ebook img

THE AMPLE CONE OF THE KONTSEVICH MODULI SPACE 1. Introduction Positive-dimensional ... PDF

14 Pages·2012·0.2 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview THE AMPLE CONE OF THE KONTSEVICH MODULI SPACE 1. Introduction Positive-dimensional ...

THE AMPLE CONE OF THE KONTSEVICH MODULI SPACE IZZETCOSKUN,JOEHARRIS,ANDJASONSTARR Abstract. We produceample,respectivelyNEF,eventually free,divisorsin the Kontsevich space M0;n(Pr;d) of n-pointed, genus 0, stable maps to Pr, givensuchdivisorsinM0;n+d. Weprovethisproducesallample,respectively NEF,eventuallyfree,divisorsinM0;n(Pr;d). Asaconsequence,weconstruct a contraction of the boundary [bkd==12c(cid:1)k;d(cid:0)k in M0;0(Pr;d) analogous to a contraction ofthe boundary[bkn==32c(cid:1)~k;n(cid:0)k inM0;n (cid:12)rstconstructed byKeel andMc Kernan. 1. Introduction Positive-dimensionalfamiliesof varietiesspecialize {non-generalvarietiesinthe family exhibit special properties. Given a parameter space, the subset parametriz- ing varieties with a special property is typically closed. Which special properties occurincodimension1,respectivelyforevery1-parameterfamilyofvarieties? More precisely, when is the associated closed subset an e(cid:11)ective divisor, respectively an ample divisor? These questions, among others, motivate the study of e(cid:11)ective and ample divisors in parameter spaces of varieties. TheparameterspacewestudyistheKontsevichmodulispaceofn-pointed,genus 0, stable maps to projective space, denoted M (Pr;d). Here we study the ample 0;n cone, and more generally the NEF and eventually free cones. In a second article [CHS], using signi(cid:12)cantly di(cid:11)erent methods and under additional hypotheses, we study the e(cid:11)ective cone. Our goal is to study families of curves in a general target X. Fortunately, this largely reduces to the study for Pr: As the Kontsevich space is functorial in the target, for every morphism X ! Pr, NEF and base-point-free divisors on M (Pr;d)giveNEFandbase-point-freedivisorsonM (X;(cid:12))(thisfunctoriality 0;n 0;n isoneofmanyadvantagesofthe KontsevichspaceovertheHilbertschemeandthe Chow variety). Here is our main result. Theorem 1.1. Let r and d be positive integers, n a nonnegative integer such that n+d(cid:21)3. There is an injective linear map, v :Pic(M )Sd !Pic(M (Pr;d)) : 0;n+d Q 0;n Q 2000 Mathematics SubjectClassi(cid:12)cation. Primary:14D20,14E99,14H10. During the preparation ofthis article the second author was partially supported by the NSF grant DMS-0200659. The third author was partiallysupported by the NSF grant DMS-0353692 andaSloanResearchFellowship. 1 The NEF cone of M (Pr;d), respectively, the base-point-free cone, is the product 0;n of the cone generated by H;T;L ;:::;L and the image under v of the NEF cone 1 n of M ==S , respectively, the base-point-free cone. 0;n+d d The action of S on M permutes the last d marked points. The map v d 0;n+d generatesNEFandbase-point-freedivisorsonthe KontsevichspacefromNEFand base-point-freedivisorson M ==S . In particular,it generatescontractionsof 0;n+d d the Kontsevich space from contractions of M ==S . 