Tangent-point repulsive potentials for a class of non-smooth \(m\)-dimensional sets in \(\mathbb R^n\). I: smoothing and self-avoidance effects (Q363200)

The authors study the energy functional \[ \mathcal{E}_q(\Sigma)=\int_\Sigma \int_\Sigma{{1}\over {R^q_{\mathrm{tp}}(x,y)}}d\mathcal{H}^m(x)d\mathcal{H}^m(y), \] defined for a class \(\mathcal{A}\) of admissible \(m\)-dimensional sets in \(\mathbb{R}^n\). For each \(\Sigma \in \mathcal{A}\) a weak counterpart of the classic tangent plane is defined almost everywhere with respect to the \(m\)-dimensional Hausdorff measure \(\mathcal{H}^m\) on \(\Sigma\). In the case of \(\mathcal{H}\), i.e., for almost each \(x\in \Sigma\) there is an \(m\)-plane \(H_x\), such that the portion of \(\Sigma \) near the point \(x\) is close to the affine plane \(x+H_x\subset \mathbb{R}^n\). The quantity \[ R_{\mathrm{tp}}(x,y)= {{|y-x|^2}\over {2\;\mathrm{dist}\;(y,x+H_x)}} \] denotes the radius of the smallest sphere tangent to the affine plane \(x+H_x\) and passing trough \(y\). The authors show that for the infinite range of exponents \( q>2m\), the finiteness of \(\mathcal{E}_q(\Sigma)\) has three sorts of consequences for any admissible set \(\Sigma \in \mathcal{A}\): measure-theoretic, topological and analytical. They present some examples of such admissible surfaces. Theorem 1.2 (uniform Ahlfors regularity): Assume that \(\Sigma \in \mathcal{A}\) is an admissible \(m\)-dimensional surface in \(\mathbb{R}^n\) with \(\mathcal{E}_q(\Sigma)\leq E,\;q>2m\). Then there exists a constant \(a_1=a_1(q,n,m)>0\), depending only on \(q,n\) and \(m\), such that \[ \mathcal{H}^m(\Sigma \cap B(x,r))\geq (1/2)\omega(m)r^m, \] for all \(x\in \Sigma\) and all radii \[ 0<r,R_l=R_l(q,n,m,E)=a_1/E^{1/(q-2m)}. \] Theorem 1.3 (uniform decay of \(\beta\)-numbers): Let \(\Sigma \in \mathcal{A}\) be an admissible \(m\)-dimensional surface in \(\mathbb{R}^n\) with \(\mathcal{E}_q(\Sigma )<E\) for some \(q>2m\). Then there exist two constants \(a_2(q,n,m)>0\) and \(A_2(q,n,m)<\infty\), both depending on \(n,m, q\), such that whenever the radius \[ d\leq R_2=R_2(q,n,m,E)=a_2(q,n,m)/E^{1/(q-2m)} \] and the bound \(\varepsilon > 0\) satisfy the balance condition \[ \varepsilon^{4m+q}d^{2m-q}\geq A_2(q,n,m)E, \] then we have \[ \beta_{\Sigma}(x,d)= \inf_{P\in G(n,m)}\biggl(\sup_ {y\in B(x,d)\cap \Sigma}\frac{\text{dist}(y,x+P)}{d}\biggr)\leq \varepsilon, \,\, x\in \Sigma, \] where \(G(n,m)\) denotes the Grassmannian of all \(m\)-dimensional subspaces in \(\mathbb{R}^n\). Theorem 1.4 (geometric Sobolev-Morrey embedding): Let \(\Sigma \in \mathcal{A}\) and \(\mathcal{E}_q(\Sigma)< \infty\) for some \(q>2m\). Then \(\Sigma\) is an embedded submanifold of class \(C^{1,\mu}\), where \(\mu =1-2m/q\). The authors obtain the values of the constants \(a_3\) and \(A_3\) assuring the validity of this theorem. Some comments and conjectures are presented.