Four-manifold invariants from higher-rank bundles (Q2494203)

The author defines SU\((N)\) Donaldson polynomial invariants. He includes the blow-up formula [\textit{P. B. Kronheimer} and \textit{T. S. Mrowka}, J. Differ. Geom. 41, No. 3, 573--734 (1995; Zbl 0842.57022)] in general, and the knot surgery formula [\textit{R. Fintushel} and \textit{R. J. Stern}, Invent. Math. 134, No. 2, 363--400 (1998; Zbl 0914.57015)] for the invariants without insertions. He also computes the invariant for the K3 surface. The papers of \textit{S. Donaldson} [J. Differ. Geom. 26, No. 3, 397--428 (1987; Zbl 0683.57005); Topology 29, No. 3, 257--315 (1990; Zbl 0715.57007)] working out these invariants in the SU(2) case are fairly long and use a number of substantial earlier works. This work is nicely explained in the book by \textit{S. K. Donaldson} and \textit{P. B. Kronheimer} [The geometry of four-manifolds, New York: Clarendon Press Oxford University (1990; Zbl 0820.57002)]. There is no way to get to these invariants without a fair bit of hard work. Thus this paper would probably be difficult to read for someone unfamiliar with the SU\((2)\) case since it packs all of the issues described in the book [loc. cit.] into 59 pages and does a fair bit more (the blow-up formula and a special case of the knot surgery formula.) Of course there are now many topological and geometric invariants defined by evaluating natural cohomology classes on a moduli space described as the solutions of a non-linear Fredholm equation (Donaldson invariant, Seiberg-Witten invariant, Gromov-Witten invariant, many flavors of Floer theory). The abbreviated description given here is just right for people who are familiar with one of these examples of non-linear Fredholm theory (especially the SU\((2)\) case). The author gives a brief outline of the definition of the invariants in the introduction and spends most of the paper on the points that are different from the SU\((2)\) case. Donaldson invariants arise as evaluations of natural cohomology classes on the moduli space of anti-self-dual connections.The first difference between the SU\((2)\) case and the SU\((N)\) case is that there are many more types of reducible connections in the SU\((N)\) case. To address these he considers connections on U\((N)\) bundles with fixed determinant. He then first considers topological restrictions on the relevant bundles to insure that the unperturbed moduli space contains no reducibles, then shows that for sufficiently small perturbation the moduli space will still contain no reducibles. To cover the general case he proves the Blow-up formula for the cases satisfying the topological restrictions and defines the invariants in general so that the the Blow-up formula will continue to hold. The second difference comes in the transversality argument. For SU\((2)\) for generic metrics the anti-self-dual connections will be regular. This is no longer true in the SU\((N)\) case. To get around this he uses the holonomy perturbations. There is an argument in the Physics literature [\textit{M. Mariño} and \textit{G. Moore}, Commun. Math. Phys. 199, No. 1, 25--69 (1998; Zbl 0921.58080)] that states that the SU\((N)\) invariants should contain no new topological information past the SU\((2)\) invariants. Justifying this mathematically is then a natural thing to do. This reviewer is happy to see that P. Kronheimer took up this task because his exposition is flawless and nicely complements the standard reference from the SU\((2)\) case that he coauthored. It is nice to have a Donaldson argument for the knot surgery formula (even in the special case). Even though there does not appear to be new topological information in the SU\((N)\) invariants there are still interesting questions related to this theory that the author lists at the end of the paper.