Lecture Notes in Mathematics 1595 Editors: A. Dold, Heidelberg B. Eckmann, Ziirich F. Takens, GroningenRobert D.M. Accola Topics in the Theory of Riemann Surfaces Springer-Verlag Berlin Heidelberg New York London Paris Tok Yo Hong Kong Barcelon: BudapestAuthor Robert D.M. Accola Department of Mathematics Brown University Providence, Rhode Island 02912, USA. Mathematics Subject Classification (1991): 30F10; 14H15 ISBN 3-540-58721-7 Springer-Verlag Berlin Heidelberg New York CIP-Data applied for This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. © Springer-Verlag Berlin Heidelberg 1994 Printed in Germany ‘Typesetting: Camera-ready out put by the author SPIN: 10130221 46/3140-543210 - Printed on acid-free paperPreface These are lecture notes for a course given during the Fall of 1988 at Brown University. The students were assumed to have had a previous course on the theory of Riemann surfaces or algebraic curves. Chapter One of these notes gives a review of most of the basic material needed later, but few proofs are given, and the reader is assumed to have some previous acquaintance with the material. Besides giving the author's point of view and notational conventions, the introduction gives a few proofs (or rather demonstrations) intended to bridge the gap between the more analytic approach to the subject as exemplified by the books of Ahlfors-Sario [4], Farkas-Kra [11] or Forster {42] and the more algebraic approach as exemplified by Walker [26]. In particular the proof of the genus formulas for a plane curve based on the Riemann-Hurwitz formula and other materials in Section 13 were inspired by a seminar given in the early '60's at Brown by S. Lefschetz. Despite the recent appearance of several excellent books on the theory of compact Riemann surfaces (which is, of course, basically the same as the theory of algebraic curves over the complex numbers), most of the material in these notes does not appear in these books. The two main subjects treated here are exceptional points on Riemann surfaces (Weierstrass points, higher-order Weierstrass points) and automorphisms of Riemann surfaces. A foundational treatment of the theory of automorphisms from the viewpoint of Galois coverings of Riemann surfaces is given in Chapters Four and Five, following and expanding to some extent the treatment of Ahlfors-Sario {4] and Seifert-Threlfall [25]. The treatment is technically different from that of A. M. Macbeath [19] and his students, although fundamentally it is the same. In the treatment here, no mention of Fuchsian groups occurs. Also a treatment of the extremely useful inequality of Castelnuovo-Severi is included, a treatment for which the author knows no reference. (It is difficult, however, to believe that anything on this venerable subject can be really new.) The treatment of all subjects is basically elementary. The author heartily thanks Natalie Ruth Johnson for an excellent job in preparing this manuscript.Preface Table of Contents Chapter |: Review of basic concepts 1.4 Coverings 1.2. Function Fields 1.3 Plane Curves 14 Divisors on Riemann surfaces 15. Linear series 1.6 Line bundles Chapter 2: Some exceptional points on Riemann surfaces 2.6 (Theorem on the sum of generalized Weierstrass weights) 2.7 Examples 28 Theorem (on the maximum weight of a generalized Weierstrass point) Chapter 3: The inequality of Castelnuovo-Severi 35 3.6 37 3.8 3.9 Statement of the inequality of Castelnuovo-Severi Applications Some Remarks Proof of Theorem 3.5 Castelnuovo's Riemann-Roch theorem 3.10 An application of Castelnuove's Riemann-Roch theorem Page Naane e S 13 ays 18 18 20 21 22 22 23 27 29vil Chapter 4: Smooth and branched coverings of Riemann surfaces 4. 4.2 43 44 45 46 47 48 49 4.10 a 4.13 4.14 al? (Some standard results) Classification of smooth coverings Lifting conformal structures Cover transformations Some remarks on permutation groups (Galois coverings) (The monodromy group) (Existence theorern) Galois closure Branched coverings (Lifting automorphisms) (Lifting automorphisms to Abelian coverings) Conjugate automorphisms groups (Some consequences) Chapter 5: Automorphisms of Riemann surfaces, | cil 5.2 53 54 55 5.6 57 58 59 5.10 SAL 512 5.13 (Theorem of H. A. Schwarz) (Galois coverings of the Riemann sphere) Groups of genus zero Groups of genus one Maximal automorphism groups for p > 2 An infinite number of (3,3,7)'s Smooth cyclic coverings of prime order N(p) < 84(p - 4) infinitely often Groups with partitions 2, x 2 2 * z x 2, Dihedral groups of automorphisms Commuting involutions 30 30 30 31 31 33 33 34 36 36 36 39 40 40 41 42 42 42 43 44 45 46 48 48 48 49 ia) 50 513 Chapter 6: When are fixed points of automorphisms exceptional in some other sense? 6.1 Lewittes’ theorem 62 Gig on, x 63 Cyclic coverings of P* of prime order: Examples 64 (Motivating Lemma) 6.6 (Lemma on G-gaps) 643 (Theorem on higher order Weierstrass points) 644 Cyclic coverings in general 6.16 Invariant divisors 619 (Theorem on G-weights) 6.20 (Automorphisms with three or more fixed points) Chapter 7: Automorphisms of Riemann surfaces, II; N(p) 72 For p22,N(p)2 8(p+4) 73 lf 3 divides p, then N(p) 2 8(p+ 3) 74 (Z, Lower bounds for some N(p!'s) 75 (Statement of Kiley’s theorem) 78 (Group theoretic result and application) 7.9 (Another group theoretic preliminary) 7.40 Proof of Theorem 7.5 7.11 (N(p) for some small genera) 7.45. (Sharpness of bounds in Theorem 7.7) 7.46 (N(n+4) for all primes n} 747 (N(Qn +1) for all primes n) References Index 52 52 52 54 58 = 62 62 64 70 71 74 74 74 75 76 80 83 84 89 93 ce. 102Chapter 1. Review of some basic concepts in the theory of Riemann surfaces. L1_Coverings. A surface X is a connected Hausdorff space which satisfies the second axion of countability and has a basis for the open sets of sets homeomorphic to open sets in €, the complex numbers. A Riemann surface is a surface with an open cover {U,| a € A}, for some index set A , and homeomorphisms (9,}, 9, :U, 7 ©, where 9,°9 _ is biholomorphic wherever it is defined. A pair (U,.9,) 18 called a chart and the set {(U,,9,)|o A} is called an atlas. 9, ‘will be called a local parameter. All Riemann surfaces are orientable. A continuous function f:X — Y between Rieman surfaces will be said to be holomorphic if it is holomorphic (analytic) when expressed in. local parameters. If { is holomorphic, injective, and surjective then f is said to be biholomorphic and X and Y are said to be conformally equivalent. A biholomorphic map of a Riemann surface onto itself will be called an automorphism, and the group (under composition) of auto- morphisms of a Riemann surface X will be denoted Aut(X). An automorphism of period 2 will be called an involution. A holomorphic (meromorphic) function on a Riemann surface X is a holomorphic mapping of X into € (P1, the projective line, or Riemann sphere). If £:X > Y isa non-constant holomorphic mapping of Riemann surfaces and x € X, then in suitable local coordinates at x and f(x), f looks like 2 > 2", na positive integer. We say f is equivalent to z— 2" at x. (In particular, f is an open mapping.) n is said to be the multiplicity of f at x,and n-1 is said to be the br or ramification of f at x, denoted ram (f). The ramified points x, x where ram (f)>0, area discrete set in X. x Theorem. Let f:X —~ Y bea non-constant holomorphic map between compact Riemann surfaces. Then there exists a positive integer n so that for any y € ¥ necard{ yy} + Dram. f(x}=y n will be called the number of sheets in the covering X > Y. If y= 1 then f isa non-constant meromorphic function on X and2 n is called the order of f , denoted o(f) . A non-constant holomorphic map of compact Riemann surfaces will often be called a covering. e li ngs. Let f:X + Y be a non-constant n-sheeted holomorphic map between compact Riemann surfaces. Let p and q be the genus of X and Y respectively. Let ram(f) = > ram, (f). Then xeX 2p - 2 = n(2q- 2) + ram(f) . Definition. if X, is a compact Riemann surface, the subscript p will always denote the genus of x) . Corollary. Let £:X, > X, bea non-constant n-sheeted holomorphic mapping of compact Riemann surfaces where n 22. Then p2q with equality possible only if p= 0 or 1. 4.2 Function Fields, If X is a Riemann surface let M(X) denote the field of meromorphic functions on X. If f:X > Y is an n-sheeted holomorphic map then {*:M(Y) > M(X) maps M(Y) onto a field which is of index n in M(X). Conversely, if X is a compact Riemann surface and K is a subfield of M(X) of index n., then there exists an n-sheeted covering f:X — Y of compact Riemann surfaces and t*(mcv)} = K. A rational function field is a field isomorphic to M(P 4) , that is, the rational functions. A meromorphic function field on a Riemann surface of genus 1 wil be called an elliptic function field A Riemann surface x p= 2, will be called hyperelliptic if M(X,) admits a rational subfield of index 2. This is equivalent to X, admitting a 2-sheeted covering X > P 1 | The interchange of the two sheets of this covering 1s an automorphism of X_ of order 2 called the hyperelliptic involution. A hyperelliptic Riemann surface, X_, is the Riemann surface for a polynomial P(z,w) « Cfz,w] where 2p+2 2 P(z,w) = wo - II (2-a,) where aja). jel A Riemann surface of genus greater than one will be called elliptic-hyperelliptic if it admits a_2-sheeted covering of a Riemann3 Surface of genus one; that is, its function field admits an elliptic function field of index 2. In general, if X is a compact Riemann surface then M(X) is the Riemann surface for some (many) irreducible polynomial P(z,w) € €lz,wl. |. Let f(x,y,z) be an irreducible homogeneous polynomial of degree n. Let C= {(x,y,2) € P20) [fey 0) =o}, an irreducible plane curve. Suppose C_ is non-singular, that is, f,06,y,2) = O= £ Gry,2) =f (,y,2) implies (a,y,2) = (0,0,0) We show that asa Riemann surface, C, has genus {9-1)(n-2) | c. has only a finite number of inflection points (where the tangent line has more than two intersections with c). Choose coordinate axes in P2 so that none of the inflectional tangent lines pass through (0,1,0) fie. no inflectional tangent line is parallel to the y-axis). Then each tangent line to C_ passing through (0,1,0) has only two intersections where it 1s tangent. Such points occur where f= Gand {= 0. At such points the projection of C_ onto the x-axis is locally two-to-one; n that is, this projection (a holomorphic map onto the x-axis (= P+)) has ramification one at each of these points. The number of such points is n(fx- 1). y Using the Riemann-Hurwitz formula we have 2p- 2=-2n+nin-4); 2p-2=n(n- 3)4 We can use the same method if C_ has singularities. Suppose C_ has an ordinary singularity of multiplicity k at (0,0,1). Dehomo- genizing f(x,y,z) to f(x,y,1) we have k f+ I (a,x + B,y) + higher order terms isl where the k tangents lines a,x +p,y = 0 arealll distinct. Thus k-1 I (8, + 6y) + higher order and so {= 0 and f= 0 have ist k(k~ 1) intersections at (0,0,1). Now assume C_ has s ordinary singularities of multiplicities ky 7 wok, Then the total ramification for the projection of C, onto the x-axis is n(n ~ 1) ~ >» ky (ky =.) Ot ky (ky~ 1) = a=i)(n-2) | Dee 2 Let us look more closely at singularities of muluplicity two at, say, (0,0,1). f = ax? + bxy + cy + higher order. £ = bx + Icy + higher order. v Local parametric equations for {> 0 are x= 2ct +o,t? tee y=-bt+ b,t? fees, Thus f(x,y) = a(4c?t? +--+) + b(-2bet? +--+) + o(b?t? + +++) + oft 3) = cldac ~ b)t? + oft) . If 4ac-b*=0,f=0 and {, have two intersections (node). if dac-b?=0.f+0 and {, = 0 have at least three intersections (cusp). If f = (ax + by)? + (ax + by)Q(z,y) + 4" degree terms and higher order, then f= 0 and f = 0 have 4 intersections (tacnode). Nodes add nothing to the ramification of the propection map. Cusps add one to this ramification and tacnodes add nothing.5 If 8 = * of nodes, x = * of simple cusps, n = * of simple tacnodes for C. and there are no further singularities then 2p- 2= -2n+n(n- 1) - 28 - 3K - 4n+K p = fn=iyin=2) -8-K-2n. For a general plane curve of degree n we have = G-1)(n-2) | P 2 8 where § will be called the S-invariant (sometimes called the number of nodes suitable counted). Ordinary k-fold singularities add K(S=1) to the 8-invariant . Simple cusps add 1 to the S-invariant . Simple tacnodes add 2 to the S-invariant . 14 _Divisors on Riemann surfaces. A divisor D on a Riemann surface X isa singular zero-chain , written D = Ss nz, for nye Z and x _ €X, The desree of D.deg D, is Dyn,, an integer. If m2 0 for all i then D is said to be positive (or integral) and we write ° D2. The zero divisor is positive. If D= Yaz, and D'= dxz, are two divisors then by the greatest common divisor of D and D', written (D,D') , we mean the divisor), min(n,.n)2, . Thus to write (E,2) = 0 1s tosay that the coefficient of z in E is zero. If (E.F) = E we will write (C126 If f is a meromorphic function on X,, the divisor of f is written (f) where (f = (zeros of £) - (poles of £) oF 0 Pp Note that deg(f) = 0. Two divisors D,E are said to be (linearly) equivalent if there exist an f¢M(X) and (f) = D-E (an equivalence relation). Notation: D=E. Definition If D is a divisor, L(D) = {fr « M(x)|() + D 20}.6 Theorem. If X is compact then L(D) is a finite dimensional vector space over €. Definition. If D a divisor on X , 2(D) = dimgL(D) , the affine dimension of L(D). r(D) = 8(D) - 1, the projective dimension of Li). A meromorphic (or abelian) differential w is a one-form that locally can be written w = flz)dz where { is meromorphic. The value of a meromorphic differential at a point of X is not well defined; however, zeros and poles are well defined so we can consider the divisor of w , (w) = (zeros of w) - (poles of w). If ®, and w, are two meromorphic differentials then @,/H, isa well defined meromorphic function on X. Since deg(«, /w,)= 0 we see that deg(w,) = deg(w,) Theorem. If is a meromorphic differential on X_ then deg(w) = 2p - 2. Definition. If D is a divisor Q(D) = {meromorphic differentials «| (w) - D 2 0} i(D) = dimension of Q(D) asa vector space over C. Lemma. If K is the divisor of a meromorphic differential then i(D) = &(K - D) Now we list some of the basic theorems concerning compact Riemann surfaces. Riemann-Roch theorem. If D is a divisor on a compact Riemann surface of genus p, then r(D) = degD - p +i(D). Brill-Noether formulation of the Riemann-Roch theorem. If D and D' are two divisors so that D + D' = (w) where w is a meromorphic differential, then degD - 2r(D) = degD' - 2r(D') . Notice that the divisors in the Brill-Noether theorem need not be positive. Definition. A meromorphic differential without poles is called holomorphic (or analytic, or regular, or an abelian differential of the first kind). Definition. If D>0 and i(D)>0 then D is said to be special and i(D) is often calied the index (of speciality) of D . Clifford's theorem. If D is special then degD - 2r(D) > 07 Definition. If D is a divisor, the Clifford index of D, denoted c(D) , is: c(D) = degD - 2r(D) = p- r(D) -i(D). ‘Strong form of Clifford's theorem. If x admits a special divisor D where degD = 0, degD = 2p-2 and c(D) = 0, then x, is hyperelliptic. Definition. Om) = {holomorphic differentials on x) . Corollary. dime Q(X,)=p. If w¢Q(X) then i((w)) = 1. 15 Linear series. Let D bea divisor on X_. Then the complete linear serigs determinant by D , denoted IDI, is Dl {D2 0|D'=p}. IDI can be empty. If D2 0 then IDI isa set of divisors parametrized by the projectification of L(D). pis {o|D-D= 0, fe LO}. Let S bea linear subspace of L(D) of dimension r+1,rsr(D). The set of divisors parametried by S (o|p-p-w tes} iscalled a linearseries and denoted g_ where n= deg and r+1=dimS. 2 If r-D- Fes), ScLo). |, bea basis for S. If xe X let B(x) = (f,60,--+,f wo) € PTE). If we pick a local parameter z at x, 1 2%) =0 then8 f-2™4g (2) where g,(0) =0. Locally z is mapped by © into (++-,2%g.(2), 8,(2),-++) and so 8 Let n= min{n,). Then 2 ( nin x Genz! a2) y1++) and at least one component is 2=0 non-zero. That is, @ is well defined on all of X. Let C= 6(X) C PY be the image of X under @. Then the hyperplane sections of © pull back via @ to the divisors of g”_. For r if Saye 0 isa hyperplane H in P* then i=0 OeHnCe Sate = 0 x isa zero of Dag, . Tf 4 then D is the pull back of the hyperphane (at ») eo If the map @ is one-to-one in general, then s is said to be simple. In this case X is (conformally equivalent to) the Riemann surface (or normalization) of C. (C can, of course, have singularities.) Each hyperplane in PY cuts C in n_ points (counting multiplicities) so we write C, for C,a curve of degree n in PY If the map © is not one-to-one in general then 7 _ is said to be composite. Then X isa t-sheeted covering (t 2 2) of (the normalization of) C and C has degree n/t, since ce has no base pomts. (Thus a linear series of dimension r 2 2 of prime degree greater than r without base points is simple.) In this case, if Y, is the normalization of C then Y, admits a e. If n/t :X 4 Y_ isthe t-sheeted covering then g? on Y_ lifts to g” P q n/t q a . The fibers of this map w are called an involution of genus ¢ on Pand denoted x? In this case we say that _g* _is compounded of the a involution a . (In the sequel, we shall be careful not to confuse the two meanings of the word “involution’.) Given g* and ©:X ~ PY, @ depends on the choice of bases {fy.