Professional Documents
Culture Documents
Mathematical Analysis (2nd Ed)
Mathematical Analysis (2nd Ed)
1, the finite union ‘This is open, since it i the union of open sets. We shall show that for some value ‘of m the union S,, covers A. For this purpose we consider the complement R* — S,, which is closed, Define a countable collection of sets (Q1, Qa,..-} a6 follows: Q, = A, and for m>i = dom Sy, ‘Thats, Qa consists of those points of 4 which ic outside of Sq. Ifwe can show that for some Yalu of the set Q is empty, then we will ave shown that for this te point of lies outside in oter words, we wil hve shown that some Sy Observe the following properties of the sets Q4: Each set Q, is close, since its the intersection ofthe closed set 4 and the closed set R" — 5. The stis Oy sate decteasin, since the Sy are increasing: that is, Qn; © Qn” The sts Qyr being subsets of 4, are all bounded. Therefore, i no set 0, is empty, we can apply the Cantor intersection theorem to cogclude that the intersection (2, Q, is also not empty. This means that chee i some point in A which i in all the Sts Qq» or, what isthe same thing, outside all theses S,. But this is impossible, since A'S Ufua Si. Therefore some 0, must be empty, and this completes the proof, mast Compacoen in Re *” 312 COMPACTNESS IN R* We have just seen that if a sct S in R¢ is closed and bounded, then any open covering of S cam be reduced toa finite covering. TL is natural to inquire whether there might be sets other than closed and bounded sets which also have this ‘property. Such sets will be called compact 5.30 Definition of « compact set. A set S in RY is said to be compact If, and only if, ‘every open covering of S contains a finite subeover, that is, a finite subcollection which also covers S. “The Heine-Borel theorem states that every closed and bounded set in R* is ‘compact. Now we prove the converse resul Theorem 331. Let § be a subset of R°. Then the following three stasements are equivalent: 8) Sis compact. b) Sis closed and bounded. .¢) very infinite subset of S has an accumutation point in S. Proof. As noted above, (b) implies (a). If we prove that (a) implies (b), that (b) implies (¢) and that (c) implies (b), this will establish the equivalence of all cree statements. ‘Assume (a) holds. We shall prove fist thar SIs bounded. Chowse 2 point p in S. The collection of m-balls B(9; 4), k = 1,2, -.-. isan open covering of By compactness a finite subcolection also covers Sand hence Sis bounded, [Next we prove that S is closed. Suppose is not closed. Then there is an accumulation point y of S such that y ¢ S. Ix ¢ S,letr, = [x ~ yhj2. Each ry is postive since y # S and the collection {B(x; r,):* € S} is an open covering of ‘S. By compactness, a finite number of these neighborhoods cover S, sty se Let r denote the smallest ofthe radii Fae... Then itis easy to prove that the ball B(y; r) has no points in common with any of the balls B(x; 10). In fact, if x€ B57), then x= yl] < rs ry and by the triangle inequality we have ly — 4h Sly — xb + bx ~ ml 80 [x= al 2 ly - x 9h = 4 = Ix vl > Hence x ¢ Bla; 4). Therefore Bly: r) 0 Sis emply, contradicting the fact that isan accumulation point ofS. This contradiction shows that Sis closed and hence ‘) implies () ‘Assume (b) holds. In this case the proof of (c) is immediate, because if T is ‘an infinite subset of S then T is bounded (since S is bounded), snd hence by the olzano-Weierstrass theorem has an accuraulation point x, say. Now x is also Box ndo ements of Pott Set Topology et 332 12 xeumolion point ofS and bene x S, nce Sis coed, Therefore () implies (€). Assume (@) holds. We shall prove (b). If S is unbounded, then for every 'm > Othere exists. a point x, in Swith [xql| > m. The collection T ~ (x3, X35...) ‘an infinite subset of S and hence, by (c), T has an accumulation point y in S, But for m > 1 + [yl we have Dy — yl Bx — ty > m= yt > ing the fact that y is am accumulation point of T. This proves that Sis cont bounded, To complete the proof we must show that Sis closed. Let x be an accurmulation point of S. Since every neighborhood of x contains infinitely many points of S, we can consider the neighborhoods B(x; 1/k), where k = 1, 2,...,and obtain a ‘countable set of distinct points, say T = {x,, x3... Js Contained in S, such that 2, B(x; 1/h). The point x is also an accumulation point of T. Since 7 is an infinite subset of S, part (c) of the theorem tells us that T must have an accumula- tion point in S. The theorem will then be proved if we show that x is the only accumulation point of 7. ‘To do this, suppose that y # x. Then by the triangle inequality we have Wy xt sly ml + by xl Oi x # y. 3. dex, ») = dy, 2). 4. dls, y) & dbx, 2) + diz, The nonnegative number d(x, y) is to be thought of as the distance From x 10 _y. In these terms the intuitive meaning of properties | through 4is clear. Property 4s called the triangle inequality. Polat Set Topology a Metre Spaces a ‘We sometimes denote « metric space by (M, d) to emphasize that both the set -M and the metric d play a role in the definition of a metric space. Examples 1. M = RYsd(x,y) = |x ~ yf. Thisizcalled the Euclidean metric. Whenever we refer te Baclidean space R*, it will be understood thatthe metric is the Euclidean metric unless another metric is specifically mentioned. 2 M = C, the complex plane; dlzy, 22) = \ty ~ 2]. AS. metric space, Cis indstin- ‘uishabie from Euclidean space R* because it has tbe same points and the same metric. 3. Many nonempty set; d(x. y) = OiCx = y, dx, ») = LiEx # y. This called the ‘discrete meri, and (M,d) is called a discrete metric space. ‘4 If (44 d) isa metic space and if Sis any nonempty subset of then (S,d) is also & retric space with the same metric or. more precsly, with the restriction of d to ‘Sx San metre. This is sometimes called the relauce metic indoced by don S, and ‘Sis called a metric subspace of M. For example, the rational numbers Q with the sets d(x, ») = Je — y1 form a metic subspace of R. SM = RE, dtu, 9) = VE, = 7) + Mey = 7a), where x= Gey) andy» (G92. "The metric space (Md) is not « metric subspace of Euclidean space RP ‘because the metric is diferent 6 Mm (xyx)ia} + do 1), the unit cide in RA: dO smaller arc joining the two points x and y on the nit circle. the length of the TM = (xy, Say 9) 294 +d + 99 ~ 1, the it aphere fa RO: dx») — the length ‘ofthe smaller are along the great crcl joining the two points x andy. BM = Ri dex) = bey — yl Hoot bo 9. M = RE; dex, YY = max fin, — Yulee se b= 34 POINT SET TOPOLOGY IN METRIC SPA‘ ‘The basic notions of point set topology can be extended to an arbitrary metric space (M. d). Tae M, the ball Bla; r) with center a and radius r > 0 is defined to be the set ofall xin M such that dix, a)0. Hence By(x: r) = Blt: S'S AS =X ws open in S. ‘Conversely, assume X is open in S. We will show that X = 4 n S for some ‘open set A in Bf. For every x in X there isa ball Bg(x: r,) contained in X. Now Bax; 7.) = Bulors 1.) 7S, $0 if we let A= U Bass rs then is open in Mf and itis easy to verify that Am S = X, Theorem 34. Let (S, d) be a metric subspace of (M, d) and let ¥ be a subset of 'S. Then ¥ is closed in Sif, and only if, ¥ — BS for some set B whichis closed in M. Proof. IY = BS, where Bis closed in M, then B = M — A where 4 is open in Ms0 Y= Sm B= Sc(M ~ A) = S~ A; hence ¥ is closed in S. Conversely, if Yis closed in S, let X = 8 — ¥. Then ¥ is open in S so X = AS, where 4 is open in Af and S-(UoS)=S— A= SOM A=SH8, where B = M — 4 is closed in M. This completes the proof. If = M, x poi x in M is called an adherent point of Sif every ball By(x: r) contains at least one point of S. If x adheres to S— {x} then x is called an accumulation point of S. ‘The closure § of $ isthe set of all adherent points of S, ‘and the derived se¢ 5" is the st ofall accumulation points of S. Thus, S = Su 5" Yes-Xx m3 (Compact Sant of « Mette Space « ‘The following theorems are valid in every metric space (M, d) and are proved exactly as they were for Euclidean space R*. Tn the proofs, the Euclidean distance ix — yl need only be replaced by the metric d(x, ») ‘Theorem 3.35. 3) The union of any collection of open sets is open, and the inter- section ofa finite collection of open sets is open. ) The union of a finite collection of closed sets is closed, andthe intersection of any collection of closed sets is closed. Theorem 3.36. If A is open and B is closed, then A ~ Bis open and B— Ais closed. ‘Theorem 337. For any subset $ of M the following statements are equivalent a) Ss closed in M. 1) S contains all its adherent points ©) S contains all ts aceurmulation points. Qs=5. Example, Let M = Q, the set of rational numbers, with the Euclidean metric of R Let $ consist ofall rational numbers i whe open interval (a, 6}, where both a and b ate inrationsl. Then 5's a closed subset of Q. Our proofs of the Bolzano-Weierstrass theorem, the Cantor intersection theorem, and the covering theorems of Lindel6f and Heine-Bore! used not only the metric properties of Euclidean space I" but also special proper ies of R not gen- erally valid in an arbitrary metric space (M, d). Further restrictions on Mf are required to extend these theorems to metric spaces. Ove of these extensions is ‘outlined in Exercise 3.34 “The next section describes compactness in an arbitrary metric space. 31S COMPACT SUBSETS OF A METRIC SPACE. Let (M, d) be a metric space and let S be a subset of M. A collection F of open subsets of Mis said 10 be an open covering of Sif $< User 4 'A subset S of Af is called compact if every open covering of $ contains finite subeover, Sis called bounded if $ = B(a; r) for some r > 0 and some ain M. Theorem 3.38. Let S be a compact subset of a metric space M. Then 4) S is elosed and bounded. ii) Every infinite subset of S has an accumulation point in S. ‘Proof. To prove (i) we refer to the proof of Theorem 3.31 and use that part ofthe ‘argument which showed that (a) implies (b). The only change is that the Euclidean distance Ix — yl is to be replaced throughout by the metric d(x, 1).