FUNCTION THEORY OF SEVERAL COMPLEX VARIABLES SECOND EDITIONFUNCTION THEORY OF SEVERAL COMPLEX VARIABLES SECOND EDITION BY STEVEN G. KRANTZ pec ey bs AMS CHELSEA PUBLISHING American Mathematical Society + Providence, Rhode Island2000 Mathematics Subject Classification, Primary 32-02, 32Bxx, 32Dxx, 32F xx, 32Hxx, 32Txx, 32Uxx, 32Wxx; Secondary 35N15. Library of Congress Cataloging-in-Publication Data Krantz, Steven G. (Steven George), 1951 Function theory of several complex variables / Steven G. Krantz.—2nd ed. p. cm, Includes bibliographical references and index. ISBN 0-8218-2724-3 (alk. paper) 1. Functions of several complex variables. I. Title. QA331.7.K74 2000 515/.94—de21 00-059363 Copying and reprinting. Individual readers of this publication, and nonprofit, libraries acting for them, are permitted to make fair use of the material, such as to copy a chapter for use in teaching or research. Permission is granted to quote brief passages from this publication in reviews, provided the customary acknowledgment. of the source is given. Republication, systematic copying, or multiple reproduction of any material in this, publication is permitted only under license from the American Mathematical Society. Requests for such permission should be addressed to the Assistant to the Publisher, n Mathematical Society, P.O. Box 6248, Providence, Rhode Island 02940-6248, ts can also be made by e-mail to reprint-permission@ans.org. © 1992 held by the American Mathematical Society. Printed with corrections, 2001. All rights reserved. The American Mathematical Society retains all rights except those granted to the United States Government. Printed in the United States of America. © The paper used in this book is acid-free and falls within the guidelines established to ensure permanence and durability. Visit the AMS home page at URL: http: //www.ams.org/ 10987654321 060504 03 0201to RandiContents Preface to the Second Edition Preface to the First Edition 0 0.1 0.2 0.3 11 1.2 1.3 1.4 1.5 An Introduction to the Subject Preliminaries 1 What is a Holomorphic Function? 4 Comparison of C! and C® 5 Some Integral Formulas The Bochner-Martinelli Formula 19 Applications of Cauchy Theory and the 0 Equation 31 Basic Properties of Harmonic Functions The Bergman Kernel 49 The Szegé and Poisson-Szegé Kernels 66 35 xiii xv 19viii 1.6 2.1 2.2 2.3 2.4 3.1 3.2 3.3 3.4 3.5 41 Contents Afterword to Chapter 1 69 EXERCISES 70 Subharmonicity and Its Applications 81 Subharmonic Functions 81 Pluriharmonic and Plurisubharmonic Functions 94 Power Series 99 Hartogs’s Theorem on Separate Analyticity 107 EXERCISES 110 Convexity 117 Many Notions of Convexity 117 Convexity and Pseudoconvexity 132 Pseudoconvexity, Plurisubharmonicity, and Analytic Discs 142 Domains of Holomorphy 151 Examples of Domains of Holomorphy and the Edge-of-the-Wedge Theorem 159 EXERCISES 164 Hormander’s Solution of the 0 Equation 175 Generalities about the 0 Problem 1754.2 4.3 44 4.6 on nm 5.2 6.1 6.2 6.3 Contents ix Unbounded Operators on a Hilbert Space 176 The Formalism of the 0 Problem 179 Some Computations 182 An Existence Theorem for the 9 Operator 190 A Regularity Theorem for the 9 Operator 194 EXERCISES 201 Solution of the Levi Problem and Other Applications of 0 Techniques 213 An Extension Problem 213 Singular Functions on Strongly Pseudoconvex Domains 216 Hefer’s Lemma and Henkin’s Integral Representation 223 Approximation Problems 228 EXERCISES 231 Cousin Problems, Cohomology, and Sheaves 243 Cousin Problems 243 Sheaves 254 Dolbeault Isomorphism 268x Contents 6.4 6.5 7A 7.2 7.3 7.4 7.5 8.1 8.2 8.3 8.4 Algebraic Properties of the Ring of Germs of Holomorphic Functions 275 Sheaf of Divisors, Chern Classes, and the Obstruction to Solving Cousin II 281 EXERCISES 286 The Zero Set of a Holomorphic Function 293 Coherent Analytic Sheaves 293 Applications of the Cartan and Oka Theorems: The Structure of Ideals 295 Zeros of One Holomorphic Function 298 Zero Sets for Different H? Spaces Are Different 302 Zero Sets of H? Functions Do Not Have a Simple Mass Distribution Characterization 307 EXERCISES 310 Some Harmonic Analysis 321 Review of the Classical Theory of H? Spaces on the Disc 321 Three Propositions about the Poisson Kernel 330 Subharmonicity, Harmonic Majorants, and Boundary Values 338 Pointwise Convergence for Harmonic Functions on Domains in RY 3448.5 8.6 8.7 8.8 9.1 9.2 9.3 10 10.1 10.2 10.3 Contents xi Boundary Values of Holomorphic Functions in Cc" 346 Admissible Convergence 352 The Lindelof Principle 362 Additional Tangential Phenomena: Lipschitz Spaces 365 EXERCISES 371 Constructive Methods 381 Story of the Inner Functions Problem 381 The Hakim/Low/Sibony Construction of Inner Functions 385 Further Results Obtained with Constructive Methods 395 EXERCISES 401 Integral Formulas for Solutions to the 0 Problem and Norm Estimates 403 The Henkin Integral Formula 403 Estimates for the Henkin Solution on Domains in C? 407 Smoothness of the Henkin Solution and the Case of Arbitrary 0-Closed Forms 411xii Contents 10.4 First Cousin Problem with Bounds and Uniform Approximation on Strongly Pseudoconvex Domains inC? 418 EXERCISES 421 11 Holomorphic Mappings and Invariant Metrics 11.1 Classical Theory of Holomorphic Mappings 429 11.2. Invariant Metrics 434 11.3 The Theorem of Bun Wong and Rosay 443 11.4 Smoothness to the Boundary of Biholomorphic Mappings 453 11.5 Concept of Finite Type 462 11.6 Complex Analytic Dynamics 476 EXERCISES 480 Appendix I Manifolds Appendix II Area Measures Appendix III Exterior Algebra Appendix IV Vectors, Covectors, and Differential Forms List of Notation Bibliography Index 429 493 495 499 503 509 515 553PREFACE TO THE SECOND EDITION It is a pleasure to take this opportunity to revise, correct, and update the original edition of Function Theory of Several Complex Variables that. was published in 1982. I find that, taking into account the developments of the last ten years, I have not fundamentally changed my view of what constitutes the foundations of this subject. Accordingly, Chapters 1-8 have been changed only to the extent of updating some references and correcting some errors. In the process, I have been able to clarify and amplify certain passages. Chapter 11 now briefly describes some of the ground-breaking work of Lempert on the Kobayashi metric. We still do not understand its full significance; but its impact on the subject is undeniable. This chapter now also includes a brief introduction to the concept of finite type and its connections with mapping theory. It concludes with a quick introduction to complex analytic dynamics. I am grateful to J. E. Forness for providing me with an outline of this latter material. The new Chapter 9 treats what I call “constructive methods.” By this I mean the solution of the inner functions problem and the attendant techniques (developed by Forstneric, Low, Rudin, and Ryll-Wojtacek) for constructing holo- morphic functions and mappings with specified characteristics. Although the initial surge of activity following the solution of the inner functions problem has subsided, I think that the long-term effects of this methodology will prove to be substantial. Although there have been exciting developments in the theory of biholo- morphic and proper