0;n+d d Theorem 1.2. For every integer r (cid:21)1 and d(cid:21)2, there is a contraction, cont:M (Pr;d)!Y; 0;0 restricting to an open immersion on the interior M (Pr;d) and whose restriction 0;0 to the boundary divisor (cid:1)k;d(cid:0)k (cid:24)= M0;1(Pr;k)(cid:2)Pr M0;1(Pr;d(cid:0)k) factors through the projection to M (Pr;d(cid:0)k) for each 1 (cid:20) k (cid:20) bd=2c. The following divisor is 0;1 the pull-back of an ample divisor on Y, bd=2c D =T + k(k(cid:0)1)(cid:1) : r;d k;d(cid:0)k kX=2 SomeconnectionbetweentheampleconeoftheKontsevichspaceandtheample cone of M is natural, and certainly not surprisingto experts. A similarconnec- 0;n tion between the Fulton-MacPherson space and M was proved in [Che]. The 0;n primaryimportanceofTheorem1.1istheprecise,simpledescriptionofv: withone exception, it maps each boundary divisor of M to the corresponding bound- 0;n+d ary divisor of the Kontsevich space. This is used to construct the contraction in Theorem 1.2, which is analogous to the \democratic" contraction of the boundary of M (cid:12)rst constructed in an unpublished note of Keel and Mc Kernan. 0;n Recently we were informed of di(cid:11)erent constructions of the contraction of The- orem 1.2 in [Par] and by Anca Musta(cid:24)ta(cid:20) and Andrei Musta(cid:24)ta(cid:20). One advantage of our proof is that it uses only the existence of the map v, which is itself a formal consequence of the de(cid:12)nition of the Kontsevich space. The proof of Theorem 1.2 alsogivesanew,veryshortconstructionofKeel-Mc Kernan’scontractionofM . 0;n Acknowledgments: We would like to thank B. Hassett and A. J. de Jong for suggesting helpful references and for useful discussions. 2. Statement of results TheKontsevichmoduli spaceM (Pr;d) compacti(cid:12)esthe schemeparameteriz- 0;n ing smooth, rational curves of degree d in Pr. Precisely, it is the smooth, proper, Deligne-Mumford stack parameterizing families of data (C;(p ;:::;p );f) of, 1 n (i) a proper, connected, at-worst-nodal,genus 0 curve C, (ii) an ordered sequence p ;:::;p of distinct, smooth points of C, 1 n (iii) and a degree-d morphism f : C ! Pr satisfying the following stability condition: every irreducible component of C mapped to a point under f contains at least 3 special points, i.e., marked points p and nodes of C. i In [Pa] R. Pandharipande gives generators of the Kontsevich space: (1) the class H of the divisor of maps whose images intersect a (cid:12)xed codimen- sion two linear space in Pr (provided r >1 and d>0), 2 (2) the class Li of the pull-back ev(cid:3)i(OPr(1)), for 1 (cid:20) i (cid:20) n, associated to the ith evaluation morphism, ev (C;(p ;:::;p );f):=f(p ), i 1 n i (3) and the classes (cid:1) of the boundary divisors consisting of maps (A;dA);(B;dB) with reducibledomains. HereAtB isanyorderedpartitionof themarked points, andd and d arenon-negativeintegerssatisfyingd=d +d . If A B A B d =0 (respectively, if d =0), we demand #A(cid:21)2 (resp. #B (cid:21)2). A B The divisor classes H and L on M (Pr;d) are NEF and base-point-free. For i 0;n d (cid:21) 2, there is another NEF and base-point-free divisor class T, the tangency divisor: Fixing a hyperplane (cid:5) (cid:26) Pr, T is the class of the divisor parametrizing stable maps (C;p ;:::;p ;f) for which f(cid:0)1((cid:5)) is not simply d reduced, smooth 1 i points of C. In terms of Pandharipande’sgenerators,the class of T equals, bd=2c d(cid:0)1 k(d(cid:0)k) T = H+ 0 (cid:1) 1: d d (A;k);(B;d(cid:0)k) kX=0 AX;B @ A Finally, the map v from Theorem 1.1 is described in Section 3. Together, all nonnegative-linearcombinationsofthesedivisorsgiveaconeinPic(M (Pr;d)) . 0;n Q Weusethe methodof test families toprovethisisthe entireconeof NEFdivisors, respectively, eventually free divisors. In other words, we (cid:12)nd morphisms from test varieties to M (Pr;d) . Since every NEF divisor, resp. eventually free divisor, 0;n Q pulls back to such a NEF divisor, resp. eventually free divisor, this constrains the NEF and eventually free divisors among all divisors. By producing su(cid:14)ciently many test families, we prove every NEF, resp. eventually free divisor, is in our cone. Hypothesis 2.1. For the rest of the