London Mathematical Society Student Texts. 34 Complex Algebraic Surfaces Second Edition Arnaud Beauville Université Paris-Sud 4 CAMBRIDGE 9 UNIVERSITY PRESSPublished by the Press Syndicate of the University of Cambridge The Pitt Building, Trumpington Street, Cambridge CB2 1RP 40 West 20th Street, New York, NY 10011-4211, USA 10 Stamford Road, Oakleigh, Victoria 3166, Australia Originally published in French as Surfaces Algébriques Complezes, Astérique 54 and ©Société Mathématique de France, Paris 1978 English translation ©Cambridge University Press 1983, 1996 Translated by R. Barlow, With assistance from N.I. Shepherd-Barron and M. Reid First published in English by Cambridge University Press 1983 as Compler Algebraic Surfaces Second edition first published 1996 Printed in Great Britain at the University Press, Cambridge Library of Congress cataloguing in publication data available A catalogue record for this book is available from the British Library ISBN 0 521 495105 hardback ISBN 0 521 49842 2 paperbackIntroduction Notation Chapter I Chapter II Chapter III Chapter IV Chapter V Chapter VI Chapter VII Chapter VIII Chapter 1X Chapter X Appendix A Appendix B Appendix C References Index CONTENTS The Picard group and the Riemann—Roch theorem Birational maps Ruled surfaces Rational surfaces Castelnuovo’s theorem and applications Surfaces with pg = 0 and g 21 Kodaira dimension Surfaces with « = 0 Surfaces with « = 1 and elliptic surfaces Surfaces of general type Characteristic p Complex surfaces Further reading vii lL 25 40 55 68 86 89 107 113 119 lal 124 127 131INTRODUCTION This book is a modified version of a course given at Orsay in 1976-77. The ‘aim of the course was to give a comparatively elementary proof of the Enriques classification of complex algebraic surfaces, accessible to a student familiar with the basic language of algebraic geometry (divisors, differential.forms, ...) as well as sheaf cohomology. I have, however, preferred to assume along the way various hard theorems from algebraic geometry, rather than resort to complicated and artificial proofs. : ‘Here is an outline of the course. ‘The first two chapters introduce the basic tools for the study of surfaces: in Chapter I we define the intersec- tion form on the Picard group, and establish its properties; assuming the duality theorem we deduce the fundamental results (the Riemann-Roch theorem, the genus formula). Chapter II is devoted to the structure of birational maps; we show in particular that every surface is obtained from a minimal surface by a finite number of blow-ups. The chapter ends with Castelnuovo’s contractibility criterion, which characterizes excep- tional curves by their numerical properties. The classification begins in Chapter III with ruled surfaces, that is, surfaces birational to P! x C. We show that (except in the rational case) their minimal models are P!-bundles over a base curve C, and we study their geometry. Chapter IV gives some examples of rational surfaces; we take a stroll through the huge menagerie collected by the geometers of the 19th century (the Veronese surface, del Pezzo surfaces, ...). The next two chapters are perhaps the keystone of the classification; they give the characterization of ruled surfaces by their numerical prop- erties - more precisely, by the vanishing of the ‘plurigenera’ P,. Sur- faces with q = 0 are treated in Chapter V, where we prove Castelnuovo’s theorem: a surface with q = P. = 0 is rational. We deduce two impor- tant consequences: the structure of minimal rational surfaces and theviii Introduction uniqueness of the minimal model of a non-ruled surface. In Chapter VI we begin the study of surfaces with g > 0. We show without too much trouble that a surface having py = 0 and q > 2 is ruled; which leaves certain non-ruled surfaces with py = 0 and g = 1. According to an idea of Enriques one can classify these surfaces very precisely, and show that they have P;2 > 0. Thus a surface is ruled if and only if Piz = 0 (Enriques’ theorem). Chapter VII, which is very short, introduces the Kodaira dimension «, which is a convenient invariant for the