set, 132 Orthonormal functions, 192, 144-146 definition of complex, 152 Orthonormal series, 182, 188, 144-146 Orthonormal sets, 132 Outer measure [sce Exterior measure} Parseval’s identity, 133, 147, 149 completeness and, 117, 134, 147 generalized Fourier series and, 149 ‘mean square error and, 147, 148 proof of, 147 Partial sums, 9 ‘of Fourier series, 130, 191, 188, 138, 199 Partition, net or mode of subdivision, 58, 57-60, 154, 155 norm of, 185 refinement of, 155 Perfect set, 27 Cantor set as example of, 27 Plane, measure and integration in the, 180-182 m4 Point sets, 6 definitions involving, 6, 15, 18 theorems on, 7, 16-18 Points, as real numbers, 2 boundary, 6, 18 exterior, 0, 15 interior, 6, 15, 16 limit or accumulation, 7, 15, 16, 21 of Euclidean spaces, 6 of L? spaces, 116 of spaces, 1 Positive integers, 2 Positive variati Poussin, de la Vallée, 151 Principal value, Cauchy, 158 Produet fonction, 25 Produets, Cartesian, 4, 12 Proper subsets, 1 Property method, 1, 10 Rademacher functions, 152 Range, of a funetion, 4, 18 Rational numbers, 2, U1 Real continuum, 5 Real function, 4 Real numbers, 2,10, 11 Real variables, 4 Rectangles, clozed and open, 180 component, 180 Refinement of a partition, 155 Regular exterior measure, 39 Riemann integrability, necessary and ‘suificient condition for, 57, 68, 155, 156 Riemann integrable funetions, 154-157 [see ‘algo Riemann integrals) example of non-, 159 ‘measurability of, 67, 68 special types of, 158, 161-163 theorems on, 156, 187 Riemann integrals, 53, 164-174 fas limits of sums, 155, 161 Cauchy principal value of, 158 change of variable in, 157 definite, 154 definition of, 154, 159 derivatives and, 95, 167, 165-168 discontinuities and, 57, 68, 156 geometrie interpretation of, 189 improper, 77, 78, $88 170, 171 integration by parts for, 157, 167, 168 lower and upper, 154, 155, 159-161 lower and upper sums for, 154, 155, 159-161 mean-value theorems for, 156, 164, 172, 173 necessary and suificient condition for ‘existence of, 51, 68, 155, 156 notation for, 54 of continuous functions, 155 of functions of bounded variation, 155 fof monotonic functions, 155Riemann integrals (cont.) ‘of sequences and series, 168-170 relationship of with Lebesgue integrals, 57, 81, 68, 77 theorems involving, 181-165 Riemann-Lebessrue theorem, 180, 189, 140 proof of, 137 Riesz-Fischer theorem, 117, 122-125 for genoralized Fourier series, 134, 148, 149 proof of, 124, 125 Right and left hand derivatives, 87 Right and left hand limits, 8, 18, 19 Rolle’s theorem, 166 Roster method, 1, 10 Russell's paradox, 2¢ Schroeder-Bernsteln theorem, 5 Schwarz’s inequality, 115 roof of, 119) ‘Second mean-value theorem, 173 Sequences, 8, 20-24, 158, 168-170 bounded, 8 Cauchy, 9, 22, 117, 122-126 convergence of, § [ace algo Convergence] in L? spaces, 117, 122-125 limit inferior and limit superior of, 9, 21 limite of, 8, 20 of functions, 9 of partial sums, 9 uniform convergence of, 9, 28, 158, 168, 169 [ace also Uniform convergence] uniformly bounded, 58 Series, 9, 158, 168-170 convergence and divergence of, 10 definition of, 9 derivatives of, 158 Fourier [see Fourier series) orthonormal, 182, 128, 144-146 partial sums of, 9 sum of, 10 uniform convergence of, 10, 23, 24, 135, 136, 158, 169 [see also Uniform eonvergenes] Set, Cantor [see Cantor set] ‘Set function, 4 length as Sets, 1, 10-12 algebra of, 4 