‘assume that no point of Sis an eccumulation point of 7, ‘Then for each point x in ‘S there isa ball B(x) which contains no point of T (if x ¢ T) or exactly one point of T(x uself, if x€ 7). As x runs through S, the union of these balls B(x) is an ‘open covering of S. Since Sis compact, a finite subcollection covers $ and henee also covers 7. But this is 2 contradiction because T'is an infinite set and each ball Contains at most one point of 7: Nore. In Euclidean space R*, each of properties () and (i) is equivalent to com- paciness (Theorem 3.31). In'a general metric space, property (i) is equivalent to ‘compactness (for a proof see Reference 3.4), but property (i) is not. Exercise 3.42 -zives an example of a metric space M in which certain closed and bounded subsets are not compact. Theorem 339. Let X bea closed subse of @ compact metre space M. Then X is compact Proof. Let F be an open covering of X, say XS User A. We wil show that a finite number ofthe sets A cover X. Since Xis closed its complement M— X is ‘open, so Fu {(M — X)} isan open covering of M. But M is compact, 30 this covering contains afte subcover which we can assume includes MX. There: fore ME ALU Ud, ~ XD This subcover also covers X and, since M — X contains no points of X, we can delete the set Mt — X from the subcover and sill cover X. Thus XA, U"" UA, 0 X is compact. 346 BOUNDARY OF A SET Definition 340. Let S be a subset of a metric space M.A point xin M is called a boundary point of S if every bell By(x; r) contains at least one point of S and at least one point of M ~ S. The set ofall boundary points of Sis called the boundary (of S and is denoted by a5. ‘The reader can easily verify that a= SaM—s. ‘This formula shows that 25 is closed in Example In R’ the boundary of a ball Ba; r) ste set of points x such hat fx — a Tn RY, the boundary of the set of rational numbers i all of R? Further properties of metric spaces are developed in the Exercises and also in Chapter 4 EXERCISES: ‘Open and clase sets in W* and? 3.1 Prove that an open interval in R* isan open set and that a clowed interval is a closed ‘32 Determine all the secumulation points ofthe following ses in R? and decide whether the sets are open or closed (or neithet). 2) All integers 1) The interval (a, 5 ©) Alloumbers ofthe form In, = 123.00 © Al rational oumbers, ©) All numbers ofthe form 2-* + 5", (mn= 1.0). 1) Allmumbers ofthe form (—1)" + (im, m= 1,2,...) 1) All numbers ofthe form (1/) + fm), (mae Q.d 1) All numbers of the form (— Y/N + Gi}, = 3,2,. 33 The same as Exercise 3.2 forthe following sets in R? 4) All complex such that [2] > 1 ') All complex z such that | 2 ©) Allcomplex numbers ofthe form (it) + Gm. (mn = 1.2. 4) All poins (x,y) such that a? ~ y? <1 ©) All points (x, ») such that x > 0. 1) All points (x, ») such that x = 0. 3.4 Prove that every nonempty open set $ a RY contains both rational and irrational numbers. ‘3.5 Prove that the ony sets in R* which are both open and closed are the empty set and 1B! itself Tea similar statement true for 2? ‘24 Prove that every closed st in Ris the intersection of« countable collection of open sets 3.7 Prove that « nonempty, bounded closed set Sin R! is elther a closed inerva, or that ‘Scan be obtained from a ciosed interval by removing a countable disjoint collection of ‘open intervals whose endpoints belong to S. ‘Open and closed sts In R* ‘3.8 Prove that open m-balls and m-imensional open intervals are open ets in RY 3.9 Prove that the interior of a set in R* is open in R". 340 IFS & RY, prove that int Sis the union ofall open subsets oft which are contained in S. This is described by saying that int S's the largest open subset ofS. AL IFS and Tare subsets of R, prove that (int 5) (int T) ~ int (SF), and int $)U Gat TY int (SUT).“ lement of Piet Set Tepoiogy 3.2 Let 5" denote the derived et and 5 the closure ofa set Sin R', Prove that 2) Sis closed in RY; that, (SY SS" DISS Tithe ss 7 Qu sur, o® ©) Sis closed in RY, 1) Sis the intersection ofall closed subsets of R* containing S. That is, Sis the smallest closed set containing S. 3.13 Lat Sand Thesubsets of RY, Prove that SAT 5 SF andinat SOT ¢ SAT If Sis open, NOTE. The statements in Exercises 39 through 3.13 are true in any metic space. 3.14 A stt Sin RY is called conve if for every pac of points x and yin Sand every real B satistying 0 < 0 < 1, wehave Ox + (I ~ Oly S. Interpret this statement grometric- ally Gn and 2) and prove that: 9) very bull in RY is conver. ) Every maimensional open interval is conven ©) The interior of a convex seis convex 4) The closure of a convex set is convex, AS Let F be a colletion of ses in RY, and Set $= User A and T= (\ser A. For fich of the following statements, either give a proof or exhibit a counterexample 8) Ifx isan accumulation point of 7, then x is an accumulation point of each set ') fxs an accumulation point ofS, then x i an accumulation point of at least one set Ain F. 346 Prove that the et $ of rational numbers inthe interval (, 1) cannot be expressed 2s the imersection of a countable collection of open set. Hint. Write S = (ry, %3)--- ‘assume = (EL, 5,, where each 5; is open, and construct a sequence {Q,) of closed {intervals such that Q,,, © Q, © Sy and such that s3¢ Q_ Then use the Cantor iner- section theorem to obiin a contradiction. ‘Covering theorems fo R* 37 IFS © R®, prove thatthe collection of isolated points ofS is countable, 3.18 Prove that the set of open disks in the sonplane with center at (x, 2) and radius x > 0, x rational, isa countable covering of the set, 9): > Oy > 0). 3.49 The collection Fof open intervals of the form (I/n, 2m), where # = 2, 3,..., isan ‘open covering of the open interval (0, 1). Prove (without using Theorem 3.31) tat no Finite subcollection of Feovers (0,1) 3.20 Give an example of a set S which is closed but not bounded and exhibit a countable ‘open covering F such that no finite subset of F covers S. 3.21 Given a set Sin RY withthe property tht for every x in there isan a-ball x) such that B(x) © Sis countable. Prove that $s countable, 1322 Prove that a collection of disjoint open sets in Rts neSessirily countable, Give an ‘example ofa collection of disjoint closed ses which i not countable, a o 3.23 Assume that $ & RY, A point xin RY is sad tobe a condensation pot of if every ‘ball BG) has the property that B(x) 7 Sie not countable, Prove that if Sis net count- Able, then there exists & point x én S such that x is a condensation point ofS. ‘324 Assume that § & R and assume that $ fs not countable. Let T denote the sot of ‘condensation points of S. Prove that ‘i 8) S ~ Tis countable, ) $0 Tie not countable, © Tis a closed set, 4) T contains no isolated points. Note that Exercise 3.23 isa special case of () 3.5 A sat in Ris called perfec if S = S, that ii Sis closed set which contains a0 ‘isolated points Prove that every uncountable closed set Fin RY can be expressed in the form F = AB, where A is perfect and 2 is countable (Canor-Bendxon theorem). Hint, Use Exercise 324. Metric wpaces| ‘3.26 In any metric space (Md), prove thatthe empty set and the whole space Mf are ‘ort open and closed ‘327 Consider the following two mets in Rt: ee) = ae tes nh lad = $v — a In each of the following metric spaces prove that the ball B(a; r) has the geometric appearance indicated: 9) Ind), 'b) In (RY, dy) square with diagooals parallel to the axes ©) A cube in (R*, di). 6) Am octahedron in (, d). ‘328 Let d; and dj be the metris of Exercise 3.27 and let [x — y] denote the usual Euclidean metric. Prove the following inequalities for all x and y in Rt square with sides paral! to the coordinate axes. x,y) < Ix = yhSdsy) and — sx. y) < Vax ~ yh < nda, 3.29 1F(M, d) isa metric space, define dex») To dee) rove that dis also a metic for M. Note that © = dx, y) < I forall x, yin ME. 3.30 Prove that every finite subset of a metric space is closed. 331 In metric space (M, d) the claed Ball of radius ¢ > 0 about a point 2 in M isthe set Bla; r) = tex, @) Se). 8) Prove that Ai; 7) sa closed set ) Give an example ofa metric space in which Ba: ‘all Ba) fen ig not the elosure ofthe open« ‘ements of Pott Set Toplogy 3.32 Ina metric space M, if subsets saify AS Ss , where Zi the closure of 4, then ‘A staid to be dense in. For example, the set Q of rational numbers i dense in R. It 1 is dense in $ and if $ is dense in T, prove that 4 it dense ia. 3.39 Refer to Exercise 1.32. A meitic space M issahd to be separable if there isa countable subset 4 whichis dense in M. For example, R i separable because the set Q of rational ‘number is countable dense subset, Prove that every Euclidean space RFs separable. 3.34 Refer to Exercise 3.33, Prove that the Lindel covering theorem (Theotem 3.28) is valid in any separable metric pace. 3.35 Refer to Exercise 3.32. IF Ais dense in Sandit Bis openin S,provethat 8 § 7AB. Hint, Exercise 3.13. 13.36 Refer to Exercise 332. each of 4 and Bis dense in $ and if Bis open in S, prove that 47 Bis dense in S, 3.37 Giveo two metric spaces (S,, dj) and (S., dz), a metric p forthe Cartesian product ‘5, » S,can be constructed from dy and din many ways. For example, ix = (ry, 22) ‘and y = (re, 7) ar in Sy % Sy. pls9) = duty. 9x) + data Ja) Prove that is a metric for 5, S, and construct further examples. (Compact subsets of a metric space Prove each ofthe following statements concerning an arbitrary metric space (M, d) and subsets, T of M. 338 Assume $s T's M, Then Sis compact ia (M. dif, and only if, Sa compact in the metic subspace (Fd). 