Scribd
Upload
509 views505 pages

Complex Analysis

a good book to learn complex analysis

Uploaded by

followrabbit
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
509 views505 pages

Complex Analysis

a good book to learn complex analysis

Uploaded by

followrabbit
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
Graduate Texts in Mathematics Serge Lang Complex Analysis Fourth Edition | g) Springer Graduate Texts in Mathematics 103 Eahtril Board [Link] EW Gebring KA. Ribet Springer BOOKS OF RELATED INTEREST BY SERGE LANG ‘Short Caleuus 2002, ISBN 0-387-95327.2 (Calculus of Several Variables, Third Baition 1987, ISBN 0-387-96405.3, Undergraduate Analysis, Second Edition 1996, ISBN 0.387-94841-8 Introduction to Linear Algebra 1997, ISBN 0387.96205.0, ‘Math Talks for Undergraduates 1999, ISBN 0-387-98749-5, (Omir Books By Lano Pustisten By SPRINGER-VERLAG ‘Math! Encounters with High School Students» The Beauty of Doing Mathematics + Geomety A High School Course» Basie Mathemstis» Sort Calculus «A First (Course in Calculus = Inteduction to Linear Algebra + Calculus of Several Variables» Linear Algebra « Undergraduate Analysis + Undergraduate Algebra * Complex. Analysis » Math Talks for Undergraduates + Algebra + Real and Functional Analysis Introduction to Diflerentiable Marifols » Fundamentals of Differential Geometry» Algebraic Number Theory « Cyelotorie Fields Tan I= Tetodution © Diophantine Approximations» SL;(R) » Spherical Inversion on SL,(R) (With Jay Jorgenson)» Elipic Functions + Ellipse Curves: Diophantne ‘Analysis + Insodeecon to Arakelov Theory * Riemana-Roch Algebra. (with Wiliam Futon) = Abelian Varieties » Inroducion to Algebric snd Abelian Functions + Complex Mulplicason + Inzodueton to Modular Forms» Modular Units (ovih Daniel Kubert) + Insodueton to Complex Hyperbolic Spaces = [Narber Theory Il + Survey on Diophantine Geometry Collected Papers I-V, including the following: Invodveton to Transcendental [Numbers in volume i, Frobenius Distributions in GL2-Extensions (with Hale Troter in volume Il, Topics in Cohomology of Groups in volume IV, Basie Analysis of Regulerizd Series and Products (with Jay Jorgensor) in volume V se Enpicit Formulas for Regularized Products and Series (with Jay Jorgenson) in volume V ‘THE FILE « CHALLENGES. Serge Lang Complex Analysis Fourth Edition With 139 Ulustrations a Springer ‘Serge Lang Department of Mathematics Yale Univers New Haten, CT 06520, Usa aor Board S. Anker FW. Ceding KA Ritet Dathemates Department Mathematics Department Mathematics Department San Francisco Site Eee Hall University of Cafornta a University Universiy of Michigan ‘Berkeley San Franeito,CA 96132 Ann Arbor MLGBIOS” Berkeley, CA 947203840, Usa USA Usa Mathematics Sujet Casifeton 200): 3001 Libra of Cong Cason in Palin Dats Tange 1 “Gopi ania Sere Lane —sth ed Poe Oratute ents abe: 3) cde pti reece on ne Ban ors088 | ype Ironton of comple araies 2. Mathai ietide sere OusinLe 99 ISON o.37 585921 Printed on se pope © 1959 Singer Sciences Busines Mei re ‘iis esse This work ayn Be ante copied in whole in prt without the tite persion othe baer Spring Scene Bunnes Medi ne283 Spr Sue ‘Rew Yr NY 1001 USA} sept for bre xc in commen it oews rly {nah Use conaton witha fom of ioenion songs ad ret cto ay ‘en compu stare, or by tna or Suamarmetlgy nom bow ober etapa orden ‘Toe utin i polenon of wade names traemarhs sve mis ad ater even ‘ibearent ented arse atte taken sr an egresion of epton to he ‘ot tyre sito propriety ih oe76s Foreword ‘The present book is meant as text for @ course on complex analysis at the advanced undergraduate level, of first year graduate level. The fist half, more or less, can be used for a one-emester course addressed to undergraduates. ‘The second half can be used for a second semester, at tither level Somewhat mote material has been included than can’ be covered at leisure in one oF tWo terms, to give opportunities for the Instructor to exerese individual taste, and to lead the course in whatever directions stikes the instructors fancy at the time as well as extra read- ing material for students on their own. A laege number of routine exer- dace are included for the more standard portions, and 2 few harder ‘exercises of striking theoretical interest are also included, but may be ‘omitted in courses addressed to less advanced students Tn some sense, I think the classical German prewar texts were the best (Hurwite-Courant, Knopp, Bieberbach ete) and 1 would recommend to anyone to look through them. More recent texts have emphasized ‘connections with real analysis, which is important, but at the cost of ‘exhibiting succinctly and clearly what is peculiar about complex analysis: the power series expansion, the uniqueness of analytic continuation, and the caloulus of residues. The systematic elementary development of for- mal and convergent power Series was standard fare in the German texs, but only Cartan inthe more recent books, includes this material, which 1 think is quite essential, eg, for diferential equations. T have written a short text, exhibiting these features, making it applicable to 2 wide vari- ety of tastes. “The book essentially decomposes into two parts. “The first por, Chapters I through VII, includes the basic properties ‘of analytic functions, esentally what cannot be let out of say, a one- 1 have no fixed idea about the manner in which Cavchy’s theorem is to be treated. In less advanced classes, of if time is lacking, the usual hand waving about simple closed cures and interiors is not entirely inappropriate. Perhaps beter would be to state precisely the homologi- ‘al version and omit the formal proof. For those who want a more thorough understanding, I include the relevant material. ‘Artin originally had the idea of basing the homology needed for com- plex variables on the winding number. T have included his proof for Caucty’s theorem, extracting, however, a purely topological Iemma of independent interest, not made explicit in Artn’s original Notre Dame notes [Ar 65] oF in Ablfor® book closely following Artin [Ah 66] 1 hhave also included the more recent proof by Dixon, which uses the ‘winding number, but replaces the topological lemma by greater use of elementary properties of analytic functions which can be derived directly {rom the local theorem. The two aspects, homotopy and homology, both ‘enter in an extential fashion for diferent applications of analytic Tunc- tions, and neither is slighted atthe expense of the other ‘Most expositions usally include some of the global geometric ties of analytic maps at an early stage. T chose to make the prefiminsries ‘on complex functions as short as possible to get quickly into the analytic part of complex function theory: power series expansions and Cauchy's {heorem. The advantages of doing this, reaching the heart of the subject rapidly, are obvious. The cost is that certain elementary global geometric considerations are thus omitted from Chapter I, fo instance, to reappear later in connection with analytic isomorphisms (Conformal Mappings, CChapter Vil) and potential theory (Harmonic Functions, Chapter VII), T think itis best for the coherence of the book to have covered in one ‘sweep the basic analytic material before dealing with these more geomet- ric global topics. Since the proof of the general Riemann mapping theo- fem is somewhat more dificult than the study of the specific cases con- sidered in Chapter VI, it has been postponed to the second part ‘The second and third parts of the book, Chapters IX through XVI, eal with further assorted analytic aspects of functions in many direc: tions, which may lead to many other branches of analysis. 1 have em phasized the possibility of defining analytic functions by an integral in- volving a parameter and differentiating under the integral sign. Some lassical functions are given to work out as exercises, but the gamma function is worked out in detail in the tex, a8 a prototype "The chapters in Part tallow considerable flexibility in the order they are covered. For instance, the chapter on analytic continuation, including the Schwarz reflection principle, and/or the proof of the Riemann map- ping theorem could be done right after Chapter VIL, and sill achieve ‘reat coherence. ‘As most of this part is somewhat harder than the frst pat it can easily be omitted from a one-term course addressed to undergraduates. In the same spirit, some of the harder exercises in the frst part have been starred, to make their omission easy ‘Comments on the Third and Fourth Editions I have rewritten some sections and have added a mumber of exercises. 1 have added some material on harmonic functions ad conformal maps, on the Borel theorem and Bore!’s proof of Picard’s theorem, as well 3s DJ Newman's short proof of the prime number theorem, which illustrates ‘many aspects of complex analysis in classical setting. 1 have made more complete the treatment of the gamma and zeta functions. 1 have also fadded an Appendix which covers some topics which T find sufisenly important to have in the book. The fist part of the Appendix recall ‘summation by parts and its application to uniform convergence. The ‘others cover material which isnot usally included in standard texts on ‘complex analysis: diference equations, analytic diferenil equations, fixed points of fractional linear maps (of importance in dynamical systems), Cauchy's formula for C= functions, and Cauehy’s theorem for locally integrable vector feds inthe plane. This material gives addtional insight fn techniques and results applied to more standard topics in the text. Some of them may have been assigned as exercises, and’ hope students will try 10 prove them before looking up the proofs in the Appendix ‘Lam very grateful to several people for pointing out the need for a ‘number of corrections, especally Keith Conrad, Wolfgang Fluch, Alberto Grunbaum, Bert Hochvald, Micha Jastrzeski, José Carlos Santos, Eres . Schlesinger, A. Vijayakumar, Barnet Weinstock, and Sandy Zabel, Finally, I thank Rami Shakarchi for working cut an answer book. New Haven 1998 Sect Lavo Prerequisites ‘We assume that the reader has had (wo years of caleulus, and fas some acquaintance with epsilon-delta techniques. For convenience, we have recalled all the necessary lemmas we need for continuous functions on Compact sets in the plane. Section §1 in the Appendix also provides some background ‘We use what is now standard terminology. A function fS=T is called injective if x # yin S implies f(x) # f(). I is called surjective if for every 2 in T there exists xe such that fle) =2. If f is surjective, then we also say that f maps S onto T. if f is both injective and surjective then we say that f is bijective. ‘Given two functions f, g defined on a set of real numbers containing arbitrarily large numbers and such that g(x) 2 0, we write Seg or fixjagls) for x0 {o mean that there exists 2 number C>0 such that forall x suifcently large, we have: yea s cate. Similarly, if the functions are defined for x near 0, we use the seme symbol < for x-+0 to mean that there exists C > 0 such that L7eo = Cate) x [PREREQUISITES for all x sufciently small (there exists 6 >0 such that if Jx] <5 then 1/01 $ Colx)). Often this relation is also expressed by writing S13) = (40s), Which is read: fly) is big ob of g(, for x-+ 20 or x-+0 a5 the case may be. ‘We use Jo, 6[ to denote the open interval of numbers aexch Similarly, [o, BE denotes the hall-open interval, ete Contents Forewors . Prerequisites PART ONE Basle Theory es 5 ‘Complex Numbers and Functions 4. Defaiton 2 Polar For 3. Complex Valued Functions Limits and Compact Set ‘Compact Sete 55. Complen Diferenitiity {6 The Cvehy-Remann Equations 7. Angles Under Holomorphic Maps 4, Formal Power Seses . 2 Convergent Power Series {Relations Between Formal snd Convergent Seis ‘Sums and Products Quotient: Composition of Series 4 Analy Functions §5. Diferentation of Power Series, a 2 1 a 2 EY a nearssss | The Inverse and Open Mapring Theorems 41. The Local Maximum Modulus Principle. ‘Cauchy's Theorem, Fest Part {1 Holomorphic Functions on Connected Sets ‘Appendix Connectedness, Integrals Over Paths a Local Primitive for 4 Holomorpic Function [Another Description ofthe Integral Along Path ‘The Homotopy Form of Cauchy's Theorem Existence of Global Primitives, Definition of te Logarithm “The Local Cauchy Formula ageree Winding Numbers and Cauchy's Theorem $1 The Winding Number sees cee ceeteeeeeee 2 The Olobal Cauchy Theorem ‘Dixons Proof of Theorem 25 (Cauchy's Formals) £8. Anins Proot ‘Applications of Cauchy's Integral Formula 1 Uniform Limits of Analytic Fun 2. Lauren Series £2. lelated Sigulcitis Removable Singuaries Essent Sngulares 5 (Calculus of Residues 1 The Residue Formula Resides of Diferenas {2 Evaluation of Definite Integrals Fourier Transforms ‘Trigonomeric Integral : 5 Metin Transforms CContormal Mappings 9. Schware Lerma 2 Analytic Automoephisms of the Disc = 1222211 (9. The Upper Half Pane : : i. Other Examples {5 Fractional Linear Tansormations 13 1 18 1a 19 156 156 tet 165 165, 16 168 m 184 1 194 7 199 210 22 20 aI Harmonte Functions oy 1. Defsiton 2a ‘Application: Pespendicaaity 246 ‘Application: Flow Lines 28 . Buamples ma 2 [5 Base Properties of Harmonic Functions 29 H. The Poison Formola m “The Poisson Integral a « Convluion m §5 Construction of Harmonie Functions 26 5. Appendix. Diflrenating Under the Inga Sig 26 PART TWO, ‘Geometric Funetion Theory... es I a Schwarz Retlection a 1. Scare ReBection (by Complex Conjugation) 23 2 Reflection Across Analytic Ares a {Application of Schwa Reiction x3 ‘The Riemann Mapping Theorem M6 1, Statement ofthe Theorem ses 2. Compact Sets in Function Spaces ue {2 Proof ofthe Riemann Mappiag Theorem : an |S. Behavior a the Boundary Bia ‘Anayie Continuation Along Curves Eo 4 Continuation Along 3 Cure 2 2. The Dilogaitim =. 3 3. Application to Pleats Theoret 35 PART THREE Various Analytic Toples a ‘evictions of the Maximum Modulus Principle and Jensen's Formula 339 1. ence’s Formula ery 2 The Piard- Borel Theorem Sas {3 Bounds by the Real Part, Borel-Carsthéodory Theowert ast The Use of Three Cire and the Ent of Small Derivatives 36 "ermite Interpolation Formula 358 {5 Entre Functions with Rational Vales : xo fe The Phrapmen Lindel and Hadamard Theorems TINT 3s Entre and Meromorphie Functions om GL Inte Products 2 eee i sn |B. Weserstrass Produce 36 |B. Functions of Finite Order 3a 'H. Metomoephie Functions, Mitag-Lefee Thee 37 Eipte Functions 31 1. The Liouill Theorems 31 {2 The Weirtrass Fenton 335 3. The Addition Theorem =. St 400 HA. The Sigma and Zeta Functions 03 ‘The Gamma and Zeta Functions. 08 1 The Diferentation Lemma ce 409) 2 The Gamma Function a3 ‘Weierstrass Produt 4B ‘The Gauss Mulpiation Formula (Distibuton Relation) 416 The (Other) Gauss Formula ag The Melin Transform a0 The Siding Formula a Proof of Stings Formula a4 {The Lerch Formula ane a 1H Zeta Functions 35 ‘The Prime Number Theorem 440 4 Basie Analytic Properties ofthe Zeta Funtion “a |B. The Main Lerma and it Application 6 4. Proof ofthe Main Lemma 49 ‘appends 4s 4. Summation by Parts and Non Absolute Convergence. 453 2 Diferene EqUations 457 {3 Analytic Difrental Equations “61 |B. Foted Points of a Fractional Lincar Tvansormation : 465 §5. Cauchy's Formula for C> Functions 4s {6 Cauety’s Theorem for Locally Integrable Veter Fiids an Sf. More on Cauchy-Riemann Loan ography pees cee a Index as PART ONE Basic Theory CHAPTER 1 Complex Numbers and Functions (Ove of the advantages of dealing with the real numbers instead of the rational numbers is that certain equations which do not have any solu- tions in the rational numbers have a solution in real numbers. For instance, x=? is such an equation However, we also know some ‘equations having no solution in real numbers, fr instance x* = — 1, or X'= 2 We define a new kind of number where sich equations have Solutions. The new kind of numbers will be called complex numbers |, §1. DEFINITION ‘The complex numbers are a set of objects which can be added and multiplied, the sum and product of two complex numbers being also a complex number, and satisfy the following conditions 1 Every real number is a complex number, and if a, fare real numbers, then their sum and product as complex numbers are the same as their sum and product as real numbers. 