Smooth 2-spheres in homology K3 surfaces (Q1336621)

An important problem in the theory of 4-manifolds has been to determine which homology classes in a 4-manifold are represented by embedded 2- spheres. The expository article by the reviewer [Expo. Math. 10, No. 4, 289-309 (1992; Zbl 0764.57020)] gives a survey of some of the results obtained on this problem before 1991. There has been much recent progress through work of \textit{P. B. Kronheimer} and \textit{T. S. Mrowka} [Topology 32, No. 4, 773-826 (1993; Zbl 0799.57007); 34, 37-97 (1995; Zbl 0832.57011); The genus of embedded surfaces in the projective plane, Math. Res. Letters 1, No. 6, 797-808 (1994)], \textit{R. Friedman} and \textit{J. W. Morgan} [Smooth four-manifolds and complex surfaces, Ergebn. Math. Grenzgeb. 3. Folge 27 (1994; Zbl 0817.14017)], \textit{J. W. Morgan}, \textit{T. Mrowka} and the author [The \(L^2\)-moduli space and a vanishing theorem for Donaldson polynomial invariants, Monographs in Geometry and Topology 2 (1994; Zbl 0830.58005)], and others applying ideas of gauge theory, and, more recently, the Seiberg-Witten invariants, to this problem and generalizations such as finding the minimal genus of a surface representing a given homology class. The paper under review shows the nonexistence of a smooth 2-sphere representing a class of positive square in a smooth spin 4-manifold sharing certain homology properties with the K3 surface. More precisely, the main theorem is: Let \(C\) be a 2-dimensional homology class in a smooth, spin 4-manifold \(W\) with \(b_2^+ (W) = 3\), \(b_2 (W) \geq 8\), and no 2-torsion in \(H_1 (W)\). If \(C\) has self-intersection number \(C^2 > 0\), then \(C\) cannot be represented by a smoothly embedded 2-sphere. The same is true if \(C = 0\) and the divisibility of \(C\) is odd. As a corollary, there is a result on embedded projective planes in a homology Enriques surface that follows from the above theorem by taking a double covering. The technique used to prove the theorem is to study the orbifold \(X\) obtained from the 4-manifold by removing a neighborhood of the embedded sphere and coning off the resulting lens space (in the case of positive square). The author then modifies the proof of \textit{S. K. Donaldson}'s Theorem C [J. Differ. Geom. 24, 275-341 (1986; Zbl 0635.57007)] to this orbifold context to prove the result. A key part of the argument is to show that potential ends of a cut down moduli space being studied coming from instantons ``bubbling off'' at the cone point do not occur. An appendix concerning spin structures on orbifolds provides details of the existence of the Dirac operator needed to construct bundles used in the analysis of the ends of the moduli space.