Last modified date: <%+ tp.file.last_modified_date() %>
- Tags:
- Refs:
-
Links:
- theta divisor
- abelian variety
- Jacobian
Abel-Jacobi map
- Let \(X \in {\mathsf{sm}}{\mathsf{Var}}_{/ {{\mathbf{C}}}} ^{\mathop{\mathrm{proj}}, \dim = n}\)
- Let \(Z^i_h(X)\) be the nullhomologous \(i{\hbox{-}}\)cycles.
- Let \(J^\ell(X)\) be ???
- Define the Chow group.
- Define the Griffiths Abel-Jacobi map: \begin{align*} \Phi_X: Z^i_h(X) \to J^{2i-1}(X) \end{align*}
- Define \begin{align*} \begin{align*} G: H_{2 n-2 i+1, B}(X, \mathbb{Z}) &\rightarrow {\operatorname{Fil}}^{n-i+1} H_{B}^{2 n-2 i+1}(X, \mathbb{C}) {}^{ \vee }\\ \alpha &\mapsto \int_\alpha \end{align*}, \end{align*} and \(\operatorname{im}G\) as the group of periods.
-
If \(z \in Z^{n-i}_h(X)\), so of real dimension \(2n-2i\), there is a \(\Gamma \in {\operatorname{Fil}}^{n-i+1} H_{B}^{2 n-2 i+1}(X, \mathbb{C}) {}^{ \vee }\) of dimension \(2n-2i+1\) with \({{\partial}}\Gamma = z\), and \(\int_\Gamma\) is well-defined even if \(\Gamma\) is not a closed cycle (key idea: choose representatives of closed forms and apply the \({\partial}{ \mkern 1.5mu\overline{\mkern-1.5mu{\partial}\mkern-1.5mu}\mkern 1.5mu}\) lemma). Then \(\Gamma\) is determined up to addition of closed chains, making \(\int_\Gamma\) defined up to elements of the form \(\int_\alpha\).
- Thus there is a well-defined map \begin{align*}\Phi_X(z) \coloneqq\int_\gamma \in {{\operatorname{Fil}}^{n-i+1} H_{B}^{2 n-2 i+1}(X, \mathbb{C}) {}^{ \vee }\over \operatorname{im}G}, \quad \in J^{2i-1}(X)?\end{align*}
- The Abel-Jacobi map factors through rational equivalence, yielding \begin{align*} \Phi_X: \mathsf{Ch}^i_h(X) \to J^{2i-1}(X) \end{align*}
\(\ell{\hbox{-}}\)adic variant
- Pick \(x\in Z^r(X_{/ {k}} )\) of dimension \(r\) where \(x\in Z^r_h(X_{/ { { \mkern 1.5mu\overline{\mkern-1.5muk\mkern-1.5mu}\mkern 1.5mu } }} )\), so a cycle on \(X\) which becomes nullhomologous after base change to $\mkern 1.5mu\overline{\mkern-1.5muX\mkern-1.5mu}\mkern 1.5mu \coloneqq X\otimes_k { \mkern 1.5mu\overline{\mkern-1.5muk\mkern-1.5mu}\mkern 1.5mu } $.
-
This determines a class
\begin{align*}
e_{Z} \in H^{1}\left(G_{K}, H_{\text {èt }}^{2 r-1}\left(X \otimes \mkern 1.5mu\overline{\mkern-1.5muK\mkern-1.5mu}\mkern 1.5mu, \mathbb{Z}_{\ell}(r)\right)\right)
\end{align*}
which only depends on the rational equivalence class of \(z\) and the image of \(z\) under the Abel-Jacobi map
\begin{align*}
C H_{h}^{r}(X) \rightarrow H^{1}\left(G_{K};\, H_{\text{ét}}^{2 r-1}\left(X \otimes \mkern 1.5mu\overline{\mkern-1.5muK\mkern-1.5mu}\mkern 1.5mu, \mathbb{Z}_{\ell}(r)\right)\right)
\end{align*}
- This map is defined on rational equivalence classes of homologically trivial codimension \(r\) cycles on \(X\)
Abel’s theorem
- Statement for codimension 1 cycles: the Abel Jacobi map is an isomorphism \begin{align*} \Phi_{X}: \mathrm{CH}_h^{1}(X) { \, \xrightarrow{\sim}\, }J^{1}(X) \end{align*}
-
Proof:
- Look at LES for exponential exact sequence, identigy \({\operatorname{CH}}^1(X) = \operatorname{Pic}(X) = \operatorname{Pic}(X^{\mathrm{an}})\) and \(\mathsf{Ch}^1_h(X) = \ker c_1 \subseteq \operatorname{Pic}(X^{\mathrm{an}})\) to get \begin{align*} 0 \rightarrow H^{1}(X, \mathbb{Z}) \rightarrow H^{1}\left(X, \mathcal{O}_{X}\right) \rightarrow \mathrm{CH}^{1}(X)_{\text {h }} \rightarrow 0\end{align*}
- Yields \(H^{0, 1}(X) = H^1(X; {\mathcal{O}}_X)\) and \(J^1(X) { \, \xrightarrow{\sim}\, }{\operatorname{CH}}^1_h(X)\); show it is the inverse of \(\Phi_X\).
Universal divisors
- If \(X \in {\mathsf{sm}}{\mathsf{Var}}_{/ {{\mathbf{C}}}} ^{\mathop{\mathrm{proj}}}\) then \(J^1(X) \in {\mathsf{Ab}}{\mathsf{Var}}\) and is a moduli of nullhomologous divisors on \(X\) modulo linear equivalence, or equivalently topologically trivial line bundles on \(X\).
-
Theorem: there is a line bundle \(P\in \operatorname{Pic}(J^1(X) \times X)\) such that for any \(t\in J^1(X)\) the associated divisor \(D_t\) is nullhomologous and \(\Phi_X(D_t) = t\). Such a divisor \(D_t\) is called a universal divisor or Poincare divisor when \(X\in {\mathsf{Ab}}{\mathsf{Var}}\).
- Note: it is not unique, but can be normalized by fixing \(x\in X\) and imposing that the restriction \({ \left.{{P}} \right|_{{H^1(X)\times \left\{{x}\right\}}} }\) is trivial.