Sobolev-Hardy space with general weight (Q2493010)

In this paper, the authors prove the following \(k\)th order Hardy inequality with general weight. Let \(\Omega\) be a bounded domain. Then, under the assumptions \((H_1)\) and \((H_2)\), for each positive integer \(k\) the inequality \[ \int_{\Omega}\phi|\nabla u|^2\,dx-\int_{\Omega}\phi\sum_{i=1}^{k}\left(\frac{h_i'}{h_i}\right)^2u^2\,dx\geq\int_{\Omega} \phi\left( \frac{h_{k+1}'}{h_{k+1}}\right)^2u^2\,dx,\tag{1} \] holds for all \(u\in H^1_0(\Omega,\phi)\). In addition, the constant \(1\) in the previous inequality is sharp. The above assumptions \((H_1)\), \((H_2)\) are stated as follows. {\((H_1)\)} \(\phi (r) \in C^1(0, a)\), where \(a\) is such that \( \Omega \subset B_a(0)\), is a positive function and \( \phi(| x| ) \in L^1_{\text{Loc}}(\Omega)\). {\((H_2)\)} For all \(i= 1,\dots,k\), \(h_i(r)\in C^1(0,a)\) is a positive function which satisfies \[ r^{N-1}\phi_i\cdot(h_i^2)'=c_i, \] where \(N\geq 2\), \(\phi_1=\phi\), \(\phi_{i+1}=\phi_i h_i^2\) and \(c_i< 0\). Moreover, we must have \(h_1^{-1}(0)=0\). As the authors notice, the inequality (1) is a true generalization of existing Hardy inequalities since if \(\phi\equiv 1\), one can obtain, among other things, with an appropriate choice of the functions \(h_i\), the inequality of \textit{S.\,Filippas} and \textit{A.\,Tertikas} [J.~Funct.\ Anal.\ 192, 186--233 (2002; Zbl 1030.26018)] or the results of \textit{Adimurthi} et al.\ in [Proc.\ Am.\ Math.\ Soc.\ 130, 489--505 (2002; Zbl 0987.35049) and Proc.\ R.\ Soc.\ Edinb., Sect.\,A 132, 1021--1043 (2002; Zbl 1029.35194)]. Next, the authors introduce quite naturally the \(k\)th Sobolev--Hardy space \(H_{0,k}^1 (\Omega,\phi)\) which is the completion of \(C_0^{\infty}(\Omega)\) with respect to the norm \(\| u \|_{1,k,\phi}\) derived from the inner product \[ (u, v)_{1, k, \phi} = \int_{\Omega}\left(\phi\nabla u\cdot\nabla v - \phi\sum_{i=1}^{k} \left(\frac{h_i'}{h_i}\right)^2 uv\right)dx. \] Then, the corresponding Poincaré inequality proven by the authors can be written as follows: Let \(N \geq 2\). Still under the assumptions \((H_1)\) and \((H_2)\), the following inequality holds for every \(u \in H^1_{0, k}(\Omega, \phi)\): \[ \| u \|^{2}_{1, k, \phi}\geq\lambda_{1}(\phi_{k+1})\int_{\Omega}\phi_{k+1}u^2\,dx, \] where \[ \lambda_{1}(\phi_{k+1}) := \inf_{u\in H_0^1(\Omega,\phi_{k+1})}\frac{\int_{\Omega}\phi_{k+1}|\nabla u | ^2\,dx}{\int_{\Omega}\phi_{k+1}u^2\,dx}. \] Finally, an interesting consequence of the last theorem presented by the authors is that \[ H_0^1(\Omega, \phi)\subset H_{0,k}^1(\Omega,\phi)\subset\bigcap_{1\leq q<2}W_0^{1,q}(\Omega,\phi). \] In addition, the space \(H_{0, k}^1(\Omega,1)\) is compactly embedded in the space \(L^p(\Omega)\) for each \(p<Nq/(N-q)\), where \(1\leq q<2 \) and \(N\geq 2 \).