This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

\setkomafont

disposition \frefformatplaindefdefinition\fancyrefdefaultspacing#1\FrefformatplaindefDefinition\fancyrefdefaultspacing#1 \frefformatplainlemmalemma\fancyrefdefaultspacing#1\FrefformatplainlemmaLemma\fancyrefdefaultspacing#1 \frefformatplaincorollarycorollary\fancyrefdefaultspacing#1\FrefformatplaincorollaryCorollary\fancyrefdefaultspacing#1 \frefformatplainremremark\fancyrefdefaultspacing#1\FrefformatplainremRemark\fancyrefdefaultspacing#1 \frefformatplainthmtheorem\fancyrefdefaultspacing#1\FrefformatplainthmTheorem\fancyrefdefaultspacing#1 \frefformatplaintabletable\fancyrefdefaultspacing#1\FrefformatplaintableTable\fancyrefdefaultspacing#1

Genus 2 Curves in Small Characteristic

Lukas Zobernig
Abstract

We study genus 2 curves over finite fields of small characteristic. The pp-rank ff of a curve induces a stratification of the coarse moduli space 2\mathcal{M}_{2} of genus 2 curves up to isomorphism. We are interested in the size of those strata for all f{0,1,2}f\in\{0,1,2\}. In characteristic 2 and 3, previous results show that the supersingular f=0f=0 stratum has size qq. We show that for q=3rq=3^{r}, over 𝔽q\mathbb{F}_{q} the non-ordinary f=1f=1 and ordinary f=2f=2 strata are of size q(q1)q(q-1) and q2(q1)q^{2}(q-1), respectively. We give results found from computer calculations which suggest that these formulas hold for all p7p\leq 7 and break down for p>7p>7.

1 Introduction

[MN07] studied supersingular genus 2 curves in characteristic 2. [How08] showed that over an algebraically closed field kk characteristic 3, the supersingular locus 𝒮2\mathcal{S}_{2} of the coarse moduli space 2\mathcal{M}_{2} is in bijection with kk via certain invariants of genus 2 curves; we will discuss these invariants in \Frefsec:moduli_and_invs.

In \Frefsec:theorems, we prove a similar result for the non-ordinary stratum of 2\mathcal{M}_{2} in characteristic 3, again in terms of certain absolute invariants. From this we obtain a complete classification of isomorphism classes of supersingular, non-ordinary, and ordinary genus 2 curves defined over a finite field 𝔽q\mathbb{F}_{q} with q=3rq=3^{r} in terms of absolute invariants, and their counts. In \Frefsec:comp_results we present and discuss further computational results for p>3p>3.

Let us briefly recall some background on the pp-torsion structure of abelian varieties, moduli and fields of definition of genus 2 curves, and previous results for those curves over finite fields of characteristic 2 and 3.

1.1 The pp-rank and aa-number of Abelian Varieties and Curves

Fix a prime pp, and an algebraically closed field kk of characteristic pp containing 𝔽p\mathbb{F}_{p}. Consider the additive group scheme 𝔾a=Speck[X]\mathbb{G}_{a}=\operatorname{Spec}k[X] and the multiplicative group scheme 𝔾m=Speck[X,X1]\mathbb{G}_{m}=\operatorname{Spec}k[X,X^{-1}]. The kernel of the relative Frobenius on these yields the finite group schemes αpSpeck[X]/Xp\alpha_{p}\cong\operatorname{Spec}k[X]/X^{p} and μpSpeck[X]/(Xp1)\mu_{p}\cong\operatorname{Spec}k[X]/(X^{p}-1), respectively. The Cartier dual of αp\alpha_{p} is itself, and the Cartier dual of μp\mu_{p} is the constant group scheme /p\mathbb{Z}/p\mathbb{Z}.

Let A/kA/k be an abelian variety of genus gg and consider its pp-torsion A[p]A[p] as a group scheme. The pp-rank of AA is given by f=dim𝔽pHom(μp,A[p])f=\dim_{\mathbb{F}_{p}}\operatorname{Hom}(\mu_{p},A[p]). Similarly, the aa-number of AA is given by a=dimkHom(αp,A)a=\dim_{k}\operatorname{Hom}(\alpha_{p},A). Thus, geometrically, A[p](k)(/p)fA[p](k)\cong(\mathbb{Z}/p\mathbb{Z})^{f}. It holds that 0fg0\leq f\leq g and 1a+fg1\leq a+f\leq g.

Example 1 ([Pri08]).

