Let $E$ be an elliptic curve over $\mathbb{Q}$. As proved by Wiles et al., its $L$-series $L(E, s)$ is entire. Set $r := \mathrm{ord}_{s = 1} L(E, s)$, a value conjecturally equal to $\mathrm{dim}_{\mathbb{Q}} (E(\mathbb{Q}) \otimes_{\mathbb{Z}} \mathbb{Q})$. There is a $c \in \mathbb{C}^\times$ such that $L(E, s) \sim c \cdot (s - 1)^r$ as $s \rightarrow 1$. Birch and Swinnerton-Dyer predict a great deal about $c$, in particular, that

  1. $c \in \mathbb{R}$,
  2. $c \in \mathbb{R}_{> 0}$,
  3. $c/\Omega_E \in \mathbb{Q}$ if $r = 0$, where $\Omega_E$ is the volume of $E(\mathbb{R})$ computed with respect to "the" Neron differential of $E$, and
  4. $c/\Omega_E \in \mathbb{Q}_{> 0}$ if $r = 0$.

Which of these predictions are known to hold?

Partial answers and answers in special cases (e.g., in the case when $E$ has potential complex multiplication) are very welcome!

  • 1
    $\begingroup$ 1 and 2 are known. 3 and 4 are not right unless $r=0$, because the height pairing doesn't take rational values. If $r=0$, then maybe 3 and 4 are known by Kolyvagin, but I am not sure. $\endgroup$ – Felipe Voloch Jul 3 '14 at 2:13
  • $\begingroup$ Thanks for your comment! Could you give a reference where 1. and 2. are discussed (or a sketch how the argument goes)? You are right that 3. and 4. are incorrect as stated; I will add the $r = 0$ condition to them. $\endgroup$ – Question Mark Jul 3 '14 at 2:34
  • 3
    $\begingroup$ The identity $L(E,\bar{s})=\overline{L(E,s)}$, which is clear for $\Re(s)>3/2$, analytically continues. $\endgroup$ – Felipe Voloch Jul 3 '14 at 3:00
  • 1
    $\begingroup$ I think #2 in the $r=0$ case is Waldspurger in essence, and then a long series of papers concluding with Guo, will track reference (cf bottom pg 722 of Iwaniec/Sarnak perspectives paper). "On the positivity of the central value of automorphic $L$-functions for $GL(2)$", Duke Math 83 1996, 1-18. As IwSa says, Guo uses the relative trace formula, rather than $\Theta$-series. projecteuclid.org/euclid.dmj/1077244251 $\endgroup$ – NAME_IN_CAPS Jul 3 '14 at 3:32
  • 2
    $\begingroup$ I don't think Kolyvagin can say anything about the sign of the L-value. His argument is prime-by-prime and for most primes the formula of BSD then holds. But that will always miss the sign it seems to me. $\endgroup$ – Chris Wuthrich Jul 3 '14 at 9:11

Let me summarise the comments above that give a full answer (correct me if I am wrong).

  1. The analytic continuation of $L(E,\bar{s})=\overline{L(E,s)}$ shows that $c\in\mathbb{R}$.

  2. If $r=0$, the fact that $c>0$ is proven in On the positivity of the central value of automorphic L-functions for GL(2) Duke Math 83 1996, 1-18. It would follow from the generalised Riemann hypothesis, too. For $r=1$, it follows from Gross-Zagier and the case $r=0$. I don't know about $r>1$.

  3. That $L(E,1)/\Omega_E$ is a rational number is a consequence of the theorem of Manin-Drinfeld on modular symbols.

  4. Is just 2+3. Note that $\Omega_E$ is defined to be positive, as it is the least positive real period of a Néron differential on $E$ (or twice that depending on your normalisation).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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