mappings, this area rapidly degenerates into a zoology of specialized results and technical methodologies. The methods of Bell are still the most far-reaching in the subject; they provide a suitable introduction to mappings for a first book on several complex variables. We have, accordingly. retained that treatment in Chapter 11. We might have said more about the d-Neumann problem and Condition R, but those are treated in detail in the forthcoming volume Partial Differential Equations and Complex Analysis by this author. The d-problem is treated in Chapters 4 and 10 of the present: book. Next, a word about Chapter 1. Just as the Cauchy integral formula is the driving force in the function theory of one complex variable, so I see integral for- mulas as occupying a central position in the theory of several complex variables. Much of the history of several complex variables could be interpreted as an ef- fort to circumvent an inadequate understanding of integral formulas (indeed, we now realize that we could have solved the Levi problem a long time ago if we xiiixiv Preface to the Second Edition had had a good grip on the Cauchy-Fantappié formalism). We now have a firm understanding of both the canonical and the constructible integral formulas on. strongly pseudoconvex domains—the most important domains in the subject. Thus I believe that integral formulas deserve a prominent, indeed a motivating, position in my exposition. But many users, especially students, have found my Chapter 1 discourag- ing. The heavy use of differential forms, together with the plethora of different integral formulas, can be confusing to the neophyte. The good news is that one can skip all of Chapter 1 (except for a glance at Sections 1.1 and 1.2) on a first reading. There are only a few portions of Chapters 2 through 6 that depend directly on the Bochner-Martinelli kernel or the Bergman projection, and the reader may refer back to Chapter 1 for key ideas when they are needed. In the end, the reader will certainly want to become well acquainted with Chap- ter 1. But Chapters 0,2,3,4,5, and 6 by themselves constitute a self-contained and logically presented first course in the function theory of several complex variables. Also a word about the references: We have made every effort to provide an extensive bibliography. This should prove especially useful to those beginning work in the subject. But it would foolish to claim that it is exhaustive. Even allowing for limitations of the author, the unfortunate barriers to communication that have existed between eastern Europe and the western world for so many years (and which, fortunately, are now being removed) have the side effect that the important contributions of the Russian and other schools tend to appear to be minimized. Sincere apologies for any omissions in this wise. It is a pleasure to thank the many friends and colleagues who have helped me with this second edition. Eric Bedford and John D’Angelo have given me permission to use exercises from courses they have taught. A number of people have brought misprints and corrections in the first edition to my attention. I mention particularly Gerardo Aladro, Andrew Balas, David Barrett, Eric Bed- ford, Steven Bell, Harold Boas, Dan Burns, David Catlin, Urban Cegrell, Kevin Clancey, John D’Angelo, Guy David, Katy Diaz, Fausto Di Biase, Peter Duren, John E. Forness, Sigi Fu, Paul Gauthier, Estela Gavosto, Robert E. Greene, Reese Harvey, Xiaojun Huang, Piotr Jakdbezak, Marck Jarnicki, Morris Kalka, S. Kobayashi, Jacob Korevaar, Mark Kruelle, John M. Lee, Eric Low, Bao Lu- ong, Daowei Ma, John Millson, Alex Nagel, Marco Peloso, Wiestaw PleSniak, John Polking, Walter Rudin, Stephen Semmes, Nancy Stanton, Lee Stout, Jim Walker, Sid Webster, Jan Wiegerinck, H. H. Wu, and Chen Zhenhua. Jiye Yu contributed several suggestions for the new edition. I am grateful to Richard Laugesen, Marco Peloso, Paul Vojta, and Tomasz Wolniewicz for reading much of the new edition inch by inch and helping to correct both the mathematics and the exposition. Responsibility for all remaining errors resides, of course, entirely with me. —S. G. K.PREFACE TO THE FIRST EDITION The strong interplay between harmonic analysis and the theory of several com- plex variables (hereinafter referred to as SCV) that has come about in the last 20 years has stimulated desire for a text that introduces the classically oriented analyst to holomorphic functions on C”. This is such a text. It is addressed to those with an interest in partial differential equations, Fourier analysis, and integral operators. The organization of the material is accordingly function-theoretic. The omission of certain topics (e.g., analytic spaces, the proofs of Cartan’s theorems A and B) also reflects the prejudice toward classical analysis. We have included a detailed chapter and a half on sheaf cohomology, written strictly for the ingénue, but this material may be safely skipped with no loss in continuity. Novelties in the text include discussions of the Szeg6 and Bergman kernels, peaking functions for algebras of holomorphic functions, the Henkin integral for- mulas and estimates for the 0 equation, foliation by complex analytic varieties, the Fefferman mapping theorem, invariant metrics, the characterization of the ball by its automorphism group, the notion of admissible convergence of H? functions, the edge-of-the-wedge theorem, nonisotropic Lipschitz spaces, exten- sion phenomena, and a detailed discussion of convexity. Since this is a text, there are many examples and exercises. Analogies and nonanalogies are drawn with the one-variable theory (e.g., there are infinitely many Cauchy integral formulas in SCV, whereas there is only one in the classical one-variable theory). Several- variable phenomena are examined from the one-variable point of view (e.g., why is every open subset of C pseudoconvex?). Chapter 0 consists of a long expo- sition of the differences between one and several complex variables. Numerous open problems (e.g., the inner functions problem, the biholomorphic mapping problem, the corona problem) are considered in both the text and the exercises. Much of the material in Chapters 5 and 7 through 10 has never appeared in book form. The exercises are of two kinds. The ones in the body of the text are fairly routine. Some consist of “details left to the reader,” whereas others contain ad- ditional ideas. The Miscellaneous Exercises at the end of each chapter are highly nonuniform in character and difficulty. Some are straightforward whereas others are fairly challenging. Some exposit results from published or unpublished pa- pers (references are included). Others are results from the mathematical folklore. Difficult exercises are distinguished by an asterisk (+). In places this book is computational: We have computed the behavior of a convex domain under a biholomorphic map, computed detailed examples of xvxvi Preface to the First Edition pseudoconvex domains, constructed smoothly varying peak functions on strong pseudoconvex domains, computed numerous integral formulas, and computed estimates for the 8 operator. The purpose of these computations is to force the reader to do some work. We hope this feature will