Whitney-type extension theorems for jets generated by Sobolev functions (Q529229)

The article deals with the following variant of a classical Whitney extension problem: (1) How can we decide whether there exists a function \(F \in L_p^m({\mathbb R}^n)\) such that \(T_x^{m-1}[f] = P_x\) for all \(x \in E\)? (2) How small can be the norms of functions \(F \in L_p^m({\mathbb R}^n)\) such that \(T_x^{m-1}[f] = P_x\)? Here, \[ T_x^{m-1}[F](y) = \sum_{|\alpha| \leq m-1} \frac1{\alpha!} \, D^\alpha F(x) (y - x)^\alpha, \qquad y \in {\mathbb R}^n, \] is the Taylor polynomial of \(F\) of degree \(m-1\) at \(x\), \(E\) a closed subset of \({\mathbb R}^n\), \(n< p \leq \infty\). The main results are characterizations of the restrictions \(\{D^\alpha F\big|_E:|\alpha| \leq m-1\}\) to \(E\) of \((m-1)\)-jets generated by functions \(F \in L_p^m({\mathbb R}^n)\). In particular, the following result is obtained: a family \({\mathbf P} = \{P_x: x \in E\}\) of polynomials of degree at most \(m - 1\) has the form \[ P_x = T_x^{m-1}[F] \quad \text{for every} \quad x \in E, \quad F \in C^{m-1} \cap L_p^m({\mathbb R}^n), \] if and only if, for some \(\gamma \geq 1\), \[ {\mathcal N}_{m,p,E}({\mathbf P}) := \sup \bigg\{\sum_{i=1}^k \sum_{|\alpha| \leq m-1} \frac{|D^\alpha P_{x_i}(x_i) - D^\alpha P_{y_i}(x_i)|^p}{\|x_i - y_i\|^{(m - |\alpha|)p - n}}\bigg\}^\frac1p < \infty. \] Here, \(\{\{x_i,y_i\}: i \in I\}\) is a family of pairs in \({\mathbb R}^n\) such that there exists a collection \(\{Q_i: i \in I\}\) of pairwise disjoint cubes in \({\mathbb R}^n\) such that \[ x_i,y_i \in \gamma Q_i \quad \text{and} \quad \text{diam}\, Q_i \leq \gamma\|x_i - y_i\| \quad \text{for all} \quad i \in I. \] Furthermore, \[ \|{\mathbf P}\|_{m,p,E} \sim {\mathcal N}_{m,p,E}({\mathbf P}), \] where \[ \|{\mathbf P}\|_{m,p,E} := \inf \big\{\|F\|_{L_p^m({\mathbb R}^n)}:F \in L_p^m({\mathbb R}^n), \;T_x^{m-1}[F] = P_x \;\;\text{for every} \;\;x \in E\big\}. \] (If \(p = \infty\), this result coincides with the classical Whitney-Glaeser theorem for \(m\)-jets.) The author's method is based on the representation of the space \(L_p^m({\mathbb R}^n)\), \(p > n\), as a union of \(C^{m,(d)}({\mathbb R}^n)\)-spaces, where \(d\) belongs to a family of metrics on \({\mathbb R}^n\) with certain ``nice'' properties.