3.99 IC Sis closed and T's compact, then 7) Tis compact. 340 The intersection of an arbitrary collection of compact subsets of M it compact. 3.41 The union of file number of compact subsets of M is compact, 342 Consider the metric space Q of rational numbers with the Boctidean metric of. Let $ consist ofall rational numbers in the open interval (a, 6). where a and 6 are ira- tional. Thea 5 isa closed and bounded subset of Q which i not compact. Minceincus properties ofthe ner td the hry 16 and dee abicary sabes of rs spce M, prov that: 20 int = MHA. 4 in (= A) = ME 2345 int Gin = int 6 5) int (Ys Ad ~ (Yea Gt 4), wher ach Ay SM 8) it (aar 4) Place Git A) if Fis a iit colton of ube of ME ©) Give an example wher: equality does not hold in (b). SAT 9) User Gat 4) 5 it User 1) Gie a example of finite coliction Fin which equality doesnot holdin (a SAN) int 2A) = 864i open oi A ie ove in 1) Give an exp in wich it (24) M. 3.49 IF int A = int B= B and i Ais closed in M, then int (4 UB) = @ 3.50 Give an example in which int A = int B = @ but int (4 U B) = M. 35184 = OM =A and 24 = AM 4, 382 UAE = 9, thea XA vB) = 24a SUGGESTED REFERENCES FOR FURTHER STUDY 34 Boas, RP, A Primer of Real Funesions. Carus Monograph No. 13. Wiley, New York, 1960. 3.2 Gleason, A., Findamental of Abstract Analysts. Addison-Wesley, Reading, 1966. 33 Kaplansky, 1, Set Theory and Metric Spaces. Allyn and Bacon, Boston, 1972 3.4 Simmoas, G. F.,Inredcton o Topology ond Modern Analysis. McGraw ‘York, 1963. NewCHAPTER 4 LIMITS AND CONTINUITY 41 INTRODUCTION “The reader i already familiar with the limit concept as introduced im clementary ‘aleulus whee, a fect, several kinds of mats are usually presented. For example, the limit of a sequence of real numbers {x,}. denoted symbolically by writing lim x, = A, means that for every number ¢ > 0 there is an integer N sul that my Al N. ‘This limit process conveys the intuitive idea that x, can be made arbitrarily close to A provided that m is sufficiently large. There is also the limit of a function, indicated by notation such as lim f(a) = A, ‘hick means that for every # > 0 there is another number § > O such that @) — Al 0 there is an integer N such that ley p) N. 7 ne (Caorreet Seqpences a Mitre Sc n We also say that {3,) converges 10 p and we write x,» p as n—» 00, oF simply ep. If there is no such pn S, the sequence (x, 1 said 10 diverge. Note, The definition of convergence implies that Xa 7p ifand only if d(x p) + 0. ‘The convergence of the sequence (dbx 9} 00 takes place in the Evelidean metric space R!. Examples 1, In Euclidean space R, a sequence (x) is called increasing if x, < xyes forall n. IF an increasing sequence is bounded above (that is, i x, < Af for wome M > O and all), then (,} converges tothe supremum ofits range, sup {xy ay... Similarly, {ra} is called decreasing if 4 for all x. Every decreasing sequence whichis ‘bounded below converges tothe infcnum ofits range. For example, {1/a} converses 100. 2. If ay) and (6, are real sequences converging 00, then (a, + b,) also converges to 0. If 0 = c= a forall m and if (a,) converges to 0, then {e,) also converges t0 0. ‘These elementary properties of sequences in R! can be used (0 simplify some of the ‘roots concerning limits in a general metric space. 3. In the complex plane C, let 2g = 1+ 7? + (2 ~ IVa. Then (2,) convernes 10 1+ 2 because Hig 1 + 27 = bg — C14 20? = Ze 50 dla 1 + 21) -+ 0, 0am = Theorem 4.2. A sequence (x,} ina metric space (S, d) can converge 10 at most one ot in S. Proof. Assume that x, p and x, +4. We will prove that p= q. By the triangle inequality we have 0 < dlp, g) < dp, x) + ty Q)- Since d(p, x,) + 0-and d(ey a) -+ 0 this implies that d(p, 9) = 0,50 p = 4. If. sequeace (3,} converges, the unigue point wo which it converges i called the limit of the sequence and is denoted by lim x, or by lim... Xy- Example. to Fuctidean space Rt we have fim, In = 0. The same sequence in the rmetisubepace T = (, 1] doesnot converge Bocse the ony candidate forthe ini is Oand0¢ 7. Thisexampie sows tha! the convergence or dverzaee of sequen depends ‘on the undetying space as well as onthe metre ‘Theorem 4.3. In a metric space (S, d), assume x, - p and let T = {x,, x3,--.} be the range of {x,). Then: 2) T is Bounded. 1) pis an adherent point of T.n ts nn Continty maa Prof. 2) Lt 1 be he near cormapndig 10 # = 1a the on of com vere, Then every wit» 2 fin te bal ps0 ey pein a ‘ies im the ball B(p; r), where ae 1+ max (de, x, 4p, Xn-))- ‘Therefore T is bounded. by Sinoe every ball B(p; «) contains a point of 7, pis an adherent point of 7. Nore. If Tis infnie, every ball B(p; ¢) contains infinitely many points of 7, 80 Pia an accumlation point of 7: ‘The next theorem provides a converse to part (b)- ‘Theorem 4.4. Given a metric space (S, d) and a subset T <= S. If a point pin S ir an adherent point of T, then there isa sequence {x,} of points in T which converges wp. Proof. For every integer m 2 1 there is a point x, in T with ap, x,) < ifn. Hence d(p, x3) + 0, 30.5, +p. ‘ Theorem 45. Ina metre space (S, 4) o oqunce {x} camerges 1 pf, and every subsequence (uy) converges 1p. Phodeny Proof. Assume + p and consider any subsequence {syq3). For every € > 0 ‘there is an W such that m > implies d(x, p) < ¢. Since (24) it a subsequence, there is an integer M such that k(a) 2 N for n= M. Hence m > M implies 8, which proves that xj -» p. The converse statement holds trivially since (,} i itself a subsequence, 43 CAUCHY SEQUENCES fa sequence {x,} converges to a limit p, its terms must ultimately become close to 1 and hence close to each other. This property is stated more formally inthe next Theorem 4. Assume that {x,} converges in # metric space (S, d). Then for every > O there is an integer N such that Mey) <6 whenever n> Nand m > N. Proof. Let p = lim xy Given ¢ > 0, let N be much that dix p) < 2/2 whenever mz. Then din) O there ison integer N such that dey xq) N. ‘Theorem 4.6 states that every convergent sequence is a Cauchy sequence. The converse is not true in a general metric space. For example, the sequence {In} is 1 Cavehy sequence in the Euctdean subspace T = (0, 1] of R*, but this sequence does not converge in T. However, the converse of Theorem 4.6 is true in every Euclidean space Rt Theorem 4.8. In Euclidean space R every Cauchy sequence is convergent. Proof. Let (xq) bea Cauchy sequence in RY and let T = (x, a...) be the range of the sequence. If Tis finite, then all except a finite number of the terms {x,) are ‘qual and hence {x,} converges 10 this common value. [Now suppose Tis infinite. We use the Bolzano-Weicrstrass theorem to show that T bas an accumulation point p, and thea we show that {x,} converges top. First we need to know that T's bounded. This follows from the Cauchy condition. In fact, when 2 = 1 there isan N such that n 2 N implies Ix, — xf <1. This ‘means that all pons x, with w 2 1 le inside a ball of radius Yabuul xy us vente, 0 T lies inside a ball of radius | + M about @, where M is the largest of the numbers [xy[,.+-, D&yl-. Therefore, since T is a bounded infinite set it hus an accumulation point p in RE (by the Bolzano-Weierstrass theorem). We show next that {x,} converges (0 P. Given ¢ > 0 there is an N such that x, ~ Xqll <¢/2 whenever m= N and ‘m= N. The ball B(y; #/2) contains a point x, with m > N. Hence ifn > N we have x ll < Ii — H+ kav < 5+ = 40 lim x, ~ p. "This completes the proof. amples 1. Theorem 48 often used for proviog the convergence of a sequence when he Hint {not Known in advance. Por example, consider the equece in B! dened by™ ‘Limits an Cotiaity Det 49 50 tm ~ se! < €usSoona5 N > Ie, Therefore {3} iba Cauchy sequence und hence ‘i converges 10 sore limit. ean be shown (see Exercise 8.18) that this limit slog, 2 fact which is not immediately obvious. Given a real sequence {a4} such that ldys2 — dyes) 5 Hogg ~ ay fOr all n @ 1 We can prove that (ay) converges without knowing it mit. Let by = lay: ~ a ‘Then bess & By/280, by induction, d,s 6/2". Hence, ~+ 8, Also ifm > n wehave Een vi lena Bed (te bens ence <2. ¥ ‘This implies that (a) Cauchy sequence, so a,} converses. 44 COMPLETE METRIC SPACES Definion 4.9. A metric space (S, a) is called complete if every Couchy sequence in S converges In S. A subset TofS sealed complete f the metric subspace (T. 4) 15 complet. Example 1. Every Euclidean space R* is complete Theorem 4). tn particular, Ris ‘compete, bot the subspace 7 = (0. 1 snot complete Example 2 The space R withthe met 0x) sax cian ‘The next theorem relates completeness with compactness 17s complete. Theorem 4.10. In any metric space (S, d) every compact subset T is complete. Proof. Let (x,} bea Cauchy sequence in Tand let A = (xy,.X3..-.} denote the range of (x). If is finite, then {x,} converges to one ofthe elements of 4, hence {x4} converges in T. If is infinite, Theorem 3.38 tells us that A has an accumulation point p in T since Tis compact. We show next that xy -+ p. Given ¢ > 0, choose N so that > Nand mz N implies d(x, x4) < o/2. The ball B(p; 6/2) contains a point 2_ With m & N. Therefore ifm > [NV the triangle inequality gives us Ain 2S dln Aad + dite <5 + Fm 6 s0.x, + p. Therefore every Cauchy sequence in T'has.a limit in 7, s0 Tis complete. 