facilitate the reader’s rapid and serious involvement in the subject. Also, we wish to provide a repository for several calculations that are not readily available elsewhere. As do all books, this one makes certain demands on the reader. We assume familiarity with the elements of real variable theory, measure theory, one complex variable, and functional analysis. We adopt the custom of using the same letter to denote constants whose specific values may change from line to line (to which fact the unaccustomed reader must become acclimated). What is perhaps more significant is that this book uses the theory of differential forms consistently and from the very outset. We expect, however, only that the reader should have an acquaintance with differential forms at the level of advanced calculus. The only way to gain facility with the language of forms is to apply it, and this is the ideal subject in which to do so. Based on experience in teaching this subject, we have included appendices (which are both ad hoc and extremely terse) on the four prerequisites with which students seem to experience the greatest discomfort: manifolds, surface measure, exterior algebra, and differential forms. ‘A few comments on the organization of the book are in order. Chapters 0 through 5 constitute a basic course in the analytic parts of several complex variables theory. Chapter 6 and Sections 7.1 through 7.3 contain the rudiments of the algebraic theory. Sections 7.4, 7.5, and Chapters 8 through 10 are dessert. They contain more advanced topics in harmonic analysis, partial differential equations, differential geometry, and holomorphic mapping theory. These topics may be read selectively and in almost. any order. Many of the results in this book have been supplied with names (suggestive of the mathematicians who proved them), and many corresponding references have been given. This bibliographic work is not based on an exhaustive search of the literature and is not meant to be complete. It is meant only to help the reader begin to explore the vast number of research papers in existence. Any omissions of references or failure to attribute theorems reflects only my ignorance and not any malicious design. —Steven G. KrantzO An Introduction to the Subject 0.1 Preliminaries The following standard notation will be used: R denotes the real numbers; C denotes the complex numbers; (x1,...,2)v) denotes an element of RY; and (21y-++52n) © (@1 + ty - + tym) © (15 Yt5-++52ns Yn) is an element of C. The partial differential operators on C” given by dee Oe 2 dx; ay 9 JS1 yn, Oe son cee = 2 dn; "Ou )? ? will prove very useful. While at first perhaps puzzling in their form, these defi- nitions are forced upon us by the logical requirements that and, when j # k,2 Chapter 0: An Introduction to the Subject Likewise, we have the differentials dz; = dx; + idy;, dz; =dx;—idyj, j=1,...,n. Ce a (ass ay a nBHEHO; 62 Gee Next, since we shall deal with a great many integrals on both RN and C”, we introduce the standard Euclidean volume form on C” (in perhaps slightly unfamiliar notation): Notice that and, when j #k, dV (z) = dV,(2) (5) (day Ada) \...\ (din A den) i (3) (ide, Ady) A...A (ide, A dyn) (dey A dy1) N...A (ditn A dyn). The symbol dV is also used to denote the usual Euclidean volume on R". We shall have occasion throughout this book to use many standard concepts from the classical function theory of one complex variable: Cauchy integrals, nor- mal families, power series, and so forth will be used virtually without comment. The classic text Ahlfors [1] is a good reference for these matters. However, all several complex variables concepts will be defined and developed here from first principles. If 2° € C",r > 0, then define the open ball B(2°,r) = {ze C": —2| 0, then D(P,r) denotes the disc {z € C: |z — P| 0. Ifa, € (Z*)", then a < # means that a; < f; for all j (and similarly for a < 8). Finally, we consider differential forms on C”. For any j > 1, let a € (Z+)’ Then dz = dza, A---Adzq, and d2* = dq, \--- Adiq,. There is a similar notation for real forms. If a = (a1,...,@;) is a multi-index used to index a form (rather than a derivative), dz“ = dzq, A---Adza,, then |a| denotes j, the length of a. Context will make clear which meaning of |a| is intended. The theory of functions of one complex variable is old and well developed. Many theorems are proved on fairly arbitrary domains. The function theory of several complex variables is a much younger one; correspondingly, we are not at a stage where we can profitably study domains with pathological boundaries. In most (but not all) contexts in this book, we shall restrict attention to bounded open sets with some boundary regularity. Our terminology is as follows. A domain is a connected open set. A domain Q in R% (resp. C”) with boundary AQ is said to have C* boundary, k > 1, if there is a & times continuously differentiable function p defined on a neighborhood U of the boundary of Q such that a. QNU = {z EU: p(z) <0}; b. Vp £0 on the boundary of 2. We call the function p a C* defining function for Q. It is an exercise with the implicit function theorem to see that, for k > 1,@ has a C* defining function if and only if 0Q is a C* manifold. See Hirsch [1] for more on these matters.4 Chapter 0: An Introduction to the Subject This definition is so important that it bears some discussion. It is intuitively appealing to think of the C* boundary of a domain 2 C R™ as locally the graph of a C* function. That is, if P € 09, then we select a neighborhood Up 3 P in R% and a Euclidean coordinate system t1,...,t on Up, so that the positive t-axis points into the domain. We say that 00 is locally the graph of the C* function @p(ti,..-,tn—1) if 820 Up = {(t1,---,tnw-1, P(t... tn-1)) € Up}. Then, on Up, the role of the defining function p is played by pp(t) = ép(ti,---,t—-1) — tw. A defining function on a neighborhood of all of OQ is obtained by using a partition of unity to patch together the local defining func- tions. The concept of defining function p allows us to avoid (the somewhat ar- tificial) reference to local coordinates when describing a boundary. Along with guaranteeing that the boundary is a C* manifold, the condition that Vp # 0 im- plies that Vp will be a nontrivial outward normal to the boundary at each point. The books J. Munkres [1] and M. Hirsch [1] give a thorough treatment of these ideas. Appendix I also treats other methods of thinking about C* boundaries. Note in passing that once a defining function is given on a neighborhood of the boundary of a domain, it is then a simple matter, using a partition of unity, to extend the defining function to all of space. It is sometimes convenient for us to assume that our defining function is given on all of space, and we do so without comment. 0.2 What is a Holomorphic Function? Now we can begin our study of the function theory of several complex variables. How shall we define the concept of holomorphic (i.e., analytic) function? Let us restrict attention to functions f : @ — C, a domain in C”, that are locally integrable (denoted f € L},.). That is, we assume that fj, |f(z)|dV(z) < 00 for each compact subset K of 2. In particular, in order to avoid aberrations, we shall only discuss functions that are distributions (to see what happens to function theory when such a standing hypothesis is not enforced, see the work of E. R. Hedrick [1]). Distribution theory will not, however, play an explicit role in this book. We now offer four plausible definitions of holomorphic function on a domain aec. DEFINITIONI A function f : 2 —+ C is holomorphic if for each j = 1 and each fixed 21,..