Construction of generically ordinary families of hyperelliptic curves

Hui June Zhu · 2026

Read Construction of generically ordinary families of hyperelliptic curves by Hui June Zhu (2026) 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-2606-31783.txt.

Introduction

Zhu settles a conjecture of Katz about families of hyperelliptic curves and when they are generically ordinary. A curve in positive characteristic is ordinary when its Jacobian has only the mildest Newton polygon slopes. Katz's 2018 lecture proposed that one specific family stays generically ordinary for all large enough primes, with a quadratic bound on the prime.

Zhu proves more. For every genus at least two and every nonzero algebraic integer alpha, the family with alpha in place of the fixed coefficient is generically ordinary at every prime above an explicit bound. The bound differs for odd and even degree, and examples show it cannot be weakened. Katz's original family follows as the special case where alpha equals minus the degree.

Catalog record

CREATORHui June Zhu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.31783

Browse the full library · About the Tower of Babel

Full text

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

Construction of generically ordinary families of hyperelliptic curves Hui June Zhu Abstract . Katz conjectured in a 2018 lecture that the family of curves y 2 = x d − d ​ x + t y^{2}=x^{d}-dx+t over the t t -line is generically ordinary for all sufficiently large primes p p . We prove that, for every g ≥ 2 g\geq 2 and every nonzero algebraic integer α \alpha , the genus- g g families 𝒞 α : y 2 = x d + α ​ x + t \mathcal{C}_{\alpha}:y^{2}=x^{d}+\alpha x+t where d ∈ { 2 ​ g + 1 , 2 ​ g + 2 } d\in\{2g+1,2g+2\} are generically ordinary at every prime p > P + ​ ( d ) p>P^{+}(d) , provided that α \alpha is nonzero modulo every prime above p p . The bound P + ​ ( d ) = d 2 − 4 ​ d + 2 P^{+}(d)=d^{2}-4d+2 if d d is odd, and P + ​ ( d ) = ( d 2 − 3 ​ d + 2 ) / 2 P^{+}(d)=(d^{2}-3d+2)/2 if d d is even. 1. Introduction and statement of results Let \mathcal{C} be a smooth projective curve of genus g over a field of characteristic p>0 . The curve \mathcal{C} is ordinary if, in terms of the zeta function of \mathcal{C} , the middle coefficient of the numerator is prime to p . For an elliptic curve, this middle coefficient is the classical Hasse invariant . In higher genus, the analogous object is the Hasse–Witt matrix, or its dual the Cartier–Manin matrix, and ordinarity is detected by the nonvanishing of its determinant. Equivalently, \mathcal{C} is ordinary if its Jacobian is an ordinary abelian variety, that is, its Newton polygon has only slopes 0 and 1 . The ordinary locus in the moduli space of curves \mathcal{M}_{g} is open, dense, geometrically irreducible, and the corresponding homomorphism \pi_{1}\rightarrow\mathrm{GL}_{g}(\mathbb{F}_{p}) is surjective, see [ FG04 , FC90 ] . Analogous statements hold for the ordinary locus in the hyperelliptic locus \mathcal{H}_{g} (see [ GP05 , AP08 ] , and also [ PZ12 ] for the characteristic- 2 case). Given a one-parameter family of hyperelliptic curves over a number field, with “large monodromy”— that is, whose topological monodromy is of finite index in \mathrm{Sp}_{2g}(\mathbb{Z}) — does it always meet the ordinary locus in \mathcal{H}_{g} for every prime p ? Katz’s family \mathcal{K}:y^{2}=x^{d}-dx+t, over (the smooth locus of) the t -line has large monodromy (as shown in [ Kat14 ] ). However, \mathcal{K} may fail to be generically ordinary for small p (see [ Kat18 ] ). Conjecture 1.1 (Katz 2018) . Let d\in\{2g+1,2g+2\}, and g\geq 2 . There exists a quadratic polynomial P^{+}(d) such that the family \mathcal{K} is generically ordinary at p for all p>P^{+}(d) . Remark (The elliptic curve case) . In the case g=1 (that is d=3,4 ) the short Weierstrass family y^{2}=x^{d}+\alpha x+t over the t -line is generically ordinary at every prime p>3 and p\nmid\alpha . The case g=2 can be proved by Will Sawin’s argument [ Saw16 ] . This paper settles this conjecture for all g\geq 2 . We use the following standing notation throughout the paper: Notation 1.2 . Let d\in\{2g+1,2g+2\} and g\geq 2 . Let \alpha\neq 0 be an algebraic integer and let \mathcal{C}_{\alpha}(t):\quad y^{2}=x^{d}+\alpha x+t. Let D_{\alpha}(t) denote the discriminant \mathrm{disc}_{x}(x^{d}+\alpha x+t) . Denote \mathcal{S}^{\circ}\coloneqq\mathrm{Spec}\mathcal{O}_{\mathbb{Q}(\alpha)}[1/2,t,D_{\alpha}(t)^{-1}]. Let P^{+}(d)=\begin{cases}d^{2}-4d+2&\mbox{ if }d=2g+1\\ (d^{2}-3d+2)/2&\mbox{ if }d=2g+2.