Maximal average along variable lines (Q1001396)

For a vector field \(\vec{v}\) on \(\mathbb{R}^2\) one defines a family of lines \(\ell(x)=x-t\vec{v}(x)\) and a maximal operator \(Mf(x)=\sup_{r>0}{1\over r} \int_0^r |f(x-t \vec{v}(x))|\, dt\). If \(v(x_1,x_2)=(1,a(x_1))\) for some \(a\) defined on \([x_0,\infty)\) then one can define \[ \mathcal{M}_a f(x) =\chi_{[x_0,\infty)} (x_1) \sup_{r>0} {1\over r} \int_0^r |f(x_1-t,x_2-a(x_1) t)|\, dt. \] \textit{A. Carbery, A. Seeger, S. Wainger} and \textit{J. Wright} [J. Geom. Anal. 9, No. 4, 583--605 (1999; Zbl 0964.42003)] proved \(L^p\)-boundedness of \(\mathcal{M}_a\) under the assumption that \(a\) is convex and the hypothesis that, for each \(\ell>0\), the logarithmic derivative of \(a\) is comparable to one over the measure of the set on which \(da/dt\approx 2^\ell\) whenever \(t\) is in this set. The purpose of this note is to remove this condition on \(a\). More precisely, it is shown that if \(da/dt:[t_0,\infty)\to [0,\infty)\) is monotone increasing and vanishing at \(t_0\) then \(\mathcal{M}_a\) is bounded on \(L^p(\mathbb{R}^2)\) for each \(p\in (1,\infty]\). The proof proceeds by establishing a weak-\(L^1\) bound on a {\textit{local}} version of \(\mathcal{M}_a\) and a direct \(L^p\) bound for a {\textit{global}} version. Connections with related results in the literature are pointed out.