Let $C$ be a curve over $\mathbb Q$ with a point $P$ on $Pic^1$. For each $\mathbb Q$-rational point $Q$, $Q-P$ is a point on the Jacobian $J$. We can use the map $H^0(\mathbb Q, J) \to H^1(\mathbb Q,J[n])$ to obtain a class in $H^1(\mathbb Q,J[n])$ associated to each rational point.

Rather than trying to find a local obstruction to existence of rational points on $C$, we could try to find a local obstruction to existence of rational points in each $H^1(\mathbb Q, J[n])$ class. A local obstruction at $\mathbb Q_p$ would occur when the class, projected down to $H^1(\mathbb Q_p, J[n])$ does not arise from any $\mathbb Q_p$-rational point of $C$.

If $C$ does not have rational points, must there be some $n$ such that each class in $H^1(\mathbb Q, J[n])$ has a local obstruction at some $p$ or at $\infty$?

  • $\begingroup$ Is this inspired at all by the descent obstruction to rational points? $\endgroup$ – stankewicz Jun 14 '13 at 19:17
  • $\begingroup$ This is inspired by work of Manjul Bhargava on hyperelliptic curves. I guess this is the universal descent obstruction for finite abelian groups? $\endgroup$ – Will Sawin Jun 14 '13 at 20:09
  • 4
    $\begingroup$ Yes. This is the finite abelian descent obstruction. It's been conjectured that it's the only obstruction to the existence of rational points. $\endgroup$ – Felipe Voloch Jun 14 '13 at 20:25

As was pointed out by Felipe Voloch in the comments, this obstruction is equivalent to the finite abelian descent obstruction [Stoll, Thm. 6.4 and the discussion before Cor. 6.2]. It is also shown that for curves, this agrees with the Brauer–Manin obstruction [Stoll, Cor. 7.3]. Finally, Stoll conjectures the following:

Conjecture [Stoll, Conj. 9.1]. Let $C$ be a smooth projective geometrically connected curve over a number field $K$. Then $\overline{C(K)} = C(\mathbf A_K)_\bullet^{\text{f-ab}}$ (see [Stoll, Def. 5.4 and §2] for notation).

That is, the finite abelian descent obstruction is the only obstruction to the Hasse principle and weak approximation.

What is known.

  • If $g = 0$, then $C$ satisfies the Hasse principle $C(K) = \varnothing \Leftrightarrow C(\mathbf A_K)_\bullet = \varnothing$. If $C(\mathbf A_K)_\bullet = \varnothing$ then there is nothing to prove, and if $C \cong \mathbf P^1_K$ then $C(K)$ is dense in $C(\mathbf A_K)_\bullet$.
  • If $g = 1$, then the conjecture is true if and only if $Ш(K,\operatorname{Pic}^0_C)_{\text{div}} = 0$ [Stoll, Cor. 6.2 and the discussion after Cor. 6.3]. In particular, this is a weak form of the Tate–Shafarevich conjecture.
  • If $g \geq 2$, then $C(K)$ is finite by Faltings, so $\overline{C(K)} = C(K)$. Thus, the conjecture says that $C(\mathbf A_K)_\bullet^{\text{f-ab}} = C(K)$. I'm not even sure if we know whether $C(\mathbf A_K)_\bullet^{\text{f-ab}}$ is finite.

Finally, Stoll proves the conjecture in the following case:

Theorem [Stoll, Thm. 8.6]. Assume $C$ admits a nonconstant map $C \to A$ into an abelian variety $A$ such that $A(K)$ is finite and $Ш(K,A)_{\text{div}} = 0$. Then $C(K) = C(\mathbf A_K)_\bullet^{\text{f-ab}}$.

This is for example the case when $K = \mathbf Q$ and $A$ is modular of analytic rank $0$. As mentioned before, the condition $Ш(K,A)_{\text{div}}$ is expected to always be true, but the condition that $A(K)$ is finite is a serious restriction on the generality of the theorem.

Recent progress?

I haven't worked on rational points for a while, so I don't know what the current status is. I would be surprised if there were a great breakthrough on the conjecture in its general form, though.


[Stoll] Stoll, Michael, Finite descent obstructions and rational points on curves, Algebra Number Theory 1, No. 4, 349-391 (2007). ZBL1167.11024.

  • $\begingroup$ Shouldn't this answer have $C(\mathbb A_K)_\bullet^{\text{f-ab}}$ in every single place where you wrote $C(\mathbb A_K)_\bullet$? $\endgroup$ – M.D. Mar 30 at 21:39
  • $\begingroup$ @M.D. you're absolutely right! Fixed now. $\endgroup$ – R. van Dobben de Bruyn Mar 31 at 2:16

This is a known difficult problem as Felipe Voloch mentions in the comments and is as far as I know still open today.

  • $\begingroup$ Reference? What is known? .... $\endgroup$ – Chris Wuthrich Jul 31 '18 at 8:08

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.