\end{cases} Theorem 1.3 . Let \alpha be a nonzero algebraic integer and let F=\mathbb{Q}(\alpha) . Then for every prime p>P^{+}(d) and p\nmid N_{F/\mathbb{Q}}(\alpha) , the family \mathcal{C}_{\alpha} of genus- g curves \displaystyle\mathcal{C}_{\alpha} \displaystyle\longrightarrow \displaystyle\mathcal{S}^{\circ} is generically ordinary at each prime \mathfrak{p}\mid p of F . Remark 1.4 . Notice that the Katz family \mathcal{K}=\mathcal{C}_{-d} is a special case; hence Conjecture 1.1 follows from Theorem 1.3 . Remark 1.5 . The bound P^{+}(d) cannot be weakened to p\geq P^{+}(d) in the theorem. The data provided by Katz in [ Kat18 ] shows that the generic ordinarity fails for p=P^{+}(d) in the following cases: (d,p)=(21,359),(31,839),(51,2399), and further computations suggest the same for (65,3967),(105,10607). In all of these cases, d is odd and p\equiv 2\bmod d . See Corollary 4.5 for a refinement of the bound in some cases. Remark 1.6 . Katz showed in [ Kat14 ] that for any Morse polynomial h(x) of degree d the family of curves y^{2}=h(x)+t has large monodromy. It is straightforward to verify by definition that h(x)=x^{d}+\alpha x in characteristic 0 is Morse if and only if \alpha is nonzero. In characteristic p it is Morse if and only if p\nmid d(d-1) and \alpha is nonzero. This shows that \mathcal{C}_{\alpha} has topological monodromy of finite index in \mathrm{Sp}_{2g}(\mathbb{Z}) . Hence Theorem 1.3 shows that part of Conjecture 1 of [ Kat18 ] holds for this family \mathcal{C}_{\alpha} . For g\geq 2 we have P^{+}(d)>d . Every prime p>P^{+}(d)>d implies p\nmid d(d-1) . At every prime \mathfrak{p}\mid p for which \alpha\not\in\mathfrak{p} , the polynomial x^{d}+\alpha x\bmod\mathfrak{p} is Morse. This implies, by Katz’s argument [ Kat14 ] , that \mathcal{C}_{\alpha} at \mathfrak{p} has large monodromy. Our result lies in the tradition of Miller’s paper [ Mil72 ] . Miller exhibited two families over the t -line \begin{cases}y^{2}&=x^{2g+1}+tx^{g+1}+x\quad\mbox{ if $p\nmid g$}\\ y^{2}&=x^{2g+2}+tx^{g+1}+1\quad\mbox{ if $p\mid g$}\end{cases} are generically ordinary. The first one was in fact proposed by J. Lubin and N. Katz (unpublished). For Miller’s families, the coefficient matrix has the support of a permutation matrix. By contrast, the matrix for \mathcal{K} is generally not monomial and can contain many nonzero entries, so its determinant need not reduce to a single product. A different and new approach is needed. This paper is organized as follows. We present in Section 2 how to evaluate a determinant \Delta and prove that prime factors of \Delta are bounded above by P^{+}(d) . In Section 3, we define the Hasse–Witt polynomial H_{p,\alpha} , of \mathcal{C}_{\alpha} over \mathcal{O}_{\mathbb{Q}(\alpha)} , and show that it is nonzero at each prime \mathfrak{p}\mid p for p>P^{+}(d) and p\nmid N_{\mathbb{Q}(\alpha)/\mathbb{Q}}(\alpha) . This is accomplished by showing its leading coefficient is, up to a p -adic unit factor, congruent to \Delta mod p . In Section 4 we prove the generic ordinarity of the family \mathcal{C}_{\alpha}(t) . Theorem 4.1 proves Theorem 1.3 . We then deduce Katz’s Conjecture 1.1 in Corollary 4.2 . As a natural application we also derive a similar result for a two-parameter family over the affine plane in Corollary 4.3 . 2. Prime factors of a determinant 2.1. Product formula for a binomial determinant The goal of this subsection is to produce a product formula for a determinant \Delta in Proposition 2.4 . The core technique in the proof is to reduce it to a Vandermonde-type determinant, which has a natural product formula. For any integer 1\leq r\leq g-1 we define a positive rational number (1) \lambda_{r}=\frac{d(2r+1)}{2(d-1)}. Lemma 2.1 . For 1\leq r\leq g-1 , \lambda_{r} is not an integer. Proof. First notice 3\leq 2r+1<d-1 . If d is odd, then (2(d-1),d)=1 . Since 2(d-1)\nmid(2r+1) , \lambda_{r}\not\in\mathbb{Z} . If d is even, then \lambda_{r}=\frac{\frac{d}{2}(2r+1)}{d-1} . Since (d-1,\frac{d}{2})=1 and d-1\nmid 2r+1 , \lambda_{r}\not\in\mathbb{Z} . ∎ Lemma 2.2 . Let P_{j}(T)=\prod_{r=1}^{j-1}(T-r)\prod_{r=j}^{g-1}(T-\lambda_{r}). Let P(x_{1},\dots,x_{g})=\det_{1\leq i,j\leq g}P_{j}(x_{i}). Then P(x_{1},\dots,x_{g})=\displaystyle\prod_{i<j}(x_{j}-x_{i})\displaystyle\prod_{r=1}^{g-1}\prod_{k=1}^{r}(k-\lambda_{r}). Proof. The determinant P

Read the complete text