Basic classes for four-manifolds not of simple type (Q5929588)

The paper under review is to extend the notion of basic classes defined by \textit{P. B. Kronheimer} and \textit{T. S. Mrowka} [J. Differ. Geometry 41, No. 3, 573-734 (1995; Zbl 0842.57022)] to all 4-manifolds with \(b_2^+ > 1\). For a class \(w \in H^2(X, Z)\), \(X\) is of \(w\)-finite type if \(D_X^w((x^2-4)^nz) = 0\) for some \(n \geq 1\), for any \(z \in A(X)= \text{Sym}^*(H_{\text{even}}(X, Z))\otimes \bigwedge^*(H_{\text{odd}}(X, Z))\), where \(D_X^w: A(X) \to C\) is the Donaldson invariant; \(X\) is of \(w\)-strong simple type when \(D_X^w((x^2-4)z)=0, D_X^w(\delta z) = 0\) for any \(z \in A(X)\) and \(\delta \in H_1(X, Z)\). Thus \(X\) is of strong simple type if \(X\) is of \(w\)-strong simple type for any \(w \in H^2(X, Z)\). The basic result is Theorem 1.2 which states that: If \(X\) is a 4-manifold with \(b_2^+ > 1\) and of \(w\)-strong simple type for some \(w\), then \(X\) is of strong simple type, there are finitely many basic classes \(K_i \in H^2(X, Z)\) and \(a_i\) rational numbers such that \(D_X^w = e^{Q/2}\sum (-1)^{(K_i \cdot w + w^2)/2}a_ie^{K_i}\); those \(K_i\)'s are integral lifts of \(w_2(X)\); for any surface \(S\) of genus \(g\) representing an integral homology class with \(S^2 \geq 0\), \(2g-2 \geq S^2 + |K_i \cdot S|\) for all \(K_i\), where \(Q\) is the intersection form of the 4-manifold \(X\). When \(X\) is simply connected, Theorem 1.2 recovers the structure theorem of Kronheimer and Mrowka. The proof follows in several steps. (1) Characterize the Donaldson invariant \(D_X^w\) for \(w\)- and \(w+ \Sigma\)-strong simple type by a solution of a differential equation and the relative invariants factoring through the Fukaya-Floer homology; (2) Using the lemma that \(X\) is of \(w\)-strong simple type if and only if \(X\) is of \(w+ \Sigma\)-strong simple type, and blow-up techniques, the author proves that \(X\) being of \(w\)-strong simple type implies that it is of \(w^{'}\)-strong simple type for any \(w^{'} \in H^2(X, Z)\); (3) There exists at least one \(w\in H^2(X, Z)\) with \(D_X^w = e^{Q/2} \sum a_{i, w}e^{K_i}\); (4) Verifying that \(K_i\) is integral lift of \(w_2(X)\) as characteristic class, obaining the formula \(D_X^{w^{'}}\) for any \(w^{'}\) via blow-up formula and spinor norm; (5) The final adjunction inequality follows from the blow-up and the argument in Kronheimer and Mrowka. The paper uses \(b_1 = 0\) to refer the simple connectedness, the generalization of \(b_1 > 0\) also is meant to have nontrivial integral classes of \(H_1(X, Z)\) and no-torsion elements in \(H_1(X, Z)\). The technique in the paper under review cannot be applied to 4-manifolds with torsion classes in \(H_*(X, Z)\) directly. Readers should be careful on the notations and their meanings.