-++,f.) un S.C LID). A second choice of basis gives another curve in PY , but the two curves are related by a projective transformation whose matrix is that which relates the two bases of S. It e 1s simple, then @(X) ¢ P® is called a plane model for X. For a st without base points @ :X ~ P! corresponds to an n-sheeted covering of P!. If x, (p 2 2) admits a sty then Xp is hyperelliptic, and conversely. If x admits ag? then p=0. If X admits a complete r g »r22,then p=1. rei The canonical curve is the image of X by @ in PP"! given by Pp the canonical series g?~! grt is base point free. It is com- 2p-2 2p-2 posite if and only if x is hyperelliptic in which case it is compounded e gi of the gi, For a non-hyperelliptic X, , the canonical curve is a non-singular plane quartic, C, where the lines in 2 cut out the canonical series. If g¥ is acompiete linear series and Deg’ then r=n-p+i(D). a a Since i(D) does not depend on the divisor chosen, we will occasionally write i(g™ ) in place of i(D) n lf gis a complete linear series and D is an integral divisor, by r lpr ~ Dl we mean the divisors E of g* containing D, minus D. n {E-D|Ees', DCE} . if the degree of D is m then @r Deere and we say10 D_imposes _t (independent, linear) conditions on cee If IDI = c= then we would write Bee aa n °m a-m gs since any divisor in g*__ will impose t conditions on g™ (Of course, m a. ‘we are assuming some divisor in g* contains D.) n Asa final remark, when considering linear series one should always keep in mind the three possibilities: simple or composite; complete or incomplete; base points or no base points. 16 Line bundles. We give now a very brief discussion of line bundles. We refer to [13] and [14] for a more complete discussion. Let {(U,,9) |x € - (0) is holomorphic OB ~ e Mg =1 cco 8) Sup By Byq = 4 - 8 8 * Bay: Example: The canonical line bundle. If for « and p we have the following diagram Ya/ \% h w= h(z) z-plane A | A, ¢ 2-plane Cc w-planen then g,,7 48 z by K. A holomorphic (meromorphic) section of a line bundle n is a set of oe « ° @, - The canonical line bundle will also be denoted holomorphic (meromorphic) functions {s,]% € A} (or P+) so that s =g s inU NU. x Fans % The set of sections of a lie bundle form a vector space over €. For example, the sections of the canonical bundle are differentials. Theorem. On a compact Riemann surface the vector space of holomor- phic sections of a holomorphic line bundle is finite dimensional Two line bundles n = e,.,) and v= fh.) are said to be equivalent if there exists a set of non-zero holomorphic functions tt, | ae A} and tg =h t aap OB 8 Thus if {s,} is a section of n then tts is a section of v. Thus the sections of equivalent line bundles can appear to be quite different. What is invariant? (a) If (s,) isa section of a line bundle then the divisor of (s,} makes sense since the transition functions are non-zero holomorphic functions. Let ({s,}) denote the divisor of the section {s,}. Then the set of divisors of holomorphic sections of a line bundle is independent of the equivalence class of the line bundle. (b) If (s,) and {s',} are two holomorphic sections of a line bundle n then {s, /s',} defines a global meromorphic function. Therefore ((s,)) = (6,2) . So the divisors of holomorphic sections of a line bundle form a linear series which is easily seen to be complete. ic) If sy+s, are r +t linearly independent sections of 1 then wecan map ¥> PF by x > (5,(2),--+,5, (0) (The curve in PY is determined up to a projective transformation of PT corresponding to different choices of bases in , the span of those sections.) The curve is independent of the equivalence class of n. (d) If D is adivisor on X we form a line bundle [D] as follows. In12 U_ let f_ bea meromorphic function whose divisor is DN ve . Let Bg Sit in us NU, , anon-zero holomorphic function. (If DA uy = 0 let fy = 1.) Then clearly ff) is a meromorphic section of (@,.,} {= (DI) , whose divisor is D. Any holomorphic section of [D] will have a positive divisor equivalent to D. Thus vector spaces of holomorphic sections of [D] will parametrize linear series of divisors equivalent to D. (e) There is a one-to-one correspondence between cornplete linear series of projective dimension greater than or equal to zero and holomorphic line bundles admitting holomorphic sections, given by DI [ID] In= e ; 1 n= (g,,) and v= (h,,} areline bundles then (g, ,h, ,) and {g"1} are also line bundles, the former denoted nv (or n@ v) called OB the (tensor) product of n and v, and the latter denoted n°! called the inverse of n. If K is the canonical line bundle then K@K@-++@K (k times) or K*, will be called the k-canonical line bundle (or written additively, kK) whose sections are called k-canon- ical differentials . With respect to a local parameter z a k-canonical differential is written f(z)(dz)* and the transition functions for kK k are (48) * 4.» referring to the diagram earlier in this section.Chapter 2. Some exceptional points on Riemann surfaces. 24. Ona Riemann surface, X_, the points where a holomorphic differential has a zero of order greater than p-1 are called Weierstrass points. There are only a finite number of such points, at most (p-41) p(p +1). If there are fewer then to each Weierstrass point can be associated a weight and the sum of all the weights is again (p- 1) p(p +41). The maximum weight at any point is p(p - 1)/2, and if the maximum weight is attained, then the Riemann surface is hyperelliptic. In this case the Weierstrass point is a fixed point for the hyperelliptic involution, that is, a branch point for the unique _ on x, The purpose of this chapter is to extend to more general linear series these results of Hurwitz [15]. (2.2. Let g* bea linear series without base points but not necessarily n simple or complete. Let 515, line bundle L determined by g¥ which generates the divisors in g’_. n a coer oe be a basis for the sections of the Fix x) in X. Ina local parameter 2 at x 2.) = 0 , consider the Wronskian W(s, ney) > 8, (2) s,@) sone s, (2) sg (s) om (2) sees s. (z) ©) ta) Ga +--+ Sh) Ga) Gere co) = as (2) , the k'® derivative of s with respect to the local ee 2. If we choose a different basis the two bases are related by a constant matrix whose determinant is A and W(s),-++,5.) = AW(ty ret). 0 r 0 r Thus the order of vanishing of Wis, . os) at x, is independent of 0 the basis chosen for g* n 2.3, We now choose a basis for the sections adapted to x, as follows. Using the local parameter z, 2(x,) = 0, we will find integers14 O-n will be a basis adapted to Xy and the son Xp equal ton, ¢tys++ order of vanishing of W(t,,.-+,t.) at x) will be r > fm, ~ J) . This last sum will be seen to be independent of the J=0 various choices made. Let ny = 0, and let t, bea section not vanishing at x, (since oo n is without base points) . Let EB = (t) a divisor disjoint from Xy: Let S, = {sections s|s(0) = 0}. Let n, be the smallest order of vanishing at 0 of sections in s,. nye 1. Choose a section u which vanishes precisely to order n, at 0. Then (t,) = myx) +B, (Hy.E,) = 0 1 nen.) * and E,€@ A> 14%) : (n,) Let S, = {sections s|s * (0) +0}. Let ny be the smallest order of vanishing at 0 for s€S,. Choose t, €S, which vanishes to order n, at z=0. njon,. 204 } }= Then (t,. = Nyx) +E), (x ,E,) = 0 and Ey € 8 y7 MgXq) (= 8 and integers Continue. We obtain sections (n, - j) . We follow Farkas-Kra (11, p. 78]. j=0 First we note that if g is holomorphic in a neighborhood of x, then egrth wee . W(sty.gty,-+,gt_) = gF°! Wity,t,,-++,t,) . For a typical column of the matrix whose determinant is W(g tyr swt) is Gt et tet g"t + 2g tiret, oe Using elementary row operations this reduces to — and the result follows. It now follows that the order of vanishing of Wis, oo 8.) at x, does not depend on the local parameter. Now to prove the order of vanishing of W(t,,-++,t.) is > (ny ~ j) we proceed by induction on r. For r= 0 the result is immediate. Suppose it is true for integers less than r . W(tosty, t sity) 2 ty) * wag or) 0 red = wh eG’). t. Now ¢ a =z *hy(2) where hy (0) = 0 0 a, apt =z °hi+ h te 3 apt apn @ Neen, +njhy. na-1 Therefore W(tg=5t,)= (2) Wiiys-+-f,) where ne fo! : 5, and g,(0) = 0. By induction W(f,,-++,f,) vanishes to order16 5 Dey-ay-G-v. a Thus Wty, tw) vanishes to order 5 ray - 1) + Dulay-ny J+) jea r = Xa,- 4 (ng = 0). j=0 25. We now show that W(s,,-++,s,) is a section of the line bundle Lrttgr(r+1)/2 where L is the line bundle corresponding to the completion of g" and K is the canonical line bundle. n In UAV. the intersection of two parametric neighborhoods of x, with local parameters z and w let g be the transition function for L. If s, isa section, let f (2) and hw) be the representatives of s, in U, and V._ respectively where w = w(z) is biholomorphic. Then £(2) = gah (w(2)) gh.) = aft w(h, (w(2)},-+-.h (w(2)}) . dh, New Br (em) wen 2 W(fysrrf) = Wighy Again by elementary row operations we see that the it® column in the determinant becomes 2 : Re ee ee (hate w ew eed row) es) Thus we see that Wtyserat,) = a GW!) O12 wee sh). r7 Since g¥*t(w')F(r*4)/2 is the transition function for LT*1KY (r+1)/2 the proof is complete. It follows that the degree of W(s, (r+ A)n + (r(r + 1)/2)(2p - 2) =(r+i)(n+r(p-1)) (9, p. 432]. 2.6. We now make some definitions and summarize the results of the last several sections in a theorem. Let g’_ bea linear series (without base points) which is a subseries n 1s.) is r of the complete linear series corresponding to a line bundle L. A point xeX, where a divisor of g" has order greater than r will be called a n generalized Weierstrass point for g’ The Weierstrass weight of g* a a at x, denoted w(g?_, x), is computed as in Section 2.4. a Theorem 2.6. The total sum > w(g", » X) is equal to xeR, (r+4)(n + r(p -1)) . The divisor of generalized Weierstrass points for g*_ (counting multiplicities) is a divisor corresponding to the line bundle Lente lert)/2 . If a is complete and non-special this total is (r + 1)2p. In particular, if g"_ is the k-canonical series , which we denote by kK (k2 2), then this total is (2k -1)?(p - 1)?p . A generalized Weierstrass point for the k-canonical series will be called a higher order Weierstrass point. It is known that the higher order Weierstrass points are dense in X_ [22] If gf = IGl, a complete linear series, we will often write w(G,x) for wg ,x). a For a linear series g* and x X, the integers nynj,-++,n_ will n 7 be called the non-gaps for g*_ at x. The complimentary integers in n (0,n] will be called the gaps for Ca at x. (For the canonical series18 note that this terminology differs from the usual terminology.) Then r nr we Dy -d* Dig) jet where g,,°++8,_. are the gaps for g’ at x. If welet A, denote a " a r non-gap greater than r (an extraordinary non-gap) and let a denote a gap less than r +1 (an extraordinary gap) then Da-8 = 4 and the number of extraordinary gaps equals the number of extraordinary non-gaps. 27 Examples. The generalized Weierstrass points for g! are the a w(8 9%) branch points for the cover X= P! given by a. The number is 2(n - p+ 4) which agrees with the Riemann-Hurwitz formula. For a simple g?__ the inflection points (flexes) of the corresponding plane model, C, are among the generalized Weierstrass points. However, one sees that cusps also contribute to the total, in particular, a simple cusp contributes 2 to the total. If the plane curve is non- singular then the total weight is 3(n + 2(p - 1)) = 3n (m - 2) since p= (n - 1)(n - 2)/2, and this counts the number of inflection points. For p= 3 and x, non-hyperelliptic, the c, corresponding to the canonical ae has 24 flexes (counting multiplicities). 28 Theorem. Let g¥ bea complete linear series without base points. The maximum weight at a point of xX is (r= n-p+i) (m-r)(n-r+1) | (p= i)(p =i +1) 2 2 The maximum is attained only on hyperelliptic Riemann surfaces and only at hyperelliptic Weierstrass points. Remarks. For a = ce this gives Hurwitz's result. The method of proof of the theorem is that of Segre [24], [9,p. 292] . Proof. For Xo € Ls we find a basis of sections for a line bundle corresponding to g’ adapted to Xy as in Section 2.3. We obtain n19 =n, ¢ On s. Then s21 and for So, - nen, special since r jzs.g (n-n)- (r= J =n-r-ny- Den- By Clifford's theorem for j 2s aon, 2 2r-j or n- Ort jz Now nyzstt 50 n-ar+sen -s2t and s22r-n+1. r Computing: a = x (a;- J) J < > (n- 2r + j) r-n+h ar c > x = Crest). 2 ket If we have equality in the last formula we have equality in all the preceeding inequalities, and so X, hyperelliptic by the strong form of Clifford's theorem. In particular if we let j= r-1 then n-n 2Qr-ar+2=2 r-1 and n-n r ore Oa Consequently IE,_jI= gt, and g” = Inxgl. W follows that 2x, € g1, and so x, is a hyperelliptic Weierstrass point. qe.Chapter 3. The inequality of Castelnuovo-Severi. ZL As motivation for this inequality we consider for hyperelliptic Riemann surfaces some properties which are part of the folklore of the subject. These properties will be generalized by the Castelnuovo-Severi inequality and also by some later results on automorphisms. 3.2 Lemma. Suppose - is hyperelliptic and f ¢ MX). Let w: XP? bea 2-sheeted cover. (Then m is a meromorphic function of order 2.) Suppose o(f) < p. Then there exists a rational function R(&) € €(%) and f= Rem andso o(f) is even. Proof. We may assume that no poles of f lie on the branch points of m. For if T is a fractional linear transformation then Tf will have the same properties as f except that by choosing T properly the poles of T° can be moved away from any finite set on X_. If the conclusion of the lernma holds for Tf it will hold for f. Let B be the image of the 2p+2 branch points of m. For ze P1-B, let wz) = (x,,x,) andlet g(z) = (F(x,) - fx,))*. Then g is meromorphic on P!-B and bounded near B and so can be extended to be meromorphic on all of P!. g , so extended, has at most 2o0(f) poles, and atleast 2p+2 zeros at the points of B. If g(z) is not identically zero then 2 olf) 2 # of poles of g = # of zeros of gz 2p +2. But off) sp, so g is identically zero. That is, f identifies the fibers of mw, and so f is the lift of a function R from P1. f= Rem where R is rational. qed. 3.3 Lemma. For p22, the sty on a hyperelliptic X, is unique. Proof, A g1_ is the set of fibers of a meromorphic function of order 2. If f has order 2 then f= Rem where 1 is the original function of order 2. Thus R must have order one in M(P'); that is, R is a fractional linear transformation. Consequently m and f have the same set of fibers as maps from X, ~ pi. qed. 3.4. Here is another result that asks to be generalized but which will not be considered further since the methods considered in these notes do not seem entirely appropriate. Lemma 3.4. Suppose x, is hyperelliptic and 1 2X o> _ is a map of24 compact Riemann surfaces. If qz2 then Y, is hyperellptic. Proof. Let f bea function of order 2 on X_. If the map XY, Pp q is n-sheeted , for almost all y in Y_ let wy) = {x,,++.x.}. Define aly) = FGx,) £06,» £0) which can be extended to be a well defined meromorphic function on Y If zeros and poles of f happen to lie in the same fiber of the map t, replace f by f- % for some }, of €. Then gly) will be a function of order 2 on Y . q qed. vo"! i, Theorem 3.5. [8]. (i) Suppose we have three compact Riemann surfaces related in the following manner where the holomorphic maps w and p have m and n sheeets respectively. (ii) Suppose further that the diagram admits no proper factorization; that is, there does not exist an x, with p'