2. There isa complex number denoted by such that # = —1 5. Every complex number can be writen uniquely in the form a + bi where a,b ae real numbers. 4, The ordinary laws of arithmetic concerning addition and multipli- cation are satised. We list these laws: I, 7 are complex numbers, then (of) = a(P), and (+B +yma4 (pen 4 ‘COMPLEX NUMBERS AND FUNCTIONS a8 We have (8+ 7) = af +a, and (B+ yx = B+ 70. We have af = fa and a + p= I is the real number one, then 1a = a is the real number zero, then Ox We have «-+(—Na=0, We shall now draw consequences of these properties With each complex number a+ bi, we asiociate the point (2,8) in the plane. Let a= ay +ayi and P= b; + byi be two complex numbers. Then ane 1 +, + (as + adh Hence addition of complex numbers is carried out “componentwise™ For example, (2+ 3) +(—I +5) = 1+ 8% nog as tieteb) Figue 1 {In multiplying complex numbers, we use the rule = —1 to simplify 8 product and to put It in the form a+ bi. For instance, let «= 2+ 3 and =~ i. Then op 2+ Nt =) = 20 =) +34 a2 +in3aD e243 HF a as4i Let am at bi be a complex number. We define # t0 be a— bi ‘Thus if a= 2+ 34 then @= 2M. The complex number @ is ealed the 6 COMPLEX NUMBERS AND FUNCTIONS oso ‘The real number b is called the imaginary part of a, and denoted by Iva). ‘We define the absolute value of a complex number 4.4, are real) to be + ag (where lal = Jat +a I we think of « as a point in the plane (2,2) then [ais the Fength of the line segment from the origin t0 a In terms of the absolute value, provided « #0. Indeed, we observe that Jal? = a Figue 2 We 1 fag, We note that ] because (24)? = oh, so Jal ead = Jama ‘Theorem 12. The absolute value of a complex number satisfies the {allowing properties. If a B are complex numbers, then Joo = lip, a+ Bl ll + U6 Proof. We have 88 = lap? 1.9) DEFINITION 7 Taking the square root, we conclude that |2|If1= lal, thus proving the frst assertion. As for the second, we have let BP = (e+ EFA (a + BEB) a8 + Pa + oP + BB la + 2Re( a) + |8® because a = J. However, we have 2 Re( 8H) $ 21881 because the real part of a complex number is < its absolute value. Hence e+ OP sal + 21pa1+ 16 Sel? + 21601 + 107® lal + 10 ‘Taking the square r00t yields the second assertion of the theorem. “The inequality et BLSial +101 is called the triangle inequality. 1t also applies to a sum of several terms, Ife, 0% Ate complex numbers then We have ley bet ayl Slad to + le Also observe that for any complex number 2, we have Ia=te Proof? |, §1. EXERCISES 1. Express the following complex numbers in the form +19, where sy ae real numbers @ rea" ® ason—9 a+nie—9 @ ¢-9e-5 f) Osx © Gada fo Wane + 9 ty) Wem 20+ 9 2. Enpress the following complex mumbers in the form x+y, where x are real numbers. 8 COMPLEX NUMBERS AND FUNCTIONS ne warm sy on 2 Leta be a complex number #0. What is the absolute value of of? What ee 4. Leta f be wo complex numbers, Show tha 3B 3B and that ape 5 sty the ssetion made inthe proof of Theorem 12, thatthe real part of complex number is is absolute value. 6 Warm a with ob cel then bi called the imaginary part of 2 and we ‘write b= je) Show that «= 2a}. Show that eo) Heo 7 Find the teal and imaginary pats of (t+ 0. ve that for any two complex numbers 2, we have: (else wi+ IM, © bsl=Iwsie— © blIulsieew 9. Lean at iband 2x + iy: Let e be eal >0. Transform the condition {nto an equation involving only xy, band ¢, and describe in a simple vray what geometric figure i represented by this equation. 10, Deserve geometrically the sts of points » sting the following conitons. @ bates=s, eal s @ nies (@ [eemist (@ Ime>o. © Imzzo @ R50 (hy Rerzo |, §2. POLAR FORM Let (33)= +1) be @ complex number, We know that any point in the plane can be represented by polar coordinates (76). We shall now see how to write our complex number in terms of such polar coordinates Let @ be areal number. We define the expression e® to be of = cos 0 + isin. ‘Ths e# is a complex number. we POLAR FORM ° For example, if Furthermore 2 and ef =i Figure 3 Let x,y be real numbers and x + iy 8 complex number. Let Ve ay. 11 (0) are the polar coordinates ofthe pint (nthe plane, then =reos0 and y=rsing. x4 1y = 7 6050 + isin O = re ‘The expression re! is called the polar form of the complex number xt. The number 6 is sometimes called the angle, or argument of 2 and we write arse “The most important property of this polar form is given in. Theo- rem 21. Tt will allow us to have a very good geometric interpretation for the product of two complex numbers. ‘Theorem 2.1. Let 8, @ be two real numbers. ‘Then Proof. By definition, we have i810) conf + 9) + Fsin(8 + 9) » couruex wonneRs AN FUNCTIONS asa sing headin formulas for sin and cosine, West that the prec Ine cen isc to cos0 on 9 sin Bsn 9 + iin Dens @ + sing 6 ‘This exact the same expeson atthe ove we obtain by malin “ (608 6 + isin @)(cos @ + isin @) (ou theorem is prove Theorem 21 justifes our notation, by showing that the exponential ‘of complex numbers saises the same formal rule as the exponential of real numbers Let a= a, + ia bea complex number. We define e* to be For instance, let a +34 Then of = ee Theorem 22. Let , be complex numbers. Then o. Proof. Let a= a, + ia, and B= by + iby. Then ertF a cevthivte thw eth Using Theorem 2.1, we see that this last expression is equal to erent w eneincbcts ‘By definition, this is equal to ete, thereby proving our theorem. ‘Theorem 22 is very useful in dealing with complex numbers, We shall now consider several examples to illustrate it Example 1. Find 2 complex number whose square is de” Let 2= 2c, Using the rule for exponentials, we see thet 2? fet, Example 2 Let n be a positive integer. Find a complex number w such that w= oe POLAR Font " is clear thatthe complex number w = e™ satisfies our requirement, In other words, we may express Theorem 22 as follows: Let 2 =e and 2, =r,e% be two complex numbers. To find the product 2,2, we multiply the absolute values and add che angles Thus In many eases, this way of visualizing the produet of complex numbers {is more useful than that coming out of the defition, Warning, We have not touched on the logarithm. As in calculus, we want to say that e* = w if and only if 2=log w. Since e2* = {for all integers Ki follows that the iverse function 2 = log w is defined only up to the addition of an integer multiple of 2ri We shall study the loga- rithm more closely in Chapter Il, §3, Chapter I, §5, and Chapter If, §6 |, §2. EXERCISES 1. Pat the following complex numbers in pola form. wre Wry? @ 3 @ a 1-2 ~Si @ -7 wy 2. Put th following complex numbers inthe ordinary form x +i. Qe em] sem Ge) nee a 3 Let be a complex number #0. Show that there are two distinct complex numbers whose square 4, Let a4 bi be @ complex number. Find real numbers x y such that expressing x, yin terms of and bh 5, Plot all the complex numbers = such that 2 forn= 23,4 and 5 6 Leta be a compen number #0. Let m be 2 poiive integer. Show tht there aren dnc complex numbers such that = a Wate Use complex numbers in pol form +1 ind the real and imsinry parts of, taking the fourth root such that ts tng es Between O and 3/2 (a) Desribe all complex numbers» such that = (8) Ler w be a compler number Lat be a complex number such that Gos Dacre al comple numbers such that = on a sect of graph paper R ‘COMPLEX NUMBERS AND FUNCTIONS 9 9. Wet = e% show that there sam integer k such that 2 = w + 2k 10. (@) 18s eal, show that cond tet fer and sind (©) Fer atiteary comples , suppose we defie cos = and sin = by replacing © with =n the above formola. Show thatthe only values of ? for which cor =O and sin 0 are the url rel vales from tgonometr. 11, Prove that for any complex number 2 #1 we have Pat 12 Using the preceding exerci, and taking real parts, prove: 1 ste Ledgers cna Sls bm 1B, Let zw be two complex numbers soch that 2 1. Prowe that Jet straits Hlal= 1 or ot (here are many ways of dong thi One way isa follows. First check that you may assume that £5 Teal, say z= 7 Foethe Fst inequality you are reduced to proving (ole) <1 — rt = Expand both sides and make caneltons to simplify the problem) |, §3. COMPLEX VALUED FUNCTIONS Let $ be a set of complex numbers. An association which to each clement of $ associates a complex number is called complex valued function, or a function for short. We denote such a function by symbols Tike F546, 1.91 COMPLEX VALUED FUNCTIONS B If zis an element of S, we write the association of the value fl) to 2 by the special arrow erste We can write Se) = we) + ite), where u(t) and o(2) are real numbers, and thus srtuleh — aevote) are real valued funetions, We cll u the real part of f; and o the fmagi- rary part of f ‘We shall usually write +m, where x, y are real, Then the values of the function f ean be written in the form fea fo +9) (9) els Yh viewing 4,» as functions of the two real variables x andy. Example, For the function fe) Ay + isingx + yy we have the real part, us ey and the imaginary part, 6, ix + 9. Example. The most important examples of complex functions are the power functions. Letn be a postive integer. Let Se=2 “Then in polar coordinates, we can wite 2 = re®, and therefore fla) = re" = (eos no + isn 96) FFor this function, the real part is r*e08n0, and the imaginary part is’ sind ry COMPLEX NUMBERS AND FUNCTIONS nel Let D be the closed dise of radius 1 centered at the origin in C. In cother words, D isthe set of complex numbers 2 such that [2] $1. It 2 is an element of D, then 2" is also an element of D, and so 2-~2" maps D jntoitsell Let be the sector of complex numbers re such that 050.20, 8 shown on Fig. 4. 2 Fue “he function ofa ral variable snaps the unit interval (0, 1 entitle The function Onno saps the interval (6, 29m) (0,2 In this way, we see thatthe function fle) {ull die ofall numbers 1 maps the sector $ onto the with 0S¢1 and B, “iy such that 1 and OS yom + iy such that 0S y 3 (no condition on xb |, §4. LIMITS AND COMPACT SETS Let a be 2 complex number, By the open dise of radius r > 0 centered at 2 we mean the set of complex numbers such that Fer the dd dis we we te condone ingen. We stat Si eh epee tes ev pet hs spa Sip ote i, deed by De Te ca ae dene by a mi U be a subset of the complex plane. We say that U is open if for coy it tim thee a oes Den) cel tana one fate’ wh tat thr die Dia) cna We hve ised on open nF 18 ‘COMPLEX NUMBERS AND FUNCTIONS os ® Figures [Note that the radius r of the dise depends on the point «As « comes closer to the boundary of U, the radius ofthe disc wil be smaller. Examples of Open Sets, The fst quadtant, consisting ofall numbers “fy with x0 and y > 0 is open, and drawn on Fig. (2) Figure 9 (On the other hand, the set consisting of the first quadrant and the vertical ané horizontal axes as on Fig. 9(6) isnot open. "The upper half plane by definition isthe set of complex numbers pexsiy with y > 0, Iti an open set. Let $ be a subset of the plane. A boundary point of S is @ point & such that every dise D(a,r) centered at a and of radius r > 0 contains both points of $ and points notin S. In the closed first quadrant of Fig. 9(0), the points on the x-axis with x 2 0 and on the y-axis with y 2 0 are ‘boundary points ofthe quedrant ‘A point a is said to be adherent (0 $ if every dise D(a.) with r>0 contains some clement of S. A point « is said to be an interior point of S if there exists a disc D(e,r) which is contained in S. Thus an adherent point can be a boundary point or an interior point ofS. A set is called 98) Lists AND CoMPACT sets 9 closed if t contains al its boundary points. The complement of a closed [Link] then open. The closure of a set S is defined to be the union of $ and all is boundary points. We denote the closure by 5. A set Sis said (0 be bounded if there exists a number C > 0 such that KSC forallzins. For instance, the set in Fig. 10 is bounded. The fist quadrant is not bounded Figure 10 ‘The upper half plane is not bounded. The condition for boundedness means that the set is contained in the closed disc of radius C, as shown on Fig. 10, Let f be a function on S, and let « be an adherent point of S. Let w be a complex number. We say that w= tim fl) it the following condition is satisfied. Given € > 0 there exists 5 > 0 such that f2@ and [z~ al <4, then Ve) ~ wl <« ‘We usually omit the symbols 2 S under the limit sign, and write merely tim fe) In some applications we $ and in some applications, a S. 2» ‘COMPLEX NUMBERS AND FUNCTIONS iro) Let eS. We say that f is continuous ato if lim fe) = fee, ‘These definitions are completely anslogous to those which you should hhave had in some analysis or advanced calculus course, so we don't spend much time on them. As usual, we have the rules for limits of sums, products, quotients asin calculus If.) (@= 1,2...) i sequence of complex numbers, then we say that ifthe following condition is satis: Given € > 0 there exists am integer N such that if n NY, then lnowlce Let $ be the set of fractions In, with n= Then Let f/m) = 2, lim :,=w ifand only if tim sl) =. ‘Thus basic properties of limits for n-+ co ate reduced to similar proper- ties for functions. Nove that inthis case, the number 0 is not an element ofS, [A sequence (z,) is sad to be a Cauchy sequence if given ¢, there exists 1 such that if mn 2, then Few <6 Write Beet ite Since ental = Ve xO. WP and sy — al Slee Dy Yul Slee we conslude that {2,) is Cauchy if and only ifthe sequences (x,) and {.) of real and imaginary parts are also Cauchy. Since we know that real Cauchy sequences converge (ie. have limits), we conclude thet com- plex Cauchy sequences also converge. ‘We note that all the usual theorems about Hmite hold for complex numbers: Limits of sums, limits of products, limits of quotients, limits us) LMITs AND Compact sers a ‘of composite functions. The proofs which you had in advanced calculus hhold without change in the present context. It is then usually easy 0 compute limits. Example. Find the limit lim See for any complex number 2 We 2=0, itis clear that the limit is 0. Suppose 2 #0, Then the ‘quotient whose limit we ate supposed to find ean be written Hence the limit of the quotient is 2/t = 1 Compact Sets ‘We shall now go through the basic results concerning compact sets. Let S be a set of complex numbers. Let {z) be a sequence in S. By a point of accumulation of {2,} we mean a complex number v such that given € (atways assumed > 0) thre exis infinitely many integers n such that We may say that given an open set U containing o, there exis infinitely many n such that 3,6 U. Similarly we define the notion of point of accumulation of an infinite setS. It is 8 complex number » such that given an open set U contain- ing, there exist infinitely many elements of $ Tying in U. In particular, 1 point of accumulation of $ is adherent to. ‘We assume that the reader is acquainted with the Welerstrass-Bolzano theorem about sets of real numbers: If $ isan infinite bounded set of real rnbers, then S has a point of accumulation ‘We define a set of complex numbers $ to be compact if every sequence of elements of § has @ point of accumulation in S. This property is ‘equivalent to the following properties, which could be taken as alternate definitions: (a) Every infinite subset of $ has & point of sccumulaion in S. 2 (COMPLEX NUMBERS AND FUNCTIONS usa (Every sequence of clements of $ has 2 convergent subsequence whose limit isin S. We leave the proof of the equivalence between the three possible Aefnitions to the reader. ‘Theorem 4A. 4 set of complex numbers is compact if and only if i is closed and bounded, Proof. Assume that Sis compact. If Sis not bounded, for each posi- tive integer there exists 2, ¢ 8 euch that lel > ‘Then the sequence (z,) does not have 2 point of accumulation, Indeed, if» is a point of accumulation, pick m> 2Jol, and note that |o| > 0. ‘Then leg — 012 be — [ol 2 me — ol > To ‘This contradicts the fact that for infinitely many m we must have 2 close ton. Hence § is bounded. To show $ is closed, let» be in its closure. ‘Given nm, there exists 2, such that pole tym ‘The sequence (s,) converges to 0, and has a subsequence converging to a limit in S because $ is assumed compact. This limit must be e, whence eS and § is closed Conversely, assume that § is closed and bounded, and let B be bound, so [21S B forall :€5. If we write roxtin then x $B and |y| < B, Let (z,} be a sequence in 5, and write 2% tie ‘These isa subsequence (:,) such that (x,} converges to & real number 1, and there is @ sub-subsequence (z,,} sich that (),,} converges (0 a real number b. Then {n= Met Dead converges to a+ ib and S is compact. This proves the theorem, ‘Theorem 42. Let S be a compact set and let S,>5;>~~ be a sequence of nonempty closed subsets such that $. S,.,. Then the imersecton ofall, forall = 1,2. 