The smallest invariant factor of elliptic curves, and coincidences

Alexander Milner, Jack Shotton · 2026

Read The smallest invariant factor of elliptic curves, and coincidences by Alexander Milner, Jack Shotton (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-2604-21601.txt.

Introduction

Alexander Milner and Jack Shotton study the smallest invariant factor of points on an elliptic curve over the rationals. They ask when the constant giving the density of primes with a given factor stays positive. For curves without complex multiplication, they use the image of the adelic Galois representation to strengthen an earlier result by Kim. Positivity, they show, under the generalized Riemann hypothesis means infinitely many such primes exist. Experiments suggest the constant vanishes only when division fields coincide, and they document coincidence families from abelian division fields.

Catalog record

CREATORAlexander Milner, Jack Shotton
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53283
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.21601

Browse the full library · About the Tower of Babel

Full text

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

The smallest invariant factor of elliptic curves, and coincidences Alexander Milner, Jack Shotton Abstract . For an elliptic curve ℰ \mathcal{E} over ℚ \mathbb{Q} and a natural number j j , Cojocaru has shown that there is an explicit constant 𝒞 ℰ , j \mathcal{C}_{\mathcal{E},j} giving (under GRH) the density of primes p p of good reduction such that the smallest invariant factor of ℰ ⁡ ( 𝔽 p ) \mathcal{E}(\mathbb{F}_{p}) is j j . For ℰ \mathcal{E} without complex multiplication, we study the question of when 𝒞 ℰ , j \mathcal{C}_{\mathcal{E},j} is positive (a necessary and, on GRH, sufficient condition for there to be infinitely many such p p ), strengthening the main result of [ 14 ] . Our arguments are group-theoretic using the image of the adelic Galois representation of ℰ \mathcal{E} . Experimentally, 𝒞 ℰ , j \mathcal{C}_{\mathcal{E},j} appears to vanish only when there is a coincidence of division fields; we document a number of families of such coincidences arising from abelian division fields. 1. Introduction Let \mathcal{E} be an elliptic curve over \mathbb{Q} without complex multiplication, and let p be a prime of good reduction for \mathcal{E} . Then \mathcal{E}(\mathbb{F}_{p}) is an abelian group that is a product of at most two cyclic groups, and so we may write \mathcal{E}(\mathbb{F}_{p})\cong\mathbb{Z}/d_{\mathcal{E},p}\mathbb{Z}\times\mathbb{Z}/e_{\mathcal{E},p}\mathbb{Z} where d_{\mathcal{E},p},e_{\mathcal{E},p}\in\mathbb{N} with d_{\mathcal{E},p}\mid e_{\mathcal{E},p} . Thus d_{\mathcal{E},p}=1 if and only if \mathcal{E}(\mathbb{F}_{p}) is cyclic. We can ask how often, as p varies, d_{\mathcal{E},p} takes a given value, and a conditional answer is given by Cojocaru. Theorem 1.1 ( [ 4 ] , Theorem 2) . Let j be a positive integer. Under the generalised Riemann hypothesis (GRH) for the Dedekind zeta functions of the division fields \mathbb{Q}(\mathcal{E}[n]) of \mathcal{E} , \#\{p\leq x:d_{\mathcal{E},p}=j\}\sim\mathcal{C}_{\mathcal{E},j}\,Li(x) where \mathcal{C}_{\mathcal{E},j}=\sum_{k=1}^{\infty}\frac{\mu(k)}{[\mathbb{Q}(\mathcal{E}[jk]):\mathbb{Q}]}. (Here \mu is the Möbius function). In fact, the same statement also holds if \mathcal{E} has complex multiplication, and the GRH is then not required (see [ 16 ] ). We will only consider the case of elliptic curves without complex multiplication. As stated, this theorem does not address the issue of whether \mathcal{C}_{\mathcal{E},j}>0 — that is, whether (conditionally) there are infinitely many p with the smallest invariant factor of \mathcal{E}(\mathbb{F}_{p}) equal to j . 1 1 1 If \mathcal{C}_{\mathcal{E},j}=0 then it is easy to show d_{\mathcal{E},p}\neq j for all but finitely many p , see Proposition 2.8 below. Indeed, it is in fact possible that \mathcal{C}_{\mathcal{E},j}=0 , even if j=1 : if \mathcal{E} has full rational 2 -torsion, then 2\mid d_{\mathcal{E},p} for all p . For j=1 this is the only obstruction: Theorem 1.2 ( [ 3 , Section 6] ) . Suppose that \mathbb{Q}(\mathcal{E}[2])\neq\mathbb{Q} . Then \mathcal{C}_{\mathcal{E},1}>0. Remark 1.3 . They also prove this in the case that \mathcal{E} has complex multiplication. Remark 1.4 . In [ 10 , Theorem 1] it is shown unconditionally that, under the same hypothesis, \mathcal{E}(\mathbb{F}_{p}) is cyclic for infinitely many p . This also implies Theorem 1.2 , by Proposition 2.8 below. Remark 1.5 . In fact, there