# Torsion of elliptic curves is finite

Let $S$ be an integral 1-dimensional scheme with function field $K$.

Let $E$ be an elliptic curve over $K$. The torsion of $E$ over $K$ is not necessarily finite. As an example consider an elliptic curve over $\mathbf C(t)$.

Now, assume the residue field of each closed point of $S$ to be finite.

Is the torsion of $E(K)$ finite?

Note that I'm not assuming $S$ to be noetherian.

Note that we may assume $S$ to be affine. To answer the above question positively, it would suffice to show the torsion embeds into the rational points of a special (not geometric) fibre of $E$.

-

First, it seems to me that you can even assume that $S$ is the spectrum of a local ring. Then, if $p$ is the resicual characteristic, the reduction map will in general not be injective on $p$--torsion. So, if I had to produce couterexamples to this, I would start with an $E$ over $\mathbb Z_p$ with supersingular reduction, and then base change $E$ to some big, totally ramified extension of $\mathbb Z_p$ where lots of $p$--power torsion points of $E$ are defined.
I think you can work backwards, choosing the subfield $K$ of an algebraic closure of $\mathbb{Q}_p$ by adjoining the coordinates of all $p$-power torsion points to $\mathbb{Q}_p$. –  S. Carnahan Dec 7 '12 at 3:59