classification of surfaces. Ruled surfaces are characterised by « = —oo; the three ensuing chapters deal with surfaces with « = 0,1, and 2. Surfaces with « = 0 fall into four classes: K3 surfaces, Enriques surfaces, Abelian surfaces, and bielliptic surfaces. The bielliptic surfaces were already listed in Chapter VI, in the context of surfaces with pg = 0 and q = 1; here we study K3 surfaces and Enriques surfaces, and give numerous examples. : In Chapter IX we show that surfaces with « =-1 have a (not necessarily rational) pencil of elliptic curves; conversely we study those surfaces with such a pencil. sho Finally Chaper X concerns surfaces with « = 2,. said to be of general type; although these surfaces are the most general, there is not véry much that we can say about them. We have limited ourselves to giving some examples and proving Castelnuovo’s inequality x(Os) > 0. In Appendix A we sketch (without proof) the classification of surfaces in characteristic p, and in Appendix B that of complex compact surfaces. Appendix C indicates some of the new results (or new approaches to old tesults) which have been obtained since the first appearance of this book. It is hard to claim any originality in a subject whose main theorems were proved at the turn of the century. I have been largely inspired by the existing literature, in particular by Shafarevich’s seminar [Sh 2]; in a historical note at the end of each chapter I have tried to describe the origins of the principal results. The exercises indicate various possible extensions to the course.NOTATION By ‘surface’ we shall mean smooth projective surface over the field C of complex numbers. Let S be a surface, and D, D’ two divisors on S. We write: D = D' if D and D’ are linearly equivalent Os(D): the invertible sheaf corresponding to D H *(S,Os(D)), or simply H*(D): the ith cohomology group of the sheaf Os(D) hi(D) = dime Hi(D) x(Os(D)) = h°(D) — h}(D) + h?(D), the Euler—Poincaré characteristic of the sheaf Os(D) |D| = the set of effective divisors linearly equivalent to D = the projective space corresponding to H°(D) Ks or K = ‘the’ canonical divisor = a divisor such that Os(K) = 2% Pic S = the group of divisors on S modulo linear equivalence = group of isomorphism classes of invertible sheaves NS(S) = the Néron-Severi group of S (1.10) Alb(S) = the Albanese variety of S (see Chapter V) a(S) or ¢ = hY(Os) = b(O4) Po(S) oF By = h*(Os) = h"(Os(K)) Pn(S) or Py = h°(Os(nK)) (for n > 1) 6;(S) or 6; = dim H'(S,R) xron(S) = E(-19'4(5)I THE PICARD GROUP AND THE RIEMANN-ROCH THEOREM Unless otherwise stated, we consider surfaces with their Zariski topol- ogy (the closed subsets are the algebraic subvarieties); ‘sheaf’ will mean ‘coherent algebraic sheaf’. This is a matter of convention: Serre’s gen- ata} theorems ([GAGA]) give a bijection between algebraic and analytic dbtiérent sheaves which preserves exactness, cohomology, etc. All our ar- gine with' coherent algebraic sheaves will remain valid in the analytic ontext. Fact I.1 The Picard group Let S be a smooth variety. Recall that the Picard group of S, Pic S, is the group of isomorphism classes of invertible sheaves (or of line bundles) on S. To every effective divisor D on S there corresponds an invertible sheaf Os(D) and a section s € H°(Os(D)), s # 0, which is unique up to scalar multiplication, such that div(s) = D. We define Os(D) for au arbitrary D by linearity. The map D ++ Os(D) identifies Pic S with the group of linear equivalence classes of divisors on S. Let X be another smooth variety and f : S + X a morphism. We can define the inverse image with respect to f of an invertible sheaf, to get a homomorphism f* : PicX — Pic S. If f is surjective, then we can also define the inverse image of a divisor, in such a way that f*Ox(D) = Os(f*(D)): just note that the inverse image of a non-zero section of Ox (D) is non-zero. If f is a morphism of surfaces which is generically finite of degree d, then we define the direct image f,C of an irreducible curve C by setting f.C=0 if f(C) is a point, £C=rt if f(C) is a curve T, the morphism C — T induced by f being finite of degree r.