Stratification for the singular set of approximate harmonic maps (Q1794592)

Let \(M\) and \(N\) be two compact Riemannian manifolds with a geodesic ball \(B_3(p) \subset M\) satisfying \[ \begin{aligned} & |\text{sec}_{B_3(p)}| \leq K_M, \,\,\, \text{inj}(B_3(p)) \geq K_M^{-1}, \\ & \partial N = \emptyset,\,\,\, |\text{sec}_N |\leq K_N, \,\,\, \text{inj}(N) \geq K_N^{-1}, \,\,\, \text{diam}(N) \leq K_N,\\ & {\dim}(M) =m, \,\,\, \dim(N) \leq n. \end{aligned} \] Also assume that \(N\) is isometrically embedded into \({\mathbb R}^n\). Consider the Sobolev space \(H^1(M, N)\) defined by isometrically embedding into \({\mathbb R}^n\), and for \(\Omega \subset M\), let \[ H^1(\Omega, N) = \left\{ f \in H^1(\Omega, {\mathbb R}^n)\,:\, f(x) \in N \,\,\, \text{for a.e.\(x\in \Omega\)}\right\}. \] A map \(u \in H^1(\Omega, N)\) is said to be (stationary) approximate harmonic if it satisfies the following in the distribution sense: \[ \begin{aligned} & \Delta u +A(u)(\nabla u, \nabla u) = f,\\ & \text{div}_j \left(|\nabla u|^2\delta_{ij} - 2\langle \nabla_i u, \nabla_j u\rangle\right) + 2\langle \nabla_i u, f\rangle =0 \end{aligned} \] for some \(f \in L^2(\Omega)\). In case \(f = 0\), we call such a \(u\) stationary harmonic map. In the paper under review, the authors prove that the results obtained in [the authors, Ann. Math. (2) 185, No. 1, 131--227 (2017; Zbl 1393.58009)] for stationary harmonic maps continue to hold for stationary approximate harmonic maps. More precisely, they show that if \(u : B_3(p) \subset M \to N\) is an approximate harmonic map, and \(f \in L^p\) for \(p > \frac{m}{2}\), then for each \(k\) the set \({\mathcal S}^k(u)\) is \(k\)-rectifiable. Here \({\mathcal S}^k(u)\) is the \(k\)-th stratification of \(u\). They also obtain estimates on the volume of \(k\)-th \((\epsilon, r)\)-stratification \({\mathcal S}^k_{\epsilon, r}(u)\) and \(k\)-th \(\epsilon\)-stratification \({\mathcal S}^k_{\epsilon}(u)\) of \(u\). In this paper, the authors simplify many of their arguments in [loc. cit.], and in particular they produce a new main covering lemma which vastly simplifies arguments in [loc. cit.].