Kodaira dimension and symplectic sums (Q1001222)

The author studies the symplectic sum \(Z=X_1\sharp_{F_1=F_2}X_2= (X_1\setminus\nu_1)\cup_{\partial\nu_1\sim_{\Phi}\partial\nu_2}(X_2\setminus \nu_2)\) of symplectic 4-manifolds \((X_1, \omega_1),\;(X_2, \omega_2)\) where \(F_1\subset X_1,\;F_2\subset X_2\) are embedded 2-dimensional submanifolds of equal area and genus with opposite self-intersection and \(\Phi: N_X{}_1F{}_1\to N_X{}_2F{}_2\) is an orientation preserving isomorphism of their normal bundles. The notion of symplectic sum is due to \textit{R. E. Gompf} [Ann. Math. (2) 142, 527--595 (1995; Zbl 0849.53027)] and \textit{J. D. McCarthy} and \textit{J. Wolfson} [Topology 33, 729--764 (1994; Zbl 0812.53033)]. The author calls the symplectic sum \(Z\) smoothly trivial if for some \(i\in \{1, 2\},\;X_i\) is the total space of an \(S^2\)-bundle of which \(F_i\) is a section. Otherwise the symplectic sum is called smoothly nontrivial. The symplectic sum is called relatively minimal if for each \(i=-1, 2\) there are no embedded symplectic spheres of square \(-1\) in \(X_i\setminus F_i\). The main result of the paper is the following: Suppose the symplectic sum \(Z\) is smoothly nontrivial and relatively minimal and the genus of the \(F_i\) positive. Then: (a) Z does not have Kodaira dimension \(-\infty\). (b) When \(Z\) has Kodaira dimension 0, the manifolds \(X_1,\;X_2\) and \(Z\) are given as follows: \((X_1, X_2)=(\mathbb CP^2\sharp((18-k)\overline{\mathbb CP}^2, \mathbb CP^2\sharp k\overline{\mathbb CP}^2)\) or \((S^2\times S^2, \mathbb CP^2\sharp17\overline{\mathbb CP}^2)\) and \(Z\) is a \(K3\) surface; \((X_1, X_2)=(\mathbb CP\sharp(9-k)\overline{\mathbb CP}^2, (S^2\times T^2)\sharp k\overline{\mathbb CP}^2),\;(S^2\times S^2, (S^2\times T^2)\sharp8\overline{\mathbb CP}^2)\) or \((\mathbb CP^2\sharp9\overline{\mathbb CP}^2, S^2\tilde{\times}T^2)\) where \(S^2\tilde{\times}T^2\) refers to the smoothly nontrivial \(S^2\)-bundle over the torus, and \(Z\) is an Enriques surface; \((X_1{}, X_2{})= (S^2\times T^2, S^2\times T^2),\;(S^2\times T^2, S^2\tilde{\times}T^2)\) or \((S^2\tilde{\times}T^2, S^2\tilde{\times}T^2)\) and \(Z\) is a \(T^2\)-bundle over \(T^2\) as described in \textit{K. Sakamoto} and \textit{S. Fukuhara} [Tokyo J. Math. 6, 311--327 (1983; Zbl 0543.55011)], that is, the manifolds obtained from a quotient of \(\mathbb R^2\times T^2\) by a \(T^2\) action on \(\mathbb R^2\) with a replaced neighborhood of a torus fiber of the projection to \(T^2\) by twisting on its boundary \(\partial D^2\times T^2\). (a) is proved from the result of the author [Internat. Math. Res. Notices, Article ID 49857, 1--17 (2006; Zbl 1110.57017)] on the Gromov-Witten invariants. To prove (b) the author shows that (Theorem 3.1) \(F_i\) are the both tori and are Poicaré dual to \(-K{}_X{}_i\) where \(-K{}_X{}_i\) is the canonical class of \((X_i, \omega_i)\). The diffeomorphism type of \(X_1,\;X_2\) and \(Z\) are enumerated in the three divided cases: (1) \(X_1\) and \(X_2\) are rational; (2) \(X_1\) is rational and \(X_2\) is irrational; (3) \(X_1\) and \(X_2\) are irrational and ruled.