paper assume that the triple (n;r;d) in M (Pr;d) satis(cid:12)es r (cid:21)1; d(cid:21)1 and n+d(cid:21)3: 0;n Pr f * p C L L L L Figure 1. The morphism (cid:11). The morphism (cid:11). There is a 1-morphism (cid:11) : M (cid:2)Pr(cid:0)1 ! M (Pr;d) 0;n+d 0;n de(cid:12)ned as follows. Fix a point p 2 Pr and a line L (cid:26) Pr containing p. To every curveC inM attachacopyofLateachofthelastdmarkedpointsanddenote 0;n+d the resulting curve by C0. Consider the morphism f : C0 ! Pr that contracts C to p and maps the d rational tails isomorphically to L (see Figure 1). Since the space of lines in Pr passing through the point p is parameterized by Pr(cid:0)1, there is an induced 1-morphism (cid:11):M (cid:2)Pr(cid:0)1 !M (Pr;d): 0;n+d 0;n Since(cid:11)isinvariantforthe actionofS permutingthe lastdmarkedpoints,the d pull-back map determines a homomorphism (cid:11)(cid:3) =((cid:11)(cid:3);(cid:11)(cid:3)):Pic(M (Pr;d))!Pic(M )Sd (cid:2)Pic(Pr(cid:0)1): 1 2 0;n 0;n+d 3 We will denote the two projections of (cid:11)(cid:3) by (cid:11)(cid:3) and (cid:11)(cid:3). 1 2 C p i L * slide p along L i Figure 2. The morphism (cid:12) . i The morphisms (cid:12) . For each 1 (cid:20) i (cid:20) n, there is a 1-morphism (cid:12) : P1 ! i i M (Pr;d) de(cid:12)ned as follows. Fix a degree-(d (cid:0) 1), (n (cid:0) 1)-pointed curve C 0;n containingallexcepttheith markedpoint. AtageneralpointofC,attachalineL. The resulting degree-d, reducible curve will be the domain of our map. The (cid:12)nal, ith markedpointisinL. Varyingp inLgivesa1-morphism(cid:12) :P1 !M (Pr;d) i i 0;n (seeFigure2). Thisde(cid:12)nitionhastobeslightlymodi(cid:12)edinthecases(n;d)=(1;1) or (2;1). When (n;d)=(1;1), we assume that the line L with the varying marked point p constitutes the entire stable map. When (n;d) = (2;1), we assume that i the map has L as the only component. One marked point is allowed to vary on L and the remaining marked point is held (cid:12)xed at a point p2L. slide attachment point Pr C C L f L L Figure 3. The morphism (cid:13). The morphism (cid:13). If d (cid:21) 2, there is a 1-morphism (cid:13) : P1 ! M (Pr;d) 0;n de(cid:12)ned as follows. Take two copies of a (cid:12)xed line L attached to each other at a variable point. Fix a point p in the second copy of L. Let C be a smooth, degree- (d(cid:0)2), genus 0, (n+1)-pointed stable map to Pr whose (n+1)-st point maps to p. Attach this to the second copy of L at p. Altogether, this gives a degree-d, n-pointed, genus 0 stable maps with three irreducible components. The n marked pointsarethe (cid:12)rstnmarkedpointsof C. Theonlyvaryingaspectof thisfamily of stablemapsistheattachmentpointofthetwocopiesofL. Varyingtheattachment point in L(cid:24)=P1 givesa stablemap parameterizedbyP1, hence thereis an induced 1-morphism (cid:13) :P1 !M (Pr;d) (see Figure 3). When (n;d) =(1;2), we modify 0;n the de(cid:12)nition by assuming that the map consists only of the two copies of the line L and the marked point is held (cid:12)xed at the point p on the second copy of L. Notation 2.2. If d(cid:21)2, denote by P the Abelian group r;n;d P :=Pic(M )Sd (cid:2)Pic(Pr(cid:0)1)(cid:2)Pic(P1)n(cid:2)Pic(P1): r;n;d 0;n+d Denote by u=u :Pic(M (Pr;d))!P the pull-back map r;n;d 0;n r;n;d u =((cid:11)(cid:3);((cid:12)(cid:3);:::;(cid:12)(cid:3));(cid:13)(cid:3)): r;n;d 1 n 4 If d=1, denote by P the Abelian group r;n;1 P :=Pic(M )Sd (cid:2)Pic(Pr(cid:0)1)(cid:2)Pic(P1)n r;n;1 0;n+d and denote by u=u :Pic(M (Pr;1))!P the pull-back map r;n;1 0;n r;n;1 u =((cid:11)(cid:3);((cid:12) ;(cid:3);:::;(cid:12)(cid:3))) r;n;1 1 n Theorem 1.1 is equivalent to the following. Theorem 2.3. The map u (cid:10)Q:Pic(M (Pr;d)) !P (cid:10)Q is an isomor- r;n;d 0;n Q r;n;d phism. The image under u (cid:10)Q of the ample cone, resp. NEF, eventually free r;n;d cone of M (Pr;d) equals the product of the ample cones, resp. NEF, eventually 0;n free cones of Pic(M0;n+d)Sd, Pic(Pr(cid:0)1), and the factors Pic(P1). This is equivalent to Theorem 1.1 because the linear map u is simply the r;n;d inverse of the product of the linear map v and the maps Q ! Pic(M (Pr;d)) 0;n Q associated to each generator H, T and L ;:::;L . 