MARKETING Ri WILL INVITE LEGAL ACT! NarosaCONTENTS Preface to the Second Edition v Preface to the First Edition vi List of Symbols xi Scheme of Numbering/Cross-Referencing xi |. PRELIMINARIES 1 11 Sets 1 1.2 Functions 3 1.3 Relations 4 1.4 Sequences 5 1.5 Countable and Uncountable Sets 5 1.6 Inequalities 6 2. INTRODUCTORY CONCEPTS 9 2.1 Definition and Examples of Metric Spaces 2.2 Open Spheres and Closed Spheres 16 2.3 Neighbourhoods 19 2.4 Open Sers 20 2.5 Equivalent Metrics 25 2.6 Interior Points 28 2.7 Closed Sets 30 2.8 Limit Points and Isolated Points 33 -29 Closure of a Set 35 2.10 Boundary Points 37 2.11 Distance between Sets and Diameter of a Set 38 y 2.12 Subspace of a Metric Space 40 2.13 Product Metric Spaces 43 2.14 Bases 44 3. COMPLETENESS a 73.1 Convergent Sequences 47 3.2 Cauchy Sequences 52 3.3 Complete Spaces 55 v7 34 Dense Sets and Separable Spaces 62 3.5 Nowhere Dense Sets 68 2xvi__Metric Spaces 3.6 Baire’s Category Theorem 70 3.7 Completion 74 4. CONTINUOUS FUNCTIONS yl 4:1 Definition and Characterisations 79 4.2 Extension Theorem 85 4.3 Uniform Continuity 86 74-4 Homeomorphism 90 4.5 Uniformly Equivalent Metrics 92 5. COMPACTNESS 5.1 Compact Spaces and Sets 95 5.2 Sequential Compactness 98 5.3 Totally Boundedness 102 5.4 Equivalence of Compactness and Sequential Compactness log | 5.5 Compactness and Finite Intersection Property 110 il 5-6 Continuous Functions and Compact Spaces 112 6. CONNECTEDNESS 4.1 Separated Sets 115 \ «A-2 Disconnected and Connected Sets 117 \ | 76.3 Components 122 j 6-4 Connected Subsets of R124 \ 6.5 Continuous Functions and Connected Sets 125 | By 7. FIXED POINT THEOREMS AND 47.1 Fixed Point Theorems 127 7.2 Application to Integral Equations 132 7.3 Application to Implicit Function Theorem i and Inverse Function Theorem 135 APPENDIX A: THE CANTOR SET. THEIRAPPLICATIONS — 127 18 | APPENDIX B: LIMITS IN METRIC SPACES 4 \ BIBLIOGRAPHY ur \ INDEX 482 INTRODUCTORY CONCEPTS In analysis, we are concerned mainly with two elementary concepts: (i) convergent sequences in IK (real or complex numbers) and (ii) continuous functions with domains and ranges in K, We note that each of these notions depend precisely on the concept of the absolute value |x ~ xo! of the difference between the numbers x and x5 in K. Many of the properties of convergent sequences and continuous functions depend only on the properties of this distance, and not directly on the algebraic nature of the real (or complex) number systems. We write d(x. y) =|x—y |, the distance between x and y. The following properties of the distance are well known: @ Ix-ylz0 Gi) Ix-yl=0ex= Gi) Lx-yl=ly—al (iv) le-ylslx-zltlz-yl We want to generalize the concept of distance by taking any non-empty set (instead of R or C) in such a way that we can generalize the notions of convergence of sequences and continuity of functions (real and complex) to a ‘more general situation. 2.1 DEFINITION AND EXAMPLES OF METRIC SPACES 2.1.1 Definition Let X be a non-empty set. A function a@XXXR is said to be a metric on X if it satisfies the following conditions @ dxyy20, VnyeX dxy=0ex=y, myeX (it) diy y)=dy,x), Vx ye X (symmetry) (iv) d(x y) Sd(x,z)+dlz.y), VuyceX (triangle inequality) Pr eres eee10 Metric Spaces The ordered pair (X, d) is called a metric space. If there is no confusion likely tg Occur we, sometimes, denote the metric space (X, d) by X. Note When we say that d is a metric on X or (X, d) is a metric space, it jg understood throughout that X is a non-empty set. Remarks 1. The triangle inequality may be interpreted as that “the length of one side