An arithmetic intersection for squares of elliptic curves with complex multiplication

Elisa Lorenzo García, Christophe Ritzenthaler, Fernando Rodríguez Villegas · 2024

Read An arithmetic intersection for squares of elliptic curves with complex multiplication by Elisa Lorenzo García, Christophe Ritzenthaler, Fernando Rodríguez Villegas (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-08738.txt.

Introduction

Elisa Lorenzo Garcia, Christophe Ritzenthaler, and Fernando Rodriguez Villegas study a class of genus two curves whose Jacobian splits as the square of a curve with extra symmetry. Their paper appeared in twenty twenty-four. The extra symmetry comes from a full ring inside a field with square roots of minus signs. The team asks which primes can show bad drop in a stable model of such a curve.

They prove any such prime must be no more than one fourth of the field mark. They also give a step-by-step method to list those primes, using a refined invariant from Kani. Worked cases for small field marks check the method against a past draft rule.

Catalog record

CREATORElisa Lorenzo García, Christophe Ritzenthaler, Fernando Rodríguez Villegas
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE102 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-08738

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

An arithmetic intersection for squares of elliptic curves with complex multiplication Elisa Lorenzo García, Christophe Ritzenthaler, Fernando Rodríguez Villegas Abstract . Let C C be a genus 2 2 curve with Jacobian isomorphic to the square of an elliptic curve with complex multiplication by a maximal order in an imaginary quadratic field of discriminant − d < 0 -d<0 . We show that if the stable model of C C has bad reduction over a prime p p then p ≤ d / 4 p\leq d/4 . We give an algorithm to compute the set of such p p using the so-called refined Humbert invariant introduced by Kani. Using results from Kudla-Rapoport and the formula of Gross-Keating, we compute for each of these primes p p its exponent in the discriminant of the stable model of C C . We conclude with some explicit computations for d < 100 d<100 and compare our results with an unpublished formula by the third author. 2010 Mathematics Subject Classification 11G20, 14Q05, 14D10, 14D20, 14H25 1. Introduction The initial motivation of this article was the following formula appearing in [ RV00 ] . The third author defined there a genus 2 curve (1.1) \displaystyle C \displaystyle:y^{2}=6^{-3}h_{1}(x)h_{2}(x) where \displaystyle h_{1}(x) \displaystyle=(7144\sqrt{-163}-151790)x^{3}+(129789\sqrt{-163}+1752597)x^{2}+ \displaystyle(-47481\sqrt{-163}+510153)x+(-1596\sqrt{-163}-37250) and h_{2}(x)=x^{3}\overline{h(-1/x)} where bar denotes the complex conjugation of the coefficients. Its discriminant factors nicely 1 1 1 As we will see, in order to avoid ad hoc factors like 2^{-12} , one needs to consider not the discriminant of C but an absolute normalized discriminant. See Example 6.6 for the computation in these terms. as (1.2) D=2^{-12}\disc(h_{1}h_{2})=(2\cdot 3^{2}\cdot 5\cdot 7\cdot 11\cdot 17\cdot 19\cdot 23)^{12}. The third author then observed that (1.3) \log D=-6\sum_{m\in\mathbb{Z}^{3}}\sum_{n|(163-Q(m))/4}\left(\frac{-163}{n}\right)\log n, where Q is the positive definite quadratic form with matrix \begin{pmatrix}24&4&6\\ 4&55&1\\ 6&1&83\end{pmatrix}. The curve C is of course special: its Jacobian is isomorphic to E^{2} , where E is an elliptic curve with complex multiplication by \mathbb{Z}[(1+\sqrt{-163})/2] . The factorization of D into a product of small primes is similar to the somehow ‘complementary’ case of simple abelian surfaces with CM studied in full details in [ GL07 , GL06 , GL12 , ABL+13 , LV15 , LV14 ] . But the formula ( 1.3 ) is closer in essence to the work of [ GZ85 ] or in a different setting [ BY06 ] . In the 90’, the third author actually developed a strategy that leads to such a formula for E^{2} where E has CM by a maximal order (see Section 6 for the explicit formula and comparison with the present results) but the work remained unfinished. However, the references and ideas it contains allowed us to shape the results in the present article and to propose an algorithm for the factorization of discriminants of genus 2 curves which Jacobian is the power of an elliptic curve with CM by a maximal order. Let us now