Visible 2-torsion in the Tate-Shafarevich group of an elliptic curve

Tom Fisher · 2026

Read Visible 2-torsion in the Tate-Shafarevich group of an elliptic curve by Tom Fisher (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-2605-31488.txt.

Introduction

Fisher proves that two 2-torsion elements drawn from an elliptic curve's Tate-Shafarevich group show up together inside a single abelian surface. Earlier work had shown this only for a single element on its own. His finding also clarifies early remarks on visibility by Cremona and Mazur, explaining part of what they first observed.

Catalog record

CREATORTom Fisher
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE42767
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.31488

Browse the full library · About the Tower of Babel

Full text

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

Visible 2-torsion in the Tate-Shafarevich group of an elliptic curve Tom Fisher Abstract . We show that every pair of 2 2 -torsion elements in the Tate-Shafarevich group of an elliptic curve are visible in the same abelian surface. This was previously only known for a single 2 2 -torsion element. Our result explains some of the original observations on visibility in the paper of Cremona and Mazur. 1. Introduction Let E be an elliptic curve defined over a number field k . A torsor (or principal homogeneous space) under E is a pair (C,\mu) where C is a smooth curve of genus one, and \mu:E\times C\to C is morphism (both defined over k ) inducing a simply transitive action on {\overline{k}} -points. The Weil-Châtelet group H^{1}(k,E) parametrises the torsors under E up to k -isomorphism. The Tate-Shafarevich group \mbox{\wncyr Sh}(E/k)=\ker\left(H^{1}(k,E)\to\prod_{v}H^{1}(k_{v},E)\right) is the subgroup of torsors that are everywhere locally soluble. This group is conjecturally finite, and when finite, Cassels [ 7 ] showed it has order a square. Its non-zero elements correspond to genus one curves with Jacobian E that are counterexamples to the Hasse principle. Mazur introduced the notion of visibility as a way to study elements of \mbox{\wncyr Sh}(E/k) . Definition 1.1 . An element \xi\in H^{1}(k,E) is visible in an abelian variety A if there is an inclusion of abelian varieties \iota:E\to A such that \xi is in the kernel of the induced map \iota_{*}:H^{1}(k,E)\to H^{1}(k,A). Equivalently, \xi is visible in A if the corresponding torsor may be realised as a coset of E inside A . Cremona and Mazur [ 13 , 2 ] gave examples, for each n\in\{2,3,4,5\} , of elliptic curves E/{\mathbb{Q}} where all of \mbox{\wncyr Sh}(E/{\mathbb{Q}})\cong({\mathbb{Z}}/n{\mathbb{Z}})^{2} is visible in the same abelian surface. Mazur [ 20 ] showed that every 3-torsion element in \mbox{\wncyr Sh}(E/k) is visible in an abelian surface. Klenke [ 19 ] showed that every 2-torsion element in H^{1}(k,E) is visible in an abelian surface. An argument using restriction of scalars, first recorded by Agashe and Stein [ 1 ] , shows that every n -torsion element in \mbox{\wncyr Sh}(E/k) is visible in an abelian variety of dimension n . Fisher [ 15 , 16 ] gave examples of 6 and 7 -torsion elements in \mbox{\wncyr Sh}(E/{\mathbb{Q}}) that are not visible in an abelian surface, and also of 7 -torsion elements that are visible in an abelian threefold. It follows from the restriction of scalars argument that \mbox{\wncyr Sh}(E/k) is finite if and only if all its elements are visible in the same abelian variety. We prove the following. Theorem 1.2 . Every pair of 2 -torsion elements in \mbox{\wncyr Sh}(E/k) are visible in the same abelian surface. The proof works by showing that certain complete intersections of two quadrics in {\mathbb{P}}^{5} satisfy the smooth Hasse principle. In Section 2 we rephrase Theorem 1.2 as a statement about 2 -Selmer groups. In Section 3 we introduce the relevant quadric intersections and describe some of their geometry. In Sections 4 and 5 we give two different proofs that these satisfy the smooth Hasse principle. We complete the proof of Theorem 1.2 in Section 6 . In the final two sections we give some examples and discuss a variant of the proof. We thank Shiva Chidambaram for sharing some data in connection with [ 4 ] which prompted us to revisit this problem. A Magma [ 6 ] file checking some of the calculations and examples in this paper is available on the author’s webpage. 