45 LIMIT OF A FUNCTION In this section we consider two metric spaces (S, d:) and (T, dy), where dy and dy denote the respective metrics. Let 4 be a subset of $ and let fi» T be a function from 4 to T ino RR RR a a man Lint ofa Function 1s Definition 4.11. If p isan accumulation point of A and if 5 © T, the notation lim fix) = 5, o is defined to mean the following: For every e > O there isa > 0 such that dy(fis),B) <2 whenever xe A, x # p, and de(x,p) < 3, ‘The symbol in (1) is read “the Timit of /(), as x tends top, i &.” oF “f(X) approaches bas x approaches p.” We sometimes indicate this by writing /(x) + & ax “The definition conveys the intuitive idea that f(x) can be made arbitrarily close to b by taking x suficiently close to p. (See Fig. 4.1.) We require that p be fn accumulation point of 4 to make certain that there will be points x in A suficiently close 10 p, with x # p. However, p need not be in the domain off, fand b need not be in the ange off Figare 4.1 Note, The definition can also be formulated in terms of balls. Thus (1) holds if and only if, for every ball B,(6), there isa ball Bs(p) such that By(p) oA is not ‘empty and such that fis) @ Brld) whenever x € Bp) 0 A, x = p. ‘When formulated this way, the definition is meaningful when p or 6 (or both) are in the extended real number system R¢ or in the extended complex number system (C*_ However, in what follows, it isto be understood that p and b are finite unless it is explicitly stated that they can be infinite “The next theorem relates limits of functions to lis Theorem 4.12. Assume p i an accumulation point of A and assume be T- Then lim fx) = 6, o ts of convergent sequences.% te nd Contioty nan Vand only ¥, lim f(x) = b, o for every sequence {x4} of points in A — {p} which converges 10 p. Proof. UC(2) holds, then for every e > 0 there is a 8 > 0 such that 4d J), 6) N implies d(x, p) < 6. Therefore (4) implies de(fcx), B) <0 for m> N, and hence {f(x} converges to b. Therefore @2) implies (3). ‘To prove the converse we assume that (3) holds and that (2) is false and arrive ‘ata contradiction. 1f (2) is false, then for some e > O and every 5 > 0 there is & xin A (where x may depend on 4) such that O< dix <6 but dr(f, 8) Be. ° Taking 6 = im, n = 1, 2,..., this means there is a corresponding sequence of points {x,} in A’— {p} such that 0 < dylty p) < Hn but dfx), Be ‘Cleary, this sequence {x} converges top but the sequence {Lx} does not con- verge 10 b contradicting (3). NoTE. Theorems 4.12 and 42 together show that a function cannot have two ifferent limits as x + p. 46 LIMITS OF COMPLEX-VALUED FUNCTIONS Let (S, d) be a metric space, let A be a subset of S, and consider two complex- ‘valued functions f and g defined on A, FAs gAse. ‘The sum f + g is defined to be the function whose value at each point x of A is ‘the complex number fx) + g(x). The diference f — 9, the product f-q, and the ‘quotient fig are similarly defined. 1t is understood that the quotient is defined only at those points x for which g(x) 0. ‘The usual cules for calculating with limits are given in the next theorem, Theorem 4.13. Let {and g be complex-calued functions defined on a subset A of @ ‘metric space (S, d). Let p be an accumulation point of A, and assume that Jim fx) = a, litt ax) = b. mae ‘Lr of Veco aloe Poneto n Then we also have: 2) Him, [/@) £ ao] = @ 2 4, ») limp /Uaot2) = ab, ©) limp FOa(2) = afb Yb # 0. Proof. We prove (b), leaving the other parts as exercises. Given with < ¢ < 1, ete’ be a second number satisfying 0 < e” < 1, which will be made to depend on tina way to be described later. There sad > Osuch that ix @ 4 and atx, p) < 8, then (fa) al 2), but we can define the sum f + g, the product if (if 4 is real) and the inner product {-g by the respec- tive formulas (6+ BE) = 1G) +269, AL) = Af), (EG) = fla)-Ht2) for each x in A. We then have the following roles for calculating with limits of vector-valued functions. Theorem 4.14. Let p be an accumulation point of A and assume that fim g(x) = b. lim fa) = a, Then we also have: 3) Tim, (x) + aia] = a+b, 1) lim, fx) = a for 2» sealar 2, ©) tim, 2) R(3) = ab, 4) Time. MDL = Hal.” Limits and Cony Deas Proof. We prove only parts (e) and (d). To prove (6) we write fix)-gix) — a°b = [fls) a}-{ata) — b) + aa) ~ ¥] + befits ~ a} ‘The triangle inequality and the Cauehy-Schware inequality give us 0 iRay-ats) ~ 9b) 5 HMC) ~ al nC) — bi + Fal fet) ~ Bl + BH EMG - at Each term on the right tends to 0 as x ~+ p, so flx)*g(x) -+ arb. This proves (©. To prove (A) note that jf(x)| ~ Hadj < Hx) — al NOTE. Let fi... fg be m real-valued functions defined on 4, and let f: 4+ RY ‘be the vector-valued function defined by the equation £2) (OAD), A) ifxe A. ‘Then fi... fy ate called the components of f, and we also write f = (fi...) (o denote this Felationship. Ifa = (a,,...,a), then for each r = 1,2,...,a we have (Wed ~ ail < Bla) — all s Do ifta) - a, ‘These inequalities show that lim, fix) = a if, and only if, lim,., f(s) = a, for each r. 48 CONTINUOUS PUNCTIONS ‘The definition of continuity presented in elementary calculus can be extended to functions from one metric space to another, Definition 4.15. Let (S. ds) and (T, d;) be metric spaces and let f:S-+T be a Suction fram $ to T. The function f 1s said 10 be comtinious at & point pin Sif for every t > O there is a> 0 such that 4(F).AP) < IFL8 continuous at every polnt of a subset A of S, we say fls continuous on A whenever dx, p) < 6, This definition reflects the intuitive idea that points close to p are mapped by {into points close to fip). It can also be stated jn terms of balls: A function fis continuous at p if and only if, for every # > 0, there is a3 > O such that MB4p:5) © BAAD: 9. Here Bs(p3.5) is a ball in S; its image under f must be contained in the ball By(s(p)i 8) in T. (See Fig. 4.2.) 1 p isan accumulation point ofS, the definition of continuity implies that lim 00) = f¢9). penetra eA m7 Contiity of Composite Fanci » Band) Beso Tage of Bai. 3) ie Figure 42 fp isan isolated point of S (a point of S which is not an accumulation point of '), then every f defined at p will be continuous at p because for sufficiently small 5 there is only one x satisfying d(x, p) < 8, namely x = p, and dy(7{p), f(2)) = 0. Theorem 4.16, Let f:S + Te a function from one metric space (S, ds) to another GF, dp), and assume p € S. Then is continuous at pif, and ony if for every sequence {54} th S convergent top, the sequence (f(x,)} nT converges to ftp); in symbols, Bn so) = Ji) ‘The proof of this theorem is similar to that of Theorem 412 and is left as an cnercte forthe reacer. (The result can also be deduced fom 4.12 but ther is & ‘minor complication in the argument du tothe act tha some terms ofthe seauence {a4} could be equal tof.) “The theorem is often described by seying that for continuous functions the limit symbol canbe interchanged with the fenction symbol. Some cae is needed in interchanging these symbols because sometimes {f(49) converges when (x) siveres Baample x, — and yy in metic space (5.4), then dls) + dz, 9) (Exercite 4.7). "The reader can verily that discontinuous on the metric space (SS, 9). where p is the metric of Exercise 3.37 with 5, = 5; = S. Nore, Continuity ofa function fata point pis called a focal property of f because it depends onthe behavior of fonly inthe immediate vicinity ofp. A property of {ohich conceras the whole domain offs called agloal property. Thus, continvity Off on ts domain is a global property. 49 CONTINUITY OF COMPOSITE FUNCTIONS Theorem 4.17. Let (S, ds), (T, dy), and (U, d,) be metric spaces. Let f:S + T and 9 =f(S) + U be factions, and let h be the compasite function defined on S by the equation mx) = affix) for xin. IF is continuous at p and if g is continuous at fip), then his continuous at p.» ‘Limits and Cotimaity mais Proof. Letb = fiz). Given & > 0, there is a5 > Osuch that Alats,9() < whenever dy(y, B) < 6. For this 5 there is a 6° such that (fC), 9) < 8 whenever dels, p) < 5 Combining these two statements and taking » = f(x), we find that a.lha), Mp) < © whenever delx, p) < 8, $0 His continuous at p. 410 CONTINOUS COMPLEX-VALUED AND VECTOR-VALUED FUNCTIONS Theorem 4.18. Let f and g be complex-valued factions continuous at a point p i 4a meri space (Sd). Then f ¥ 9, f ~ a. and f-g are each cominuous at p. ‘The ‘quotient fg is abo continuous at p if gp) + 0. Proof, The result i trivial if pis an isolated point of S. If p isan accumulation point of S, we obtain the result from Theorem 4.13, ‘There is, of course, a corresponding theorem for vector-valued functions, which is proved in the same way, using Theorem 4.14, Theorem 419. Let $ and be functions continuous at a point p in a metric space (5, d), and assume that { and g have values in RY. Thon each of the following is continuous at p: the sum £ + g, the product if for every real i, the inner product fog, and the norm (fh Theorem 4.20. Let f,,..., f, be n reabealued functions defined om a subset A of @ meric space (S, ds), ad let f = (f,,.... Jy). Then £ is continuous at a point p of A if ond only if each of the functions f,,., f,i8 continuous at p. Proof. If pis an isolated point of 4 there is nothing to prove. Ip is an accumula on point, we note that fx) Hn) asx = pif amd only is) - foreach 4.11 EXAMPLES OF CONTINUOUS FUNCTIONS. Let S = C, the complex plane. Its a trivial exercise to show that the following complex-valued functions are continuous on C: 8) constant functions, defined by f(z) = for every z in C; ») the identity function defined by f(z) = = for every z in C. Repeated application of Theorem 4.18 establishes the continuity of every poly nomial: Fie) = ay + aye boast Hoot a, the o, being compe numbers. man Cootiny and Inverse Tages of Open oe less Sts If $ is a subset on C on which the polynomial f does not vanish, then Iif is ‘continuous on $. Therefore a rational function gif, where g and fare polynomials, is continuous at those points of C at which the denominator does not vanish ‘The famiiar al-valued functions of elementary calculus, such as the ex: ‘ponential, trigonon etric, and logarithmic functions, are all continuous wherever they are defined. The continuity ofthese elementary functions justifies the common practice of evaluating certain limits by substituting the limiting value of the “independent variable”; for example, lim et =o = 1. ‘The continuity of the complex exponential and trigonometric functions is @ ‘consequence of the continuity of the corresponding real-valued functions and Theorem 4.20. 