8 not emp. 0.8) LIMITS AND COMPACT SETS 2B Proof. Let 2465,. The sequence {2,} has 2 point of accumulation in S. Call it” Then v is also a point of accumulation for each sub- sequence {24} with 2m, and hence is in the closure of S, foreach 1, But Sis atoumed dosed and Bence ve, forall TRS proves the ‘Theorem 43. Let $ be a compact set of complex numbers, and let f be 4 continuous function on S. Thon the image of fe compact. Proof. Let (1,} be a sequence in the image of J; so that =) for 268. ‘The sequence {25} has 2 convergent subsequence (2), with a fir S. Since f is continuous, we have Hence the given sequence (w,} has @ subsequence which converges in JS) This proves that f(S) is compact. ‘Theorem 4. Let S be a compact set of complex numbers, and let FSR be a continuous function, Then {has a maximum on S, that is, there ‘exists 0S such that fle) S fe) forall 26S. Proof. By Theorem 43, we know that f{S) is closed and bounded. Let b be its least upper bound Then b is adherent to JS}, whence in $15) because (5) is closed. So there is some veS such that fle) = b. ‘This proves the theorem Remarks, In practice, one deals with a continuous function f:5—+€ and one applies Theorem 44 to the absolute value of f, which is also ‘continuous (compesite of two continuous functions “Theorem 45. Let S be a compact set, and let f be a continuous {function on $. Then f & uniformly continous, Le. gen ¢ there exists 5 ‘uch that whenever 2, w 6 S and 2 — | < 8 then If) flv)] <€ ‘Proof. Suppose the astertion of the theorem i false. Then there exits «and for each m there exists a pair of elements ,, 6S such that lep— ml < tn but Lfle,)— fm) >« ™ ‘COMPLEX NUMRERS AND FUNCTIONS na “There ie an infinite subset J, of positive integers and some ve such that 3)-+0 for n> and me, There isan infinite subset J, of Jy and eS such that wu for n->co and ney. Then, taking the limit for n+ and neJ, we obtain Iu |= 0 and w= because wal sl aha — ml +e, — eh Hence fi) ~ fle) =0. Furthermore, se) Hr Sf) ~ $10] + Lf) ~ fe + Le) fl Again taking the limit as n+ 0 and me J, we conelude that Sie) ~ fl) approaches 0. This contradicts the assumption that Lien) — fol >« and proves the theorem. Let A, B be two sets of complex numbers. By the distance between them, denoted by d(A, B), we mean (A, B= elle — where the greatest lower bound gb. is taken overall elements z¢ A and we. IB consists of one point, we also write d(4,w) instead of d(A, B). ‘We shall leave the next two results as easy exercises, ‘Theorem 46, Let $ be a closed set of complex numbers, and let v be a complex monber. There exists a point we such that ats, [Hints Let be a closed dsc of some suitable and consider the function 2-+[2 ~ o|for se SE) ius, centered at 0, ‘Theorem 47. Let K be a compact set of complex numbers, and let $ be closed set, There exist elements 59 K and wo © such that aK, 5) [Utint; Consider the function 21-45, 2) for 2¢ K:] 0.89) LIMITS AND COMPACT SETS 2s Theorem 48. Let $ be compact. Let r be a real number > 0. There exists a finite number of open discs of radius r whose union contains 5. Proof. Suppose this is false. Let 2, ¢5 and let D, be the open dise of radius r centered at 2,. Then D, does not contain, and there is some 56S, 2, #2,. Proceeding inductively, suppose we have found open discs D,,...D, of radius r centered at points 2, ..,2,tespectively, soch that 2,4, does not ie in D,u-~"w Dy, We can then find 2,., which docs ot fie in D,U~-UD,, and we let Bs, be the disc of radius r centered at yay. Let v be a point of accumulation of the sequence (z,). By 4efttion, there exist positive integers m, k with k > m such that lyrol 0 centered at'w. Let N be so latge that 2/N <7. There exists n> N such 26 ‘COMPLEX NUMBERS AND FUNCTIONS us that le wls UN. Any point of D, is then at a distance <2/N from w, and hence Dy is contained in D, and thus contained in U,. This contradicts the hypothe- sis made on D,, and proves the theorem. 1, §4. EXERCISES 1. Late be a complex number of absolte valve <1 What i im a"? Proof? 2. Wel > 1, does lin 2 exit? Why? ~ 2 Show tht for any complex number z# 1, we have mn (ete 4. Let be the faction defined by 1s it posible to define fl) when [1 = 1 in such way to make f continvous? f= bears (@) What i the domain of dfaition off, that for which complex numbers = does the init ext? (©) Give expliity the wales of fe) forthe vari 1 Show tht the series 39 ‘COMPLEX DIFFERENTIABILITY ” converges t© 1A ~ 2} or [el <1 and to 1s(1~2) for el > 1 Prove that the convergence is uniform for [> ¢ <1 inthe fst case and [a] 28> Vin the second. [Hint Muliply and divide each term by I and do 8 pata fraction decompostion, geting a telescoping fet) |, §5. COMPLEX DIFFERENTIABILITY In studying differentiable functions of a real variable, we took such func- tions defined on intervals For complex variables, we have 10 select ‘domains of definition in an analogous manner, Let U be an open set, and let 2 be @ point of U. Let f be a function fon U, We say that fis complex dferendable at» ifthe limit Se+h 7 ‘exists. This limit i denoted by (2) 0 afd In this section, differentiable will alvays mean complex differentiable ‘The usual proofs of frst course in caleulus concerning basic proper- ies of diferentiabilty are valid for complex diferentiablity, We shall run through them again, We note that if f is differen because le at F then fis continuous at = tn +H = fle ® im (fle + )~ fa) and since the limit of a product isthe product of the limits the limit on the rightchand side is equal to 0 ‘We let fg be functions defined on the open set U, We assume that Sg are diferentiable at. ‘Sum. The sum f+ 9 is difereniable at 2, and Ur are)=se)+ oe. Proof. This is immediate from the theorem thatthe limit of & sum is the sum ofthe init. Product. The product fo i differentiable at 2, and arte = fe)06) + fle’ 2B ‘COMPLEX NUMBERS AND FUNCTIONS Proof. To determine the limit ofthe Newton quotient She + Wale +) feel) Re \we write the numerator in the form Sle + Rg + I) flee + W) + fede +H) = fal ‘Then the Newton quotient is equal to a sum LEAN Fe yy got N=) “Taking the limits yields the formula, ‘Quotient. If (2) 0, then the qutien ff is differentiable at z, amd geyte)— flas'e) ate Proof. This is again proved as in ordinary calculus. We fist prove the diferentibility of the quotient function 1/g. We have __oe+h-o@) he Mate) Taking he nt ils 1 “ae which isthe usual value. The general formula for a quotient from this by writing obtained Sla= So. and using the rules forthe derivative of a product, and the derivative of Wo Examples. As in ordinary calculus, fom the formula for a product and induction, we see that for any postive integer n, ast nee, 85) » ‘The rule for 2 quotient also shows that this formula remains valid when nis a negative integer, "The derivative of #2 — 1) is (@2— N22 @-F ‘This formula is valid for any complex number # such that 2¢—1 #0, More generally, et Se) = Prev where P, Q are polynomials. Then f is differentiable at any point 2 where O(2)# 0 Last comes the chain rule. Let U, V be open sets in C, and let FU9V and gV4C be functions, such thatthe image of fis contained in V. Then we can form the composite function ge f such that oof olse)- (Chain Rule. Let w = fle). Assume that fs diferentiabe at 3, and g i= diferentiable at w. Then 9s f 1s diferenlabe at 2, and (oo sve=o(enrer Proof. Again the proof is the same ss in ealeulus, and depends on cexpresing diferentiabiity by an equivalent property not involving de- nominator, as Follows. Suppose that fis diferentiable ats, and let y= LEAN=IE py thea o Neh =f0)= 10h hes at @ Lim oh) = 0 30 (COMPLEX NUMBERS AND FUNCTIONS 1.95) Furthermore, even though g is at first defined only for sufficiently small ‘hand #0, we may also define g(0)= 0, and formula (I) remains valid for h~0, Conversely, suppose that there exists a function @ defined for sufi- ciently small hand a number a such that 0 J +H) J) = ah + ho) and e Dividing by h in formula (I') and taking the imit as h—+0, we see that the limit exists and & equal to a Thus f(a) exsts and is equal to a Hence the existence of a function o satisfying (I, (2) is equivalent (0 diferentabiity. We apply this to proof of the chain rule. Let w = fie) and km fet m-flo, s0 that af see +) — a Je) = atw + 4) — ato, ‘There exists function yh) such that tim Y(A) = 0 and ator + K)~ gto) = gk + vk) = waste + Wy ~ fle) + (e+ flaw Dividing by h yields 2 a) geste + h)= h [As h0, we note that k +0 also by the continuity off, whence yh) +0 by assumption. Taking the limit of this last expression a8 h»0 proves the chain rule [A function f defined on an open set U is said to be differentiable if it is differentiable at every point. ‘We then also say that f is holomorphic ‘on U. The word holomorphic is usually used in order not to have to spocly complex diferentability as distinguished from real difleretibilty. 1.96) “THE CAUCHY-RIEMANW EQUATIONS 3 1m line with general terminology, a holomorphic function f0av from an open set into another is called holomorphic isomorphism if there exists a holomorphic function avau such that gis the inverse of f that is, Gof ily and fog iy ‘A holomorphic isomorphism of U with itself is called a holomorphic automorphism. In the next chapter we discuss this notion in connection With functions defined by power series. |, §6. THE CAUCHY-RIEMANN EQUATIONS Let f be a function on an open set U, and write f in terms of its re and imaginary parts, Sl i) = ws 9) + ils 9h It is reasonable to ask what the condition of diferentiailty means in lerms of u and o. We shall analyze this situation in detail in Chapter VIL, but both for the sake of tradition, and because there is some need psychologically to see right away what the answer is, we derive the ‘equivalent conditions on u,v for f to be holomorphic. ‘Ata fixed 2, let (2) =a + Dt Let w= hi, with hk real, Suppose that Sle + w)~ fle)= Few + own, ‘where lim et) = 0. ‘Then ren ‘On the other hand, let FUR? 2 ‘COMPLEX NUMBERS AND FUNCTIONS 0.661 be the map (often ealled vector field) such that Fes) (ux yh ot 9D We call F the (real) vector field associated with f. Then Fle + hy + Fx, 9) = ah — bk, Bh + ab) + 00h, D+ 0908 where oy(h. kl ooh k) are functions tending to 0 as (hk) tends to 0. Hence if we assume that fis holomorphic, we conclude that F is difer- centiable in the sense of real variables, and that its derivative is repre- sented by the (Jacobian) matrix, du au zy 2 B® ‘This shows that sed ae ‘These are called the Cauchy-Riemann equations. Conversely, let u(x.) and vfx,y) be two functions satisfying. the CCauchy-Riemann equations, and continuously diflerentiable in the sense of real functions. Define sl e+ iy) = ws, 9) its 9 ‘Then it is immediately verified by reversing the above steps that f is comples-diferentiable, ie, holomorphic. “The Jacobian determinant Ay of the associated vector field F is sneer a(S) (5) -G) +) as ANGLES UNDER HOLOMORPHIC MAPS 2 Hence 4, 2 0, and is #0 if and only if (2) #0. We have Syl. = IFO ‘We now érop these considerations until Chapter VIUL ‘The study of the real part of a holomorphic function and its relation with the function itself will be carried out more substantially in Chapter VIL It i important, and much of that chapter depends only on elemen- tary facts. However, the mest important part of complex analysis at the present level les in the power series aspects and the immediate applica- tions of Cauchy’s theorem. The real part plays no role in these matters ‘Thus we do not wish to interrupt the straightforward flow of the book snow towards these topes. However, the reader may read immediately the more elementary pars 91 and §2 of Chapter VIII, which can be understood already at this point |, §6. EXERCISE 1. Prove in detail that ifm, w satisfy the Cauchy-Riemann equation, then the function F121) = wa 9 +o 9 is holomorphic. |, §7. ANGLES UNDER HOLOMORPHIC MAPS In this section, we give a simple geometsic property of holomorphic maps. Roughly speaking, they preserve angles. We make this precise as fellows Let U be an open set in C and let fa] 0 be a curve in U, so we write = x04 190. ‘We assume that y is differentiable, so its derivative is given by O=XO+ HO. “ ‘COMPLEX NUMBERS AND FUNCTIONS 97 Let f:U-+€ be holomorphic. We let the reader verify the chin rule a Fs) = sow. We interpret '() as a vector in the diretion ofa tangent vector at the pointy) This derivative (not 0, defines the direction ofthe curve Bt the point » © yn 1 awe 11 Consider two curves y and » passing through the same point x. Say Za te) = ns) Then the tangent vectors »(4) and 1(,) determine an angle @ which is ‘defined to be the angle berween the curves. Figure 12 Applying J, the curves fo7 and fn pass through the point flee) and. by the chain rule tangent vectors of thee image curves are Seobvtto) and fea) Theorem 7. If f'(zo) A 0 then the angle between the curves 7. at 20 Is the same asthe angle beeween the cures fe fn at flceh Proof. Geometrically speaking, the tangent vectors under fare changed by multiplication with f'(a) which can be represented in polar coordinates asa dilation and a rotation, so preserves the angles, ae ANGLES UNDER HOLOMORPHIC MAPS 3s We shall now give a more formal argument, dealing with the cosine and sine of anges, Let 2, w be complex numbers, +bi and woes, where 2b, 6, dare real Then Ba ee 4 bd ibe — ad), Define the scalar product o ew elem) ‘Then, ¢2,W) is the ordinary scalar product ofthe vectors (a,b) and (6,4) in R*. ‘Let 62, ») be the angle between 2 and w. Then om @ cos 02,9) = J em fell Since sin @ = cos(@—"), we can define bnsn(0-$)- ween iwy Tew o sin (0) = ‘This gives us the desired precise formulas for the cosine and sine of an angle, which determine the ange Let fz) =a Then o tam) = Re(aa8) = a8 Rete = al*m, then the term A has all coefcents of order Sm equal to 0, ‘Thus we may define the m-th coefficient of @ t0 be the mth coefcint of the finite sum Dah Tt is then easy to verify that (mead he ER Bm) is equal t0 14 a power serie of arbitrarily high order, and consequently is equal to 1. Hence we have found the desired inverse for f. Example. Let ror coo Ta1-T 47 be the formal power series whose coefficients are the same as for the ‘Taylor expansion of the ordinary cosine function in elementary calculus We want to write down the first few terms ofits (formal) inverse 1 wT Up to terms of order 4, these il be