Let A/kA/k be of genus 2, i.e. an abelian surface, then we have the following possible types:

ff aa A[p]A[p] Type Codim.
22 0 L2L^{2} ordinary 0
11 11 LI1,1L\oplus I_{1,1} non-ordinary 1
0 11 I2,1I_{2,1} supersingular 2
0 22 I1,1I1,1I_{1,1}\oplus I_{1,1} superspecial 3

Here L=/pμpL=\mathbb{Z}/p\mathbb{Z}\oplus\mu_{p} is the pp-torsion of an ordinary elliptic curve, whereas I1,1I_{1,1} is the pp-torsion of a supersingular elliptic curve. It is also the unique BT1BT_{1} group scheme of rank pp and fits into the following non-split exact sequence: 0αpI1,1αp00\to\alpha_{p}\to I_{1,1}\to\alpha_{p}\to 0. Similarly, I2,1I_{2,1} is the unique BT1BT_{1} group scheme of rank p2p^{2} with pp-rank 0 and aa-number 1. Note that a superspecial abelian surface is in particular supersingular. The codimension of the associated strata in the full moduli space of abelian surfaces 𝒜2\mathcal{A}_{2} is given as well.

We define the pp-rank and the aa-number of a genus g curve CC as the corresponding invariants of its Jacobian Jac(C)\operatorname{Jac}(C) as an abelian variety.

Assume char(k)2\operatorname{\mathrm{char}}(k)\neq 2 and consider a genus gg hyperelliptic curve CC defined by an equation y2=f(x)y^{2}=f(x) for f(x)k[x]f(x)\in k[x] of degree 2g+12g+1 or 2g+22g+2. Let cic_{i} denote the coefficient of xix^{i} in the expansion of f(x)(p1)/2f(x)^{(p-1)/2}, and define for =0,,g1\ell=0,\dots,g-1 the g×gg\times g matrix AA_{\ell} with entries (A)i,j=(cipj)p(A_{\ell})_{i,j}=(c_{ip-j})^{p^{\ell}}. Following [Yui78] and heeding [AH19] we call A0A_{0} the Cartier-Manin matrix of the curve CC and define the matrix M=Ag1A1A0M=A_{g-1}\cdots A_{1}A_{0}. Then we have the following lemma.

Lemma 1.

Let CC be a genus g hyperelliptic curve defined by y2=f(x)y^{2}=f(x).

  1. 1.

    The pp-rank of CC is fC=rank(M)f_{C}=\operatorname{\mathrm{rank}}(M).

  2. 2.

    The aa-number of CC is aC=grank(A0)a_{C}=g-\operatorname{\mathrm{rank}}(A_{0}).

1.2 Moduli Space and Invariants

Points in the coarse moduli space 2\mathcal{M}_{2} of genus 2 curves up to isomorphism have various different descriptions. Consider the weighted projective space Sk=(2,4,6,8,10)(k)S_{k}=\mathbb{P}(2,4,6,8,10)(k) over an arbitrary field kk. [Igu60] gave a set of invariants [J2:J4:J6:J8:J10]Sk[J_{2}:J_{4}:J_{6}:J_{8}:J_{10}]\in S_{k} associated to every genus 2 curve CC defined over kk. Two curves are isomorphic over kalgk^{\text{alg}} if any only if their Igusa-invariants are the same in SkS_{k}. If char(k)2\operatorname{\mathrm{char}}(k)\neq 2, we can assume that CC is given by a model of the form y2=f(x)y^{2}=f(x), where f(x)k[x]f(x)\in k[x] is a polynomial of degree 5 or degree 6. Then 212J102^{12}J_{10} is simply the discriminant of f(x)f(x).

For fields of characteristic different from 2, Cardona, Quer, Nart, and Pujolàs [CQ05, CNP05] gave the absolute “G2” invariants

(g1,g2,g3)={(J25J10,J23J4J10,J22J6J10)ifJ20,(0,J45J102,J4J6J10)ifJ2=0,J40,(0,0,J65J103)ifJ2=J4=0.(g_{1},g_{2},g_{3})=\begin{cases}\left(\frac{J_{2}^{5}}{J_{10}},\frac{J_{2}^{3}J_{4}}{J_{10}},\frac{J_{2}^{2}J_{6}}{J_{10}}\right)&\text{if}\;J_{2}\neq 0,\\ \left(0,\frac{J_{4}^{5}}{J_{10}^{2}},\frac{J_{4}J_{6}}{J_{10}}\right)&\text{if}\;J_{2}=0,J_{4}\neq 0,\\ \left(0,0,\frac{J_{6}^{5}}{J_{10}^{3}}\right)&\text{if}\;J_{2}=J_{4}=0.\end{cases} (1)

Again, two curves are isomorphic over kalgk^{\text{alg}} if any only if their absolute invariants are the same. Hence, the kk-points 2(k)\mathcal{M}_{2}(k) are in bijection with the set of points (g1,g2,g3)𝔸3(k)(g_{1},g_{2},g_{3})\in\mathbb{A}^{3}(k) as above.

