Let $X$ be an abelian variety over $k$, $X^t = \text{Pic} _{X/k}^0$ its dual, and $\mathscr{P}$ be the Poincare bundle on $X \times X^t$.

View $\mathscr{P}$ as a family of line bundles on $X^t$ parameterized by $X$. This gives a morphism $\kappa : X \to X^{tt}$

How exactly does $X$ parameterize the family of line bundles? Do we take $x \in X$ and consider $\mathscr{P} \mid _{\{x\} \times X^t}$?

Another possible interpretation I have in mind, although less intuitive:

$\text{Pic}_{X/k}$ is actually a functor. By shorthand notation, if it is representable, then the same notation is used to denote the group variety representing it. Making this explicit, we have that giving a morphism $Y \to \text{Pic}_X$ the representing object is equivalent to giving an element of the $\text{Pic}_X (Y)$, where now it is being considered as a functor. The latter is essentially $\text{Pic}(X \times Y)$, the group of line bundles on $X \times Y$. So given a line bundle $L$ on $X \times Y$, there is a corresponding $Y \to \text{Pic}_{X/k}$. Do we then say that viewing $L$ as a family of line bundles on $X$ parameterized by $Y$ gives $Y \to \text{Pic}_{X/k}$? This seems the likely approach based on how the theory is set up. However, it is not conducive to computations that one would like to make, especially eventually showing that $\kappa$ is an isomorphism. See the following paragraph.

Finally, is there a more explicit description of $\kappa$? Namely, either a description of what it does on points or perhaps a geometric intuition for it. An example I would like to compute is that $\kappa (0) = 0$, i.e. that it is a homomorphism. Using my previous interpretation of "parameterization" this seems equivalent to saying that $\mathscr{P} \mid _{0 \times X^t}$ is trivial.


It turns out that the two approaches I had in mind are equivalent. Picking a fiber over a point is equivalent to choosing such an isomorphism for the representation. So every line bundle on a product gives a map on one of the factors to $\text{Pic}$ by the description in the question. Tracing through, this is exactly pulling back the bundle over a fiber of a point $x_0$, namely taking the restriction to $\{x_0\} \times Y$. The latter interpretation is obviously better for explicit computations.

With this in mind, we can compute $\kappa ( 0 )$ by taking $\mathscr{P} \mid _{0 \times X^t} $ which is indeed just the structure sheaf. This is what it means for $\kappa (0) = 0$.


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.