the same as rir re +o-Gt tat wd (Ghd) reste “This gives ws the coefficients of 1fe0s T up to order 4 2 PoweR SERIES a 5 ‘The substitution of J in the gcometric series used to find an inverse ‘can be generalized. Let f-Eate be a power series, and let wnt bbe a power series whase constant term i 0, So ord h& 1. Then we may “Substitute” rin f to define a power series foh or 4, by (Fo BNT) = J(HCT) = fo h= ag + ash agh + aah? + way. Indeed, the finite sums ay taht ah are defined by the ordinary sum and product of power series. If m>n, then aqh™ has order >; in other words, it is a power series starting with ‘non-zero terms of order >n Consequently we can define the power series fc as that series whose th coefficient isthe nth coefficient o ag taht taht This composition of power series, ke addition and multiplication, can therefore be computed by working only with polynomials In fat, it is useful to discuss this approximation by polynomials @ litle more ‘ystematical We say that two power series f= ¥a,T* and g = Sb,T* are com srucnt mod 7" and write f= g (mod 7) if aad for n= NA ‘This means that the terms of order << N — I coincide forthe two power series. Given the power series f, there is a unique polynomial P(T) of degree SN ~ 1 such that S(T) = PT) (mod 7, namely the polynomial Pury fo TH+ Oy TH ost) FORMAL POWER SERIES 48 Wf, = fy ond 9, = 93 (0d T), then LF Oto, and — fy, foo (mod 7, If ys hy are power series with zero constant term, and 4, by (mod 7"), then AlbT) = fT) (mod 7%). Proof. We leave the sum and product to the reader. Let us Took at the proof for the composition fel, and f,ehy. Fist suppose h has zero constant term. Let F., be the polynomials of degree N — 1 such that Lee R and f= R (mod TH. Then by hypothesis, A, isthe same polynomiat, and Si) = Pd) = BAA) = f(b) (0d 7") [Next let Q be the polynomial of degree V1 such that aan (7) = Q(T) (mod 7", Write P= ay + ay T-$- 4+ aya T*. Then Ph) = a9 + ayy "ay 9H a, ay 5 yt oghy t+ ay hf Plh) (mod T") ‘This proves the desired property that fy oh = fy hy (mod 7) ‘With these rues we can compute the coefficients of various operations bbetneen power series by reducing the computations to polynomial opera- tions, which amount to high-school algebra. Indeed, two power series Sg ate equal if and only if S20 (mod 7%) for every positive integer WV. Verifying that f= (mod T*) can be “4 POWER SERIES os0 done by working entirely with polynomials of degree < W. YW Sas Sy are power series, then H+ £10 = full) + 0 GLI = HOMER, and — Ye Lf Hd = Sul If ord fy = 0. If gk have constant terms equal 100, then Sots) Proof. In each case, the proof is obtained by reducing the statement to the polynomial case, and secing that the requited properties hold for polynomials, which is standard. For instance, for the associativity of ‘composition, given a positive integer N, let P, Q, R be polynomials of degree SN I such that fo ai f R (mod 7". ‘The ordinary theory of polynomials shows that P(Q(R)) = (Pe OR. ‘The lefthand side is congruent to fth)), and the righthand side is congruent to (fe 9)(h) (mod 7") by the properties which have already been proved, Hence Hath) =F © anh) (mod 7" ‘This is true foreach N, whence f(oth) = (Fe ohh a desired. In applications itis useful to consider power series which have a finite umber of terms involving 1/2, and this amounts also to considering arbitrary quotients of power series as follows. Just as Fractions mj are formed with integers mm and n #0, we can form quotients Sla= ST NgT) of power series such that 90. Two such quotients fig and f,/o, are regarded as equal if and only if Sa = fa Which is exactly the condition under which we regard two rational num- 91) FORMAL POWER SERIES 4 bers mix and my/n, a8 equal We have then defined for power series all the operations of arithmetic. Let JT) = 0g ges TH par te 8 power weis wih o£ 0. We may then write fin the form Sma T™(l + MT). wer HT has zero constant term. Conequnty if haste form 14 eres) ‘We know how to invert 1+ W(7), say (+ ncnyt = 14,7 +6, 7 ‘Then 1/f(T) has the shape I is a power Series with a finite number of terms having negative powers of T. In this manner, one ses that an arbitrary quotient can always be expressed a8 a power series of the form +t te teotaT batt t Ear If con 0, then we call —m the order of fg. It is again verified as for power series without negative terms that if en fe ad %=flo ord 9, = ord 9 + ord 9, Find the terms of order 3 in the power series for sin T sin T= T= TIL TES! — = TU Teste Test ~~) 46 POWER SERIES asa Hence mo 1 Set" TIP ae Pe 1479/31 71/314 (75/2 + higher teams =f) 1 ire “This does what we wanted, I, §1. EXERCISES We shall wrte the formal power sees in terms of = because that's the way they arise in practice. The series for sin = co87, cele are to be viewed as formal 1. Give the terms of order 3 inthe power svies: () esinz 0) Gn2¥eos2) 10 tobe edd if 2, = 0 for n ere, Verity that fis even if and ony i f=2)= fe) and f is od i and only if Sa afe. | Define the Beroal amber 8 by the power seis em. Prove the recursion formula oft et, lo itn. os CONVERGENT POWER SERIES " Replace z by 2: to show that Express the power series for tan, skin, scot jn terms of Bemoul umber. 6 Dillernce Equations). Given complex numbers ay, a, 1, My define oy for nE2by wits tater we have s factorization Wy Tome (—oKT =a), and att, show thatthe numbers a, a6 given by ith abe A B. Fad A.B inte oy 8 Conier he ome rn= Sar: ‘Show tha it represents a ratonal fanetion, and give its partial fraction decom position 7. More penerly et a9, ite be gen complex mimbers. Let oy aly BE complex number such thai the polynomial PUT = TT 4) Ins itn 008 0... Define 4 for mz by A= Hiden tH lee ‘Show that there exit numbers Ay,..A, such that for alm ae At A i, §2. CONVERGENT POWER SERIES ‘We fist recall some terminology about series of complex numbers. Let (25) be a sequence of complex numbers. Consider the series «8 PoweR SERIES a2) pet a ‘We say thatthe series converges if there is some w such that exists, in which case we say that w is equal to the sum of the series, that is then the sum and produet converge. Namely, A+ BaD +8) AB = Bien Sy Let 64) be a series of real numbers ¢, 2 0. If the partial sums fare bounded, we recall from calculus that the series converges, and that the least upper bound of these partial sums is the limit Let Fa, be a series of complex numbers. We shall say that this series converges absolutely ifthe real postive series Diet converges: Ifa series converges absolutely, then it converges. Indeed, let be the partial sums. Then for ms» we have oan Fo a (u, 621 CONVERGENT PowER SERIES 0 whence tu -suis Ste Assuming absolute convergence, given « there exists NV such that ifm, ‘ZN, then the righthand side of this lat expression is <<, thereby proving that the partial sums form a Cauchy sequence, and hence that the series converges. "We have the usual test for convergence: Let 5c, be a series of rel numbers 20 which converges. If lag S for alin, then the series Soy converges absolutely. Procf. Te pastil sums av turd yas, wc he pr ss Eis ha sr so ound a te ale omer te, In the sequel we shall also assume some standard facts about abso- Tutely convergent series, namely (Ifa series Ta is absoluely convergent, then the series obtained by ‘any rearrangement of the terms 1s also absolutely convergent, and converges to the same lit (i Ifa double series EE) is absolutly convergent, then the order of summation can be inter- changed, and the series so oblained is absolutely convergent, and converges to the same value ‘The proof is easily obtained by considering approximating partial sums (nite sums), and estimating the tail ends. We omit it "We shall now consider series of functions, and deal with questions of sniformity. Let be 2 set, and f bounded function on S. Then we define the 0 POWER SERIES 621 sup norm Us Is = sup ister, ‘where sup means least upper bound. I€ is @ norm in the sense that for two functions f, 9 we have Lf + al 1/l+ lal, and for eny number © wwe have Lef= el. Alo f= 0 if and only if If =O. Let {f) (n= 1,2,..) be a Sequence of functions on S. We shall say that this Sequence converges aniformly on Sif there exists a function f on 5 satisfying the fllowing properties. Given «, there exists W such that if REN, then U-Sl 0 such the the series Elale converges. Then the series Taye" converges absolutely and uniformly Jorlisr Proof. Special case of the comparison test. Example. For any r> 0, the series Lent ‘converges absolutely and uniformly for |z| Proof. Suppose thatthe series does not converge abeoltely forall 2 Let be the least upper bound of those numbers s 2 0 such that lads converges Then F Jat verges if el > 1, and converges i ol <¥ by the comparison te sour aston Obie “The number r in Theorem 25 is called the radius of convergence of the power series, Ifthe power series converges absolutely forall, then we Say that its radius of convergence is infinity. When the radius of conver- fetoe is 0, then the series converges absolutely only for z= 0. TT power serice har a nonzero radius of convergence, then itis called a convergent power series, I D is a disc centered atthe orgin and Contained in the disc D(Q,r, where ris the radius of convergence, then wwe say that the power series converges on D. ‘The radius of convergence can be determined in terms of the co- efficients. Lett, be @ sequence of real numbers. We recall that a point of fsecumblation of this sequence is @ number ¢ such that, given ¢, there exist Es PoweR SERIES 0,9 infinitely many indices such that In other words, infinitely many points of the sequence tie in a given interval centered at t. An elementary property of real numbers asserts that every bounded sequence has a point of accumulation (Weiestrass~ Baleano theorem) ‘Assume now that (t5} is a bounded sequence. Let S be the set of points of accumulation, $0 that S looks ike Fig. pet} tet ert gue 1 We define the limit superior, lim sup, of the sequence to be a im sup f= least upper bound ofS. ‘Then the reader will verily at once that 2 is ite a point of accumulation of the sequence, and is therefore the largest such point. Purthermore, 2 has the following properties: Given ¢, here exist only finitely many n suck that 1,246. There exis infinitely many m such that Proof. If there were infinitely many m such that ¢, 2 2-4 6 then the sequence {1,} would have 2 point of accumulation zlterd, contrary to assumption. On the other hand, since 2 itself is a point of fsccumulation, given the interval about 2, "there have to be infinitely many n such that t lies in this «interval, thus proving. the second ‘We leave it to the reader to verily that if @ mmber 2 has the above properties, then iti the lim sup of the sequence. For convenience, if (t,} is not bounded from above, we define its lim sup to be infinity, wetien oo. 2] CONVERGENT POWER SERIES 58 ‘As an exercise, you should prove: Let (tgs {5} Be Sequences of real mumbers 20. Let r=limsept, and sup sy Then lim spt, +5) S45 1 140, then lim suptess) $s 1 len ty exists, then t= lin ty ‘This last statement says that if the sequence has an ordinary limit, then that limit isthe lim sup of the sequence. “The second statement is often applied in case one sequence has 2 lim sup, and the other sequence has a limit 70. The hypothesis 1 0 is ‘made only to allow the possibility that s= co, in which ease 1s is under stood to be co. If sc, and ¢# co, and lim, exist, then it is true unrestriciedly that Him supts) ‘Theorem 26. Let Taz" be a power series, and let r be its radius of convergence. Then im suplag! If r= 0, this relation isto be imerpreted as meaning that the sequence (lasit) ie not Bounded. If r= oo, i to be interpreted os meaning that tim supa" = 0. Proof. Lat «= lim supla?™. Suppose fist that #0, 0. Given € > 0, there exist only a finite number of such tha |!" 2+ € ‘Thus for all but a finite number of m, we have lalst+e Whence the series T.,2" converges absolutely if [el < Mle +6} by com- parison with the geometric series. Therefore the radius of convergence r Satsfies FB ICH for every € > 0, whence r 2 it 56 owen SERIES a Conversely, given there exist infinitely many n such that a)!" 2 and therefore lad 20-9" Hence the series 3 a,2" does not converge if [z|= 1/(¢~ e, because its ‘mth term does net even tend to 0. ‘Therefore the radius of convergence r Satishies r 5 Ye — 0 for every «> 0, whence rs 1/t. This concludes the proof in ease 1 0, 2, The case when £=0 oF oo will be let to the reader. The above arguments apply, even with some simplifications. Corollary 2.7. If tm a" = ¢ exists, then r = It Proof. I the limit exists, then + is the only point of accumulation of ‘the sequence [a and the theorem states that ¢ = Ir Corollary 28. Suppose that Ta,e" has « radius of concergence greater than O. Then there exists a postive number C such that f A> Ur then lal SCA" forall n Proof. Let s theorem: In the next examples, we shall use a weak form of St namely YA so 0- 0. IEA= 6, one ean simply replace the ters onthe ef the Incas by 0 throughoat. Example (The Binomial Series). Let «be any complex number #0. Define the binomial coefficients a5 usual, 6) +0 8 POWER SERIES on, 21 And the binomial series astrenin= §(°)r: @)- Te ral of comer of Hamel sre 1 8 mt al ovine 0, By convention, Proof, Under the stated assumption, none of the coefficients 0, are 0, and we have ‘The limit is 1 as n+ 20, $0 we can apply the ratio test. Warning. Let r be the radius of convergence of the series fle) [Nothing has been said about possible convergence if |2| = r. Many cases ‘an occur concerning convergence er non-convergence on this circle. See Exercises 6 and 8 for example UM, §2. EXERCISES 1. Let [al <1. Express the stm of the geometric series Sey 2” i it usual simple form. 2. Letr be a real mumber, 0: r< 1. Show thatthe series ret ana Se conver (0 is el). Expres the series in simple terms wing the uel Touma ors geomet eis 4. Show thatthe usual power seis for log + 2) 0 log ~ 2) fom elementary alee converges absolutly for [ 0. [Hint Foe this problem andthe next, se summa See Appendix, $1] 9, (Abels Theorem). Let $2 a,2* be a power series with radius of convergence by pars, BAL Assume tha the series F oy conver. Let 0 Sx <1. Prove that mnbor-Ee ee Th anc nite sto as en wat ne 0 (CE Theovem 35 of Chapter VIE in my Underraducte Analysis, Springer Vesag, 1983 10 Lat Faye" and Fa" be two power sre, with radius of convergence rsd ss respetvely. What cn you say about the rade of convergence of the @) Dette © Daher 11, Let a Pte complex numbers with al < Ia Let $e) = Dee — spe. POWER seRiEs 0, 81 Determine the radius of convergence off) 12. Let {a} be the sequence of real numbers defined bythe conditions: Beha = and gma tons for mee Deter the radius of convergence ofthe power series ar. [ints What isthe general soation ofa diference equation? CI. Exercie 6 ogi] 13, More genealy let, Be complex numbers sch the polynomial > fay Tay (P= aT 0) has two dine roots witha 0 such that f(2) #0 forall * with 21 Ss, and 2 #0. (6) Let J(7)=3Sa47" and g(T) = SbyT* be two convergent power series, Suppose that f(x) = g(x) forall points x in an infinite set fetng 0 a pit of acm, ‘hen (7) = 47), tat is Proof. We can write Je) is higher terms, and 0, £0 mace hye thet to) e+ HO, where Me) = bz +2? + is a power series having a non-zero radius fof convergence, and zero constant term. For al sufiiently small [s,the value [h(i small, and hence 1m) #0. 