Skip to main content

Section 1.7 Line Bundles and the Dual Abelian Variety (Angus)

Subsection 1.7.1 Introduction

Meta-goal.

Understand line bundles on abelian varieties.

Setup.

\(A\) an abelian variety \(/k\text{.}\)

Last time.

For \(L\) a line bundle on \(A\) we get a map

\begin{align*} \phi_L\colon A(K) \amp\to \Pic(A)\\ a\amp\mapsto t_a^* L\otimes L^{-1} \end{align*}

where

\begin{equation*} \Pic(A) = \{\text{line bundles on } A\}/\sim\text{.} \end{equation*}

This a is a group homomorphism (by the theorem of the square 1.5.15). We define

\begin{equation*} K(L)(k) = \ker(\phi_L) = \{a\in A(k) : t_a^* L \simeq L\}\text{.} \end{equation*}
Today.

We are going to package these into a big map

\begin{align*} \phi\colon \Pic(A)\amp\to \Hom(A(k), \Pic(A))\\ L \amp\mapsto \phi_L\text{.} \end{align*}
1.
\begin{align*} \phi_{L\otimes M}(a) \amp = t_a^*(L\otimes M) \otimes(L\otimes M)^{-1}\\ \amp = t_a^*L\otimes L^{-1} t_a^*M\otimes M^{-1}\\ \amp = \phi_L\otimes \phi_M \end{align*}
2.
\begin{align*} \phi_{t_b^*L}(a) \amp = t_a^*(t_b^*L) \otimes(t_b^*L)^{-1}\\ \amp = t_{a+b}^*L \otimes(t_b^*L)^{-1}\\ \amp = t_{a}^*L \otimes t_b^*L \otimes L^{-1} \otimes (t_b^*L)^{-1}\\ \amp = \phi_L(a) \end{align*}

by the theorem of the square 1.5.15

Definition 1.7.2.
\begin{align*} \Pic^0(A) \amp = \ker(\phi)\\ \amp = \{ L \in \Pic(A) : \phi_L = 0\}\\ \amp = \{ L \in \Pic(A) : t_a^* L\simeq L \ \forall a\in A(k)\}\\ \amp = \{\text{translation invariant line bundles}\}/\sim \end{align*}
Goals.

Study \(\Pic^0(A)\text{,}\) give it an abelian variety structure, solve a moduli problem, demonstrate some duality.

Subsection 1.7.2 Aside: alternate description of \(\Pic^0(A)\)

Definition 1.7.3. Algebraic Equivalence.

Two line bundles \(L_1,L_2\) on an abelian variety are algebraically equivalent if there exists a variety \(Y\) with line bundle \(L\) on \(A\times Y\) and points \(y_1y_2 \in Y\) s.t. \(L|_{A\times\{y_1\}} \simeq L_1, L|_{A\times\{y_2\}} \simeq L_2\text{.}\)

Remark 1.7.4.

This looks like homotopy.

Subsection 1.7.3 See-Saws

Remark 1.7.7.

In fact \(M = p_{2*}L\text{.}\)

Subsection 1.7.4 Properties of \(\Pic^0 A\)

1.
\begin{equation*} (m^*L\otimes (p_1^*l)^{-1} \otimes (p_2^*l)^{-1})|_{A\times\{a\}} = t_a^*L \otimes L^{-1} = \sheaf O_A \end{equation*}
\begin{equation*} (m^*L\otimes (p_1^*l)^{-1} \otimes (p_2^*l)^{-1})|_{\{a\}\times A} = t_a^*L \otimes L^{-1} = \sheaf O_A \end{equation*}

by see-saw 1.7.6 whole thing is trivial on \(A\times A\text{.}\)

2.
\begin{equation*} (f+g)^*L\cong (f\times g)^* m^*L \cong (f\times g)^*(p_1^*L \otimes p_2^*L) \cong f^*L \otimes g^*L \end{equation*}
3.

Induction of 3.

4.
\begin{equation*} \phi_{\phi_L(a)} = \phi_{t^*_a L}\otimes L^{-1} = \phi _{t_a^*L}\otimes L^{-1} = \phi_L \otimes \phi_{L^{-1}} = 0 \end{equation*}

If \(H^0(A,L) \ne 0\text{,}\) we would have a nontrivial section \(s\) of \(L\) then \(\lb -1\rb ^* s\) is a nontrivial section of \(\lb-1\rb^* L = L^{-1}\text{.}\) But if both \(L\) and \(L^{-1}\) have a nontrivial section then \(L \cong \sheaf O_A\text{.}\) So since \(L\) is nontrivial \(H^0(A,L) = 0\text{.}\) Now assume \(H^i(A,L) = 0\) for all \(i\lt j\text{.}\) Consider

\begin{gather*} A\xrightarrow{\id \times 0} A\times A \xrightarrow m A\\ a\mapsto (a,0)\mapsto a \end{gather*}

this gives

\begin{equation*} H^j(A, L) \to H^j(A\times A, m^*L) \to H^j(A,L) \end{equation*}

which composes to the identity.