of a triangle can not exceed the sum of the lengths of the other two sides” Equivalently, the distance from x to y via any intermediate point z can not be shorter than the diiect distance from x to y. Zz d(x,2) a(z.y) d(x y) Fig.2.1 » 2 The triangle inequality can be generalised for any number of additional POINS 2, 2.04% inX; Le., d(x, y) S$ d(x, z,) + (z), 23) + (Zp, Y) 2.1.2 Examples 1. Let X= R, the set of all real numbers. For x, y € X, define (x,y) =1x-yl ‘Then (X, d) is a metric space. This is called t usual metric and we denote it by R,. 2 LetX= C, the set ofall complex numbers. For x, ye X, de d(x,y) =1x-yI Then (X, d) is a metric space, This is called usual metric and we denote it by C, the metric space R with the fine the metric space C with the Note If there is no confusion likel (resp., C) in place of R, (resp., C,). 3, LAX be an arbitrary non-empty Set. For x, ye X, define d by “ 0, x= 'y to occur, we may simply write Then (X, d) is a mettic space. The metric the space (X, d) is called discrete Xp dis called the discrete metric and metric space and is denoted bYIntroductory Concepts II 4, Let X = Q, the set of all rational numbers. For x, y € X, define a (x,y) =Ix-yl Then (X, d) is a metric space. 5, Let X=[0, If. Forx, y € X, define - (x,y) =Ix-yl Then (X, d) is a metric space, 6 Let X =, the set of all points in the coordinate plane. For x = (x1, x;) and ‘y= Qj, ¥2) in X, define 3 8G) FG, y) = max (1xy-y,h 1x2-y! } Gi) d"(@, y) = 1-1 + lay — yy! Then each of the spaces (X, d), (X, d°) and (X, d™) is a metric space. 7. Let X = R®, For x = (x, x)) and y = (y}, y2) in X, define ax, = ils ney Jai[+|n [+] — 92) tye Then (X, d) is a metric space. 8 Let X=R", the set of all ordered n-tuples of real numbers. For x= (a, 2. wm. @,) and y = (B;, By, ..., By) in X, define dlx,y) = & (a, - rf! ial Then (X, d) is a metric space. In view of Minkowski’s Inequality 1.6.6, for p = 2, we note that atx.) -{ (a, -ay} (a; -n| +d(z,y), where z = (9j, 7, «-» %) is in X. This verifies the triangle inequality. The other conditions for d to be a metric on R" are quite easy to check. 9. LetX=C", the set of all ordered n-tuples of complex numbers. Fora = (a, @y, ..., ,) and y = (By, Bos + By) in X, define12 Metric Spaces 1 ' 2 d(s.y) = is le; -a'} ‘Then (X, d) is a metric space. Note The metric space in Example 8 is called the Euclidean Nesp; e and is denoted by R" while the mettic space in Example 9 is called the unitary n-space (or Complex Euclidean n-space) and is denoted by. C". 10. Let X= K" (R" or £"), Define @ dy) = max 1a; 6 len @ May=Y 10,-B)1 a ‘Then, each of the spaces (X, d") and (X, d’) is a metric space. 1. LetX=K"(R" or ©"), Define 1 4,(x.y) = & le, a} ss p eT 2.2 OPEN SPHERES AND CLOSED SPHERES 2.2.1 Definition Let (X, d) be a metric space. Let xe X and r > 0 be a real number. The open shpere with centre x and radius r, denoted by S,(x), is the subset of X given by S,(x) = (ye X: d(x, y) 0. Remark . Every open shpere in the usual metric space BR, is an open interval. But the converse is not true; for instance, J~ ©, e[ is an ope? interval in R but not an open sphere. Fig. 2.4Introductory Concepts_17 the usual metric space C, the open sphere $,(cy) is the circular disc Iz-zgl0. 