240, then ay2" #0. This proves the fist part of the theorem, For part (b) let hi) = f(T) ~ g( 7) = S(@,~b,)T*. We have x)= 0 for an infinite set of points x having 0 a8 point of accumulation. By part (a) this implies that A(T) is the zero power series, $0 a, =, for all n, thus proving the theorem. eampl. Thre ett st ct ne convert power set Lar eh th or sone inenal (ee) ne have Jo) eer or Ae Ecey Tis proves he uniques of ny power sates cen fs Capone fion fo a Somos number Silty oar he te test ofthe powers credng he Soe and Sento (HL §3] RELATIONS BETWEEN FORMAL AND CONVERGENT SERIES 63 Furtemore et ex) = E2!. Then Zeer ‘Summing over even m and odd 1, we find that expt respectively. Hence e* for real 6 coincides with expli@) as given by the power series expansion, Quite generally if (7), A(T) are power series with O constant term, then exp(o(T) +4(T)) = (expa(T)(exPh(7)). Proof. On ove hand, by defrition, and on the other hand, Shantan exvainyfepKin) = SS FEM ag In particular, for complex numbers wwe have ex then substitute T'= 1, Thus we se that the exponent ‘in Chapter I has the same values as the function defined by the usual powe ‘Theorem 32 also allows us to conclude that any polynomial relation between the elementary functions which have a convergent Taylor expsn- son at the origin also holds for the extension of these functions as complex power series, where sins = ‘are defined by the usual power serie Indeed, the has infinite radiws of comergence, and has “Theorem 32 implies that there is at mest ope series having this property, and that is the series 1, as desired. It would be disagreeable to show dreety thatthe formal power series forthe sine and o owen sentes m1. cosine satisly this relation. I is easier to doit through elementary calculus a8 above. Example Let m be a postive inter. We have seen in §2 tha the vison ies »e)=2(°} with «= I/m has @ radius of convergence 2 I, and thus converges ab- solely for |2| <1. By elementary calculus, we have Bor = 14x When x is real, and |x] <1 (or even when [x is sufficiently smal). There fore BUTY" isthe unique formal power series such that Buy = 14x forall sufciently small real x, and therefore we conclude that AUP =14T. Tn this manner, we see that we ean take mth roots by the binomial series when |: <1 Quorients In our discussion of formal power series, besides the polynomial rela tions, we dealt with quotients and also compesition of series. We still hhave to relate these to the convergent case. It will be convenient to introduce a simple notation to estimate power series. Let f(T) = 3. a,T" be a power series. Let oT) =EeT bbe a power series with real coeficiemts c, 20. We say that f is domi- rated by 9, and write f O such that alsa mB (We can take C by picking A large enough) Then 1 ’ Fin Teen + M+ WO ut an<§ ar At Therefore A(T) = g(T) sates aT, ure AP * = ATF But (AT +247 + ATH #> 7 0, and that > Ois @ number such thot Liltalst Sr. Let 9 = StH) be the formal power series obtained by composition, xa(Zar) ‘Then 9 converges absolutly for el $s, and for such 2, at at (hte). Proof. Let g(T)= Fc, 7". Then o(T) is dominated by the series arr $ tai(Z irs) and by hypothesis, the series on the right converges absolutely for |e! = 50 oft) converges absolutly for lel $s. Let SATY= 054 0,7 4 04-4 be the polynomial of degsee V1 beginning the power series f. Then seer) ~ ftier) « $a (Eta and f(3(T)) = o(7) by definition. By the absolute convergence we con- clude: Given ¢, there exits No such that iN 2 Ng and [2] = 5 then tote) — Chie) 0 such that the series converges absolutely for lz ~zol <”, pd such that for such z, we have Stem Zoe ~ so. tH ga ANALYTIC FUNCTIONS. @ Suppose f is a fonction on an open set U. We say that f is analytic on Uff is analytic at every point of U. In the light of the uniqueness theorem for power series, Theorem 3.2, ‘we see that the above power series expressing fin some neighborhood of 2 is uniquely determined. We have flzs)=0 ifand only it a)=0. [A point 20 such that fgg) =0 is called a zero of f. Instead of saying that fis analytic at zo, we aso say that f has a power series expansion at z¢ (meaning that the values of f() for 2 near 25 are given by an ab- Solutely convergent power series a& above). IES is an arbitrary st, not nacessarly open, itis useful to make the convention that a fusetion is analyte on Sif it isthe restriction of an analytic function on an open set containing S, This i useful, for in- stance, when $ isa closed die. ‘The theorem concerning sums, products, quotients and composites of convergent pover series now immeditely imply: If f.9 are analytic on U, s0 are f + 9, fo. Also fig is analytic on the open subset of 2€U such hat gfe) ¥ 0. Ig: U-+V is analytic and f: V-+C is analytic, then fog is analytic. For this las assertion, we note that if U and g(ze) = Wo, 50 -afz) = wot So bole — 20)" and fw) = Yo ant — wo)" for w near wy, then (2) — Wo is represented by a power series {2 — 24) without constant term, so that Theorem 34 applies: We can “substitute” Slated = ¥ aote)— wo to pet the power series representation for f(g) in a neighborhood of a “The next theorem, although easy to prove, requires being stated. It ives us in practice a way of finding power series expansion for a function at a point Theorem 4. Let f() = ¥. 0,3" be a power series whose ras of con tergence ir. Then f i analyte onthe open die DO.) Proof, We have to show that f has a power series expansion st an arbitrary point zp of the dis, $0 |zol0 be such that 70 POWER SERIES Um, $4] lil +5 3) eae (60 disega + 26-14-17 (+je-n) “( [rsde ‘These are the desired terms ofthe expansion, Remark. Making 2 translation, the theorem shows that if f has a power series expansion on a disc D(z, that is, Late 20" for |z ~z4| 0. A function f on the disc for which there exists a power series J-a,2" having a radius of convergence 2r and such that Sie) = Daa" for ail = in the dsc is said to admit a power series expansion on this dsc We shall now see that such a function is holomorphic, and that its 0 be such that [el +6-<7. We consider complex numbers hich that In<6. We have fee MaDaleeir Lal em he PR) iH, §5) DIFFERENTIATION OF POWER SERIES B where A, h) is a polynomial in x and h, with positive inteper coef cents, in fact Reh [Note that we have the estimate: inemis § (1) Subtracting series, we find Se +H) l2)~ ¥,nage™*h m Hay Pe Mh, and since the series on the left is absolutely convergent, 0 is the series fon the right. We divide by 10 get ern EAN=IO 5 gg ADEA M. For [hl <4, we have the estimate IE ght s Sell. 5 DlanlPa(l21,8)- ‘This lst expression is fixed, independent of h. Hence AY hl OE la ee 3- ‘Ash approaches 0, the right-hand side approaches 0, and therefore ° im mY ath ‘This proves that is diferentable, and that its derivative at» is given by the series Fna,2"", a5 was to be shown. Remark. Conversely, we shall see ater Cauchy's theorem that a fune- tion which is diferentiable admits power series expansion—a very re rmarkable fact, characteristic of comple diferentability. From the theorem, we se thatthe Ath derivative of J given by the SOE) = Hag +h ™ owen senies m5) where hy is 2 power series without constant term. Therefore we obtain the standard expression for the coeficints of the power series in terms of, the derivatives, namely 0), I we deal with the expansion at point zo, namely Sle) = 99+ 0,(2 ~ 29) + ayle — 20h 4% then we find egw fi. 1 is utterly trivial thatthe formally integrated series bate has radius of convergence atleast r, because its coefficients are smaller in absolute value than the coeficienls of f. Since the derivative of this integrated series # exactly the series for f, it follows from Theorem 61 thatthe integrated series has the same radius of convergence as f ‘Let be a funetion on an open set U. If gis a holomorphic function ‘on U such that g’ = f, then g is called a primitive for f- We see that a funetion which has a” power series expansion on a dise always has a primitive on that diss. In other words, an analytic function has a local Primitive at every point Example. The function 1/2 is analytic on the open set U consisting of the plane ffom which the origin has been deleted Indeed, or 25 #0, we have the power series expansion rot 1 1 27 gt % toll Ft) S(-te-men) converging on some disc [2 ~ 2] <7 Hence l/s has a primitive on such 4 dis, and this primitive may be called log = m5) DIFFERENTIATION OF POWER SERIES 6 Ii, §5. EXERCISES In Exercies 1 through 5, also determine the rave of convergence of the given 1 ut Prove that (6). Are+ue+se=0, ‘5. For any poe integer et. - acral)” t+ ene EAE Prove that 6 (3) For[z~11-< show that the derivative of the fnction a gos logs = oat +(2— ise (0) Let 2940. For |e ~ ral < 1, define fe) = DI-1H(e~ za. Show that Per= We % owen senies 0,59 UN, §6. THE INVERSE AND OPEN MAPPING THEOREMS Let J be an analytic function on an open set U, and let f(U) = ¥. We shall say that fi an analytic Bomorphism if V is open and there exists tie function aVou such that fog = idy and ge f= functions to each oth ‘We say that fis local analytic isomorphism or i locally invertible at 2 point 2» if there exists an open set U containing %9 such that f is an Analytic isomorphism on U, lovin other words, f and g are inverse Remark. The word “inverse” is used in two senses: the sense of §1, ‘when We consider the reciprocal I/f of a function f, and in the current ‘sense, which may be elled the composition inverse, ie. an inverse for the ‘composite of mappings. ‘The context makes clear which is meant. In this ‘section, we mean the composition inverse. ‘Theorem 61. (@) Let f(T) = .a,T+ higher terms be a formal power series with 4y #0. Then there exists @ unique power series g(T) such that {GolT) = T. This power series also saises @/(T)) ©) If is4 convergent power series, sos 0 (© Let be on analytic function on an open set U containing 29 Suppose that f(z) # 6. Then f isa local analytic lsomorphism at Proof. We fist deal with the formal power series problem (ah and we find first a formal inverse for f(T). For convenience of notation below we write f(T) inthe form siry=a,7~ aT". We seck @ power series such that ‘The solution to this problem is given by solving the equations in terms (11,6) THE mvveRse AND OPEN MAPPING THEOREMS n ofthe coefcients of the power series 4, 6(7) ~ aot ~ ‘These equations are of the form by — Blea en Base 0 and ah where P, is a polynomial with positive integer coefficients (generalized binomial coefficients). In fact, one sees at once that Plo, niBerbyy bead 0gPaalPy scab) #24 a Pala yoenerh where again A. is a polynomial with positive coefficients, In this man- rer we can solve recursively forthe coefficients Dd since, appears linaly with coefficient, #0 in these equations, and the offer terms donot contain fy This ‘shows that formal lovee clsts abd is uniguly determined Next we prove that g(f(7)) = T. By what we have proved already, there existe © power series (7) = eT higher tems that o((T)) = 7. Then using fla) AMD) = al stath(TI) = (HET) = hich proves the desired formal elation. “Assume next that fis convergent ‘We must now show that g() is absolutely convergent on some dis. ‘To simplify the number of symbols used, we assume that a, = 1. This loses no generality, because if we find a convergent inverse power series for a;¥/(2) we immediately get the convergent inverse power series for Sle ist i SN~T~ Fatt be a power series with of real 20 such that Ja < of forall m, Let (7) be the formal inverse of *(T) so oM= Lat ‘Then we have 7 POWER SERIES OH, $61 6 Phooey Guu) = 0 ‘with those same polynomials F, as before. By induction, it therefore Follows that ¢, is real 2 0, and also that (als since by = Ras... byaa) Tt sufces therefore to pick the series /* so that it has an easily computed formal inverse @ which is easly verified to have a postive radius of convergence. It is now a simple matter to carry out this idea, and we pick for J* a geometric series. There exists A'> 0 such that forall we have lal s 4% (We can omit 2 constant C in front of A* by picking A sufficiently large) Then or a SU=7- Fan The power series o(T) is such that f*(9(T)) = T; namely hich is equivalent with the quadratic equation (A? + ATH ~(1 + ATI(T) + T= 0. ‘This equation has the solution AT ~ Je ATP aS ot = Lt AT = VIC ATP = aT = Hh Mae Ay ‘The expresion under the radical sign is of the form (ATU a i ary(s ATA) Seana ATA? + A)\ wan Tea) ‘We use the binomial expansion to find the square root of a series of the (UU, 69] THE IWVERSE AND OPEN MAPPING THEOREMS ” form 1 + H(T) when K(T) has 220 constant term. Tt is now clear that ‘0(T) is obtained by composition of convergent power series, and bence has a non-zero radius of convergence, This proves that the power series (7) also converges, Finally, for (e, suppose first that 20 = 0 and f(z) = 0, s0 fi analytic fon an open set containing 0, This means that f has a convergent power series expansion at 0, so we view J a5 being defined on its open disc of convergence f:D9e Let Y bean open disc centered at 0 such tat Yp is contained in the disc convergence of g, and such that Vg) < D. Sich » neighborhood of 0 eniss simply because 9 8 continuous. Let Uy =J-¥(Vg) be the St of all 26D such that fhe Ne. Let Sei to-*Yo te the restiction off to Up. We cai that fo 6 an analytic isomer- phism. Note that gH) © Uy because for we V, we have ftw) =w by ‘Theorem 34, so we consider the restriction gp of 9 to Ho 38 mapping 04: Yo Us Again by Theorem 34, for z€ Up we have go(fol)) = 2. which proves that fo and gp ae inverse to each other, and concludes the proof of “The general case is reduced to the above case by translation, as one says. Indeed, for an arbitrary f, with fle)= 3: a@ ~ 29), change vari ables and let wnt nte Foie sdf Sant ‘Then we may apply the previous special case to F and find a local inverse G for F. Let w= fle) and let 960) = G0 — W) +20 “Then gis loal inverse for J, thus finishing the proof of Theorem 61 “There are (at least) four ways of proving the inverse function theorem, 1. The way we have just gone through, by estimating the formal in verse {0 show that it converges. 0 POWER SERIES aga 2. Reproducing the ral variable proof for real functions of class C' By the contraction principle, (ehrinking lemma), one first shows that the ‘map is locally surjective, and one constructs a local inverse, which is shown to be diflerentiabe, and whose derivative satisfies, for w = fh the relation 9) = UFO ‘The reader should be able to copy the proof from any standard book on analysis, (certainly from my Undergraduate Analysis [La $3). 