Energy gap for Yang-Mills connections. II: Arbitrary closed Riemannian manifolds (Q526360): Difference between revisions

The paper extends the 4D energy gap result of Part I [Zbl 1456.58014] to arbitrary dimensions \(d\geq2\). The precise statement is this. Let \(G\) be a compact Lie group, and \(P\) be a principal \(G\)-bundle over a closed smooth \(d\)-dimensional Riemannian manifold \((X,g)\). Then there is \(\varepsilon=\varepsilon(d,g,G)>0\) such that any smooth Yang-Mills connection on \(P\) with \(||F_A||_{L^{d/2}(X)}\leq\varepsilon\) is flat. Previous \(L^{d/2}\) result of Gerhardt, and \(L^\infty\) result of Bourguignon-Lawson-Simons imposed a positivity constraint on a combination of the Riemannian and Ricci curvatures, \(Ric_g\wedge I+2Riem_g\). Nakajima proved a similar \(L^2\) result without assuming positivity, but only for spheres \(S^d\), albeit with arbitrary metrics. The idea of proof is quite different from the previous approaches. Intuitively, without the gap there would be non-minimal Yang-Mills connections with \(L^{d/2}\) energy arbitrarily close to zero, which should violate the analyticity of the Yang-Mills \(L^2\) energy functional. This last property is expressed by the Łojasiewicz-Simon gradient inequality established by Råde for \(d=2,3\), and generalized by the author and Maridakis to \(d\geq2\). The remaining ingredients of the proof are due to Uhlenbeck: a priori estimate for the curvature of a Yang-Mills connection; local existence of a Coulomb gauge for a connection with curvature small in \(L^{d/2}\) ; existence of a flat connection \(\Gamma\) given a Sobolev connection \(A\) with small in \(L^p\) curvature (for \(p>\frac12\dim X\)), together with a global gauge transformation of \(A\) to Coulomb gauge with respect to \(\Gamma\), and a Sobolev estimate for the distance between \(A\) and \(\Gamma\). Other approaches and prospects for extending the main theorem from closed to complete Riemannian manifolds are discussed. Appendices contain simplified proofs of many technical results under additional assumptions.