4.12 CONTINUITY AND INVERSE IMAGES OF OPEN OR CLOSED SETS ‘The concept of inverse image can be used to give two important global descriptions of continuous functions. 421 Definition of inverse image. Let f ST be a function from a set S10 a Set T. IP Y is a subset of T, the inverse image of ¥ under f, denoted by f-*(Y), és defined to Be the largest subset of S wick f maps into Y: that is, PMY) = ferxeS and fire Y). ove, Hffhas an inverse function /~*, the inverse image of ¥ under fis the same the image of ¥ under f~1, and in this case there is no ambiguity in the notation FHL), Note also that 4) = f-B)iPA S BST. Theorem 4.22. Let f: 8+ T be a fiewtion from Sto T. If XG Sand YET, then we have: 8) X= F741) implies OX) = ¥. b) ¥ = f(X) implies X = f~Y). ‘The proof of Theorem 4.22 is a direct translation of the definition ofthe sym- bole f-1(Y) and (UX), and is let to the reader, It should be observed that: in general, we cannot conclude that ¥ = UX) implies ¥ = /~¥(Y), (See the example in Fig. 43.)2 ‘Li and Continty man [Note that the statements in Theorem 4,22 can also be expressed as follows: SUH XS SLU). [Note also that {4A v B) = f-*(4) Uf-4(B) forall subsets A and 2 of 7. Theorem 4.23. Let f 1S» T bea function from one metric space (S, ds) to another (Z, d)). Then fis continuous on 5 if, and only if, for every open set Y in T, the inoerse image J-"(Y) s open in S. Proof. Let {be continuous on $, let ¥ be open in 7, and let p be any point of £-4(D). We will prove that p isan interior point of f""(¥). Let y = f(g). Since Yis open we have By(y; 6) © ¥for some e > 0. Since f is continuous at p, there is a8 > Osuch that (Bx(9; 8)) & By(¥: 6). Hence, BAP: ESM BAP: HN} SSBC; 9} =F", 40 p isan interioe point of 401), Conversely, assume that /~1(Y) is open in $ for every open subset Y in 7. ‘Choose pin S and let y = fp). We will prove that fis continuous at p. For every > 0, the ball By(y; 6) is open in T, 50 f7"(B(y; 0) is open im S. Now, Pef~1(By(y 8) s0 there is a8 > O-such that By(p: 5) < f~"(Byly 6). There- fore, (Bx(p; 8) = By( ye) 50,/s continuous a p. ‘Theorem 4.24, Let f:S + T be a function from one metre space (S, d;) 1 another (1, dy). Then § i continuous on S if, and only if, for every closed set Y in T, the inverse image f"(¥) is closed in S, Proof. If Y is closed in T, then T — Y'is open in Tand SAT N= s- FM, Now apply Theorem 423. ‘Examples. The image of an open set under 2 continuous mapping is not necessarily open. A simple counterexample is a constant function which maps al of $ onto a single point in R?. Similarly, he image ofa closed st under a continuous mapping need not be closed. Foc example the eatnived function () = aan maps R onto the Open Iva 413 FUNCTIONS CONTINUOUS ON COMPACT SETS ‘The next theorem shows that the continuous image of a compact set is compact. ‘This is another global property of continuous functions. Theorem 425. Let: T be a function from one metric space (S, ds) 10 another (7, dy). If Fis continuous on a compact subset X of S, then the image f(X) is a compact subset of T; in particular, (X) is coed and bounded in T. nao ‘Fectiat Coatiwoat on Compact Sets s Proof. Let Fbe an open covering of f(X), $0 that f(X) S User 4. We will show that finite number of the sts 4 cover OX). Since fis continuous om the metric subspace (X, de) we can apply Theorem 4.23 to conclude that each set £~1(4) is copen in (X, dy). The sets f*(4) form an open covering of X and, since X is compact, a finite number of them cover X, say XE f-WA) v=" U SHA, Hence FO) SIU MAYO MAN) = SUF MAN) 9 STF *ADY Sau so f(X)is compact. Asa corollary of Theorem 3.38, we sce that /(X) is closed and bounded. Definition 4.26, A function £: S—+ R* is called bounded on S if there is @ positive umber M such that f(x) 0. For every ¢ > Othere is a.5 > Ouch that IQ = 6 < fla) < fl) +e whenever xe BE; 8) 0S. ‘Take the 8 corresponding to ¢ = f{C)/2 (this e is positive). Then we have A) < J) < fle) whenever xe Ble; 8) > S, 10 f{) has the same sign as f(6) in Ble; 8) - S. ‘The proof is similar if f(e) < 0, except that we take ¢ = 4/6). Theorem 4.52 Bolzano). Let f be real-valued and continuous on a compact intereal [a, 6] in R, and suppose that f(0) and f(b) have opposite signs; that is, assume Hf) <0. Then there is atleast one point cin the open interval (a,b) such that Fo = 0. Proof. For definiteness, assume f(a) > 0 and f(b) < 0, Let A= (e:xe[e0} and fa) 20). “Then A i nonempty since @ ¢ 4, and A ix hounded above hy b. Let 6 = sup A. ‘Then a < ¢ 0, there are points x > ¢ at which flx) > 0, contradicting the definition of e. Ifflc) < 0, then ¢ ~ 4/2 8 an upper bound for A, again contradicting the definition of ¢. Therefore we must have fle) = 0. From Bolzano’s theorem we can easily deduce the intermediate value theorem for continuous functions. Theorem 433. Assume f is real-valued and continuous on a compact interoal S in R. Suppose there are two points a < Bin S suck that fla) (B). Then f takes ccery abe between fe) and f(B) inthe interoal (2, 6). Proof. Let k be a number between fiz) and /(8) and apply Bolzano's theorem to the function g defined on (3, 6] by the equation gtx) = fx) ~ k. “The intermediate value theorem, together with Theorem 4.28, implies that the continuous image of a compact interval S under a real-valued function is another ‘compact interval, namely, Link 18), sup f(5)). fis constant on S, this will be a degenerate interval) The next section extends this property to the more general setting of metric spaces.“ ‘Ln tot Continaty D4 436 CONNECTEDNESS ‘This section describes the concept of connectedness and its relation to continuity. Definition 434. A metric space S is called disconnected if S$ = A B, where A and B are disjoint nonempty open sets in S. We call S connected if itis not die Nome A subset X of a metric space S is called connected if, when regarded as & ‘metric subspace of S, itis a connected mettic space. Fxamgles “The metric space S = R ~ {0} with the vsual Euclidean metrics disconnected, since its the union of two disjoint nonempty open set, the positive real numbers and the negative eal numbers, 2 very open interval in Ris connected. This was proved in Section 34 as a conse- quence of Theorem 3.11, 3. The st Q of rational numbers, regarded as a metric subspace of Euclidean space R, is disconnected, In fact, Q ~ AUB, wheee A consists ofall rational numbers < Vand Bof al rational numbers > V2. Similarly, every ballin Q is disconnected, ‘4 Every metric space $ contains nonempty connected subsets. In fact, for each pin S the set (p} i connected. Zo relate connectedness with continuity we introder the concept of» two-valued function. Definition 435. 4 real-oalued function {whi said to be two-valued on S if f(S) = (0, 1). In other words, a two-valued function is a continous function whose only possible values are Qand 1. ‘This can be regarded as a continuous function from S to the metric space T = {0, 1}, where T has the discrete metric. We recall that every subset of a discrete metric space Tis both open and closed im T. is continuous on a metric space Sis Theorem 436 A meric space S i comected If, and oly I, cery twos function on $ is constant. ae Proof. Assume $ is connected and let fbe a two-valued function on S. We must show that f is constant. Let A = f-1({0}) and B = f~¥((1}) be the inverse images of the subsets {0} and (1). “Since {0} and (1) are open subsets of the discrete metric space {0, 1}, both A and B are open in S. Hence, $= AU B, where A and B are disjoint open sets. But since Sis connected, either 4 isempty and B = S, or ese Bis empty and A = S. In either case, fis constant on S. Conversely, assume that S is discoanected, so that § = 4 u B, where A and Bare disjoin! nonempty open subsets of S. We will exhibit a two-valued function ‘on S which is not constant. Let nf fred, mont toes nas Compras os Metre Space " Since 4 and B are nonempty, f takes both values 0 and 1, s0 fi not constant. ‘Also, fit continuous on S because the inverse image of every open subsct of (0,1) is openin 5, “Next we show thatthe continuous image of connected st i connected. Theorem 437. Let {15 -» M be a function from a metric space $ to another metric space M. Let X be a connected subset of S. If 8 continucus om X, then GO) 18.0 comected subset of M. Proof, Let g bea twowalved function on /(X). We will show that gis constant. ‘Consider the composite function h defined on X by the equation h(x) = of). Then his continuous on X and can only take the values 0 and 1, so his two-valued com, Since ¥ is connected, hs constant on X and this implies that g is constant on fUX). Therefore (2) is connected, Example. Since an iterval X in Ris connected, every continous image ected, 1 has ral valu, the image /(X) is aotber interval. Iffbas vals in RY, the Image LX) felled a care in RTs, every curve in Ris connected. ‘Asa corollary of Theorem 4.37 we have the following extension of Bolrano's theorem. Theorem 438 (Intermedite-vale theorem for real contiawous functions). Let f be real-valued and continuous on a connected subset Sof R*. Iff takes on two diferent aus nS, say o-and b, then foreach real e between a and b there exsse pont x in S such that f) = €. Proof. The image /(5) is a connected subset of R!, Hence, /(S) is an interval containing «and b sce Exercise 438). If some value c between a and 6 were not in f(S), thes f(S) would be disconnected. 