3._ Assuming the theorem for C* real functions. One can show (and we shall do so later when we diseus the real aspects of an analytic function) that an analytic function is C°, as a funetion of (x), writing pexty. ‘The hypothesis f"(z) 4 O (namely a, # 0) is then seen to amount to the property that the Jacobian ofthe real function of two variables has non- 2210 determinant, whence fas aC inverse leally by the real theorem. Tt is then an easy matter to show by the chain rule that this inverse ‘satisfies the Cauchy-Riemann equations, and is therefore holomorphic, whence analytic by the theory which follows Cauchy's theorem, 4. Giving an argument tated on more complex function theory, and carried out in Theorem 17 of Chapter VI All four methods are important, and are used in various contexts jn analysis, both of functions of one variable, and functions of seve variables, Let U be an open set and let f be a function on U. We say that f is fan open mapping if for every open subset U" of U the image s(U') is open. ‘Theorem 6.1 shows that the particular type of function considered there, ic. with non-zero first coeficient in the power series expansion, i locally open. We shall now consider arbitrary analytic functions, fist at the origin, Let Je) = Zag" bea convergent non-constant power series, and let m Se int" + higher terms, oy #0 = ag2"(1 + HG), a $6] THE INVERSE AND OPEN MAPPING THEOREMS at ‘where 2) is convergent, and has zero constant term. Let a be a com plex number such that a” = 2, Then we can write ft) in the form Sle) = (a(t +b). where /iy(2) is @ convergent power series with 2ero constant term, ob- tained from the binomial expansion (+ ay = LC, and SiC ah + ho) 1 + a2hy() is power series whose coefficient of 2 is «#0. Theorem 61 therefore applies to (2, whichis therefore locally open at the origin. We have Sta) = fr. Let U be an open disc centered at the origin on which J, converges ‘Then fi(U) contains an open disc V. The image of V under the map isa disc. Hence f(U) contains an open disc centered atthe origin. ‘Theorem 62. Let f be analytic on an open set U, and assune that for cach point of U, f is not constant on a gen neighborhood of that point. Then fis an open mopPing. Proof. We apply the preceding discussion to the power series expan- sion of fat a point of U, s0 the proof is obvious in the light of what we Ihave already done, “The construction in fact yielded the following statement which itis ‘worthwhile extracting as @ theorem. ‘Theorem 63. Let f be analytic at point 20, Steyn a9 + ¥ ae 20, with m & 1 and dy #0. Then there exist at O such that Local analytic isomorphism Ste) 19+ 220) 2 POWER SERIES 5) We interpret Theorem 63 a5 follows. Let wUaY be an analytic isomorphism. We write w= ya). We may view yas a change of coordinates, from the coordinate 7 10 the coordinate w. In “Theorem 63 we may therefore write Sl = a5 + Ww", where w= @(e ze). The expansion for f in terms of the coordinate wis therefore much simpler than in terms of the coordinate 2 We also get a criterion for function to have an analytic inverse on a whole open set. ‘Theorem 64. Let f be analytic on an open set U, and assume that f is injective. Let V= f(U) be its image. Thon f:U-»V is an analytic ‘isomorphism, and f") #0 forall 6 U: Proof. The function f between U and V is bijective, so we can define an inverse mapping g: V+ U. Let zo be a point of U, and let the power Series expansion of f at 25 be as in Theorem 63. If'm> I then we see that f cannot be injective, because the mth power function in a neigh: Dorhood of the origin is not injective (it wraps the dsc m times around), Hence m= 1, and Theorem 6. now shows that the inverse function gis analytic at lg) This proves the theorem. Example 1 Let fle) = 3 ~ Se + higher terms. Then f(0)= 3, nd LO= a, Hence J isa local analytic isomorphism, or locally invertible, at 0. S40, Example 2 Let fz) = 2-22 42% We want to determine whether f is locally invertible at 2= 1. We write the power series expansion of fat 1, namely Se)= +(e WF = 1+ ale — 1. Here we have a, = 0. Hence fis not locally invertible at z= Example 3. Let fe) at 20. In this cae 2. Determine whether fi locally invertible Sle) =1— 5 + higher terms, 50 a) =O and f is not locally invertible U1 §7] THE LocaL waxnwuss MODULUS PRINCIPLE 8 Example 4. Let fle)= 2°. Then fe) = 32% and (0)=0. Thus f is not locally invertible at 0. On the other hand, {"e) 0 ifs #0. Hence if 29 #0 then f is locally invertible at zo. However, let U be the open set obtained by deleting the origin from C. Then f is not invertible on U. (Why Ii, §6. EXERCISES ermine which of the following functions ar Tcl analytic isomorphism at the ‘ren point Give the rearon for your shaver L fine ara 2. Je) =siote) at A. feyate~ IMe=2)at = 4 fe) = (in at 2-0. 5. fl)=coez at 2 {6 Linear Dferetal Equations. Prove: Theorem. Let ach yl) Be analyte fmctions in a neightorhod of 0. Assure that (0) #0. Given munbers co iy thre es @ ugue anes Iie fiction f at 0 such that PY = 6 for waht (nd sch that D4) + 0,1 Ya) ++ ase) =O [Hint First you may assume os) = 1 (why) Then save for J by a formal Domes seven Then prove ths formal series conver] 7. Ordinary Diferential Equatoes. Prove: Theorem. L219 be only at 0. There exis a nique analytic fineton fat O saisyng f0)=0, ad ferme. (iin: Aesin find formal soliton, a then prove that i converses] Wore: You wil find the above two problems worked cut in the Append, $3, but please try too them fiat before looking up the solutions} Ii, §7. THE LOCAL MAXIMUM MODULUS PRINCIPLE. “This principe i an immediate application of the open mapping theorem, and so we give it here, to emphasize is direct dependence with the preceding section. On the other hand, we wait for a lster chapter fr less 8 Power senies os) basic applications mostly for psychological reasons. We want to alt rate the formal operations with power series and the techniques which will arise from Cauchy's theorem. The later chapter could logically be read almost in its entirety after the preset section, however ‘We say that a function fis locally constant at a point 2 if there exists fn open set D (or 8 dise) containing tp such that fs constant on D. ‘Theorem TA. Let f be analytic on an open set U. Let 26U be a maxima for fh thot is Vez Le, fora se. Then J slay constant at 2, Proof. The function J has a power series expansion at zo, $0) = a5 + a2 20)+ 1 fis not the constant ag = fz) then by Theorem 62 we know that f js an open mapping in neighborhood of z, and therefore the image of Ff consis & se Diz.) of radius «> 0, centered al ay. Hence the st ‘of numbers [f(a for’ z in a neighborhood of zo, contains an open interval around ag, s0 [ft] > [fe] for some z. Hence fo) = aol cannot be a maximum for f,a contradiction which proves the theorem. Corollary 72. Let f be analytic on an open set U, and let 96 U be a mmaximun for the rel part Re f, tht is, Re fles)ZRefleh forall 260. Then fis aly constant ot +0 Proof. Te function e! is analytic on U, and if St) = we) +i) isthe expression of fin terms fits el and iain parts, then Jef) = er, Hence a maximum for Ref is also @ maximum for le/l, and the corollary fellows from the theorem, U1 §7] THE LocaL waxiMuM MODULUS PRINCIPLE 85 ‘The theorem is often applied when f is analytic on an open set U and is continuous at the boundery of U. Then a maximum for [/( necessarily oocurs on the boundary of U. For this one needs that U is connected, and the relevant form of the theorem will be proved as ‘Theorem 1.3 of the next chapter We shall give here one more example of the power of the maximum ‘modulus principle, and postpone to 2 later chapter some of the other applications “Theorem 73. Let J) = 05 + 042-40 + ayet be a polynomial, nor constant, and say ay #0, Then has some com= plex zero, ie. a mumber 2 such that fl) Proof. Suppose otherwise, so that IUf(e) is defined for all 2, and efines an analyte function. Writing 2 (rhe), im tle) = cone sees that Let « be some complex number such that fle) #0. Pick a positive number R large enough such that ll < R, and if el R, then aoe Tran = ea Let $ be the closed dise of radius R centered at the origin, Then $ is closed and bounded, and 1/1/() is continuous on S, whence has 2 maxi ‘mum on S, say at Zo. By construction, this point ze cannot be on the ‘boundary of the disc, and must be an interior point’ By the maximum ‘modulus principle, we conclude that 1/f() is locally constant at zo. This Js obviously impossible since f itself is not locally constant, say from the expansion Sie) Fb ze) to be With suitable coefficients by, ..Jy and by ¥ 0. This proves the theorem, CHAPTER Ill Cauchy’s Theorem, First Part I, §1. HOLOMORPHIC FUNCTIONS ON CONNECTED SETS Let [2,8] be a closed interval of real numbers. By a curve y (defined on ‘this interval) we mean a function 7 La,B]-¥€ which we assume to be of cass C' ony — NJ Figue 1 ‘We recall what this means, We write 10 = nO i100, where 7, isthe real part of 7, and y, is its imaginary part. For instance, the curve 6) =c0s0-+ising, 0S0S2n isthe unit cccle, OF class C! means that the functions 74(0, 15(0 have continuous derivatives in the ordinary sense of calculus. We have drawn 86 UII $1] HoLomorPiic FUNCHIONS ON CONNECTED sets a7 ‘curve in Fig. 1. Thus a cure isa parametrized curve, We call (a) the beginning point, and (0) the end point of the curve. By a point on che carve we mean'a point w such that w = 9() for some rin the interval of Sefnition of We define the derivative (9 inthe obvious way, namely 7O= n+ G0 1 is easily verifed as usual that the rules for the derivative of a sum, product, quotient, and chain rule are valid in this case, and we lave this as an exercise. Io fat, prove systematically the folowing statements Let F:[a,b]-+C and G:(e,b]-»C be complex valued diferentable functions, defined on the same interval. Then F+oy-F +a, (FG) = Fo" + FG, [63 = (GF ~ FoyG* (this quotient rule being valid only om the set where G(0) # 0} “Let ¥:(e,d]- [a5] be a diferentiable function. Then yy is die cntiabe, and evi = WOW. as illustrated on Fig. 260. metre (2) Figure 26) Finally suppose is a curve in an open set U and Use is holomorphic function. Then the composite fy is dire 4 function of the real variable ¢) and Fono=FGon'e 8 ‘CAUCHY’s THEOREM, FIRST PART om, 81) as ilstrated on Fi. 20 ros Figure 20) It is technically convenient to deat with a generalization of curves. By 4 path we shall mean a sequence of curves, } (60 each curve 7 is C*) such that the end point of 7 is equal to the beginning point Of yy,» If i defined on the interval (2,8), this means that fo) We have drawn a path on Fig. 3, where =; is the end point of 3. We call (a, the beginning point of y, and y(t) the end point of 7. The path is sad to ie im an open set U if each curvey, lis in U, ie. foreach f, the Point 1 es in U. Heya We define an open set U to be connected if given two points « and fi in U, there exists path {7,,..-%e} in U such that a is the beginning point of 7 and ff isthe end point of y; in other words, if there exists & path in U which joins « to f. In Fig 4 we have drawn an open set Which is not connected. In Fig. 5 we have dravin a connected open set (11 $1 HoLomoneutc FuNcTIONS ON CONNECTED SETS 9 (The definition of connected applies of course equally well to a set which is mot necessarily open. It is usually called pathwise connected, but for ‘open ses, this coincides with another possible definition. See the appen- 0 contained in U. ‘Then every tlement of this disc can be joined to 2, by a line segment inthe disc, and can therefore be joined to zy by a pth in U, so V is open. We assert further that V i closed. To Se this, let {2,} bea sequence in V convers- ing to a point w in U. Since U is open, there exists 2 dise Dlu,”) of, radius r>0 contained in U. For some n the point 2, lies in Dur. ‘Then there isa line segment in D(u,) joining w and z,, and so u ean be joined by a path to zp. This proves that V is closed. Hence Vis both ‘open and closed, and by assumption, V= U. ‘This proves that U is pathwise connected, and concludes the proof of Theorem 1.6. Warning, ‘The equivalence of the two notions of connectedness for fopen sets may not be valid for other types of sets. For instance, consider the set consisting of the horizontal positive xaxis, together with vertical Segments of length 1 above the points 1, 1/2, 1, Ifa. and also above 0. Now delete the origin. The remaining set is topologically ‘connected but not pathwise connected, Draw the picture! Also compare ‘with inaccessible points as in Chapter X, 4 Ml, §1. EXERCISES 1, Prove Lemme 1, 2 Let U be a bounded open connected set, (f,} a sequence of continuous fanetions on the elesus of U, analytic on U. Assume tht (f.) converges tfotly on the boundary of U. Prove that (f, converses wiformly on U 3, Let oy, dy be points on the nit cite. Prove that there exists a point: on the ani Gleao thatthe product ofthe distances fom 70 theo atleast, 1 (vou may ue the maximum principle) 9 ‘CAUCHY’S THEOREM, FIRST PART cm, 92) Ul, §2. INTEGRALS OVER PATHS Let F: (a,b) +€ be 2 continuous function, ‘Write F in terms ofits real and imaginary parts, say FI = (+ il. Deine the indefinite integral by Verify that integration by parts ie valid (assuming that F* and G" exist and are continuous), namely (The proof is the same as in ordinary calculus, from the derivative of a product) ‘We define the integral of F over [a,b] to be [rose Pan ares[ aaa ‘Thus the integral is defined in terms of the ordinary integrals of the real functions w and v. Consequently, by the fundamental theorem of calculus the fanetion mffrve 1 diferemiable, and its derivative is F(0), because this assertion is true if wwe replace F by w and o, respectively. Using simple properties of the integral of real-valued functions, one wien ['r04 2 Work it out as Exercise 11 Let f be a continuous function on an open set U, and suppose that is a curve in U, meaning that all values 9() lie in U for ast sb. We [lene cu, 32) INTEGRALS OVER PATHS 9s efine the tegral of {along 7 to be i Stoo ae. ‘We want to find the value ofthe integral off over the circle, fie 50.05 6S 2x. By definition, this integral is equal to [ober fare ‘Asin calculus, we have defined the integral over parametrized curves. In practice, we tometimes describe a curve without giving an explicit parametrization. The context should always make it dear what is meant. Furthermore, one can also easily see thatthe integral is independent of the parametrization, in the following manner Let 6:0, 81+Le,d be a C? function, such that g(a) =, of) d and let eledoc ners oie) aaa seve, 96 (CAUCHY'S THEOREM, FIRST PART cm, 2 tea curve. Then we may form the composed curve 0 = Hol). We find [r- [oor f seommaereroe [sore fe ‘Thus the integral of f along the curve is independent of the parametrza- (srt) i @ path, then we define fe {be the sum of the integrals of f over each curve 1 ofthe path ‘Theorem 2.1. Let f be continuous on an open set U, and suppose that f ‘has a primitive 9, that is, gis holomorphic and g' = f. Let & B be two points of U, and lt'y be a path in U Jotning a to B. Then fs-a0 ae and in particular this itera depends only onthe beainning ond end ein ofthe path: Tes Independent ofthe path Prot Asume fst hat the path carve. Then froe=[ eooroa 3y the chain rue, he expreion under theirs i the derivate Sata. a 2 INTEGRALS OVER PAT ” Hence by ordinary callus, the integral is equal to (vt) = of) ~ afore), which proves the theorem in this case. In general ifthe path consists of curves 7,5 sty and 2; the end point of , then by the case we have just settled, we find fe = oles) ote) + olen) — ates) + + alta) —olea-id = ale.)