(n- 1)? then any a without base Points is unique on X, . (iv) Suppose p > 9 and xX admits two complete atys without base points. Then with q= r= 0 and m=n-=4 wesee that the diagram admits a proper factorization and X_, admits two complete atys. Consequently p'=1. Thus X, is elliptic-hyperelliptic; that is, a 2-sheeted covering of a Riemann surface of genus one. BZ Some remarks, An equivalent way of stating hypothesis (ii) of Theorem 355 is the following. [x*(ex) »pr(Mox)}] = McK) where the bracket indicates the field generated by the two fields inside. The theorem of the primitive element then assures us that Mx.) = p*(MOxX))Ial where g is some function in n(Mex)- It follows that for almost all yeX, p(y) isaset of n distinct points on which g takes ni distinct values. From this it follows that almost all divisors in x4 n are composed of n distinct points.23 3.8 Proof of Theorem 35. Let got be a complete, non-special linear series on x, n-ltg without base points. If IDI = g™t let f+ n-itg a LD). got tag lifts to X, to become a got nit B necessarily complete). Let f. = f be a basis of pn m(n-i+q) (no longer pom Me ye X let pty) = (x,,-++,4 } and define h(y) as follows. f op 2 le (rx )erees 2 lf, Ox) fG) £,(rx,) £(rx) bly) = = f ix )er ssf) if (rx,) +--+ (rx) ni non a 4 a a We prove the inequality in 3 steps: (i) We show gt can be chosen so h(y) is not identically re teg zero. (Gi) We then show that the number of poles of h(y) is less than or equal to 2m(n-1+4q). (iii) We show the numbers of zeros of h(y) is greater than or equal to ramp = 2p-2-n(2r- 2). Since the number of poles of h equals the number of zeros we have 2m(n~-1+q) 2 2p-2-n(2r- 2) which is the inequality of Castelnuovo-Severi. (i) Suppose that h(y) is identically zero (h(y) = 0). Then every divisor in st, must lie in a divisor of get beg: Frit hly) = 0, Ley) = fx. joe : on P (y) = &,, ox), then there exist aye v4, € ©, not all zero, p and > ajfjrx,) = 0 forall i. Thus (wx, ,-++,1x,} is part of et the divisor24 p= {ze X, | > ajf@=o}. 1 opt = tee is Now fix y, €X_ so that w+ p"l(y,) ( (nx, wx) is a set of n distinct points, and no x isa branch point of p. Let D_ bea 1 q non-special integral divisor on x of degree q, and mx, €D,; Djl=g°. Let ID, twx,te-enx, leg” n q’ a ce q 7 ‘n-4 n-l+q ” Since this linear series is non-special r = n- 1. Moreover, gt ~ Wx, -+++- mx | = D_. Thus no divisor of gt contains n-1 q os nitq ot teq WK tet WX Ot UK . 1 a-ln (ii) and (iii) We will combine these proofs in a lemma. Lemma 3.8.1. Let p: x, > x be an n-sheeted covering of compact Riemann surfaces (n 2 2). Let D bea divisor on X_ so that IDI = aR where Ren-1. Let f,,...f, be m independent functions in L(). For ye X, , let pty) =, x}. Define hy) by f,@) i@) 2 hly) = . . 1,%,) 1, @) If h(y) is not identically zero then Nz p-1-n(r- 1). Proof. Since R21 we can assume that D is a divisor of N distinct points which are disjoint from the set of branch points of p. Then if k points of D arein p-1(y) one sees that h has a pole of order at most 2k at y. Consequently h has at most 2N poles. Now we show that if xy¢X, and ram, p=s then ‘0 x 0 contributes at least s-1 to the zeros of h(y). Xy (Section 2.3). Call this new basis shysersh > . Ina parametric neighborhood of Xo U, with Choose a basis of adapted to local parameter t (tx) = 0) the map p is equivalent to t+ tS=z,25 where z isa local parameter at p(x,) €X_ (2(p(x,)) = 0). Using these local parmeters we have pl(z) NU = {t,tt,-+,t7 3h} where t= &%/S and tS =z. Then eit h(t) = . kK, oe h@) st k) " A (t)=t™k (t) n a where Os 8c cor cb jst, and K,(0) = 0. Using these parameters we have s-1 2 h hid hy (re) s-1 h(t) h(t) y(t 1) h(z) = a2 hott) a (rt) hire *t) (A = 0) sol hl how) hy (ett) h(t) h (tt) h (ttt) in a a Each ss determinant from the first s rows hasa zero (asa function of t) of order at least O+1+2+++-+(s-1) =s(s- 1)/2, Thus as a function of t the square of the determinant is divisible by +s) But t8=2,50 h hasa zero of order at least s- 1 coming from the first s rows of the determinant. Thus h(y) has at least ramp zeros and ramp = 2p - 2-n(2r- 2). Since h(y) has an equal number of zeros and poles the result follows. qed.26 This completes the proof of the Castelnuovo-Severi inequality. Now suppose we have equality. Then every inequality in the proof becomes an equality. In this case the zeros of h(y) arise only from the ramification points of p. Any g®“! tog for Which the n-1+q corresponding h(y) is not identically zero has no divisor containing a divisor of 1 for that would add another zero to hy). If a gui tog 18 chosen so that one divisor of si is in one of its divisors neLeq n then every divisor of 1 lies in some divisor of g™! q since ly) a m1 a Now pick D, € v1 so that D_ has distinct points . Choose G to a be a non-special divisor of distinct points of degree q. Then IG+D |= g™ and we can assume it is without base points. a 8 neg Since [Gl = g? , the n points of D_, impose independent conditions q on g®_ ; that is, if nt Dixy te te tk, then a oes pe ene Now g™ givesa map @: x, > P". The statement that D. ntq imposes independent conditions on g" 4 translates into the statement: nt @(x,),--+,8(x,) span a linear space of dimension n-1 in PP. This is an open condition; that is, for E close to D (degE=n,E20) Ewill also impose independent conditions on g™ 4 ne Lemma 3.8.2. (Castelnuovo.) If gE ev! , then E imposes independent a conditions on g" | if andonly if E =D, . ne Proof. Let G=P, +--+ +P, ,q distinct points. Consider no =P = [ID + Pl tee egnl 8 ea P,=ID,+P, +P See: Consequently g®t 1, Contains all divisors of x1 Suppose neq-! a , which contains D,,7 Ee x! imposes independent conditions on g™ n ntq n-1 lke Stet gE HL Thus a + Bg en He TP! n Ee 0 Bg En Mg Paes : since E, imposes independent conditionson g" —_. Therefore neq Pp, eg™ -E.. But there was nothing unique about P.- Thus ntq Peg? -E j 87 7B, forall j gh -E =P, +.++P =G ntqg ont 4 eg" =(D +Gl=|E +Gl ng n a or D#E. ees Conversely, if D =E then g? -E = ,s0 E_ imposes n n ntg aq i independent conditions on g” neq qed. We complete the proof that equality in the Castelnuovo-Severi inequality implies all divisors in y!_ are linearly equivalent by a a connectivity argument. Let x‘) = {all integral divisors of degree n on X,) and let q #4 = (divisors in x1 consisting of n distinct points). Then is a a a connected subset of X) when the latter is given the natural topology. Fix D_¢ #1. Let S= {E ¢ 1 |B =D}. Then S is aoa 0 On easily seen to be closed in ¥1_ . But the lemma of Castelnuovo shows a that S is open, that is, S = B. Since §1 = y1_ - {finite set of nn divisors} the proof is complete. te ' the Theorem 3.9, [9, p. 324], [61 Suppose sis an algebraic series of28 index one and genus q on X,; thatis, 31 is the set of fibers of n an n-sheeted map X,~ X,. Let ef be a complete linear series on X where R2n-1. Suppose N-Rép-nq. Then any divisor in a will not impose independent conditions on en: that is, if Dey! then R _ pe gR-ntlee bey Dies ce where ¢20 Remarks. Notice that aa may be composite and that we may assume on is without base points without loss of generality. Clearly g 7 must be special. For q = 0 the theorem is a consequence of the Riemann-Roch theorem. Proof. For the proof we will assume g 7 is without base points. Also for technical convenience we will prove the theorem only for divisors Dex! made up of n distinct points. a Let D =x, +++ +x, bea divisor in xt made up of n distinct a points. Suppose D_ imposes n independent conditions on _ We will arrive at a contradiction. Let Dp =x, teeta |. If Ren wecan finda divisor Y of R-n+41 distinct points so that a +Y imposes R conditions on an , and the unique divisor in g containing Y+D,_, does not contain a] R _yjegnt q Now 1gR-Yi= gt. Since N-R+en-1 A isa homeomorphism. Definition: Suppose f : ¥ > X is a local homeomorphism between surfaces. Let ¥:[0,1] ~ X be a continuous map (path), and suppose 5(0) = x). Suppose ¥:(0,1] > Y isa pathso that ¥(0) * Yq: flyp) = xy ,and fe ¥=y. Then F will be called a lift of ¥ starting at yy Lemma If ¥ and ¥' are lifts of ¥ starting at y, then (uniqueness of lifts). Lemma. If f: Y > X isa local homeomorphism between surfaces, then the following are equivalent: (a) f£ is a covering; (b) If y:[0,1] +X isa path and fly) = x(0), then there exist a lift of ¥ starting at yy. Monodromy Theorem. Let £: Y > X bea covering. Let ¥,,%, be paths in X so that (0) = ¥, (0), ¥)(1) = 8,(1), and, is homotopic to 4, . Let i and 5 be lifts of ¥ and ¥, starting at the same point above ¥,(0) . Then % is homotopic to 5 , and in particular Rw = RM. Theorem. Let £:Y > X bea covering. Then card(f~!(x)} isa constant function on X , possibly infinite. This integer (or ) is called the number of sheets of the covering. 4.2 Classification of smooth coverings. Let X bea surface, x,€X. A triple (Y,f,yq) will be called a covering of (Xx) if f£: Y > X is a covering of surfaces, Yo € Y,and fly) = %- If (Y,£,¥5) and (Z,g,2)) are coverings of (Xx) » they will be said to be equivalent if there exists a homeomorphism h: Y ~ Z so that hly,) = 2 and goh=f.34 Suppose (Y,f,yq) is a covering of (X,x,). Let [¥] be an equivalence class of paths in 1, (Kx) - Let Dc 1, (X%) be the paths [¥] so that the lift of ¥ starting at Yq isa closed path. By the monodromy theorem D isa well defined subgroup of 1, (X,x,) . In fact if we define f=, Qay9) > (Km) by flo] = fea] then D= £,(n,(¥,¥Q))- Existence Theorem. If D is a subgroup of 1, (Kx) , there exists a covering (Y,f,y,) of (X,x,) so that D = £,(n,%yQ)) - Theorem. Suppose Dc D'c (Xx) and let (Yfy9) and ete r¥'y) be coverings as in the existence theorem. Then there exist a covering g:Y ~ Y' so that (Y,g,yq) is a covering of (Y",y',) and f=fog Corollary. (Y,f,y9) and (Y',f',y'9) are equivalent coverings of (X,x9) if and only if D= D’ Suppose for each subgroup DC m, (X,x)) we pick one covering (Lp rty¥p) as in the existence theorem. We define a partial ordering pifpr¥p)? Wp. fy ¥py) if there exist a covering g:Y) > Yp, so that (Y),g, %»? is a covering of (Yp,,¥p,) and {y+ fp.eg. Then (Yp.fp.¥p)? Yp. fp. Yp.) ifand only if D ¢ D’. Thus the map D > (Wp sfp»¥p) is an order reversing lattice isoror- of these covers as follows: (Y, phism. A covering (Y),fy,yp) corresponding to D = C m,(X,x)) LY; DD is called a (the) universal covering of X and Yo is simply connected. 4.3. Lifting conformal structures. If f: Y > X is a covering of surfaces and X is a Riemann surface then there is a unique conformal structure on Y so that f is holomorphic. For if xX and A isa small simply connected open neighborhood which is mapped by h biholomorphically onto {lzl< 1} then on f-!(A) = Ua, het restricted to each A, gives the conformal structure for Y. If X is the universal covering of X then & is a simply connected Riemann surface. Uniformization Theorem. Every simply connected Riemann surface is conformally equivalent to P!,€,or D (the unit disk). 44. Cover transformations. Notational convention for Sections 4.4, 4.5, 4.6, and 4.7. We will write all functional notation with the argument on the left; eg. x (x)f. Then the composition of f followed by g will be written32 fg or fg. This notational convention will help distinguish between isomorphisms and anti-isomorphisms which occur naturally in discussing cover transformations. Suppose f: Y > X is a covering. A homeomorphism g:Y ~ Y will be called a cover transformation if g°f =f. Under composition the set of cover transformations is a group denoted G(Y/X) and is called the Galois group of the covering. If X and Y are Riemann surfaces and f is holomorphic the cover transformations are biholomorphic. 4. A cover transformation with a fixed point is the identity. Proof. Suppose g € G(Y/X) and (y,)g = y,- If y, isany point on Y let_y, = (yg. Let ¥ be a path from Yq to y,- Then Fog isa path from y, to y,. But (¥)f = (¥og)f=¥,apathin X starting at x. ¥ and ¥og are lifts of y starting at ¥q- By the uniqueness of 0 lifts ¥= ¥eg andso y qed. Theorem 44. Let (Y,f,y9) bea covering of (X,x)). Let D=£,(m,(Y,y9)) ¢ ,(X,x5). Then G(Y/X) = N(D)/D where N(D) is the normalizer of D in m,(X,x,). Proof. Fix g ¢€G(Y/X). Let o bea pathin Y connecting Yq *0 (yp)8- Let o= & ef. Now the covering (¥,£,6)8) is equivalent to (Y,f,yq) . Since g ef =f wesee that f, maps both m,(¥,y,) and t,(¥,(¥9)g) isomorphically onto D. But at ~ mY, )8) © F mY) 6. Thus D=o Do. Now consider the map from G(Y/X) > N(D)/D > Do, where S, connects yy to (y,)g. We must check several things about this map. : (i) The map is well defined. If 7, connects yy to (yq)g then ~ xx cte 3,2 78,°8,. But % 5, is a closed path starting at y, and so maps onto D by f,. Therefore +, € Do, . gs 1 (ii) The map is injective. If oe = Do, then a9 / ED:33 %, a is a closed path on Y. Thus (18, = (1), or (yp)h = (¥))8- Consequently h = g. (iii) The map is surjective. If o"!Do =D let & be the liftof o starting at y,. Let y, = (0G. Then wj(Vy,)= 5 my (¥y9) 6. Since , maps both of these groups onto D we see that (Y.f,yq) is equivalent to Why) - The homeomorphism g:Y ~ Y so that ¥, = (ig)8 isa cover transformation and o = 6 (iv) The map is an anti-isomorphism. A path connecting Yo to (yeh 1s seen to be 6, (oh which projects by f onto 19, Thus o = 0,0 gh “h's qed. Remark. If y:G—H isan anti-isomorphism then ¥:G~>H, (g) = (g74)@ is an isomorphism. 45. Some remarks on permutation groups. Let 20 bea set and let 8(2°) be the group of permutations of 2. if De (1,2,---yn} CZ then (2) will be denoted 8. A subgroup Hc 38(3)) is said to be transitive if for all x,y ¢ D. there isa he H so that (x)h=y. A subgroup H is said to be regular if the only element in H which fixes any element of 2. is the identity. If H is regular and he H,h=e, then when we write h asa product of disjoint cycles, all cycles have the same length. Examples. The right and left Cayley representations. If G is a group the right Cayley representation T:G > 8(G) is given by g > T, where (h)T = hg. Then T is an injective homomorphism and (G)T C &(G) is a regular transitive subgroup. The left Cayley representation t:6 ~ S(G) is given by g +t, where (h)t, = gh. +, isan injective anti-homomorphism and again (G)t is a regular transitive subgroup of 8(G). The two representations are the same if and only if G is abelian. 4.6. Suppose (Yi fy9) > (Xx) is a covering of Riemann surfaces. Then G(Y/X) acting on the fiber of { over x, isa regular group of permutations since only the identity cover transformation has a fixed point. Let this fiber, (x) be denoted by 1.34 From the proof of Theorem 4.4 we see that G(Y/X)|| is transitive if and only if D isnormalin 1, (X,x,) (D = (w,(¥,y,))f,).. In this case the covering is said to be Galois or normal. Assuming the covering is Galois, for x € X, y € (x)f-?_ we have (x)f7! = {ors|s €G(Y/X)}. Denote this set, the G(Y/X)-orbit of y,, by ly]. Let {lyily«¥} be denoted by Y/G(Y/X) . Since each G(Y/X)-orbit [y] can be identified with (y)f. ¥/G(Y/) can be given the structure of a Riemann surface and Y/G(Y /Z) =X. Lemma 4.6. Suppose (Y,f,y,) is a Galois covering of (X,x,) and H is a subgroup of G(Y/X). Then there is a factoring of the covering f:Y>X,Y~>Z—>X, where Y > Z isa Galois covering and G(Y/Z) = H. H is normal in G(Y/X) if and only if the covering Z > X is a Galois covering, In this case GIZ/X) = GY/®) / GIY/Z) . Proof: Let D be the image of 1,(Y,y)) under f,. Then D is normal in m,(X,45) and 1, (X,x,)/D = G(Y/X). Let @ be the quotient map of m,(X,x,) onto G(Y/X). Then Dc (H) pt c m,(X,x)). Let Z be the Riemann surface which is the covering of X corresponding to (#)g!. Since D is normal in (H) g” the covering ¥ ~ Z is Galois with Galois group isomorphic to (H) g!/D , which is isomorphic to H. If H is normal in G(Y/X) then (H) @”* is normal in 1, (X,x,) and G(Z/X) & 1, (X,x5)/D gt & GUY /X)/H. qed. If D= then X, the universal covering surface of X, isa Galois cover with G(X/X) = 1, (X,x5). If X is isomorphic to the unit disk D, then G(X/X) is a group of fixed-point-free conformal self-maps of D (fractional linear transformations). 47. The monodromy group of the cover (Y,f,yq) > (X,x9) (25, p. 198]. = aS Let 1= (4) £7! = (yy,yy,¥p.eee}. Suppose [4] € 1, (X,x)) - (Hereafter we will write ¥ instead of [x] since all statements will depend only on the homotopy class of ¥.) We obtain a permutation (x) in 8(D as follows. Let § be the lift of y starting at yy; Then define (y,)(s) uw = (1) i, :35 (i) By the monodromy theorem u is well defined as a function from m,(X,x)) into 3(). Gi) is seen to be a homomorphism. (iii) Since Y is connected the image of u is a transitive subgroup of (I). It is called the monodromy group of the covering and will be denoted M(Y/X). tiv) (x won = {ven X.2)|G)@H= ¥,)- w) (ny yt = {el,p0u y,} . Call this group D, (vi) Let 5, ie foie lift, 5, , to Y starting at Yo ends at 1 { = ee Then ¥~ mY) 5 = m0) 7 nee = Po 5, D 7 (vil) 1, (X,x)) = Doty U Dot, U Dot, U +++, @ right coset decom- position of (KX) Dy 5 is the set of paths which when lifted to startat y, endat y.. 0 j (wii) keru=(1d,- 10 *Dyo. This is the largest J) cen Xx) normal subgroup of 1, (X,x9) lying in D, . Now if the cover Y > X is Galois then D, = Kern and M(Y/X) is a regular sub- group of 8(I). (ix) Thus if Y + X is Galois we have two regular transitive subgroups of 8(I) namely M(Y/X) and G(Y/X) | 1. Both groups are isomorphic to 1m, (X,x,)(n,(Y.yQ))f, - (x) How are these two subgroups of 8(I) related? Let G=r 1% x, fy) /Dy * (Dy Bq sDy ue s+}. Identify y, with Dy. If p is the monodromy map then G&)@ ue neg, 8). That is, (8) acting on Ds, is Dey = D¥,Dy. The monodromy map is thus the right Cayley representation. If g, € G(Y/X) where (y,)g,* ¥, then (ye; . Orns . (See the proof of Theorem 4.4.) That is, 8, acting on D¥, is DY se Ds Dy, Thus G(Y/X)|I is the left Cayley representation. Theorem 4.7. Suppose Y,f.¥9) > (Xx) is a Galois covering. Then the36 Galois group acting on Gyr (=D and the monodromy group are regular transitive subgroups of &(I). If we identify I with the right cosets (D¥,} of D in 1, (X,x,) as above (D = (n,(¥.y,))f,),, then M(Y/X) acting on I is the right Cayley representation and G(Y/X) is the left Cayley representation. Corollary. The two representatons are the same if and only if G(Y/X) is abelian. 4.8. We now abandon the notational convention adopted at the begin- ning of Section 4.4. Theorem. (Hurwitz). Let G bea finite group. Then there exists a compact Riemann surface Y which admits a group of automorphisms isomorphic to G. Proof. Let By 8, be generators for G. Let X bea Riemann AS By surface of genus s. Let A, B, be a canonical basis for m4(X,%). Then TT{A;,B,] = e, and this is the only defining relation. Then 1: m,(X,x)) ~ G can be defined where u(A,) = 6, and 1%+%q) onto G. The covering Riemann surface Y corresponding to the kernal of 1 has G(Y/X) =G. u(B,) =e and 4 is a homomorphism from 1 qed. 49 Galois (or normal) closure. Let (Y,f,¥q) ~ (X,x4) be a covering of surfaces where D=f,(my(¥.y9)). Let N= 1 "tbo which is the een, (Xx) kernel of the monodromy map. The covering Y,, corresponding to N is called the Galois (or normal) closure for the covering Y ~ X. Note that GY /) = M(Y/X). 