Continuous embedding between harmonic Hardy spaces and mixed-norm spaces (Q1367167)

Let \(D\subset \mathbb{R}^n\) and let \(\lambda\) be a defining function, i.e. such that \(\lambda\in C^{\infty}(\mathbb{R}^n)\), \(D=\{x\in \mathbb{R}^n;\lambda (x)<0\}\) is bounded, \(|\nabla\lambda|\neq 0\) in some neighbourhood of \(\partial D\). For \(r>0\) sufficiently small, set \(D_r=\{x\in \mathbb{R}^n;\lambda (x)<-r\};\) let \(d\sigma_r\) and \(d\sigma \) be the elements of \((n-1)\)-dimensional Hausdorff measures on \(\partial D_r\) and \(\partial D\), respectively. For \(f\) continuous on \(D\) and \(0<p<\infty ,\) the \(p\)-th integral mean of \(f\) on \(\partial D_r\) is defined by \[ M_p(f, r)=\Biggl[\int_{\partial D_r}|f(x)|^pd\sigma_r(x) \Biggr]^{1/p}. \] Let \(h(D)\) be the space of all complex valued harmonic functions on \(D\) and let \[ |\text{grad}_mf|=\biggl[\sum_{|\alpha|=m}|D^{\alpha}f|^2(\alpha !)^{-1} \biggr]^{1/2}, \] where \((D^{\alpha} f)(x)=\displaystyle{{\partial ^{|\alpha|}f}\over{\partial x^{\alpha_1}_1\dots\partial x^{\alpha_n}_n}}\), \(f\in h(D)\), \(x\in D\), \(\alpha =(\alpha_1,\dots,\alpha_n)\) \((\alpha_j\geq 0\) integers, \(j=1,\dots,n)\), \(|\alpha|=\alpha_1+\dots+\alpha_n\) [cf. e.g. \textit{A. P. Calderon} and \textit{A. Zygmund}, Studia Math. 24, 211-226 (1964; Zbl 0168.37002)]. \textit{F. Beatrous} [Studia Math. 98, 53-71 (1991; Zbl 0734.42011)] proved that there exist a neighbourhood \(U\) of \(\partial D\) and a diffeomorphism \(\Phi :\partial D\times [-2, 2]\rightarrow {\overline U}\) such that (i) \(\Phi (x, 0)=x\;\forall x\in \partial D\); (ii) \(\Phi \{\partial D\times (0, 2]\} ={\overline U}\cap D\) and \(\Phi\{\partial D\times [-2, 0)\} ={\overline U}\setminus{\overline D}\). The harmonic Hardy space \(H^p(D)\), \(0<p<\infty \) consists of all \(f\in h(D)\) satisfying \[ |f|_{H^p}= \Biggl[\int_{\partial D}f^*(x)^pd\sigma (x) \Biggr]^{1/p}<+\infty, \] where \(f^*(x)=\sup\{|f\circ\Phi (y, s)|;(y, s)\in\Gamma (x)\}\), and \(\Gamma (x)=\{ (y, s)\in\partial D\times (0, 1]; |y-x|<s\}\). In the note under review, the author asserts (without proof) that there exists an \(\varepsilon >0\) such that, for \(m\in \mathbb{N}\) and \(f\in h(D)\): \[ \int_0^{\varepsilon}r^{mp-1}M^p_p(\text{grad}_mf, r)dr\leq C|f|^p_{H^p}\quad \text{for}\quad 2\leq p<\infty, \] \[ |f|^p_{H^p}\leq C\Biggl[\int_0^{\varepsilon}r^{mp-1}M_p^p(\text{grad}_mf, r)dr+\sum_{j=0}^{m-1}|\text{grad}_jf(x_0)|^p \Biggr] \quad \text{for }x_0\in D,\;0<p\leq 2, \] \[ \int_0^{\varepsilon}r^{(n-1)k(\frac{1}{p}-\frac{1}{q})+mk-1}M_q^k(\text{grad}_mf, r)dr\leq C|f|^k_{H^p}\text{ for } 0<p<q<\infty ,\;p\leq k<\infty , \] \[ \int_D|\text{grad}_mf(x)|^q\rho (x)^{(n-1)\frac{q}{p}+mq-n}dm(x)\leq C|f|^q_{H^p}\quad\text{for}\quad 0<p<q<\infty , \] \[ \int _D|\text{grad}_mf(x)|^p\rho (x)^{mp-1}dm(x)\leq C|f|^p_{H^p}\quad\text{for}\quad 2\leq p<\infty, \] \[ |f|^p_{H^p}\leq C\Biggl[\int _D|\text{grad}_mf(x)|^p\rho (x)^{mp-1}dm(x)+\sum_{j=0}^{m-1}|\text{grad}_jf(x_0)|^p\Biggr]\text{ for } x_0\in D,\;0<p\leq 2, \] where \(\rho (x)\) is the Euclidean distance from \(x\in D\) to \(\partial D\). Similar results have been obtained e.g. by \textit{F. Beatrous} [Indiana Univ. Math. J. 40, 915-966 (1991; Zbl 0726.32005)] for holomorphic functions on general bounded domains in \(\mathbb{C}^n\).