~ otto Which proves the theorem, Example 2. Let fle) = 2%. Then f has a primitive gf) = 24/4 Hence the integral off ftom 2-31 to 1 ~i over any path is equal to a of +s aa Example 3. Let fle) =e. Find the integral of f from 1 to in taken cover a line segment. Here again f'(2) = fle so f has a primitive. Thus the integral is independent ofthe path and equal toe" —e! = —1—e By 2 dosed path, we mean 2 path whose beginning point is equal 10 its end point. We may now give an important example ofthe theorem If 18 a continous function on U admitting a holomorphic primitive g, ‘nd y 1s any closed path in U, then J Ts fiber se nm ge = Try aa a es ew cet [es ‘This ig true because 2* has the primitive "Yn + 1). [When m is nega tive, we have (o assume that the closed path does not pass through the fovigin, because the function is then not defined at the origin] 98 ‘CAUCHY'S THEOREM, FIRST PART cm, s21 Puiting this together with Example 1, we have the following tabula- tion. Let Cy be the circle of radius centered at the origin oriented ‘counterclockwise. Let » be an integer. Then: [preety fc OF course, in Example 1 we did the computation when R=, but you ‘can check that one gets the same value for arbitrary R. In the exercises, you can check similar values for the integral around a circle centered Around any point 2 ‘We shall sce later that holomorphic functions are analytic. In that cas, in the domain of convergence & power series Ease — zo)" ‘can be intgrated term by term, and thus integrals of holomorphic func- tions are reduced to integrals of polynomials. This is the reason why there is no need hereto give further examples. ‘Theorem 22. Let U be a connected open set, and let f be a continuous function on U. If the intepral of f along any closed path in U is equal to 0, then f has a primitive g on U, that is, a function g which is holomorphic such that g' = J. here the integral is taken along any path from z9 to in U. If 7, 9 ae {wo such paths, and 1” is the reverse path of (ck Exercise 9}, then {ron} isa closed path, and by Exercise 9 we know that Dele ‘Therefore the integral defining gis independent ofthe path from zo to =, and defines the function. We have gle +W~ ote) 7h sO ae, tm, 2) INTEORALS OVER PATHS ° and the integral from 2 to 2+ h ean be taken along. segment in U from Florth Write LO) = fe) + 00), sere lim of) = (this can be done by the contnsty of fats} Then robe [woe ‘The length of the interval from z to z +h is [hl Hence the integral on the right is estimated by (se below, Theorem 23) t tm max otC jl maxlo(h Where the max is taken for { on the interval. This max tends to 0 as n-+0, and this proves the theorem Remarks, The reader should recognize Theorems 2.1 and 22 as being the exact analogues for (complex) dilleentiable functions ofthe standard theorems of advanced calculus concerning the relation between the eis fence of a primitive (potential function for a vector field) and the inée- pendence of the integral (of a vector Feld) from the path, We shall sec later that 2 holomorphic function is infinitely complex diferentable, and therefore that f itself is analytic Let y be a curve, 7:[2,b] +C, assumed of dass C! as always. The speed is defined as usual t0 be fy(Oh and the length L(y is defined to be the integral ofthe speed, Loy | told (rs --ste) 8 a path, then by definition n= £0 Let f be @ bounded function on a set S. We let I be the sup norm, 100 CAUCHY’s THEOREM, FIRST PART com, 62] written If ifthe reference to $ needs to be made for clarity, so that wa ps 1s the least upper bound of the values [/(2] for z € S. Let f be continuous on an open set U. By standard results of eemen- tary real analysis, Theorem 43 of Chapter [, the image of a curve or a path 7 is closed and bounded, ie. compact. Ifthe curve isin U, then the function ee s610) is continuous, and hence f is bounded on the image of y. By the ‘compactness of the image of 7, we ean always find an open subset of U ‘containing y, on which f 8 bounded. If is defined on [a 8], we let U7, = max Lf10). Theorem 23. Let {be continuous function on U. Let y be a path in I[-| Proof. Uy is curve, then 4 ULL6. [tooo =f vowntrenee SILO), 1s was to be shown, The statement for a path follows by taking an appropriate sum. ‘Theorem 24 Let {f,} be a sequence of continuous fictions on U, converging uniformly 0 a function f. Then tin f n= [ Eis a series of continuous functions converging uniformly on U, cm, 92) INTEGRALS OVER PATHS 101 then Proof. The frst assertion is immediate from the inequality [[4-[ele[u-nen-neo ‘The second follows from the fist because uniform convergence of series is defined in terms of the uniform convergence ofits partial sums, See Site the proves the theorem. Example 5 Let f be analytic on an open set containing the closed ise B(O,R) of radius R centered at the origi, except possibly at the origin. Suppose f has a power series expansion fle) = SE hays tae possibly with negative terms, such that the series with nom-negatve terms £ has a radius of convergence > R. Let Cy be the circle of radius R centered atthe orgin. Then “This isa special case of Theorem 24 and Example 4, by letting Esch f, isa finite sum, 50 the integral off, isthe sum of the integrals of the individual terms, which were evaluated in Example 4 02 CCAUCHY’S THEOREM, FIRST PART com, $2] I, §2. EXERCISES 1. (2) Given an artitary point zy, let C be a cite of rads r > O centered at 2, oBented counterclockwise. Find the integral forall integers pote or negative (01 Suppor f has « power sees expansion se= ae) hich ie absoltely convergent on ¢ dis of radius > R centered at 2 Tet Cy be the circle of radios R centered at zy Find the integral [roe 2 Find the integral of fl) =e" from —3 to 3 taken along a semicircle. 1 this integral eieret from the integral taken over the ne segment between the two pots? 3. Sketch the following curves with O15 1 @ 0=1+i © 10 © 10 @ Walser 4. Find the integral of each one of the following functions over each one ofthe curves in Exe 3. @ feos? © feo=2 © Fea sr fore (0) om the print he pnt 14 2, en lng aight te mp 09 Fete eta y= «ret eat [acre {rom the origin tothe point 1+, taken along the parabola om, 2 INTEORALS OVER PATHS 103 7 Late bea vera segment, say parametried by ainsthe, -1stsh, where fy 62 xed complex number, and cis» fixed real number > 0. (Draw the Piet) Let a 29-4 x anda” = 25 ~ x, Where xis eal postive. Find ts (Gaede (Draw the picture) Warning: The answer isnot 0! Let x> 0. Find the limit: we [Gata 9, Let yi fat]-+€ bea curve. Define the reverse or opposite curve to be ville such that YW) = a+ b= A Show that 10, Let (8) and (<4) be two intervals (got reduced {0 a point) Show that there fea fonction g()=rt +s such tat g is sly increasing, oe) = ¢ and (90) =a. Thus a curve canbe parametrized by any given interval. 11, Let F be continous complex-valued funtion on the interval [2,8 Prove [eal fio tin: Let P= [0 Gory sty 8) be a pation of (2,8) From the fenton of integrals with Riemann sums, the integral [Pros esi yet en Eri whenever manne, — 0 is sail and Cena semi y Ets “The proof sconce by sing the tian inequality] 108 CCAUCHY’S THEOREM, FIRST PART cm, $31 Ml, §3. LOCAL PRIMITIVE FOR A HOLOMORPHIC FUNCTION Let U be 2 connected open set, and let f be holomorphic on U. Let ze€U. We want to define a primitive for f on some open disc centered at to, ie. locally at, The natural way i to define such a primitive by an integral, wo [08 taken along some path from zp to 2 However, the integral may depend fn the path Te turns out that we may define g locally by using only a special type of path. Indeed, suppose U is a dise centered at zy. Let 26U. We Select for a path from zp to = the edges of a rectangle as shown on Fig 8 Figure 8 ‘We then have restricted our choice of path to two possible choices as shoven. We shall see that we get the samme value for the integrals in the to cases. It will be shown afterwards that the inlegral then gives us @ primitive By a rectangle R we shall mean a rectangle whose sides are vertical or horizontal, and R is meant as the set of points inside and on the bound ary of the rectangle, so R is assumed to be closed. The path describing the boundary of the rectangle taken counterclockwise will be also called the boundary of the rectangle, and will be denoted by ar, If is an arbitrary set of points, we say that a function f is holo- ‘morphic on $ iF tis holomorphic on some open Set containing 3. LL, $3] LocaL privirive FoR A HoLoMORPHIC FUNCTION 105, Theorem 3.1 (Goursa). Let R be a rectangle, and let f be a futon Jolomorpie ot Re Then free ‘Proof. Decompose the rectangle into four rectangles by bisecting the sides, as shown on Fig. 8 Then Consequently, and there is one rectangle, say R", among Ry, Rp, Ry, Ry such that If_.leélf.+4 [Next we decompose Rito four rectangles, again bisecting the sides of R' as shown on Fig. 10. FFor one of the four rectangles thus obtained, say 20), we have the st at ded 106 CAUCHY'S THEOREM, FIRST PART cm, 91 Figure 10 ‘We continue inthis way, {0 obtain a sequence of rectangles ROD RMD RD f../l8 Kdedla (On the other hand, let Ly be the length of 2R™. Then such that ‘Then 0 that by induction, 4 plo where Lo = length of 2 ‘We contend that the intersection Aro consists of a single point zp. Since the diameter of R tends to 0 a8 n Becomes arg, itis immediate tat there iat most one pein inthe inersecion. Leta, be the center of R Then the sequee {o,) a LIT, $3] Loca prwerive FoR A HoLOMORPHIC FUNCTION 107 Cauchy sequence, because given <, let N be such that the diameter of ROMs Tess than €. In, m ZN, then dy cy fe in RO" and 50 ley — al Siam RO” <6 Let 29 =lim 2. Then to His in each rectangle, because each rectangle is closed, Hence zp lies'in the intersection of the rectangles R™ for N= 1,2, a8 desired. (See also Theorem 42 of Chapter I) Since "i diferentiable at zo, there is a disc V centered at zo such that forall se V we have Je) = + Fleahe ~ te) + ~ ze), where [tose f, smaseseafensva By Example 4 of §2, we know thatthe fist two integrals on the right of this equality sign are 0. Hence [=f vonne and we obtain the inequalities Alleles 1 sunlit) Ato diam R° supthe t e Where the sup i taken for all 26.84, But diam R= (1/2) diam R. Thea fu Sly diam R suplhe) 108 CCAUCHY’S THEOREM, FIRST PART cm, $81 Seer ee ues We carry out the program outlined at the beginning of the section to find a primitive locally a5 was to be shown. ‘Theorem 32. Let U be a disc centered at a point rq. Let f be cominuous on U, and assume that for each recangle R contained in U ve he Jon For each point 2, inthe dis, define whee the incral is taken long the sides of « rectangle R wtose Srposte tries ae rad. Thongs flonerpic U and isa tive for nel owas Prof. We have ate, += ate [pana Figure 11 UL, $3] LocAL PRIMITIVE FOR A HOLOMORPHIC FUNCTION 108 ‘The integral between 2; and 2 +I is taken over the bottom side hy and vertical side hy of the rectangle shown in Fig. Il, Since fis continuous at 2), there exists a function (2) such that and Ses) + ve) yf ae We divide by h and take the limit as hr0. The length of the path from 43, to- +h is bounded by [hi] + [isk Hence we get a bound h ‘where the sup i taken for z on the path of integration. ‘The expression fon the right therefore tends to 0 as hn+0. Hence alent = ol) im A IE = ey), as was to be shown Knowing that a primitive for f exists on 2 dsc U centered at Z, we can now conclude thatthe integral off along any path between z and 2 in U is independent of the path, according to Theorem 21, and we find: ‘Theorem 33. Let U be a disc and suppose that fis holomorphic on U. ‘Then f has a primitive on U, and the imegral of f along any closed path in U is 0. Remark, In Theorem 72 we shall prove that a holomorphic fonction js analytic Applying this result to the function g in Theorem 32, we Shall conclude that the function f in Theorem 32 is analytic. See Theo- rem 77. 10 CCAUCHY’S THEOREM, FIRST PART cm, 641 Il, §4, ANOTHER DESCRIPTION OF THE INTEGRAL ALONG A PATH Knowing the existence of @ local primitive for a holomorphic function allows us to describe its integral along a path in a way which makes no use ofthe diferentability of the path, and would apply to a continuous path as well, We start with curves Lemma 41. Let 7: [0,8] + U be a continuous curve nan open set U. Then there i some positive number r> 0 such that every point on the curoe lies at distance ™ r from the complement of U. Figure 12 Proof. The image of is compact. Consider the function oo ni 0) — wh ‘where the minimum is taken for all w in the complement of U. This minimum exists because it sufces to consider w lying inside some big, rele, Then 9(0) is easly verified 10 be a continuous function of ‘whence @ has a minimum on [2,6], and this minimum cannot be 0 because U is open. ‘This proves our assertion. Let P = [2o,.--1] be a partition of the interval [a 8]. We also write Pin the form ao $4 Sa; S-" Sa, Let {Do,.o-sDy) be a sequence of discs. We shall say that this sequence of discs is connected by the curve along the partition if D, contains the image y({a, a45}) The following figure illustrates this ‘One can always find a partition and such a connected sequence of ites. Indeed, let ¢ > 0 be a postive number such that € 0. Let) =e, What isthe image UD Is IU spy connected? Ill, §6. EXISTENCE OF GLOBAL PRIMITIVES. DEFINITION OF THE LOGARITHM In §3 we constructed locally a primitive for a holomorphic function by integrating We now have the means of constructing primitives for & ‘much wider class of open sets. ‘Theorem 6.1. Let be holomorphic on a simply connected open set U. Let 2€ U. For any point 2€U the integral ‘is independent of the path in U from 20 to 2, and g is a primitive for f, Proof. Let jus 7p be two paths in U from %— to 2 Let yF be the reverse path of fy, from # to 29. Then y= tn) is a closed path, and by the fist form of Cauchy's theorem, freleele Since the integral of f over ny is the negative of the integral of f over rh, we have proved the first assertion. "As to the second, to prove the diferentiability of 9 at point 2, if2 is near z,, then we may select a path from zo to 7 by passing through i that is ata mate ff 120 CAUCHY’s THEOREM, FIRST PART com, $6) snd we have slteady seen that cis latter integral defines a local primitive for f in a neighborhood of 21. Hence e@)=Se) 8 desired, Example. Let U be the plane from which @ ray starting from the ‘origin has been deleted. Then U is simply connected. Proof. Let 7 be any closed path in U. Por simplicity, suppose the ray is the negative x-axis, a5 on Fig. 17. Then the path may be described in terms of polar coordinates, wO= rine, astsb with x <0( 0 Such that the closed disc D(z, R) centered at zy and of radius R is Contained in U. We are therefore reduced to proving the following theo- fem, which will give us even mote information concerning the power series expansion off at Zo 128 CAUCHY’S THEOREM, FIRST PART om, $71 ‘Theorem 73. Let f be holomorphic on a closed dise Dito, R), R>0. Let Cy be the circle bounding the disc. Then f has a power series ‘expansion fle) = Yase~ 20)" whose coefciens a, are given by the fermula 1 _ £0 |= er Furthermore, if fx denotes the sup norm of f on the circle Cy, then wwe have the emiate Jeg) S U/Uad RY {In particular, the radius of convergence of the series is & R. Proof. By Theorem 71, for all inside the circle Cy, we have fe) Let 0.c. Thie i true for every >, and hence fis holomorphic for Rez>t. In the same example, we have Sie) = “28 By Theorem 1.2, it follows that in this same region, V, §1. EXERCISES 1. Let f be analytic on an open set U, let 24 €U and fz) #0. Show that Fes” Lara wee Cis smal ie ered Weierstrass theorem for areal interval [4,0] states that continuous fue tion an be uniformly approximated by polynomials. Ts this conclusion stil teu forthe closed sit ds, Le can ever) continuous funtion onthe dis be Uunormlyapproximited by polynomials? Let a> 0, Show that each ofthe following series represents a holomorphic function co Sem ornerso; i) Licew ta wo Schy orteso hor 4. Show that each of the two series converses uniformly on each closed dive bisewihO ey +e for ere 20. Show thatthe sees defies an analytic fnction in this half pane ‘The number o is called the abscis of comergence ‘The next exercises give expressions and estimates fr an analytic fnetion in terms of integral 1. Lat f be analytic on closed dite D of radius 6>0, centered at 2p, Show that Jeffjaerintsre tus (Hie: Use polar coordinates and Cauchy formula, Without loss of gen- eraiy, you may asume that z= 0. Why) Let D be the unit dis and let S be the unit square, that i, the st of ‘complex numbers = such that O-) (v2) LAURENT SERIES 18 ‘We can then integrate term by term, and the desired expansion falls out To show the uniqueness of the coefcients, integrate the series 37 ante" against ¢~" for a given integer k, term by term from 0 to 2x All terms op out except for n= k, showing that the kth coefficient i determined by f. ‘An example of 2 function with a Laurent series with infinitely many negative terms is given by e', that i, by substituting I/: in the ordinary exponential series, fan annulus is centered at @ point zo, then one obtains a Laurent series at 29 ofthe form (On the other hand, suppose we want the Laurent series for |e] > 1. Then ‘whence §2. EXERCISES 1, Prove that the Laurent series can be difrntated term by tem in the wsual Iranner to give the derivative of fon the ann 2 Let f be holomorphic on the annulus A, defined by 03 () emeve ou Wel Z me neety trieetet Seore—ir weet 1. Expand (a) cas, (sin in 8 power sre about 22 Fi the