Serre--Taubes duality for pseudoholomorphic curves (Q1400998)

Let \((X, \omega)\) be an integral symplectic \(4\)-manifold. A Lefschetz pencil is a map \(f: X\backslash\{q_1,\dots, q_d\}\to\mathbb S^2\) (the \(2\)-sphere), for finitely many points \(q_i\in X\), which is a submersion outside of finitely many critical points \(\{p_1,\dots, p_r\}\), all in distinct fibers of \(f\), such that 1) at each point \(p_i, f\) has the local model \((z_1, z_2)\mapsto z_1z_2 \); 2) at each \(q_j, f\) has the local model \((z_1, z_2)\mapsto z_1/z_2\). For any sufficiently large integer \(k\), there exists a Lefschetz pencil \(f\) such that the closed fibers are Poincaré dual to \(k[\omega]/2\pi\). This was proved by \textit{S. Donaldson} and the author [Topology 42, 743-785 (2003; Zbl 1012.57040)]. \(k\) is called a degree of \(f\). After blowing up at each \(q_j\), we have a manifold \(X'\) fibered over \(\mathbb S^2\) with smooth \(2\)-surfaces of some fixed genus as fibers over the points of \(\mathbb S^2\backslash\{f(p_i)\}\) and with critical fibers surfaces with a single ordinary double point. The exceptional spheres of the blow-ups form distinct sections of the Lefschetz fibration \(X'=X\sharp d\overline{\mathbb C P}\to\mathbb S^2\). A choice of an almost complex structure \(J\) on \(X'\) defines a smooth map \(\phi_f\) from \(\mathbb S^2\) to the Deligne-Mumford moduli space \(\bar M_g\) of stable curves of genus \(g\). \(\phi_f\) induces, from the universal relative Hilbert scheme, a fiber bundle \(F: X_r(f)\to\mathbb S^2\) with fibers which are the \(r=(2g-2)\)th symmetric products of fibers of \(f\). \(X_r(f)\) is a smooth compact manifold. There is a natural injection \(\phi\) from the set of homotopy equivalence classes of \(F\) to \(H_2(X; \mathbb Z)\). The standard surface count \(\mathcal J_{(X, f)}(\alpha)\) is the number of the holomorphic sections of \(F\) in the unique class \(\widetilde\alpha\) with \(\phi(\widetilde\alpha)=\alpha\in H_2(X; \mathbb Z)\). Main results of the paper are as follows: if \(b_+(X)>1+b_1(X)\) then \(\pm\mathcal J_{(X, f)}(\alpha)=\mathcal J_{(X, f)}(\kappa-\alpha)\) (Serre-Taubes duality for \(\mathcal J_{(X, f)}\)) where \(\kappa\) is Poicaré dual to the canonical class of the symplectic structure; if \(b_+(X)=1\) and \(b_1(X)=0\) and \(\alpha\cdot[\omega]>0, \alpha^2>K_X\cdot\alpha\) then \(\mathcal J_{(X, f)}(\alpha)=\pm 1\), where \(K_X\) is the first Chern class of a compatible almost complex structure on \(X\). The broad strategy is as in the paper by \textit{S. Donaldson} and the author stated above. After setting up enough theory to define relevant Gromov invariants, the author uses almost complex structures adapted to diagonal loci in symmetric products, obtained by Brill-Noether theory, and then uses almost complex structures adapted to the geometry of the Abelian-Jacobi map \(u: X_r(f)\to P_r(f)\) to perform explicit computations, where \(P_r(f)\) is the fiber bundle pulled back by \(\phi_f\) from the universal relative Picard scheme. The author sketches how to show that if \((X, \omega)\) contains no embedded torus of square zero, then the modulo two reductions of \(\mathcal J_{(X, f)}\) and the Gromov invariants \(Gr_X\) are equal. The author conjectures that if \(f\) is of sufficiently high degree, then \(\mathcal J_{(X, f)}\) is independent of \(f\) and equal to \(Gr_X\) and he is expecting it to be equal to the Seiberg-Witten invariant.