On the inverse problem of determining a connection on a vector bundle (Q4949596)

A compact Riemannian manifold \((M,g)\) with boundary is called simple if the boundary is strictly convex w.r.t. the metric \(g\) and for every points \(x,y\) in \(M\), there exists a unique geodesic joining the points which depends smoothly on these points. First, the author gives the following formulation of a conjecture due to Uhlmann. Let \(\eta \) be a smooth complex vector bundle over the simple compact Riemannian manifold \((M,g)\), \(\nabla '\) and \(\nabla ''\) be two connections on \(\eta \). Given points \(x,y \) in \(M\), let \(I_{x,y}' (I_{x,y}''): \eta _x \rightarrow \eta_y \) be the parallel transport with respect to the connection \(\nabla'\) (respectively \(\nabla''\)) along the geodesic joining the points, where \(\eta _x \) and \(\eta _y\) are the fibers over \(x\) and \(y\). Assume that \(I_{x,y}'= I_{x,y}''\) for every points on the boundary of \(M\). Does this imply the existence of an automorphism of the bundle \(\eta \) which is the identity on the boundary and transforms the connection \(\nabla'\) to \(\nabla''\)? The author gives a positive answer to a local version of this problem proving the conjecture when \(\nabla'\) and \(\nabla''\) are close enough to a given connection \(\nabla \) under some assumption of smallness of the curvature tensors of the connection \(\nabla \) and of the metric \(g\). The proof is quite complicated, by estimating the terms in a Pestov identity.