Yang-Mills fields on cylindrical manifolds and holomorphic bundles. I (Q1919266): Difference between revisions

[The following review covers both Part I and Part II of the paper.] The problem of finite energy instantons on infinite cylinders arises naturally in Floer theory and also when trying to produce surgery formulae for the Donaldson invariants of a smooth 4-manifold. Works of Morgan, Mrowka, Ruberman and Taubes have established many important facts about these instantons and the way they are organized in moduli spaces. For many applications, an explicit description of the moduli space of these tunneling instantons is highly desirable and at the same time, very difficult to obtain. The papers under review have precisely this goal in mind on a special but extremely useful situation. More precisely, the author is looking for a concrete description of finite energy instantons on cylinders \(X=\mathbb{R}\times Y\), where \(Y\) is the total space of a degree \(d\) \(S^1\)-bundle over a Riemann surface \(\Sigma\). The manifold \(X\) has a natural complex structure and admits a natural compactification as a ruled surface \(S\) over \(\Sigma\). Naturally, the first thing that comes in mind is the Kobayashi-Hitchin correspondence in the compact case according to which minimal Yang-Mills fields are in a bijective correspondence with the moduli space of stable rank 2 holomorphic bundles. This algebraic object can be studied using the very developed machinery of algebraic geometry. In the compact case, the Kobayashi-Hitchin correspondence is established using heat equation techniques. This is the approach taken by the author in this noncompact case. The circle bundle \(Y\) has natural product-like metrics and by fixing such a metric one obtains a product metric on the cylinder \(\mathbb{R}\times Y\). This is the metric the author chose to work with. It is not a Kähler metric but it is conformal to one. The loss of Kähler features of the cylindrical metric on \(X\) is compensated by the nice spectral properties of the complex Laplacian. In this way, the apparently hopeless heat equation approach works remarkably well in this noncompact situation as well. The finite energy instantons \(A(t)\) over \(\mathbb{R}\times Y\) have well-defined asymptotic limits \(A(\pm\infty)\) which are flat connections on an \(SU(2)\) bundle over \(Y\). The flat connections on \(Y\) are uniquely determined by their holonomy \(\rho\): \(\pi_1(Y)\to SU(2)\). The holonomy along a fiber classifies these flats in several types. Thus, the instantons can be classified according to the types of its asymptotic limits as well. Part I is devoted to the case when the fiberwise holonomy of the asymptotic limits is \(\pm1\in SU(2)\). Part II [ibid., 777-788 (1996; see the following review Zbl 0857.58011)] studies the case when this holonomy is an element of \(\mathbb{Z}_{|d|}\subset SU(2)\) other than \(\pm1\). The \(-\infty\) limit on \(X\) should be regarded as the behavior near the zero section \(\Sigma_0\) in \(S\supset X\) while the \(+\infty\) limits should be regarded as the behavior near the \(\infty\) section \(\Sigma_\infty\) in the ruled surface \(S\) compactifying \(X\). The main conclusion of Part I of this work is that there exists an injection \(\iota\) between the moduli space of finite energy instantons (with irreducible flat limits and trivial fiber holonomy) and a space of stable holomorphic bundles over \(S\) having special (and explicit) behaviors near the zero and \(\infty\)-section in \(S\). The author conjectures that the injection \(\iota\) is actually surjective and provides evidence for its validity. Part II deals with the case when the asymptotic limits have nontrivial holonomy along fibers. Although the logical structure of the proof is similar to the compact case, the non-compact setting makes the heat equation approach highly nontrivial. Besides the obvious analytical difficulties, one has to cope with geometric problems specific to this noncompact case, such as the removal of a (complex) codimension 1 singularity (near \(\Sigma_0\) and \(\Sigma_\infty\)), which (in the case of trivial holonomy) allows to extend objects on \(X\) to the compactification \(S\). When the fiber holonomy of the limiting flat connection is nontrivial, one cannot extend the corresponding object over the zero and \(\infty\)-section and instead one has to work in an orbifold context.