2. The 2 -Selmer group Let E be an elliptic curve defined over a number field k , and let n\geqslant 2 be an integer. We write E[n] for the n -torsion subgroup. The n -Selmer group S^{(n)}(E/k)=\ker\left(H^{1}(k,E[n])\to\prod_{v}H^{1}(k_{v},E)\right) parametrises the n -coverings of E that are everywhere locally soluble. It is the middle term in the Kummer exact sequence 0\longrightarrow E(k)/nE(k)\longrightarrow S^{(n)}(E/k)\longrightarrow\mbox{\wncyr Sh}(E/k)[n]\longrightarrow 0. Lemma 2.1 . Let F/k be a second elliptic curve and \pi:E[n]\to F[n] an isomorphism of {\operatorname{Gal}}({\overline{k}}/k) -modules. If \xi,\eta\in S^{(n)}(E/k) are in the kernel of the composition H^{1}(k,E[n])\stackrel{{\scriptstyle\pi_{*}}}{{\longrightarrow}}H^{1}(k,F[n])\longrightarrow H^{1}(k,F) then their images in \mbox{\wncyr Sh}(E/k) are visible in the same abelian surface. Proof. We identify E[n] and F[n] via \pi and write \Delta=E[n]=F[n] . We embed \Delta diagonally in E\times F and let A=(E\times F)/\Delta . There is a commutative diagram \textstyle{H^{1}(k,\Delta)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \textstyle{H^{1}(k,F)\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \textstyle{H^{1}(k,E)\ignorespaces\ignorespaces\ignorespaces\ignorespaces} \textstyle{H^{1}(k,A)\hbox to0.0pt{.\hss}} We are given that \xi and \eta are in the kernel of the top horizontal map. By the commutativity of the diagram, their images in \mbox{\wncyr Sh}(E/k) are in the kernel of the bottom horizontal map, and so visible in A . ∎ We now take n=2 . Elements of the 2 -Selmer group S^{(2)}(E/k) may be represented by binary quartics. See for example [ 3 , 5 , 10 , 12 , 26 ] . Definition 2.2 . A binary quartic is a homogeneous polynomial f\in k[x,z] of degree 4 . The invariants of the binary quartic f(x,z)=ax^{4}+bx^{3}z+cx^{2}z^{2}+dxz^{3}+ez^{4} are (1) \displaystyle I \displaystyle=12ae-3bd+c^{2}, \displaystyle J \displaystyle=72ace-27ad^{2}-27b^{2}e+9bcd-2c^{3}. A binary quartic is non-singular if it has no repeated roots (i.e. linear factors) over {\overline{k}} , equivalently if its discriminant 4I^{3}-J^{2} is non-zero. A non-singular binary quartic f defines a smooth projective curve of genus one with affine equation y^{2}=f(x,1) . The projective curve is defined either by gluing affine pieces, or directly as the curve with equation y^{2}=f(x,z) in the weighted projective plane {\mathbb{P}}(1,2,1) . This curve is a 2 -covering of its Jacobian (2) E_{I,J}:\quad y^{2}=x^{3}-27Ix-27J. A binary quartic is soluble if the corresponding genus one curve has a k -point, and everywhere locally soluble if it has a k_{v} -point for all places v . Definition 2.3 . Let K/k be any field extension. Binary quartics f and g are properly K -equivalent if f(x,z)=\frac{1}{(ps-qr)^{2}}g(px+rz,qx+sz) for some p,q,r,s\in K with ps-qr\not=0 . Binary quartics that are properly k -equivalent have the same invariants. The converse is true for non-singular binary quartics over an algebraically closed field. We may write any elliptic curve E defined over k in the form E_{I,J} for some I,J\in k with 4I^{3}-J^{2}\not=0 . The 2 -Selmer group S^{(2)}(E_{I,J}/k) may then be identified with the set of proper k -equivalence classes of everywhere locally soluble binary quartics with invariants I and J . The genus one curve defined by a non-singular binary quartic f is a double cover of {\mathbb{P}}^{1} ramified over 4 points. These 4 points are the fibre of the 2 -covering above the identity on E , and form a torsor under E[2] . In Section 6 we will use the fact that the class of f in H^{1}(k,E[2]) is the class of this torsor. 3. Quadric intersections We work over a field k with \operatorname{char}(k)\not=2,3 , with algebraic closure {\overline{k}} . Let F(x_{1},\ldots,x_{n}) and G(x_{1},\ldots,x_{n}) be a pair of quadratic forms in n variables. Let A and B be the n\times n symmetric matrices of second partial derivatives of F and G . Then \det(\lambda A+\mu B) is a binary form of degree n . By abuse of notation we write this binary form as \det(\lambda F+\mu G) . The next two lemmas are well known. The first is [ 23 , Proposition 2.1] . Lemma 3.1

Read the complete text