-,2)-1;2)41+++.2n the function C9 FZ ee Bay Gs Spt 2n)0.3. Comparison of C! and C5 is holomorphic, in the classical one-variable sense, on the set N21, 0665 Bas Sst ees Zn) S(O EC: (00 B15 6 Stay e+ Zn) € DH: In other words, we require that f be holomorphic in each variable separately. DEFINITION IIA function f : 2 + C that is continuously differentiable ir each complex variable separately on @ is said to be holomorphic if f satisfies the Cauchy-Riemann equations in each variable separately. This is just another way of requiring that f be holomorphic in each variable separately. DEFINITION III A function f : 2 — C is holomorphic (viz. complex ana- lytic) if for each 2° € Q there is an r = r(z°) > 0 such that D”(2°,r) CQ and f can be written as an absolutely and uniformly convergent power series $2) = Doaale- 2°)" for all z € D"(z°,r). DEFINITION IV Let f : 2 — C be continuous in each variable separately and locally bounded. The function f is said to be holomorphic if for each w € Q there is an r = r(w) > 0 such that D"(w,r) C 2 and eieLiiat Hee ST (Gi tests Gs) an FO) = GRAF Sic, enter f, Gea) for all z € D"(w,r). Fortunately Definitions I-IV, as well as several other plausible definitions, are equivalent. This matter is developed in detail in the next several sections. We ask, for now, that the reader accept this fact in order to appreciate the forthcoming overview of some of the important questions of the subject of several complex variables. In Section 1.2 we fix one formal definition of holomorphic function of several variables and proceed logically from that definition. dn 0.3 Comparison of C! and C" At a quick glance, one might be tempted to think of the analysis of several com- plex variables (or several real variables, for that matter) as being essentially one variable theory with the additional complication of multi-indices. This percep- tion turns out to be incorrect. Deep new phenomena and profound (and as yet6 Chapter 0: An Introduction to the Subject FIGURE 0.1 FIGURE 0.2 unsolved) problems present themselves in the theory of several variables. ‘The purpose of this section is to give the reader an appreciation for some of these new features. 0.3.1 Domains of Holomorphy Call an open set U CC” a domain of holomorphy if the following property holds: There do not exist nonempty open sets U;, U2, with Up connected, Up ¢ U, U; © U,NU, such that for every holomorphic function h on U there is a holomorphic hy on Up such that h = hy on Uj. (Refer to Figure 0.1). The definition of domain of holomorphy is complicated because we must allow for the possibility (when dealing with an arbitrary open set U rather than a smooth domain 2) that OU may intersect itself; see Figure 0.2. Setting aside this technicality for the moment, we see that an open set U is not a domain of holomorphy if there is an open set U that properly contains U and such that every holomorphic function h on U analytically continues to a holomorphic function h on U. Note that the open set U is not supposed to depend on h. In the theory of holomorphic functions of one complex variable, every open set is a domain of holomorphy. Here is a simple proof that the unit dise D C C0.3. Comparison of C! and C" 7 is a domain of holomorphy. Let fo(6) = S296”. j=0 Then the series defining fy converges absolutely and uniformly on D by the Weierstrass M-test. Therefore, it is continuous on D and holomorphic on D. However the boundary function oo On ae Date j=0 is the Weierstrass nowhere differentiable function (see Y. Katznelson [1]), so that fo cannot be extended even differentiably, much less analytically, to a larger open set. Since there exists one analytic function on D that cannot be continued to a larger open set, it follows that D is a domain of holomorphy. Now if 2 is a simply connected, smoothly bounded domain then, by the Riemann mapping theorem, there is a conformal mapping © : @ — D. By an old result of Painlevé (see M. Tsuji [1] or S. Bell and S. G. Krantz [1]), the mapping & continues to be continuously differentiable from 2 to D. Likewise the mapping 6~! extends to be continuously differentiable from D to ©. Then fy 0 ® is holomorphic on Q and continuous on 9, yet could not continue to be even differentiable on any strictly larger open set. It follows that Q is a domain of holomorphy. For the general case, let U be any connected open subset of C. Let {z;} CU be a sequence of points that has no accumulation point in U but that accumulates at every boundary point of U (the reader may construct such a sequence by exploiting the distance function to the boundary). By Weierstrass’s theorem (see subsection 0.3.4) there is a nonconstant analytic function h on U that vanishes at each 2; and nowhere else. If h were to analytically continue to an analytic function h on some strictly larger open set U, then it would necessarily hold that U contains a boundary point P of U. But then h would have zero set with interior accumulation point P. It follows that h = 0, which is a contradiction. Thus U is a domain of holomorphy. The situation in several complex variables is decidedly different. Some open sets are trivially domains of holomorphy; if 2 = D?(0,1) CC’, then let fo be the function on the disc that we constructed above and set (21,22) = fo(21) + fo(z2). This function is clearly analytic in Q and does not analytically continue to any strictly larger open set. [f this simple product con- struction were the only way to construct domains of holomorphy in several vari- ables, then there would be little justification for studying this subject. But, in8 Chapter 0: An Introduction to the Subject fact, the characterization of domains of holomorphy in several complex variables is quite subtle. It turns out that the unit ball B(0,1) C C” is a domain of holomorphy, but is not a product domain. The proof of this statement requires considerable ingenuity, and we shall explore it later. The important lesson now is that not all domains in several complex variables are domains of holomorphy. We now explain this important discovery of F. Hartogs [1]. Let Q = D*(0,3) \ D?(0,1) € C?. We claim that every holomorphic function on Q analytically continues to the do- main Q = D?(0,3). This assertion is a special case of a more general phenomenon (the Kugelsatz) that we explore later. For now we give an ad hoc proof, using only elementary one-variable facts about power series, of the continuation result. Now let h be holomorphic on . For 2; fixed, |z1| <3, we write oo he, (22) = A(zi22) = > aj(a)d, (0.3.1.1) I where the coefficients of this Laurent expansion are given by ig had alan f., Ga &. In particular, aj(z1) depends holomorphically on z; (by Morera’s theorem, for instance). But a;(21) = 0 for j < 0 and 1 < |z1| < 3. Therefore, by analytic continuation, a;(z;) = 0 for j < 0. But then the series expansion (0.3.1.1) becomes Saja) j=0 and this series defines a holomorphic function h on all of D*(0,3) that agrees with the original function h on Q. Thus Q is not a domain of holomorphy- all holomorphic functions on Q continue to the larger domain D?