Remark 1.

Consider a point of moduli P=[C]2(k)P=[C]\in\mathcal{M}_{2}(k) defined over kk. Then [Mes91], and [CQ05, Theorem 2 ] showed that it is only the generic case Aut(C)C2\operatorname*{\mathrm{Aut}}(C)\cong C_{2} where there exists an obstruction to the point PP being represented by a curve CC^{\prime} defined over kk, which is an element in Br2(k)\operatorname{Br}_{2}(k) (i.e. a non-trivial 2-torsion element). If the automorphism group Aut(C)≇C2\operatorname*{\mathrm{Aut}}(C)\not\cong C_{2}, then there always exists a curve CC^{\prime} defined over kk with CCC\cong C^{\prime} (geometrically). Hence, if kk has trivial Brauer group (for example when kk is a finite field, or an algebraic closure thereof), then for every (g1,g2,g3)𝔸3(k)(g_{1},g_{2},g_{3})\in\mathbb{A}^{3}(k) there exists a curve CC defined over kk with those invariants.

1.3 Transformation of Sextic Forms

In most generality, a genus 2 curve is given by an equation of the form y2=f(x)y^{2}=f(x) with f(x)=c6x6+c1x+c0f(x)=c_{6}x^{6}+\cdots c_{1}x+c_{0} for coefficients c0,,c6kc_{0},\dots,c_{6}\in k. After homogenising, we find F(X,Z)=c6X6+c1XZ5+c0Z6F(X,Z)=c_{6}X^{6}+\cdots c_{1}XZ^{5}+c_{0}Z^{6}. Then the group GL2\operatorname{GL}_{2} acts on F(X,Z)F(X,Z) via the following transformation: Let M=(αβγδ)GL2M=\begin{pmatrix}\alpha&\beta\\ \gamma&\delta\end{pmatrix}\in GL_{2} and let

X=αX+βZ,\displaystyle X=\alpha X^{\prime}+\beta Z^{\prime},
Z=γX+δZ.\displaystyle Z=\gamma X^{\prime}+\delta Z^{\prime}.

We find F(X,Z)=c6X6+c1XZ5+c0Z6=G(X,Z)F(X,Z)=c_{6}^{\prime}X^{\prime 6}+\cdots c_{1}^{\prime}X^{\prime}Z^{\prime 5}+c_{0}^{\prime}Z^{\prime 6}=G(X^{\prime},Z^{\prime}) for new coefficients c0,,c6c_{0}^{\prime},\dots,c_{6}^{\prime}. Hence, we may call F(X,Z)F(X,Z) and G(X,Z)G(X^{\prime},Z^{\prime}) to be GL2\operatorname{GL}_{2}-conjugates. Let g(x)=c6x6+c1x+c0g(x^{\prime})=c_{6}^{\prime}x^{\prime 6}+\cdots c_{1}^{\prime}x^{\prime}+c_{0}^{\prime} be the dehomogenisation of G(X,Z)G(X^{\prime},Z^{\prime}). Then the two curves C:y2=f(x)C:y^{2}=f(x) and C:y2=g(x)C^{\prime}:y^{\prime 2}=g(x^{\prime}) are isomorphic via

(x,y)(x,y)=(αx+βγx+δ,y(γx+δ)3).(x^{\prime},y^{\prime})\mapsto(x,y)=\left(\frac{\alpha x^{\prime}+\beta}{\gamma x^{\prime}+\delta},\frac{y^{\prime}}{(\gamma x^{\prime}+\delta)^{3}}\right).

This is actually an if and only if relation: Two genus 2 curves C,CC,C^{\prime} are isomorphic if any only if their associated sextic forms are conjugate under GL2\operatorname{GL}_{2}-action. Thus we can also study points in the moduli space 2\mathcal{M}_{2} by considering sextic forms up to GL2\operatorname{GL}_{2}-conjugacy, see [Mes91].