\begin{equation*} H^j(A\times A, m^*L) = H^j(A\times A, p_1^*L \otimes p_2^*L) = \bigoplus_{i=0}^j H^i(A,L) \otimes H^{j-i}(A,L) \end{equation*}

by Künneth. The RHS is 0 by the inductive hypothesis. So the identity on \(H^j(A,L)\) factors through 0, hence the group is 0.

We now think of \(\phi_L\) as a map \(\phi_L \colon A(k) \to \Pic^0(A)\) with kernel \(K(L) (k)\text{.}\)

Idea is to study

\begin{equation*} \Lambda (L) = m^* L \otimes (p_1^* L)^{-1} \otimes (p_2^* L)^{-1}\text{.} \end{equation*}

Given an ample line bundle \(L\) on \(A\) we now have an isomophism of groups

\begin{equation*} A(k)/K(L)(k) \cong \Pic^0(A) \end{equation*}

the LHS allows us to put an abelian variety structure on \(\Pic^0(A)\text{.}\)

Subsection 1.7.5 The Dual Abelian Variety

(Sketch) (characteristic 0) Cover \(A\) by affine opens \(U_i = \Spec R_i\) such that for all \(a \in A\) the orbit \(K(L)a \subseteq U_i\) for some \(i\text{.}\) We can do this because abelian varieties are projective. Then we say \(U_i / K(L) = \Spec(R^{K(L)}_i)\) then glue. (details in Mumford, II sec, 6 appendix). Since we are in characteristic 0, the quotient scheme is in fact a variety.

Definition 1.7.13. Dual abelian varieties.

The dual abelian variety is

\begin{equation*} \hat A = A/K(L)\text{.} \end{equation*}
Remark 1.7.14.
  • \begin{equation*} \hat A(K) = \Pic^0(A) \end{equation*}
  • We have an isogeny
    \begin{equation*} \phi_L\colon A \to \hat A\text{.} \end{equation*}

Subsection 1.7.6 Dual morphisms

Let \(f\colon A\to B\) be a homomorphism of abelian varieties. Let \(\sheaf P_A,\sheaf P_B\) be the Poincaré bundles on \(A\) and \(B\text{.}\) Consider \(M= (F\times \id_{\hat B})^* \sheaf P_B\) on \(A\times \hat B\text{,}\) then

  1. \begin{equation*} M|_{A\times \{x\}} \in \Pic^0(A) \end{equation*}
  2. \begin{equation*} M|_{\{0\} \times \hat B} = 0 \end{equation*}

thus by the universal property we get a unique morphism

\begin{equation*} \hat f\colon \hat B \to \hat A \end{equation*}

satisfying

\begin{equation*} (\id_A\times \hat f)^* \sheaf P_A = (f\times \id_{\hat B})^*\sheaf P_B\text{.} \end{equation*}
Definition 1.7.16. Dual morphisms.

\(\hat f\) as above is called the dual morphism.

Remark 1.7.17.
  • \begin{equation*} \hat f\colon \hat B = \Pic^0(B) \to \hat A(k) = \Pic^0(A) \end{equation*}
    \begin{equation*} L\mapsto f^*L \end{equation*}
  • \begin{equation*} \hat{\lb n_A\rb} = [n_{\hat A}] \end{equation*}

Consider the Poincaré bundle \(\sheaf P_{\hat A}\) on \(\hat A \times \hat{\hat{A}}\text{,}\) now think of \(\sheaf P_A\) as living on \(\hat A \times A\text{.}\) By the universal property of \(\sheaf P_{\hat A}\) get a unique morphism

\begin{equation*} \operatorname{can}_A\colon A\to\hat{\hat A}\text{.} \end{equation*}
Definition 1.7.21. Symmetric morphisms, (principal) polarizations.

A morphism \(f\colon A \to \hat A\) is symmetric if \(f = \hat f\circ \operatorname{can}_A\)

A polarization is a symmetric isogeny \(f\colon A \to \hat A\) s.t. \(f= \phi_L\) for some ample line bundle \(L\) on \(A\text{.}\)

A principal polarization is a polarization of degree 1, i.e. an isomorphism.

Remark 1.7.22.

Elliptic curves always admit principal polarization.

If one wishes to mimic the theory of elliptic curves, one should study principally polarized abelian varieties.