1 n Notation 2.4. Denote by n the set f1;:::;ng. Denote by 4 = 4 the set of n;d 4-tuples ((A;d );(B;d )) of an ordered partition AtB of n and an ordered pair A B of nonnegative integers (d ;d ) such that d +d = d and #A (cid:21) 2 (#B (cid:21) 2) if A B A B d = 0 (d = 0, respectively). Denote by 40 the subset of 4 of data such that A B #A+d (cid:21)2 and #B+d (cid:21)2. A B Recall that the groupPic(M ) is generatedby boundary divisors (cid:1)~ , where 0;n A;B AtB is an ordered partition of n with #A (cid:21)2 and #B (cid:21)2. Let (cid:1)~ denote k;n(cid:0)k the sum of the boundary divisors (cid:1)~ , where the sum runs over pairs (A;B) A;B (A;B) suchthat #A=k and#B =Pn(cid:0)k. The groupPic(M0;n+d)Sd is generated by boundary divisors (cid:1)~ , where ((A;d );(B;d )) 2 (cid:1)0. The divisor (A;dA);(B;dB) A B (cid:1)~(A;dA);(B;dB)denotestheSd-invariantsumofboundarydivisors (A0;B0)(cid:1)~(A;A0);(B;B0), where the sum runs over pairs (A0;B0) such that A0 tB0 is a paPrtition of the last d points and #A0 =d and #B0 =d . A B To apply Theorem 2.3, we need to express the images of the standard gener- ators of Pic(M0;n(Pr;d)) in terms of the standard generators for Pic(M0;n+d)Sd, Pic(Pr(cid:0)1) and Pic(P1) factors. Proposition 2.5. (i) Assume d(cid:21)2 so that (cid:13) is de(cid:12)ned. Then (cid:13)(cid:3)T =O (2); (cid:13)(cid:3)H=0; (cid:13)(cid:3)L =0; for 1(cid:20)i(cid:20)n: P1 i The pull-back (cid:13)(cid:3)(cid:1) = 0 unless (#A;d ) or (#B;d ) is equal to (0;1) (A;dA);(B;dB) A B or (0;2). Moreover, if (n;d)6=(0;3), (cid:13)(cid:3)(cid:1) =O (4) and (cid:13)(cid:3)(cid:1) =O ((cid:0)1): (;;1);(n;d(cid:0)1) P1 (;;2);(n;d(cid:0)2) P1 If (n;d)=(0;3), then (cid:13)(cid:3)(cid:1) =O (3). (;;1);(n;d(cid:0)1) P1 (ii) Assume n(cid:21)1 so that (cid:12) ;:::;(cid:12) are de(cid:12)ned. Then 1 n (cid:12)(cid:3)H=0; (cid:12)(cid:3)L =O (1); (cid:12)(cid:3)L =0 if j 6=i; and (cid:12)(cid:3)T =0: i i i P1 i j i For every 1 (cid:20) i (cid:20) n, the pull-back (cid:12)(cid:3)(cid:1) equals O (1) if (n;d) 6= (1;2), i (;;1);(n;d(cid:0)1) P1 and equals O (2) if (n;d) = (1;2). If (n;d) 6= (1;2), then (cid:12)(cid:3)(cid:1) P1 i (fig;1);(figc;d(cid:0)1) equals O ((cid:0)1). And (cid:12)(cid:3)(cid:1) equals 0 if neither (A;d ) nor (B;d ) equal P1 i (A;dA);(B;dB) A B (;;1) or (fig;1). (iii) (cid:11)(cid:3)H=(0;OPr(cid:0)1(d)); (cid:11)(cid:3)Li =0; for 1(cid:20)i(cid:20)n; (cid:11)(cid:3)T =0: 5 Divisors in M (Pr;d) (cid:11)(cid:3) (cid:11)(cid:3) (cid:12)(cid:3) (cid:13)(cid:3) 0;n 1 2 i T 0 0 0 O (2) P1 H 0 OPr(cid:0)1(d) 0 0 L 0 0 O (1) 0 i P1 L 0 0 0 0 j6=i (cid:1)(;;1);(n;d(cid:0)1) c OPr(cid:0)1((cid:0)d) OP1((cid:0)1) OP1(4) (cid:1) (cid:1)~ 0 0 O ((cid:0)1) (;;2);(n;d(cid:0)2) (;;2);(n;d(cid:0)2) P1 (cid:1) (cid:1)~ 0 O ((cid:0)1) 0 (fig;1);(figc;d(cid:0)1) (fig;1);(figc;d(cid:0)1) P1 (cid:1) (cid:1)~ 0 0 0 (A;dA);(B;dB) (A;dA);(B;dB) all others Figure 4. The pull-backs of the standard generators If #A+d ;#B+d (cid:21) 2, then (cid:11)(cid:3)(cid:1) equals (cid:1)~ . The pull- A B (A;dA);(B;dB) (A;dA);(B;dB) back (cid:11)(cid:3)(cid:1)(;;1);(n;d(cid:0)1) equals (c;OPr(cid:0)1((cid:0)d)), where c is the class (cid:0)1 c= d (d +#B)(d +#B(cid:0)1)(cid:1)~ : (n+d(cid:0)1)(n+d(cid:0)2) A B B (A;dA);(B;dB) ((A;dAX);(B;dB)) 2(cid:1)0 Proof. Items (i) and (ii) follow fromLemma 3.5 and Lemma 4.1. Item (iii), except for the computation of c, is straightforward. The class c equals (cid:0) d . To i=1 n+i rewrite this as above, use [Pa, Lemma 2.2.1] (cf. also [dJS, LemmaP6.10]). (cid:3) With the exceptions of (n;d) =(0;3), (1;2), and (1;3), Proposition 2.5 is sum- marized by Figure 4. The phrase \all others" means, all pairs ((A;d );(B;d )) A B such that neither ((A;d );(B;d )) nor ((B;d );(A;d )) already occur in the ta- A B B A ble. The lines (cid:13)(cid:3) and (cid:1) apply only if d (cid:21) 2. The lines L , L and (;;2);(n;d(cid:0)2) i j (cid:1) only apply if n(cid:21)1. (fig;1);(figc;d(cid:0)1) In [Kaw], Kawamata associated an e(cid:11)ective, NEF Q-Cartier divisor L on M 0;n to every n-tuple of rational numbers, (d ;:::;d ) satisfying 0 < d (cid:20) 1 and d + 1 n i 1 (cid:1)(cid:1)(cid:1)+d =2. In an unpublished note Keel and Mc Kernan proved the following. n Theorem 2.6 (Keel-Mc Kernan). The Q-Cartier divisor L is eventually free. In particular, when d =(cid:1)(cid:1)(cid:1)=d =2=n, the divisor class of L equals (1=n(n(cid:0) 1 n 1))D , where n bn=2c D = k(k(cid:0)1)(cid:1)~ : n k;n(cid:0)k kX=2 6 This is the divisor class giving the \democratic" contraction of the boundary of M , cf. [Has, x2.1.2]. One application of Theorem 2.3 is the construction of the 0;n analogous contraction in Theorem 1.2 as well as a new, short construction of the democratic contraction. Theorem 2.7. For every integer n(cid:21)4, there is a contraction cont:M !Y 0;n restricting to an open immersion on the interior M and whose restriction to 0;n the boundary divisor (cid:1) = M (cid:2)M factors through projection to k;n(cid:0)k 0;k+1 0;n+1(cid:0)k M for each 3 (cid:20) k (cid:20) bn=2c. The divisor D is the pull-back of an ample 0;n+1(cid:0)k n divisor on Y. It follows easily that for every rational number b satisfying 2 2 2 <b< resp. b= ; (n(cid:0)1) bn=2c bn=2c settingB =b(cont) ((cid:1)~ ),K +B isanampleQ-Cartierdivisor,and(Y;B)is (cid:3) 2;n(cid:0)2 Y Kawamata log terminal, resp. log canonical (for this one only needs the existence of the contraction and the formula for L). 3. The splitting homomorphism In this section we de(cid:12)ne a map v : Pic(M0;n+d)Sd ! Pic(M0;n(Pr;d))(cid:10)Q that mapstheNEFdivisorsinPic(M0;n+d)Sd toNEFdivisorsinM0;n(Pr;d). Themap v givesasplittingof the map(cid:11)(cid:3) de(cid:12)nedinthe introductionandisessentialforthe 1 proof of Theorem 2.3. Let (cid:5)(cid:26)Pr be a hyperplane not containing the point p used to de(cid:12)ne the mor- phisms (cid:11) and (cid:13). Assume that the degreed(cid:0)1 curveused to de(cid:12)ne the morphisms (cid:12) is not tangent to (cid:5), and none of the marked points on this curve are contained i in (cid:5). Finally, assume that the degreed(cid:0)2 curveused to de(cid:12)ne the morphism (cid:13) is not tangent to (cid:5) and none of the marked points are contained in (cid:5). Denote by M (Pr;d) the open substack of M (Pr;d) parameterizing 0;n+d 0;n+d stable maps with irreducible domain. Let ev :M (Pr;d)!(Pr)d n+1;:::;n+d 0;n+d be the evaluation morphism associated to the last d marked point. Denote by M (Pr;d) the inverse image of (cid:5)d and by M (Pr;d) the closure of 0;n+d (cid:5) 0;n+d (cid:5) M (Pr;d) in M (Pr;d). 0;n+d (cid:5) 0;n+d M (Pr;d) is S -invariant under the action of S on M (Pr;d) per- 0;n+d (cid:5) d d 0;n+d muting the last d marked points. Denote by (cid:25) :M (Pr;d)!M (Pr;d) 0;n+d 0;n the forgetful 1-morphism that forgets the last d marked points and stabilizes the resulting family of prestable maps. This is S -invariant. Denote by d (cid:26):M (Pr;d)!M 0;n+d 0;n+d the1-morphismthatstabilizestheuniversalfamilyofmarkedprestablecurvesover M (Pr;d). This is S -equivariant. 