1.4 Genus 2 Curves in Characteristic 2 and 3

[MN07], and [How08] studied supersingular genus 2 curves in characteristic 2 and 3, respectively. [How08, Theorem 2.2] states that in characteristic 3, the coarse moduli space 𝒮2\mathcal{S}_{2} of supersingular genus 2 curves is isomorphic to the affine line 𝔸1\mathbb{A}^{1}: The absolute invariants (0,0,g3)(0,0,g_{3}) correspond to the curve y2=x6+g32x3+g33x+g34y^{2}=x^{6}+g_{3}^{2}x^{3}+g_{3}^{3}x+g_{3}^{4} if g30g_{3}\neq 0 and the point (0,0,0)(0,0,0) corresponds to y2=x5+1y^{2}=x^{5}+1. Hence, over the finite field 𝔽q\mathbb{F}_{q} where q=3rq=3^{r}, the subspace 𝒮2(𝔽q)\mathcal{S}_{2}(\mathbb{F}_{q}) of 2(𝔽q)\mathcal{M}_{2}(\mathbb{F}_{q}) corresponding to isomorphism classes of supersingular genus 2 curves contains qq elements.

2 Theoretical Results

The goal of this section is to show that over the finite field 𝔽q\mathbb{F}_{q} where q=3rq=3^{r}, there are q2(q1)q^{2}(q-1) many ordinary, q(q1)q(q-1) many non-ordinary, and qq many supersingular genus 2 curves.

Theorem 1.

Let kk be a finite field of characteristic 33. Let CC be a genus 2 curve defined over kk and let (g1,g2,g3)(g_{1},g_{2},g_{3}) be its absolute invariants. Then CC is non-ordinary if and only if g1=0g_{1}=0, g2k×g_{2}\in k^{\times}, and g3kg_{3}\in k.

Proof.

We can assume that CC has a model over kk of the form dy2=x6+c5x5+c4x4+c3x3+c2x2+c1x+c0dy^{2}=x^{6}+c_{5}x^{5}+c_{4}x^{4}+c_{3}x^{3}+c_{2}x^{2}+c_{1}x+c_{0}. Then the matrices A0A_{0} and A1A_{1} of CC are given by

A0=(c2c1c5c4),A1=(c23c13c53c43).A_{0}=\begin{pmatrix}c_{2}&c_{1}\\ c_{5}&c_{4}\end{pmatrix},\;A_{1}=\begin{pmatrix}c_{2}^{3}&c_{1}^{3}\\ c_{5}^{3}&c_{4}^{3}\end{pmatrix}.

Thus, the Hasse-Witt matrix MM of CC becomes

M=A1A0=(c13c5+c24c13c4+c1c23c2c53+c43c5c1c53+c44).M=A_{1}A_{0}=\begin{pmatrix}c_{1}^{3}c_{5}+c_{2}^{4}&c_{1}^{3}c_{4}+c_{1}c_{2}^{3}\\ c_{2}c_{5}^{3}+c_{4}^{3}c_{5}&c_{1}c_{5}^{3}+c_{4}^{4}\end{pmatrix}.

By \Freflemma:prank_anumber, C has pp-rank 1 if rank(M)=1\operatorname{\mathrm{rank}}(M)=1. To make analysis easier, we can alternatively require rank(A0)=1\operatorname{\mathrm{rank}}(A_{0})=1, and exclude the supersingular cases dy2=x6+c5x5+c3x3+c0dy^{2}=x^{6}+c_{5}x^{5}+c_{3}x^{3}+c_{0} and dy2=x6+c3x3+c1x+c0dy^{2}=x^{6}+c_{3}x^{3}+c_{1}x+c_{0} we find from the constraint M=0M=0 (again by \Freflemma:prank_anumber). The remaining cases correspond to the following seven possible A0A_{0} matrices:

(c2000),(000c4),(c20c50),(0c10c4),(c20c20),(0c50c5),(c2c5c2c5).\begin{pmatrix}c_{2}&0\\ 0&0\end{pmatrix},\begin{pmatrix}0&0\\ 0&c_{4}\end{pmatrix},\begin{pmatrix}c_{2}&0\\ c_{5}&0\end{pmatrix},\begin{pmatrix}0&c_{1}\\ 0&c_{4}\end{pmatrix},\begin{pmatrix}c_{2}&0\\ c_{2}&0\end{pmatrix},\begin{pmatrix}0&c_{5}\\ 0&c_{5}\end{pmatrix},\begin{pmatrix}c_{2}&c_{5}\\ c_{2}&c_{5}\end{pmatrix}.

