The Cox Ring of $\bar{M}_{0,6}$
Ana-Maria Castravet

TL;DR
This paper proves that the Cox ring of the moduli space of stable rational curves with six marked points is finitely generated by specific boundary and hyperelliptic divisors, providing insight into its algebraic structure.
Contribution
It establishes finite generation of the Cox ring of ar{M}_{0,6} using divisors related to boundary components and hyperelliptic loci, offering a new proof of the effective cone structure.
Findings
Cox ring of ar{M}_{0,6} is finitely generated.
Generators include boundary divisors and pull-backs of hyperelliptic divisors.
Effective cone is generated by these divisors.
Abstract
We prove that the Cox ring of , the moduli space of stable, rational curves with 6 marked points, is finitely generated by sections corresponding to the boundary divisors and divisors which are pull-backs of the hyperelliptic locus in , the moduli space of stable, genus 3 curves, via morphisms that send a 6-pointed rational curve to a curve with 3 nodes by identifying 3 pairs of points. In particular, this gives a self-contained proof of Hassett and Tschinkel's result about the effective cone of being generated by the above mentioned divisors.
Click any figure to enlarge with its caption.
Figure 2Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAlgebraic Geometry and Number Theory · Advanced Algebra and Geometry · Cryptography and Residue Arithmetic
The Cox Ring of
Ana-Maria Castravet
Department of Mathematics, University of Massachusetts at Amherst, Amherst 01003
Abstract.
We prove that the Cox ring of the moduli space , of stable rational curves with marked points, is finitely generated by sections corresponding to the boundary divisors and divisors which are pull-backs of the hyperelliptic locus in via morphisms that send a -pointed rational curve to a curve with nodes by identifying pairs of points. In particular this gives a self-contained proof of Hassett and Tschinkel’s result about the effective cone of being generated by the above mentioned divisors.
Mathematics Subject Classification: Primary 14E30, 14H10, 14H51, 14M99
Keywords: Cox rings, Mori Dream Spaces, moduli spaces of stable curves
1. Introduction
A question of Fulton about the moduli space , of stable, -pointed, rational curves, is whether the cone of effective cycles of codimension in is generated by -strata, i.e., loci in corresponding to reducible curves with at least nodes. While the case when (i.e., the cone of effective curves) is completely open (and an affirmative result would imply, by results of Gibney, Keel and Morrison [GKM], the similar statement for the moduli space , of stable, -pointed, genus curves, thus determining the ample cone of ) the case when (i.e., the cone of effective divisors) was settled independently by Keel (unpublished, a refference to this may be found in [GKM], p.277) and Vermeire [V]: Fulton’s question has a negative answer when (and therefore for any ). Hassett and Tschinkel prove in [HT] that the Keel-Vermeire divisors (pull-backs of the locus of hyperelliptic curves in the moduli space , via morphisms sending a -pointed rational curve to a curve with nodes by identifying pairs of points) together with the -strata (the boundary) generate the cone of effective divisors in . The proof in [HT] is based on a computer check. In this paper we give a proof of Hassett and Tschinkel’s result, by proving a stronger statement: we show that the sections corresponding to the above divisors generate the Cox ring of .
Recall that if is a smooth projective variety with Picard group freely generated by divisors , then the Cox ring (or total coordinate ring) of is the multi-graded ring:
[TABLE]
The Cox ring being finitely generated has strong implications for the birational geometry of ( is a so-called Mori Dream Space): the effective cone and the nef cone are both polyhedral and there are finitely many small modifications of (i.e., varieties isomorphic in codimension one to ) such that any moving divisor on (i.e., a divisor whose base locus has codimension at least ) is nef on one of the varieties (see [HK] for the precise statements). It has been conjectured by Hu and Keel [HK] that any log-Fano variety has a finitely generated Cox ring. This has been recently proved in the groundbreaking paper [BCHM]. In [HK] Hu and Keel ask the following question:
Question 1.1**.**
Is the Cox ring of finitely generated?
As pointed out in [KM], the moduli space is log-Fano only for .
We answer Question 1.1 for by finding explicit generators. Our hope is that our method for finding generators, that proved to be useful in other circumstances (see [CT]), will eventually help answer Question 1.1 for larger as well.