410. Branched coverings. The coverings of surfaces we have considered so far in this chapter will often be called smooth coverings in contradistinction to branched coverings. The latter will be continuous maps between surfaces which locally are equivalent to z > 2. An example of a branched covering is a non-constant holomorphic map between two compact Riemann surfaces. Consider the map from {0} + D - {0}, f(2) = 2", a smooth covering of Riemann surfaces. Any finite sheeted smooth covering of37 D - {0} is of this type. f can obviously be extended to a holomorphic map of D> D by sending 0 — 0. f so extended is now a branched covering. (i) Let X bea Riemann surface. Let B bea discrete set in X. (if X is compact B is finite) Let X*= X-B andlet {:Y* + X* be a finite sheeted smooth covering. If by €B,let AC X beapara- metric disk centered at by so that AN B= {b.}. f(A - (by) isa collection of finite sheeted coverings of A - {by}. To each component of £71(A - {by}) we add a point to Y* above b, and extend f to these points to map onto by . We do this for each point by € B. Let Y = Y*U (points added above B) and {°:Y > X is f extended Then {*:¥ + X is a finite-sheeted holomorphic map branched (if at all) only above B. Again consider by €A,b, €B as above. Apath ven,(X*,x)), 1 xq € X*, will be said to “circle” by if ¥= oto"! where o isa path from x, to x, €A- {by} and t isa path in A - {by} beginning and ending at x, and winding once (positively) around by. If u is the monodromy representation for the cover Y* — X* then u(¥) is a permutation, each cycle of which corresponds to a component of £1(A - {by}); that is, each cyle of u(¥) corresponds to a point of Y above by. The length of a cycle in 11(s) will be the multiplicity of f° at the corresponding point of Y. (ii) Also any cover transformation g of G(Y*/X*) extends toa biholomorphic map g® of Y > Y,andsince fog =f wehave f%og®=f%. Such a map of Y > Y will be called a cover trans- formation of Y and the group of such cover transformations will be denoted G(Y/X). (iii) Let X be a compact Riemann surface, B a finite set tb, obi, s X*=X-B. Let D beanormal subgroup of 1, (X*,x5) 4) €X", and let Y* be the corresponding Galois cover of X*,f:Y* > X* We then extend Y* to Y,f.to f®,and G(Y*/X*) to G(Y/X). Then G(Y/X) is isomorphic to 1, (X*,x)/D. Now for each b, € By38 let. y be a path circling b . Then Hs) is a regular permutation j of length, say, v.. The ramification of f° above b, is then seen to be (n/v,)(v,- 1) where n is the number of sheets in the covering. We obtain The Riemann-Hurwitz formula for Galois coverings 5 -2. 2 -i apy 2. n(2py - 2) + Maa vy) : Jet Lemma 4.10, Let G bea finite group of automorphisms on a Riemann. surface X. Suppose there is an XE X so that a(x,) =X for all geG. Then G isa cyclic group. Proof. Let A bea parametric disk centered at x,. Let 4 be the 0 component of 1 gA containing xq. Then Ag is simply connected geG and invariant under G. By the uniformization theorem we may assume that 6 acting on A, is a finite group of fractional linear transformation of D -* D which leaves zero fixed. It now follows that G is cyclic. (iv) Let Y be a Riemann surface with a finite group automorphisms G. Let Y/G be the set of orbits of G. If the only element of G with a fixed point is the identity then Y/G is made a Riemann sur- face X by giving it the quotient topology and requiring that the natural map {:Y — X be holomorphic. in fact, f is a smooth covering, Suppose G admits non-identity elements with fixed points. For y EY let Stly)= {g¢ 6] gly) = y), the stabilizer of y. I heG then St(h(y)) = h St(y)h7!. Thus the fixed points of non-identity elements of G are orbits of G Let Y* = {y|St(y) =e}. Then G- (e) is fixed-point-free on Y*. Let X* = Y*/G. By the previous discussion f: ¥* + X* can be extended to {°: Y + X where we add to X (points corresponding to) the orbits of points in Y with stabilizers larger than . f° is, now holomorphic and branched precisely over X - X*. We now summarize this discussion in a theorem. Theorem 4.10. If G isa finite group of automorphisms on a Riemann surface Y , then the orbit space Y/G is naturally a Riemann surface so that the projection f: Y > Y/G is a holomorphic map. The branch-39 ing of f occurs precisely at the fixed points of non-identity elements of G. The number of sheets in the covering Y — Y/G is the order of G. (With a slight modification of the above discussion one sees that this theorem is true for countable groups of automorphisms, G .) 441. We saw in Lemma 46 that if Y > Z is a Galois covering and H is normal in G(Y/Z) then there exist a factoring Y ~ X > Z where both intermediate coverings are normal. We want to reverse this procedure. If Y ~ X and X ~Z are Galois coverings, when do we know that Y — Z is Galois? This then is the Riemann surface (or function field) version of a basic problem in group extensions. Theorem 4.11. (A. M. Macbeath [18]). Let X bea surface. Let D bea normal subgroup of ™,(X,x,) and let (¥,f,¥)) > (X,x9) be the cor- responding smooth Galois covering. Let @ be a homeomorphism of X (possibly with fixed points). Let « be a path in X joining Xy to otx,) . Define 9, (Xx) = 1, (XK x9) by 9 (8) = & olsyat. Suppose g, (D) = D. (Since D is normal this criterion is independent of « .) Let vy be any point above (x) _, Then there exists a homeomorphism @ of Y onto Y so that Gly) = yy and f° G= oof. Thatis, @ preserves the fibers of f Consequently @ normalizes G(Y/X) . Proof. Let « beapathin Y connecting Yo to Yy- Let a = f(a), connecting x) to g(x,). If ye Y let & connect a a= £(G) let x= o(1). plc) connects o(a,) to glx). Let $ be the lift of @(o) starting at y,. Define Gly) = s(t). To show that @ is well defined, suppose T connects Yo toy. Then + = {(%) connects x) to x. Now Gt”! isa closed path in Y so ot! is aclosed path in D. Let ¢ be the lift of g(t) starting at y, - Now 9, (ot) isin D. Therefore, the lifts of «g(o) and g(t) have the same endpoint in Y; that is, {(1) = $(1). This shows that @ is well defined. That @ is locally f"1 » gf shows that @ is a local homeomorphism. That g has an inverse sending %y to Yo is seen by inverting the steps in the proof.40 412. If @ is an element in a group, let o(g) denote its order. If o() is finite in the previous theorem, it will always be the case that o(p) divides o(@) , but equality is not necessary. We now give a simple criterion for o() = 0(@) . (In the language of group extensions this means that the extension O- K+ K,@ > [p(x)] takes X/G onto X/pGg?. 415. We now include some consequences of the previous paragraphs which will be of use later. 4.15.1. A Riemann surface of genus 2 does not admit an automor- phism of order 7. A Riemann surface of genus 3 does not admit an automorphism of order 5 . 4.15.2. A Riemann surface of even genus does not admit an autormor- phismn group isomorphic to Z, Z, x Z, where Z, is the cyclic group of order 2. 415.3. Let Y + X be a Galois covering of compact Riemann surfaces where G(Y/X) is abelian. If the covering is not smooth, then it must be branched above at least 2 points of X. The first three results follow from the Riemann-Hurwitz formula for Galois covers. The fourth follows from the observation that the commu- tator subgroup of an abelian group is the identity.Chapter 5. Automorphisms of Riemann surfaces, I. 5.4. If X is a Riemann surface let Aut(X) stand for the group of automorphisms of X. We now reproduce, with slight modifications, Hurwitz's proof that a group of automorphisms on a compact Riemann surface of genus greater than one is finite [15]. Lemma 5.1. Let X be a compact Riemann surface of genus p. Let 9 bean automorphism of X. Then @ has at most 2p +2 fixed points. Proof. Suppose @ has s fixed points. Let f be a meromorphic func- tion with precisely p+1 poles at p +1 distinct points, none of which are fixed points for @ and so that none of the poles of f and fe@ coincide. Then f-fe@ has 2p +2 poles and at least s zeros. There- fore 2p+22zs. qed. Theorem 5.1. (H. A. Schwarz [23]). If x is a compact Riemann surface with genus p22, then Aut(X)) isa finite group. Proof. For p> 2 the total weight of the 3K points (higher order Weierstrass points for the tri-canonical series) is 25(p-1)2p. The maximum weight at any point is p(p + 1)/2. Thus there are at least 25(p - 1)*p /(p(p + 1)/2) distinct 3K points. For p22 this number is greater than 2p +2. Any automorphism permutes the 3K-points. We thus have a map from Aut(X_) + 8(3K-points) . The kernel of this map fixes the 3K- points , and so by Lemma 5.1 the kernel is the identity. Thus Aut(X) is isomorphic to a subgroup fo a finite permutation group. qed. Definition. For p22 let N(p) be the largest order of an automor- phism group on a Riemann surface of genus p. Many applications of what follows will give information about the nature of this function, N(p) . 5.2 Let f:Y 7 P! bea Galois covering of n-sheets , Y compact. Suppose the ramification occurs above B = (2, ...,2.} el. Let % be a path circling 2 so that iS . mw, (PSB, 2) = (8,4. Syt8, [Byte =e). Let v, be the ¥, A , i ramification of a single point over 2,. Then %, =e. If43 wen, (Pt -B,2)) > 8, is the monodromy map then M(Y/P2) is generated by ney), rw(¥.) If H@) sae a, then the ays satisfy the following: a,' =e, i= 1,2,-++,8,and ajay-++a,-e. These 2 generators obviously can be expected to satisfy other relations. Definition. A finite group generated by elements aya, So that s+ a, =e will be denoted (v,,+-+,¥,)- s In the discussion preceeding the definition both G(¥/P1) and M(y/P1) are (wv 't41%s) groups since the two groups are isomorphic. If p is the genus of Y then the Riemann-Hurwitz formula is: 3 s ES -Lyenl(s-2-Y2 (*) 2p-2= amen D (1 Heals 2 De). kel kel When G(¥/P*) is abelian, in the identification of G(Y/P+) with M(¥/P1) the permutation u(¥,) is equal to the stabilizer of one (and therefore all) of the points in Y above 2, in P!. This will, in fact, be the case for arbitrary G(Y/P 1) whenever u(8,) is central in M(y/P}). The observations of this paragraph are valid for any Galois cover Y > X and do not depend on the genus of X being zero. All these remarks follow from Theorem 47. 5.3. Groups of genus zero . In 5.3 assume that p= 0, thatis Y= P!. Definition. A finite group fo automorphism of P! will be called a group of genus zero. In the fomula (*) of 5.2 we see that s - 2- at must be kei negative, equal to -2 , in fact. Since v, 22 we see that s must be less than or equal to 3. If s=2 G(Pt/Pt) isan (n,n) , a cyclic group of order n. If s = 3 we list the possibilities in the following table:44 c(ptypty order description (2,2,n) 2n Dihedral group D_ (2,3,3) 12 Alternating group A, (2,3,4) 24 Symmetric group 8, (2,3,5) 60 Alternating group A, This allows us to completely describe hyperelliptic automorphism groups Theorem 5.3. Let X, bea compact hyperelliptic Riemann surface, p22. Let H be the hyperelliptic involution. Then the subgroup of order 2, P! the Riemann-Hurwitz formula is s (-2-Sa)-0 kai * ‘with no obvious condition on n. The four possibilities are s=4 (2,2,2,2) s=3 (2,3,6) (2,4,4) (3,3,3) In analogy with hyperelliptic Riemann surfaces we define a Riemann surface to be elliptic-hyperelliptic if it is a _2-sheeted branched covering of a Riemann surface of genus one. An involution whose orbits are the45 fibers of this 2-sheeted covering will be called an elliptic-hyperelliptic involution. Later in this chapter we will show that an elliptic-hyper- elliptic Riemann surface of genus greater than 5 has a unique elliptic- hyperelliptic involution. (This also follows from the Castelnuovo-Severi inequality.) Then we can repeat the proof of Theorem 5.3 to conclude that for genus greater than 5 , an elliptic-hyperelliptic automorphism group has a normal (central) subgroup of order two , and the quotient group is a group of genus one . 5.5. Maximal automorphism groups for p> 2. We now reproduce Hurwitz's proof that for p22 N(p) < 84(p ~ 1) (15). Lemma 5.5. Let Y + X be a Galois covering of n sheets of Riemann surfaces. Assume p_,the genus of Y, is greater than one. If Py the genus of X , is positive then o(stv/%) < 40, - 1). Proof. The Riemann-Hurwitz formula for the covering is 2p, n(2p,, ae (0-4), if p> 7 e - If p22 then 2p, 22 n(2p, 2)2 2n, or nsp, 1. s If p,= 4 then ap,-2-n p(1-4) dn. i ft qed. Thus to obtain large automorphism groups we must have P, 0. s Then t,-2=n(e-2- D4). _ If s25 then 2p, -22n(3-3)-4n. 9-3). i= Since Py >1,not all the v, can equal 2. The order of G(Y/X) , n, attains its maximum when \,, V4 13,¥, 'q) = (2,2,2,3) and 3° n=12(p -1 (p ye fs=4 then 2p, -46 3 - 1 Hf s*3 then 2p,-2=n(t 2+) vo In order to assure that p_ > 1, we must exclude those values of (v,,¥),0_) which occur for groups of genus zero and one. Thus wwe choose (v,,0,,045) so that (1- >,+) is positive but as small as possible. The appropriate values are (2,3,7) in which case n= 84(p - 1). Theorem 5.5. (Hurwitz). For p22,N(p) < 84(p - 1). If a Riemann surface of genus p admits a group G of automorphisms of order 84(p-1) then G isa (2,3,7) and the quotient map X_~ P! is branched over 3 points of P! above which the multi- plicities of the branch points are 2,3, and7. Conversely, if G is an abstract (2,3,7) , then there exists a Riemann surface X_ admitting an automorphism group isomorphic to G of order 84(p- 1). Proof. To prove the statement in the second paragraph, let G be an abstract (2,3,7),G= where a= b?= (ab)” =e. Define uin,(P4-(0,1,0),2) 46 by w,)=a, u(t.) =b and ws) = (ab)"!. The kernel of 41 corresponds to a Galois covering Y* of P* (= P1-{0,1,00}) with G(Y*/P*) = G. Filling in the branch points over 0,1, and gives a branched Galois covering Y > P! and NO) = 84(py- 1). qed. Thus the problem of finding Riemann surfaces admitting automor- phism groups of maximum order is the purely group theoretic problem of finding finite (2,3,7)'s . To show that N(p) = 84(p - 1) for an infinite number of p's , one aust find an infinite number of (2,3,7)'s . This was done by G.A. Miller [21] in 1902, about 10 years after Hurwitz's original work. 56. An infinite number of (2,3,7)'s. We shall follow A. M. Macbeath's discussion of (2,3,7)'s [18]. We first show the existence of a famous group. Lemma5é. There exist a (2,3,7) of order 168. Proof. Let G = GL(3,2) , the group of all 3x3 non-singular matricesa7 over the field K, of two elements. By counting the number of ordered bases in the 3-dimensional vector space over kK, we see that the order of G is 168 = 7-6-4. Now consider G as the group of automorphisms of the elementary abelian group of order 8,G,. Then G, = ‘a,b,c? or G, = {e,a, b, ¢, ab, be, ca, abc} + (€,8, 8,884 Bp & & } where we relable the group elements g. in the manner indicated; i eg. g, = be. Let A be the automorphism a>b,b~a,c— abc Using the subscripts of the g's we can describe A by the permutation (12)(37) . Let B = (145)(267). Then C = AB = (1673245) . If we let = H then H isa (2,3,7) and we want to show that the order of H is 168. Since 3 and 7 divide the order of H, we show H contains a subgroup of order 8. is such a subgroup since AB™! AB = (1372)(46) aoe = Dy a (2,2,4). qed. With a little more work one shows that G is simple. Applying Theorem 5.5 gives us Klein's Riemann surface of genus 3 admitting 168 automorphisms. Theorem 6.7. (Macbeath [18,19]. There exist a sequence p> = and Nip.) = 84(p, - 1). Proof. Let x, be Klein's Riemann surface. Let HL be the characteristic subgroup of ,(X,,x,) containing the commutator subgroup and all nth powers. H_ is the kernel of the surjective homomorphism (Xx) > 2x2 *Z x2 *ZxZ. Since H_ is characteristic every element of Aut(X,) leaves H. invariant. The Riemann surface Y_ corresponding to H_ is a smooth, ie A A n°-sheeted Galois covering of x and all of Aut (%) lifts to i . The genus of Y, is 2n® +1, and the order of Aut(Y,) is 168n® qed. This method of Macbeath’s, that is, lifting automorphisms to Galois coverings corresonding to characteristic subgroups of 1. hi (Xx o) is very useful.48 5.7_Smooth cyclic coverings of prime order. Let n bea prime integer. We need a criterion which assures that an automorphism of order n on a Riemann surface is fixed point free. Lemma 57. Let S, be a cyclic group of prime order n ona Riemann surface X,. Suppose p- 1=sn and n>2s+3. Then the covering Xp > Xp/Sq(= Y) is smooth. Proof. Suppose not. The Riemann-Hurwitz formula for the covering is 2p, -25 n(2py - 2)+t(n- 1) where t is the number of branch points and t>0. Then t(n- 1) = 2p, - 2-n(2py - 2) = 2n(s+1-py). Therefore n- 1] 2(s+4 Py) or n-1< 2s+2. This is the desired contradiction. qed. 58. N(p) < 84(p - 1) infinitely often. Theorem. (Macbeath [19]). Let n bea prime greater than 84. Let pen+i1. Then N(p)< 84(p- 1). Proof. Suppose X, admits group G of order 84(p - 1) = 64n. Let S_ be an n-Sylow subgroup of G, acyclic group of order n Since n> 84,S_ is anormal subgroup of G (Sylow’s theorem). By Lemma 57 X > x s, is a smooth covering and x 1s, isa Riemann surface of genus 2 admitting a group of automorphisms of order 84 (Lemma 4.6). But a Riemann surface of genus 2 does not admit an automorphism of order 7 (4.15.1). qed. We shall return to this function N(p) in Chapter 7. 