(0,3). This completes the proof of the “Hartogs extension phenomenon.” One of the principal jobs for us in a basic course of several complex variables is to solve the Levi problem: to characterize the domains of holomorphy in terms of some geometric properties of the boundary (which should make no reference to holomorphic functions on 2). This highly nontrivial problem was first solved by K. Oka [1] for n = 2 and by H. Bremerman [1], F. Norguet [1], and K. Oka. [3] for n > 3. There are many useful approaches to the problem, including the sheaf-theoretic methods of the authors just mentioned and the functional0.3. Comparison of C! and C"™ 9 analysis methods of A. Andreotti and H. Grauert [1], L. Ehrenpreis [1], Grauert [2], and R. Narasimhan [2] (to name only a few). The Levi problem for complex analytic manifolds and for analytic spaces has been solved by H. Grauert [1], R. Narasimhan [2], R. Remmert and K. Stein [1], and others. These results are both of theoretical interest and are extremely useful in applications. Further progress on the Levi problem has been made by H. Skoda [2], J. P. Demailly [1], Y. T. Siu [1], and others. A number of important problems still remain open. The present text will present the partial differential equations approach to the problem of characterizing domains of holomorphy. This technique has been developed primarily by C. B. Morrey [1], J. J. Kohn [1], and L. Hérmander [1], although modern contributions have been made by J. E. Fornaess [1], R. M. Range 3], and others. For the record, we note that the difficult part of char- acterizing domains of holomorphy—namely, showing that a certain differential geometric condition on the boundary is sufficient for a domain to be a domain of holomorphy—is known as the Levi problem, in honor of E. E. Levi. This problem will be enunciated in great detail in Chapters 3 and 5. It tums out that all convex domains are domains of holomorphy. This would be a lovely necessary and sufficient condition, for it is purely geometric and makes no reference to holomorphicity. However convexity is not preserved under holomorphic mappings: the domain D C C is convex, but its image un- der the mapping ¢ + (4+)! is not. Thus some less rigid geometric condition is required to characterize domains of holomorphy. We shall learn that there is a biholomorphically invariant version of convexity, known as pseudoconvex- ity, which characterizes domains of holomorphy. The notion of convexity will command much of our attention in Chapters 3, 4, and 5. That is why we have spent so much time discussing it now. It is the paradigm for a number of posi- tivity conditions that play an important role in the complex analysis of several variables. We conclude this section by giving rather more brief coverage to several other problems that distinguish several complex variables from one complex vari- able. 0.3.2 Biholomorphic Mappings The Riemann mapping theorem says that every proper, simply connected open subset Q of C is biholomorphic to the disc: that is, there is a one-to-one, onto, holomorphic mapping @ : Q — D. (It is redundant to point out, but we do so for the record, that such a mapping automatically has a holomorphic inverse.) There is no analogue for this result in C”,n > 2. Indeed “most” (in the sense of category) domains in C” that are close to the ball—in any topology you like, say, the C® topology—are not biholomorphic to the ball. In fact the set of biholomorphic equivalence classes, formed among domains close to the ball in any reasonable sense, is uncountable in number (see D. Burns, S. Shnider, and R. Wells (1], and R. E. Greene and S. G. Krantz [1, 2]).10 Chapter 0: An Introduction to the Subject Historically, the first result that suggested the failure of a Riemann map- ping theorem in several complex variables was proved by Poincaré; he proved that, in any dimension n > 2, the ball B(0,1) and the polydisc D"(0,1) are not biholomorphic. Poincaré’s technique was to calculate the group of biholo- morphic self-maps of each domain and to demonstrate that these groups are not. isomorphic as groups. We shall treat that proof and also some more modern, geometric proofs. The discussion in the preceding two paragraphs implies that there is no canonical topologically trivial domain in C",n > 2, as there is in C! (namely, the disc). This means that the analysis of holomorphic functions of several variables depends on the domain in question. Even for topologically complex domains in C!, the uniformization theorem (see H. Farkas and I. Kra [1]) often enables one to reduce analytic questions on a planar domain to analytic questions on the disc. This approach is explored in detail in §. Fisher [1]. But there is also no uniformization theorem in several complex variables. In spite of the complex state of affairs discussed in the last paragraph, there has been some heartening progress made in the search for substitutes in several variables for the Riemann mapping theorem. We mention particularly the work of Fridman [3], L. Lempert [2], R. E. Greene and S. G. Krantz [1, 2], Daowei Ma {i], and S. Semmes [1]. The subject of biholomorphic and, more generally, proper holomorphic mappings has been an active area of research for the last 20 years. The most fundamental question is the following: If 91,2 C C” are open sets in C” with smooth boundaries and if ® : 2, + Q» is a biholomorphic (resp. proper holomor- phic) mapping, that is, 6(z) = (®,(z),...,®,(z)) where each &; is holomorphic and @ is one-to-one and onto (resp. each ©; is holomorphic and &~!(K) is com- pact in Q, for each compact K in 9), is it true that © extends to be a C° mapping of 21 to Q2? The answer to this last question is known to be yes in a number of important cases. There are no known counterexamples. We discuss some of these matters in detail in Section 11.4. Once the answer to this basic question is known to be yes, then one might hope to calculate differential boundary invariants for biholomorphic mappings. In turn, this will, in principle, allow one to identify domains that are biholomorphically inequivalent, at least locally. There has been some heartening progress in this program of Poincaré (see 8. S. Chern and J. Moser [1], and C. Fefferman [2]), and we shall touch on it in Chapter 11. 