3, Let xy be any point in the discrete metric space X, (Example 2.1.2(3)). Then spug = feok erst 4, Inthe metric space in Example 2.1.2(5), 510) = ls sl 140) = fOi, roa 5, In the metric space ( the origin is given by ) of Example 2.1.2(6), the unit sphere centred at S(O.) = {04.49 2x7} +5 < 1} n3 Fig. 2.5 \S 4 6A the metric space (R?, d*) of Example 2.1.2(6), the unit sphere centred at the origin is given by S\((0,0)) = (Oty) Lay < Lb h< 1} pe _ % (1,0) Fig. 2.618 Metric Spaces 7. In the metric space (R?,d") of Example 2.1.2(6), the unit sphere centred at the origin is given by S,(,0)) = {(x,x,) :Lyl + lag h< Lb Fig. 2.7 8. In the metric space Cla, b] (Example 2.1.2(15)) the open sphere 5,(x) Consists of all functions x in Cla, b] whose graphs lie within the shaded band of vertical width 2r centred on the graph of x. 2.2.3 Definition Let (X, d) be a metric space. Let xe X and r>0, The closed sphere with centre x and radius r, denoted by 5,{x], is the subset of sphere with cent X given by SX) = ly eX: dl, y) 0. 3. In the discrete metric space X, (Example 2. 1.2(3)), the closed sphere S,l9] is given by +0 1 5,L%] = {os at Remarks In the metric space X,, observe that () For0 1, So) = S)L%ol =X Silty) = (Xo), Slo] =X 2.3. NEIGHBOURHOODS 2.3.1 Definition Let (X, d) be a metric space and x € X. A set Nc X is said to be a neighbourhood (nbd) of x if 3 an open sphere centred at x and contained in N, i.e, if $,(2) CN, for some r> 0. 2.3.2 Examples 1. The open interval Ja, b[ is a nbd of each of its points. 2. The set R of real numbers is a nbd of each of its points. 3. The closed interval [a, b] is a nbd of each point of Ja, b{ but is not a nbd of the end points a and b. 4. The set N, Z or Q is not a nbd of any of its points. 5, Inadiscrete metric space X,, a subset YC X is a nbd of each of its points. 2.3.3 Theorem Let (X, d) be a metric space. A set N.C X is a nbd of a point p € X if and only if Jan open sphere 5,(x) such that p € S,(x) CN. Proof First we assume that Vis a nbd of a point p © X. Then, 3 an > 0 such that p € S,(p) CN. This proves that 3 an open sphere S,(p) containing p and contained in N.20_ Metric Spaces Conversely, assume that p € Nand 3 an open sphere 5,(x) such that p © §,¢) 0. Let y € S,,(p). Then d(y, p) d(x, p) + d(y, p) MONEN, (b) Ne Nand MoN=>MeEN, Proof (@) We have MN Ny = 3 ry ry >0 such that S,.(x) CM and 5,,(x) ¢N = SQ) CM and Sx) CN, where r= min{r,, 75} => SQ) CMAN = MONisanbd of x = MANEN: (b) We have NeN,=> Janr>0 such that S,(2) CN > SG) cM (2 MDN) > MeN.O 2.4 OPEN SETS 2.4.1 Definition Let (X, d) be a metric space. A set G of each of its points, (Equivalently, a set G CX is sai O such that 5,(x) © G). © Xis said to be an open ser if it is a nbd id to be an open set if for each xe G,3anr> 2.4.2 Examples 1, In the usual metric space R, (a) Ris an open set.Introductory Concepts _21 (©) 10, I is an open set (©) (0, If is not an open set. (d) The sets N, Z and Q are not open sets, (©) The set of all irrational numbers is not an open set. ial © {nd + is not an open set (@ (x), x R, is not an open set. (h) The Cantor set C is not an open set. 2. In the metric space of Example 2.1.2(5), (0, @{, @< 1, is an open set. 3, In the metric space of Example 2.1.2(4), the set (a) Qis an open set. ) {h 4. In the discrete metric space Xj, every set G c X is an open set. In particular, every singleton set in X, is open. 2.4.3 Theorem Let (X, d) be a metric space. Then, the empty set @ and the whole space X are open sets. Proof In order to prove that g is an open set, we need to verify that each point of @ is the centre of some open sphere contained in g, But 9 contains no point, Therefore, the requirement is automatically satistied For the second part, let x € X be arbitrary. Then, 3 an open sphere S,(x) X. This is possible since any open sphere centred on a point of X cannot go beyond X. Thus X is a nbd of x. But x € X is arbitrary. Therefore X is a nbd of eaeh of its