Old and new motivic cycles on Abelian surfaces

Let {\mathcal Q}_0 be a rational curve in {\mathds P}^2 which determines a certain Noether-Lefschetz divisor as in Secton ?. Let A be an Abelian surface whose moduli point does not lie on that divisor. Let {\mathcal Z}_{{\mathcal Q}_0,P,R} be the cycle in H^3_{{\mathcal M}}(A,{\mathds Q}(2)) determined by {\mathcal Q}_0, a node P and an auxilliary point R as in Theorem ?. Then there is a hyperellipic curve {\tilde{D}} mapping to a curve D on A and points P_1,P_2 and \tilde{R} on {\tilde{D}} such that under the induced map from the Jacobian \pi:J({\tilde{D}})\rightarrow A

where Z_{P_1,P_2,\tilde{R}} is the cycle constructed by Collino in CH^{g}(J({\tilde{D}}),1)=H^{2g-1}_{{\mathcal M}}(J({\tilde{D}}),{\mathds Z}(g)) as in section ?.

Suppose C is a nodal rational curve on a surface X with node P. Let \pi:\tilde{X}\rightarrow X denote the blow up of X at P such that the strict transform \tilde{C} of C no longer has a node at P. Let P_1 and P_2 be the two points lying over P on \tilde{C}. If E_P is the exceptional fibre it is a rational curve and it meets the strict transform at the points P_1 and P_2. Let f be a function with \operatorname{div}(f)=P_1-P_2 on \tilde{C} and let g be a function on E_P with divisor \operatorname{div} (g)=P_2-P_1. Then

is a cycle in H^{3}_{{\mathcal M}}(X,{\mathds Z}(2)).

Let A be an Abelian surface and {\mathds P}^2_A the corresponding configulation of {\mathds P}^2 with six lines l_i tangent to a conic. Let q_{ij}=l_i \cap l_j denote the fifteen nodal points of the degenerate sextic S_A. Then A corresponds to a point on H_{5} if and only if there exist a conic {\mathcal Q} passing through five of the points q_{ij} and tangent to the remaining line.

Let {\mathcal Q} be a rational curve on {\mathds P}^2_A which meets the sextic S_A only at points of even multiplicity or at the double points q_{ij} as in the theorems of Birkenhake and Wilhelm. Let \psi:{\mathcal C} \longrightarrow Q be the double cover. Then {\mathcal C} is not irreducible and is the union of two rational curves {\mathcal C}_1 and {\mathcal C}_2.

Let n_d be the number of rational curves of degree d passing through 3d-1 points in general position. Then n_d is finite and non-zero.

Let 0 \leq k \leq 3d-1. Let n_{d,k} be the number of rational curves passing through k points and meeting (3d-1-k) lines tangentially. Then n_{d,k} is finite and non-zero.

Consider pairs (S,{\mathcal Q}) where S=\bigcup_{i=1}^6 L_i is a degenerate sextic in {\mathds P}^2 given by a product of six lines and {\mathcal Q} is a rational curve of degree d. In general they will meet at 6d points. Suppose there exists a degenerate sextic S_0=\bigcup_i L_{0,i} and a rational curve {\mathcal Q}_0 of degree d such that they meet at \leq 3d+1 points of the following type

• k nodes of the form q_{0,ij}=L_{0,i} \cap L_{0,j} of S_0

• 3d-1-k other points tangent to the lines L_{0,i} (which could coincide to be of even multiplicity).

• 2 other points (which could coincide and be of one of the above types).

Then, for any degenerate sextic S=\bigcup_i L_i the number N_{d,k} of rational curves {\mathcal Q} of degree d passing through the k nodes and meeting the lines L_i tangentially at 3d-1-k points is finite and non-zero.

For z \notin P_{\Delta}, let

Then {\mathcal Z}_{{\mathcal Q}_0}(z) is an element of the group H^3_{{\mathcal M}}(\hat{K}_{A_z},{\mathds Q}(2)). Further, it defines an element of the motivic cohomology group of the generic fibre H^3_{{\mathcal M}}(\hat{K}_{A_{\eta}},{\mathds Q}(2)) which is indecomposable and has boundary a non-zero multiple of \hat{{\mathcal C}}_1(z)-\hat{{\mathcal C}}_2(z) in the fibres over P_{\Delta}, up to the boundary of a decomposable element.

Let D be a curve of genus g and P a point on D. Let \eta: D \rightarrow A be a map from D to an Abelian variety A such that \eta(P)=0. Then there is a unique homomorphism \tilde{\eta}:J(D) \longrightarrow A such that

where \iota_P:D \longrightarrow J(C) is the map

Let D be a curve of genus g and \mu:D \rightarrow A be a map such that \mu(D)=Q for some point Q on A. Then there is a morphism \tilde{\mu}:J(C) \rightarrow A such that

Further, if P' is another point such that \mu(P')=Q, then \mu=\tilde{\mu} \circ \iota_{P'} as well.

The points lying over the node P under the map \tilde{{\mathcal D}} \rightarrow {\mathcal D} are ramification points of the map \tilde{\phi}:{\tilde{{\mathcal D}}}\rightarrow \tilde{\mathcal C}. To reduce notation, we will continue to call them P_1,P_2.

Let Z_{P_1P_2,R_1} be the Collino cycle corresponding to P_1 and P_2, where R_1 is a Weierstrass point mapping to the point R on {\mathcal D}. Then

where \tilde{\mu} is the map J({\tilde{{\mathcal D}}}) \longrightarrow A obtained by Corollary 4.2.

Recall that there is an extension class \bar{e}^4_{P_1P_2,R} in \operatorname{Ext}_{MHS}({\mathds Z}(-g),H^{2g-2}(J({\tilde{{\mathcal D}}}))) corresponding to Collino’s cycle Z_{P_1P_2,R_1}. From the theorem above, the class \frac{1}{2}\tilde{\mu}_*(\bar{e}^4_{P_1P_2,R}) is the extension class corresponding to the cycle \Xi_{P,R} in \operatorname{Ext}_{MHS}({\mathds Z}(-2),H^2(A)).