0.3.3 Zero Sets of Analytic Functions If Q C C™ is a domain, f : 2 + C is holomorphic, and No(f) = {2 € 2: f(z) = 0}, then what properties (geometric, topological, or otherwise) does the set No(f) have? In one complex variable, the Weierstrass theorem says that any (denumerable) set $ C © without interior accumulation point is the zero set of a holomorphic function. The converse is true as well and is rather trivial. A0.3. Comparison of C! and C® 11 more interesting question in one complex variable is to characterize the zero sets of holomorphic functions satisfying a certain growth condition. For instance, let H®(D) be the collection of bounded, holomorphic functions on the unit disc DCC. Then a necessary and sufficient condition for {2;}32, C D to be the zero set of an f € H®(D) is that }>(1 — |z;|) < 00. Matters are vastly more complicated in the theory of several complex vari- ables. The zeros of a holomorphic function f : C" — C, n > 2, are never isolated. [If f had an isolated zero at z = 0, then 1/f would be holomorphic on D"(0,3) \ D*(0,1) yet not continuable to D(0,3), contradicting the Hartogs phenomenon.] Roughly speaking, the zero set of a holomorphic function of sev- eral complex variables is no more complicated than the zero set of a holomorphic polynomial of several variables (the Weierstrass preparation theorem, treated in Chapter 6, makes this statement precise). Yet the study of zero sets of polyno- mials of several variables is the principal subject of algebraic geometry, one of the deepest: branches of mathematics. Speaking somewhat imprecisely, we may say that the zero set of a holo- morphic function on a domain 2 C C” is locally (but for an exceptional set of lower dimension) a complex manifold of complex dimension (n —1) (think about what this says in dimension one!). To amplify and clarify this true, but rather vague, description of a complex analytic variety is hard work (see R. C. Gunning (2). We shall say more about this matter in Chapters 6 and 7. Recently some results have been obtained about zero sets for holomorphic functions satisfying a growth condition, but they are hard to state and much more so to prove (see G. M. Henkin {6], H. Skoda [3], N. Th. Varopoulos [1, 2, 3], W. Rudin [3]). 0.3.4 Divisors Recall the Mittag-Leffler theorem of one complex variable theory. THEOREM 0.3.1 (Mittag-Leffler) Let 2 C C be a proper open subset and {z;}9, C Qa discrete set (no interior accumulation points). Let {f;} be meromorphic in a neighborhood of z;, each j. Then there exists a meromorphic function f on Q such that f is analytic on Q \ {z; : j = 1,2,...} and f — fy is analytic near z; for each j. Likewise, the Weierstrass theorem, which we mentioned earlier in this chap- ter, is as follows: THEOREM 0.3.2 (Weierstrass) Let 2, {z;}92, C © be as in the preced- ing theorem. Let {kj} C Z. There is a meromorphic function f on @, analytic and nonzero off {z; : j = 1,2,...}, such that f(z)/(z — z;)*» is analytic and nonzero near z;, each j. It is of interest to have analogues of these theorems in the function theory of several complex variables. Since we know from subsection 0.3.3 on zero sets of holomorphic functions that the singularities of holomorphic functions are never12 Chapter 0: An Introduction to the Subject isolated, it follows that there is difficulty in even formulating analogues of the theorems. Here is another way to phrase these types of problems. Let S C Q have the property that it is locally the zero set of an analytic function; that is, to each w € S is associated a small open set 2, C 2 and a holomorphic function f,, on Qw such that SNQw = Na, (fw) Does it follow that there is a holomorphic function F on Q with S = No(F)? Problems formulated in this way lend themselves to solution by means of (the known solutions of) the Cousin problems. The Cousin problems were a precursor of sheaf theory. Both sheaf theory and the Cousin problems will enter into our formulations and our proofs of analogues of the Weierstrass and Mittag-Leffler theorems in Chapters 6 and 7. 0.3.5 Factoring Holomorphic Functions Let H*(D) be the space of bounded analytic functions on the unit disc. If f ¢ H™(D), then (Section 8.1) it is possible to write f = F - B, where 1. F,B are analytic on D; 2. B is bounded by 1; 3. lim,_.;- B(re®) exists and has modulus 1 for almost every 6 € (0,27); 4. F is nonvanishing. Stated crudely, F carries the “size” of f, whereas B carries the zeros. It turns out that B is a Blaschke product, a special case of the concept of “inner function.” It is known (see L. A. Rubel and A. Shields [1]) that a factorization like f = F'- B is impossible in several complex variables. However there do exist nonconstant inner functions: bounded analytic functions on the ball in C” with unimodular radial boundary limits almost everywhere. In fact, it was an open problem, studied intensely from 1965 until 1982, whether nonconstant inner functions exist in dimensions greater than one. Most people, including the original formulators of the problem (Rudin and Vitushkin), were convinced that inner functions could not exist. A fairly large body of work was amassed to support that belief. Rather surprisingly, Aleksandrov [1] and Low [1] constructed nonconstant inner functions in 1982. While inner functions have proved a useful tool for constructing holomor- phic functions with prescribed behavior (see W. Rudin [9] for an exhaustive treatment), they have not provided the rich structure that inner functions and Blaschke products have given us in dimension one (see K. Hoffman [1] for the classical theory). In Chapter 9 we shall construct inner functions of several com- plex variables and present related techniques in constructive function theory.0.3. Comparison of C! and C™ 13 A related problem, also in the spirit of factoring holomorphic functions, is ascribed to A. Gleason. For a domain 2 € C”, any n, let A(Q) denote the space of functions that are continuous on 2 and holomorphic on Q. If f € A(D) and f(0) = 0, then of course we can write f(z) = z-g(z), with g also an element of A(D). Gleason proposed that the analogue in C” should be that if f € A(B) and f(0) =0, then we can write f(z) = 2191 (2) + +++ + 2ngn(2) with g; € A(B),j = 1,...,n. This result was proved not only for the ball, but for convex domains with C? boundary, in Leibenzon [1]. It was proved for more general domains in N. Kerzman [2], N. Kerzman and A. Nagel (1), N. Sibony and J. Wermer [1], and J. E. Forneess and N. Ovrelid {1]. We will treat these matters in Chapters 5, 7, and 10. 0.3.6 Domains of Convergence for Power Series Consider the power series : Sas j=0 for 2 € C. It is elementary to see that there is an r,0 r. Observe that the domain of convergence is always a disc. Matters are not so simple, geometrically speaking, in several variables. The series at 2) j=0 converges absolutely and uniformly for |z; +22| < 1 and diverges for |21 +22| > 1. However, the series Ya +22) j=0 converges absolutely and uniformly for |21-22| < 1 and diverges when |z1+z9| > 1. Finally, the series Mae converges absolutely for |2| < 1,22 arbitrary and diverges for |z| > 1, 2 arbitrary.14 Chapter 0: An Introduction to the Subject Whereas in one complex variable the domain of convergence of a power series is always a disc, in several complex variables the domain of convergence does not have a simple geometric description. Since power series are central to the way that we represent holomorphic functions, it is clearly of interest to understand this matter. We shall address it in Chapter 2. 