Both approaches yield the following seven possible types of models:

dy2\displaystyle dy^{2} =x6+c3x3+c2x2+c0,\displaystyle=x^{6}+c_{3}x^{3}+c_{2}x^{2}+c_{0},
dy2\displaystyle dy^{2} =x6+c4x4+c3x3+c0,\displaystyle=x^{6}+c_{4}x^{4}+c_{3}x^{3}+c_{0},
dy2\displaystyle dy^{2} =x6+c5x5+c3x3+c2x2+c0,\displaystyle=x^{6}+c_{5}x^{5}+c_{3}x^{3}+c_{2}x^{2}+c_{0},
dy2\displaystyle dy^{2} =x6+c4x4+c3x3+c1x+c0,\displaystyle=x^{6}+c_{4}x^{4}+c_{3}x^{3}+c_{1}x+c_{0},
dy2\displaystyle dy^{2} =x6+c3x3+c2(x2+x)+c0,\displaystyle=x^{6}+c_{3}x^{3}+c_{2}(x^{2}+x)+c_{0},
dy2\displaystyle dy^{2} =x6+c5(x5+x4)+c3x3+c0,\displaystyle=x^{6}+c_{5}(x^{5}+x^{4})+c_{3}x^{3}+c_{0},
dy2\displaystyle dy^{2} =x6+c5(x5+x4)+c3x3+c2(x2+x)+c0.\displaystyle=x^{6}+c_{5}(x^{5}+x^{4})+c_{3}x^{3}+c_{2}(x^{2}+x)+c_{0}.

These types are all pairwise isomorphic over kalgk^{\text{alg}} either by simply switching affine patches via the transformation x1/xx\mapsto 1/x followed by a rescaling of yy, or by a more general GL2(kalg)\operatorname{GL}_{2}(k^{\text{alg}}) transformation of the associated sextic forms, see \Frefsec:sextic_forms_transf.

Hence, without loss of generality, assume that CC has a model dy2=x6+c3x3+c2x2+c0dy^{2}=x^{6}+c_{3}x^{3}+c_{2}x^{2}+c_{0}. For its Igusa invariants we find [0:c23:c03c0c23+c36:c26:c0c26]\left[0:c_{2}^{3}:-c_{0}^{3}-c_{0}c_{2}^{3}+c_{3}^{6}:-c_{2}^{6}:-c_{0}c_{2}^{6}\right]. Using \Frefeq:g2invariants, the absolute invariants are

(0,c23c02,c03+c0c23c36c0c23)=(0,g2,g3).\left(0,\frac{c_{2}^{3}}{c_{0}^{2}},\frac{c_{0}^{3}+c_{0}c_{2}^{3}-c_{3}^{6}}{c_{0}c_{2}^{3}}\right)=(0,g_{2},g_{3}).

Since the discriminant 212J10=c0c262^{12}J_{10}=-c_{0}c_{2}^{6} has to be non-zero for the model to define a smooth curve, we have that c0,c20c_{0},c_{2}\neq 0 and so g2k×,g3kg_{2}\in k^{\times},g_{3}\in k.

On the other hand, choose g2k×,g3kg_{2}\in k^{\times},g_{3}\in k, and set c2=g23c_{2}=\sqrt[3]{g_{2}}, c3=1+g2g2g36c_{3}=\sqrt[6]{1+g_{2}-g_{2}g_{3}}. Then by \Frefrem:field_of_mod_def the curve y2=x6+c3x3+c2x2+1y^{2}=x^{6}+c_{3}x^{3}+c_{2}x^{2}+1 is geometrically isomorphic to a curve defined over kk with absolute invariants (0,g2,g3)(0,g_{2},g_{3}), and is non-ordinary by \Freflemma:prank_anumber. ∎

The proof of \Frefthm:nonord_classification immediately implies the following corollary.

Corollary 1.

Let kk be a finite field of characteristic 3. For g2k×g_{2}\in k^{\times} and g3kg_{3}\in k, the curve

y2=x6+1+g2g2g36x3+g23x2+1y^{2}=x^{6}+\sqrt[6]{1+g_{2}-g_{2}g_{3}}x^{3}+\sqrt[3]{g_{2}}x^{2}+1

defined over kalgk^{\text{alg}} is non-ordinary and has absolute invariants (0,g2,g3)(0,g_{2},g_{3}).