associative, commutative and distributive laws for, 3 Borel, 33 boundary point and boundary of, 6, 7, 15, bounded, 2 bounded infinite, 7 ‘eardinal numbers of, 5, 25 closed [see Closed sets] closure of, 7 compact, 7, 18 countable, 5 covering of, 7, 22, 94 denumerable, 5 derived, 7, 15 198 Sets (cont,) difference of, 8 disjoint, 8 clements or members of, 1 ‘empty or mull, 1 ‘equivalent, 5 exterior point and exterior of, 6, 7, 15 finite, 5 inflate, 5 interior point and interior of, 6, 15, 18, intersections of, 8 lengths of, 34 ‘measurable [sce Measurable sets) ‘measure of [see Measure] non-measurable, 34, 40, 41 open [see Open sets] operations with, 8 ‘orthonormal and orthogonal, 182 perfect, 27 point, 6, 7, 16-18 singleton, 10 subsets of, 1 test, 31 theorems involving, 3, 4 unbounded, 78, 77, 01 universal, 1 jgnum 2 (sgn (2)), 152 Sine series, Fourier, 150 Singleton set, 10 Space, 1 Buclidean, 6, 29 Hilbert, 115) L?,115-119 Sphere, open and closed, 6 Spherical neighborhood, 6 Square integrable functions, 116 Step functions, 51 Subcovering of a set, 7 Subsequences, 117, 118, 126 Subsets, 1 Summability of Fourier series, 182, 175-179 Sum of an infinite series, 10 Sums, lower and upper [see Upper sums and Lower sums] Superior limit, 9, 21 Supremum [sup], 2 Symmetric property, for distance in metrie spaces, 6 Ternary expansions, 15 Cantor set and, 15, 26 Test sets, 31 ‘Tonelli, 182, Total variation, 96, 108 Transcendental numbers, 28 ‘Transfinite numbers, § {see also Cardinal number) Translation invariance of measure, 30 ‘Triangle inequality for metric spaces, 6 Trigonometrie series, 150 which are not Fourier series, 144 ‘Triple integral, 182 Truncated function, 78194 INDEX ‘Unbounded functions, 12, 13, 77, 78 Lebesgue integrals for, 72-04 in the plane, 181 Riemann integrals for Upper limit of a sequence [eee Limit [see Improper Riemann integrals] superior} Unbounded sets, Lebesgue integrals on, Upper Riemann integral, 154 76, 77, 91 ‘theorems on, 154, 155, 159-161 Riemann integrals on (see Improper Riemann integrals] Uniform continuity, 8 19, 20 definition of, 8 Uniform convergence, definition of, 2, 23 Egorov's theorem and, 45, 50 of sequences, 9, 28, 158, 168, 169 of sequences’ and series of continuous functions, 9, 23 of series, 10, 23, 24, 195, 126, 158, 168, 169 Weierstrass M test and, 10 Uniformly bounded sequences of functions, 56 Unions, associative and commutative Upper Lebesgue integral, 54 Variables, real, 4 change of for Lebesgue integrals, 99 change of for Riemann integrals, 157 Variation, bounded [gee Functions of bounded variation) positive, negative and total, 96, 103, Venn diagrams, 8, 11, 12 Vitali covering, 34, 39 Vitali’s covering theorem, 83, 34, 89, 40 proof of, 39, 40 Volume, 29, 182 ‘Weak convergence, 128, 129 laws for, 3 of closed sets, 7,17 of open sets, 7, 16 of sets, 3 of sets represented as unions of mutually Weierstrass-Bolzano theorem, 1, 16 root of, 21 relationship to Heine-Borel theorem, 7 Weierstrass funetion, 113, Weierstrass M test, 10, 23, 24 disjoint set, 18 roof of, 23 Universal set, 1 Universe, 1 zy plane, 4 Venn diagrams for, 3 Universe of discours Upper boundary function, 44 Zero measure, 32, 38, 87, 156 Upper bounds, 2, 11 Zero property, for distance in metric least, 2, 11, 21 spaces, 6 ero length, 29Schaum s OUTLINE SERIES ISBN 0-07-0b0221-2 wigan SCHAUNS REAL VARIABL WIMIMIEAIN + +250