417 COMPONENTS OF A METRIC SPACE “Tis section shows that every metric space Scan be expressed in a unique way as 4 union of connected “pieces” called components. Fist we prove the following: Theorem 439. Let F be a collection of connected subsets of @ metre space S such ‘thatthe intersection T= (Yau 18 mot empty. Then the union U = ay A is connected Proof, Since T# 0, here is some tin T. Let f be a two-valued function on U. ‘We will show that fis constant on U by showing that tx) ~ /() for all x in U. It xe U, then xe 4 for some A in F. Siace A is connected, fis constant on A and, since £6 4, (2) = f(0). [Every point x in a metric space S belongs to atleast one connected subset of ‘5, namely’ (x). By Theorem 439, the union of all the connected subsets which ‘contain xis also connected. We call this union a component of S, and we denote it by UG). Thus, U(2) ithe maximal connected subset of S which contains x.a ‘Lint a Continty mae Pm yr of ai ac mig er Component ofS. nor word the components ofS forma election of do sets whose union is S. . = ” ‘Prof Two dint component canot contin pint there (oy Theorem ‘Spyies woud be rps conarted we conaiing 438 ARCWISE CONNECTEDNESS “This section describes 8 special property, called arewise comectedness, which is possessed by some Gat not al) connected sets in Euclidean space R= Definition 441: A set Sin RY ix called arcwie connected if for any oo points & adbinS theresa comin faclon te Sache et 1-2 ond Kad rors, Such a funtion is called path from a to b. 1° (0) # {(), the image of [0,1] under ¥ is called an orc joining a and b. Ts, 3% acwise counected i every par of distinct point in can be joined by an are lying in S.Arcwise connected sets are alo called potwise comecied. FM) = tb + (1 ~ 1a for O's Fs 1, the carve joining and bs called line segment emis 1. Eyer comes set in in arciconneicd, sae tein semen nog two point cofsuch act hein the st Tn parr, every al breve comes. 2. the seta Fi. 44 (anion of to tangent com dss) arise connected Figure 44 3. The st in Fig. 45 consists of those points on the curve described by y = sin (I/n), 0 < 1 1, along with the points onthe horizontal segment ~1 = <0. This ct is connected but not urewise connected (Exercise 446) 7 Figure 45 ef. 40s Arenae Comectdness » ‘The next theorem relates arcwise connectedness with connectedness. Theorem 4.42, Every arcise connected set § in R° is connected. Proof. Let g be two-valued on S, We will prove that g is constant on S, Choose a point «in S. Ifxe S, join ato x by an arc lying in S. Since T is connected, is constant on so g(x) = 9(4). But since xis an arbitrary point ofS, this shows that g is constant on S, 50 Sis connected. We have already noted that there are connected sets which are not arcwise ‘connected. However, the coucepts are equivalent for open sets. Theorem 4.43. Every open connected set in R* is arewise connected. Proof. Let S be an open connected set in R" and assume x ¢ S. We will show that x can be joined to every point y in Sby an arc lying in S. Let A denote that subset Of S which can be so joined to x, and let B= S~ 4. Then S = AU 8, where “A and 8 are disjoint, We will show that A and B are both open in R°. “Assume that a 4 and join a to x by an arc, say [lying in S. Since ae ‘and Sis open, there is an n-ball Bla) & S. Every y in Bia) can be joined to a by f line segment (in S) and thence to x by T. Thus ye if ye Bia). That is, Bla) & A, and hence A is open. "To see that Bis also open, assume that be B. Then thereis an n-ball BY) © S, since Sis open. But if @ pointy in B(b) could be joined wo x by an arc, say I", Sying in S, the point b itself could also be so joined by frst joining B to y (by a Sine segment in B(W)) and then using’. But since b ¢ A, no point of B(b) can be in A. That is, B(b) © B, so Bis open. ‘Therefore we have a decomposition S= AU B, where A and if are disjoint ‘open sets in RY. Moreover, A is not empty since x ¢ A. Since S is connected, it follows that B must be emply, so S= A. Now 4 is clearly arcwise connected, because any two ofits points can be suitably joined by first joining each of them {0 1X. Therefore, S is arcwise connected and the proof is complete ove. A path f:[0, 1] + $ is said to be polygonal ifthe image of (0, 1 under £ is the union ofa finite number of Fine segments. The same argument used to prove ‘Theorem 4.43 also shows that every open connected set in R° is polyyonally con- nected. "That is, every pait of points in the set can be joined by a polygonal arc tying im the set. Theorem 444. Every open set S in R* can be expressed in one and only one way as a ‘countable disjoint union of open connected sets. Proof. By Theorem 440, the components of S form a collection of disjoint sets ‘whose union is S, Each component T of S is open, because if x € T then there is ‘an m-ball B(x) contained in S. Since B(x) is connected, B(x) © T, so T is open, By the Lindelof theorem (Theorem 3.28), the components of S form a countable collection, and by Theorem 4.40 the decomposition into components is unique. Definition 448, A set in R* is called region if tts the union of an open connected tet with some, none, o alts Boundary points. If nome of the boundary points are* {nits and Cootaiey Det 46 ‘included, the region is called an open region, If all the boundary points are included, the region is called a closed region, Note. Some authors use the term domain instead of open region, especialy in the complex plane. UNIFORM CONTINUTTY Suppose fis defined on metric space (S, d), with values.in another metric space (7, dy), and assume that fis continuous on a subset 4 of S. Then, given any point Pin A and any £ > 0, there is a 6 > 0 (depending on p and on ¢) such that, if xe A, then AA), (D)) < whenever delx, 9) < 5, In gencral we cannot expect that fora fixed ¢ the same value of 8 will serve equally well for every point p in A. “This might happen, however. When it does, the function is called uniformly continuous on A. Definition 4.46. Leif: S -» Tbe a function from one metric space (S, ds) to another (Fg) Thon sao be nfo conta nt mbes of 5 oon condition holds: aad —— and p A then : 4F0),AP)) <¢ whenever dlp) < 6. o To emphasize the dterence between continuity on 4 aud uniform continuity ‘we consider the folowing esaunpes of reab-valued functions. Examples 1. Let (0) = Ips for x > @ and take = (, 1]. This function is continuous on A tut not uniform continuous on A. To prove this, lt « = 10, and suppose we could find a,0 < 5 < 1, to satisfy thecoodition ofthe definition, Takingx = 8, p = 3]t1, ‘we obtain [x ~ pl '< 3 and _0 $876 enc, for thse two points we would always have (72) ~ fp) > 10, contradicting the deition of norm sontcay. YO 10) 2 Let fle) = 2° ifxeR! and take A = (0,1] 25 above. This function is uniformly continuows on 4. To prove ti, oberve that Fes) S19) = bet = w= Me ~ kx + DL < Bx ~ a It fx ~ pl < 8, then L7cx) ~ f(p)] < 26, Hence, i i given, we need only take 4 = «(2 to guarantee that [/(2) ~ flp)| < « for every pai x, p with Ir — pl < d. ‘This shows that fis uniformly continuous on (Fo) ~ 09) > 10. man Uniform Contimity and Compact Sts ” ‘An instructive exercise isto show that the function in Example 2 is not uni {ormly continuous on R 420 UNIFORM CONTINUITY AND COMPACT SETS Uniform continuity on a set A implies continuity on A. (The reader should verily this) The converse is also true if 4 is compact. Theorem 4.47 (Heine). Let f :S + T be a function from one metric space (S, ds) to another (T,d;). Let A be a compact subset of S and assume that fés continuous on A. Then is uniformly continuous on A. Proof. Let 6 > be given. Then each point a in A has associated with ita ball ‘Ba; r), with» depending on a, such that Aff), flO) < a whenever x € Ba; 7) 0 A. ‘Consider the collection of balls By(a; 1/2) each with radius 1/2. These cover 4 and, since 4 is compact, a finite number of them also cover A, sty ac aa: ) In any ball of twice the radius, B(a we have 4f00.sl00) < whenever Bian) 9 Let 8 be the smallest ofthe numbers 1/2, .-.4fa/2. We shall show that this 5 works in the definition of uniform continuity. or this purpose, consider two points of A, say x and p with dy, p) <6. By the above diseusion there is some ball Bay; r/2) containing x, so 49.09) <5 [By the triangle inequality we have 49.0) 5 4.9) + dfssa) S, 80 we also have dr{/(p), f(a) < 2/2. Using the triangle inequality once more we find 40). AD) 5 dy(F0), flap) + deffo). JCP) < F + 5 = 6 ‘This completes the proof.a ‘Limits a Cootinty mas 421 FIXED-POINT THEOREM FOR CONTRACTIONS, Let f:.8 — 5 be a function from a metric space (S, d) into itself, A point p in ‘Sis called a fixed poinr of fif f\p) = p. ‘The function fis called a contraction of ‘Sif there is & positive number a < 1 (called « contraction constant), such that A{ft2), f)) < dex, ») for all x yin, o Clearly, « contraction of any mettc space is uniformly continuous on S. Theorem 4.48 (Ficed-point theorem). A contraction f of @ complete metric space S has a unique fixed point p. Proof. It p and p’ are two fixed points, (7) implies d(p, p') < dip, p'), $0 s0 dip, ') ~ Oand p = p’. Hence f has at most one fixed point : To prove it has one, take any point x in and consider the sequence of iteraten: % fe Meo). ‘That is, define a sequence (p,} inductively as follows Po Pest SP, 9 = O12 ‘We will prove that {pz} converges 10a fixed point of First we show that (p,} is 4 Cauchy sequence. From (7) we have - Ares s5 Pd) = MEDIA Py) & AA Pm Path 0, by induction, we find Pass PS AD, Po) = Oo, where ¢ = d¢p, po). Using the triangle inequality we find, for m > 1, dw 2) Ee Ayu nd se See = LEE ctw, Since «* ~ as n -+ co, this inequality shows that (p,} is a Cauchy sequence, But ‘Sis complete so there isa point p in S such that p, —» p. By continuity off, Ao) = (im) = fn Sr 50 p isa fined point off. This completes the proof. Many important existence theorems in analysis are easy analysis ‘consequences of the fixed point theorem. Examples are given in Exercises 7.36 and 7.37. Reference 44 gives applications to numerical analysis. 