Embedding theorems for generalized Sobolev spaces (Q1577165)

Let \((X,d)\) be a metric space with finite diameter, let \(\mu\) be a finite positive Borel measure, and let \(1<p\leq \infty\). The author considers the Sobolev spaces \(L^1_p(X,d,\mu)\) and \(W^1_p(X,d,\mu)\) defined by \textit{P.~Hajłasz} [Potential Anal. 5, No. 4, 403-415 (1996; Zbl 0859.46022)]. It is supposed that the measure \(\mu\) satisfies the doubling condition, i.e., there is a constant \(c\) such that, for all balls, \(\mu (B(x,2r))\leq c\mu(B(x,r))\). The author proves the following theorem: Suppose that \(1<p<\infty\), \(0<\alpha \geq 1\), and \((1-\alpha)p<n\). If the metric \(\rho\) admits an estimate \(\rho(x,y)\geq cd(x,y)^{\alpha}\) then the space \(L^1_p(X,d,\mu)\) is continuously embedded into \(L^1_q(X,\rho,\mu)\), where \(\frac 1q=\frac 1p-\frac{1-\alpha}{n}\). An analog of the classical embedding theorem is also established. In the second section, the author considers an embedding theorem of various measures that is associated with the problem of describing the traces of functions. Let \(E\subset X\) and let a measure \(\nu\) be such that \(\nu\bigl(B(x,r)\bigr)\leq cr^{-\alpha}\mu\bigl(B(x,r)\bigr)\) for every ball \(B(x,r)\) centered at \(x\in E\). Here \(0<\alpha<n\). The following theorem is proven: Suppose that \(1<p<\infty\) and \(0<\alpha<\min (n,p)\). Then the space \(L^1_p(X,d,\mu)\) is continuously embedded into the space \(L^1_q(E,d^{1-\gamma})\), where \(\alpha/p<\gamma<\min(1,\alpha,n/p)\) and \(q\leq \frac{p(n-\alpha)}{n-\gamma p}\).