Combining \Frefsec:prev_results and \Frefthm:nonord_classification yields the following classification for the absolute invariants of an ordinary genus 2 curve over a finite field of characteristic 3.

Corollary 2.

Let kk be a finite field of characteristic 33. Every ordinary genus 2 curve defined over kk has absolute invariants (g1,g2,g3)(g_{1},g_{2},g_{3}) with g1k×g_{1}\in k^{\times} and g2,g3kg_{2},g_{3}\in k.

For a finite field k=𝔽qk=\mathbb{F}_{q} with q=prq=p^{r}, denote by 𝒮2(k)\mathcal{S}_{2}(k), 𝒩2(k)\mathcal{N}_{2}(k), and 𝒪2(k)\mathcal{O}_{2}(k) the supersingular, non-ordinary, and ordinary strata of the coarse moduli space 2(k)\mathcal{M}_{2}(k). Finally, we can use [How08, Theorem 2.2] and \Frefthm:nonord_classification to describe the sizes of these pp-rank strata defined over 𝔽q\mathbb{F}_{q} where q=3rq=3^{r}.

Theorem 2.

Let kk be a finite field of characteristic 3 with qq elements. Then

  1. 1.

    #𝒪2(k)=q2(q1)\#\mathcal{O}_{2}(k)=q^{2}(q-1),

  2. 2.

    #𝒩2(k)=q(q1)\#\mathcal{N}_{2}(k)=q(q-1),

  3. 3.

    #𝒮2(k)=q\#\mathcal{S}_{2}(k)=q.

Proof.

The discussion in \Frefsec:moduli_and_invs implies that there are #𝔸3(k)=q3\#\mathbb{A}^{3}(k)=q^{3} many geometric isomorphism classes of curves defined over kk. Then the sizes of 𝒪2(k)\mathcal{O}_{2}(k), 𝒩2(k)\mathcal{N}_{2}(k), and 𝒮2(k)\mathcal{S}_{2}(k) immediately follow from [How08, Theorem 2.2] and \Frefthm:nonord_classification. ∎

3 Computational Results for Small Primes

We have computed the pp-rank of a curve corresponding to each absolute invariant (g1,g2,g3)𝔸3(𝔽q)(g_{1},g_{2},g_{3})\in\mathbb{A}^{3}(\mathbb{F}_{q}) for small finite fields 𝔽q\mathbb{F}_{q} of characteristic 2,3,52,3,5, and 7. \Freftable:strata_sizes_exact summarises the results we have found.

q #𝒮2(k)\#\mathcal{S}_{2}(k) #𝒩2(k)\#\mathcal{N}_{2}(k) #𝒪2(k)\#\mathcal{O}_{2}(k)
2 2 2 4
222^{2} 222^{2} 2232^{2}\cdot 3 2432^{4}\cdot 3
3 3 323\cdot 2 3223^{2}\cdot 2
323^{2} 323^{2} 3283^{2}\cdot 8 3483^{4}\cdot 8
333^{3} 333^{3} 33263^{3}\cdot 26 36263^{6}\cdot 26
5 5 545\cdot 4 5245^{2}\cdot 4
525^{2} 525^{2} 52245^{2}\cdot 24 54245^{4}\cdot 24
7 7 767\cdot 6 7267^{2}\cdot 6
727^{2} 727^{2} 72487^{2}\cdot 48 74487^{4}\cdot 48
Table 1: Strata sizes of 2\mathcal{M}_{2} over finite fields of characteristic 2,3,52,3,5, and 77.

As expected, in characteristic 22, #𝒮2\#\mathcal{S}_{2} follows the results of [MN07]. Similary, characteristic 33 behaves as predicted by \Frefthm:strata_sizes. Inspired by the computational results, we conjecture that this behaviour also holds in characteristic 5 and 7 and for the other strata in characteristic 2.

Denote by Δ0=#𝒮2(k)q\Delta_{0}=\#\mathcal{S}_{2}(k)-q, Δ1=#𝒩2(k)q(q1)\Delta_{1}=\#\mathcal{N}_{2}(k)-q(q-1), and Δ2=#𝒪2(k)q2(q1)\Delta_{2}=\#\mathcal{O}_{2}(k)-q^{2}(q-1) the difference in number of points to what one would expect if the size of the strata followed the small prime behaviour. We have computed \Freftable:strata_sizes_small_char determining the pp-rank of a curve corresponding to each absolute invariant (g1,g2,g3)𝔸3(𝔽q)(g_{1},g_{2},g_{3})\in\mathbb{A}^{3}(\mathbb{F}_{q}) for finite fields 𝔽q\mathbb{F}_{q} with q>7q>7.