is a small error in [ 3 , Section 6] , which we explain and correct in Section 5 . Our main result also gives an alternative proof of Theorem 1.2 . In [ 14 ] , Kim proves 2 2 2 Kim states that this result is conditional on the GRH, but in fact his result only requires Theorem 1.2 which holds unconditionally if \mathcal{C}_{\mathcal{E},j} is defined as an infinite sum. that the conclusion of Theorem 1.2 holds as long as j is coprime to the adelic level m_{\mathcal{E}} of \mathcal{E} , the minimal integer such that the image of the adelic Galois representation of \mathcal{E} is the full pre-image of a subgroup of GL_{2}(\mathbb{Z}/m_{\mathcal{E}}\mathbb{Z}) . If A(\mathcal{E}) is the product of the primes such that the p -adic Galois representation associated to \mathcal{E} is not surjective and N_{\mathcal{E}} is the conductor of \mathcal{E} , then the prime factors of m_{\mathcal{E}} divide 2N_{\mathcal{E}}A(\mathcal{E}) ; see Proposition 2.9 . 3 3 3 This definition of A(\mathcal{E}) has the same prime factors as that in [ 5 ] and differs from that of [ 2 ] by (perhaps) factors of 30. It is unclear what definition of A(\mathcal{E}) is used in [ 14 ] , but what is actually proved is the result as stated here in terms of the adelic level. In Corollary 3.8 , we prove the following generalisation of Kim’s result. Theorem 1.6 . Suppose that j is coprime to 2A(\mathcal{E}) and that \mathbb{Q}(\mathcal{E}[2])\not\subset\mathbb{Q}(\mathcal{E}[j]) . Then \mathcal{C}_{\mathcal{E},j}>0. In fact, we prove something stronger which allows us to handle some cases where j and 2A(\mathcal{E}) have common factors. See Theorem 3.2 and the examples that follow. It is an interesting question, related to ‘entanglements’ and ‘coincidences’ of division fields, to classify all pairs \mathcal{E},j with \mathcal{C}_{\mathcal{E},j}=0 . We give some examples in Section 4 . These examples all arise from coincidences \mathbb{Q}(\mathcal{E}[j])=\mathbb{Q}(\mathcal{E}[jp]) with p=2 or p=3 and we conjecture that this is the only way that \mathcal{C}_{\mathcal{E},j} can vanish. Coincidences of division fields have recently been studied in (for example) [ 6 ] , [ 20 ] , and [ 7 ] . In Section 4 we define a p -coincidence to be one of the form \mathbb{Q}(\mathcal{E}[j])=\mathbb{Q}(\mathcal{E}[jp]) and conjecture that, for non-CM elliptic curves over \mathbb{Q} , p -coincidences occur only for p=2 and p=3 . In Section 4.1 we document some infinite families of examples of p -coincidences stemming from elliptic curves with abelian mod n Galois representations for n=2,3,4 , and 8 ; in Section 4.2 we give a few sporadic examples that we found interesting. We restrict attention to elliptic curves over \mathbb{Q} , although many of our results would continue to hold over number fields containing no nontrivial abelian extension of \mathbb{Q} . The vanishing of \mathcal{C}_{\mathcal{E},1} for elliptic curves over number fields has been studied in [ 1 ] , while coincidences over number fields are the subject of [ 20 ] . 1.1. Notation We summarise some of the key notation used throughout this paper: • We let \mathcal{E} be an elliptic curve over \mathbb{Q} without complex multiplication. • We write N_{\mathcal{E}} for the conductor of \mathcal{E} , and m_{\mathcal{E}} , A(\mathcal{E}) for its adelic level and Serre constant (defined in Section 2 below). • We let \rho_{n} (for n\in\mathbb{N} ) and \rho be the mod n and adelic Galois representations associated to \mathcal{E} (defined in Section 2 below). • We write \mu for the Möbius function. • For a positive integer b , let \rad(b) be its radical and let \langle b\rangle be the set of integers all of whose prime factors divide b . • For a group G let [G,G] be its commutator subgroup. • The cardinality of the group GL_{2}(\mathbb{Z}/n\mathbb{Z}) is given by \psi(n)=|GL_{2}(\mathbb{Z}/n\mathbb{Z})|=n^{4}\prod_{p|n}\left(1-\frac{1}{p}\right)\left(1-\frac{1}{p^{2}}\right). 1.2. Acknowledgments The first author was funded by a London Mathematical Society Undergraduate Research Bursary and by Durham University. We thank Harris Daniels for useful conversations on the topic of this paper. 2. Galois images Associated to \mathcal{E} , we have the adelic Tate module T(\mathcal{E})=\varprojlim_{n}\mathcal{E}[n]. It is a free \hat{\mathbb{Z}} -module of rank 2, and we will fix a choice of basis. We then obtain a continuous representation \rho:\mathrm{Gal}(\bar{\mathbb{Q}}/\mathbb{Q})\to GL_{2}(\hat{\mathbb{Z}}) where \mathrm{Gal}(\bar{\mathbb{Q}}/\mathbb{Q}) is the absolute Galois group

Read the complete text