Consider the Kapranov description of the moduli space . If are points in linearly general position in , then is the iterated blow-up of along and along the proper transforms of the lines for all . If is a general point in , there is a unique twisted cubic in that contains the points . Then is a -pointed rational curve, hence an element of . The point corresponds to the ’th marking (the so-called moving point).
Denote by the hyperplane class on and by and the exceptional divisors in corresponding to the points and the lines .
Notation 1.2**.**
Let be the class of the proper transform of the plane :
[TABLE]
If and , let be the boundary divisor in with general element a curve with two irreducible components with the partition of the markings given by . In the Kapranov description, the boundary divisors have the following classes:
[TABLE]
[TABLE]
Notation 1.3**.**
Let be the class of the proper transform of the unique quadric that contains all the points and the lines :
[TABLE]
The divisor classes are exactly the divisors considered by Keel and Vermeire: for example, if one considers the map given by identifying the pairs of points , then the class of the pull-back of the hyperelliptic locus in is computed in [HT] to be the class of . We call the divisors the Keel-Vermeire divisors. We prove the following:
Theorem 1.4**.**
The Cox ring of is generated by the sections (unique up to scaling) corresponding to the boundary divisors (i.e., and the exceptional divisors and ) and the Keel-Vermeire divisors .
The paper is divided as follows: Section 2 explains the strategy of proof, there are two main cases, the details of each are given in Section 3, respectively Section 4. The remaining sections contain auxiliary results needed in the proof. Section 5 contains proofs for some basic inequalities, while Section 6 contains some general multiplicity estimates needed for Case II. Section 7 contains the proof of Lemma 2.21 (needed for Case II) that states that the Cox ring of the blow-up of in seven (non-general) points is generated by sections corresponding to and curves. Section 8 gives necessary and sufficient conditions for a divisor on , the iterated blow-up of in four general points and lines through them, to have sections. Finally, in Section 9 we compute the restrictions of an arbitrary divisor to all the boundary divisors and Keel-Vermeire divisors on . Moreover, we derive some necessary conditions for these restrictions to be effective (an assumption in our main proof).
Acknowledgements. I thank Jenia Tevelev and Sean Keel for useful comments.
2. Plan of Proof
Consider an arbitrary divisor class on :
[TABLE]
In all that follows we assume .
Notation 2.1**.**
Let be the class of the proper transform in of a general line in . Let be the class of a general line in . Let be the class of the proper transform of a general cubic that passes through :
[TABLE]
The curves with class cover a dense set of ; hence, for any effective divisor .
Definition 2.2**.**
Let be the sections (unique up to scalar) corresponding to the divisors:
[TABLE]
Definition 2.3**.**
We call a section a distinguished section if
[TABLE]
where are non-negative integers.
To show that is generated by distinguished sections, we do an induction on . Note that we may assume that contains none of the divisors (2.1) in its base locus, i.e., equivalently, if for any of the divisors in (2.1), one has . To see this, note that if is an effective divisor, say is the zero locus of a section , then there is an exact sequence:
[TABLE]
If then any is of the form , where . If in addition is a divisor in (2.1) then we may replace with and with . (Clearly, if is generated by distinguished sections, then is too.) Therefore, we may assume:
Assumption 2.4**.**
for all divisors in (2.1).
Denote by the restriction to :
[TABLE]
To prove Theorem 1.4 it is enough to prove the following:
Main Claim. Let be a divisor on :
[TABLE]
such that and that satisfies Assumption 2.4. Up to a renumbering, we may assume that , for . If for all , then we may assume that the maximum of the ’s for all is attained for for some . Let . Then for any , there is , generated by distinguished sections, such that .
To see how the Main Claim implies Theorem 1.4, note that the kernel of the restriction is and the map is given by multiplication with . If , then , where . If is generated by distinguished sections, then to show that is generated by distinguished sections it is enough to show that is generated by distinguished sections. We may replace with , and continue the procedure. Since is always among the ’s, note that and is generated by distinguished sections by induction. The process has to stop as for any effective divisor . (In particular, note that decreases also when we substract from any of the divisors in (2.1) for which .)
Notation 2.5**.**
Given any divisor on we denote by the restriction of to . By (9.3) one has:
[TABLE]
Let