5.9. Groups with partitions [2]. We now generalize the hyperelliptic situation. Definition. A finite group G, is said to admit a partition if there is a collection of subgroups G,, i= 1,2,-+-,8,823, so that : (1) 6y= U6, and (2) if O. The examples of such groups that will interest us are groups in which every element has order 2, and dihedral groups. Theorem 5.9. Let X be a compact Riemann surface of genus p. Suppose X admits a finite group of automorphisms, , , where 6) admits a partition. Let the pertinent subgroups be G Gorey49 Let the order of G be n. let Xi X/G. and let P, be the genus of X/G for i= 0,1,2 . Then: (S- L)p+ngPy = Da, ie. Proof. Let the Riemann-Hurwitz formula for X > Xx/G, be ix 2p - 2 = n,(2p,- 2) + ram, : we stow i rat, «ra i= leave x fixed form a cyclic group whose generator lies in one of the Gis It follows that St(x) lies completely in this G. Consequently, ip: For x€ X the elements in G& which the contribution of St(x) to ram, contributes to one and only one of the ram,’s The theorem now follows by summing equation («) as i runs from 1 to s and comparing to equation («) with i= 0, noting that s it qed. The theorem 1s of interest because the ramifications do not enter into the formula. For a beautiful generalization of the material in this section see the paper of Ernst Kani "Relations between the genera and between the Hasse- Witt invariants of Galois coverings of curves’ Canadian ulletin Vol. 28 (3), 1985, pp. 321-327. $40, Z,«Z, We first apply Theorem 5.9 to the non-cyclic group of order 4 (four group). There are 3 subgroups of order 2 so in this case s= 3. The formula becomes P+ 2p) =P, +P, *Pz- We use this formula to deduce a folk theorem that appears to be due to Enriques [10]. Lemma 5.10. Let xX, be a covering of an X,. Then X, is hyperelliptic. Proof. X, > X, is smooth and 2-sheeted. Let T be the involution in Aut(X,) so that X,/ = X,. The hyperellipic involution on X, lifts to an automorphism S on % of order two with some fixed points. S50 normalizes ; consequently, isa four group on X3 with XS, 0 and one >= P!. Apply the formula above with p=3, Py of the ps equal to 2. The other pis must be 0 and 1. qed. Similarly, one shows thatan X, (or X,) which isa two-sheeted covering of a hyperelliptic X, (or X,) must be hyperelliptic or elliptic-hyperelliptic. SAL Z,%2,*Z, There are 7 subgroups of Z,*Z,%Z, of order 2. Theorem 5.9 gives 3p + 4p, Lemma 5.11. A Riemann surface of genus 3 admittinga Z, 4% Zz 9 Zz, _ is hyperelliptic. Eroof. If X, is not hyperelliptic then all p's above (iz 1) must be one , whereas the left-hand side is at least 9 qed. In fact, by relabeling, if necessary, it can be arranged that =O. PL = Py = Py=t,and a i A hyperelliptic Riemann surface cannot admit Z,Z,*Z,*Z, as a group of automorphisms because Z,xZ,xZ, is not a group of genus zero (Theorem 5.3). 5.12. Dihedral groups of automorphisms. Let D, bea dihedral group of order 2n. Let R generate the cyclic subgroup of order n. If V isanelement of order 2 not in then V,= RIV (= 1,2,-++,n) are the elements in D_ notin . If Dy is realized as a group of automorphismns on a Riemann surface X of genus p, let X= X/, X= X/Vp, is ten andlet X) = K/D, with genera pp.Py.:++,P,.Py respectively. Theorem 5.9 now gives np + 2npy= npp +2 ‘i trl if n is odd then all the groups are conjugate and so p, =51 for i= 2,+++,R (Lemma 4.14). Thus we obtain P+ 2p) = Pp + 2p, If n is even then is conjugate to if and only if J 12 J (mod2) . Let p, and p, be the two genera in this case. Then \ = * &) D+ 2p = Ppt Pi +Py- / This last formula holds for n odd or even with n odd implying Pio Pao 5.43. Commuting involutions. By the Castelnuovo-Severi inequality we know that if p> 4p) +1 and X_ admits an involution T so that the genus of X/ is Py then this property makes T unique. If, however, ps 4p)+1 then T need not be unique. The next application is an extension of these ideas. Lemma 5.13. Let p, and p, be non-negative integers. Let X bea Riemann surface of genus p so that 2p 2 3p, + 3p, +3 Let X admit two distinct involutions $ and S, so that X/ has genus p,,i1,2. Then S, and S, commute Proof, Let G, = a dihedral group of order In. Set S, = V, and S,=V,. Then P+ 2p) = Py * Py * Pp We wish to show n= 2. Suppose n> 3. The Riemann-Hurwitz formula for X > X/ is 2p- 2 = n(2p, - 2) + ram 2 3(2p_ - 2) or Spy sp +2. Since py? 0 we have 3p < 3p + 6p, = 3p, + 3p, + 3p, This contradiction proves the lemma. 5 3p, + 3p, +p +2. qed. If Py * Py the above result can be strengthened since n = 2 implies 24; for in this case S, and S, cannot be conjugate.Chapter 6. When are fixed points of automorphisms exceptional in some other sense? 641. Lewittes’ Theorem. The following theorem is due to J. Lewittes [17]. Theorem 6.1. Let X_ bea Riemann surface of genus p22. Let T be an automorphism with 5 or more fixed points. Then each fixed point is an ordinary Weierstrass point. This theorem will follow from the following lemma. Lemma 61. Let f: - > x, be an n-sheeted branched covering of Riemann surfaces. Suppose the total ramification of this covering is Greater than 4(n- 1). If z is a branch point of f of multiplicity n then z is an ordinary Weierstrass point. Proof. We show that X_ admits a meromorphic function whose only pole is at 2 and whose order is less than or equal to p. Let y = f(z). Let g be a meromorphic function on X_ whose only pole isat y and whose order is less than or equal to q+1. The Riemann-Hurwitz formula for the cover is: 2p - 2 = n(2q- 2) + ram > 2nq- 2n+4n-4 or p+i1>n(q+1). Consequently, g ° f is meromorphic on X_; its only pole is at 2 and its order is less than or equal to n(q +1). Consequently its order is less than or equal to p. qed. Proof of Lewittes' Theorem. Let T be the automorphism of order n with 6 or more fixed points. Since the ramification of the cover X)~> X/ is 5(n~ 1) or more, the lemma shows every fixed point to be a Weierstrass point. 62 G. on X,- (Some folklore.) Let X, bea Riemann surface of genus 3 admitting the finite (2,3,7) group G,.. of automorphisms of order 168. Consider the Galois covering W, > P! of 168 sheets. Let _ P! be the point over which there are 84 branch points of multiplicity 2; let 2, € pi be the point over which there are 56 branch points of multiplicity 3; let 2, ¢P 1 be the point over which there are 24 branch points of multiplicity 7. We claim that the 24 points above z, are the 24 ordinary53 Weierstrass points in X,. For the weights of points ina single G-orbit must all be the same. But the total Weierstrass weight is 24. The only place the Weierstrass points can fit is over 2, , all other orbits having more than 24 points. Consider the points over 2,. The stabilizer of one of these points is an automorphism of order 3; callit T. If the genus of W,/ is zero, there would be 5 fixed points, by the Riemann-Hurwitz formula, and so all would be ordinary Weierstrass points, by Lewittes’ theorem, a contradiction to the previous paragraph. Consequently, the genus of x LT? (= x) must be one, and the Galois covering of 3 sheets must have 2 fixed points. Let 2x + 2y be the divisor of branch points in the covering X, ~ X, . Then this is a canonical divisor on X, , and corresponds to the intersection of a line with g(X,) where g:X, ~ p2 is the map given by the canonical series onto a non-singular plane quartic C, 4° Thus 2x + 2y corresponds to the two points of contact of a bitangent to C,. Since there are 8¢ points above 2, , these points are the 56 points of contact of the 28 bitangent to C, . We now locate the 2°4 order Weierstrass points (which we call 2K points). The total weight of the 2K points is 108. Let w, Was and w. be 2 o 1y2, and 2, respectively. (Since all other G-orbits have order 168 , none of these points can have positve 2K weight .) Thus rs) 84w, + S6w, + 24w, = 108. the 2K weight of a single point above z The only possible solution in positive integers to this equation is w,=1,w,=0,and w,=1. The total weight of the kK points is (2k - 1)?+12. Let wy be , 1 : the kK weight above points of P. {2),2,,2,}- Thus if WoW + and w, are the kK weights above 2, ,and 2, we have z 27°3 168w, + 84w, + S6w, + 24w, = (2k- 1)2+ 12. If we reduce this equation module7 we have: 3w, = 5(2k - 1)? (mod7) or w, = 4(2k-1)? (mod7).54 Since 0< Wy $ 5 we see that w. is the smallest non-negative residue 7 of 4(2k - 1)? (mod?) . If we let w,(k) (k 22) be the kK weight of a point above 2), we see that w, (2) ei, w, (3) =2 ow, (4) =0 yw, (5) = 2, w,(6) = 1, w,(7) = 4,w,(8) = 4, and w,(k) is periodic in k with period 7 for k= 2. 6.3. Cyclic coverings of P! of prime order: Examples. (More folklore.){3}. The title of this section refers to the Riemann surface for the s polynomial P(x,y) = y"- I (>a, =0 f ) where n is prime, s 0 P° is smooth jel over ©. Call the n points on X, over eo, 20, ,.++,00,. Thus the function x~- x) on P! lifts toa function with n zero above X and simple poles at ee dx lifted hasa zero of order n-1 ateach A. and n poles of order 2 at each point i55 o),y = 3/T(x-a,)! sothat at each =, y hasa pole of order xs, /n. Thus over @ ,y hasa total of (f/m) D4, poles. Near Aj, y= (x-a,) k(x) where k(a,) = 0 above . Near gt A local parameter t at Ay can be choosen so that t® = x - ays so we see that y hasa Zero of order f at A We can summarize this information about x,y and dx on x, by the following table: [ Ay_ [Ao |<: | As | totalover =| _totalover ot (xo) dx n-i[a-i ~[ a -T] 2a y fy fy seed fy 2, =m [0 [0 0 [cn 2 As an example consider the polynomial = y>-(x- 5 . -a)? So P(x,y) = y?-(% a) ay) (x a,)(x a,) (p= 4) K= 9% Ai | Az | Ag | Aq | totalover dx a{[4[4]4 | -0 y 1 1 [1 2 5 x-ay 5 0 0 0 5 wpdery? 2 2 2 0 0 0p: (x-ag) daly? 1 r{a]3 0 ag: (x-aq) ax/y! 0 of of 1 5 og: Ge-ag) daly! o{ of of 6] o tog: (ea x-ag) dy! s | of o| 1] 0 The last 5 rows correspond to holomorphic one-forms on X, , Go, ,W,,+++,00,. We see that all Aj’s are ordinary Weierstrass points with A,,A,,and A, having weight 2 and A, having weight 4. By multiplying together two holomorphic one-forms we can compute56 non-gaps for quadratic differentials as follows: 2K= tp Ay | Ag | Ag| Aq | totalover « oy 4 {4 ]4]01]0 wp 2a 2p ee |e 0 os o jo © | jo 4 o fo | o} 12] 0 2 os 108) Om InOb ls 21 |0 Ux-ag)/(x-ay 10%, Be | OlLalio aay 3 {[s[3] 3] 0 31 95 7 [2][2] i[o 205, e {[1[i] 4[ 0 O304 o [ofo| 7[5 2204 io The 2K non-gaps at A, are 0,1,2,3,4,5,6,7,10; weight 2. The 2K non-gaps at A, are 0,1,2,3,4,6,7,9,12; weight 8. By multiplying one-forms with quadratic differentials we can obtain the 3K non-gaps at the A's. And so forth. The table on the following page for a slightly more complicated case is worked out. It is readily seen that the kK weights at each A. are f periodic with period 7 for k 22. Several features of this table will later be shown to be special cases of more general phenomena. A final useful observation is the following. s 7 Let y" -Tl (x- ay)’ define a Riemann surface whose function field is C(x,y). Then letting s 5 z-y T]«-a jet we see that €(x,y) = €(x,2) and ,8# 0 (modn)57 kxonongaps for y7 = (x-a,)(x-a9)(x-a3)"(x-ay)?— (p=6) nongaps weignt kel ay Of 2e Go 4 T 2 a, ° 5 8 4 ay ° 2 aos 9 6 le 2a oe) 90 13014 15 3 BS oo oO 13 16 18 8 a, 0-8 9 lo wd 134M 18 6 k= 3 a, 0-18 19 20 21 22 23 24 ° a3 0-18 19 20 22 23° «24 26 4 ay 0-18 «1920 22 23 26 27 8 ee) tere ee ey 33 35 37 6 a3 0-28 29 31 32 34 36 39) 12 a, 0-28 29 30 31 32 35 36 a k= 5 a) 0-38 39 HO Wl 42 43 44 ° a3 0-38 39 40 41 42 44 47 4 a, 0-38 «39 Hon 44 4s 48 8 k= 6 a, 0-48 49 50 51 52 53 37 3 a; 0-48 49 50 52 54 55 51 a, 0-48 49 50 52 53 54 57 k=7 a, 0-58 59 60 62 63 64 70 8 a3 0-58 59 60 oe ee 65 70 10 ay 0-58 59 6. 62 63 66 70 12 k= 8 a, 0-68 70 71 72 73 74 17 7 a; 0-68 70 71 72 73 15 18 do ay 0-68 To 72 72 m4 15 79 3258 s sf, +ne, dj J Il (e-a) i= defines the same covering of P!. Thus y= (x- ax ~ ax ~ ay)® 72 (x 2 - 2 3 y? = (x-a)?(x- a,)*(x - a.) 7. _ 3 (x - 3 = y? = Gr~ a,)3 (x - a,)3 (x - ay) y? aera Ae a4 x-a,)? ete. all define the same covering x : X, > pt. s f 64, Lemma. Let y" = II (x= a)? where n is prime, jet O [#0 (modn). Let A, be the point on the Riemann surface X) above a,. Then n(2p-2)A, = nK for all j and so winK,A)> 0 for all j. Proof. Let g1 be the fibers of the cover x p1. Then by n Theorem 2.6 we have (B is the branched locus) K=-2g! +B 2 where we write the result additively. Now g! = nA, , and a s B= Dea-va,. yl s at a- Dna, pt =n(-2n+s(n-41)) A, = n(2p - 2) Ay. Thus nk = -2ng! an” A, + s(a- 1) nA, qed This lemma motivates most of the work in this chapter. 65, We first review some material form Chapter 2, with slightly different notation. Let IGI = g® y bea complete linear series without59 base point where G is an integral divisor on X. For 2X, there are R+1 G-non-gaps for ay at 2,0=nyiny con, sN. The N-R (=) G-gaps at z are g,¢g)¢-++¢g, 0. Therefore there exists an integral divisor E so that G-(g+1)2+E=K or G+E=(g+1)z+K. Now z isa fixed point of IE] ¢ r(E- z)=r(E) # i(G - gz) = i(@ - (g + 1)z) . Therefore z is not a fixed point of IE|. Now reverse the argument. ed. (Note: If G = K we get the usual results about ordinary Weierstrass points.) 62. Lemma. Let i, = i(G). Then i, = i(G~g,2) and i(6-G, - 1)2) ip td da A QT. Proof. Because G, is the first gap we have r(G-g,2)=R~-g,.60 a Therefore IG - gyzl = g Neo and it is complete; thus it has 1 the same index as IG] . By the previous lemma, if n isa non-gap then i(G- nz) = i(G- (2 +1)z). Thus the index of G~- nz changes only at the gaps, and at each gap the change is one. qed. 68 Lemma. If [GI is not special then for m_a positive integer w(G+mz,z) = w(G,z). Proof. It suffices to prove the lemma for m = 1 (and then use induction). Foranyj Genz+E, (€,.z)*0 G+z5(n +1)2+E,. i ij Rel Net Ongti nytt yny tt and the (G+ z)-gaps are Sth gp tt. Now [6 +2l , 80 the (G+ z)-non-gaps are g Thus the sum |), fi~ @ is unchanged. 69. Lemma, The smallest G-gap at 2,8, , satisties: g,2N-2p+1. Equality occurs @ G = (g, +1)2+K (in which case g, +1 is. a non-gap). Proof. to iG - 8,2) » It iy >0 then N< 2p- 2 and there is nothing to prove. Therefore suppose i, = 0. Then i(G-(g, +1)z)=1 s0 deg(6 - , +1)z)s 2p-2 or _— Equality occurs @ G- (g, + 1)z = K since K is characterized by having degree 2p- 2 and index 1. £10. Lemma If mz 2 then cetera 2) = w{K,z) +p. Proof. Let G=mz+K. Then [l= g"_ where N=m+2p-2,N-R=p. 7 g,-m-1. Let Mgrsml_, be the K-non-gaps at z. Since K is fixed-point-free there are p G-non-gaps6 mein jr Otjeypot Thus the m+p-41 G-non-gaps are 0,1,-+-,m-2 smtnysyme+n ect Thus w(G,z) = Dmen'y-(m-1+39) ° -1 =Ppt Se, ~ J) = p+wiK,z). j=0 641. Lemma. Suppose G = Nz where lee, N-R=p. Then (1) g=N- 4 and (2) w(G.z) = p+w(K.z). Proof. (1) Since i, = 0 Lemma 67 gives, for j= p i{(N- g -1)z) +p. Since K is fixed-point-free this implies N- 8, 0 5 Dov-p+)-gj-t ( De+1-J) = 2 =w(G,2)-p. (N-p=R) 612. Lemma, Let G = Nz, degH = M,M22, and |G- HI is not special. Then w(G - Hz) = w(H +K,z) = w(G@+H+K,z). Proof. Since G= Nz the second equality follows form Lemma 68. For the first equality, let g bea (G- H)-gap. By Lemma 6. there exists an integral divisor E, ,(E,,2) = 0 and G-H+E, =, +1)2+Kk or (N-g,-De+E, = H+K. Thus N-g,+4 isan (H+ K)-non-gap. Now g, sN-M62 so M-1sN-g-1. IH+Kl= Mee? app There are M+p-1 (H+ K)-non-gaps of +2p~ which the first M-1 are 0,1,-++,M~- 2 (Lemma 69). The other (H + K)-non-gaps are N- a> 1,j=1,2,-++,p. Compute w(H + K,z) for non-gaps 2 M-1. DB w+ K.2)- > [ov-g,--(-1+G-0)] gel P =D (N-M-G-0-g) j=1 But w(G - H,z) = Sov-m-pss-g) jal qed. 643. In the light of Lemma 64 the following theorem explains many features of the results tabulated in Section 6.3. Theorem 613. Suppose for 2¢ X we have m(2p- 2)2 = mK . (1) If m= 1, then forall @ = 2,3,4,-+- we have w(tK,z) = w(K,z) +p. (2) If k,022 and k= (modm) then w(kK,z) = w(tK,z) . (3) If m22,k+l=4 (modm),and k,22 then w(kK,z) = w(8K,z) . (4) If m22,0=4,2,-++ then w(tmK,z) = w{(tm + 1)K,z) = w(K,z) +p Proof. (1) follows from Lemma 6.11. (2) follows from Lemma 68. (3) follows from Lemma 6.12. For any t21 and k so that 2skstm-~2 let G=tmK and H=kK in Lemma 612 (4) follows from Lemma 6.11 and (3) above. 6414, Cyclic converings in general. We now show how to generalize the results of Theorem 6.13 to cyclic coverings of prime order where the quotient is no longer required to be the Riemann sphere. Thus we consider a Riemann surface X_ of genus P with an automorphism T of prime order n and x = xD . The brariched covering 9 : x = x, will have n-sheets. If MK) is the63 field of meromorphic functions on X_ then we will denote o(mox)) by M*(X_) a subfield of M(X) of index n. Lemma 614 There exists y ¢ MOC) - M*(X,), with y® « M*(X) and MIX.) = M*(X ly. Proof. Pick y « M(X.)-M*(X,). Since n is prime, = MOK) = MOK DIy]. Let te". Then SH oo i m=0 ke sn-1. For k=O0,t,-++n-4 let yy = Myo T™). m=0 nari Thus y, °T > Tela) Tmt) a at m=0 n-1 noi n-1 aie Qos BB) ye k=0 m=0 Therefore, there isa y, , not in M"(X,), so that yr e M*(X,) i eT) =y oT = since (y, ¢ T) ae Tey 7 qed. We can and will assume that y is chosen as in this lemma so that yeT=ty. Let y2=FeM*(X_). Then F= Fog fora function F ¢ MX) . The divisor of F can be written (F) vi, f. ay +nEy where Ey is a divisor on Wy and O f, = 0 (modn). (Some of the a/s can also occur in E,.) Since X_ is the Riemann surface for the polynomial X"- F « Mex, )IX] we see that the covering g: x, oad a is branched over each a, and only there. (If (F) = nE,then the covering is smooth.) The ramification over each a is n- 1 so the Riemann-Hurwitz a formula for the covering is 2p - 2 = n(2q- 2) +s(n- 1). Let A, be the point on X_ above a, on X. Then the divisor of y is and Di 0 (modn). 