Obstruction theory for the approximation and the deformation problems for Sobolev mappings (Q2486680)

The author begins with a good history of the problem that puts prior work and his own work on obstruction and deformation into context. Let \(M,N\) be closed Riemannian manifolds of dimensions \(m,n\), respectively. We may assume, by Nash's imbedding theorem, that \(N\) is isometrically embedded in \(\mathbb{R}^k\) for some large \(k>n\). He considers the Sobolev space \[ W^{1,p}(M, N) = \{ f \in L^p(M, \mathbb{R}^k): \nabla u \in L^p(M), f \in N {\text{ for a.e.}} \;x \in M \}. \] and the associated weak and strong sequential closures of the \(C^{\infty}(M,N)\) functions in \(W^{1,p}(M, N)\),\linebreak \(H^{1,p}_S(M, N), H^{1,p}_W(M, N)\). It is clear that \[ H^{1,p}_S(M, N) \subseteq H^{1,p}_W(M, N) \subseteq W^{1,p}(M, N). \] Cases of inequality and equality for special choices of \(M\) and \(N\) are known. The deformation problem considers whether a smooth map obtained by deforming \(f\) has additional geometric structure, and the approximation problems considers when a function in one of the classes can be approximated by smooth functions and the properties of the approximating sequence. First, we consider the approximation problem. \textit{F.~Bethuel} [Acta Math.\ 167, No.~3/4, 153--206 (1991; Zbl 0756.46017)] had proved that for \(W^{1,p}(D^m, N)\), where \(D^m\) is the unit ball in \(\mathbb{R}^m\), that \(C^{\infty}(D^m,N)\) is strongly dense in \(W^{1,p}(D^m, N)\) if and only if \(\pi_{[p]}(N) = 0\), and for a long time it was thought that this local obstruction would be the only obstruction to approximation. However, \textit{F.~Hang} and \textit{F.~Lin} [Acta Math.\ 191, No.~1, 55--107 (2003; Zbl 1061.46032)] showed that there is also a global obstruction by giving an example of a manifold which satisfies \(\pi_{[p]}(N) = 0\), but for which there is a map in \(W^{1,p}(M, N)\) that cannot be approximated by smooth maps. They also gave a necessary and sufficient condition for the global obstruction, completing the work of Bethuel. If \(1 \leq p < \dim M\), where \(M\) and \(N\) are as above, smooth maps between \(M\) and \(N\) are dense in \(W^{1,p}(M, N)\) if and only if \(\pi_{[p]}(N) = 0\), and \(M\) satisfies the \([p] - 1\)-extension property with respect to \(N\). Here \([p]\) denotes the greatest integer in \(p\). \(M\) satisfies the \(k\)-extension property with respect to \(N\) if for some CW-structure (\(M^j)_{(j \geq 1)}\) of \(M\) \((M^j\) being the \(j\)-skeleton of \(M\)), for every \(f \in C(M^{k + 1},N), f| _{M^k}\) has a continuous extension to \(M\). This shows that the obstruction to the global approximation problem is purely topological. The author shows that the obstruction to the global approximation problem (which has a local and a global part) can be described in terms of whether the given function \(f\) has a well-defined weak-\([p]\) homotopy type (local obstruction) and whether there is a Lipschitz CW-structure which is generic for \(f\) and such that for \(p \notin \mathbb{Z}, f| _{M^{[p]}}\) has a continuous extension to \(M\), or if \(p \in \mathbb{Z}\), the well-defined homotopy class \(\left[f| _{M^{[p]}}\right]\) has a continuous extension to \(M\) as a map of \(M \rightarrow N\). The deformation problem was first studied by \textit{H.~Brezis} and \textit{Y.~Li} [J.\ Funct.\ Anal.\ 183, No.~2, 321--369 (2001; Zbl 1001.46019)] who proved several interesting results and conjectured that any map in \(W^{1, p}(M, N)\) is connected to a smooth map by a continuous path in \(W^{1, p}(M, N)\). Hang and Lin [op.\ cit.] showed that this is not true in general, proving that if \(1 < p < m\) and \(u, v \in W^{1, p}(M, N)\), then \(u\) is connected to \(v\) by a continuous path in \(W^{1, p}(M, N)\) iff \(u\) is \([p] - 1\) homotopic to \(v\) (denoted \(u \sim_p v\)). The present author shows that it is possible to describe the obstruction to the deformation problem in terms of an element of a certain cohomology group, for the precise description of which we refer the interested reader to the paper. The author works out the abstract theory for the concrete case of \(N = S^2\).