0;n+d d 7 Lemma 3.1. The 1-morphism (cid:25) :M (Pr;d) !M (Pr;d) is (cid:19)etale. Denot- 0;n+d (cid:5) 0;n ing the image by O , the morphism (cid:25) :M (Pr;d) !O is an S -torsor. (cid:5) 0;n+d (cid:5) (cid:5) d Proof. Let (C;(p ;:::;p ;q ;:::;q );f) be a stable map in M (Pr;d) . Then 1 n 1 d 0;n+d (cid:5) (C;(p ;:::;p );f) satis(cid:12)es, 1 n (i) C is irreducible, (ii) f(cid:0)1((cid:5)) is a reduced Cartier divisor, and (iii) none of the marked points p is contained in f(cid:0)1((cid:5)). i Conversely, for every stable map (C;(p ;:::;p );f) satisfying (i){(iii), for every 1 n labeling of f(cid:0)1((cid:5)) as q ;:::;q , (C;(p ;:::;p ;q ;:::;q );f) is a stable map in 1 d 1 n 1 d M (Pr;d) . Thus O is the open substack of stable maps satisfying (i){(iii) 0;n+d (cid:5) (cid:5) andM (Pr;d) istheS -torsoroverO parameterizinglabelingsofthe(cid:12)bers 0;n+d (cid:5) d (cid:5) of f(cid:0)1((cid:5)). (cid:3) Denote by q : M ! M =S the geometric quotient. The composition 0;n+d 0;n+d d q(cid:14)(cid:26):M (Pr;d) !M =S isS -equivariant. BecauseM (Pr;d) is 0;n+d (cid:5) 0;n+d d d 0;n+d (cid:5) an S -torsor over O , there is a unique 1-morphism (cid:30)0 : O ! M =S such d (cid:5) (cid:5) (cid:5) 0;n+d d that (cid:30)0(cid:14)(cid:25) =q(cid:14)(cid:26). De(cid:12)nition 3.2. De(cid:12)ne U to be the maximal open substack of M (Pr;d) over (cid:5) 0;n which (cid:30)0 extends to a 1-morphism, denoted (cid:5) (cid:30) :U !M =S : (cid:5) (cid:5) 0;n+d d De(cid:12)neI tobethenormalizationoftheclosureinM (Pr;d)(cid:2)M =S ofthe (cid:5) 0;n 0;n+d d image of the graph of (cid:30)0 , i.e., I is the normalization of the image of ((cid:25);q (cid:14)(cid:26)). (cid:5) (cid:5) De(cid:12)ne I to be the normalization of the image of ((cid:25);(cid:26)) in M (Pr;d)(cid:2)M . (cid:5) 0;n 0;n+d Finally, de(cid:12)ne U to be the inverse image of U in I . e (cid:5) (cid:5) (cid:5) e e There is a pull-back map of S -invariant invertible sheaves, d (cid:26)(cid:3) :Pic(M )Sd !Pic(I )Sd; 0;n+d (cid:5) whichfurtherrestrictstoPic(U(cid:5))Sd. After(cid:19)etalebasee-changefromU(cid:5) toascheme, the morphism U ! U is the geometric quotient of U by the action of S . (cid:5) (cid:5) e (cid:5) d Therefore the peull-back map Pic(U(cid:5)) ! Pic(U(cid:5))Sd is ane isomorphism after ten- soringwith Q; in fact, both the kerneland cokernelareannihilated by d!. Because e M =S is a proper scheme and because M (Pr;d) is separated and normal, 0;n+d d 0;n bythevaluativecriterionofpropernessthecomplementofU hascodimension(cid:21)2. (cid:5) The smoothness of M (Pr;d) and [Ha, Prop. 6.5(c)] imply that the restriction 0;n map Pic(M (Pr;d))!Pic(U ) is an isomorphism. 0;n (cid:5) De(cid:12)nition3.3. De(cid:12)nev :Pic(M0;n+d)Sd !Pic(M0;n(Pr;d))(cid:10)Qtobetheunique homomorphism commuting with (cid:26)(cid:3) via the isomorphisms above. The map v is independent of the choice of (cid:5), hence it sends NEF divisors to NEF divisors. S Lemma 3.4. For every base-point-free invertible sheaf L in Pic(M0;n+d) d, v(L) is base-point-free. In particular, for every ample invertible sheaf L, v(L) is NEF. Thus, by Kleiman’s criterion, for every NEF invertible sheaf L, v(L) is NEF. 8 Proof. For every [(C;(p ;:::;p );f)] in M (Pr;d), there exists a hyperplane (cid:5) 1 n 0;n satisfying the conditions above and such that f(cid:0)1((cid:5)) is a reduced Cartier divisor containing none of p ;:::;p . By Lemma 3.1, (C;(p ;:::;p );f) is contained in 1 n 1 n U . Since L isbase-point-free,thereexistsa divisorD in the linearsystemjLj not (cid:5) containing(cid:30) [(C;(p ;:::;p );f)]. Bytheproofof[Ha,Prop. 6.5(c)],theclosureof (cid:5) 1 n (cid:30)(cid:0)1(D) is in the linear system jv(L)j; and it does not contain [(C;(p ;:::;p );f)]. (cid:5) 1 n (cid:3) Lemma 3.5. (i) The images of (cid:11), (cid:12) and (cid:13) are contained in U . i (cid:5) (ii) The morphisms (cid:30) (cid:14)(cid:12) and(cid:30) (cid:14)(cid:13) areconstantmorphisms. Therefore (cid:12)(cid:3)(cid:14)v (cid:5) i (cid:5) i and (cid:13)(cid:3)(cid:14)v are the zero homomorphism. (iii) The composition of (cid:11) with (cid:30) equals q(cid:14)pr . Therefore (cid:5) M 0;n+d (cid:11)(cid:3)(cid:14)v :Pic(M )Sd !Pic(M )Sd (cid:2)Pic(Pr(cid:0)1) 0;n+d 0;n+d is the homomorphism whose projection on the (cid:12)rst factor is the identity, and whose projection on the second factor is 0. Proof. (i): The image of (cid:11) is contained in O . Denote by q the intersection point (cid:5) of L and (cid:5). The image (cid:12) (L(cid:0)fqg) is contained in O . The stable map (cid:12) (q) sends the ith i (cid:5) i markedpointinto(cid:5). Uptolabelingthedpointsofthe inverseimageof(cid:5), thereis onlyone(n+d)-pointedstablemapinM (Pr;d) thatstabilizestothisstable 0;n+d (cid:5) map. Itisobtainedfrom(cid:12) (q)byremovingthe ith markedpointfromL,attaching i a contracted component C0 to L at q, containing the ith marked point and exactly one of the last d marked points, and labeling the d(cid:0)1 points in C \(cid:5) with the remaining d(cid:0)1 marked points. Similarly, (cid:13)(L(cid:0)fqg)is containedin O . The stablemap (cid:13)(q) has twocopies of (cid:5) L attached to each other at q. This appears to be a problem, because the inverse image of (cid:13)(q) in M (Pr;d) is 1-dimensional, isomorphic to M . The stable 0;n+d (cid:5) 0;4 maps have a contracted component C0 such that both copies of L are attached to C0 and 2 of the d new marked points are attached to C0. The remaining d(cid:0)2 marked points are the points of C \(cid:5). However, the map (cid:26) that stabilizes the resulting prestable (n+d)-marked curveis constant on this M . Indeed, the (cid:12)rst 0;4 copy of L has no marked points and is attached to C0 at one point. So the (cid:12)rst stepin stabilizationwill pruneL reducingthe numberof special pointson C0 from 4 to 3. (ii): In the family de(cid:12)ning (cid:12) , only the ith marked point on L varies. After i adding the d new marked points, L is a 3-pointed prestable curve; marked by the node p, the ith marked point, and the point q. For every base the only family of genus0, 3-pointed, stable curvesis the constant family. So upon stabilization, this family of genus 0, 3-pointed, stable curves becomes the constant family. Inthefamilyde(cid:12)ning(cid:13),onlythe attachmentpointofthe twocopiesofLvaries. The (cid:12)rst copy of L gives a family of 2-pointed, prestable curves; marked by q and the attachment point of the two copies of L. This is unstable. Upon stabilization, the(cid:12)rstcopyofLisprunedandthemarkedpointqonthe(cid:12)rstcopyisreplacedbya markedpointonthesecondcopyattheoriginalattachmentpoint. Nowthesecond copy of L gives a family of 3-pointed, prestable curves; marked by the attachment pointpofthesecondandthirdirreduciblecomponents,theattachmentpointofthe (cid:12)rst and second components, and q. For the same reason as in the last paragraph, this becomes a constant family. 