No label defined (Q4713490)

The authors study a family \(T_ \alpha\) of operators on the sphere \(S^{n-1}\), \(n\geq 3\), where \(T_ \alpha\) is defined by \[ T_ \alpha f(x')= \frac{1}{\Gamma(1-\alpha)}\int_{S^{n-1}}| x'\cdot y'|^{-\alpha}f(y')\,dy',\quad x'\in S^{n-1} \] for \(0\leq \Re \alpha<1\). The definition of \(T_ \alpha\) is then extended to \(\Re \alpha\geq 1\) by analytic continuation. Such operators have been studied by \textit{R. Strichartz} [Trans. Am. Math. Soc. 148, 461--471 (1970; Zbl 0199.17502)], who proved, among others, the boundedness of \(T_ \alpha\) from \(L^{n/(n-\alpha)}\) into \(L^{n/\alpha}\) with \(0<\alpha<n/2\) by different methods. In the paper under review the authors regard the sphere \(S^{n-1}\) as the homogeneous space \(G/K\), where \(G=\mathrm{SO}(n)\) and \(K\) denotes the isotropy subgroup of \(G\) at the north pole \({\mathbf{1}}=(0,\dots,0,1)\). For each function \(f\) on \(S^{n-1}\), \(f^{\natural}\) is defined by \(f^{\natural}(x)=f(x{\mathbf{1}})\) for \(x\in G\). Then the (spherical) convolution of two functions \(f\) and \(g\) on \(S^{n-1}\) is defined by \[f*g(x{\mathbf{1}})=f^{\natural}*g^{\natural}(x)\quad\text{for }x\in G.\] So, \(T_ \alpha f=f*K_ \alpha\) with \(K_ \alpha(y')=| y'\cdot{\mathbf{1}}|^{-\alpha}/\Gamma(1-\alpha)\). Since \(K_ \alpha\) is zonal, it has an orthogonal expansion as \(K_ \alpha=\sum^ \infty_{k=0} b_{k,\alpha}Z^{(k)}_ 1\), where \(Z^{(k)}_ 1\) is the suitably normalized zonal spherical harmonic of degree \(k\). It follows that \(T_ \alpha f\sim\sum^ \infty_{k=0} b_{k,\alpha}{\mathcal P}_ k f\), \(f\in L^ 2(S^{n-1})\), where \({\mathcal P}_ k\) denotes the orthogonal projection to the space of spherical harmonics of degree \(k\). The authors begin with an exact expression for coefficients \(b_{k,\alpha}\). This enables them to have precise estimate for the growth rate of the coefficients and thus obtain the region to which \(T_ \alpha\) is analytically continued. It is then seen that \(T_ \alpha\) exists for \(\alpha\) belonging to the strip \(S=\{\alpha\in\mathbb{C}: 0\leq \Re \alpha\leq n/2\}\) and that \(T_ \alpha\) is a bounded operator from \(L^{n/(n-\Re \alpha)}\) into \(L^{n/\Re \alpha}\) with \(\alpha\in S\). This is similar to Strichartz's result but the present method seems to have wider application. The authors proceed further to obtain main boundedness results. For this purpose they use \textit{L. Colzani}'s theory of Hardy classes \(H^ p(S^{n-1})\) on the sphere \(S^{n-1}\), developed in [Boll. Unione Mat. Ital., VI. Ser., C 4, 219--244 (1985; Zbl 0585.42021)], in which the Hardy classes are shown to be defined via atomic decompositions as well as maximal functions. To state the main result, let \(P\equiv(1/p_0, 1/q_0)\) be in the first quadrant \(Q\) of the plane. An operator \(T\) is said to be bounded at \(P\) if it is a bounded map from \(H^{p_ 0}(S^{n-1})\) to \(L^{q_ 0}(S^{n-1})\) in case \(0<p_0\leq 1\) and from \(L^{p_ 0}(S^{n-1})\) to \(L^{q_ 0}(S^{n-1})\) in case \(p_0>1\). \(T\) is said to be bounded in \(E\subset Q\) if it is bounded at each point in \(E\). For each number \(\alpha\in S\) we set \[ A_\alpha\equiv\left(\frac{n-1-\Re \alpha}{n-2},\frac{n-1-\Re \alpha}{n-2}\right),\qquad B_\alpha\equiv\left(1-\frac{\Re \alpha}{n}, \frac{\Re \alpha}{n}\right), \] \[ C_\alpha\equiv(1- \Re \alpha,0),\qquad D_\alpha\equiv\left(\frac{\Re \alpha - 1}{n-2},\frac{\Re \alpha-1}{n-2}\right) \] and \[ S_\alpha\equiv\bigl\{(x,y)\in \mathbb R^2: x=0,\ y\geq\max\{(\Re \alpha-1)/(n-2),0\}\bigr\}. \] Define \(E_\alpha\) to be the closed convex hull generated by \(A_\alpha\), \(B_\alpha\), \(C_\alpha\), \(S_\alpha\) if \(0\leq \Re \alpha<1\), and by \(A_\alpha\), \(B_\alpha\), \(D_\alpha\), \(S_\alpha\) if \(\Re \alpha\geq 1\). The main result (Theorem 2) states that, for any \(\alpha\) belonging to \(S\), the operator \(T_\alpha\) is bounded in \(E_\alpha\). If \(\Re \alpha=1\), then boundedness at \(D_\alpha\) is interpreted as a map from \(L^\alpha\) to BMO. A discussion concerning the optimality of this result is also included.