0
$\begingroup$

Let $$L(C,s)=\sum_{n=1}^\infty \frac{a_n}{n^s}$$ be the Dirichlet series of the Hasse--Weil L-function of an elliptic curve $C$ over $ℚ$. The modularity theorem implies that $L(C,s)$ is the $L$-function of a holomorphic cusp form for a congruence subgroup and it is entire function and have a holomorphic continuation. The order of vanishing (the analytic rank) of $L(C,s)$ at $s=1$ is denoted by $m$ (the minimal integer $m≥0$ such that $L(C,s)^{(m)}(1)≠0$) and the algebraic rank of $C(ℚ)$ is denoted by $r$.

My question is: Is there is a known relation or expression containing the algebraic rank $r$? I am looking for any kind of relations (equalilities, inequalities, etc...). In particular, I have this one: $$2^{r}=(|Imα||Imα′|)/4$$

for some well defined maps $α$ and $α′$. See: Rational Points on Elliptic Curves by Alexandru Gica (2006).

$\endgroup$

1 Answer 1

7
$\begingroup$

First, is the reference supposed to be to the book "Rational Points on Elliptic Curves" by John Tate and myself? The formula $$2^r=\frac{\#\alpha(\Gamma)\cdot\#\overline\alpha(\Gamma)}{4}$$ appears at the bottom of page 91 of our book. So maybe Dr. Gica's notes are based on that?

Anyway, this is all a special case of the standard descent on an elliptic curve. For an isogeny $\phi:C\to C'$ there is an injection $$C'(K)/\phi(C(K)) \hookrightarrow H^1(\hbox{Gal}(\overline K/K),\hbox{Ker}(\phi)).$$ If $\hbox{Ker}(\phi)$ is defined over $K$ and is (say) cyclic of order $n$, or more generally if the kernel is isomorphic as a Galois module to $\mathbf{\mu}_n$, then $H^1(\hbox{Gal}(\overline K/K),\hbox{Ker}(\phi))\cong K^*/(K^*)^n$ by Hilbert Theorem 90. So one gets $$\alpha:C'(K)/\phi(C(K)) \hookrightarrow K^*/(K^*)^n.$$ Doing the same thing with the dual isogeny gives $$\alpha':C(K)/\hat\phi(C'(K)) \hookrightarrow K^*/(K^*)^n.$$ Combining these and a little information about rational $n$-torsion, one can get a formula relating $2^{\text{rank} C(K)}$ to $\#\text{Image}(\alpha)\cdot\#\text{Image}(\alpha')$.

Of course, the hard part is computing those images, especially if the Tate-Shafarevich group has non-trivial $n$-torsion.

BTW, this is all explained in Chapter 10 of my book The Arithmetic of Elliptic Curves.

$\endgroup$
2
  • $\begingroup$ @ Joe Silverman: Thank you very much for the nice answer. So the given formula is true for all elliptic curves without any restriction. $\endgroup$ Jun 1, 2014 at 8:47
  • 1
    $\begingroup$ @Shpigle It depends what you mean "without any restriction". If you take an arbitrary elliptic curve $E/K$ and want $\alpha$ and $\alpha'$ to have values in $K^*/(K^*)^n$, you may need to replace $K$ by an extension field. So for example, for $n=2$, you need $E(K)$ to contain a $K$-rational $2$-torsion point. But if you don't mind $\alpha$ and $\alpha'$ having images in the Selmer group, then it should work in complete generality. $\endgroup$ Jun 1, 2014 at 12:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.