points. Hence X is an open set. 0 2.44 Theorem Let (X, d) be & metric space. Then, each open sphere in X is an open set. Proof Let S,%p) = (4 © X: dlx, Xo) 0. Consider S00) = (8 € Xd yo) <7} Let y€ §, (yo) be arbitrary. Then d(y, yo) <7. Now AC. ¥) S$ dXoy Yo) + de Y). (by triangle inequality) << dly, Yo) +722_Metric Spaces > y € S,(%) Consequently S00) © S00) ; ‘ This proves that 5,(x) is a nbd of yp. But yy € S,(%) is nebiay. Therefore, $,(%) isa nbd of each of its points. Hence S,(x9) is an open set. SY 1 < dX, Yo) 11 > Xo, Yo) (a) (e) Fig.2.9 Remark Converse of the above theorem need not be true, For instance, the interval }~s, co[ in the usual metric space R,, is an open set while it is not open: sphere. 2.4.5 THEOREM Let (X, d) be a metric space and Gc X. Then, G is an open set if and only if itis the union of open sphere: Proof Let G be an open set. Then, each point of G is the centre of some open sphere which is contained in G. Clearly, union of all such open spheres is precisely the set G. Conversely, let G be the union of ‘open spheres. Let % be the family of these open spheres. Let x € G be arbitrary. Then, x belongs to some open sphere, say, Silt) €%. Since every open sphere is an open set, x is the centre of an open sphere S,,(x) such that $100) € S,054) But Silay) CG Therefore SQ) cG This proves that G is a nbd of x. But. € G is arbitrary, Therefore G is a nbd of each of its points. Hence G is an open set, OIntroductory Concepts _ 23 2.4.6 Theorem Let (X. d) be a metric space. Then: (a) Arbitrary union of open sets in X is open. (b) Finite intersection of open sets in X is open. Proof (a) Let (Ga) gen be a family of open'sets in X. We shall prove that _U Gyis aca open. Since each Gg, is open, it is union of open spheres for each ae A. ‘Then, Gq is the union of unions of open spheres. Hence, by Theorem it is open. (b) Let (Gj: i= 1, 2....,) be the finite family of open sets in X. We shall prove that 9G, is open. Let x € G, be arbitrary. Then x€G; foreach i=1,2,.47 = Ban r,>0 such that S, (x) Gy (= Ly em) +: Each G; is open) Write Then SX) CS, QS (621, 2M) > 5) AG, il Hence 7) G; is an open set. 0 a Remark Arbitrary intersection of open sets need not be open. 2.4.7. Example In the usual metric space R,, consider the family {} 4 4 ne n} of open Yin e x} {0} which is not an open set. sets. Then 0 { : n en sets are among the most important concepts in geometry and ow often new ideas are defined in terms of open sets d theorems which give necessary and sufficient s of open sets. Furthermore, the basic Remark Op: analysis. One may note hi and there are results ani conditions for these new ideas in term: foundation of topology lies on the concept of open sets.24 Metric Spaces 2.4.8 THEOREM —— . ‘on of Every non-empty open set in the usual metric space R, is the union of a countable disjoint class of open intervals. - iT 1s Proof Let G be anon-empty open set inR,. Letx € G. she ison eae centre of a bounded open interval contained in G. Let J, cs eve intervals which contain x and is contained in G. We note the ; (@ 1, is an open interval which contains x such that 1, ¢ G. (If is anon i A Gi,isan empty class of intervals on the real line such that > 1, # then 41; interval.) . ined io 6 Gi) £, contains every open interval which contains x and is contained in G. Gi) Ifye Z,, then I, = I. : i (iv) Ifx and y are two distinct points in G, then either f, = 1, or f, 0.1, = 0: for if there is a point z €/, 0 /,, then zel, and = 1,31, and = Lek Now, consider the class /of all distinct sets of the form J, for points x € G. Clearly, Vis