0.3.7 Extending Holomorphic Functions Let B C C? be the unit ball and define a-an{iajetia=3}. Of course d is just a copy of the disc D1(0, \/3/4). Suppose that ¢ is a holo- morphic function of one complex variable on D(0, \/3/4). Then, by defining f(z1,4) = 6(a1), we can construct an analytic function on d. The question is: Is $ necessarily the restriction to d of a function F on all of B? If ¢ is bounded, can we find an F that is bounded? In any program to construct holomorphic functions with specified prop- erties, an affirmative answer to these questions would be useful. There is an affirmative answer on the ball, and on any domain of holomorphy, but a great deal of the machinery of several complex variables must be used to provide a construction. [Note, in the example posed in the last paragraph, that the trivial extension F(z1, 22) = f(z1, 4) cannot work since it makes no sense for |zi| > ~2.] We shall explore methods for treating extension problems in Chapter 5. Observe that the types of extension problems being discussed here are trivial in complex dimension one. 0.3.8 Approximation Problems Let 2 € C" be a domain and f : 2 — C a holomorphic function. Let K CQ be compact. Is f the uniform limit, on K, of a sequence of functions holomorphic on a neighborhood of 2? This is an example of what we call an approximation problem. In one complex variable, much is known. Two examples of classical results of this type are as follows. THEOREM 0.3.3 (Runge) — Let 2 C C be adomain. Let f be holomorphic on Q C Cand let K C Q be compact. Let 7 > 0. Then there is a rational function (quotient of polynomials) r(z) with poles in °K such that sup |f(¢) —r(Q)| <0. K0.3. Comparison of C! and C® 15 THEOREM 0.3.4 (Mergelyan) Let K be a compact subset of C whose complement in C is connected. Let f : K — C be continuous on K and holo- morphic on the interior of A. Then f is the uniform limit on K of holomorphic polynomials. Notice that the hypotheses of both these theorems are purely topological in nature. In several complex variables, to the limited extent that we have any results of this type, the hypotheses depend on differential geometric data of and K and also on how Q and K are imbedded in space. Sections 5.4 and 10.4 treat these matters briefly. 0.3.9 The Cauchy-Riemann Equations Let @ be a continuously differentiable function on a domain @ in C and taking values in C. Then ¢ is holomorphic if and only if (8/92)4(¢) = 0. For if we write ¢ = $1 + id2—the decomposition of ¢ into its real and imaginary parts—then a6 1/a oN... . ae i(ae tig) + ide) _ 1/86, 62) _1.(0¢1 , Ade = a(S) (Be). Thus the condition that (0/0¢)¢(¢) = 0 corresponds to the simultaneous van- ishing of the real and imaginary parts of the displayed equations. But that is just the classical Cauchy-Riemann equations for ¢. It is of interest to be able to solve the inhomogeneous equation du _y, FY when 7 is a given function on C, that is, say, C! with compact support. It turns out (as we shall prove in the next chapter) that the formula _ 1s © uo=-2 [evo always supplies a solution to this problem. However, even though v is compactly supported, no solution of the equation du/9¢ = wu can be compactly supported if [ ¥(C)dV (6) # 0. For if there were a compactly supported u, say u(¢) = 016 Chapter 0: An Introduction to the Subject when |¢| > R, that satisfied the equation, then 0 = di $98 (Stokes) Ou = % Ion FE) aE Na = 2% W(€) dV i [..% ©) x 0 if D(0, R) contains the support of yy. That is a contradiction. Now the analogous problem in several complex variables is as follows: Let V1,-++;Un be given functions that are in C}(C") (that is, C) with compact support in C"). We seek a C? function u such that Ou ae 7M J (0.3.2) Since Ou Pu 020%, 024.02;’ it is clearly necessary, for (0.3.2) to have a solution, that Duy _ Ov = OPE al jk= a 03° * 7 aye Subject to this compatibility condition, we shall prove in Section 1.1 that the system is always solvable and a solution is given by veya f Mensano for any choice of j = 1,...,n. This solution is always compactly supported when the dimension n > 1. Suppose for simplicity that j = 1. Now du/0% = ve = 0, all €, when z is large. Thus u is holomorphic in each variable separately, hence holomorphic, for z large—say when |z| > R. But, looking at the definition of u, we see that u(z) itself must be zero when z is large. By analytic continuation, u(z) = 0 on the unbounded component of “(Ujsupp%;) In other words, wu is compactly supported. Although it is not evident now, it will become plain later that many of the differences between the the function theories of one and several complex variables can be accounted for by the support behavior of solutions of the inhomogeneous Cauchy-Riemann equations. These equations will be a principal tool for us in0.3. Comparison of C! and C" 17 constructing holomorphic functions with specified properties. Since the theory of one complex variable has so many other tools for constructing functions— Blaschke products, Weierstrass products, Runge and Mittag-Leffler theorems, and so on—the Cauchy-Riemann equations have played a less prominent role in that subject. But in several complex variables these equations are central. We now terminate this list of introductory problems, since the intended point has been established—namely, we have seen that there are significant dif- ferences between the theories of one and several complex variables, and we have introduced a number of the problems to be considered in succeeding chapters. We will see later that these problems are by no means disjoint; many of the same tools are used to attack all of them. This is a reflection both of the beauty of the subject and of the limited nature of our knowledge.1 Some Integral Formulas Our mission in this chapter is a bit eccentric when compared to the mission of other books on the subject. But it is well motivated: Whereas many classical treatments of the function theory of one or several complex variables imply by their presentation that complex analysis is a subject unto itself, with own language and its own notation, our point of view is rather different. We demonstrate in this chapter that Stokes’s theorem is the first step and that all our basic analytic tools follow from it. Thus our presentation will entail the repetition of a bit of classical textbook material; for instance, we derive Green’s theorem in N dimensions directly from Stokes’s theorem. We derive several important integral formulas using first principles directly from Green’s theorem. We show that the classical Cauchy integral formula, and some of its generalizations to several complex variables, is an aspect of Stokes’s theorem. Although many of the steps of our treatment may be found in some form in some other text (generally in a much more abstract setting), it is our purpose here to present the entire development in a self-contained fashion, beginning with ideas of advanced calculus. Since the function theory of several complex variables can become rather abstract rather quickly, it is well to begin our journey with concrete material such as this. For another quite concrete development, see W. Rudin [7]. The reader who is impatient may safely skip Sections 1.3-1.6, referring back to this material when necessary. We do encourage the impatient reader to look over the material in Sections 1.1 and 1.2 before proceeding to Chapter 2. 1.1 The Bochner-Martinelli Formula If w is a differential form on 2 C R%, then, in coordinates, w can be written as a finite sum of terms of the form wadx, where a is a multi-index and wa is a smooth function. Differential forms on C” may be written in this fashion also, since C” is canonically identified with R?”