9 (iii): Eachstable map in (cid:11)(M (cid:2)Pr(cid:0)1) is obtained froma genus0, (n+d)- 0;n+d pointed,stablecurve(C ;(p ;:::;p ;q ;:::;q ))andalineLinPr containingpby 0 1 n 1 d attaching for each 1(cid:20)i(cid:20)n, a copy C of L to C where p in C is identi(cid:12)ed with i 0 i q in C . The map to Pr contracts C to p, and sends each curve C to L via the i 0 0 identity morphism. Denoting by r the intersection point of L and (cid:5), the inverse image of (cid:5) consists of the d points r ;:::;r , where r is the copy of r in C . 1 d i i The component C is a 2-pointed, prestable curve: marked by the attachment i point p of C and by r . This is unstable. So, upon stabilization, C is pruned i i i and the marked point r is replaced by a marking on C at the point of attach- i 0 ment of C and C , namely q . Therefore, up to relabeling of the last d marked 0 i i points, the result is the genus 0, (n+d)-pointed, stable curve we started with, (C ;(p ;:::;p ;q ;:::;q )). (cid:3) 0 1 n 1 d 4. More divisors In the previous section we constructed a map (see De(cid:12)nition 3.3) v :Pic(M )Sd !Pic(M (Pr;d))(cid:10)Q: 0;n+d 0;n In this section we prove that the image of v, the divisor classes H, T and the tautological divisors L , generate Pic(M (Pr;d))(cid:10)Q. i 0;n The divisor class H , [Pa, Prop. 1] is the class of stable maps whose image (cid:3) intersects a (cid:12)xed codimension 2 linear space (cid:3) of Pr. This is de(cid:12)ned to be the empty divisor if r =1. For convenience, assume (cid:3) is contained in (cid:5) and does not intersect L or the curves C used to de(cid:12)ne (cid:12) and (cid:13). If n (cid:21) 1, the divisors L , i i;(cid:5) i=1;:::;n,[Pa,Prop. 1]arethepull-backbyev oftheCartierdivisor(cid:5). Ifd(cid:21)1, i the last divisoris T , [Pa, x2.3];the divisorof stable maps (C;(p ;:::;p );f) such (cid:5) 1 n thatf(cid:0)1((cid:5))isnotareduced,(cid:12)nitesetofdegreed. Thisisde(cid:12)nedtobetheempty divisorifd=1. In[Pa]PandharipandeprovesthatH ,L andT areirreducible (cid:3) i;(cid:5) (cid:5) Cartier divisors (when they are nonempty). Lemma 4.1. (i) The Cartier divisors T , L and H are NEF. (cid:5) i;(cid:5) (cid:3) (ii) The pull-backs (cid:11)(cid:3)(T ) and (cid:11)(cid:3)(L ) are zero. The pull-back (cid:11)(cid:3)(H ) equals (cid:5) i;(cid:5) (cid:3) (0;OPr(cid:0)1(d)) in Pic(M0;n+d)Sd (cid:2)Pic(Pr(cid:0)1); if r =1, then OPr(cid:0)1(1) is the trivial invertible sheaf. (iii) Assume n (cid:21) 1 so that (cid:12) is de(cid:12)ned for 1 (cid:20) i (cid:20) n. The pull-backs (cid:12)(cid:3)(T ) i i (cid:5) and (cid:12)(cid:3)(H ) are zero. For 1 (cid:20) j (cid:20) n di(cid:11)erent from i, (cid:12)(cid:3)(L ) is zero. i (cid:5) i j;(cid:5) Finally, (cid:12)(cid:3)(L ) is O (1). i i;(cid:5) P1 (iv) Assume d (cid:21) 2 so that (cid:13) is de(cid:12)ned. The pull-backs (cid:13)(cid:3)(H ) and (cid:13)(cid:3)(L ) (cid:3) i;(cid:5) are zero, and (cid:13)(cid:3)(T ) is O (2) in Pic(P1). (cid:5) P1 Proof. (i): By an argument similar to the one in Lemma 3.4, these divisors are base-point-free(whenevertheyarenon-empty). ThedivisorH isbigifr(cid:21)2,and (cid:3) T is big if d(cid:21)2. The divisors L are not big. (cid:5) i (ii): Bythe proof of Lemma 3.5, the imageof (cid:11) isin O , which isdisjoint from (cid:5) T . Also,ev (cid:14)(cid:11)istheconstantmorphismwith imagep,sothe inverseimageofL (cid:5) i i is empty. Finally, the pull-back of H equals the pull-back under the diagonal (cid:1) (cid:5) of the Cartier divisor d pr(cid:0)1((cid:3)) in (Pr(cid:0)1), where (cid:3) is considered as a divisor j=1 j in Pr(cid:0)1 via projectionPfrom p. 10

Description:
2000 Mathematics Subject Classification. The NEF cone of M0,n(Pr,d), respectively, the base-point-free cone, is the product of the cone generated
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.