422 DISCONTINUITIES OF REAL-VALUED FUNCTIONS ‘The rest of this chapter is devoted to special properties of real-valued functions defined on subintervals of R, Det 408 scott of Res Valued Functions 33 Let f be defined on an interval (a, 6). Assume ce (a, 6). IF f(x) -+ A as 2x +e through values greater than ¢, we say that 4 isthe righthand limit of f at € ‘and we indicate this by writ im foe) “The righthand limit 4 ial denoted by /(e+). tn thee, 5 terminology this means that for every € > O there is a 3 > Osuch that Ve) — fet O there is a 5 > 0 such that eb ¥. The only alternative is and this means that f“* i titty increasing. “Theorem 4.52, together with ‘Theorem 4.29, now gives us Theorem 4.53. Let f be siridly increasing and continuous on a compact interval [a,b Then fis continuous and strictly increasing on the intereal {fd f0)) NOTE Theorem 4.53 tells us that a continuous, strictly increasing function is a ‘topological mapping. Conversely, every topological mapping of an interval [2,6] ‘onto an interval [c, d] must be a strictly monotonic function. ‘The verification of this fact wil be an instructive exercise for the seader (Exercise 4.62).Limits of sequences 4.1 Prove each of the following statements about sequences in C. a 60 fle] < 15 (2) diverges if [zl > 1 ) IF z, + Oand if {c,} is bounded, then {e243 +0, (€) s4{n! + 0_ for every complex z. ) Ia, = Vi? +2 ~ nm, then a, +0, Az econ Cas + anf For alin & Lshow that ay + 2038. Mit tun 2 ~ tans Hay ~ a4 43 10 < xy <1 and if tye, = 1 — V7 — x for all m = 1, prove that (x) is a decreasing sequence with limit 0, Prove also that X,44/x,~+ 4 4.4 Two sequences of positive integers {2,) and (b,) are defined recursively by taking 4 = by = Land equating rational and irrational pacts in the equation y+ V2 = ogay + bye 2F form 2 2, Prove that a3 — 23. = 1 for n > 2. Deduce that a4, + V2 through values > V2, that 25,Ja, ~+ 2 through values < V2. 45 A reat sequence () satisfies 7xy4) = x2 + Gform > 1. If-xy = 4, prove that the seaence inerenrer and find ie limit, "What happens if; = } or ify ~ 4? 46 lagi <2 and \dny2 ~ Ones! * Slade ~ a for all n> 1, prove that fo,) ‘converges. 4.7 In a metric space (S, 4), assume that x, > x and yy» y. Prove that dry 33) de 9. 43 Prove that in a compact metric space (S, 2), every sequence in S has a subsequence which converges in S. This property also implies thal S is compact but you aresnot re- quired to prove this. (For a proof see either Reference 4.2 or 4.3.) 49 Let d bea subset ofa metric space $. If Ais complete, prove that A is closed, Prove thatthe converse also holds if is complete Limits of fnetons Nore, In Exercises 4.10 through 4.28, all functions are real-valued. 4.10 Let /be defined on an open interval (a) and assume x € (2,8). Consider the wo statements 2) im fe + 8) = Sea) = 05 by im [fur + A foe ~ = 0. Prove that (a) always implies (band give an example in which (b) holds but (a) des not. 411 Let fbe defined on R.A and ifthe one-dimensional limits ny f(x 9) and lin fx, 9) both exist, prove that im lim ts. 9M = Le Ain i Fes, 9} xerces ” [Now consider the functions f defined 00 R? 25 follows: ofan =, if Ge, 7) # (0,0), (0,0) = 0 ay wr? ofan if # 0,03,/00,00= FEM aes aE ©) fx, ») = Lin ory) Mx # 0,40, 9) = ¥. x + Yosin sin Up) iE # Oand y #0, ofan= lo itr towyee fin x ~ sin y tans 4 tan, orn zai om ittan = tan y. In each of the preceding exumples, determine whether the following limits exist and fe thowe lis that do exist im LiF 99} 5 in i x, 15 FD 4.12 If 6 [0,1] prove that the Following limit exists, ln (li cs?" (mt x2) and that its valve is 0 or 3, according to whether x 8 trational or rattonal Continuity of rea-valoed functions 443 Let f be continuous on fa, 6] and let f(x) = 0 when x is rational. Prove that Hla) = O for every xin f, 6) ‘414 Let f be continuous at the point a = (ayy a3... .204) ia RE Keep ty yo. dy Fixed and define a new function g of one rel variable by the equation 0) = $05, 35-004 09. Prove that ¢ is continuous at the point x ~ a. (This is sometimes stated as follows: ‘A continuous function of sariabes is continuous tn each variable separately.) ‘AIS Show by an example that the converse ofthe statement in Exereise 4.14 isnot true in general 436 bath df be defined on [0,1] 2 follows: “7be) = obs) Hx) = 0, whenever is irational; JG) = Land g(x) = x, whenever i rational; 1x) = Un if isthe rational number mya (in lowest terms); HO) = Prove that fis not continuous anywhere in [0,1] that gis continuous only at x = 0, aed ‘that fs continuows only atthe ieational points in. (, 1} “x7 For cach x i [0, 1} let f(x) = x if x @ rational, and let f(x) = 4 — x if x is inrational. Prove that 2) fUfte) = x forall xia [0,12 0) 00) + £0 ~ a) = 1 forall xin [0,1]es Limit ad Contaicy ©) is continuous only atthe point x = 4, 1) Fsssumes every value between 0 and 1 Mle + 9) ~ F0) ~ FY rational for al x and y in [0,11 4.18 Let /be defined on R and assure tha there exist atleast one point in at which ‘Fiscontinuous. Suppose also that, for every x snd y tn R, satiahes the equation Sle + 9 = S00 fOr Prove that there exists a constant a such that f(s) = ax for all. 4.19 Let /be continuous on fa, bJand define gas follows: gf) ~ fla)and, fora < x 0. Prove tha there isan mball Ap r) such that Cz) > O for every x inthe ball, 422 Let /be defined and continuous ona closed set Sin R, Let A= (rxeS and fia) = 0), Prove that i a closed subset of R 423 Given a function f:R ~ R, define two sets A and Bin R? as follows: A= ny < fon, (univ > fo rove that ff continuows on Rit, and only f, both 4 and Bate open subsets of R2 4424 Let /te defined and bounded on a compact interwal Sin R. If T= 5, the number 247) = 0 Vo) — fo): xe Hy eT) {is called the ascilarion (or span) of fon T. If + € , he oxilation of fat + is defined to be the number ods) 8m 21BEiN 4 8). Prove that this limit always exists and that a2) = Of and only if, fis continuous at x ‘425 Let f be comiinvous on u compact interval [2,6]. Suppose that fhas a local max: imum atx, and a local maximum atx Show that there aust be a thitd point between x, and.cs where fhas a local minimath, ware, ‘To say that fas # local maximum at x, means that there isa {all B¢x,) such that fx) fs) forall xin Bx,) [a,b]. Local minimum is simulaly defined. 426 Let fea real-valued function, continuous on {0.1}, with the following property For every realy, either there is no x in [0,1] for which f(x) = y or there Is exactly one such +. Prove that fs strictly monotonic on (0,1). 427 Let /be a function defined on (0, 1] with the following property: For every reat number» ether there is no x in 0, 1] for which f(x) = yor there are exactly wo values of xin (0, 1] for which fx) = y. eas » 4) Prove that feannot be continaous on (0, 1) 1) Construct a function f which hus the abowe property ‘6, rove that any function with this property has infiaitely many discontinuities on fo} 428 In each cate, give an example of a function 41S) = T, oF else explain why there can be no such, continuous on S and such that )S= 00, Tr DS= OD, Tr OS= RY r OS Wi B3, 7 OSH O1}x Ih 7 HS |B} x 1, T= Ox OD. DSH=ONXGD, THR. ‘Coctimaty in etre spaces In Exercises 4.29 through 4.33, we assume that f: S'-+ Tis a function from one metric space (S,d) 10 another (T, dy). 429 Prove that fs continuous on $ if, and only i, F-Mint B) Sim 74D) for every miter B of 7. 430 Prove thats continuous on Sif, and only if, Ay STA) for every subset A of S. 4s continuous on $ if, and only if, Fs continuous on every compact Subset of S. Bint. Hhxg + pin S, the set {p, x1, Xp...) iscompact. ia closed in 7 432 A function f: § + Tia called a closed mapping on Sif the image /(4) is closed Tor every closed subset A of §. Prove that fis cootiquous and closed on $f, and only i, A) = F(A for every subset A of S. 433 Give an example of 2 continuous fand a Cauchy sequence (4, in some metic space S for which {f(r} is not a Cauchy sequence in 7: 434 Prove that the interval (—1, 1) ia Ris homeomorphic 10 R'. This shows that neither boundedness noc completeness sa topological property, 4.35 Section 9.7 contains an example of a fonction f, continuous on {0.1} with JK, 1) = {0.1} * [B, 1]. Prove that no such fan be one-to-one on [0, 1} Comectoaness 436 Prove that a metric space Sis disconnected if, and ony if, there fs nonempty subset (4of S, A # S which is both open and closed in S. {437 Prove that « metric space Sis connected if, and only if, the only subsets of S which ‘are both open und closed in $are tbe empty set and S itself. 4.38 Prove thatthe anly connected subsets of Rare (a) the empty et, () sets consisting ‘of a single point, and (2) intervals (open, closed, hal-open, oF infinite),100 ‘Lint a Comtity 429 Let X be a connected subset of a metric space S. Let Y be a subuet of $ such that XS YS X.where isthe closure of X. Prove that ¥ is ako connected. In particular, this shows that is connected. 4.40 If is « point in a metric space S, let U(x) be the component of S containing x Prove that U(x) is closed in. 441 LetS be an open subset of R. By Theorem 3.11, Sis the union ofa countable dis. jin collection of open intervals in R. Prove tha each of these open interval isa com nent of the meric subspace S. Explain why this does not contradict Exercise 4.40. 442 Given a compact set in R™ withthe following property: For every par of points 1 and b in $ and for every # > 0 there exists © fnte set of poinis (p.Xi,---s it With p = @and x, = b such that In - Prove or disprove: Sis connected. 4463 Prove that a metric space S is connected if, and only if, every nonempty proper subset of Shas « nonempty boundary. 