2
$\begingroup$

Let $E/K$ be a number field. For quadratic field extensions $L/K$, it is known that $\operatorname{Ш}(E/L)[2]$ can be arbitrarily large (cf. P. L. Clark and S. Sharif, "Period, index and potential Sha," Algebra and Number Theory 4, No. 2, 151–174, 2010).

What is known about elements of order 4? More specifically, for a quadratic extension $L/K$, can the number of order 4 elements in ${\operatorname{Ш}}(E/L)$ grow arbitrarily large, or is it bounded? If any results on this question are known, I would appreciate references.

$\endgroup$
3
  • 3
    $\begingroup$ ANT's landing page, the PDF itself, and MSN all think that the title of the Clark–Sharif paper is "Period, index and potential, III". However, the arXiv (and the content of the paper itself!) makes it clear that the title of the paper is supposed to be as you've said. What a strange gaffe by the publisher. $\endgroup$
    – LSpice
    Commented Nov 13 at 15:04
  • 4
    $\begingroup$ It follows from work of Alex Smith that the 4-Selmer rank can be arbitrarily small with the 8- Selmer rank vanishing and this implies the algebraic rank is 0 and hence the 4-Sha rank is large. $\endgroup$
    – Will Sawin
    Commented Nov 13 at 17:24
  • $\begingroup$ @Will Sawin Thank you very much. I am having trouble identifying which reference and theorem number is relevant. I thought it might be Theorem 4.3 of arxiv.org/abs/1607.07860, is this the result you had in mind? $\endgroup$
    – Duality
    Commented Nov 18 at 15:55

1 Answer 1

3
$\begingroup$

This follows from Theorem 1.5 of Alex Smith's paper "The distribution of $\ell^\infty$-Selmer groups in degree $\ell$ twist families I" which states

Suppose $A/\mathbb Q$ is an elliptic curve that fits into eitehr case (1) or (3) of Assumption (1.1). Given any nonicnreasing sequence $r_2 \geq r_4 \geq \dots \geq r_{2^k} \geq \dots$ of nonnegative integers, we have $$ \lim_{H \to\infty} \frac{ \# \{d \in \mathbb Z^{\neq 0} \colon |d|<H \textrm { and } r_{2^K} (A^d) = r_{2^k} \textrm{ for all }k\geq 1\}}{2H} = P^{\textrm{Alt}}(r_2\mid \infty) \cdot \prod_{k=2}^{\infty} P^{\textrm{Alt} }( r_{2^k} \mid r^{2^{k-1}})$$

Choose $r_2 = 2N$ for a large $N$, $r_4 =2N$, $r_8=0$, and thus $r_{2^k}=0$ for all $K>3$. The definition of $P^{\textrm{Alt}}(j\mid n))$ makes it clear that this probability is nonzero as long as $j$ and $n$ have the same parity and $1$ if $j=n=1$, and it can also be checked that $P^{\textrm{Alt}}(j\mid \infty))$ is nonzero for all nonnegative integers $j$, so the product on the right-hand side is zero. Hence the limit on the left-hand side is nonzero. In particular, the numerator on the left-hand side must be nonzero for some $H$, and therefore there exists a $d$ such that the $2$-Selmer rank $r_{2}$ of the quadratic twist $A^d$ is $2N$, the $4$-Selmer rank is $2N$, and the $8$-Selmer rank is $0$.

This implies the algebraic rank is $0$, and so the 4-Sha rank of the Tate-Shafarevich group is at least $2N-2$ (the loss of $2$ coming from the possible $2$-torsion of $A$, though we may choose $A$ to have no $2$-torsion).

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .