Regularity of harmonic maps from polyhedra to CAT(1) spaces (Q1703134)

Let a \textit{normalized half-space} in \({\mathbb R}^n\) mean a connected component of \({\mathbb R}^n\setminus h\), where \(h\) is a linear subspace of codimension 1. An \textit{\(n\)-dimensional \(\nu\)-wedge} \(W\) will then mean the closure of the intersection of \(\nu\) normalized half-spaces, whose normals are linearly independent \((\nu\leq n)\). The intersections of \(W\) with its determining linear subspaces of codimension 1 are called \textit{\((n-1)\)-dimensional faces}. Intersections with more than one of these linear subspaces form the \textit{lower-dimensional faces} of \(W\). The union of all \(l\)-dimensional faces is called \(l\)-skeleton of \(W\). Now, a \textit{dimension-\(n\), codimension-\(\nu\) local model} \(B\) is defined as a disjoint union of \(n\)-dimensional \(\nu\)-wedges \(W_1,\ldots,W_k\) modulo an equivalence relation which identifies the points of each \((n-1)\)-dimensional face of each \(W_i\) with the points of exactly one \((n-1)\)-dimensional face of another \(W_j\) via a linear isometry. A local model \(B\) is called \textit{admissible}, if \(B\setminus F\) is connected for every \((n-2)\)-dimensional face \(F\) of \(B\). (Here, \(F\) is understood as the equivalence class of a face of a wedge \(W_i\).) On the other hand, a complete metric space \((Y,d)\) is called \textit{geodesic} if every two of its points \(P, Q\) can be joined by a curve of length \(d(P,Q)\) (a ``geodesic''). Such a space is called \textit{CAT(1)}, if for every geodesic triangle \(PQR\) (with perimeter less than \(2\pi\)) and a comparison triangle (i.e., of respectively equal sides) \(\tilde{P}\tilde{Q}\tilde{R}\) on the Euclidean sphere of radius 1, it holds that \(d(P_t,R_s)\leq d(\tilde{P}_t,\tilde{R}_s)\), where \(P_t\) (\(R_s\)) divides the geodesic segment \(PQ\) (\(RQ\)) in the ratio \(\frac{t}{1-t}\) (\(\frac{s}{1-s}\)) (analogously \(\tilde{P}_t, \tilde{R}_s\)). The authors use the notion of energy density for maps from Riemannian manifolds into metric spaces as defined in [\textit{N. Korevaar} and \textit{R. Schoen}, Commun. Anal. Geom. 1, 561--659 (1993; Zbl 0862.58004)], to extend it with the help of Proposition 2.1 in [\textit{G. Daskalopoulos} and \textit{C. Mese}, Commun. Anal. Geom. 18, 257--337 (2010; Zbl 1229.53068)] to the case where the domain of definition is a (ball in a) local model \(B\) and the target space CAT(1). They focus on energy-minimizing maps and study their regularity properties. Among other things they prove that if the (in a natural sense, Riemannian) metric on \(B\) is Lipschitz, then every energy minimizing map is Hölder regular with Hölder constant and exponent depending on the total energy of the map and the metric. At points away from the \((n-2)\)-dimensional skeleton the regularity can be improved to locally Lipschitz. Finally, they give lower bounds for the Hölder exponent in the neighborhoods of points on lower-dimensional skeleta by applying techniques of [Daskalopoulos and Mese, loc. cit.].