rewind and give a more formal introduction. Let E be an elliptic curve with CM by the maximal order \mathcal{O} of an imaginary quadratic field K of discriminant -d<0 . We consider a principally polarised abelian surface A=E^{2} with an indecomposable polarization \lambda , i.e. there does not exists E_{1},E_{2} two elliptic curves such that (A,\lambda)\simeq(E_{1}\times E_{2},\lambda_{0}) where \lambda_{0} is the product polarization. Therefore (A,\lambda) is the Jacobian of a genus 2 curve C and one can prove (see beginning of Section 3 ) that all these objects are defined over the Hilbert class field H of K . Even better, given a prime p\in\mathbb{N} and a prime \mathfrak{p} over p in the ring of integers of H , we can construct the stable model of all these objects over its local completion \mathbb{Z}_{\mathfrak{p}} in H_{\mathfrak{p}} . The model \mathcal{E} of E is smooth since E has CM and therefore the special fiber (\bar{E}^{2},\bar{\lambda}) of (\mathcal{E}^{2},\lambda) is a principally polarised abelian surface. However, when \bar{E} is supersingular, the polarization \bar{\lambda} may become decomposable or equivalently the special fiber of the stable model of C is singular. This shows the deep connection between decomposability and primes of bad reduction. We will first focus on writing down an algorithm to compute the primes \mathfrak{p} for which the polarization becomes decomposable. We call such a prime \mathfrak{p} a prime of decomposable reduction . The classical source for studying decomposition of abelian surfaces (up to isogeny) over \mathbb{C} is [ Hum99 ] . It was revisited and extended to any characteristic in a series of papers by Kani and our main reference will be [ Kan16 ] . For any principally abelian surface (A,\lambda) , he defines the so-called refined Humbert invariant as an integral positive definite quadratic form q_{(A,\lambda)} on the quotient L^{Kani}:=\Ends(A,\lambda)/\mathbb{Z}\cdot 1_{\End(A)} of the symmetric endomorphism of A (i.e. the endomorphisms which are invariant under the Rosati involution associated to \lambda ). The crucial result for us is Theorem 2.10 , [ Kan16 , Prop.6] : (A,\lambda) is decomposable if and only if q:=q_{(A,\lambda)} represents 1 . We give in Section 2 a summary of the results we need about this quadratic form which may be a good introduction for people interested in the prolific work of Kani. However, instead of working with L^{Kani} , we introduce a sublattice L of \Ends(A,\lambda) , analogue to the Gross lattice for elliptic curve in [ Gro87 ] such that (L^{Kani},q) and (L,q/4) are isometric, but this second space allows a more direct proof of Theorem 2.10 . With this form at hand, a first piece of the algorithm becomes straightforward. In Section 3 , given • an explicit E/H with CM by \mathcal{O} , • a indecomposable principal polarization given as \lambda=\lambda_{0}P where P\in M_{2}(\mathcal{O}) , • a prime p\in\mathbb{N} which is inert or ramified in \mathcal{O} , • a prime ideal \mathfrak{p}\in H over p , we compute explicitly \bar{L}\subset\Ends(\bar{E}^{2},\lambda_{0}P) and its Gram matrix with respect to q/4 . It is then enough to look for the existence of an element of norm 1 in the 5-rank lattice \bar{L} to check whether \mathfrak{p} is a prime of decomposable reduction. This is summarised in Algorithm 1 . The second step is to find a bound on the primes p\in\mathbb{N} over which a decomposable reduction can occur. We use a key argument of [ KR00 ] by looking at the image of our lattice \bar{L} into the kernel of the p -power Frobenius map on \bar{E}^{2} . Because the dimension drops, on can show that p has to divide the determinant of the Gram matrix T of a rank-4 lattice generated by the rank-3 Gross lattice L\subset\Ends(E^{2},\lambda) and a vector of norm 1, necessarily in \bar{L}\setminus L . As one knows the discriminant of the Gram matrix of L to be 4d , one can then bound p by 4d . Adding some necessary congruence conditions on the remaining entries of the potential matrices T , one eventually gets the sharp bound p\leq d/4 (Theorem 4.7 ). Combining this two steps, one could then algorithmically obtain all the primes \mathfrak{p} of decomposable reduction for a given (E^{2},\lambda) . In Section 5 , we prefer to get a formulation free of the choice of E among all Galois conjugate curves with CM by

Read the complete text