# Is there an analog of the Birch/Swinnerton-Dyer conjecture for abelian varieties in higher dimensions?

I am wondering if there is a multi-dimensional analog of the Birch/Swinnerton-Dyer (BSD) conjecture. The recent famous result inching toward resolution of that conjecture is:

Bhargava, Manjul, and Christopher Skinner. "A positive proportion of elliptic curves over $\mathbb{Q}$ have rank one." 2014. arXiv link.

As I understand it: "66.48% of elliptic curves satisfy the (rank part of the) BSD Conjecture."

I see there are books such as

Miyaoka, Joichi, and Thomas Peternell, eds. Geometry of higher dimensional algebraic varieties. Vol. 26. Springer, 1997.

that extend elliptic curves to "higher dimensional algebraic varieties." So,

Q. Is there an analog of the Birch/Swinnerton-Dyer conjecture for abelian varieties in higher dimensions?

Obviously this is a naive question. Thanks for educating me!

• Yes. J. Tate, On the conjectures of Birch and Swinnerton-Dyer and a geometric analog. Séminaire Bourbaki, Vol. 9, Exp. No. 306,(1965) 415–440 Sep 24, 2014 at 23:47
• @JoeSilverman I am not sure. Ask him, next time you see him. Sep 25, 2014 at 0:15
• Yes, the generalization to abelian varieties was due to Tate (in his February 1966 Bourbaki talk). Now, it would not be so difficult to make the generalization, but at the time few understood heights on abelian varieties. And only Tate was capable of showing that the statement is compatible with isogenies, which is convincing evidence that he got the factors correct.
– anon
Sep 25, 2014 at 11:58
• In your question, you ask about Abelian Varieties and on that direction Joe Slverman's answer is indeed very complete (and accepted, as I see). But "most" of BSD has been generalized to more general varieties, without the need of any group structure, mainly by Bloch and Kato. So I wonder whether you are intentionally focusing on Abelian Varieties or if you were interested in generalizations to any sort of higher domensional gadhets. Sep 25, 2014 at 22:37
• Bloch calls the generalised rank equality for smooth projective algebraic varieties a "Recurring Fantasy" (which ranks above Idle Speculation) in eudml.org/doc/152633 Sep 26, 2014 at 5:03

Yes, there is a (well-known) analogue for abelian varieties of all dimensions. I was going to suggest looking at the Wikipedia article on the Birch/Swinnerton-Dyer conjecture and was surprised to see that it only talks about the elliptic curve case. (Clearly an opportunity for someone to add a section on generalizations.) There are also further major generalizations due to Tate (concerning the order of vanishing), Beilinson (concerning the transcendental factor in the leading coefficient, analogous to the real period) and Bloch and Kato (concerning the form of the algebraic part of the leading coefficient, analogous to the SHA*Regulator/(torsion)$^2\cdot (c_p$ factors)). [And I'm probably missing some other names here.]
The abelian variety version is undoubtedly described in many places, but since I have a copy handy, I'll mention the statement over $\mathbb Q$ is in my book with Hindry, Diophantine Geometry: An Introduction, Conjecture F.4.1.6 (page 462). It says:
Conjecture Let $A/\mathbb Q$ be an abelian variety, and assume that $L(A/\mathbb Q,s)$ has an analytic continuation to $\mathbb C$. Then $$\operatorname{ord}_{s=1} L(A/\mathbb Q,s) = \operatorname{rank} A(\mathbb Q)$$ and if the rank is $r$, then $$\lim_{s\to1} \frac{L(A/\mathbb Q,s)}{(s-1)^r} = \Omega_A \frac{\#\text{SHA}(A/\mathbb Q,s)\cdot\text{Reg}(A/\mathbb Q,s)} {\#A(\mathbb Q)_{\text{tors}}\cdot\#\hat A(\mathbb Q)_{\text{tors}}} \cdot \prod_{p} c_p.$$ Most of the terms have the same meaning as for elliptic curves. Notice the appearance of the torsion in the dual $\hat A$. An elliptic curve is self-dual, that's why $\#(E(\mathbb Q)_{\text{tors}})^2$ appears in the elliptic curve version. Also, the regulator is the pairing between a basis for the free parts of $A(\mathbb Q)$ and $\hat A(\mathbb Q)$ relative to the Poincare bundle on the product $A\times\hat A$. The fudge factors $c_p$ are for the primes of bad reduction, with $c_p=\#A(\mathbb Q_p)/A^0(\mathbb Q_p)$, where $A^0$ is the identity component of the Neron model.
Alternatively, as suggested by user52824, if one lets $\mathcal A_p$ be the reduction of the Neron model of $A$ at $p$ and $\mathcal A_p^0$ its identity component, then $c_p = \#(\mathcal A_p/ \mathcal A^0_p)(\mathbb F_p)$.
• For $c_p$ you meant to use $\mathbf{Q}_p$-points rather than $\mathbf{Q}$-points (and might be more natural, though equivalent, to define it to be $\#A_p(\mathbf{F}_p)/\#A_p^0(\mathbf{F}_p) = \#(A_p/A_p^0)(\mathbf{F}_p)$ where $A_p$ is the mod-$p$ reduction of the Neron model at $p$ and $A_p^0$ is its identity component, the second equality due to Lang's theorem). Sep 25, 2014 at 2:19