Averaging operators over nondegenerate quadratic surfaces in finite fields (Q2260258)

Let \(V\) be an algebraic variety in \(\mathbb F_q^d\). For a function \(f : \mathbb F_q^d\to \mathbb C\) define an averaging operator by: \[ Af(x) =\frac{1}{|V|}\sum_{y\in V} f(x-y). \] Let \(dx\) be the normalized counting measure on \(\mathbb F_q^d\). The \textit{averaging problem} for \(V\) is to determine \(p\) and \(r\), \(1\leq p, r\leq\infty\), such that there exists a constant \(C\), independent of \(f\) and \(q\), such that: \[ ||Af||_r\leq C||f||_p,\tag{*} \] where \(||f||_p\) denotes the \(L^p\)-norm, with respect to \(dx\), of \(f\). The author considers the case where \(q\) is odd and \(V\) is the zero set of a non-degenerate quadratic form in \(d\) variables. In a recent work, the author and \textit{C.-Y. Shen} [Proc. Edinb. Math. Soc., II. Ser. 56, No. 2, 599--614 (2013; Zbl 1279.42007)] solved the averaging problem for \(d\geq 3\) odd. Now, suppose \(d\geq 4\) is even and that \(V\) contains a \(d/2\)-dimensional subspace. The author and Shen showed that a necessary condition for (*) is that \((1/p, 1/r)\) lies in the convex hull of \((0,0), (0,1), (1,1)\) and \[ \left(\frac{d^2-2d+2}{d(d-1)}, \frac{1}{d-1}\right), \quad \left(\frac{d-2}{d-1}, \frac{d-2}{d(d-1)}\right). \] Here, the author shows that this is also sufficient, thus solving the averaging problem. For \(d\geq 6\) the proof depends on estimates on the \(L^2\)-norm and \(L^{\infty}\)-norm for the convolution \(E*\hat K\) and then applying interpolation. Here, \(E\) is a characteristic function and \(\hat K\) is the Fourier transform of the Bochner-Riesz kernel. However, interpolation cannot be used for the \(d=4\) case which requires numerous, delicate estimates.