10.07.2015 Views

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

184 Chapter 6. Complex multiplication in higher generaAlgorithm 6.3: Computation of the Igusa class polynomials of a non-maximal or<strong>de</strong>rInput: An or<strong>de</strong>r O in a CM quartic number field KOutput: The Igusa class polynomials H 1 (X), H 2 (X) and H 3 (X) corresponding to thePicard group of O1 Compute a set M of representatives of the isomorphism classes of p.p.a.v. with CM by Ogiven by invertible i<strong>de</strong>als2 foreach Triple (Φ, a, ξ) ∈ M do3 Compute a symplectic basis of a with respect to the polarization <strong>de</strong>fined by ξ4 Reduce the corresponding period matrix Ω into a fundamental domain5 Evaluate the Igusa invariants i 1 , i 2 and i 3 with sufficient precision through analyticevaluation of theta constants on Ω6 Reconstruct the polynomials H 1 (X), H 2 (X) and H 3 (X) with rational coefficients fromtheir respective roots7 return H 1 (X), H 2 (X), H 3 (X)This method can as well be exten<strong>de</strong>d to the case of non-maximal or<strong>de</strong>rs. In<strong>de</strong>ed, we know thatσ ∈ Aut(C/K r ) also acts as multiplication by the idèle N Φ r(s −1 ) for general or<strong>de</strong>rs. To <strong>de</strong>scribethat action explicitly as multiplication by an invertible i<strong>de</strong>al, it is sufficient that N Φ r(s −1 ) ≡ 1mod f, which can always be achieved up to multiplication by a principal idèle. Equivalently, theaction of an i<strong>de</strong>al a of O K r is given by multiplication N Φ r(a −1 ) ∩ O provi<strong>de</strong>d that N Φ r(a −1 ) isco-prime to f. This can also always be achieved up to multiplication by a principal i<strong>de</strong>al.6.4.4 Going furtherThe previous <strong>de</strong>scription of the algorithms is very scarce and many <strong>de</strong>tails are omitted. The missingsteps for the principal case can be found in the Ph.D. thesis of Streng [252] and extend naturallyto the non-principal case. We now list a few of them as well as some practical optimizations.A first concrete complication in comparison with the genus 1 case is that Igusa class polynomialsare no more integral, but only rational. Moreover, computing bounds on the value of the<strong>de</strong>nominator is a hard problem. It was recently achieved by Goren and Lauter [119, 120] eventhogh their result is not very sharp.Furthermore, once the isomorphisms classes and the corresponding period matrices havebeen computed, they should be reduced, i.e. moved to a fundamental domain, to obtain fasterconvergence as is done for genus 1 [252, II.5].Then, the theta constants should be computed using Borchardt sequences as proposed byDupont [75]. This led to some record computations ma<strong>de</strong> by Dupont, Enge and Thomé [76].The complex analytic method is not the only one which extends from genus 1. Eisenträgerand Lauter <strong>de</strong>veloped a CRT method [80], implemented by Freeman [103] and later improved byBröker, Gruenewald and Lauter [29] and Lauter and Robert [166]. As far as the p-adic methodsare concerned, Gaudry et al. [112] <strong>de</strong>veloped a 2-adic one and Carls, Kohel and Lubicz [43] a3-adic one.To conclu<strong>de</strong>, it should be mentioned that it is possible to <strong>de</strong>fine smaller class invariants thanthe Igusa invariants as was already the case in dimension 1. Such an approach, based on thehigher-dimensional Shimura reciprocity, is being conducted by Streng [254, 255, 253]. Relatedi<strong>de</strong>as can also be found in the Ph.D. thesis of Uzunkol [269].

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!