q #𝒮2(k)\#\mathcal{S}_{2}(k) Δ0\Delta_{0} #𝒩2(k)\#\mathcal{N}_{2}(k) Δ1\Delta_{1} #𝒪2(k)\#\mathcal{O}_{2}(k) Δ2\Delta_{2}
11 9 -2 101 -9 1221 11
11211^{2} 117 -4 14403 -117 1757041 121
13 20 7 149 -7 2028 0
13213^{2} 330 161 28231 -161 4798248 0
17 25 8 264 -8 4624 0
19 26 7 335 -7 6498 0
23 36 13 494 -12 11637 -1
29 49 20 851 39 23489 -59
31 54 23 970 40 28767 -63
37 102 65 1229 -103 49322 38
41 70 29 1794 154 67057 -183
47 109 62 2308 146 101406 -208
53 155 102 2843 87 145879 -189
61 186 125 3775 115 223020 -240
67 210 143 4093 -329 296460 186
71 146 75 5770 800 351995 -875
73 269 196 4949 -307 383799 111
79 216 137 6838 676 485985 -813
83 259 176 7529 723 563999 -899
89 226 137 9053 1221 695690 -1358
97 408 311 8726 -586 903539 275
101 347 246 11784 1684 1018170 -1930
103 443 340 11394 888 1080890 -1228
107 357 250 12443 1101 1212243 -1351
109 412 303 11750 -22 1282867 -281
113 417 304 12834 178 1429646 -482
127 570 443 17592 1590 2030221 -2033
131 409 278 20931 3901 2226751 -4179
137 576 439 18198 -434 2552579 -5
139 516 377 20745 1563 2664358 -1940
Table 2: Strata sizes and differences of 2\mathcal{M}_{2} over finite fields of small characteristic.

It would be interesting to determine how exactly the differences Δ1\Delta_{1} and Δ2\Delta_{2} of the non-ordinary and ordinary strata, respectively, depend on qq. Alternatively we would like to say something about their distribution depending on the characteristic pp. This seems to require some Sato-Tate style argument depending on the reduction of the moduli space 2\mathcal{M}_{2} at various primes.

3.1 The Supersingular Locus

In this section we recall some theory about the size of the supersingular locus 𝒮2\mathcal{S}_{2}. [IKO86], [KO87], and [Kob75] show that the supersingular locus of 2\mathcal{M}_{2} is a union of projective lines with the singular points of this union corresponding to the superspecial points. Denote the finite set of superspecial points by 𝒮𝒫2\mathcal{SP}_{2}.

Let BB be the definite quaternion algebra over \mathbb{Q} with discriminant DD and let 𝒪\mathcal{O} be a maximal order of BB. Write D=D1D2D=D_{1}D_{2} for two positive integers D1,D2D_{1},D_{2} and set Ln(D1,D2)L_{n}(D_{1},D_{2}) the set of left 𝒪\mathcal{O}-lattices in BnB^{n} which are equivalent to (𝒪p)n(\mathcal{O}\otimes\mathbb{Z}_{p})^{n} at pp if pp does not divide D2D_{2}, and otherwise to the other local equivalence class if pp divides D2D_{2}, see [IKO86]. Denote by Hn(D1,D2)H_{n}(D_{1},D_{2}) the number of global equivalence classes in Ln(D1,D2)L_{n}(D_{1},D_{2}). In particular, for a prime number pp, denote by Hn(p,1)H_{n}(p,1) the class number of the principal genus and by Hn(1,p)H_{n}(1,p) the class number of the non-principal genus.

Let kk be a finite field of characteristic pp. On the one hand, it is known that every principally polarised superspecial abelian surface AA over kalgk^{\text{alg}} arises from choosing a principal polarisation on a product E×EE\times E for a supersingular elliptic curve EE. Let B=End(E)B=\operatorname*{\mathrm{End}}(E)\otimes\mathbb{Q} the endomorphism algebra of EE, which is the definite quaternion algebra ramified at pp. Then the number of principally polarisations on E×EE\times E over kalgk^{\text{alg}} up to automorphisms is equal to H2(p,1)H_{2}(p,1) for BB.

