Maximal functions for convex curves (Q1110760)

The Hilbert transform \({\mathfrak h}\) and maximal operator \({\mathfrak m}\) associated with a plane curve \(\Gamma (t)=(t,\gamma (t))\) are defined by \[ {\mathfrak h}f(x)=p.v.\int^{\infty}_{-\infty}f(x-\Gamma (t))t^{- 1}dt,\quad {\mathfrak m}f(x)=\sup_{r\geq 0}r^{-1}\int^{r}_{0}| f(x-\Gamma (t))| dt. \] The authors are assuming that \(\gamma\) is convex for \(t\geq 0\), continuous and satisfies \(0=\gamma (0)=\gamma '(0)^+\) (the right hand derivative of \(\gamma\) at 0). They had previously shown that when \(\gamma\) is odd, the boundedness of \({\mathfrak h}\) depends on \(h(t)=t\gamma '(t)-\gamma (t),\) which is the negative of the y-intercept of tangent line to the curve \(\Gamma\), namely, \({\mathfrak h}\) is bounded on \(L^ 2({\mathbb{R}}^ 2)\) if and only if h has bounded doubling time, i.e., there is a \(D>1\) such that h(Dt)\(\geq 2h(t)\) (\(\gamma\) odd) [the authors, ibid. 50, 735-744 (1983; Zbl 0524.44001)]. Here they show that if h has bounded doubling time, \({\mathfrak m}\) is bounded on \(L^ 2({\mathbb{R}}^ 2)\). The converse to this is known to be false [\textit{H. Carlsson} and \textit{S. Wainger}, Indiana Univ. Math. J. 34, 815-823 (1985; Zbl 0587.42011)].