615, Lemma Suppose g < M(X_) where @:X, > X, is an n-sheeted covering of compact Riemann surfaces. Suppose there is a divisor of degree zero on X,,D,, so that (g) = @”M(D.). Then g® ¢ o* M(X,). (We do not assume that n_ is prime or that the covering is Galois.) Proof. For y ¢ X,, let gly) = Ge, ,-++.4 Define 2 Gey) = [] atx) .awel detined function on X,. Then (6) = nDp.. jal So Ge and g® have the same divisor. qed. $16. Invariant divisors. Definition. Let H be a group of automorphisms on a Riemann surface X A linear series g® N will be said to be invariant under _H if for all h eH and all divisors D in g®,we have hDe gh. For example the k-canonical series will be invariant under any subgroup of Aut(X) . Lemma£i6. Let g®\ bea complete linear series invariant under a cyclic group of prime order. Then there exists a divisor G, in BR so that TG, =G). The proof of this lemma will depend on the following general fact. Proposition 6.16. Let IGl = ay be a complete linear series invariant under a group H « Aut(X) . Consider a map65 e:X+PR given by R +1 independent functions in L(G). Then there is a group of projective transformations in P® , H, isomorphic to H and @(X) is invariant under H. Moreover, H acts on the hyperplanes of P® in the same manner that H acts on the divisors in IG] . Proof Let 0}. Then isa basis of L(TG) where Te H. Since aR is invariant under T, there exists F ¢ M(X) and (F)=@-T@ Thus (f, ° T)F ¢ L(G) and so (R+1)x(R+1) ODF @,) Define @:X + PR by x~(f,(x),---,f,(0). Then Tx maps into (f° TO), -++fp TH) “(ote Fol + T(0) F(x) : (erate 2 2 Rj f,). Thus T induces on PP a projective transformation T given with respect to this basis by the matrix @) . Everything now follows. aed. Proof of Lemma 6.16. Suppose a is invariant under . Then T acting on P® alsoactson PR*, the hyperplanes in P® Suppose the action of T on PR* is given by the matrix M. Thus M"-al,xe€. If M=-LM then M"=1; that is, we can lift My the cyclic group to €R*1 - {0} over PR. Now we proceed as before. For ve €R*1 define66 k and so My =T Ye and ave Dy: Therefore there is a v7 0. The line through the origin and v, €R*1 corresponds to a point in P2* , which in turn corresponds to an invariant hyperplane H in P® . The pull back to X of the hyperplane section 9(X) NH will be a divisor in IG] invarant under T (but not necessarily pointwise invariant). qed. 617, We continue the previous notation, some of which we now recapitulate. T is an automorphism of X_ of prime order n, and ox, > x, is the quotient map. y generates M(x) over M*(X,) where y® e M*(X.) s “1 Y= > fyAyte (E,) jet where E) is a divisor on X_. We will not assume s > 0, although most q of the complications which follow are the result of assuming T has fixed points. Let a be a complete linear series invariant under . We will assume for convenience that ry is without base points; although this assumption is not necessary . Let Gy be an invariant divisor in GR that is TG) = G,. Then -1 2 BAy+e Oy) where 0sg, 8, (modn) . J For example it gy = K, , the canonical series on X,,, then s “4 > (n-DA,+9 (Ky) jet67 It ae ~ IKK] then for isksn 6, Dy BA, #9 (ek, +e ay +o). a= lA). Let L= L0G.) = (F eMQ&)|(F)+6, 20}. Then L(G,) is invariant 0 Pp 0 0 under T and dimL= R+1. Let L = (FeL[Fet-<™F},m-= 0,1, m on. Let d_ =dimL . ™m ‘m 5 —— Then L= Dib, and Reis Did, m=0 m=0 Let 8. = {(F)+6,|FeL} where 8 = if 4. = 0. m 0 m.: m m We want a more explicit description of the divisors in an : 5 -1 Note that (y™) = 2A, +9 (mE). i If FeL then (F/y™)oT = F/y™; thatis F/y™e M*(X,). Suppose (F)=G_-G, where G@ «8 . m 0 m m (F/y™) = g71(D) for some divisor D in x, ~G =(y™ al thus G_ (y™) +9 7 (D) Gt ym sot). -1 1 -1 Gy =D, Bj +O Dy) + D, MEA, +P omEQ) #9 "(D) A. = Dimt,+)a,+9 (D") — =D tnyAyt 970") where t,,, = mf, +g, , the smallest non-negative residue of mf.+g (modn),D' and D" are divisors on 2 ,and D">0. Lemma 6.17. If an is non-empty then an is the set of all divisors68 oat | : of the form > tay Aj (pq) where a) the divisor is linearly equivalent to G), and b) D__ is integral. Proof. That any divisor with these proeprties is in 8 follows by reversing the above argument. Theorem £47, For each m= 0,4,-++,n-1 there exists on X, a divisor D_, not nessarily integral, so that s . 4 a) Ge > tmp Ay (D,,) and jet b) a = dimID I 7 Proof. For part a) we unwind the previous calculation. -1 -1 ee Daaee (y) - DGrmiya-o (B) 2 -1 oat .- 2 miajome Ey) +9 Oy,) a -1 =-™) +9 D,) where D_ = mE, +D)~B and B is a divisor supported in fa,j++.a,). Thus oe: S= Dts @,,)- j Now dimiD, |= 0 if and only if there is no integral divisor equi~ valent to D_. By the lemma this is equivalent to 8 = @ , which in tum is equivalent to d_ = 0. If dimiD_ 12 1, let D° bea fixed divisor in IDI. Then fix FocL_ where m 0 1,0 (F)= > taj Ayt® Di.) ~So- We define an isomorphism of L(D° ) onto L_ as follows. m ‘m69 If ()+D° >0 let D' = (f)+D°_. Then m m m : 0 eee (Fe) = Das Ast (D),) - &: ira that is, F°@*f is in L- The map f > F*f is the required isomor- phism. For if FeL_ then F/F° is seen to be y*f for some fel? ). qed. 618. Lemma. Let r = s(n ~ 1), the ramification of the covering n-1 os Xoo ont 9X, X,. Then > > ts ng. m=0 j=l Proof. As m runs through the integers 0,1,---,n-4 ny 7 jt mf, does also since f # 0 (modn). Thus s n-1 s S Saye Seger sar ‘mj 2 2 2 j=1 m=0 Jel qed. s In Theorem 6.17 degD,, = (N~ x ) /n. Applying the jst Riemann-Roch theorem to ID. | on X,, we have dimID, |= degD-q+i+i, where i,, is the index of ID_ |. Summing this over m gives Rete Ya-n- CO D tay) Jao mta- + yi, ™ m 4 ™ N-5-n(q-1)+ Ya ™ noi Sree m=0 by the Riemann-Hurwitz formula. Another application of the Riemann- Roch theorem on X_ gives this P m70 n-i Theorem 618. i(g%y)- > i m=0 Corollary 6.18. If g® x '8 Not special then : dag (W~ Do te,)/a- att jet a non-negative number. 619. We now assume that a is non-special and we fix A, . Assuming an = @ we ask how the divisors in an contribute to the g® iy RoN-gapsat Ay. Fixa Gin 8 where a Ga Qa tmyAyt® Dp) J if D,,=n'a,+D',D'20, ne : R . (a,,D) 0, then tap tan isa g w nee gap at A,. Thus the divisors in 8. contribute d_ non-gaps at A, and all of these qn non-gaps are congruent to t_, (modn) . Since there are Reis > d,, non-gaps at A, we see that the only non-gaps at A, congruent to t|, come from divisors in 8, If a, is nota t D point (to be expected in general) then the non-gaps congruent tot, at A, are n,t ,¢@nyerryt +d - tn. ‘me ‘mt m t + me” me Definition. Let d_, be the number of integers in 0,1,++-,R congruent to t. , (modn) . n-1 n-1 (We remark that > day 7 R+1= > an) m=0 m=0n Writing w(G,A,) as > n,~ J we obtain: R R wGAY= Dny- Di 370 j=0 act tnt act fmt - Cue t00- 2 Qe Gage 15 m=0 370 m=0 j=0 act 4m? : > Cae 5m) “emt j : 3 3 . 12 (‘me + 4g1Mt tne tn OR) (4a - de). = Since S (4, ~ djy)) = 9 we obtain the following. m Theorem 6.19. If w(D_,a,) = 0 for all m then act 4 wiGa)) : 2 > [2 at * n(d, dip )] Gap ~ dae) . m=0 Definition. The right-hand side of the equation in this theorem will be called the expected weight and denoted w,(G,A,) . Corollary 619. (1) w(G,A,) 2 wiGa) (2) wiG,A)) 7 wGA,) (modn) Proof. If a, isa D_ypoint for some m_, the contribution of an to w(G,A,) will increase by a multiple of n over that computed in the theorem. qed. 6.20 Lemma, The formula for w,(G,A,) in Theorem 6.19 depends only wee Ris none on Uf, rf.8,¢+%48, provided g® is non-special. Proof. We show w,(G,A,) does not depend on N or q.2 Suppose first that q is fixed and a. has the same g,,-- as a Now N= xs (modn) so N= N' (modn). If Ren N'=N+n then a =g . Since n 3 dy = (N- D tmj)/m- 441. dey is increased by one. 4... the number of integers congruent to t_, in {0,R] is also increased by one. Since > (da, - dy) = 0 the result follows. m Now suppose q is increased by one while N is fixed. Then 4 is decreased by one. p is increased by n,R is decreased by n and so d__, is decreased by one . qed. We now consider the extent to which fixed points of automorphisms are higher order Weierstrass points, that is, kK-points for k 22. Theorem 6.20, Suppose p22 and T is an automorphism of prime order n with three or more fixed points. Then each fixed point is an ’K-point for 222 and !=0 (modn) or 2 =1 (modn) (The case ! = 1 (modn) appears in Farkas-Kra [11]). Proof. Consider the cover XX, (= XT») By the above lemma we may apply the results of the case Xango P1 provided p-nq2 2. (Since s23,p-nq21.) But then the results of Theorem 6.13 (4) apply. The two cases when p-nq=1 are (i) n=2,s*4 and (i) n= 3 and s=3. Then the two cases to consider are (i) W, > W, where T? = 1d and (i) W,>W, where TS = 1d. Here we use the forrnula in Theorem 6.19 directly. We will only work out the second case W, > Ww, is efefe= = git In this case f, fy 1. For 3K, (=g 18) we may take 6, = 6) =6,=0. Then d=6,d,-5,and d,=4 while a = 5. Then doy = yy73 w(3Ky,Ay) -4(2 +0 +316 +5) )(6 - 5) +4(2 : 1 +36 +5))(-5) +$(2-2+34+5)) 4-5) et. Actually, as soon as we have dy > Soy we know there is an extra- ordinary 3K ‘qnon-gap at A, and so A isa 3K, qpoint - Similarly, w, (4K,,A,) = 4 qed.Chapter 7. Automorphisms of Riemann surfaces, IJ; N(p) . 1 In section 5.5 through 5.8 we discussed autornorphism groups of order 84(p- 1) on Xp. If Xp admits a group of automorphisms of order 84(p- 1) then we immediately know that N(p) = 84(p - 1) If, however, N(p) < 84(p - 1) then determining N(p) involves two problems: the existence problem, to show there is a Riemann surface of genus p admitting a group of the conjectured order; and the eli- mination problem, to show that no Riemann surface of genus p admits a group of order greater than that conjectured. The problem of determining the function N(p) for all p appears to be extremely difficult. Here we shall show how to find some infinite sequences of genera, Py , where N(pq) can be determined. 22 In showing the existence of a group of automorphisms whose order competes for N(p) , very often we will be extending an abelian group by a group of genus zero. In using Macbeath’s method of lifting automorphisms, Theorem 4.13 simplifies matters considerably. Theorem 7.2, For p> 2,N(p)28(p +1). i/(2p+' Proof, Fix an integer p22. Let =e"??? Ler B= (tk]k = 0,1,..., 2p +1), the vertices of a regular (2p + 2)-gon inscribed in the unit circle. Let D,, be the dihedral group in pt generatd by Vz= 27! and Rz= 72. Dypag Permutes the set B. If ¥) isa path circling = 4, let 5, = RY, , a little path circling xk, Let Z, =

. Let usm, (Pt - B,2) ~ 22 take each ¥, onto g. The kernel of u in H,(P?-B,Z) is 2pet (> old c, = 0 (mod 2)} 0 a subgroup of index 2 which is invariant under D,.,,. The cor- responding 2-sheeted covering, with points filled in above the 2p +2 points of B, is a hyperelliptic Riemann surface Xp. Daps2 lifts to Xp which together with the hyperelliptic involution yields a group of automorphisms of order 8(p +1). qed. 23, We continue in the same vein. Theorem 7.3. If 3 divides p then N(p) 2 8(p +3).75 Proof. Consider the symmetric group S, acting as the orientation preserving symmetries of the cube inscribed in the Riemann sphere. Lable the 8 vertices a 4,2, so that any two of the vertices with odd indices lie at opposite ends of a diagonal of one of the square faces of the cube. Let Ay = {a,,a,,a,,a)} and A, = {a,.a4,a,.a9) . The even indexed vertices have the same Property. Let A= A, UA). For 2) A and any positive integer _m r2gr define pL: (Pt ~ A,2,) > Z,, (= ) by sending paths circling the odd indexed vertices into g and paths circling the even indexed vertices into o”!. The elements of S 3, Permute the sets A,,A, so that S, lifts to the Riemann surface X_ covering P! corresponding to the kernel of. By the Riemann-Hurwitz formula 2p-2=-2m+8(m- 1) or p= 3(m-4). The order of the group on X, is o(S,)o(Z,,) = 24m. Thus X, admits a group of order 8(p +3). qed. (The results of Theorems 7.2 and 7.3 together with the result that these bounds are sharp infinitely often were obtained independently by Colin Maclachlan [20] and the author [1]). 24. We shall soon see that the lower bound 8(p + 1) s N(p) is sharp infinitely often, so the following definition is justified. Definition. A group G of automorphism on a Riemann surface X, will be called big if 0(G)>8 (p+ 1). In this section we shall show the existence of many sequences of big groups whose orders lie in arithmetic progressions. These groups will all be (2,4,A)'s . So let us look ata (2,4,d) = where a? = b4=4=e,c* ab, and the order of the group is wr, where UL is the index of the cyclic group . Let N be the smallest normal subgroup containing b2. Then is an extension by a dihedral group of an elementary abelian 2-group. Theorem. (W.T. Kiley [16]). Let & be a positive integer. If p=(1~- 2") (mod 221) then N(p) 2 8(p-1+ 24).76 Proof, Fix positive integers ¢ and m. Let Ae (th[ x = e2mi/2tm 2 0,1,-+-, 2tm- 1) the vertices of a regular 2#m-gon inscribed in the unit circle. Let A,=(tklk j (modt), j=0,4,...,8-4. There are 2m points in each set A,. Let D,,,, be generated by R and V where Rz= 72 ae i and Vz=27!. Then Dy 4 Permutes the sets {Aj}. Let a circle t* as before, Define U1: m,(P*-A,29) = (Zp)" (eq £ A) where ee y= 9.90%? bY u(s,) = @_ where k =k (mod). e et oe In H,(P°- A,Z) kerp = oy Sat, | > $20 kaj kaj which is invariant under D, © (mod2}} , The Riemann surface corresponding to ker U, X, , has a group of _ i automorphisms of order o(D,)_)o((Z,)!) = 4¢m - 2 By the Riemann-Hurwitz formula 2p-2=-2-2+20m- 2th, Thus 4-2'-tm= 8(p-1+2') and p=1-2!+m- 202! for all imei. qed. For ¢ = 1 we have the result of Theorem 7.2. For t= 2 we have: if p= 1 (mod4) then N(p) > 8(p +3) a result different from Theorem 7.3. ZS. We now state a theorem which shows that the lower bounds on N(p) discussed in the last three sections are sharp infinitely often. Theorem 7.5. (W.T. Kiley [16]. Suppose we have positive integers s and t sothat 522 and 4|st. Then there are an infinite number of genera p so that 1) N(p) s8(@-1+5) and 2) p= (1-5) (mod(st)/4) . {In Theorem 7.4 s= 2' and t= 28, t= 1,2,....) The proof of this theorem is an elaboration of the basic idea in Theorem 5.8 where Macbeath showed that N(p) < 84(p - 1) for infinitely many p. There p= n+1 where n was a large prime.7 Here we shall assume p- 1 = sn where again n isa large prime satisfying further conditions. Moreover, a closer examination of the steps in the proof will allow us to actually compute N(p) where p=n+i,or 2n+41, where n is prime. The proof of Theorem 7.5 will occupy the next several sections. 78. - From now on n will denote a prime integer and p will denote the genus of a Riemann surface X, where the subscript may be omitted if no confusion follows. In the sequel we will need to recall Lemmas 5.5 and 5.7. From elementary group theory we recall that the group of automorphisms of Z_ is Z__, . Also by using the transfer one proves that if a finite group G admits a central abelian subgroup S of index i, then the transfer homomorphism from G into S$ is g— g!. If S is a Sylow subgroup the map is surjective with kernel of order i [27]. Other results from group theory will be proven in the propositions below. Now we catalogue the big groups. If X admits a big group of automorphism, G , then the genus of X/G is zero (Lemma 5.5). Recalling the discussion of section 5.5 one sees that the possible big groups on a Riemann surface X_ (p22) are (for convenience we will divide them into 3 types): Type (I) (2,5,2) 5 84s; (ii) n divides 0(6) . Then G admits a normal subgroup H where G/H is a cyclic group which is one of the following 4 possibilities: ( (2,3,6) G/H has order 6 and X,/H has genus one ; (ii) (2,4,4) G/H has order 4 and x /H has genus one ; (ii) (3,3,3) G/H has order 3 and ae has genus one ; (iv) (2,5,10) G/H has order 10 and x fa has genus two. tf 3] @-4) then o(6) < 24(p- 4). if 3 (4) and 4[™-1) then 0(6)< 16(p - 1). If (12,n-4)+2 then 5|t-4) and 0(G) < 10(p - 1). In particular, ((n - 1)/2,30) > 4 Proof. Let S_ be then-Sylow subgroup of G. Now 84(p - 1) = 84sn2 0(6).. Consequently S_ is a normal cyclic subgroup, by Sylow’s theorem. By Lemma 5.7 X,— X /S_ isa smooth covering, Consider the homomorphism © :G — Aut(S_) given by @(g)(h) = ghg. Let H be thekernelof ©. Then SCH and H is the centralizer of Sin G. $_ is thus a central subgroup of H, and 4 is a cyclic group whose order divides n- 1. The transfer map H > s. since Aut(S_) = can , G/H is isomorphic to a subgroup of 2 G/H is surjective and the kernel of the transfer, N , is a normal complement to S. in H; thatis,H=NS_ where NOS. = . We thus have n 2 the following diagram. X_/G (genus 0) P79 Now H/N#S_ and S, acting on X, is without fixed points. Thus the covering X_/N > X_ /H is also a smooth covering of n sheets, Pp Consequently, the genus of X./H is positive. By Lemma 5.5, o(H) < 4(p- 1). Since G is big, o(G/H) > 2. Now G isa (v,.U,,..,0,) group (s = 3 or 4) and so G/H isa yr (Hy,02.