. However it is much more convenient to use the complex notation introduced in Chapter 0. Thus if Q C C” and w is a differential form on Q, then w is a sum of terms of the form wagdz% A d29, 1920 Chapter 1: Some Integral Formulas where @, 8 are multi-indices with lal 1, and learn what consequences it has for the function theory of both one and several variables. DEFINITION 1.1.2 On C” we let w(z) = dy Adz A---Adin, n m2) = S21 Hayden A+++ Adzj-1 A dajy1 Ave Adin. j=l The form 7) is sometimes called the Leray form. We will often write w(Z) to mean di, \--- Adz, and, likewise, n(Z) to mean D7 ,(—1))*!z;)d% A--- A dzj-1 A 2541 A+++ A d2q. The genesis of the Leray form is explained by the following lemma. LEMMA 1.1.3 For any 2 € C", any € > 0, we have [ nla) Aw anf w(2) Nw(2). 'OB(2°,¢) B(2",€) Proof Notice that d7j(2) = dn(2) = nw(2). By Stokes’s theorem, Foo pM netear= fl. aln(2) Awe) ‘OB(2",e) B(2°,e) Of course the expression in | _] is saturated in dz’s, so, in the decomposition d=0+9, only the term will not die. Thus the last line equals Ff. , Wena awe) =n fl ule) Aula) o B(2%,e) B(2,6) Remark: Notice that, by change of variables, Pont root ) en Jig yt Ae A straightforward calculation shows that: [ w(Z) Aw(z) = (—1)™ - (2i)" « (volume of the unit ball in C” © R?”), B(0,1) where q(n) = [n(n — 1)]/2. We denote the value of this integral by W(n).22 Chapter 1: Some Integral Formulas THEOREM 1.1.4 (Bochner-Martinelli) Let © C C” be a bounded do- main with C! boundary. Let f € C1(@). Then, for any 2 € , we have 1 L(Qn(C = 2) Awd) f= GW Joo Cae 7 af(¢) ~ i aie foe Ae -D A0t0. Proof Fix z € 9. We apply Stokes’s theorem to the form = 2) Awl LQ = Unie oo © on the domain 2... = 2\B(z, e), where € > 0 is chosen so small that B(z,e) C 9. Note that Stokes’s theorem does not apply to forms that have a singularity; thus we may not apply the theorem to L, on any domain that contains the point z in either its interior or its boundary. This observation helps to dictate the form of the domain Q,,.. As the proof develops, we shall see that it also helps to determine the outcome of our calculation. Notice that A(Qz,<) = ONUAB(z,€) but that the two pieces are equipped with opposite orientations (Figure 1.1). Thus, by Stokes, fb0- 7 | f,,.b a dg(Lz(¢))- (1.1.4.1) Notice that we consider z to be fixed and ¢ to be the variable. Now OcL2(6) IFO An — 2) Aw(6) = +h(Q)- Eee *)| w(Q Ault). (11.4.2) deL2(6)1.1, The Bochner-Martinelli Formula 23 FIGURE 1.1 Observing that &(4)-or- a ie = aR) C= a we find that the second term on the far right of (1.1.4.2) dies and we have OFC) An - 2) Aw(6) aL, (¢) = ic Substituting this identity into (1.1.4.1) yields _f FG) AnG= 2) Aw(6) [0-0 20> fe —ar : (1.1.4.3) Next we remark that —2Z)A Osa an OB(z,<) ae oe (f() = F(2)) nC - 2) Aw(d) a i \¢ - 2)? = T+tk. (1.1.4.4) Since | f(6)— f(z)| < C|¢—z| (and since each term of n(¢—2) has a factor of some (¢; —3;)), it follows that the integrand of Ty is of size O(|¢ — z|)~2"+? = E72"? Since the surface over which the integration is performed has area © €?”~1, it follows that Tz + 0 as € > 0*.24 Chapter 1: Some Integral Formulas By Lemma 1.1.3, we also have n= ere f — E-2 Awd) B(2,6) See 7 = neste) f oo ON = nW(n)f(2). (1.1.4.5) Finally, (1.1.4.3)-(1.1.4.5) yield that (f, 2-0) -awenge ron = ff aron| Since n(¢ - 2) = aF = O16 - 2), the last integral is absolutely convergent as « — 0+ (remember that Of is bounded). Thus we finally have a £0)= Spay fH aay [M0 ALB roo. This is the Bochner-Martinelli formula. o COROLLARY 1.1.5 If QC C is a bounded domain with C! boundary and if J €C1(Q), then, for any 2 € Q, eee LQ Lf (af(Q/80) 1-55 f Powe acoso dé Ade. Proof It is necessary only to note that, when n= 1, w(Q)=de, n(C-2)=C-2, and nW(n) = 2ri. a COROLLARY 1.1.6 With hypotheses as in Corollary 1.1.5 and the additional assumption that Of = 0 on 2, we have 1-3, f 2 «. 2mi Jon G—21.1. The Bochner-Martinelli Formula 25 Remark: Corollary 1.1.6 contains the familiar Cauchy integral formula from analysis of one variable. It is a wonderful fluke of one-variable analysis that when n = 1, the expression (¢; — 2;)/|¢ — z|?” becomes (¢ — 2)/|¢ — z|? = 1/(¢ - 2). In particular, the kernel of the classical Cauchy integral formula is holomorphic in z. This provides an important technique for creating holomorphic functions (namely, integrating this kernel against a measure) in one-variable analysis that at this stage is missing from multi-variable analysis. a COROLLARY 1.1.7 If 2 C C” is bounded and has C! boundary and if f € C1(Q) and Jf = 0 on Q, then 1 F(Qn(¢ = 2) a= | AY) ng. Liz. f(z) nvm) Iya [e= 2 (0) (1.1.7.1) It is rather surprising that the nonholomorphic kernel in (1.1.7.1) somehow manages to reproduce 8-closed, or holomorphic, functions. What makes (1.1.7.1) of limited utility is that it does not create holomorphic (d-closed) functions. Although it is true that the form n= 2), icon () is O-closed in z for z # ¢ (in fact, this kernel is a fundamental solution for the 8 operator—see R. Harvey and J. Polking [1]), it is not the case that this expression is holomorphic in the free variable z unless n = 1. In other words, the form is 0-closed away from the singularity, but its coefficients, as functions, are not holomorphic in z. It is a fairly recent development, primarily due to G. M. Henkin [2], E Ramirez [1], and H. Grauert and I. Lieb [1], that fairly explicit holomorphic n- dimensional reproducing kernels have been constructed (a nonconstructive way to find such kernels is given in Sections 1.4 and 1.5). We consider these con- structions in Chapters 5 and 10. In spite of its limitations, a few useful facts can be gleaned from Corollary 1.1.7. In particular, one may check by differentiation under the integral sign that a 0-closed function is in fact C°, indeed real analytic, on its domain of definition because the integral kernel has these properties. Now we derive the explicit integral solution formula for the inhomogeneous Cauchy-Riemann equations that we discussed in subsection 0.3.9. We will see that the holomorphicity of the one-dimensional Cauchy kernel will play an im- portant role in the derivation of this formula. First we give a careful definition of the support of a form. DEFINITION 1.1.8 If) is a differential form on R™ or C”, then the support of \, written supp A, is the complement of the union of all open sets on which the coefficients of \ vanish identically.26 Chapter 1: Some Integral Formulas THEOREM 1.1.9 Let € C}(C). The function defined by = _ ¥(§) ud=—g5 f PO aenae=-F f PO we satisfies Ee ou 7 BE Ou() = S5(¢) dC = ¥(¢) de. aC Proof Let D(0,R) be a large disc that contains the support of. Then 10 wg) Ou ae) = “ari de Jo ¢ aNd — 129 foet+o a _ ge ENd Eto) pee eee REO) a imi € db nds (6) bE Ad. 1 Qri low e-¢ By Corollary 1.1.5, this last equals 10-55 bo pet k= 40. Here we have used the support condition on y. This is the result that we wish to prove. a Recall now that the convolution of two L! functions on R* is defined by (f *9)(«) -{ f(x — t)g(t) dV(t) | g(a —t) f(t) dV(t). RN RN The same notion applies in C" ~ R2", Let K(€) = —1/(n€) on C. Then Theorem 1.1.9 asserts that for any € C}, it holds that u(¢) = K * Y(¢) satisfies In other words,