Weighted Fourier Transform Inequalities for Radially Decreasing Functions

From MaRDI portal





The authors consider weighted Fourier transform estimates \(\|\widehat f\|_{q,u}\leq C\| f\|_{p,v}\), \(0<p, q<\infty\), where \(u\), \(v\) are positive weight functions and \[ \widehat f(x)=\int_{\mathbb{R}^ n} e^{-2\pi ix\cdot y}f(y)dy,\quad x\in\mathbb{R}^ n. \] Here, \(\| f\|_{p,v}=\bigl(\int_{\mathbb{R}^ n}| f(x)|^ p v(x)dx\bigr)^{1/p}\). Results are known when \(u\) and \(v\) satisfy monotonicity conditions. The authors, extending work of the second author [Teubner-Texte Math. 119, 42-85 (1990; preceding review)], consider the case in which \(f\) decreases radially to zero and \(u(x)=v(x'/| x|)| x|^{n(p-2)}\), where \(x=| x| x'\) defines \(x'\) as an element of the unit \(n\)-sphere. J. J. Benetetto, the second author and the reviewer showed in one dimension that if \(v\) is a monotone increasing function, the inequality \[ \int v(1/x)| x|^{p-2}| \widehat f|^ p dx\leq C^ p\int v(x)| f(x)|^ pdx \] holds if and only if \(v\) is a weight for which the Hardy-Littlewood maximal operator is bounded from \(L_{p,v}\) to \(L_{p,v}\). For \(n=1\), the authors consider the case in which \(v\) is arbitrary and show that the corresponding inequality is valid for the set of even functions decreasing to zero if and only if the averaging operator \(Af(x)={1\over x}\int^ x_ 0 f(t)dt\) is bounded from \(L_{p,v}\) to \(L_{p,v}\). (The Hardy-Littlewood maximal function of a decreasing function is the averaging operator applied to that function.) The authors show that if \(1\leq p<\infty\) and \(v\) is a positive weight, then \[ \int_{\mathbb{R}^ n} v(x'/x)| x|^{n(p- 2)}|\widehat f(x)|^ p dx\leq C^ p\int_{\mathbb{R}^ n} v(x)| f(x)|^ p dx \] holds for all positive radial functions \(f(x)=f_ 0(| x|)\) with \(g(t)=t^{n-1} f_ 0(t)\) decreasing to zero iff \[ \int_{| x|\geq r} | x|^{-np} v(x)dx\leq Br^{-p} \int_{| x|\leq r} | x|^{-p(n-1)} v(x)dx. \] In each case necessity follows by taking \(f(x)=| x|^ \alpha\chi_{B_ r}\) for suitable \(\alpha\). Sufficiency is proved by use of the second mean-value theorem to convert the Fourier transform, which is a cosine transform, into a sine transform. Similar, although more complicated, results are given for the two- weighted case, with \(u\) of the form \(u(1/x)| x|^{q-2}\), when the index is in one of the ranges \(1<p\leq q<\infty\), \(0<q<p<\infty\), \(p>1\) or \(0<p<1<q\). They give necessary and sufficient conditions for a one-weighted estimate for the Hankel operator \[ H_ \lambda f(x) = \int^ \infty_ 0 (xt)^{1/2} J_ \lambda(xt) f(t)dt \] for \(\lambda>1/2\), where \[ J_ \lambda(z) = C_ \lambda z^ \lambda \int^ 1_ 0 (1-t^ 2)^{\lambda-1/2}\cos(zt)dt, \] and give two- weighted estimates for the K-transform, where the Bessel function \(J_ \lambda\) is replaced by a modified Bessel function of the third kind. This latter result gives results for the Laplace transform, which is \(\sqrt{{2\over\pi}} K_{\pm 1/2} f(x)\).











This page was built for publication: Weighted Fourier Transform Inequalities for Radially Decreasing Functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4032581)