Laurent sees or J: (a) the die [<1 (©) In the annulus 1 tel <2. (@) To the eon 2 < [et 9. Find the Lawent series for (£4 INle—1) in the region (@) [el<1s @ri>t 10. Find the Laurent series for 124 ~2) inthe regions: (eke he ll> 1 1H. ind the power sere expansion of 1 fonts round the point = and fd the rads of convergence of this series va ISOLATED SINGULARITIES 16s 12, Find the Laurent expansion of 10°F fort 0, (Hit: Show that there isan snalytie function f* fn 8 dis from which the origin is deeted such that ren fe. “What i the Laurent series for f°7 Abbrevite a = 115, Assumptions being a8 in Exercie 14, sppose in adton that here ents ju>0 toch that J(=)= flx + 19) 16 Bounded in the domain 2 Jp. Prove that the coeffens e, ae eal to O for mc Ts the converse tre? Proot? V, §3. ISOLATED SINGULARITIES Let zo be a complex number and let D be an open dise centered at zo, [Let U be the open set obtained by removing 2 from D. A function f ‘which is analytic on U is said to have an isolated singularity at 29. We suppose this isthe case. Removable Singularities ‘Theorem 31. If f & bounded in some neighborhood of 2, then one can define flze) in a unique way such thatthe function is also analytic at 166 APPLICATIONS OF CAUCHY'S INTEORAL FORMULA EV, 63] By §2, we know that f has a Laurent expansion Hd ~ Boat Saat Proof. Say 2 for OO and let N > 2R, Write S10) = gfe) + Med, 168 _APPLIcATIONS oF CAUCHY’S INTEGRAL FoRMULA — [V, $3] ‘Then g is a rational function, and is therefore meromorphic on C. Furthermore, fom its expression as a finite sum, we see that g has simple poles atthe integers n such that In] =. Fr the infinite series defining #, we apply Theorem 1.1 and prove that ‘he series is uniformly convergent. For [2] < R we have the estimate ‘The denominator satisfies 1 (Rm? 22 for n>N>2R. Hence 2 | aR stals$e for wean ‘Therefore the series for h converges uniformly in the disc [2] < R, and h is holomorphic in this dise. This proves the desired assertion. Essential Singuaries I the Laurent series has 2 infinite number of negative terms, then we say ‘that 2g is an essential singularity off. Example. The function fe) = e™* has an essential singularity at because its Laurent series is ‘Theorem 32 (Casorati-Welersras). Let O be an escential singularity of the function f, and let D be a disc centered at O on which f is Ilomorphic excépe at 0. Let U be the complement of O in D. Then {§{U) 15 dense inthe complex munbers. In other words, he values of f ‘on U come arbirarly close to any complex umber. Proof. Suppose the theorem is false. There exists complex number « and a positive number s > 0 such that Ue)—a)>s forall re U, ws ISOLATED SINGULARITIES 169 ‘The funetion ate) Te is then holomorphic on U, and bounded on the die D, Hence 0 is a removable singularity of g, and 9 may be extended to a holomorphic function on all of D. It then follows that I/a(2) has at most pole at 0, ‘which means that f(z) a has at most a pole, contradicting the hypothe sis that f(2) has an essential singularity (infinitely many terms of negative ‘order in its Laurent series) This proves the theorem. Actually, it was proved by Picard that f not only comes arbitrarily close to every complex number, but takes on every complex value except ssibly one, The function e* omits the value 0, so it is necessary t0 low for this one omission. See Chapter XI, 3 and Chapter Xl, §2 ‘We recall that an analytic isomorphism f09V from one open set to another is an analytic function such that there exists another analytic hinetion oVau satistying fea ly» and ge. idg, where id is the identity function, An analytic automorphism of U is an analyte isomorphism of U with isl Using the Casorati-Weierstrass theorem, we shall prove: ‘Theorem 33. The only analytic automorphisms of C are the functions (of the form fle) = a2 +b, where a,b ae constants, 0 0. Proof. Let J be an analytic automorphism of C. After making 2 translation by —f(O, we may assume without loss of generality that ‘fO)=0. We then have to prove that fle) = ar for some constant & Let e)= file) for = #0. ‘Then h is defined for all complex numbers excet for the origin. We fst prove that h cannot have an essential singularity atO. Since f i a local “analytic isomorphism at 0, f gives bijection between an open neighbor hood of 0 with some open neighborhood of 0. Since fis also an analytic 170 APPLICATIONS OF CAUCHY'S INTEGRAL FoRMULA LV, $3] ‘somorphisin of C, it follows that there exists 8 > 0 and ¢ > O such that if Il > 1/8 then [fiw] >c. Let 2=Ifw or w= iz. Then [h()| >e for Is|<4. If 0 is an essential singularity, this contradicts the Casorati- Weierstrass theorem, Let /(2) = 5° aye" $0 A(z) =D aa(1/2)". Since 0 is not an essential sSingulaity of hy it foliows that ihe series for h, hence for J, has only a finite number of terms, and Sl) mag aye bboy ia polynomial of degree W for some Nf has two distinct roots, then J cannot be injective, contradicting the fet that J hasan inverse ane tion Hence fas only one root and fe)= a2" for some zp. ITN > I, iti then clear that fis not injective so we must hhave N= I, and the theorem is proved, V, §3. EXERCISES 1. Show that the following sree define a meromorphic Funston on C and dtermine the st of poles, and ther orders pcr z 1 © Zara Laree © Sats Or £ 2 Show that the function “ 0° Sarre 's dened and continuous forthe rel values of. Determine the region of| the complespline in which this fonction is analy. Determine its poles iG) eines a analytic fneion on a dis of rads 1 centered at 5. Show that the series 4. Let {a} bea sequence of distinct compler numbers sch that Lv. $3) ISOLATED SINGULARITIES m rove thatthe series Elea-a) oe mewn no Wha ps ths ton? 5 Let f be meromorphic on C but not entire. Leto) not teromorphie on C tals 6 Let f be a non-constant entze funtion, i. function analytic on al fC. Show tht the fnage off dense in © 1. Let f be meromorphic on an open st U. Let vou te an analyte isomorphism. Suppose za) = vps and f has order at Show that Fe has order » at ry. In other words, the order i invariant ‘under analytic omorphisms. (Here mi posite or negate integer) 8A meromorphic function f is said 10 be peradic with period w it Sen) =fle) for al ze. Letfbe 2 meromotpie function, dnd suppose is period with tree peiods ws, yy wich are nearly dependent over the rational numer. Prove thai fis constant. (dit: Prove that thee est ‘laments w which ae integral near combinations of m, wy, m3 nd ata ‘sal in absolute value) The exponential funtion it an example of singly periodic fonction. Examples of duly period fencions wil be given in Chapter XI. 9. Let f be meromorpic on C, and suppose i Uae Prove that i rational fenton. (You cannot assume as given that f has ‘only a fire namber of poles) 10. (Mie Riemann Sphere). Let S be the union of C and a singe other point Senoted by co, and called init. Let [bea unetion on Let t= Hf and ‘stine on = 09 for 0, ca, We say that f hasan elated singularity (espe meromorphic ‘esp holomorphic) at iat fg has an slated singly (sp is me ‘morphi, esp s holomorphic) at 0. The order of g atO il also be called the erder off at ia. Ig has 2 removable singuasity at 0, and so can be sine asa lomorphi function ina acighborhood ef 0, then we ay that fs bolomerphi titi We say that f is meromorphic on 5 iff i meromorphic on C and it ao meromorphic at infinity. We say tht f ie holomorphic on S Hf i hoo ‘morphic on C and also holomorphic at afin 172 apptications oF caucuy's ivreonat. rormuta —[¥, §3] Prove: The only meromorphic fnctions on S ore the rational functions, tha 5, quo tients of polynomials The only holomorphic functions on are the conta If ts halmorphe on C ond has @ pole efit, then f Ye polyno In this lest ease, how would you deseibe the order of fat infinity in terms ofthe polynomial? 1, Let f be a meromorphic function on the Riemann sphere, s0 4 rational function by Exercise 10. Prove that Lord, s=0, luhere the sum is taken over ll points P which are either points of C, oF Prove that thee exists & meromorphic function f on the Riemann sphere sich that ord fm for tor and ord, f= 011 # A, CHAPTER VI Calculus of Residues We have established all the theorems needed to compute integrals of analytic functions in terms of their power series expansions. We frst five the general statements covering this situation, and then apply them to examples. Vi, §1. THE RESIDUE FORMULA Let soe § aor have & Laurent expansion at a point zp. We eal a oy and write the residue of fat ay eRe, f Theorem I. Let 29 be an isolated singularity of f, and let C be a small circle oriented counterclockwise, centered at 29 such that f is holomorphic on C and its interior, except possibly at zy. Then J forde = ania, = 20 Re, roo Sine the re fr f() converges eft and abcely for Gon the circle, we may iniegrate it term by term. The integral of 1 m4 ‘CALCULUS OF RESIDUES iit] (— 20) over the circle is equal to 0 for all values of m except possibly When m= —1, in which case we know that the value is 2ni, cf: Examples 1 and 4 of Chapter Ill 2. This proves the theorem. From this local result, we may then deduce a global result for more ‘eneral paths, by using the reduction of Theorem 24, Chapter 1V. ‘Theorem 1.2 (Residue Formula). Let U be an open set, and y a closed chain bs U such that y is homologous to 0 in U. Let f be analytic on U except ata finite number of points 21,2. Let m= Wz). Then [rn2sATi ms Res, f Proof. Immediate by plugging Theorem 1.1 in the above mentioned theorem of Chapter IV. ‘Theorem 1.2 is used most often when U is simply connected, in which case every closed path is homologous to 0 in U, and the hypothesis on y need not be mentioned expicily. In the applications, U will be a disc or the inside ofa rectangle, where the simple connectedness is obvious. Remark. The notation «/=T is the standard device used when we don't want to confuse the complex number {with an index We shall give examples how to find residues. [A pole ofa function f is said to be simple i itis of order 1, in which case the power series expansion off is of type $0) = 8+ 9+ higher terms, and a #0. Lemma 13. (2) Let f have a simple pole at to, and let g be holomorphic at 2 Then Re (0) = 20) Res, (0) Suppose flzs)=0 but fee) #0. Then I/f has a pole of order 1 fat 25 and the residue of Uf af 2 i8 I/F (2) (vist) ‘THE RESIDUE FORMULA 195 Proof. (a) Let fle) = a.,/{2~ 29) + higher terms, Say z6 = 0 for sim plicty of notation. We have Slat) (Ge )eorne eo Po nigher terms, $0 our assertion is dear. () Let fiz) =0 but fz) #0. Then fl) terms, and a, #0. Say 29 =0 for simplicity. Then le — 29) + Bigher fle) =ayx1+h) with od hz 1, Soha erm a 0 restlif)= = Uf"), as was to be shown Remark. Part (a) of the lemma merely repeats what you should have seen before, to make this chapter more systematic. Example. We give an example for part (b) ofthe lemma Let fle) = sin, Then f has a simple zero at z= x, because J'(2)= cos? and Fn) = 140. Hence Ife) has a simple pole at 2 = x, and Example, Find the residue of f= > To do this, we write $0 e EH Note that g()= s%A2-1) is holomorphic at 1, and that the residve of Ife Dis. Hence Res, f= ot 2. 176 ‘CALCULUS oF RESIDUES ris Example, Find the residue of (in s/s? at ++ higher terms We have Hence the desired residue is 1 eres t= Example. Find the residue off ‘We note thatthe function oye is holomorphic at == 1, and has an expansion of ype (2) = be + ble — 1) + higher terms. Then oe) to SOE and therefore the residue of f at 1 is by, which we must now find. We Dy soa L42e- nee FT" 2+ he= Inverting by the goometric series gives Ze{(iefe—o-) Therefore 1 ae tei * whence Res, f= 3 (vist) THE RESIDUE FORMULA i” Example. Let C be a circle centered at 1, of radius 1, Let wine fs he in ny ei 1 I ee maces ey meat ane et ican a a ea oma set [raed Figwe 1 IC is the boundary of the rectangle as shown on Fig 2, then we also find Sn tri-h Figure 2 1% ‘CALCULUS oF nestDUES (v1, 611 Example. Let f{e) = 2 — 22 +3 Let C be @ rectangle as shown on Fig 3, oriented clockwise. Find 1 te lf) ein? wt ‘The roots of fte) found by the quadratic formula to be 22 J8 2 = i/B. The rectangle goes around these two points, in the lockwise direction, The residue of I/fte) at, 1 {ley ~ 23) because f has a simple pole at zy. The residue of 1/flz) at zs is 1A; — 2) forthe same reason. The desied integral is equal to ~2rfeum of the residues) Example. Let f be the same function asin the preceding example, but now find the integral of 1/f over the rectangle as shown on Fig 4. The rectangle is oriented clockwise. In this case, we have seen that the rei- due at | 12's ‘Therefore the integral over the rectangle is equal to a ~ Preside) anit 21 v1.6) THE RESIDUE FORMULA 119 Figure & Next we give an example which has theoretical significance, besides ‘computational significance Example. Let f have a power series expansion with only a. finite number of negative terms (so st most @ pole) say a the origin, See mat + higher terms, a4 #0, and m may be positive or negative, Then we can write Se) M(H, where Me) 6 @ power series with zero constant term. For any two Functions fg we know the derivative ofthe product, (fo ‘9+ fas so that dividing by fo yields and H(2y(t + H() is holomorphic at 0. Consequently, we get: 180 ‘CALCULUS oF RESIDUES (vist Lemma LA, Let f be meromorphic at 0. Then Rese ff = ord f and for any point za where f has at most « pole, Res, ff = Orb, f ‘Theorem 15. Let» be a closed chain in U, homologous to 0 in U. Let Sf be meromorphic on U, with only a finite munber of zeros and pole, Say at the pois 25 vty tone of which le on 7. Let m= W029 Then In applications, is frequently equal to @ circle C, or a rectangle, and the points 2, wry At inside C. Suppose that the zetos of f inside C and the poles are ‘Then in the ease, [sp aeJ=A(E ont — Fm, ) We follow our convention whereby the multiplicity of a pole isthe nega tive ofthe order of fat the poe, s0 that sul, f= ~ordy f by definition. Ione counts zeros and poles with their multiplicities, one may te- phrase the above formula inthe more suggestive fashion: Let C be a simple closed curse, and let f be meromorphic on C and its Interior. Assume that {has no zero or pole on C. Then (een visa THE RESIDUE FORMULA ral where number of zeros = number of zeros of fin the interior of C, and umber of poles = number of poles off in the interior of C. Of course, we have not proved that a simple closed curve has an “interior”. The theorem is applied in practice only when the curve is 50 cxpliciy given (a with a tele or rectangle) that itis clear what “ine Besides, one can (not s0 artificially) formalize what is needed of the notion of “interior” so that one can use the standard language. Let y be 2 closed path. We say that 7 has an interior if (72) = 0 oF { for every complex number « which does not ie on y. Then the set of points « such that 17( 2) = fil be called the interior of y. 1's that simple. “Theorem 1.6 (Rovehé’s Theorem). Ler y bea closed path homologous 10 © in U and assume hat 7 has am interior. Let f, 9 be analytic on U, and Lye) — gta < te Jor 2 ony. Then f and g have the same number of seros in the Itrior of. Proof. Note that the assumption implies automatically that fg have: no zer0 on y. We have for zon y. Then the values of the function g/f are contained in the open dise with center I and radius I. Let F'= gf. Then Fy isa closed path contained in that dist, and therefore WFen0) because 0 lies outside the disc. Ify is defined on (a, 8) then on meen ome -[re = [ am ru ‘What we want now follows rom Theorem 1.5, as desired 182 ‘CALCULUS OF RESIDUES (vist) Example. Let P(e) = 2" — 5242-2. We want to find the number of roots ofthis polynomial inside the unit crcl. Let St itis immediate that 52, 19) = Poa = |= Ped 0 such that We) 215 KieP Let G, be the circle of radi r, and let Jal 1/2 and Kr?

You might also like