-0.9,) group where ¥ - Since G/H is eyclicand X_/H has positive genus, 3 of the y's. must be larger than one. By consulting the list of big groups one sees that the 4 possibilities for G/H listed in the statement of the theorem are the only possibilities. Consequently 3,4, or 5 must be a factor of n-1. lt 3|(n- 1) then G/H could bea (2,3,6) of order 6. Then ae has genus one so that 0(G) < 6-4(p - 1). The remaining parts in the statement of the theorem are proved similarly. Note that if thegenusof X_/H is 2 then o(H) < p- 1. (Seethe proof of Lemma 5.5.) qed. Corollary 7.7. Let s bea positive integer andlet n bea prime. Let P=snt1 andlet G bea big group on X.. Suppose ( nz 84s (ii) (23,30) =1 (ie n= 23,47, or 59 (mod 60)). Then n _ 0(G) and G is not a big group of Type |. Proof. That n ff 0(6) follows immediately from the theorem, Now consider the 35 possibilities for the big groups of Type I. The80 (2,2,2,3) is excluded since it has order 12(p- 1) and n[p- 1 The other 34 big groups are (a,p,8)'s. By the Riemann-Hurwitz 1 1 formula 2p - 2 = 0(6)(1 - 2 - $7 P)- We see that 2p - 2 = 0(G) 4 where A and B are integers. Since n f 0(6) it follows that n]A. By working out these 34 cases one sees that the largest prime dividing an A is 47. Since n> 84, this rules out all Type I big groups. qed. 72.8. The following group theoretic proposition will eliminate the (2,3,n)'s. Proposition 7.8. Let G bea non-cyclic (2,3,A) of order Ud. Then where a? = b? = ch= d\c=ab and d=ba. Since de= b?, G= . Now N is normal, in fact, central in G. Consider the following diagram «o> a Ne «er G / alk xe where ku is the index of in G and k}]r. Now ku. For in G' = G/ N = the cyclicat subgroups and intersect only in C¢e> and so G is cyclic. Similarly, 3|ku The transfer t:G > isthemap g— g*#. But 6[ ku; therefore,a and b map into the identity; that is, g > g*" is the trivial homomorphism, Consequently ck# =e or 4|ku. Thus ds kus p?. That 4] k(d,u) is now immediate. qed. Theorem 7.8. Assume the hypotheses of Corollary 7.7. Assume, moreover, that n > 216s? = 6°s?. Then G isa (2,4,) group. Proof. If G isa (2,3,X) of order ud, then by the Riemann- Hurwitz formula ans = 2p-2=0(6) (1-3-3 -F)= ua(é-2 = EQ - 6) or ans = ur - 6). Since nf’ u we see that nl a-6) or \-6=7n. Thus 12s= th and = 128/17 < 12s. By the proposition 0(G) < (12s)>. But G is a big group. Therefore 0(G) > 8(p + 1) > 8(p- 1) = 8ns 2 (12s)°, a contradiction. qed.a2 1.9. Now we prove another group theoretic proposition, very similar to that in 7.8, to help eliminate the unwanted (2,4,A)'s. Proposition 7.9. Let G bea (2,4,d) group of order yr where u>4,4|u and »> yu? (0(G)> 3). Then there is a positive integer k so that 3 where a’ We show first that G+ . @=e,c=ab,d=ba. Suppose G= . Since 4] it follows exactly as in Proposition 7.8 that if the index of M is ku, the transfer homomorphism G6 > N, g > g*™ is the identity and consequently A]km. Also k< Ul as before, so x < u2. But this is excluded by hypothesis. Now is normal (since aca = d) and must now have index 2 in G. Consider the diagram «o> 7 A / 9 = we see as in Proposition 78 that ks u/2. If k= 41 and so w= 2, a contradiction.83 If k= 2 G/cc> N isa (2,4,2), of order 8 andso u=4,a contradiction. If k= 3 G/N 216s”. Since 60|¢ we see that n= 59 (mod60). Also n> 84s. Let p=sn+i1. Then p= 1-s4#s(n+1) =(1-s) (modc) = (1-5) (mod()) . By Dirichet’s theorem on primes in an arithmetic series, there are an infinite number of such genera p. It remains to show that N(p) < 8(p-1+5). Suppose G is. big group on X,. Thus by Corollary 7.7 and Theorem 78 G isa (2,4,d) group of order, say WA and n does not divide 0(G). By the Riemann-Hurwitz formula 8p- 8 = w- 4) Now n divides p-1 but n does not divide 0(6) . Therefore n|(A- 4). Let X- 4 =n. Then ans = 8(p- 1) = wnt or w= 8s/t. Since 0(G) = wr = 8p- 8 + 4u it remains to show Us 2s or T2484 We eliminate the other possibilities for 1. Now hent+4=tint+4)+(4-7). Letting t= 1,2 or 3 weseethat 4-1 =3,2 or 1. We apply Proposition 7.9 to produce a k so that 3cksu/2(: 45/1) and k|x. If k= 3 then u=8. By the choice of n wealso have k[(n+1). Thus k divides 1,2 or 3, all cases which are impossible (k= 3 + t= 1 > wz 16). Thus t2 4. aed. LAL In order to apply the proof of Theorem 7.5 to produce genera p where N(p) is known, applying condition (a) in Section 7.10 together with n> 216s? quickly leads to rather large numbers. We conclude this chapter by proving several theorems which allow us to compute N(p) for some genera which are relatively small. Theorem 7.11. Let n bea prime integer,n > 84 (p-1=n). fa) if 3](-4) then Nim +4) = tan; ie. N(p) = 12(p- 1). () If 3 (n- 1) and §|(n- 1), then N(p) = 10n; ie. N(p) = 10(p- 1). (c) If (n~ 1,15) = 4, then Nin +1) = 8(n+4); ie Nip) = 8(p +3). The proof of this theorem is derived by examining more closely Theorem 7.7 and 7.8 when s= 1. ‘212. Theorem. Let n bea prime integer,n > 84,p=n+1 If 3]{- 4), then N(p) s 12(p- 1).a5 if 3 @-1) and 5 | (n- 1), then N@) < 10(p~ 1). If (n- 1,15) = 1 and G isa big group, then G is not of Type and nf 0(G) . Proof. Again consider the diagram in the proof of Theorem 7.7 when. assuming that alot) where G is a big group. Ls Zo xn ee \ 7% X /H Pp X,/G Genus 0) Since X — X _/S_ is smooth we see that the genus of X /S_ is 2. P pon pn The genus of X./H is one or 2. A Galois cover X, > X, must have precisely 2 sheets so the order of H is mn or 2n. If the genus of X_/H is one then G/H isa cyclic group of genus one and therefore has order bounded by 6 , and the order divides n - 1. Thus if 3[(n~1) the maximum order for G is (6)(2n) = 12n. (If 0(G) = 12m one sees that G must bea (2,6,6)) . If 3} (n - 1) and the genus of X,/H is one, the maximum86 order for G/H is 4 and 0(G) s 8n which is not big. Soif 3f n-1 and G is big then the genus of X/H is 2. In this case G/H must be a cyclic group of genus 2,a (2,5,10) which has order 10 Thus if 3 f (n- 1) and 5 | (n- 1) then the maximum order of G is ton. If 3 (n-4) and | (n- 1) then the situation illustrated in the diagram is impossible and n ! 0(G). Then we eliminate groups of Type I by the same argument as in the proof of Corollary 7.7. qed. 243. We now complete the proof of Part (c) of Theorem 7.11. We assume p=nt+i,nisprime,nz 84, and (n-1,15)=1. X) admits a big group and n ! 0(G). By examining the proofs in Section 7.8 we show that without any further restrictions G cannot bea (2,3,r). If G isa (2,3,d) of order ud then thereisa k so that k[y,ksu and a[k(O,u). Since p-1=n, the Riemann-Hurwitz formula gives 2p- 2= wa(g - 4) or 12n = ulr- 6). As before n|\~6 so that »-6= cn and 12= ut. The possi- bilities for + are 1,2,3,4 and 6 and correspondingly w= 12,6,4,3 and 2. If t=1,p=412,=n+6 then (u,a)=1. Thus »|k; thatis Asks we 12. Consequently n+6 12. This contradicts n2 84 Thus t cannot be 1.a7 If t=2,u=6,2=+2n+6, then (u,d) = 2..Thus 4] 2k and so Qn +6 s 12, again a contradiction . For t2 3, <4 so directly by Proposition 78 0(G) < 64. Thus our big group is nota (2,3,%) . We conclude that G must bea (2,4,4) group of order War. Again the Riemann-Hurwitz formula gives 8n = W(A- 4). Again »-4=tn and ut=8. Therefore o(G) = na = S(tn + 4) s 8+ 4) = 8D + 3). But if (n- 1,15) = 4 then n= 2,8 or 14 (mod15). Consequently p=n+1=0 (mod3). By Theorem7.3 N(p) = 8(p+3). The proof of Part (c) of Theorem 7.11 is complete. 2.14, To complete the proof of Parts (a) and (b) of Theorem 7.11 we first exhibit two groups of genus 2 which must act on X,/S, in the diagram of Section 7.12. The (2,6,6) is Z, x Z, = where a? = bo =e. The (2,5,10) is 249 = 9) where @ generates Z,,. if 3|(2-4) (a prime) we want to extent Z_ by the above (2,6,6) toobtaina (2,6,6) of order 12n. This will complete the proof of Part (a) of Theorem 7.11. if 5|(2~4) (n prime) we want to extend 2, by the above (2,5,10) to obtain a (2,5,10) of order 10n. This will complete the proof of Part (b) of Theorem 7.11.88 Proposition 7.14 Let G bean («,8,%) where the lcm. [a,p) divides ¥. Let G= where a% = bP = cl =e (c= ab). Assume «,p,f2 2 and at most one of these integers is 2. (i) Suppose there exists a surjective homomorphism 8:65 Zhe8] so that 0(@(a)) = « and o(@(b)) =p. (i) Suppose n is prime and [«,8] |@- 1) but nfo). Then there exists G’, again an («,8,%) of order o(G)n,and G@ satisfies hypothesis (i) . Proof. @ maps G onto Z which is a subgroup of Aut(Z,). Let [x.e] G' be the semi-direct product of G and Za . That is, G' = where G = , =Z_ and 0 = , together with the following relations: aga’ = 1 (mod n),x #1 (mod n); ppb! = @” ,y®=1 (mod n),y #1 (mod n) ——D—™rté—<—~—OO——S—, a subgroup of G', isan («,p,%). To show that G = we need only show € . Consider the commutator [a,bg] €. lapel = = cpa Mby)t * Pca tp lgy = gv¥&Dig,b} [a,bg}?(S) = gy&- el) Now y(x- 1)0(G) # 0 (mod n),s0 g € . Since is normal in G' and G'/<@> = G, the proof of the Proposition is easily completed. qed. The proof of Theorem 7.11 is completed by applying the proposition to the two cases discussed above, (2,6,6) and (2,5,10) . 15. We now show that the two other bounds given in Theorem 7.7 are sharp infinitely often, by letting s = 2 in the theorem. Theorem 7.15. Suppose n is a prime integer,n2 168. Let p= 2n+1 if 3|(-4) then Nip) = 24(p- 1). Hf 3 Pm-1) and 4|m-1) then Nip) = 16-1). Proof. It suffices to show the existence of two groups satisfying the hypotheses of Proposition 7.14, namely: (i) a (2,3,12) of order 48 and genus 3 witha (2,3,6) of order 6 as a homomorphic image. (i) a (2,4,8) of order 32 and genus 3 witha (2,4,4) of order 4 as a homomorphic image. (i) Let Ay = (2,3,3) be represented as the symmetries of the regular tetrahedron inscribed in the Riemann sphere. Let90 B= (b, b,b,} be the 4 vertices which are fixed points for 1 1Pys . a (ptepp) o> elements of order 3 in A,. Define w:n,(P!-B,b)— Z, (= ) by (3) = @ for each of the 4 paths circling the b/s The kernel of [1 is associated to a 3-sheeted cover of P' with total ramification 12, that is, a Riemann surface X,. A, lifts to X, and so Xs admits a (2,3,12) of order 48. If the group is G = , a = b® = (ab)! = e, it is not difficult to see that the commutator subgroup of G is (= G') and G' is isomorphic to the quaternion group of order 8. Thus G/G' = Z, and G/G' is a (2,3,6). (ii) Let Zz, x Z, = ,a (2,4,8) of order 32. The subgroup is seen to be normal and G/egy® = Z,*Z,,a (2,44). Thus G/ , ag? = Z,+a (2.4.4). All the Riemann surfaces occuring in Theorem 7.15 are elliptic-hyperelliptic. If G = ,c =ab then c® (resp. c4)1 is the elliptic-hyperelliptic involution in the case of (2,3,12) (resp. (2.4,8)) . 1.16. We continue this discussion of N(p) by showing that the results of Theorem 7.11 hold for any prime integer n> 11, n= 13, Macbeath (19] showed that N(14) = (13)(84). The following remarks will be useful. If a Riemann surface of genus p admits an automorphism group of order m(p - 1) and m 2 24, then the only possible values for m are 24, (132/5), 30, 36, 40, 48, and 84. The reader will see that part of the following discussion depends on looking fairly carefully at the 35 Type I groups. This is slightly tedious but quite easy. Also a useful table of non-cyclic simple groups of low order is appended to this chapter, Lemma 7.16. Let n bea prime integer,n> 11, n= 13 and n¢84, Let p=n+4. Let G bea big group on Wp and suppose n| 0(6). Then nf 0(G) , Gis solvable, and the n-Sylow subgroup is normal in G. Proof. The only big automorphism group of order m(p-1) where m is prime is a (2,3, 78) of order 13(p- 1). Consequently, n°’ 0(6). Suppose G is not solvable. Then G has a non-cyclic simple factor in any composition series, If o(G) = mn, we may assume m < 84, and so 0(6) <(84)” (+ 7056). By considering the non-cyclic simple groups of order less than 7056, we see that the relevant prime factors of the orders of these simple groups (11, 17, 19, 23) have indices (eg. 60, 144, 180, etc.) inappropriate for automorphism groups. We conclude that G is solvable. Since G is solvable, in any composition series there is a group H with a subgroup N, normal in H , so that H/N = zy Consider the92 coverings: W > WIN W_/H. Pp Pp Pp The covering W_/N > W /H isan unramified Z_ covering, and P P n so the genus of W, /H is positive, Consequently, off) s 4(p ~ 1) = 4n. Therefore, o(N) s 4 and S_ (the n-Sylow subgroup) is normal in H(n2 11). S_ isa characteristic subgroup of H and so is normal 2 in the next subgroup of the composition series and so on until we reach G. qed. Now that we know that Ss. is normal, the proof of Theorem 7.11 proceeds exactly as before until last sentence, where we consider Type ! groups, G, with n f/o(G). Such a group has order (B/A) (2p - 2) = (2Bn/A) where n must divide A. The possible values for A are 11, 17, 19, 23, 29, 31, 41, and 47, and so these are the values of n to consider. In all but two cases the (xp, 8) Type! group is perfect ((«,p) = (6, 8) = (8,0) = 1) and so must have a non-cyclic simple group is any composition series. We then see that, with one exception, 0(6) is not divisible by the order of a non-cyclic simple group and so cannot occur, The exceptional perfect group is a (2, 5, 9) group of order (180/17). However, Paul Hewitt showed the author that there does not exist a (2,5, 9) group of order 180. The two non-perfect Type I candidates fail because the group would be too small. For example, a (2, 5, 16) group of order (160/19) on Woo would yield a group of order 160 which is less than 8(p + 1) (= 168). Thus Type I groups are eliminated from the93 competition to achieve N(p). An examination of the proofs in Section 7.13 shows that they work equally well for the primes now under consideration and so Theorem 7.11 does indeed hold for all primes nz 11,n = 13. p = 12 is the smallest genus for which N(p) = 8(p + 3). 42. Now we consider N(p) for p= 2n+1,n primeand n2 11 We will prove the following. Theorem 7.17. Let n bea prime integer greater than or equal to 11. Let p= n+1. a) If 3 | (n~ 1) then N(p) = 240-0) b) if 3f (n-1) and 4|(n- 4) then N(p) = 16(p - 1) & If (m- 41), 12) 4 then N(p) = 8(p + 1) Remarks. p= 23 is the smallest genus for which N(p) = 8(p + 1). Paul Hewitt showed this by finding all previously unknown N(p) for p < 25. Notice that we have lower bounds in parts a) and b) of the theorem by Theorem 7.15 The integers n and p will always have the above meaning in this section, We now adapt Theorem 7.7 to the present situation. Lemma 7.17. Suppose 0(G) = N(p) and n | 0(6). Then a nfo. b) For n< 168, G is solvable. ©) For alln,S_ (the n-Sylow subgroup) is normal in G. Proof. 0(G) = m(p - 1). The only case where m is prime is a (2, 3, 78) of order 13(p ~ 1) . This is excluded for p = 23 since 394 does not divide the order of the group. Thus n’f{ 0(@). If 0(@) = m(2n) then ms 84, and 0(6) s 168n. If n> 168 then s, is normal. Hf n¢168 then 0(6) < (168)" (= 28, 224), But all non-cyelic simple groups of order less than 28, 224 have the relevant cyclic Sylow subgroups with too great an index to be an automorphism group of a Riemann surface, with the possible exception of a simple group of order 660 (= 22.30) We consider the possibility of a (2, 3, 10) on Was: Sy in this group is unramified and by Sylow’s theorem has index 5 in its normalizer. Thus W,,/S,, admits an automorphism of order 5. We now reach the contradiction by observing that the genus of Wo5/S4, 38 3. Thus G cannot be simple. In fact, it must be solvable since a non-cyclic simple group cannot be fitted into a composition series. We now use the argument in Lemma 7.15 to show that S_ is normal. qed. We can now apply the argument of Theorem 7.7 which results from $_ being normal in G In the figure accompanying the proof of that theorem w,/s, has genus 3. Consequently, possibility (iv) in the statement of the theorem is ruled out because, again, a W, does not admit a a . Thus in the light of Theorem 7.15, the conclusion of Theorem 7.7 for our n and p is:95 It 3 | (n~1) then 0(6) = 24(p- 1). it 3 (@-1) and 4|(m- 1) then 0(6) = 16(p- 1). In any case, (n - 1, 12) # 1. Now Corollary 7.7 reads: If G is big and (n- 1, 12) = 4 then nf 0) We now eliminate Type I groups by observing that their orders lose out in the competition or they are impossible because they are perfect. To eliminate Type II groups we proceed as in Proposition 78. @ is a (2,3, a) of order ur where nf of). 