Inversion of potential-type operators with symbols degenerate on hyperboloids and paraboloids (Q2460488)

It is well-known that the Riesz potential with a density \(\varphi\) in \(L^p\) may be left inverted by means of hypersingular integral operator. There were a variety of papers devoted to such an inversion of more general potential type operators. In the case of translation-invariant potential operators \[ K_\theta^\alpha \varphi=\int_{R^n}k(x-y)\varphi(y)\,dy, \quad k(x)=\frac{\theta(x)}{| x| ^{n-\alpha}} \] with rather smooth and bounded function \(\theta(x)\), the inversion is known to be constructed also in terms of the hypersingular integral operators with a certain characteristic explicitly depending on \(\theta(x)\); in the elliptic case, see, for instance, the book of the reviewer [``Hypersingular integrals and their applications (Analytical Methods and Special Functions 5, London:\ Taylor \& Francis) (2002; Zbl 0998.42010)]. Besides the method of hypersingular operators, for such an inversion there is also known to be used the method of approximative inverse operators, constructed as limits of convolution operators with ``nice'' kernels. The latter method, in particular, is more effective in non-elliptic cases when the symbol of the operator (that is, the Fourier transform of the kernel) may vanish. Although \textit{M.\,M. Zavolzhenskii} and \textit{V.\,A. Nogin} [Dokl.\ Ross.\ Acad.\ Nauk 324, 738--741 (1992; Zbl 0808.47026)] showed that the approximative inverse operator may be constructed in the case of an arbitrary set of degeneracy of measure zero, this general construction is rather awkward and difficult to use for concrete cases of the degeneracy. In the present paper, the authors deal with potential operators whose symbol may degenerate on a hyperboloid or a paraboloid. Namely, they consider the operators defined by their symbols \(\hat{k}(\xi)=\frac{\xi_1^2-\xi_2^2-\cdots-\xi_n^2+a}{| \xi| ^{\alpha+2}},\;0<\alpha<n-2,\;a>0\) and \(\hat{k}(\xi)=\frac{\xi_n-| \xi^\prime| ^2}{| \xi| ^{\alpha+2}},\;0<\alpha<n-1\), where \(\xi^\prime=(\xi_1,\xi_2,\dots,\xi_{n-1})\). In these two cases, they construct the left inverse operators on the range \(K_\theta^\alpha(L^p)\) as the limit of convolution operators with integrable kernels which may be effectively constructed in terms of elementary or special functions.