4.44 Prove tha every convex subset of RY is connected, 445 Given a function f: RS +R which is one-to-one and continuous on B®, If ‘pen and disconnected in R*, prove that f(4) is open and dlscenneted in £087). 4M Ltd = (Ox :0 << 1, y=sintix), Be Gx,):y=0, -1 sx <0), and lt S'= 41 Prove that Sie connasted but not arcvise counested. (See Fs 4h Section 4.18) 407 Let F = (Fi, Fa... } be a countable collection of connected compact seis in R° such that Fas SF; (or each & 1. Prove thatthe intention (i. is connected and closed. ‘448 Let 5 be an open connected set in RY. Let Tbe a component of Rt — 5. Prove that RY — Tis connected, 449 Let (S, d) be connected mesic space which is not bounded. Prove that for every in Sand every ¢ > 0, the set (x: d(x, ) ~ 1) is nonempty Pee fork = tem Uniform cootnaity 450 Prove tha a function which is uniformly continuous on $ is aso continuous on S. 451 IF/(0) = 2 for x im R, prove that fis vot uniformly continuous on BR. 4.52 Assume that fs uniformly continuous on a bounded set Sin RY. Prove be bounded on S, . 453 Let be a function defined on a set Sin RY and assume that {(5) SR Let g be defined on £5) with value in RS, and lt b denote the compenite function defined by ‘nex) = a{f(2)]ifx« S. If Fisunifoemly cominuous on S and fis uniformly continuous (on £(5), show that b i uniformly continuous on 5. 454 Assume f: $+ Tis uniformly continuous on S, where S and T are metric spaces. 1 x) is any Canchy sequence nS, prove that (f(4,)} Cauchy sequence in. (Com: pare with Exercise 4.33) frst Kerio. 0 5+ Tbe « function fom a mete ace 50 another netic ace T ‘sone (sania onion ter Ao Sanda T compte. Prove Dat thre ungue encnon oF /10 4 whi som conics on 456 Inn ei spe (Sd), let A be a none wb! of $. Defies fneson fa S > R by the equation 14a) = 8 Hs 276A) foreach nS. The numer (2 eal the san fom «A {Prova j,i wsifermy continous 00S ©) Prow that = (ese ant 146) ~ Ob = ona ee 457 Ina merc apce (St A a Be int closed sess ofS, Proves Sin dijout open ats Vand ¥ of Seah at a'© U and Bs PH Lat If) ~ fs, he ootation of Exeee 436 and coer gM 0) and 90, +00). Discontosites 4.58 Locate and classify the discontinuities of the functions f defined on R! bythe follow. ing eavations: 2) fl) = (in ax ite # 0,f0) = 0. b fa = ix ¥0,f0 ~ 0. Df) =e + sin in) if #0 AO) ~ 0. Ofoy= 1K KOFO=0 4.59 Locate the points in R? at which each ofthe functions in Exercise 4.1 is not con Male tne 40 Lt be didn he open intel 1) and asa tha for enh ior on ott nse sea BO) ch / 8 nering Prove ha sa eI incon roughest 8) . conincu n compact inal 648 sine ta domo ae ical imum ors lal inna at any intro piace The ove folowing rece 425) Deve tat mt be monotone on b {4611/1 one-one an continuo bl, prove tha mat be sialy monte Safi) Tht wove ht eer opll apeng tf ot an eral fe 4] sma be say monotone 1A Let foe an retingfcton defined on (ob nd ey. Be oi Teimcrornchdate Coy Sarees fee 2) show hat Fh Vee) = fl 9) 570-9 ~ fe) 3) Dodie tom prt tha th of dacs of oonta, ©) Prove that f has points of continuity in every open subinterval of [a, 6) 1464 Give an example ofa function f, defined and strictly increasing on ase Sin R, such that J~" isnot continuous on f(S),102 Limits and Cotizaty 4.65 Let fbe strictly increasing on a subset $ of R. Assume that the image /(S) has one fine ftowing reer: (0S) open; ()/(S) isonet (3) /) lee Prove that f must be continuous on S. Metric spaces and fixed pints 4.66 Let B(S) denote the set ofall real-valued functions which are defined and bounded ‘on a nonempty set 5. Ife BOS), let EL = sup en. The number 1/| is called the “sup norm” off 8) Prove thatthe formula 4,0) = f° ~ a defines u metic d on B(S). ) Prove that the metric space (AS). d) is complete. Hint. If (/;} is a Cauchy Sequence in B(S), show that (62) isa Cauchy sequence of real numbers for each x in S. 4.67 Refer to Exercise 4.66 and let C(S) denote the subset of B(S) consisting of all func tions continous and bounded on 5, where now S's 2 metric space 8) Prove that C(S) isa closed subset of B(S). ') Prove that the metric subspace C(S) i complete ‘4.68 Refer tothe proof of the fixed point theorem (Theorem 4.48) for notation. 8) Prove that dip.) = a(x, Feo)ePIU ~ a). ‘This inequality, which ie weil in numerical work, provides on com p, to the fied point p. An example is given in (). ) Take f(x) = 4(x + 2/2), $ = [1, +20). Prove that /is a contraction of S with contraction constant « = j and fixed point p = V2. Form the sequence (p,) starting with x = pp = Vand show that |p, ~ 2} < 2- 469 Show by counterexamples that the fixed-point theorem for contractions need not hold if either (a) the underlying metric space is not complete, ot (b) the contraction ‘constant « 2 1. 470 Let: 5 ~ be a function from a complete metric space (S,d) ito itself, Assume ‘there isa real sequence (ay) which converges to 0 such that d/*x), *)) << eylts, 9) forall = 1 and al x iS, where "is the mth iterate off that is, PG) =F), FG) = LD) orn = 1. Prove that has a unique fixed point. iat. Apply the fixed-point theorem to f™ for a sate m. AN La f:5~ Stew imate for the distance notion from a metric space (S,d) into itself such that APSO) < Aly) whenever + # y. 2) Prove that Fhus at most one fixed point, and give an example of uch an fwith no fixed point, by IF S is compact, prove that f has exactly one fined point. Hint. Show that ‘) ~ d(x.) attains its minimum on 5: {Give am example with $ compact in which fs not a contraction, Rerecse 103 472 Assume that f satsies the condition in Exercise 471. I xe S, ket po = x, Pert ® Pd, 208 €y = Ay Pros) fOr n = 0. 1) Prove that {c} isa decreasing sequence, and let ¢ = lim cy by Assume there is x subsequence (pus) which converges 10 8 point @ in S. Prove “ € = dla, = d(fa) FF). Dedvce that gis. xd point of fand that 7,» SUGGESTED REFERENCES FOR FURTHER STUDY 4.1 Boas, R. Pa A Primer of Real Functions. Carus Monograph No, 13. Wiley, New York, 1960. 42 Gleason, A., Fundamentals of Abstract Analysis, Addison-Wesley, Reading, 1966. 43 Simmons, G. F Iniroduction 10 Topology and Modern Anaiysis. MeGiraw-Hil, ‘New York, 1963, 4A Todd, J, Seroey of Numerical Analyst. McGraw-Hill, New York, 1962CHAPTER 5 DERIVATIVES 54 INTRODUCTION ‘This chapter treats the derivative, the central concept of differential calculus. Two fiffcrent types of probiem—the physical problem of finding the instantaneous velocity of a moving particle, and the geometrical problem of finding the tangent line to a curve at a given point—both lead quite naturally to the notion of deriva- tive, Here, we shall not be concerned with applications to mechanics and geometry, ‘but instead will confine our study to general properties of derivatives. ‘This chapter deais primarity with derivatives of functions of one real variable, specifically, real-valued functions defined on intervals in R. It also discusses briefly derivatives of vector-valued functions of one real variable, and partial derivatives, since these topics involve no new ideas. Much of this material should ‘be familiar to the reader from elementary calculus, A more detailed treatment of erivative theory for functions of several variables involves significant changes and is dealt with in Chapter 12. ‘The last part of the chapter discusses derivatives of complex-valued functions ‘of a complex variable. 52 DEFINITION OF DERIVATIVE If fis defined on an open interval (a, 5), then for two distinet points x and ¢ in (4, 6) we can form the difference quotient Sos) ~ Heo) We keep c fixed and study the behavior of this quotient a -+«. Definition 5.1. Let f be defined on an open interval (a, b) and assume that ¢ € (a,b) Then fs said tobe differentiable at © whenever the limit fx) = fo) cexlsts, The limit, denoted by Jc) is called the derisaive of fa ‘This limit process defines a new function f°, whose domain consists of those points in (a, 6) at which /'is differentiable. "The Function f” is called the first 108 masa Derivative od Contnaity es derivative of f. Similarly, the nih derivative off, denoted by #™, is defined to be the first derivative of f°, for m= 2,3,.... (By our definition, we do not ‘consider #© unless °° is defined on an open interval) Other notations with ‘which the reader may be familiar are ly where y = fC), Zl, Umneey = Jo so = ofa = 4 (= or similar notations. The function f itself is sometimes written /*". The process ‘which produces /" from fis called differentiation. 3. DERIVATIVES AND CONTINUITY “The next theorem makes it possible to reduce some of the theorems on derivatives 10 theorems on continuity. Theorem 5.2. If f is defined on (a,b) and differentiable at a point ¢ in (a,b), then there isa function f* (depending on f and on ¢) which is continuous at ¢ and which satisfies the equation a Sea) = fle) = be — AF", o for alt x in (a, b), with f*(c) = fc). Conversely, if there is a function f*, con ima hich sts) ten Ps iene ond) = Proof. WF (6 exists, letf* be defined on (a, #) a follows: Py FOAM tree, Mo=Ke. ‘Then f* is continuous at ¢ and (1) holds forall x in (a,b). Conversely, if (1) holds for some /* contiquous atc, then by dividing by x ~ and letling x —> ¢ we ste that fe) exists and equals /*(c). ‘As an immediate consequence of (1) we obtain: Theorem 5,3. If fis differentiable atc, then fis continuous at c. Proof. Let x + ein (N), wore. Equation (1) fas a geometric interpretation which helps us gai insight ots meaning, Since /¥ continuous at f°) i nearly equal 0 /*(e) = f°) wis mere. Replacing /*() by JC it (L) we obain the equation Pe = KO) + PKe ~ — cis smal. In other words, if fis “which should be approximately correct when x — cis small. In other words, ifferentiable atc, then fis approximately a linear funetion near ¢. (See Fig. 5.1) Differential calculus continually exploits this geometric property of functions.