2p~- 2 = (1W/6)( ~ 6). Conse- quently, there is an integer t which divides 24 and = tn +6, and w= 24/1. Finally there is an integer k dividing A so that »|kKO,w and ks ph If t= 1 wesee that (s,u) = 1 since n is prime, andso d= k. It follows that n<18 and so » has the possible values 17, 19 and 23. Since all these groups would be perfect, they are eliminated. If t= 2 weseethat (A, u)=4,andso 4<48,or n<21. The possible values for n are 19, 17, 13 and 11, and the possible values for A are 44, 40, 32 and 28. The first three possibilities lead to groups that are too small. Thus we are led to consider the possibility of a (2, 3, 28) of order 336 ona Wos- If we denote Gby with c= ab and d= ba, then G = and the normal subgroup <> 1 must have order 4, The quotient of G by this normal subgroup would then have to be a (2, 3, 7) of order 84 , a contradiction. The larger values of t are easily eliminated and we see that our96 big group must be a (2, 4, d). For these groups we have: 2p-2= 4n= u(r-4)/4 or 16n = nlr- 4) since nj 0(G) we see that there is an integer t so that 4-4= tn, andso = tn+4, 1 =16/t ando(@ = 8(2n+(8/t)). The proof will be completed by showing that t= 1 or 2 is impossible. We let G = «a,b», c= ab andd = ba, as before. We must first show that G = . Assuming the contrary, we use the transfer onto the central subgroup N to find an integer k , dividing A,k |uk. uw is always a power of 2. If t= 1,2 is odd. Now G is simultaneouly a (2, 4, ) and a (A, 2, 2). Thus the abelianized G must be cyclic with order dividing 2 and » Thus we reach the contradiction that G is perfect. If t = 2 , we see that (A, w = 2,2 < 16 andso n_ is too small. We conclude that the index of in G is 2. Now refer to the diagram accompanying Proposition 7.9. First assume t= 2. = 2n+4,and so 2 is not divisible by 4. Consequently, k= 2 or 3. Butif k= 2,G/ M isa (2, 4, 2) of order 16, a contradiction. If k = 3then 3 | (2n +4) which implies that 3 | (n - 1), in which case the Type Ill groups are not in the competition. Now assume t= 1,u+16,A-n+ 4,and k <8. Since n is prime, k must be odd. Let G' = G/ NM . k = 3. implies that G is a (2, 4, 3) of order 48, a contradiction. k = 5 implies that G' isa (2, 4, 5) of order 80 , again a contradiction since it implies that a W, admits an automorphism of order 5. If k = 7, we havea (2, 4, 7)7 of order 112. Let cand d’ be the elements in G’ corresponding to c and d in G. o() is 56, and since S_ is not normal in Ke, d>,S, is. ofS, = 8 and being characteristic in , it is normal in 6’. S, has as non-identity elements - 6}, and so S, is seen to be an elementary abelian group of order 8. G'/S, acts faithfully on S, as a dihedral group of order 14. This is the final contradiction since the automorphisms of Ss, form the simple group of order 168, a group which does not have as a subgroup a dihedral group of order 14 This completes the proof of Theorem 7.17. With the work of Paul Hewitt we know N(p) forall p=n+1, or 2n+4 wheren isa prime integer. Orders of ae simple groups < (168)° = 28224 60 = 2°35 168 = 2s 360 = 2°3°5 504 = 2°3°7 660 = 23511 1092 = 2°3,743 244g = 23°47 2520 = 2°3°57 3420 = Pees 4080 = 2135.17 5616 = ae 13 5 6048 = 23° 76072 7800 7920 9828 12180 14880 20160 25308 25920REFERENCES (1] Accola, R.D.M, "On the number of automorphisms of a closed Riemann surface’; Transactions of the American Mathematical Society, Vol. 131 (1968), pp. 398-408. [2] Accola, R.D. M3 "Riemann ae with automorphism groups admitting partitions"; Society, Vol. 21 (1969), pp. 477- 402, [3] Accola, R.D,M,; “On generalized Adsphtid points on Riemann surfaces’; Wu s in University of Pittsburg, Pittsburg, Pennsylvania (i978), pp. 1-19. [4] Ahlfors, L. V. and Sario, Leo; Riemann Surfaces; Princeton University Press, 1960. [5] Arbarello, E., Cornalba, M., Griffiths, P. A., Harris, J.; Geometry of Algebraic Curves, Vol. 1, Springer-Verlag, New York, 1984. [6] Castelnuovo, G; “Alcune osservazion sopra le serie irrazionali di gruppi di punti appartenenti ad una curva algebrica";_Rendiconti della R. Accademia dei Lincei, Series 4, VII, 1891 (Memorie scelte p. 71). [7] Castelnuovo, G; “Sui multiple di una serie lineare di gruppe di punti appartenente ad una curve algebrica"; Rendiconti del Circolo Matematico di Palermo, VII (1893), pp. 89-110 (Memorie scelte p. 95). {8] Castelnuovo, G; “Sule serie algebriche di gruppi di punti appartenenti ad una curve algebrica’; Rendiconti della R. Accademia dei Lincei Series 5, XV, 1906 (Memorie scelte p. 509). [9] Coolidge, J. L; A Treatise on Algebraic Plane Curves; New York, Dover Publications, 1959. [10] Enriques, F; "Sopra le superficie che posseggono un fascio ellittico 0 di genere due di curve razionali’; Rendiconti (5), 7,, (1898), pp. 281-286.100 (11] Farkas, H.M. and Kra, 1; Riemann Surfaces; New York, Heidelberg, Berlin, Springer-Verlag, 1980. {12] Forster, 0; Lectures on Riemann Surfaces; New York, Heidelberg, Berlin, Springer-Verlag, 1981. {13] Griffiths, P. and Harris, J; Principles of Algebraic Geometry, New York, John Wiley and Sons, 1978. [14] Gunning, R.C; Lectures on Riemann Surfaces; Princeton Mathematical Notes, 1966. [15] Hurwitz, A; “Uber algebraische Gebilde mit eindeutigen Transformationen in sich", Mathematischen Annalen (Vol. 41) (1893), pp. 403-442. {16] Kiley, W. T; “Automorphism groups on compact Riemann surfaces’; i , Vol. 150 (1970), pp. 57-863. [417] Lewittes, J; "Automorphisms of compact Riemann surfaces’; American Journal of Mathematics, Vol. 85 (1963), pp. 734-752. {18] Macbeath, A.M; “On a theorem of Hurwitz"; Proceedings of the Glasgow Mathematical Association, Vol. 5 (1961), pp. 90-96. [19] Macbeath, A.M; Fuchsian Groups; Queen's College Dundee, University of St. Andrews, 1961. (20] Maclachlan, C; "A bound for the number of automorphism of a compact Riemann surface’; Society, Vol. 44 (1969), pp. 265-272. [21] Miller, G.A; "Groups defined by the order of two generators and the order of their product”; American Journal of Mathematics, Vol. 24 (1902), pp. 96-100. [22] Olsen, B.A; “On higher order Weierstrass points’; Annals of Mathematics (2), Vol 95 (1972), pp. 357-364. [23] Schwarz, H. A; Gesammelte Mathematische Abhandlungen, Springer-Verlag, Vols. 1 and II (1890), Berlin.101 [24] Segre, C; "Intorno ai punti di Weierstrass"; Atti della Reale Academia dei Lincei, Series 5, Vol. viii (1899), pp. 89-91. [25] Seifert, H. and Threlfall, W.; Lehrbuch der Topologie, Chelsea, New York, 1945. [26] Walker, R. J; Algebraic Curves, New York, Dover Publications, Inc., 1962. [27] Zassenhaus, H.; The theory of groups, Chelsea, New York, 1949.Index adapted basis affine dimension of an L(D) algebraic series atlas Automorphism Aut (X) base points of a linear series big group biholomorphic function branched covering branching Brill-Noether theorem canonical curve canonical line buncle canonical series Castelnuove's Riemann Roch theorem Cayley representations chart Clifford index Clifford's theorem Clifford's theorem, strong form commuting involutions complete linear series composite linear series compounded of an involution conformal equivalence conjugate automorphism groups cover transformation covering cusp cyclic coverings AX oe ases Pee wivove Ss 4 32,3: Zaw103 S-invariant degree of a divisor dihedral groups differential, meromorphic divisor divisor of a differential divisor of a function elliptic function field elliptic-hyperelliptic involution elliptic-hyperelliptic Riemann surface equivalent coverings equivalent divisors equivalent line bundles expected weight extraordinary gap extraordinary non-gap fixed points of a linear series flex function fields Galois closure Galois group gap generalized Weierstrass point greatest common divisor groups of genus one groups of genus zero groups with partitions G-weight Gy) higher order Weierstrass point holomorphic differential holomorphic function holomorphic section hyperelliptic involution hyperelliptic Riemann surface S hagPavanangan BN oo 18 36 32,34 17 17 44 43 48 59 32104 incomplete linear series index of speciality inequality of Castelnuove-Severi integral divisor inverse line bundle involution (automorphism) involution of genus q k-canonical line bundle kK points LD) lift line bundle linear series linear series invariant under a group imposing linear conditions local parameter maximal automorphism groups meromorphic function monodromy theorem monodromy group of a covering multiplicity of branching M(x) M(Y/X) Ww 4.) group NO) number of sheets node non-gap normal covering o(f) ordinary singularity 12 53 30 10 64 10 45 30 35 35 43 42 17 34 be105 plane curve plane model positive divisor product of line bundles projective dimension of L(D) ramification rational function field regular permutation group Riemann-Hurwitz formula for coverings Riemann-Hurwitz formula for Galois coverings Riemarn-Roch theorem Riemann surface AE) section simple linear series singularities of plane curves smooth covering special divisor surface tacnode transition function transitive permutation group Type I, II, If groups uniformization theorem universal covering Weierstrass point r w@ x), wG,x) n w(G, A) e 2, oh aoe 10 33 77 31 31 13 17 7 41Lecture Notes in Mathematics For information about Vols. 1-1414 please contact your bookseller or Springer-Verlag Vol 1415: M. Langevin, M. Waldschmidt (Eds, Cinguante Ans de Polyndmes. Fifty Years of Polynomials Proceedings, 1988. IX, 235 pages.1990. Vol, 1416: C, Albert (Ed.), Géométrie Sympiectique et Mécanique. Proceedings, 1988, V, 289 pages, 1990, Vol. 1417: A.J. Sommese, A. Biancofiore, E.L. Livorni (Eds.), Algebraic Geometry. Proceedings, 1988. V. 320 pages. 1990, Vol. 1418: M. Mimura (Ed, ), Homotopy Theory and Related Topics. Proceedings, 1988. V, 241 pages. 1990, Vol. 1419: P.S. Bullen, P.Y. Lee, J.L. Mawhin, P. Muldowney, W.F. Pfeffer (Eds.), New Integrals Proceedings, 1988, V. 202 pages, 1990, Vol. 1420: M. Galbiati, A. Tognoli (Eds.), Real Analytic Geometry. Proceedings, 1988. IV, 366 pages. 1990. Vol, 1421: H.A. Biagioni, A Nonlinear Theory of Generalized Functions, XIl, 214 pages. 1990. Vol. 1422: V. Villani (Ed.), Complex Geometry and Analysis, Proceedings, 1988. V, 109 pages. 1990, Vol. 1423: $.0. Kochman, Stable Homotopy Groups of ‘Spheres: A Computer-Assisied Approach, VIII, 330 pages. 1990, Vol. (424: FE, Burstall, LH. Rawnsley, Twistor Theory for Riemannian Symmetric Spaces. Ill {12 pages. 1990. Vol. 1425: R.A. Piccinini (Ed.), Groups of Self Equivalences and Related Topics. Proceedings. 1988, V, 214 pages. 1990. Vol. 1426: 1, Azéma, P.A. Meyer. M. Yor (Eds,), Séminaire de Probabilités XXIV, 1988/89. V, 490 pages. 1990. Vol, 1427: A, Ancona, D, Geman, N. tkeda, Bcole d°Eé de Probabilités de Saint Flour XVIII, 1988, Ed: PLL. Hennequin, Vil, 330 pages. 1990. Vol, 1428: K. Erdmann, Blocks of Tame Representation Type and Related Algebras. XV. 312 pages. 1990. Vol. 1429: 8, Homer, A. Nerode, R.A. Platek, G.E, Sacks, AA. Scedrov, Logic and Computer Seience. Seminar, 1988, Editor: P, Odifreddi. V, 162 pages. 1990. Vol. 1430: W. Bruns, A. Simis (Eds.), Commutative Alge- bra. Proceedings. 1988. V, 160 pages. 1990 Vol. 1431: J.G. Heywood, K. Masuda, R. Rautmann, V.A. Solonnikov (Eds.), The Navier-Stokes Equations - Theory and Numerical Methods. Proceedings, 1988, Vil, 238 pages. 1990, Vol, 1432: K, Ambos-Spies, G.H. Muller, G.E, Sacks (Eds,), Recursion Theory Week. Proceedings, 1989. VI, 393 pages. 1990, Vol, 1433: $. Lang, W. Cherry, Topies in Nevanlinna ‘Theory. I, 174 pages.1990, ‘Vol. 1434: K. Nagasaka, E, Fouvy (Eds. Analytic Number ‘Theory. Proceedings, 1988, VI, 218 pages, 1990. Vol. 1435: St. Ruscheweyh, EB. Saff, LC. Salinas, RS. Varga (Eds.), Computational Methods and Function ‘Theory. Proceedings, 1989. V1, 211 pages. 1990. Vol. 1436: S, Xambé-Descamps (Ed,), Enumerative Geometry, Proceedings, 1987, V, 303 pages. 1990. Vol. 1437: H. Inassaridze (Ed.), K-theory and Homological Algebea, Seminar, 1987-88. V, 313 pages. 1990. ‘Vol. 1438: P.G. Lemarié (Ed.) Les Ondelettes en 1989, Se sminar. LV. 212 pages. 1990. Vol. 1439: E, Bujalance, J.J. Etayo, J.M. Gamboa, G. Gromadzki, Avtomorphism Groups of Compact Bordered Kiein Surfaces: A Combinatorial Approach, XII, 201 Pages. 1990. Vol. 1440; P. Latiolais (Ed). Topology and Combinatorial ‘Groups Theory. Seminar, 1985-1988. VI, 207 pages. 1990. Vol. 1441: M. Cooenaest, T. Delzant. A. Papadopoutos. Géoméirie ef théorie des groupes. X, 165 pages. 1990. Vol. 1442: L. Accardi, M. von Waldentels (Eds.), Quan- tum Probability and Applications V. Proceedings, 1988, VL, 413 pages. 1990, Vol. 1443: K.H. Dovermann, R. Schultz, Equivariant Surgery Theories and Their Periodicity Properties. VI,227 pages. 1990. Vol. 1484: H. Korezliogtu, A. Ustunel (Eds.), Stochastic Analysis and Related Topics Vi. Proceedings, 1988, V, 268 pages, 1990. Vol. 1445: F. Schulz, Regularity Theory for Quasilinear Elliptic Systems and ~ Monge Ampére Equations in Two Dimensions. XV, 123 pages. 1990. Vol. 1446: Methods of Nonconvex Analysis, Seminar, 1989. Editor: A. Cellina. V, 206 pages. 1990. Vol. 1447; JG, Labesse, J. Schwermer (Eds), Cohomology of Arithmetic Groups and Automorphic Forms. Proceedings. 1989. V, 358 pages. 1990, Vol, 1448: S.K. Jain, S.R. Lopez-Permouth (Eds,), Non- Commutative Ring Theory. Proceedings, 1989. V, 166 pages. 1990. Vol. 1449: W. Odyniec, G. Lewicki, Minimal Projections in Banach Spaces. Vill, 168 pages. 1990. Vol. 1450: H. Fujita, T. tkebe, $.T. Kuroda (Eds.), Functional-Analytic Methods for Paria Differential Equa- tions, Proceedings, 1989. VII, 252 pages. 1990. Vol. 1451: L, Alvarez-Gaumé, E. Arbaretlo, C. De Concini, NJ. Hitchin, Global Geometry and Mathematical Physics. Montecatini Terme 1988. Seminar. Editors: M. Francaviglia, F. Gherardelli IX, 197 pages, 1990. Vol, 1452: E. Hawks, RF. Tily (Bds,), Number-Theoretic Analysis. Seminar, 1988-89. V, 220 pages. 1999. Vol. 1453: Yu.G. Borisovich, Yu.E. Gliklikh (Eds), Globat ‘Analysis Studies and Applications LV. V, 320 pages. 1990.Vol, 1454: F. Baldassari, S. Bosch, B, Dwork (Eds.), p> adic Analysis. Proceedings, 1989. V, 382 pages. 1990, Vol. 1455: J.-P. Frangoise, R. Roussari (Eds. Bifurcations (of Planae Veetor Fields. Proceedings, 1989. V1, 396 pages. 1990, Vol, 1486: LG. Kovles (Ed,), Groups ~ Canberra 1989, Proceedings. Xl, 198 pages. 1990. Vol. 1457: 0. Axelsson, L.Yu. Kolotilina (Eds.) Preconditioned Conjugate Gradient Methods, Proceedings. 1989. V, 196 pages. 1990. Vol. 1458: R, Schaaf, Global Solution Branches of Two Point Boundary Value Problems. XIX, 141 pages. 1990. Vol. 1459: D. Tiba, Optimal Control of Nonsmooth Distributed Parameter Systems. VIL, [59 pages. 1990. Vol. 1460: G. Toseani, V. Boffi, S, Rionero (Eds.). Mathematical Aspects of Fluid Plasma Dynamics, Proceed- ings, 1988. V, 221 pages. 1991, Vol, 1461: R. Gorenflo, S. Vessella, Abel Integral Equations. Vil, 215 pages. 1991, Vol. 1462: D. Mond, J. Montaldi (Eds), Singularity Theory and its Applications. Warwick 1989, Par, VII, 405 pages. 1991 Vol. £463: R, Robers, I, Stewart (Eds), Singularity Theory and its Applications. Warwick 1989, Part Il. VIII, 322 pages. 1991 Vol. 1464: D. L. Burkholder, E. Pardoux, A. Sznitman, Ecole d°Et¢ de Probabilités de Saint- Flour XIX-1989 Editor; PL, Henneguin. V1, 286 pages, 1991 Vol. 1465: G, David, Wavelets and Singular Integrals Curves and Surfaces. X. 107 pages. 1991 Vol. 1466: W. Banaszezyk, Additive Subgroups of ‘Topological Vector Spaces. Vil, 178 pages. 1991 Vol: 467: W. M. Schmid, Diophantine Approximations and Diophantine Equations, VII, 217 pages. 1991 Vol. 1468: J. Noguchi, T. Ohsawa (Eds.), Prospects in Complex Geometry. Proceedings, 1989. Vil, 421 pages. 1991 Vol, 1469: J. Lindensteauss, V. D, Milman (Eds.), Geometric Aspects of Functional Analysis, Seminar 1989- 90. X1, 191 pages. 1991 Vol. 1470; E. Odell, H. Rosenthal (Eds.}, Functional Analysis, Proceedings, 1987-89. Vil, 199 pages. 1991, Vol. 1471: A. A. Panchishkin, Non-Archimedean Lx Functions of Siegel and Hilbert Modular Forms, VII, 157 pages, 1991 Vol. 1472: T. T. Nielsen, Bose Algebras: The Complex and Real Wave Representations. V, 132 pages. 1991 Vol. 1473: Y. Hino, S. Murakami, T. Naito, Functional Dif= ferential Equations with infinite Delay. X,317 pages. 1991 Vol. 1474: S, Jackowski, B. Oliver, K. Pawalowski (Eds), Algebraic Topology, Poznari 1989. Proceedings. VIII, 397 pages, 1991 Vol. 1475: S. Busenberg. M. Maneli (Eds). Delay Dif= ferential Equations and Dynamical Systems. Proceedings, 1990, Vill, 249 pages. 1991, Vol. 1476: M, Bekkali, Topics in Set Theory. Vil. 120 pages. 1991 Vol. 1477: R. Jajte, Strong Limit Theorems in Noncommutative L,-Spaces. X, 113 pages. 1991 Vol, 1478: M.-P. Malliavin (Ed.). Topics in Invariant ‘Theory. Seminar 1989-1990. V1, 272 pages. 1991 Vol. 1479: S. Bloch, 1. Doigachev, W. Fulton (Eds.), Algebraic Geometry. Proceedings, 1989. VII, 300 pages 1991 Vol. 1480: F. Dumortier, R, Roussarie, J. Sotomayor, H. Zoladek, Bifurcations of Planar Vector Fields: Nilpotent Singularities and Abelian Integrals. Vill, 226 pages. 1991. Vol. 1481: D, Ferus, U, Pinkall, U. Simon, B. Wegner (Eds), Global Differential Geometry and Global Analysis, Proceedings. 1991. VII, 283 pages. 1991 Vol. 1482: J. Chabrowski, The Dirichlet Problem with L2- Boundary Data for Elliptic Linear Equations. V1, 173 pages, 1991 Vol. 1483: E, Reithmeier, Periodic Solutions of Nonlinear Dynamical Systems, Vi, 171 pages. 1991 Vol. 1484: H. Deifs, Homology of Locally Semialgebraic Spaces. IX. 136 pages. 1991. Vol. 1485:1. Anéima, P_A. Meyer, M. Yor (Eds.), Séminaire de Probabilités XXV. VIM, 440 pages. 1991 Vol. 1486: L, Ammold, H. Crauel, JP, Eckmann (Eds.), Lyapunov Exponents, Proceedings, 1990. Vill, 365 pages. 991 Vol. 1487: E. Freitag, Singular Modular Forms and Theta Relations. Vi, 172 pages. 1991 Vol. 1488: A. Carboni, M,C. Pedicchio, G. Rosolini (Eds), Category Theory. Proceedings, 1990. Vil. 494 pages. 1991 Vol. 1489: A. Micike, Hamiltonian and Lagrangian Flows fon Center Manifolds. X, 140 pages. 1991 ‘Vol. 1490: K. Metsch, Linear Spaces with Few Lines. Xiil 196 pages, 1991 Vol. 1491:B, Livis-Puebla,J-L. Loday, H. Gillet, C, Soul, V. Snaith, Higher Algebraic K-Theory: an overview. 1X, 164 pages. 1992. Vol. 1492: K. R. Wicks, Fractals and Hyperspaces. VILL 168 pages. 1991 Vol. 1493: E, Benoit (Ed.), Dynamic Bifurcations. Proceedings. Luminy 1990. VII, 219 pages. 1991 Vol. 1494: M.-T. Cheng. X.-W. Zhou, D.-G. Deng (Eds.), Harmonic Analysis, Proceedings, 1988. 1X, 226 pages. 1991 Vol. 1495: J. M. Bony, G. Grubb, L. Hormander, H. Komatsu, J. Sjostrand, Microlocal Analysis and Applications. Montecatini Terme, 1989. Editors: L. Cattabriga, L. Rodino. VII, 349 pages. 1991 Vol. 1496; C. Foias, B. Francis, J. W. Helton, H, Kwakemaak, J. B. Pearson, H,,-Control Theory. Como, 1990, Editors: E. Mosca, Pandolfi. VII, 336 pages, 1991, Vol. 1497: G, T. Herman, A, K, Louis, F. Natterer (Eds), Mathematical Methods in Tomography. Proceedings 1990. X, 268 pages. 1991. Vol. 1498: R. Lang, Spectral Theory of Random Schrodinger Operators. X, 125 pages. 1991 Vol. 1499: K. Taira, Boundary Value Problems and Markov Processes. IX, 132 pages. 1991 Vol. 1500: J.-P. Serre, Lie Algebras and Lie Groups. Vil 168 pages. 1992, ‘Vol. 1501: A. De Masi, E. Presutti, Mathematical Methods for Hydrodynamic Limits, IX, 196 pages. 1991