On the other hand, as a principally polarised abelian variety, either A=E1×E2A=E_{1}\times E_{2} for two supersingular elliptic curves E1E_{1} and E2E_{2}, or A=Jac(C)A=\operatorname{Jac}(C) for a superspecial genus 2 curve CC. Hence the number of superspecial genus 2 curves over kalgk^{\text{alg}} is given by Hp=H2(p,1)hp(hp+1)/2H_{p}=H_{2}(p,1)-h_{p}(h_{p}+1)/2, where hp=H1(p,1)h_{p}=H_{1}(p,1) is the number of isomorphism classes of supersingular elliptic curves over kalgk^{\text{alg}}, see [IKO86, Corollary 2.12 ].

The number HpH_{p} is finite; every superspecial genus 2 curve can be defined over 𝔽p2\mathbb{F}_{p^{2}}. This allows us to determine the finite contribution of superspecial genus 2 curves to #𝒮2(k)\#\mathcal{S}_{2}(k) when kk contains 𝔽p2\mathbb{F}_{p^{2}}. For small primes pp we find \Freftable:ssing_sspec_sizes; we have also determined the number of superspecial curves defined over 𝔽p\mathbb{F}_{p}.

pp 11 13 17 19 23 29 31 37 41
#𝒮2(𝔽p)\#\mathcal{S}_{2}(\mathbb{F}_{p}) 9 20 25 26 36 49 54 102 70
HpH_{p} 2 3 5 7 10 18 20 31 40
#𝒮𝒫2(𝔽p)\#\mathcal{SP}_{2}(\mathbb{F}_{p}) 2 3 5 5 8 12 12 9 22
Table 3: Number of 𝔽p\mathbb{F}_{p} points of 𝒮2\mathcal{S}_{2} and 𝒮𝒫2\mathcal{SP}_{2} in small characteristic.

The number of irreducible components of the supersingular locus of 2\mathcal{M}_{2}, i.e. the aforementioned projective lines, is equal to the class number of the non-principal genus H2(1,p)H_{2}(1,p), see [KO87, Theorem 5.7 ].

Acknowledgements

We thank Yan Bo Ti for helpful discussions and Everett Howe for helpful comments.

References

  • [AH19] Jeffrey D. Achter and Everett W. Howe “Hasse–Witt and Cartier–Manin matrices: A warning and a request” In Arithmetic geometry: computation and applications, Contemp. Math 722, 2019, pp. 1–18
  • [CNP05] Gabriel Cardona, Enric Nart and Jordi Pujolàs “Curves of genus two over fields of even characteristic” In Mathematische Zeitschrift 250.1 Springer, 2005, pp. 177–201
  • [CQ05] Gabriel Cardona and Jordi Quer “Field of moduli and field of definition for curves of genus 2” In Computational Aspects of Algebraic Curves, 2005, pp. 71–83
  • [How08] Everett W. Howe “Supersingular genus-2 curves over fields of characteristic 3” In Computational arithmetic geometry 463, Contemp. Math. Amer. Math. Soc., Providence, RI, 2008, pp. 49–69
  • [Igu60] Jun-Ichi Igusa “Arithmetic variety of moduli for genus two” In Annals of Mathematics 72.3 Annals of Mathematics, 1960, pp. 612–649
  • [IKO86] Tomoyoshi Ibukiyama, Toshiyuki Katsura and Frans Oort “Supersingular curves of genus two and class numbers” In Compositio Mathematica 57.2, 1986, pp. 127–152
  • [KO87] Toshiyuki Katsura and Frans Oort “Families of supersingular abelian surfaces” In Compositio Mathematica 62.2, 1987, pp. 107–167
  • [Kob75] Neal Koblitz pp-adic variation of the zeta-function over families of varieties defined over finite fields” In Compositio Mathematica 31.2, 1975, pp. 119–218
  • [Mes91] Jean-François Mestre “Construction de courbes de genre 2 à partir de leurs modules” In Effective Methods in Algebraic Geometry Birkhäuser Boston, 1991, pp. 313–334
  • [MN07] Daniel Maisner and Enric Nart “Zeta functions of supersingular curves of genus 2” In Canadian Journal of Mathematics 59.2 Cambridge University Press, 2007, pp. 372–392
  • [Pri08] Rachel Pries “A short guide to p-torsion of abelian varieties in characteristic p” In Computational arithmetic geometry 463, Contemp. Math. Amer. Math. Soc., Providence, RI, 2008, pp. 121–129
  • [Yui78] Noriko Yui “On the Jacobian varieties of hyperelliptic curves over fields of characteristic p > 2” In Journal of Algebra 52.2, 1978, pp. 378–410