On fractional degrees of certain differential operators commuting with dilatations (Q1897925)

The authors consider the Riesz potential like operators associated with the differential operator \(\pm\sum^n_{j=1} x^2_j {{\partial^2} \over {\partial x^2_j}}\) on the spaces \[ \text{Ł}_p (\mathbb{R}^n_+)= \Biggl\{f(x): |f|^p_{\text{Ł}_p}= \int_{\mathbb{R}^n_+}|f(x)|^p\;{dx_1 \over x_1}\cdots {dx_n \over x_n}<+ \infty\Biggr\}, \qquad 1\leq p<\infty, \] and for \(\gamma\in \mathbb{R}\), \[ \text{Ł}_{p,\gamma} (\mathbb{R}^n_+)= \left\{ f(x): |f|^p_{\text{Ł}_{p,\gamma}}= \int_{\mathbb{R}^n_+}|f(x)|^p \left(1+ \sqrt{\sum^n_{j=1} (\ln x_j)^2} \right)^{-\gamma} {dx_1 \over x_1} \cdots {dx_n \over x_n}<+ \infty\right\}, \] \(1\leq p<c\), where \(\mathbb{R}^n_+= \{x\in \mathbb{R}^n\mid x_j>0\), \(j=1,\dots, n\}\). The fractional integrals are defined in terms of the Mellin transform by \[ {\mathcal M}^*I^\alpha_\pm \phi(\xi)= \Biggl(\mp|\xi|^2\pm i\sum^n_{j=1} \xi_j\Biggr)^{-\alpha/2} {\mathcal M}\phi(\xi), \] and have a direct integral representation as a (multiplicative) convolution operator \[ {}^*I^\alpha_- \phi(x)= \int_{\mathbb{R}^n_+} i_{\alpha,-}(t) \phi(x/t) {dt_1 \over t_1}\cdots {dt_n \over t_n}. \] Their main results are a Sobolev type theorem on the spaces above for the operators \({}^*I^\alpha_\pm\), a constructive formula for the inverse of the operators \({}^*I^\alpha_\pm\) and results about the images of the spaces \(\text{Ł}_p\) under \(({}^*I^\alpha_\pm)^{-1}\). The problems are transferred to \(\mathbb{R}^n\) via the map \(Q\phi(x)=\phi(\overline{e^x})= \phi(e^{x_1},\dots, e^{x_n})\), which sends \({}^*I^\alpha_\pm\) into the operator \({}^*J^\alpha_\pm\) with symbol \[ \Biggl( \mp|\xi|^2\pm i\sum^n_{j=1} \xi_j\Biggr)^{-\alpha/2} \] via the formula \[ {}^*I^\alpha_\pm \phi=Q^{-1} {}^*J^\alpha_\pm Q\phi. \] The authors' main result is: Theorem. If \(0< \text{Re } \alpha<n+1\), then the operator \({}^*I^\alpha_-\) is defined on functions from \(\text{Ł}_p\), \(1\leq p<{{2n} \over {n+\text{Re } \alpha-1}}\), and it is a bounded operator from \[ \text{Ł}_p\to \text{Ł}_{p,\gamma}, \quad \gamma> {{n+\text{Re } \alpha-1} \over 2} p, \qquad 0<\text{Re } \alpha<n-1; \] for \(\text{Re } \alpha=n-1\), from \[ \text{Ł}_1\to \text{Ł}_{1,\gamma}, \qquad \gamma> {{n+\text{Re } \alpha-1} \over 2}; \] and from \[ \text{Ł}_p\to \text{Ł}_q, \quad 1<p< {{2n}\over {n+\text{Re } \alpha-1}}= {n\over{n-1}}, \quad q={{np}\over {n-p(n-1)}}. \] Finally, if \(n-1<\text{Re } \alpha<n+1\), the operator is bounded from \(\text{Ł}_p\to \text{Ł}_q\), for an interval of \(q\) given \(1\leq p<{{2n}\over {n+\text{Re } \alpha-1}}\) (see the paper for the precise interval). The Sobolev like results are seen in that \({}^*I^\alpha_-\) is a bounded operator from \(\text{Ł}_p\) into \(\text{Ł}_q\) if \(p\neq 1\) is as above, and \(1/q= 1/p- {{n+\text{Re } \alpha-1} \over{2n}}\), with a similar result if \(n-1<\text{Re } \alpha<n\), \(1<p< {n\over {\text{Re } \alpha}}\) namely \({}^*I^\alpha_-\) maps \(\text{Ł}_p\) into \(\text{Ł}_q\), where \(1/q= 1/p- {{\text{Re } \alpha}\over n}\). The connection with Sobolev is reinforced by the relation \({}^*J^\alpha_\pm \phi(x)= ({2\over {\sqrt{n}}})^\alpha R^{-1} G^\alpha_\pm R\phi\), where \(G^\alpha\) is the usual Bessel potential operator acting on \(\mathbb{R}^n\), and \(R\phi(x)= \exp(- {{x_1+\dots+x_n} \over {\sqrt n}})\phi ({2\over {\sqrt n}}x)\). The relation with \(G^\alpha\) is used to give the inversion formula. The inversion formula, which is given by a principal value integral, is too complicated to give here. It is used to characterize \(({}^*I^\alpha_\pm)^{-1}\) of the above spaces.