a disjoint class of open intervals such that Gzul ley It remains now to show that -¥ is countable. Let G, be the set of all rational points in G. Clearly, G, is non-empty. Define f: G,—> Y by fx) = Iiixe G,), The function fis well defined since for each rational xin G, , there is a unique J, in c¥ {the disjoint intervals in eeannot contain the same rational). Also fis surjective since each interval in ¥ contains a rational in G,. Since the set G; is countable. the family cY is countable (the range of a countable set is countable), L 2.4.9 Theorem Eel (X. a) be a metric space. Then, for every pair of distinct points ¥ and yin, 3 disjoint open sets U and V such that xe U and ye V. Proof | Let (X, d) be a metric space and let x, y be two distinct points of X, Since 44 ¥ Uy) > 0. Let dts, ») =3r, U = §,(4) and V = $( fhen.ve Wand ye V, Clearly. UW V= ¢. not, let p © UAV. Then p € U and Pe V. Therefore, PEU = Axper and PEV => Uypyer Now Ay) Sx, p) + dly, p) 0 such that . x€ S.Q@)cG and as such (i) holds. Conversely, if (i) holds and G ¢ c% then for each x € G, 5 d’-open sphere 5,.(x) such that . xe SQ)CG => e Gev Sa) cG => G= US) x66 ‘Thus G is an open set in the metric space (X, d°) showing there by that GeF’. Hence Fo¥'.O 2.5.3. Corollary ‘The metrics d and d” on a non-empty set X are equivalent if and only if @ for each d-open sphere $,(x), 3 an r” > 0 such that S,4(x) & 5,(a), and Gi for each d’-open sphere Sj.(x"), 3 an r > O such that S,(x°) ¢ S,4 x"). Proof The result follows if we establish that (i) and (ii) in Theorem 2.5.2 are equivalent, respectively, to (i)’ and (ii)’ of Corollary 2.5.3. It is enough to show that (i) € (i) since the other one can be obtained similarly by interchanging the roles of d and d’. ‘The assertion (i) => (i)’ follows trivially since an open sphere is an open set in a metric space. In order to establish the reverse implication (i)’ => (i), let us consider a G € % If x € G, then 3 an r > 0 such that S,(x) < G. Therefore, by (', 3 an 7° > 0 such that SQ) CS) 0 such that Ayd(x, y) Sd", y) $ kqd(x, y), @ for all x, y € X, then the metrics d and d" on X are equivalent. Proof Let r>0 be given. Take r” = k;'r. Then, (1), for each x €X, we have Sfx) CS,A(x) by using the first inequality inIntroductory Concepts _27 which, in fact, is Condition (ii)’ in Corollary 2.5.3. Similarly, by using the second inequality in (1), one can establish condition (i)’ in Corollary 2.5.3. Hence. the result follows. O Notes 1, Theorem 2.5.4 is quite useful for establishing certain metrics on a set to be equivalent. 2. The condition in Theorem 2.5.4 is sufficient for two metrics on a set to be equivalent. However, this condition is not necessary. 3, We shall discuss more about the above condition in Chapter 4. 2.5.5 Examples 1. The three metrics defined on R? given in Example 2.1.2(6), €.8., as.y) = Yler= yi) #62 = 92 fey) dy) are equivalent. Note that ax { 1xy-yyhbay—yol) xy-yy bt Lay yal d'(ny) $day) < V2 d'y) Vanye R? and d'(ay) $d") $2d', 9), Vanye R ‘Thus, the metrics d, d’ and d’ on R? are equivalent in view of Theorem 254. 2. The metrics defined on R" as in Examples 2.1.2(8), 2.1.2(10), €.8 dix,y) = {s (a, - os} d'(x,y) = max 10-8)! ten dey) = > lon - Bil are equivalent. — The equivalence of the metrics d, d" and d* on R" follows immediately in view of the following inequalities: doy) Sdiny) S Vandy) Vanye RY and d'(uy) Sd y) Snd' ys Vx ye RY 3, The usual metric and the discrete metric on B are not equivalent. Indeed, the set (0} is not open in the usual metric space R,, while it is open in the discrete metric space Ry.